Genetic diversity of Puccinia striiformis from cereals in Alberta, Canada

Authors


Abstract

Stripe rust of wheat caused by Puccinia striiformis f. sp. tritici has recently become a production problem on wheat in Alberta, Canada, and stripe rust of barley caused by Pstriiformis f. sp. hordei occurs regularly. A total of 261 isolates of Pstriiformis were collected from wheat, barley, Hordeum jubatum and triticale plants in Alberta, Canada from 2007 to 2012, and compared to isolates from other provinces and the USA. The genetic diversity of the pathogen was assessed using 11 simple sequence repeat (SSR) markers and by examining a length polymorphism in the ribosomal DNA (rDNA) intergenic spacer 1 (IGS1) region. A total of 28 SSR genotypes were detected within Alberta. The 13 genotypes common on wheat (Pstriiformis f. sp. tritici) were distinct from the 15 genotypes common on barley (Pstriiformis f. sp. hordei). Four SSR genotypes, two within each forma specialis, represented 85% of the isolates recovered. Genotypic diversity was low, population genetic analysis indicated a clonal structure, and the genotypes were widely dispersed. In both formae speciales, the dominant genotype varied between years. The second most common Pstriiformis f. sp. hordei genotype was found to be more closely related to older Pstriiformis f. sp. tritici genotypes from the USA than to other Pstriiformis f. sp. hordei genotypes.

Introduction

Stripe (yellow) rust on small grain cereals is caused by Puccinia striiformis and occurs worldwide. Puccinia striiformis is a functionally obligate, biotrophic, heteroecious rust fungus. Until recently it was thought to be microcyclic, but it was shown to successfully complete sexual reproduction on Berberis chinensis and Mahonia aquifolium (Jin et al., 2010; Wang & Chen, 2013). Stripe rust has traditionally been divided into formae speciales based on the host it primarily infects. Puccinia striiformis f. sp. tritici (Pst) mainly occurs on wheat, but also includes some barley, rye and triticale cultivars as hosts, and is the most economically important and widely studied of the stripe rust pathogens. Barley stripe rust is caused by Pstriiformis f. sp. hordei (Psh). A new forma specialis, pseudo-hordei (Psp-h), causing barley grass stripe rust was first described in Australia (Wellings et al., 2000). Psp-h occurs mainly on wild Hordeum species and is avirulent towards most wheat and barley cultivars tested (Wellings et al., 2000; Golegaonkar et al., 2013).

The genetic structure of Pst populations has been examined on multiple continents using virulence markers and molecular markers such as simple sequence repeats (SSRs) and amplified length polymorphism (AFLP), and is typically found to show low diversity and often to be dominated by several clones (Hovmøller et al., 2002; Bahri et al., 2009b). Despite this, there is evidence that sexual reproduction as well as hybridization do occur within some Pst populations (Cheng & Chen, 2009; Mboup et al., 2009). Less attention has been paid to Psh, with the few studies examining its populations focusing on isolates from the USA (Chen et al., 1995; Spackman et al., 2010).

In North America, Pst was first identified in the USA in 1915 and then in British Columbia, Canada in 1916 (Humphrey et al., 1924). Two years later it was found in Edmonton, Alberta on the wild barley grass species Hordeum jubatum (Fraser & Conners, 1925). This appears to be near the northern limit of the pathogen's North American range. Recent studies of Pst in North America have often resulted in the separation of the formae speciales into three distinct groups: the old population that was present prior to 2000, an invasive population that entered around 2000, and a group of pathotypes believed to be the result of recombination between the pre- and post-2000 populations (Markell & Milus, 2008; Chen et al., 2010). The invasive post-2000 population was first detected in the southeastern USA, and was found to be very dissimilar from the pre-2000 population based on AFLP markers (Markell & Milus, 2008). Members of this post-2000 population have been reported to be more aggressive, particularly at higher temperatures (Milus et al., 2009) and since their introduction have become dominant across the USA (Chen et al., 2010). After 2000, Pst began to cause significant levels of disease in regions where it had previously been negligible (Chen, 2005). In 2007, new distinct pathotypes were detected in the western USA. These pathotypes were considered to potentially be the result of recombination between isolates belonging to the pre-2000 and the post-2000 populations (Chen et al., 2010).

The first documented incursion of Psh into the Americas was in 1975 in Colombia and by 1982 the pathogen had spread to almost all commercial barley-producing regions in South America (Dubin & Stubbs, 1986). The disease was first identified in North America in 1987 in Mexico and reached Washington, USA, south of the Canadian border in 1995 (Chen et al., 1995).

The Pacific Northwest region of the USA is believed to act as the primary centre of diversity for the species in North America. The greatest diversity in pathotypes is detected in this region and new pathotypes typically originate from there (Chen et al., 2010). This is attributed primarily to the diversity of cultivars grown in the region as well as to the cropping system and a favourable climate that allow year-round survival of the pathogen (Chen, 2005; Wan & Chen, 2012). Additionally, the only known alternative host of Pstriiformis native to North America, Maquifolium, is common in the region. However, the pathogen's sexual stage has not yet been detected on Maquifolium in the wild (Wang & Chen, 2013). The Pstriiformis inoculum that arrives in Canada is believed to come from the Pacific Northwest and Great Plains of the USA (Chen, 2005). As the winter conditions in Alberta are considered too adverse to allow for the survival of the pathogen, local extinction occurs annually followed by recolonization the subsequent year. Such extinction–recolonization cycles are typical of cereal rusts in North America and Pstriiformis on other continents (Brown & Hovmøller, 2002).

The aims of the present study were to: (i) examine the population structure of Pstriiformis in Alberta, Canada; and (ii) differentiate the two most common formae speciales, Pst and Psh.

Materials and methods

Isolate collection and preservation

A total of 261 uredinial isolates (Table S1) of Pstriiformis were collected during 2007–2012 in central and southern Alberta, Canada. One hundred and fifty-six collections were made from spring or winter wheat, 89 from spring barley, three from triticale, and 13 from Hjubatum located in research plots or commercial fields. Because of variability in the severity and distribution of the disease between years, sampling was performed in an ad hoc manner. Thirty-one isolates were collected in 2007, 29 in 2008, 15 in 2009, 66 in 2010, 68 in 2011 and 52 in 2012. Single-uredinial isolates were derived from the collections, multiplied and stored until use as detailed by Kumar et al. (2012). An additional four Pst isolates from Manitoba (2005) and two Pst isolates from British Columbia, Canada (2009) were added to the isolate set. The majority of Canadian isolates (165 of 267) were characterized previously based on their virulence on cultivars of their host species (Kumar et al., 2012; Holtz et al., 2013). DNA from five pathotypes of Pst from the USA were also included for comparison with the genotypes detected within Alberta. Two Pst pathotypes first identified in the USA before 2000 (PST-035 & PST-045), two common widespread pathotypes first detected after 1999 (PST-078 & PST-114) and one pathotype first detected in 2007 in the western USA (PST-127) were used (Chen et al., 2010).

DNA extraction

DNA was extracted from each isolate using the following procedure. Urediniospores (c. 5–25 mg) were added to 400 μL lysis buffer (50 mm Tris, pH 8·0, 50 mm EDTA, 3% SDS, 1% mercaptoethanol). Spores were ground using either a micropestle attached to an electric drill or with an equal volume of acid-washed sand and three steel balls in an oscillating mill (MM 301, Retsch) at 15 Hz for 3 × 30 s. An equal volume of extraction buffer (100 mm Tris, pH 8·0, 10 mm EDTA, 10 mm NaCl, 2% PVP, 1% sarkosyl) was added. The mixture was incubated at 65°C for 30 min. The lysate was extracted with 700 μL phenol: chloroform: isoamyl alcohol (25:24:1 v/v/v). The supernatant was transferred to a new tube and an equal volume of isopropanol and a 0·1 volume of 3 m sodium acetate was added. The DNA was precipitated at −20°C for 20 min then centrifuged at 10 000 g for 10 min. The DNA pellet was washed with 70% ethanol, dried and then resuspended in 100 μL TE (10 mm Tris, 0·1 mm EDTA, pH 8·0). The concentration of the extracted DNA was determined by spectrophotometry (NanoDrop 1000, Thermo Scientific). DNA was diluted in sterile deionized water to 5 ng μL−1 before PCR analysis.

SSR genotyping

Eleven microsatellite primers developed for Pst were used to characterize the Pstriiformis isolate collection: RJ4, RJ13, RJ17, RJ18, RJ21, RJ24 (Enjalbert et al., 2002), RJ2N, RJ8N, RJ10N, RJ11N and RJ13N (Bahri et al., 2009a). PCR reaction conditions were 94°C for 2 min, followed by 35 cycles consisting of 30 s at 94°C, 30 s at 52°C and 30 s at 72°C, with a final extension step performed for 5 min at 72°C. The PCR reactions contained 1 × PCR buffer, 1·5 mm MgCl2, 0·2 mm dNTP, 0·4 μm each primer, 10 ng genomic DNA and 0·4 U Taq DNA polymerase. Amplification products were separated by electrophoresis in 6% polyacrylamide gels stained with ethidium bromide. Fragments were visualized under UV light and allele sizes were determined by comparison to a 20 bp ladder (Lonza) and isolates with previously characterized allele sizes with the software AlphaView (Cell Biosciences).

Intergenic spacer length polymorphism

Length polymorphisms in the intergenic spacer 1 (IGS1) region of the ribosomal DNA (rDNA) were examined for all isolates using the primers Q and Y (Fox et al., 1995). PCR was performed in 20 μL volumes with 1 × PCR buffer, 1·5 mm MgCl2, 0·2 mm dNTP, 0·4 μm each primer, 10 ng genomic DNA and 0·4 U Taq DNA polymerase using the cycling conditions described by Fox et al. (1995). The amplification products were separated on 1% TBE agarose gels stained with ethidium bromide and visualized under UV light. Allele sizes were determined by comparison to a 100 bp ladder (Lonza) with AlphaView. Reactions were repeated on replicate DNA extractions of isolates that had rare IGS1 alleles.

Data analysis

Genotypic and genetic diversity assessment

For the purpose of data analysis, isolates were considered to be either Pst or Psh based on their previous assessment on differential sets. For the isolates that had not previously been characterized on differentials, they were considered to be the same forma specialis as isolates that had been characterized and shared the same SSR genotype or a very similar genotype.

Standard single locus statistics, number of alleles (Na), effective number of alleles (Ne), number of private alleles, expected (He) and observed (Ho) heterozygosity and deviations from Hardy–Weinberg equilibrium (HWE) were calculated for Albertan isolates of each forma specialis separately and combined using the program GenAlEx v. 6.5 (Peakall & Smouse, 2006, 2012). Genotypic diversity was calculated using R = ( 1)/(N − 1), where G is the number of genotypes in the sample, and N is the number of isolates in the sample (Dorken & Eckert, 2001).

Test of genetic recombination

Tests for departure from random mating were done for both formae speciales individually and combined, using both the total data set and the clone-corrected data set, where only a single representative of each genotype was retained. The index of association corrected for the number of loci (math formula) was used (Agapow & Burt, 2001). Deviations from the null hypothesis of math formula = 0 (complete panmixia) were tested with 1000 randomizations of the data sets using multilocus v. 1.3 (Agapow & Burt, 2001).

Group assignment and test of genetic differentiation based on genetic distance

Cluster analysis was performed on the clone-corrected SSR data set. The shared allele distance (DAS; Jin & Chakraborty, 1994) was calculated using populations v. 1.2.32 (Langella, 2002). A UPGMA tree based on the resulting distance matrix was then generated. The resulting tree was visualized using FigTree v. 1.4 (Rambaut, 2012).

The program structure v. 2.3.3 (Pritchard et al., 2000) was used to implement a Bayesian model-based clustering method to identify the number of populations (K) present in the species. The analysis was performed on the clone-corrected version of the total SSR data set. The admixture model with independent allele frequencies among populations and no a priori forma specialis identification or geographic information was used. A burn-in of 50 000 Markov chain Monte Carlo (MCMC) iterations and a run of 100 000 MCMC iterations was used with K set as 1–8 with 10 repetitions. The value of K best supported by the model was established by determining the value of ∆K over the 10 repetitions using the method described by Evanno et al. (2005) implemented in the program structure harvester (Earl & von Holdt, 2012). The structure output for the best value of K was visualized with distruct v. 1.1 (Rosenberg, 2004). structure analysis does assume random mating and linkage equilibrium within subpopulations, which often does not occur in Pstriiformis. Despite this, the method has been shown to be capable of identifying the population structure of clonal Pstriiformis populations (Bahri et al., 2009b).

Principal coordinate analysis (PCoA) based on the genetic distances (Smouse & Peakall, 1999) was performed in GenAlEx v. 6.5 to visualize the genetic pattern among all SSR genotypes detected. PCoA does not rely on assumptions of random mating or linkage equilibrium and can therefore be used to determine if the violation of assumptions made by structure produced erroneous results.

Association between SSR genotype and virulence phenotype

The relationships between the genotype and virulence phenotypes of the isolates were examined with Mantel tests. The pathotypic distance between isolates was calculated using the simple matching (SM) distance between isolates. This was based on the isolates' virulence on the 41 cultivars (23 wheat, 17 barley and one triticale) consistently used to characterize the isolates previously (Kumar et al., 2012; Holtz et al., 2013). The calculations were performed using Virulence Analysis Tool (Schachtel et al., 2012). The SM distance matrices were calculated for both formae speciales separately and combined. All 105 Pst and 60 Psh isolates that were phenotyped were included. Matching genetic distance (DAS) matrices were calculated. Mantel tests, with 999 permutations of the data sets, were calculated in GenAlEx v. 6.5. The relationship between genetic and pathotypic distance for the 67 Pst examined by Holtz et al. (2013) was also determined using the virulence data from the 41 cultivars described above and seven additional wheat cultivars from the USA that were included in that study.

Results

A total of 28 SSR genotypes were detected among the samples from Canada. An additional two genotypes were present in the reference Pst samples included from the USA (PST-035 and PST-045). Eleven of the genotypes (T1–T6, T8, T9 and T11–T13) occurred in isolates previously identified as Pst. Two other genotypes (T7 and T10) were assumed to be Pst because of their similarity to other Pst genotypes and they were recovered exclusively from wheat (Table 1). Twelve of the genotypes (H1, H2, H4–H11, H13 and H14) occurred in isolates previously identified as Psh. Three additional genotypes (H3, H12 and H15) were assumed to belong to Psh isolates because their SSR genotypes clustered with other Psh genotypes and they were only recovered from cultivated barley (Table 1). Of the 176 Pst isolates collected in Alberta, all were recovered from wheat with the exception of three from triticale, 10 from barley and eight from Hjubatum. All Psh isolates were recovered from cultivated barley except three from wheat and five from Hjubatum. The genotypic diversity of the samples was low: 0·104 overall, 0·0693 in Pst and 0·163 in Psh.

Table 1. Abundance, year of collection and host species of Puccinia striiformis genotypes recovered in Alberta, Canada
Genotype200720082009201020112012Total
WheatBarleyTriticaleWheatBarleyWheatBarleyWheatBarley Hordeum jubatum WheatBarleyTriticale Hordeum jubatum WheatBarley
T12318193304312231118
T2112
T3123
T411
T5112
T611
T722
T8330437
T911
T1011
T1144
T1211
T1311
H1514142641
H2112
H333
H422
H511
H6124119624
H711
H8112
H922
H1011
H1133
H1211
H1311
H1422
H1511

All 11 microsatellite markers were polymorphic. Allelic diversity tended to be relatively low, ranging from two to eight alleles per locus with a total of 42 alleles identified (Table 2). The Ho was higher than the He for all loci, except RJ2N in Pst and Psh and RJ11N in Psh (Table 2). Most loci were almost fixed for heterozygosity. However, RJ8N was monomorphic in Pst and isolates appeared to be homozygotic at RJ2N. All loci except RJ8N, and RJ13N in Psh, deviated from HWE (Table 2). Although originally designed for Pst, all SSRs amplified multiple alleles within Psh. Thirty-two alleles were present in Canadian Pst isolates and 34 in Psh. Eighteen of the alleles were private to Canadian isolates of the formae speciales, with the majority occurring within Psh (Table 2). Two of the private Canadian Psh alleles were shared with the pre-2000 USA Pst isolates. Seven of the private alleles, four in Psh and three in Pst, were rare occurring only once or twice in the sample. The average Na per locus was near three for both formae speciales and the average Ne per locus was near two for both formae speciales (Table 2).

Table 2. Characteristics of the microsatellite loci used on Puccinia striiformis recovered in Alberta, Canada
LocusAll isolates (= 261)Puccinia striiformis f. sp. tritici (= 174)Puccinia striiformis f. sp. hordei (= 87)
N a a N e b H e c H o d N a N e P e H e H o N a N e P H e H o
  1. aNa: number of observed alleles; bNe: effective number of alleles; cHe: expected heterozygosity; dHo: observed heterozygosity; eP: number of private alleles.

  2. Significance levels for departure from Hardy–Weinberg equilibrium: *< 0·05; ***< 0·001; ns, not significant.

RJ421·900·4740·774***22·0000·4990·966***21·4600·3140·391*
RJ1322·000·5000·966***22·0000·4990·966***22·0000·4990·966***
RJ1722·000·5000·966***22·0000·4990·966***22·0000·4990·966***
RJ1843·810·7370·989***32·5520·6081***22·0010·4990·966***
RJ2182·640·6210·966***42·0220·5050·966***62·1740·5390·966***
RJ2464·160·7600·962***62·7220·6320·960***43·9200·7450·966***
RJ2N41·920·4790***21·0710·0670***31·0520·0450***
RJ8N21·210·1730·192 ns11·0000021·6910·4100·575***
RJ10N42·960·6620·973***22·0000·5000·977***43·8120·7370·966***
RJ11N64·250·7650·793***62·6610·6250·994***52·2700·5600·391***
RJ13N21·780·4390·651***22·0000·4990·966***21·0200·0230·023 ns
Mean (SE)3·82 (0·63)2·602 (0·316)  2·91 (0·51)2·002 (0·169)   3·1 (0·44)2·126 (0·288)   

The index of association was high for the combined analysis of all isolates (math formula = 0·58, < 0·001) as well as in each forma specialis (Pst, math formula = 0·65, < 0·001; Psh, math formula = 0·35, < 0·001). In all cases, the null hypothesis of random mating was rejected. After clone-correcting, the null hypothesis of random mating was still rejected (both ff. sp. math formula = 0·22, < 0·001; Pst, math formula = 0·55, < 0·001; Psh, math formula = 0·10, < 0·001).

Both formae speciales were dominated by two clones. Within Pst, a single genotype, T1, represented 68% of samples collected from Alberta as well as all samples from British Columbia and Manitoba (Tables 1 & S1). The two post-2000 USA samples, PST-078 and PST-114, also had this genotype. This genotype was the dominant Pst genotype recovered from 2007 to 2011, and the only genotype recovered in all years of sampling for either forma specialis (Table 1). A second genotype (T8) first recovered in 2011 represented 21% of the total Albertan Pst sample and was the dominant genotype in 2012 (Table 1). The third most common genotype (T11) was represented by four isolates recovered in 2011 as well as the USA pathotype PST-127. The 10 other Pst genotypes recovered were represented by only one to three collections and were recovered in only one or two years. The Psh genotype H1 represented 47% of all Psh collections and was recovered from 2007 to 2011. The second most common Psh genotype, H6, was identified in 28% of Psh collections and recovered from 2009 to 2012. The 13 remaining Psh genotypes were recovered only one to three times, and only in a single year with the exception of H2. For both formae speciales, all genotypes recovered more than twice were found in multiple counties (Table S1). All genotypes recovered more than once were predominantly recovered from the primary host of the forma specialis (Table 1).

Primers Q and Y produced five amplicons in the IGS1 region of the rDNA in each isolate, with each isolate possessing one to four amplicons (Fig. 1). All isolates with SSR genotypes T1, T2, T3, T4, and PST-035 amplified two bands of size 1353 and 1570 bp. Isolates with SSR genotypes T5, T6, T7, T8, T9, T10, and PST-045 amplified three bands of sizes 1353, 1438 and 1570 bp. Four fragments of sizes 1353, 1438, 1554 and 1660 bp were amplified in isolates with genotypes T11, T12 and T13. In Psh, all isolates with genotypes H6, H7, H8, H9, H10, H11 and H13 amplified a single band of size 1353 bp. Genotypes H2, H3 and H5 had fragments of size 1353 and 1438 bp. Isolates with genotype H1 and H4 had either bands 1353 and 1438 bp or only 1438 bp. The isolates with genotype H12, H14 and H15 had fragments of sizes 1554 and 1660 bp.

Figure 1.

PCR amplification of the intergenic spacer 1 (IGS1) of the ribosomal DNA of Puccinia striiformis isolates. Lane 1: genotypes T1–T4 and isolate PST-035; lane 2: genotypes T5–T10 and isolate PST-045; lane 3: genotypes T11–T13; lane 4: genotypes H6–H11 and H13; lane 5: genotypes H2, H3 and H5 and some isolates with genotypes H1 and H4; lane 6: some isolates with genotypes H1 and H4; lane 7: genotypes H12, H14 and H15.

The Bayesian clustering model implemented in structure indicated that K = 2 was the optimal grouping of the stripe rust genotypes. All Psh genotypes grouped together along with Pst genotypes PST-035 and PST-045 (Fig. 2). The remaining Pst genotypes were in the other cluster, Pst genotypes T11, T12 and T13 appeared to be significantly admixed between the two groups.

Figure 2.

Population structure of Puccinia striiformis inferred by model-based Bayesian cluster analysis of SSR data (calculated by structure and visualized by distruct). Each vertical bar represents a genotype and its proportional assignment into one of two clusters (K = 2). Genotypes are arranged by forma specialis (above) and SSR genotypes are listed below.

The first three eigenvectors from the PCoA explained over 82% (1st 43·8%, 2nd 23·0% and 3rd 15·3%) of the variation in the SSR data. The low diversity between most Pst genotypes recovered in Alberta was indicated by their tight clustering on the PCoA plot (Fig. 3). The Pst genotypes T11, T12 and T13 that appeared admixed by the structure analysis were well separated from the other Pst genotypes. The Psh genotypes appeared to be more diverse, with more diffuse clustering. The two reference samples of pre-2000 Pst grouped amongst several of the Psh genotypes.

Figure 3.

Principal coordinate analysis based on the multilocus genotypes at 11 SSR loci of Puccinia striiformis f. sp. tritici (squares) and Puccinia striiformis f. sp. hordei (circles). Percentages of the total variance explained by each axis are in brackets.

The UPGMA separated genotypes belonging to Pst isolates from Psh isolates with the exception of the two pre-2000 USA Pst pathotypes PST-035 and PST-045, which clustered amongst the Psh isolates, similar to the clustering of structure and PCoA (Fig. 4). Within formae speciales there were several clusters present, typically composed of SSR genotypes that shared the same IGS genotype. Cluster A and B contained the majority of Pst isolates recovered. Cluster C contained the three Pst genotypes that appeared to be admixed in the structure analysis and are related to PST-127. Cluster D contained the two USA samples representing the pathotype present before 2000. Cluster E contained only Psh isolates and were closely related to the Pst isolates of cluster D. Clusters F and G were also composed of Psh isolates. All the above clusters contained only a single IGS type with the exception of cluster D and F where two IGS types were present.

Figure 4.

UPGMA dendrogram of 30 Puccinia striiformis genotypes based on the shared allele distance calculated from the clone-corrected data from 11 SSR loci. Puccinia striiformis f. sp. tritici genotypes detected in Canada are labelled T1–T13. Two additional USA Pstriiformis f. sp. tritici isolates (PST-035 and PST-045) are shown. Puccinia striiformis f. sp. hordei genotypes detected in Canada are labelled as H1–H15. Number of isolates is given in brackets. A diagrammatic representation of the intergenic spacer 1 haplotypes present in the genotypes is given. The presence of a haplotype in all isolates (|) or only some isolates (:) of a genotype is indicated.

There was a strong positive relationship between genetic and pathotypic distance when both formae speciales were analysed together (= 0·866, < 0·001). Within each forma specialis, the relationship between genetic and pathotypic distance was strong in Psh (= 0·66, < 0·001) but weak in Pst (= 0·161, = 0·044). For the subset of 67 Pst isolates that were examined on a larger number of wheat cultivars the correlation was stronger (= 0·53, < 0·001).

Discussion

This study was undertaken to characterize the genetic variation in the populations of Pstriiformis collected within Alberta, Canada. The results indicated that the Pstriiformis population is dominated by a few widespread clones and there was no differentiation between the areas sampled. There were shifts in the frequency of the recovery of specific genotypes between years, and genotypes are associated with particular host species.

Multiple parameters indicate that stripe rust in Alberta is exclusively asexual or recombination effects are too rare to be detected. The frequent resampling of genotypes, low genotypic diversity, strong linkage disequilibrium, and the departures from HWE are all indicative of asexual reproduction by urediniospores. Additionally, the common occurrence of heterozygotes at microsatellite loci is indicative of the Meselson effect where alleles mutate independently over time in predominantly asexual species (Halkett et al., 2005). Studies in other regions have also indicated that stripe rust is typically clonal (Hovmøller et al., 2002; Bahri et al., 2009b).

The lack of differences between the regions sampled here is not surprising. The largest distance between sampling locations in any year was 433 km. Puccinia striiformis has been shown to establish clonal populations covering much larger areas in Europe (Hovmøller et al., 2002). In both formae speciales, there were rapid changes in frequency of the dominant genotypes near the end of the study. The reason for this is unclear. There was no change in sampling method and no known large shifts in cultivar use within Alberta. The changes in the local population probably reflect shifts in the source population further south in North America. Within the USA, multiple shifts in the frequency of Pst pathotypes have been detected over the past decade (Chen et al., 2010; Wan & Chen, 2012). Both the lack of regional differentiation and shifts in genotypes are probably in part as a result of the local climatic conditions. Local extinction is expected in winter, necessitating the area to be recolonized each successive year. Annual recolonization of the area would easily lead to founder effects and resulting genetic drift. This would prevent regional differentiation and increase the possibility of genotype shifts occurring on an annual basis.

Isolates previously identified as Pst always had genotypes different from isolates previously identified as Psh. The separation between the two formae speciales was not as clear as expected though. Bayesian and UPGMA clustering and PCoA all showed the grouping of the two pre-2000 USA pathotypes (PST-035 and PST-045) with Psh. Previous studies found that the two formae speciales form separate groups based on molecular markers (Chen et al., 1995; Spackman et al., 2010). It is possible that the Psh genotypes (cluster E) that clustered near the pre-2000 Pst isolates are the result of gene flow between Pst and Psh. In Australia, Psp-h has been shown to be more closely related to Pst than Psh using SSR markers (Spackman et al., 2010). However, SSR analysis of Pstriiformis found that Psp-h isolates had private alleles not found in Psh or Pst, and combinations of alleles at loci not found in Pst or Psh. Despite six of the SSR loci used by Spackman et al. (2010) being employed here, this was not the case. No alleles were private to cluster E and all combinations of alleles at each locus had been detected in 2007 or 2008 in Psh or Pst, suggesting these isolates are not Psp-h. This suggests that some form of gene flow has occurred between Pst and Psh and these isolates are the result. All isolates in cluster E that were phenotyped were found to be virulent to multiple wheat and barley differentials (Holtz et al., 2013). This led to the hypothesis that these isolates may be the result of hybridization between members of the two formae speciales. This apparent hybridization of Pst and Psh was detected independently earlier in the USA (Cheng & Chen, 2009). Although no specific attempts were made to determine the parentage of these Psh isolates, their closer association in the clustering analyses with the two genotypes that were included to represent the pre-2000 USA Pst population suggests that they are probably the result of gene flow between Pst and Psh.

Evidence of intraspecific hybridization in Pstriiformis is not new, but previously was thought to occur only within a single forma specialis (Little & Manners, 1969; Newton et al., 1986). The ability of gene flow to occur between the two formae speciales could potentially alter the evolution of the pathogen. Currently, the incidence of Pst has been much more severe than that of Psh in Alberta. This has been attributed to Pst strains that are significantly more aggressive. If it is possible for the genes that cause this increased aggressiveness to be transferred to Psh, it could result in more severe barley stripe rust.

Pst was less diverse than Psh. This lack of diversity is a result of the Pst population in Alberta being dominated by the more aggressive post-2000 strains and the absence of any pre-2000 genotypes related to PST-035 or PST-045. Initial studies of the invasive strains found them to be very low in molecular polymorphisms and extremely hard to distinguish. The presence of closely related genotypes within cluster A indicated that the invasive stripe rust has continued to mutate within North America. The sudden decline in the most common Pst genotype (T1) in 2012 and the dominance of genotype T8, that was first detected in 2011, is unexplained. It is probably as a result of genetic drift arising from a small founding population. The genotypes similar to T8 in cluster B also did not occur until 2011 or 2012 and all shared an IGS type distinct from that in genotype T1 and the other genotypes in cluster A. Members of both clusters have been pathotyped, but there was no clear differentiation between the pathotypes in each cluster (Holtz et al., 2013).

Of the three apparently admixed Pst genotypes, one (T11) was the genotype of PST-127, the other two were quite similar. The appearance of admixture is not surprising. It has been suggested, based on its virulence phenotype, that PST-127 is the result of recombination between the members of the pre-2000 USA Pst population and the invasive strains that entered around 2000 (Chen et al., 2010). PST-127 and similar pathotypes have increased dramatically in frequency in the USA in recent years (Wan & Chen, 2012). Although not detected until 2007, PST-127 and similar pathotypes accounted for over 28% of Pst isolates sampled in the western USA in 2009 (Wan & Chen, 2012). Despite this, genotypes similar to this pathotype were much rarer in Alberta. They were not detected until 2011 and were not detected again in 2012. This may suggest that a significant portion of the inoculum present in Alberta does not arrive from the Pacific Northwest as assumed or that PST-127 and similar pathotypes are poorly adapted to conditions in Alberta and therefore are under-represented. Unfortunately, as no sampling could be performed further south in the source population of stripe rust, it is not possible to determine if the genotype frequencies found in Alberta are representative of the source population.

The presence of multiple IGS1 haplotypes within individual stripe rust isolates has been demonstrated several times (Roose-Amsaleg et al., 2002; Spackman et al., 2010; Wang et al., 2012). The easily resolved IGS1 polymorphisms between the formae speciales and subgroups within the formae speciales could be of use in the diagnosis or monitoring of stripe rust populations, as has been suggested elsewhere (Spackman et al., 2010; Wang et al., 2012). The intensity of the IGS1 haplotypes amplified within some isolates varied. For many isolates with genotype H1, the larger IGS1 band amplified much more strongly than the smaller IGS1 band. This could indicate a difference in copy number within the isolates or a mutation in the primer binding site reducing the amplification of one of the haplotypes. Despite their ability to distinguish the formae speciales here, this does not necessarily mean that they would work in other stripe rust populations. If hybridization or sexual recombination occurs within a stripe rust population the IGS1 haplotype patterns may not be stable over time. Obviously, this necessitates the use of additional markers to characterize a population reliably.

Comparison of the genetic and pathotypic distances showed a very strong correlation when members of both formae speciales were analysed at the same time. This is expected as both the genetic markers used here and previous characterization based on virulence clearly separated Pst from Psh (Kumar et al., 2012; Holtz et al., 2013). Similarly, the strong correlation within Psh reflects previously detected similarity between the genetic clustering of isolates and their virulence phenotypes. The ‘hybrid’ Psh isolates in cluster E have been shown to be virulent on fewer barley cultivars and more wheat and triticale cultivars then the Psh isolates of cluster F (Holtz et al., 2013). In contrast, the Psh isolates in cluster G are virulent on more barley cultivars than the other Psh isolates, but are unable to infect any wheat cultivars (Holtz et al., 2013). The low correlation between genetic and virulence data within the entire phenotyped Pst sample is not surprising. Almost all Pst isolates analysed had highly similar genotypes and virulence phenotypes, neither of which appear related. The stronger relationship within Pst when the isolates from 2010 and 2011 were reanalysed is probably a result of the presence of the isolates in cluster C, the additional wheat cultivars used to determine virulence and the reduced number of other Pst isolates in the analysis. The additional wheat cultivars, some of which were only susceptible to isolates from cluster C, were added specifically to identify newer pathotypes that were known to exist in the USA (Holtz et al., 2013).

In conclusion, analysis of the genetic diversity of Pstriiformis showed that the pathogen is highly clonal with low diversity. The contemporary populations of the two formae speciales were easily distinguished from each other, although a significant proportion of the Psh isolates appeared to be closely related to the historical pre-2000 Pst population.

Acknowledgements

The authors are grateful to L. Langford for technical assistance with the multiplication and isolation of the pathogen and M. Wilson for assistance with the DNA extraction. The authors thank Dr Xianming Chen for providing the DNA of isolates from the USA, Dr Dean Spaner for the isolates from British Columbia and Dr Brent McCallum for the isolates from Manitoba. The funding provided by Alberta Crop Industry Development Fund for this project is gratefully acknowledged.

Ancillary