Probability Models Based on Soil Properties for

Microb Ecol (2017) 74:550–560 DOI 10.1007/s00248-017-0958-2

FUNGAL MICROBIOLOGY

Probability Models Based on Soil Properties for Predicting Presence-Absence of Pythium in Soybean Roots Kimberly K. Zitnick-Anderson 1 & Jack E. Norland 2 & Luis E. del Río Mendoza 1 & Ann-Marie Fortuna 3 & Berlin D. Nelson Jr 1

Received: 1 June 2016 / Accepted: 24 February 2017 / Published online: 6 April 2017 # Springer Science+Business Media New York 2017

Abstract Associations between soil properties and Pythium groups on soybean roots were investigated in 83 commercial soybean fields in North Dakota. A data set containing 2877 isolates of Pythium which included 26 known spp. and 1 unknown spp. and 13 soil properties from each field were analyzed. A Pearson correlation analysis was performed with all soil properties to observe any significant correlation between properties. Hierarchical clustering, indicator spp., and multiresponse permutation procedures were used to identify groups of Pythium. Logistic regression analysis using stepwise selection was employed to calculate probability models for presence of groups based on soil properties. Three major Pythium groups were identified and three soil properties were associated with these groups. Group 1, characterized by P. ultimum, was associated with zinc levels; as zinc increased, the probability of group 1 being present increased (α = 0.05). Pythium group 2, characterized by Pythium kashmirense and an unknown Pythium sp., was associated with cation exchange capacity (CEC) (α < 0.05); as CEC increased, these spp. increased. Group 3, characterized by Pythium heterothallicum and Pythium irregulare, were associated with CEC and calcium carbonate exchange (CCE); as CCE increased and CEC Electronic supplementary material The online version of this article (doi:10.1007/s00248-017-0958-2) contains supplementary material, which is available to authorized users. * Berlin D. Nelson, Jr [email protected] 1

Department of Plant Pathology, North Dakota State University, Fargo, ND, USA

2

Natural Resources Management, North Dakota State University, Fargo, ND, USA

3

Department of Soil Science, North Dakota State University, Fargo, ND, USA

decreased, these spp. increased (α = 0.05). The regression models may have value in predicting pathogenic Pythium spp. in soybean fields in North Dakota and adjacent states. Keywords Pythium . Soil properties . Hierarchical clustering . Indicator species . Probability models

Introduction The genus Pythium contains numerous species that when present, are of importance because of the economic losses that result from reductions in yield associated with the pathogens. Pythium spp. typically cause pre- and post-emergence damping-off, a devastating agricultural and horticultural disease. Pythium as a soil-borne pathogen does not produce aerial spores for long-distance dispersal, and the life cycle occurs within the soil [1]. The pathogen infects plants primarily through the root system. The primary inoculum generally consists of zoospores, and activity of these flagellated spores is limited by the amount of moisture within the soil. Zoospores need water for dissemination and depending on the soil composition, can only travel as far as the capillaries and pore space within the soil allows. There has been limited research on the effects of soil properties on the occurrence and growth of Pythium and the diseases caused by Pythium. Texture, organic matter, and certain metals have been observed to have positive to adverse effects on Pythium [2, 3]. The characteristics of soil texture indirectly describe the amount of moisture a soil can retain. Soils greater in clay content retain more moisture than soils with more sand or silt [4]. When soil moisture decreases, the motility of the zoospores is negatively affected because the spores require free water to move [5]. However, a recent study suggested that the opposite was true for a diverse Pythium population

Pythium probability models based on soil properties

detected on soybean (Glycine max) in the Ohio Valley [6]. A decrease in disease incidence as clay increased could be due to the presence of organic matter, which was also observed to have a suppressive relationship with disease incidence [6]. Multiple studies have attempted to use organic matter for control of damping-off caused by Pythium [7–10]. Research has suggested that the nutrients in fresh or less decomposed organic matter are not readily accessible to Pythium [9]. The organic matter provides more nutrients to Pythium when the decomposition is more advanced [8]. Ultimately, the rate of organic matter decomposition is more important than the amount of organic matter [8]. Similarly, metals can hinder or support the basic critical functions of Pythium growth. The metal nickel has been observed to increase a plants ability to directly inhibit Pythium prior to contact with the plant root system [11]. Iron is another metal that indirectly suppresses Pythium by stimulating siderophore formation within Pseudomonas fluorescens, a commonly abundant soil organism [12]. In contrast, zinc has been reported to be critical in the formation of oogonia and the vegetative growth of Pythium [13]. Only recently have there been in-depth studies evaluating the relationships between Pythium community composition, species diversity, and soil properties. Broders et al. [6] examined differences in soil properties between Pythium groups in Ohio fields planted to corn (Zea mays L.) and soybean using principal component analysis to determine which soil properties accounted for variability in disease incidence between locations. They identified pH, calcium, magnesium, and the field capacity of a soil as the most important factors in the separation of five Pythium groups defined by cluster analysis. The Pythium communities were defined by the presence of specific spp. The first and largest community was characterized by the presence of Pythium inflatum and Pythium irregulare; the second by P. irregulare and P. inflatum in the presence of Pythium ultimum var. ultimum, Pythium dissotocum, and Pythium torulosum; and the third by P. irregulare and P. inflatum in conjunction with P. ultimum var. sporangiiferum, P. dissotocum, and P. torulosum [6]. The fourth community did not have the presence of P. inflatum and had Pythium pleroticum and an unknown spp., while the fifth community was characterized by the absence of P. irregulare and the presence of P. inflatum, P. ultimum var. ultimum, P. pleroticum, P. torulosum, and an unknown spp. [6]. In all of the Pythium communities, P. irregulare and P. inflatum were predominantly paired together [6]. There was a strong association between the soil components and the structure of the Pythium groups as well as Pythium diversity. As pointed out by Broders et al. [6], the composition of Pythium groups and the soil factors that affect them has received limited study. A recent study by Rojas et al. [14] examined the Pythium community in soybean roots over 11 major soybean

551

producing states and correlated community composition to environmental and edaphic factors. They found 15 factors that correlated with community composition. Soil parameters such as cation exchange capacity (CEC), pH, clay content, and bulk density were significant factors affecting communities. For example, they reported that Pythium heterothallicum increased in abundance at pH of 7 to 8 and a CEC of 30–40 meg/100 g soil while Pythium sylvaticum abundance decreased under those conditions. Environmental factors such as precipitation and temperature were also highly important to community composition. The objectives of this current study were to examine Pythium groups and indicator species in soybean roots from North Dakota using cluster analysis, then determine if soil properties of the fields where the plants were collected were associated with those Pythium indicator species. Logistic regression analysis was then employed to develop statistical models to calculate the probability of the presence of Pythium indicator species based on soil characteristics. Such information could be useful in understanding the ecology of Pythium in soil and enhance our understanding of diseases caused by Pythium.

Methods Collection of Plants and Soils The collection and identification of the spp. of Pythium used in this research is described by Zitnick-Anderson and Nelson [15] in a companion paper. Briefly, in June of 2011 and 2012, seedlings and soil samples were arbitrarily collected from 83 soybean fields in 20 counties in the eastern half of North Dakota, the primary soybean production area in the state. Ten seedlings with roots at the first trifoliate leaf stage were collected at random from each field and Pythium was isolated from roots. Seedlings were sampled regardless of presenting symptoms of disease. The emphasis was to determine Pythium populations in soybean roots. Pythium species were identified using both morphological and molecular methods [15]. The species and number of isolates of each spp. for each field site were recorded. Approximately 500 g of soil were collected to a depth of 25 cm from each field and analyzed for 13 total soil properties. Cation exchange capacity (CEC) and particle size analysis (sand, silt, and clay) were conducted using the Bower [16] and Hydrometer [17] methods, respectively. The other nine soil properties were analyzed by the Soil Testing Laboratory at North Dakota State University using standard methodology; P, K, pH, electrical conductivity (EC), organic matter (OM), Zn, Fe, Cu, and calcium carbonate exchange (CCE) [16, 18–24].

552

Isolation and Identification of Isolates Nine (2 cm long) cuttings were made at random from the roots of each plant and plated onto selective medium PARP + B (primaricin, sodium ampicillin, rifampicin, pentachlo ronitrobenzene, and benomyl) water agar (WA) using the under- the-block-technique [25]. Pythium can be present on seedlings regardless of the presence of disease symptoms, therefore justifying the random root cuttings of every plant collected. Root samples were incubated at 23 ± 2 °C in an incubator for 72 h then examined at ×10 to ×20 with a BX43 clinical microscope (Olympus, Center Valley, PA) under phase one lighting for presence of Pythium-like growth. Defining characteristics, such as coenocytic hyaline hyphae with flowing protoplasm, were used as criteria to select cultures for transfer. Sub-cultures were taken from the tips of hyphae and plated onto another selective medium, P 10 VP V8 agar, containing pentachloronitrobenzene, primaricin, and vancomycin [26]. Subcultures were grown for 3 days and then transferred to potato-dextrose agar (PDA Difco Laboratories). Cultures colonized PDA for 4 to 5 days in the incubator at 23 °C prior to attempts at identification. Wet mounts of each isolate were made for morphological descriptions. Isolates that did not readily produce sexual structures on PDA were cultured using a grass leaf culture technique that was modified from Abad et al. [27] by ZitnickAnderson and Nelson [15]. Briefly, tap water was used instead of deionized water to boil a combination of grass clippings (tall fescue, cv. Grande II and Kentucky Bentgrass); the water was allowed to cool and was decanted off and saved. Agar plugs with mycelium were placed in the water used to boil the grass clippings in 100 × 20 mm petri dishes and incubated at room temperature for 3 to 5 days. Sexual structure production for heterothallic spp. (Pythium diclinum, Pythium intermedium, Pythium kashmirense, Pythium attrantheridium, P. sylvaticum, P. heterothallicum, and P. inflatum) was accomplished by combining multiple isolates of the same species into one petri dish. All morphological structures for each isolate were examined as described previously. All morphological features were photographed and recorded using an Infinity 2 digital camera and, Infinity analysis computer program (Lumenera Corp., Ottawa, Canada.). Morphological features were compared to descriptions of spp. listed in the identification keys by Plaats-Niterink [28] and Dick [29]. These keys do not include a number of newly described species. When an isolate could not be identified using either key, DNA sequence analysis was used as described below to obtain a putative identification. From those potential species identities, the original publications describing the species were consulted and morphological features of the unknown were compared to those described in the literature. In addition to morphological features, DNA sequences were also used to identify isolates to species. The internal transcribed

Zitnick-Anderson K. K. et al.

spacer (ITS) sequence is a widely used DNA region that has good resolution and is a method accepted by the mycology community for spp. identification [30]. The primers ITS1 (TCCGTAGGTGAACCTGCGG) and ITS4 (TCCTCCGC TTATTGATATGC) were used to amplify a section of the 18S region, ITS1, the 5.8S region, ITS2, and a section of 28S region of ribosomal DNA [25]. The DNA extraction and PCR methods were as stated in Broders et al. [25]. The DNA extraction and PCR were performed on all isolates three times to confirm the molecular identification. The DNA sequence data were compared to known sequences that had been deposited in the National Center for Biotechnology Information (NCBI) nonredundant database to confirm morphological identification or to assist in the identification of isolates where using morphological identification was not attainable. The BLAST parameters for the sequences were sequence lengths, e values, maximum identity match, and query coverage. The sequence lengths were approximately 800 bp or greater. Identities were selected based on e values of 0.0, maximum identity match of 95% or greater, and a query coverage of 98% or greater. Species Diversity, Evenness, Spearman, and Pearson Correlation The species diversity and evenness were calculated for the entire Pythium collection using the Shannon index and evenness index E5 [31]. The Shannon index is described as H′ = Σ pi ln pi, where H′ is the species diversity score and pi is the proportion of individuals in the ith species [32]. The evenness equation is described as E5 = (((1/λ)-1)/ e H′-1), where λ is Simpson’s index [31]. Simpson’s index is a measure of diversity that accounts for number of species present and the abundance of each species [31, 33]. Species diversity and evenness were not calculated for each field because only one to two species on average were identified per field. The purpose of determining species diversity was to see how soil types affect the overall species diversity. Therefore, abundance data from all fields were aggregated to determine the overall diversity and evenness indices. However, it is important to note that aggregating all data could artificially inflate the species diversity index. The data used for abundance included the number of isolates for each species per field. Relationships between species diversity and the 13 soil properties were evaluated using Spearman’s correlation analysis (SAS version 9.1; SAS Institute, Cary, NC). A Pearson correlation analysis was performed for all 13 properties to observe any significant correlation between properties. Cluster Analysis, Indicator Species Analysis, and Multi-Response Permutation Procedure Hierarchical clustering, indicator species, multi-response permutation procedures (MRPP), and logistic regression analysis were used to determine the value of soil properties as

Pythium probability models based on soil properties

predictors of the presence-absence of different Pythium spp. PC-ORD version 6 was used to perform hierarchical clustering analysis [34]. Raw data was transformed using square root transformation because compared to several other transformations (logistic, arcsine), that transformation resulted in the most normal distribution of the data [34]. Indicator species analysis in combination with the hierarchical grouping of sample units (field numbers) was a method used to describe the structure of Pythium groups [35]. The Pythium groups can be formed using cluster analysis as shown in previous studies [6, 36]. Indicator species analysis is a method where the environmental differences can be conceptualized as groups of sample units [35]. These groups can be defined by categorical environmental variables, levels of disturbance, experimental treatments, and in the case of this study the geographical distribution based on presence-absence of a target species. Indicator species analysis combines data on the species abundance in a particular group and the exclusivity of a species in a particular group [35]. The perfect indicator species would be a species that is always present within a specific group and does not appear in any other group [35]. The indicator species analysis can be used as an objective criterion to determine the most ecologically meaningful point to prune a dendrogram from cluster analysis [34]. According to Legendre and Legendre [37], the use of indicator species Bprovide criteria to compare typologies derived from data analysis, to identify where to stop dividing clusters into subsets, and to point out the main levels in a hierarchical classification of sites.^ Hierarchical clustering analysis was performed using the relative Euclidean distance measurement and Ward’s method to group the fields together based on species abundance and frequency [34]. The dendrogram was used to define and determine the optimum number of groups implementing the Dufrêne and Legendre [38] method of pruning based on indicator species analysis [34]. Group membership at each step of cluster formation was entered into the PC-ORD 6 program where indicator values were calculated for each species at each level of grouping. The p values (generated using the Monte Carlo test) were averaged across all species; this procedure was repeated for all steps of clustering. The cluster step with the smallest averaged p value was determined to be the most informative level in the dendrogram [34]. In addition to the averaged p values, the number of species determined to be significant indicators (α ≤ 0.05) were tallied for each cluster step. The more species determined to be significant indicators with the lowest averaged p value was the criterion used for determining the clusters (groups) of Pythium spp. [34]. The cluster analysis using the criterion described indicated that three groups of Pythium were present. In addition, a MRPP was performed to ascertain that the three groups were in fact dissimilar. The test statistic (T) was used to verify separation between the three Pythium groups. The more negative the T value, the stronger the separation between groups [39]. The p value was used in conjunction

553

with the chance-corrected within-group agreement value (A). The p value is useful in evaluating the likelihood that an observed difference between data sets is due to chance. The smaller the p value, the less likely the observed differences in groups are due to chance [39]. The A value describes within-in group homogeneity, compared to the random expectation. When the A value is close to one, the items are identical within groups. When the A value is close to zero, heterogeneity within groups equals expectation by chance [39]. Logistic Regression Logistic regression analysis was performed using the presence-absence data from the three groups characterized by the indicator species and 12 soil properties using stepwise selection to generate multiple candidate models for presence of indicator species. The analysis was performed to create an accurate model that could be used to calculate probabilities of the presence and absence of each group based on statistically significant differences in soil properties. Analysis was conducted using SAS version 9.1. Clay content was eliminated from the data set since the Pearson correlation analysis showed a significant correlation with CEC and due to the fact that clay content and CEC are directly related. CEC is derived from negative charges on clays, and to a lesser extent soil organic matter, therefore the clay variable was contained within the CEC variable [3]. Logistic regression analysis requires at least 30 data points for results to be accurate in SAS version 9.1. Because not all Pythium spp. were present in at least 30 different fields, logistic regression analysis was not conducted on individual species, justifying the generation of the Pythium groups using cluster and indicator species analysis. The Akaike information criterion (AIC), and c (a variant of Somer’s D) value, plus the Hosmer and Lemeshow test were used to select the most appropriate model and evaluate the fitness and relative quality of each model for the data. The c value is similar to the R-square value of linear regression. In the absence of the R-square value, the c value can be used to highlight the accuracy of the models. The c value represents the Barea under the curve^ that is explained by the models. Since the c value ranges from 0 to 1, it can be interpreted similarly to the R-square value. After computing the y values from the logistic models, probabilities were calculated using the same formula (P = ey/ (1 + ey)), where P is probability and ey linearizes the logistic y values.

Results Isolation and Identification A total of 2877 isolates of Pythium were recovered from 83 fields in 20 counties during the months of June in the eastern

554

Zitnick-Anderson K. K. et al.

half of North Dakota. Overall, 26 species (P. attrantheridium, P. aristosporum, P. arrhenomanes, P. coloratum, P. debaryanum, P. diclinum, P. dissotocum, P. heterothallicum, P. hypogynum, P. inflatum, P. intermedium, P. irregulare, P. kashmirense, P. lutarium, P. minus, P. nunn, P. oopapillum, P. orthogonon, P. periilum, P. perplexum, P. rostratifingens, P. sylvaticum, P. terrestris, P. ultimum, P. viniferum, and P. violae) of Pythium and one unidentified spp. were identified using both DNA sequence analysis and morphological features. The unidentified spp. represented 16% of the total isolates.

Species Diversity, Evenness, Spearman, and Pearson Correlation The species diversity index was 2.45. When multiple species are present and equally abundant within a dataset, the diversity index approaches the value of five, the maximum value for the Shannon diversity index. When abundance data are dominated by a single species, despite the presence of other species, the Shannon diversity index will be closer to zero. The evenness index was 0.69, the ratio of the number of abundant species to the number of rarer species [31]. The closer the index is to zero, the less evenness between species is observed; a value close to zero would be indicative of a data set that had little diversity. In contrast, the opposite is true when the value is closer to one [31]. The evenness index indicates that Pythium was evenly distributed throughout the fields sampled. The Spearman correlation analysis between the Pythium diversity and the 13 soil properties provided 4 statistically significant (α = 0.05) correlations between diversity and CEC, CCE, OM, and Zn (Table 1). Although there were significant correlations between species diversity and other soil properties such as P, K, pH, EC, Fe, Cu, sand, and silt, the impact or percent correlation was below 13% and considered negligible. Although clay was also significantly (α = 0.0001) correlated with diversity, the property was shown to have a positive correlation with CEC in the Pearson correlation analysis (Table 2). Therefore, due to that strong biological relationship and the significant correlation, clay was not included in the logistic regression analysis. The Pearson correlation analysis (Table 2) also showed that pH, EC, Fe, and Cu also correlated with CEC, but these soil properties are not directly related to one another and are mainly Table 1

derived from the soil parent material [3]. Although the percent sand and OM had a significantly strong negative and positive relationship, respectively, with CEC, these properties were included in the logistic regression. The addition or exclusion of the percent sand and OM in the logistic regression analysis had no effect on the probability models. Cluster Analysis Three major groups were defined using cluster analysis, indicator species analysis, and MRPP. The indicator species analysis showed that pruning the cluster analysis dendrogram at cluster step three had the smallest averaged p value across all species (at α = 0.30) and the greatest number of species determined to be significant indicators (five; P. ultimum, P spp. unknown, P. kashmirense, P. heterothallicum, and P. irregulare) (Tables 3 and 4). The results of the MRPP had a test statistic (T) value of −32.01, α = 0.05, and an A value of 0.16 (Table S1). The MRPP results indicated that the three groups (Tables 3 and 4) were distinctly separate from one another, i.e., differences in the groupings were not due to chance, and these groups were heterogeneous (equaled the expectations by chance). Group 1 was characterized by the indicator species P. ultimum. Group 2 was characterized by an unknown Pythium spp., and P. kashmirense. Group 3 was characterized by P. heterothallicum and P. irregulare. Other Pythium spp. were found within each group; however, these species were not significant indicators as shown by the indicator species analysis (Tables 3 and 4). Logistic Regression The AIC values measure the amount of information loss for each model developed for each group. The model that was selected for each group had the smallest AIC value of all the possible candidate models developed for each group. The rank correlation of ordinal variables (c value), and the accuracy of the model for fit to the data (Hosmer and Lemeshow test) for group 1 characterized by P. ultimum were 0.86 and 0.8, respectively. The values were close to one, indicating that the logistic regression model extrapolated from the analysis was a good fit for the data set modeled. The logistic regression model for group 1 characterized by P. ultimum was logit (y) = −2.05 + 0.55 (Zn). Probabilities were then calculated showing that, as

Spearman’s correlation analysis of relationships between Pythium spp. diversity and soil properties

Spearman’s p Probabilityb

CECa

P

K

pH

EC

OM

Zn

Fe

Cu

CCE

Sand

Silt

Clay

−0.30 0.01

−0.07 0.00

−0.06