Sclerotinia sclerotiorum is an important pathogen of many crop plants which also infects wild hosts. The population structure of this fungus was studied for different crop plants and Ranunculus acris (meadow buttercup) in the UK using eight microsatellite markers and sequenced sections of the intergenic spacer (IGS) region of the rRNA gene and the elongation factor 1-alpha (EF) gene. A total of 228 microsatellite haplotypes were identified within 384 isolates from 12 S. sclerotiorum populations sampled in England and Wales. One microsatellite haplotype was generally found at high frequency in each population and was distributed widely across different hosts, locations and years. Fourteen IGS and five EF haplotypes were found in the 12 populations, with six IGS haplotypes and one EF haplotype exclusive to buttercup. Analysis of published sequences for S. sclerotiorum populations from the USA, Canada, New Zealand and Norway showed that three of the IGS haplotypes and one EF haplotype were widely distributed, while eight IGS haplotypes were only found in the UK. Although common microsatellite and IGS/EF haplotypes were found on different hosts in the UK, there was evidence of differentiation, particularly for one isolated population on buttercup. However, overall there was no consistent differentiation of S. sclerotiorum populations from buttercup and crop hosts. Sclerotinia sclerotiorum therefore has a multiclonal population structure in the UK and the wide distribution of one microsatellite haplotype suggests spatial mixing at a national scale. The related species S. subarctica was also identified in one buttercup population.
Sclerotinia sclerotiorum is an important necrotrophic plant pathogen with worldwide distribution and a host range of more than 400 species, including many agricultural and horticultural crops (Boland & Hall, 1994). The fungus is a haploid, multinucleate ascomycete and produces sclerotia which can survive in soil for several years. Infection of the majority of host plants is by ascospores released from apothecia produced through carpogenic germination of the sclerotia, although direct infection by myceliogenic germination can also occur. Sclerotinia sclerotiorum is predominantly homothallic and hence sexual reproduction to form apothecia through self-fertilization should result in a clonal population structure. This has been observed for S. sclerotiorum populations from a range of crops in Alaska, Australia, Canada, Iran, New Zealand, Turkey and the USA using DNA fingerprinting (Kohn et al., 1991; Kohn, 1995; Cubeta et al., 1997; Carbone et al., 1999; Carpenter et al., 1999; Carbone & Kohn, 2001; Hambleton et al., 2002; Phillips et al., 2002) or microsatellites (Sexton & Howlett, 2004; Sexton et al., 2006; Winton et al., 2006; Mert-Turk et al., 2007; Hemmati et al., 2009). In such studies, DNA fingerprints or microsatellite alleles were often closely associated with mycelial compatibility, an independent marker also used to genotype isolates and thought to be under multigenic regulation (Schafer & Kohn, 2006). The typical local distribution (often a single field) of S. sclerotiorum clones within a population is such that one or a small number of clones is sampled at high frequency with the remaining clonal genotypes found only once or a few times (Kohn, 1995). In some cases, the S. sclerotiorum clones sampled at high frequency at the local scale have also been found repeatedly over several years in the same locality and also over a wider geographic area. For instance, clones identified by DNA fingerprinting found repeatedly in western Canada on oilseed rape (Kohn et al., 1991; Kohli et al., 1992, 1995) were also identified on bean/soyabean in Ontario and Quebec in Canada in 1999–2000 (Hambleton et al., 2002). However, using the same DNA fingerprinting methodology none of these Canadian clones were found in S. sclerotiorum populations from the USA on oilseed rape from Georgia and Alabama (Phillips et al., 2002), on cabbage from North Carolina and Louisiana (Cubeta et al., 1997), or on lettuce from California or pea and lentil from Washington (Malvarez et al., 2007). The studies also showed that none of the USA populations shared DNA fingerprints with each other (Phillips et al., 2002; Malvarez et al., 2007). The distribution of most S. sclerotiorum clones is therefore often restricted geographically, with little or no sharing of genotypes between locations, resulting in genetically distinct subdivided populations (Malvarez et al., 2007). This was also found to be the case in Australia for S. sclerotiorum populations from oilseed rape fields in different areas 400 km apart (Sexton & Howlett, 2004). Population divergence in S. sclerotiorum has been examined through nested clade and coalescent analysis using multilocus sequence data from S.sclerotiorum populations derived from different hosts and locations. Genealogy based on the intergenic spacer region (IGS) of the rRNA gene was particularly informative and four main populations were originally identified: ‘subtropical’, ‘temperate’, ‘temperate subtropical’ and ‘wild’ (Carbone & Kohn, 2001). More recently, populations from Californian lettuce and Washington pea and lentil were identified as further genetically differentiated populations (Malvarez et al., 2007).
Although the vast majority of evidence from population studies suggests predominant clonality in S. sclerotiorum, there have been some reports of outcrossing and genetic exchange between isolates. For instance, the hypothesis of random mating based on linkage disequilibrium measures of molecular data could not be rejected for some S. sclerotiorum populations from lettuce in California (Malvarez et al., 2007), potato in Washington State (Atallah et al., 2004) and oilseed rape in Australia (Sexton & Howlett, 2004) and Iran (Hemmati et al., 2009). Although this is not direct evidence for recombination, there was also in some cases complete lack of association of molecular markers with mycelial compatibility group (MCG; Atallah et al., 2004). The most direct evidence of outcrossing was where sibling ascospores from a single apothecium collected in the field were found to belong to more than one MCG (Atallah et al., 2004; Malvarez et al., 2007).
Although the population structure of S. sclerotiorum has been well examined on crop plants in Canada and the USA, there have been no such studies in the UK. In addition, many wild plants are also susceptible to S. sclerotiorum and could potentially act as reservoirs of inoculum for crops or as a source of new genotypes. Only one population study has been carried out on a wild host. This involved populations from Ranunculus ficaria (lesser celandine) from two areas (Sandvika, Vestfold) in Norway (Kohn, 1995) which constituted the ‘wild’ population in the IGS genealogy of Carbone & Kohn (2001). In these populations, ascospores from individual apothecia collected in the field were often from different MCGs and there was incongruence between DNA fingerprint and MCG, potentially indicating outcrossing. These wild populations of S. sclerotiorum also exhibited low diversity (seven genotypes within 300 isolates), a more complex DNA fingerprint pattern and evidence of strong spatial substructuring compared to agricultural populations. Kohn (1995) concluded that populations of S. sclerotiorum on R. ficaria were inbred and isolated compared to those on agricultural crops.
The major aim of this study was to examine the population structure of S. sclerotiorum in the UK, not only for several agricultural crop hosts (carrot, lettuce, oilseed rape, pea and celery) but also for the wild host Ranunculus acris (meadow buttercup). It was found that S. sclerotiorum could regularly be isolated from dying buttercup flowers in meadows both close to and remote from UK agricultural production. In New Zealand, the natural occurrence of S. sclerotiorum on R. acris has resulted in the pathogen being developed as a mycoherbicide (Cornwallis et al., 1999; Verkaaik et al., 2004). Meadow buttercup is a good model wild host as it is geographically widespread in natural and semi-improved grassland in the UK and Europe and forms dense stands amenable to transect-based sampling within single meadows on a local scale analogous to sampling single fields of crop plants. Moreover, the present study also identifies ‘ancient’ buttercup meadows (both close to and remote from agricultural production), which had been out of cultivation for more than 100 years and which could potentially harbour a different S. sclerotiorum population structure as a result of host-based ecological speciation. The objectives of this study were therefore to determine: (i) if a ‘typical’ clonal population structure exists for S. sclerotiorum in the UK; (ii) if there is evidence of population differentiation of different hosts or locations; and (iii) if there are common haplotypes between wild and crop host populations.
Materials and methods
Sampling of S. sclerotiorum isolates from different crop hosts
Sclerotinia sclerotiorum isolates were obtained from different crops (carrot, lettuce, oilseed rape, celery and pea) and locations in England and Wales between 2005 and 2009 (Table 1). For each crop, sclerotia were collected from between one and five infected plants at points 8–10 m apart along transects. A minimum of 40 transect points were sampled for each crop and the sclerotia collected from different plants stored separately. Cultures of S. sclerotiorum were obtained from individual sclerotia by surface-sterilizing them in 50% v/v sodium hypochlorite and 70% ethanol for 4 min with agitation followed by two washes in sterile distilled water (SDW) for 1 min. Sclerotia were then bisected, placed on potato dextrose agar (PDA; Oxoid) and incubated at 20°C. After 3–4 days, the actively growing mycelium for each isolate was subcultured onto PDA (using agar plugs from the leading edge) and after approximately 6 weeks the sclerotia that had formed were collected and stored at 5°C. These stock sclerotia were used to initiate new cultures as required. Sclerotinia sclerotiorum isolates were obtained from a total of six populations: two from oilseed rape crops grown in 2005 and 2007 in adjacent fields in Herefordshire and one from each of the other crop plants in different locations (Table 1). For each population, 32 S. sclerotiorum isolates representing the spatially separated transect points were selected at random for molecular characterization.
Table 1. Location, host, sampling date and meadow age for Sclerotinia sclerotiorum populations from England and Wales, at locations between 5 and 367 km apart
Meadow age (years)a
aMeadow age estimated using the method of Warren (2009). Not estimated where Ranunculus repens absent.
Carrot (Daucus carota) cv. Nairobi
25 August 2005
Lettuce (Lactuca sativa) cv. Silverado
08 July 2005
Oilseed rape 2005
Oilseed rape (Brassica napus) cv. Winner
01 July 2005
Preston Wynn, Herefordshire
Oilseed rape 2007
Oilseed rape (B. napus) cv. Lioness
08 August 2007
Preston Wynn, Herefordshire
Celery (Apium graveolens) cv. Victoria
09 July 2009
Pea (Pisum sativum) cv. Setchey
17 July 2009
Sutton St Nicholas, Herefordshire
Meadow buttercup (Ranunculus acris)
31 May 2007
Meadow buttercup (R. acris)
23 May 2008
Deans Green 2008
Meadow buttercup (R. acris)
23 June 2008
Deans Green, Warwickshire
Deans Green 2009
Meadow buttercup (R. acris)
21 May 2009
Deans Green, Warwickshire
Meadow buttercup (R. acris)
15 June 2009
Michaelchurch Escley, Herefordshire
Elan Valley 2009
Meadow buttercup (R. acris)
25 June 2009
Elan Valley, Powys
Sampling of S. sclerotiorum isolates from meadow buttercup
Preliminary work had demonstrated that S. sclerotiorum could be regularly isolated from prematurely dead and dying flowers of the wild host meadow buttercup (R. acris). Four meadows in different locations with dense stands of buttercup were sampled in May/June between 2007 and 2009 (Table 1). Flowers from five plants showing symptoms of S. sclerotiorum infection were collected from a minimum of 40 points at 10-m intervals along transects (flowers from each individual plant stored separately). To obtain S. sclerotiorum cultures, the flowers were incubated on damp tissue paper in sealed plastic boxes at 20°C. After approximately 2–3 weeks, fungal growth emerging with morphology typical of S. sclerotiorum (white fluffy mycelium/sclerotia) was transferred onto PDA amended with 0·02 g L−1 chlortetracycline and incubated at 20°C. After 3–4 days, the actively growing mycelium for each isolate was subcultured to provide stock sclerotia as before. Sclerotinia sclerotiorum isolates were obtained from a total of six buttercup populations: four from two meadows in Warwickshire in different years (Holywell 2007, 2008; Deans Green 2008, 2009) and two meadows in other locations (Michaelchurch, Herefordshire and Elan Valley, Powys; Table 1). The age of each buttercup meadow was estimated using the method of Warren (2009) which is based on counting the number of flowers of Ranunculus repens (creeping buttercup) with additional petals to the standard number of five. The more plants with more than five petals, the older the meadow.
Molecular characterization of S. sclerotiorum isolates
Sclerotinia sclerotiorum isolates from all populations (n = 384) were characterized using eight microsatellite markers (Sirjusingh & Kohn, 2001) and sequencing of part of the intergenic spacer (IGS) region of the rRNA gene and part of the translation elongation factor 1-alpha (EF) gene. Sclerotinia sclerotiorum cultures derived from stock sclerotia were subcultured onto PDA and incubated at 20°C. Three agar blocks from the resultant actively growing mycelium were used to inoculate Petri dishes containing 20 mL potato dextrose broth (PDB, Formedium) and incubated for 3 days at 20°C. The agar plugs were then removed and the mycelial mat rinsed three times in molecular-grade water before being blotted dry and lyophilized. Total genomic DNA was extracted from approximately 0·01 g lyophilized mycelium using the DNeasy Plant Mini Kit (QIAGEN), following the manufacturer’s protocol.
Fluorescent-labelled primer pairs (Applied Biosystems) were used in multiplex PCR amplifications for sets of three (7-2, 8-3, 92-4) and five (13-2, 17-3, 55-4, 110-4, 114-4) microsatellite loci (Sirjusingh & Kohn, 2001). Multiplex PCR was performed in 10-μL reaction mixtures consisting of 1 × QIAGEN Multiplex PCR Master Mix, 0·5 × Q solution, forward and reverse primer pairs (0·2 μmol L−1 each) and c. 10 ng DNA template as described by Winton et al. (2006). Thermal cycling parameters were 95°C for 15 min, 35 cycles of 94°C for 30 s, 55°C for 90 s and 69°C for 75 s, followed by 69°C for 75 s. The size of PCR products was determined on an ABI Prism 3100 Genetic Analyser and allele sizes assigned using GeneMarker v. 1.6 (SoftGenetics). FlexiBin (Amos et al., 2007) was used to bin allele sizes and estimate the relative number of sequence repeats for individual S. sclerotiorum isolates at each of the eight loci. Between two and six separate PCR amplifications per locus were carried out for each isolate to ensure the reproducibility of amplicon sizes, the greater number of repeats being used for rare allele sizes.
PCR primers were designed to amplify an 834-bp (approx.) fragment of the IGS of the rRNA gene corresponding to variable sites (nos 38–47) which distinguished six S. sclerotiorum lineages (A–F) as described by Carbone et al. (1999). PCR amplification was carried out in 25-μL reaction mixtures consisting of 12·5 mL 0·5 × REDTaq ReadyMix PCR reaction mix (Sigma-Aldrich), IGS2F (5′-TTACAAAGATCCTCTTTCCATTCT-3′) and IGS2R (5′-GCCTTTACAGGCTGACTCTTC-3′) primers (4 μmol L−1) and c. 10–30 ng DNA template. Thermal cycling parameters were 94°C for 2 min, 40 cycles of 94°C for 30 s, 57°C for 30 s and 72°C for 2 min, followed by 72°C for 10 min. PCR products were visualized on a 1·5% agarose gel to confirm amplification and then purified using the QIAquick PCR purification kit (QIAGEN), according to the manufacturer’s protocol. Sequencing of the purified products was carried out using BigDye v. 3.1 terminator chemistry (Applied Biosystems) according to the manufacturer’s instructions and an ABI Prism 3100 Genetic Analyser. PCR amplification was also carried out for a 300-bp portion of the EF gene in 25-μL reaction mixtures consisting of 12·5 μL 0·5 × REDTaq ReadyMix PCR reaction mix, EF1-728F and EF1-986R primers (4 μmol L−1; Carbone et al., 1999) and c. 10–30 ng DNA template. Thermal cycling parameters were 94°C for 2 min, 40 cycles of 94°C for 15 s, 60°C for 20 s and 72°C for 1 min, followed by 72°C for 5 min. PCR amplification was confirmed by gel electrophoresis and products purified and sequenced as described before.
Analysis of microsatellite data
Diversity, linkage disequilibrium and population differentiation
arlequin software (Excoffier et al., 2005) was used to calculate the haplotype frequency for each S. sclerotiorum population (based on the relative number of repeats at each of the eight microsatellite loci), identify shared haplotypes between populations and calculate Nei’s unbiased gene diversity (expected heterozygosity). Genotypic diversity was calculated as , where pi is the frequency of the ith genotype (Stoddart & Taylor, 1988). This diversity index has a minimum value of 1 where all isolates have the same genotype and a maximum value of the sample size for each population; it is hence an indication of how evenly genotypes are distributed. As each S. sclerotiorum population in this study had the same sample size (32) genotypic diversity was comparable between populations. Significant differences in genotypic diversity between populations were determined using a t-test (Chen et al., 1994).
Multilocus indices of disequilibrium were determined for each population data set and also for clonally corrected data sets where only one representative of each microsatellite haplotype was included using the software multilocus (Agapow & Burt, 2001). The program computes both the index of association IA and an alternative measure rd which is less dependent on the number of loci. The significance of these indices was tested by comparing observed values with those expected under the null hypothesis of random mating based on 1000 randomizations of the sample. If IA or rd was significantly different from 0, then the hypothesis of random mating was rejected.
The extent of population subdivision was estimated using arlequin through pairwise comparisons of RST (Slatkin, 1995), a statistic analogous to Wright’s FST (Wright, 1951), but which uses a stepwise mutation model appropriate for microsatellites. Significance was tested by permuting (1023) haplotypes between populations.
arlequin was also used to perform an unstructured hierarchical analysis of molecular variance (amova) to partition the variance between and within all S. sclerotiorum populations. amova was also carried out with a structure where S. sclerotiorum populations were grouped according to whether they were sampled from buttercup or one of the five agricultural hosts to explore the hypothesis that these groups were different. Significance of the resulting F-statistics was tested using the inbuilt non-parametric permutation approach (1023 permutations).
A preliminary investigation of population structure was made by cluster analysis of the microsatellite data based on Nei’s unbiased minimum distance (Nei, 1987) using Tools for Population Genetic Analysis (tfpga; Miller, 1997) and construction of a dendrogram with bootstrapping of data conducted using 1000 permutations. A Bayesian approach to infer population structure through clustering genetically homogenous groups based on microsatellite allele frequencies was also implemented using structure (Falush et al., 2003). This approach allowed the identification of groups without assuming any predefined structure based on geographical location or host plant. Although structure can deal with mixture and admixture linkage disequilibrium, it can potentially overestimate the number of clusters if there is strong background linkage disequilibrium as a result of markers that are very closely linked (Falush et al., 2003). Hence both fstat (Goudet, 1995) and genepop (Rousset, 2008) were used to test linkage between pairwise combinations of the microsatellite loci. Where significant linkage was found, the structure analysis was re-run excluding the data associated with either one of the linked loci. In addition, IGS and EF haplotype data were also added to the complete microsatellite data and the analysis run again to confirm the number of groups found using the microsatellite data alone. For all the structure analyses the admixture model using the correlated allele frequencies option was used with a 20 000 burn-in period and 100 000 Markov chain Monte Carlo (MCMC) iterations for 20 replicate runs assuming the true number of populations (K) ranged from 1 to 20. The standard deviation in the posterior probability L(K) was calculated for each value of K as well as first- and second-order (ΔK) rates of change of the likelihood distribution for successive values of K according to the procedure outlined by Evanno et al. (2005). With this procedure, the best estimates of K are those associated with the highest values of ΔK (Evanno et al., 2005). structure also calculates the probability of individuals belonging to each of the K populations.
Analysis of sequence data
Sequences obtained for IGS and EF regions were aligned using the clustalW algorithm implemented in mega v. 4 (Tamura et al., 2007) and imported into the snap workbench software for further analysis. snap map (Aylor et al., 2006) was used to produce a haplotype map and frequencies for both IGS and EF by collapsing sequences with phenotype (assigned as sampling location/year) excluding indels and infinite site variations. DnaSP v. 5 (Librado & Rozas, 2009) was used to calculate haplotype diversity (Hd) for IGS and EF sequences where , fi is the frequency of the ith genotype and n is the number of individuals sampled. DnaSP was also used to examine the extent of subdivision between populations using pairwise comparisons of the nearest neighbour statistic (Snn) with significance testing carried out using the inbuilt permutation scheme (1000 permutations; Hudson, 2000). To visualize IGS and EF phylogenies and haplotype frequencies, a median joining network of haplotypes (Bandelt et al., 1999) was constructed as implemented in network v. 4.6 (Fluxus Technology, USA). These phylogenies were further expanded using IGS and EF sequence data from GenBank for S. sclerotiorum populations from different hosts, locations and years in Canada, the USA, New Zealand and Norway as described by Carbone & Kohn (2001; Table S1).
To examine migration between populations, the ratio (M) of immigration rate (m) to mutation rate (μ) (= m/μ), the number of migrants per generation (= 2Nem, where Ne is the effective population size) and theta (θ = 2Neμ) were estimated using the Bayesian method in migrate for microsatellite and IGS sequence data with and without a geographic distance matrix implemented (Beerli & Felsenstein, 2001). To reduce the number of estimated parameters and improve consistency of results, symmetric migration was assumed and population data from the same location (but from different years) were pooled (OR1 + OR2, HO1 + HO2, DG1 + DG2) as these were not differentiated in other tests. This gave a total of nine populations from different locations for analysis. The Brownian motion approximation to the stepwise mutation model was used for the microsatellite data. For both microsatellite and sequence data a Bayesian inference search strategy was performed using a constant mutation rate and an exponentially distributed prior. A slice-sampling MCMC algorithm was used to search the proposal distribution with a static heating scheme. One long chain was used with a burn-in of 10 000 iterations and 10 000 recorded steps. Runs were repeated to ensure consistency of results. Shapiro–Wilk tests for normality were performed on the output migration rate distributions and Mann–Whitney U-tests were performed to determine whether the migration rates calculated from each of the runs were equally distributed.
Mycelial compatibility testing of S. sclerotiorum isolates
In addition to molecular characterization, S. sclerotiorum isolates from five of the populations (CA1, LE1, OR1, OR2 and HO1) were also grouped according to mycelial compatibility tests following the methodology described by Schafer & Kohn (2006). Briefly, this involves confronting two isolates on agar and observing their interaction. Compatible isolates grow over each other to form a continuous colony whereas an incompatible reaction results in dead cells or a clear barrage zone between the colonies. Based on this, isolates can be assigned to MCGs and previous work has suggested that these are generally associated with a particular S. sclerotiorum microsatellite haplotype (e.g. Sexton & Howlett, 2004). Sclerotinia sclerotiorum isolates were initially tested for compatibility within each of the five population samples of 32 isolates. This was done by dividing each population of 32 isolates into eight groups of four which were paired in all combinations. If two or more isolates within these groups were compatible, then only one representative isolate of that MCG was used in the next round of tests where each of the initial groups was combined with another for pairing the isolates. This scheme was continued until all isolates had been assigned to an MCG. Finally, isolates representing MCGs identified in each of the five S. sclerotiorum populations were tested against each other. Pairs of isolates which resulted in compatible reactions were retested at least twice and control self–self pairings were also carried out at each round of testing.
Detection of Sclerotinia subarctica
While generating microsatellite data and IGS sequences for buttercup population MI1 from Herefordshire, 15 of the original 32 isolates selected at random for genotyping had some microsatellite alleles outside the expected size range and also had IGS sequences very different to those previously observed for S. sclerotiorum. These isolates had typical morphology of S. sclerotiorum in culture although sclerotia were generally larger. Sequencing of the rRNA ITS regions and the lack of an intron in the large subunit (LSU) rDNA compared with S. sclerotiorum (Holst-Jensen et al., 1998) identified them as S. subarctica (Clarkson et al., 2010). Hence a further 15 S. sclerotiorum isolates which were available from population MI1 were selected at random for IGS and microsatellite analysis after their identity was confirmed by ITS sequencing (Clarkson et al., 2010) to give a total of 32 isolates.
Analysis of microsatellite data
Diversity, linkage disequilibrium and population differentiation
Microsatellite analysis of all the S. sclerotiorum isolates resulted in 4–17 polymorphic alleles per locus, with loci 7-2, 8-3, 13-2, 17-3, 55-4, 92-4, 110-4 and 114-4 having 4, 7, 17, 13, 15, 6, 9 and 15 alleles, respectively. In total, 228 microsatellite haplotypes were found within the 384 isolates from the 12 S. sclerotiorum populations (Fig. 1) and of these, 22 were shared between two or more populations. Overall, one haplotype was found at a much higher frequency than the rest (haplotype 1, 42 isolates; Fig. 1) and was also the most common haplotype within each individual population, with the exception of buttercup DG1 and celery CE1 where it was found at low frequency, and buttercup EV1 where it was completely absent. The second most frequent haplotype (haplotype 2, 21 isolates; Fig. 1) was found in all buttercup populations with the exception of HO2, but in only one agricultural population (LE1). The total number of haplotypes found in the six agricultural populations was 125 compared to 117 in the six buttercup populations, with 14 haplotypes shared. Between one and five microsatellite haplotypes were found to persist from 1 year to the next in populations from the three locations where sampling was done in consecutive years (OR1 and OR2, HO1 and HO2, DG1 and DG2). Within each population sample of 32 isolates, the number of microsatellite haplotypes ranged from 13 (buttercup EV1) to 29 (buttercup HO2), while the number of unique haplotypes ranged from nine (EV1) to 23 (HO2; Table 2). Gene diversity within each population was in the range 0·51–0·74 with the exception of buttercup EV1 which had a much lower diversity of 0·29 (Table 2). Genotypic diversity, which essentially measures how the proportional abundances are distributed among the different genotypes within populations, varied between 7·8 and 23·2, with the values for populations DG1, EV1 and CE1 being particularly low (7·8, 9·3 and 10·0, respectively) and significantly different from populations HO2, OR2, MI1 and LE1 (Table 2).
Table 2. Microsatellite haplotype frequency, diversity and linkage disequilibrium measures for Sclerotinia sclerotiorum populations from crop plants and buttercup (32 isolates per population)
No. unique haplotypesa
NS: not significant.
aNumber of haplotypes not found in any other population.
bGenotypic diversity (Stoddart & Taylor, 1988). Values followed by the same letters are not significantly different (P ≤ 0·1) in t-tests.
cIndices of association IA and rd calculated using the software multilocus (Agapow & Burt, 2001).
The values of the index of association IA ranged from 0·042 (buttercup EV1) to 2·263 (buttercup DG1) with corresponding values of rd from 0·008 to 0·325 (Table 2) when all the data for each population were used. Significance testing showed that the hypothesis of random mating was rejected for all populations except for buttercup EV1 and this was confirmed when clone-corrected data sets were also tested (data not shown). The fixation index (RST) values indicated significant differentiation of some S. sclerotiorum populations (Table 3). Buttercup population EV1 was differentiated from all other populations, while celery population CE1 was also differentiated from the majority of the other populations with the exception of lettuce LE1, buttercup DG1 and pea PE1. No differentiation was detected for S. sclerotiorum populations sampled from the same site in different years; oilseed rape OR1 and OR2, buttercup HO1 and HO2 and buttercup DG1 and DG2.
Table 3. Fixation index (RST) valuesa for pairwise comparisons of Sclerotinia sclerotiorum populations from crop plants and buttercup
RST-based amova analysis with no population structure assigned showed that 93% of the molecular variance was within S. sclerotiorum populations and 7% between populations. Imposing the structure of grouping populations according to whether they were from buttercup or crop hosts, RST-based amova showed that 93% of the variance was within populations with only 1% assigned among these two defined groups (not significant) and 6% between populations. However, if FST-based amova was used, 2·6% of the variation was attributable to differences between buttercup and agricultural populations which was significant (P = 0·04).
Analysis based on Nei’s minimum genetic distance using tfpga grouped the S. sclerotiorum populations into two main clusters. Cluster 1 contained carrot CA1, oilseed rape OR1 and OR2, pea PE1 and buttercup DG2 and cluster 2 contained lettuce LE1, buttercup DG1, MI1, HO1 and HO2. The remaining populations, celery CE1 and buttercup EV1, were not clustered with any other population (Fig. 2). Whilst buttercup EV1 was clearly not in clusters 1 and 2, the bootstrap confidence for CE1 being different from clusters 1 or 2 was low. There was no clear clustering of buttercup or agricultural populations, although cluster 1 had a higher proportion of the latter.
The Bayesian cluster analysis using structure with microsatellite data suggested that the number of genetically distinct ancestral populations for S. sclerotiorum was K = 3 based on the fact that this value was associated with the highest value of ΔK. Significant linkage was found between marker 17-2 and the majority of the other microsatellite loci, but re-running structure excluding these data again resulted in K = 3 based on the highest value of ΔK as was the case when IGS and EF haplotype data were added. All results were consistent for replicate runs of the structure program. Plotting the probability of each individual S. sclerotiorum isolate belonging to each of the three populations, shown as blue, green or red in Figure 3, revealed that all individuals from buttercup EV1 were assigned to the blue population. Individuals from the remaining buttercup populations and also those from lettuce LE1 and celery CE1 were largely assigned to the blue population, whereas those from the crop hosts OR1, OR2, PE1 and CA1 were largely assigned to the green population. Few individuals from each S. sclerotiorum sample were generally assigned to the red population, but where this did occur they were assigned with high probability. OR1 and OR2 also had very similar proportions of individuals assigned to the red, blue and green populations.
Analysis of sequence data
In total, 14 IGS and five EF haplotypes were identified within the 384 isolates from the 12 S. sclerotiorum populations (Table 4). Sequences representing each haplotype were deposited in GenBank (IGS, JQ219943JQ219956; EF, JQ219957JQ219961). The number of haplotypes in each population ranged from three to eight for IGS and from two to four for EF. Haplotype diversity ranged from 0·341 (buttercup EV1) to 0·819 (buttercup HO1) for IGS and from 0·272 (celery CE1) to 0·718 (buttercup EV1) for EF (Table 4).
Table 4. IGS and EF haplotype frequency and diversitya for Sclerotinia sclerotiorum populations from crop plants and buttercup (32 isolates per population)
The IGS network and haplotype frequency data (Fig. 4, Table 4) revealed that 13 of the 14 IGS haplotypes in the UK were represented in populations originating from buttercup compared to eight in agricultural populations. IGS1, IGS2 and IGS3 were the most abundant haplotypes comprising 32, 110 and 188 isolates respectively (Table 4). Six IGS haplotypes each comprising either one or two isolates (IGS8, 9, 10, 11, 12 and 13) were found exclusively within buttercup populations. IGS6 (eight isolates) and IGS4 (14 isolates) were also made up almost exclusively of buttercup isolates except for one or two lettuce isolates respectively. The majority of these buttercup IGS haplotypes had one base substitution change compared to the more common haplotypes (IGS2, 3 and 5) found in both crop plants and buttercup. Generally, isolates with the same microsatellite haplotype also had the same IGS haplotype. Isolates comprising the high frequency microsatellite haplotype 1 (42 isolates) all had the predominant IGS haplotype (IGS3). Isolates (21) from the second most frequent microsatellite haplotype 2 were also assigned to IGS3, with the exception of one from buttercup MI1, which was assigned to IGS4. Overall, nine of the 46 shared microsatellite haplotypes comprised isolates from more than one IGS haplotype, with a maximum of three IGS haplotypes (IGS2, 3 and 14) for microsatellite haplotype 13.
When the other sequence data sets from Canada, the USA, New Zealand and Norway were included, three further IGS haplotypes were added to the network (IGS15, 16 and 17), with IGS15 taking the position of the hypothesized haplotype linking IGS3, IGS5 and IGS14 for the UK data. IGS1, 2 and 3 were common in the UK, Canada, the USA and New Zealand, whereas IGS15 (Alabama, Georgia, Louisiana, North Carolina) and IGS16 (Alabama, Georgia, Louisiana) were only found in the USA. IGS6 consisted mainly of isolates from R. ficaria and R. acris in Norway and the UK, whilst IGS17 was exclusively from the R. ficaria population in the Sandvika region in Norway. IGS haplotypes 4, 8, 9, 10, 11, 12, 13 and 14 were exclusive to the UK.
The EF network and haplotype frequency data (Fig. 5, Table 4) showed that for the UK, each of the five EF haplotypes was represented in S. sclerotiorum populations originating from buttercup compared to four in agricultural populations. EF4 was made up almost exclusively of buttercup isolates except for one lettuce isolate. EF5 was exclusively from the buttercup population EV1. As for IGS, isolates with the same microsatellite haplotype also had the same EF haplotype. Isolates comprising the high-frequency microsatellite haplotype 1 were all assigned to EF2 and the majority of microsatellite haplotypes 2 and 3 were assigned to EF3 and EF1, respectively. Overall, only six of the 46 shared microsatellite haplotypes comprised isolates from more than one EF haplotype. When the other sequence data sets from Canada, the USA, New Zealand and Norway were included in the phylogenetic network, three further EF haplotypes were added (EF6, 7 and 8). EF1 was found in all countries, whilst EF7 was exclusively from Georgia, USA and EF8 from R. ficaria in Norway.
The nearest neighbour statistic (Snn) values showed that there was significant subdivision between many of the pairwise comparisons of the S. sclerotiorum populations for both IGS and EF (Table 5). However, the celery CE1 and buttercup EV1 populations were the only ones significantly differentiated from all the other populations for IGS and EF, with the exception of CE1 vs CA1 for EF. No differentiation was detected for S. sclerotiorum populations sampled from the same site in different years: oilseed rape OR1 and OR2, buttercup HO1 and HO2 and buttercup DG1 and DG2.
Table 5. Nearest neighbour statistic (Snn)a values for pairwise comparisons of Sclerotinia sclerotiorum populations from crop plants and buttercup for IGS (below diagonal) and EF (above diagonal) sequences
Analyses using migrate did not indicate great differences in migration between S. sclerotiorum populations in the nine different UK locations using either microsatellite or sequence data (data not shown). The number of migrants per generation (= 2Nem = θM) ranged from 0·5 to 1·3 for microsatellite data and from 0·6 to 1·7 for IGS sequence data, suggesting moderate gene flow between populations (Beerli & Palczewski, 2010). Shapiro–Wilk tests for normality on the migration rates for each run led to rejection of the null hypothesis that the collective rates were normally distributed. Independent sample Mann–Whitney U-tests, and not t-tests, were therefore performed, leading to acceptance of the null hypothesis that samples from different runs with or without the geographic distance matrix implemented were drawn from the same distribution. This indicated that runs were consistent and that geography had no effect on migration rates.
Mycelial compatibility testing of S. sclerotiorum isolates
A total of 105 mycelial compatibility groups (MCGs) were identified within the 160 isolates from the five S. sclerotiorum populations tested (CA1, LE1, OR1, OR2 and HO1) and only seven of these MCGs contained isolates from more than one population. This compared with a total of 111 microsatellite haplotypes found in the same 160 isolates, with six found in more than one population. Isolates in the same MCG always had the same IGS and EF haplotype and generally the same microsatellite haplotype. The frequency distribution of microsatellite haplotypes and MCGs was therefore very similar for each S. sclerotiorum population. However, there were some exceptions to the linkage between microsatellites and MCGs. For instance, while 21 of the 23 isolates that comprised the most common MCG had an identical microsatellite haplotype (corresponding to the high-frequency haplotype 1 in Fig. 1), one of the remaining two isolates (from buttercup HO2) differed in allele size at one microsatellite locus while the other (from lettuce LE1) differed at four loci. In all, six of the 23 MCGs that comprised more than one S. sclerotiorum isolate contained isolates that differed in allele size at between one and seven of the eight microsatellite loci. However, there was only one instance where an isolate with the same microsatellite haplotype was not in the same MCG (from lettuce LE1).
The population structure of S. sclerotiorum was examined in the UK for the first time on both crop plants and a wild host, meadow buttercup (R. acris). During the study, S. subarctica (Sclerotinia sp. 1) was discovered in buttercup population MI1 (Herefordshire), in sympatry with S. sclerotiorum, the first report of this pathogen in the UK (Clarkson et al., 2010). Whilst S. subarctica has been reported in Norway and Alaska (Holst-Jensen et al., 1998; Winton et al., 2006), its presence in the UK indicates that it is not confined to high latitudes as previously suggested (Winton et al., 2006). However, S. subarctica was only found in one location and not in any crop plant samples, suggesting it is not widespread in the UK.
The microsatellite analysis revealed that each UK S. sclerotiorum population comprised multiple haplotypes, but with one sampled at a much higher frequency than the rest, and the majority of other haplotypes sampled only once or a few times. The most common microsatellite haplotype was distributed widely and at high frequency across different hosts, locations and years, as it was present in 11 of the 12 populations sampled at distances of up to 367 km apart. Sclerotinia sclerotiorum populations in the UK therefore have the same multiclonal structure, consisting of one or a few high-frequency haplotypes, as observed in other parts of the world (see Introduction). The reasons for the success of certain S. sclerotiorum clones have yet to be understood but could be the result of traits conferring successful reproduction and spread, such as aggressiveness, sclerotial survival and germination or efficacy of ascospore dispersal. However, evidence suggests that aggressiveness is not correlated with clone frequency (Kull et al., 2004), whilst ascospores are generally only dispersed locally within crops or in neighbouring fields (Wegulo et al., 2000; Hammond et al., 2008). The occurrence of shared microsatellite haplotypes between different S. sclerotiorum populations, and particularly the presence of a predominant haplotype in all populations except for buttercup EV1, suggests contemporary or historic exchange and spatial mixing of haplotypes from both wild and agricultural hosts over a wide area. This result is in contrast to Kohn (1995) on the wild host R. ficaria in Norway, where no S. sclerotiorum genotypes were found in common with agricultural isolates.
Although one microsatellite haplotype predominated and a total of 22 were shared between two or more populations, 59% (228) of the 384 isolates had distinct haplotypes. Other S. sclerotiorum population studies using microsatellites have reported haplotype proportions of 89% for potatoes in Washington State, USA (Atallah et al., 2004), 51% and 52% for two studies of oilseed rape in Australia (Sexton & Howlett, 2004; Sexton et al., 2006), 63% for oilseed rape in Turkey (Mert-Turk et al., 2007), 29% for oilseed rape in Iran (Hemmati et al., 2009), but only 11% for various vegetable crops in Alaska (Winton et al., 2006). The present study also observed more alleles at loci in common with the USA, Alaskan and Turkish studies. Although these results are not directly comparable because of differences in the range of shared microsatellite markers and sample numbers, they suggest a comparatively high level of gene diversity for S. sclerotiorum in the UK. Genotypic diversity based on the relative abundance of microsatellite haplotypes in each population ranged from 24 to 72% of the maximum. This was comparable with the population studies in Australia (29–80%, Sexton & Howlett, 2004; 28–68%, Sexton et al., 2006), but was a larger range than found in Iran (21–45%, Hemmati et al., 2009).
The IGS and EF sequence data provided a different, lower level of phylogenetic resolution than the microsatellite data, with 14 IGS and five EF haplotypes identified within the 12 S. sclerotiorum populations. The occurrence of different IGS haplotypes followed a similar pattern to the microsatellites, with some predominant haplotypes (IGS1, 2 and 3) found at a higher frequency than the rest. Six haplotypes (IGS8, 9, 10, 11, 12 and 13) were found exclusively in UK buttercups, suggesting the possibility of host adaptation. In preliminary tests S. sclerotiorum isolates representing these ‘buttercup’ IGS haplotypes were capable of causing disease on oilseed rape plants, although the majority were less aggressive than common ‘crop’ IGS haplotypes (data not shown). However, complementary pathogenicity tests on buttercup plants with a range of S. sclerotiorum isolates from different host origins have yet to be carried out. Although variation in aggressiveness between different S. sclerotiorum isolates or MCGs is commonly reported (e.g. Kohli et al., 1995; Ziman et al., 1998; Cornwallis et al., 1999; Kull et al., 2004) there is so far little or no evidence to suggest host specificity for certain S. sclerotiorum genotypes.
Sclerotinia sclerotiorum isolates with the same microsatellite haplotype generally had the same IGS and EF haplotype and were in the same MCG (for the five populations where MCG was determined). The linking of different markers with few instances where they were de-coupled leads to the conclusion that reproduction in UK S. sclerotiorum populations is predominantly clonal with low levels of outcrossing and was also supported by measures of linkage disequilibrium for the microsatellite data. This is in agreement with many other studies as outlined in the Introduction.
The IGS and EF sequence data published for S. sclerotiorum populations from the USA, Canada, New Zealand and Norway (Carbone & Kohn, 2001) allowed comparison with UK populations, although conclusions relating to haplotype frequency are problematic because of different sample sizes. Nonetheless, the most common IGS (IGS2 and 3) and EF haplotypes in the UK (EF1) were found in most other countries. However, IGS6 was found only in R. ficaria from Norway and predominantly R. acris from the UK, whilst a further eight haplotypes (IGS4, 8, 9, 10, 11, 12, 13 and 14) were exclusive to the UK. Although the available IGS sequence data are limited, the frequency and relationship of different IGS haplotypes observed in the phylogenetic network suggest that a few, common haplotypes (IGS1, 2 and 3) have been distributed globally, probably through anthropogenic factors such as artificial movement of soil, plant material and seeds contaminated with sclerotia (Malvarez et al., 2007). It is likely that the lower-frequency IGS haplotypes have then emerged locally from the more common haplotypes (e.g. IGS 4, 9, 12, 13 from IGS3, and IGS8 and 11 from IGS2 in the UK). The microsatellite data support this hypothesis at the local scale with the more widely distributed microsatellite haplotypes having the most common IGS haplotypes. However, further data need to be gathered for populations from different countries in order to fully understand the distribution of S. sclerotiorum globally.
The finding that there are S. sclerotiorum microsatellite and IGS haplotypes common to both UK buttercup and crop populations suggests that wild hosts can potentially act as sources of inoculum for crop plants (and vice versa). Other researchers have also proposed a role for wild plants in initiating sclerotinia disease in nearby crops as well as enabling the survival and proliferation of the pathogen in the absence of a susceptible host (e.g. Morgan, 1971; Hims, 1979; Phillips, 1992). There are little or no data on how often wild hosts become infected, although this study was able to find diseased R. acris flowers in all the sampling locations irrespective of year and their proximity to susceptible crops. In addition, S. sclerotiorum infections were occasionally found in weeds such as fat hen (Chenopodium album), nettle (Urtica spp.) and thistle (Cirsium spp.), both within the crop and in the field margins.
Although common haplotypes were found in different S. sclerotiorum populations, including buttercup and crop hosts, there was also evidence of differentiation based on both microsatellite and sequence data, particularly for buttercup population EV1. Microsatellite, IGS and EF haplotype frequencies also varied between years in populations sampled from the same locations, but these were not significantly differentiated. Changes in haplotype frequency from year to year in the same field have been observed previously and probably reflect a mixing of ascospore inoculum from both resident and immigrant sources (Kohli et al., 1995) or perhaps differential responses of isolates to environmental conditions or host plants present in a particular year. Although analysis of microsatellite data using tfpga and structure to a certain extent grouped the S. sclerotiorum buttercup and crop-plant populations separately, this was not consistent, as, for example, the lettuce and celery populations LE1 and CE1 more closely resembled a ‘buttercup’ population for both analyses. The FST-based amova of the microsatellite data suggested a significant difference between buttercup and crop populations, whereas RST-based amova did not. There was therefore only weak evidence for consistent differences between S. sclerotiorum populations from buttercup and crop hosts based on the microsatellite data. In contrast, a greater diversity of IGS haplotypes was found in the majority of buttercup populations (with the exception of EV1) than in those from crop plants. Moreover, six ‘buttercup’ haplotypes were exclusive to one or more buttercup populations, suggesting greater levels of diversity in the wild host for this particular marker. Genetic diversity or differentiation in buttercup did not appear to be related to meadow age.
Buttercup population EV1 was consistently distinct from all other S. sclerotiorum populations as: (i) it was the only population where the predominant microsatellite haplotype was absent; (ii) it was significantly genetically differentiated from all other populations for both microsatellite and sequence data; (iii) it clustered separately from all other populations for microsatellite data using tfpga and had a completely different population allocation profile in structure; (iv) it was the only population where the hypothesis of random mating was not rejected using measures of linkage disequilibrium, suggesting possible outcrossing; (v) it had fewer microsatellite haplotypes and much lower genotypic diversity than all other populations; and (vi) it had a much lower IGS haplotype diversity. This evidence suggests that buttercup EV1 is an isolated population and hence represents a similar situation to that observed by Kohn (1995) for S. sclerotiorum populations on R. ficaria from Sandvika and Vestfold in Norway. The Sandvika population was reported to be ‘well away from regions of agricultural production’ while the Vestfold population was ‘in a hedgerow between fields in cereal–pasture rotation’, presumably where agricultural hosts of S. sclerotiorum were largely absent. Data from the UK Land Cover Map 2000 database (LCM2000) suggests that the Elan Valley site is similarly isolated from crop production as it was the only location where there was little or no agriculture (arable or horticulture) within 20 km (data not shown). Geographically, the meadow is in an isolated valley. This population merits further investigation to establish whether it has other attributes associated with isolation.
In conclusion, S. sclerotiorum populations from meadow buttercup and crop hosts in the UK have a similar clonal population structure with wide distribution of one common high-frequency microsatellite haplotype. Overall, there was only weak evidence for differentiation of S. sclerotiorum populations from crop plants and the wild buttercup R. acris, suggesting there is no ecological specialization in response to host adaptation. However, disentangling the factors that promote differentiation is problematic and to do so conclusively would require more consistent sampling than was possible here. Ideally, S. sclerotiorum isolates from each host (preferably in close proximity) would be required from every sampling location in different years. However, the distribution of buttercup meadows and different cropping areas, as well as the requirement for rotations, means that this is extremely difficult in practice. Sclerotinia disease incidence can also vary considerably from year to year and hence infected crop plants are not always available. It is not surprising, therefore, that few plant pathogen population studies comparing both wild and crop hosts have been carried out. However, a study of sympatric populations of Botrytis cinerea on grapevine and bramble in different locations (Fournier & Giraud, 2008) demonstrated significant differentiation and restricted gene flow between these agricultural and wild hosts, providing evidence that divergence in response to host specialization is possible in generalist plant pathogens such as S. sclerotiorum.
We thank the Warwickshire and Herefordshire Wildlife Trusts for allowing access to buttercup meadows for sampling and J. Warren (Aberystwyth University) for help with buttercup identification and meadow age estimation. We also thank Faye Ritchie (ADAS) for collecting the pea isolates, Julie Smith (ADAS), Howard Hinds Crop Consultancy and G’s Growers for identifying crops with sclerotinia disease and Dez Barbara for useful comments on the manuscript. The work was funded by the Department for Environment, Food and Rural Affairs.