Genetic diversity of the wheat yellow rust population in Pakistan and its relationship with host resistance

Authors


E-mail: bochra.bahri@hotmail.fr

Abstract

Yellow rust, caused by Puccinia striiformis f.sp. tritici (PST), is an important disease that threatens wheat production in Pakistan. This study was designed to explore the virulence and simple sequence repeat (SSR) diversity of the Pakistani PST population and the ongoing selective pressures of widely grown wheat cultivars. Analyses of 49 isolates sampled from the North-West Frontier Province of Pakistan led to the identification of 12 distinct pathotypes. The virulence frequencies of v2 (virulent against Yr2), v6, v7, v9, vSU and v27 ranged from 63% to 100%. Virulences v3, v4, v17 and vSD were uncommon, whilst v5, v10, v15, v24, v32 and vSp were not detected. The pathotypes thus described were then classified into 27 distinct genotypes. Bayesian structure analysis clustered these genotypes into five groups (in addition to one hybrid isolate) which represent three distinct lineages of the SSR-based phylogenetic tree. Of the studied isolates, 80%, represented by three predominant pathotypes (P1–P3), belonged to the same characteristic Pakistani lineage, whilst the other isolates were close to either a Mediterranean lineage or a Northern European lineage. Genetic recombination was detected within P2 isolates. Resistance genes postulated in 40 Pakistani wheat cultivars indicated the high frequency of Yr2, Yr6, Yr7, Yr9, Yr27 and YrSU. Only 11 cultivars were found to be resistant to P1–P3. Migration and varietal diversity factors might contribute to maintaining the currently high genetic diversity in Pakistani PST, and have serious regional implications for wheat improvement programmes.

Introduction

Wheat yellow rust is a disease that is present worldwide and causes high yield losses if it is not controlled by resistant cultivars or the application of fungicides. Molecular studies of Northwest (NW)-European and Australian populations identified two important biological features of Puccinia striiformis f.sp. tritici (PST): the pathogen shows clonal behaviour, but at the same time displays rapid adaptation to host resistance genes through mutations in virulence factors (Enjalbert et al., 2005; Hovmøller & Justesen, 2007; Wellings, 2007). If these PST populations are characterized by limited genetic variability, contrastingly high diversity has been observed for virulence and/or molecular markers in Central and West Asia (Hovmøller et al., 2008), with clear evidence of recombination in China (Mboup et al., 2009). How these differences in population structure are related to the geographic origin and migratory history of PST still needs to be elucidated. In that context, the study of Central and West Asian populations is of particular interest because not only is this the region of origin of cereals (accounting for more than 20% of the world’s harvested wheat (Singh et al., 2004)), but yellow rust has been recognized for decades as a major threat in the region.

In Pakistan, wheat is cultivated on more than eight million hectares, 70% of which are prone to yellow rust. Infestation is severe in the foothills in the north, but is also present in the central region and western upland areas (Hassan, 1968). Primary foci appear along the foothills in January, after which their spread and development are confined to the cooler parts of western Punjab and the North-West Frontier. Thirteen epidemics have been recorded since 1948. Twelve yellow rust epidemics were described by Afzal et al. (2008), whilst the most recent was reported by Duveiller et al. (2007). Numerous PST pathotypes have been reported in the country by different scientists (Ahmad, 2001; Hussain et al., 2004). Four major yellow rust epidemics were recorded in 1978, 1997–98 and 2005 and caused respective losses of US$244 million, $33 million and $100 million to the Pakistan economy (Hussain et al., 2004; Duveiller et al., 2007). The two more recent epidemics were caused by the successive emergence of pathotypes presenting new virulences, v9 and v27 (Duveiller et al., 2007), and corresponded to successive resistance breakdowns. The strong links between seed suppliers and public research institutions had led to the successive deployment of a few elite lines, which resulted in millions of hectares being planted with a limited number of cultivars. This situation provided ideal conditions for Pirsabak-85 and Pak81 (Yr9 breakdown in 1995: Ehsan et al., 2003) or for the selection of new virulences and the subsequent breakdown of cultivar resistance, as in the case of cv. Inquilab-91 (Yr27 breakdown in 2005: Duveiller et al., 2007). More generally, the resistance genes Yr2, Yr6, Yr7 and Yr9, either singly or in combination and postulated in old commercial cultivars of Pakistan (Kirmani et al., 1984; Perwaiz & Johnson, 1986), were no longer effective.

Even though the recurrence of devastating yellow rust epidemics has matched the resistance breakdown dynamic observed in most countries, the diversity of virulence is one of the broadest in the world (Ehsan et al., 2003), which places in doubt a narrow clonal origin for Pakistani PST isolates. To address this question, an analysis was performed of genetic diversity of virulences and molecular markers, concentrating on the North-West Frontier Province (NWFP) of Pakistan, which is prone to yellow rust epidemics because of its cool climatic conditions (Gupta & Paul, 2002). In addition, yellow rust resistance genes were postulated in 40 modern wheat cultivars and advanced lines, in order to identify the ongoing selective pressures on avirulence in Pakistan.

Materials and methods

Sampling and spore multiplication

The sampling of PST isolates was performed in 2006 in NWFP, an important wheat-growing area of Pakistan. This province covers an area of 74 521 km2 where fertile valleys are isolated by high mountains. The climate is extremely diverse in this province. The south is one of the hottest areas in the Indian subcontinent, whilst the weather in the mountainous northern region is temperate in summer and intensely cold in winter. The isolates were collected from two zones where yellow rust occurs regularly: the Peshawar valley of NWFP (i.e. Peshawar (coordinates: 34°00′N, 71°32′E) and Nowshara (coordinates: 34°04′N, 71°58′E)) and the uplands of NWFP (i.e. Abbotabad (coordinates: 34°09′ N, 73°13′ E) and Mingora (coordinates: 34°50′ N, 72°22′ E)).

During the growing season (March–April), infected leaves were sampled from farmed fields as well from the experimental plots of breeders (differential lines and released cultivars) and shipped by express mail service. Each sample was stored in a separate paper envelope to prevent contamination. All the regions were well represented, with 18 isolates from Peshawar, 13 from Nowshara and 17 from Abbotabad, although only one isolate was obtained from the Mingora area.

As low rates of sample recovery were experienced using single-stripe (linear uredinia) leaves in 2005, leaves covered with sporulating lesions were collected in 2006, and recovery success was almost complete. To prevent any mixing of isolates, each sample was inoculated at a low density. Plants bearing a single rust pustule on a leaf were isolated in a cellophane bag and spores collected from the single uredinium were increased. Spore production was performed in a high-containment, spore-proof and climate-controlled room to prevent any leakage of exotic spores into the environment. Diseased leaves were rubbed on the first leaf of 7-day-old seedlings of a mix of susceptible cultivars – Michigan Amber and Victo. The inoculated plants were then incubated in a dew chamber at 8°C for 16 h in the dark to ensure successful infection, and then transferred to a climate chamber (day: 16 h, 300 μmol quanta m−2, 17°C; night: 8 h, 14°C). High-intensity light treatment was applied to the seedlings prior to inoculation for at least 8 h in order to maximize infection success (de Vallavieille-Pope et al., 2002). A week after inoculation, each pot was sealed within a cellophane bag to avoid cross-contamination. Eighteen days post-inoculation, urediniospores were collected, dried in a desiccator at 4°C for 3 days, and stored in microtubes at −80°C. Multiplications were repeated until 15 mg of spores per isolate had been collected, this being the quantity necessary to perform pathotype and molecular analyses.

Phenotypic analyses

Pathotype analysis

Virulence combinations of yellow rust isolates were determined using the European and world sets of 15 differential varieties (Johnson et al., 1972), to which were added 17 wheat lines: Kalyansona (Yr2), Federation × 4 Kavkaz (Yr9), Clement (Yr9 + ), VPM1 (Yr17 + ), TP981 (Yr25), Anza (YrA), Early Premium, Jubilejina 2 and Australian isogenic lines (Yr1, Yr5, Yr6, Yr7, Yr8, Yr15, Yr24, Yr26 and Yr27) in the Avocet background (see http://www.ars.usda.gov/SP2UserFiles/ad_hoc/36400500Resistancegenes/Yrgene.xls).

This differential set was able to discriminate between 23 virulence factors. For each isolate, 5 mg urediniospores suspended in 300 μL mineral oil (Soltrol) were sprayed onto five seedlings (two-leaf stage) of each variety, and incubated as described above. The plants were scored individually 15–17 days after inoculation using a 0–9 scale based on the presence of necrosis and chlorosis and the intensity of sporulation (McNeal et al., 1971). Virulence reactions were defined as infection types 7–9. Virulence against one YrX gene was postulated when at least one cultivar among all those possessing YrX presented with a susceptible reaction (Table 1). In the text below, virulence against the YrX resistance gene is referred to as vX (AvX for avirulence).

Table 1.   Characteristics of Pakistani wheat yellow rust pathotypes sampled during 2005–6 in North-West Frontier Province (NWFP)
PathotypeNumber of isolatesVirulence factorsastructure groupNumber of genotypesGenotype diversitybPathotype diversitybSpecific SSR allelescPrevious reports
185(RJ21); 372(RJ9N)233 (RJ5N)354(RJ3N); 173(RJ21); 324 (RJ9N); 306 (RJ9N)300 (RJ20h)207 (RJ4); 304 (RJ20h)181 (RJ21)170 (RJ21); 268 (RJ4Nb); 451 (RJNh); 173 (RJ2N)188 (RJ2N)231 (RJ10N)175 (RJ21)
  1. aVirulence factors for the pathotypes. A, SU, SD and V refer to resistance genes of cvs Anza, Suwon92xOmar, Strubes Dickkopf and Victo, respectively. Two Chinese cultivars, Early Premium (EP) and Jubiliena (JU), carrying unknown resistance genes, were added.

  2. bSimpson’s diversity index.

  3. cNumber indicates the allele position (pair of bases) of the SSR marker given in brackets. Crosses indicate the specific presence of listed SSR alleles.

P1191267927SU3100·920·74XXXXXPakistan in 1969–76 (Ahmad, 2001)
P214267827ASU56XXEthiopia in 1990–2001 (Hakim et al., 2001), India in 1995–2004 (Prashar et al., 2007)
P31267827SU3/51XXXX
P32267827SU42XXXXPakistan in 1969–76 (Ahmad, 2001), India in 1995–2004 (Prashar et al., 2007)
P41267892527AEPV14XXXXNorth Africa, Iran in 2005−6 (Bahri et al., 2009b)
P52267925AEPVXXXXIran, Cyprus in 2005–6 (Bahri et al., 2009b)
P622678925AEPVXXXXLebanon, Turkey, Israel, Cyprus, Iran in 2005–6 (Bahri et al., 2009b)
P7126792527AEPVXXXXIran 2005 (Bahri et al., 2009b)
P8126782527AEPVXXXXNorth Africa, Middle East (Bahri et al., 2009b), Ethiopia, Pakistan in 1973 and 1979 (Hakim et al., 2001)
P91126792527ASUEPVXXXX
P10112367892527ASUEPJUV24XXX
P11112792527ASUEPJUVXXX
P121123469172527SUSDEPJUVXXXAlgeria in 2005–6 (Bahri et al., 2009b), NW Europe (de Vallavieille-Pope et al., 1990)
P121267927SUXXX
N1 (237E141-v17) 1234691725SUSDEPJUV2              
S1 (6E16-Av9) 2678V1              

Resistance gene postulation

Resistance genes were postulated for 40 bread wheat cultivars from Pakistan, which included 19 cultivars released during the period 1991–2006, whilst the remaining 21 were candidate varieties (Table 2). Inqilab-91 and Bakhar-2002 are the predominant cultivars grown in Pakistan, whilst Bakhtawar-92 and Tatara occupy the major acreage in NWFP.

Table 2.   Postulation of the resistance genes for 40 Pakistani bread wheat cultivars based on infection types of resistance (R) or susceptibility (S) to a set of 12 French wheat yellow rust pathotypes (N1–N10: northern France; S1 and S2: southern France) and three predominant Pakistani pathotypes (P1, P2 and P3)
CultivarOriginYears of releaseReactions to the pathotypes inoculatedaPostulated resistance gene
Northern FranceSouthern FrancePakistan
N1 (2)N2 (2)N3 (2)N4 (2)N5 (2)N6 (2)N7 (2)N8 (2)N9 (2)N10 (2)S1 (1)S2 (1)P1 (3)P2 (5)P3 (4)
  1. aNumber in brackets represents the SSR genetic group determined under structure software of the inoculated pathotypes. N1: 237E141-v17; N2: 40E8; N3: 43E138; N4:106E139; N5: 109E141; N6: 169E136-v17; N7: 232E137; N8: 233E169-v17; N9: 237E141; N10: 239E143-v17; S1: 6E16-Av9; S2: 6E16-v9 (see Materials and methods for virulence spectra).

  2. bPostulation not confirmed by the P2 Pakistani pathotype.

  3. cPresence of Yr3 or YrSD but not confirmed by the S2 southern France pathotype.

  4. dPresence of YrSU but not confirmed by the S2 southern France pathotype.

  5. -: Unidentified resistance gene or combination of genes.

  6. U: presence of unknown additional resistance gene.

  7. I: Cultivars from public breeding programmes based in NWFP; II: cultivars from the Punjab and Sindh provinces of Pakistan.

  8. MT: Candidate bread wheat lines which have completed or are undergoing 2 years’ mandatory testing at the national level and are expected to be released as future cultivars.

  9. R/S: Intermediate reaction observed in cultivars after inoculation of the PST pathotype.

Bhakkar 2002II2000–2SSSSSSSSSSSSSSSSusceptible to all pathotypes
Kirin-95II1991–6SSSSSSSSSSSSSSS
Blue SilverIIMTSSSSSSSSSSSSRSRSusceptible to all French pathotypes
Soughat-90II1991–6SSSSSSSSSSSSRSS
KT-2003I2000–6SSSSSSSSSSSSSR/SR
CT-00231IIMTSSSSSSSSSSSSRRR
Wafaq-2001II2000–6SSSSSSSSSSSSRSR
V-01078IIMTSRSSSSSSSSSSSRRYr2 + U
DN-47IMTSRSSSSSSSSSSSSR
Watan-94II1991–6SRSSSSSSSSSSRSR
V-002493IIMTSRSSSSSSSSRSRSR
7_03IIMTSRRSSRSSSSRRSSRYrSU + U
TW-0135IIMTSRRSSRSSSSRRSSS
CT-00062IMTRRRSRRRRRSRRRRRYr7 + Yr4
Bakhtawar-92I1991–6RRRRRRRRRSRRRRRYr7 + Yr9 + U
Saleem 2000I2000–6RRRRRRRRRSRRRRR
SARC-5IIMTRRRRRRRRRSRSSSRYr7 + Yr9b
SD-66II1991–6RRRRRRRRRSSSRRRYr6 + Yr7 + U
TararaI1991–6RRRRRRRRRSSSRRR
Suleman-96I1991–6SRRRSRRRRSSSSSSYr6 + U
PR-84IMTSRRRRRRRRSRSSRRYr6 + Yr9
V-9021IIMTSRRRRSSSSSRSSRRYr9 + U
V-00055IIMTRRRRRRRRRRRSRRRResistant to all Northern France pathotypes; Yr27 possible
DiamondIIMTRRRRRRRRRRRSRRR
RWM-9313IIMTRRRRRRRRRRRSSRR
Kohsar-93II1991–6RRRRRRRRRRRSRSR
KT-2000I2000–6RRRRRRRRRRRSRRR
Pirsabak-91I1991–6RRRRRRRRRRRSSRR
91BT010-84IIMTRRRRRRRRRRRSSSR
Inqilab-91I1991–6RRRRRRRRRRRSSSS
Nowshehra-96I1991–6RRRRRRRRRRRSSSS
V-02192IIMTRRRR/SRRRRRRSSSSS-
V-01180IIMTRRRSSRRRRRRSSSR-
SarsabzIIMTSRRSRSSSR/SSRSSSR-
Lakki J03I2000–6R/SR/SR/SSRSSSR/SSSRSSR-
Marwat J01II2000–6SRRRRRRRRRRSRRR-
99B4012IIMTRRRRRRRRRRSSSSR-
V-00125IIMTSSSSSSSSSSRSRRR-c
Shaheen-94II1991–6SRRRSRRSSSRSSSR-d
99B2278IIMTSRRSSRSSSSRSSSS-d

Resistance genes were postulated on the basis of cultivar reaction to a set of 10 northern France (N1–N10), two southern France (S1–S2) and three predominantly Pakistani (P1–P3) PST pathotypes displaying complementary virulences: 237E141-v17 (coded N1: v1, 2, 3, 4, 6, 9, 17, SD, SU), 40E8 (coded N2: v3, SD), 43E138 (coded N3: v1, 2, 3, 7, SD), 106E139 (coded N4: v2, 3, 4, 7, SD, SU), 109E141 (coded N5: v1, 2, 3, 4, 6, SD, SU), 169E136-v17 (coded N6: v1, 2, 3, 9, 17, SD, A), 232E137 (coded N7: v2, 3, 4, 9, SD, SU), 233E169-v17 (coded N8: v1, 2, 3, 4, 9, 17, 32, SD, SU), 237E141 (coded N9: v1, 2, 3, 4, 6, 9, SD, SU), 239E143-v17 (coded N10: v1, 2, 3, 4, 6, 7, 9, 17, SD, SU) (de Vallavieille-Pope et al., 1990), 6E16-Av9 (coded S1, v2, 6, 7, 8), 6E16-v9 (coded S2: v2, 6, 7, 8, 9, 25, 27, A), P1 (v1, 2, 6, 7, 9, 27, SU), P2 (v2, 6, 7, 8, 27, A, SU) and P3 (v2, 6, 7, 8, 27, SU). As most of these pathotypes present more than one avirulence factor, precise resistance gene combinations could not always be inferred. The pathotype codes and virulence spectra of the isolates were determined using the European and world sets of differential cultivars (Johnson et al., 1972) and additional cultivars with specific resistance genes: VPM1 (Yr17), Kalyansona (Yr2), TP981 (Yr25) and Avocet-Yr27 as described above for pathotype determination.

Molecular analyses

Genomic DNA was extracted from 10 mg urediniospores using a modified CTAB protocol (Enjalbert et al., 2002). After quantification on agarose gel, the isolates were genotyped for five previously reported simple sequence repeat (SSR) markers (RJ4, RJ18, RJ20, RJ21 and RJ24: Enjalbert et al., 2002) and 10 newly developed from a P. striiformis EST library published in GenBank (RJ2N, RJ3N, RJ4N, RJ5N, RJ6N, RJ8N, RJ9N, RJ10N, RJ11N and RJ13N: Bahri et al., 2009a). Two French pathotypes were genotyped and used as a reference for the Mediterranean and NW-European PST populations: S1 from southern France and N1 from northern France. Microsatellite PCR amplifications were carried out as described by Enjalbert et al. (2002), using 10 ng genomic DNA. The cycling protocol was: 1 × 94°C (2 min), 35 × (30 s at 94°C, 30 s at 52°C, 30 s at 72°C) and 1 × 72°C (2 min).

Data analysis

Population structure

Genotypic data were analysed using structure software (Pritchard et al., 2007). This method is based on the Bayesian Monte Carlo Markov Chain (MCMC) approach that clusters individuals into K distinct populations in a way that minimizes Hardy-Weinberg disequilibrium and gametic phase disequilibrium between loci within groups. The MCMC scheme was run for 500 000 iterations after an initial burn-in period of 500 000. structure was executed for K ranging from 1 to 10 and repeated at least 10 times to check for the convergence of likelihood values for each value of K. The number of populations that best represented the data observed using the model implemented was determined by maximizing the estimated log-likelihood of the data for different values of K. Analyses were performed on clone-corrected data, with one individual per genotype. It should be noted that structure analysis assumes panmixia in subpopulations, whilst PST is known for its strong clonality. However, many recent studies have demonstrated the efficiency of this approach in describing population structure in clonal or partly clonal species (Bahri et al., 2009b; Lo et al., 2009).

Phylogenetic tree

The populations program (Langella, 2008) was used to calculate Nei and Dice distances between genotypes from the SSR dataset. A phylogenetic tree was reconstructed using the neighbour-joining (NJ) method, and the stability of tree branches was tested by bootstrapping both individuals and loci (500 times).

Genetic diversity

The numbers of genotypes and private alleles were estimated using genetix software (Belkhir et al., 2004). Pathotypic and genotypic diversities were estimated using Simpson’s index.

Clonality test

The presence of recombination in the Pakistani population was assessed using two approaches. First, the normalized index of association (Smith et al., 1993) was estimated and the null hypothesis rd = 0 was tested by 10 000 randomizations using the multilocus program (Agapow & Burt, 2001). Secondly, the parsimony tree length permutation test (PTLPT) was used as proposed by Burt et al. (1996) with the phylip package (Felsenstein, 1989). The length of the most parsimonious tree estimated from the observed dataset was compared with the lengths of the most parsimonious trees estimated from 1000 datasets randomized for each locus (null hypothesis).

Results

Genetic variability in Pakistani PST populations

Diversity in virulence factors

Forty-nine isolates were recovered from the samples from NWFP (Table 1) and 12 pathotypes were identified. The virulence frequencies of v6, v7, v9, v27 and vSU ranged from 63% to 98% in the total sample, whilst v2 was fixed in all pathotypes. No virulences to Yr5, Yr10, Yr15, Yr24, YrCV (=Yr32) or YrSP were detected. Virulences v3, v4, v17, vSD (v25 + ), which are common in NW Europe, were only present in one or two pathotypes (2–4% of the population).

Two predominant pathotypes, P1 and P2, representing 67% of the population, shared virulences v2, v6, v7, vSU and v27, but differed regarding four other virulences (additional virulences v1 + v9 for P1, and vA + v8 for P2). The remaining pathotypes were represented by one to three isolates. P1, P2 and P3 displayed patterns of virulence that contrasted with the other pathotypes: all were avirulent on cultivars TP981 (Yr25) and Early Premium, as well as on the very susceptible cv. Victo, for which no resistance gene has been determined so far using Northern European, Mediterranean and Chinese isolates. Moreover, virulence vSU was not associated with virulence v4 in these three pathotypes, a pattern never observed to date in NW-European pathotypes. One case of dissociation between vSU and v4 was observed in the USA (de Vallavieille-Pope & Line, 1990).

Another important group of related isolates consisted of pathotypes P4–P9, which shared virulences v2, v6, v7, v9, v25, vA and vEP, but differed in terms of single virulences (v1, v8, v27 or vSU).

P12 displayed a particular virulence pattern, characterized by the virulences v4, v17 and vSD, a combination identical to the reference race N1 from northern France (237E141-v17) with an additional virulence against Yr27. The P8 pathotype had a virulence combination close to the S1 (6E16-Av9) reference race from southern France, with an additional virulence against Yr27. The P4 pathotype, found in one Pakistani isolate, had a virulence combination close to the S2 (6E16-v9) reference race recently found to be spread throughout the western Mediterranean region (Bahri et al., 2009b) and seen in southern France.

Diversity of SSR markers

Within the 12 pathotypes described in NWFP, 62 SSR polymorphic alleles were identified and defined 27 distinct genotypes. The structure analysis of the data clustered the 27 genotypes into five groups (Table 1 and Fig. 1). The groups thus observed were in good agreement with the pathotypes described above: the two predominant pathotypes, P1 and P2, clustered in groups 3 and 5, respectively, whilst pathotypes P4–P9 corresponded to group 1. Group 2 corresponded to the rare pathotypes P10, P11 and P12, and group 4 corresponded to the infrequent pathotype P3. It should be noted that one pathotype P3 isolate presented a hybrid structure between groups 3 and 5 (see Fig. 1). The reference pathotypes for the southern France (S1: 6E16-Av9) and NW-European population (N1: 237E141-v17 pathotype) were assigned to groups 1 and 2, respectively. The latter two groups corresponded to the rare Pakistani pathotypes. The majority of the isolates studied (80%: groups 3, 4 and 5) had genotypes distinct from both the NW-European and Mediterranean isolates, whilst most of the remainder were close to the Mediterranean isolates (group 1).

Figure 1.

 Repartition of Pakistani wheat yellow rust (Puccinia striiformis f.sp. tritici) population into genetic groups based on SSR markers determined by structure software. (a) Localization of the five wheat yellow rust genotype groups and pathotypes (given into brackets) in the four locations sampled in North-West Frontier Province, Pakistan (in yellow in the general map of Pakistan). (b) Summary plot of estimates of the posterior probability for an individual from the Pakistani wheat yellow rust population to belong to the five inferred genotype groups (Q) using structure software. Each individual is represented by a single horizontal line broken into five coloured segments, with lengths proportional to Q. All the genotypes are accurately assigned to one of the five inferred groups, except for one individual (no. 27) presenting a clear hybrid structure (50/50 yellow-red individual corresponding to the group 3/5 hybrid).

A phylogenetic tree constructed from the SSR genotypes revealed that these five groups represented three different lineages (Fig. 2). Groups 1 and 2 were in two different lineages, whereas groups 3, 4 and 5 defined a third lineage. The P2 (group 5) and P3 (group 4) pathotypes seemed to be derived from pathotype P1 (group 3). This lineage grouped in particular the genotypes and pathotypes (P1, P2 and P3) specific to Pakistan. The isolate from P3 with a hybrid structure (group 3/5) was genetically close to pathotypes P1 and P2. Minimum spanning network analysis (fluxus-engineering.com) of the Pakistani pathotypes based on the differences observed in virulence spectra produced a similar phylogenic tree (data not shown).

Figure 2.

 Phylogenetic tree of the Pakistani wheat yellow rust (Puccinia striiformis f.sp. tritici) population constructed with 15 SSR markers (populations). The first character represents the location (A: Abbotabad; P: Peshawar; N: Nowshera; M: Mingora). The second and third characters represent the pathotype. The number in brackets represents the number of isolates sampled. Two French pathotypes were used as references of the Mediterranean (S1: 6E16-Av9) and NW-European (N1: 237E141-v17) PST populations.

Differences between the five groups could partly be explained by the existence of private alleles (Table 1): groups 1 and 4 had two and four specific alleles, respectively, whilst one allele was specific to group 3 and one specific to group 5. However, the true privacy of the alleles could not be concluded from the small number of isolates analysed during this study. The good concordance between SSR diversity and virulence patterns was also confirmed when the individual-based phylogenetic tree was considered in more detail. In addition, SSR markers were able to discriminate up to 10 different genotypes within one pathotype (P1, Fig. 2).

In line with the existence of a strong structure in the NWFP isolates analysed, both PTLPT and the normalized index of association tests rejected the hypothesis of random mating in the sample. A normalized association index rd of 0·26 was observed, a value differing significantly (P < 10−4) from that expected (zero). The tree estimated from the observed SSR dataset displayed a much shorter length (126 steps, P < 10−3) than that expected under panmixia (randomized dataset: 248 steps), but it was also much longer than what might be anticipated under strict clonality (63 steps, data not shown). As the existence of substructures in a sample is a major bias for such tests, the two predominant pathotypes, P2 (group 5) and P1 (group 3), for which an acceptable number of individuals were available, were re-analysed. For pathotype P1, both tests led to rejection of the null hypothesis for free recombination (34 steps observed in the most parsimonious tree and 42 steps obtained for a randomized dataset). However, for pathotype P2, the most parsimonious tree displayed a length of 11 steps, corresponding to the mean value obtained for the randomized dataset. As the minimum value expected under strict clonality (six steps) was not observed in the randomized datasets, it was concluded that the hypothesis for strict clonality could be rejected. Normalized association index analysis produced similar results (rd = −0·003, P = 0·92) (Fig. 3).

Figure 3.

 Distribution of the normalized index of association rd frequency for Pakistani wheat yellow rust (Puccinia striiformis f.sp. tritici) genotypes of the pathotype P2 estimated by 10 000 randomizations using multilocus software.

Postulated resistance genes in Pakistani wheat cultivars

Postulation of the Yr genes based on the 10 northern France pathotypes (N1–N10) and the previously predominant southern France pathotype (S1) led to the conclusion that there is a specific Yr gene in most of the cultivars. However, using the three predominant Pakistani pathotypes (P1–P3) and the S2 pathotype that predominates in the western Mediterranean region (Bahri et al., 2009b), postulation was more difficult, leading to nine cultivars with an undetermined postulation because these pathotypes might carry other unknown virulences. The resistance genes postulated in Pakistani wheat cultivars, i.e. Yr2, Yr6, Yr7, Yr9, Yr27 and YrSU, were most frequently observed in bread wheat from this area. Yr4 was rarely identified (Table 2).

At the seedling stage, only 11 cultivars were resistant to the three Pakistani pathotypes P1, P2 and P3. Among the 21 lines expected to be released as future cultivars, only five (CT-00231, CT-00062, V-00055, V-00125 and Diamond) were totally resistant to P1 and P2, the predominant pathotypes currently identified in Pakistan. Furthermore, out of the 40 cultivars tested, 34 were susceptible to S2 and similar to P4, which is rare in Pakistan but frequent in the western Mediterranean region (Bahri et al., 2009b), and only eight were susceptible to all NW-European pathotypes (Fig. 4, Table 2).

Figure 4.

 Frequencies of yellow rust resistance genes postulated in 40 Pakistani bread wheat cultivars (a) and frequencies of virulence genes observed in 49 Pakistani Puccinia striiformis f.sp. tritici isolates (b).

Discussion

Previous virulence surveys of wheat yellow rust epidemics in Pakistan depicted a large number of pathotypes (Hakim et al., 2001), contrasting with the low diversity present in other parts of the world; e.g. in NW Europe and Australia. The aim of the present work was thus to compare the genetic diversity of the Pakistani PST population with the well-documented NW-European population, with particular focus on the North-West Frontier Province (NWFP), an area where yellow rust is endemic. Despite the limited sample size (49 purified Pakistani PST isolates), major differences in genetic diversity could be determined.

In the studied sample, 12 pathotypes were described in NWFP, with three pathotypes already reported in Pakistan (Hakim et al., 2001). This pathotype diversity (0·76) represents a major difference from the European situation, where only two to four pathotypes are present on a similar regional scale. Using 15 microsatellite markers, 27 different genotypes were described (genotypic diversity = 0·92). Once again, molecular diversity was greater than that observed in NW Europe: all the pathotypes identified in NW Europe during the past 20 years display a single SSR genotype, and pooling the Western Mediterranean population only adds five additional SSR genotypes (Bahri et al., 2009b). Pathotype divergence was observed in accordance with molecular divergence (Fig. 2).

structure Bayesian clustering of the SSR dataset revealed that the population is substructured into five genetic groups representing three different lineages. One lineage grouped six related pathotypes (P4–P9) as well as the southern France reference (S1), and corresponded to a genetic group extending over the Mediterranean region for both virulences (Hakim et al., 2001) and SSR profiles. The presence of migrations between the Mediterranean region and Pakistan is not surprising, as virulence v9 invaded Southern Asia after successive migrations from the Red Sea to the Middle East and thence to Iran and Pakistan (Singh et al., 2004). The second lineage (group 2) grouped three rare pathotypes (P10, P11 and P12) and the reference isolate for northern France (N1 or 237E141-v17) which belongs to a more extended NW-European population. Despite the high divergence of SSR genotypes in group 2, these results suggest that pathotypes P10 and P11, never reported to date in Pakistan or in any other country, are related to the NW-European population. The presence of two pathotypes (P1 and P12) with a genotype identical to N1 illustrates the importance of windborne dispersal or human activity to the introduction and emergence of new pathogens, as previously reported in Europe and Australia (Wellings, 2007; Hovmøller et al., 2008). The third lineage included the two predominant pathotypes in Pakistan (P1 and P2) along with the rare pathotype P3. The predominance of P1 and P3 had been reported in Pakistan between 1969 and 1976 under the names 67E0 and 66E16, respectively. P2 only differs from P3 in terms of virulence vA and virulence on cv. Heines Kolben (Yr6 + 2+), and could thus have originated by mutation from the old reported pathotype P3. In Pakistan, this virulence was detected in 1977 for the first time (Hussain et al., 2004), soon after introduction into the country of the resistance gene YrA in old wheat cultivars (1971: Blue Silver; up to 1980: Pari 73, Sandal 73, Yecora 70 and Zamindar 80; Wellings et al., 1988). These particular pathotypes (P1, P2 and P3) may be specific to the Central Asian region as they have also been reported in the North Hills region of India (Prashar et al., 2007).

As demonstrated here, the PST population in Pakistan is highly variable and many pathotypes and genotypes can contribute to epidemics during a single season. One source of diversity is certainly the successive selection of novel pathotypes arising from stepwise mutations. Pathotypes differing by a single virulence, but sharing an unique SSR genotype, are good candidates for recent evolution through virulence mutation, as has been found in European populations (Hovmøller et al., 2008). However, most pathotypes, such as P1, are made up of numerous distinct SSR genotypes (Fig. 1), a situation that contrasts markedly with European data. The examination of host structure and epidemic events in the area provided some indications regarding the reasons for the high degree of PST genetic diversity in Pakistan.

First, the NWFP is considered to serve as a gateway for the entry into Pakistan of new pathotypes of rusts originating from Central and Western Asia. These migrations may have contrasting effects on local diversity, depending on their strength and nature. The spread of new virulent strains over large areas has often been reported and may contribute to homogenization and lower genetic diversity. Such spreads have been reported in Central Asia, West Asia, and North Africa (CWANA), and have been attributed to uniform use of the same resistance genes. Indeed, the widespread use of Yr2, Yr6 and Yr7 in the CWANA region has resulted in successive breakdowns of these resistance genes. In the late 1980s, pathotype 6E16-Av9, combining virulences v2, v6, v7 and v8, was the most widespread and recurrent in East Africa (Ethiopia), Yemen, North Africa, the Middle East (El-Daoudi et al., 1996; Hakim et al., 2001) and in Pakistan, where it was responsible for a major epidemic in 1973 (Ahmad, 2001). The 1B/1R translocation carrying Yr9 resistance was intensively introgressed in International Maize and Wheat Improvement Center (CIMMYT) lines during the 1980s. These lines, developed under the name of the Veery 5 family, were very popular in different Central and West Asian countries as well as in South Asia, but became susceptible within a few years, as a new race, 6E16-v9, which possessed virulence v9, was spreading. Detected for the first time during 1986 in Ethiopia (Badebo et al., 1990), the migration of this new pathotype from the Eastern African highlands to South Asia, via West Asia, was reported by Singh et al. (2004). The breakdown of Yr9 led to the introduction of a number of Kauz-derived cultivars (in Pakistan, India, Syria, Iran and Turkey), which present a Yr9 and Yr27 resistance combination. This release was at the origin of the current emergence of pathotypes with both v9 and v27 virulences, observed in the present study as well as in the Middle East (Afshari, 2008; Bahri et al., 2009b). Despite the breakdown of Yr9 15 years ago, it is interesting to note that this resistance was still found to be frequent in the Pakistani wheat cultivars studied.

The observations above depict Pakistan as a key step in the spread of new virulences from the Middle East to the Indian subcontinent, but the high diversity described in this study moderates this picture. Indeed, if Pakistan was simply a recipient of pathotypes emerging from East Africa, one might expect a low variability as a result of founder effects and clonal diffusion of the most fit strains. Migrations from very distant places might explain part of the observed diversity, and three contrasted lineages characteristic of the Western Mediterranean (group 1), Northern Europe (group 2) and Central Asia (groups 3, 4 and 5) were identified in the same area in Pakistan. In contrast to this migration hypothesis, the PST populations in Pakistan also display ‘old’ pathotypes such as P1, P3 and P8, with virulence patterns that have been reported for decades in the area. The P1 pathotype was predominant from 1969 to 1976, re-emerging recently to become predominant again in 2005–6. Whether the re-emerging pathotype has been maintained locally or is the result of convergent evolution caused by the selective pressures on different lineages, could not be determined from this study. However, the high SSR diversity found within the P1 pathotype favours the hypothesis of a long-term existence of the lineage, rather than that of recent selection. The persistence of pathotypes over a long period of time can be explained by the presence in the area of unchanged host populations, such as old wheat cultivars and landraces. In Pakistan, alongside the elite cultivars developed by the CIMMYT (jointly with the International Center for Agricultural Research in the Dry Areas (ICARDA) and local institutions), about 15% of the wheat- growing area is sown with old cultivars and landraces (such as Khattak Wal and local white: Ehsan et al., 2003). Thus, ‘old’ pathotypes might be best adapted to and predominant on this traditional germplasm.

The high climatic and agronomic diversity from north to south, or between uplands and valleys in Pakistan, might also contribute to the pathogen variability observed. These contrasted environmental conditions may select for local adaptation, each agroecological area being infected by specific PST subpopulations with contrasting environmental optima. Indeed, an adaptation to high temperature was recently observed in PST (Milus et al., 2009), and suggested as an explanation for the spatial differentiation between northern and southern French PST populations (Mboup, 2008). However, within the limits of the present study, the absence of a strong spatial structure did not provide strong evidence to support the hypothesis of local adaptation, and no tests were performed on temperature adaptation. Another effect of the agro-environmental diversity prevailing in Pakistan is the presence of areas with conditions that are highly conducive to both PST epidemics and survival. At present, the strong over-summering of yellow rust reported at some sites in NWFP (Hassan, 1968) may result in a large effective population size, and thus high genetic diversity (Kimura & Crow, 1964). Yellow rust over-summers in the Himalayas, Hindu Kush and Suleman ranges (Chaube & Pundhir, 2005). The existence of a green bridge based on off-season wheat culture in Pakistan, which avoids the recurrent bottlenecks caused by host absence during the intercrop season, may further contribute to the genetic diversity of local populations.

Finally, the presence of a recombination signature in the P2 pathotype (group 5) in NWFP was observed. However, part of the PTLPT and rd values could be explained by homoplasy in SSR markers (similarity in allelic size but no common ancestry), caused by their high mutation rate (Schug et al., 1998). The amount of homoplasy in a population is not only dependent on the mutation rate, but will also be affected by the demographic history of the population (the larger the effective size, the stronger the homoplasy; Estoup et al., 2002). It should be noted that a unique multilocus SSR genotype was found during several decades in NW Europe (Enjalbert et al., 2005), a fact that argues against a strong impact of homoplasy in PST populations. More individuals and more markers would improve quantification of the intensity of recombination in genetic group 3, but the additional hybrid genotype described between groups 3 and 5 also demonstrates that genetic recombination is probably at the origin of some of the genotypes sampled in Pakistan. The recent description of Berberis chinensis as an alternate host for PST under laboratory conditions (Jin et al., 2010) is in accordance with a sexual origin of these recombinant genotypes, rather than through a parasexual cycle, the hypothesis that prevailed until now. Various Berberis spp. occur in Pakistan, including B. lycium and B. vulgaris (Perveen & Qaiser, 2010). The putative role of the alternate host in wheat yellow rust epidemics in Pakistan still needs to be investigated.

In conclusion, the high degree of genetic diversity associated with evidence of recombination raises new questions about the origin of the observed polymorphisms for virulence and molecular markers. In addition to demography and migration, the presence of recombination is an important factor that could contribute to the variability of PST in Pakistan. The existence of recombination implies important consequences for breeders, because it enhances population variability for virulence combinations, and thus the speed of breakdown for new resistance combinations. Joint studies on world yellow rust populations and their comparison with known clonal populations (such as those in NW Europe and the Mediterranean) are necessary to assess the importance of recombination in PST populations on a worldwide scale, and to determine whether a sexual cycle exists, with a potential mechanism of action through the control of an aecial host.

To understand the origin of genetic diversity in these structurally complex Pakistani PST populations it is necessary to disentangle the effects of local adaptation to the host/environment, recombination, migrations and founder effects in a country at the crossroads of PST migration routes. This can only be achieved by refining the geographic structure of PST populations in Pakistan with relation to host/climate conditions and neighbouring countries through pluri-annual surveys of PST populations.

Acknowledgements

We would like to thank Laurent Gérard and Nathalie Galet for their technical support during urediniospore propagation and virulence assessments, as well as Julie Dawson for helping us to improve the manuscript. This work was supported by the European Integrated Project BIOEXPLOIT (FOOD-CT-2005-513959) and by ANR (Agence Nationale de la Recherche) through the EMERFUNDIS project (ANR 07-BDIV-003). The first author was supported by a grant from the Institut Français de Coopération in Tunisia.

Ancillary