LlU Tai-guo, GE Run-jing, MA Yu-tong, LlU Bo, GAO Li, CHEN Wan-quan

State Key Laboratory for Biology of Plant Diseases and Insect Pests, Institute of Plant Protection, Chinese Academy of Agricultural Sciences, Beijing 100193, P.R.China

Abstract Puccinia triticina, the causal agent of wheat leaf rust, is one of the most devastating rust fungi attacking wheat worldwide.Seventy-six isolates of the wheat leaf rust pathogen from Yunnan, Sichuan, Gansu and Henan provinces, China, were tested on wheat leaf rust differentials and the population structure was analyzed using four presumably neutral partial sequence markers such as elongation factor-1α (EF-1α), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), β-tubulin (TUB) and the second largest RNA polymerase subunit (RPB2). The phenotypic diversity of Yunnan and Sichuan populations was higher than that of Gansu and Henan populations. The four populations were separated into two clusters based on the pathogenic data. A total of 12 single nucleotide polymorphisms (SNPs) and 32 haplotypes were identified among the four sequences.The 32 haplotypes were divided into two clusters in a neighbor-joining tree. Bayesian analyses also identified two clusters.Pairwise FST between populations in different regions were significantly different (P＜0.05). Analysis of molecular variance(AMOVA) indicated that 68% of the total genetic variation was within populations.

Keywords: population structure, polymorphism, virulence, wheat leaf rust, Puccinia triticina

1. lntroduction

Puccinia triticina, the causal agent of wheat leaf rust, can be a devastating pathogen under favorable conditions. Leaf rust is a common and widely distributed disease, occurring almost worldwide.P.triticinais an obligate parasite and needs an alternate host to complete its full macro cyclic life cycle (Ordo?ez and Kolmer 2007; Kolmer 2013), but its survival under agricultural conditions is almost entirely clonal. The damage caused by leaf rust on wheat is less than that caused by yellow rust and stem rust, but it possibly causes more yield losses on an annual basis because of its prevalent and widespread distribution (Huerta-Espinoet al.2011). Yield reductions caused byP.triticinaeach year can be about 5–15%, and sometimes more (Kolmer 1996; Ordo?ez and Kolmer 2007). In China, leaf rust mainly occurs in the southwest and northwest, and middle and lower reaches of the Changjiang River, and south of the Yellow River, Huai River and Hai River. Destructive epidemics occurred in 1969, 1973, 1975 and 1979 (Liet al.2010;Huerta-Espinoet al.2011).

Resistant cultivars are the most effective, economic, and environmentally friendly way to control leaf rust (Dadrezaieet al.2013).P.triticinapopulations are genetically diverse both in terms of pathogenicity (avirulence/virulence) and genetic markers, and mutation is the main source of variation(Kolmer and Ordo?ez 2007). A new resistant cultivar may lose its effectiveness within a few years post-release due to selection of previously uncommon variants of the brown rust or due to a new virulence caused by mutation (Ordo?ezet al.2010).P.triticinaundergoes several cycles of clonal reproduction by means of urediniospores every growing season. Urediniospores are air-borne and the spread is wind-assisted, sometimes up to hundreds of kilometers,and important means of spread of new races from one region to another (Kolmer and Ordo?ez 2007). Thus, an understanding of the genetic diversity, population structure,and relationship between pathogenic variation and genetic diversity withinP.triticinais essential for resistance cultivars breeding and leaf rust control.

Population analysis ofP.triticinacan be differentiated by classical pathogenicity (race) surveys or by molecular markers. Race surveys allow determination of the frequencies of avirulence or virulence with respect to wheat genotypes carrying known or unknown genes for resistance.Avirulence or virulence phenotypes are affected by environment (Van de Wouwet al.2010), and actual human interpretation of the contrasting phenotypes. Moreover,sampling involves a tiny portion of a potentially very large clonally propagating population (Kolmeret al.1995). With the development of molecular markers, population analysis based on various marker types, such as restriction fragment length polymorphism (RFLP), random amplified polymorphic DNA (RAPD), sequence characterized region (SCAR), simple sequence repeat analysis (SSR), and amplified fragment length polymorphism (AFLP) became possible (Kolmer 2013), while single nucleotide polymorphism (SNPs) is now being increasingly utilized following recent developments in sequencing technology (Johnson 2009). The first five molecular techniques depending on electrophoresis are effective for DNA polymorphisms, but it is not that utility in phylogenetic analysis because common ancestries are not indicated by the same bands. While SNPs can infer the historical process of contemporary populations by using in more complex population genetic model (Parkset al.2009).Compared to image-based technology, multi-locus sequence analysis is rapid and usually depends on housekeeping genes that are under stable selective pressure. At present,the SNPs researches on animal (Adamset al.2014), insect(Taerumet al.2013), bacteria (Bakhshiet al.2014) and fungi(O’Donnellet al.2013; Woudenberget al.2014) are well documented. Considerable research on fungi demonstrates the use of multi-locus sequences for phylogenetic analysis.Huanget al.(2009) analyzed partial sequences of theβ-tubulin gene and demonstrated thatβ-tubulin sequence markers were more effective in phylogeny studies inSaccharomycesisolates than 26S rDNA. In a study of the wheat powdery mildew fungus, partial sequences of the oxidase, protein kinase A, protein phosphatase type 2A andβ-tubulin genes were analyzed to examine geographic subdivision and the recent common ancestries.Blumeria graminisf. sp.triticipopulations in eastern United States of America, the United Kingdom and Israeli were divided into four populations consistent with geographic origin andB.graminisf. sp.triticipopulations of east USAwas compromised of northern and southern subpopulations,which indicated that ‘New populations’ were isolated from the‘Old ones’ by genetic drift (Parkset al.2009). The different geographic distributions were clearly revealed by dataset of internal transcribed spacer (ITS), nuclearLSU,rpb2andmtSSUmarkers, which was very useful for redefining morphological characters of subgroup ofEntolomaand the new insights to the classical phylogeny were also provided.In addition, new taxa were described, and new combinations were made by these multigene locus analysis at some cases(Morgadoet al.2013). InFusariumphylogenetic diversity and evolutionary relationships between 20Fusariarelatives were analyzed by partial sequences of RNA polymerase II large subunits (RPB1andRPB2) (O’Donnellet al.2013). To our knowledge, phylogenetic analysis using multi-sequence variation has few reports inP.triticinapopulations before 2013 (Liuet al.2013, 2014).

In this study, the population structure of 76P.triticinaisolates collected from Yunnan, Sichuan, Gansu and Henan provinces of China were analyzed using four partial gene sequences of the elongation factor-1α (EF-1α),glyceraldehyde-3-phosphate dehydrogenase (GAPDH),β-tubulin (TUB) and the second largest RNA polymerase subunit (RPB2) along with avirulence/virulence information.Two hypotheses were made. First, we hypothesized that the structure of theP.triticinapopulation might vary geographically, and that most differences would be between northern and southern populations. The second hypothesis was that a close relationship existed between avirulence/virulence phenotype and molecular genotype of theP.triticinaisolates. Our objectives therefore included:(i) to determine the pathogenic diversity of theP.triticinapopulations from different provinces; (ii) to examine the association of pathogenic variation and molecular genotype in the sampled populations; (iii) to determine if there are‘hot-spot’ areas for long-term survival ofP.triticinain China similar to that shown forP.striiforms(Chen and Duan 2014).

2. Material and methods

2.1. P. triticina isolates and pathogenicity test

Seventy-sixP.triticinasamples were collected from Yunnan(n=18), Gansu (n=16), Sichuan (n=22) and Hennan (n=20) in 2012 (Appendix A). For the samples from Yunan Province,three were from Dehong District, 10 from Lincang and five from Kunming on the lines of trap nursery plant. There was one sample collected from Nanchong, one sample collected in Meishan, 19 samples collected in three counties of Mianyang, two samples collected in Guangyuan of Sichuan Province when the investigation was carried out by our colleagues or collaborators. In Gansu Province, one sample was collected in Pingliang, five samples were collected from Dingxi, 10 collected from three counties of Tianshui, where was one of the hotspot areas in stripe rust epidemiology. In Henan Province, a central part of China, two samples were collected from Nanyang, one collected from Jiaozuo, seven samples from Luohe, 10 from three counties in Sanmenxia.Single pustule-derived cultures were established and increased according to Liu and Chen (2012) except that we used 3–5 mg of urediniospores and Isopar L as the dispersal agent. The differentials compromised Thatcher near-isogenic lines containingLr1,Lr2a,Lr2b,Lr2c,Lr3,Lr3bg,Lr3ka,Lr9,Lr10,Lr14a,Lr14b,Lr15,Lr16,Lr17,Lr18,Lr19,Lr23,Lr24,Lr26,Lr29,Lr30,Lr32,Lr33,Lr36andLr38, and several cultivars with known leaf rust resistance genes, including Hussar, Thew, Transec, Gatcher, CM2D-2M which carryLr11,Lr20,Lr25,Lr10+(Lr27+31),andLr28,respectively. The wheat differential sets were planted in a greenhouse and infection types were recorded 10–12 days post inoculation on a 0–4 scale with infection type (IT) 0–2+recorded as low and IT 3–4 considered to be high (Liu and Chen 2012). Races were designated according to the North American system (Long and Kolmer 1989).

2.2. DNA extraction

Uredeniospores of the differentP.triticinaisolates were used for DNA extraction by a modified CTAB/SDS method.The procedure was as follows: 10 mg urediniospores were put into 2-mL tubes and ground by shaking with 0.4 g(Φ=1 mm) and 0.2 g (Φ=0.1 mm) glass beads in a FastPrep 24 shaker (MP Biomedicals, USA) for 1 min at the speed of 6.5 m s–1. CTAB (1.54 mol L–1NaCl, 0.165 mol L–1Tris-HCl(pH=8), 66 mmol L–1EDTA (pH=8), 600 μL 1.1% CTAB,60 μL of 10% SDS, and 8 μL of 20 mg mL–1proteinase K were added, mixed gently and incubated in water at 65°C for 1 h (the tubes were gently mixed every 20 min). After adding of 330 μL chloroform/isoamyl (24:1) and 330 μL phenols, the mixture was inverted for 2 min and centrifuged at 12 000 r min–1for 10 min at 4°C. The supernatant was moved to a new Eppendorf tube and extracted a second time with 500 μL chloroform/isoamyl (24:1). The aqueous phases were transferred to 1.5-mL tubes and 5 μL RNase solution was added after centifugation. The mixture was kept at 37°C for 1 h and then twice extracted with chloroform/isoamyl(24:1). A two-thirds volume of cold isopropanol was added and kept in –20°C overnight before centrifuging for 10 min at 12 000 r min–1and 4°C. The pellet was twice washed with 70% cold ethanol in 1 000 μL. After drying, 30 μL dd H2O at 4°C was added to dissolve the DNA. The stock DNA solution was diluted to 50–100 ng μL–1for PCR amplification.

2.3. Primer design, DNA amplification, cloning and sequencing

Sequences obtained from the Broad Institute, Cambridge,MA, USA (http://www.broadinstitute.org/) and NCBI (http://www.ncbi.nlm.nih.gov/) websites were used in Primer Premier 5 to design. One or more introns were included for increasing sequence polymorphisms (Table 1). PCR amplification was performed using the following reagents: 100 to 200 ng DNA,10 μmol L–1primers, 10 μL 5× PCR buffer, 2.5 mmol L–1each dNTP and 1 U Primerstar HS DNA polymerase, with ddH2O added to 50 μL. The amplification program included one cycle for 15 s at 98°C, 35 cycles for 10 s at 98°C, 30 s at 55°C, 3 min at 72°C, and a final cycle of 10 min at 72°C.The amplified PCR products were separated in 2% agarose gel by electrophoresis. Selected bands were recovered by an EasyPure Quick Gel Extraction Kit (TransGen Biotech,Beijing). Recovered products were connected to an pEASYTM-Blunt Cloning Vector and transferred to Trans-T1 Phage Resistant Chemically Competent Cells (TransGen Biotech). Three clones of each sample were selected for sequencing (Sangon Biotech, China).

2.4. Data analysis

After removing the vector sequence, the four gene sequences were combined. The combined sequences were converted to a multiple alignment using BIOEDIT 7.0.9.0 (Hall 1999). All polymorphic positions were then visually inspected to edit the final alignments. With large numbers of genetic polymorphisms in natural populations,neutrality tests can be used to detect whether individual polymorphisms in DNA sequence are consistent with neutral model of selection. Neutrality tests statistics of the four regions including Tajima’sD, Fu and Li’sD,Fu and Li’sF,and Fu’sFSwere calculated by DNASP v5.0 (Librado and Rozas 2009).

Estimates of the number of variable sites (S), number of haplotypes (H), nucleotide diversity (Pi) and haplotype diversity (Hd), and distribution of haplotypes in each province were computed by DNASP v5.0 (Librado and Rozas 2009). Phylogeny based on haplotypes was analyzed by the neighbor-joining (NJ) method (Saitou and Nei 1987) in MEGA 4.0 (Tamuraet al.2007). One thousand replicationswere run to support the analysis. According to the Euclidean squared distance matrix calculated by ARLEQUIN v3.11(Excoffieret al.2005), a minimum spanning trees was computed to construct a haplotype network (Kruskal 1956;Prim 1957) using HAPSTAR (Teacher and Griffiths 2011) in order to analyze the evolutionary relationships among the generated haplotypes.

Table 1 Primer sequences for amplifying the single-nucleotide polymorphisms in the four reference candidate genes

The possible population structure was deduced from sequence data by STRUCTURE v2.2 (Pritchardet al.2000;Falushet al.2003) according to a Bayesian clustering model, and that the population is in linkage equilibrium in accordance with the Hardy-Weinberg law (Pritchardet al.2000). As a result of potential migration from one place to anotherviawind, the ‘a(chǎn)dmixture model’ was used in calculation (Falushet al.2003; Lindeet al.2010) and the allele frequency correlated by 100 thousand burn-in periods and 500 thousand Monte Carlo Markov Chain (MCMC)reactions were repeated in order to determine the optimalKand ΔKvalues based on Evannoet al.(2005). The number of subpopulationsKwas defined as 1–10 and 10 interactions were run for eachK. The graphical plots were generated by Distruct (Rosenberg 2004).

Analysis of molecular variance (AMOVA) was computed by ARLEQUIN v3.11. The genetic variance can be differentiated within or among a population when AMOVA is calculated (Excoffieret al.1992). The four populations were divided into two regions (a southern region including the Yunnan and Sichuan populations, and a northern region including the Henan and Gansu populations) to test the genetic diversity among regions, among populations within regions, and within populations. 1 023 haplotype permutations were compared in computing the variance components and Wright’sFSTa standard parameter for testing differences in genetic variance among populations.Population migrations were estimated using Wright’s method based on the formula Nm=(1–FST)/4FST(Wright 1949). To evaluate correlations between pathogenicity and sequence genotype of the fourP.triticinapopulations the co-efficients of genetic differentiation according to virulence and fixation index based on sequence differences between populations were used to do correlation analysis in SAS 8.1 (Statistical Analysis Institute, Inc., Cary, NC).

3. Results

3.1. Pathogenicity of P. triticina

In total, 41 avirulence/virulence (race) phenotypes were detected on 31 differentials (Appendix B). Virulence frequencies forLr9,Lr19,Lr24,Lr25,Lr28,Lr29,Lr38andLr42were lower than 30% in all four populations. A large number of isolates were virulent on differentials possessingLr2b,Lr2c,Lr10,Lr14a,Lr14bandLr33at frequencies greater than 75%.Virulence frequencies were more variable between populations in regard to differentials containingLr2a,Lr3a,Lr11,Lr18,Lr20,Lr27+31andLr36(Appendix B). Based on the collected samples, there were 39 races identified in total, and 14 in Yunnan, 18 in Sichuan, 11 in Gansu and 14 in Henan (Appendix A). TheP.triticinapopulations had high virulence diversity (Table 2). The Sichuan population had the highest virulence diversity, and the lowest virulence diversity was detected in the Henan population. Nei’s unbiased genetic distance and genetic identity indicated that Yunnan and Sichuan populations had a high genetic identity on virulence; isolates in these two populations exhibited similar virulence frequencies for different resistance genes. Gansu showed a high genetic difference to the other populations (Table 3). The UPGMA dendrogram (Fig. 1) according to Nei’s genetic distance clearly revealed the virulence distances separating the fourP.triticinapopulations.

3.2. Neutrality tests and molecular diversities of P. triticina

Neutrality tests of SNP markers showed that Tajima’sD,Fu and Li’sDandF(excess of ancient mutations) were not significantly greater than zero in all four populations,but Fu’sFS(excess of recent mutations or rare alleles) in the four populations was significant (Table 4). The four genes were amplified and successfully sequenced for all 76 cultures. Twelve SNPs were detected in the four genes among which one was identified inEF-1α, three were inGAPDH, four inTUBand four in theRPB2. They were all information-providing single-nucleotide mutations, and occurred in at least two haplotypes. No insertions and deletions contributed to the variation. Haplotype diversities were all high but nucleotide diversities were low in all four populations (Table 5).

Table 2 Virulence diversity parameters of Puccinia triticina populations1)

Table 3 Nei’s unbiased genetic distances and genetic identities based on virulence data

Fig. 1 UPGMA dendrogram based on virulence.

There were 32 haplotypes in all, with 11, 12, 12 and 13 documented in Yunnan, Sichuan, Gansu and Henan,respectively. H7, H8 and H9 were detected only in Yunnan;H13, H16 and H17 were detected only in Sichuan; H19, H20,H21, H24, H25, H27, H28 and H29 were detected only in Gansu; and H30, H31 and H32 were detected only in Henan.No haplotype was common to all four provinces. Yunnan and Sichuan shared five, Sichuan and Henan shared five,and Gansu and Henan shared four; lower numbers were shared between other pairs of provinces (Table 6, Fig. 2).

PairwiseFSTbetween two populations indicated no significant genetic differentiation between Yunnan and Sichuan or between Gansu and Henan populations,whereas other two-way comparisons’ between populations were significant. Where populations were not significantly differentiated since Nm was low, and apparently high gene flow was detected between populations (Table 7).

NJ phylogeny showed that the 32 haplotypes were divided into two main groups. H10, H11, H16, H5, H13, H17,H1, H6, H9, H8, H7, and H32 clustered in one clade, and the remaining haplotypes were in another. The haplotype phylogeny showed a partial relationship with geographical origin. The majority of haplotypes in the second clade was from Gansu and Henan, and only seven (H3, H14, H2, H12,H15, H18 and H4) were from Yunnan and Sichuan and were shared with Henan or Gansu. In the first cluster, only H32 was from Henan; the rest were all from Yunnan and Sichuan populations (Fig. 2). There was an obvious link betweengeographic location and related haplotypes.

Table 4 Neutrality tests of Puccinia triticina populations

Table 5 Molecular diversity indices of Puccinia triticina populations

Table 6 Haplotype patterns in Puccinia triticina populations from different locations

To further test the relationships of the 32 haplotypes, a HAPSTAR network was built (Fig. 3). This network showed a star-like structure with H29 and H14 in the center. Black dots in the figure represent the haplotypes not found, but probably existing in the population, thus indicating multiple mutational steps between identified haplotypes. Haplotypes in the second clade of NJ phylogeny were distributed across the entire network, whereas most of the haplotypes in the first clade were in the center of the figure. In general, most of the haplotypes in Gansu and Henan populations were ancient haplotypes and most of the Yunnan and Sichuan haplotypes were young. These indicating that the potential selection on resistant cultivars was more important in the south, for example in Sichuan and Yunnan, where CIMMYT materials containing resistant genes might have had a greater influence on yield and quality breeding (Zouet al.2007; Zhanget al.2011) and perhaps on the dispersal or migration of the pathogen. And there were also many derivatives of 1BL.1RS which was introduced in the 1970s(Heet al.2001).

STRUCTURE analysis based on Bayesian modeling indicated that the optimumK-value was 2, which was strongly supported by ΔK. MCMC convergence and consistency among runs were verified. Individuals with mixed genotypes from different clusters indicated remnant genetic similarity between recently diverged populations or immigrant ancestry (Figs. 4 and 5). Populations from Yunan and Sichuan constituted one cluster, and those from Gansu and Henan formed another. The results of correlation analysis showed that theR2was 0.66463, and theP-value was greater than 0.05 indicating that the correlation between avirulence/virulence phenotype and molecular genotype was not significant.

Fig. 2 Neighbor-joining phylogenetic tree of multi-locus sequence haplotypes of Puccinia triticina samples.

4. Discussion

4.1. Pathogenicity tests showed virulence of P. triticina in Sichuan and Yunnan population differed with that of Gansu and Henan

Avirulence/virulence with respect to resistance genesLr2a,Lr3,Lr11,Lr18,Lr20,Lr27+31andLr36was polymorphic across all four populations. Therefore none of these genes would provide effective control of leaf rust if used alone incultivars, but might play a role if deployed in combinations with other genes. All four populations had high pathogenic diversity, but it was higher in the Sichuan and Yunan populations than that in the Gansu and Henan populations.This was probably due to the climate was cool in summer and warm in winter in Yunnan and Sichuan which was beneficial to the occurrence and development of wheat leaf rust. Yunnan and Sichuan were the areas thatP.triticinaoften occurs. The south has a long-term permanent population because of the environment whereas the northern population is more transient. It was reported that virulence diversities originated from universal resistances,especially for the biotrophic fungi wheat interactions as gene-for-gene examples (Thrall and Burdon 2003).

Table 7 Pairwise FST and Nm between geographical Puccinia triticina populations

4.2. Molecular diversities and populations structures were clearly demonstrated by multi-locus sequences analysis

FSTwas not different between the Yunnan-Sichuan and Henan-Gansu populations, in contrast to significant differentiation reported between other populations caused largely by geographical barriers. Kolmer and Ordo?ez(2007) showed that geographically based genetic differentiation existed inP.triticinapopulations. The distance between Henan and Gansu is further than the distance between Gansu and Sichuan populations although a genetic differentiation existed between Gansu and Sichaun and no genetic differentiation existed between Gansu and Henan indicating geographical distance was not the main effector inducing geographical barrier.

Nm was detected among all theP.triticinapopulations.It was reported that genetic diversity existed among populations when Nm＜1 (Allendorf 1983). Except for Yunnan-Sichuan and Henan-Gansu, all the Nm comparisons were below 1. Thus the infrequent exchange between populations was a factor in population divergence.

AMOVA showed that the largest component variance(68.13%) was within populations, whereas about one-third came from between populations. However, both variance components were significant (Table 8). This was consistent with Dadrezaieet al.(2013), who examined population structure of 100 isolates collected from 14 provinces in Iran using AFLP markers. Thus, the geographical barrier and infrequent exchange were not only the main elements causing genetic differentiation.

Fig. 3 Minimum spanning network of Puccinia triticina haplotypes.represent haplotypes that were not found in this study, but may exist in the population.

Fig. 4 Comparison of ΔK values for the optimal K-value from structure.

Fig. 5 The clustering of 76 isolates by STRUCTURE 2.2.3.Each vertical bar represents one isolate and is partitioned into K colored parts indicating the probability that the isolate is the part of K clusters. The optimal K-value was 2.

Sub-populations of theP.triticinapopulation showed a high level of genetic differentiation. The MEGA 4.0 program divided all the haplotypes into two clades and STRUCTURE indicated an optimalK-value of 2. The northern populations were more clearly differentiated than the southern populations. Genetic structure of populations is affected by mutation, population size, random genetic drift,method of reproduction, gene flow and selection (McDonald and Linde 2002). Mutation directly results in variation in DNA sequence and thus generates new alleles/haplotypes,which are the basic sources of genetic variation. Ordo?ez and Kolmer (2009) found that mutation was important inP.triticinapopulations, leading to both SSR allele and single virulence differences among isolates. Population size also influences gene diversity through gene drift, in which large population size will more likely cause potential new mutations than small populations (McDonald and Linde 2002). A mixed reproduction existed inP.triticina, including both sexual and asexual reproduction. In Morocco, leaf rust isolates had sexual and asexual production stages, and sexual stage is on the alternative host Anchusa (Bouftasset al.2010). But the sexual stage did not play a significant role in disease spread and source of genetic variation for the lack of suitable alternative hosts (Kolmer 2005). The relative high Nm values within regions and low values between regions was another factor inducing the subpopulations ofP.triticina. Morever, selection by individual host genotypes has a significant effect on genetic diversity inP.triticinawhich was higher on common wheat than on durum (Ordo?ez and Kolmer 2007) probably due to more widespread deployment of single resistance genes in wheat.

Table 8 Analysis of molecular variance (AMOVA) of wheat leaf rust populations (Yunnan and Sichuan were regarded as the southern region, and Gansu and Henan were regarded as the northern region)

4.3. Haplotype network showed that Gansu and Henan populations of P. triticina were more ancient than the other populations

The network of haplotypes revealed that most of the haplotypes in Gansu and Henan populations were ancient haplotypes and most of Yunnan and Sichuan haplotypes were more recent. This suggests thatP.triticinain the four populations had the same origin. H14 and H29 were older than other haplotypes.Haplotype diversity based on DNA sequence was high, but nucleotide diversity was low in all four populations, consistent with a population bottleneck and rapid population growth (Althoff and Pellmyr 2002). Neutrality tests showed that Fu’sFSwas significant,indicating population expansion (Fu 1997).

Although the correlation for virulence and molecular genotype was high, no significant correlation between virulence phenotypes and sequence genotypes was found based on our data. Chenet al.(1993) reported that the evolution of the whole genome of the pathogen was faster than that of the individual genes controlling pathogenicity. The four partial DNA sequences may have different evolutionary rates relative to pathogenicity genes. In addition, the virulence phenotype was based on the interaction between host and pathogen in the certain environment and probably influenced by human factors,whereas the four DNA sequences were not pathogenicity genes affected by host cultivars. Furthermore, virulence phenotype was determined by pathogenicity and only represented a small part of genetic variation, it could not reflect the genetic relationship of isolates, however,neutral DNA marker could reveal the difference in nature.A correlation was found between the pathogenicity and the SSR genotype for the different individuals when it was analyzed for the wheat leaf rust populations from the Central Asia and Caucasus (Kolmer and Ordo?ez 2007).The different results compared to the present study might relate to sample location, number of collections and host genotypes.

5. Conclusion

The wheat leaf rust populations were highly diversified, both pathogenically and molecularly. The Yunnan and Sichuan populations had the higher pathogenic diversity than the Gansu and Henan populations, and the Gansu population stood alone based on the Nei’s genetic distance in the UPGMA dendrogram. Two clusters were generated by Bayesian analysis based on DNA sequence data. Significant genetic differentiation existed among populations from different regions. No significant relationship was found between pathogenicity and the molecular genotype in our study. We obtained no evidence to support a migration pathway forP.triticina; however, further study based on a larger number of samples is required for confirmation in next step. In order to better control the wheat leaf rust disease, it was important to trace the pathway of leaf rust epidemics.

Acknowledgements

We thank Prof. Robert McIntosh from the University of Sydney, Australia, and Dr. Anna Berlin from Swedish University of Agricultural Sciences for the earlier version and critical reviews of this manuscript. The financial supports by the National Natural Science Foundation of China(31671967), the National Key Research and Development Program from the Ministry of Science and Technology,China (2016YFD0300705), the National GMO New Variety Breeding Project, China (2014ZX0801101B) and the earmarked fund for China Agriculture Research System(CARS-3) were gratefully acknowledged.

Appendicesassociated with this paper can be available on http://www.ChinaAgriSci.com/V2/En/appendix.htm