Time for change: Towards the implementation of new ...

THÈSE Pour obtenir le grade de

DOCTEUR DE LA COMMUNAUTE UNIVERSITE GRENOBLE ALPES Spécialité : Biodiversité, Écologie, Environnement Arrêté ministériel : 25 mai 2016

Présentée par

Kálmán TAPOLCZAI Thèse dirigée par Agnès BOUCHEZ, DR, INRA UMR CARRTEL, et codirigée par Frédéric RIMET, IE, INRA UMR CARRTEL préparée au sein du Laboratoire INRA UMR CARRTEL dans l'École Doctorale SISEO

Time for change: Towards the implementation of new approaches in diatom-based ecological quality assessment for rivers Thèse soutenue publiquement le 15 décembre 2017 devant le jury composé de :

Mme Judit PADISÁK Professor, MTA-PE Limnoecology Research Group and Department of Limnology, University of Pannonia, Hungary (présidente)

M Jean-Nicolas BEISEL Professor, Ecole Nationale du Génie de l'Eau et de l'Environnement de Strasbourg (ENGEES), Laboratoire Image Ville Environnement (LIVE) (rapporteur)

M Gábor BORICS Professor, Department of Tisza Research, Danube Research Institute, Centre for Ecological Research, Hungarian Academy of Sciences, Hungary (rapporteur)

M Rainer KURMAYER Professor, Research Department for Limnology, University of Innsbruck, Austria (rapporteur)

Mme Viktória BÁCSI-BÉRES Research fellow, Department of Tisza Research, Danube Research Institute, Centre for Ecological Research, Hungarian Academy of Sciences, Hungary (membre)

M Kristian MEISSNER Research programme manager, Jyväskylä office, Freshwater Centre, Finnish Environmental Institute, Finland (membre)

THÈSE Pour obtenir le grade de

DOCTEUR DE LA COMMUNAUTE UNIVERSITE GRENOBLE ALPES Spécialité : Biodiversité, Écologie, Environnement Arrêté ministériel : 25 mai 2016

Présentée par

Kálmán TAPOLCZAI Thèse dirigée par Agnès BOUCHEZ, DR, INRA UMR CARRTEL, et codirigée par Frédéric RIMET, IE, INRA UMR CARRTEL préparée au sein du Laboratoire INRA UMR CARRTEL dans l'École Doctorale SISEO

Time for change: Towards the implementation of new approaches in diatom-based ecological quality assessment for rivers Thèse soutenue publiquement le 15 décembre 2017 devant le jury composé de :

Mme Judit PADISÁK

Professor, MTA-PE Limnoecology Research Group and Department of Limnology, University of Pannonia, Hungary, (présidente)

M Jean-Nicolas BEISEL

Professor, Ecole Nationale du Génie de l'Eau et de l'Environnement de Strasbourg (ENGEES), Laboratoire Image Ville Environnement (LIVE) (rapporteur)

M Gábor BORICS

Professor, Department of Tisza Research, Danube Research Institute, Centre for Ecological Research, Hungarian Academy of Sciences, Hungary (rapporteur)

M Rainer KURMAYER

Professor, Research Department for Limnology, University of Innsbruck, Austria (rapporteur)

Mme Viktória BÁCSI-BÉRES Research fellow, Department of Tisza Research, Danube Research Institute, Centre for Ecological Research, Hungarian Academy of Sciences, Hungary (membre)

M Kristian MEISSNER Research programme manager, Jyväskylä office, Freshwater Centre, Finnish Environmental Institute, Finland (membre)

Centre Alpin de Recherche sur les Réseaux Trophiques et Écosystèmes Limniques

Institut National de la Recherche Agronomique

Université Savoie Mont Blanc

INRA UMR CARRTEL 75 bis avenue de Corzent CS 50511 74203 THONON LES BAINS cedex Tel. : +33(0)4 50 26 78 00 http ://www.dijon.inra.fr/thonon

Because you'll always find life growing there where the rivers flow — Attila and Friends, Rivers

First of all, I would like to thank my thesis director, Agnès Bouchez and my supervisors, Frédéric Rimet, Judit Padisák and Csilla Stenger-Kovács their effort in supervising my work, their support, help and encouragement. I am grateful to my thesis committee, Philippe Usseglio-Polatera, Floriane Larras and Estelle Lefrançois for the helpful committee meetings. I thank AFB (Agence Française pour la Biodiversité) for funding the project within which this work was done and INRA (Institut National de la Recherche Agronomique) and Université Savoie Mont Blanc where this work was done. I would also thank the teams at BRGM (Bureau de Recherches Géologiques et Minières) and at DEAL (Direction de l’Environnement de l’Aménagement et du Logement) in Mayotte, Nathalie Mary from ETHYCO (ETudes des HYdrosystèmes Continentaux) and Gilles Gassiole for their help, for the data they provided and the discussions during the sampling mission on the field. I thank my friends and colleagues, Valentin Vasselon and François Keck for this great collaboration between PhD students that resulted in several interesting papers published. Special thanks to Valentin for the molecular data and his expertise that helped me to understand and be able to use such data in my work. I would like to express my gratitude to everyone at the hydrobiological station of INRA UMR CARRTEL for these years, their kind welcome, the fantastic atmosphere, and their patience for my development in the French language. It was a real pleasure to work here. I am very grateful to all my friends who I met in France and to friends from before, to my family, and to Diána for all their supports during this time. Thank you all.

i

é

é

Les lacs et les rivières sont exposés à une forte pression due aux activités anthropiques. La dégradation des eaux met en danger non seulement l’approvisionnement en eau mais aussi tout l'écosystème aquatique. Plusieurs méthodes ont été développées pour surveiller la qualité de ces masses d'eau afin de proposer des solutions pour réduire l’impact anthropique. Aujourd'hui, la directivecadre européenne sur l'eau (DCE) est un cadre législatif pour l'Union européenne (UE) au sein duquel les États développent leurs stratégies nationales de biosurveillance des écosystèmes aquatiques. La qualité écologique des masses d’eau est évaluée au travers de différents groupes d’organismes : les éléments de qualité biologique, parmi lesquels le phytobenthos. Les diatomées sont un groupe de microalgues riche en espèces et largement majoritaire dans la communauté phytobenthique, utilisées comme proxy du phytobenthos. Elles ont plusieurs caractéristiques qui en font d’efficaces bioindicateurs : réponse rapide aux changements environnementaux et facilité d'échantillonnage et d'analyse. Pour satisfaire aux exigences de la DCE, la majorité des pays ont importé des indices diatomiques déjà existants. Cependant, ces indices comportent plusieurs problèmes : approche non holistique, hypothèse d’une répartition gaussienne de la réponse des espèces le long des gradients de pollution, espèces rares négligées, besoin d'experts pour l’identification précise des espèces, sensibilité aux erreurs d'identification, temps d’analyse en microscopie long. Cette thèse donne une vue d’ensemble des méthodes existantes pour l'évaluation de la qualité écologique basée sur les diatomées, présentant leurs possibles évolutions, leurs avantages et inconvénients. Tout d’abord, les approches basées sur les traits et les classifications écologiques du phytobenthos sont discutées dans une review, comme méthode alternative de bioindication. Ensuite, deux études ont été réalisées sur les données du réseau de biomonitoring des rivières d’un nouveau département français, l'île de Mayotte, pour développer des indices de qualité écologique alternatifs ou complémentaires. La première de ces études a permis de tester une approche de bioindication basée sur les traits et la seconde une approche basée sur le metabarcoding ADN, utilisant directement des OTUs (Operational Taxonomical Unit). Ces deux approches ouvrent des perspectives pour remplacer les indices classiques : (i)

L'indice basé sur les traits est efficace pour indiquer le niveau en matière organique, matières en suspension et nutriments ;

ii (ii)

L'indice basé sur les traits présente l’avantage, par rapport à celui basé sur les espèces,

de

nécessiter

moins

de

connaissances

taxonomiques

pour

l’identification, d’être plus rapide à mettre en œuvre tout en étant plus sensible à la pollution ; (iii)

L’indice basé sur les OTUs fournies par le séquençage moléculaire haut-débit d’amplicons permet de ne pas utiliser de bibliothèques de référence de barcodes ; ces dernières étant souvent incomplètes ;

(iv)

L'effet du seuil de similarité entre séquences (SST) utilisé pour regrouper cellesci en OTUs a été testé pour des valeurs allant de 80 à 99%, ce seuil étant un proxy de la résolution taxonomique;

(v)

Le pouvoir de discrimination, l'efficacité et le nombre de paramètres environnementaux prédictibles par l’indice OTU augmentent avec le SST, atteignant un plateau à 91%. Cette étude révèle la nécessité de faire un compromis entre pouvoir de discrimination et stabilité de l’évaluation.

(vi)

L'indice basé sur les OTUs fournit une évaluation écologique précise, économique et rapide pour laquelle aucune connaissance taxonomique n’est requise.

Enfin, nous discutons ces résultats et montrons que les méthodes basées sur les traits et sur le metabarcoding peuvent améliorer les outils d'évaluation de la qualité écologique et qu’elles ouvrent la voie à de nouvelles perspectives pour un outil d'évaluation holistique, robuste et précis basé sur le phytobenthos.

iii

Rivers and lakes are exposed to severe pollution and alteration from anthropogenic sources. The impairment of our waters endangers not only our freshwater supply but also the organisms inhabiting them and the entire ecosystem. Several methods were developed over time to monitor water bodies and evaluate their ecological quality in order to propose solutions to reduce anthropogenic impacts. Today, the European Water Framework Directive (WFD) is an international legislative framework for the water bodies of the European Union (EU) within which member states develop their national biomonitoring strategies. The quality value of water bodies is assessed using different groups of organisms: the biological quality elements from which one is the phytobenthos. Diatoms are a group of microalgae largely diversified and are dominant in the phytobenthic communities of rivers, thus used as a proxy to the entire phytobenthos. Several features make diatom efficient bioindicators: rapid response to environmental changes at community level and ease of sampling and analysis. A majority of EU countries imported already existing diatom indices to meet the requirements of the WFD. However, these indices carry some drawbacks: lack of holistic approach, supposed Gaussian response of species along pollution gradients, rare species neglected, need of experts for a precise identification of species, sensitivity to misidentifications, quite long analysis time under microscope. This thesis provides an overview of existing methods in diatom-based ecological quality assessment, presenting their potential evolutions together with their advantages and drawbacks. First, the trait-based approaches and ecological classifications of phytobenthos are reviewed in the framework of bioassessment and as an alternative method. Then, two studies were carried out on data from the biomonitoring network of the rivers of the new French department, Mayotte Island, to develop alternative or complementary ecological quality indices. The first study provided a test of a trait-based approach, while the second tested a molecular-based approach, the taxonomy-free DNA metabarcoding with direct use of OTUs (Operational Taxonomical Unit). Both approaches are opening interesting perspectives for the drawbacks of classical indices: (i)

The trait-based index was efficient to indicate organic matter, suspended solids and nutrient levels;

(ii)

The trait-based index had several advantages compared to the species-based one: easily recognisable traits, less taxonomic knowledge is required, faster to implement while being more sensitive to pollution;

iv (iii)

The taxonomy-free index based on solely OTUs provided by highthroughput molecular sequencing of amplicons is a way to avoid the use of reference barcoding libraries which are often incomplete;

(iv)

The effect of the sequence similarity threshold (SST) used to group sequences in OTUs has been tested for values from 80 to 99%, this threshold being a proxy of taxonomical resolution;

(v)

The discrimination power, efficiency and the number of predictable environmental parameters increase with the SST, reaching a plateau at 91%. There is an important trade-off between discrimination power and stability of quality assessment.

(vi)

The OTU-based index provides a precise ecological assessment, which is time- and cost-effective, and does not require taxonomical knowledge.

Finally, we discussed these results and show that both trait- and the molecular-based methods can improve tools for ecological quality assessment, and these methods also open the door towards new perspectives for a holistic, robust and precise evaluation tool based on the phytobenthos.

v

vi

vii

viii

ix

SCIENTIFIC PUBLICATIONS TAPOLCZAI K., VASSELON V., BOUCHEZ A., STENGER-KOVÁCS C., PADISÁK J., RIMET F. (under review) Taxonomy-free DNA biomonitoring for rivers: How to choose the sequence similarity threshold? Molecular Ecology Resources VASSELON V., BOUCHEZ A., RIMET F., JACQUET S., TROBAJO R., CORNIQUEL M., TAPOLCZAI K., DOMAIZON I. (under review) Avoiding quantification bias in metabarcoding: application of a cell biovolume correction factor in diatom molecular biomonitoring. Methods in Ecology and Evolution RIMET F., ABARCA N., BOUCHEZ A., KUSBER W-H., JAHN R., KAHLERT M., KECK F., KELLY MG., MANN DG., PIUZ A., TROBAJO R., TAPOLCZAI K., VASSELON V., ZIMMERMANN J. (under press) The potential of high throughput sequencing (HTS) of natural samples as a source of primary taxonomic information for reference libraries of diatom barcodes. Fottea TAPOLCZAI K., BOUCHEZ A., STENGER-KOVÁCS C., PADISÁK J., RIMET F. (2017) Taxonomy- or trait-based ecological assessment for tropical rivers? Case study on benthic diatoms in Mayotte island (France, Indian Ocean). Science of the Total Environment 607-608: 1293-1303 VASSELON V., RIMET F., TAPOLCZAI K., BOUCHEZ A. (2017) Assessing ecological status with diatoms DNA metabarcoding: Scaling-up on a WFD monitoring network (Mayotte island, France). Ecological Indicators 82: 1-12 KECK F., VASSELON V., TAPOLCZAI K., RIMET F., BOUCHEZ A. (2017) Freshwater biomonitoring in the Information Age. Frontiers in Ecology and the Environment 15(5): 266-274 TAPOLCZAI K., BOUCHEZ A., STENGER-KOVÁCS C., PADISÁK J., RIMET F. (2016) Trait-based ecological classifications for benthic diatoms: review and perspectives. Hydrobiologia 776: 1-17.

SCIENTIFIC COMMUNICATIONS BOUCHEZ A., KERMARREC L., REYJOL Y., TAPOLCZAI K., VASSELON V., RIMET F. (2017) On the way to implementation of ecogenomic indices for river biomonitoring: A French progress report for diatoms. 10th Symposium for European Freshwater Sciences (SEFS 10), Olomouc (Czech Republic), 2-7 July (communication). TAPOLCZAI K., BOUCHEZ A., STENGER-KOVÁCS C., PADISÁK J., VASSELON V., RIMET F. (2017) Diatom-based ecological assessment on the rivers of the tropical island, Mayotte (France) using different approaches. 10th Symposium for European Freshwater Sciences (SEFS 10), Olomouc (Czech Republic), 2-7 July (communication). BOUCHEZ A., FRANC A., BLANCHER P., CHAUMEIL P., FRIGERIO J.M., KECK F., KERMARREC L., MONNIER O., REYJOL Y., SALIN F., TAPOLCZAI K., VASSELON V., RIMET F. (2017) Diatom DNA metabarcoding & WFD: where are we? COST DNAqua-net kick-off meeting, Essen (Germany), 78 March (communication). RIMET F., VASSELON V., TAPOLCZAI K., BOUCHEZ A. (2017) R-Syst::diatom, a reference library for diatoms: overview, uses and perspectives. COST DNAqua-net kick-off meeting, Essen (Germany), 7-8 March (communication). VASSELON V., BOUCHEZ A., CORNIQUEL M., JACQUET S., RIMET F., TAPOLCZAI K., DOMAIZON I. (2017) Quantitative diatom metabarcoding: a correction factor inferred from cell biovolume. COST DNAqua-net kick-off meeting, Essen (Germany), 7-8 March (poster). VASSELON V., DOMAIZON I., RIMET F., TAPOLCZAI K., BOUCHEZ A. (2017) Optimization of diatom DNA metabarcoding: application to Mayotte streams monitoring network. COST DNAqua-net kick-off meeting, Essen (Germany), 7-8 March (communication).

x RIMET F., VASSELON V., KECK F., CHARDON C., TAPOLCZAI K., PIUZ A., BOUCHEZ A. (2016) Bases de référence de barcodes-ADN diatomées : comment les compléter rapidement à faible coût ? 35 ème Colloque de l’Association des Diatomistes de Langue Française (ADLaF), Belvaux (Luxembourg), 13 - 15 September, (communication). TAPOLCZAI K., BOUCHEZ A., VASSELON V., KECK F., STENGER-KOVÁCS C., PADISÁK J., RIMET F. (2016) L’évaluation de la qualité des cours d’eau de Mayotte basé sur un indice classique et un indice trait. 35ème Colloque de l’Association des Diatomistes de Langue Française (ADLaF), Belvaux (Luxembourg), 13-15 September, (communication). BOUCHEZ A., CHARDON C., KECK F., RIMET F., TAPOLCZAI K., VASSELON V. (2016) Metabarcoding and High-Throughput Sequencing for assessing river ecological quality with diatom indices at the scale of a regular monitoring network. Society for Freshwater Science 2016 annual meeting, Sacramento (USA), 22-26 May, (communication). RIMET F., VASSELON V., KECK F., CHARDON C., TAPOLCZAI K., PIUZ A., BOUCHEZ A. (2016) Diatom DNA-barcoding databases: how to fill them quickly at low cost? 10th Central European Diatom Meeting, Budapest (Hungary), 20-23 April (communication). TAPOLCZAI, K., BOUCHEZ A., VASSELON V., KECK F., STENGER-KOVÁCS, C., PADISÁK J., RIMET F., (2016) Species- and trait-based quality evaluation methods for the rivers of Mayotte (France, Southeast Africa). 10th Central European Diatom Meeting, Budapest (Hungary), 20-23 April (communication). BOUCHEZ A., RIMET F., CHAUMEIL P., FRIGERIO J.M., KECK F., TAPOLCZAI K., VASSELON V., FRANC A. (2015) Potentiel du metabarcoding et de la phylogénie des diatomées pour la bioindication 34 ème Colloque de l’Association des Diatomistes de Langue Française (ADLaF), Bordeaux (France), 7-10 September (communication). TAPOLCZAI K., BOUCHEZ A., RIMET F., STENGER-KOVÁCS, C., PADISÁK J. (2015) Les premières étapes sur l'analyse de cours d'eaux de Mayotte. 34ème Colloque de l’Association des Diatomistes de Langue Française (ADLaF), Bordeaux (France), 7-10 September (communication). TAPOLCZAI K., BOUCHEZ A., RIMET F., STENGER-KOVÁCS, C., PADISÁK J. (2015) Creating new diatom-based evaluation metrics for a universal and facilitated river biomonitoring – first steps. 9th Use of Algae for Monitoring RIvers and comparable habitats (9th UAMRIch) and International Workshop on Benthic Algae Taxonomy (InBAT), Trento (Italy), 15-19 June (poster). RIMET F., CHAUMEIL P., FRIGERIO J.M., TAPOLCZAI K., KECK F., VASSELON V., FRANC A., BOUCHEZ A. (2015) Potential of diatom metabarcoding and phylogeny for ecological assessment. 9th Use of Algae for Monitoring RIvers and comparable habitats (9th UAMRIch) and International Workshop on Benthic Algae Taxonomy (InBAT), Trento (Italy), 15-19 June (communication).

SCIENTIFIC REPORTS BOUCHEZ A., RIMET F., MONTUELLE B., TAPOLCZAI K., VASSELON V., FRANC A., CHAUMEIL P., FRIGERIO J.M., SALIN F., MARY N., USSEGLIO-POLATERA P. (2016) Développement d’outils de bio-indication « phytobenthos » et « macroinvertébrés benthiques » pour les eaux de surface continentales de Mayotte. Rapport INRA-ONEMA, Partenariat 2015, Fonctionnement des écosystèmes aquatiques et changements globaux – Action n°10, 102 pp. BOUCHEZ A., RIMET F., MONTUELLE B., TAPOLCZAI K., VASSELON V., FRANC A., CHAUMEIL P., FRIGERIO J.M., SALIN F. (2014) Développement d’outils de bio-indication « phytobenthos » et « macroinvertébrés benthiques » pour les cours d’eau de Mayotte. Rapport INRA-ONEMA, Partenariat 2014, Fonctionnement des écosystèmes aquatiques et changements globaux – Action n°10 : Développement d’outils de bio-indication « phytobenthos » et « macroinvertébrés benthiques » pour les cours d’eau de Mayotte. 47 pp.

xi

Global map of threat to human water security and biodiversity on a scale from 0 (no threat) to 1 (most severe threat). Threats are more important in highly populated and industrialised regions (China, India, Europe, United States) (figure from VÖRÖSMARTY et al., 2010). .......................................................................... 4 The river basin districts of Europe (2012 — direction générale de l’environnement). ............... 6 Simplified scheme of ecological quality assessment. EQR is calculated based on each BQEs for a particular site, then thw overall status is given using the „one-out-all-out” principle (figure by K. Tapolczai). .............................................................................................................................................................................. 9 Examples for the variety of chloroplast number and morphology among diatoms. Pinnularia spp. usually have two big plate-like chloroplasts lying along the girdle side but only one can be observed from girdle view (a).Two plate-like chloroplasts can be observed too, in Navicula spp from valve view (b). Diatoma mesodon with several chloroplasts per cell (c). Stephanodiscus neoastraea with several discoid chloroplasts (d) (Photos by K. Tapolczai (a-b) & F. Rimet (c-d)............................................................................................................ 12 Scanning electron microscope (SEM) photos of the centric Stephanodiscus alpinus Hustedt (a — valve view) and the pennate Rhopalodia hirundiniformis O. Müller (b — girdle view, sample from Mayotte). S. alpinus have radial symmetry, striae organised radially, several pores can be also observed in the annulus but less organised. Siliceous spines are present all along the edge of the valve with which the cells are organised into chain colonies. R. hirundiniformis have a strong dorsiventral symmetry so one can observe the valve surface with dense striae pattern in girdle view. The three girdle bands of the epitheca overlapping the hipotheca can be observed too (photos from MANN et al., 2016). .............................................................................................. 12 Diatom life cycle. During the asexual reproduction, the thecae of the mother cell became the epitheca in the daughter cells that leads to a decrease of average cell sie in the population. After a size limit, cells go through a sexual reproduction during which an auxospore is formed, which will become a vegetative cell with the original cell size (figure by K. Tapolczai). ............................................................................................... 14 Variety of diatom colonies. Zig-zag colony of Diatoma sp (a): cells are attached by mucilage pads produced at different pole of the cells. Encyonema sp. cells are located in a mucous tube inside which cells can moce (b). Tabellaria flocculosa (Roth) Kützing colony with star-like and zig-zag patterns (c). Fragilaria crotonensis Kitton forms long ribbon colonies to reduce sinking rate (d) (photos by K Tapolczai and F. Rimet) ................................................................................................................................................................................................ 16 Calculation of indicator and sensitivity value of species. Weighted average (wa) and weighted standard deviation (wSD) is calculated from the unimodal distribution of speces’ abundance values along an environmental gradient. These data for each species give the database for indices that will be later used when indice is calculated for a site. ............................................................................................................................................ 24 The process of ecological quality evaluation via a classical autecological diatom index. (1) Sampling of the biofilm using a toothbrush and preserving samples in ethanol. (2) Laboratory preparation. (3) Microscopic identification and counting. (4) Floristic list with species’ taxonomic names and their relative abundance. (6) Calculation of the diatom index using the abundance values (4) and the indicator and sensitivity values of species from the index’s database (5). ............................................................................................................ 25 The process of ecological quality evaluation using diatom traits. (1) Sampling of the biofilm using a toothbrush and preserving samples in ethanol. (2) Laboratory preparation. (3) Microscopic identification and counting. (4) Floristic list with species’ taxonomic names and their relative abundance. (5)Assignment several traits with trait classes to taxa. (6) Traits’ abundance list obtainded from the floristic list and the taxato-traits database. This trait abundance list can then be used to find significant relations with environmental data that can be later used in the development of a multimetric index. ................................................................... 28 The process of ecological quality evaluation via metabarcoding. (1) Sampling of the biofilm using a toothbrush and preserving samples in ethanol. (2) DNA extraction. (3) DNA amplification – PCR. (4) Sequencing step using HTS platform. (5) List of DNA barcodes after a bioinformatics process. (6) Assignment of taxonomic names to DNA barcodes via a reference database. (7) Creation of floristic list using the barcode list and the reference database. ......................................................................................................................................... 29

xii

The process of ecological quality evaluation with OTUs. (1) Sampling of the biofilm using a toothbrush and preserving samples in ethanol. (2) DNA extraction. (3) DNA amplification – PCR. (4) Sequencing step using HTS platform. (5) List of DNA barcodes after a bioinformatics process. (6) Creation of OTU abundance lists avoiding the assignment of OTUs to taxonomic names. ..................................................... 30 Number of papers related to the trait-based and ecological classification approach for benthic diatoms. Data is from Web of Science, September 2017. The searched keywords were the following: ‘‘diatom(s)’’ or ‘‘phytobenthos’’ or ‘‘benthic alga(e)’’ in the title and ‘‘river(s)’’ or ‘‘stream(s)’’ in the topic, additionally with one of the following in the topic: ‘‘guild(s)’’, ‘‘functional group(s)’’, ‘‘adaptive strategy(ies)’’, ‘‘life(-)-form(s)’’, ‘‘growth(-)form(s)’’, ‘‘trait(s)’’, ‘‘life-strategy(ies)’’. Updated version of Fig. 1 in Tapolczai et al. (2016). ................................................................................................................................................................................... 50 Conceptual framework of defining ecological groups; definition of functional groups (FGs) are based on a species-trait database using statistical methods or expert knowledge. Datasets of FGs and environmental parameters are used to define the ecology of FGs via multivariate statistical methods.............. 67 Conceptual framework of defining ecological groups; environmental data are used to define habitat types either with statistical methods or expert knowledge. Each habitat types possess dominant species with adaptive traits. Then, an interpretation of the trait–environment relation is required .................................. 68 Location of Mayotte and the river sampling sites. White, grey and black dots represent REF (reference), RCS (Le Réseau de Contrôle de Surveillance – Regular monitoring network), and POLL (polluted) sites, respectively. ................................................................................................................................................................ 81 Results of CCA analysis with all physico-chemical parameters. Distribution of sampling sites of the three networks (A); RCS (circles), REF (squares), and POLL (triangles), and the relative contribution of environmental factors (B). Parameters used to describe the two gradients, nutrient (bold), and organic/turbidity (bold-italic) gradient. .......................................................................................................................................................... 85 CCAs showing the distribution of samples of the three networks; REF (squares), RCS (circles), POLL (triangles) and the relative contribution of environmental factors of the nutrient- (NO3− – nitrate, DIN/DIP – dissolved inorganic nitrogen-to-dissolved inorganic phosphorus ratio, NH4+ – ammonium, NO2− – nitrite, TP – total phosphorus, PO43 − – phosphate, A), and the organic pollution gradient (turb – turbidity, SS – suspended solids, TOC – total organic carbon, DOC – dissolved organic carbon, B). ............................... 85 Pearson's correlation of the taxonomy-based quality sub-indices with the nutrient (A) and the organic pollution (B) gradients, practically the site locations along the 1st axes of the CCA analyses (Idx.M_nutr p < 0.05, r = − 0.91 with a CI95 = (− 0.98, − 0.86) and Idx.M_org p < 0.05, r = − 0.32, with CI95 = (− 0.74, 0.23), respectively) of the test dataset. Without the two outliers, the correlation coefficients are r = − 0.88 and r = − 0.77, respectively. ..................................................................................................................................................... 87 Pearson's correlation between the trait-based sub-indices with the nutrient (A) and the organic/turbidity (B) gradients, practically the site locations along the 1 st axes of the CCA analyses (Idx.M_nutrtrait p < 0.05, r = − 0.47, CI95 = (− 0.83, − 0.10) and Idx.M_orgtrait p < 0.05, r = − 0.54, CI95 = (− 0.71, − 0.34), respectively). Without the outliers, the correlations are stronger, r = − 0.80 and r = − 0.63, respectively. .......................................................................................................................................................................... 89 Pearson's correlation between the classical, taxonomy-based (Idx.M) and trait-based (Idx.Mtrait) indices (p 150,000 people. Although a system for the disposal and treatment of communal sewage and for municipal waste exists, most households are not connected or do not use it. Thus, the wastewaters are often released directly into the streams along with communal waste. Moreover, the streams are often contaminated with washing powder from laundry containing 4A zeolite as detergent and released from the intensive washing-by-hand activity by the women in the rivers. Industrial activity is not present in the island, but there is agricultural activity on many small fields of local farmers. A regular monitoring network (RCS — Réseau de Contrôle de Surveillance) is present for the streams of Mayotte since 2008. During this thesis, in order to cover a wider pollution gradient range, two additional monitoring networks were set up: a reference network from 2013 to 2015 and a polluted network from 2014 to 2015. In the section Material and methods of the self-edited version of two articles (Chapter 3-4), the sampling

- 32 -

sites, sampling method, sampling preparation, data availability and analyses are described in details. Photos of typical pollutions in the rivers of Mayotte are presented in Appendix A.

- 33 -

AFNOR., 2016. – NF T90 354 - Qualité de l’eau - Échantillonnage, traitement et analyse de diatomées benthiques en cours d’eau et canaux., 1‑79 p. AGOSTINHO A. A., THOMAZ S. M. & GOMES L. C., 2004. – Threats for biodiversity in the floodplain of the Upper Paraná River: effects of hydrological regulation by dams. International Journal of Ecohydrology & Hydrobiology, 4 (3) : 267–280. ANDERSEN R. A., 2004. – Biology and systematics of heterokont and haptophyte algae. American Journal of Botany, 91 (10) : 1508–1522. ANDERSON S., 1994. – Area and Endemism. The Quarterly Review of Biology, 69 (4) : 451‑471 doi : 10.1086/418743. APOTHELOZ-PERRET-GENTIL L., CORDONIER A., STRAUB F., ISELI J., ESLING P. & PAWLOWSKI J., 2017. – Taxonomy-free molecular diatom index for high-throughput eDNA biomonitoring. Molecular Ecology Resources, doi : 10.1111/1755-0998.12668. ARMBRUST E. V., 2009. – The life of diatoms in the world’s oceans. Nature, 459 (7244) : 185‑192 doi : 10.1038/nature08057. BAILEY R. C., NORRIS R. H. & REYNOLDSON T. B., 2001. – Taxonomic resolution of benthic macroinvertebrate communities in bioassessments. Journal of the North American Benthological Society, 20 (2) : 280–286. BAIRD D. J. & HAJIBABAEI M., 2012. – Biomonitoring 2.0: a new paradigm in ecosystem assessment made possible by next-generation DNA sequencing. Molecular ecology, 21 (8) : 2039‑2044. B-BERES V., LUKACS Á., TÖRÖK P., KOKAI Z., NOVAK Z., T-KRASZNAI E., TOTHMERESZ B. & BACSI I., 2016. – Combined eco-morphological functional groups are reliable indicators of colonisation processes of benthic diatom assemblages in a lowland stream. Ecological Indicators, 64 : 31‑38 doi : 10.1016/j.ecolind.2015.12.031. B-BERES V., TÖRÖK P., KOKAI Z., KRASZNAI E. T., TOTHMERESZ B. & BACSI I., 2014. – Ecological diatom guilds are useful but not sensitive enough as indicators of extremely changing water regimes. Hydrobiologia, 738 (1) : 191–204 doi : 10.1007/s10750-014-1929-y. BENNETT E. M., CARPENTER S. R. & CARACO N. F., 2001. – Human Impact on Erodable Phosphorus and Eutrophication: A Global PerspectiveIncreasing accumulation of phosphorus in soil threatens rivers, lakes, and coastal oceans with eutrophication. BioScience, 51 (3) : 227‑234 doi : 10.1641/00063568(2001)051[0227:HIOEPA]2.0.CO;2. BESSE-LOTOTSKAYA A., VERDONSCHOT P. F. M., COSTE M. & VAN DE VIJVER B., 2011. – Evaluation of European diatom trophic indices. Ecological Indicators, 11 (2) : 456‑467 doi : 10.1016/j.ecolind.2010.06.017. BIGGS B. J. F. & THOMSEN H. A., 1995. – Disturbance of Stream Periphyton by Perturbations in Shear Stress: Time to Structural Failure and Differences in Community Resistance1. Journal of Phycology, 31 (2) : 233‑ 241 doi : 10.1111/j.0022-3646.1995.00233.x. BLONDEL J., 2003. – Guilds or functional groups: does it matter? Oikos, 100 (2) : 223‑231 doi : 10.1034/j.16000706.2003.12152.x. BOETIUS A., ALBRECHT S., BAKKER K., BIENHOLD C., FELDEN J., FERNANDEZ-MENDEZ M., HENDRICKS S., KATLEIN C., LALANDE C., KRUMPEN T., NICOLAUS M., PEEKEN I., RABE B., ROGACHEVA A., RYBAKOVA E., SOMAVILLA R., WENZHÖFER F. & PARTY R. P. A.-3-S. S., 2013. – Export of Algal Biomass from the Melting Arctic Sea Ice. Science, 339 (6126) : 1430‑1432 doi : 10.1126/science.1231346. BORJA Á., 2005. – The European water framework directive: A challenge for nearshore, coastal and continental shelf research. Continental Shelf Research, 25 (14) : 1768‑1783 doi : 10.1016/j.csr.2005.05.004. BORJA Á. & RODRIGUEZ J. G., 2010. – Problems associated with the ‘one-out, all-out’ principle, when using multiple ecosystem components in assessing the ecological status of marine waters. Marine Pollution Bulletin, 60 (8) : 1143‑1146 doi : 10.1016/j.marpolbul.2010.06.026.

- 34 -

BOYD C. & GRADMANN D., 2002. – Impact of osmolytes on buoyancy of marine phytoplankton. Marine Biology, 141 (4) : 605‑618 doi : 10.1007/s00227-002-0872-z. BROOK A. & JOHNSON L., 2002. – Order Zygnematales. The Freshwater Algal Flora of the British Isles. An Identification Guide to Freshwater and Terrestrial Algae. Cambridge University Press, Cambridge, : 479‑ 593. BUTCHER R. W., 1947. – Studies in the Ecology of Rivers: VII. The Algae of Organically Enriched Waters. The Journal of Ecology, 35 (1/2) : 186 doi : 10.2307/2256507. CAMIZ S., ALTIERI A. & MANES F., 2008. – Pollution Bioindicators: Statistical Analysis of a Case Study. Water, Air, and Soil Pollution, 194 (1‑4) : 111‑139 doi : 10.1007/s11270-008-9702-3. CANTE M. T., DE STEFANO M., GIUDICE F., TOTTI C. & RUSSO G. F., 2008. – Marine gastropod shells as selective microenvironments for diatom communities. . CARPENTER S. R., STANLEY E. H. & VANDER ZANDEN M. J., 2011. – State of the World’s Freshwater Ecosystems: Physical, Chemical, and Biological Changes. Annual Review of Environment and Resources, 36 (1) : 75‑99 doi : 10.1146/annurev-environ-021810-094524. CEN., 2003. – Water Quality – Guidance Standard for the Routine Sampling and Pretreatment of Benthic Diatoms from Rivers. Geneva: Comité European de Normalisation. –––., 2016. – Water quality - Technical report for the routine sampling of benthic diatoms from rivers and lakes adapted for metabarcoding analyses., 7 p. CHONOVA T., KECK F., LABANOWSKI J., MONTUELLE B., RIMET F. & BOUCHEZ A., 2016. – Separate treatment of hospital and urban wastewaters: A real scale comparison of effluents and their effect on microbial communities. Science of The Total Environment, 542 : 965‑975 doi : 10.1016/j.scitotenv.2015.10.161. CHUA L. H. C., LO E. Y. M., SHUY E. B. & TAN S. B. K., 2009. – Nutrients and suspended solids in dry weather and storm flows from a tropical catchment with various proportions of rural and urban land use. Journal of Environmental Management, 90 (11) : 3635‑3642 doi : 10.1016/j.jenvman.2009.07.001. COHN S. A. & WEITZELL R. E., 1996. – Ecological considerations of diatom cell motility. I. Characterization of motility and adhesion in four diatom species. Journal of Phycology, 32 (6) : 928‑939 doi : 10.1111/j.00223646.1996.00928.x. CORING E., SCHNEIDER S., HAMM A. & HOFMANN G., 1999. – Durchgehendes Trophiesystem auf der Grundlage der Trophieindikation mit Kieselalgen. DVWK Materialien, 6 (1999) : 1‑219. COSTE M., 1982. – Étude des méthodes biologiques d’appréciation quantitative de la qualité des eaux. Rapport Cemagref QE Lyon-AF Bassin Rhône Méditerranée Corse, . COSTE M., BOUTRY S., TISON-ROSEBERY J. & DELMAS F., 2009. – Improvements of the Biological Diatom Index (BDI): Description and efficiency of the new version (BDI-2006). Ecological Indicators, 9 (4) : 621 ‑650 doi : 10.1016/j.ecolind.2008.06.003. COX E. J., 1996. – Identification of Freshwater Diatoms from Live Material. 1 edition., London ; New York : Springer, 168 p. CROSSETTI L. O., STENGER-KOVACS C. & PADISAK J., 2013. – Coherence of phytoplankton and attached diatom-based ecological status assessment in Lake Balaton. Hydrobiologia, 716 (1) : 87‑101 doi : 10.1007/s10750-013-1547-0. CUMMINS K. W., 1974. – Structure and Function of Stream Ecosystems. BioScience, 24 (11) : 631‑641 doi : 10.2307/1296676. DARWIN C., 1859. – On the origin of species. J. Murray, London, . DELL’UOMO A., 1996. – Assessment of water quality of an Apennine river as a pilot study for diatom-based monitoring of Italian watercourses. Use of Algae for Monitoring Rivers, : 65‑72. DELL’UOMO A. & TORRISI M., 2011. – The Eutrophication/Pollution Index-Diatom based (EPI-D) and three new related indices for monitoring rivers: The case study of the river Potenza (the Marches, Italy). Plant Biosystems - An International Journal Dealing with all Aspects of Plant Biology, 145 (2) : 331‑341 doi : 10.1080/11263504.2011.569347. DU RIETZ G. E., 1931. – Life-forms of terrestrial flowering plants. Almqvist & Wiksell.

- 35 -

DUELLI P. & OBRIST M. K., 2003. – Biodiversity indicators: the choice of values and measures. Agriculture, Ecosystems & Environment, 98 (1) : 87‑98 doi : 10.1016/S0167-8809(03)00072-0. DURUIBE J. O., OGWUEGBU M. O. C., EGWURUGWU J. N. & OTHERS., 2007. – Heavy metal pollution and human biotoxic effects. International Journal of Physical Sciences, 2 (5) : 112–118. EDGAR R. C., 2013. – UPARSE: highly accurate OTU sequences from microbial amplicon reads. Nature Methods, 10 (10) : 996‑998 doi : 10.1038/nmeth.2604. ELBRECHT V. & LEESE F., 2015. – Can DNA-Based Ecosystem Assessments Quantify Species Abundance? Testing Primer Bias and Biomass—Sequence Relationships with an Innovative Metabarcoding Protocol. PLOS ONE, 10 (7) : e0130324 doi : 10.1371/journal.pone.0130324. EUROPEAN COMMISSION., 2000. – Directive 2000/60/EC of the European Parliament and of the Council of 23rd October 2000 establishing a framework for Community action in the field of water policy. Official Journal of the European Communities, 327 : 1‑72. FADILA K., HOURIA D., RACHID R. & REDA D. M., 2009. – Cellular response of pollution biondicator model (Ramalina farinacea) following treatment with fertilizer (NPKs). Am-Eurasian J Toxicol Sci, 1 : 69–73. FINLAY B. J., 2002. – Global Dispersal of Free-Living Microbial Eukaryote Species. Science, 296 (5570) : 1061 ‑1063 doi : 10.1126/science.1070710. FJERDINSTAD E., 1950. – The Microflora of the River Mølleaa: With Special Reference to the Relation of the Benthal Algae to Pollution. Gleerupska univ.-bokh. GARCIA M., 2003. – Psammic Diatoms in Southern of Brazil. Journal of Coastal Research, : 363‑368 doi : 10.2307/40928783. GELENCSER A., KOVATS N., TUROCZI B., ROSTASI Á., HOFFER A., IMRE K., NYIRO-KOSA I., CSAKBERENYIMALASICS D., TOTH Á., CZITROVSZKY A., NAGY A., NAGY S., ÁCS A., KOVACS A., FERINCZ Á., HARTYANI Z. & POSFAI M., 2011. – The Red Mud Accident in Ajka (Hungary): Characterization and Potential Health Effects of Fugitive Dust. Environmental Science & Technology, 45 (4) : 1608‑1615 doi : 10.1021/es104005r. GERHARDT A., 2002. – Bioindicator species and their use in biomonitoring. Environmental monitoring, 1 : 77– 123. GOEL P. K., 2006. – Water Pollution: Causes, Effects and Control. New Age International, 19 p. GOMEZ N. & LICURSI M., 2001. – The Pampean Diatom Index (IDP) for assessment of rivers and streams in Argentina. Aquatic Ecology, 35 (2) : 173‑181. GRIMM N. B. & FISHER S. G., 1989. – Stability of Periphyton and Macroinvertebrates to Disturbance by Flash Floods in a Desert Stream. Journal of the North American Benthological Society, 8 (4) : 293‑307 doi : 10.2307/1467493. GUIRY M. D., 2012. – How many species of algae are there? Journal of Phycology, 48 (5) : 1057‑1063 doi : 10.1111/j.1529-8817.2012.01222.x. HAINES-YOUNG R. & POTSCHIN M., 2010. – The links between biodiversity, ecosystem services and human well-being. Ecosystem Ecology: a new synthesis, : 110–139. HART D. D. & FINELLI AND C. M., 1999. – Physical-Biological Coupling in Streams: The Pervasive Effects of Flow on Benthic Organisms. Annual Review of Ecology and Systematics, 30 (1) : 363‑395 doi : 10.1146/annurev.ecolsys.30.1.363. HEBERT P. D., CYWINSKA A., BALL S. L. & OTHERS., 2003. – Biological identifications through DNA barcodes. Proceedings of the Royal Society of London B: Biological Sciences, 270 (1512) : 313–321. HERING D., BORJA A., CARSTENSEN J., CARVALHO L., ELLIOTT M., FELD C. K., HEISKANEN A.-S., JOHNSON R. K., MOE J., PONT D., SOLHEIM A. L. & DE BUND W. VAN., 2010. – The European Water Framework Directive at the age of 10: A critical review of the achievements with recommendations for the future. Science of The Total Environment, 408 (19) : 4007‑4019 doi : 10.1016/j.scitotenv.2010.05.031. HOLT E. A. & MILLER S. W., 2011. – Bioindicators: using organisms to measure environmental impacts. Nature Education Knowledge, 3 (10) : 8. HUMBOLDT A. VON., 1806. – Ideen zu einer Physiognomik der Gewächse. 1. Auflage., Tübingen : Cotta.

- 36 -

HÜRLIMANN J. & NIEDERHAUSER P., 2002. – Méthode d’analyse et d’appréciation des cours d’eau en Suisse, Diatomées, niveau R région. Office fédéral de l’environnement, des forêts et du paysage OFEFP, Berne, Version provisoire du, 2 (2002) : 111. HUSE S. M., WELCH D. M., MORRISON H. G. & SOGIN M. L., 2010. – Ironing out the wrinkles in the rare biosphere through improved OTU clustering. Environmental Microbiology, 12 (7) : 1889‑1898 doi : 10.1111/j.1462-2920.2010.02193.x. HUTCHINSON G. E., 1957. – Concluding Remarks. Cold Spring Harbor Symposia on Quantitative Biology, 22 : 415‑427 doi : 10.1101/SQB.1957.022.01.039. INAGAKI Y., DACKS J. B., DOOLITTLE W. F., WATANABE K. I. & OHAMA T., 2000. – Evolutionary relationship between dinoflagellates bearing obligate diatom endosymbionts: insight into tertiary endosymbiosis. International Journal of Systematic and Evolutionary Microbiology, 50 (6) : 2075–2081. INSEE., 2016. – Estimation de la population au 1er janvier par région, département, sexe et âge de 1975 à 2015. http://www.insee.fr/fr/themes/detail.asp?reg_id=99&ref_id=estim-pop Consulté le 8/6/2016. JAIN A., SINGH B. N., SINGH S. P., SINGH H. B. & SINGH S., 2010. – Exploring Biodiversity as Bioindicators for Water Pollution. Dans : Banaras Hindu University, National Conference on Biodiversity, Developpement and Poverty Alleviation. . JARVIE H. P., NEAL C. & WITHERS P. J. A., 2006. – Sewage-effluent phosphorus: A greater risk to river eutrophication than agricultural phosphorus? Science of The Total Environment, 360 (1) : 246‑253 doi : 10.1016/j.scitotenv.2005.08.038. JEAN PRYGIEL P. C., 2002. – Determination of the biological diatom index (IBD NF T 90–354): results of an intercomparison exercise. Journal of Applied Phycology, 14 (1) : 27‑39 doi : 10.1023/A:1015277207328. JONSSON B. G. & JONSELL M., 1999. – Exploring potential biodiversity indicators in boreal forests. Biodiversity and conservation, 8 (10) : 1417–1433. JULIUS M. L. & THERIOT E. C., 2010. – The diatoms: a primer. Dans : The diatoms: Applications for the environmental and earth sciences. Cambridge University Press, p. 8‑22. KAHLERT M., ALBERT R.-L., ANTTILA E.-L., BENGTSSON R., BIGLER C., ESKOLA T., GÄLMAN V., GOTTSCHALK S., HERLITZ E., JARLMAN A., KASPEROVICIENE J., KOKOCINSKI M., LUUP H., MIETTINEN J., PAUNKSNYTE I. ET AL., 2009. – Harmonization is more important than experience—results of the first Nordic–Baltic diatom intercalibration exercise 2007 (stream monitoring). Journal of Applied Phycology, 21 (4) : 471‑482 doi : 10.1007/s10811-008-9394-5. KAHLERT M., KELLY M., ALBERT R.-L., ALMEIDA S. F. P., BESTA T., BLANCO S., COSTE M., DENYS L., ECTOR L., FRANKOVA M., HLUBIKOVA D., IVANOV P., KENNEDY B., MARVAN P., MERTENS A. ET AL., 2012. – Identification versus counting protocols as sources of uncertainty in diatom-based ecological status assessments. Hydrobiologia, 695 (1) : 109‑124 doi : 10.1007/s10750-012-1115-z. KATI V., DEVILLERS P., DUFRENE M., LEGAKIS A., VOKOU D. & LEBRUN P., 2004. – Testing the Value of Six Taxonomic Groups as Biodiversity Indicators at a Local Scale. Conservation Biology, 18 (3) : 667‑675 doi : 10.1111/j.1523-1739.2004.00465.x. KATTGE J., DIAZ S., LAVOREL S., PRENTICE I. C., LEADLEY P., BÖNISCH G., GARNIER E., WESTOBY M., REICH P. B., WRIGHT I. J., CORNELISSEN J. H. C., VIOLLE C., HARRISON S. P., VAN BODEGOM P. M., REICHSTEIN M. ET AL., 2011. – TRY - a global database of plant traits: TRY - A GLOBAL DATABASE OF PLANT TRAITS. Global Change Biology, 17 (9) : 2905‑2935 doi : 10.1111/j.1365-2486.2011.02451.x. KECK F., VASSELON V., TAPOLCZAI K., RIMET F. & BOUCHEZ A., 2017. – Freshwater biomonitoring in the Information Age. Frontiers in Ecology and the Environment, doi : 10.1002/fee.1490. KELLY M., 2006. – A comparison of diatoms with other phytobenthos as indicators of ecological status in streams in northern England. Dans : Proceedings of the 18th International Diatom Symposium. Bristol : Biopress. KELLY M., 2013. – Data rich, information poor? Phytobenthos assessment and the Water Framework Directive. European Journal of Phycology, 48 (4) : 437–450 doi : 10.1080/09670262.2013.852694. KELLY M. G., CAZAUBON A., CORING E., DELL’UOMO A., ECTOR L., GOLDSMITH B., GUASCH H., HÜRLIMANN J., JARLMAN A., KAWECKA B., KWANDRANS J., LAUGASTE R., LINDSTRØM E.-A., LEITAO M.,

- 37 -

MARVAN P. ET AL., 1998. – Recommendations for the routine sampling of diatoms for water quality assessments in Europe. Journal of Applied Phycology, 10 (2) : 215–224. KELLY M. G., GOMEZ-RODRIGUEZ C., KAHLERT M., ALMEIDA S. F. P., BENNETT C., BOTTIN M., DELMAS F., DESCY J.-P., DÖRFLINGER G., KENNEDY B., MARVAN P., OPATRILOVA L., PARDO I., PFISTER P., ROSEBERY J., SCHNEIDER S. & VILBASTE S., 2012. – Establishing expectations for pan-European diatom based ecological status assessments. Ecological Indicators, 20 : 177‑186 doi : 10.1016/j.ecolind.2012.02.020. KELLY M. G., KING L., JONES R. I., BARKER P. A. & JAMIESON B. J., 2008. – Validation of diatoms as proxies for phytobenthos when assessing ecological status in lakes. Hydrobiologia, 610 (1) : 125‑129 doi : 10.1007/s10750-008-9427-8. KELLY M. G. & WHITTON B. A., 1995. – The trophic diatom index: a new index for monitoring eutrophication in rivers. Journal of Applied Phycology, 7 (4) : 433–444. KELLY M. G. & ZGRUNDO A., 2013. – Potential for cross-contamination of benthic diatom samples when using toothbrushes. Diatom Research, 28 (4) : 359‑363 doi : 10.1080/0269249X.2013.806959. KELLY M., KING L. & NI CHATHAIN B., 2009. – THE CONCEPTUAL BASIS OF ECOLOGICAL-STATUS ASSESSMENTS USING DIATOMS. Biology & Environment: Proceedings of the Royal Irish Academy, 109 (3) : 175‑189 doi : 10.3318/BIOE.2009.109.3.175. KERMARREC L., BOUCHEZ A., RIMET F. & HUMBERT J.-F., 2013a. – First evidence of the existence of semicryptic species and of a phylogeographic structure in the Gomphonema parvulum (Kützing) Kützing complex (Bacillariophyta). Protist, 164 (5) : 686‑705. KERMARREC L., FRANC A., RIMET F., CHAUMEIL P., HUMBERT J. F. & BOUCHEZ A., 2013b. – Next-generation sequencing to inventory taxonomic diversity in eukaryotic communities: a test for freshwater diatoms. Molecular Ecology Resources, 13 (4) : 607‑619 doi : 10.1111/1755-0998.12105. KHATRI N. & TYAGI S., 2015. – Influences of natural and anthropogenic factors on surface and groundwater quality in rural and urban areas. Frontiers in Life Science, 8 (1) : 23‑39 doi : 10.1080/21553769.2014.933716. KOCIOLEK J. P., HAMSHER S. E., KULIKOVSKIY M. & BRAMBURGER A. J., 2017. – Are there species flocks in freshwater diatoms? A review of past reports and a look to the future. Hydrobiologia, 792 (1) : 17‑35 doi : 10.1007/s10750-016-3075-1. KOCIOLEK J. P. & WILLIAMS D. M., 2015. – How to define a diatom genus? Notes on the creation and recognition of taxa, and a call for revisionary studies of diatoms. Acta Botanica Croatica, 74 (2) doi : 10.1515/botcro-2015-0018. KOELMANS A. A., BESSELING E. & SHIM W. J., 2015. – Nanoplastics in the Aquatic Environment. Critical Review. Dans : Marine Anthropogenic Litter. Springer, Cham, p. 325‑340. doi : 10.1007/978-3-319-165103_12. KOLKWITZ R., 1908. – Ökologie der pflanzlichen Saprobien. Berrichten der Deutschen Botanischen Gesellschaft, 26 : 505‑519. KOOISTRA W. H. C. F., GERSONDE R., MEDLIN L. K. & MANN D. G., 2007. – The Origin and Evolution of the Diatoms: Their Adaptation to a Planktonic Existence. Dans : Evolution of Primary Producers in the Sea. Elsevier, p. 207‑249. doi : 10.1016/B978-012370518-1/50012-6. KOOISTRA W. H. C. F. & MEDLIN L. K., 1996. – Evolution of the Diatoms (Bacillariophyta). Molecular Phylogenetics and Evolution, 6 (3) : 391‑407 doi : 10.1006/mpev.1996.0088. LAMPERT W. & SOMMER U., 2007. – Limnoecology: The Ecology of Lakes and Streams. 2 edition., Oxford : New York : Oxford University Press, 336 p. LANGE K., TOWNSEND C. R. & MATTHAEI C. D., 2016. – A trait-based framework for stream algal communities. Ecology and Evolution, 6 (1) : 23‑36 doi : 10.1002/ece3.1822. LAW R. J., ELLIOTT J. A. & THACKERAY S. J., 2014. – Do functional or morphological classifications explain stream phytobenthic community assemblages? Diatom Research, 29 (4) : 309–324 doi : 10.1080/0269249X.2014.889037. LEE J. J., 2006. – Algal symbiosis in larger foraminifera. Symbiosis (Rehovot), 42 (2) : 63‑75.

- 38 -

LEE R. E., 2008. – Phycology. Cambridge University Press, 534 p. LEESE F., ALTERMATT F., BOUCHEZ A., EKREM T., HERING D., MEISSNER K., MERGEN P., PAWLOWSKI J., PIGGOTT J., RIMET F., STEINKE D., TABERLET P., WEIGAND A., ABARENKOV K., BEJA P. ET AL., 2016. – DNAqua-Net: Developing new genetic tools for bioassessment and monitoring of aquatic ecosystems in Europe. Research Ideas and Outcomes, 2 : e11321 doi : 10.3897/rio.2.e11321. LENG M. L., LEOVEY E. M. K. & ZUBKOFF P. L., 1995. – Agrochemical Environmental Fate State of the Art. CRC Press, 428 p. LENOIR A. & COSTE M., 1996. – Development of a practical diatom index of overall water quality applicable to the French National Water Board Network. Use of Algae for monitoring rivers, : 29‑43. LI W. C., 2014. – Occurrence, sources, and fate of pharmaceuticals in aquatic environment and soil. Environmental Pollution, 187 : 193‑201 doi : 10.1016/j.envpol.2014.01.015. LITCHMAN E. & KLAUSMEIER C. A., 2008. – Trait-Based Community Ecology of Phytoplankton. Annual Review of Ecology, Evolution, and Systematics, 39 (1) : 615‑639 doi : 10.1146/annurev.ecolsys.39.110707.173549. LITCHMAN E., KLAUSMEIER C. A. & YOSHIYAMA K., 2009. – Contrasting size evolution in marine and freshwater diatoms. Proceedings of the National Academy of Sciences, 106 (8) : 2665‑2670 doi : 10.1073/pnas.0810891106. LOMAN N. J., MISRA R. V., DALLMAN T. J., CONSTANTINIDOU C., GHARBIA S. E., WAIN J. & PALLEN M. J., 2012. – Performance comparison of benchtop high-throughput sequencing platforms. Nature Biotechnology, 30 (5) : 434‑439 doi : 10.1038/nbt.2198. MANN D. G., 1999. – The species concept in diatoms. Phycologia, 38 (6) : 437‑495 doi : 10.2216/i0031-888438-6-437.1. –––., 2011. – Size and Sex. Dans : Seckbach J, Kociolek P. The Diatom World. Dordrecht : Springer Netherlands, p. 145‑166. doi : 10.1007/978-94-007-1327-7_6. MANN D. G., CRAWFORD R. M. & ROUND F. E., 2016. – Bacillariophyta. Dans : Archibald JM, Simpson AGB, Slamovits CH, Margulis L, Melkonian M, Chapman DJ, Corliss JO. Handbook of the Protists. Cham : Springer International Publishing, p. 1‑62. doi : 10.1007/978-3-319-32669-6_29-1. MANN D. G., SATO S., TROBAJO R., VANORMELINGEN P. & SOUFFREAU C., 2010. – DNA barcoding for species identification and discovery in diatoms. Cryptogamie. Algologie, 31 (4) : 557–577. MANN D. G. & VANORMELINGEN P., 2013. – An Inordinate Fondness? The Number, Distributions, and Origins of Diatom Species. Journal of Eukaryotic Microbiology, 60 (4) : 414‑420 doi : 10.1111/jeu.12047. MARTIN W. & KOWALLIK K. V., 1999. – Annotated English translation of Mereschkowsky’s 1905 paper ‘Über Natur und Ursprung der Chromatophoren im Pflanzenreiche’. European Journal of Phycology, 34 (3) : 287 ‑295. MC COY M. A. & BALZER I., 2001. – Algal symbiosis in flatworms. Dans : Symbiosis. Springer, p. 559–574. MEDLIN L., KOOISTRA W. H. C. F., GERSONDE R., SIMS P. A. & WELLBROCK U., 1997. – Is the origin of diatoms related to the end-Permian mass extinction? Dans : EPIC3Nova Hedwigia, 65, pp. 1-11. , p. 1‑11. MERESCHKOWSKY C., 1905. – Über Natur und Ursprung der Chromatophoren im Pflanzenreiche. Biol Centralbl, 25, : 593‑604. MEYBECK M., 2003. – Global analysis of river systems: from Earth system controls to Anthropocene syndromes. Philosophical Transactions of the Royal Society B: Biological Sciences, 358 (1440) : 1935‑1955 doi : 10.1098/rstb.2003.1379. MOROZOVA G. S., 2005. – A review of Holocene avulsions of the Tigris and Euphrates rivers and possible effects on the evolution of civilizations in lower Mesopotamia. Geoarchaeology, 20 (4) : 401‑423 doi : 10.1002/gea.20057. MUCHUWETI M., BIRKETT J. W., CHINYANGA E., ZVAUYA R., SCRIMSHAW M. D. & LESTER J. N., 2006. – Heavy metal content of vegetables irrigated with mixtures of wastewater and sewage sludge in Zimbabwe: Implications for human health. Agriculture, Ecosystems & Environment, 112 (1) : 41‑48 doi : 10.1016/j.agee.2005.04.028.

- 39 -

MULHOLLAND P. J. & ROSEMOND A. D., 1992. – Periphyton Response to Longitudinal Nutrient Depletion in a Woodland Stream: Evidence of Upstream-Downstream Linkage. Journal of the North American Benthological Society, 11 (4) : 405‑419 doi : 10.2307/1467561. NIEMI G. J., AXLER R. P., BRADY V., BRAZNER J., BROWN T., CIBOROWSKI J. H., DANZ N., HANOWSKI J. M., HOLLENHORST T., HOWE R. & OTHERS., 2015. – Environmental Indicators for the US. Great Lakes Coastal Region. OKI T. & KANAE S., 2006. – Global hydrological cycles and world water resources. science, 313 (5790) : 1068– 1072. OPPENHEIM D. R., 1990. – A Preliminary Study of Benthic Diatoms in Contrasting Lake Environments. Dans : Antarctic Ecosystems. Springer, Berlin, Heidelberg, p. 91‑99. doi : 10.1007/978-3-642-84074-6_9. PADISAK J., BORICS G., GRIGORSZKY I. & SOROCZKI-PINTER É., 2006. – Use of Phytoplankton Assemblages for Monitoring Ecological Status of Lakes within the Water Framework Directive: The Assemblage Index. Hydrobiologia, 553 (1) : 1–14 doi : 10.1007/s10750-005-1393-9. PADISAK J., SCHEFFLER W., SIPOS C., KASPRZAK P., KOSCHEL R. & KRIENITZ L., 2003a. – Spatial and temporal pattern of development and decline of the spring diatom populations in Lake Stechlin in 1999. Archiv für Hydrobiologie Beiheft Advances In Limnology, 58 : 135–155. PADISAK J., SOROCZKI-PINTER É. & REZNER Z., 2003b. – Sinking properties of some phytoplankton shapes and the relation of form resistance to morphological diversity of plankton–an experimental study. Hydrobiologia, 500 (1‑3) : 243–257. PARMAR T. K., RAWTANI D. & AGRAWAL Y. K., 2016. – Bioindicators: the natural indicator of environmental pollution. Frontiers in Life Science, 9 (2) : 110–118. PASSY S. I., 2007. – Diatom ecological guilds display distinct and predictable behavior along nutrient and disturbance gradients in running waters. Aquatic Botany, 86 (2) : 171–178 doi : 10.1016/j.aquabot.2006.09.018. PASSY S. I., BODE R. W., CARLSON D. M. & NOVAK M. A., 2004. – Comparative Environmental Assessment in the Studies of Benthic Diatom, Macroinvertebrate, and Fish Communities. International Review of Hydrobiology, 89 (2) : 121‑138 doi : 10.1002/iroh.200310721. PASSY S. I. & LARSON C. A., 2011. – Succession in Stream Biofilms is an Environmentally Driven Gradient of Stress Tolerance. Microbial Ecology, 62 (2) : 414‑424 doi : 10.1007/s00248-011-9879-7. PATERSON D. M., 1989. – Short-term changes in the erodibility of intertidal cohesive sediments related to the migratory behavior of epipelic diatoms. Limnology and Oceanography, 34 (1) : 223–234. PATIN N. V., KUNIN V., LIDSTRÖM U. & ASHBY M. N., 2013. – Effects of OTU Clustering and PCR Artifacts on Microbial Diversity Estimates. Microbial Ecology, 65 (3) : 709‑719 doi : 10.1007/s00248-012-0145-4. PINHO P., BERGAMINI A., CARVALHO P., BRANQUINHO C., STOFER S., SCHEIDEGGER C. & MAGUAS C., 2012. – Lichen functional groups as ecological indicators of the effects of land-use in Mediterranean ecosystems. Ecological Indicators, 15 (1) : 36‑42 doi : 10.1016/j.ecolind.2011.09.022. POTAPOVA M. G., CHARLES D. F., PONADER K. C. & WINTER D. M., 2004. – Quantifying species indicator values for trophic diatom indices: a comparison of approaches. Hydrobiologia, 517 (1) : 25–41. POULIN M., UNDERWOOD G. J. C. & MICHEL C., 2014. – Sub-ice colonial Melosira arctica in Arctic first-year ice. Diatom Research, 29 (2) : 213‑221 doi : 10.1080/0269249X.2013.877085. RANDALL B. M. & RANDALL R. M., 1984. – Algae on Jackass Penguins (Spheniscus demersus). The Auk, 101 (4) : 880‑882 doi : 10.2307/4086917. RAUNKIAER C., 1934. – The life forms of plants and statistical plant geography; being the collected papers of C. Raunkiaer. The life forms of plants and statistical plant geography; being the collected papers of C. Raunkiaer., . REYNOLDS C. S., 2006. – Ecology of Phytoplankton. First., New York : Cambridge University Press, 535 p. RIMET F., BALLORAIN K., CARPENTIER A., RIVERA S. F., VASSELON V., WETZEL C. E., ECTOR L. & BOUCHEZ A., in prep. – DNA metabarcoding and microscopic analyses of sea turtles biofilms: complementary to understand turtles behaviors. .

- 40 -

RIMET F. & BOUCHEZ A., 2012. – Life-forms, cell-sizes and ecological guilds of diatoms in European rivers. Knowledge and Management of Aquatic Ecosystems, (406) : 01 doi : 10.1051/kmae/2012018. RIMET F., BOUCHEZ A. & TAPOLCZAI K., 2016. – Spatial heterogeneity of littoral benthic diatoms in a large lake: monitoring implications. Hydrobiologia, 771 (1) : 179‑193 doi : 10.1007/s10750-015-2629-y. RIMET F., ECTOR L., CAUCHIE H.-M. & HOFFMANN L., 2009. – Changes in diatom-dominated biofilms during simulated improvements in water quality: implications for diatom-based monitoring in rivers. European Journal of Phycology, 44 (4) : 567‑577 doi : 10.1080/09670260903198521. RIVERA S. F., VASSELON V., JACQUET S., BOUCHEZ A., ARIZTEGUI D. & RIMET F., in press. – Metabarcoding of lake benthic diatoms: from structure assemblages to ecological assessment. Hydrobiologia, . ROBINSON N. J., MAJEWSKA R., LAZO-WASEM E. A., NEL R., PALADINO F. V., ROJAS L., ZARDUS J. D. & PINOU T., 2016. – Epibiotic Diatoms Are Universally Present on All Sea Turtle Species. PLoS ONE, 11 (6) : e0157011 doi : https://doi.org/10.1371/journal.pone.0157011. ROOT R. B., 1967. – The Niche Exploitation Pattern of the Blue-Gray Gnatcatcher. Ecological Monographs, 37 (4) : 317‑350 doi : 10.2307/1942327. ROTT E., 1991. – Methodological aspects and perspectives in the use of periphyton for monitoring and protecting rivers. Use of algae for monitoring rivers, : 9‑16. ROTT E., HOFMANN G., PALL K., PFISTER P. & PIPP E., 1997. – Indikationslisten für Aufwuchsalgen, Teil 1: Saprobielle Indikation (Indication lists for periphytic algae. Part 1: Saprobic indication). Bundesministerium für Land-und Forstwirtschaft (Federal Ministry of Agriculture and Forestry), Wien, . ROTT E., PIPP E., PFISTER P., VAN DAM H., ORTLER K., PALL K. & BINDER N., 1999. – Indikationslisten für Aufwuchsalgen in österreichischen Fliessgewässern. Teil 2: Trophie-indikation sowie geochemische Präferenz; taxonomische und toxikologische Anmerkungen. Bundesministerium für Land-und Forstwirtschaft, Wasserwirtschaftskataster, Wien, . ROUND F. E., CRAWFORD R. M. & MANN D. G., 1990. – The Diatoms: Biology & Morphology of the Genera. Cambridge University Press, 768 p. ROVIRA L., TROBAJO R., SATO S., IBAÑEZ C. & MANN D. G., 2015. – Genetic and Physiological Diversity in the Diatom Nitzschia inconspicua. The Journal of Eukaryotic Microbiology, 62 (6) : 815‑832 doi : 10.1111/jeu.12240. RUCK E. C. & THERIOT E. C., 2011. – Origin and Evolution of the Canal Raphe System in Diatoms. Protist, 162 (5) : 723‑737 doi : 10.1016/j.protis.2011.02.003. RUMEAU A. & COSTE M., 1988. – Initiation à la systématique des diatomées d’eau douce. Pour l’utilisation pratique d’un indice diatomique générique. Bulletin Français de la Pêche et de la Pisciculture, (309) : 1‑69 doi : 10.1051/kmae:1988009. SALMASO N., NASELLI-FLORES L. & PADISAK J., 2015. – Functional classifications and their application in phytoplankton ecology. Freshwater Biology, 60 (4) : 603‑619 doi : 10.1111/fwb.12520. SARTHOU G., TIMMERMANS K. R., BLAIN S. & TREGUER P., 2005. – Growth physiology and fate of diatoms in the ocean: a review. Journal of Sea Research, 53 (1) : 25‑42 doi : 10.1016/j.seares.2004.01.007. SATO S., BEAKES G., IDEI M., NAGUMO T. & MANN D. G., 2011. – Novel Sex Cells and Evidence for Sex Pheromones in Diatoms. PLoS ONE, 6 (10) : e26923 doi : 10.1371/journal.pone.0026923. SCHNEIDER S. C., LAWNICZAK A. E., PICINSKA-FALTYNOWICZ J. & SZOSZKIEWICZ K., 2012. – Do macrophytes, diatoms and non-diatom benthic algae give redundant information? Results from a case study in Poland. Limnologica - Ecology and Management of Inland Waters, 42 (3) : 204‑211 doi : 10.1016/j.limno.2011.12.001. SHOKRALLA S., SPALL J. L., GIBSON J. F. & HAJIBABAEI M., 2012. – Next-generation sequencing technologies for environmental DNA research. Molecular Ecology, 21 (8) : 1794‑1805 doi : 10.1111/j.1365294X.2012.05538.x. SIMS P. A., MANN D. G. & MEDLIN L. K., 2006. – Evolution of the diatoms: insights from fossil, biological and molecular data. Phycologia, 45 (4) : 361‑402 doi : 10.2216/05-22.1. SINGH G., 1971. – The Indus Valley Culture. Archaeology and Physical Anthropology in Oceania, 6 (2) : 177‑ 189 doi : 10.1002/j.1834-4453.1971.tb00134.x.

- 41 -

SMEETS E. & WETERINGS R., 1999. – Environmental indicators: Typology and overview. European Environment Agency Copenhagen. SMOL J. P. & STOERMER E. F., 2010. – The Diatoms: Applications for the Environmental and Earth Sciences. Cambridge University Press, 687 p. SOININEN J., JAMONEAU A., ROSEBERY J. & PASSY S. I., 2016. – Global patterns and drivers of species and trait composition in diatoms. Global Ecology and Biogeography, 25 (8) : 940‑950 doi : 10.1111/geb.12452. SOMMER U., GLIWICZ Z. M., LAMPERT W. & DUNCAN A., 1986. – The PEG-model of seasonal succession of planktonic events in fresh waters. Arch. Hydrobiol, 106 (4) : 433–471. SPRINGE G., SANDIN L., BRIEDE A. & SKUJA A., 2006. – Biological quality metrics: their variability and appropriate scale for assessing streams. Hydrobiologia, 566 (1) : 153‑172 doi : 10.1007/s10750-006-0099y. STANLEY J.-D., KROM M. D., CLIFF R. A. & WOODWARD J. C., 2003. – Short contribution: Nile flow failure at the end of the Old Kingdom, Egypt: Strontium isotopic and petrologic evidence. Geoarchaeology, 18 (3) : 395‑402 doi : 10.1002/gea.10065. STEELE T. D., KRALISCH S., KLEIN D. & FLÜGEL W. A., 2008. – A comparative evaluation of selected aspects of the EU’s Water Framework Directive versus the US Clean Water Act. Dans : 13th IWRA World Water Congress. . STEINBERG C. & SCHIEFELE S., 1987. – Biological indication of trophy and pollution of running waters. Wasser Abwasser-Forsch, 21 : 227‑234. STEVENSON R. J., 1997. – Scale-Dependent Determinants and Consequences of Benthic Algal Heterogeneity. Journal of the North American Benthological Society, 16 (1) : 248‑262 doi : 10.2307/1468255. STEVENSON R. J., PAN Y. & VAN DAM H., 2010. – Assessing environmental conditions in rivers and streams with diatoms. Dans : The Diatoms: Applications for the Environmental and Earth Sciences. Cambridge : Cambridge University Press, p. 57‑85. STODDARD J. L., LARSEN D. P., HAWKINS C. P., JOHNSON R. K. & NORRIS R. H., 2006. – Setting Expectations for the Ecological Condition of Streams: The Concept of Reference Condition. Ecological Applications, 16 (4) : 1267‑1276 doi : 10.1890/1051-0761(2006)016[1267:SEFTEC]2.0.CO;2. SZABO B., PADISAK J., SELMECZY G. B., KRIENITZ L., CASPER P. & STENGER-KOVACS C., 2017. – Spatial and temporal patterns of benthic diatom flora in Lake Stechlin, Germany. TURKISH JOURNAL OF BOTANY, 41 : 211‑222 doi : 10.3906/bot-1606-41. TAO S., LI B. G., HE X. C., LIU W. X. & SHI Z., 2007. – Spatial and temporal variations and possible sources of dichlorodiphenyltrichloroethane (DDT) and its metabolites in rivers in Tianjin, China. Chemosphere, 68 (1) : 10‑16 doi : 10.1016/j.chemosphere.2006.12.082. TAPOLCZAI K., BOUCHEZ A., STENGER-KOVACS C., PADISAK J. & RIMET F., 2016. – Trait-based ecological classifications for benthic algae: review and perspectives. Hydrobiologia, 776 (1) : 1‑17 doi : 10.1007/s10750-016-2736-4. TAPOLCZAI K., VASSELON V., BOUCHEZ A., STENGER-KOVACS C., PADISAK J. & RIMET F., under review. – Taxonomy-free DNA biomonitoring for rivers: How to choose the sequence similarity threshold? Molecular Ecology Resources, . TSAKIRIS G., 2015. – The Status of the European Waters in 2015: a Review. Environmental Processes, 2 (3) : 543‑557 doi : 10.1007/s40710-015-0079-1. UEHLINGER U., KAWECKA B. & ROBINSON C. T., 2003. – Effects of experimental floods on periphyton and stream metabolism below a high dam in the Swiss Alps (River Spöl). Aquatic Sciences, 65 (3) : 199‑209 doi : 10.1007/s00027-003-0664-7. ULEN B., BECHMANN M., FÖLSTER J., JARVIE H. P. & TUNNEY H., 2007. – Agriculture as a phosphorus source for eutrophication in the north-west European countries, Norway, Sweden, United Kingdom and Ireland: a review. Soil Use and Management, 23 (s1) : 5–15. U.S. CONGRESS., 1972. – Federal Water Pollution Control Act Amendments. .

- 42 -

VANORMELINGEN P., VERLEYEN E. & VYVERMAN W., 2008. – The diversity and distribution of diatoms: from cosmopolitanism to narrow endemism. Biodiversity and Conservation, 17 (2) : 393‑405 doi : 10.1007/s10531-007-9257-4. VASS M., REVAY Á., KUCSERKA T., HUBAI K., ÜVEGES V., KOVACS K. & PADISAK J., 2013. – Aquatic hyphomycetes as survivors and/or first colonizers after a red sludge disaster in the Torna stream, Hungary. International Review of Hydrobiology, 98 (4) : 217‑224 doi : 10.1002/iroh.201301540. VASSELON V., DOMAIZON I., RIMET F., KAHLERT M. & BOUCHEZ A., 2017a. – Application of high-throughput sequencing (HTS) metabarcoding to diatom biomonitoring: Do DNA extraction methods matter? Freshwater Science, 36 (1) : 162‑177 doi : 10.1086/690649. VASSELON V., RIMET F., TAPOLCZAI K. & BOUCHEZ A., 2017b. – Assessing ecological status with diatoms DNA metabarcoding: Scaling-up on a WFD monitoring network (Mayotte island, France). Ecological Indicators, 82 : 1‑12 doi : 10.1016/j.ecolind.2017.06.024. VENKATACHALAPATHY R. & KARTHIKEYAN P., 2015. – Application of Diatom-Based Indices for Monitoring Environmental Quality of Riverine Ecosystems: A Review. Dans : Ramkumar M, Kumaraswamy K, Mohanraj R. Environmental Management of River Basin Ecosystems. Cham : Springer International Publishing, p. 593‑619. doi : 10.1007/978-3-319-13425-3_28. VIOLLE C., NAVAS M.-L., VILE D., KAZAKOU E., FORTUNEL C., HUMMEL I. & GARNIER E., 2007. – Let the concept of trait be functional! Oikos, 116 (5) : 882‑892 doi : 10.1111/j.0030-1299.2007.15559.x. VÖRÖSMARTY C. J., MCINTYRE P. B., GESSNER M. O., DUDGEON D., PRUSEVICH A., GREEN P., GLIDDEN S., BUNN S. E., SULLIVAN C. A., LIERMANN C. R. & DAVIES P. M., 2010. – Global threats to human water security and river biodiversity. Nature, 467 (7315) : 555‑561 doi : 10.1038/nature09440. VOULVOULIS N., ARPON K. D. & GIAKOUMIS T., 2017. – The EU Water Framework Directive: From great expectations to problems with implementation. Science of The Total Environment, 575 : 358‑366 doi : 10.1016/j.scitotenv.2016.09.228. WARMING E., 1908. – Om planterigets livsformer. G.E.C. Gad, 100 p. WATANABE T., ASAI K. & HOUKI A., 1986. – Numerical estimation to organic pollution of flowing water by using the epilithic diatom assemblage ----- diatom assemblage index ( DAIpo ) ----. Science of The Total Environment, 55 : 209‑218 doi : 10.1016/0048-9697(86)90180-4. WERNER P., ADLER S. & DRE?LER M., 2016. – Effects of counting variances on water quality assessments: implications from four benthic diatom samples, each counted by 40 diatomists. Journal of Applied Phycology, 28 (4) : 2287‑2297 doi : 10.1007/s10811-015-0760-9. YU D. W., JI Y., EMERSON B. C., WANG X., YE C., YANG C. & DING Z., 2012. – Biodiversity soup: metabarcoding of arthropods for rapid biodiversity assessment and biomonitoring: Biodiversity soup. Methods in Ecology and Evolution, 3 (4) : 613‑623 doi : 10.1111/j.2041-210X.2012.00198.x. ZELINKA M. & MARVAN P., 1961. – Zur präzisierung der biologischen klassifikation der reinheit flie\s sender gewässer. Arch. Hydrobiol, 57 (3) : 389–407. ZORBA M. A., JACOB P. G., AL-BLOUSHI A. & AL-NAFISI R., 1992. – Clams as pollution bioindicators in Kuwait’s marine environment: metal accumulation and depuration. Science of the total environment, 120 (3) : 185–204. 2003. – Bioindicators & biomonitors: principles, concepts, and applications. Amsterdam ; Boston : Elsevier, 997 p.

- 43 -

- 44 -

Chapter 2

This chapter is the self-edited version of the following article: Tapolczai, K., A. Bouchez, C. Stenger-Kovács, J. Padisák & F. Rimet, 2016. Trait-based ecological classifications for benthic algae: review and perspectives. Hydrobiologia 776: 117. doi: 10.1007/s10750-016-2736-4

- 46 -

Abstract A high number of species often represents a relevant redundancy in terms of ecological adaptation strategies. Collecting species to groups based on their functional adaptations can handle this redundancy and obtain the ‘‘real’’ functional complexity of ecosystems. Functional traits are proxies of adaptation strategies under particular environmental conditions, and a set of functional traits are interpreted as life-strategies. Organisms with life-strategies occupying a similar niche can be collected in ecological groups (functional group/guild). In this study, we review the latest traitbased approaches and existing attempts at functional classifications in phytobenthos studies. Advantages and shortcomings of these classifications are discussed with perspectives of their utility in ecological status assessment.

Keywords Benthic algae, Diatoms, Ecological groups, Functional groups, Guilds, Life-forms, Traits - 47 -

Since its introduction, the binomial nomenclature of Linnaeus (1758) has remained the basic classification system of species in biology. Thus, species are the basic units of the taxonomical hierarchy and, consequently, of ecological studies. Although the concept of species is well established (De Queiroz, 2007), delimitation criteria and methods remain under continuous development. Especially among simple organisms (e.g. algae), the classical morphology-based identification is under change into a phylogeny-based delimitation stimulated by the rapid development of molecular techniques. At present, their taxonomy is unstable and quickly changing. Estimations on the total number of algal species vary from 30,000 to 1 million species. Mann (1999) mentions several tens of thousands species for diatoms alone. Later this number dropped to around 10,000 species (Mann & Vanormelingen, 2013). This diversity and the wide geographical and environmental distribution make these organisms useful tools for ecological assessment. Quality assessment methods based on benthic algae (with a strong bias towards diatoms) rely on taxonomic units (species or genus). Autecological indices [e.g., BDI (Prygiel & Coste, 1998), IPS (Cemagref, 1982), TDI (Kelly & Whitton, 1995), PIT (Schneider & Lindstrøm, 2011), AIP (Schneider & Lindstrøm, 2009)] are the most common method for ecological quality assessment used in the Water Framework Directive (European Commission, 2000). However, these indices carry uncertainties. Those thousands of taxa included in the databases of the indices comprise a number of rare species with hardly definable ecological profiles (Rimet & Bouchez, 2012a). In addition, the different European indices use different ecological profiles for the same species (probably because their profiles were defined from different ecoregions with limited range of environmental variables); this practice reduces the robustness of the estimations, especially for species present with low abundance per site and low frequency of occurrence (Besse-Lototskaya et al., 2011). Since indices are developed specifically for a particular ecoregion, their use in another ecoregion should be carried out with caution. Several “unreliable” species have been identified, i.e. they indicate trophic state changes from oligotrophic to hypertrophic, depending on the index (Besse-Lototskaya et al., 2011). Both taxonomic misidentification and species with different ecological optima can result in false assessments. Additionally, species’ response to environmental parameters may depend on geographic or habitat-dependent distributions, resulting in different responses of the same species in different ecoregions. The rationale of such diatom indices have been often questioned (Kelly, 2013). As Kelly (2011) posed the question in a comment paper to Besse- 48 -

Lototskaya et al. (2011): “Do we need diatom trophic metrics in Europe?”. The question rose from the recognition that while the main debate between diatomists is about taxonomic issues, end-users of quality evaluation methods do not get clear answers for their emergent environmental problems. His other main point is that often factors other than trophic condition acts as an underlying factor for casual relationship; however, this may correlate with nutrients. At first sight, it does not change the result but can bear problems when one must give advice or solutions to end-users (Kelly, 2011). The theoretical advantage of diversity metrics compared to autoecological indices is that they quantify the impact of pressures; in practice, this consists of mainly nutrient enrichment (e.g. eutrophication, organic pollution) on the structure of the entire community. They were already used successfully to indicate organic pollution (Stevenson & Bahls, 2002) and stream order (Stenger-Kovács et al., 2013b). There are also promising results on new generation diversity metrics that are proved to be sensitive and precise indicators (e.g. trophic level or pH) (Stenger-Kovács et al., 2016). On the other hand, studies on such metrics often contradict theoretical predictions, resulting in weak correlations and unclear patterns (Blanco et al., 2012). It suggests a more complex mechanism than simple correlations of how pressures affect diatom composition: response often evolves non-monotonic stressor gradients (Stevenson, 2014). The fact that both autecological and diversity indices are based on taxonomic units (species or genus) involves technical and theoretical issues. Accurate species-level identification is not always insured since it requires high-level experts in diatom identification to follow the continuously changing taxonomy, and it is time consuming (Berthon et al., 2011; Kermarrec et al., 2014). These issues entail the problem of disharmony in identification accuracy: variation in both space (differences between labs) (Kahlert et al., 2009, 2012) and time (Straile et al., 2013). The structuring impact of different ecoregions sets another challenge (Rimet et al., 2007). It is a particularly difficult task for countries with oversea departments under the EU legal system (France, Spain, Portugal) to apply their evaluation system to these regions with highly different geographical location and climate regimes. These regions may have unique algae flora and environmental conditions that require specific and robust assessment metrics. Species-based classification carries further drawbacks. The role of a member in an ecosystem depends on the morphological, physiological properties where it belongs to in order to adapt and compete in a particular habitat. These traits may include phylogenetically close species as well as distant ones. Even different strains or ecotypes of the same species can possess different traits. A well-known example is the toxic

- 49 -

and non-toxic strains of cyanobacteria (Neilan et al., 1995). Regarding the selective factors in an environment and the possible adaptive strategies, one can see a high redundancy at the species level (Kelly, 2013). Ecological groups cluster species with similar adaptive strategies corresponding to the real compartments of an ecosystem to potentially simplify its complexity (Salmaso et al., 2015). This concept among benthic algae has been promising, and the number of studies in this field in the last decade has increased (Figure 2.1). The aim of this review is to provide an overview of trait-based approaches and the ecological group concept in studies of benthic algae. We show a critical analysis of the status of existing ecological classifications and present their advantages and drawbacks (Table 1), including identification criteria, trait response to environmental factors, and utility. We propose perspectives that could be envisaged to improve trait-based ecological classification and its application in biomonitoring and quality assessment.

Number of papers related to the trait-based and ecological classification approach for benthic diatoms. Data is from Web of Science, September 2017. The searched keywords were the following: ‘‘diatom(s)’’ or ‘‘phytobenthos’’ or ‘‘benthic alga(e)’’ in the title and ‘‘river(s)’’ or ‘‘stream(s)’’ in the topic, additionally with one of the following in the topic: ‘‘guild(s)’’, ‘‘functional group(s)’’, ‘‘adaptive strategy(ies)’’, ‘‘life(-)-form(s)’’, ‘‘growth(-)form(s)’’, ‘‘trait(s)’’, ‘‘life-strategy(ies)’’. Updated version of Fig. 1 in Tapolczai et al. (2016).

- 50 -

- 51 -

1

21

0

Morphological, functional

Morphology

Morphological, physiological, behavioural, lifehistory based

Morphological, functional

Guilds

Combined CSR strategy

Trait-based conceptual framework

Ecomorphological functional groups

20

3

Criteria

Functional groups

2

21 in 7 categories

3

1

Number Number of of traits groups

Diatoms

34

91

78

Diatoms+ non diatoms

Diatoms+ non diatoms

104

Artificial

Natural

Artificial

Natural

Number Substrate of taxa

Diatoms

Algae groups

Already existing functional classification for benthic algae discussed in this review

Lange et al. (2016)

Ecologically well-justified traits showing good results, but no ecological groups are defined Several sites in the Manuherikia River catchment area (New Zeland)

B-Béres et al. (2016)

Law et al. (2014)

Few new information considering the number of hypothetical groups Several points in Wyre and Loud tributaries (UK)

Highlights "hidden" correlations within one guild

Passy (2007)

All guilds show habitat indication (however later studies shows contradictions)

Several points in White Creek (USA) and Mesta River (Bulgaria)

Tócó stream (Hungary)

Reference Efficiency

Origin of data

Traits are the basic units of developing any kind of ecological classification of organisms. According to Violle et al.’s (2007) definition, traits are “any morphological, physiological or phenological measurable feature at the individual level”. This seems to be valid for all organisms regardless the study objects (e.g., animals, terrestrial plants, phytoplankton, diatoms). In the case of complex and physiognomically diverse organisms such as vascular plants, a large set of easily identifiable traits exists. Now, several extended databases are accessible for terrestrial plants including hundreds of traits measured by standardized methods that allow for comparative studies (Kattge et al., 2011). Regarding unicellular algae, applying trait-based approaches in the field of their ecological study has been a great challenge for scientists. Their simple structure, microscopic size, and potential observational difficulties hamper the identification of a large set of determinable traits with clearly associated ecological functions. However, trends show that this challenge can be accomplished. Phytoplankton studies already provide the knowledge of several categorized functional traits (e.g. morphological, physiological, behavioral, lifehistory) (Litchman & Klausmeier, 2008). Linking these traits to their appearance along environmental gradients was the basis of functional classification of phytoplankton (Margalef, 1978; Reynolds et al., 2002). The deficiency compared to plant functional groups is the lack of a global trait database with guidelines and standardised measurements (Litchman & Klausmeier, 2008). A comprehensive database of traits is missing for benthic algae as well, but the terminology is used to refer to both simple measurable features (e.g. biovolume, size-classes) and the more complex life-strategies (e.g. life-forms, guilds) (Virtanen et al., 2011; Laine et al., 2014). This trait-based approach is the basis of defining the so-called ecological groups (guilds or functional groups). The concept for phytobenthos has been under progress, and there are already some attempts for a possible complex ecological classification, as summarized in Table 1.

The most widely known ecological classification is the ecological guilds of Passy (2007). Practically, the term “guild” is used as the synonym of “functional group” that is historically

- 52 -

more preferred in plant biology than the former, which is preferred in animal biology (Blondel, 2003). “Guild” refers to a group of species that exploits resources in a similar way, resulting in stronger competition within the guilds than between them. In contrast, the base of the definition of functional groups is the similarity in the ecosystem functioning rather than in resource sharing. The criteria are more process oriented than structural. Since the way of resource utilising is more diverse and apparent in animals (i.e. a set of feeding strategies), the term “guild” became more common in animal studies (Blondel, 2003). Since such attempts for classifications in phytobenthos studies import concepts and methods from phytoplankton studies where “functional group” is the accepted definition, we suggest keeping this terminology or the use of the more neutral “ecological groups”. The guild concept of diatoms states that the great diversity of benthic diatom growth forms (i.e. life-forms) shows high redundancy (i.e. niche overlap) along the main structuring environmental pressures: nutrient availability and flow disturbance. Using this redundancy, species were classified into three ecological guilds with distinct features in the changing habitat. Unlike in the pelagic, in benthic habitats there is a steep vertical gradient of resources, i.e. nutrients and light within the biofilm characterized by canopy. The distance between individuals is spatially much closer than in the rather “dilute” phytoplankton. Passy’s classification can be regarded functional, including the way species attached to the substrate, thus how they cope with disturbance (e.g. flow velocity, grazing) and the way they utilise resources. Low-profile species positioned on the bottom layer of the biofilm attached strongly to the substrate with the whole valve surface, while the big, erected, or colonial highprofile species represent the canopy layer of the biofilm. Passy’s study on the guilds has been cited 150 times (Web of Science, 2017), indicating a strong interest on this topic. However, studies often show different results and interpretations of their findings that can be attributed to some deficiencies of clarity in the original guild classification. The theoretical background of Passy’s findings about the negative correlation between resources and low-profile dominance is that due to their vertical position, lowprofile species are exposed to resource limitation in a thick biofilm. Thick biofilm can develop under high nutrient values, and high-profile species have adaptive advantage to reach light and nutrients in the biofilm. The fact that in that case low-profile species are suppressed should not mean that they have an advantage under low-resource circumstances since their dominance under nutrient-poor conditions can be explained on several ways. First, if there is no difference in nutrient requirement between low- and high profile guild, the adaptive

- 53 -

strategy of low-profile species against flow velocity can favour their growth even under low flow velocity. Second, low-profile species are frequently small species, and this “linked” trait may mean more effective nutrient uptake due to higher surface ratio or faster growth rate. Passy is also inconsequent in this question. While she states that a low-profile guild is “likely to be resource-stressed but disturbance-free, i.e. it experiences resource limitations”, in the next statement, she states that this guild has “the ability to withstand resource limitation”. On the other hand, the “disturbance-stressed” high-profile guild is suppressed in high-disturbance habitats. Several studies tested the response of guilds to environmental parameters, mainly, nutrients and physical forces (i.e. water-flow) (Table 2.2.). Rimet et al. (2015) examined the seasonal guild succession in the littoral benthic diatom assemblage in Lake Geneva (FranceSwitzerland). Their explanation of dominance of low and motile species driven by nutrient availability and grazing pressure supports Passy’s (2007) concept and is consistent with other studies (Berthon et al., 2011). However, the suggested reason of high-profile dominance during the nutrient-limited period by their competitive ability is somewhat in contradiction with Passy. In one of the cases, the same adaptation (i.e. competitive advantage for nutrients in a dense biofilm) results in their dominance in nutrient-rich habitats (Passy, 2007); in other cases, adaptation results in their dominance in nutrient-poor habitats (Rimet et al., 2015). Leira et al. (2015) suggested that high-profile forms might have advantages under low irradiance level caused by sediment resuspension and suppress the a priori shade tolerant low-profile species. Additionally, they showed that even under low resources and light availability when the development of a three-dimensional biofilm was prevented, high-profile species dominated (Leira et al., 2015). Stenger-Kovács et al. (2013a) found an increasing trend in the abundance of low-profile guilds with the increasing irradiance due to seasonal change coupled with the removal of high-profile guilds due to floods. They argue that prostrate forms due to their vertical location in the biofilm utilise weaker irradiance better than highprofile species. The relative abundance of motile guild, however, correlated negatively with the irradiance. It is clear that a strong improvement is needed in defining how environmental factors affect diatoms in order to sort them into ecological groups. Regarding the resources, at least the separation between nutrients and light is essential. In a thick biofilm, both have gradients towards the same direction, but at larger scales, this is not the case. An interesting observation from Vilar et al. (2015) is that while low-profile species dominated low-nutrient, clear water, they were absent in an enclosure with artificially low turbidity. This is because low-profile species could colonise first in natural habitats after a strong disturbance event

- 54 -

due to their resistance to flushing. Their dominance was due to the mass effect from the predisturbance period and not to their adaptation to the new environment. The motility of the motile guild enables them to find the best place in the microhabitat to avoid disturbance, i.e. resistance against moderate water discharge (Lengyel et al., 2015b) or reach the best position to acquire nutrients. Passy’s use of the term disturbance is confusing: it is used to describe both the effect of water velocity and grazing. However, it would be welcome to make a distinction between stress and disturbance when one tries to classify ecological groups based on adaptive features. The term “disturbance-stressed” that is used in her study is meaningless and is not used elsewhere in the scientific literature. This lack of separation is unfortunate especially because the study addressed to draw an analogy between the guild classification and Grime’s (1974) CSR strategy classification, which clearly defines that stress restricts production via the shortages of resources (nutrients, temperature stress, light limitation, etc.). Disturbance affects organisms through events that cause damage to the vegetation (e.g. grazing, floods, wind). The same factor can act both as stress and as disturbance. As discussed by Borics et al. (2013), it is the temporal frequency that differentiates them. While stress is a continuously acting pressure of the physical environment limiting resource utilisation, growth rate, or reproduction of organisms (Grime, 1989), disturbance is considered as an unpredictable, stochastic event that interferes with the community development towards an ecological climax (Reynolds et al., 1993). Continuous nutrient limitation, high-velocity water flow, and grazing pressure all act as stress that enable the development of stress-adaptation strategies, while occasional disturbance events shift the community into an earlier successional phase, promoting recolonization. The remarkable difference in their effect on the community is that while stress decreases diversity, the effect of disturbance events on compositional diversity depends on its frequency and intensity, as formulated in the intermediate disturbance hypothesis (Hardin, 1960; Padisák, 1993; Lengyel et al., 2015a, b). Strongly stressed habitats represent ideal study sites for studying stress tolerance. For example, Central European saline lakes are characterised by high conductivity, high pH, and low light-availability due to inorganic turbidity, fluctuating water level, and high daily temperature variation. These extreme conditions showed correlation with the dominance of the motile guild (StengerKovács et al., 2014) that seems to be efficiently adapted to stressed environment, i.e. freemoving, shade-tolerance (Padisák, 2003). The adaptive advantage of low-profile species to high flow velocity circumstances is ecologically well founded. Passy (2007) found the strongest correlation between guilds and this factor, and subsequent studies confirmed it

- 55 -

(Mackay et al., 2012; Stenger-Kovács et al., 2013a; Tang et al., 2013). The same morphological trait, i.e. strong attachment, being adnate helps to avoid grazing (Passy, 2007; Gottschalk & Kahlert, 2012) and thus hampers the distinction of these two pressures on the guild composition.

Traits already used for ecological classification, and factors for which they were tested. Ticks designate already tested relations Categorized traits

Nutrients

Water flow

Light

Conductivity

Grazing

Morphology







Biovolume





 

 

Greatest axial length dimension





Attachment mechanism





Surface-to-volume ratio







Life-forms







Profile guilds

























 





Behavior Motility guilds Physiology Nitrogen fixation Life-history

1

- 56 -

 

Main reproductive techniques



Spore formation







  



  

 

Organic pollution



  





 





Another attempt to use the CSR classification (Grime, 1977; Reynolds, 2006) on benthic algae was made by Law et al. (2014). They used simple morphometric features to categorize benthic taxa: the surface area-to-volume ratio and the greatest axial linear dimension (GALD) of the cell. These features can be regarded as the proxy of adaptation to different resource levels; nutrients, light or against water flow (Table 2.2). Colonists (C) are favoured by higher level of nutrients and light, stress-tolerants (S) can withstand low level of nutrients, and ruderals (R) that can withstand low light level. The study combined these three categories with the life-forms used by Berthon et al. (2011), resulting in 21 variations. The use of lifeforms is ecologically well justified since they represent easily measurable morphological traits that are good proxy of adaptive strategies. Similarly to other concepts, this also originates from terrestrial plant studies (Humboldt, 1806; Raunkiaer, 1934; Gómez-Aparicio, 2009) and phytoplankton (Pianka, 1970; Margalef, 1978; Crossetti & Bicudo, 2008; Dunck et al., 2013). A coherent classification of life forms exists for diatoms (Round et al., 1990; Rimet & Bouchez, 2012b) that is based on their cellular structure (unicellular or colonial), attachment (e.g. not attached, adnate, attached by mucilage pad), and the type of aggregation (e.g. chain, ribbon-, arbuscular colonies). This classification uses easily determinable traits (from living sample) with adaptive meanings (flow-resistance, nutrient uptake). As an example, tubeforming diatoms appear to be effective indicators of low organic and trophic levels (Berthon et al., 2011). Other studies also confirm that species prone to tube-forming are found mainly in oligotrophic habitats (Rumeau & Coste, 1988; Leira et al., 2009). A similar relationship was shown for stalked diatoms (Berthon et al., 2011) with the interpretation of Pringle (1990) that these species are less adapted to uptake nutrients absorbed on the substratum but well adapted to exploit dissolved nutrients. This hypothesis was confirmed by an experimental study (Rimet et al., 2009). Although this classification involves diatoms only, recognition of simple life-forms would not involve special sample preparation; diatoms and non-diatoms could be classified together, since the interpretation of these traits, in this term, is not taxa specific, as shown by Law et al., (2014) and Lange et al. (2015). However, in the study of Law et al. (2014), life-forms alone did not give interpretable results tested against environmental factors that are, according to the authors, due to the potential of species utilising more than one life-form. Tests with the CSR classification showed that S-category species with low surface-to-volume ratios and short GALD were associated to eutrophic conditions that is

- 57 -

surprisingly the opposite of what is shown for phytoplankton (Reynolds, 1988). Colonists with higher surface area-to-volume ratios with short GALD were found in every treatment and ruderals with long GALD, and high surface-to-volume ratios dominated oligotrophic conditions due to their competitive abilities. It is notable that while importing such concepts from phytoplankton studies can be successful, the two communities are quite different; in plankton, the organisms are relatively far from each other and the interaction between them is much less important than in the phytobenthos, where it is more important. Additionally, in the phytobenthos, a steep vertical gradient of the environmental constraints is present. Such differences raise limits in such direct application of the CSR classification. A combination of CSR strategies and life-forms gave better results (Law et al., 2014). Under eutrophic conditions, R-category motile species dominated, while under oligotrophic conditions, the S-category colonial species were abundant. Although this combination of classifications theoretically results in 21 groups, the study showed that most of them can exist only theoretically, and only two groups could indicate environmental gradients.

The application of trait categories presented first by Litchman & Klausmeier for phytoplankton (2008) was applied for benthic algae by Lange et al. (2016). The base is a matrix, where several traits grouped in trait categories are paired with their adaptive advantage category (e.g. resource acquisition, resistance to disturbance, predator avoidance). The advantage of this system is that it is applied for all benthic algae, not only diatoms. Cell size is one of the easiest measurable features with several ecological adaptive meanings proven by several former studies (Table 2.2). Large, erected cells are more sensitive to physical disturbances (e.g. flush, floods). Another example is that high surface-to-volume ratios related to small size promote efficient nutrient uptake (Reynolds, 2006). Cattaneo (1987) and Morin et al. (2001) showed a significant positive correlation between cell size and nutrient concentration on environmental data, and a similar relation was shown by an experimental study (Carrick & Lowe, 1989). A major part of the studies, however, focuses on only diatoms; thus, they demonstrate less convincing results. Lavoie et al. (2006, 2010) found no significant relation in the size (e.g. biovolume, surface) distribution along the P gradient; thus, they do not suggest this trait for assessment purposes. However, they refer to former studies on coastal waters (Busse & Snoeijs, 2002, 2003; Snoeijs et al., 2002) where a significant effect of salinity and wave movement on diatom size was shown (Table 2.2). Berthon et al. (2011) used size classes and showed some effect of trophic level and organic

- 58 -

pollution on them but unfortunately without a clear ecological interpretation. Even if nutrients have no clear effect on diatom size, other factors are more relevant. Grazing proved to be a strong selective factor on diatom size (length), and the selected sizes strongly depended on the grazer species (Tall et al., 2006). Another study showed that watercolour (as a proxy of dissolved organic carbon) explained a major part of the size distribution of diatoms in Canadian rivers (Wunsam et al., 2002). A recent study examined the cell size structure of the two main phytobenthos algal group: desmids and diatoms in peatlands (Neustupa et al., 2013). While the cell size of desmids was strongly affected by the ombrominerotrophic gradient, pH, and Ca ion concentration, diatom cell size was weakly related to these factors. In contrast, both biovolume and surface area of diatom cells were strongly correlated with conductivity. This example clearly suggests that benthic algae other than diatoms can provide additional information in a perspective of habitat assessment. Lange et al. (2016) tested their defined traits against farming intensity (as a proxy of nutrients) and water abstraction (effect of streamflow). Results on cell size showed dominance of small size cells under low nutrient level, but increasing of nutrients induced the dominance of large, filamentous forms. An effect of the interaction of water abstraction and farming intensity was also shown. At high farming intensity (high nutrient concentration), high levels of water abstraction favoured the development of small cells. They also demonstrated that farming intensity favoured the development of non-attached but filamentous algae and that water abstraction with the risk of drying out the stream increased the dominance of small, resilient, and motile taxa (Lange et al., 2016). The tests on life-forms showed a positive correlation of unicellular algae with water abstraction (i.e. low water flow stress) at high nutrient levels is in contrast with the presupposition that filamentous forms dominated under such circumstances. Probably, other factors overcame that effect; single-cell organisms have advantage under increased sedimentation, and they also have a greater chance to enter crevices in substratum particles (Lange et al., 2016). Nitrogen fixation as a physiological trait has been also tested and successfully indicated N-limited conditions. Two traits (i.e. reproduction techniques and spore formation) formed the category of life-history traits, and both proved to be successful indicators. The dominance of fragmentation over fission was present under high nutrient levels, which is explained with the dominance of filamentous forms under such circumstances. Spore formation showed negative response to nutrients but positive response to water abstraction.

- 59 -

The ecological classification by B-Béres et al. (2016) is a simple combination of the three ecological guilds of Passy (2007), which also adds a fourth guild of planktic species (Rimet & Bouchez, 2012b) and five size classes (Berthon et al., 2011), resulting in 20 combinations. The study was made in the framework of a colonisation process and analysed the effect of a disturbance event on the benthic diatom community as well. The study showed that the ecological guilds were not correlated significantly with the tested environmental factors. In contrast, in the combined eco-morphological groups, size classes highlighted differences within the same guilds. For example, small, low-profile species were present in the beginning of the colonisation, which stemmed from a fragmented mature biofilm and settled. Small species from the same guild dominated after a strong disturbance event (heavy raining) as the first colonisers. The study showed that the further refinement of existing classification could detect new niches.

Two of the four classifications involved non-diatoms in their study (Table 1), and in both cases it provided important additional information. Diatomists often forget about other taxa and draw their conclusions for the whole phytobenthos based only on diatom data. Even though diatoms can be often used as a proxy for the entire benthic algal community, in a perspective of assessment, studies on all groups may provide important additional information (Denicola and Kelly, 2014). The use of only diatoms has the practical advantage of a standardised sampling and preparation methods after which species can be easily identified based on clear morphological features. The question remains whether this advantage can compensate the loss of other information. Kelly et al. (2008) tested the species-environment responses based on three kind of datasets: only diatoms, diatoms and non-diatoms, and only non-diatoms. Results based on only diatoms were similar to results based on diatom and non-diatom data together, but both gave better correlation to environmental drivers (i.e. total phosphorus, dissolved inorganic carbon, conductivity, and calcium concentration) than non-diatoms alone. Even if nondiatoms represent a wider ecological scale, their indicator value is low due to the lower species richness (Kelly et al., 2008). On the other hand, it is clear that for a better understanding of

- 60 -

the benthic communities, study of algae other than diatoms is inevitable. They have an important contribution, especially in eutrophic waters, and often dominate the algal community (Denicola et al., 2004). Although for now, most European countries use only diatom-based metrics in ecological quality assessment based on “macrophytes and phytobenthos” required by the WFD, there are countries using indices, including nondiatoms: Austria and Germany (Rott et al., 1997, 1999; Schaumburg et al., 2004), Czech Republic, and Norway (Schneider & Lindstrøm, 2009, 2011). The Norwegian examples demonstrate the utility of non-diatom benthic algae (mainly filamentous chlorophytes) as the indicator of trophic level (Periphyton index of trophic status, PIT — Schneider & Lindstrøm, 2011) and acidity (acidification index periphyton, AIP — Schneider & Lindstrøm, 2009). Schneider et al. (2012) showed that including non-diatom algae can provide additional information of the habitat. They found that while non-diatoms were mainly influenced by the channel substrate parameters, the diatom assemblage was influenced by both the substrate parameters and the riverbank characteristics. The authors explain those changes with the different dispersal characteristics. While filamentous Cyanobacteria or Chlorophyceae attach strongly to the substrate, diatoms generally disperse more easily due to the water flow. After the transport of diatom cells, habitat selection depends on the riverbank morphology, while non-diatom benthic algae are more dependent on the local substrates. Such important ecological differences have to be considered for the functional characterisation of the phytobenthos. A simple measurable trait like the filament width of Oedogonium was found to positively correlate with the TP concentration (Schneider & Lindstrøm, 2011). Abundant appearance of Mougeotia indicates acidification (Graham et al., 1996a, b). It is possible that several already mentioned contradictions in studies trying to understand phytobenthos based on only diatom data were derived from such lack of information. The number of groups is a critical point of functional classification. Comparative studies on phytoplankton functional groups show that the two most effective classification in terms of covering habitat diversity are the FG (Reynolds et al., 2002; Padisák et al., 2009) and MFG classifications (Salmaso & Padisák, 2007). They give similar results with their 40 and 31 groups, respectively. We suggest that a number of 20-40 groups would be ideal to cover habitat-diversity. Although benthic diatom assemblages in rivers are different from planktic communities in lakes, a similar conclusion is considered valid: only a few groups are insufficient to cover the main habitat types. However, the existence of each group has to be clearly justified. Although the combination of the CSR strategies and life-forms resulted in

- 61 -

21 groups, most of them remained hypothetical. The study of Lange et al. (2016) does not define groups, but the number of ecologically meaningful traits has the potential to define several ecological groups. The criteria of group definition is morphological in three cases (Table 1). It has the advantage of easy measurements and use, but it clearly limits the potential of defining a sufficient number of groups. An important task is collecting as much information as possible about the possible traits and their functional roles, as was previously done by Lange et al (2016). In further studies, it is important not only to look for correlations in environmental data but to confirm them by experiments where we can see not only correlations but causations too. Laboratory experiments on the effect of water flow, grazing, temperature, light intensity, nutrients, conductivity, etc. (e.g. Lange et al., 2011; Svensson et al., 2014; Cochero et al., 2015; Lengyel et al., 2015a) on particular species can provide useful information about the species preference that can be than built in a trait database and help to define more realistic ecological groups. A particular symbiotic relation is represented between some species of cyanobacteria and species of diatoms (e.g. the genera Epithemia and Rhopalodia) (Janson, 2002). Having these symbiotic cyanobacteria is a very important trait of these diatoms. These heterocytous cyanobacteria can fix atmospheric nitrogen, which provides a clear competitive advantage in N-limited habitats (Stancheva et al., 2013; Lange et al., 2016). Unfortunately, at present, due to the applied protocol for diatom sampling and preparation, we lack important information from monitoring data. Lack of data about nondiatom benthic algae can cause shortcomings regarding the functionality of the benthic community. During analysing the samples, we know neither which cells were alive in the time of sampling nor which cells were already dead. Obviously, originally dead cells do not represent the conditions in the time of sampling. Although one study showed no difference between involving this information or not in assessing habitat conditions (Gillett et al., 2008), which was probably based on the habitat type (i.e. the current effect that washes away dead cells), results can change. We lose all information visible only in unprepared samples (e.g. type of attachment to substrate, colony-forming) after preparation. Even if we have information about this for several species, we can never be sure since some species can change their traits. Some Cymbella species can be unicellular and motile once and yet attached with a peduncle another time (Rimet & Bouchez, 2012b). Encyonema silesiacum can be found motile and tube-dwelling or colonial, and Amphora lybica can be attached with

- 62 -

entire valve surface or stalked too (Law et al., 2014). This means that one species can be represented in two ecological groups depending on the environmental conditions in which they exist. In a particular case, a potential shift could be detected in ecological groups but not at the species level. Another example from phytoplankton is the planktic Cylindrospermopsis raciborskii that can be classified into two functional groups depending on whether it develops near the surface or forms a deep layer population (Padisák et al., 2009). For the mentioned reasons, more studies on traits are welcome, and data from investigation of unprepared samples are also necessary. In some cases where traits are missing for several species, phylogeny can also serve with solutions. Only if we have evidence that a particular trait is phylogenetically related can we assign this trait to all the taxa of that phylogenetic level (Keck et al., in press; Larras et al., 2014). We already possess knowledge of traits and their usability under particular conditions (Table 2.2). Most of our information is based on studies tested with nutrients and physical forces (flow, grazing), since they are the most common features that shape the benthic community. This information can be used to develop a conceptual framework similar to the one of Lange et al. (2016) and test them on a diverse dataset. One of the weakest points of the eco-morphological classification (B-Béres et al., 2016) is that their dataset originates from artificial substrata at one single sampling station containing only 34 diatom species. In contrast, the study of Lange et al. (2016) covers several sites in the catchment area of a river, with samples of diatoms and non-diatoms from natural substrate, containing 91 taxa in total. Some shortcomings presented above can be derived from the problematics presented by Kelly (2012). We gain our information on the phytobenthos after a set of technical processes (e.g. sampling, sample preservation, preparation, microscopic examination, use of transfer function for quantifications); after that, the real picture of the community becomes an abstraction understood only by the experts. Kelly’s proposition is a more holistic view with the help of ‘guiding’ images that not only provide a method for generating a more realistic view on the phytobenthos but also strengthens the bridge between scientists and end-users.

We propose two basic, general methods for the development of functional groups. The first is based on an assignation of traits to species (Fig. 2.2). The chosen traits must be ecologically meaningful and justified by literature or experiments. These data are used to define groups of species that possess similar traits. The definition of such groups can be carried out by

- 63 -

statistical methods, e.g. ordination techniques, clustering (Margalef, 1978; Usseglio-Polatera et al., 2000; Kruk et al., 2010; Law et al., 2014), or by expert knowledge (Passy, 2007; Salmaso & Padisák, 2007; Centis et al., 2010; B-Béres et al., 2016). The use of expert knowledge requires strong background knowledge in order to define ecologically meaningful groups. This concept was used for the ecological guilds (Passy, 2007) and the morpho-functional diatom groups (MFDG) (Centis et al., 2010), which is an adaptation of the morphofunctional group (MFG) classification developed for phytoplankton (Salmaso & Padisák, 2007) on planktic diatoms. The eco-morphological functional groups of B-Béres et al. (2016) is also based on presupposed combination of traits. Then, these groups must be tested on environmental data to see if they represent separated niches of the environment. This can be easily done by multivariate analyses, e.g. canonical correspondence analysis (B-Béres et al., 2016). The other approach we propose (Fig. 3) is based on the Functional Group classification for phytoplankton (Reynolds et al., 2002). It has a phytosociological base in analogy to “associations” of terrestrial plants. A particular habitat is represented by a set of environmental characteristics, to which the occurring species are adapted, i.e. they possess functional traits that make the species competitive and therefore abundant there. A similar approach has not yet been tested for benthic algae flora. This approach also requires welldefined trait-environmental factor relations so that a new species can be classified into a functional group. The definition criteria for phytoplankton FGs is not only morphological but also structural, functional, ecological (e.g. trophic preferences), habitat-based, and taxonomical. The advantage of using trait-based classifications in ecological assessment is already recognised, and there is a trend in developing and using them for the purposes of the WFD (Hering et al., 2010; Reyjol et al., 2014). Several studies address developing such groups for the different “Biological Quality Elements” of the WFD: macrophytes (Orfanidis et al., 2003; Wells et al., 2007), fish (Logez et al., 2013), macroinvertebrates (Dolédec & Statzner, 2008; Borja et al., 2009), and phytoplankton (Padisák et al., 2006). The assemblage index to evaluate ecological status of lakes with their phytoplankton (Padisák et al., 2006) composition is based on the functional classification by Reynolds (Reynolds et al., 2002; Padisák et al., 2009). The index classifies Hungarian lakes into eight lake types according to their typology. Each phytoplankton codon has a factor number between 1 and 5 for each lake type indicating how favourable the presence of this codon in the particular habitat type. The final value is a simple average of the biomass contribution of each codon weighted by their factor value. A possible

- 64 -

first step of integrating the approach in the Water Framework Directive can be the testing of already ecologically justified traits on the European river typology. There is a potential for defining particular trait compositions for the typological categories that can be further specified with further analyses. For example, we have already good knowledge of the trait vs. nutrient or physical force, while less is known about how benthic algae communities are affected by the geochemical properties of the water body. Another crucial point is the question of seasonality. Especially from the point of view of applied assessment, it would be necessary to standardise the sampling period. The most important requirement is representativeness. We have to look for the nearly competitively selected equilibrium period when the benthic community is the most representative. For example, for phytoplankton in lakes, Padisák et al., (2006) suggest the stable late summer period when the phytoplankton community reach a near steady-state condition, and this period also integrates the preceding events. Benthic communities are less intensively studied in terms of seasonality (King et al., 2006; Lengyel et al., 2015b). Although colonisation time of the substrate by benthic algae strongly varies based on the environmental conditions, it can be measured in weeks. Generally, colonization experiments suggest four weeks for having a densely colonised substrate that we can sample. Therefore, sampling times should be long enough (~ 4 weeks) after the last known strong disturbance event that can reverse the successional phases. It is important to avoid the comparison of habitats assessed by communities in different colonisation phases, because we may detect the differences between the successional stages and not the habitats (King et al., 2006). Higher water temperature and light intensity enhance succession speed (Hoagland et al., 1982); hence, they practically can help to find an appropriate sampling time. Since in winter (under temperate climate and average altitude), the low temperature and light stress the community, these circumstances result in potentially low diversity assemblages with stress-tolerant species. This period is clearly not appropriate if the aim is to detect differences between the habitats caused by processes that are more complex. The spring period with its strong floods represent a likewise pressure avoiding the development of a mature biofilm, resulting in similar communities with different habitats. Hoagland et al. (1982) carried out a study investigating the successional and colonisation process of a benthic algal community on artificial substrates in two reservoirs. They showed that the densest biofilm appeared in the summer, and the two reservoirs differed most in the summer based on their benthic community. Thus, we suggest that under a temperate climate, the summer-autumn period appears to be the appropriate period for sampling. When algal succession is fast, the biofilm is dense, and the

- 65 -

algal community may reach the competitively selected equilibrium state. We have to consider that sampling time is also climate-dependent. Based on our own experiences on the tropical island of Mayotte, located 200 kilometres east of Madagascar, the second half of the dry season (July-August) proved to be the best for assessing environmental conditions. In their paper, Hoagland et al. (1982) also showed that filamentous non-diatom algae become apparent in the last phase of succession that confirms the importance of investigating nondiatom benthic algae in a functional classification. If we suggest an analogy between the species-based autecological indices and the functional group indices, it would be worth considering using functional diversity metrics with species-based diversity metrics (e.g. richness, Shannon-diversity, evenness). The first attempts on these metrics, which have been tested on virtual animal trait database, are promising (Schleuter et al., 2010). The upcoming challenge for diatomists is to define ecologically meaningful functional traits that will be used for the development of adequate number functional groups for diatoms covering as many different habitat types as possible. This classification can be the basis of a new quality evaluation system that is more robust and general, since it is based on traits and adaptations instead of species.

- 66 -

Conceptual framework of defining ecological groups; definition of functional groups (FGs) are based on a species-trait database using statistical methods or expert knowledge. Datasets of FGs and environmental parameters are used to define the ecology of FGs via multivariate statistical methods

- 67 -

Conceptual framework of defining ecological groups; environmental data are used to define habitat types either with statistical methods or expert knowledge. Each habitat types possess dominant species with adaptive traits. Then, an interpretation of the trait– environment relation is required

- 68 -

This study was funded by ONEMA (Office National de l’Eau et des Milieu Aquatiques).

B-Béres, V., Á. Lukács, P. Török, Z. Kókai, Z. Novák, E. T-Krasznai, B. Tóthmérész & I. Bácsi, 2016. Combined eco-morphological functional groups are reliable indicators of colonisation processes of benthic diatom assemblages in a lowland stream. Ecological Indicators 64: 31–38. Berthon, V., A. Bouchez & F. Rimet, 2011. Using diatom life-forms and ecological guilds to assess organic pollution and trophic level in rivers: a case study of rivers in south-eastern France. Hydrobiologia 673: 259– 271. Besse-Lototskaya, A., P. F. M. Verdonschot, M. Coste & B. Van de Vijver, 2011. Evaluation of European diatom trophic indices. Ecological Indicators 11: 456–467. Blanco, S., C. Cejudo-Figueiras, L. Tudesque, E. Bécares, L. Hoffmann & L. Ector, 2012. Are diatom diversity indices reliable monitoring metrics? Hydrobiologia 695: 199–206. Blondel, J., 2003. Guilds or functional groups: does it matter? Oikos 100: 223–231. Borics, G., G. Várbíró & J. Padisák, 2013. Disturbance and stress: different meanings in ecological dynamics? Hydrobiologia 711: 1–7. Borja, A., A. Miles, A. Occhipinti-Ambrogi & T. Berg, 2009. Current status of macroinvertebrate methods used for assessing the quality of European marine waters: implementing the Water Framework Directive. Hydrobiologia 633: 181–196. Busse, S. & P. Snoeijs, 2002. Gradient responses of diatom communities in the Bothnian Bay, northern Baltic Sea. Nova Hedwigia 74: 501–525. Busse, S. & P. Snoeijs, 2003. Gradient responses of diatom communities in the Bothnian Sea (northern Baltic Sea), with emphasis on responses to water movement. Phycologia 42: 451–464. Carrick, H. J. & R. L. Lowe, 1989. Benthic algal response to N and P enrichment along a pH gradient. Hydrobiologia 179: 119–127. Cattaneo, A., 1987. Size distribution in periphyton. Canadian Journal of Fisheries and Aquatic Sciences 44: 2025–2028. Centis, B., M. Tolotti & N. Salmaso, 2010. Structure of the diatom community of the River Adige (NorthEastern Italy) along a hydrological gradient. Hydrobiologia 639: 37–42. Cochero, J., M. Licursi & N. Gómez, 2015. Changes in the epipelic diatom assemblage in nutrient rich streams due to the variations of simultaneous stressors. Limnologica – Ecology and Management of Inland Waters 51: 15–23. Cemagref, 1982. Etude des méthodes biologiques quantitative d’appréciation de la qualité des eaux. Rapport Q.E.Lyon-A.F.Bassin Rhône-Méditerranée-Corse: 218 pp. Crossetti, L. O. & C. E. de M. Bicudo, 2008. Adaptations in phytoplankton life strategies to imposed change in a shallow urban tropical eutrophic reservoir, Garças Reservoir, over 8 years. Hydrobiologia 614: 91–105. Denicola, D. M. & M. Kelly, 2014. Role of periphyton in ecological assessment of lakes. Freshwater Science 33: 619–638. Denicola, D. M., E. de Eyto, A. Wemaere & K. Irvine, 2004. Using epilithic algal communities to assess trophic status in Irish lakes. Journal of Phycology 40: 481–495. De Queiroz, K., 2007. Species concepts and species delimitation. Systematic Biology 56: 879–886. Dolédec, S. & B. Statzner, 2008. Invertebrate traits for the biomonitoring of large European rivers: an assessment of specific types of human impact. Freshwater Biology 53: 617–634.

- 69 -

Dunck, B., J. C. Bortolini, L. Rodrigues, L. C. Rodrigues, S. Jati & S. Train, 2013. Functional diversity and adaptative strategies of planktonic and periphytic algae in isolated tropical floodplain lake. Brazilian Journal of Botany 36: 257–266. European Commission, 2000. Directive 2000/60/EC of the European Parliament and of the Council of 23rd October 2000 establishing a framework for Community action in the field of water policy. Official Journal of the European Communities 327: 1–72. Gillett, N., Y. Pan & C. Parker, 2008. Should only live diatoms be used in the bioassessment of small mountain streams? Hydrobiologia 620: 135–147. Gómez-Aparicio, L., 2009. The role of plant interactions in the restoration of degraded ecosystems: a metaanalysis across life-forms and ecosystems. Journal of Ecology 97: 1202–1214. Gottschalk, S. & M. Kahlert, 2012. Shifts in taxonomical and guild composition of littoral diatom assemblages along environmental gradients. Hydrobiologia 694: 41–56. Graham, J. M., P. Arancibia-Avila & L. E. Graham, 1996a. Effects of pH and selected metals on growth of the filamentous green alga Mougeotia under acidic conditions. Limonology and Oceanography 41: 263–270. Graham, J. M., P. Arancibia-Avila & L. E. Graham, 1996b. Physiological ecology of a species of the filamentous green alga Mougeotia under acidic conditions: light and temperature effects on photosynthesis and respiration. Limonology and Oceanography 41: 253–262. Grime, J. P., 1974. Vegetation classification by reference to strategies. Nature 250: 26–31. Grime, J. P., 1977. Evidence for the existence of three primary strategies in plants and its relevance to ecological and evolutionary theory. American Naturalist 111: 1169–1194. Grime, J. P., 1989. The stress debate: symptom of impending synthesis? Biological Journal of the Linnean Society 37: 3–17. Hardin, G., 1960. The competitive exclusion principle. Science 131: 1292–1297. Hering, D., A. Borja, J. Carstensen, L. Carvalho, M. Elliott, C. K. Feld, A.-S. Heiskanen, R. K. Johnson, J. Moe, D. Pont, & others, 2010. The European Water Framework Directive at the age of 10: a critical review of the achievements with recommendations for the future. Science of the Total Environment 408: 4007–4019 Hoagland, K. D., S. C. Roemer & J. R. Rosowski, 1982. Colonization and community structure of two periphyton assemblages, with emphasis on the diatoms (Bacillariophyceae). American Journal of Botany 69: 188–213. Janson, S., 2002. Cyanobacteria in Symbiosis with Diatoms Cyanobacteria in Symbiosis. Springer, New York: 1–10. Kahlert, M., R.-L. Albert, E.-L. Anttila, R. Bengtsson, C. Bigler, T. Eskola, V. Gälman, S. Gottschalk, E. Herlitz, A. Jarlman, J. Kasperoviciene, M. Kokociński, H. Luup, J. Miettinen, I. Paunksnyte, K. Piirsoo, I. Quintana, J. Raunio, B. Sandell, H. Simola, I. Sundberg, S. Vilbaste & J. Weckström, 2009. Harmonization is more important than experience—results of the first Nordic-Baltic diatom intercalibration exercise 2007 (stream monitoring). Journal of Applied Phycology 21: 471–482. Kahlert, M., M. Kelly, R.-L. Albert, S. F. P. Almeida, T. Bešta, S. Blanco, M. Coste, L. Denys, L. Ector, M. Fránková, D. Hlúbiková, P. Ivanov, B. Kennedy, P. Marvan, A. Mertens, J. Miettinen, J. PicinskaFałtynowicz, J. Rosebery, E. Tornés, S. Vilbaste & A. Vogel, 2012. Identification versus counting protocols as sources of uncertainty in diatom-based ecological status assessments. Hydrobiologia 695: 109–124. Kattge, J., S. Díaz, S. Lavorel, I. C. Prentice, P. Leadley, G. BöNisch, E. Garnier, M. Westoby, P. B. Reich, I. J. Wright, J. H. C. Cornelissen, C. Violle, S. P. Harrison, P. M. Van Bodegom, M. Reichstein, B. J. Enquist, N. A. Soudzilovskaia, D. D. Ackerly, M. Anand, O. Atkin, M. Bahn, T. R. Baker, D. Baldocchi, R. Bekker, C. C. Blanco, B. Blonder, W. J. Bond, R. Bradstock, D. E. Bunker, F. Casanoves, J. Cavender-Bares, J. Q. Chambers, F. S. Chapin Iii, J. Chave, D. Coomes, W. K. Cornwell, J. M. Craine, B. H. Dobrin, L. Duarte, W. Durka, J. Elser, G. Esser, M. Estiarte, W. F. Fagan, J. Fang, F. FernáNdez-MéNdez, A. Fidelis, B. Finegan, O. Flores, H. Ford, D. Frank, G. T. Freschet, N. M. Fyllas, R. V. Gallagher, W. A. Green, A. G. Gutierrez, T. Hickler, S. I. Higgins, J. G. Hodgson, A. Jalili, S. Jansen, C. A. Joly, A. J. Kerkhoff, D. Kirkup, K. Kitajima, M. Kleyer, S. Klotz, J. M. H. Knops, K. Kramer, I. KüHn, H. Kurokawa, D. Laughlin, T. D. Lee, M. Leishman, F. Lens, T. Lenz, S. L. Lewis, J. Lloyd, J. Llusià, F. Louault, S. Ma, M. D. Mahecha, P. Manning, T. Massad, B. E. Medlyn, J. Messier, A. T. Moles, S. C. MüLler, K. Nadrowski, S. Naeem, Ü. Niinemets, S. NöLlert, A. NüSke, R. Ogaya, J. Oleksyn, V. G. Onipchenko, Y. Onoda, J. OrdoñEz, G. Overbeck, W. A. Ozinga, S. PatiñO, S. Paula, J. G. Pausas, J. PeñUelas, O. L. Phillips, V. Pillar, H. Poorter,

- 70 -

L. Poorter, P. Poschlod, A. Prinzing, R. Proulx, A. Rammig, S. Reinsch, B. Reu, L. Sack, B. Salgado-Negret, J. Sardans, S. Shiodera, B. Shipley, A. Siefert, E. Sosinski, J.-F. Soussana, E. Swaine, N. Swenson, K. Thompson, P. Thornton, M. Waldram, E. Weiher, M. White, S. White, S. J. Wright, B. Yguel, S. Zaehle, A. E. Zanne & C. Wirth, 2011. TRY – a global database of plant traits: TRY – a global database of plant traits. Global Change Biology 17: 2905–2935. Keck, F., F. Rimet, A. Franc, & A. Bouchez, 2015. Phylogenetic signal in diatom ecology: perspectives for aquatic ecosystems biomonitoring. Ecological Applications. 26 (3): 861-872. doi:10.1890/14-1966. Kelly, M., 2011. The Emperor’s new clothes? A comment on Besse-Lototskaya et al. 2011. Ecological Indicators 11: 1492–1494. Kelly, M., 2012. The semiotics of slime: visual representation of phytobenthos as an aid to understanding ecological status. Freshwater Reviews 5: 105–119. Kelly, M., 2013. Data rich, information poor? Phytobenthos assessment and the Water Framework Directive. European Journal of Phycology 48: 437–450. Kelly, M. G. & B. A. Whitton, 1995. The trophic diatom index: a new index for monitoring eutrophication in rivers. Journal of Applied Phycology 7: 433–444. Kelly, M. G., L. King, R. I. Jones, P. A. Barker & B. J. Jamieson, 2008. Validation of diatoms as proxies for phytobenthos when assessing ecological status in lakes. Hydrobiologia 610: 125–129. Kermarrec, L., A. Franc, F. Rimet, P. Chaumeil, J.-M. Frigerio, J.-F. Humbert & A. Bouchez, 2014. A nextgeneration sequencing approach to river biomonitoring using benthic diatoms. Freshwater Science 33: 349– 363. King, L., G. Clarke, H. Bennion, M. Kelly & M. Yallop, 2006. Recommendations for sampling littoral diatoms in lakes for ecological status assessments. Journal of Applied Phycology 18: 15–25. Kruk, C., V. L. M. Huszar, E. T. H. M. Peeters, S. Bonilla, L. Costa, M. Lürling, C. S. Reynolds & M. Scheffer, 2010. A morphological classification capturing functional variation in phytoplankton. Freshwater Biology 55: 614–627. Laine, M., S. Morin & J. Tison-Rosebery, 2014. A multicompartment approach – diatoms, macrophytes, benthic macroinvertebrates and fish – to assess the impact of toxic industrial releases on a small French river. PLoS One 9: e102358. doi:10.1371/journal.pone.0102358. Lange, K., A. Liess, J. J. Piggott, C. R. Townsend & C. D. Matthaei, 2011. Light, nutrients and grazing interact to determine stream diatom community composition and functional group structure: diatom responses to light, nutrients and grazing. Freshwater Biology 56: 264–278. Lange, K., C. R. Townsend & C. D. Matthaei, 2016. A trait-based framework for stream algal communities. Ecology and Evolution 6: 23–36. Larras, F., F. Keck, B. Montuelle, F. Rimet & A. Bouchez, 2014. Linking diatom sensitivity to herbicides to phylogeny: a step forward for biomonitoring? Environmental Science & Technology 48: 1921–1930. Lavoie, I., S. Campeau, M.-A. Fallu & P. J. Dillon, 2006. Diatoms and biomonitoring: should cell size be accounted for? Hydrobiologia 573: 1–16. Lavoie, I., J. Lento & A. Morin, 2010. Inadequacy of size distributions of stream benthic diatoms for environmental monitoring. Journal of the North American Benthological Society 29: 586–601. Law, R. J., J. A. Elliott & S. J. Thackeray, 2014. Do functional or morphological classifications explain stream phytobenthic community assemblages? Diatom Research 29: 309–324. Leira, M., G. Chen, C. Dalton, K. Irvine & D. Taylor, 2009. Patterns in freshwater diatom taxonomic distinctness along an eutrophication gradient. Freshwater Biology 54: 1–14. Leira, M., M. L. Filippi & M. Cantonati, 2015. Diatom community response to extreme water-level fluctuations in two Alpine lakes: a core case study. Journal of Paleolimnology 53: 289–307. Lengyel, E., A. W. Kovács, J. Padisák & C. Stenger-Kovács, 2015a. Photosynthetic characteristics of the benthic diatom species Nitzschia frustulum (Kützing) Grunow isolated from a soda pan along temperature-, sulfateand chloride gradients. Aquatic Ecology 49: 401–416. Lengyel, E., J. Padisák & C. Stenger-Kovács, 2015b. Establishment of equilibrium states and effect of disturbances on benthic diatom assemblages of the Torna-stream, Hungary. Hydrobiologia 750: 43–56.

- 71 -

Linneaus, C., 1758. Systema naturae per regna tria naturae: secundum classes, ordines, genera, species, cum characteribus, differentiis, synonymis, locis. Laurentius Salvius, Stockholm. Litchman, E. & C. A. Klausmeier, 2008. Trait-based community ecology of phytoplankton. Annual Review of Ecology, Evolution, and Systematics 39: 615–639. Logez, M., P. Bady, A. Melcher & D. Pont, 2013. A continental-scale analysis of fish assemblage functional structure in European rivers. Ecography 36: 80–91. Mackay, A. W., T. Davidson, P. Wolski, S. Woodward, R. Mazebedi, W. R. L. Masamba & M. Todd, 2012. Diatom sensitivity to hydrological and nutrient variability in a subtropical, flood-pulse wetland. Ecohydrology 5: 491–502. Mann, D. G., 1999. The species concept in diatoms. Phycologia 38: 437–495. Mann, D. G. & P. Vanormelingen, 2013. An inordinate fondness? The number, distributions, and origins of diatom species. Journal of Eukaryotic Microbiology 60: 414–420. Margalef, R., 1978. Life-forms of phytoplankton as survival alternatives in an unstable environment. Oceanologica acta 1: 493–509. Morin, A., N. Bourassa & A. Cattaneo, 2001. Use of size spectra and empirical models to evaluate trophic relationships in streams. Limnology and Oceanography 46: 935–940. Neilan, B. A., D. Jacobs & A. E. Goodman, 1995. Genetic diversity and phylogeny of toxic cyanobacteria determined by DNA polymorphisms within the phycocyanin locus. Applied and Environmental Microbiology 61: 3875–3883. Neustupa, J., J. Veselá & J. Št’astný, 2013. Differential cell size structure of desmids and diatoms in the phytobenthos of peatlands. Hydrobiologia 709: 159–171. Orfanidis, S., P. Panayotidis & N. Stamatis, 2003. An insight to the ecological evaluation index (EEI). Ecological Indicators 3: 27–33. Padisák, J., 1993. The influence of different disturbance frequencies on the species richness, diversity and equitability of phytoplankton in shallow lakes. Hydrobiologia 249: 135–156. Padisák, J., 2003. Phytoplankton. In O’Sullivan, P. E. & C. S. Reynolds (eds), The lakes handbook, Vol. 1. Blackwell Science Ltd, Hoboken: 251–308. Padisák, J., G. Borics, I. Grigorszky & É. Soróczki-Pintér, 2006. Use of phytoplankton assemblages for monitoring ecological status of lakes within the water framework directive: the assemblage index. Hydrobiologia 553: 1–14. Padisák, J., L. O. Crossetti & L. Naselli-Flores, 2009. Use and misuse in the application of the phytoplankton functional classification: a critical review with updates. Hydrobiologia 621: 1–19. Passy, S. I., 2007. Diatom ecological guilds display distinct and predictable behavior along nutrient and disturbance gradients in running waters. Aquatic Botany 86: 171–178. Pianka, E. R., 1970. On r-and K-selection. American Naturalist 104: 592–597. Pringle, C. M., 1990. Nutrient spatial heterogeneity: effects on community structure, physiognomy, and diversity of stream algae. Ecology 71: 905. Prygiel, J. & M. Coste, 1998. Mise au point de l’Indice Biologique Diatomée, un indice diatomique pratique applicable au réseau hydrographique français. L’Eau, l’industrie, les nuisances 211: 40–45. Raunkiaer, C., 1934. The Life Forms of Plants and Statistical Plant Geography. The Clarendon Press, Oxford. Reyjol, Y., C. Argillier, W. Bonne, A. Borja, A. D. Buijse, A. C. Cardoso, M. Daufresne, M. Kernan, M. T. Ferreira, S. Poikane, N. Prat, A.-L. Solheim, S. Stroffek, P. Usseglio-Polatera, B. Villeneuve & W. van de Bund, 2014. Assessing the ecological status in the context of the European Water Framework Directive: where do we go now? Science of The Total Environment 497–498: 332–344. Reynolds, C. S., 1988. Functional Morphology and the Adaptive Strategies of Freshwater Phytoplankton. Growth and Reproductive Strategies of Freshwater Phytoplankton. Cambridge University Press, Cambridge: 388–433. Reynolds, C. S., 2006. Ecology of Phytoplankton. Cambridge University Press, New York.

- 72 -

Reynolds, C. S., J. Padisák & U. Sommer, 1993. Intermediate disturbance in the ecology of phytoplankton and the maintenance of species diversity: a synthesis. Hydrobiologia 249: 183–188. Reynolds, C. S., V. Huszar, C. Kruk, L. Naselli-Flores & S. Melo, 2002. Towards a functional classification of the freshwater phytoplankton. Journal of Plankton Research 24: 417–428. Rimet, F. & A. Bouchez, 2012a. Biomonitoring river diatoms: implications of taxonomic resolution. Ecological Indicators 15: 92–99. Rimet, F. & A. Bouchez, 2012b. Life-forms, cell-sizes and ecological guilds of diatoms in European rivers. Knowledge and Management of Aquatic Ecosystems 406: 1–14. doi:10.1051/kmae/2012018 Rimet, F., J. Gomà, J. Cambra, E. Bertuzzi, M. Cantonati, C. Cappelletti, F. Ciutti, A. Cordonier, M. Coste, F. Delmas, J. Tison, L. Tudesque, H. Vidal & L. Ector, 2007. Benthic diatoms in Western European streams with altitudes above 800 M: characterisation of the main assemblages and correspondence with ecoregions. Diatom Research 22: 147–188. Rimet, F., L. Ector, H.-M. Cauchie & L. Hoffmann, 2009. Changes in diatom-dominated biofilms during simulated improvements in water quality: implications for diatom-based monitoring in rivers. European Journal of Phycology 44: 567–577. Rimet, F., A. Bouchez & B. Montuelle, 2015. Benthic diatoms and phytoplankton to assess nutrients in a large lake: complementarity of their use in Lake Geneva (France–Switzerland). Ecological Indicators 53: 231– 239. Rott, E., G. Hofmann, K. Pall, P. Pfister, & E. Pipp, 1997. Indikationslisten für Aufwuchsalgen, Teil 1: Saprobielle Indikation (Indication lists for periphytic algae. Part 1: Saprobic indication). Bundesministerium für Land-und Forstwirtschaft (Federal Ministry of Agriculture and Forestry), Wien. Rott, E., E. Pipp, P. Pfister, H. Van Dam, K. Ortler, K. Pall, & N. Binder, 1999. Indikationslisten für Aufwuchsalgen in österreichischen Fliessgewässern. Teil 2: Trophie-indikation sowie geochemische Präferenz; taxonomische und toxikologische Anmerkungen. Bundesministerium für Land-und Forstwirtschaft, Wasserwirtschaftskataster, Wien. Round, F. E., R. M. Crawford & D. G. Mann, 1990. The Diatoms: Biology & Morphology of the Genera. Cambridge University Press, Cambridge. Rumeau, A., & M. Coste, 1988. Initiation à la systématique des diatomées d’eau douce. Pour l’utilisation pratique d’un indice diatomique générique. Bulletin Français de la Pêche et de la Pisciculture 309: 1–69. Salmaso, N. & J. Padisák, 2007. Morpho-Functional Groups and phytoplankton development in two deep lakes (Lake Garda, Italy and Lake Stechlin, Germany). Hydrobiologia 578: 97–112. Salmaso, N., L. Naselli-Flores & J. Padisák, 2015. Functional classifications and their application in phytoplankton ecology. Freshwater Biology 60: 603–619. Schaumburg, J., C. Schranz, J. Foerster, A. Gutowski, G. Hofmann, P. Meilinger, S. Schneider & U. Schmedtje, 2004. Ecological classification of macrophytes and phytobenthos for rivers in Germany according to the water framework directive. Limnologica – Ecology and Management of Inland Waters 34: 283–301. Schleuter, D., M. Daufresne, F. Massol & C. Argillier, 2010. A user’s guide to functional diversity indices. Ecological Monographs 80: 469–484. Schneider, S. & E.-A. Lindstrøm, 2009. Bioindication in Norwegian rivers using non-diatomaceous benthic algae: the acidification index periphyton (AIP). Ecological Indicators 9: 1206–1211. Schneider, S. C. & E.-A. Lindstrøm, 2011. The periphyton index of trophic status PIT: a new eutrophication metric based on non-diatomaceous benthic algae in Nordic rivers. Hydrobiologia 665: 143–155. Schneider, S. C., A. E. Lawniczak, J. Picińska-Faltynowicz & K. Szoszkiewicz, 2012. Do macrophytes, diatoms and non-diatom benthic algae give redundant information? Results from a case study in Poland. Limnologica – Ecology and Management of Inland Waters 42: 204–211. Snoeijs, P., S. Busse & M. Potapova, 2002. The importance of diatom cell size in community analysis 1. Journal of Phycology 38: 265–281. Stancheva, R., R. G. Sheath, B. A. Read, K. D. McArthur, C. Schroepfer, J. P. Kociolek & A. E. Fetscher, 2013. Nitrogen-fixing cyanobacteria (free-living and diatom endosymbionts): their use in southern California stream bioassessment. Hydrobiologia 720: 111–127.

- 73 -

Stenger-Kovács, C., E. Lengyel, L. O. Crossetti, V. Üveges & J. Padisák, 2013a. Diatom ecological guilds as indicators of temporally changing stressors and disturbances in the small Torna-stream, Hungary. Ecological Indicators 24: 138–147. Stenger-Kovács, C., L. Tóth, F. Tóth, É. Hajnal & J. Padisák, 2013b. Stream order-dependent diversity metrics of epilithic diatom assemblages. Hydrobiologia 721: 67–75. Stenger-Kovács, C., E. Lengyel, K. Buczkó, F. Tóth, L. Crossetti, A. Pellinger, Z. Zámbóné Doma & J. Padisák, 2014. Vanishing world: alkaline, saline lakes in Central Europe and their diatom assemblages. Inland Waters 4: 383–396. Stenger-Kovács, C., É. Hajnal, E. Lengyel, K. Buczkó & J. Padisák, 2016. A test of traditional diversity measures and taxonomic distinctness indices on benthic diatoms of soda pans in the Carpathian basin. Ecological Indicators 64: 1–8. Stevenson, J., 2014. Ecological assessments with algae: a review and synthesis. Journal of Phycology 50: 437– 461. Stevenson, R. J., & L. L. Bahls, 2002. Periphyton protocols Rapid Bioassessment Protocols for Use in Streams and Wadeable Rivers: Periphyton, Benthic Macroinvertebrates, and Fish. EPA: 1–23, http://water.epa.gov/scitech/monitoring/rsl/bioassessment/ch06main.cfm. Straile, D., M. C. Jochimsen & R. Kümmerlin, 2013. The use of long-term monitoring data for studies of planktonic diversity: a cautionary tale from two Swiss lakes. Freshwater Biology 58: 1292–1301. Svensson, F., J. Norberg & P. Snoeijs, 2014. Diatom cell size, coloniality and motility: trade-offs between temperature, salinity and nutrient supply with climate change. PLoS One 9: e109993. Tall, L., L. Cloutier & A. Cattaneo, 2006. Grazer-diatom size relationships in an epiphytic community. Limnology and Oceanography 51: 1211–1216. Tang, T., S. Q. Niu & D. Dudgeon, 2013. Responses of epibenthic algal assemblages to water abstraction in Hong Kong streams. Hydrobiologia 703: 225–237. Usseglio-Polatera, P., M. Bournaud, P. Richoux & H. Tachet, 2000. Biological and ecological traits of benthic freshwater macroinvertebrates: relationships and definition of groups with similar traits. Freshwater Biology 43: 175–205. Vilar, A. G., J. A. Vonk, S. Bichebois, H. van Dam, W. Admiraal & H. G. van der Geest, 2015. Suspended organic particles drive the development of attached algal communities in degraded peatlands. Hydrobiologia 744: 211–221. Violle, C., M.-L. Navas, D. Vile, E. Kazakou, C. Fortunel, I. Hummel & E. Garnier, 2007. Let the concept of trait be functional! Oikos 116: 882–892. Virtanen, L. K., P. Kongas, S. Aitto-Oja & J. Soininen, 2011. Is temporal occurrence of diatoms related to species traits, local abundance, and regional distribution? Journal of Phycology 47: 1445–1453. von Humboldt, A., 1806. Ideen zu einer Physiognomik der Gewächse. Cotta, Tübingen. Wells, E., M. Wilkinson, P. Wood & C. Scanlan, 2007. The use of macroalgal species richness and composition on intertidal rocky seashores in the assessment of ecological quality under the European Water Framework Directive. Marine Pollution Bulletin 55: 151–161. Wunsam, S., A. Cattaneo & N. Bourassa, 2002. Comparing diatom species, genera and size in biomonitoring: a case study from streams in the Laurentians (Québec, Canada). Freshwater Biology 47: 325–340.

- 74 -

Chapter 3

This paragraph is the self-edited version of the following article: Tapolczai, K., A. Bouchez, C. Stenger-Kovács, J. Padisák & F. Rimet, 2017. Taxonomy- or trait-based ecological assessment for tropical rivers? Case study on benthic diatoms in Mayotte island (France, Indian Ocean). Science of the Total Environment 607-608: 12931303. doi: 10.1016/j.scitotenv.2017.07.093

- 76 -

Abstract Diatom-based ecological quality assessment methods have been implemented and used regularly in the Water Framework Directive. These indices use the species' abundance profiles along a specific environmental gradient, which they aim to assess. However, this approach has several problematic issues including the unstable and fast-changing diatom taxonomy. The use of traits can be a solution if their responses to the environmental pressure are well-defined. Here, we developed taxonomy-based and trait-based diatom assemblage indices to assess the ecological status of riverine sites on a tropical island. The two indices are based on two sub-indices that measure the diatom assemblage response to a nutrient and organic matter/turbidity gradient. Both taxonomy- and trait-based indices correlated significantly with the selected environmental gradients of the test database, which was not used during index development. We showed that traits could be used for quality assessment of the Mayotte rivers and require much less effort than taxonomy-based indices. There were differences between the two types of indices, which are discussed in this paper. As a perspective for further studies, tests of trait-based indices among different eco-regions would be challenging.

Keywords Bacillariophyta, Biomonitoring, Morphological traits, Quality index, Water Framework Directive

- 77 -

Graphical abstract

Several features of benthic diatoms make them efficient and practical indicators of water quality. They display large species diversity (Mann and Vanormelingen, 2013), respond directly and sensitively to environmental changes (Kelly, 1998; McCormick and Stevenson, 1989; Squires et al., 1979), and inhabit almost all aquatic ecosystems but with habitat-specific assemblages (Stevenson et al., 2010). Diatoms are a good proxy for the entire benthic algal community (Kelly et al., 2008), and indices based solely on the diatom community are routinely used to fulfill the requirements of the Water Framework Directive (European Commission, 2000) for quality assessment by phytobenthos. Autecological indices, based on the Zelinka-Marvan equation (Zelinka and Marvan, 1961) have been developed to assess different pressures on water bodies (eutrophication, organic pollution, etc.). These include the Pollution Sensitivity Index (IPS, Coste, 1982), the Biological Diatom Index (BDI, Prygiel and Coste, 1998), and the Trophic Diatom Index (TDI, Kelly and Whitton, 1995). These indices entail several uncertainties. Misidentifications of diatom species (Werner et al., 2016) are common and often lead to disharmony between labs (Kahlert et al., 2009, 2012) or to disharmony in time, e.g., when the analysis of a long-term monitoring program is inconsistent (Straile et al., 2013). Diatom taxonomy is unstable and rapidly changing; precise determination is time-consuming requiring great effort and expertise. However, with the rapid development of DNA metabarcoding techniques, the accurate determination of species has become faster and more reliable (Groendahl et al., 2017; Vasselon et al., 2017; Zimmermann et al., 2015).

- 78 -

Wetzel et al. (2015) showed how a set of species suffered a taxonomic shift over time because of a confused history in their nomenclature. Moreover, rare or uncommon species lack well-established and robust ecological profiles (Besse-Lototskaya et al., 2011). Thus, these species are irrelevant when such indices are used. A further obstacle in using a general taxonomy-based index is the structuring impact of eco-regions (Rimet et al., 2007) - this calls for the development of unique assessment systems. Former studies on organisms other than diatoms, e.g. phytoplankton (Kruk et al., 2010; Padisák et al., 2006; Reynolds et al., 2002) and macroinvertebrates (Mondy et al., 2012; Usseglio-Polatera et al., 2000, 2001) have already proved that trait-based approaches can successfully indicate environmental conditions during ecological quality assessment. These traits can be based on several morphological, physiological, or phenological measurable features at the individual level (Violle et al., 2007). The presence and the abundance of taxa possessing these features then indicate a particular environmental condition. This approach shows an increasing trend in the study of benthic algae as well - especially diatoms (see review by Tapolczai et al., 2016)). Several studies have shown how diatom traits (e.g. ecological guild, size, life-form, etc.) can indicate nutrient supply, organic pollution, shear stress of water-flow, grazing pressure, etc. (B-Béres et al., 2016; Berthon et al., 2011; Lange et al., 2016; Passy, 2007; Soininen et al., 2016). Thus, trait-based metrics can potentially be applied for quality assessment purposes avoiding the weaknesses of taxonomy-based methods and resulting in a less labor intensive and more robust index. Our first objectives was to collect data from a French overseas department, Mayotte (300 km NW from Madagascar) and to then develop and test a classical diatom index based on the abundance and the specific indicator values (optimum and tolerance) based on the ecological profile of the taxa. Our second objective was to define diatom traits (biovolume, surface-to-volume ratio, length-to-width ratio, N-fixing capacity, and ecological guilds) to develop and test a trait-based index. Our third objective was to compare these two indices for several aspects: correlation with the pressure gradients, their sensitivity, and the required working effort to carry out them. Our hypothesis is that morphological and functional traits of diatoms (biovolume, surface-to-volume ratio, length-to-width ratio, capacity for N2 fixing, ecological guilds) indicate environmental conditions and can be used for river ecological quality assessment by requiring much less effort in terms of taxonomical skills than the taxonomy-based index.

- 79 -

Mayotte is a tropical island with a surface area of 374 km2. The island is located in the Indian Ocean northwest of Madagascar and east of Mozambique (12°50′35″S 45°08′18″E). Geologically, it is part of the Comoros archipelago and consists of two main parts: the smaller Petite-Terre (11 km2) and the Grande-Terre (363 km2) where this study was performed. The main pressure on its rivers is related to the fast growing and dense population (226,915 habitants in 2015; Insee, 2016). Moreover, there is an important clandestine immigration estimated at > 150,000 people. Although a system for the disposal and treatment of communal sewage and for municipal waste exists, most households are not connected or do not use it. Thus, the wastewater is often released directly into the streams along with communal waste. The streams are often contaminated with washing powder from laundry containing 4A zeolite as detergent coming from the intensive washing-by-hand activity by the citizens. Industrial activity is not present in the island, but there is agricultural activity on many small fields of local farmers. During the dry season (May to October), the air temperature varies from 22 to 25 °C, and the water level in the rivers is stable and shallow (5–20 cm). The small streams of Mayotte have a typical width of 1–2 m at the middle section. Mayotte is the oldest island of the Comoro archipelago; it was formed 9 million years ago. Thus, it is now highly eroded without high altitudes. The source of most of the streams we sampled is located at an altitude of 100–200 m. Altogether, we sampled 47 sites in 34 rivers covering all of GrandeTerre between 2008 and 2015, (Fig. 3.1) (Bouchez et al., 2016). To cover a wide range of environmental gradients, the original monitoring network (RCS) was complemented with a “reference” network (REF) in 2013 and a “polluted” network (POLL) in 2014. This classification of sites is based, in general, on the visible conditions of the area following the approach of Hughes et al. (1986) or Stoddard et al. (2006). The reference sites were located either upstream or on very sparsely inhabited areas with dense vegetation. Conversely, polluted sites are downstream and/or in urbanized areas.

- 80 -

Location of Mayotte and the river sampling sites. White, grey and black dots represent REF (reference), RCS (Le Réseau de Contrôle de Surveillance – Regular monitoring network), and POLL (polluted) sites, respectively.

Physical and chemical parameters (Supplement 1) and diatoms were sampled from 2008 to 2015 during the dry season (July–August). Sites were samples once a year but the majority of sites were revisited in one year to another. In total, 158 samples were collected. Analyses of physical and chemical parameters followed the APHA standards (APHA., 1976). At each sampling site, the diatoms were removed from five randomly chosen stones collected from the streambed using one toothbrush per site. The resulting material was mixed. Ethanol (90%) was used to preserve the samples until preparation. Sample preparation used hot hydrogen-peroxide and Naphrax for the permanent slides following the European standards (Afnor, 2007, 2014a). During microscopy analyses, a minimum of four hundred valves were identified to the species or genus level (Afnor, 2014b, 2016) in each sample using current

- 81 -

literature (Lange-Bertalot, 2000; Lange-Bertalot et al., 2011a, 2011b; Lange-Bertalot and Metzeltin, 2002; Levkov, 2009; Levkov et al., 2014; Metzeltin et al., 2005; Metzeltin and Lange-Bertalot, 2007).

For each taxon, three morphological traits (metrics) were assigned (four size-classes, five surface-to-volume ratio classes, and six length-to-width ratio classes) as well as one functional trait (four diatom ecological guilds - Rimet and Bouchez, 2012) and one physiological trait (capability for fixing of atmospheric nitrogen) (Table 3.1, Supplement 3). The biovolume and surface values were calculated based on the geometrical shapes used by Leblanc et al. (2012). The dimension values were assigned from an already existing database (Rimet and Bouchez, 2012) and were not directly measured from the samples. For the ecological guild classification, we used the terminology of Passy (2007) modified by Rimet and Bouchez (2012). The trait “N-fixing” was assigned to the species of two genera (Epithemia and Rhopalodia) that possess atmospheric nitrogen-fixing endosymbiotic cyanobacteria (Lange et al., 2016; Stancheva et al., 2013). For the trait database, we converted the frustule-countbased relative abundance values into relative biovolume values

Traits and their classification used in the study

1

Morphological Size classes (biovolume) s1 =< 100 µm3 100 µm3< s2 =< 1000 µm3 1000 µm3< s3 =< 10 000 µm3 s4 > 10 000 µm3

Functional

Physiological

Ecological guilds

N-fixing

Low-profile

Yes

1.5 < lw2 =< 3

High-profile

No

1.5 < sv3 =< 2

3 < lw3 =< 6

Motile

2 < sv4 =< 3 sv5 > 3

6 < lw4 =< 10 10 < lw5 =< 20 lw6 > 20

Planktic

Surface-to-volume ratio sv1 =< 1

Length-to-width ratio lw1 =< 1.5

1 < sv2 =< 1.5

For the index development, we used 75% (118) of our data. The remaining 25% (40) were used to test it. The selection of datasets was random, and the three sampling networks were selected by visible conditions and not via measurable parameters. The final composition of

- 82 -

the two datasets was quite balanced. The test dataset contained 24, 28 and 19% of the total number of samples in the “REF”, “RCS” and “POLL” networks, respectively. Preliminary analyses and definition of environmental gradients used canonical correspondence analysis (CCA). The same training and test datasets were used for the different environmental gradients. Optimum and tolerance values of all taxa have been identified based on their ecological profile (abundance values along the gradients) (Supplement 2). All analyses and calculations used the R statistical software (R Development Core Team, 2008). The ZelinkaMarvan equation (Zelinka and Marvan, 1961) was used to develop the classical diatom index: 𝐼𝑑𝑥. 𝑀 =

∑𝑛𝑗=1 𝑎𝑗 𝑠𝑗 𝑣𝑗 ∑𝑛𝑗=1 𝑎𝑗 𝑣𝑗

a: relative abundance of taxon j s: sensitivity value of taxon j, calculated from the optimum of the taxon along the environmental gradient v: indicator values of taxon j, calculated from the tolerance of the taxon along the environmental gradient The ecological preference was defined for all taxa that were present in > 5% of the samples (101 taxa from the total of 382). The index values were then calibrated on a scale from 0 to 20 where increasing values indicate better quality.

For the trait-based index, the same training and test datasets were used to ensure comparability between indices. Each studied trait-class showed a distribution along the defined environmental gradients on which a logit model was fitted: 𝑒 𝐵0 +𝐵1𝑥 𝑦= 1 + 𝑒 (𝐵0 +𝐵1 𝑥) From which we get:

𝑥=

𝑦 𝑙𝑛 (−𝑦 + 1) − 𝐵0 𝐵1

x: quality value (the value along the environmental gradient) y: relative biovolume of the particular trait

- 83 -

B0 and B1 are standard values of the logit models These values then were calibrated on a scale from 0 to 20 where increasing values indicate better quality and were calculated for the selected trait-classes. A weighted average of these values for each trait-class was then calculated using the pseudo-r2 values of the logit models as weights to get a trait-based index: 𝐼𝑑𝑥. 𝑀𝑡𝑟𝑎𝑖𝑡 =

∑𝑛𝑖=1 𝑥𝑖 ∗ 𝑅𝑖 ∑𝑛𝑖=1 𝑅𝑖

xi: ecological value calculated from the logit model for i trait-class Ri: pseudo-R2 values of the logit model of i trait-class

Fig. 3.2 shows the results of a canonical correspondence analysis: the distribution of sites and their relation with the environmental parameters. Using these results, we defined two environmental gradients (nutrient and organic carbon together with suspended solids and turbidity). These are related to human-indicated pressures. The first axis of the CCA explained 14.18% of the total inertia along which dissolved organic carbon (DOC), total organic carbon (TOC), suspended solids (SS), and turbidity were the important factors. Along the second axis (explaining 11.68% of the total inertia), nutrients (total phosphorus (TP), inorganic phosphate (PO43−), nitrite (NO2−), nitrate (NO3−), ammonium (NH4+), dissolved inorganic nitrogen to dissolved inorganic phosphorus ratio (DIN:DIP)) were the most discriminative factors. The three sampling networks are well separated from each other - especially sites of the “Polluted” network. These sites possess a higher level of N-forms and parameters related to organic pollution and a wide dispersion due to the higher variability of values. The RCS and “Reference” sites are more mixed and more separated along the 2 nd axis (Fig. 3.2).

- 84 -

Results of CCA analysis with all physico-chemical parameters. Distribution of sampling sites of the three networks (A); RCS (circles), REF (squares), and POLL (triangles), and the relative contribution of environmental factors (B). Parameters used to describe the two gradients, nutrient (bold), and organic/turbidity (bold-italic) gradient.

CCAs showing the distribution of samples of the three networks; REF (squares), RCS (circles), POLL (triangles) and the relative contribution of environmental factors of the nutrient- (NO3− – nitrate, DIN/DIP – dissolved inorganic nitrogen-to-dissolved inorganic phosphorus ratio, NH4+ – ammonium, NO2− – nitrite, TP – total phosphorus, PO43 − – phosphate, A), and the organic pollution gradient (turb – turbidity, SS – suspended solids, TOC – total organic carbon, DOC – dissolved organic carbon, B).

- 85 -

Thus, two new CCAs were run with the parameters assigned to the two gradients (Fig. 3.3A–B). Factors that are more related to geology and/or not to pressures were omitted from these analyses (temperature, pH, eH, SO42−, Cl−, Na, conductivity, K+, Mg2+, Ca2+, SRSi). In each case, the 1st axis of the analysis can be interpreted as the pressure gradient of either nutrients or parameters related to organic matter and turbidity (organic/turbidity gradient). Sites of the “Reference” network are more associated with low levels of nitrogen forms, organic carbon, turbidity, and suspended solids. However, phosphorus forms levels are higher at these sites. “Reference network” is well separated from “Polluted” network on these figures. These sites have high levels of the above-mentioned parameters. RCS sites have transitional conditions, but they are more mixed with “Reference” sites than “Polluted” network sites (Fig. 3.3).

Using the indicator and sensitivity values of the taxa and their abundance, two sub-indices (Idx.M_nutr/Idx.M_org) were developed separately to assess nutrient and organic/turbidity pollution. These indices were then tested on the test database, i.e. samples that were not included in the index development (Fig. 3.4A–B). The index values correlated significantly with the nutrient (Idx.M_nutr p < 0.05, r = − 0.91) and the organic/turbidity pollution (Idx.M_org p < 0.05, r = − 0.32) gradients (Table 3.2). Bootstrapping was carried out to study the effect of the two points, which appeared to be outliers in the correlation that gave a confidence interval of the correlation coefficients at 95% (CI95) of (− 0.98, − 0.86) and CI95 = (− 0.74, 0.23), respectively. By removing the outliers, the correlation coefficients are r = − 0.88 and r = − 0.77, respectively, with a confidence interval of CI95 = (− 0.98, − 0.81) for the nutrient gradient and CI95 = (− 0.90, − 0.62) for the organic/turbidity gradient. This shows that the outliers strongly biased the correlation of Idx.M and the organic/turbidity gradient.

- 86 -

Pearson's correlation of the taxonomy-based quality sub-indices with the nutrient (A) and the organic pollution (B) gradients, practically the site locations along the 1st axes of the CCA analyses (Idx.M_nutr p < 0.05, r = − 0.91 with a CI95 = (− 0.98, − 0.86) and Idx.M_org p < 0.05, r = − 0.32, with CI95 = (− 0.74, 0.23), respectively) of the test dataset. Without the two outliers, the correlation coefficients are r = − 0.88 and r = − 0.77, respectively.

Correlation coefficients and confidence interval after bootstrapping of the different indices (Idx.M_nutr, Idx.M_org, Idx.M_nutrtrait, Idx.M_orgtrait) and pressure gradients (nutrient, organic/turbidity).

1

a

Original R

CI95

Ra

CI95a

Idx.M_nutr

-0.91, p< 0.05

(-0.98, -0.86)

-0.88, p< 0.05

(-0.98, -0.81)

Idx.M_org

-0.32, p< 0.05

(-0.74, 0.23)

-0.77, p< 0.05

(-0.90, -0.62)

Idx.M_nutrtrait

-0.47, p< 0.05

(-0.83, -0.10)

-0.80, p< 0.05

(-0.96, -0.70)

Idx.M_orgtrait

-0.54, p< 0.05

(-0.71, -0.34)

-0.63, p< 0.05

(-0.82, -0.49)

indicates correlation coefficients and CI95 without outliers.

For each environmental gradient, the response of the selected traits against the two gradients was observed and the logit model was fitted on the distribution. The pseudo-R2 values of the logit regressions were defined (Table 3.3). Generally, the trait classes showed better correlation with the organic matter gradient than with the nutrients. Among ecological traits, motile and low-profile guilds correlated significantly both with the organic matter and the nutrient gradient. Alone, the size- and length-to-width ratio classes did not give interpretable results, but we got significant results when the classes were combined together based on their direction of change along the gradients. The results show the dominance of motile, smallsized (s3, s4) and quite elongated (lw3-lw6) taxa at polluted sites. Because we found strong

- 87 -

autocorrelation between biovolume and surface-to-volume ratio, the latter was not included in the index development (Supplement 4). We observed that high proportion of species hosting N-fixing cyanobacterial endosymbionts occurred under low total nitrogen values, but this was not significant (p = 0.12). Thus, this trait was not used in the index development (Supplement 5). Two trait-based quality indices were then developed for the two gradients (Idx.M_nutrtrait, Idx.M_orgtrait) based on the same metrics: the relative biovolume of the motile guild, the low-profile guild, the sum of s1 and s2 size classes, and the sum of lw1 and lw2 length-to-width classes. The equations of the quality values based on the metrics and calibration equations with standard values are presented in Supplement 6. In the computation, we only used the relative biovolume of s1 and s2 but not s3 and s4. Similarly, we used the trait-class of lw1 and lw2 but not lw3 to lw6. This is because it would be redundant to calculate the ecological value based on e.g. s1 and s2 and also on s3 and s4 because they are complementary and would give the same results. The quality values given by the individual metrics were then used to calculate the two trait-based sub-indices as presented in Section 2.4.2. Finally, the two indices were tested on the test database providing significant results both for the nutrient (Idx.M_nutrtrait p < 0.05, r = − 0.47) and the organic/turbidity pollution (Idx.M_orgtrait p < 0.05, r = − 0.54) gradients (Table 3.2, Fig. 3.5). Two sites - the same ones biased in the correlation of the Idx.M above - also appeared here as outliers. After bootstrap analysis, we got CI95 = (− 0.83, − 0.10) for Idx.M_nutrtrait and CI95 = (− 0.71, − 0.34) for Idx.M_orgtrait. Without the outliers, the original correlations are Idx.M_nutrtrait p < 0.05, r = − 0.80 for the nutrient gradient and Idx.M_orgtrait p < 0.05, r = − 0.63 for the organic/turbidity gradient). Bootstrapping gave a value of CI95 = (− 0.96, − 0.70) for Idx.M_nutrtrait and CI95 = (− 0.82, − 0.49) for Idx.M_orgtrait without the outliers. These two outliers show a weak nutrient contamination but strong organic/turbidity pollution; however, their quality value is around 10 in each case.

- 88 -

Pseudo-R2 values of the models fitted on trait-environmental gradient relationships.

1

Nutrients Traits

Guilds

Size-classes

Length-towidth ratio

Trait classes Low High Motile Planktic s1 s2 s3 s4 lw1 lw2 lw3 lw4 lw5 lw6

R2 -0.15* 2.4x10-4 0.17* -6.6x10-4 0.01 0.04 -0.01 -0.06 -0.06* -0.01 0.05 3x10-3 8x10-3 7x10-3

R2 (merged classes)

0.12* -0.12* -0.16*

0.16*

Organic/ turbidity R2 (merged R2 classes) -0.45* 5x10-4 0.34* 4.5x10-4 0.04 0.32* 0.11* -4.9x10-3 -0.32* -0.37* -0.06* -0.50* -0.36* 0.03* 0.08* 0.50* 8.1x10-3 4.6x10-3

Values in bold indicate relationships that were used in the index development. a indicates significant relationship.

Pearson's correlation between the trait-based sub-indices with the nutrient (A) and the organic/turbidity (B) gradients, practically the site locations along the 1 st axes of the CCA analyses (Idx.M_nutrtrait p < 0.05, r = − 0.47, CI95 = (− 0.83, − 0.10) and Idx.M_orgtrait p < 0.05, r = − 0.54, CI95 = (− 0.71, − 0.34), respectively). Without the outliers, the correlations are stronger, r = − 0.80 and r = − 0.63, respectively.

- 89 -

The sub-indices that were developed for the two types of pressures (organic and nutrient) were then combined into one taxonomy-based and one trait-based final index (Idx.M, Idx.Mtrait). The contribution of each sub-index depended of the explanation power of the parameters related to the two gradients in the CCA analyses (Fig. 3.3A–B). 𝐼𝑑𝑥. 𝑀 =

37.85 ∗ 𝐼𝑑𝑥. 𝑀𝑛𝑢𝑡𝑟 + 56.86 ∗ 𝐼𝑑𝑥. 𝑀𝑜𝑟𝑔 37.85 + 56.86

where Idx.M: value of the taxonomy-based quality index Idx.M_nutr: value of the taxonomy-based sub-index for the nutrient gradient Idx.M_org: value of the taxonomy-based sub-index for the organic matter gradient and 𝐼𝑑𝑥. 𝑀𝑡𝑟𝑎𝑖𝑡 =

37.85 ∗ 𝐼𝑑𝑥. 𝑀𝑛𝑢𝑡𝑟 𝑡𝑟𝑎𝑖𝑡 + 56.86 ∗ 𝐼𝑑𝑥. 𝑀𝑜𝑟𝑔

𝑡𝑟𝑎𝑖𝑡

37.85 + 56.86

where Idx.Mtrait: value of the trait-based quality index Idx.M_nutrtrait: value of the trait-based sub-index for the nutrient gradient Idx.M_orgtrait: value of the trait-based sub-index for the organic matter gradient The two indices were then compared and correlated with each other (Fig. 3.6). The relation is significant and strong (p < 0.05, r = 0.67). The distribution of the values, however, differs markedly at the two indices: a concentration of Idx.M values can be observed between 10 and 15, and the Idx.Mtrait values are distributed quite evenly along the entire range (0 − 20).

Even if the implementation of the WFD in Mayotte is required, its rivers are very different from those of the European continent for geographical reasons. In particular, we observed an opposite concentration gradient of the two main nutrient forms: phosphorus and nitrogen. The concentrations of all nitrogen forms (TN, NO3−, NO2−, NH4+) were high when concentrations of phosphorus forms (TP, PO43−) were relatively low. Unimpacted upstream sites commonly had low nitrogen but elevated phosphorus concentrations, and thus a low - 90 -

N-to-P ratio. One reason is that volcanic soils supply nutrient rich water, which becomes exploited by different mechanisms. P is usually the limiting nutrient in aquatic ecosystems and the aquatic vegetation downstream might utilize it. Nevertheless, we rejected this hypothesis since aquatic vegetation including microphytobenthos has a low productivity in Mayotte streams as shown by very thin biofilms.

Pearson's correlation between the classical, taxonomy-based (Idx.M) and trait-based (Idx.Mtrait) indices (pOff. J. Eur. Communities 327, 1–72.

- 98 -

Groendahl, S., Kahlert, M., Fink, P., 2017. The best of both worlds: a combined approach for analyzing microalgal diversity via metabarcoding and morphology-based methods. PLoS One 12, e0172808. http://dx.doi.org/10.1371/journal.pone.0172808. Hajnal, É., Padisák, J., 2008. Analysis of long-term ecological status of Lake Balaton based on the ALMOBAL phytoplankton database. Hydrobiologia 599:227–237. http://dx.doi.org/10.1007/s10750-007-9207-x. Hughes, R.M., Larsen, D.P., Omernik, J.M., 1986. Regional reference sites: a method for assessing stream potentials. Environ. Manag. 10:629–635. http://dx.doi.org/10.1007/BF01866767. Insee, 2016. Estimation de la population au 1er janvier par région, département, sexe et âge de 1975 à 2015 [WWW Document]. URL. http://www.insee.fr/fr/themes/ detail.asp?reg_id=99&ref_id=estim-pop (accessed 6.8.16). Kahlert, M., Albert, R.-L., Anttila, E.-L., Bengtsson, R., Bigler, C., Eskola, T., Gälman, V., Gottschalk, S., Herlitz, E., Jarlman, A., Kasperoviciene, J., Kokociński, M., Luup, H., Miettinen, J., Paunksnyte, I., Piirsoo, K., Quintana, I., Raunio, J., Sandell, B., Simola, H., Sundberg, I., Vilbaste, S., Weckström, J., 2009. Harmonization is more important than experience—results of the first Nordic–Baltic diatom intercalibration exercise 2007 (stream monitoring). J. Appl. Phycol. 21:471–482. http://dx.doi.org/10.1007/s10811-008-9394-5. Kahlert, M., Kelly, M., Albert, R.-L., Almeida, S.F.P., Bešta, T., Blanco, S., Coste, M., Denys, L., Ector, L., Fránková, M., Hlúbiková, D., Ivanov, P., Kennedy, B., Marvan, P., Mertens, A., Miettinen, J., PicinskaFałtynowicz, J., Rosebery, J., Tornés, E., Vilbaste, S., Vogel, A., 2012. Identification versus counting protocols as sources of uncertainty in diatom-based ecological status assessments. Hydrobiologia 695: 109– 124. http://dx.doi.org/10.1007/s10750-012-1115-z. Kalff, J., Knoechel, R., 1978. Phytoplankton and their dynamics in oligotrophic and eutrophic lakes. Annu. Rev. Ecol. Syst. 9, 475–495. Kelly, M.G., 1998. Use of the trophic diatom index to monitor eutrophication in rivers. Water Res. 32:236– 242. http://dx.doi.org/10.1016/S0043-1354(97)00157-7. Kelly, M.G., Whitton, B.A., 1995. The trophic diatom index: a new index for monitoring eutrophication in rivers. J. Appl. Phycol. 7, 433–444. Kelly, M.G., King, L., Jones, R.I., Barker, P.A., Jamieson, B.J., 2008. Validation of diatoms as proxies for phytobenthos when assessing ecological status in lakes. Hydrobiologia 610:125–129. http://dx.doi.org/10.1007/s10750-008-9427-8. Kruk, C., Huszar, V.L.M., Peeters, E.T.H.M., Bonilla, S., Costa, L., Lürling, M., Reynolds, C.S., Scheffer, M., 2010. A morphological classification capturing functional variation in phytoplankton. Freshw. Biol. 55:614– 627. http://dx.doi.org/10.1111/j.1365-2427.2009.02298.x. Lange, K., Townsend, C.R., Matthaei, C.D., 2016. A trait-based framework for stream algal communities. Ecol. Evol. 6:23–36. http://dx.doi.org/10.1002/ece3.1822. Lange-Bertalot, H., 2000. Diatom Flora of Marine Coasts, Iconographia Diatomologica. Koeltz Scientific Books. Lange-Bertalot, H., Metzeltin, D., 2002. Diatoms from the “Island Continent”Madagascar, Iconographia Diatomologica. Koeltz Scientific Books. Lange-Bertalot, H., Bak, M., Witkowski, A., Tagliaventi, N., 2011a. Eunotia and Some Related Genera, Diatoms of Europe. Gantner Verlag. Lange-Bertalot, H., Hofmann, G., Werum, M., 2011b. Diatomeen im Süßwasser – Benthos von Mitteleuropa.: Bestimmungsflora Kieselalgen für die ökologische Praxis. Über 700 der häufigsten Arten und ihre Ökologie. Gantner, A R, Ruggell. Larras, F., Coulaud, R., Gautreau, E., Billoir, E., Rosebery, J., Usseglio-Polatera, P., 2017. Assessing anthropogenic pressures on streams: a random forest approach based on benthic diatom communities. Sci. Total Environ. 586: 1101–1112. http://dx.doi.org/10.1016/ j.scitotenv.2017.02.096. Leblanc, K., Arístegui, J., Armand, L., Assmy, P., Beker, B., Bode, A., Breton, E., Cornet, V., Gibson, J., Gosselin, M.-P., Kopczynska, E., Marshall, H., Peloquin, J., Piontkovski, S., Poulton, A.J., Quéguiner, B., Schiebel, R., Shipe, R., Stefels, J., van Leeuwe, M.A., Varela, M., Widdicombe, C., Yallop, M., 2012. A global diatom database–abundance, biovolume and biomass in the world ocean. Earth Syst. Sci. Data 4:149–165. http://dx.doi.org/10.5194/essd-4-149-2012.

- 99 -

Lengyel, E., Padisák, J., Stenger-Kovács, C., 2015. Establishment of equilibrium states and effect of disturbances on benthic diatom assemblages of the Torna-stream, Hungary. Hydrobiologia 750:43–56. http://dx.doi.org/10.1007/s10750-014-2065-4. Levkov, Z., 2009.Amphora Sensu Lato, Diatoms of Europe. Gantner Verlag. Levkov, Z., Metzeltin, D., Pavlov, A., Lange-Bertalot, H., 2014. Luticola and Luticolopsis, Diatoms of Europe. Gantner Verlag Mackay, A.W., Davidson, T., Wolski, P., Woodward, S., Mazebedi, R., Masamba, W.R.L., Todd, M., 2012. Diatom sensitivity to hydrological and nutrient variability in a sub-tropical, flood-pulse wetland. Ecohydrology 5:491–502. http://dx.doi.org/10.1002/eco.242. Mann, D.G., Vanormelingen, P., 2013. An inordinate fondness? The number, distributions, and origins of diatom species. J. Eukaryot. Microbiol. 60:414–420. http://dx.doi.org/10.1111/jeu.12047. Mary, N., 2016. Sites de référence et sites “pollué”- cours d'eau de Mayotte. Campagne d'etiage (2015). McCormick, P.V., Stevenson, R.J., 1989. Effects of snail grazing on benthic algal community structure in different nutrient environments. American Benthological Society] –> J. N. Am. Benthol. Soc. 8:162–172. http://dx.doi.org/10.2307/1467634. Metzeltin, D., Lange-Bertalot, H., 2007. Tropical Diatoms of South America, II, Iconographia Diatomologica. Gantner Verlag. Metzeltin, D., Lange-Bertalot, H., Garcia, F., 2005. Diatoms of Uruguay, Iconographia Diatomologica. Koeltz ScientificBooks. Mondy, C.P., Villeneuve, B., Archaimbault, V., Usseglio-Polatera, P., 2012. A new macroinvertebrate-based multimetric index (I2M2) to evaluate ecological quality of French wadeable streams fulfilling the WFD demands: a taxonomical and trait approach. Ecol. Indic. 18:452–467. http://dx.doi.org/10.1016/j.ecolind.2011.12.013. Morin, A., Bourassa, N., Cattaneo, A., 2001. Use of size spectra and empirical models to evaluate trophic relationships in streams. Limnol. Oceanogr. 46:935–940. http://dx.doi.org/10.4319/lo.2001.46.4.0935. Nõges, P., Mischke, U., Laugaste, R., Solimini, A.G., 2010. Analysis of changes over 44 years in the phytoplankton of Lake Võrtsjärv (Estonia): the effect of nutrients, climate and the investigator on phytoplankton-based water quality indices. Hydrobiologia 646:33–48. http://dx.doi.org/10.1007/s10750010-0178-y. Padisák, J., Borics, G., Grigorszky, I., Soróczki-Pintér, É., 2006. Use of phytoplankton assemblages for monitoring ecological status of lakes within the water framework directive: the assemblage index. Hydrobiologia 553:1–14. http://dx.doi.org/10.1007/s10750-005-1393-9. Passy, S.I., 2007. Diatom ecological guilds display distinct and predictable behavior along nutrient and disturbance gradients in running waters. Aquat. Bot. 86:171–178. http://dx.doi.org/10.1016/j.aquabot.2006.09.018. Potapova, M.G., Charles, D.F., Ponader, K.C., Winter, D.M., 2004. Quantifying species indicator values for trophic diatom indices: a comparison of approaches. Hydrobiologia 517, 25–41. Prygiel, J., Coste, M., 1998. Mise au point de l'Indice Biologique Diatomée, un indice diatomique pratique applicable au réseau hydrographique français. L'Eau, l'industrie, les nuisances 40–45. R Development Core Team, 2008. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Reavie, E.D., Kireta, A.R., Kingston, J.C., Sgro, G.V., Danz, N.P., Axler, R.P., Hollenhorst, T.P., 2008. Comparison of simple and multimetric diatom-based indices for Great Lakes coastline disturbance. J. Phycol. 44:787–802. http://dx.doi.org/10.1111/j.1529-8817.2008.00523.x. Reynolds, C.S., 1980. Phytoplankton assemblages and their periodicity in stratifying lake systems. Ecography 3:141–159. http://dx.doi.org/10.1111/j.1600-0587.1980.tb00721.x. Reynolds, C.S., 2006. Ecology of Phytoplankton. First. ed. Cambridge University Press, NewYork. Reynolds, C.S., Huszar, V., Kruk, C., Naselli-Flores, L., Melo, S., 2002. Towards a functional classification of the freshwater phytoplankton. J. Plankton Res. 24, 417–428. Rimet, F., Bouchez, A., 2012. Life-forms, cell-sizes and ecological guilds of diatoms in European rivers. Knowl. Manag. Aquat. Ecosyst. 01.http://dx.doi.org/10.1051/kmae/2012018.

- 100 -

Rimet, F., Gomà, J., Cambra, J., Bertuzzi, E., Cantonati, M., Cappelletti, C., Ciutti, F., Cordonier, A., Coste, M., Delmas, F., Tison, J., Tudesque, L., Vidal, H., Ector, L., 2007. Benthic diatoms in western European streams with altitudes above 800 M: characterisation of the main assemblages and correspondence with Ecoregions. Diatom Res. 22:147–188. http://dx.doi.org/10.1080/0269249X.2007.9705702. Schneider, S.C., Hilt, S., Vermaat, J.E., Kelly, M., 2016. The “Forgotten” Ecology Behind Ecological Status Evaluation: Re-Assessing the Roles of Aquatic Plants and Benthic Algae in Ecosystem Functioning. Springer, Berlin Heidelberg, Berlin, Heidelberg. Soininen, J., Jamoneau, A., Rosebery, J., Passy, S.I., 2016. Global patterns and drivers of species and trait composition in diatoms: global compositional patterns in stream diatoms. Glob. Ecol. Biogeogr. http://dx.doi.org/10.1111/geb.12452 (n/a-n/a). Squires, L.E., Rushforth, S.R., Brotherson, J.D., 1979. Algal response to a thermal effluent: study of a power station on the Provo River, Utah, USA. Hydrobiologia 63, 17–32. Stancheva, R., Sheath, R.G., Read, B.A., McArthur, K.D., Schroepfer, C., Kociolek, J.P., Fetscher, A.E., 2013. Nitrogen-fixing cyanobacteria (free-living and diatom endosymbionts): their use in southern California stream bioassessment. Hydrobiologia 720:111–127. http://dx.doi.org/10.1007/s10750-013-1630-6. Stenger-Kovács, C., Lengyel, E., Crossetti, L.O., Üveges, V., Padisák, J., 2013. Diatom ecological guilds as indicators of temporally changing stressors and disturbances in the small Torna-stream, Hungary. Ecol. Indic. 24:138–147. http://dx.doi.org/10.1016/j.ecolind.2012.06.003. Stenger-Kovács, C., Lengyel, E., Buczkó, K., Tóth, F., Crossetti, L., Pellinger, A., Zámbóné Doma, Z., Padisák, J., 2014. Vanishing world: alkaline, saline lakes in Central Europe and their diatom assemblages. Inland Waters 4:383–396. http://dx.doi.org/10.5268/IW-4.4.722. Stevenson, R.J., Bahls, L.L., 2002. Periphyton protocols. Rapid Bioassessment Protocols for Use in Streams and Wadeable Rivers: Periphyton, Benthic Macroinvertebrates, and Fish. US, EPA, pp. 1–23. Stevenson, R.J., Pan, Y., Van Dam, H., 2010. Assessing environmental conditions in rivers and streams with diatoms. The Diatoms: Applications for the Environmental and Earth Sciences. Cambridge University Press, Cambridge. Stoddard, J.L., Larsen, D.P., Hawkins, C.P., Johnson, R.K., Norris, R.H., 2006. Setting expectations for the ecological condition of streams: the concept of reference condition. Ecol. Appl. 16, 1267–1276 (doi:10.1890/1051-0761(2006)016[1267:SEFTEC]2.0.CO;2). Straile, D., Jochimsen, M.C., Kümmerlin, R., 2013. The use of long-term monitoring data for studies of planktonic diversity: a cautionary tale from two Swiss lakes. Freshw. Biol. 58:1292–1301. http://dx.doi.org/10.1111/fwb.12118. Tall, L., Cloutier, L., Cattaneo, A., 2006. Grazer-diatom size relationships in an epiphytic community. Limnol. Oceanogr. 51, 1211–1216. Tang, T., Niu, S.Q., Dudgeon, D., 2013. Responses of epibenthic algal assemblages to water abstraction in Hong Kong streams. Hydrobiologia 703:225–237. http://dx.doi.org/10.1007/s10750-012-1362-z. Tapolczai, K., Bouchez, A., Stenger-Kovács, C., Padisák, J., Rimet, F., 2016. Trait-based ecological classifications for benthic algae: review and perspectives. Hydrobiologia 776: 1–17. http://dx.doi.org/10.1007/s10750-016-2736-4. Usseglio-Polatera, P., Bournaud, M., Richoux, P., Tachet, H., 2000. Biomonitoring Through Biological Traits of Benthic Macroinvertebrates: How to Use Species Trait Databases? 422/423 pp. 153–162 Usseglio-Polatera, P., Richoux, P., Bournaud, M., Tachet, H., 2001. A functional classification of benthic macroinvertebrates based on biological and ecological traits: application to river condition assessment and stream management. Arch. Hydrobiol. 139, 53–83 Supplement band. Monographische Beiträge. Utermöhl, H., 1958. Zur Vervollkommnung der quantitativen Phytoplankton-Methodik. Mitt. Int. Ver. Theor. Angew. Limnol. 9, 1–38. Vasselon, V., Domaizon, I., Rimet, F., Kahlert, M., Bouchez, A., 2017. Application of high-throughput sequencing (HTS) metabarcoding to diatom biomonitoring: do DNA extraction methods matter? Freshw. Sci. 36:162–177. http://dx.doi.org/10.1086/690649. Violle, C., Navas, M.-L., Vile, D., Kazakou, E., Fortunel, C., Hummel, I., Garnier, E., 2007. Let the concept of trait be functional! Oikos 116:882–892. http://dx.doi.org/10.1111/j.0030-1299.2007.15559.x.

- 101 -

Werner, P., Adler, S., Dreßler, M., 2016. Effects of counting variances on water quality assessments: implications from four benthic diatom samples, each counted by 40 diatomists. J. Appl. Phycol. 28:2287– 2297. http://dx.doi.org/10.1007/s10811-015-0760-9. Wetzel, C.E., Ector, L., Van de Vijver, B., Compere, P., Mann, D.G., 2015. Morphology, typification and critical analysis of some ecologically important small naviculoid species (Bacillariophyta). Fottea 15:203–234. http://dx.doi.org/10.5507/fot.2015.020. Zelinka, M., Marvan, P., 1961. Zur präzisierung der biologischen klassifikation der reinheitflie\s sender gewässer. Arch. Hydrobiol. 57, 389–407. Zhang, B., Wu, D., Wang, C., He, S., Zhang, Z., Kong, H., 2007. Simultaneous removal of ammonium and phosphate by zeolite synthesized from coal fly ash as influenced by acid treatment. J. Environ. Sci. 19:540– 545. http://dx.doi.org/10.1016/S1001-0742(07)60090-4. Zhang, Y.Y., Wang, L.P., Du, E.D., Chen, Y.Z., 2011. The removal of phosphorus in wastewater by modified 4A zeolite. Adv. Mater. Res. 356–360:502–505. http://dx.doi.org/10.4028/www.scientific.net/AMR.356360.502. Zimmermann, J., Glöckner, G., Jahn, R., Enke, N., Gemeinholzer, B., 2015. Metabarcoding vs. morphological identification to assess diatom diversity in environmental studies. Mol. Ecol. Resour. 15:526–542. http://dx.doi.org/10.1111/1755-0998.1233

- 102 -

Chapter 4

This chapter is the self-edited version of the following article: Tapolczai, K., V. Vasselon, A. Bouchez, C. Stenger-Kovács, J. Padisák & F. Rimet, 2017. Taxonomy-free DNA biomonitoring for rivers: How to choose the sequence similarity threshold? Submitted article to Molecular Ecology Resources.

- 104 -

Abstract High Throughput sequencing (HTS) and DNA-metabarcoding enables to assess rapidly the ecological status of aquatic ecosystems. However, this technology lacks standardization and still carries biases. One of the most crucial being the reference libraries which incompleteness disable a proper identification of all environmental sequences. To overcome this limitation, taxonomy-free indices can be developed in order to use all environmental sequences. Our objectives were to (i) develop and test a solely OTU-based index without taxonomic assignment and to (ii) assess the impact on biomonitoring efficacy of different Sequence Similarity Thresholds (SST: 80%-99%) used to cluster sequences into OTUs. In our study, we used diatom samples from a rivers monitoring network (Mayotte Island, France, Indian-Ocean). The taxonomy-free index showed significant correlations with the pollution gradient whatever the SST used. Nevertheless, this efficiency reached a plateau at 91% SST that is below the currently considered species-level delimitation for diatoms (95-97%). We showed an important trade-off between index stability (more stable for coarse resolution) and discrimination power along the pollution gradient (better discrimination for finer resolution) that may set a challenge for water managers when deciding which tool to develop for routine bioassessment with metabarcoding.

Keywords Diatoms, High throughput sequencing, OTU, Quality assessment, Water Framework Directive

- 105 -

Benthic diatom communities are widely used as ecological indicators of water bodies as they have short generation times, large species diversity (Mann & Vanormelingen, 2013) and sensitivity to many kinds of environmental changes (M. G. Kelly, 1998), etc. This is the reason why phytobenthos, of which benthic diatoms are good proxy, is one of the five biological quality elements (BQEs) required by the European Water Framework Directive (WFD) to assess ecological quality of water bodies (European Commission, 2000). Diatombased indices mostly use the Zelinka-Marvan equation (Zelinka & Marvan, 1961) based on the species’ abundance profiles along particular environmental gradients (Coste, 1982; Coste, Boutry, Tison-Rosebery, & Delmas, 2009; M. G. Kelly, 1998). Routine methodologies require microscopic species identification based on their morphological attributes, which is a time-consuming process requiring experienced analysts. Moreover, microscopic species identifications often carries uncertainties that can bias the ecological quality assessment (Kahlert et al., 2012). Misidentifications are common causing disharmony in datasets between labs thus requiring regular harmonization (Kahlert et al., 2009). Additionally, inconsistencies in species identification are common because of e.g. changes in the taxonomic literature or/and the taxonomic expertise of the counting staff during a long-term monitoring project (Straile, Jochimsen, & Kümmerlin, 2013). Finally, many species can even suffer a taxonomic shift (species splitting, lumping, or shifts in their taxonomic boundaries) over their nomenclature history (Wetzel, Ector, Van de Vijver, Compere, & Mann, 2015) and this can have a direct effect on the results of ecological assessment. The rapid development of molecular techniques and the set-up of DNA barcoding brought a great change in identification accuracy (Hebert, Cywinska, Ball, & others, 2003). This method uses standard markers to identify taxa-specific DNA sequences (DNA barcodes) extracted from organisms. DNA-barcoding as a complementary method have been often used in studies where identification based on morphological features was problematic (Taberlet, Coissac, Pompanon, Brochmann, & Willerslev, 2012). The rapidly changing taxonomy of diatoms is a very good example for the advantage of using such techniques in their identification (Mann, 1999). The further development of High Throughput Sequencing (HTS) technologies, together with DNA-metabarcoding, makes the simultaneous analysis of multiple taxa from multiple environmental samples possible (Taberlet et al. 2012a; b). DNA-based methods have been shown to avoid misidentifications that come from phenotypic plasticity, morphologically cryptic taxa or different life-stages (that latter one has less relevance among algae). In addition, DNA-metabarcoding with HTS - 106 -

technologies makes it a fast, cost-effective technique permitting routine analyses of monitoring network samples. The huge amount of information (Big data) gained by this method compared to classical morphology-based identification has the potential of revolutionizing ecological quality assessment (Baird & Hajibabaei, 2012; Keck, Vasselon, Tapolczai, Rimet, & Bouchez, 2017). However, the reliability of molecular techniques depends on several factors that should be adjusted and standardized before their routine use in quality assessment. First is the choice of an appropriate DNA barcode, as shown by Kermarrec et al. (2013). In their study, three gene markers (SSU rDNA, rbcL, cox1) were tested with pyrosequencing (454) and it was found that rbcL from the chloroplast genome proved to be the most efficient for the identification of diatoms at species level. Second is the choice among HTS technologies, which also differ in several aspects: price, cost per run, minimum throughput (read length), run time, error rate, etc. (Loman et al., 2012; Shokralla, Spall, Gibson, & Hajibabaei, 2012). Third, a standardization of the wet-lab processes is in need, among which DNA extraction (Vasselon, Domaizon, Rimet, Kahlert, & Bouchez, 2017). In a final step comes the bioinformatics process of the HTS sequence data from which the list of DNA sequences from each sample is obtained. The huge amount of sequences need to be clustered into Molecular Operational Taxonomic Units (MOTU/OTU) based on a pre-defined sequence similarity threshold (SST), showing in what percent the sequences are identical, in order to avoid the analyses of sequencing artifacts and also to reduce data volume (Coissac, Riaz, & Puillandre, 2012; Patin, Kunin, Lidström, & Ashby, 2013). The SST can be also considered as a proxy of taxonomy resolution of our data. The commonly used SST values are between 95 and 97% (Edgar, 2013; Elbrecht & Leese, 2015; Huse, Welch, Morrison, & Sogin, 2010; Patin et al., 2013; Yu et al., 2012) or even 99% (Apothéloz-Perret-Gentil et al., 2017a) that is considered as a quasi-species delimitation level (Birtel, Walser, Pichon, Bürgmann, & Matthews, 2015; P. D. Schloss & Handelsman, 2005). However, the effect of this threshold on biomonitoring efficiency value has never been studied in detail. In order to use such data for biomonitoring purposes, two main approaches exist. In the first one, OTUs are assigned to taxonomic names using a reference database such as RSyst::diatom (Rimet et al., 2016). Since OTUs are groups of sequences, usually the most abundant sequence in an OTU is used to pair it with a taxonomic name (Patin et al., 2013). Several studies used this method (Groendahl, Kahlert, & Fink, 2017; Vasselon, Rimet, Tapolczai, & Bouchez, 2017; Visco et al., 2015), however always facing the same problem:

- 107 -

the incompleteness of reference libraries. Therefore, only part of the data can be used which do not cover the entire diversity representing in the study site (Rivera et al., 2017). Thus, the index calculation is based on only a part of the species while the others are unidentified among which dominant species may be encompassed that are important for ecological assessment. The other approach is to use of OTUs directly in quality indices without taxonomic assignment. This alternative approach in diatom-based quality assessment efficiently avoids the problematics with incomplete reference libraries (Apothéloz-PerretGentil et al., 2017a). Finally, to assess a quality index value to a sampled site, in the first approach, ecological profiles defined for each species are used, while in the second approach, values have to be defined for each OTU. In our study, we test a diatom-based quality index based on the ecological profile of OTUs (IdxOTU), ignoring their taxonomic affiliation. We hypothesize that the choice of a correct SST (e.g. 80%, 95% or 99%) to group sequences into OTUs is important in order to get the right taxonomic resolution of environmental diatom communities for bioindication purposes. In case it is too low, it would result in a low number of OTUs, corresponding to a low taxonomic resolution, which could results in low biomonitoring efficiency. At the opposite, too high SST would result in a huge number of OTUs that may be even under species level (e.g. population level) with little interest for biomonitoring purpose. The developed index was tested at twenty SSTs from 80% to 99%, on 90 samples from a WFD river monitoring network. We compared the effect of these similarity thresholds on the final quality index in several aspects: (i) number of OTUs and taxonomic resolution, (ii) number of environmental factors the index can predict, (iii) correlation between the index values calculated for a test database and the environmental gradient, (iv) discrimination power of index and (v) index stability.

The French oversea department, Mayotte, is located in the Indian Ocean, as a part of the Comoros archipelago, to the north-west of Madagascar (12°50′35″S 45°08′18″E) (Supplement 1). After the change in its legal status in 2011, the application of the Water Framework Directive (WFD) became obligatory on its water bodies. For this purpose, three monitoring networks (regular monitoring network – RCS; polluted sites network – POLL; and reference sites network – REF) have been monitored (Tapolczai, Bouchez, Stenger-

- 108 -

Kovács, Padisák, & Rimet, 2017; Vasselon, Rimet, et al., 2017). Our data come from these three monitoring networks where 90 samples of phytobenthos were collected from 51 river sites in 2014 and 2015, along with their supporting physical and chemical data.

The phytobenthos sampling procedure followed French and European standards (Afnor, 2014, 2016) and was carried out in the dry season (July-August) when conditions are more stable than in the rainy season with its strong floods. The samples were collected from at least five stones in the lotic part of the river. A clean toothbrush was used to brush the biofilm from the surface of the stones. The samples were then preserved by adding 99% ethanol for a final ethanol concentration of about 70%. Sampling and analysis of physical and chemical parameters was carried out in the same time period following APHA standards (APHA., 1976). Parameters used in the analyses of our study were those that are related to the nutrients and organic matter which were described as the main environmental gradients in Mayotte (Tapolczai et al., 2017); total phosphorus (TP), phosphate (PO43-), ammonium (NH4+), nitrate (NO3-), nitrite (NO2-), suspended solids (SS), turbidity, total organic carbon (TOC), dissolved organic carbon (DOC).

Total DNA was extracted from 2 mL of phytobenthos samples using the GenEluteTM-LPA method DNA extraction. Detailed protocol can be found in former studies (Chonova et al., 2016; L. Kermarrec et al., 2013). This method was recommended for diatom metabarcoding (Vasselon, Domaizon, et al., 2017) because it uses various lysis mechanisms (mechanical, enzymatic, heat), which facilitate diatom cell disruption and DNA recovery. A short region of the rbcL gene (312 bp) was used as DNA marker and amplified by PCR using an equimolar mix of the modified version of the Diat_rbcL_708F forward and the R3 reverse primers (Rimet et al., 2018; Vasselon, Rimet, et al., 2017). PCR amplification of each DNA sample was performed in triplicate using 1µL of extracted DNA in a final reaction volume of 25 µL. Detailed PCR reaction mix and PCR amplification conditions are summarized in Supplement 2. The three PCR replicates obtained for each DNA sample were pooled and purified using Agencourt AMPure beads (Beckman Coulter, Brea, USA). Quality and quantity of

- 109 -

purified amplicons were checked using the 2200 TapeStation (Agilent technologies, Santa Clara, USA). Following library preparation method described by Vasselon et al. (2017a), individual A-X tag adapter (Ion ExpressTM Barcode adapters, Life Technologies, Carlsbad, USA) were added to amplicons by ligation using the NEBNext® Fast DNA Library Prep set for Ion TorrentTM (BioLabs, Ipswich, USA). Samples libraries were pooled in two mixes corresponding to Mayotte 2014 and 2015 sampling campaigns that contained 49 and 41 samples, respectively. Each mix was prepared with a final concentration of 100 pm and sequenced independently on 2 Ion 318TMChip Kit V2 (Life Technologies, Carlsbad, USA) using the PGM Ion Torrent machine. Sequencing was performed by the “Plateforme Génome Transcriptome” (PGTB, Bordeaux, France) that provided one fastq file per sample for the 90 libraries with demultiplexed DNA reads. Quality trimming was applied to remove low quality DNA reads as described by Vasselon et al. (2017b). The 90 trimmed files were merged in order to treat together all the samples following the bioinformatics process described in Vasselon et al. (2017b) using the Mothur software (Patrick D. Schloss et al., 2009). In addition to this bioinformatics process, DNA reads were dereplicated in order to obtain Independent Sequence Units (ISUs). Abundance of ISUs, corresponding to the number of sequence replicate per ISU, was used to remove ISUs with less than one replicate. Retained ISUs were then clustered in OTUs using different SSTs ranging from 80% to 99%. Finally, 20 OTU lists, corresponding to each threshold and including the number of DNA reads within the 90 samples were produced. A taxonomy was assigned to each OTU using the consensus taxonomy of DNA reads determined by Mothur (classify.otu command, Schloss et al. 2009) and the R-Syst::diatom library (Rimet et al. 2016, 13-02-2015: R-Syst::diatom v3, http://www.rsyst.inra.fr/en) with a consensus confidence threshold over 80%. Fastq files with demultiplexed DNA reads, the final OTU lists (80-99%) (including DNA reads proportion, DNA representative sequence of each OTU and the OTU taxonomic assignment) as well as the sampling site description are available for all the samples on the Zenodo repository website (http://doi.org/110.5281/zenodo.802608).

A schematic flowchart (Fig 1) summarizes the process of the OTU-based index development. We first produced 20 different OTU inventories for each samples (90 samples from 51 river sites per table) using 20 different SSTs (from 80% to 99%) to group sequences into OTUs. Rare OTUs that were present in less than the 5% of the samples were removed

- 110 -

from the OTU lists in order to have enough data points to define a stable ecological profile for each OTU. We then produced 20 different samples-to-OTU list tables. In order to define the pressure gradient, a canonical correspondence analysis (CCA) was run for each of these 20 datasets using the R vegan package (Jari Oksanen et al., 2016; R Development Core Team, 2008). Forward selection (pstep= 1000) on the environmental parameters was run for each CCA, in order to keep only the relevant ones. The 1st axis of each CCA represents the pressure gradient. Then each of the 20 datasets was randomly divided into a training dataset on which an OTU-index was developed (75% of the samples) and a test dataset on which this OTU-index was tested (25% of the samples). The randomization however respected the ratio of the three sampling networks in order to keep a reasonable range of the pressure gradients. For each SST, the randomization and index development process were executed 100 times to avoid falling into extreme cases. The whole process results in 100 indices tested for each of the 20 (SSTs) dataset; 2000 indices in total.

Schematic representation of the statistical analyses and index development process.

- 111 -

The Zelinka-Marvan equation (Zelinka & Marvan, 1961) was adapted on our data to define IdxOTU: 𝐼𝑑𝑥𝑂𝑇𝑈

∑𝑛𝑗=1 𝑎𝑗 𝑠𝑗 𝑣𝑗 = ∑𝑛𝑗=1 𝑎𝑗 𝑣𝑗

where aj = relative abundance of OTU j, sj = sensitivity value or optimum of OTU j and vj = indicator value or tolerance of OTU j in the sample. Sensitivity and indicator values for each OTU were calculated by plotting their ecological profile, i.e. their relative abundance in function of the position of sites along the 1st axis of the CCA analysis. Only data from the training dataset was used to define these profiles. Weighted average (optimum) as sensitivity value and the reciprocal of weighted standard deviation (tolerance) value as indicator value were calculated for each OTU. IdxOTU was calculated for each site in the test dataset and correlated with their position on the pressure gradient (1st axis of CCA).

Among the 20 OTU lists that were defined, the number of OTUs increased exponentially with the increasing SST (Fig 2A). The number of OTUs range from 159 at 80% SST to 15296 at 99% SST. Common OTUs, those that were present in >5% of the samples showed similar trend, however, the ratio of the removed rare OTUs increased too: at 99%, more than 60% of the OTUs, whereas at 80% only 18% of the OTUs were removed. By assigning taxonomy to the OTUs, we could see that the taxonomic resolution changed strongly in function of the SST (Fig 2B). From 80% to 93%, the proportion of “unclassified” OTUs varied between 45-50%, and then showed a strong decrease for SST higher than 93% (Fig 2B). The proportion of OTUs identified at species level showed an opposite trend. Up to 90%, its proportion was around 25%, followed by a strong increase and reached about 50% at 99% of SST (Fig 2A). The proportion of OTUs identified at genus level did not show such a clear pattern and varied between 25 and 30% along the gradient (Fig 2A).

- 112 -

Number of OTUs (A) at each SST before (black) and after (grey) omitting rare species (present in less than 5% of the samples). Taxonomic affiliation of the OTU lists at each SST (B): proportion of OTUs identified at species and genus level and the proportion of unclassified OTUs (not present in the reference database or identified at a taxonomic level higher than genus).

For each SST, a CCA was carried out with a forward selection process to define an environmental gradient. The set of significant environmental parameters varied according to SST used to define the set of OTUs (Fig 3). Nutrient forms were the most important factors in the CCA analyses, with the phosphorus forms and total nitrogen as the first two important ones that were significant at each SST, except for TN at 89%. Beside nutrients, dissolved organic carbon and turbidity appeared significant in two (80%, 82% SST) and one (89% SST) cases respectively. There is an increase in the number of significant parameters with increasing SST. While under 93% of SST, the number of significant factors vary between three and four, above 93%, it is constantly five (Fig 3).

- 113 -

Parameters considered significant in the CCA analyses after forward selection at each SST.

Indices were developed based on each of the 20 training dataset. These indices were then used with the corresponding test dataset and correlated with site positions along the environmental gradient. A linear model was fitted on these relations and their linear regression coefficients (R2) represent the index’s efficacy (Fig 4). Since at each tested SST, the training and test database were randomly selected 100 times, each boxplot contains the R2 value of 100 linear regressions performed at the same SST. The range of the R² values is always high, due to a few extreme cases occurring during the randomization process. The index’s efficacy shows an increase between 85 and 92-93%, and then it reaches a plateau with a median R2 value of about 0.56. The number of cases when the model’s p-value falls in a particular range is also presented on Fig 4 (upper table). Except a few cases at lower SST (8087%), the relation between index values on test data and their position on the pressure gradient was significant. Above 90% SST, the models were strongly significant (p< 0.01) in 95 to 97% of the cases.

- 114 -

Index values calculated on the test database were correlated with their position along the environmental gradient. Linear regression coefficients (R 2) are presented for each SST (bottom). Labels on the top of the boxplots show significance; same letters indicate non-significant difference. Top table presents the number of cases of significance (ns – non-significant, p < 0.05 – significant).

The index values, giving the site quality assessment, were ordered for all sites according to increasing median index values (all SST and all randomization set for each site) (Fig 5A). Polluted sites appear to be better discriminated than the others. The effect only of the 100 randomization step on the index value of each site obtained for each SST is presented on Fig 5B. There is an instability in site assessment for the most polluted samples and the reference samples. Mainly the polluted sites showed a strong variation (standard deviation – SD) in their index values among the randomization steps but this variation was also higher when higher SSTs were used to define OTUs. Three groups of SST with different variations can be recognized: 80-85%, 86-92% and 93-99%. Variability of site assessment in the middle of the pressure gradient was not remarkable, except a few sites. The index’s ability to discriminate sites according to their position on the pressure gradient is presented on Figure 6. SD of index values was calculated as a proxy of discrimination power among sites at each randomization step and SST. An increase can be observed along the SST gradient with a steep transitional zone at 86-92% after which it

- 115 -

reaches a plateau and stops increasing (Fig 6A). However, together with the increase of discrimination power, an increase of its interquartile range (IQR) can be observed, too (Fig 6A). This value can be represented as the index’s robustness/stability against the randomization process. It shows an important trade-off between discrimination power and reliability of the index. Figure 6B shows the ratio of discrimination power-to-IQR and a peak can be observed at 89% where the index has a relatively high discrimination power and can be still considered stable.

Calculated IdxOTU values on the test-dataset ordered by median for each site within all SST and randomization step (A). Variability (SD) of index values for each site and SST (B). Darker cells represents higher variability.

- 116 -

Discrimination power (standard deviation of index values at each standardization step calculated for the test-dataset) in function of SST (A). Median-to-interquartile range ratio as a proxy of discrimination power-to-instability (B).

The correlation between OTU richness (number of OTUs) and index variability (SD) is presented on Figure 7. There is a negative correlation at each SST and the relation is stronger when the SST is higher. The sites with high variability are typically the polluted or reference sites with low OTU richness. This can explained by what was observed on Figure 5.

In our study, we developed a diatom-index, based on the same principles as the classical ones, e.g. PSI (Coste, 1982), BDI (Coste et al., 2009), TDI (M. G. Kelly, 1998). However, we used directly the ecological profile of OTUs, instead of species’ profiles avoiding the problematics of incomplete DNA reference libraries. Though the number of species in the DNA reference libraries are increasing (Rimet et al., 2016, 2018), the proportion of OTUs that can be assigned to species level is still far from satisfying (e.g. 35.7% in Vasselon et al. (2017b), 35% in Apothéloz-Perret-Gentil et al. (2017), 23% in Rivera et al. (2017)). Our approach is similar to those of (Apothéloz-Perret-Gentil et al., 2017a; Cordier et al., 2017)

- 117 -

but the method to define the OTUs’ indicator and sensitivity values used in the ZelinkaMarvan equation (Zelinka & Marvan, 1961) is different. In the mentioned studies, sites were pre-classified using the original Swiss morphology-based index and the OTU indicator and sensitivity values depended on their frequency in these classes. In our study, we used directly the environmental pressure gradient of physical and chemical parameters defined by a multivariate analyses. Our method thus is completely independent from morphological taxonomy-based methods. The disadvantage however, is that rare OTUs had poor ecological profiles and had to be removed in order to keep only OTUs with robust ones.

Correlation between OTU richness (number of OTUs) (x-axis) and the standard deviation of index values due to randomization (y-axis) for each SST.

- 118 -

The first result of our study is that the OTU-based index performs well. Indeed, the correlation of index values on the test database correlated significantly with the pressure gradient, avoiding the taxonomic assignment step. Several studies referred to the problematics of having incomplete DNA reference libraries, causing serious biases (Apothéloz-Perret-Gentil et al., 2017a; Groendahl et al., 2017; Lenaïg Kermarrec et al., 2014; Rivera et al., 2017; Visco et al., 2015; Zimmermann, Glöckner, Jahn, Enke, & Gemeinholzer, 2015).

The main purpose of this study was to analyse the effect of the SST on a taxonomy-free OTU-based index from different biomonitoring aspects (Fig 8). We could define three different ranges of SST (80-86, 87-92 and 93-99%), which showed distinct properties. Four parameters out of six (number of environmental parameters the index assesses, model efficacy, index’s discrimination power between reference/polluted sites, index instability) showed a sigmoid pattern with a strong increase within the 87-92% SST range. The number of OTUs (both all and common) and the taxonomic resolution (proportion of OTUs identified at species level and proportion of unclassified OTUs) however, increased exponentially from 80 to 99%. The relevance of our study is that the choice of the SST has an important impact on different properties of biomonitoring efficiency. These different properties are discussed in detail in the following paragraphs. The choice of the correct taxonomic resolution in ecological studies is a common and active debate in biomonitoring whatever kind of biota are studied e.g. macroinvertebrates or microalgae (Bowman & Bailey, 1997; Greffard, Saulnier-Talbot, & Gregory-Eaves, 2011; Schmidt-Kloiber & Nijboer, 2004; Thompson & Townsend, 2000). However, the question is more relevant for microscopic organisms where microscopic identification at species level is difficult and/or labor intensive (Carneiro, Bini, & Rodrigues, 2010; Lavoie, Dillon, & Campeau, 2009; Rimet & Bouchez, 2012). The strongest argument for a precise taxonomic resolution (i.e. species-level) is that since species are the basic units of ecosystems, they thus potentially give the most valuable information if their ecological niches are clearly defined (Salmaso, Naselli-Flores, & Padisák, 2015). DNA-barcoding brought a big change in identification accuracy of such organisms, since much more OTUs are obtained than the number of species identified by microscopy (Keck et al., 2017; Vasselon, Rimet, et al., 2017; Zimmermann et al., 2015). Commonly, a sequence similarity of

- 119 -

95% is used for species-level delimitation for metabarcoding diatoms studies (Vasselon, Domaizon, et al., 2017). In this study, 1239 OTUs at 95% similarity were identified which is clearly much more than the 382 species identified in microscopy (Tapolczai et al., 2017). This difference is greatly due to the recognition of cryptic diversity that is common in diatoms; indeed it was shown for several species that genetic diversity was richer than the morphological diversity (Evans, Wortley, Simpson, Chepurnov, & Mann, 2008; Mann et al., 2004; Souffreau et al., 2013). Moreover, the recognition of such cryptic species can be important because their ecological niches may differ even if they live in sympatry (Martyn G. Kelly, Trobajo, Rovira, & Mann, 2015; Rovira, Trobajo, Sato, Ibáñez, & Mann, 2015; Vanelslander et al., 2009). Our study however, showed that the IdxOTU efficiency reached a plateau at 91%, thus applying the commonly used SST of 95% did not improve our results (Fig 4). Several previous studies showed that species-level accuracy is often not necessary in bioassessment. In their study, Rimet & Bouchez (Rimet & Bouchez, 2012) did not find significant difference in the correlation between diatom assemblage and environmental parameters when changing taxonomic resolution from species to order level. Similar results in (Lavoie et al., 2009) showed that a diatom genus-based index was able to separate reference sites from impacted ones as well as a species-based index. This suggests that the extra information gained by improving taxonomic resolution is needed when the aim is the indication of the effect of e.g. high level of habitat heterogeneity. As presented in the section Results, the number of predictable factors increased until 93% where it reached the maximum (five predictable physical and chemical parameters). The increasing discrimination power of the IdxOTU was probably due to the extra information gained from the increasing taxonomic resolution. However, similarly to the efficiency, it also reached its maximum value at 92% above which increasing the SST did not have any effect. On the other hand, stability of IdxOTU against the randomization process decreased with SST. It is due to the fact that with the exponentially increasing number of OTUs, many of them become less frequent, with less robust ecological profiles. The impact of this is that a higher number of rare OTUs have to be removed from the training dataset. It makes the dataset very sensitive to the random selection of training and test datasets. On the contrary, coarse taxonomic resolution generates less OTUs with wider but more robust ecological profiles and this results in a more stable IdxOTU but with weaker discrimination power. This instability was even more important when sites presented low OTU richness. Low OTU richness was observed for strongly polluted sites, where only a few resistant species could

- 120 -

survive (Archibald, 1972; Blanco et al., 2012; R. J. Stevenson, Pan, & Van Dam, 2010; R. Jan Stevenson, Hill, Herlihy, Yuan, & Norton, 2008), and in reference sites where nutrient limitation selects only for a limited number of taxa (Blanco et al., 2012; R. Jan Stevenson et al., 2008). A technical drawback of using high similarity threshold is the higher risk of giving biological sense to artifacts/error sequences in our data, biasing our results (Patin et al., 2013).

Summary representing the change of the different parameters that have been analysed in the study in three ranges of SST.

We can conclude that our OTU-based index to indicate pollution on Mayotte rivers reached its maximal efficiency at a coarser taxonomic resolution (91% SST) than species-level (9597% SST). An important trade-off was found between the discrimination power and the stability of the index value. This study clearly shows that a compromise must be found between discrimination power (given for the highest SST) and robustness and stability (given for lower SST) of ecological assessment. Water managers have to choose between advantages. In our case, the best compromise is at 89% SST but this value may change from one study region to another or with other pressures to indicate. The index we developed avoids taxonomic assignment to OTUs, however speciesspecific data carry important ecological or evolutionary information (e.g. recognition of species flocks; (Kociolek, Hamsher, Kulikovskiy, & Bramburger, 2017). For instance, the identification of diatom traits requires microscopic analysis. Trait studies also reduce

- 121 -

taxonomic resolution by grouping species based on their functional traits that indicate certain environmental pressures (B-Béres et al., 2017, 2016, Tapolczai et al., 2017, 2016). While the trait-based approach groups species regardless their phylogenetical position, the above studied OTU-based method relies on the genetic similarity. Thus, it is important to state that the aim is not to replace microscopic analyses of species with molecular data. Studies combining the advantages of the both approaches, ecologically meaningful classification of species and their accurate identification, should provide important information for future bioassessment tools. Supplementary data for this chapter is in Appendix C

The study was funded by ONEMA (French National Office for Water and Aquatic Ecosystems) in the context of the 2013 − 2018 “Développement d'outils de bioindication « phytobenthos » et «macroinvértébrés benthiques » pour les eaux de surface continentales de Mayotte”program.

Afnor. (2014). NF EN 13946. Qualité de l’eau – Guide pour l’échantillonnage en routine et le prétraitement des diatomées benthiques de rivières et de plans d’eau. Afnor. (2016). NF T90 354 - Qualité de l’eau - Échantillonnage, traitement et analyse de diatomées benthiques en cours d’eau et canaux (pp. 1–79). APHA. (1976). Standard Methods for the Examination of Water and Wastewater 14ed. APHA American Public Health Association. Apothéloz-Perret-Gentil, L., Cordonier, A., Straub, F., Iseli, J., Esling, P., Pawlowski, J., 2017. Taxonomy-free molecular diatom index for high-throughput eDNA biomonitoring. Molecular Ecology Resources. doi:10.1111/1755-0998.12668 Apothéloz-Perret-Gentil, L., Cordonier, A., Straub, F., Iseli, J., Esling, P., & Pawlowski, J. (2017b). Taxonomyfree molecular diatom index for high-throughput eDNA biomonitoring. Molecular Ecology Resources. doi:10.1111/1755-0998.12668 Archibald, R. E. M. (1972). Diversity in some South African diatom associations and its relation to water quality. Water Research, 6(10), 1229–1238. Baird, D. J., & Hajibabaei, M. (2012). Biomonitoring 2.0: a new paradigm in ecosystem assessment made possible by next-generation DNA sequencing. Molecular Ecology, 21(8), 2039–2044. B-Béres, V., Lukács, Á., Török, P., Kókai, Z., Novák, Z., T-Krasznai, E., … Bácsi, I. (2016). Combined ecomorphological functional groups are reliable indicators of colonisation processes of benthic diatom assemblages in a lowland stream. Ecological Indicators, 64, 31–38. doi:10.1016/j.ecolind.2015.12.031 B-Béres, V., Török, P., Kókai, Z., Lukács, Á., T-Krasznai, E., Tóthmérész, B., & Bácsi, I. (2017). Ecological background of diatom functional groups: Comparability of classification systems. Ecological Indicators, 82, 183–188. doi:10.1016/j.ecolind.2017.07.007

- 122 -

Birtel, J., Walser, J.-C., Pichon, S., Bürgmann, H., & Matthews, B. (2015). Estimating Bacterial Diversity for Ecological Studies: Methods, Metrics, and Assumptions. PLOS ONE, 10(4), e0125356. doi:10.1371/journal.pone.0125356 Blanco, S., Cejudo-Figueiras, C., Tudesque, L., Bécares, E., Hoffmann, L., & Ector, L. (2012). Are diatom diversity indices reliable monitoring metrics? Hydrobiologia, 695(1), 199–206. doi:10.1007/s10750-0121113-1 Bowman, M. F., & Bailey, R. C. (1997). Does taxonomic resolution affect the multivariate description of the structure of freshwater benthic macroinvertebrate communities? Canadian Journal of Fisheries and Aquatic Sciences, 54(8), 1802–1807. doi:10.1139/f97-085 Carneiro, F. M., Bini, L. M., & Rodrigues, L. C. (2010). Influence of taxonomic and numerical resolution on the analysis of temporal changes in phytoplankton communities. Ecological Indicators, 10(2), 249–255. doi:10.1016/j.ecolind.2009.05.004 Chonova, T., Keck, F., Labanowski, J., Montuelle, B., Rimet, F., & Bouchez, A. (2016). Separate treatment of hospital and urban wastewaters: A real scale comparison of effluents and their effect on microbial communities. Science of The Total Environment, 542, Part A, 965–975. doi:10.1016/j.scitotenv.2015.10.161 Coissac, E., Riaz, T., & Puillandre, N. (2012). Bioinformatic challenges for DNA metabarcoding of plants and animals: BIOINFORMATIC FOR DNA METABARCODING. Molecular Ecology, 21(8), 1834–1847. doi:10.1111/j.1365-294X.2012.05550.x Cordier, T., Esling, P., Lejzerowicz, F., Visco, J., Ouadahi, A., Martins, C., … Pawlowski, J. (2017). Predicting the ecological quality status of marine environments from eDNA metabarcoding data using supervised machine learning. Environmental Science & Technology. doi:10.1021/acs.est.7b01518 Coste, M. (1982). Étude des méthodes biologiques d’appréciation quantitative de la qualité des eaux. Rapport Cemagref QE Lyon-AF Bassin Rhône Méditerranée Corse. Coste, M., Boutry, S., Tison-Rosebery, J., & Delmas, F. (2009). Improvements of the Biological Diatom Index (BDI): Description and efficiency of the new version (BDI-2006). Ecological Indicators, 9(4), 621–650. doi:10.1016/j.ecolind.2008.06.003 Edgar, R. C. (2013). UPARSE: highly accurate OTU sequences from microbial amplicon reads. Nature Methods, 10(10), 996–998. doi:10.1038/nmeth.2604 Elbrecht, V., & Leese, F. (2015). Can DNA-Based Ecosystem Assessments Quantify Species Abundance? Testing Primer Bias and Biomass—Sequence Relationships with an Innovative Metabarcoding Protocol. PLOS ONE, 10(7), e0130324. doi:10.1371/journal.pone.0130324 European Commission. (2000). Directive 2000/60/EC of the European Parliament and of the Council of 23rd October 2000 establishing a framework for Community action in the field of water policy. Official Journal of the European Communities, 327, 1–72. Evans, K. M., Wortley, A. H., Simpson, G. E., Chepurnov, V. A., & Mann, D. G. (2008). A MOLECULAR SYSTEMATIC APPROACH TO EXPLORE DIVERSITY WITHIN THE SELLAPHORA PUPULA SPECIES COMPLEX (BACILLARIOPHYTA)1. Journal of Phycology, 44(1), 215–231. doi:10.1111/j.1529-8817.2007.00454.x Greffard, M.-H., Saulnier-Talbot, É., & Gregory-Eaves, I. (2011). A comparative analysis of fine versus coarse taxonomic resolution in benthic chironomid community analyses. Ecological Indicators, 11(6), 1541–1551. doi:10.1016/j.ecolind.2011.03.024 Groendahl, S., Kahlert, M., & Fink, P. (2017). The best of both worlds: A combined approach for analyzing microalgal diversity via metabarcoding and morphology-based methods. PLOS ONE, 12(2), e0172808. doi:10.1371/journal.pone.0172808 Hebert, P. D., Cywinska, A., Ball, S. L., & others. (2003). Biological identifications through DNA barcodes. Proceedings of the Royal Society of London B: Biological Sciences, 270(1512), 313–321. Huse, S. M., Welch, D. M., Morrison, H. G., & Sogin, M. L. (2010). Ironing out the wrinkles in the rare biosphere through improved OTU clustering. Environmental Microbiology, 12(7), 1889–1898. doi:10.1111/j.1462-2920.2010.02193.x Jari Oksanen, Guillaume Blanchet, Roeland Kindt, Pierre Legendre, Peter R. Minchin, R. B. O’Hara, … Helene Wagner. (2016). vegan: Community Ecology Package. R package version 2.3-5. Retrieved from http://CRAN.R-project.org/package=vegan

- 123 -

Kahlert, M., Albert, R.-L., Anttila, E.-L., Bengtsson, R., Bigler, C., Eskola, T., … Weckström, J. (2009). Harmonization is more important than experience—results of the first Nordic–Baltic diatom intercalibration exercise 2007 (stream monitoring). Journal of Applied Phycology, 21(4), 471–482. doi:10.1007/s10811-008-9394-5 Kahlert, M., Kelly, M., Albert, R.-L., Almeida, S. F. P., Bešta, T., Blanco, S., … Vogel, A. (2012). Identification versus counting protocols as sources of uncertainty in diatom-based ecological status assessments. Hydrobiologia, 695(1), 109–124. doi:10.1007/s10750-012-1115-z Keck, F., Vasselon, V., Tapolczai, K., Rimet, F., & Bouchez, A. (2017). Freshwater biomonitoring in the Information Age. Frontiers in Ecology and the Environment. doi:10.1002/fee.1490 Kelly, M. G. (1998). Use of the trophic diatom index to monitor eutrophication in rivers. Water Research, 32(1), 236–242. doi:10.1016/S0043-1354(97)00157-7 Kelly, M. G., Trobajo, R., Rovira, L., & Mann, D. G. (2015). Characterizing the niches of two very similar Nitzschia species and implications for ecological assessment. Diatom Research, 30(1), 27–33. doi:10.1080/0269249X.2014.951398 Kermarrec, L., Franc, A., Rimet, F., Chaumeil, P., Frigerio, J.-M., Humbert, J.-F., & Bouchez, A. (2014). A next-generation sequencing approach to river biomonitoring using benthic diatoms. Freshwater Science, 33(1), 349–363. doi:10.1086/675079 Kermarrec, L., Franc, A., Rimet, F., Chaumeil, P., Humbert, J. F., & Bouchez, A. (2013). Next-generation sequencing to inventory taxonomic diversity in eukaryotic communities: a test for freshwater diatoms. Molecular Ecology Resources, 13(4), 607–619. doi:10.1111/1755-0998.12105 Kociolek, J. P., Hamsher, S. E., Kulikovskiy, M., & Bramburger, A. J. (2017). Are there species flocks in freshwater diatoms? A review of past reports and a look to the future. Hydrobiologia, 792(1), 17–35. doi:10.1007/s10750-016-3075-1 Lavoie, I., Dillon, P. J., & Campeau, S. (2009). The effect of excluding diatom taxa and reducing taxonomic resolution on multivariate analyses and stream bioassessment. Ecological Indicators, 9(2), 213–225. doi:10.1016/j.ecolind.2008.04.003 Loman, N. J., Misra, R. V., Dallman, T. J., Constantinidou, C., Gharbia, S. E., Wain, J., & Pallen, M. J. (2012). Performance comparison of benchtop high-throughput sequencing platforms. Nature Biotechnology, 30(5), 434–439. doi:10.1038/nbt.2198 Mann, D. G. (1999). The species concept in diatoms. Phycologia, 38(6), 437–495. doi:10.2216/i0031-8884-386-437.1 Mann, D. G., McDonald, S. M., Bayer, M. M., Droop, S. J. M., Chepurnov, V. A., Loke, R. E., … du Buf, J. M. H. (2004). The Sellaphora pupula species complex (Bacillariophyceae): morphometric analysis, ultrastructure and mating data provide evidence for five new species. Phycologia, 43(4), 459–482. doi:10.2216/i0031-8884-43-4-459.1 Mann, D. G., & Vanormelingen, P. (2013). An Inordinate Fondness? The Number, Distributions, and Origins of Diatom Species. Journal of Eukaryotic Microbiology, 60(4), 414–420. doi:10.1111/jeu.12047 Patin, N. V., Kunin, V., Lidström, U., & Ashby, M. N. (2013). Effects of OTU Clustering and PCR Artifacts on Microbial Diversity Estimates. Microbial Ecology, 65(3), 709–719. doi:10.1007/s00248-012-0145-4 R Development Core Team. (2008). R: A language and Environment for Statistical Computing. Vienna, Austria: R Fondation for Statistical Computing. Retrieved from http://www.r-project.org Rimet, F., Abarca, N., Bouchez, A., Kusber, W.-H., Jahn, R., Kahlert, M., … Zimmermann, J. (2018). The potential of high throughput sequencing (HTS) of natural samples as a source of primary taxonomic information for reference libraries of diatom barcodes. Fottea, 1. Rimet, F., & Bouchez, A. (2012). Biomonitoring river diatoms: Implications of taxonomic resolution. Ecological Indicators, 15(1), 92–99. doi:10.1016/j.ecolind.2011.09.014 Rimet, F., Chaumeil, P., Keck, F., Kermarrec, L., Vasselon, V., Kahlert, M., … Bouchez, A. (2016). RSyst::diatom: an open-access and curated barcode database for diatoms and freshwater monitoring. Database, 2016. doi:10.1093/database/baw016 Rivera, S. F., Vasselon, V., Jacquet, S., Bouchez, A., Ariztegui, D., & Rimet, F. (2017). Metabarcoding of lake benthic diatoms: from structure assemblages to ecological assessment. Hydrobiologia, Accepted paper.

- 124 -

Rovira, L., Trobajo, R., Sato, S., Ibáñez, C., & Mann, D. G. (2015). Genetic and Physiological Diversity in the Diatom Nitzschia inconspicua. The Journal of Eukaryotic Microbiology, 62(6), 815–832. doi:10.1111/jeu.12240 Salmaso, N., Naselli-Flores, L., & Padisák, J. (2015). Functional classifications and their application in phytoplankton ecology. Freshwater Biology, 60(4), 603–619. doi:10.1111/fwb.12520 Schloss, P. D., & Handelsman, J. (2005). Introducing DOTUR, a Computer Program for Defining Operational Taxonomic Units and Estimating Species Richness. Applied and Environmental Microbiology, 71(3), 1501– 1506. doi:10.1128/AEM.71.3.1501-1506.2005 Schloss, P. D., Westcott, S. L., Ryabin, T., Hall, J. R., Hartmann, M., Hollister, E. B., … Weber, C. F. (2009). Introducing mothur: open-source, platform-independent, community-supported software for describing and comparing microbial communities. Applied and Environmental Microbiology, 75(23), 7537–7541. doi:10.1128/AEM.01541-09 Schmidt-Kloiber, A., & Nijboer, R. C. (2004). The effect of taxonomic resolution on the assessment of ecological water quality classes. In Integrated Assessment of Running Waters in Europe (pp. 269–283). Springer. Retrieved from http://link.springer.com/chapter/10.1007/978-94-007-0993-5_16 Shokralla, S., Spall, J. L., Gibson, J. F., & Hajibabaei, M. (2012). Next-generation sequencing technologies for environmental DNA research. Molecular Ecology, 21(8), 1794–1805. doi:10.1111/j.1365294X.2012.05538.x Souffreau, C., Vanormelingen, P., Van de Vijver, B., Isheva, T., Verleyen, E., Sabbe, K., & Vyverman, W. (2013). Molecular evidence for distinct Antarctic lineages in the cosmopolitan terrestrial diatoms Pinnularia borealis and Hantzschia amphioxys. Protist, 164(1), 101–115. doi:10.1016/j.protis.2012.04.001 Stevenson, R. J., Hill, B. H., Herlihy, A. T., Yuan, L. L., & Norton, S. B. (2008). Algae–P relationships, thresholds, and frequency distributions guide nutrient criterion development. Journal of the North American Benthological Society, 27(3), 783–799. doi:10.1899/07-077.1 Stevenson, R. J., Pan, Y., & Van Dam, H. (2010). Assessing environmental conditions in rivers and streams with diatoms. In The Diatoms: Applications for the Environmental and Earth Sciences (2nd ed., pp. 57– 85). Cambridge: Cambridge University Press. Retrieved from http://www.ces.iisc.ernet.in/energy/stc/biomonitoring_of_wetlands/diatom_lake_river.pdf Straile, D., Jochimsen, M. C., & Kümmerlin, R. (2013). The use of long-term monitoring data for studies of planktonic diversity: a cautionary tale from two Swiss lakes. Freshwater Biology, 58(6), 1292–1301. doi:10.1111/fwb.12118 Taberlet, P., Coissac, E., Hajibabaei, M., & Rieseberg, L. H. (2012). Environmental DNA. Molecular Ecology, 21(8), 1789–1793. doi:10.1111/j.1365-294X.2012.05542.x Taberlet, P., Coissac, E., Pompanon, F., Brochmann, C., & Willerslev, E. (2012). Towards next-generation biodiversity assessment using DNA metabarcoding. Molecular Ecology, 21(8), 2045–2050. Tapolczai, K., Bouchez, A., Stenger-Kovács, C., Padisák, J., & Rimet, F. (2016). Trait-based ecological classifications for benthic algae: review and perspectives. Hydrobiologia, 776(1), 1–17. doi:10.1007/s10750016-2736-4 Tapolczai, K., Bouchez, A., Stenger-Kovács, C., Padisák, J., & Rimet, F. (2017). Taxonomy- or trait-based ecological assessment for tropical rivers? Case study on benthic diatoms in Mayotte island (France, Indian Ocean). Science of The Total Environment, 607–608, 1293–1303. doi:10.1016/j.scitotenv.2017.07.093 Thompson, R. M., & Townsend, C. R. (2000). Is resolution the solution?: the effect of taxonomic resolution on the calculated properties of three stream food webs. Freshwater Biology, 44(3), 413–422. doi:10.1046/j.1365-2427.2000.00579.x Vanelslander, B., Créach, V., Vanormelingen, P., Ernst, A., Chepurnov, V. A., Sahan, E., … Sabbe, K. (2009). Ecological differentiation between sympatric pseudocryptic species in the estuarine benthic diatom Navicula phyllepta (Bacillariophyceae). Journal of Phycology, 45(6), 1278–1289. doi:10.1111/j.15298817.2009.00762.x Vasselon, V., Domaizon, I., Rimet, F., Kahlert, M., & Bouchez, A. (2017). Application of high-throughput sequencing (HTS) metabarcoding to diatom biomonitoring: Do DNA extraction methods matter? Freshwater Science, 36(1), 162–177. doi:10.1086/690649

- 125 -

Vasselon, V., Rimet, F., Tapolczai, K., & Bouchez, A. (2017). Assessing ecological status with diatoms DNA metabarcoding: Scaling-up on a WFD monitoring network (Mayotte island, France). Ecological Indicators, 82, 1–12. doi:10.1016/j.ecolind.2017.06.024 Visco, J. A., Apothéloz-Perret-Gentil, L., Cordonier, A., Esling, P., Pillet, L., & Pawlowski, J. (2015). Environmental Monitoring: Inferring the Diatom Index from Next-Generation Sequencing Data. Environmental Science & Technology, 49(13), 7597–7605. doi:10.1021/es506158m Wetzel, C. E., Ector, L., Van de Vijver, B., Compere, P., & Mann, D. G. (2015). Morphology, typification and critical analysis of some ecologically important small naviculoid species (Bacillariophyta). Fottea, 15(2), 203– 234. doi:10.5507/fot.2015.020 Yu, D. W., Ji, Y., Emerson, B. C., Wang, X., Ye, C., Yang, C., & Ding, Z. (2012). Biodiversity soup: metabarcoding of arthropods for rapid biodiversity assessment and biomonitoring: Biodiversity soup. Methods in Ecology and Evolution, 3(4), 613–623. doi:10.1111/j.2041-210X.2012.00198.x Zelinka, M., & Marvan, P. (1961). Zur präzisierung der biologischen klassifikation der reinheit flie\s sender gewässer. Arch. Hydrobiol, 57(3), 389–407. Zimmermann, J., Glöckner, G., Jahn, R., Enke, N., & Gemeinholzer, B. (2015). Metabarcoding vs. morphological identification to assess diatom diversity in environmental studies. Molecular Ecology Resources, 15(3), 526–542. doi:10.1111/1755-0998.12336

- 126 -

Chapter 5

- 128 -

Improvements in diatom-based ecological quality assessment of water bodies are more than timely. This thesis overviews the advantages and drawbacks of methods that have been used for a long time and proposes several perspectives towards their improvement with the implementation of new approaches mainly from an ecological scope (Chapter 1-2). Diatom traits were successfully used to develop a sensitive index handling species redundancy, workeffort and misidentifications (Chapter 3). Using DNA sequencing, we developed another index directly using OTUs, avoiding taxonomic inventories (Chapter 4). Additionally, the effect of taxonomic resolution (OTU sequence similarity threshold) was tested on metabarcoding data and showed that there is a high redundancy when we use high sequence similarity threshold for assessing pollution (Chapter 4). The work is a part of a larger project aiming to develop new tools in bioassessment, proposing new ecological approaches as well as new molecular-based methods (Figure 5.1). In the following chapter, we discuss our results and propose several perspectives for further studies, opening a larger context of innovative approaches in diatom-based bioassessment.

Since diatom species are not necessarily delimitated according to their adaptive strategies, their number is much higher than is necessary for the assessment of some particular environmental pressure. This result in a species redundancy, which means that there are many species with the same ecological functions, and in particular, there are many species with the same pollution sensitivities. The issues of this species redundancy in term of environmental assessment and misidentifications of species are relevant in both a theoretical or ecological aspect and a practical aspect. The first issue is that there are several way for delimitating species. According to the Darwinian concept, the driving force of speciation is the adaptation of individuals to different environmental conditions by their functional traits (DARWIN, 1859; MALLET, 2010). Thus, theoretically, each diatom species would indicate a specific habitat and could be used for complex bioassessment purposes. This theory would be true if we analyse mature biofilms in a climax successional phase where the best adapted species would have already

- 129 -

outcompeted the less adapted ones , so the number of species equalled the limiting factors (HARDIN, 1960). It is however a quasi-theoretical state because the environment is constantly changing, promoting a more diverse community as suggested by the intermediate disturbance hypothesis (CONNELL, 1978). Consequently, there are probably many species that can be grouped together in order to handle the redundancy in species, because they are affected in a similar way by those two pressure gradients that we used in our study (nutrient and organic pollution together with turbidity) . Thus, the relevance of this species redundancy according to bioassessment depends on the complexity of the conditions we aim to evaluate. Finally, diatom species are not delimitated based on their adaptation strategies. A detailed review of the evolution of species concepts in diatoms was published by MANN (1999). For a long time diatom taxonomy developed without any conceptual basis, by using only morphological characteristics of the frustules without the attempt to understand how and why differences in morphology evolved or what are their adaptive significance (MANN, 1999). This weak foundation led to an instability in diatom taxonomy with continuously changing number of species and reclassification of species or genera. An early study estimated the number of diatom species as 200,000 (MANN & DROOP, 1996) but another studies reduce it to much lower numbers of 100,000 (MANN & VANORMELINGEN, 2013) or 20,000 (GUIRY, 2012). The second issue is the problem of misidentification based on morphological criteria that is also very common among diatoms. Phenotypic plasticity of several species add a complexity in the morphological-based identification. The development of light microscopes (LM) and the use of electron microscopes (SEM) introduced more precise morphological identification but also affected the taxonomy, especially in the case of small species, which are difficult to identify with LM. Thus, they suffer taxonomic shifts over time (WETZEL et al., 2015). Additional sources of misidentifications can be the person in charge of the analysis under microscope, the calibration of the microscope, the use of different identification literature, etc. (KAHLERT et al., 2009).

- 130 -

Schematic visualisation of this thesis in a larger context of our work on improving tools in diatom bioassessment. Boxes with no line represents articles without the contribution of this thesis’s author. Dashed line represent articles with a contribution as co-author. Solid line represent first author articles (figure by K. Tapolczai).

- 131 -

During the microscopic analysis of our samples, we often met the problem that species level identification was not possible. Out of a total 382 taxa, 69 were only identified to genus level (Chapter 3). Both traits and molecular data handled this problem but in a different way. Traits cluster species based on features representing their adaptive strategies, regardless the phylogenetic relatedness. While some traits can occur for phylogenetically distant species, others can be shared by phylogenetically close ones (NAKOV et al., 2014). For instance, while size class was a phylogenetically not related trait among diatoms in Mayotte rivers, the motile guild or the capacity of N-fixing occurred in phylogenetically closely related species. Molecular techniques enable a more precise identification of specimens or from environmental sample. With its rapid development, molecular techniques have been revolutionising the way to identify organisms and the amount of information that can be used for biomonitoring. Not even two decades ago, HEBERT et al. (2003) introduced the identification of species through identifying taxa-specific DNA sequences (DNA-barcodes) using DNA extracted from the organisms. At that time, it was a costly and time-consuming process, thus it was mainly used to help or confirm morphology-based identification and to register taxonomical knowledge for future generations. Today, the simultaneous analysis of multiple species (DNA-metabarcoding) from bulk or environmental samples is available with high-throughput sequencing from which OTUs can be defined. When we look at the number of OTUs that we got at a SST percentage of 95% (considered as species-level delimitation) — 1239 OTUs — it does not look like we handled species redundancy of the 382 morphospecies identified via microscope. As presented in Chapter 4, this higher number of taxonomic units is probably due to the presence of cryptic species and to the intraspecific genetic diversity that is often richer than morphospecies diversity. Our study showed that, according to several criteria (efficiency, stability vs discrimination power, predictable factors) our OTU-based index reached a plateau at 91% that can be considered as an important threshold (Chapter 4). At this similarity threshold, the number of OTUs used for the index development was only 420, still more, but not far from the number of morphospecies identified by microscopy. Several studies were published comparing metabarcoding and morphological identification regarding the diatom community or diatom-based assessment. ZIMMERMANN et al. (2015) identified almost three times more species via metabarcoding (270) than by microscopy (103). However, almost all the taxa that were identified via microscope were also found by metabarcoding, those that were missing, were those not present in the reference database (ZIMMERMANN et al., 2015). Another advantage of

- 132 -

metabarcoding is that it is able to find less abundant or rare species too. The higher number usually found via metabarcoding than by microscopy clearly shows that genetic variability is higher than the morphological. One of our study (VASSELON et al., 2017, see Figure 5.1) compared the results of a diatom index (IPS) for assessing the ecological status of Mayotte rivers based on both approaches. In this case, less taxa were identified via metabarcoding (66) compared to microscopy (204) but it was due to the incompleteness of reference database used to assign species names to OTUs. We can conclude that, while ecological traits handle species redundancy to have as many ecological meaningful groups of species as necessary to indicate particular pollutions, metabarcoding handles a more precise identification of specimens and may allow the exploration of the hidden genetic diversity behind the morphological one.

From a practical point of view, the identification of a few morphological traits is much easier than the species-level determination based on the frustule features. It was shown in Chapter 3 that using four traits (size classes, length-to-width ratios, low-profile and motile guilds) for all 382 observed taxa resulted in just as efficient results as the classical index based on the ecological profile of the 101 taxa (among 382) that have ecological profiles. However, other traits as e.g. ecological guilds cannot be observed in a microscopy-prepared sample based on dead frustules, thus it requires the use of a reference database to attribute such trait to each taxon. Regarding the molecular method, the fast development of the technique may clearly results in a labour- and cost-effective biomonitoring. It is especially welcome in the field of biomonitoring where the aim is the assessment of many river sites across a large spatial (sampling network) and temporal (long-term monitoring) scale (KECK et al., 2017). The analyses of samples tend to adapt to the increasing needs; metabarcoding techniques with sample multiplexing, the development of HTS platforms, bioinformatics tools permit to obtain and analyse more and more data in shorter time, and more cost-efficiently (Figure 5.2).

- 133 -

Estimation of price and time in function of the number of samples using microscopic (yellow) and molecular (blue) identification of diatoms. Data refer only to the laboratory of INRA UMR CARRTEL in 2017, based on Illumina sequencing and multiplexing of 280 samples (VASSELON, 2017)).

During the development of the trait-based index, we weighted the count data of species with their biovolume, i.e. we used relative biovolume rather than relative abundance of the frustules. We argue that it is ecologically well-founded since it is based on the resource partitioning principle. Larger specimens use more resources than small ones, thus the difference in size has to be taken into account in bioassessment (PADISÁK et al., 2006). This is exceptionally important for organisms like algae where differences in biovolume can be 4 orders of magnitude (MANN et al., 2016) or even more. It is surprising that this aspect has never been taken into account in diatom studies while it is a standardised method for phytoplankton counting for several decades (UTERMÖHL, 1958; EDLER & ELBRÄCHTER, 2010). Moreover, phytoplankton counting method permits the quantification of phytoplankton concentration in a water sample, concentration that is useful information for bioassessment. Absolute quantification of the biofilm is more difficult because this community forms heterogeneous layers on substrates. Additionally, the biomass is very heterogeneous in benthic habitats at a microscale, that prevents an easy quantification (CAZAUBON et al., 1995). Some studies on phytobenthos attempt to evaluate the quantity of

- 134 -

the biofilm (e.g. SABATER & SABATER, 1992; MARTIN, 1999; VINEBROOKE & LEAVITT, 1999), but this remains quite challenging. However, the use of relative biovolume of taxa would be still possible and the fact that the currently used diatom indices skim over such important ecological detail raise the question: Why this ecological aspect from a field so close has never been implemented in diatom studies? We think that it may be due to historical reasons. The feature of the siliceous cell wall of diatoms and the standard methods to analyse frustules in details put the main focus of diatom studies on the identification of species and on taxonomy. Phytoplankton is a multi-taxa complex assemblage where the species determination may be less precise but it is based on diverse features (colonies, presence of flagella, presence of mucilage, etc.) that can be easily connected to their ecological functionality. In contrast to phytoplankton, in a prepared diatom sample, the basis of identification is solely the shape and the striation patterns of the frustule. This method focus the attention on the identification but probably neglects the functional and life forms aspects. While phytoplankton studies moved towards to trait- and functional ecology, diatomists made less effort in this direction (Figure 5.3).

(A) Number of publications after a search using the keywords “diatom(s)” with “taxonomy” and “diatoms” with “trait(s)” (dark bars) and using the keywords “phytoplankton” with “taxonomy” and “phytoplankton” with “trait(s)” (lighter bars). (B) Proportion of publications with the keywords “diatom(s)” with “taxonomy” and “diatoms” with “trait(s)” compared to all publications with the keyword “diatom(s)” (dark bars). Proportion of publications with the keywords “phytoplankton” with “taxonomy” and “phytoplankton” with “trait(s)” compared to all publications with the keyword “phytoplankton” (lighter bars). (Web of Science, September 2017)

- 135 -

In our study, we corrected diatom counts by their biovolume but it may carry uncertainties within the current methods. Firstly, we used different databases (LECOINTE et al., 1993; RIMET & BOUCHEZ, 2012a; LEBLANC et al., 2012) to assign the cell dimensions to species found in the samples. Since the standard method does not require the measurement of cell dimensions this was the only solution to estimate biovolumes for the samples during the whole study period (2008-2015). Some species have an important variability in their biovolume among specimens partly due to their morphological plasticity, partly due to their particular life-cycle during which the average cell size of the population decreases in time. The second bias is also due to the standard microscopical analysis method (AFNOR, 2016). Following a transect on the microscope slide, a minimum of 400 diatom frustules are counted with a 1000x magnification. If we found an appropriate dilution during the samples preparation, the 400 frustules and the 1000x magnification give a representative picture of the community for small species. However, large species which are usually not abundant may be seen or not, just by chance. As the biomass is taken into account in the index developed for Mayotte rivers, when large species are found, they contribute significantly to the final biomass of the sample, giving a final picture of the community that is biased. It is because of the incidental appearance of large species in the field of view in the microscope. For this reason, if we want to take the biovolume into account for biomonitoring, it appears important to have the better view as possible of the relative abundance of large species which are often rare ones. To that aim, the counting protocol could be adapted with a separate counting of large specimens at a lower magnification. Sequencing techniques can be a solution to these problems. For diatom studies, we use the rbcL marker gene during the metabarcoding process. This gene is in the chloroplast genome of which several copies exist. In a study, we showed (VASSELON et al., under review) that the copy number of this genome well correlates with the size of the chloroplast that itself depends on cell biovolume/biomass (Figure 5.4). The effect of this phenomenon on HTS results is that DNA reads related to small cells contribute less in the diatom assemblage than those from big cells with a higher number of gene copy. In other words, the final results are weighted by the species biovolumes, similarly to the method we used for the trait-based index (Chapter 3) with an over-estimation of large specimens. Ironically, the mentioned study (VASSELON et al., under review) developed a correction factor to reduce the number of DNA reads of large species in order to get similar results as the classical IPS diatom index based on microscopy. Indeed, the number of reads of each species is divided by a correction factor that is proportional to the species biovolume. Ecologically however, it would make

- 136 -

sense to do the opposite, and to make use of this proportion of reads that is correlated with the biovolume, as it carries some ecological functioning meaning.

Correlation between the diatom cell biovolume and the rbcL gene copy number per cell after log(x+1) transformation (VASSELON et al., under review)

Two important characteristics of a quality index are its sensitivity and its stability. In Chapter 3, we argued that the trait-based index is more sensitive than the taxonomy-based one, i.e. values vary more on the same scale thus differences between sites are more expressive and possible to discriminate (Figure 3.6). The reason is probably the different distribution models that the two approaches use. The classical, taxonomy-based index supposes a unimodal distribution of species abundance along environmental gradients and thus use weighted mean and weighted standard deviation to estimate ecological optimum and – tolerance, respectively (ZELINKA & MARVAN, 1961; BROWN, 1984; BIRKS et al., 1990). When the distribution of species are left- or right skewed however, this estimation in not correct (POTAPOVA et al., 2004) and always moderate the optimum values when they would be more

- 137 -

extreme in the reality (Figure 5.5). This provides an index where values are always rather moderate thus stable, but less sensitive.

Ecological profile and the weighted average (WA) of two species, Gomphonema bourbonense E.Reichardt (A) and Gomphonema parvulum (Kützing) Kützing (B), along the organic/ turbidity gradient from the rivers of Mayotte (see Chapter 3). WA is a good estimation of the ecological optimum for G. bourbonense but not for G. parvulum (figure by K. Tapolczai).

The construction of the trait-based index depends on a logit model since the distribution of trait-classes showed sigmoid patterns rather than unimodal ones along the environmental gradients. Here, the increasing or decreasing trend of abundance of the traitclasses along the environmental gradient determines the quality value. The reason of several 0 and 20 values for this index is that, when the abundance of a trait-class is lower (or higher) than the abundance for the model-estimated minimum (or maximum) quality value, the final value will be 0 (or 20) for that trait class (Figure 5.6). Additionally, for this index, we weighted the diatom counts with their biovolume, thus using relative biovolume and not relative abundance of frustules for species (species being then classified into traits). For ecological reasoning, see the related section above. According to the currently used counting protocol, the chance of seeing some big specimen under the microscope during sample counting can be very incidental and, if found, overestimate the real proportion of this taxon in the

- 138 -

community. In extreme cases, only one big specimen among many small ones can strongly bias the final value, and so has an impact on the stability of the index value.

Distribution of low-profile species abundance along the organic/ turbidity gradient in Mayotte rivers. The logit model shows a sigmoid pattern along which the quality value can be determined by the observed abundance. Values are automatically 0 (bottom) or 20 (top) in the shaded areas.

We also tested the effect of the SST, as a proxy of taxonomic resolution on the OTU-based index’s stability and sensitivity (or discrimination power) in Chapter 4. The IdxOTU at all SST is based on the same calculation as the classical taxonomy-based one, thus here the differences are not due to the different methods in the index development but only to the different SSTs, i.e. taxonomic resolution. The two parameters, which change with the taxonomic resolution, are the number of OTUs and their ecological profiles. The higher the taxonomic resolution, the higher the number of OTUs but the occurrence of each OTU is lower. Increasing the resolution provides OTUs with more distinct and precise ecological profiles and this increases the efficiency and the discrimination power of the index. Then, it reaches a plateau from which the further delimitation does not provide additional pollution assessment discrimination. Furthermore, the lower occurrence of the increasing number of OTUs results in weaker and less robust ecological profiles. This makes the index more sensitive to the initial

- 139 -

randomisation process when selecting training and test datasets. It thus becomes less stable. We also showed that sites with lower OTU richness were more sensitive to this phenomenon. This is due to the eventuality coming from the fact that the effect of the not robust enough profiles is stronger in the case of only a few OTUs, while at OTU-rich sites with higher number of OTUs can moderate the effect of some extreme cases. This shows the importance of the taxonomic resolution to handle the mentioned trade-off between index’s stability and sensitivity. Index’s stability and sensitivity are two properties that water managers have to well balance in their choice of assessment tools. For instance, an index that is less sensitive may not indicate quality changes in the environment because it could be too stable whereas intervention would be needed. On the other hand, if an index is too sensitive, it may show quality differences that can be due to its instability and not due to real changes.

A great advantage of DNA-metabarcoding combined with HTS methods is the massive quantity of data provided. This will surely boost biomonitoring where the analysis of a large quantity of samples over a long time from diverse sampling networks is the objective. Due to the rapid development of information technologies, the bioinformatic analysis of these data has become even more time- and cost efficient (KECK et al., 2017). This sudden increase in the obtainable data is often referred as "big data" and the handling of these enormous datasets is becoming a separate field of study. The advantages of DNA metabarcoding with HTS compared to the classical, morphological approach, together with the improvement in the science of big data can potentially revolutionise biomonitoring (BAIRD & HAJIBABAEI, 2012). Currently, the metabarcoding approach is mainly used to renovate classical biomonitoring methods. For instance, the ongoing studies in diatom-based bioassessment via metabarcoding focus on the standardisation of each step in order to be comparable to classical methods but in a more rapid and cost-effective way, and with a precise identification accuracy. The introduction of molecular techniques is very welcome in the field of biomonitoring and a strong improvement could be achieved in terms of throughput and cost-efficiency. However, it would be also welcome to use this massive quantity of data to improve the ecological meaning of bioassessment tools. From our metabarcoding study on diatom communities from Mayotte rivers, we could highlight a few tracks:

- 140 -

(i)

Regarding the sequence similarity threshold to calculate OTUs, studies usually use 95 or 97% with which they try to target a quasi-species level resolution. Our study was the first analysing the effect of the SST as a proxy for taxonomic resolution (see Chapter 4) and drew the same conclusion as other studies, based on microscopy, examining the effect of the taxonomic resolution: species level identification is not necessary for bioassessment purposes (RIMET & BOUCHEZ, 2012b).

(ii)

We provided and alternative solution for the problematics of incomplete reference databases where sequences could be assigned to taxonomic names. In our study, we developed an index (Chapter 4) that completely avoids taxonomic assignment, i.e. the use of reference database. Instead, we defined directly the ecological profile with ecological optimum and –tolerance of the OTUs. To ensure a minimum robustness of these profiles, all OTUs that were present in less than 5% of the samples, were removed. This database with the sensitivity and indicator values of each OTUs were then used in order to calculate the index.

(iii)

The calculation background of the OTU-based index was the same as for the taxonomy-based one, i.e. the weighted average and –standard deviaton were used as proxies for the optimum and tolerance values (ZELINKA & MARVAN, 1961). Thus, it integrates the same uncertainties coming from the unimodal profile model. It can be however corrected by using alternative models representing the ecological profiles the best (POTAPOVA et al., 2004).

(iv)

Another ecological aspect is that the DNA barcode for amplification and sequencing is chosen to detect mainly diatoms and only a few phylogenetic neighbouring groups (e.g. see Material and methods in VASSELON et al., 2017). The original reason of using only diatoms in the classical, microscopical method is the ease of the sampling and the preparation, and the ease of identification of the species based on their cell wall structure, while the determination of nondiatom algae would require different process and more labour especially for benthic habitats. By using an appropriate target DNA barcode, DNAmetabarcoding sequencing results could cover the whole phytobenthic community and thus provide an ecologically more meaningful bioassessment (SCHNEIDER et al., 2016) tool. However, the problem of incomplete reference database will still exist for diatoms (RIMET et al., 2016) and should be even more acute with other algae classes.

- 141 -

A way to measure the efficiency of a diatom index is to check if it is correlated to particular environmental parameters (e.g. nutrients, organic matter, pH). When a good correlation is found, the index is considered as efficient, otherwise, it is less or not efficient at all. The same concept was used in our work even if it is sometimes questionable ecologically (KELLY, 2011; SCHNEIDER et al., 2016). Indeed, it would be worth analysing this concept from different aspects: (i) The concentration of an environmental parameter, e.g. nutrients, can be both the cause and the effect of the present algal community. The most typical example is eutrophication that is often measured by the nutrient concentration, neglecting the fact that it is a process triggered by nutrient loading that affect concentration together with the aquatic flora that is also affected by the nutrient concentration (SCHNEIDER et al., 2016). It does not give an appropriate evaluation for eutrophication though. In Mayotte, there is no sign of eutrophication but downstream sites had higher nitrogen-forms concentration, organic matter and turbidity. Phosphorus however has a particular longitudinal pattern with higher concentration upstream than downstream that lead us to consider the importance of going beyond the results from simple correlations. Although we could not test it, our experience and further publications strongly suggest that the lower concentration of phosphorus downstream was due to the A4 Zeolite found in the “eco-friendly” washing powder constantly used for washing the laundry in the rivers by locals; the dominating structuring factor explaining diatom communities was suspended matter and organic matter concentrations (Chapter 3). (ii) In some cases, biotic interactions can indirectly affect correlations between the chemical parameters and the biological indicator and this makes things more complicated for bioassessment than via simple correlations. For instance, epiphyton can regulate the growth of macrophytes by a shading effect by reducing light availability (KÖHLER et al., 2010). This interaction cannot be seen in a simple biota vs. parameter correlation, thus ecological interpretation is not possible, and the status assessment is strongly biased (SCHNEIDER et al., 2016). Additionally, when correlating phytobenthic metrics with environmental parameters, the top-down effect of e.g. grazers (e.g. JONES & SAYER, 2003) on benthic algae is neglected. Our study did not examine the top-down effect neither but we had information about the macroinvertebrate communities at some sites and we hypothesised that they may have an effect on diatom size selection (Chapter 3).

- 142 -

(iii) An interesting question was raised by SCHNEIDER et al. (2016): if the test for ecological metrics is to correlate them with an environmental parameter, then what is the point of using that metric instead of measuring the environmental parameter? The straightforward answer is of course the pressure-integration offered by the bioindicator. However, one other answer is that the ecological response along an environmental gradient can be described by different types of functions (Figure 5.7) and it may affect the appropriate management intervention. If the relationship differs from a linear response, a threshold value need to be set for a particular range of environmental pressure. In our study on traits (Chapter 3), we showed that traits had a threshold/sigmoid response to the environmental gradient, thus a more steep response occurred in a particular precise pollution range than on the entire pollution gradient (Figure 5.6).

Potential relationships between a measured variable (e.g. water total phosphorus concentration) and an ecological response (e.g. species composition of aquatic flora); the figures exemplify a linear, threshold, asymptotic and exponential response (from left to right); vertical dashed lines exemplify where critical values of the measured variable may be set, such that they lie before or after steep parts of the ecological response (SCHNEIDER et al., 2016).

Our work aimed to contribute to the improvement of diatom-based bioassessment and to enlighten some ecological aspects and several drawbacks of the currently used methods. Traits represent a direct and functional link between species and the environment in which they are present, thus they can be successfully used for biomonitoring purposes. It serves a solution for the problematics of species redundancy or misidentifications. Molecular techniques provide a fine and accurate identification and we showed the effect of the taxonomic resolution on a molecular-based diatom index. However, the final objective in long term would be a more holistic approach in which the structure and composition of the

- 143 -

phytobenthos would be directly linked to the integrated effect of all environmental factors describing the habitat (TAPOLCZAI et al., 2016).

The WFD introduced a new paradigm to freshwater assessment, management and legislation: the ecological quality status (EUROPEAN COMMISSION, 2000). The ecological status should represent the ecological integrity of the site by comparing the BQE communities’ structure, composition, functioning of the impacted site to those of a hypothetical reference site without (or with minimal) anthropogenic impact (PARDO et al., 2012). The pre-existing diatom indices were implemented to fulfil the needs of WFD to evaluate water bodies, however, they (i)

use only diatom taxonomic lists without regard to their functionality, traits, interactions, etc., and

(ii)

aim to evaluate only specific pollutions (e.g. trophic index) instead of the overall ecological status.

Clearly, a more holistic approach with diverse metrics is lacking in this field to measure the “health” of a habitat that could integrate the numerous environmental parameters coupling the habitat and its communities. The innovative concept of “ecological quality” stated in the WFD was finally not followed by a renewal of methods (KELLY et al., 2009).

Many studies argued for the usefulness of diatom communities as a proxy of the entire phytobenthos when assessing specific pollutants (e.g. KELLY et al., 2008). This has practical advantages because standard and simple sampling and preparation techniques are available for diatoms, while the analyses of other taxa would require more work-effort. In other cases however, non-diatom taxa make up a large part of the phytobenthos and the necessity of their analysis is inevitable. Their contribution is especially important under eutrophic conditions as they are likely to dominate the benthic algal community (DENICOLA et al., 2004). Among the numerous diatom-based quality indices implemented in the WFD, some includes non-diatoms too. The Norwegian Periphyton Index of Trophic status (PIT; SCHNEIDER & LINDSTRØM, 2011) and Acidification Index Periphyton (AIP; SCHNEIDER &

- 144 -

LINDSTRØM, 2009) successfully use mainly filamentous benthic chlorophytes to assess trophic and acidity levels. Austria and Germany also use indices including non-diatoms (ROTT et al., 1997; ROTT et al., 1999; SCHAUMBURG et al., 2004) and the Czech Republic as well (MARVAN et al., 2011). For a holistic approach, we clearly suggest that the analysis of non-diatom taxa is important (Chapter 2). Due to the traditional and applied protocol of diatom sample preparation, another important metric is lost preventing an ecologically relevant and holistic view on the phytobenthic community: the proportion of frustules that contain living or dead cells. The importance of this poorly studied metric can highly varies among sites. In streams, where the flow can wash away dead cells, this effect is less important (GILLETT et al., 2008) than for example in lowland, slow-flowing streams where it has important effect (B-BÉRES, personal discussion). This aspect was not studied in our work but surely should be explored in future research. We can though hypothesise that this metric could be more important at downstream than at upstream sites. Molecular techniques can serve with solutions for both problems in order to provide a better picture. First, it is based on the DNA extracted from the chloroplast, thus it does not detect the presence of dead and empty frustules, though it focuses on the living part of the community. Secondly, with the choice of an appropriate target DNA barcode for amplification, we could be able to cover the phytobenthos at a larger taxonomic scale, not solely diatoms. Consequently, DNA-based methods could have a very important role in the development of new assessment methods with a strong ecological foundation.

In the Perspectives of Chapter 2, we proposed different trait-based approaches for ecological classification of benthic algae. Our work on traits (Chapter 3) partly followed our first proposition (Figure 2.2) using statistical methods to discover trait-environment relationships and we managed to develop an efficient trait-based diatom index for Mayotte rivers. However, the small size of the island, its homogenous geological composition, the reduced number of ecological traits due to the mentioned drawbacks of diatom sampling and preparation protocols (see above, 5.3.2) prevented the establishment of a complex functional classification for benthic algae. Further risk of this approach is that the distribution of a trait along an environmental gradient can be biased either by the effect of other environmental

- 145 -

parameters or by the effect of other traits that are not observed in that particular correlation. For instance, B-BÉRES et al. (2016) showed that, while diatom guilds (PASSY, 2007) did not correlate significantly with the tested environmental factors, a combination of guilds and size classes of BERTHON et al. (2011), i.e. the Combined Eco-Morphological Functional Groups (CEMFGs; B-BÉRES et al., 2016; B-BÉRES et al., 2017) highlighted differences within the same guild and provided ecologically meaningful interpretation for some of these combined groups. The combination of guilds and size classes however, were carried out without preassumptions about their ecological response to the environmental parameters. It is the same a priori method as in LAW et al. (2014) where different traits ( 3 CSR and 7 life forms) were combined but finally only 2 from the 21 groups were able to indicate environmental gradients. Our future plan is to follow our second approach (Figure 2.3) based on a phytosociological approach - very similar to that one developed by REYNOLDS et al. (2002) for phytoplankton. It starts with the definition of habitats that can be done by different ways: expert knowledge, clustering or the use of well-described river typologies, etc. We think that the advantage of this approach lies in its holistic point of view in the sense that habitats here are defined by a set of different environmental parameters. In this approach, traits will not be correlated with single factors, but to a more complex ecological habitat. Similarly, dominant species or species associations in these habitats possess a set of traits, i.e. a combination of traits that are not randomly assembled but include species adapted to the habitat where they are present. Finally, the interpretation of the traits occurring under particular circumstances is needed. It means, we have to analyse which are the specific traits related to a specific environment and in what way they provide adaptation advantage to the species possessing it. It is a long process and in order to get a better understanding a change in the traditional protocols and several supporting ecophysiological studies are needed. In our studies (Chapter 2-4), we always emphasised that taxonomical lists, trait data and molecular data have to be used in a way to complement and not to replace each other in order to have a more holistic view on the benthic algal communities. Such a study would provide a strong base for further improvement of an ecologically meaningful phytobenthosbased bioassessment.

- 146 -

AFNOR., 2016. – NF T90 354 - Qualité de l’eau - Échantillonnage, traitement et analyse de diatomées benthiques en cours d’eau et canaux., 1‑ 79 p. BAIRD D. J. & HAJIBABAEI M., 2012. – Biomonitoring 2.0: a new paradigm in ecosystem assessment made possible by next-generation DNA sequencing. Molecular ecology, 21 (8) : 2039‑2044. B-BÉRES V., LUKÁCS Á., TÖRÖK P., KÓKAI Z., NOVÁK Z., T-KRASZNAI E., TÓTHMÉRÉSZ B. & BÁCSI I., 2016. – Combined eco-morphological functional groups are reliable indicators of colonisation processes of benthic diatom assemblages in a lowland stream. Ecological Indicators, 64 : 31‑38 doi : 10.1016/j.ecolind.2015.12.031. B-BÉRES V., TÖRÖK P., KÓKAI Z., LUKÁCS Á., T-KRASZNAI E., TÓTHMÉRÉSZ B. & BÁCSI I., 2017. – Ecological background of diatom functional groups: Comparability of classification systems. Ecological Indicators, 82 : 183‑188 doi : 10.1016/j.ecolind.2017.07.007. BERTHON V., BOUCHEZ A. & RIMET F., 2011. – Using diatom life-forms and ecological guilds to assess organic pollution and trophic level in rivers: a case study of rivers in south-eastern France. Hydrobiologia, 673 (1) : 259–271 doi : 10.1007/s10750-011-0786-1. BIRKS H. J. B., LINE J. M., JUGGINS S., STEVENSON A. C. & BRAAK C. J. F. T., 1990. – Diatoms and pH Reconstruction. Philosophical Transactions of the Royal Society B: Biological Sciences, 327 (1240) : 263‑ 278 doi : 10.1098/rstb.1990.0062. BROWN J. H., 1984. – On the Relationship between Abundance and Distribution of Species. The American Naturalist, 124 (2) : 255‑279 doi : 10.1086/284267. CAZAUBON A., ROLLAND T. & LOUDIKI M., 1995. – Heterogeneity of periphyton in French Mediterranean rivers. Hydrobiologia, 300 (1) : 105–114. CONNELL J. H., 1978. – Diversity in tropical rain forests and coral reefs. Science, 199 (4335) : 1302–1310. DARWIN C., 1859. – On the origin of species. J. Murray, London, . DENICOLA D. M., EYTO E. DE., WEMAERE A. & IRVINE K., 2004. – Using epilithic algal communities to assess trophic status in Irish lakes. Journal of Phycology, 40 (3) : 481‑495 doi : 10.1111/j.1529-8817.2004.03147.x. EDLER L. & ELBRÄCHTER M., 2010. – The Utermöhl method for quantitative phytoplankton analysis. Microscopic and molecular methods for quantitative phytoplankton analysis, 110 . EUROPEAN COMMISSION., 2000. – Directive 2000/60/EC of the European Parliament and of the Council of 23rd October 2000 establishing a framework for Community action in the field of water policy. Official Journal of the European Communities, 327 : 1‑72. GILLETT N., PAN Y. & PARKER C., 2008. – Should only live diatoms be used in the bioassessment of small mountain streams? Hydrobiologia, 620 (1) : 135‑147 doi : 10.1007/s10750-008-9624-5. GUIRY M. D., 2012. – How many species of algae are there? Journal of Phycology, 48 (5) : 1057‑1063 doi : 10.1111/j.1529-8817.2012.01222.x. HARDIN G., 1960. – The Competitive Exclusion Principle. Science, 131 (3409) : 1292–1297. HEBERT P. D., CYWINSKA A., BALL S. L. & OTHERS., 2003. – Biological identifications through DNA barcodes. Proceedings of the Royal Society of London B: Biological Sciences, 270 (1512) : 313–321. JONES J. I. & SAYER C. D., 2003. – Does the Fish–Invertebrate–Periphyton Cascade Precipitate Plant Loss in Shallow Lakes? Ecology, 84 (8) : 2155‑2167 doi : 10.1890/02-0422. KAHLERT M., ALBERT R.-L., ANTTILA E.-L., BENGTSSON R., BIGLER C., ESKOLA T., GÄLMAN V., GOTTSCHALK S., HERLITZ E., JARLMAN A., KASPEROVICIENE J., KOKOCIŃSKI M., LUUP H., MIETTINEN J., PAUNKSNYTE I. ET AL., 2009. – Harmonization is more important than experience—results of the first Nordic–Baltic diatom intercalibration exercise 2007 (stream monitoring). Journal of Applied Phycology, 21 (4) : 471‑482 doi : 10.1007/s10811-008-9394-5. KECK F., VASSELON V., TAPOLCZAI K., RIMET F. & BOUCHEZ A., 2017. – Freshwater biomonitoring in the Information Age. Frontiers in Ecology and the Environment, doi : 10.1002/fee.1490.

- 147 -

KELLY M., 2011. – The Emperor’s new clothes? A comment on. Ecological Indicators, 11 (5) : 1492‑1494 doi : 10.1016/j.ecolind.2011.02.014. KELLY M. G., KING L., JONES R. I., BARKER P. A. & JAMIESON B. J., 2008. – Validation of diatoms as proxies for phytobenthos when assessing ecological status in lakes. Hydrobiologia, 610 (1) : 125‑129 doi : 10.1007/s10750-008-9427-8. KELLY M., KING L. & NÍ CHATHÁIN B., 2009. – THE CONCEPTUAL BASIS OF ECOLOGICAL-STATUS ASSESSMENTS USING DIATOMS. Biology & Environment: Proceedings of the Royal Irish Academy, 109 (3) : 175‑189 doi : 10.3318/BIOE.2009.109.3.175. KÖHLER J., HACHOŁ J. & HILT S., 2010. – Regulation of submersed macrophyte biomass in a temperate lowland river: Interactions between shading by bank vegetation, epiphyton and water turbidity. Aquatic Botany, 92 (2) : 129‑136 doi : 10.1016/j.aquabot.2009.10.018. LAW R. J., ELLIOTT J. A. & THACKERAY S. J., 2014. – Do functional or morphological classifications explain stream phytobenthic community assemblages? Diatom Research, 29 (4) : 309–324 doi : 10.1080/0269249X.2014.889037. LEBLANC K., ARÍSTEGUI J., ARMAND L., ASSMY P., BEKER B., BODE A., BRETON E., CORNET V., GIBSON J., GOSSELIN M.-P., KOPCZYNSKA E., MARSHALL H., PELOQUIN J., PIONTKOVSKI S., POULTON A. J. ET AL., 2012. – A global diatom database – abundance, biovolume and biomass in the world ocean. Earth System Science Data, 4 (1) : 149‑165 doi : 10.5194/essd-4-149-2012. LECOINTE C., COSTE M. & PRYGIEL J., 1993. – “Omnidia”: software for taxonomy, calculation of diatom indices and inventories management. Dans : Twelfth International Diatom Symposium. Springer, p. 509– 513. MALLET J., 2010. – Group selection and the development of the biological species concept. Philosophical Transactions of the Royal Society B: Biological Sciences, 365 (1547) : 1853‑1863 doi : 10.1098/rstb.2010.0040. MANN D. & DROOP S., 1996. – Biodiversity, biogeography and conservation of diatoms. Dans : Biogeography of freshwater algae. Springer, p. 19‑32. MANN D. G., 1999. – The species concept in diatoms. Phycologia, 38 (6) : 437‑495 doi : 10.2216/i0031-888438-6-437.1. MANN D. G., CRAWFORD R. M. & ROUND F. E., 2016. – Bacillariophyta. Dans : Archibald JM, Simpson AGB, Slamovits CH, Margulis L, Melkonian M, Chapman DJ, Corliss JO. Handbook of the Protists. Cham : Springer International Publishing, p. 1‑62. doi : 10.1007/978-3-319-32669-6_29-1. MANN D. G. & VANORMELINGEN P., 2013. – An Inordinate Fondness? The Number, Distributions, and Origins of Diatom Species. Journal of Eukaryotic Microbiology, 60 (4) : 414‑420 doi : 10.1111/jeu.12047. MARTIN G., 1999. – Distribution of phytobenthos biomass in the Gulf of Riga (1984-1991). Hydrobiologia, 393 : 181‑190. MARVAN P., OPATRILOVA L. & FRÁNKOVÁ M., 2011. – Phytobenthos assessment methods for river ecological status evaluation in the Czech Republic and neighbouring countries (in Czech). Vodohospodářské technicko-ekonomické informace 0322-8916, 53 : 1‑4. NAKOV T., THERIOT E. C. & ALVERSON A. J., 2014. – Using phylogeny to model cell size evolution in marine and freshwater diatoms. Limnology and Oceanography, 59 (1) : 79‑86 doi : 10.4319/lo.2014.59.01.0079. PADISÁK J., BORICS G., GRIGORSZKY I. & SORÓCZKI-PINTÉR É., 2006. – Use of Phytoplankton Assemblages for Monitoring Ecological Status of Lakes within the Water Framework Directive: The Assemblage Index. Hydrobiologia, 553 (1) : 1–14 doi : 10.1007/s10750-005-1393-9. PARDO I., GÓMEZ-RODRÍGUEZ C., WASSON J.-G., OWEN R., VAN DE BUND W., KELLY M., BENNETT C., BIRK S., BUFFAGNI A., ERBA S., MENGIN N., MURRAY-BLIGH J. & OFENBÖECK G., 2012. – The European reference condition concept: A scientific and technical approach to identify minimally-impacted river ecosystems. Science of The Total Environment, 420 : 33‑42 doi : 10.1016/j.scitotenv.2012.01.026. PASSY S. I., 2007. – Diatom ecological guilds display distinct and predictable behavior along nutrient and disturbance gradients in running waters. Aquatic Botany, 86 (2) : 171–178 doi : 10.1016/j.aquabot.2006.09.018.

- 148 -

POTAPOVA M. G., CHARLES D. F., PONADER K. C. & WINTER D. M., 2004. – Quantifying species indicator values for trophic diatom indices: a comparison of approaches. Hydrobiologia, 517 (1) : 25–41. REYNOLDS C. S., HUSZAR V., KRUK C., NASELLI-FLORES L. & MELO S., 2002. – Towards a functional classification of the freshwater phytoplankton. Journal of plankton research, 24 (5) : 417–428. RIMET F. & BOUCHEZ A., 2012a. – Life-forms, cell-sizes and ecological guilds of diatoms in European rivers. Knowledge and Management of Aquatic Ecosystems, (406) : 01 doi : 10.1051/kmae/2012018. RIMET F. & BOUCHEZ A., 2012b. – Biomonitoring river diatoms: Implications of taxonomic resolution. Ecological Indicators, 15 (1) : 92–99 doi : 10.1016/j.ecolind.2011.09.014. RIMET F., CHAUMEIL P., KECK F., KERMARREC L., VASSELON V., KAHLERT M., FRANC A. & BOUCHEZ A., 2016. – R-Syst::diatom: an open-access and curated barcode database for diatoms and freshwater monitoring. Database: The Journal of Biological Databases and Curation, 2016 doi : 10.1093/database/baw016. ROTT E., HOFMANN G., PALL K., PFISTER P. & PIPP E., 1997. – Indikationslisten für Aufwuchsalgen, Teil 1: Saprobielle Indikation (Indication lists for periphytic algae. Part 1: Saprobic indication). Bundesministerium für Land-und Forstwirtschaft (Federal Ministry of Agriculture and Forestry), Wien, . ROTT E., PIPP E., PFISTER P., VAN DAM H., ORTLER K., PALL K. & BINDER N., 1999. – Indikationslisten für Aufwuchsalgen in österreichischen Fliessgewässern. Teil 2: Trophie-indikation sowie geochemische Präferenz; taxonomische und toxikologische Anmerkungen. Bundesministerium für Land-und Forstwirtschaft, Wasserwirtschaftskataster, Wien, . SABATER S. & SABATER F., 1992. – Longitudinal changes of benthic algal biomass in a Mediterranean river during two high production periods. Archiv für Hydrobiologie, 124 (4) : 475‑487. SCHAUMBURG J., SCHRANZ C., HOFMANN G., STELZER D., SCHNEIDER S. & SCHMEDTJE U., 2004. – Macrophytes and phytobenthos as indicators of ecological status in German lakes — a contribution to the implementation of the water framework directive. Limnologica - Ecology and Management of Inland Waters, 34 (4) : 302‑314 doi : 10.1016/S0075-9511(04)80003-3. SCHNEIDER S. C., HILT S., VERMAAT J. E. & KELLY M., 2016. – The “Forgotten” Ecology Behind Ecological Status Evaluation: Re-Assessing the Roles of Aquatic Plants and Benthic Algae in Ecosystem Functioning. Berlin, Heidelberg : Springer Berlin Heidelberg. SCHNEIDER S. C. & LINDSTRØM E.-A., 2011. – The periphyton index of trophic status PIT: a new eutrophication metric based on non-diatomaceous benthic algae in Nordic rivers. Hydrobiologia, 665 (1) : 143‑155 doi : 10.1007/s10750-011-0614-7. SCHNEIDER S. & LINDSTRØM E.-A., 2009. – Bioindication in Norwegian rivers using non-diatomaceous benthic algae: The acidification index periphyton (AIP). Ecological Indicators, 9 (6) : 1206‑1211 doi : 10.1016/j.ecolind.2009.02.008. TAPOLCZAI K., BOUCHEZ A., STENGER-KOVÁCS C., PADISÁK J. & RIMET F., 2016. – Trait-based ecological classifications for benthic algae: review and perspectives. Hydrobiologia, 776 (1) : 1‑17 doi : 10.1007/s10750-016-2736-4. UTERMÖHL H., 1958. – Zur Vervollkommnung der quantitativen Phytoplankton-Methodik. Mitt. int. Ver. theor. angew. Limnol., 9 : 1 – 38. VASSELON V., 2017. – Barcoding and Biomonitoring : Assesment of stream water quality using benthic diatom metabarcoding. Université Savoie Mont Blanc. VASSELON V., BOUCHEZ A., RIMET F., JACQUET S., TROBAJO R., CORNIQUEL M., TAPOLCZAI K. & DOMAIZON I., under review. – Avoiding quantification bias in metabarcoding: application of a cell biovolume correction factor in diatom molecular biomonitoring. Methods in Ecology and Evolution, . VASSELON V., RIMET F., TAPOLCZAI K. & BOUCHEZ A., 2017. – Assessing ecological status with diatoms DNA metabarcoding: Scaling-up on a WFD monitoring network (Mayotte island, France). Ecological Indicators, 82 : 1‑12 doi : 10.1016/j.ecolind.2017.06.024. VINEBROOKE R. D. & LEAVITT P. R., 1999. – Phytobenthos and Phytoplankton as Potential Indicators of Climate Change in Mountain Lakes and Ponds: A HPLC-Based Pigment Approach. Journal of the North American Benthological Society, 18 (1) : 15‑33 doi : 10.2307/1468006.

- 149 -

WETZEL C. E., ECTOR L., VAN DE VIJVER B., COMPERE P. & MANN D. G., 2015. – Morphology, typification and critical analysis of some ecologically important small naviculoid species (Bacillariophyta). Fottea, 15 (2) : 203‑234 doi : 10.5507/fot.2015.020. ZELINKA M. & MARVAN P., 1961. – Zur präzisierung der biologischen klassifikation der reinheit flie\s sender gewässer. Arch. Hydrobiol, 57 (3) : 389–407. ZIMMERMANN J., GLÖCKNER G., JAHN R., ENKE N. & GEMEINHOLZER B., 2015. – Metabarcoding vs. morphological identification to assess diatom diversity in environmental studies. Molecular Ecology Resources, 15 (3) : 526‑542 doi : 10.1111/1755-0998.12336.

- 150 -

Appendix A

Two reference sites: (A) Dapani (upstream, 2015) and (B) Mapouhera (upstream, 2015) in Mayotte. Reference sites are located in the forest surrounded by dense vegetation and containing a high quantity of dead leaves being decomposed (photos by K. Tapolczai and F. Rimet).

(A) Locals wash their clothes in the rivers, a typical environmental pressure mainly at downstream sites that are more urbanised (Chajou downstream, 2014). (B) Packs of the “ecofriendly” washing powder thrown away are very common in the rivers (Combani downstream 2014), (C) A high diversity of garbagaes can be found in the downstream sections of rivers (Doujani 2014). (D) Pipes can be observed in the walls of several households from which wastewater flows directly in the rivers (Mouala downstream 2015)

(A) Hydrosera triquetra forms a dense filamentous biofilm at the reference site of Longoni upstream (2015). (B) The same species after preparation under light microscope.

Appendix B

Mean and standard deviation (SD) values of physico-chemical parameters at the three sampling networks (Ref, RCS, Poll).

Environmental parameter

Mean (±SD) REF

RCS

POLL

Temperature (°C)

23.43 ± 1.07

24.44 ± 1.99

24.80 ± 1.46

pH

7.70 ± 0.39

7.63 ± 0.36

7.35 ± 0.25

eH (mV)

187.64 ± 47.17

161.41 ± 111.54

15.60 ± 111.51

Conductivity (µS cm-1)

284.70 ± 107.09

230.50 ± 87.26

884.88 ± 2668.10

Oxygen saturation (%)

84.72 ± 13.83

78.60 ± 25.92

37.59 ± 27.33

Oxygen concentration (mg L-1)

7.17 ± 1.20

6.58 ± 2.16

3.18 ± 2.34

Turbidity (NFU)

2.86 ± 2.87

2.55 ± 2.48

14.65 ± 15.60

Suspended matter (mg L-1)

2.66 ± 1.99

4.60 ± 4.42

12.20 ± 18.16

68.26 ± 46.73

42.44 ± 1.99

69.50 ± 43.62

1.75 ± 1.70

2.66 ± 1.24

3.17 ± 3.34

Total organic carbon (mg L-1)

1.70 ± 1.86

2.40 ± 1.22

5.26 ± 6.00

Total nitrogen (mg L-1)

0.21 ± 0.11

0.19 ± 0.17

0.97 ± 1.27

Total phosphorus (mg L-1)

0.15 ± 0.11

0.09 ± 0.05

0.07 ± 0.11

Silica (mg L-1)

41.99 ± 8.81

34.30 ± 9.36

37.9 ± 12.74

Nitrite (mg L-1)

0.01 ± 0.00

0.02 ± 0.02

0.07 ± 0.15

Nitrate (mg L-1)

0.39 ±0.25

0.52 ± 0.23

1.19 ± 2.30

Ammonium (mg L-1)

0.02 ± 0.01

0.04 ± 0.02

0.72 ± 1.90

Phosphate (mg L-1)

0.40 ± 0.26

0.22 ± 0.14

0.15 ± 0.35

Calcium (mg L-1)

21.00 ± 10.94

16.51 ± 6.38

21.82 ± 19.68

Magnesium (mg L-1)

12.43 ± 4.97

10.49 ± 3.79

23.72 ± 54.82

Sodium (mg L-1)

22.62 ± 14.16

16.95 ± 8.07

117.46 ± 438.60

Potassium (mg L-1)

3.04 ± 1.90

2.42 ± 0.82

7.88 ± 17.24

Chloride (mg L-1)

18.62 ± 7.74

15.18 ± 4.41

219.61 ± 941.33

Sulphate (mg L-1)

3.12 ± 1.16

2.58 ± 0.91

32.33 ± 133.70

19.09 ± 12.35

25.58 ± 11.98

24.56 ± 15.89

Chemical oxygen demand (mg L-1) Dissolved organic carbon (mg L-1)

Flow velocity (cm s-1)

Taxa's sensitivity and indicator values for the two sub-indices of the taxonomy-based index.

Species code

Nutrients

Organic/turbidity

Sensitivity

Indicator

Sensitivity

Indicator

ACOP

0.25830194

2.93233276

1.29147162

1.94748519

ADEG

0.20330081

0.54930395

0.0531638

0.51086148

ADMI

-0.83275695

0.76885595

-0.63901197

0.7688188

ADSH

0.75566813

1.61509094

0.88469449

1.57994106

ADSP

0.04901836

1.953301

-0.22305365

1.84180468

AMIA

1.40101907

1.35929859

0.82891009

2.15031288

AMPS

0.11396039

2.94988067

1.94876582

1.09264719

AMSM

2.75205423

0.92074713

1.61792305

1.17672401

APED

0.61773743

2.86294945

1.01450765

2.20254128

HSTU

0.36684177

3.67575737

0.32729146

1.96454841

CMEN

-0.32544314

1.01620501

-0.247272

0.99237108

CPLA

-0.59142017

0.89350494

-0.162451

0.51574921

CPLE

0.46880486

1.51351835

0.77988893

1.47423477

CPLI

0.29797884

0.83566024

0.35813795

0.65133537

DCOF

-0.30248535

0.90801434

-0.14762488

0.85196228

ENMS

-0.15000245

1.65156487

0.88159545

1.0135364

EOMI

-0.73150332

0.62362921

-0.84112986

0.542964

EOS1

0.15552279

2.09137672

0.28225909

2.05648582

EOS2

0.32602187

1.59282463

0.5635467

1.19824257

EOSP

0.367738

1.50273544

0.38234064

1.40545994

EORU

0.82099558

0.79269824

1.54003868

1.27377567

ESLE

-0.10192211

1.07589359

-0.20199988

1.17964166

EUN1

1.35962335

2.23650175

0.72281202

3.49636299

EUN2

0.98991578

1.15912983

0.52178983

1.98306165

EUNS

0.19328093

1.25758822

0.53848259

1.1846127

FMER

0.61803654

1.26193925

0.45073336

0.87046526

GAFF

-0.63045276

0.84041235

-0.73441831

0.66628712

GANG

-1.55723755

0.72298637

-0.97133988

0.88916481

GBOB

0.45885687

1.40085419

0.20046785

1.35311081

GBPA

0.1744218

1.00261435

0.51475908

1.08106109

GCLA

0.73556705

1.30680304

0.10382509

1.35392783

GCLE

0.21713187

1.86490119

0.05667609

1.36316481

GDES

0.02187602

1.82034249

-0.04437253

1.53814729

GGRA

-0.22469057

1.58819222

-0.6293819

0.90787974

GMIN

-0.06184666

2.17192701

0.27154901

1.72453327

GOMS

0.21019852

2.71532762

-0.09156017

1.25325696

GOS1

0.01060387

3.1186773

0.09376857

2.66492296

GPAR

-2.1249248

0.63651375

-2.44163132

0.552174

GPPS

0.01250605

1.81327464

0.29785281

2.34415315

GTNR

0.74480418

1.01933303

-0.38054423

0.62790591

HGHA

0.31502248

1.33656233

0.08167448

0.81663343

HLMO

-0.61110359

1.28662362

-0.20984541

0.91189638

HUCO

0.64604861

1.62677593

0.11859251

1.31535669

HUPA

0.26349096

2.23314695

0.52744375

1.15614826

LMUT

-0.62441958

0.75781164

-0.74747575

0.71190976

MPMI

0.25810959

2.3087737

0.16776841

1.4589151

NAFR

-0.01279898

1.38019217

0.46518935

0.97486648

NAMP

-0.09792614

0.71159325

0.00054093

0.68246909

NASP

-0.27652304

1.67776712

0.2787682

1.17691121

NCLA

-0.3169149

0.93240712

-0.0947643

1.08248973

NMCV

1.0927023

1.73249848

0.7521831

2.24516208

NCRX

-0.24592746

1.87350934

-0.20432555

2.50523524

NCRY

-0.6236393

0.96852663

-0.46922676

0.82258882

NCTE

-0.05880656

1.18349685

0.02539992

1.31380484

NCXM

0.18433666

1.4963295

0.09692593

1.37003365

NDAB

0.76175088

2.60732875

0.80373857

1.54544622

NDMA

-0.58359505

1.72979803

0.48111299

1.04858051

NERI

0.16279177

1.08834154

0.4450764

1.30161583

NESC

-0.18904057

1.12454415

0.1352761

1.17757496

NIFR

-0.40120884

0.92477164

-1.62118741

0.81923029

NIGE

-0.25624339

2.05928584

0.38248641

1.02244668

NINC

0.13778915

1.38794569

-0.0887378

1.13697926

NINT

-0.1057252

1.34844663

-0.39420277

1.02759339

NIS1

0.21736994

1.19014602

-0.08592648

1.19144719

NIS2

0.20963077

2.30578576

-0.05876786

1.70504918

NJAC

-0.60520332

1.31207651

0.07579254

1.28798093

NLIN

0.15372628

1.52829855

0.11893614

1.80476573

NLOR

0.22358044

1.91215411

0.17570978

3.99395739

NLUN

-0.71214384

2.23959575

-0.07214786

1.43373771

NNGO

0.31741873

1.54929309

0.20037915

1.59398561

NNOT

-0.09097505

2.1911816

0.17256191

0.86053692

NPAL

-2.11853229

0.60577139

-2.50631755

0.53717516

NQDJ

-0.5553273

1.10305222

-0.1782571

0.94749815

NROS

-0.51505617

1.46518459

-0.73906724

1.32711367

NSIA

0.05542487

1.21782479

0.09015215

1.41092633

NSLC

0.60410039

1.6149314

1.33192309

2.08260761

NTRO

-0.10003289

1.02528685

0.96262546

0.94265429

NUPS

0.08055923

1.48248075

0.52219243

0.92680161

NVIP

-0.14446247

1.14234134

-0.14225681

0.97845171

NZSS

0.31305619

1.81123691

0.65806006

1.40788336

PGIB

-2.34223959

0.90319285

-2.4887085

0.74634165

PINS

0.03052088

0.58077655

0.0354555

0.56757982

PLBI

0.50384018

2.76724465

0.8566843

1.63834889

PLFR

-0.08336512

1.9577574

0.2865272

1.76654446

PLHU

0.17872264

2.72224422

1.30264336

1.47207629

PRBU

0.43512913

1.68221938

0.57671116

3.20026698

PRST

0.77166871

1.30613183

0.65487538

1.23370237

RGBL

-0.33120096

1.96136947

-0.02055367

1.50909091

RHOS

0.42075351

0.82883051

0.03396943

1.07096008

RMUS

0.5890225

2.58544434

1.16322668

1.01916656

SELS

-0.39612988

0.61003061

-0.64388027

0.52448635

SMST

0.00380205

1.87860679

0.09909953

1.53498071

SPUP

-2.70553428

0.49426837

-3.3152771

0.45570401

SSEM

-1.20785627

0.62746757

-1.67985946

0.56415846

STAS

0.53908447

1.25695414

1.20315127

1.26388016

TDEB

0.42499197

1.02889258

0.68495366

1.18116438

TLEV

0.27750582

2.69386395

0.66889587

2.352944

TMUS

0.30128844

2.49050118

1.0260858

1.75150625

TWEI

-0.69537257

1.37793568

-0.56951414

1.33061577

UBIC

-0.32398005

2.2176863

0.06669816

1.46988083

UULN

-0.51707953

0.75118152

-0.23431252

0.65697709

guilds

bryophila

17.5

3.8

2.0

4.6

11

104.5

171.4

1.64

motile

0

ACHS

Achnanthes

sp.

15.0

3.5

1.8

4.3

11

72.2

133.3

1.85

low

0

ACOA

Achnanthes

coarctata

16.0

3.5

1.2

4.6

11

52.8

124.7

2.36

low

0

ACOP

Amphora

copulata

50.0

24.0

24.0

2.1

17

10053.1

3069.1

0.31

low

0

ACS1

Achnanthes

sp.1

15.0

3.5

1.8

4.3

11

72.2

133.3

1.85

low

0

ADCS

Achnanthidium

sp.

15.0

3.5

1.8

4.3

11

72.2

133.3

1.85

low

0

ADCT

Achnanthidium

catenatum

12.5

3.5

1.8

3.6

11

60.1

112.7

1.87

high

0

ADEG

Achnanthidium

exiguum

12.0

6.5

3.3

1.8

11

199.1

217.0

1.09

low

0

ADEU

Achnanthidium

eutrophilum

11.0

3.5

1.8

3.1

11

52.9

100.3

1.90

low

0

ADMA

Achnanthidium

macrocephalum

10.0

3.0

1.5

3.3

11

35.3

77.8

2.20

low

0

ADMI

Achnanthidium

minutissimum

15.0

3.5

1.8

4.3

11

72.2

133.3

1.85

low

0

ADMS

Adlafia

minuscula

12.0

4.5

1.5

2.7

11

63.6

123.7

1.94

motile

0

ADPY

Achnanthidium

pyrenaicum

14.0

4.5

2.3

3.1

11

111.3

164.3

1.48

low

0

ADSA

Achnanthidium

saprophilum

11.5

3.5

1.8

3.3

11

55.3

104.5

1.89

low

0

ADSH

Achnanthidium

subhudsonis

11.5

4.0

2.0

2.9

11

72.3

121.0

1.67

low

0

ADSP

Adlafia

sp.

18.5

3.0

1.5

6.2

11

65.4

137.8

2.11

motile

0

ADSU

Achnanthidium

subatomus

10.0

4.5

2.3

2.2

11

79.5

121.9

1.53

low

0

AFON

Amphora

fontinalis

57.0

13.0

13.0

4.4

17

3362.6

1789.7

0.53

low

0

AGON

Achnanthes

gondwana

92.5

20.5

10.0

4.5

11

14893.1

4753.6

0.32

low

0

AINA

Amphora

inariensis

19.0

4.5

4.5

4.2

17

134.3

206.8

1.54

low

0

AINF

Achnanthes

inflata

63.0

14.0

10.0

4.5

11

6927.2

2595.0

0.37

high

0

AMIA

Amphora

minutissima

23.5

6.0

6.0

3.9

17

295.3

342.3

1.16

low

0

AMID

Amphora

indistincta

13.0

3.0

3.0

4.3

17

40.8

94.2

2.31

low

0

AMIS

Achnanthes

minuscula

8.0

3.5

1.5

2.3

11

33.0

71.1

2.15

low

0

AMPS

Amphora

sp.

50.0

24.0

24.0

2.1

17

10053.1

3069.1

0.31

low

0

AMSM

Amphora

subatomus

14.5

2.5

2.5

5.8

17

31.6

86.7

2.74

low

0

AMUS

Adlafia

muscora

15.0

3.5

1.5

4.3

11

61.9

126.1

2.04

motile

0

APED

Amphora

pediculus

11.5

3.0

3.0

3.8

17

36.1

83.9

2.32

low

0

ARPU

Achnanthes

rupestoides

13.5

6.0

3.0

2.3

11

190.9

219.1

1.15

low

0

BPAX

Bacillaria

paxillifera

105.0

6.0

4.9

17.5

10

3111.0

2356.3

0.76

motile

0

BRSP

Brachysira

sp.

23.0

4.0

2.5

5.8

13

115.0

106.6

0.93

motile

0

CACM

Craticula

accomodiformis

32.5

10.0

2.0

3.3

13

325.0

342.0

1.05

motile

0

CAER

Caloneis

aerophila

19.5

4.0

2.0

4.9

11

122.5

196.3

1.60

motile

0

CAEX

Cymbella

excisa

27.0

8.0

3.1

3.4

15

520.7

336.9

0.65

low

0

CAL1

Caloneis

sp.1

31.5

6.5

2.6

4.8

11

410.8

474.1

1.15

motile

0

CAL3

Caloneis

sp.3

31.5

6.5

2.6

4.8

11

410.8

474.1

1.15

motile

0

CALS

Caloneis

sp.

31.5

6.5

2.6

4.8

11

410.8

474.1

1.15

motile

0

CAMB

Craticula

ambigua

61.5

17.0

3.0

3.6

13

1568.3

1093.4

0.70

motile

0

species

N_fix

ratioSV

Adlafia

Genus

width

ABRY

Code

length

surface

biovolume

shape_code (Leblanc et al. 2012)

ratioLW

thickness

Taxa list with their traits.

CAS1

Caloneis

sp.1

31.5

6.5

2.6

4.8

11

410.8

474.1

1.15

motile

0

CAS2

Caloneis

sp.2

31.5

6.5

2.6

4.8

11

410.8

474.1

1.15

motile

0

CBAC

Caloneis

bacillum

31.5

6.5

2.6

4.8

11

410.8

474.1

1.15

motile

0

CDUB

Cyclostephanos

dubius

20.0

20.0

3.0

1.0

4

942.5

816.8

0.87

planktic

0

CFON

Caloneis

fontinalis

20.0

4.5

3.0

4.4

11

212.1

256.8

1.21

motile

0

CHMC

Chamaepinnularia

muscicola

11.0

2.5

1.5

4.4

11

32.4

75.0

2.32

motile

0

CHYA

Caloneis

hyalina

21.0

5.0

2.5

4.2

11

206.2

267.0

1.30

motile

0

CLAU

Caloneis

lauta

36.5

7.0

3.0

5.2

11

602.0

606.3

1.01

motile

0

CLCT

Caloneis

lancettula

38.0

7.0

3.0

5.4

11

626.7

629.9

1.01

motile

0

CMEN

Cyclotella

meneghiniana

24.0

24.0

3.0

1.0

4

1357.2

1131.0

0.83

planktic

0

CMLF

Craticula

molestiformis

16.0

4.0

1.9

4.0

13

59.5

71.7

1.20

motile

0

CMOL

Caloneis

molaris

45.5

7.5

3.0

6.1

11

804.1

785.8

0.98

motile

0

COCE

Cyclotella

ocellata

15.5

15.5

2.5

1.0

4

471.7

499.1

1.06

planktic

0

COCS

Cocconeis

sp.

28.0

21.0

2.0

1.3

11

923.6

1077.6

1.17

low

0

COPL

Cocconeis

pseudolineata

23.0

15.0

3.5

1.5

11

960.5

753.5

0.78

low

0

CPEA

Cocconeis

placentula

28.0

21.5

3.0

1.3

11

1418.4

1178.9

0.83

low

0

CPLA

Cocconeis

placentula

44.0

24.0

3.5

1.8

11

2902.8

2032.6

0.70

low

0

CPLE

Cocconeis

placentula

28.0

21.5

3.0

1.3

11

1418.4

1178.9

0.83

low

0

CPLI

Cocconeis

placentula

45.0

23.0

3.0

2.0

11

2438.7

1946.2

0.80

low

0

CPTG

Cocconeis

placentula

44.0

24.0

3.5

1.8

11

2902.8

2032.6

0.70

low

0

CSMU

Chamaepinnularia

submuscicola

13.0

3.5

2.0

3.7

11

71.5

123.3

1.73

motile

0

CTRO

Cymbella

tropica

39.5

11.0

3.0

3.6

15

1023.8

578.7

0.57

low

0

CYMS

Cymbella

sp.

27.0

8.0

3.1

3.4

15

520.7

336.9

0.65

high

0

CYPS

Cymbellopsis

persantosana

24.5

5.5

3.0

4.5

15

317.5

250.0

0.79

motile

0

DCFD

Diadesmis

confervaceoides

18.5

7.0

3.0

2.6

11

305.1

323.6

1.06

motile

0

DCOF

Diadesmis

confervacea

18.5

7.0

3.2

2.6

11

326.7

332.1

1.02

motile

0

DCTG

Diadesmis

confervacea

18.5

7.0

3.2

2.6

11

326.7

332.1

1.02

motile

0

DDSP

Diadesmis

sp.

17.0

4.0

1.9

4.3

11

101.3

169.4

1.67

motile

0

DEHR

Diatoma

ehrenbergii

75.0

7.5

3.7

10.0

11

1621.1

1359.1

0.84

high

0

DIAM

Diadesmis

sp.

17.0

4.0

1.9

4.3

11

101.3

169.4

1.67

motile

0

DIPL

Diploneis

sp.

19.0

6.0

2.0

3.2

11

179.1

257.6

1.44

motile

0

DIPS

Diploneis

sp.

19.0

6.0

2.0

3.2

11

179.1

257.6

1.44

motile

0

DPSO

Diploneis

pseudovalis

23.5

11.5

6.8

2.0

11

1447.5

799.4

0.55

motile

0

DPUE

Diploneis

puella

19.0

11.0

4.0

1.7

11

656.6

516.8

0.79

motile

0

DSBO

Diploneis

subovalis

42.5

16.0

4.0

2.7

11

2136.3

1435.7

0.67

motile

0

DTEN

Denticula

tenuis

28.0

5.0

2.4

5.6

11

264.7

344.7

1.30

motile

0

DVUL

Diatoma

vulgaris

41.5

12.5

5.8

3.3

11

2382.1

1310.8

0.55

high

0

DWOL

Discostella

woltereckii

7.0

7.0

2.0

1.0

4

77.0

121.0

1.57

planktic

0

EBOT

Eunotia

botuliformis

24.0

3.4

3.4

7.1

10

277.4

349.5

1.26

high

0

EBST

Eunotia

biseriata

34.8

6.6

4.4

5.3

10

1001.7

820.1

0.82

high

0

ECIS

Epithemia

cistula

47.5

9.5

8.0

5.0

17

1288.4

967.0

0.75

low

1

EFAB

Eunotia

faba

38.0

7.0

7.0

5.4

10

1862.0

1162.0

0.62

high

0

EGRC

Epithemia

gracilis

42.5

9.5

8.0

4.5

17

1152.8

868.2

0.75

low

1

EIMP

Eunotia

implicata

30.0

4.5

4.5

6.7

10

607.5

580.5

0.96

high

0

EINC

Eunotia

incisa

34.0

5.0

5.0

6.8

10

850.0

730.0

0.86

high

0

ELEP

Eolimna

lepidula

11.0

4.5

2.3

2.4

11

87.5

132.5

1.52

motile

0

EMIN

Eunotia

minor

40.0

6.5

2.9

6.2

10

755.0

790.1

1.05

high

0

EMON

Eunotia

monodon

127.5

10.5

10.0

12.1

10

13387.5

5437.5

0.41

high

0

ENJV

Encyonema

jemtlandicum

42.0

10.0

4.0

4.2

11

1319.5

986.5

0.75

high

0

ENMI

Encyonema

minutum

15.0

5.5

2.6

2.7

11

167.3

212.7

1.27

high

0

ENMS

Encyonema

neomesianum

50.0

10.5

3.0

4.8

11

1237.0

1109.8

0.90

high

0

ENSP

Encyonema

sp.

15.0

5.5

2.6

2.7

11

167.3

212.7

1.27

high

0

EOMI

Eolimna

minima

11.5

3.5

2.2

3.3

11

69.1

114.7

1.66

motile

0

EOMT

Eolimna

minima

11.5

3.5

2.2

3.3

11

69.1

114.7

1.66

motile

0

EORH

Eolimna

rhombelliptica

7.0

3.0

1.5

2.3

11

24.7

56.5

2.29

motile

0

EORU

Eolimna

ruttneri

12.0

3.0

1.5

4.0

11

42.4

91.9

2.17

motile

0

EOS1

Eolimna

sp.1

11.5

3.5

2.2

3.3

11

69.1

114.7

1.66

motile

0

EOS2

Eolimna

sp.2

11.5

3.5

2.2

3.3

11

69.1

114.7

1.66

motile

0

EOS3

Eolimna

sp.3

11.5

3.5

2.2

3.3

11

69.1

114.7

1.66

motile

0

EOSP

Eolimna

sp.

10.0

4.5

1.8

2.2

11

62.8

111.2

1.77

motile

0

EPIS

Epithemia

sp.

10.8

10.0

5.0

1.1

17

281.4

189.5

0.67

low

1

ESBM

Eolimna

subminuscula

10.0

5.0

2.2

2.0

11

88.0

131.3

1.49

motile

0

ESIO

Eunotia

siolii

35.0

5.0

2.5

7.0

10

437.5

550.0

1.26

high

0

ESLE

Encyonema

silesiacum

29.0

8.0

4.0

3.6

11

728.8

596.9

0.82

high

0

ESOL

Eunotia

soleirolii

75.0

6.5

4.0

11.5

10

1950.0

1627.0

0.83

high

0

ESUM

Encyonopsis

subminuta

17.5

4.0

1.0

4.4

11

55.0

143.7

2.61

low

0

EUN1

Eunotia

sp.1

80.0

10.0

26.0

8.0

10

20800.0

6280.0

0.30

high

0

EUN2

Eunotia

sp.2

36.0

9.0

16.0

4.0

10

5184.0

2088.0

0.40

high

0

EUN3

Eunotia

sp.3

20.0

5.0

7.0

4.0

10

700.0

550.0

0.79

high

0

EUN4

Eunotia

sp.4

26.0

8.0

12.0

3.3

10

2496.0

1232.0

0.49

high

0

EUN5

Eunotia

sp.5

36.0

8.0

8.0

4.5

10

2304.0

1280.0

0.56

high

0

EUNS

Eunotia

sp.

34.0

3.5

3.5

9.7

10

416.5

500.5

1.20

high

0

EVCF

Eolimna

verecundaeformis

9.5

3.5

1.8

2.7

11

45.7

88.0

1.92

motile

0

EVIO

Eunotia

viola

50.0

6.7

4.0

7.5

10

1330.0

1118.2

0.84

high

0

EVUL

Encyonema

vulgare

44.0

10.5

5.0

4.2

11

1814.3

1153.7

0.64

high

0

FCAP

Fragilaria

capucina

30.0

4.5

2.0

6.7

11

212.1

320.4

1.51

high

0

FCRS

Frustulia

crassinervia

42.5

10.3

4.0

4.1

11

1368.6

1015.7

0.74

motile

0

FCVA

Fragilaria

capucina

30.0

4.5

2.5

6.7

11

265.1

347.5

1.31

high

0

FGOU

Fragilaria

goulardii

82.0

8.8

4.0

9.4

11

2254.1

1697.2

0.75

high

0

FINS

Fallacia

insociabilis

14.5

6.0

2.0

2.4

11

136.7

201.1

1.47

motile

0

FMER

Fallacia

meridionalis

14.0

6.0

3.0

2.3

11

197.9

226.2

1.14

motile

0

FMOC

Fallacia

monoculata

15.0

5.0

2.5

3.0

11

147.3

196.3

1.33

motile

0

FPEM

Fragilaria

perminuta

32.5

3.5

1.0

9.3

11

89.3

235.2

2.63

high

0

FPYG

Fallacia

pygmaea

36.0

15.0

3.0

2.4

11

1272.3

1088.6

0.86

motile

0

FRAS

Fragilaria

sp.

30.0

4.5

3.0

6.7

11

318.1

374.6

1.18

high

0

FRS1

Frustulia

sp.1

60.0

11.5

3.0

5.2

11

1625.8

1420.8

0.87

high

0

FRS2

Frustulia

sp.2

60.0

11.5

3.0

5.2

11

1625.8

1420.8

0.87

high

0

FRSP

Frustulia

sp.

60.0

11.5

3.0

5.2

11

1625.8

1420.8

0.87

high

0

FRUM

Fragilaria

rumpens

30.0

4.0

2.0

7.5

11

188.5

295.3

1.57

high

0

FSAP

Fistulifera

saprophila

6.0

3.0

1.0

2.0

11

14.1

42.4

3.00

motile

0

FSAX

Frustulia

saxonica

55.0

16.0

3.0

3.4

11

2073.5

1716.9

0.83

high

0

FSBH

Fallacia

subhamulata

18.5

5.5

3.0

3.4

11

239.7

272.9

1.14

motile

0

FSLU

Fallacia

sublucidula

9.5

4.5

2.0

2.1

11

67.2

111.1

1.65

motile

0

FTEN

Fragilaria

tenera

85.0

2.5

1.5

34.0

11

250.3

540.0

2.16

planktic

0

FVUL

Frustulia

vulgaris

60.0

11.5

3.0

5.2

11

1625.8

1420.8

0.87

high

0

GACU

Gomphonema

acuminatum

70.0

11.0

5.0

6.4

21

465.1

787.2

1.69

high

0

GAFF

Gomphonema

affine

65.0

9.5

4.0

6.8

21

299.1

628.0

2.10

high

0

GANG

Gomphonema

angustatum

33.5

7.5

4.0

4.5

21

119.7

262.8

2.20

high

0

GANT

Gomphonema

angustum

71.0

7.5

4.0

9.5

21

260.2

544.0

2.09

high

0

GARV

Gomphonema

archaevibrio

97.5

11.8

7.0

8.3

21

976.7

1180.6

1.21

high

0

GBOB

Gomphonema

bourbonense

25.0

4.5

2.4

5.6

21

32.5

116.6

3.59

high

0

GBPA

Gomphonema

brasiliense

26.0

4.0

4.0

6.5

21

50.3

113.1

2.25

high

0

GCLA

Gomphonema

clavatum

57.5

10.0

5.0

5.8

21

346.1

592.8

1.71

high

0

GCLE

Gomphonema

clevei

31.0

6.5

3.0

4.8

21

72.2

207.8

2.88

high

0

GCUN

Gomphonema

cuneolus

25.0

4.5

2.5

5.6

21

33.8

117.0

3.46

high

0

GDCL

Geissleria

declivis

20.5

6.5

2.0

3.2

11

209.3

294.1

1.41

motile

0

GDEC

Geissleria

decussis

21.0

7.5

2.5

2.8

11

309.3

359.3

1.16

motile

0

GDES

Gomphonema

designatum

25.0

4.5

2.4

5.6

21

32.5

116.6

3.59

high

0

GESP

Geissleria

sp.

21.0

7.5

2.5

2.8

11

309.3

359.3

1.16

motile

0

GEXL

Gomphonema

exilissimum

30.0

6.0

3.0

5.0

21

64.6

186.4

2.88

high

0

GGRA

Gomphonema

gracile

60.0

7.5

3.0

8.0

21

164.2

455.6

2.77

high

0

GINO

Geissleria

ignota

18.5

4.5

2.5

4.1

11

163.5

221.1

1.35

motile

0

GLIP

Gomphonema

lippertii

40.5

9.5

7.5

4.3

21

343.0

421.4

1.23

high

0

GMIN

Gomphonema

minutum

22.5

6.0

3.0

3.8

21

47.8

141.4

2.96

high

0

GOAH

Gomphosphenia

oahuensis

28.0

4.5

2.0

6.2

21

30.4

128.7

4.23

high

0

GOLI

Gomphonema

olivaceum

26.5

8.5

4.0

3.1

21

105.0

236.5

2.25

high

0

GOM1

Gomphonema

sp.1

26.5

8.5

4.0

3.1

21

105.0

236.5

2.25

high

0

GOM5

Gomphonema

sp.5

26.5

8.5

4.0

3.1

21

105.0

236.5

2.25

high

0

GOMS

Gomphonema

sp.

26.5

8.5

4.0

3.1

21

105.0

236.5

2.25

high

0

GOP1

Gomphosphenia

sp.1

7.0

3.0

2.0

2.3

21

4.8

23.8

4.97

high

0

GOS1

Gomphonema

sp.1

26.5

8.5

4.0

3.1

21

105.0

236.5

2.25

high

0

GPAR

Gomphonema

parvulum

23.0

6.0

3.0

3.8

21

48.9

144.4

2.95

high

0

GPLI

Gomphosphenia

lingulatiformis

44.5

8.0

4.0

5.6

21

171.2

367.4

2.15

high

0

GPPS

Gomphosphenia

sp.

13.5

3.5

2.0

3.9

21

11.2

50.1

4.49

high

0

GPRC

Gomphonema

procerum

31.0

4.5

2.2

6.9

21

36.8

142.9

3.88

high

0

GPRI

Gomphonema

pumilum

25.0

4.5

2.4

5.6

21

32.5

116.6

3.59

high

0

GPUM

Gomphonema

pumilum

25.0

4.5

2.4

5.6

21

32.5

116.6

3.59

high

0

GSPP

Gomphonema

saprophilum

23.0

6.0

3.0

3.8

21

48.9

144.4

2.95

high

0

GTER

Gomphonema

tergestinum

33.5

8.5

4.0

3.9

21

134.7

296.0

2.20

high

0

GTNR

Gomphosphenia

tenerrima

29.0

9.0

4.0

3.2

21

121.9

271.9

2.23

high

0

GUTA

Gomphonema

utae

31.5

6.0

4.0

5.3

21

90.7

200.1

2.21

high

0

GYAC

Gyrosigma

acuminatum

120.0

14.5

2.4

8.3

13

2118.5

1813.6

0.86

motile

0

GYOB

Gyrosigma

obtusatum

77.5

13.0

1.7

6.0

13

836.0

1040.1

1.24

motile

0

GYRS

Gyrosigma

sp.

98.8

13.8

2.0

7.2

13

1389.9

1408.8

1.01

motile

0

HAMP

Hantzschia

amphioxys

32.5

6.0

2.4

5.4

10

468.0

574.8

1.23

motile

0

HBRE

Humidophila

brekkaensis

32.5

4.0

4.0

8.1

10

520.0

552.0

1.06

motile

0

HGHA

Halamphora

ghanensis

25.5

5.3

5.3

4.8

17

250.0

325.2

1.30

motile

0

HLMO

Halamphora

montana

16.0

3.8

3.8

4.2

17

80.6

147.1

1.82

motile

0

HNOR

Halamphora

normanii

30.0

13.0

13.0

2.3

17

1769.8

985.8

0.56

motile

0

HSTU

Halamphora

subturgida

20.0

4.0

4.0

5.0

17

111.7

192.3

1.72

motile

0

HTRQ

Hydrosera

triquetra

60.0

60.0

80.0

1.0

10

288000.0

0.09

high

0

HUCO

Humidophila

contenta

9.0

2.5

2.5

3.6

10

56.3

26400. 0 102.5

1.82

motile

0

HUPA

Humidophila

pantropica

19.5

3.0

2.0

6.5

10

117.0

207.0

1.77

motile

0

HUPC

Humidophila

paracontenta

10.0

4.0

2.0

2.5

10

80.0

136.0

1.70

motile

0

HVEN

Halamphora

veneta

25.5

5.5

5.5

4.6

17

269.3

338.0

1.26

motile

0

KASU

Karayevia

suchlandtii

11.0

5.0

2.5

2.2

11

108.0

149.2

1.38

low

0

KOBG

Karayevia

oblongella

13.5

6.0

3.0

2.3

11

190.9

219.1

1.15

low

0

KOSU

Kobayasiella

subtilissima

28.0

5.0

2.0

5.6

10

276.0

410.1

1.49

low

0

LAEQ

Luticola

aequatorialis

13.0

8.0

5.0

1.6

11

408.4

328.3

0.80

motile

0

LHUN

Lemnicola

hungarica

25.5

6.0

3.0

4.3

11

360.5

388.8

1.08

low

0

LMUT

Luticola

mutica

18.0

6.5

2.3

2.8

11

212.1

272.6

1.29

motile

0

LSIM

Luticola

simplex

19.3

6.8

4.0

2.9

11

408.2

367.5

0.90

motile

0

LUSP

Luticola

sp.

37.5

10.5

3.0

3.6

11

927.8

844.7

0.91

motile

0

MAAT

Mayamaea

atomus

11.0

5.0

2.1

2.2

11

91.1

139.4

1.53

motile

0

MAEX

Mayamaea

excelsa

14.0

6.0

2.3

2.3

11

150.0

203.4

1.36

motile

0

MAFO

Mayamaea

fossalis

11.0

5.0

2.2

2.2

11

93.5

140.8

1.51

motile

0

MALC

Mayamaea

alcimonica

7.5

4.0

2.2

1.9

11

51.8

86.9

1.68

motile

0

MAYA

Mayamaea

sp.

7.5

4.0

2.2

1.9

11

51.8

86.9

1.68

motile

0

MPMI

Mayamaea

permitis

7.5

4.0

2.2

1.9

11

51.8

86.9

1.68

motile

0

MVAR

Melosira

varians

21.5

9.0

9.0

2.4

4

3267.5

1334.0

0.41

high

0

NAAM

Navicula

amphiceropsis

36.5

9.0

3.3

4.1

11

861.6

754.7

0.88

motile

0

NACD

Nitzschia

acidoclinata

29.0

3.5

1.1

8.3

11

90.3

217.3

2.41

motile

0

NAFR

Nitzschia

amphibia

28.0

5.0

2.4

5.6

11

262.3

343.6

1.31

motile

0

NAGN

Nitzschia

agnita

29.0

4.0

1.4

7.3

11

125.7

253.7

2.02

motile

0

NAMP

Nitzschia

amphibia

28.0

5.0

2.4

5.6

11

262.3

343.6

1.31

motile

0

NANT

Navicula

antonii

20.5

7.0

3.0

2.9

11

340.9

356.1

1.04

motile

0

NAS1

Navicula

sp.1

11.5

3.5

2.2

3.3

11

69.1

114.7

1.66

motile

0

NAS2

Navicula

sp.2

11.5

3.5

2.2

3.3

11

69.1

114.7

1.66

motile

0

NAS3

Navicula

sp.3

11.5

3.5

2.2

3.3

11

69.1

114.7

1.66

motile

0

NASP

Navicula

sp.

11.5

3.5

2.2

3.3

11

69.1

114.7

1.66

motile

0

NAV2

Navicula

sp.2

35.0

6.7

4.0

5.2

11

736.7

630.4

0.86

motile

0

NBCL

Nitzschia

bacillum

16.0

2.8

2.0

5.8

11

69.1

128.0

1.85

motile

0

NBRE

Nitzschia

brevissima

36.0

5.5

2.0

6.5

11

311.0

441.4

1.42

motile

0

NCHR

Navicula

chiarae

27.0

5.5

3.0

4.9

11

349.9

386.4

1.10

motile

0

NCLA

Nitzschia

clausii

37.5

4.0

2.0

9.4

11

235.6

366.0

1.55

motile

0

NCPL

Nitzschia

capitellata

45.0

5.5

2.2

8.2

11

431.2

564.7

1.31

motile

0

NCPR

Navicula

capitatoradiata

35.0

8.5

2.4

4.1

11

553.7

629.2

1.14

motile

0

NCRX

Navicula

crassuliexigua

10.5

4.5

2.5

2.3

11

92.8

133.1

1.43

motile

0

NCRY

Navicula

cryptocephala

30.0

6.0

2.4

5.0

11

338.5

418.1

1.24

motile

0

NCTE

Navicula

cryptotenella

27.0

6.0

2.4

4.5

11

303.2

378.0

1.25

motile

0

NCTO

Navicula

cryptotenelloides

13.5

4.0

1.9

3.4

11

78.5

135.7

1.73

motile

0

NCXM

Navicula

cruxmeridionalis

18.5

4.0

2.0

4.7

11

114.8

185.3

1.61

motile

0

NDAB

Naviculadicta

absoluta

15.0

5.5

2.0

2.7

11

129.6

194.0

1.50

motile

0

NDCM

Naviculadicta

cosmopolitana

14.3

4.3

2.0

3.3

11

97.6

156.2

1.60

motile

0

NDES

Nitzschia

desertorum

18.5

4.5

1.8

4.1

11

119.4

196.7

1.65

motile

0

NDIF

Navicula

difficillima

11.5

3.5

1.4

3.3

11

43.2

95.4

2.21

motile

0

NDIS

Nitzschia

dissipata

49.0

5.5

2.3

8.9

11

490.9

621.9

1.27

motile

0

NDMA

Nitzschia

dissipata

49.0

6.0

2.7

8.2

11

616.5

692.5

1.12

motile

0

NDVI

Naviculadicta

vitabunda

15.0

5.5

2.0

2.7

11

129.6

194.0

1.50

motile

0

NEAF

Neidium

affine

50.0

11.5

3.0

4.3

11

1354.8

1193.0

0.88

motile

0

NEAM

Neidium

ampliatum

70.0

19.0

3.0

3.7

11

3133.7

2508.6

0.80

motile

0

NECH

Navicula

eichorniaephila

36.0

6.0

3.0

6.0

11

508.9

537.2

1.06

motile

0

NEDG

Navicula

eidrigiana

40.0

7.0

2.4

5.7

11

529.2

617.5

1.17

motile

0

NEGE

Navicula

egregia

12.5

3.5

1.5

3.6

11

51.5

106.4

2.06

motile

0

NERI

Navicula

erifuga

30.0

6.0

2.4

5.0

11

338.5

418.1

1.24

motile

0

NESC

Navicula

escambia

41.5

8.0

3.0

5.2

11

782.3

754.8

0.96

motile

0

NEXI

Navicula

exilis

30.0

6.0

2.4

5.0

11

338.5

418.1

1.24

motile

0

NFIC

Nitzschia

filiformis

32.5

4.0

1.9

8.1

11

198.7

315.8

1.59

motile

0

NFLX

Nitzschia

flexoides

71.5

3.3

2.0

22.0

11

365.0

599.8

1.64

motile

0

NFON

Nitzschia

fonticola

37.5

4.0

2.3

9.4

11

270.2

385.1

1.43

motile

0

NGER

Navicula

germainii

33.0

8.0

2.0

4.1

11

423.3

546.2

1.29

motile

0

NGRE

Navicula

gregaria

27.5

7.5

2.4

3.7

11

380.9

453.3

1.19

motile

0

NHAN

Nitzschia

hantzschiana

29.0

4.0

1.8

7.3

11

165.7

276.5

1.67

motile

0

NIAR

Nitzschia

archibaldii

27.5

2.5

2.5

11.0

11

132.7

223.8

1.69

motile

0

NIFR

Nitzschia

frustulum

32.5

3.5

2.3

9.3

11

202.6

306.9

1.51

motile

0

NIGE

Nitzschia

ingenua

435.0

7.5

3.5

58.0

11

8968.3

7557.5

0.84

motile

0

NIGR

Nitzschia

gracilis

65.0

3.5

0.9

18.6

11

163.4

455.7

2.79

motile

0

NILA

Nitzschia

lacuum

15.0

2.5

1.2

6.0

11

35.3

91.9

2.60

motile

0

NINC

Nitzschia

inconspicua

12.5

3.5

2.0

3.6

11

69.9

119.8

1.71

motile

0

NINT

Nitzschia

intermedia

95.0

5.5

2.7

17.3

11

1114.5

1249.5

1.12

motile

0

NIOG

Nitzschia

oligotraphenta

37.5

3.5

1.5

10.7

11

154.6

302.8

1.96

motile

0

NIPF

Nitzschia

paleaeformis

29.5

5.0

1.6

5.9

11

179.9

315.8

1.76

motile

0

NIPU

Nitzschia

pusilla

20.5

4.0

2.3

5.1

11

145.3

215.6

1.48

motile

0

NIS1

Nitzschia

sp.1

42.5

4.0

2.3

10.6

11

307.1

435.0

1.42

motile

0

NIS2

Nitzschia

sp.2

60.0

3.0

2.0

20.0

11

282.7

480.7

1.70

motile

0

NIS3

Nitzschia

sp.3

42.5

4.0

2.3

10.6

11

307.1

435.0

1.42

motile

0

NISC

Nitzschia

scalpelliformis

65.0

6.0

2.9

10.8

11

885.9

935.2

1.06

motile

0

NISO

Nitzschia

solita

34.0

5.0

2.0

6.8

11

267.0

389.6

1.46

motile

0

NJAC

Navicula

jacobii

33.0

6.0

3.0

5.5

11

466.5

494.8

1.06

motile

0

NJUA

Navicula

juanitalinda

31.0

5.5

3.0

5.6

11

401.7

439.8

1.09

motile

0

NLAL

Nitzschia

labella

12.5

2.9

1.5

4.4

11

42.0

92.1

2.20

motile

0

NLAN

Navicula

lanceolata

49.0

10.5

2.4

4.7

11

963.7

1031.1

1.07

motile

0

NLBT

Nitzschia

liebetruthii

32.5

3.5

2.0

9.3

11

178.7

291.8

1.63

motile

0

NLGI

Neidium

longiceps

34.0

6.5

2.0

5.2

11

347.1

474.4

1.37

motile

0

NLIB

Navicula

libonensis

32.5

7.0

3.0

4.6

11

536.0

543.5

1.01

motile

0

NLIN

Nitzschia

linearis

131.0

5.5

2.3

23.8

11

1275.5

1615.0

1.27

motile

0

NLOR

Nitzschia

lorenziana

113.5

5.0

3.0

22.7

11

1337.1

1449.8

1.08

motile

0

NLST

Navicula

leptostriata

36.5

55.0

0.3

0.7

11

419.4

3191.6

7.61

motile

0

NLSU

Nitzschia

linearis

56.5

3.5

2.0

16.1

11

310.6

499.1

1.61

motile

0

NLUN

Navicula

lundii

24.0

5.0

2.0

4.8

11

188.5

279.6

1.48

motile

0

NMCV

Navicula

medioconvexa

14.5

4.0

2.0

3.6

11

91.1

149.2

1.64

motile

0

NMIC

Nitzschia

microcephala

13.0

3.0

2.4

4.3

11

73.0

121.2

1.66

motile

0

NNAN

Nitzschia

nana

77.5

4.0

2.0

19.4

11

486.9

743.0

1.53

motile

0

NNGO

Naviculadicta

nanogomphonema

14.0

5.0

2.5

2.8

11

137.4

184.6

1.34

motile

0

NNOT

Navicula

notha

25.5

5.0

2.0

5.1

11

200.3

296.1

1.48

motile

0

NPAD

Nitzschia

palea

37.0

4.0

1.6

9.3

11

184.6

334.7

1.81

motile

0

NPAE

Nitzschia

paleacea

31.5

3.0

2.3

10.5

11

167.3

270.6

1.62

motile

0

NPAL

Nitzschia

palea

42.5

4.0

2.3

10.6

11

307.1

435.0

1.42

motile

0

NPBY

Navicula

pseudobryophila

22.0

5.5

3.0

4.0

11

285.1

319.7

1.12

motile

0

NPNU

Navicula

perminuta

13.0

3.0

1.5

4.3

11

45.9

99.0

2.15

motile

0

NQDJ

Navicula

quasidisjuncta

23.5

6.0

2.5

3.9

11

276.9

337.3

1.22

motile

0

NRCH

Navicula

reichardtiana

16.0

5.5

2.0

2.9

11

138.2

205.8

1.49

motile

0

NRDI

Navicula

radiosiola

45.0

7.5

4.0

6.0

11

1060.3

860.0

0.81

motile

0

NREC

Nitzschia

recta

67.5

5.5

2.3

12.3

11

684.1

852.2

1.25

motile

0

NREV

Nitzschia

reversa

135.0

4.5

1.8

30.0

11

843.5

1341.7

1.59

motile

0

NROS

Navicula

rostellata

42.0

8.5

2.4

4.9

11

670.7

750.5

1.12

motile

0

NRUC

Navicula

ruttneri

50.0

8.0

2.4

6.3

11

758.7

848.3

1.12

motile

0

NSBR

Navicula

subrotundata

10.0

5.0

2.0

2.0

11

78.5

125.7

1.60

motile

0

NSGG

Navicula

supergregaria

38.0

8.0

4.0

4.8

11

955.0

766.5

0.80

motile

0

NSHR

Navicula

schroeteri

60.5

11.0

1.1

5.5

11

563.1

1166.4

2.07

motile

0

NSIA

Navicula

simulata

30.0

5.0

3.0

6.0

11

353.4

400.6

1.13

motile

0

NSIG

Nitzschia

sigma

999.0

14.5

0.2

68.9

11

1767.1

0

Nitzschia

sigmoidea

325.0

11.5

2.5

28.3

11

7283.0

13.0 2 0.99

motile

NSIO

23001. 1 7182.3

motile

0

NSLC

Navicula

salinicola

13.5

3.5

1.5

3.9

11

55.7

114.3

2.05

motile

0

NSMU

Navicula

submuralis

10.0

5.5

2.0

1.8

11

88.0

136.0

1.55

motile

0

NSOC

Nitzschia

sociabilis

40.0

4.0

2.4

10.0

11

307.1

420.2

1.37

motile

0

NSOL

Nitzschia

solgensis

20.0

5.5

2.4

3.6

11

203.4

267.1

1.31

motile

0

NSRH

Navicula

subrhynchocephala

37.5

7.5

2.0

5.0

11

441.8

583.2

1.32

motile

0

NSUA

Nitzschia

subacicularis

50.0

2.5

0.7

20.0

11

64.4

250.4

3.89

motile

0

NTEN

Navicula

tenelloides

17.5

3.5

2.2

5.0

11

107.6

170.0

1.58

motile

0

NTER

Nitzschia

terrestris

70.0

4.0

2.0

17.5

11

434.3

669.4

1.54

motile

0

NTPT

Navicula

tripunctata

50.0

8.0

2.4

6.3

11

758.7

848.3

1.12

motile

0

NTRO

Nitzschia

tropica

37.5

4.0

1.6

9.4

11

190.9

341.2

1.79

motile

0

NTUB

Nitzschia

tubicola

42.0

5.0

2.0

8.4

11

329.9

477.5

1.45

motile

0

NUP1

Nupela

sp.

14.0

3.0

1.0

4.7

11

33.0

92.7

2.81

motile

0

NUPS

Nupela

sp.

12.0

4.0

2.0

3.0

11

75.4

125.7

1.67

motile

0

NUS1

Nupela

sp.

14.0

3.0

1.0

4.7

11

33.0

92.7

2.81

motile

0

NUS2

Nupela

sp.

14.0

3.0

1.0

4.7

11

33.0

92.7

2.81

motile

0

NUSP

Nupela

subpallavicini

13.5

4.0

2.0

3.4

11

84.8

139.8

1.65

motile

0

NVDM

Navicula

vandamii

20.5

4.5

2.0

4.6

11

144.9

223.4

1.54

motile

0

NVEN

Navicula

veneta

21.5

5.5

2.4

3.9

11

219.1

285.8

1.30

motile

0

NVIP

Navicula

vilaplanii

14.5

3.0

1.1

4.8

11

37.7

98.7

2.62

motile

0

NVLC

Nitzschia

valdecostata

17.5

4.0

2.0

4.4

11

110.0

177.5

1.61

motile

0

NWIL

Navicula

wildii

36.5

7.0

3.5

5.2

11

702.3

640.5

0.91

motile

0

NXAS

Navicula

associata

16.0

6.5

2.0

2.5

11

163.4

234.0

1.43

motile

0

NXSH

Navicula

schmassmannii

8.0

2.8

2.0

2.9

11

34.6

68.3

1.98

motile

0

NZAB

Nitzschia

abbreviata

11.0

2.0

1.5

5.5

11

25.9

65.2

2.52

motile

0

NZIT

Nitzschia

inconspicua

12.5

3.5

2.0

3.6

11

69.9

119.8

1.71

motile

0

NZLB

Nitzschia

lange-bertalotii

35.0

4.5

1.6

7.8

11

194.8

345.1

1.77

motile

0

NZLT

Nitzschia

linearis

18.5

4.0

2.4

4.6

11

138.2

200.3

1.45

motile

0

NZSS

Nitzschia

sp.

42.5

4.0

2.3

10.6

11

307.1

435.0

1.42

motile

0

NZSU

Nitzschia

supralitorea

17.5

3.5

1.4

5.0

11

67.5

142.5

2.11

motile

0

PACR

Pinnularia

acrospheria

88.0

13.5

6.0

6.5

10

7128.0

3594.0

0.50

motile

0

PBOR

Pinnularia

borealis

33.0

9.0

4.2

3.7

10

1258.0

949.8

0.76

motile

0

PFTN

Pseudofallacia

tenera

18.0

6.5

3.0

2.8

11

275.7

299.2

1.09

low

0

PGIB

Pinnularia

gibba

94.0

10.0

5.0

9.4

10

4700.0

2920.0

0.62

motile

0

PGLC

Pinnularia

graciloides

68.5

12.0

6.0

5.7

10

4932.0

2610.0

0.53

motile

0

PGRO

Pinnularia

graciloides

83.5

11.0

6.7

7.6

10

6136.3

3099.7

0.51

motile

0

PGTR

Pinnularia

graciloides

93.5

12.0

6.0

7.8

10

6732.0

3510.0

0.52

motile

0

PIN1

Pinnularia

sp.1

33.0

9.0

4.2

3.7

10

1258.0

949.8

0.76

motile

0

PINS

Pinnularia

sp.

33.0

9.0

4.2

3.7

10

1258.0

949.8

0.76

motile

0

PLAS

Placoneis

sp.

27.5

7.5

3.0

3.7

11

486.0

488.9

1.01

motile

0

PLBI

Planothidium

biporomum

18.0

7.5

2.8

2.4

11

292.2

322.4

1.10

low

0

PLEV

Pleurosira

laevis

142.5

40.0

23.5

3.6

11

105177.4

0.15

high

0

PLFR

Planothidium

frequentissimum

17.0

6.0

2.1

2.8

11

172.0

15688. 6 237.8

1.38

low

0

PLHU

Platessa

hustedtii

13.5

6.0

2.4

2.3

11

155.5

202.1

1.30

low

0

PLRS

Pleurosira

sp.

8.5

4.3

2.0

2.0

11

56.7

96.8

1.71

high

0

PLSS

Pleurosigma

sp.

95.0

15.0

7.0

6.3

13

4987.5

1593.3

0.32

motile

0

PMIC

Pinnularia

microstauron

55.0

9.0

4.0

6.1

10

1980.0

1502.0

0.76

motile

0

POBF

Pinnularia

obscuriformis

24.5

6.0

2.1

4.1

10

310.0

422.6

1.36

motile

0

PRBU

Planothidium

robustius

23.0

7.5

2.8

3.1

11

373.1

402.9

1.08

low

0

PRST

Planothidium

rostratum

15.0

6.0

3.0

2.5

11

209.7

239.2

1.14

low

0

PSAL

Pleurosigma

salinarum

95.0

15.0

7.0

6.3

13

4987.5

1593.3

0.32

motile

0

PSEL

Pinnularia

subcapitata

42.0

6.0

2.9

7.0

10

733.0

783.2

1.07

motile

0

PSHO

Pinnularia

schoenfelderi

28.0

6.0

2.0

4.7

10

330.0

469.6

1.42

motile

0

RABB

Rhoicosphenia

abbreviata

42.5

5.5

3.0

7.7

11

550.8

593.4

1.08

low

0

RGBL

Rhopalodia

gibberula

105.0

11.0

11.0

9.5

17

4434.9

2738.3

0.62

motile

1

RGIB

Rhopalodia

gibba

161.0

10.0

10.0

16.1

17

5620.0

3802.3

0.68

motile

1

RHIR

Rhopalodia

hirundiniformis

107.3

30.0

5.0

3.6

10

16095.0

7811.0

0.49

motile

1

RHOS

Rhopalodia

sp.

161.0

24.0

20.0

6.7

17

27627.6

8170.2

0.30

motile

1

RHPS

Rhoicosphenia

sp.

38.7

6.0

4.0

6.4

11

728.8

645.1

0.89

low

0

RMUS

Rhopalodia

musculus

46.0

25.0

20.0

1.8

17

8281.6

2546.5

0.31

motile

1

ROPE

Rhopalodia

operculata

35.0

19.5

15.0

1.8

17

3714.8

1475.7

0.40

motile

1

RRUP

Rhopalodia

rupestris

35.0

6.5

4.0

5.4

17

354.1

402.5

1.14

motile

1

RSIN

Reimeria

sinuata

24.5

6.5

2.5

3.8

11

318.9

374.3

1.17

low

0

SCON

Staurosira

construens

19.5

7.0

3.2

2.8

20

346.4

354.0

1.02

high

0

SEL1

Sellaphora

sp.

11.5

3.5

2.2

3.3

11

69.1

114.7

1.66

motile

0

SELS

Sellaphora

sp.

11.5

3.5

2.2

3.3

11

69.1

114.7

1.66

motile

0

SHAN

Stephanodiscus

hantzschii

17.5

17.5

2.4

1.0

4

577.3

613.0

1.06

planktic

0

SIDE

Simonsenia

delognei

11.0

2.0

2.4

5.5

11

41.6

83.8

2.01

motile

0

SJOU

Sellaphora

joubaudii

10.0

4.0

1.6

2.5

11

51.1

98.6

1.93

motile

0

SMST

Seminavis

strigosa

39.5

5.8

5.8

6.9

17

455.9

541.2

1.19

motile

0

SPIN

Staurosirella

pinnata

19.0

4.0

4.0

4.8

11

238.8

263.9

1.11

high

0

SPOS

Staurophora

sp.

34.7

8.7

5.0

4.0

11

1179.0

812.2

0.69

motile

0

SPUP

Sellaphora

pupula

50.0

12.5

1.9

4.0

11

929.1

1167.6

1.26

motile

0

SSEM

Sellaphora

seminulum

12.0

3.5

1.6

3.4

11

54.2

106.0

1.96

motile

0

SSLE

Staurosira

leptostauron

15.0

12.5

5.1

1.2

20

755.6

524.2

0.69

high

0

SSPE

Staurosira

sp.

7.0

4.5

10.0

1.6

20

247.4

245.8

0.99

high

0

STAN

Stauroneis

anceps

49.5

12.0

3.0

4.1

11

1399.6

1222.9

0.87

motile

0

STAS

Stauroneis

sp.

49.5

12.0

3.0

4.1

11

1399.6

1222.9

0.87

motile

0

STHE

Stauroneis

thermicola

12.5

4.0

1.6

3.1

11

62.0

119.5

1.93

motile

0

STKR

Stauroneis

kriegeri

20.5

5.0

2.0

4.1

11

161.0

241.1

1.50

motile

0

SURS

Surirella

sp.

28.0

10.0

3.7

2.8

11

812.1

660.3

0.81

motile

0

SVER

Sellaphora

verecundiae

14.5

5.5

2.3

2.6

11

142.2

196.6

1.38

motile

0

TAPI

Tryblionella

apiculata

39.0

7.0

2.4

5.6

10

656.0

767.1

1.17

motile

0

TBCO

Tryblionella

constricta

39.0

7.0

3.0

5.6

10

819.0

822.0

1.00

motile

0

TCAL

Tryblionella

calida

48.5

8.5

2.0

5.7

10

824.5

1052.5

1.28

motile

0

TDEB

Tryblionella

debilis

19.5

8.5

2.0

2.3

10

331.5

443.5

1.34

motile

0

TLEV

Tryblionella

levidensis

41.5

15.5

2.3

2.7

10

1479.0

1548.6

1.05

motile

0

TMUS

Terpsinoe

musica

90.0

40.0

90.0

2.3

11

254469.0

0.09

high

0

TWEI

Thalassiosira

weissflogii

18.0

18.0

2.0

1.0

4

508.9

24033. 2 622.0

1.22

planktic

0

UBIC

Ulnaria

biceps

455.0

7.5

3.7

60.7

10

12752.0

0.81

high

0

UDEA

Ulnaria

delicatissima

340.0

3.5

0.3

97.1

10

303.0

10281. 6 2554.9

8.43

high

0

UULN

Ulnaria

ulna

313.5

5.5

2.7

57.0

10

4724.0

5196.5

1.10

high

0

Pearson's correlation between the log transformed biovolume and the surface-tovolume ratio (p < 0.05, r = − 0.85).

The abundance of N-fixing species along the total nitrogen gradient (p = 0.12, r = −0.13).

Calculation of quality values based on the trait classes and the final trait-based index (Idx.Mtrait). Equation

𝑥𝑠𝑖𝑧𝑒

B0_nutr

B0_org

B1_nutr

B1_org

𝑠12 ln − 𝐵0 −𝑠12 + 1 = 𝐵1

-0.23

-0.21

0.66

1.38

0.12

0.32

𝑥_𝑐𝑎𝑙𝑠𝑖𝑧𝑒 = 𝑎 ∗ 𝑥𝑠𝑖𝑧𝑒 + 𝑏

𝑙𝑤12 − 𝐵0 −𝑙𝑤12 + 1 𝐵1

-0.22

-0.42

-0.96

-2.44

-0.16

-0.50

𝑥_𝑐𝑎𝑙𝑙𝑤 = 𝑎 ∗ 𝑥𝑙𝑤 + 𝑏

𝑥𝑙𝑤 =

𝑥𝑙𝑜𝑤 =

𝑥𝑚𝑜𝑡𝑖𝑙𝑒 =

ln

ln

ln

PseudoR2_nutr

PseudoR2_org

Calibration (0-20)

anutr

5.85

𝐺𝑙𝑜𝑤 − 𝐵0 −𝐺𝑙𝑜𝑤 + 1 𝐵1

-0.56

-0.80

-0.98

-2.27

-0.15

-0.45

𝑥_𝑐𝑎𝑙𝑙𝑜𝑤 = 𝑎 ∗ 𝑥𝑙𝑜𝑤 + 𝑏

𝐺𝑚𝑜𝑡𝑖𝑙𝑒 − 𝐵0 −𝐺𝑚𝑜𝑡𝑖𝑙𝑒 + 1 𝐵1

-0.45

-0.46

0.80

1.42

0.17

0.34

𝑥_𝑐𝑎𝑙𝑚𝑜𝑡𝑖𝑙𝑒 = 𝑎 ∗ 𝑥𝑚𝑜𝑡𝑖𝑙𝑒 + 𝑏

aorg

4.76

bnutr

5.88

1

xsize: ecological value based on the size metric s12: is the sum relative biomass of taxa assigned to the s1 and s2 groups, xlw: ecological value based on the length-to-width ratio metric lw12: sum relative biomass of taxa assigned to the lw1 and lw2 groups xlow: ecological value based on the low profile guild metric Glow: relative biomass of taxa assigned to the low-profile guild B0 and B1 are standard values of the model. x_calsize: calibrated value of xsize in a range of 0-20 x_callw: calibrated value of xlw in a range of 0-20 x_callow: calibrated value of xlow in a range of 0-20 x_calmotile: calibrated value of xmotile in a range of 0-20 a and b are the standards of the linear correlation used in the calibration of ecological values on a scale of 0-20

borg

9.20

Appendix C

The map of Mayotte with the sampling sites and rivers indicated

rbcL primers, reactions mix and condition used during the PCR amplification of the rbcL 312bp fragment. Information is provided for 1 reaction in a final volume of 25μL.

Primer name Forward

Reverse

Diat_rbcL_708F_1

Primer sequence (5' - 3') AGGTGAAGTAAAAGGTTCWTACTTAAA

Length (bp) 27

Diat_rbcL_708F_2 Diat_rbcL_708F_3 R3_1 R3_2

AGGTGAAGTTAAAGGTTCWTAYTTAAA AGGTGAAACTAAAGGTTCWTACTTAAA CCTTCTAATTTACCWACWACTG CCTTCTAATTTACCWACAACAG

27 27 22 22

Reagents dNTP TaKaRa LA Taq polymerase H2O molecular grade Buffer Forward (Diat_rbcL_708F_1 + _2 + _3) Reverse (R3_1 + R3_2) Bovine Serum Albumin (BSA) DNA

Initial conc. 2.5 mM 5U 10X 10 µM 10 µM 10 mg/mL 25 ng/µL

Final conc. 0.2 mM 0.75 U 1X 0.5 µM 0.5 µM 0.5 mg/mL 2 ng/µL

Step 1

Time (s) 900

Temperature (°C) 95

2 3 4 5

45 45 45 300

95 55 72 72

Cycles

X 30

Volume (µL) 2 0.15 15.6 2.5 1.25 1.25 1.25 1

Appendix D

CONCEPTS AND QUESTIONS

Freshwater biomonitoring in the Information Age

1

François Keck1,2*, Valentin Vasselon1, Kálmán Tapolczai1, Frédéric Rimet1, and Agnès Bouchez1 Freshwaters worldwide face serious threats, making their protection increasingly important. Freshwater monitoring has historically produced valuable data and continues to develop. Rapid improvements to ­biomolecular techniques are revolutionizing the way scientists describe biological communities and are bringing about major changes in biomonitoring. Combined with high-­ throughput sequencing, DNA metabarcoding is fast and cost-­effective, generating massive amounts of data. In a world with numerous ­ecological threats, “big data” constitute a tremendous opportunity to improve the efficiency of biological monitoring. These fundamental changes in biomonitoring will require freshwater ecologists and environmental managers to reconsider how they handle large amounts of data. Front Ecol Environ 2017; doi:10.1002/fee.1490

H

uman activities have broadly affected freshwater   ecosystems, especially since the Industrial Revolution. Over the past 50 years, however, policy makers and citizens have become more attuned to ­ ­environmental issues. This has led to the development of important governmental programs to assess and limit ecological impacts of human activities (Figure 1). In this context, one objective of environmental managers is to evaluate how water quality changes over time. Bioindicator organisms are commonly used for this pur­ pose, based on the premise that the presence or absence of certain biological communities at a given site reflects its environmental quality. Freshwater biomonitoring has a long tradition in the field of ecology. A century of research has led to substan­ tial improvements in understanding how human distur­ bances can shape biological communities. Based on this knowledge, many approaches have been developed to estimate environmental quality from the richness, diver­ sity, structure, and functioning of these communities

In a nutshell: • DNA metabarcoding and high-throughput sequencing methods produce massive quantities of data and will mark­ edly change freshwater biomonitoring • Molecular methods propel biomonitoring into the Information Age and bring exciting new opportunities to make ecological monitoring more effective and relevant • Genetic “big data” challenge scientists to think differently about the way that biological monitoring information is analyzed; we propose and discuss alternatives to the classical taxonomic affiliation approach to process bioassessment metabarcoding data

1

UMR CARRTEL, Institut National de la Recherche Agronomique, Université Savoie Mont Blanc, Thonon-Les-Bains, France; 2 Department of Aquatic Sciences and Assessment, Swedish University of Agricultural Sciences, Uppsala, Sweden *([email protected]) © The Ecological Society of America

(Jørgensen et al. 2010). These widely used methods are based on solid theoretical grounds and are known to ­perform quite well. Most of them commonly require a taxonomical description of the community. Hence, fresh­ water biomonitoring essentially consists of collecting individual organisms, performing taxonomic identifica­ tion, and using inventories to estimate the environ­ mental condition of a given site. However, traditional biomonitoring also faces recurrent criticisms, mainly ­ related to taxonomic identification relying on mor­ phological criteria, a process that is time-­ consuming, complex, and technically demanding (Mandelik et al. ­ 2010). These limits inevitably restrict the number of sites that can be monitored and the frequency of controls. During the past decade, the idea arose that DNA anal­ yses (Figure 2) could advantageously replace morphologi­ cal methods to identify species (Hebert et al. 2003). Metabarcoding was developed as a set of techniques to identify multiple taxa simultaneously from an environ­ mental sample with standard genetic markers (Taberlet et al. 2012; Panel 1 and Figure 2). This has led to the idea of “Biomonitoring 2.0”, which offers novel perspectives for monitoring environmental communities (Baird and Hajibabaei 2012). In this paper, we explain why and how metabarcoding will profoundly change the nature of data produced by biomonitoring. We examine these changes in the general context of massive data production – so-­ called “big data”, a topic that is the subject of increasing interest in biology (Marx 2013). We show why this big data revolution holds promise for ecological assessment purposes. Finally, we highlight three challenges posed by big data for metabarcoding and propose a framework that takes them into account. We illustrate our point with examples taken from freshwater monitoring, where metabarcoding is developing rapidly (Hajibabaei et al. 2011; Kermarrec et al. 2014). Nevertheless, the ideas ­discussed could be extended and applied to a broader ­context. www.frontiersinecology.org

Freshwater biomonitoring in the Information Age

2

(a)

F Keck et al. (b)

Figure 1. Two streams included in the river monitoring network of Mayotte Island, France. (a) A pristine upstream site (Longoni River) and (b) a polluted site located downstream of village waste (Majimbini River). The biological assessment of Mayotte’s rivers currently relies on benthic diatom communities studied using both classical morpho-­taxonomical and metabarcoding approaches.

can now be identified through the use of DNA barcodes (Hebert et al. 2003); for definitions of selected specialist Characterizing ecological quality from biological entities terms used throughout, see Panel 1. The introduction of has produced important sources of data since the first high-­throughput sequencing (HTS; Shokralla et al. 2012) attempts to do so at the beginning of the 20th century. coupled with the development of extended reference This is because biomonitoring largely consists of sam­ databases (Ratnasingham and Hebert 2007; Benson et al. pling, identifying, enumerating, and reporting biological 2008) and efficient bioinformatics tools (eg Schloss et al. organisms. The saprobic system for organic pollution 2009) have enabled the production of reliable and cost-­ assessment developed by Kolkwitz and Marsson (1908, effective community inventories from environmental 1909) is often cited as the first bioassessment tool in DNA (Chariton et al. 2015; Gibson et al. 2015; Pawlowski freshwaters and uses 298 plant species and 527 animal et al. 2016). While numerous issues and technical limita­ species as indicator organisms. Methods soon diversified tions remain (DNA spatial transfer and persistence over thereafter, and specific biological groups (fishes, mac­ time, polymerase chain reaction [PCR] amplification roinvertebrates, algae) have been employed. Increasing biases, sequencing errors, chimeras, quantification; see stringency in precision requirements has led to more also Coissac et al. 2012 and Shokralla et al. 2012), meth­ powerful and sophisticated tools, based on hundreds ods are improving quickly and metabarcoding is expected of families and thousands of species. to be an increasingly important component of biomoni­ The amount of data produced has increased rapidly toring in the future. because biomonitoring is rarely done in isolation, but The progressive adoption of metabarcoding for taxonom­ instead is replicated across space (through a network of sites; ical identification will substantially increase the volume of eg along a river, within a watershed) and over time (long-­ data produced by biomonitoring activities and modify the term monitoring). Since the 1970s, general awareness of characteristics of these data (Dafforn et al. 2016). It is often ecological issues has grown, and biomonitoring has been stated that characteristics of big data fulfill five “Vs”: vol­ increasingly implemented and incorporated into legal frame­ ume, velocity, variety, variability, and value (Fan and Bifet works for fresh waters, such as the Clean Water Act (CWA, 2013). Biomonitoring data will likely meet these five crite­ 1972) in the US and the Water Framework Directive ria in unprecedented ways in the coming years. (WFD, 2000) in Europe. This guarantees the abundant pro­ duction of data with respect to recognized standards. Volume However, biomonitoring methods are expected to change considerably in coming years. After a century of The amount of data acquired from biomonitoring is classifying taxa based on morphological criteria, species expected to increase very quickly. HTS techniques are JJ Biomonitoring

as a source of massive data

www.frontiersinecology.org

© The Ecological Society of America

F Keck et al.

Freshwater biomonitoring in the Information Age

3

Panel 1. Biomonitoring and metabarcoding The biological monitoring of freshwater systems is traditionally based on the morphological identification of indicator species, which provides information on the ecological status of their environment. Instead of relying on morphological features (eg size, shape) to perform species identification, which requires specialized knowledge of taxonomic groups, small DNA fragments – about 300 base pairs in length, known as DNA barcodes – can be used (Hebert et al. 2003). This identification approach is termed DNA barcoding. Existing DNA barcode reference databases are based on different genes (including CO1, 18S, and rbcL) and link species taxonomy to DNA barcodes. While DNA barcoding is useful for identifying individual specimens, its application to community-­level samples (ie multiple species) was difficult because it required sorted samples or even iso- Figure 2. Several steps are required to perform DNA metabarcoding: (i) the sampling of lating and cultivating i­ndividuals. This environmental samples (eg sediment, biofilm, water) or the creation of bulk samples (mix challenge was overcome through a of individual specimens); (ii) the extraction of the DNA; (iii) the amplification of a metagenomic method called metabar- DNA barcode specific to the targeted community using polymerase chain reaction (PCR) coding, which allows for the detection techniques; (iv) the sequencing of the amplicons (amplified DNA barcodes); and (v) the of all species found in one sample taxonomical assignment of the DNA reads (amplicon sequences) using bioinformatics directly from their DNA barcode seand a reference database (database connecting DNA barcode sequences to their quences using a single workflow. The DNA is extracted directly from the taxonomic identity). Total environmental DNA comprises “endogenous” DNA from sample, followed by the amplification living organisms and “exogenous” free DNA. and sequencing of the targeted DNA (Taberlet et al. 2012) and facilitate access to uncultured taxa. barcode (Figure 2). Using bioinformatics tools, DNA barcodes For example, diatom molecular inventories can be used to calare compared to those contained in a reference database to culate a quality index that indicates the ecological status of the identify the species composition within the sample. sampled river (Kermarrec et al. 2014; Visco et al. 2015). PreciEnvironmental DNA was defined by Taberlet et al. (2012) as sion and reliability of the species list obtained from DNA methe “DNA that can be extracted from environmental samples tabarcoding depend on the completeness and reliability of the (such as soil, water, or air), without first isolating any target reference database. organisms”. This includes DNA from microorganisms and free The development of high-­ throughput sequencing (HTS) DNA. The free part of environmental DNA may be used to enables the rapid and inexpensive sequencing of hundreds ­ detect the presence of invasive species (Ficetola et al. 2008) of environmental samples at a time, making the incorporaor to monitor rare and indicator species (Mächler et al. 2014). tion of the DNA metabarcoding into biomonitoring programs Microorganisms present in environmental samples (eg bacteria, ­possible. fungi, and diatoms) enable the use of longer DNA barcodes

developing rapidly and have extremely high-­throughput (Figure 3d). With the development of standardized protocols, the processing rate will also probably increase considerably and allow more sites to be surveyed and with greater frequency. Finally, assessments that rely on morphological criteria alone tend to underestimate species diversity, whereas the level of diversity detected by genetic methods tends to be much higher, especially for microbial communities (Caron et al. 2009), leading to larger inventory tables. © The Ecological Society of America

Velocity

Traditional monitoring requires experts to undertake a long and laborious process of taxonomically identifying collected biota. Consequently, one site is typically monitored seasonally or yearly. With metabarcoding and HTS techniques, however, the identification process is automated and faster. This will allow sites to be monitored at a finer time scale and to approach real-­ time monitoring. www.frontiersinecology.org

Freshwater biomonitoring in the Information Age

4

(a)

(c)

(b)

(d)

F Keck et al.

with big data, including monitoring over space and time; examining multi-­trophic food web structure; and assessing the effects of pollution, environmental restoration, and in­ vasive species. Moreover, biomoni­ toring data are often exploited by ecologists for purposes other than environmental assessment, such as studying biodiversity patterns or val­ idating theoretical models (Lovett et al. 2007; Lindenmayer and Likens 2010). JJ Modern

techniques and big

data Increasing the number of indicators Figure 3. The Information Age is characterized by rapid technological developments exponentially increasing scientists’ capacity to produce, store, and process data. (a) Storage capacity of commercialized computer hard drives in gigabytes (dots) and average price of a gigabyte (dashed line). (b) Microprocessor performance (dots) in millions of floating-­point operations per second (MFLOPS) and average price of MFLOPS (dashed line). (c) Number of entries in the open-­access nucleotide sequence database GenBank. (d) The throughput and read length evolution of high-­throughput sequencing technologies. Variety

Biomonitoring elicits multiple types of data. Community inventories generally come in the form of presence– absence or count data tables. Environmental managers often prefer to rely on multiple biological indicators (eg fishes and macroinvertebrates) to monitor multiple sources of impairment. Moreover, assessment methods commonly integrate physical and chemical data, which may also constitute big data, especially when recorded with remote sensors and with high frequency. Metabarcoding will also make it possible to work with genetic data and phylogenies (Hajibabaei et al. 2007). Variability

Biomonitoring data are valuable when there is variability in community structures between reference and impacted sites (Jørgensen et al. 2010). With the use of DNA, finer-­ scale taxonomic characterization of communities can be achieved. Thus, with appropriate analyses, it will be possible to differentiate communities in a subtler way (Stein et al. 2014a) and to gain capacity in dis­ tinguishing the effects of various pressures. Value

Data produced by biomonitoring are used to assess en­ vironmental quality. Many applications could be enhanced www.frontiersinecology.org

The modern concept of biomonitor­ ing – as implemented in the WFD and CWA – is to use biological indicators accompanied by hydromor­ phological and physicochemical measurements (Ibáñez et al. 2010). For example, the WFD’s bioindicators (or biological quality elements [BQEs]) are fishes, macrophytes, macroinvertebrates, ben­ thic diatoms, and phytoplankton. Each of these indicators presents advantages (eg diversity, ubiquity, ecological importance) and disadvantages (sampling difficulties, lack of metrics) (Resh 2008). Each BQE can indicate different pressures and provide complementary information (Passy et al. 2004; Figure 4). Thus, the overall quality assessment of an aquatic ecosystem is based on the results of all BQEs. In the WFD, the “one out all out” (OOAO) rule states that the worst status of the BQEs used in the assessment determines the final status of the eco­ system. However, in practice, using all BQEs for a sampled site is seldom or only partly achieved because of both financial and logistical constraints (Birk et al. 2012). There is a trade-­off between the ease of sampling and the ease of identifying organisms with respect to the aver­ age size of different BQEs (Figure 4). Groups of organisms with larger individual body size (typically fishes) are more difficult to sample representatively and collect, whereas smaller or microscopic organisms such as macroinverte­ brates or benthic diatoms are relatively easy to collect by sampling the substrate directly. On the other hand, larger organisms are easier to manipulate and identify. For fishes and macrophytes, identification is performed in situ, whereas macroinvertebrates, benthic diatoms, and phyto­ plankton require arduous laboratory-­based work (chemi­ cal treatment, microscopy). Modern molecular techniques appear to offer a promising solution to the trade-­ off between the ease of sampling and identifying organisms. © The Ecological Society of America

F Keck et al.

Freshwater biomonitoring in the Information Age

5

Covering a larger diversity

In traditional biomonitoring, taxo­ nomical identification is rarely per­ formed at the most precise levels of specificity because doing so is cost-­ prohibitive. DNA metabarcod­ ing, however, could reveal diversity at the finest level for a fraction of that cost. With appropriate libraries, DNA barcodes can be linked to a Linnaean taxonomic name. The precision of taxonomic affiliation depends on the selected barcode and the availability of data in the ref­ erence libraries. By using correctly populated libraries, it is possible to reach the species level (eg Hajibabaei et al. 2011; Kermarrec et al. 2014) with less ambiguity and discrepancy than with classical microscopy, where species-­ level identification is often extremely laborious and even im­ possible at some development stages. However, data derived from DNA Figure 4. Gradients, trade-­offs, and complementarity between body size, ease of carry much more information than sampling, ease of identification, and the indicated pressures of the five indicators included taxonomic names alone. Baird and in the Water Framework Directive. Hajibabaei (2012) emphasized that genetic techniques have far more potential for identifying obtain an improved and integrated view of environmental taxa than the traditional approach of relying on mor­ quality, researchers must augment the number of sampling phological characteristics. DNA-­based techniques should sites to account for the spatial heterogeneity of the facilitate working at the infra-­species level and ultimately broader area. This increases the resolution of the grid of at the nucleotide level. It will therefore be possible to sampled sites and enables better interpolations among the disentangle cryptic species complexes and to perform nodes of the monitored network. For a given site, the fre­ population-­level analyses. Having the capacity to monitor quency of sampling is also important. A more frequent diversity at so many levels should also promote the sampling protocol gives a more reliable picture of the tem­ development of very sensitive tools to monitor the effects poral evolution of the site’s environmental quality. This is especially relevant for microscopic communities, which of specific types of pollution on various biota. change extremely quickly with changes in the environ­ ment. Thus, sampling plans with higher spatial and tem­ Enforcing and extending monitoring networks poral resolution should enable the development of more High-­throughput sequencing and the evolution of lab­ complex spatiotemporal models and increase the capacity oratory methods have made metabarcoding much more to detect the effects of local and diffuse pollution. cost-­effective (Stein et al. 2014b), and prices continue to decrease as technologies develop (van Dijk et al. JJ Taking advantage of the data deluge: a proposed 2014). DNA-­based methods are also much faster than framework traditional methods. Sample processing can be serialized and automated with the aid of robots (Chapman 2003). From morphology to genetics: beyond the classical Reductions in cost and processing time should boost concept of species sampling efforts by making it possible to increase the number of sites being monitored and the sampling Conventional taxonomy aims to classify biological organ­ frequency. This is an advantageous consequence of using isms in different groups based on shared traits. These metabarcoding, because biomonitoring often lacks spatial groups correspond to the different taxonomic levels, with the species level as a central unit. Even if still under and temporal representativeness. One specific site will poorly represent an entire ecosys­ debate (De Queiroz 2007), the concept and definition tem, particularly when habitats therein are heterogeneous of species provides scientists with a unit of reference for and when bioindicators are micro-­habitat dependent. To ecological studies. With the rise of molecular methods, © The Ecological Society of America

www.frontiersinecology.org

Freshwater biomonitoring in the Information Age

6

F Keck et al.

challenge is to develop new, high-­ quality indices based on DNA reads and environmental information. Three alternative but complemen­ tary approaches are described below and are represented in Figure 5. Developing MOTU-­based indices

Biomonitoring assumes that the presence or absence of particular taxa at a site of interest is indicative of distinct environmental condi­ tions at that site. Thus, in traditional biological assessments, an ecological profile associated with each taxon is required. Pawlowski et al. (2016) suggested calibrating MOTU-­based indices with traditional indices ­computed from simultaneously con­ ducted morphology-­based identifica­ tions. However, the traditional indices could be easily adapted to the new molecular approach by computing the indices directly from the reads clustered in MOTUs Figure 5. Flowchart introducing a new framework to process bioassessment (Steele et al. 2011). This approach metabarcoding data. Genetic reads can be interpreted without taxonomic affiliation using would require databases associating reads/MOTUs-­based indices, phylogenetic modeling, or machine learning. reads, MOTUs, and their responses to environmental stressors (Fig­ure 5). the DNA sequence has appeared as a promising alternative Thus, the MOTU-­ based indices approach is expected unit. Scientists have tried to integrate genetic sequences to be fully functional when ecological profiles for clus­ in the classical taxonomy, with varying degrees of success ters of reads are estimated directly from previous mo­ (Padial et al. 2010). However, in the context of biomon­ lecular inventories; this will require substantial work itoring, the question remains, whether the traditional in addition to data compilation and sharing. As a first Linnaean binomial species name affiliation still makes step, known ecological profiles for taxa can be trans­ sense within a full molecular approach. ferred to MOTUs. Typically, DNA reads provided by HTS are clustered into molecular operational taxonomic units (MOTUs), Using phylogeny to include rare species which are in turn converted to species units through the use of a bioinformatic workflow and a DNA reference DNA metabarcoding can reveal a wealth of diversity, database. The conversion from DNA reads to species but the lack of taxon–stressor response libraries is prob­ units is not without drawbacks: for instance, selected bar­ lematic. Given that ecological profiles are usually estimated codes may be associated with incorrect taxonomic affilia­ from in situ observations of general disturbances or from tions, genetic information may be lost (unaffiliated reads laboratory bioassays for specific substances, such libraries are discarded), and rare species are often insufficiently are restricted to common species and to a few types of studied. This approach is suitable if the reference data­ disturbances. Rare species are often ignored (Guénard base is sufficiently comprehensive, but this is rarely the et al. 2011), and the effects of specific compounds remain case because of the high species diversity and the time poorly understood (Schwarzenbach et al. 2006). One elegant way to solve these problems could involve and effort required to sequence organisms’ barcodes. Previously undescribed species are also frequently phylogenetic methods harnessing the principle that spe­ detected from genetic data, while formal taxonomic cies’ tolerances are the legacy of evolution (Keck et al. description can be a very long process (Goldstein and 2016). The increasing availability of DNA sequences and DeSalle 2011). Moving to full molecular biomonitoring computational power (Figure 3) should allow for the will allow for much more data to be used, beyond that establishment of large and robust phylogenies. Then, if limited strictly to taxonomic assignments. The greatest adequately long and informative (thereby excluding short www.frontiersinecology.org

© The Ecological Society of America

F Keck et al.

fragments and degraded DNA), reads can be inserted in the reference phylogeny using a posteriori replacement algorithms (Matsen et al. 2010; Berger et al. 2011). Finally, recent approaches to predict species’ tolerances based on information available from other species and their respective phylogenetic positions (Guénard et al. 2013) could be used to estimate an ecological profile for a given read (Figure 5). Routine inclusion of such phylogenetic-­ based methods in biomonitoring would help to account for the immense diversity uncovered by DNA barcoding and the thousands of toxicants in the environment. Machine learning techniques for ecological assessment

Analyzing and extracting valuable information from massive datasets can be extremely challenging. This has encouraged the development of machine learning meth­ ods, which use a set of statistical algorithms designed to recognize complex patterns in vast quantities of data. These methods include modern algorithms for classifi­ cation, such as random forest, gradient boosting, support vector machines, and neural networks (Hastie et al. 2009). Machine learning approaches are fully data-­driven and do not rely on any theoretical models (Breiman 2001). This system fits particularly well with the goals of biomonitoring, where the first aim is not necessarily to understand and explain the ecological processes leading to a given observation. In an applied context, correlation approaches are interesting because the final aim is to assess the state of the environment. This does not imply that machine learning should be used indiscriminately, but that these techniques are fully compatible with the ecological monitoring philosophy. Machine learning methods have a broad range of appli­ cations. In biomonitoring, they may be used with differ­ ent kinds of inputs for site classification, analyses of spa­ tial networks of sites, and time-­ series forecasting. However, the most anticipated application of machine learning for biomonitoring is the processing of genetic

Freshwater biomonitoring in the Information Age

data. The ultimate aim is for algorithms to classify a new site directly from the bulk of DNA reads just by identify­ ing genetic patterns learned from previous experience. The same data can be interpreted in various ways if analyzed by different algorithms programmed with differ­ ent training for different purposes (eg detection of eutrophication, effects of toxicants, or changes in flow regime). A set of sophisticated algorithms should enable scientists to monitor the effects of complex combinations of stressors on the environment. Such approaches are needed in view of multiple global threats (Vörösmarty et al. 2010). Furthermore, these methods should be ­implemented for massive datasets and communicate with holistic and integrative algorithms for automated and autonomous monitoring systems. In contrast to other more established fields in biology (Marx 2013), bioassess­ ment is just beginning to face the problems associated with massive datasets. Scientists will need to begin col­ laborating more closely with experts in computer science and applied mathematics to benefit from big data, and to develop new ways to communicate results to managers (Panel 2). JJ Conclusions

With the development of DNA metabarcoding, tradi­ tional environmental monitoring is experiencing a period of transformation, one outcome of which will be the need to deal with unprecedented amounts of data. Ascertaining the technical requirements to obtain and analyze data is just a part of the challenge. In contrast to scientists from other disciplines, ecologists have a relatively poor culture of data sharing, despite oppor­ tunities for making big data more accessible (Reichman et al. 2011; Hampton et al. 2013). However, there are signs that this is starting to change. Making biomon­ itoring big data freely available will potentially allow a range of new applications such as meta-­analyses and large-­scale analyses of biodiversity. Metabarcoding data are particularly relevant in this case because genetic data are highly comparable. Scientists and resource

Panel 2. Communication with managers Molecular methods constitute a new paradigm in freshwater ecosystem assessment. Environmental managers who are accustomed to traditional biological assessments and who are not familiar with genetics and molecular methods may be initially reluctant to adopt these approaches or may need training in order to do so. The widespread use of metabarcoding in biomonitoring depends on how these new tools will be implemented in future environmental assessment programs. Thus, new ways to communicate with resource managers must be developed. Communication should emphasize the benefits of metabarcoding, as well as explain the basics of genetics and the vocabulary of metabarcoding and HTS to managers in order to empower them to understand, interpret, communicate,

© The Ecological Society of America

and benefit from the results of metabarcoding. However, we must also acknowledge difficulties, such as the challenges associated with machine learning. Although it is important that biomonitoring tools are derived from sound theoretical concepts in ecology, because machine learning often operates as a black box (ie the user does not understand how the algorithm works), it might be hard to relate results to environmental health and key stressors. The implementation of such new environmental assessment frameworks will therefore take time and require a close collaboration between scientists and managers. Knowledge and experience gained over many years must not be lost and traditional approaches should continue to be used, at least for the purposes of comparison and discussion.

www.frontiersinecology.org

7

Freshwater biomonitoring in the Information Age

8

F Keck et al.

managers must work together to create effective networks and to develop dedicated sharing platforms. Indeed, the technical solutions discussed in this paper require sub­ stantial quantities of data and supporting infrastructures. Sharing platforms should be accessible to citizens and ecologists and would provide both raw and processed data as well as metadata. Raw data can be re-­ used with new bioinformatic workflows and statistical methods, while processed data are important for non-­ specialists and to help inform citizens (Soranno et al. 2015). If we can make public – and make sense of – the tera­ bytes of data that ecological assessments will produce in the foreseeable future, the entry of biomonitoring into the Information Age will be a genuine success.

Guénard G, Legendre P, and Peres-Neto P. 2013. Phylogenetic eigenvector maps: a framework to model and predict species traits. Methods Ecol Evol 4: 1120–31. Guénard G, von der Ohe PC, de Zwart D, et al. 2011. Using phy­ logenetic information to predict species tolerances to toxic chemicals. Ecol Appl 21: 3178–90. Hajibabaei M, Shokralla S, Zhou X, et al. 2011. Environmental barcoding: a next-­generation sequencing approach for bio­ monitoring applications using river benthos. PLoS ONE 6: e17497. Hajibabaei M, Singer GAC, Hebert PDN, and Hickey DA. 2007. DNA barcoding: how it complements taxonomy, molecular phylogenetics and population genetics. Trends Genet 23: 167–72. Hampton SE, Strasser CA, Tewksbury JJ, et al. 2013. Big data and the future of ecology. Front Ecol Environ 11: 156–162. Hastie T, Tibshirani R, and Friedman J. 2009. The elements of statistical learning: data mining, inference, and prediction. 2nd JJ Acknowledgements edn. New York, NY: Springer. Hebert PDN, Cywinska A, Ball SL, and deWaard JR. 2003. Biological identifications through DNA barcodes. P Roy Soc We thank A Franc for constructive comments and Lond B Bio 270: 313–21. I Domaizon for insightful discussion on metabarcoding Ibáñez C, Caiola N, Sharpe P, and Trobajo R. 2010. Ecological terminology. indicators to assess the health of river ecosystems. In: Jørgensen SE, Xu F-L, and Costanza R (Eds). Handbook of ecological JJ References indicators for assessment of ecosystem health. Boca Raton, FL: CRC Press. Baird DJ and Hajibabaei M. 2012. Biomonitoring 2.0: a new para­ Jørgensen SE, Xu F-L, Salas F, and Marques JC. 2010. Application digm in ecosystem assessment made possible by next-­generation of indicators for the assessment of ecosystem health. In: DNA sequencing. Mol Ecol 21: 2039–44. Jørgensen SE, Xu F-L, and Costanza R (Eds). Handbook of Benson DA, Karsch-Mizrachi I, Lipman DJ, et al. 2008. GenBank. ecological indicators for assessment of ecosystem health. Boca Nucleic Acids Res 36: D25–30. Raton, FL: CRC Press. Berger SA, Krompass D, and Stamatakis A. 2011. Performance, Keck F, Rimet F, Franc A, and Bouchez A. 2016. Phylogenetic accuracy, and web server for evolutionary placement of short signal in diatom ecology: perspectives for aquatic ecosystems sequence reads under maximum likelihood. Systems Biol 60: biomonitoring. Ecol Appl 26: 861–72. 291–302. Kermarrec L, Franc A, Rimet F, et al. 2014. A next-­generation Birk S, Bonne W, Borja A, et al. 2012. Three hundred ways to sequencing approach to river biomonitoring using benthic dia­ assess Europe’s surface waters: an almost complete overview of toms. Freshwater Sci 33: 349–63. biological methods to implement the Water Framework Kolkwitz R and Marsson M. 1908. Ökologie der pflanzlichen Directive. Ecol Indic 18: 31–41. Saprobien. Ber Deut Bot Ges 26: 505–19. Breiman L. 2001. Statistical modeling: the two cultures. Stat Sci 16: Kolkwitz R and Marsson M. 1909. Ökologie der tierischen 199–231. Saprobien. Beiträge zur Lehre von der biologischen Caron DA, Countway PD, Savai P, et al. 2009. Defining DNA-­ Gewässerbeurteilung. Int Rev Ges Hydrobiol Hydrogr 2: 126–52. based operational taxonomic units for microbial-­ eukaryote Lindenmayer DB and Likens GE. 2010. The science and applica­ ecology. Appl Environ Microb 75: 5797–808. tion of ecological monitoring. Biol Conserv 143: 1317–28. Chapman T. 2003. Lab automation and robotics: automation on Lovett GM, Burns DA, Driscoll CT, et al. 2007. Who needs envi­ the move. Nature 421: 661–66. ronmental monitoring? Front Ecol Environ 5: 253–60. Chariton AA, Stephenson S, Morgan MJ, et al. 2015. Mächler E, Deiner K, Steinmann P, and Altermatt F. 2014. Utility Metabarcoding of benthic eukaryote communities predicts the of environmental DNA for monitoring rare and indicator mac­ ecological condition of estuaries. Environ Pollut 203: 165–74. roinvertebrate species. Freshwater Sci 33: 1174–83. Coissac E, Riaz T, and Puillandre N. 2012. Bioinformatic chal­ Mandelik Y, Roll U, and Fleischer A. 2010. Cost-­efficiency of bio­ lenges for DNA metabarcoding of plants and animals. Mol Ecol diversity indicators for Mediterranean ecosystems and the 21: 1834–47. effects of socio-­economic factors. J Appl Ecol 47: 1179–88. Dafforn KA, Johnston EL, Ferguson A, et al. 2016. Big data oppor­ Marx V. 2013. Biology: the big challenges of big data. Nature 498: tunities and challenges for assessing multiple stressors across 255–60. scales in aquatic ecosystems. Mar Freshwater Res 67: 393–413. Matsen F, Kodner R, and Armbrust EV. 2010. pplacer: linear time De Queiroz K. 2007. Species concepts and species delimitation. maximum-­likelihood and Bayesian phylogenetic placement of Syst Biol 56: 879–86. sequences onto a fixed reference tree. BMC Bioinformatics 11: Fan W and Bifet A. 2013. Mining big data: current status, and 538. forecast to the future. SIGKDD Explorations 14: 1–5. Padial JM, Miralles A, la Riva ID, and Vences M. 2010. The inte­ Ficetola GF, Miaud C, Pompanon F, and Taberlet P. 2008. Species grative future of taxonomy. Front Zool 7: 1–14. detection using environmental DNA from water samples. Biol Passy SI, Bode RW, Carlson DM, and Novak MA. 2004. Lett 4: 423–25. Comparative environmental assessment in the studies of ben­ Gibson JF, Shokralla S, Curry C, et al. 2015. Large-­scale biomoni­ thic diatom, macroinvertebrate, and fish communities. Int Rev toring of remote and threatened ecosystems via high-­ Hydrobiol 89: 121–38. throughput sequencing. PLoS ONE 10: e0138432. Pawlowski J, Lejzerowicz F, Apotheloz-Perret-Gentil L, et al. 2016. Goldstein PZ and DeSalle R. 2011. Integrating DNA barcode data Protist metabarcoding and environmental biomonitoring: time and taxonomic practice: determination, discovery, and descrip­ for change. Eur J Protistol 55: 12–25. tion. Bioessays 33: 135–47. www.frontiersinecology.org

© The Ecological Society of America

F Keck et al. Ratnasingham S and Hebert PDN. 2007. BOLD: the Barcode of Life Data system (www.barcodinglife.org). Mol Ecol Notes 7: 355–64. Reichman OJ, Jones MB, and Schildhauer MP. 2011. Challenges and opportunities of open data in ecology. Science 331: 703–05. Resh VH. 2008. Which group is best? Attributes of different bio­ logical assemblages used in freshwater biomonitoring programs. Environ Monit Assess 138: 131–38. Schloss PD, Westcott SL, Ryabin T, et al. 2009. Introducing mothur: open-­ source, platform-­ independent, community-­ supported software for describing and comparing microbial communities. Appl Environ Microbiol 75: 7537–41. Schwarzenbach RP, Escher BI, Fenner K, et al. 2006. The chal­ lenge of micropollutants in aquatic systems. Science 313: 1072–77. Shokralla S, Spall JL, Gibson JF, and Hajibabaei M. 2012. Next-­ generation sequencing technologies for environmental DNA research. Mol Ecol 21: 1794–805. Steele JA, Countway PD, Xia L, et al. 2011. Marine bacterial, archaeal and protistan association networks reveal ecological linkages. ISME J 5: 1414–25.

© The Ecological Society of America

Freshwater biomonitoring in the Information Age Stein ED, White BP, Mazor RD, et al. 2014a. Does DNA barcoding improve performance of traditional stream bioassessment met­ rics? Freshwater Sci 33: 302–11. Stein ED, Martinez MC, Stiles S, et al. 2014b. Is DNA barcoding actually cheaper and faster than traditional morphological methods? Results from a survey of freshwater bioassessment efforts in the United States. PLoS ONE 9: e95525. Soranno PA, Cheruvelil KS, Elliott KC, and Montgomery GM. 2015. It’s good to share: why environmental scientists’ ethics are out of date. BioScience 65: 69–73. Taberlet P, Coissac E, Hajibabaei M, and Rieseberg LH. 2012. Environmental DNA. Mol Ecol 21: 1789–93. van Dijk EL, Auger H, Jaszczyszyn Y, and Thermes C. 2014. Ten years of next-­generation sequencing technology. Trends Genet 30: 418–26. Visco JA, Apothéloz-Perret-Gentil L, Cordonier A, et al. 2015. Environmental monitoring: inferring the diatom index from next-­ generation sequencing data. Environ Sci Technol 49: 7597–605. Vörösmarty CJ, McIntyre PB, Gessner MO, et al. 2010. Global threats to human water security and river biodiversity. Nature 467: 555–61.

www.frontiersinecology.org

9

Ecological Indicators 82 (2017) 1–12

Contents lists available at ScienceDirect

Ecological Indicators journal homepage: www.elsevier.com/locate/ecolind

Research paper

Assessing ecological status with diatoms DNA metabarcoding: Scaling-up on a WFD monitoring network (Mayotte island, France)

MARK

⁎

Valentin Vasselon , Frédéric Rimet, Kálmán Tapolczai, Agnès Bouchez CARRTEL, INRA, Université de Savoie Mont Blanc, 74200, Thonon-les-Bains, France

A R T I C L E I N F O

A B S T R A C T

Keywords: DNA metabarcoding Diatoms Biomonitoring Water framework directive Freshwater monitoring network

Diatoms are excellent ecological indicators of water quality because they are broadly distributed, they show high species diversity and they respond rapidly to human pressures. In Europe, the Water Framework Directive (WFD) gives the legal basis for the use of this indicator for water quality assessment and its management. Several quality indices, like the Specific Polluosensitivity Index (SPI), were developed to assess the ecological quality status of rivers based on diatom communities. It is based on morphological identifications and count of diatom species present in natural biofilms using a microscope. This methodology requires high taxonomic skills and several hours of analysis per sample as 400 individuals must be identified to species level. Since several years, a molecular approach based on DNA metabarcoding combined to High-Throughput Sequencing (HTS) is developed to characterize species assemblages in environmental samples which is potentially faster and cheaper. The ability of this approach to provide reliable diatom inventories has been demonstrated and its application to water quality assessment is currently being improved. Despite optimization of the DNA metabarcoding process with diatoms, few studies had yet extended it at the scale of a freshwater monitoring network and evaluated the reliability of its quality assessment compared to the classical morphological approach. In the present study we applied DNA metabarcoding to the river monitoring network of the tropical Island Mayotte. This island is a French département since 2011 and the WFD has to be applied. This offered the opportunity to scale up the comparison of molecular and morphological approaches and their ability to produce comparable community inventories and water quality assessments. Benthic diatoms were sampled following WFD standards in 45 river sites in 2014 and 2015 (80 samples). All samples were submitted in parallel to the molecular and the morphological approaches. DNA metabarcoding was carried out using Genelute DNA extraction method, rbcL DNA barcode and PGM sequencing, while microscopic counts were carried out for the classical methodology. Diatom community structures in terms of molecular (OTUs) and of morphological (species) were significantly correlated. However, only 13% of the species was shared by both approaches, with qualitative and quantitative variation due to i) the incompleteness of the reference library (82% of morphological species are not represented in the database), ii) limits in taxonomic knowledge and iii) biases in the estimation of relative abundances linked to diatom cell biovolume. However, ecological quality status assessed with the molecular and morphological SPI values were congruent, and little affected by sequencing depth. DNA metabarcoding of diatom communities allowed a reliable estimation of the quality status for most of the rivers at the scale of the full biomonitoring network of Mayotte Island.

1. Introduction Biological indicators are commonly used by environmental agencies for water quality assessment (Ibáñez et al., 2010; Birk et al., 2012). Diatoms, a group of microalgae, are known to be efficient indicators of river ecological quality and are required to be monitored in rivers by transnational directives as the Water Framework Directive in Europe (Directive 2000/60/EC, 2000). Their indicator efficiency relies on their ⁎

huge taxonomical diversity and their species ecological preferences to particular pollution levels (Pandey et al., 2017). After collecting natural diatom communities from benthic biofilms, the relative frequencies of diatom taxa are used together with their ecological optimum and tolerance values to compute biotic indices often derived from the Zelinka and Marvan formula (Zelinka and Marvan, 1961). Some indices are based on a restricted list of taxa, adapted to local diatom biodiversity and to type of pressures. For example, the French

Corresponding author at: Valentin Vasselon, 75 bis Avenue de Corzent, 74200, Thonon-les-Bains, France. E-mail addresses: [email protected], [email protected] (V. Vasselon), [email protected] (F. Rimet), [email protected] (K. Tapolczai), [email protected] (A. Bouchez). http://dx.doi.org/10.1016/j.ecolind.2017.06.024 Received 23 March 2017; Received in revised form 2 June 2017; Accepted 4 June 2017 1470-160X/ © 2017 Elsevier Ltd. All rights reserved.

Ecological Indicators 82 (2017) 1–12

V. Vasselon et al.

Fig. 1. Location of Mayotte island (France) and the 45 river sites of the three monitoring networks: Reference sites network (REF – white), Regular WFD monitoring network (RCS – grey) and Polluted sites network (POLL – black).

much larger set of diatom species (over 2000). Due to its large taxonomical and ecological base, its efficiency to assess ecological quality in a large range of rivers in Europe has been demonstrated (e.g. Kelly, 2013). Though, SPI is the index currently used to apply WFD to rivers in several European countries like in Portugal, Belgium, Bulgaria, Netherlands, Sweden, Luxembourg or Spain (Kelly, 2013). Even in regions

WFD index, BDI (Lenoir and Coste 1996), is based on a list of 800 taxa with their associated autecology. The Swiss DI-CH (Hürlimann and Niederhauser, 2007) is based on a restricted list of only 188 taxa. At the opposite, the SPI (Specific Polluosensitivity Index, Cemagref, 1982; Coste, 1986) was developed to evaluate overall water quality in terms of organic pollution and nutrient levels and is though encountering a

2

Ecological Indicators 82 (2017) 1–12

V. Vasselon et al.

archipelago located in the Indian Ocean, in the north-west of Madagascar, in the Mozambique Channel (12°50′35″S 45°08′18″E, Fig. 1). Large ecological gradients are encountered in Mayotte rivers where pollution originated from two main anthropogenic pressures. First, a major part of households is not connected to sewage, thus their wastewater is often released directly in the rivers, contributing to an elevated organic matter concentration. Second, clothing is usually washed directly in rivers resulting in high turbidity and suspended solids values due to washing powders. To that aim, a river monitoring network has been set up recently. In Mayotte, a regular WFD monitoring network (Réseau de Contrôle de Surveillance – RCS) was monitored since 2008. This network groups 14 sites under intermediate conditions, generally just upstream agglomerations. Complementary sampling sites were added in 2014 in order to represent better the different environmental conditions in the rivers of Mayotte, based on their general conditions. Upstream sites (17 sites) under good conditions were grouped in the Reference sites network (REF). Downstream sites (14 sites) under degraded conditions and outside the zone of influence of tides were grouped in the Polluted sites network (POLL). Samples were taken from 45 sampling sites of 33 different rivers of the main island “Grande Terre” (363 km2) in 2014 and 2015 (Fig. 1). The relevance of this a priori classification was confirmed by Tapolczai et al., (2017) which showed a strong gradient of organic pollution, turbidity and suspended solids, increasing from REF to RCS and then to POLL networks (Fig. 2).

with a weaker taxonomical and ecological knowledge about diatom species, SPI is used as a reference index to reveal quality gradients in urban rivers of boreal regions (e.g. Teittinen et al., 2015) or in Chinese rivers (e.g. Yang et al., 2015). However, applying such indices requires time and a high level of taxonomical expertise. Current standardized methods (e.g. CEN) call for the determination of diatoms until a minimum 400 individuals (diatom valves) at species or sub-species level using light microscopy and several tens of iconographical books. The development of DNA metabarcoding and High-Throughput Sequencing offered a solution, allowing to investigate prokaryote and eukaryote biodiversity present in environmental samples (Creer 2010). Assessing taxonomic inventories of environmental communities of macroinvertebrates based on DNA metabarcoding has been shown by Hajibabaei et al. (2011) as a promising alternative to morphological methodologies for biomonitoring (Keck et al., 2017). Same hopes were raised for diatoms testing DNA metabarcoding on mock communities (Kermarrec et al., 2013b) and environmental communities (Kermarrec et al., 2014). The SPI index values calculated for each diatom taxonomic inventory based on metabarcoding data enabled authors to assign quality classes to the environmental samples with the same ranking than using morphological methodologies. Later studies (Zimmermann et al., 2015; Visco et al., 2015; Vasselon et al., 2017) confirmed the possibility of using DNA metabarcoding of diatom communities for environmental studies and biomonitoring. Current genes used for diatoms barcoding are the V4 region of the genomic gene 18S (Zimmermann et al., 2015; Visco et al., 2015) and the plastid gene rbcL (Kermarrec et al., 2014; Vasselon et al., 2017). According to Kermarrec et al. (2013b, 2014) who compared both barcodes, rbcL polymorphism proved to be compatible with a detection at species level, while 18S was more efficient at genus level. The recent release of the R-syst::diatom barcoding library (Rimet et al., 2016) in open-access (http://www.rsyst.inra.fr/) offers an expert and curated data on rbcL with more than 2500 rbcL sequences related to their taxonomic identity, which represents more than 900 species and 200 genera (20-03-2017: R-syst::diatom v6). However, to date studies confirming the validation of the DNA metabarcoding approach for water quality assessment were done only at small scales (Kermarrec et al., 2014: 4 samples, Zimmermann et al., 2015: 7 samples, Visco et al., 2015: 27 samples, Vasselon et al., 2017: 8 samples) and at a regional scale (Apothéloz-Perret-Gentil et al., 2017: 2 Switzerland cantons with 87 samples). We propose here to scale up the test to a large monitoring network, including larger gradients of ecological status. Mayotte Island, a tropical island part of Comoros archipelago in Mozambique Channel became a French département in 2011, therefore it is now subject to the European regulations, despite its distance to mainland Europe. We took advantage of the river monitoring network set up to develop WFD indices for Mayotte in order to scale up the DNA metabarcoding approach for diatoms. A total of 80 samples collected at 45 sites were used to compare the molecular and the classical morphological approaches. First, we compared both approaches through (i) taxonomic composition (using OTUs, species, genus and family levels), (ii) diversity (Shannon index), (iii) richness (Chao richness estimator), (iv) structure of diatom community (Bray-Curtis dissimilarity index). Second, biases that may affect the molecular SPI values were investigated, including i) the incompleteness of the reference library, ii) the importance of the taxonomic knowledge, iii) the diatom cell biovolume, and iv) the sequencing depth. Finally, the ability of DNA metabarcoding to produce congruent ecological quality status at the scale of a monitoring network was evaluated.

2.2. Diatoms sampling Diatoms were sampled following the French (Afnor, 2016) and European (Afnor, 2014a) standards and were carried out once a year during the dry season (July-August). Briefly, benthic diatoms were collected from at least 5 stones from the lotic parts of the sampling sites in order to limit local effect on diatom community (e.g. flow velocity, water depth) and mix into a unique vial. The upper surface of the stones was scrubbed with a clean toothbrush. The samples were preserved by adding 99% ethanol for a final ethanol concentration > 70%, in order to preserve DNA. For each site, 2 subsamples were taken from the vial with the pooled biofilm sample for the molecular and morphological approaches. 2.3. DNA metabarcoding 2.3.1. DNA extraction DNA extraction was performed using 2 mL of the preserved sample. After centrifugation at 13,000 rpm during 30 min, supernatant containing ethanol was removed and the pellet used as starter for DNA extraction. Total genomic DNA was isolated using a non-commercial method based on Sigma-Aldrich GenElute™-LPA DNA precipitation, as described in previous studies (Kermarrec et al., 2013b; Chonova et al., 2016). This method combined various lysis mechanisms in order to disrupt diatom cell (mechanical, enzymatic, heat) and was recommended for diatom metabarcoding (Vasselon et al., 2017). 2.3.2. PCR amplification PCR amplification was performed on rbcL plastid gene targeting a 312 bp barcode. For amplifying this region, the primer pair Diat_rbcL_708F (Stoof-leichsenring et al., 2012) and R3 (Bruder and Medlin 2007) was slightly modified. Using an alignment of 1602 rbcL reference sequences from 638 diatom species, the degeneracy of the primers was increased in order to amplify a broader diversity of diatoms as follow: the forward primer combined an equimolar mix of Diat_rbcL_708F_1 (AGGTGAAGTAAAAGGTTCWTACTTAAA), Diat_rbcL_708F_2 (AGGTGAAGTTAAAGGTTCWTAYTTAAA) and Diat_rbcL_708F_3 (AGGTGAAACTAAAGGTTCWTACTTAAA); the reverse primer combined an equimolar mix of R3_1

2. Material and methods 2.1. Mayotte island monitoring network Mayotte is a French tropical island (374 km2), part of the Comoros 3

Ecological Indicators 82 (2017) 1–12

V. Vasselon et al.

Fig. 2. Main environmental pressure gradients in rivers of Mayotte: dissolved organic carbon (DOC), total organic carbon (TOC), turbidity and suspended solids (SS). Boxplots present the log transformed values for each parameter at each of the three monitoring networks (REF, RCS, POLL).

Génome Transcriptome” (PGTB, Bordeaux, France).

(CCTTCTAATTTACCWACWACTG) and R3_2 (CCTTCTAATTTACCWACAACAG). For each DNA sample, PCR amplification was performed in triplicate in a final volume of 25 μL. Each PCR mix was composed by 1 μL of extracted DNA, 0.75 U of Takara LA Taq® polymerase, 2.5 μL of 10X Buffer, 1.25 μL of 10 μM of primers Diat_rbcL_708F_1_2_3 and R3_1_2, 1.25 μL of 10 g/L BSA, 2 μL of 2.5 mM dNTP, and completed with molecular biology grade water. The PCR reaction conditions were initiated by a denaturation step at 95 °C for 15 min followed by a total of 30 cycles of 95 °c for 45 s (denaturation), 55 °C for 45 s (annealing), and 72 °C for 45 s (final extension).

2.3.4. Bioinformatic processing The sequencing platform performed demultiplexing and provided a fastq file for each of the 80 libraries. A first quality filtering step excluded DNA reads below 250 bp read length, with a Phred quality score below 23 over a moving window of 25 bp, with more than one mismatch in the primer sequence and homopolymer over 8 bp, or with ambiguous base. All the fastq files were then treated together following the bioinformatics process described in Vasselon et al. (2017) using the Mothur software (Schloss et al., 2009). DNA reads were clustered in OTUs using a distance similarity threshold of 95% (Mangot et al., 2013), and singletons were then removed. All samples were normalized to the same read number (using the smallest read abundance obtained for 1 sample) in order to allow inter-sample comparison. Diatom molecular inventories were obtained using the R-syst::diatom library (Rimet et al., 2016, 13-02-2015: R-syst::diatom v3, http://www.rsyst. inra.fr/en) for taxonomic assignment of OTUs, and the consensus taxonomy of DNA reads with a consensus confidence threshold over 80%. Fastq files with demultiplexed DNA reads, the final OTU (95%) list (including DNA reads proportion, DNA representative sequence of each OTU and the OTU taxonomic assignment) as well as the sampling site description are available for all the samples on the Zenodo repository website (http://doi.org/10.5281/zenodo.400160).

2.3.3. Sample libraries and HTS PCR products of the 3 PCR replicates prepared for each DNA sample were pooled and cleaned with Agencourt AMPure beads (Beckman Coulter, Brea, USA). Quality and quantity of purified amplicon were checked using the 2200 TapeStation (Agilent technologies, Santa Clara, USA). Ligation of tags to amplicons and library preparation were performed as described in Vasselon et al. (2017) using the NEBNext® Fast DNA Library Prep set for Ion Torrent™ (BioLabs, Ipswich, USA) and A-X tag adapter provided in Ion Express™ Barcode adapters (Life Technologies, Carlsbad, USA). Finally, 42 samples libraries (2014 campaign) and 38 samples libraries (2015 campaign) were pooled in 2 mix at a final concentration of 100pm per mix and sequenced independently. Each mix was sequenced on an Ion 318™Chip Kit V2 (Life Technologies, Carlsbad, USA) on a PGM Ion Torrent machine by the “Plateforme 4

Ecological Indicators 82 (2017) 1–12

V. Vasselon et al.

2.4. Morphological analysis

Amphora pediculus and the Mayotte endemic Nitzschia sp.1 (Table A.1).

Parallel to their molecular analysis, the diatom benthic samples were treated for microscopy analysis, according to the European standard (Afnor, 2014a), using hot H2O2 and Naphrax to mount permanent slides. A minimum of 400 valves were counted and determined to species level (or genus level when not possible) according to Afnor (2014b) using classical European floras (e.g., Krammer and LangeBertalot 1986, 1988, 1991a, 1991b; Krammer, 2000, 2001, 2002, 2003) and literature dedicated to tropical areas (e.g. Bourrelly and Manguin, 1952; Metezeltin and Lange-Bertalot, 1998, 2007; Tudesque et al., 2008).

3.2. HTS analysis The PGM sequencing produced a total of 6,076,529 of DNA reads for the 80 libraries that were sequenced. Based on similar quality levels in the 2 sequencing runs, all sequence data could be analyzed together. After the first bioinformatics step of quality filtering, 1,562,321 reads were retained and clustered into 3381 OTUs (95% similarity threshold) with a mean of 354 OTUs per sample. To allow inter-sample comparison, all samples were rarefied to 5710 reads (lowest read abundance obtained for one sample) for a total of 456,800 reads corresponding to 2754 OTUs with a mean of 233 OTUs per sample (min = 123, max = 432). After taxonomic assignment of OTUs using R-syst::diatom, 69.2% of OTUs at Family level (75.1% of total reads), 62.2% at Genus level (72% of total reads), and 35.7% at Species level (40.7% of total reads) were successfully assigned. Unclassified proportion of reads per sample at species level varied from 1.3% to 97%. Successful taxonomic assignment of OTUs resulted in a diatom taxonomic list of 23 families, 39 genera and 66 diatom species (Table A.2) with a mean of 16 species per sample (min = 6, max = 41). The most abundant species detected by HTS among all samples were Ulnaria ulna, Amphora pediculus, Gomphonema parvulum and G. bourbonense.

2.5. Morphological and molecular SPI The ecological quality status of the different river sites was assessed based on the diatoms biological quality element, using the SPI (Cemagref, 1982). Morphological and molecular SPI were determined using species taxonomic lists (or genus level if the species level was not reached) obtained by microscopy or HTS (relative abundance of DNA reads), respectively. The OMNIDIA 5 software (Lecointe et al., 1993, library 5.3 2015) was used for SPI calculation. As no water quality classes have not yet been defined for Mayotte rivers following WFD recommendation (Tapolczai et al., 2017), the general pollution gradient of freshwater rivers was divided into 5 ecological status corresponding to different water quality classes used by the French standard (Afnor 2007): high (SPI: 17–20), good (SPI: 13–17), moderate (SPI: 9–13), poor (SPI: 5–9), bad (SPI: 1–5). Effect of sequencing depth on the molecular SPI values was checked on all the samples by reducing in silico the number of reads used to obtain diatom taxonomic list, to a minimum of 50 reads per sample, using random subsampling. SPI values calculated with subsampling data were correlated to optimal SPI values obtained using all available DNA reads.

3.3. Comparison of the diatom communities obtained by molecular or morphological assignment The morphological and the molecular taxonomic compositions of diatom communities were compared using Venn diagrams (Fig. 3). The taxonomic composition between both inventories was similar at 80.8% for family level, at 59% for genus level, and at 13% for species level. 82% of the diatom species detected only by microscopy were absent from the molecular reference library R-syst::diatom. The correlation between Shannon indices calculated with the 2 approaches for all samples were highly significant and stable when comparing the molecular to morphological indices for families (r = 0.37, p < 0.001), for genera (r = 0.33, p < 0.01), and between molecular OTUs and morphological species (r = 0.34; p < 0.01). Only the molecular and morphological Shannon index based on species was not significantly correlated (p = 0.16). Regarding the Chao index, all the correlation factors were significant when comparing the family (r = 0.40, p < 0.001), genus (r = 0.41, p < 0.001), species (r = 0.25, p = 0.02) and OTU/species (r = 0.26, p = 0.02) levels. Bray-Curtis dissimilarity indices were calculated based on OTU lists for the molecular approach and on species lists for the morphological one. Mantel’s test revealed a significant correlation between the dissimilarity matrices obtained from HTS and morphological inventories (r = 0.43, p = 0.01).

2.6. Statistical analysis Relative abundance of each diatom species within each site was determined for molecular and morphological inventories. The correlation coefficients and their associated p-value for the richness (Chao estimator), diversity (Shannon) and SPI indices comparisons were determined with the “Pearson correlation” available on R (R Development core team, 2013). Correlations were visualized using linear regression. For each index (SPI, Chao, Shannon), the effect of DNA reads subsampling on index values calculation was checked by comparing together the index values obtained for all samples and all subsampling using one-way Anova analysis. The molecular OTUs and morphological species lists were used to compute 2 separate Bray-Curtis distance matrices, which were compared together using Mantel test. Non-metric multidimensional scaling (NMDS) was used to visualize the Bray-Curtis matrices based on molecular OTUs data and morphological data. An Anosim analyses was used to compare the similarity between the water quality status assessed by morphological and molecular SPI.

3.4. Morphological and molecular SPI calculation SPI index values calculated based on the morphological (205 taxa) and the molecular diatom (84 taxa) inventories ranged from 5.1 to 19.5 and from 4.4 to 15.8 for the molecular and the morphological approaches respectively. The mean SPI values obtained for the 3 monitoring networks were congruent with their expected water quality status with both the molecular (POLL = 11.7, RCS = 14.9, REF = 15.9) and the morphological (POLL = 8.1, RCS = 11.6, REF = 12.4) approaches (Fig. 4). The POLL network was significantly different than the RCS and REF ones with both approaches (p < 0.001 in both cases). A significant correlation (r = 0.72) was observed between the SPI values for all samples obtained by both approaches (Fig. 5). When comparing the water quality classes deduced from the SPI values of both approaches, 27.5% of all the samples shared an identical

3. Results 3.1. Morphological analysis A total of 24 families, 58 genera and 204 species of diatoms were identified among all the samples including 16.2% of tropical species with 5.4% of species endemic to Mayotte island. The number of diatoms species identified per sample varies from 6 to a maximum of 54 species (mean = 25 species per sample). The most abundant species (> 5%) identified among all samples were Cocconeis placentula var. euglypta, Gomphonema bourbonense, Gomphonema parvulum, Nitzschia inconspicua, 5

Ecological Indicators 82 (2017) 1–12

V. Vasselon et al.

Fig. 5. Correlation between the diatom Specific Pollution Index (SPI) based on molecular (y axis) and morphological (x axis) inventories for all 80 samples. The linear regression model is represented by the dotted line, r and p-value are indicated. SPI values are in the range from 1 (bad quality status) to 20 (high quality status).

quality status, 60% had 1 class of difference, 10% had 2 classes of difference, and 2.5% had 3 classes of difference. The NMDS plots presented in Fig. 6 (based on OTUs similarity) showed that the samples distribution is driven by their respective water quality level both for molecular and morphological approaches. Anosim analysis indicated that the water quality classes explained 22.7% (p = 0.001) and 29.0% (p = 0.001) of the total variance for the morphological and the molecular approaches respectively. On average, the molecular SPI values were 3.6 points higher than the morphological SPI values (min difference = 0.1, max difference = 9.8) with a mean of 3.7 (sd = 2.1), 3.7 (sd = 2.4) and 3.4 (sd = 2.1) for the sites belonging to REF, POLL and RCS networks respectively. We observed positive correlations between the difference of SPI obtained by both approaches with the proportion of taxa from Eunotia genus and with the proportion of DNA reads that could not be classified at genus level in the molecular inventories (Fig. 7). Fig. 3. Venn diagrams comparing the diatom inventories assigned at family, genus and species levels either by the molecular (right circles) or by the morphological (left circles) approach (80 river samples). Taxa assigned by both approaches are represented by the overlapping region in the middle (black). For taxa detected only by one of the two methods, pie charts indicate the number of taxa with relative abundance < 0.5% and > 0.5%. For taxa only detected in microscopy, the number of taxa absent from the reference database is also indicated.

3.5. Impact of sequencing depth on richness, diversity and the water quality index Regarding the richness index (Chao), values were affected by the different subsampling with a drop of the correlation from 0.787 to 0.322 (“5710 reads vs all” and “50 reads vs all” respectively) (Fig. 8). Fig. 4. Distribution of the values of the diatom Specific Pollution Index (SPI) based on molecular (left) and morphological (right) inventories for all 80 samples within the 3 monitoring networks (POLL, RCS, REF). Different letters indicate significant difference between SPI means (T-test, p < 0.05).

6

Ecological Indicators 82 (2017) 1–12

V. Vasselon et al.

Fig. 6. Three-dimensional NMDS plots of Bray-Curtis dissimilarity based on OTU composition of all the 80 samples. Colours correspond to the quality class deduced from either molecular (left) or morphological (right) SPI values assessed to each sample. Quality classes: high (SPI: 17–20), good (SPI: 13–17), moderate (SPI: 9–13), poor (SPI: 5–9), bad (SPI: 1–5).

SPI values was detected (p = 0.99), whatever the size of the tested subsamples. The average of the difference between the SPI values calculated with 50 reads per sample and those calculated with all reads was 0.66. 4. Discussion 4.1. Community structure and ecological quality status inferred by molecular and by morphological approaches are congruent Mayotte is a small Island where rivers have the same typology due to their common geological substratum and their short length from the source to the sea pool and (≈10 km). Thus, diatom species diversity observed in morphological inventories (204 species for 58 genera) is comparable to other small tropical Islands as determined by Gassioles (doctoral dissertation, 2014) for Reunion Island (343 species for 61 genera) and by Gueguen et al. (2015) for Martinique (324 species for 59 genera) and Guadeloupe (352 species for 57 genera) islands. However, diatom community structures obtained with the DNA metabarcoding approach (based on OTU) and with the morphological approach (based on species) are correlated. This confirms previous observations that DNA metabarcoding combined to HTS is a good approach to evaluate diatom diversity in freshwater ecosystems (Zimmermann et al., 2015). Mayotte is a relatively complex field study for freshwater quality assessment, mainly due to the presence of endemic and tropical diatom species. Indeed, ecological preferences of these species have not been defined yet, thus making the applicability of many diatom indices (e.g. European diatom indices) uncertain in this part of the world. Previous studies showed that diatom indices developed in a particular geographical area are less effective when applied elsewhere (Rott et al., 2003; Potapova and Charles 2007). This is why a diatom index dedicated to Mayotte rivers quality assessment is currently developed in the framework of the WFD (Tapolczai et al., 2017). Despite that, Bellinger et al. (2006) and Bere (2016) showed that European indices can be applied to remote countries as an initial approach when undescribed diatom taxa are not the most abundant taxa. In our study, 50% of tropical and endemic species identified have unknown ecological preferences but correspond to low abundant taxa (< 1%) and therefore have a low impact on SPI index calculation. Furthermore, the dominant taxa were cosmopolitan diatoms (e.g. Cocconeis placentula, Gomphonema bourbonense, G. parvulum, Nitzschia inconspicua, Amphora pediculus) for which ecological preferences are well described in the literature and included in the SPI index calculation. Moreover, the SPI includes more than 2000 species, some being tropical. The results of the water quality

Fig. 7. Correlation between the ΔSPI (difference between the molecular SPI and the morphological SPI values) and the proportion in molecular inventories of (A) Eunotia taxa and (B) unclassified reads at Genus level, for all samples. The linear regression model is represented by the dotted line, r and p-value are indicated.

When all reads were used, an average of 555 OTUs per sample was estimated by the Chao index (min = 186; max = 937) while for 50 reads, the average of Chao index was only 45 OTUs per sample (min = 10; max = 111). The Anova analysis showed a significant effect of the subsampling on Chao index (p < 0.001) and this directly at the first subsampling of 5710 reads per sample (p < 0.001). The richness diversity index (Shannon) was affected in a similar manner by the different subsampling with a drop of the correlation from 0.99 to 0.89 (Fig. 8). The Anova analysis did reveal a significant effect of the subsampling on the Shannon index (p < 0.001) when the subsampling was done with 100 reads and 50 reads. The SPI values calculated with the molecular inventories based on all the DNA reads available per sample and with the different subsamples of reads per sample were all significantly correlated (Fig. 8). An Anova analysis indicated that no significant effect of subsampling on 7

Ecological Indicators 82 (2017) 1–12

V. Vasselon et al.

Fig. 8. Impact of the sequencing depth on molecular SPI values (left), OTUs richness (Chao estimator, middle), OTUs diversity (Shannon index, right) evaluated on all samples by performing random subsampling of DNA reads (subsampling decreasing from 5710 to 50 reads per sample). Correlations are presented between values obtained with subsamples of DNA reads (vertical axis) and values obtained using all available DNA reads for each sample (horizontal axis). All correlation values are significant.

these two approaches. Our results showed a strong correlation between the molecular and the morphological based SPI indices, supporting previous smaller scales observations made by Kermarrec et al. (2014, 4 sites) in France and Visco et al. (2015, 27 sites) in a regions of Switzerland. However, when comparing the taxonomic inventories at the species level between both approaches, important discrepancies appeared. This low correspondence was shown to be mainly due to the incompleteness of the DNA-barcode reference library. Despite those shortcomings, most of the abundant taxa were identified by both approaches at species level or at least at genus level. The SPI calculation is based on the Zelinka and Marvan (1961) formula where rare species have low impact on the final value, while abundant species drive it. This explains the good correlation between molecular and morphological based SPI. Indeed, Bigler

status inferred by the SPI based on both morphological and molecular taxonomic inventories were congruent with the expected ecological status based on observed pressures (physical-chemical parameters) at the monitored sites. The lowest SPI values were obtained for river sites belonging to the polluted network (POLL) and the highest for river sites belonging to the reference network (REF). Even if the morphological approach allow a better discrimination between polluted and reference sites than the molecular one, the latter remains highly efficient to discriminate the 2 networks. Therefore, even if optimizations are required for an accurate and well-adapted water quality assessment, the SPI index is a good basis to compare the molecular (DNA metabarcoding) and the classical (morphological) approaches as monitoring tools for diatoms. Moreover, the river monitoring network developed in Mayotte offered the opportunity of an unprecedented large scale comparison of 8

Ecological Indicators 82 (2017) 1–12

V. Vasselon et al.

taxonomical assignation we used R-syst::diatom, an expert library dedicated to diatoms (Rimet et al., 2016) which is up-dated with all available sequences allowing a reliable assignment of DNA reads at the family and genus levels. Hence, diversity indices obtained for these 2 taxonomic levels by both approaches were significantly correlated. However a major part of the sequences still remained unclassified at species level and consequently could not be included in the molecular SPI calculation, which leads to an absence of correlation for diversity indices at the species level. Among missing taxa are those often abundant which are present in most of the samples like Navicula quasidisjuncta, Achnanthidium subhudsonis, Amphora copulata, Planothidium rostratum, Navicula escambia or Gomphonema designatum (17.7% of total microscopic counts). Such lacks in the reference library contribute to differences between molecular and morphological SPI values, despite their good correlation. The incompleteness of DNA barcode reference libraries is a recurrent bias for European diatom flora already discussed in the literature (; Visco et al., 2015, Vasselon et al., 2017Zimmermann et al., 2014; Visco et al., 2015, Vasselon et al., 2017). This bias is getting even more acute for remote places like Mayotte island with tropical and endemic flora. Indeed, for diatoms, isolating living cells from fresh field samples, cultivating monoclonal strains and consequently identifying and sequencing them is the best way to add reliable new DNA barcode references in libraries. However, this time consuming approach suffers from a low success rate because it requires the isolation of living cells able to survive in culture, which depends on culture conditions (e.g. growth media composition, temperature, light). Although 24 strains from Mayotte samples had been isolated and sequenced in a previous study (Kermarrec et al., 2013a) and added to R-syst::diatom library, many endemic and tropical taxa could not be cultivated and remained absent from the library. The single-cell PCR method has been proposed to obtain sequences from uncultured diatoms (Hamilton et al., 2015; Khan-Bureau et al., 2016), however, taxonomic identification performed on living cells can lead to incomplete or inaccurate taxonomic identification. Alternative approaches based on OTU co-abundance networks (Irannia and Chen 2016) were proposed to predict the taxonomy of unknown OTU but predictions were limited to the phylum or class taxonomic levels. Finally, Rimet et al. (2017) propose recently to use environmental sequences from HTS runs, to relate them with morphological observations and to integrate them into libraries after setting several quality criteria. We also observed that increasing proportion of unclassified DNAreads increases the difference between SPI values obtained from molecular and morphological approaches. Up to now, tests of water quality assessment based on diatom DNA metabarcoding has always mimicked classical morphological approach by i) using biotic indices initially developed for morphological inventories (Kermarrec et al., 2013b, 2014; Lejzerowicz et al., 2015; Visco et al., 2015; Zimmermann et al., 2015; Vasselon et al., 2017) and ii) addressing ecological values of morphological species to OTUs through taxonomic assignment. A way to overcome this limit linked to the taxonomical assignment, as suggested by Pawlowski et al. (2016), is to connect directly OTUs with environmental data in order to determine their ecological preferences, thus morphological species in conventional indices could be replaced by OTUs. By this way, molecular indices will take into account all the molecular data, be it assigned or not, be it rare or not. This includes hidden diversity like cryptic diversity or unknown diatom taxa which have hardly been taken into account up to now. The development of such molecular indices can be a good option Apothéloz-Perret-Gentil et al. (2017), especially for remote regions with recent biomonitoring initiatives, like Mayotte, suffering from a lack of taxonomic and environmental knowledge.

et al. (2009) showed that removing taxa with relative abundance below 5% have low impact on diatom indices value, as well as Lavoie et al. (2009) who showed that rare diatom taxa have little interest for ecological assessment. The presence of rare taxa can be considered as poorly informative for water quality assessment and for that reason they are often excluded from some indices calculation (e.g. Biological Diatom Index − Prygiel et al., 2002). Though, these results show that diatom DNA metabarcoding could be a reliable tool to derive accurate diatom quality indices for regulatory use at the scale of a biomonitoring network. Another solution is to use a taxonomy-free approach to calculate molecular water quality index directly from metabarcoding data without any taxonomic assignment (Apothéloz-Perret-Gentil et al., 2017). Even if such approach prevent doing the link with ecological and historical knowledge based on morphological data, it would be suitable for Mayotte biomonitoring network where such knowledge is limited. 4.2. Biases explaining differences between molecular and morphological based diatom indices DNA-based indices are currently mimicking conventional indices using two kinds of information about diatom communities: i) a qualitative information which is a list of species usually based on taxonomic assignment of OTUs, and ii) a quantitative information based on the proportion of DNA reads per taxa. Both kinds of information can be affected during the DNA metabarcoding workflow by biases linked to different steps: the DNA extraction method (Deiner et al., 2015; Vasselon et al., 2017), the targeted gene for DNA-barcode (Kermarrec et al., 2013b; Valentini et al., 2016), the set of primers for DNA metabarcoding (Elbrecht and Leese 2015), the PCR amplification protocol (Kebschull and Zador, 2015), the sequencing technology (Quail et al., 2012), the bioinformatics data processing (Schmidt et al., 2015), and the variation of cell biomass among taxa (Thomas et al., 2016). For diatoms DNA metabarcoding, previous studies started to identify the importance of those biases in order to optimize the choice of the DNA barcode (Kermarrec et al., 2013b) or the DNA extraction (Vasselon et al., 2017). Recommendations of these 2 studies were taken into account to set up the experimental design of the present study: use of a 312 bp rbcL DNA-barcode and use of Genelute method for DNA extraction. Elbrecht and Leese (2017) shown that the use of well-developed primers, using specialized tool like PrimerMiner (Elbrecht and Leese 2016), reduces bias in macroinvertebrates metabarcoding inventories. The increase of degenerated bases in our rbcL primers allows to reduce primer bias, as discussed by Elbrecht and Leese (2017), however further investigation will have to be done to validate them for water quality assessment. In this study, the comparison between the morphological and molecular approaches was performed using 1 subsample of each sample per approach, which can be a source of variability. However, in the Vasselon et al. (2017) study, 2 subsamples per environmental sample were sequenced and detection of abundant species and SPI calculation were not affected. Similar results were shown for the morphological approach (Lavoie et al., 2005). Thus, we consider that variability linked to subsampling will have limited impact on quality index calculation and method comparison compare to other biases. We will thereafter focus on four other biases that could specifically affect qualitative and quantitative information obtained from DNA metabarcoding of diatom communities in Mayotte: the incompleteness of the reference library, the limits of current taxonomic knowledge, the variation of cell biovolume among diatom taxa and the sequencing depth. 4.2.1. Bias related to reference library incompleteness Despite the good correlation between the SPI values obtained by both approaches, molecular based SPI values differed from morphological ones for all samples. The molecular diatom inventories used for quality index calculation were only based on the part of the molecular data to which a taxonomical identity could be assigned. For

4.2.2. Bias related to limits in taxonomic knowledge (example of para/ polyphyletic taxa) The species Nitzschia inconspicua, indicator of poor quality rivers, was observed in morphological inventories in most of samples but was 9

Ecological Indicators 82 (2017) 1–12

V. Vasselon et al.

sequencing hundreds of environmental samples at once, given sufficient unique tags are available for sample multiplexing. Increasing the number of samples in one sequencing run decreases the number of available DNA reads per sample. One important question for further accurate application of this approach to field biomonitoring is to know how many DNA reads per sample are required to have a good description of the community diversity and a good evaluation of the ecological status. Previous study showed that increasing the sequencing depth can improve the ecological inference from HTS data more than increasing the number of PCR replicates (Smith and Peay, 2014). Of course, depending on the local community and its diversity, the required minimum number of DNA reads will vary from one sample to another. Lundin et al. (2012) shown that a minimum of 1000 and 5000 denoised DNA reads were sufficient to respectively describe trends in β and α diversity of bacterial communities from sediment and water samples. In the present study we focused on benthic diatom diversity, which is known to be lower than diversity of bacterial community. Thus, the α diversity (Shannon index) was not significantly affected when using only a restricted amount of reads (500 DNA reads per sample). In the same manner, reducing drastically the number of DNA reads to 500 does not change SPI values and the final water quality assessment. Although a high sequencing depth is required to detect rare species (Valentini et al., 2016), this is not required for water quality assessment purposes where rare species have low impact on index values. This indicates that it may be possible to increase the number of samples in one HTS run, reducing the cost of the DNA metabarcoding approach which tends to make it economically suitable compared to the classical approach. Complementary to our in silico subsampling, analysis in real laboratory condition are needed to confirm the minimum sequencing depth required for water quality assessment in order to propose a standard like it was done for the WFD morphological approach (with a minimum of 400 morphological counts, Afnor, 2014b).

not detected by the molecular approach. Thus, this taxon is responsible for part of the divergence between molecular and morphological SPI values, molecular SPI being around 3 points higher than morphological SPI for all samples. The incompleteness of the reference library was not in cause in that case as R-syst::diatom (v3) contains 9 sequences of N. inconspicua, among which 3 were obtained from strains isolated from Mayotte. However, N. inconspicua is a paraphyletic species (Rovira et al., 2015), making precise taxonomic assignment very difficult at the genus/species level as discussed previously by Vasselon et al. (2017). Its absence from molecular inventories tends to produce higher SPI values with metabarcoding than with the morphological approach, the later including this low-quality taxon for SPI calculation while the former does not. As the DNA reference library construction is based on traditional taxonomy, the efficiency of the OTU taxonomic assignment relies on the reliability and extent of taxonomic knowledge’s. However, nomenclature of diatoms is far from static and evolves over time (Cox, 2009; Jahn and Kusber 2009; Kociolek and Williams 2015), which creates problems of taxonomic harmonization in the reference libraries and consequently biases in taxonomic assignation of OTUs. To improve the consistency and accuracy of reference libraries for metabarcoding purposes, it is crucial to up-date them with evolving taxonomical knowledge as well as with new DNA barcode references, which is done regularly for the open-access R-syst::diatom library (Rimet et al., 2016). 4.2.3. Bias related to diatom biovolume Read proportions were observed to differ from cell proportions in some cases, which may impact derived molecular and morphological SPI values respectively. This was the case for the Eunotia genus for which variations between molecular and morphological SPI values were high and positively correlated to the proportion of DNA reads assigned to this particular genus. The read proportion of a taxon can be affected by different technological (e.g. DNA extraction, PCR amplification, DNA sequencing, bioinformatics filtering) and biological (e.g. cell biomass) factors. The sum of those factors could result in a significant variation between DNA read and diatom cell proportions. In the case of Mayotte samples, the Eunotia genus was overrepresented in the molecular inventories compared to the morphological ones and was often a dominant genus. The different species of Eunotia observed in Mayotte, using microscope, are characterised by a large cell biovolume (around 19,000 μm3). Previous studies showed the existence of a clear correlation between the SSU rDNA gene copy number and the diatom cell size and biovolume (Zhu et al., 2005; Godhe et al., 2008). If such a relationship exists between rbcL copy number and diatom cell size, it can explain why the Eunotia genus is overestimated in our molecular inventories and why its presence affects the SPI calculation. Should this hypothesis be confirmed, a general correction factor based on diatoms species biovolume could be envisaged and would help to improve the comparability between molecular and morphological indices. However, Tapolczai et al., (2017) have shown that the biovolume itself may be important to consider to assess quality status. They proposed to use the biovolume to weight diatom counts as it is routinely done for quality assessment based on other ecological quality elements (e.g. phytoplankton). The theory behind is that the biomass partition of species shows better how resources are capitalized by the species (Kalff and Knoechel 1978; Padisák et al., 2006; Reynolds 1980). The bias we observed here due to biovolume may in the end prove to be interesting to improve bioassessement accuracy in future indices that may be developed directly from sequence data. Angly et al. (2014) already proposed to use a correction factor based on 16S rRNA genome number variation to correct the molecular data and provide more reliable microbial community profiles. Further investigation is required to propose more reliable molecular inventories and adapted quality indices.

4.3. Conclusion and perspectives The use of DNA metabarcoding and HTS for diatoms appears to be a promising approach for freshwater quality assessment. Our study confirmed at a larger scale previous observations that it is possible to infer ecological quality status of rivers based on molecular inventories (e.g. Kermarrec et al., 2014; Visco et al., 2015). However, this approach still requires optimisations and to define standards for key steps of the metabarcoding workflow in order to produce reliable and inter-comparable data between laboratories and platforms. Both HTS quantitative and qualitative information can be improved by working on the major biases like the use of a correction factor for the variation of rbcL copy number (based on diatom cell biovolume for example) or the completion of the DNA barcode reference library. While the creation of correction factor required new experiments and investigations, different solutions can be applied to complete the reference library. Availability of an expert and open-access reference library as R-syst::diatom (Rimet et al., 2016) allows sharing and centralizing knowledge’s from different laboratories, increasing the number of available references. The use of HTS data to complete DNA barcoding libraries, recently proposed by Rimet et al. (2017, in press), could be an efficient way to access to DNA barcode of uncultivable diatom taxa and improve database completion, especially for taxa that play an important part in water quality assessment. In order to consider the progressive implementation of molecular approaches into large scale biomonitoring networks and to meet their requirements (e.g. WFD, CWA), it will be essential that scientists, environmental stakeholders and managers work hand in hand to validate new methods and to propose of implementation scenarios. The creation of international scientific networking groups, like the European DNAqua-Net Cost action (Leese et al. 2016, www.dnaqua.net), is an efficient way to improve scientific knowledge by addressing collectively

4.2.4. Bias related to sequencing depth DNA metabarcoding combined to HTS allows multiplexing and 10

Ecological Indicators 82 (2017) 1–12

V. Vasselon et al.

Deiner, K., Walser, J.-C., Mächler, E., Altermatt, F., 2015. Choice of capture and extraction methods affect detection of freshwater biodiversity from environmental DNA. Biol. Conserv. 183, 53–63. Directive 2000/60/EC, 2000. Water framework directive of the european parliament and the council, of 23 october 2000, establishing a framework for community action in the field of water policy. Off. J. Eur. Commun. L327, 1–72. Elbrecht, V., Leese, F., 2015. Can DNA-based ecosystem assessments quantify species abundance? testing primer bias and biomass—sequence relationships with an innovative metabarcoding protocol. PLoS One 10, e0130324. Elbrecht, V., Leese, F., 2016. Primer Miner: an R package for development and in silico validation of DNA metabarcoding primers. Methods Ecol. Evol. 8, 622–626. Elbrecht, V., Leese, F., 2017. Validation and development of COI metabarcoding primers for freshwater macroinvertebrate bioassessment. Front. Environ. Sci. 5, 1–11. Gassiole, G., 2014. Diatomées épilithiques des cours d’eau pérennes de l’île de la Réunion: taxinomie ?écologie (Doctoral dissertation). Biodiversité et Ecologie. Université de Bordeaux, Français. Retrieved from HAL (HAL Id: tel-01187627 – version 1, https:// tel.archives-ouvertes.fr/tel-01187627). Godhe, A., Asplund, M.E., Harnstrom, K., Saravanan, V., Tyagi, A., Karunasagar, I., 2008. Quantification of diatom and dinoflagellate biomasses in coastal marine seawater samples by real-time PCR. Appl. Environ. Microbiol. 74, 7174–7182. Gueguen, J., Eulin, A., Lefrançois, E., Boutry, S., Tison Rosebery, J., Coste, M., Delmas, F., 2015. Production of an Improved Version of the Indice Diatomique Antilles (IDA-2), Use for the Evaluation of the Ecological Status of Rivers in the French Caribbean: Final Version, 2015-03-12. IRSTEA Scientific Report. Pages 185. http://cemadoc. irstea.fr/cemoa/PUB00044101. Hajibabaei, M., Shokralla, S., Zhou, X., Singer, G. a C., Baird, D.J., 2011. Environmental barcoding: a next-generation sequencing approach for biomonitoring applications using river benthos. PLoS One 6, e17497. Hamilton, P.B., Lefebvre, K.E., Bull, R.D., 2015. Single cell PCR amplification of diatoms using fresh and preserved samples. Front. Microbiol. 6. Hürlimann, J., Niederhauser, P., 2007. Méthodes d’Analyse et d’Appréciation des Cours d’Eau. Diatomées Niveau R (région); Etat de l’environnement no 0740. Office Fédéral de l’Environnement, Berne 132p. Ibáñez, C., Caiola, N., Sharpe, P., Trobajo, R., 2010. Ecological indicators to assess the health of river ecosystems. In: Jørgensen, S.E., Xu, L., Costanza, R. (Eds.), Handbook of Ecological Indicators for Assessment of Ecosystem Health, 2nd. ed. CRC Press, Boca Raton, Florida, pp. 447–464. Irannia, Z.B., Chen, T., 2016. TACO: taxonomic prediction of unknown OTUs through OTU co-abundance networks. Quant. Biol. 4, 149–158. Jahn, R., Kusber, W.-H., 2009. A key to diatom nomenclature. Diatom Res. 24, 101–111. Kalff, J., Knoechel, R., 1978. Phytoplankton and their dynamics in oligotrophic and eutrophic lakes. Annu. Rev. Ecol. Syst. 9, 475–495. Kebschull, J.M., Zador, A.M., 2015. Sources of PCR-induced distortions in highthroughput sequencing data sets. Nucleic Acids Res. 43, gkv717. Keck, F., Vasselon, V., Tapolczai, K., Rimet, F., Bouchez, A., 2017. Freshwater Biomonitoring in the Information Age. Frontiers in Ecology and the Environment. Kelly, M., 2013. Data rich, information poor? Phytobenthos assessment and the Water Framework Directive. Eur. J. Phycol. 48, 437–450. Kermarrec, L., Bouchez, A., Rimet, F., Humbert, J.-F., 2013a. First evidence of the existence of semi-cryptic species and of a phylogeographic structure in the gomphonema parvulum (Kützing) Kützing complex (Bacillariophyta). Protist 164, 686–705. Kermarrec, L., Franc, A., Rimet, F., Chaumeil, P., Humbert, J.F., Bouchez, A., 2013b. Next-generation sequencing to inventory taxonomic diversity in eukaryotic communities: a test for freshwater diatoms. Mol. Ecol. Resources 13, 607–619. Kermarrec, L., Franc, A., Rimet, F., Chaumeil, P., Frigerio, J.-M., Humbert, J., Bouchez, A., 2014. A next-generation sequencing approach to river biomonitoring using benthic diatoms. Freshwater Sci. 33, 349–363. Khan-Bureau, D.A., Morales, E.A., Ector, L., Beauchene, M.S., Lewis, L.A., 2016. Characterization of a new species in the genus Didymosphenia and of Cymbella janischii (Bacillariophyta) from Connecticut, USA. Eur. J. Phycol. 262, 1–14. Kociolek, J.P., Williams, D.M., 2015. How to define a diatom genus? Notes on the creation and recognition of taxa, and a call for revisionary studies of diatoms. Acta Bot. Croat. 74, 195–210. Krammer, K., Lange-Bertalot, H., 1986. Bacillariophyceae 1. Teil: Naviculaceae. Süßwasserflora von Mitteleuropa. Gustav Fischer Verlag, Stuttgart edn: 876 pages. Krammer, K., Lange-Bertalot, H., 1988. Bacillariophyceae 2. Teil: Bacillariaceae, Epithemiaceae, Surirellaceae. Süßwasserflora von Mitteleuropa. Gustav Fischer Verlag, Stuttgart edn: 610 pages. Krammer, K., Lange-Bertalot, H., 1991a. Bacillariophyceae 3. Teil: Centrales, Fragilariaceae, Eunotiaceae. Süßwasserflora von Mitteleuropa. Gustav Fischer Verlag, Stuttgart edn: 598 pages. Krammer, K., Lange-Bertalot, H., 1991b. Bacillariophyceae 4. Teil: Achnanthaceae. Kritische Ergänzungen zu Navicula (Lineolatae) und Gomphonema. Gesamtliteraturverzeichnis Teil 4. Süßwasserflora von Mitteleuropa. Gustav Fischer Verlag, Stuttgart edn: 437 pages. Krammer, K., 2000. The Genus Pinnularia. Gantner Verlag, Ruggell 703 pages. Krammer, K., 2001. Navicula Sensu Stricto, 10 Genera Separated from Navicula Sensu Stricto, Frustulia. Gantner Verlag, Ruggell 526 pages. Krammer, K., 2002. Cymbella. Gantner Verlag, Ruggell 584 pages. Krammer, K., 2003. Cymbopleura, Delicata, Navicymbula, Gomphocymbellopsis, Afrocymbella. Gantner Verlag, Ruggell 530 pages. Lavoie, I., Somers, K.M., Paterson, A.M., Dillon, P.J., 2005. Assessing scales of variability in benthic diatom community structure. J. Appl. Phycol. 17, 509–513. Lavoie, I., Dillon, P.J., Campeau, S., 2009. The effect of excluding diatom taxa and reducing taxonomic resolution on multivariate analyses and stream bioassessment. Ecol. Indic. 9, 213–225.

current issues and to share knowledge with stakeholders in order to consider together the implementation of these new approaches. A better understanding of the DNA metabarcoding pros and cons, together with more scaling up, is still required prior any standardization and deployment. One major challenge will be to make the link between historical biomonitoring methods applied over decades and new molecular methods, in order to have continuity in water quality assessment. Acknowledgments This paper was produced as part of the program for the development of biomonitoring network of Mayotte rivers and was funded by the French National Agency for Water and Aquatic Environments (ONEMAAFB). This work was supported by the European COST action DNAquaNet (CA 15219). We thank Philippe Chaumeil (INRA Biogeco) who helped us developing the rbcL primers, Sonia Lacroix who participates to the preparation of HTS libraries, Franck Salin and Christophe Boury who performed HTS sequencing (INRA-PGTB sequencing platform). We also thank people in BRGM, DEAL Mayotte, Asconit and ONEMA for their great support and contribution during the sampling campaigns in Mayotte. Special thanks to Gilles Gassiole who performed sampling and microscopy inventories for the RCS network. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.ecolind.2017.06.024. References Afnor, 2007. NF T90 354 − Qualité de l'eau − Détermination de l’Indice Biologique Diatomées (IBD). Afnorpp. 1–73. Afnor, 2014a. N. F EN 13946 − Qualité de l'eau − Guide pour l'échantillonnage en routine et le prétraitement des diatomées benthiques de rivières et de plans d'eau. Afnorpp. 1–18. Afnor, 2014b. N. F EN 14407 − Qualité de l'eau − Guide pour l'identification et le dénombrement des échantillons de diatomées benthiques de rivières et de lacs. Afnorpp. 1–13. Afnor, 2016. NF T90 354 − Qualité de l'eau − Échantillonnage, traitement et analyse de diatomées benthiques en cours d'eau et canaux. Afnorpp. 1–79. Angly, F.E., Dennis, P.G., Skarshewski, A., Vanwonterghem, I., Hugenholtz, P., Tyson, G.W., 2014. CopyRighter: a rapid tool for improving the accuracy of microbial community profiles through lineage-specific gene copy number correction. Microbiome 2, 11. L. Apothéloz-Perret-Gentil , A. Cordonier , F. Straub , J. Iseli , P. Esling , J. Pawlowski . Taxonomy-free molecular diatom index for high-throughput eDNA biomonitoring Mol. Ecol. Resources 2017; 8: 622-626 DOI: 10.1111/1755-0998.12668 Bellinger, B.J., Cocquyt, C., O’Reilly, C.M., 2006. Benthic diatoms as indicators of eutrophication in tropical streams. Hydrobiologia 573, 75–87. Bere, T., 2016. Challenges of diatom-based biological monitoring and assessment of streams in developing countries. Environ. Sci. Pollut. Res. 23, 5477–5486. Bigler, C., Gälman, V., Renberg, I., 2009. Numerical simulations suggest that counting sums and taxonomic resolution of diatom analyses to determine IPS pollution and ACID acidity indices can be reduced. J. Appl. Phycol. 22, 541–548. Birk, S., Bonne, W., Borja, A., Brucet, S., Courrat, A., Poikane, S., Solimini, A., van de Bund, W., Zampoukas, N., Hering, D., 2012. Three hundred ways to assess Europe’s surface waters: an almost complete overview of biological methods to implement the Water Framework Directive. Ecol. Indic. 18, 31–41. Bourrelly, P., Manguin, E., 1952. Algues d'eau douce de la Guadeloupe et dépendances, Paris, France. 235 pages. Bruder, K., Medlin, L.K., 2007. Molecular assessment of phylogenetic relationships in selected species/genera in the naviculoid diatoms (Bacillariophyta). I. The genus Placoneis. Nova Hedwigia 85, 331–352. Cemagref, 1982. Étude des méthodes biologiques quantitative d’appréciation de la qualité des eaux. Bassin Rhône-Méditerranée-Corse. Centre National du Machinisme Agricole, du Génie rural, des Eaux et des Forêts, Lyon, France. Chonova, T., Keck, F., Labanowski, J., Montuelle, B., Rimet, F., Bouchez, A., 2016. Separate treatment of hospital and urban wastewaters: a real scale comparison of effluents and their effect on microbial communities. Sci. Total Environ. 542, 965–975. Coste, M., 1986. Les méthodes microfloristiques d’évaluation de la qualité des eaux. Cemagref, Bordeaux 15 pp + 46 annexes. Cox, E.J., 2009. What’s in a name? Diatom classification should reflect systematic relationships. Acta Bot. Croat. 68, 443–454. Creer, S., 2010. Second-generation sequencing derived insights into the temporal biodiversity dynamics of freshwater protists. Mol. Ecol. 19, 2829–2831.

11

Ecological Indicators 82 (2017) 1–12

V. Vasselon et al.

Bouchez, A., 2016. R-syst::diatom: an open-access and curated barcode database for diatoms and freshwater monitoring. Database 2016, baw016. Rimet, F., Abarca, N., Bouchez, A., Kusber, W.H., Jahn, R., Kahlert, M., Keck, F., Kelly, M., Mann, D., Piuz, A., Trobajo, R., Tapolczai, K., Vasselon, V., Zimmermann, J., 2017. 2017. The potential of high throughput sequencing (HTS) of natural samples as a source of primary taxonomic information for reference libraries of diatom barcodes. Fottea (in press). Rott, E., Pipp, E., Pfister, P., 2003. Diatom methods developed for river quality assessment in Austria and a cross-check against numerical trophic indication methods used in Europe. Algol. Stud. 110, 91–115. Rovira, L., Trobajo, R., Sato, S., Ibáñez, C., Mann, D.G., 2015. Genetic and physiological diversity in the diatom nitzschia inconspicua. J. Eukaryot. Microbiol. 62, 815–832. Schloss, P.D., Westcott, S.L., Ryabin, T., Hall, J.R., Hartmann, M., Hollister, E.B., Lesniewski, R.A., Oakley, B.B., Parks, D.H., Robinson, C.J., Sahl, J.W., Stres, B., Thallinger, G.G., Van Horn, D.J., Weber, C.F., 2009. Introducing mothur: opensource, platform-independent, community-supported software for describing and comparing microbial communities. Appl. Environ. Microbiol. 75, 7537–7541. Schmidt, T.S.B., Matias Rodrigues, J.F., von Mering, C., 2015. Limits to robustness and reproducibility in the demarcation of operational taxonomic units. Environ. Microbiol. 17, 1689–1706. Smith, D.P., Peay, K.G., 2014. Sequence depth, not PCR replication, improves ecological inference from next generation DNA sequencing. PLoS One 9, e90234. Stoof-leichsenring, K.R., Epp, l. S., Trauth, M.H., Tiedemann, R., 2012. Hidden diversity in diatoms of Kenyan Lake Naivasha: a genetic approach detects temporal variation. Mol. Ecol. 21, 1918–1930. Tapolczai, K., Bouchez, A., Stenger-Kovács, C., Padisák, J., Rimet, F., 2017. Species- or trait-based ecological assessment for tropical rivers? Case study of benthic diatoms in Mayotte island (France, northern Mozambique Channel). Stoten (Submitted). Teittinen, A., Taka, M., Ruth, O., Soininen, J., 2015. Variation in stream diatom communities in relation to water quality and catchment variables in a boreal, urbanized region. Sci. Total Environ. 530–531, 279–289. Thomas, A.C., Deagle, B.E., Eveson, J.P., Harsch, C.H., Trites, A.W., 2016. Quantitative DNA metabarcoding: improved estimates of species proportional biomass using correction factors derived from control material. Mol. Ecol. Resources 16, 714–726. Tudesque, L., Rimet, F., Ector, L., 2008. A new taxon of the section Nitzschiae Lanceolatae Grunow: Nitzschia costei sp. nov. compared to N. fonticola Grunow, N. macedonica Hustedt, N. tropica Hustedt and related species. Diatom Res. 23, 483–501. Valentini, A., Taberlet, P., Miaud, C., Civade, R., Herder, J., Thomsen, P.F., Bellemain, E., Besnard, A., Coissac, E., Boyer, F., Gaboriaud, C., Jean, P., Poulet, N., Roset, N., Copp, G.H., Geniez, P., Pont, D., Argillier, C., Baudoin, J.-M., Peroux, T., Crivelli, A.J., Olivier, A., Acqueberge, M., Le Brun, M., Møller, P.R., Willerslev, E., Dejean, T., 2016. Next-generation monitoring of aquatic biodiversity using environmental DNA metabarcoding. Mol. Ecol. 25, 929–942. Vasselon, V., Domaizon, I., Rimet, F., Kahlert, M., Bouchez, A., 2017. Application of highthroughput sequencing (HTS) metabarcoding to diatom biomonitoring: do DNA extraction methods matter? Freshwater Sci. 36, 162–177. Visco, J.A., Apothéloz-Perret-Gentil, L., Cordonier, A., Esling, P., Pillet, L., Pawlowski, J., 2015. Environmental monitoring: inferring the diatom index from next-generation sequencing data. Environ. Sci. Technol. 49, 7597–7605. Yang, Y., Cao, J.-X., Pei, G.-F., Liu, G.-X., 2015. Using benthic diatom assemblages to assess human impacts on streams across a rural to urban gradient. Environ. Sci. Pollut. Res. 22, 18093–18106. Zelinka, M., Marvan, P., 1961. Zur präzisierung der biologischen klassifikation der reinheit fließender gewässer. Archiv für Hydrobiologie 57 (3), 389–407. Zhu, F., Massana, R., Not, F., Marie, D., Vaulot, D., 2005. Mapping of picoeucaryotes in marine ecosystems with quantitative PCR of the 18S rRNA gene. FEMS Microbiol. Ecol. 52, 79–92. Zimmermann, J., Abarca, N., Enk, N., Skibbe, O., Kusber, W.-H., Jahn, R., 2014. Taxonomic reference libraries for environmental barcoding: a best practice example from diatom research. PLoS One 9, e108793. Zimmermann, J., Glöckner, G., Jahn, R., Enke, N., Gemeinholzer, B., 2015. Metabarcoding vs. Morphological Identification to Assess Diatom Diversity in Environmental Studies.

Lecointe, C., Coste, M., Prygiel, J., 1993. Omnidia: software for taxonomy, calculation of diatom indices and inventories management. Hydrobiologia 269–270, 509–513. Leese, F., Altermatt, F., Bouchez, A., Ekrem, T., Hering, D., Meissner, K., Mergen, P., Pawlowski, J., Piggott, J., Rimet, F., Steinke, D., Taberlet, P., Weigand, A., Abarenkov, K., Beja, P., Bervoets, L., Björnsdóttir, S., Boets, P., Boggero, A., Bones, A., Borja, Á., Bruce, K., Bursić, V., Carlsson, J., Čiampor, F., Čiamporová-Zatovičová, Z., Coissac, E., Costa, F., Costache, M., Creer, S., Csabai, Z., Deiner, K., DelValls, Á., Drakare, S., Duarte, S., Eleršek, T., Fazi, S., Fišer, C., Flot, J.-F., Fonseca, V., Fontaneto, D., Grabowski, M., Graf, W., GuÐbrandsson, J., Hellström, M., Hershkovitz, Y., Hollingsworth, P., Japoshvili, B., Jones, J., Kahlert, M., Kalamujic Stroil, B., Kasapidis, P., Kelly, M., Kelly-Quinn, M., Keskin, E., Kõljalg, U., Ljubešić, Z., Maček, I., Mächler, E., Mahon, A., Marečková, M., Mejdandzic, M., Mircheva, G., Montagna, M., Moritz, C., Mulk, V., Naumoski, A., Navodaru, I., Padisák, J., Pálsson, S., Panksep, K., Penev, L., Petrusek, A., Pfannkuchen, M., Primmer, C., Rinkevich, B., Rotter, A., Schmidt-Kloiber, A., Segurado, P., Speksnijder, A., Stoev, P., Strand, M., Šulčius, S., Sundberg, P., Traugott, M., Tsigenopoulos, C., Turon, X., Valentini, A., van der Hoorn, B., Várbíró, G., Vasquez Hadjilyra, M., Viguri, J., Vitonytė, I., Vogler, A., Vrålstad, T., Wägele, W., Wenne, R., Winding, A., Woodward, G., et al., 2016. DNAqua-net: developing new genetic tools for bioassessment and monitoring of aquatic ecosystems in Europe. Res. Ideas Outcomes 2, e11321. Lejzerowicz, F., Esling, P., Pillet, L., Wilding, T.A., Black, K.D., Pawlowski, J., 2015. Highthroughput sequencing and morphology perform equally well for benthic monitoring of marine ecosystems. Sci. Rep. 5, 13932. Lenoir, A., Coste, M., 1996. Development of a practical diatom index of overall water quality applicable to the French National Water Board Network. In: Whitton, B.A., Rott, E. (Eds.), Use of Algae for Monitoring Rivers II. International Symposium, Volksbildungsheim Grilhof Vill, AUT, 17–19 September 1995. Universität Innsbruck, pp. 29–43. Lundin, D., Severin, I., Logue, J.B., Östman, Ö., Andersson, A.F., Lindström, E.S., 2012. Which sequencing depth is sufficient to describe patterns in bacterial α- and β-diversity? Environ. Microbiol. Rep. 4, 367–372. Mangot, J.-F., Domaizon, I., Taib, N., Marouni, N., Duffaud, E., Bronner, G., Debroas, D., 2013. Short-term dynamics of diversity patterns: evidence of continual reassembly within lacustrine small eukaryotes. Environ. Microbiol. 15, 1745–1758. Metezeltin, D., Lange-Bertalot, H., 1998. Tropical diatoms of South America I. Iconographia Diatomologica 5, 1–695. Metezeltin, D., Lange-Bertalot, H., 2007. Tropical diatoms of South America II. Iconographia Diatomologica 18, 1–877. Padisák, J., Borics, G., Grigorszky, I., Soróczki-Pintér, É., 2006. Use of phytoplankton assemblages for monitoring ecological status of lakes within the water framework directive: the assemblage index. Hydrobiologia 553, 1–14. Pandey, L.K., Bergey, E.A., Lyu, J., Park, J., Choi, S., Lee, H., Depuydt, S., Oh, Y.-T., Lee, S.-M., Han, T., 2017. The use of diatoms in ecotoxicology and bioassessment: insights, advances and challenges. Water Res. 118, 39–58. Pawlowski, J., Lejzerowicz, F., Apotheloz-Perret-Gentil, L., Visco, J., Esling, P., 2016. Protist metabarcoding and environmental biomonitoring: time for change. Eur. J. Protistol. 55, 12–25. Potapova, M., Charles, D.F., 2007. Diatom metrics for monitoring eutrophication in rivers of the United States. Ecol. Indic. 7, 48–70. Prygiel, J., Carpentier, P., Almeida, S., Coste, M., Druart, J.C., Ector, L., Guillard, D., Honoré, M.A., Iserentant, R., Ledeganck, P., Lalanne-Cassou, C., Lesniak, C., Mercier, I., Moncaut, P., Nazart, M., Nouchet, N., Peres, F., Peeters, V., Rimet, F., Rumeau, A., Sabater, S., Straub, F., Torrisi, M., Tudesque, L., Van de Vijver, B., Vidal, H., Vizinet, J., Zydek, N., 2002. Determination of the biological diatom index (IBD NF T 90–354): Results of an intercomparison exercise. J. Appl. Phycol. 14, 27–39. Quail, M., Smith, M.E., Coupland, P., Otto, T.D., Harris, S.R., Connor, T.R., Bertoni, A., Swerdlow, H.P., Gu, Y., 2012. A tale of three next generation sequencing platforms: comparison of Ion torrent, pacific biosciences and illumina MiSeq sequencers. BMC Genomics 13, 341. R Development core team, 2013. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Reynolds, C.S., 1980. Phytoplankton assemblages and their periodicity in stratifying lake systems. Ecography 3, 141–159. Rimet, F., Chaumeil, P., Keck, F., Kermarrec, L., Vasselon, V., Kahlert, M., Franc, A.,

12

Avoiding quantification bias in metabarcoding: application of a cell biovolume correction factor in diatom molecular biomonitoring Valentin Vasselon*, Agnès Bouchez*, Frédéric Rimet *, Stéphan Jacquet*, Rosa Trobajo†, Méline Corniquel*, Kálmán Tapolczai *, Isabelle Domaizon* *CARRTEL,

French National Institute for Agricultural Research (INRA), University of Savoie Mont Blanc, 75 bis Avenue de Corzent, 74200, Thonon-les-Bains, France, † Aquatic Ecosystems, Institute for Food and Agricultural Research and Technology (IRTA), Crta de Poble Nou Km 5.5, Sant Carles de la Ràpita, Catalunya, Spain.

Corresponding author: Valentin Vasselon; 75 bis Avenue de Corzent, 74200, Thonon-lesBains, France; +33 (0)4 50 26 78 29; [email protected] Running headline: Improvement of diatom HTS quantification

Abstract 1. In recent years, remarkable progress has been made in developing environmental DNA metabarcoding. However, its ability to quantify species relative abundance remains uncertain, limiting its application for biomonitoring. In diatoms, although the rbcL gene appears to be a suitable barcode for diatoms, providing relevant qualitative data to describe taxonomic composition, improvement of species quantification is still required. 2. Here, we hypothesized that rbcL copy number is correlated with diatom cell biovolume (as previously described for the 18S gene) and that a correction factor (CF) based on cell biovolume should be applied to improve taxa quantification. We carried out a laboratory experiment using pure cultures of 8 diatom species with contrasted cell biovolumes in order to (i) verify the relationship between rbcL copy numbers (estimated by qPCR) and diatom cell biovolumes, and (ii) define a potential CF. In order to evaluate CF efficiency, five mock communities were created by mixing different amounts of DNA from the 8 species, and were sequenced using HTS and targeting the same rbcL barcode. 3. As expected, the correction of DNA reads proportions by the CF improved the congruence between morphological and molecular inventories. Final validation of the CF was obtained on environmental samples (metabarcoding data from 80 benthic biofilms) for which the application of CF allowed differences between molecular and morphological water quality indices to be reduced by 47 %. 4. Overall, our results highlight the usefulness of applying a CF factor, which is effective in reducing over-estimation of high biovolume species, correcting quantitative biases in diatom metabarcoding studies and improving final water quality assessment. Keywords: Benthic diatom, Biovolume correction factor, Freshwater ecosystems, Gene copy number variation, Quantitative metabarcoding

Introduction DNA metabarcoding allows species present in an environmental sample to be detected using a short DNA marker specific for a particular taxonomic group (Taberlet et al. 2012). Combined with High-Throughput Sequencing (HTS), hundreds of samples can be analyzed at the same time, offering an alternative to microscopy that is potentially cheaper and faster, and has high resolution (Stein et al. 2014). This is particularly interesting for freshwater biomonitoring, in which thousands of river samples have to be analyzed annually and management actions applied quickly (Keck et al. 2017). The European Water Framework Directive (WFD, European Council 2000) has implemented the use of benthic diatoms, among other biological indicators (fishes, macroinvertebrates, phytoplankton), for the assessment of aquatic ecosystem integrity. The different biotic diatom indices that have been developed are based on the relative abundances and the ecological values (sensitivity and tolerance to pollutants) of the species observed in river systems (e.g. Rimet 2012). Different studies have already revealed the potential application of diatom metabarcoding in freshwater quality assessment (Kermarrec et al. 2014; Visco et al. 2015; Vasselon et al. 2017a,b; ApothélozPerret-Gentil et al. 2017). However, discrepancies between DNA metabarcoding and microscopy have been observed in species composition and relative abundance (Zimmermann et al. 2015). This drawback is likely to affect the congruence between morphological and DNA metabarcoding quality index values and, in fine, the ecological assessment. With respect to qualitative aspects, the incompleteness of the reference databases, the choice of the DNA marker and the efficiency of the PCR primers have been identified as important biases affecting species detection using DNA metabarcoding (Pawlowski et al. 2016). For benthic diatoms, the rbcL gene has proved to be an appropriate taxonomic marker for biomonitoring (Mann et al. 2010; Kermarrec et al. 2013, 2014, Vasselon et al. 2017a,b) and a well-curated barcode reference library is already available in open-access to assign species names to rbcL sequences (R-Syst::diatom, Rimet et al. 2016). However, no clear relationship has yet been demonstrated between the relative species abundances obtained by DNA metabarcoding with the rbcL barcode and those obtained by morphological observations (Rimet et al. 2014). As quantification of diatom species is required by the WFD for quality index calculation (Hering et al. 2010), more investigation is needed to understand and correct biases affecting diatom quantification based on HTS data. Species quantification based on HTS data can be estimated from the number of DNA sequences (i.e. reads) assigned to each species, from which relative abundances can be calculated. Previous studies have documented a variety of problems that may affect the proportions of DNA reads obtained with HTS (Amend, Seifert & Bruns 2010; Deagle et al. 2013; Tan et al. 2015; Thomas et al. 2016; Pawlowski et al. 2016), including biological biases (e.g. gene copy number variation per cell, tissue cell density, cell biovolume), technical biases (e.g. DNA extraction method, PCR amplification), and biases linked to HTS itself (e.g. library construction, HTS technology used, bioinformatics treatments). Variation of gene copy number per cell constitutes a major bias known to affect the proportion of DNA-read found for each species present in complex assemblages; this has been demonstrated for macroinvertebrates (Elbrecht, Peinert & Leese 2017), fish, amphibians (Evans et al. 2016), oligochaetes (Vivien, Lejzerowicz & Pawlowski 2016), foraminifera (Weber & Pawlowski 2013), and microbial communities (Angly et al. 2014). However, to the best of our knowledge, no study has yet evaluated gene copy number variation bias on diatom metabarcoding quantification. While tissue cell density and species biomass are major biases likely to affect DNA metabarcoding quantification of multicellular organisms like macroinvertebrates (Elbrecht & Leese 2015) or fish (Evans et al. 2016), diatoms are unicellular organisms for which gene copy number is mainly affected by the number of genomes and the number of gene copies per genome. This may be particularly true for non-nuclear markers like the

chloroplast-encoded rbcL gene. Godhe et al. (2008) reported a clear correlation between the 18S gene copy number per cell with diatom cell length and biovolume, suggesting that the cell biovolume could be a proxy for the gene copy number. Keeping in mind that diatom biovolume varies from 101 to 109 µm3 (Snoeijs, Busse & Potapova 2002), gene copy number may vary greatly between the smallest and the biggest diatom species, affecting metabarcoding quantification. For all the reasons mentioned above, we hypothesized that a quantification correction factor (CF) based on diatom cell biovolume should be necessary to correct DNA read proportions to provide species quantification more comparable to microscopical counts. In order to confirm this hypothesis, we firstly conducted experiments on 8 pure diatom cultures to examine whether variation in rbcL gene copy number per cell correlates with morphological characteristics (e.g. biovolume, cell length), from which a CF might be calculated. Secondly, the efficiency of the proposed CF was tested on (i) mock communities made by mixing known proportions of the 8 diatom species cultures, and (ii) environmental diatom communities from rivers previously sequenced (Vasselon et al. 2017b) and for which data are available online (Vasselon et al. 2017b dataset, http://doi.org/10.5281/zenodo.400160). Last, the capacity of the CF to improve the ecological assessment of rivers was tested by comparing new water quality index values calculated from molecular data with corrected abundances to those calculated from classical morphological abundances.

Methods Evaluation of the quantification bias and development of a quantification correction factor (CF) To evaluate whether the rbcL copy number per cell varies between diatom species, strains from 8 freshwater diatom species were selected from the Thonon Culture Collection (TCC; http://www6.inra.fr/carrtel-collection_eng/) (Table 1). The 8 species were chosen for their contrasted morphological (size and cell biovolume), cytological (e.g. chloroplast number) and phylogenetic characteristics. Moreover, their rbcL genes were already sequenced and referenced in the RSyst::diatom database (Rimet et al. 2016) allowing taxonomic assignment of DNA reads from HTS data (Table 1). Cell dimensions (width, length, thickness) of the 8 diatom species were measured under light microscopy (1000× magnification) using a minimum of 10 specimens per species. Then, appropriate geometrical models were applied to calculate their cell biovolume (Sun & Liu 2003) (Table 1). The 8 diatom cultures were cultivated in triplicate in 40 mL sterile DV medium (Rimet et al. 2014) using 50 mL Nunc™ EasYFlasks™ (Thermo Fisher Scientific, Waltham, Massachusetts). Flasks were placed in a controlled thermostatic room at 21 ± 2°C and under a light/dark cycle (14h/10h) with light intensity of ca. 100 µmol quanta m-2 s-1. In order to provide a homogeneous distribution of light to all cultures, cultivation flasks were placed on a platter rotating at 4 rpm. Flasks were inoculated in order to reach a concentration of ≈ 100 cells/mL at the beginning of the experiment for each species, except for Ulnaria ulna for which a concentration of ≈ 1000 cells/mL was used (due to its low growth rate). The growth of the 8 diatom cultures was followed during 40 days, except for Pinnularia viridiformis for which the survey lasted 73 days, due to its low growth rate. Cell concentrations, proportions of live/dead cells and rbcL gene copy concentrations per mL of media were measured for each culture at 7 sampling times (referred to as T0 to T6) (Fig. 1). Diatom cell concentrations and proportions of live/dead cells were obtained by counting at least 400 specimens using inverted microscopy (at ×1000 magnification) and the standard Utermöhl technique (European Committee for Standardization (CEN) 2006) (Fig. 1).

The proportion of live/dead cells was estimated by considering cells without visible intracellular contents as dead. Only living cells were taken into account to calculate the diatom cell concentration per mL of media. Flow cytometry using Sytox-Green was also used to confirm the microscopical data (not shown). RbcL copy number per mL was estimated by qPCR. From each cultivation replicate, 10 mL of culture was centrifuged at 17,000 x g for 30 min (Fig.1). Total DNA was extracted from the resulting pellet using a protocol based on GenEluteTM-LPA DNA precipitation (SigmaAldrich, St Louis, Missouri) as previously described (Chonova et al. 2016; Vasselon et al. 2017a). Because the robustness of the diatom silica wall is known to reduce the efficiency of DNA extraction and can thus affect rbcL quantification, pellets remaining after DNA extraction were observed under the microscope and extraction was considered successful when no further cellular content was observed within the cells. Then, qPCR assays were performed for each of the 8 diatom species on DNA extracted at all 7 sampling times and with each of the 3 replicates, using the QuantiTect SYBR Green PCR Kit (Life Technologies, Carlsbad, USA) and the Rotor-Gene Q (Qiagen, Hilden, Germany). A short 312 bp region of the rbcL gene (the same as was used for HTS sequencing) was targeted using primers used by (Vasselon et al. 2017b) and described in Table S1. qPCR reactions were performed following the method used by Vasselon et al. (2017a), using a final volume of 25 µL using mix preparation and reaction conditions as described in Table S1. A fluorescence threshold of 0.01 was used to allow comparison of qPCR assays, denoising and determination of the cycles’ threshold (Ct). Data analysis was performed using the Rotor-Gene Q Series software (version 2.3.1) and the rbcL copy per mL of media was determined. Finally, the number of rbcL gene copies per diatom cell was calculated for the 8 diatom species by dividing the rbcL concentration (qPCR data) by the living cell concentration (microscopy data). A Kruskal-Wallis test was performed using R (R Development core team 2013) to determine if the rbcL gene copy number per diatom cell varied significantly between the 8 diatom species. Then, we tested the level of correlation between the number of rbcL gene copies per diatom cell and several morphological characteristics of the diatom cells: the length, the width, the thickness and the biovolume (Table 1). Variables that did not approximate normal distributions were log transformed. Pearson correlation coefficients were calculated between the gene copy number per cell and the diatom cell morphological characteristics. This correlation was represented by a linear model.

Validation of the quantification CF to mock and environmental HTS data Mock communities The calculated CF was applied to metabarcoding data obtained from controlled diatom mock communities. 5 mock communities (M1 to M5) were created by mixing DNA extracted from each of the 8 diatom species sampled during their exponential growth phase, and for which the correspondence between cell abundances (microscopy) and qPCR counts was known. For each of the 5 mock communities, the volume of DNA used for 7 species was kept unchanged (1 µL) and only the volume of DNA of P. viridiformis varied as followed: M1 = 0.2 µL, M2 = 0.4 µL, M3 = 0.8 µL, M4 = 1.6 µL, M5 = 3.2 µL. This resulted in contrasted rbcL proportions of the 8 species among the 5 mock communities. Then, HTS sequencing of the rbcL 312 bp fragment was performed on 3 replicates of the 5 mock communities. The 15 corresponding libraries were prepared following the method described by Vasselon et al. (2017a) with the same primers and PCR reaction conditions as those used for rbcL qPCR (Table S1), changing only the cycle number to 30. Each library was diluted to 100 pm and all 15 were pooled together for one HTS run performed on the PGM Ion Torrent machine by the “Plateforme Génome Transcriptome” (PGTB, Bordeaux, France). The sequencing platform provided a unique fastq file for each of the 15 libraries containing demultiplexed DNA reads without the sequencing adapters. Quality filtering of

DNA reads was performed using the Mothur software (Schloss et al. 2009) and bioinformatics process described previously (Vasselon et al. 2017a,b). Finally, a taxonomy was assigned to each DNA read with the “classify.seqs” command (Mothur) using default parameters with a confidence threshold of 85% and the R-Syst::diatom library (Rimet et al. 2016, version updated in January 2015 and available upon request) as a rbcL reference library. A molecular taxonomic list with the associated read numbers assigned to each of the 8 diatom species was obtained for each of the 5 mock communities and used for subsequent analysis. The quantification CF defined for the rbcL gene was then applied to the molecular taxonomic lists for the 5 mock communities by dividing the read number for each species by its corresponding CF. Both the uncorrected and corrected HTS relative abundances of species from the 5 mock communities were then compared to the relative abundances obtained using microscopy. Environmental diatom communities To evaluate the efficiency of the CF to improve metabarcoding quantification from environmental samples, we used rbcL HTS data obtained from (Vasselon et al. 2017b), corresponding to 80 benthic diatom samples collected from rivers in tropical island of Mayotte, Indian Ocean (Vasselon et al. 2017b dataset, http://doi.org/10.5281/zenodo.400160). A CF was calculated for each species (or genus when the species level was not reached) detected in molecular inventories of the rivers of Mayotte island using the morphological (e.g. biovolume, length) information available from the Rsyst::diatom library, and applied to HTS data. Corrected molecular inventories were produced for all the 80 river samples using the CF. The impact of the CF on diatom taxa abundance rank in the molecular inventories was assessed by comparing original and corrected molecular diatom inventories. Then, the Specific Pollution-sensitivity Index (SPI) used for ecological assessment was calculated for each sample based on the corrected diatom molecular inventories using the Omnidia 5 software (Lecointe, Coste & Prygiel 1993, library 5.3 2015) and compared to the morphological SPI values for all river samples (Vasselon et al. 2017b). Pearson correlation was used to evaluate the strength of correlations between original or corrected molecular SPI values and the morphological SPI values. Wilcoxon Signed Rank tests were conducted to determine whether the difference between the molecular and the morphological SPI (SPI) varied significantly when using the original or the corrected molecular data for the molecular SPI calculation.

Results Variation of rbcL gene copy number between diatom species Cell and rbcL gene concentrations were measured, by inverted microscopy and qPCR respectively, for the 8 diatom species at different cultivation stages corresponding to 7 sampling points (T0 to T6). Information has been summarized in Tables S2 and S3. As the 8 diatom species reached the beginning of the stationary phase at the sampling time T2 (i.e. between 13 and 31 days of cultivation), only the [cell] and the [gene copy] values obtained for the T0, T1 and T2 sampling times were used for further analysis. The calculated mean values of the rbcL gene copy number per cell for each diatom species varied between 0.5 and 130 copies per cell (Fig. 2). The Kruskal-Wallis test revealed that the rbcL copy number per cell was significantly different (p < 0.001) between the 8 diatom species. Development of quantification CFs The rbcL copy number per cell was highly correlated with cell biovolume (r = 0.97, p < 0.001), length (r = 0.82, p < 0.001), width (r = 0.94, p < 0.001) and thickness (r = 0.96, p < 0.001). The correlation between the rbcL copy number per cell and the cell biovolume followed a linear model (Fig. 3). Assuming that this linear relation is applicable to all diatom species, the equation of this model allows calculation of an estimate of the relative rbcL copy

number per cell as soon as the biovolume of the cell is known, and thus to define a CF specific to each species. Such quantification CFs were calculated for each of the 8 diatom species of the mock communities (Table 2) and varied from 0.6 for Achnanthidium minutissimum to 78.5 for P. viridiformis. For each of the diatom taxa found in the environmental samples, CFs were also calculated using the biovolume information available for each taxa (from Rsyst::diatom library) (Table S4) and varied over a wider range, from 0.03 for Fistulifera saprophila to 649.8 for Rhopalodia gibba. Application of CFs to mock and environmental HTS data 953,082 DNA reads were produced from the 15 libraries corresponding to the 5 DNA mock communities (3 replicates per mock). Following the bioinformatics quality filtering steps, 385,367 DNA reads were retained. A molecular taxonomic list was then created by removing DNA reads which remained unclassified (0.43 % of the reads) or assigned to different taxa than the 8 diatom species present in the mock communities (0.004 % of the reads) (Table S5). The proportions of P. viridiformis reads in the 5 mock communities varied from 9 % in M1 to 57 % in M5 (Fig. 4A) while observed cell proportions were lower; ≈ 0.03 % in M1 and 0.55 % in M5 (Fig. 4B). The application of the CF on DNA reads counts of the 8 species changed their relative abundances in the 5 mock communities (Fig. 4A). The rank of the 8 species was also affected; for example, in M5 the application of the CF changed the proportion of P. viridiformis from 57 % to 4 % and the proportion of A. minutissimum from 4 % to 42 %. The correspondence between morphological and molecular relative abundances was highly improved by applying the CF on the HTS data (Fig. 4A, 4B). From the 80 environmental samples previously sequenced (Vasselon et al. 2017b), a molecular taxonomic list based on assigned DNA reads was produced including 23 families (75.1 % of total reads assigned), 39 genera (72 % of total reads assigned) and 66 diatom species (40.7 % of total reads assigned). From this list, 84 diatom taxa, including taxa assigned at the genus and the species level, were used to calculate the SPI freshwater quality index. CFs calculated from cell biovolumes were then applied to correct the quantification of the environmental molecular inventories (Table S4). The proportions and ranks of the dominant taxa were affected by the application of the CFs (Fig. 5). For example, the application of CFs reduced the relative abundances of Eunotia and Ulnaria from 31.9 % to 3.3 % and 11.7 % to 2.3 %, respectively, making them more congruent with cell proportions observed with microscopy (3.1% for Eunotia and 0.4 % for Ulnaria). The correlation between the morphological and the molecular SPI values for all river samples previously described (r = 0.72, p < 0.001) was slightly improved using SPI values based on inventories with corrected abundances (r = 0.77, p < 0.001). The application of the CF to correct the HTS quantification reduced significantly (p < 0.001) the differences between the molecular and morphological SPI values by 47 % (SPI reduced to 1.9 on average compared to 3.6 before correction) (Fig. 6).

Discussion Correlation between rbcl gene copy number and diatom cell biovolume: impacts HTS quantification Since its first application to river biomonitoring by (Hajibabaei et al. 2011), DNA metabarcoding combined with HTS has appeared to be a promising and powerful approach for environmental biomonitoring surveys (Baird & Hajibabaei 2012; Shokralla et al. 2012). However, biases associated with such an approach can affect both the species detection and the estimation of species abundance from DNA read proportions, limiting its integration into existing monitoring programs. For benthic diatoms, a lot of effort has been expended during recent years to reduce the biases affecting species detection with metabarcoding, in particular to identify suitable DNA markers (Kermarrec et al. 2013), provide a reliable

reference database (Rimet et al. 2018), and validate the choice of the DNA extraction methods (Vasselon et al. 2017a). In this study we focused on the gene copy number variation bias that is likely to affect diatom quantification based on DNA reads proportions. The number of copy of the rbcL gene present in one diatom cell is affected by 3 parameters: the number of chloroplasts per cell, the number of genomes per chloroplast and the number of copies of the gene per chloroplast genome (Ersland, Aldrich & Cattolico 1981; Treusch et al. 2012). For benthic diatoms, the chloroplast number per cell is quite stable at the genus level with variations ranging from 1 to ≈ 8 chloroplast(s) per cell (Round, Crawford & Mann 1990) and there is only 1 copy of the rbcL gene per chloroplast genome (e.g. Sabir et al. 2014, like in higher plants (Gutteridge & Gatenby 1995) . Regarding the chloroplast genome number per cell, higher plants can contain up to thousands of copies of chloroplast genome per cell (Bendich 1987; Rauwolf et al. 2010) while unicellular algae generally exhibit a lower number of copies, for example, Olisthodiscus luteus (Raphidophyceae), Chlamydomonas reinhardtii (Chlorophyceae), the pennate diatom Phaeodactylum tricornutum and the centric diatom Thalassiosira pseudonana contain around 650, 80, 137 and 55 genome copies per cell, respectively (Ersland, Aldrich & Cattolico 1981; Koop et al. 2007; Gruber 2008; von Dassow et al. 2008). Thus, the rbcL copy number may vary from tens to hundreds of copies per diatom cell. Our estimations are within this range with a maximum of 130 copies estimated for P. viridiformis. However, our method underestimates the rbcL gene copy number since 0.5 copy per cell was estimated for A. minutissimum (so implying that some cells have no rbcL copy). This may result from certain variability inherent to the estimation of gene copy number by qPCR and the quantification of cells by microscopical counts. Our results demonstrate, however, that the rbcL copy number varies significantly between the 8 diatom species used in this study, according to the different diatom cell characteristics tested. In particular, we found a significant linear relationship between the rbcL copy number and the cell biovolume. Although the size of the chloroplasts could not be estimated in this study, we assume that the increase of the cell biovolume is accompanied by an increase of the chloroplast biovolume (as shown by Okie, Smith & Martin-Cereceda 2016), inducing an increase of DNA quantity and chloroplast genome copies per chloroplast as shown by Rauwolf et al. (2010). The correlation we found between the rbcL copy number and the diatom cell biovolume suggests that the relative abundance of diatom species with high cell biovolume is likely to be over-represented in metabarcoding data compared to microscopical counts. This is confirmed by the HTS data obtained for the mock communities, where diatom species with high cell biovolume are over-represented (e.g. P. viridiformis) and diatom species with low cell biovolume are under-represented (e.g. A. minutissimum). The relative abundance of P. viridiformis in the mock communities was negligible compared to other species, and doubling its proportion did not change its rank: the species remained the least abundant taxon within the morphological inventory. However, due to its high cell biovolume (104 µm3) and relatively high rbcL copy number per cell, a marked over-representation of this species within the molecular inventory was observed. A CF was thus defined to correct these quantitative biases and was verified on mock communities and environmental samples.

Current potential and limits of the quantification CF The use of the same rbcL primers for the qPCR assays and the HTS enabled us to generate a specific CF well suited to correct rbcL metabarcoding quantifications. Its application to the HTS data of the mock communities allowed us to obtain comparable species proportions in morphological and molecular based approaches of mock communities. This was also confirmed with the Mayotte river samples, for which the quantification CF resulted in a better congruence between DNA reads and cells proportions, reducing the over-

representation of high biovolume Eunotia and Ulnaria species. Furthermore, SPI calculation based on corrected metabarcoding data gives SPI values more comparable to SPI values obtained from morphological data, suggesting that it may be possible to replace morphological by molecular monitoring for the ecological assessment of Mayotte rivers. In the same way, (Vivien, Lejzerowicz & Pawlowski 2016) have shown that application of a CF to correct DNA reads proportions allows a more accurate estimation of oligochaete proportions, improving quality index calculation and quality assessment of watercourse sediments. Our results confirm that water quality index based on diatom metabarcoding and DNA read proportions are directly affected by gene copy number variation, and show the potential value of integrating CFs into molecular SPI calculation. However, as the biovolume–copy number relationship was based on only 8 diatom species and the efficiency of the resulting CFs validated on only one HTS dataset, further experiments including more species and larger datasets will be required to develop and fully validate CFs for use in molecular biomonitoring. The CF developed in the present study assumes that gene copy number is constant in each taxon. However, gene copy number may vary with the physiological status of the cell and stage of the life cycle, since in most diatoms cell volume decreases during the vegetative phase. The physiological status varies with cell cycle progression; additionally several factors may affect the physiological status of diatoms like changes in environmental conditions (e.g. nutrient availability, pollutants, temperature …) (Pandey et al. 2017). Altered physiological status of a given population is generally characterized by a higher proportion of damaged cells. The compromised/damaged cells are characterized by alteration of membrane integrity, degradation of the photosynthetic pigments or fragmentation of genomic DNA (Zetsche & Meysman 2012; Znachor et al. 2015). Variations of DNA integrity and chloroplast physiology between cells of a given population can impact directly the rbcL gene copy number per cell and thus DNA metabarcoding quantification. (Eberhard, Drapier & Wollman 2002) showed that chloroplast genome copy number is reduced when the green alga Chlamydomonas reinhardtii is cultivated under phototrophic conditions compared to cultivation in mixotrophic conditions. Limitation by mineral nutrients may also have an impact; for instance iron limitation can reduce the number of the chloroplast per cell (from 4 to 2) and their size in the marine diatom Thalassiosira oceanica (Hustedt) Hasle et Heimdal (Lommer et al. 2012). Variation of the cell physiological state was not taken into account in developing CFs for diatom metabarcoding. However, during our experiments we discriminated live and dead cells; we observed that their respective proportions did not affect significantly the correlation between the gene copy number per cell and the cell biovolume (Fig. S1). Further experiments should be performed to evaluate the impact on the final CFs of rbcL gene copy number variation linked to physiological status. The biovolume of each diatom species is required to apply the CF and hence correct the quantification in metabarcoding datasets. Several reference databases provide biovolume information for a lot of marine (e.g. Leblanc et al. 2012) and freshwater (e.g. Gosselain et al. 2005; Rimet et al. 2016) diatom species, but they do not generally account for biovolume variability, which is a complicating factor in diatoms because of the peculiarities of the life cycle. Diatom cell size within a population is not constant due to the method of vegetative reproduction, which leads to a progressive cell size reduction of the population (Crawford 1981), followed by restoration of cell size via a sexual event. For this reason, different cell sizes can be observed in the same diatom population, either in pure cultures of (e.g. in the marine diatom Thalassiosira weissflogii Grunow: Armbrust & Chisholm 1992) or in environmental populations (e.g. the freshwater species Sellaphora pupula (Kützing) Mereschk: (Mann, Chepurnov & Droop 1999). However, alhough the range of cell sizes within a given diatom population may vary by a factor of 2 to 5 in the environment (Hense & Beckmann 2015), natural populations usually have a rather narrow range of sizes and larger cells form a

negligible fraction of the population (Mann 2011). Furthermore, the distribution of cell size within environmental populations is often close to being normal (Mann, Chepurnov & Droop 1999; Spaulding et al. 2012). The balance between small and big individuals in the same population will therefore limit errors associated with the use of a mean biovolume. Hence, we propose to use the mean of biovolume to calculate CFs; without considering other potential HTS quantification biases, its application to DNA reads of environmental material should allow a good correction of their proportions.

Conclusions In the present study, we showed that the rbcL gene copy number variation is a major factor affecting diatom species quantification based on DNA read proportions, leading to the over-representation of high biovolume diatom species relative to cell counts. This generates discrepancies with the classical morphological approach, which is known to underestimate large species (Snoeijs, Busse & Potapova 2002). However, we demonstrated that a quantification CF can be applied, based on cell biovolume, to obtain DNA read proportions congruent with cell proportions. The application of this CF to environmental samples limits the over-estimation of large-celled species, such as in Eunotia and Ulnaria, in molecular inventories. Finally, we conclude that correcting the quantification of metabarcoding data allows quality index values to be obtained that are more comparable between morphological and molecular approaches, therefore providing continuity with classical biomonitoring and maintaining valuable ecological and historical knowledge already acquired about aquatic ecosystems. Development and standardization of water quality indices specifically adapted to diatom metabarcoding data will have to be addressed in further studies.

Acknowledgments The authors declare no conflict of interest. Funding provided by the French National Agency for Water and Aquatic Environments (ONEMA-AFB) and supported by the European COST action DNAqua-Net (CA 15219). We thank the Sequencing platform team (INRA-PGTB sequencing platform), particularly Franck Salin and Christophe Boury, who performed HTS sequencing. A special thanks to David G. Mann for the constructive discussions that helped to improve the manuscript.

Data accessibility All PGM raw sequence data are available for the 15 libraries, corresponding to the 5 DNA mock communities with 3 replicates, on the Zenodo repository website (http://doi.org/10.5281/zenodo.807178). Detailed description of the DNA reads composition of the 5 DNA mock communities after bioinformatics treatments is available in the Table S5 (Supporting information). We used the Mayotte HTS data of the 80 benthic diatom samples from rivers produced and archived by Vasselon et al. (2017b) on the Zenodo repository website (Vasselon et al. 2017b dataset, http://doi.org/10.5281/zenodo.400160).

Author contributions V.V., A.B., F.R., S.J., M.C., K.T., I.D contributed to the study designed. V.V., M.C and S.J. conducted the laboratory work. V.V. analyzed the data and wrote the manuscript. All the authors contributed to the discussions and to manuscript editing.

References Amend, A.S., Seifert, K.A. & Bruns, T.D. (2010). Quantifying microbial communities with 454 pyrosequencing: does read abundance count? Molecular Ecology, 19, 5555–5565. Angly, F.E., Dennis, P.G., Skarshewski, A., Vanwonterghem, I., Hugenholtz, P. & Tyson, G.W. (2014). CopyRighter: a rapid tool for improving the accuracy of microbial community profiles through lineage-specific gene copy number correction. Microbiome, 2, 11. Apothéloz-Perret-Gentil, L., Cordonier, A., Straub, F., Iseli, J., Esling, P. & Pawlowski, J. (2017). Taxonomy-free molecular diatom index for high-throughput eDNA biomonitoring. Molecular ecology resources, (in press). Armbrust, E.V. & Chisholm, S.W. (1992). Patterns of cell size change in a marine centric diatom: variability evolving from clonal isolates. Journal of Phycology, 28, 146–156. Baird, D.J. & Hajibabaei, M. (2012). Biomonitoring 2.0: A new paradigm in ecosystem assessment made possible by next-generation DNA sequencing. Molecular Ecology, 21, 2039–2044. Bendich, A.J. (1987). Why do chloroplasts and mitochondria contain so many copies of their genome? BioEssays, 6, 279–282. Chonova, T., Keck, F., Labanowski, J., Montuelle, B., Rimet, F. & Bouchez, A. (2016). Separate treatment of hospital and urban wastewaters: A real scale comparison of effluents and their effect on microbial communities. Science of The Total Environment, 542, 965–975. Crawford, R.M. (1981). The Siliceous Components of the Diatom Cell Wall and Their Morphological Variation. Silicon and Siliceous Structures in Biological Systems, pp. 129–156. Springer New York, New York, NY. von Dassow, P., Petersen, T.W., Chepurnov, V.A. & Virginia Armbrust, E. (2008). Inter- and Intraspecific relationships between nuclear DNA content and cell size in selected members members of the centric diatom genus Thalassiosira (Bacillariophyceae). Journal of Phycology, 44, 335–349. Deagle, B.E., Thomas, A.C., Shaffer, A.K., Trites, A.W. & Jarman, S.N. (2013). Quantifying sequence proportions in a DNA-based diet study using Ion Torrent amplicon sequencing: which counts count? Molecular Ecology Resources, 13, 620–633. Eberhard, S., Drapier, D. & Wollman, F.-A. (2002). Searching limiting steps in the expression of chloroplastencoded proteins: relations between gene copy number, transcription, transcript abundance and translation rate in the chloroplast of Chlamydomonas reinhardtii. The Plant Journal, 31, 149–160. Elbrecht, V. & Leese, F. (2015). Can DNA-Based Ecosystem Assessments Quantify Species Abundance? Testing Primer Bias and Biomass—Sequence Relationships with an Innovative Metabarcoding Protocol (M. Hajibabaei, Ed.). Plos One, 10, e0130324. Elbrecht, V., Peinert, B. & Leese, F. (2017). Sorting things out: Assessing effects of unequal specimen biomass on DNA metabarcoding. Ecology and Evolution, 1–16. Ersland, D.R., Aldrich, J. & Cattolico, R. a. (1981). Kinetic Complexity, Homogeneity, and Copy Number of Chloroplast DNA from the Marine Alga Olisthodiscus luteus. Plant Phisiology, 68, 1468–1473. European Committee for Standardization (CEN). (2006). Water quality - Guidance standard on the enumeration of phytoplankton using inverted microscopy (Utermöhl technique). European Standard, EN 15204, 1–42. European Council. (2000). Directive 2000/60/EC of the European Parliament and of the Council of 23 October 2000 Establishing a Framework for Community Action in the Field of Water Policy. Office for official publications of the European Communities, Brussels. Evans, N.T., Olds, B.P., Renshaw, M.A., Turner, C.R., Li, Y., Jerde, C.L., Mahon, A.R., Pfrender, M.E., Lamberti, G.A. & Lodge, D.M. (2016). Quantification of mesocosm fish and amphibian species diversity via environmental DNA metabarcoding. Molecular Ecology Resources, 16, 29–41. Godhe, A., Asplund, M.E., Härnström, K., Saravanan, V., Tyagi, A. & Karunasagar, I. (2008). Quantification of diatom and dinoflagellate biomasses in coastal marine seawater samples by real-time PCR. Applied and environmental microbiology, 74, 7174–82. Gosselain, V., Coste, M., Campeau, S., Ector, L., Fauville, C., Delmas, F., Knoflacher, M., Licursi, M., Rimet, F., Tison, J., Tudesque, L. & Descy, J.-P. (2005). A large-scale stream benthic diatom database. Hydrobiologia, 542, 151–163.

Gruber, A. (2008). Molecular Characterisation of Diatom Plastids (PhD thesis). University of Konstanz. Gutteridge, S. & Gatenby, A. (1995). Rubisco Synthesis, Assembly, Mechanism, and Regulation. The Plant Cell Online, 7, 809–819. Hajibabaei, M., Shokralla, S., Zhou, X., Singer, G. a C. & Baird, D.J. (2011). Environmental Barcoding: A NextGeneration Sequencing Approach for Biomonitoring Applications Using River Benthos (C.R. Voolstra, Ed.). PLoS ONE, 6, e17497. Hense, I. & Beckmann, A. (2015). A theoretical investigation of the diatom cell size reduction–restitution cycle. Ecological Modelling, 317, 66–82. Hering, D., Borja, A., Carstensen, J., Carvalho, L., Elliott, M., Feld, C.K., Heiskanen, A.-S., Johnson, R.K., Moe, J. & Pont, D. (2010). The European Water Framework Directive at the age of 10: A critical review of the achievements with recommendations for the future. Science of The Total Environment, 408, 4007–4019. Keck, F., Vasselon, V., Tapolczai, K., Rimet, F. & Bouchez, A. (2017). Freshwater biomonitoring in the Information Age. Frontiers in Ecology and the Environment, 15, 266–274. Kermarrec, L., Franc, A., Rimet, F., Chaumeil, P., Frigerio, J.-M., Humbert, J. & Bouchez, A. (2014). A nextgeneration sequencing approach to river biomonitoring using benthic diatoms. Freshwater Science, 33, 349–363. Kermarrec, L., Franc, A., Rimet, F., Chaumeil, P., Humbert, J.F. & Bouchez, A. (2013). Next-generation sequencing to inventory taxonomic diversity in eukaryotic communities: a test for freshwater diatoms. Molecular Ecology Resources, 13, 607–619. Koop, H.-U., Herz, S., Golds, T.J. & Nickelsen, J. (2007). The genetic transformation of plastids. Stress-Activated Protein Kinases, pp. 457–510. Leblanc, K., Arístegui, J., Armand, L., Assmy, P., Beker, B., Bode, A., Breton, E., Cornet, V., Gibson, J., Gosselin, M.P., Kopczynska, E., Marshall, H., Peloquin, J., Piontkovski, S., Poulton, a. J., Quéguiner, B., Schiebel, R., Shipe, R., Stefels, J., van Leeuwe, M. a., Varela, M., Widdicombe, C. & Yallop, M. (2012). A global diatom database – abundance, biovolume and biomass in the world ocean. Earth System Science Data Discussions, 5, 147–185. Lecointe, C., Coste, M. & Prygiel, J. (1993). ‘Omnidia’: software for taxonomy, calculation of diatom indices and inventories management. Hydrobiologia, 269–270, 509–513. Lommer, M., Specht, M., Roy, A.-S., Kraemer, L., Andreson, R., Gutowska, M.A., Wolf, J., Bergner, S. V, Schilhabel, M.B., Klostermeier, U.C., Beiko, R.G., Rosenstiel, P., Hippler, M. & LaRoche, J. (2012). Genome and low-iron response of an oceanic diatom adapted to chronic iron limitation. Genome Biology, 13, R66. Mann, D.G. (2011). Size and Sex. The Diatom World (ed E. J Seckbach & JP Kociolek), pp. 145–166. Springer, Dordrecht. Mann, D.G., Chepurnov, V.A. & Droop, S.J.M. (1999). Sexuality, incompatibility, size variation, and preferential polyandry in natural populations and clones of Sellaphora pupula (Bacillariophyceae). Journal of Phycology, 35, 152–170. Mann, D.G., Sato, S., Trobajo, R., Vanormelingen, P. & Souffreau, C. (2010). DNA barcoding for species identification and discovery in diatoms. Cryptogamie, 31, 557–577. Okie, J.G., Smith, V.H. & Martin-Cereceda, M. (2016). Major evolutionary transitions of life, metabolic scaling and the number and size of mitochondria and chloroplasts. Proceedings of the Royal Society B: Biological Sciences, 283, 20160611. Pandey, L.K., Bergey, E.A., Lyu, J., Park, J., Choi, S., Lee, H., Depuydt, S., Oh, Y.T., Lee, S.M. & Han, T. (2017). The use of diatoms in ecotoxicology and bioassessment: Insights, advances and challenges. Water Research, 118, 39–58. Pawlowski, J., Lejzerowicz, F., Apotheloz-Perret-Gentil, L., Visco, J. & Esling, P. (2016). Protist metabarcoding and environmental biomonitoring: Time for change. European Journal of Protistology, 55, 12–25. R Development core team. (2013). R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Rauwolf, U., Golczyk, H., Greiner, S. & Herrmann, R.G. (2010). Variable amounts of DNA related to the size of chloroplasts III. Biochemical determinations of DNA amounts per organelle. Molecular Genetics and

Genomics, 283, 35–47. Rimet, F. (2012). Recent views on river pollution and diatoms. Hydrobiologia, 683, 1–24. Rimet, F., Abarca, N., Bouchez, A., Kusber, W.H., Jahn, R., Kahlert, M., Keck, F., Kelly, M., Mann, D.G., Piuz, A., Trobajo, R., Tapolczai, K., Vasselon, V. & Zimmermann, J. (2018). The potential of high throughput sequencing (HTS) of natural samples as a source of primary taxonomic information for reference libraries of diatom barcodes. Fottea, (in press). Rimet, F., Chaumeil, P., Keck, F., Kermarrec, L., Vasselon, V., Kahlert, M., Franc, A. & Bouchez, A. (2016). RSyst::diatom: an open-access and curated barcode database for diatoms and freshwater monitoring. Database, 2016, baw016. Rimet, F., Trobajo, R., Mann, D.G., Kermarrec, L., Franc, A., Domaizon, I. & Bouchez, A. (2014). When is Sampling Complete? The Effects of Geographical Range and Marker Choice on Perceived Diversity in Nitzschia palea (Bacillariophyta). Protist, 165, 245–259. Round, F.E., Crawford, R.M. & Mann, D.G. (1990). Diatoms: Biology and Morphology of the Genera (Cambridge University Press, Ed.). Sabir, J.S.M., Yu, M., Ashworth, M.P., Baeshen, N.A., Baeshen, M.N., Bahieldin, A., Theriot, E.C. & Jansen, R.K. (2014). Conserved gene order and expanded inverted repeats characterize plastid genomes of Thalassiosirales. PloS one, 9, e107854. Schloss, P.D., Westcott, S.L., Ryabin, T., Hall, J.R., Hartmann, M., Hollister, E.B., Lesniewski, R. a., Oakley, B.B., Parks, D.H., Robinson, C.J., Sahl, J.W., Stres, B., Thallinger, G.G., Van Horn, D.J. & Weber, C.F. (2009). Introducing mothur: Open-source, platform-independent, community-supported software for describing and comparing microbial communities. Applied and Environmental Microbiology, 75, 7537–7541. Shokralla, S., Spall, J.L., Gibson, J.F. & Hajibabaei, M. (2012). Next-generation sequencing technologies for environmental DNA research. Molecular Ecology, 21, 1794–1805. Snoeijs, P., Busse, S. & Potapova, M. (2002). The importance of diatom cell size in community analysis. Journal of Phycology, 38, 265–281. Spaulding, S. a., Jewson, D.H., Bixby, R.J., Nelson, H. & McKnight, D.M. (2012). Automated measurement of diatom size. Limnology and Oceanography: Methods, 10, 882–890. Stein, E.D., Martinez, M.C., Stiles, S., Miller, P.E. & Zakharov, E. V. (2014). Is DNA barcoding actually cheaper and faster than traditional morphological methods: results from a survey of freshwater bioassessment efforts in the United States? (M. Casiraghi, Ed.). PloS one, 9, e95525. Sun, J. & Liu, D. (2003). Geometric models for calculating cell biovolume and surface area for phytoplankton. Journal of Plankton Research, 25, 1331–1346. Taberlet, P., Coissac, E., Hajibabaei, M. & Rieseberg, L.H. (2012). Environmental DNA. Molecular Ecology, 21, 1789–1793. Tan, B., Ng, C., Nshimyimana, J.P., Loh, L.L., Gin, K.Y.-H. & Thompson, J.R. (2015). Next-generation sequencing (NGS) for assessment of microbial water quality: current progress, challenges, and future opportunities. Frontiers in microbiology, 6, 1027. Thomas, A.C., Deagle, B.E., Eveson, J.P., Harsch, C.H. & Trites, A.W. (2016). Quantitative DNA metabarcoding: improved estimates of species proportional biomass using correction factors derived from control material. Molecular Ecology Resources, 16, 714–726. Treusch, A.H., Demir-Hilton, E., Vergin, K.L., Worden, A.Z., Carlson, C.A., Donatz, M.G., Burton, R.M. & Giovannoni, S.J. (2012). Phytoplankton distribution patterns in the northwestern Sargasso Sea revealed by small subunit rRNA genes from plastids. The ISME Journal, 6, 481–492. Vasselon, V., Domaizon, I., Rimet, F., Kahlert, M. & Bouchez, A. (2017a). Application of high-throughput sequencing (HTS) metabarcoding to diatom biomonitoring: Do DNA extraction methods matter? Freshwater Science, 36, 162–177. Vasselon, V., Rimet, F., Tapolczai, K. & Bouchez, A. (2017b). Assessing ecological status with diatoms DNA metabarcoding: Scaling-up on a WFD monitoring network (Mayotte island, France). Ecological Indicators, 82, 1–12. Visco, J.A., Apothéloz-Perret-Gentil, L., Cordonier, A., Esling, P., Pillet, L. & Pawlowski, J. (2015). Environmental

Monitoring: Inferring the Diatom Index from Next-Generation Sequencing Data. Environmental Science & Technology, 49, 7597–7605. Vivien, R., Lejzerowicz, F. & Pawlowski, J. (2016). Next-generation sequencing of aquatic oligochaetes: Comparison of experimental communities. PLoS ONE, 11, 1–14. Weber, A. a-T. & Pawlowski, J. (2013). Can abundance of protists be inferred from sequence data: a case study of foraminifera. PloS one, 8, e56739. Zetsche, E.-M. & Meysman, F.J.R. (2012). Dead or alive? Viability assessment of micro- and mesoplankton. Journal of Plankton Research, 34, 493–509. Zimmermann, J., Glöckner, G., Jahn, R., Enke, N. & Gemeinholzer, B. (2015). Metabarcoding vs. morphological identification to assess diatom diversity in environmental studies. Molecular Ecology Resources, 15, 526– 542. Znachor, P., Rychtecký, P., Nedoma, J. & Visocká, V. (2015). Factors affecting growth and viability of natural diatom populations in the meso-eutrophic Římov Reservoir (Czech Republic). Hydrobiologia, 762, 253–265.

Tables Table 1 Characteristics of the 8 diatom species selected in the Thonon Culture Collection (TCC) and used in this study. Table 2 – CF calculated for the 8 diatom species using their respective cell biovolume (Table 1) and the linear equation between the rbcL copy number and the cell biovolume (Fig. 3).

Figures Figure 1 Experimental design applied to the 8 diatom species. After the inoculation of 21 flasks containing 40mL of DV media, diatom culture growth was followed at 7 sampling time (from T0 to T6) and analysis was performed in triplicate (3 flasks per sampling time). Figure 2 Estimation of the rbcL copy number per diatom cell for the 8 diatom species. Mean values calculated using the gene and the diatom cell concentrations obtained respectively by qPCR and inverted microscopy at T0, T1 and T2 sampling points (n = 9). Figure 3 Correlation between the diatom cell biovolume and the rbcL gene copy number per cell after log(x+1) transformation. Figure 4 Relative abundances of the 8 diatom species in the 5 DNA mock communities based (A) on mean of HTS DNA reads without (left) and with (right) correcting quantification using the biovolume correction factor and (B) on mean of morphological counts from inverted microscopy. Figure 5 Dominant taxa (relative abundance > 0.5 %) obtained in HTS Mayotte molecular inventories without (left) and with (right) application of the biovolume correction factor. All samples (n=80) are considered. Figure 6 Distribution of the differences between the molecular and the morphological SPI (SPI) for all Mayotte samples using original molecular SPI values (left) and new molecular SPI values based on molecular inventories corrected with the biovolume CF (right).

Supporting Information Table S1 rbcL primers, qPCR reactions mix and condition used for the qPCR assays. Information is provided for 1 reaction in a final volume of 25µL. Table S2 Estimation of the diatom cell concentration and the live/dead cell proportion per mL of media, based on microscopy counts, for the 8 diatom species at each sampling time and for the 3 replicates (A, B, C). Mean values of cell concentration per mL of media, which only take into account living cells, is provided and used for the calculation of rbcL copy number per diatom cell (bold values). Table S3 Estimation of the rbcL copy number per mL of media determined by qPCR for the 8 diatom species at each sampling time and for the 3 replicates (A, B, C). Mean values of rbcL concentration per mL of media is provided and used for the calculation of rbcL copy number per diatom cell (bold values). Table S4 CF calculated for the 84 diatom taxa detected in Mayotte environmental samples. Calculation performed using the respective cell biovolume of each taxa (available in the

Rsyst::diatom library) and the linear equation between the rbcL copy number and the cell biovolume produced in the Fig. 3. Table S5 Number of DNA reads assigned to the 8 species in each of the 5 DNA mock communities. A, B, and C represent the 3 replicates. Figure S1 Correlation between the diatom cell biovolume and the rbcL gene copy number per cell after log(x+1) transformation based on live (black) or live/dead (grey) microscopical counts. Linear equation of the model and the Pearson correlation coefficient (r) with is associated p-value are indicated.

Species Achnanthidium minutissimum (Kützing) Czarnecki Nitzschia palea (Kützing) W.Smith Ulnaria ulna (Nitzsch) Compère Pinnularia viridiformis (Nitzsch) Ehrenberg Diatoma tenuis Kützing Nitzschia inconspicua Grunow Fragilaria perminuta (Grunow) Lange-Bertalot Cyclotella meneghiniana Kützing

Chloroplast Length Width TCC code nb./cell (µm) (µm) TCC667 1 7.1 3.2 TCC139-1 2 22.7 4,0 TCC670 2 54.6 7.9 TCC890 2 51.4 14.3 TCC861 ≈8 42.4 4.8 TCC488 2 8.1 4.3 TCC753 2 11.1 4.2 TCC690 ≈ 20 12.1

Thickness Biovolume (µm) 2.5 4,0 9.5 17.8 4.8 3.6 3.7 4.7

Table 1 – Characteristics of the 8 diatom species selected in the Thonon Culture Collection (TCC) and used in this study.

Species A. minutissimum N. inconspicua N. palea P. viridiformis D. tenuis F. perminuta U. ulna C. meneghiniana

Calculated CF 0.6 1.7 3.3 78.5 11.1 2.4 39.6 8.3

Table 2 – CF calculated for the 8 diatom species using their respective cell biovolume (Table 1) and the linear equation between the rbcL copy number and the cell biovolume (Fig. 3).

(µm3) 45 183 4087 10282 769 98 135 539

Figure 1 – Experimental design applied to the 8 diatom species. After the inoculation of 21 flasks containing 40mL of DV media, diatom culture growth was followed at 7 sampling time (from T0 to T6) and analysis was performed in triplicate (3 flasks per sampling time).

Figure 2 – Estimation of the rbcL copy number per diatom cell for the 8 diatom species. Mean values calculated using the gene and the diatom cell concentrations obtained respectively by qPCR and inverted microscopy at T0, T1 and T2 sampling points (n = 9).

log (rbcL copy + 1)

3 y = 0,73x - 1,02 r = 0.97 p < 0.001

P. viridiformis

2

U. ulna C.meneghiniana D. tenuis

1

F. perminuta N. palea N. inconspicua A. minutissimum

0

0

1

2

3

4

5

log (cell biovolume + 1) Figure 3 – Correlation between the diatom cell biovolume and the rbcL gene copy number per cell after log(x+1) transformation.

Figure 4 – Relative abundances of the 8 diatom species in the 5 DNA mock communities based (A) on mean of HTS DNA reads without (left) and with (right) correcting quantification using the biovolume correction factor and (B) on mean of morphological counts from inverted microscopy.

Figure 5 – Dominant taxa (relative abundance > 0.5 %) obtained in HTS Mayotte molecular inventories without (left) and with (right) application of the biovolume correction factor. All samples (n=80) are considered.

Figure 6 – Distribution of the differences between the molecular and the morphological SPI (SPI) for all Mayotte samples using original molecular SPI values (left) and new molecular SPI values based on molecular inventories corrected with the biovolume CF (right).

Table S1 – rbcL primers, qPCR reactions mix and condition used for the qPCR assays. Information is provided for 1 reaction in a final volume of 25µL. Primer name Forward

Reverse

Diat_rbcL_708F_1

Primer sequence (5' - 3') AGGTGAAGTAAAAGGTTCWTACTTAAA

Length (bp) 27

Diat_rbcL_708F_2 Diat_rbcL_708F_3 R3_1 R3_2

AGGTGAAGTTAAAGGTTCWTAYTTAAA AGGTGAAACTAAAGGTTCWTACTTAAA CCTTCTAATTTACCWACWACTG CCTTCTAATTTACCWACAACAG

27 27 22 22

Reagents Sybr MIX

Initial conc. 2X

Final conc. 1X

Volume (µL) 12.5

H2O molecular grade Forward (Diat_rbcL_708F_1 + _2 + _3) Reverse (R3_1 + R3_2) Bovine Serum Albumin (BSA) DNA

10 µM 10 µM 10 mg/mL 25 ng/µL

0.5 µM 0.5 µM 0.5 mg/mL 2 ng/µL

6.75 1.25 1.25 1.25 2

Step 1

Time (s) 900

Temperature (°C) 95

2 3 4 5

45 45 45 1° every 5s

95 55 72 60 to 95

Cycles

X 40

Table S2 – Estimation of the diatom cell concentration and the live/dead cell proportion per mL of media, based on microscopy counts, for the 8 diatom species at each sampling time and for the 3 replicates (A, B, C). Mean values of cell concentration per mL of media, which only take into account living cells, is provided and used for the calculation of rbcL copy number per diatom cell (bold values). Species Cmen

Npal

Uuln

Ninc

Dten

Pvir

Fper

Sampling time T0 T1 T2 T3 T4 T5 T6 T0 T1 T2 T3 T4 T5 T6 T0 T1 T2 T3 T4 T5 T0 T1 T2 T3 T4 T5 T6 T0 T1 T2 T3 T4 T5 T0 T1 T2 T3 T4 T5 T0 T1 T2

Days after inoculation 5 10 13 20 25 31 38 5 10 13 17 25 34 40 5 10 13 20 31 38 5 10 12 17 25 34 40 12 17 20 25 34 38 13 20 31 34 40 73 12 17 20

[cell.mL-1] per replicate A B C 3.7E+02 4.1E+02 4.2E+02 8.2E+03 6.0E+03 8.5E+03 1.2E+04 1.1E+04 2.0E+04 1.2E+05 5.7E+04 1.3E+05 2.6E+05 4.1E+05 3.2E+05 2.0E+05 2.4E+05 2.3E+05 4.6E+05 5.5E+05 2.6E+05 1.8E+04 2.1E+04 3.9E+04 6.1E+05 4.9E+05 4.1E+05 4.6E+05 4.8E+05 5.2E+05 4.8E+05 3.9E+05 6.2E+05 4.1E+05 4.3E+05 9.4E+05 6.2E+05 7.2E+05 7.0E+05 1.3E+06 1.1E+06 6.4E+05 8.2E+03 7.9E+03 1.5E+04 1.3E+04 1.2E+04 1.5E+04 1.2E+04 3.4E+04 7.9E+03 1.8E+04 1.6E+04 2.6E+04 1.6E+04 1.1E+04 9.2E+03 8.6E+03 9.2E+03 3.6E+04 3.9E+03 8.7E+03 5.8E+03 3.5E+05 3.9E+05 4.3E+05 4.3E+05 2.6E+05 1.1E+06 4.1E+05 6.4E+05 1.1E+06 1.7E+06 1.4E+06 9.9E+05 1.6E+06 1.3E+06 1.4E+06 1.3E+06 1.9E+06 1.7E+06 1.2E+04 4.3E+04 2.6E+04 1.1E+05 9.4E+04 1.0E+05 1.8E+05 2.2E+05 1.3E+05 4.9E+05 2.3E+05 1.4E+05 2.7E+05 2.0E+05 2.1E+05 4.1E+05 2.4E+05 1.6E+05 6.0E+02 4.7E+02 4.1E+02 9.6E+02 7.2E+02 1.1E+03 1.5E+03 1.7E+03 3.1E+03 2.0E+03 2.0E+03 2.4E+03 2.7E+03 2.2E+03 3.6E+03 4.9E+03 2.7E+03 2.6E+03 6.0E+04 3.4E+04 3.3E+04 2.7E+05 1.1E+05 1.7E+05 2.2E+05 1.6E+05 1.2E+05

% of dead cell A B C 9.8 3.8 4.0 6.1 14.5 11.9 13.5 16.1 12.0 20.2 13.9 19.6 53.9 51.2 56.3 59.9 50.4 48.2 59.1 55.1 57.1 0.0 0.0 1.0 0.0 0.0 0.0 0.9 1.9 1.0 5.9 4.0 6.7 15.4 9.9 8.1 23.5 30.3 25.0 46.6 38.5 54.1 3.8 2.8 0.7 5.4 7.8 7.2 14.3 13.6 10.5 27.1 23.8 23.9 83.3 74.4 63.9 82.8 84.8 82.2 0.5 1.0 0.0 0.0 0.2 0.2 0.6 0.2 0.7 6.9 7.0 5.1 11.6 10.4 12.6 7.2 9.9 6.7 21.4 28.9 30.8 0.2 0.0 0.3 5.7 7.0 5.5 6.9 5.7 5.5 6.3 8.8 8.2 26.5 35.8 43.1 48.3 49.3 45.5 8.0 7.5 11.7 12.0 9.3 7.3 14.3 17.3 18.5 16.5 23.5 29.6 26.1 22.6 26.7 83.7 75.8 66.8 0.7 0.7 1.3 14.6 12.6 11.6 23.4 24.1 19.7

Mean (living cells) (cell.mL-1) 3.8E+02 6.8E+03 1.2E+04 8.1E+04 1.5E+05 1.0E+05 1.8E+05 2.6E+04 5.1E+05 4.8E+05 4.7E+05 5.3E+05 5.0E+05 5.6E+05 1.0E+04 1.2E+04 1.5E+04 1.5E+04 3.0E+03 3.1E+03 6.1E+03 3.9E+05 5.9E+05 6.8E+05 1.2E+06 1.3E+06 1.2E+06 2.7E+04 9.6E+04 1.7E+05 2.7E+05 1.5E+05 1.4E+05 4.5E+02 8.3E+02 1.8E+03 1.6E+03 2.1E+03 7.7E+02 4.2E+04 1.6E+05 1.3E+05

Amin

T3 T4 T5 T6 T0 T1 T2 T3 T4 T5 T6

25 31 34 40 12 17 25 31 34 38 40

1.5E+05 6.6E+05 1.2E+06 2.9E+05 1.8E+03 3.0E+04 5.5E+05 1.3E+06 2.1E+06 2.7E+06 2.8E+06

1.8E+05 3.0E+06 3.2E+05 5.8E+05 6.2E+03 7.4E+04 4.0E+05 1.0E+06 2.9E+06 1.2E+06 2.7E+06

1.6E+05 4.4E+05 2.6E+05 5.4E+05 3.7E+03 8.4E+04 1.4E+06 5.2E+05 6.7E+05 5.6E+05 1.7E+06

62.2 69.3 78.5 82.5 0.7 4.1 4.7 13.1 11.6 15.2 16.2

65.4 73.8 74.8 75.8 1.7 3.7 7.7 13.1 10.5 11.4 11.5

62.3 65.5 76.8 73.5 1.0 3.0 4.6 10.2 13.8 16.9 17.5

6.0E+04 3.8E+05 1.3E+05 1.1E+05 3.9E+03 6.0E+04 7.5E+05 8.3E+05 1.7E+06 1.3E+06 2.0E+06

Table S3 – Estimation of the rbcL copy number per mL of media determined by qPCR for the 8 diatom species at each sampling time and for the 3 replicates (A, B, C). Mean values of rbcL concentration per mL of media is provided and used for the calculation of rbcL copy number per diatom cell (bold values). Species Sampling time Cmen T0 T1 T2 T3 T4 T5 T6 Npal T0 T1 T2 T3 T4 T5 T6 Uuln T0 T1 T2 T3 T4 T5 Ninc T0 T1 T2 T3 T4 T5 T6 Dten T0 T1 T2 T3 T4 T5 Pvir T0 T1 T2 T3 T4 T5 Fper T0 T1 T2

Days after inoculation 5 10 13 20 25 31 38 5 10 13 17 25 34 40 5 10 13 20 31 38 5 10 12 17 25 34 40 12 17 20 25 34 38 13 20 31 34 40 73 12 17 20

[rbcL] (copy.mL-1) A B 7.1E+03 7.3E+03 4.8E+04 2.2E+04 4.6E+04 2.6E+04 1.2E+05 1.1E+05 6.1E+05 6.2E+05 4.3E+05 2.3E+06 9.4E+05 1.0E+06 3.8E+04 3.4E+04 1.3E+06 1.5E+06 2.5E+06 2.8E+06 2.5E+06 2.7E+06 3.3E+06 2.3E+06 1.7E+06 1.9E+06 1.1E+06 1.3E+06 1.4E+05 2.5E+05 4.0E+05 3.3E+05 1.2E+05 1.1E+05 4.9E+05 1.8E+05 1.2E+05 1.4E+05 7.5E+04 5.6E+04 1.1E+04 1.3E+04 3.5E+05 7.2E+05 8.1E+05 6.3E+05 9.9E+06 8.6E+06 4.7E+06 5.1E+06 7.3E+06 7.8E+06 4.8E+06 4.5E+06 4.7E+05 2.1E+05 1.5E+06 2.0E+06 7.6E+05 1.4E+06 1.3E+06 5.0E+05 4.3E+05 2.3E+05 3.2E+05 4.5E+05 7.9E+04 4.6E+04 1.2E+05 1.2E+05 2.0E+05 1.3E+05 1.6E+05 2.6E+05 2.6E+05 2.2E+05 3.1E+05 5.5E+05 1.4E+04 3.4E+03 3.0E+05 2.4E+05 8.3E+05 7.1E+05

C 1.1E+04 1.3E+04 2.4E+04 1.9E+05 7.6E+05 5.3E+05 7.3E+05 7.0E+04 9.1E+05 2.7E+06 2.6E+06 3.0E+06 2.2E+06 8.4E+05 1.6E+05 3.1E+05 7.5E+04 2.6E+05 2.2E+05 5.7E+04 1.5E+04 5.2E+05 1.1E+06 7.5E+06 6.3E+06 8.1E+06 3.2E+06 3.0E+05 1.1E+06 2.8E+06 4.6E+05 5.3E+05 2.3E+05 7.1E+04 1.2E+05 2.0E+05 3.0E+05 3.8E+05 4.8E+05 9.4E+03 3.6E+05 9.2E+05

Mean (copy.mL-1) 8.4E+03 2.8E+04 3.2E+04 1.4E+05 6.6E+05 4.8E+05 9.1E+05 4.7E+04 1.3E+06 2.6E+06 2.6E+06 2.9E+06 2.0E+06 1.1E+06 1.8E+05 3.5E+05 1.0E+05 3.1E+05 1.6E+05 6.3E+04 1.3E+04 5.3E+05 8.3E+05 8.7E+06 5.4E+06 7.7E+06 4.2E+06 3.3E+05 1.6E+06 1.7E+06 7.5E+05 4.0E+05 3.4E+05 6.6E+04 1.2E+05 1.8E+05 2.4E+05 2.9E+05 4.5E+05 9.0E+03 3.0E+05 8.2E+05

Amin

T3 T4 T5 T6 T0 T1 T2 T3 T4 T5 T6

25 31 34 40 12 17 25 31 34 38 40

8.1E+05 4.4E+05 6.7E+05 4.0E+05 1.8E+03 3.4E+04 1.9E+05 1.2E+05 2.6E+05 2.8E+05 5.1E+05

4.8E+05 4.8E+05 6.8E+05 4.7E+05 2.8E+03 1.5E+04 1.7E+05 1.6E+05 3.6E+05 3.2E+05 3.6E+05

1.4E+06 4.4E+05 1.0E+06 4.5E+05 3.6E+03 2.7E+04 2.2E+05 1.6E+05 3.1E+05 2.6E+05 2.0E+05

8.8E+05 4.5E+05 8.0E+05 4.4E+05 2.7E+03 2.6E+04 1.9E+05 1.5E+05 3.1E+05 2.9E+05 3.6E+05

Table S4 – CF calculated for the 84 diatom taxa detected in Mayotte environmental samples. Calculation performed using the respective cell biovolume of each taxa (available in the Rsyst::diatom library) and the linear equation between the rbcL copy number and the cell biovolume produced in the Fig. 3. Diatom taxa Achnanthes_coarctata Achnanthidium_helveticum Achnanthidium_minutissimum Achnanthidium_sp. Amphora_pediculus Amphora_sp. Caloneis_silicula Caloneis_sp. Cocconeis_placentula Craticula_cuspidata Craticula_molestiformis Cyclotella_sp. Cymbella_excisa Cymbella_heterogibbosa Cymbella_sp. Cymbopleura_naviculiformis Encyonema_minutum Encyonema_muelleri Encyonema_silesiacum Encyonema_sp. Eolimna_subminuscula Epithemia_sp. Eunotia_bilunaris Eunotia_minor Eunotia_pectinalis Eunotia_sp. Fallacia_pygmaea Fistulifera_saprophila Fragilaria_sp. Frustulia_vulgaris Frustulia_sp. Gomphonema_acuminatum Gomphonema_affine Gomphonema_bourbonense Gomphonema_clevei Gomphonema_parvulum Gomphonema_sp. Halamphora_montana Halamphora_sp. Hydrosera_sp. Lemnicola_hungarica Luticola_sparsipunctata Mayamaea_permitis Navicula_cryptocephala Navicula_cryptotenella Navicula_lanceolata Navicula_radiosa Navicula_rostellata Navicula_sp.

Biovolume (µm3) 53 316 76 76 72 20096 1994 523 2963 2850 119 328 520 5817 520 1148 213 12784 821 213 112 5967 617 755 4219 15700 1229 14 294 1625 1625 1860 926 270 484 331 510 161 161 500 436 176 66 431 386 1227 1852 854 88

Calculated CF 0.7 5.3 1.3 1.3 1.2 128.3 23.1 8.1 31.2 30.3 2.1 5.5 8.1 51.5 8.1 15.1 3.8 92.1 11.7 3.8 2.0 52.5 9.3 10.9 40.6 107.1 16.0 0.03 5.0 19.8 19.8 21.9 12.8 4.6 7.6 5.5 8.0 2.9 2.9 7.8 7.0 3.1 1.0 6.9 6.3 15.9 21.9 12.0 1.5

Navicula_symmetrica Navicula_tripunctata Navicula_veneta Neidium_sp. Nitzschia_amphibia Nitzschia_filiformis Nitzschia_fonticola Nitzschia_inconspicua Nitzschia_lorenziana Nitzschia_palea Nitzschia_sp. Nitzschia_tubicola Pinnularia_divergens Pinnularia_subanglica Pinnularia_subgibba Pinnularia_sp. Placoneis_clementis Placoneis_elginensis Planothidium_sp. Rhopalodia_gibba Rhopalodia_sp. Sellaphora_minima Sellaphora_pupula Sellaphora_seminulum Sellaphora_sp. Seminavis_robusta Staurosira_elliptica Staurosira_sp. Stephanodiscus_hantzschii Surirella_sp. Tabellaria_flocculosa Terpsinoe_musica Tryblionella_sp. Ulnaria_ulna Ulnaria_sp.

818 966 279 240 334 737 344 89 1362 391 307 336 3908 1188 3454 1258 1123 1266 267 185472 185472 88 1183 69 88 5308 29 315 670 1034 500 10563 655 4724 5260

11.6 13.2 4.8 4.2 5.6 10.7 5.7 1.5 17.3 6.4 5.2 5.6 38.3 15.6 35.0 16.3 14.9 16.3 4.6 649.8 649.8 1.5 15.5 1.1 1.5 48.1 0.1 5.3 9.9 14.0 7.8 80.0 9.7 44.1 47.8

Table S5 – Number of DNA reads assigned to the 8 species in each of the 5 DNA mock communities. A, B, and C represent the 3 replicates. Mock 1 Species

A

B

B

A

B

C

A. minutissimum 2828 1934 2410 1785 2129 2109 1837 1900 1882 2025

1342

1683

1202

1273

1332

N. inconspicua

5480 3484 4648 3673 4533 4083 3777 3622 3824 3920

3074

3741

2462

2588

2571

N. palea

1452 1059 1126 912

715

888

695

567

634

P. viridiformis

2573 1966 2066 2372 2823 2999 6440 7861 7461 11586 10430 11722 18424 16703 14159

D. tenuis

5311 3423 4552 3286 4461 3172 4578 3377 3376 4013

2679

3442

2206

2861

2522

F. perminuta

5817 3796 4452 3484 3844 3549 3492 3569 3341 3427

2449

3083

2117

2318

2226

U.ulna

4486 3037 3863 3303 3893 3561 3259 3343 3449 3321

2412

2897

2395

2053

1992

807

1126

994

869

779

1037 718

B

896

C

904

A

Mock 5 A

850

C

Mock 4 C

984

A

Mock 3

B

C. meneghiniana 1360 844

C

Mock 2

899

1202 1344 1204 1129 1113 1235 1137

Figure S1 – Correlation between the diatom cell biovolume and the rbcL gene copy number per cell after log(x+1) transformation based on live (black) or live/dead (grey) microscopical counts. Linear equation of the model and the Pearson correlation coefficient (r) with is associated p-value are indicated.

log (rbcL copy + 1)

3

Live

Live + dead y = 0,72x - 0,99 r = 0.94 p < 0.001

y = 0,73x - 1,02 r = 0.97 p < 0.001

P. viridiformis

2 D. tenuis 1

U. ulna

C.meneghiniana F. perminuta N. inconspicua

N. palea

A. minutissimum

0

0

1

2

3

log (cell biovolume + 1)

4

5