Numerical taxonomy contributes to delimitation of

Phytotaxa 367 (2): 101–119 http://www.mapress.com/j/pt/ Copyright © 2018 Magnolia Press

ISSN 1179-3155 (print edition)

Article

PHYTOTAXA

ISSN 1179-3163 (online edition)

https://doi.org/10.11646/phytotaxa.367.2.1

Numerical taxonomy contributes to delimitation of Iranian and Turkish Hesperis L. (Brassicaceae) species ATENA ESLAMI FAROUJI1, HAMED KHODAYARI1*, MOSTAFA ASSADI2, BARIŞ ÖZÜDOĞRU3, ÖZLEM ÇETIN4, KLAUS MUMMENHOFF5 & SAMIK BHATTACHARYA5 Department of Biology, Faculty of Science, Lorestan University, 5 Km Khorramabad toward Tehran, Khorramabad, Iran; e-mail: [email protected], [email protected] 2 Department of Botany, Research Institute of Forests and Rangelands, Agricultural Research Education and Extension Organization (AREEO), P. O. Box 13185-116, Tehran, Iran; e-mail: [email protected] 3 Department of Biology, Faculty of Science, Hacettepe University, Beytepe, Ankara 06800, Turkey; e-mail: [email protected] 4 Department of Biotechnology, Selçuk University, Turkey; e-mail: [email protected] 5 Department of Biology/Botany, Osnabrück University, D-49076 Osnabrück, Germany; e-mail: [email protected], [email protected] *Corresponding author: email: [email protected] 1

Abstract Taxonomic descriptions of Iranian and Turkish Hesperis (Brassicaceae) species are generally insufficient and partly incomplete, which makes the species delimitation ambiguous. In order to clarify species circumscription, we scored 57 morphological descriptors (MDs) in 121 operational taxonomic units (OTUs) of Hesperis from Iran and Turkey and performed a multivariate analysis. The dendrogram was created from Gower’s distance matrix using Unweighted Pair Group Method with arithmetic mean (UPGMA) algorithm. The dendrogram clearly separates the 121 OTUs of Hesperis into five main phenons, which significantly deviate from the classical taxonomic treatment (sectional assignments) of the genus. Similar distinct delineation among the five phenons was revealed by a Principal Coordinate Analysis (PCoA), highlighting the resolving power of the multivariate analyses of quantitative and qualitative morphological characters. While there were significant variations among the OTUs for 57 MDs, the most distinctive morphological descriptors delimiting the phenons were estimated to be fruit, petal, stem, and leaf by a de-trended correspondence analysis (DCA). We also present a comparative discussion between the classical taxonomy and the delimitation of taxa revealed in our study. Keywords: Brassicaceae, Hesperis, morphological descriptor, phenon, variation

Introduction Hesperis Linnaeus (1753: 663) (Brassicaceae, tribe Hesperideae) is distributed naturally in southern, eastern and central Europe, Southwest Asia, Caucasus and Transcaucasia, and to a lesser extent, in northern and central Asia (Duran 2008, Aras et al. 2009). In a biogeographic context, the genus occurs in the Irano-Turanian, Mediterranean, and Euro-Siberian geographical regions (Duran 2008) and shows distinct floral diversity (Fig. 1). Hesperis can be readily distinguished from the rest of the Brassicaceae by having stalked glands with uniseriate stalks terminated with a unicellular gland (Al-Shehbaz et al. 2006). However, there is much need of a thorough systematic study due to the intra-genus ambiguity about the number of sections and sub-genera described in literature from 1821 to recent times (e.g., Andrzejowski in Candolle 1821, Boissier 1867, Cullen 1965, Dvořák 1968, Duran et al. 2003, Duran in Sanad & Dašić 2016). As for example, although Hesperis and Tchihatchewia Boissier (Tchichatscheff 1860: 292) represent tribe Hesperideae (Al-Shehbaz 2012) in a phylogenetic context and placed in lineage III in the family-wide analyses (Couvreur et al. 2010, Huang et al. 2016), recent molecular phylogenetic study from our lab (unpublished data) prompted Tchihatchewia to be regarded as a synonym of Hesperis (German & Al-Shehbaz 2018). This highlight the ambiguity is a result of narrow taxonomic evaluation of the fruit characters, which traditionally separated them as two genera and are demonstrated to be homoplastic in various Brassicaceae genera, therefore, should not be used to define lineages (Franzke et al. 2011 and references therein). Accepted by Karol Marhold: 25 July. 2018; published: 4 Sept. 2018

101

FIGURE 1. Hesperis species collected from native habitats in Iran and Turkey and their visible morphological diversity. Map locations were generated from GPS coordinates in R (package: ‘ggmap’) and yellow lines indicate the geo-location of the representative species. a, H. kotschyi; b, H. pisidica; c, H. kuerschneri; d, H. bicuspidata; e, H. pendula subsp. pendula; f, H. breviscapa; g, H. microcalyx; h, H. straussii-Photos by B. Özüdoğru (a, b, f & g), H. Yıldırım (c & e), H. Altınözlü (d) & A. Eslami Farouji (h). See Table 1 and suppl. Table S1 for more details.

TABLE 1. Distribution and collection details of the Hesperis specimens (OTUs) selected for the multivariate analysis. Classical section1 Taxon1 Pachycarpos Hesperis Diaplictos Hesperis Diaplictos Cvelevia Hesperis Hesperis Pachycarpos Delicatae Hesperis

Cvelevia

H. anatolica Duran H. armena Boiss. H. balansae E. Fourn. H. bicuspidata (Willd.) Poiret H. bottae Fourn. H. breviscapa Boiss. H. buschiana Tzvelev H. dvorakii D. A. German H. hamzaoglui Duran H. hedgei Davis & Tan H. hyrcana Bornm. & Gauba

Abbreviation1 H. ana.

Accession numbers2, 3 Field Herbarium 4498*(a), 5335***(b)

Collection year Acronym of Field Herbarium herbaria4 1999–2000 HUB

H. arm. H. bal.

-

-

2002 1990–2002

HUB HUB

H. bic.

2016

1981–2008

HUB

H. bot. H. bre.

BÖ4121, BÖ3996 BÖ4499

5879(a), 5982(b) 2720(a), 4537(b), 4583(c), 5174(d) 8158(a), 842(b), 1840(c), 844(d), 4802(e) 5182(a), 4652(b) 4825(a), 5018(b)

2016

1999–2000 1999

HUB HUB

H. bus.

-

5237

-

2000

HUB

H. dvo.

-

6920(a), 6210(b)

-

2005–2008

HUB

H. ham.

-

5694*(a), 5758***(b), 5285(c)

-

2001

HUB

H. hed.

-

5185(a), 5511(b)

-

2000

HUB

H. hyr.

-

20521(a), 17107(b), 6884(c), 35032(d), 73565(e), 51663(f), 73273(g), 81428(h), 21597(i), 24(j), 22046(k) 13451(a), 6193(b), 1352(c), 5709(d), 1666(e), 1120(f), 3632(g), 57092(h)

-

1972–1998

TARI

2015– 1950–2006 2016

HUB

H. kotschyi Boiss. H. kot.

BÖ3928, BÖ3588

...continued on next page

102 • Phytotaxa 367 (2) © 2018 Magnolia Press

ESLAMI FAROUJI ET AL.

TABLE 1. (Continued) Classical section1 Taxon1 Diaplictos Hesperis

Mediterranea Diaplictos Diaplictos Diaplictos

H. luristanica. Dvořák H. matronalis L.

H. microcalyx E. Fourn. H. nivalis Boiss. & Hausskn. H. novakii Dvořák H. odorata Dvořák H. ozcelikii Duran H. pendula DC.

Abbreviation1 H. lur.

Accession numbers2, 3 Field Herbarium AEF983

H. mat.

-

H. mic.

BÖ3588

H. niv.

-

8164(a), 8162(b), 1226(c), 1873(d), 5295(e), 5352(f), 4733(g), 4742(h), 4734(i), 5010(j), 6531(k), 4851(l), 1937(m) 4322(a), 5552(b), 2884(c), 5213(d) 54356(a), 8610(b), 8573(c)

H. nov. H. odo.

-

H. ozc. H. pen.

BÖ3651

Collection year Acronym of Field Herbarium herbaria4 2015 Lorestan University 1978–1999 HUB

2016

1976–2002

HUB

-

1974–1986

TARI

5037(a), 4544(b), 2732(c) 16900(a), 16732(b)

-

1972–2000 1975

HUB TARI

2015

1999 1978–2005

HUB HUB

2016

1972–2009

TARI, Lorestan University

2016

1972–2007

TARI, Lorestan University

-

HUB

2016

1995 2001 2001 2000 2000 1999 1983–2016

2016

1975–1989

Pachycarpos

H. persica Boiss. subsp. kurdica

H. kur.

AEF1000

Pachycarpos

H. persica Boiss. subsp. persica

H. per.

AEF1001, AEF1002

Hesperis

H. pisidica Hub.Mor. H. podocarpa Boiss. H. quadrangula Boiss. H. sp. nov.

H. pis.

-

868* 1754(a), 3715(b), 1463(c), 4518(d), 596(e) 32871(a), 46808(b), 17432(c), 4701(d) 64795(e), 27408(f), 64394(g), 97083(h), 9156(i) 956(j) 6822(a), 2865(b), 36258(c), 6372(d), 24624b(e), 4722(f), 2449(g), 65716(h), 65511(i), 90653(j) 5763(a), 5802(b)

H. pod.

-

5746(a), 5623(b), 5317(c)

-

H. qua.

-

5033(a), 4530(b)

-

H. sp. nov.

102981*, 77372(a), 2539(b), 102984(c) 102979**, 102980, 102983 AEF1003 16754(a), 2332(b), 2359(c)

Hesperis Pachycarpos

Pachycarpos Diaplictos Hesperis

Hesperis

H. straussii Bornm.

H. str.

HUB HUB TARI

TARI, Lorestan University HUB

5318(a), 5626(b) 2000–2001 H. syriaca (DC.) H. syr. Dvořák H. thy. 5549(a), 3077(b), 5231(c) 2000–2011 HUB Contorta H. thyrsoidea Boiss. H. tos. 5822(a), 5657(b) 2001 HUB Hesperis H. tosyaensis Duran H. tri. 5157 2000 HUB Mediterranea H. tristis L. H. tur. 3482* 2003 HUB Hesperis H. turkmendağhensis Duran & Ocak 1 Classical sections (with respective color codes), taxon name and abbreviations are mentioned, which are used in Fig. 3. 2 Different type specimens are marked with asterisks. *= Holotype, **= Isotype, ***= Paratype 3 More than one accession from one species are listed by letters, mentioned within parenthesis in column ‘accession numbers’. The respective letters are also used in Fig. 3 with the species name. Field collections (2015–2016) by the authors are shown in italicized text. 4 Herbaria acronyms. HUB= Hacettepe University Biology Herbarium, TARI= Research Institute of Forests and Rangelands Diaplictos

Furthermore, Hesperis is a taxonomically critical genus and this is reflected by different estimations on species number ranging from 30 (Rollins 1993) up to 56 species (Cullen 1965, Davis et al. 1988, Duran et al. 2003, Duran & Ocak 2005, Al-Shehbaz 2012, Al-Shehbaz et al. 2006). Although Hesperis is represented by 11 species in Iran (Dvořák IRANIAN AND TURKISH HESPERIS L.

Phytotaxa 367 (2) © 2018 Magnolia Press • 103

1968), in recent field studies by the authors, three species [1, H. thyrsoidea Boissier (1867: 234); 2, H. borbasii Dvořák (1968: 270); and 3, H. novakii Dvořak (1965: 22)] could not be retrieved after a thorough search (also supported by search report from Valyollah Mozaffarian, personal communication). Cullen (1965) and Davis et al. (1988) reported 31 Hesperis species in Turkey. However, a recent study revised this number to 33 taxa representing 28 species, of which 27 taxa (82%) are endemic in Turkey (Duran in Sanad & Dašić 2016). Species descriptions, especially from Turkey, are generally very short and incomplete and thus species delimitation is very difficult or impossible (Duran et al. 2003). Furthermore, several studies outlined that the key characters used for species delimitation, i.e., hair type, varies with different environmental conditions (Aras et al. 2009, Dorofeyev 2013; Duran in Sanad & Dašić 2016), causing confusion in taxonomy. Among palynological, cytological and trichome characters, Duran et al. (2003) used traditional morphological characters, e.g., life form, stem height, leaf shape, and diverse fruit characters for sectional classification. All of these underscore that the taxonomic history of genus Hesperis has been controversial because a well-defined system of classification has not been created so far (see references above). Therefore, in this current study, we aimed to thoroughly evaluate and catalog the morphological characters relevant to the sectional classification of the genus. Here we used, for the first time, multivariate analyses of 30 qualitative and 27 quantitative morphological descriptors for 121 OTUs (including 33 Hesperis species) collected from Iran and Turkey. The main aim of this paper is to clarify species delimitation of critical Hesperis species from Iran and Turkey and to evaluate sub-generic classification based on numerical taxonomy.

Material and Methods Plant material: Samples comprising 33 Anatolian and Iranian Hesperis species, distributed throughout the major native habitat of the species over a region of > 1,000,000 km2 in Iran and Turkey known (Fig. 1), were used in this multivariate numerical taxonomic analysis. For most of the species, several specimens were included from dispersed populations to contain the inter-and intra-species diversity (Table 1). However, for the few rare species, which only occur in patchy habitat, we studied limited specimen and compensated this limitation by carefully recording all details from the available samples. Specimens were initially assigned names following the description of the taxa based upon Flora Iranica, Flora of Turkey, and other floras (Dvořák 1968, 1980, Cullen 1965, Duran et al. 2003). While several crucial morphological parameters relevant to the species identification (e.g., stem height, stem and fruit orientation, and color parameters) were distinctly recorded during our recent extensive field visits (2015–2016), complete information on all of the 57 selected morphological parameters could not be gathered (see the next section of methods for the morphological descriptors). Therefore, only the best-preserved, recent herbarium samples (1972– 2011) were considered for our numerical taxonomic study, where complete details and fidelity of characters during data accumulation are paramount. Herbarium samples were carefully and thoroughly studied in different herbaria such as Central Herbarium of Iran, Research Institute of Forests and Rangelands (TARI), Lorestan University Herbarium, FUMH (Herbarium of the Ferdowsi University of Mashhad), Hacettepe University Biology Herbarium (HUB) and Shiraz University (See Table 1 for the collection details). All characters recorded from the herbaria specimens were substantiated with the available field data (Table 1) for their reliability. Morphological descriptors scoring: Fifty-seven morphological descriptors (MD, 27 quantitative and 30 qualitative, see Table 2 for details) were carefully selected from different floras including Flora Iranica (Dvořák 1968), Flora of Turkey (Cullen 1965) and revisions (Duran et al. 2003) for critical assessment of the species delimitation. Altogether 121 specimens were selected among 307 studied specimens based on the ‘completeness’ of data on the 57 characters. A cut-off value of “3” was assigned in the selection of those 121 samples; meaning a specimen where data could not be measured for more than three MDs were rejected from the study. The exceptions to this cut-off rule e.g., i) samples selected with > 3 missing data and ii) samples rejected with ≤ 3 missing data are explained in details for relevant samples in the Supplementary Table 1. Briefly, in scenario (i), the samples were retained in the study due to their representation of wide geographical regions and considerable diversity in recorded characters; while in scenario (ii), the samples were rejected from the study due to incomplete voucher information, damaged/juvenile plant parts, and their redundancy in geographical distribution with no significant character variations. 104 • Phytotaxa 367 (2) © 2018 Magnolia Press


TABLE 2. Morphological descriptors, the character attributes for numerical analysis (qualitative characters), and polymorphism information content (PIC) scored for the multivariate analysis of Iranian and Turkish Hesperis species. For each descriptor, the average of several random specimens was scored all over the plant. Quantitative characters Stem

Branch Leaf

Sepal Petal

Pedicel

Anther/Androecium Fruit

Seed

Abbreviation

PIC

Stem height (cm)

STH

0.68

Stem width (cm)

SW

0.61

Stem hair length (mm)

STHL

0.51

Branch number

BN

0.59

Branch length (cm)

BL

0.66

Leaf length (cm)

LL

0.67

Lamina width (cm)

LW

0.65

Petiole (cm)

P

0.55

Position of the ontogenetically lowest stem leaf from the base of rosette (cm) Leaf hair length (mm)

LSL

0.16

LHL

0.69

Sepal length (cm)

SL

0.59

Sepal hair length (mm)

SHL

0.75

Petal length (cm)

PL

0.42

Petal claw length (mm)

PCL

0.69

Petal hair length (mm)

PHL

0.11

Number of petals Pedicel length at flowering (mm) Pedicel length after flowering (mm) Anther length (cm)

NP PLF

0 0.68

PLAF

0.70

AL

0.38

Fruiting pedicel length (cm)

FPL

0.64

Fruiting pedicel wide (mm)

FPW

0.77

Fruit length (cm)

FL

0.68

Fruit wide (mm)

FW

0.46

Fruit hair length (mm)

FHL

0.23

Fruit pedicel hair length (mm) Seed length (mm) Seed width (mm)

FPHL

0.71

SEL SEW

0.54 0.21

Qualitative characters

Abbreviation PIC

Character attributes

Life form Stem

LF SO

0.37 0

1-perennial; 2-biennial 1-erect; 2-decumbent

SCB

0.48

1-purple; 2-green

SH PSH

0 0.13

1-present; 2-absent 1-all over the stalk; 2-absent or only in lower part

SHT

0.72

1-simple, bifurcate, stellate, glandular; 2-rarely to moderately simple, bifurcate, glandular; 3-simple, bifurcate, rarely glandular; 4-simple, rarely bifurcate, glandular; 5-simple, rarely bifurcate, rarely glandular; 6-rarely simple, rarely bifurcate, mostly glandular; 7-simple, mostly glandular; 8-bifurcate and stellate; rarely glandular; 9-simple, rarely bifurcate, glandular; 10-simple or absent ...continued on next page

Life form Stem orientation Stem color at basal part Stem hairs Position of stem hairs Stem hair type

IRANIAN AND TURKISH HESPERIS L.


TABLE 2. (Continued) Quantitative characters Leaf

Abbreviation

Leaf shape

LS

0.71

Leaf hairs Leaf hair type

LH LHT

0 0.65

Sepal

Sepal color

SC

0.54

Petal

Sepal hairs Petal color

SEH PC

0 0.76

Petal hairs Petal hair density Type of petal hair Hairs at base of petals Pedicel hair type in flowering

PH PHD

0.37 0.40

TPH

0.39

HP

0.02

PHTDF

0.73

Pedicel

PIC

1-entire; 2-subentire; 3-entire to slightly dentate; 4-slightly dentate; 5dentate to slightly dentate; 6-dentate to entire; 7-dentate; 8-sinuate to irregularly dentate 1-present; 2-absent 1-simple, bifurcate, stellate, glandular; 2-simple, bifurcate, glandular or gland; 3-simple, bifurcate, rarely glandular; 4-simple, bifurcate, rarely to moderately stellate; 5-bifurcate, stellate, glandular; 6-bifurcate, stellate, rarely glandular; 7-simple, glandular, rarely to moderately bifurcate; 8bifurcate, rarely glandular; 9-simple, bifurcate; 10-simple 1-purple; 2-green to purple;3-green to yellow; 4-green; 5-yellow; 6-violet; 7-green or slightly purple; 8-dark green or light purple; 9-green or purple; 10-reddish violet; 11-very light purple; 12-green or light lilac 1-present; 2-absent 1-purple nearer to red; 2-purple; 3-lilac; 4-livid; 5-pinkish mauve; 6purplish to violet; 7-purple with white veins; 8-lilac with white veins; 9-lilac to white; 10-indistinctly purple; 11-whitish lilac to deep violet; 12sordid or brown; 13-sordid or deeply blue; 14-golden yellow; 15-brown or yellowish brown; 16-dirty yellow with dark violet veins; 17-reddish violet with whitish margins; 18-purple ± brown or yellow; 19-brownish, bronzy green or yellowish brown with purplish veins; 20-greenish light yellow or rarely light pinkish yellow; 21-white; 22-violet veined light purple 1-present; 2-absent 1-dense; 2-moderate; 3-low; 4-absent 1-simple, bifurcate, glandular; 2-simple and glandular or gland; 3glandular or gland; 4-simple; 5-absent 1-present; 2-absent

Inflorescence

Inflorescence

I

0.59

Gynoecium

Stigma shape Ovary hairs Pollen grain shape Fruit orientation Fruit shape Fruit hairs Fruit hair density Fruit hair type

SS OH PGS

0.56 0.56 0.16

1-simple, bifurcate, stellate, glandular; 2-simple, bifurcate, glandular; 3-rarely simple, bifurcate, glandular; 4-rarely to moderately bifurcate, rarely to moderately stellate, glandular; 5-rarely simple, mostly glandular; 6-bifurcate, rarely to moderately glandular or gland; 7-simple, bifurcate; 8-glandular; 9-bifurcate, rarely stellate; 10-simple; 11-absent 1-receme bracteate; 2-raceme ebracteate; 3-raceme bracteate and ebracteate; 4-raceme 1-globose; 2-weakly globose; 3-elliptic 1-present; 2-absent; 3-present or absent 1-compressed; 2-not compressed

FO

0.38

1-erect; 2-slightly erect; 3-pendulous; 4-pendulous and contorted

FS FH FHD

0.74 0.20 0.72

FHT

0.71

Width of fruit valves Fruit pedicel hairs Fruit pedicel hair density Fruit pedicel hair type

WFV

0.43

FPH

0.05

1-thickened; 2-slightly thickened; 3-thin; 4-very thin 1-present; 2-absent 1-dense; 2-moderate; 3-moderate to dense; 4-scattered; 5-very scattered; 6-absent 1-simple, bifurcate, glandular; 2-bifurcate, stellate, glandular; 3-simple, glandular; 4-rarely simple, rarely bifurcate, glandular; 5-rarely simple, bifurcate, stellate; 6-simple; 7-rarely simple or absent 1-wider than septum; 2-slightly wider than septum; 3-narrower than septum; 4-equal to septum 1-present; 2-absent

FPHD

0.58

FPHT

0.63

Anther/ Androecium Fruit


1-dense; 2-moderate; 3-moderate to dense; 4-scattered; 5-very scattered; 6-absent 1-simple, bifurcate, glandular; 2-simple, bifurcate, stellate; 3-simple, glandular; 4-rarely simple, rarely bifurcate, glandular; 5-bifurcate, glandular; 6-simple, bifurcate; 7-bifurcate, stellate; 8-glandular; 9-simple; 10-absent


The selected specimens were considered as separate independent operational taxonomic units (OTU), the basic unit of comparison in numerical phenetics (Ward 1993). Details of observed MDs and the specific incremental values assigned to the qualitative variations within each MD are described in Table 2 and in Supplementary Table S1. Quantitative data were scored with an electronic digital caliper, a ruler and a stereomicroscope with integrated scales. For each of the quantitative MDs, mean values obtained from several observations (at least five) in each of the studied specimens were used as the representative data for the OTUs. The variations of morphological descriptors among species of Hesperis were calculated in R (ver. 3.3.1, package “ggplot2”) and PAST (ver. 3.14, Hammer & Harper 2006) and demonstrated as box-plots. Moreover, the relative discriminatory power of each character was estimated as the polymorphism information content (PIC) of each MD which is calculated by the following equation

where Pij is the frequency of the jth value for the ith descriptor and summed over n no. of values (Das et al. 2007, Ni et al. 2002, Table 2). Multivariate analysis of data scoring: The quantitative and qualitative interval data among the OTUs were used without any modification or standardization to compute the Gower’s (1971) distance coefficient matrix with unweighted pair-group method of arithmetic averages (UPGMA) algorithm (Sneath & Sokal 1973) in PAST software for constructing the dendrogram to show average taxonomic distance (Radford 1986, Ward 1993). The multivariate analyses with Gower’s (1971) coefficient are suitable for the analysis of mixed characters, e.g., qualitative (binary and ordinal) and quantitative characters to generate a distance/dissimilarity matrix. Gower distance is calculated by the average of the difference over all variables and each term normalized for the range of that variable with the following formula

where the distance between the ith and jth unit were computed considering the kth variable’s contribution to the absolute distance measured between x[j,i] and x[k,i]. UPGMA is a well-accepted, frequently used, and accurate method to deduce similarity/dissimilarity among OTUs (Romesburg 1984, Radford 1986, Ward 1993). Based on the Gower’s coefficient among the OTUs, we also constructed 2D score plot from the principal coordinate analysis (PCoA) to deduce the delimitation among the major sections/phenons of the genus. Convex hulls were drawn around the delimiting OTUs for each of the groups to visualize the segregation of phenons over the first two resolving coordinates (PC1, PC2). Furthermore, the most resolving MDs for each of the phenons were deduced by a de-trended correspondence analysis (DCA). An expected unimodal response of OTUs to the underlying parameters are calculated using the DECORANA algorithm (Hill & Gauch 1980) with a two-step normalization procedure involving ‘straightening out’ of points lying in an arch and ‘spreading out’ the points to avoid clustering at the edges. All above-mentioned multivariate analyses were performed by either PAST software package or R with numerical taxonomy packages “ape” and “ggtree”.

Results The 57 morphological descriptors (MDs) revealed considerable diversity among the selected 121 OTUs of Hesperis (Fig. 2) studied from Iran and Turkey (Table 1) as revealed by the high PIC values for the distinctive quantitative (PICmax=0.77) and qualitative (PICmax=0.76) MDs (Table 2). Gower’s (1971) distance coefficient matrix was calculated from the 27 quantitative and 30 qualitative interval data among the 121 OTUs (online supplemental Table S2) and a consensus dendrogram were obtained with UPGMA cluster analysis (Fig. 3). Groups of similar species are clustered in five phenons, underlining the power of multivariate analyses with quantitative and qualitative morphological characters in the genus Hesperis (Fig. 3). The five phenons were also clearly separated in a 2D score plot of principal coordinate analysis (PCoA, Fig. 4, online supplemental Table S3) with all morphological descriptors.



FIGURE 2. Variation of each morphological descriptor among 33 species of the Iranian and Turkish Hesperis. The absolute values of the 27 quantitative descriptors are represented as colored boxes and whisker plots for different phenons and subphenons as follows: phenon I (P1, orange), subphenon II-A (P2a, brown), subphenon II-B (P2b, green), subphenon II-C (P2c, aquamarine), phenon III (P3, blue), phenon IV (P4, violet), phenon V (P5, pink). The box delimits the first and third quartiles of the data; the solid line within the box represents the second quartile/median, whiskers as upper and lower fences, and dots as outliers. The vertical axes shown absolute quantitative range of variables. See Table 1 for the explanation of the character abbreviations.



FIGURE 3. Dendrogram generated from Gower distance matrix with UPGMA clustering algorithm of qualitative and quantitative morphological characters of 121 Hesperis accessions from Iran and Turkey. OTUs grouped within the same phenon are marked by vertical black bar adjacent to their names and those within sub-phenon are included in colored boxes. The most distinctive features for each phenon are highlighted with representative pictures of various morphological parts from selected species (a-o). Traditional sectional classifications for the OTUs are mentioned with colored circles as follows (also in Table 1): Sect. Cvelevia (red), Sect. Hesperis (blue), Sect. Mediterranea (light green), Sect. Delicatae (pink), Sect. Diaplictos (violet), Sect. Pachycarpos (dark green), Sect. Contorta (light purple) (Dvořák 1968, Duran et al. 2003, Duran in Sanad & Dašić 2016). Species abbreviations are mentioned in Table 1 along with the letters following the species names, which represent different accessions. The distance scale is presented at the top of the dendrogram (See Supplementary Table S2 for the distance matrix). Scale bars within each of the pictures (a-o) represents 1 mm, except in f (2 mm). The pictures are taken by Atena Eslami Farouji and Burcu Şağbanoğlu.



Phenon I: The first phenon consists of H. breviscapa Boissier (1842: 67), which has shorter stem height (STH, 12.5±0 cm) and branch length (BL, 5.15±0 cm), and wider fruits (FW, 3±2.12 cm) than rest of the Hesperis species (Fig. 3). However, a de-trended correspondence analysis (DCA) revealed fruit hair type (FHT) and fruit pedicel hair type (FPHT, Table 2) as the most distinctive morphological descriptor, which distinguished this phenon from the other four phenons (Fig. 5A).

FIGURE 4. The Principal Coordinate Analysis (PCoA) revealed a clear separation of the five phenons. Each convex hulls (conjoint lines) representing each phenon, delineate taxa within them (represented by each dots) from other phenons. The Gower distance measures were used to compute Eigenvalues and scores (Supplementary Table S3).

Phenon II: The second phenon is composed of three subphenons (Fig. 3, mentioned with colored bounding boxes). Subphenon A forms a discrete group with exclusively biennial life form (LF), mostly purple stem color at basal part (SCB), and always exhibit simple and mostly glandular stem hair (SHT). In contrast to other phenons, stigma shape (SS) is always elliptic in subphenon A. H. anatolica Duran (2008: 454), H. thyrsoidea, and H. dvorakii German (2012: 13) are placed in subphenon A for their distinctive features as revealed by a DCA analysis (Fig. 5B). The short stem hair length (STHL, 1.53±0.39 mm) is characteristic of subphenon B, which included H. hamzaoglui Duran (2008: 456), H. podocarpa Boissier (1842: 65), H. bicuspidata (Willdenow 1800: 519) Poiret (1813: 195), H. matronalis Linnaeus (1753: 663), H. hyrcana Bornmüller & Gauba (1940: 254) and H. turkmendaghensis Duran & Ocak (2005: 240). However, other vegetative and reproductive characters were also indicated as discriminatory features for this subphenon (Fig. 5C). Ovary hair (OH) is always present in subphenon C, while it is present or absent in subphenon A and B. Moreover, some specific anther, fruit characters are associated with the members of this subphenon (Fig. 5D), which include H. armena Boissier (1842: 63), H. kotschyi Boissier (1856: 21), H. pisidica Huber-Morath (1965: 295), H. nivalis Boissier & Haussknecht in Boissier (1888: 45), H. novakii, H odorata Dvořák (1968: 272), H. microcalyx Fournier (1868: 352) and H. luristanica Dvořak (1966a: 29). Phenon III: The third phenon includes H. balansae Fournier (1868: 338), H. kuerschneri Parolly & Tan (2006: 851), H. bottae Fournier (1868: 352), H. pendula Candolle (1821: 457), and H. persica Boissier (1842:64) subsp. kurdica (Dvořák & Hadač 1964: 307) Dvořák (1971: 239) (Fig. 3). Petal hairs are always present in the members of this phenon as revealed by DCA analysis apart from other stem (STHL), leaf (LHT), and fruit (FW) features (Fig. 5E). 110 • Phytotaxa 367 (2) © 2018 Magnolia Press


Phenon IV: The fourth phenon is characterized by their longest seed length (SEL, 3.35±0.22 mm), the width of fruit valve (WFV) that almost equals to the septum are the characteristic features of this phenon (Fig. 5G). This fourth phenon includes H. ozcelikii Duran (2009: 578), H. tristis Linnaeus (1753: 663), H. syriaca (Candolle 1821: 185) Dvořák (1973: 267) and H. quadrangula Boissier (1842: 67). Moreover, this phenon has the highest sepal length (SL, 1.14±0.55 cm), anther length (AL, 1.29±0.59 cm), and fruit length (FL, 9.82±2.98 cm) among the other phenons.

FIGURE 5. The de-trended correspondence analysis (DCA) clarified the most important characters of the five phenons. The characteristic features of each phenon are delimited by the ellipse (P