Genetic structure of Mycosphaerella graminicola populations in Iran




To provide insight into the genetic structure of Mycosphaerella graminicola populations in Iran, a total of 221 isolates were collected from naturally infected wheat fields of five major wheat-growing provinces and analysed using AFLP markers and mating-type loci. All populations showed intermediate to high genotypic diversity. In the Golestan and Ardabil populations two mating types were found at near-equal frequencies, whilst all populations were in gametic disequilibrium. Moreover, clonal haplotypes were identified in different sampling sites within a field in both the Khuzestan and Fars provinces, demonstrating that pycnidia are probably the primary source of inoculum. All five populations had low levels of gene diversity and had private bands. Low levels of gene flow and high genetic differentiation were observed among populations and different clustering methods revealed five genetically distinct groups in accordance with the sampling areas. The Golestan and East Azarbaijan populations were more genetically differentiated than the others. Random genetic drift, selection and geographic barriers may account for the differentiation of the populations. The results of this study indicate a population structure of M. graminicola in Iran contrasting to that of most other countries studied.


Mycosphaerella graminicola (anamorph Septoria tritici) is a haploid fungus with a bipolar heterothallic mating system, and causes septoria tritici blotch (STB), a major disease of cultivated wheat (King et al., 1983; Kema et al., 1996). Under favourable conditions the disease can cause significant yield losses ranging from 30 to 50% (Eyal et al., 1987).

Previous studies reported that sexual reproduction plays a significant role in genetic variability of M. graminicola populations, resulting in high biological fitness (Chen & McDonald, 1996), and it is also assumed to be the main driving force for the regular occurrence of disease epidemics (Shaw & Royle, 1989). Ascospores are wind dispersed and potentially can be blown over long distances (Shaw & Royle, 1989). It was shown that airborne ascospores released from pseudothecia in infected plant tissues serve as primary inoculum at the start of the growing season, and contribute to disease spread during the growing season (Kema et al., 1996; Zhan et al., 1998). After disease establishment in the early season, pycnidia are considered the major dissemination force, releasing a huge number of pycnidiospores that are dispersed locally by rain splash and physical contact (King et al., 1983). It was shown that the role of these asexual spores in developing epidemics is much greater than that of the sexual ascospores in the early season, whilst the contribution of the teleomorph increases progressively as leaves age later in the growing season (Eriksen & Munk, 2003).

Traditionally, several strategies have been used to control STB, including cultural practices such as crop rotation and stubble management, chemical control, and the use of disease-resistant cultivars (Eyal, 1999). Employment of resistant cultivars in cropping and incorporation of major resistant genes in breeding programmes are considered to be the most effective, economical and environmentally safe strategy for controlling this disease (Eyal et al., 1987). Nevertheless, this strategy has limited applications for pathogens with high genetic flexibility, such as M. graminicola, which are able to evolve to overcome resistance genes as a response to selection pressure imposed by resistant cultivars. Therefore, successful strategies in breeding programmes rely heavily on knowledge of the biology, population dynamics and existing genetic structure of a particular pathogen (McDonald & Linde, 2002).

Genetic structure refers to the amount and distribution of genetic diversity within and among populations. It is the outcome of the interactions among the major evolutionary forces, including mutation, random genetic drift, gene flow, reproduction/mating system, and selection. These determinants can, over time, cause substantial changes in overall pathogen populations, culminating in potential risk to the host target. Understanding the role of each factor and how it influences the emergence of a newly adapted population is essential to predict the risk of a pathogen in order to improve control strategies. This knowledge is useful for optimizing the management of resistance genes and fungicides and for employing a successful breeding strategy (McDonald & Linde, 2002).

The genetic structure of M. graminicola populations originating from different geographical regions has been analysed using various molecular markers, such as random amplified polymorphic DNA (RAPD) (Czembor & Arseniuk, 1999), restriction fragment length polymorphism (RFLP) (Linde et al., 2002; Zhan et al., 2003), simple sequence repeat (SSR) (Razavi & Hughes, 2004; Banke & McDonald, 2005) and amplified fragment length polymorphism (AFLP) (Schnieder et al., 2001; Kabbage, 2007). Low genetic differentiation was observed among populations of M. graminicola on both regional (Boeger et al., 1993; Schnieder et al., 2001) and global (Linde et al., 2002; Zhan et al., 2003) scales, suggesting gene flow between populations has been a common event. It was shown that studied populations of M. graminicola undergo a high degree of sexual recombination (Chen & McDonald, 1996; Zhan et al., 1998) and both mating types occur at approximately equal frequencies (Zhan et al., 2002a).

Septoria tritici blotch has intensified recently in some wheat-growing areas of Iran, mainly because of the introduction and monoculture of susceptible CIMMYT-derived wheat cultivars. Severe epidemics have occurred in major wheat-growing provinces, including Khuzestan, Fars, Golestan and Ardabil. Therefore, development of resistant wheat cultivars to replace the susceptible ones is needed. Knowledge of the genetic structure of M. graminicola in Iran, especially in hotspots of the disease, could be useful for improving the management of fungicides and resistance genes, and developing a breeding strategy to achieve durable disease resistance. Iran is located in the Fertile Crescent, which is considered the centre of origin of both wheat and M. graminicola (Banke et al., 2004). However, to date no information is available on the genetic structure of natural populations of this devastating pathogen in Iran. A previous study on the genetic structure of M. graminicola in Iran was conducted based on two populations that were eventually found to be artificially inoculated fields (Jürgens et al., 2006).

This study was conducted to determine the genetic structure of M. graminicola populations collected from naturally infected fields of five different provinces, representing the major wheat-growing areas of Iran. These provinces have different climates, cultivars and wheat-growing seasons. The specific objectives were to (i) estimate genetic variation among M. graminicola populations in Iran, and (ii) determine frequencies of mating types and linkage disequilibrium to test for the signatures of possible sexual recombination in the populations.

Materials and methods

Population sampling

A total of 221 M. graminicola isolates were collected from naturally infected wheat fields of five major wheat-growing provinces of Iran: Khuzestan, Fars, Golestan, Ardabil and East Azarbaijan, which were considered as separate populations (Table 1, Fig. 1). Isolates of the first three populations were collected according to the hierarchical sampling method (Razavi & Hughes, 2004) in 2006, whereas the last two populations were sampled randomly in 2007. In hierarchical sampling, isolates were obtained from six to eight locations approximately 10 m apart within individual fields. Within each field location, a single leaf showing STB symptoms was collected from each of three to four different plants. Only one isolate was obtained from each of the two to three distinct lesions per leaf.

Table 1.   Number of isolates (number of haplotypes) (n), number of bands (B), percentage of polymorphic loci (P), number of private bands (BP), gene diversity (H), genotypic (Ĝ/N) diversity and clonal fraction (C) of Mycosphaerella graminicola populations in Iran
ProvincenBP (%)HaĜ/NbCBP
  1. aNei, 1978.

  2. bStoddart & Taylor, 1988.

Khuzestan82 (37)15359·350·19200·558
Fars68 (37)13852·340·17350·469
Golestan37 (25)12440·190·12470·3217
Ardabil17 (15)12343·930·16810·127
East Azarbaijan17 (13)10130·370·11630·2414
Overall221 (127)21491·120·24300·4355
Figure 1.

 Sampling locations of Mycosphaerella graminicola populations in Iran (provinces in brackets). Asterisks and open circles indicate sample sites and capitals of sampling provinces, respectively. Dark and light grey areas represent major mountainous regions and deserts, respectively.

Fungal isolation and DNA extraction

To obtain M. graminicola isolates, leaf segments were attached to glass slides with tape and kept under high-humidity conditions until the pycnidia produced cirri containing pycnidiospores. These were then transferred onto YMDA (4 g yeast extract, 4 g malt extract, 4 g dextrose and 15 g agar L−1) plates supplemented with streptomycin (50 mg L−1) using a fine needle. Single-spore preparation and other culturing procedures were performed as described by Eyal et al. (1987). Yeast-like spores produced on YMDA were scraped off and genomic DNA of harvested spores was extracted according to Dellaporta et al. (1983), except that the potassium acetate precipitation step was replaced with two steps of chloroform:isoamyl alcohol (24:1) extraction.

AFLP analysis

AFLP was performed according to Vos et al. (1995), with some modifications. Genomic DNA (∼200 ng) of all 221 isolates was digested with restriction enzymes EcoRI and Tru1I (an isoschizomer of MseI) and ligated to the corresponding adaptors. Pre-amplification was carried out using non-selective EcoRI and Tru1I primers. Initially, a total of eight primer combinations were screened for the highest polymorphisms and five primer combinations –EcoRI-AC/Tru1I-GG, EcoRI-AC/Tru1I-CA, EcoRI-CT/Tru1I-CC, EcoRI-CT/Tru1I-AT and EcoRI-CT/Tru1I-CA – were used for the selective amplification step. PCR conditions of both pre-amplification and selective amplification steps were as described by Vos et al. (1995). The amplified fragments were separated by electrophoresis on a denaturing 6% polyacrylamide sequencing gel and were visualized by silver staining according to Creste et al. (2001). The reproducibility of the AFLP profiles was examined for a subset of isolates (n = 10) using DNA extracted independently. These tests indicated a high reliability of AFLP patterns (overall repeatability >99%).

Mating-type determination

The mating-type idiomorphs of the isolates were determined by multiplex PCR amplification of the two mating-type loci as described previously (Waalwijk et al., 2002). PCR reactions were performed in 25-μL volumes containing 10 ng genomic DNA, 1× PCR buffer, 2 mm MgCl2, 0·2 mm dNTPs, primers at 0·6 μm each and 1 U Taq DNA polymerase. The PCR conditions were 94°C for 3 min, followed by 35 cycles of 94°C for 1 min, 60°C for 30 s and 72°C for 1 min, plus a final extension at 72°C for 10 min. Amplified products of 340 and 660 bp indicated MAT1-1 and MAT1-2, respectively.

Data analysis

AFLP profiles were visually observed and each DNA band generated by each primer combination was considered as a unit character (marker) and numbered sequentially. Only fragments with high intensity were scored for presence (1) or absence (0). The data were transferred into a binary matrix and subsequently subjected to statistical analyses. Isolates were considered members of the same haplotype if they had 99% identical bands and the same mating type.

Genotype diversity and clonal fraction for all populations were calculated as described before (Stoddart & Taylor, 1988; Zhan et al., 2003). Genotypic diversity was measured using genodive software (Meirmans & Van Tienderen, 2004) and the resulting index was further divided by sample size (N) so that this index could be compared for samples of different sizes. This allowed calculation of the percentage of maximum possible diversity obtained. All further analyses were conducted on clone-corrected datasets (Zhan et al., 2003).

The relationship between individuals were calculated using multidimensional scaling (MDS) based on the Jaccard similarity coefficient, implemented in Paleontological Statistics (past) software version 1.74 (Hammer et al., 2001).

The percentage of polymorphic loci present at frequencies >1% and Nei’s unbiased gene diversity (Nei, 1978) were calculated using Tools for Population Genetic Analysis (tfpga) version 1.3 (Miller, 1997). The number of private bands amplified in each population was calculated in genalex 6 (Peakall & Smouse, 2006).

The relative magnitude of genetic differentiation among populations was estimated using different fixation indices. Nei’s GST (Nei, 1973), implemented in popgene version 1.32 (Yeh et al., 1999), was used to examine genetic differentiation among populations. Weir & Cockerham’s (1984)θ, an unbiased estimator of FST, was calculated using multilocus version 1.3 (Agapow & Burt, 2001). The significance of estimated θ values were tested using jackknifing over loci with 10 000 replications using a confidence interval (CI) of 95%. The Bayesian approach of Holsinger et al. (2002) was also used, which allows a direct estimate of FST from dominant markers. The program hickory version 1.1 was used to estimate θB, the Bayesian analogue of FST. The data were run with default sampling parameters (burn-in = 5000, number of samples = 105, thinning factor = 20) using an f-free model. Genetic variation within and among populations was further partitioned by analysis of molecular variance (amova) using the software genalex 6. Variance components and the Φ-statistic were tested by 999 permutations to obtain significance levels. The average gene flow (Nm) among populations was estimated as Nm =  (1 − FST)/2FST (Slatkin & Barton, 1989) by substituting FST with GST, θ, or θB, and the gene flow between any two populations was calculated based on pairwise θ values.

Nei’s unbiased genetic distances (Nei, 1978) among populations were computed using tfpga. The pairwise genetic distances were subjected to cluster analysis using the upgma algorithm. Bootstrap support values were calculated over loci with 1000 repetitions.

The population structure was also examined by the program structure version 2.2 (Pritchard et al., 2000). This program implements a model-based clustering method using Markov Chain Monte Carlo estimation. By comparing the likelihood of the data estimated in different runs for different numbers of populations (K) it is possible to identify the optimal K. Individuals are assigned (probabilistically) to the clusters defined by allele frequencies at each locus. The data were analysed with K ranging from one to five, three replicate runs for each K, and a burn-in period of 10 000 and subsequent run length of 105 iterations. The admixture model and uncorrelated allele frequencies were assumed for the analysis. The log probability of the given K [ln P(X|K); Pritchard et al., 2000] for each level of K was computed, and these values were used to calculate the magnitude of ΔK (Evanno et al., 2005) as a criterion to infer the most probable number of clusters represented by samples. This approach reduces the risk of overestimating K and, therefore, provides a correct estimation of the number of clusters.

Two approaches were applied to test whether the observed pattern of genetic variation was consistent with the null hypothesis of sexual reproduction. First, significant departures from the expected 1:1 ratio in mating-type frequencies were tested with a χ2 test. Secondly, the index of association (IA) and an unbiased estimate of multilocus linkage disequilibrium (inline image) were used to analyse gametic disequilibrium. IA and inline image values were calculated by multilocus software, and 1000 artificially recombined datasets were used to determine the statistical value of the test. To determine if departures from gametic disequilibrium were the result of population admixture, admixed individuals identified by structure were removed from the populations and multilocus measures of association were calculated for populations.


Of the eight primer combinations initially tested, five combinations showed high polymorphism in AFLP analysis, resulting in 214 scorable bands, of which 195 (91·12%) were polymorphic across the dataset (Table 1). The total number of DNA bands recorded in populations varied from 101 to 153 in isolates from East Azarbaijan and Khuzestan, respectively. The percentage of polymorphic bands ranged from 30·37% in East Azarbaijan to 59·35% in Khuzestan. The number of bands and the percentage of polymorphic bands tended to be highest in the Khuzestan and Fars populations.

A total of 127 distinct haplotypes were found among the 221 M. graminicola isolates analysed (Table 1). All clonal haplotypes identified by AFLP profiles showed the same mating type. The mean clonal fraction ranged from 0·12 to 0·55 in the Ardabil and Khuzestan populations, respectively. The Ardabil population showed the highest genotypic diversity (81%), whilst the Khuzestan population had the lowest genotypic diversity (20%). Among 82 isolates of the Khuzestan population, 37 distinct haplotypes (45%) were observed, with one to 15 isolates per haplotype. Three out of the 37 haplotypes were isolated from two or three locations 10–60 m apart from each other. In the Fars population of 68 isolates, there were 37 distinct haplotypes (54%). In this population, different haplotypes occurred one to seven times, with five haplotypes dispersed in two or four locations. Distances between these locations were 10 m and in one case 50 m. In the Golestan population 25 unique haplotypes were detected and identical haplotypes were usually found among isolates originating from the same leaf, or sometimes from different leaves from the same location. Among the clonal haplotypes identified in Khuzestan, Fars and Golestan, 22 (36·7%), 14 (31·1%) and 11 (57·9%) isolates, respectively, were identified from the same leaf.

The mean gene diversity across all populations was 0·24 (Table 1). Among the five populations examined, the Khuzestan population had the highest gene diversity (0·19), whilst the East Azarbaijan population had the lowest (0·11).

When the relative magnitude of genetic differentiation among five populations was estimated using Nei’s GST, a very high proportion of genetic diversity was revealed among them (GST = 0·39). Significant overall population differentiation was found among the populations based on Weir & Cockerham’s θ (θ = 0·38, < 0·01) (Table 2). Pairwise comparisons of θ revealed that the lowest differentiation was between the Khuzestan and Fars populations (θ = 0·2, < 0·01), whereas the highest was between the Golestan and East Azarbaijan populations (θ = 0·63, < 0·01). The mean θB, the Bayesian analogue of FST, was 0·34 among five populations, agreeing well with the values of GST and θ. amova analysis showed that 38·97% (< 0·01) of genetic variation could be attributed to differences among populations and 61·03% to differences among individuals within populations. Clearly, all different statistical approaches were consistent in indicating that >34% of the genetic diversity occurred among the populations, while <66% occurred within each population. The overall gene flow among populations was calculated as Nm = 0·78 using GST, Nm = 0·87 using θ and Nm = 0·97 using θB. The highest Nm value (2) was obtained in pairwise comparisons between the Khuzestan and Fars populations (Table 2).

Table 2.   Gene flow (Nm, below diagonal) and population differentiation (θ, above diagonal) for pairwise comparisons among five Mycosphaerella graminicola populations in Iran, plus overall values for all populations
 KhuzestanFarsGolestanArdabilEast Azarbaijan
  1. *Indicates significant values at < 0·01.

East Azarbaijan0·660·690·290·52
θ (all populations)0·38*    
Nm (all populations)0·87    

The MDS, based on the Jaccard similarity coefficient (Fig. 2), detected genetic divergence among populations. With a few exceptions, isolates of each population were grouped together. The isolates from Khuzestan, Fars and Ardabil were more aggregated, whilst the Golestan and East Azarbaijan isolates had the most divergent positions and formed two distinct groups.

Figure 2.

 Multidimensional scaling (MDS) of individual AFLP haplotypes of Mycosphaerella graminicola from five Iranian populations based on Jaccard similarity coefficients.

Pairwise measures of Nei’s unbiased genetic distance ranged from 0·06 to 0·19. Branches of the upgma tree were well supported by bootstrap values (Fig. 3), suggesting the reliability of the pattern found. Genetic distances between the East Azarbaijan and Golestan populations, as well as their distances from all other populations, were comparatively high (0·16–0·19). Genetic distance values between the remaining populations were lower and varied from 0·06 to 0·09.

Figure 3.

upgma phenogram based on Nei’s genetic distance (1978) among Iranian Mycosphaerella graminicola populations. Bootstrap values are given at the nodes.

Bayesian grouping of isolates provided a confirmation of the above results. In the structure analysis, five clusters were inferred based on the magnitude of ΔK (Fig. 4), consistent with the highest posterior probability [ln P(X|K)] calculated by the software (data not shown). The assignment analysis revealed that >90% of the individuals were assigned to the population from which they were sampled, suggesting that each of sample site formed a distinct population (Fig. 5). In East Azarbaijan no immigrant isolate was detected, whilst in Golestan one probable immigrant was identified. The immigrant isolate detected in Golestan population was assigned with low probability to the Ardabil and Fars populations. All other isolates of these populations were assigned with high probability to their original population. In the Khuzestan, Fars and Ardabil populations, four, five and two likely immigrant isolates were identified, respectively (even though they did not necessarily represent independent migration events). The probable immigrant genotypes detected in each of the three populations could have originated from the other two populations, suggesting that there might be some degree of gene flow between these three populations. The only exception was one isolate from Ardabil which may have migrated either from East Azarbaijan or from Fars.

Figure 4.

 Magnitude of ΔK calculated for each level of K. Maximum ΔK indicates the most likely number of Mycosphaerella graminicola populations (= 5).

Figure 5.

 Membership coefficient of multilocus AFLP genotypes of Mycosphaerella graminicola from five provinces in Iran estimated by the software structure. Clusters of isolates based on prior-defined populations are represented by five shading patterns. Each vertical line represents one multilocus genotype and shades within a vertical line represent the inferred posterior probability that the isolate belongs to a particular prior-defined population.

The results from grouping analyses were reflected by the distribution of private bands among populations; some of the private bands were fixed in the populations (Table 1). The Golestan and East Azarbaijan populations had higher numbers of private bands than the other populations, suggesting a high level of genetic divergence in these populations. The other three populations were also characterized by several, but lesser, private bands.

Both mating types were present in all populations; their frequencies departed significantly (< 0·05) from the 1:1 ratio in Khuzestan, Fars and East Azarbaijan (Table 3). Multilocus measures of association (IA and inline image) were significant (< 0·01) for the five populations (Table 3) and removing admixed genotypes did not change the results (data not shown).

Table 3.   Tests of multilocus association and mating-type frequencies of clone-corrected Mycosphaerella graminicola populations in Iran
PopulationNo. isolatesχ2aIAinline imaged
  1. aχ2 value based on 1:1 ratio and 1 degree of freedom.

  2. *Significant at < 0·05.

  3. **Significant at < 0·01.

East Azarbaijan2116·24*3·11**0·05**


In earlier studies on the genetic structure of M. graminicola, most populations of this pathogen were reported to be in linkage equilibrium, with approximately equal frequencies of mating types, a high level of genetic variation within a field, and a low level of differentiation among populations (Chen & McDonald, 1996; Zhan et al., 1998, 2002a, 2003; Schneider et al., 2001; Linde et al., 2002; Banke et al., 2004; Razavi & Hughes, 2004). In Iran, wheat is widely grown in four major environmental zones: warm and humid at the Caspian Sea coast (including Golestan province and the Moghan valley of Ardabil province), high altitudes with cold winters (including East Azarbaijan province), temperate with mild winters (including some parts of Fars province) and warm and dry (including Khuzestan province and some parts of Fars province). These zones, which are generally separated by geographic barriers (e.g. mountains and deserts) and characterized by different climates, cropping systems, wheat cultivars and growing seasons, may explain why, in contrast to populations studied elsewhere in the world, populations of M. graminicola in Iran are genetically differentiated between regions.

The FST values estimated by different approaches, as well as MDS and Bayesian analysis, clearly revealed five groups, corresponding to five geographically distinct populations. Similar results were found for natural M. graminicola populations from Australia (Jürgens et al., 2006) and the USA (Kabbage, 2007). Differentiation in Australian populations was thought to be the result of genetic drift and geographic isolation, whilst differentiation between Kansas and California populations was attributed to the selection imposed by different climate, wheat cultivars, crop rotation patterns and cultural practices. As sampling sites in this study had drastically different climatic conditions and wheat cultivation systems, it is likely that selection exerted by these differences may account for the differentiation observed in the pathogen populations. Likewise, the genetic differentiation detected among five M. graminicola populations may have resulted from random genetic drift, geographic barriers and selection imposed by wheat cultivars.

The Fertile Crescent is considered the centre of origin of M. graminicola (Banke et al., 2004). Thus, the fungal populations sampled from the Middle East have significantly higher gene diversity than populations in America and Europe (Zhan et al., 2003). However, the five M. graminicola populations collected from Iran appeared to possess lower gene diversity than the AFLP gene diversity of this pathogen in the USA and Germany (Schneider et al., 2001; Kabbage, 2007). The presence of private AFLP bands coupled with low gene diversity provided additional evidence for restricted gene flow, high genetic differentiation and ancient divergence, and the occurrence of genetic drift in these populations. Low levels of gene diversity and detection of population-specific AFLP fragments in all natural M. graminicola populations may suggest random genetic drift with bottlenecks causing the loss of alleles and subsequent loss of genetic variation (Nei et al., 1975). Coalescence analysis provided evidence that the divergence of the wheat-adapted M. graminicola from an ancestral population infecting wild grasses in Iran occurred approximately 10 000 years ago. This event coincided with the beginning of agriculture-based societies in the Fertile Crescent and the domestication of wild grasses (Stukenbrock et al., 2007). Therefore, the high level of differentiation observed among the Iranian populations may not be unexpected, because wheat and M. graminicola have been present for thousands of years in Iran and genetic drift may have had much more time to operate than in other areas where M. graminicola has been recently introduced. Local wheat landraces have been cultivated in relatively small fields during this period in Iran. In such agroecosystems, with patchy distribution of the host and the pathogen, populations of the pathogen are not so large and bottlenecks probably occur frequently. In addition, monitoring of STB during recent years has shown that the disease is sporadic in Iran, and in most years conditions are unfavourable to disease development. In such years, disease can be found at very low severity in a few fields, hence the isolates from Ardabil and East Azarbaijan were the only samples collected from these regions. Such conditions could severely reduce population size, resulting in bottlenecks. Nevertheless, during the long period of co-evolution of M. graminicola and wheat, gene flow could decrease population differentiation and counterbalance the erosion of genetic diversity, as gene flow plays the leading role in decreasing population subdivision (Zhan & McDonald, 2004). In the present case, sampling sites were separated by mountain ranges (e.g. the Zagros and Elburz mountains) or desert (e.g. the Salt desert) (Fig. 1). These geographic barriers may have caused patchy distribution of populations and limited gene flow among them. Additional comparative analyses using Iranian as well as European and/or American populations are suggested for further confirmation of the lower gene diversity of Iranian M. graminicola populations.

Regionally adapted host-genotype cultivation over the years can cause directional selection leading to local adaptation of M. graminicola populations. The sampling sites were located in regions in which different wheat genotypes were cultivated. This situation may have provided conducive conditions for pathogen selection and differentiation of populations, in agreement with previous reports that host adaptation is the main force behind population subdivision for this fungus (Zhan et al., 2002b; Zhan & McDonald, 2004).

The distinct gene pools for the five populations may suggest that migration events rarely occurred among these populations. Indeed, the estimates of gene flow and results of assignment analysis demonstrated that low levels of gene flow had occurred among the Khuzestan, Fars and Ardabil populations. A striking differentiation was observed between two populations closely located in Ardabil and East Azarbaijan. The sampling location in East Azarbaijan province was located at high altitude with cold winters and mild summers, where rainfed wheats are cultivated in limited areas on hills surrounded by mountains. By contrast, the sampling location in Ardabil province was located in the Moghan valley with mild winters and warm summers, and irrigated wheat cultivation systems. Geographic barriers, differences in cropping system, and selection imposed by different climates and wheat cultivars may have contributed to genetic differentiation of the two closely located populations. The highly significant genetic differentiation and limited gene flow among Iranian M. graminicola populations contrasts considerably with other reports on regional (Boeger et al., 1993; Schnieder et al., 2001) and global (Linde et al., 2002; Zhan et al., 2003) scales. Although the data from the present study suggest that gene flow has rarely occurred among Iranian populations, migration pattern analysis of a global set of M. graminicola isolates showed that the pathogen has migrated asymmetrically from its centre of origin (Middle East) to Europe and then to the other continents (Banke & McDonald, 2005). Therefore, it is necessary to analyse migration patterns of Iranian populations in relation to regional and global populations in order to determine the extent of gene flow among the Middle Eastern populations and their relationship to those of other continents, in order to infer the most likely ancestral populations.

The Golestan, Ardabil and East Azarbaijan populations had intermediate to high genotypic diversity. In Golestan and Ardabil, the two mating types were found at approximately equal frequencies, whilst tests for mulilocus associations showed that all three populations were in gametic disequilibrium. The high genotypic diversity and near-equal frequencies of mating types in Ardabil and especially Golestan may suggest that ascospores play an important role in initiating and maintaining epidemics of the pathogen in these regions. This is in accordance with previous reports of equal mating-type frequencies in global populations of M. graminicola (Zhan et al., 2002a). The unequal mating-type frequencies and higher genotypic diversity observed in the Ardabil and East Azarbaijan populations could be the result of insufficient sampling.

The levels of genotypic diversity and associated clonal fractions for the Khuzestan and Fars populations were intermediate. In addition, gametic disequilibrium indices were high in both populations and mating-type frequencies differed significantly from the expected ratio. Gametic disequilibrium observed in the two populations may not be attributable to the lack or low rate of sexual recombination, as recombination is only one of several evolutionary processes that affect gametic equilibrium. Processes such as linkage, population admixture, random genetic drift and selection can all cause gametic disequilibrium (Milgroom, 1996). The mere linkage of AFLP markers may account for gametic disequilibrium observed in the populations. Therefore, possible linkage of markers cannot be ruled out in drawing a firm conclusion, although the low rate of recombination in the populations can reduce allelic associations (Milgroom, 1996). Population admixture also cannot explain the observed gametic disequilibrium because removing admixed genotypes from the populations did not change the results. Selection and/or random genetic drift may account for the significant disequilibrium in these as well as other populations. The significant differences between the frequencies of the mating types found in the Khuzestan and Fars populations may suggest predominant asexual reproduction. However, skewed mating-type ratios may also result from processes unrelated to the reproduction mode of the fungus (Milgroom, 1996). It is well documented that mating-type genes may also function in the maintenance of cell wall integrity, virulence and other cellular processes (Zhan et al., 2002b). In these cases, selection exerted by different variables (e.g. resistant cultivars or fungicide applications) on a mating type or tightly linked genes, might favour the propagation of a certain mating- type idiomorph in a population. Additional research with larger sampling locations and using other marker systems, such as SSR markers, may enable firmer conclusions to be reached on the occurrence and rate of sexual recombination in Iranian populations.

In this study, identical haplotypes were found in more than one sampling location in the fields of Khuzestan and Fars. It is generally agreed that ascospores are primary inoculum sources for establishment of STB in the early season and that ascospores as well as pycnidiospores contribute to disease spread during the growing season (Kema et al., 1996; Zhan et al., 1998). It was suggested that uniform distribution of infection by M. graminicola within a field may be an indication of distant airborne ascospores acting as primary inoculum (Shaw & Royle, 1989). Asexual pycnidiospores are usually disseminated via rain-splash and hence, their potential for long-distance movement is limited. For this reason, in almost all natural M. graminicola populations analysed, isolates with the same haplotype have been found in the same leaf or same sampling site in the field (Linde et al., 2002). The proportions of pycnidiospores and ascospores depend on environmental conditions in a specific year or in a specific wheat growing area. Favourable weather conditions during summer, particularly rains, are necessary for the development of pseudothecia (Eyal et al., 1987; Eriksen & Munk, 2003). A possible explanation for dispersal of clones in these two fields might be that pycnidiospores survived on stubble during the winter, were disseminated through tillage and initiated infection as primary inoculum. This hypothesis is corroborated by the fact that Khuzestan and southern parts of Fars province are in the warm and dry zone. This zone is generally characterized by rainless and warm summers (up to 50°C in Khuzestan), mild winters and no freezing temperatures during winter. Therefore, observation of aggregated infections in sampling fields of Khuzestan and Fars on the one hand, and unfavourable summer conditions for development of the sexual state on the other hand, indicate that pycnidia are the probably primary source of inoculum.

It is concluded that the studied M. graminicola populations, particularly in Golestan and East Azarbaijan, are highly differentiated. There is limited gene flow between the populations and bottlenecks have reduced their genetic diversity. The pathogen has mixed reproductive modes in different regions and pycnidia on field stubble are probably the primary source of inoculum in Khuzestan and Fars. Based on the risk model framework (McDonald & Linde, 2002), the Iranian M. graminicola populations fall into the medium-risk category for pathogen evolutionary potential, suggesting the need for caution when applying fungicides or breeding for major-gene resistance. For this category of pathogens, the use of quantitative resistance in combination with monogenic resistance on a regional basis is proposed to achieve durable disease resistance (McDonald & Linde, 2002). As STB epidemics in Iran occur within regions independent of other regions, in addition to the strategy mentioned above, it would also be useful to increase the diversity of resistant cultivars and generate disruptive selection by spatial patterning of resistance genes within a region. In regions such as Khuzestan and Fars provinces, where pycnidia on stubble probably initiate disease epidemics as primary inoculum, cultural practices such as crop rotation and stubble management can play an important role in controlling the disease. These practices that aim to keep pathogen population sizes down can be particularly effective in Iran where the size of pathogen populations is small and therefore, their gene diversity is low. In these circumstances fungal adaptation to resistant cultivars and fungicide applications is expected to be lower than in large populations of this pathogen elsewhere (McDonald & Linde, 2002). Therefore, taking into account the genetic structure of M. graminicola and sporadic development of the disease in Iran, the management strategies are expected to be efficient in controlling epidemics and minimizing the application of fungicides.