Population structure of South African and Australian Pyrenophora teres isolates


E-mail: cengen@lantic.net


There are two recognized forms of the disease net blotch of barley: the net form caused by Pyrenophora teres f. teres (PTT) and the spot form caused by P. teres f. maculata (PTM). In this study, amplified fragment length polymorphism analysis was used to investigate the genetic diversity and population structure of 60 PTT and 64 PTM isolates collected across Australia (66 isolates) and in the south-western Cape of South Africa (58 isolates). For comparison, P. tritici-repentis, Exserohilum rostratum and Bipolaris sorokiniana samples were also included in the analyses. Both distance- and model-based cluster analyses separated the PTT and PTM isolates into two strongly divergent genetic groups. Significant variation was observed both among the South African and Australian populations of PTT and PTM and among sampling locations for the PTT samples. Results suggest that sexual reproduction between the two forms is unlikely and that reproduction within the PTT and PTM groups occurs mainly asexually.


Net blotch, caused by the fungus Pyrenophora teres, is a serious production problem for the barley (Hordeum vulgare) industry in Australia, South Africa and elsewhere (Jonsson et al., 2000; Manninen et al., 2000; Gupta et al., 2003; Leisova et al., 2005). Two forms of net blotch exist: one is the net form (NFNB) caused by P. teres f. teres (PTT) and the other is the spot form (SFNB) caused by P. teres f. maculata (PTM) (Smedegård-Petersen, 1971). Lesions of NFNB are characterised by narrow, dark brown longitudinal streaks with transverse lines, giving the lesions a net-like appearance (Parry, 1990). Lesions may be surrounded by areas of chlorosis and large areas of dead tissue can be present. Lesions of SFNB are dark brown and elliptical in shape and may be surrounded by a chlorotic halo (Parry, 1990). As it can be difficult to distinguish between spot- and net-form lesions a number of polymerase chain reaction (PCR)-based assays have been developed that differentiate spot-form and net-form isolates (Williams et al., 2001; Keiper et al., 2008). A real-time quantitative PCR assay to differentiate between the two forms of the disease and to quantify the pathogen load in infected barley leaves was also developed (Leisova et al., 2006). It was previously suggested that recombination between the two forms can occur under field conditions (Campbell et al., 2002). However, a recent study of isolates collected mainly from Sardinia tested the patterns of sequence divergence and haplotype structure at the mating-type (MAT) locus of P. teres. The results suggested long genetic isolation between the net and spot forms of P. teres and that hybridization is rare or absent under field conditions (Rau et al., 2007). That study concluded that the two forms should be considered as different species when studying host resistance.

Although the net blotches can cause high yield losses in South Africa and have recently been observed more frequently, there are relatively few published studies of the disease (Louw et al., 1996; Campbell et al., 2002). Whilst work to date indicates that chemical control is not always effective and requires multiple applications (Campbell & Crous, 2002), limited breeding for resistance has been undertaken. To date, no one has determined the effectiveness of P. teres-resistant sources in South Africa, nor have virulence profiles of P. teres been studied.

In Australia 13 different pathotypes of PTT were identified among 81 isolates (Platz et al., 2000), whereas four pathotypes were identified when Canadian-derived barley lines were tested with eight PTM isolates from five geographically distinct regions, including Australia (Wu et al., 2003). The adoption of reduced- or zero-tillage practices has significantly increased the incidence of spot and net forms of net blotch in recent years (McLean et al., 2009).

The genetic structure of fungal pathogen populations is a key indicator of how rapidly a pathogen is evolving and can be used to predict how long a control measure or resistance source is likely to be effective (Campbell et al., 2002; Serenius et al., 2007). A number of studies have used molecular markers, such as amplified fragment length polymorphism (AFLP) and random amplified polymorphic DNA (RAPD), to investigate the genetic variation of P. teres isolates (Jonsson et al., 2000; Campbell et al., 2002; Rau et al., 2003; Leisova et al., 2005; Bakonyi & Justesen, 2007; Serenius et al., 2007). Most of these studies have used distance-based clustering methods to determine the degree of variation among accessions. This matrix is then used to construct a dendogram and accession clusters are identified manually (Pritchard et al., 2000). Pritchard et al. (2000) developed a model-based cluster method that is implemented in the software package structure. This program infers population structure and assigns individuals to populations. The advantages and disadvantages of distance- versus model-based clustering approaches have been discussed widely (Pritchard et al., 2000; Lu et al., 2005; Stajner et al., 2008).

This study examines the hypothesis that P. teres populations in South Africa and Australia are genetically distinct. DNA samples were analysed from 120 P. teres isolates collected from across South African (SA) and Australian (AUS) barley-growing regions. Following AFLP analysis, both the distance- and model-based clustering methods, as well as amova, were employed to determine the genetic diversity and population structure of the collected isolates. Genetic variation was examined within and between pathogen populations from the two geographical regions, South Africa and Australia.

Materials and methods

Fungal isolates and single-spore production

Fifty-eight SA isolates were obtained from barley leaf samples collected in 2007 from the major barley-growing region in the south-western Cape in an area of about 520 000 ha (Table 1). The original field identifications based on lesion appearance are listed in the second column of Table 1. Some of these SA samples were tentatively identified as showing symptoms of spot blotch (SB), caused by Bipolaris sorokiniana (column 2, Table 1). Once the field identifications had been confirmed (or otherwise) by diagnostic molecular markers (see below), the SA isolates were given the prefix PTT or PTM for net form and spot form, respectively. The 66 AUS P. teres isolates were obtained from the Department of Employment, Economic Development and Innovation (DEEDI), Hermitage Research Station (HRS) Queensland, from Dr Hugh Wallwork at the South Australian Research and Development Institute (SARDI) and from Dr Sanjiv Gupta at Murdoch University, Western Australia. Information on the origin, year of collection and host source of each isolate used is listed in Table 2. For comparison, six B. sorokiniana isolates (including isolates used in the study by Knight et al., 2009), one P. tritici-repentis isolate (PYSSc2) and one Exserohilum rostratum (previously known as Drechslera rostrata; DROSc3) isolate were included. The AUS isolates were given the prefix NB for net form, SNB for spot form and SB for spot blotch (B. sorokiniana) according to the diagnostic molecular marker classification (Table 2).

Table 1.   List of South African Pyrenophora teres isolates collected in 2007 in the south-western Cape from barley leaves, with leaf symptoms and geographic origin
IsolateSymptomsaTown/Sampling location
  1. aField identification based on lesion appearance; PTT, Pyrenophora teres f. teres; PTM P. teres f. maculata; SB, spot blotch (Bipolaris sorokiana).

Table 2.   List of Australian isolates used in this study, with symptoms on leaves and roots (Bipolaris only), geographic origin, year of collection and host
IsolateSymptomsaRegion/Sampling locationStateTownYearHost
  1. aSymptoms observed on leaves, or on leaves and roots (Bipolaris sorokiana only). PTT, Pyrenophora teres f. teres; PTM P. teres f. maculata; SB, spot blotch (B. sorokiana); CRR, common root rot (B. sorokiana); PYS, Pyrenophora tritici-repentis; DRO, Exserohilum rostratum

NBSA49/07PTTSouthSAYorke Peninsula2007Barley
NBHRS08119PTTSouthSAYorke Peninsula2007Barley
NB029PTTWestWAWongan Hills1985Barley
NB026PTTWestWANew Norcia1978Barley
NB063PTTWestWA15 km N of Williams1994Barley
NB090PTTWestWAWongan Hills1995Barley
NB132PTTWestWAWongan Hills1995Barley
NB335PTTWestWAWongan Hills2008Barley
NB336PTTWestWASouth Perth2008Barley
NB150PTMWestWA33 km E of Lake Grace1995Barley
NB154PTMWestWA22 km N of Nyabing1995Barley
NB160PTMWestWA25 km N of Katanning1995Barley
SNB175PTMSouthSAArno Bay1996Barley
SNB258PTMSouthSAArno Bay1996Barley
SNB049PTMSouthVICSwan Hill1993Barley
SNB167PTMWestWAMt Ridley1995Barley
SNB171PTMWestWAPalinup River1995Barley
SNB172PTMWestWAMt Barker1995Barley
SNB340PTMWestWAShenton Park2007Barley
SB08014SBNorthQLDAcacia Plateau2008Triticale
SB20004SBNorthNSWCasino2004Prairie Grass

Surface-sterilized leaf samples or sections of leaf lesions were placed on water-agar plates and incubated at room temperature and normal day/night light conditions for 2–3 days for production of conidia. Single conidia were transferred to potato dextrose agar (PDA; Biolab Merck; 39 g L−1) plates supplemented with streptomycin sulphate (0·3 mL Solustrep L−1) and then subcultured onto new PDA plates.

DNA extractions

Fungal mycelium was harvested from the single-spore cultures grown on PDA plates at 25°C for 1 week. A cetyl trimethyl ammonium bromide (CTAB) DNA extraction method was used to extract the fungal DNA (Saghai-Maroof et al., 1984). Extracted DNA was quantified using an Implen NanoPhotometer (Integrated Sciences). For reasons of maintaining quarantine between South Africa and Australia, fungal DNA was extracted in South Africa and sent to Australia for analysis.

Form-specific marker amplification

The forms of P. teres were verified using the two form-specific PCR markers of Williams et al. (2001) and seven diagnostic markers (hSPT2_4tcac, hSPT2_24tcac, hSPT2_6tcac, hSPT2_13agtg, hSPT2_4agac, hSPT2_13tcac and hSPT2_3agtg) of Keiper et al. (2008).

AFLP analysis

The AFLP procedure described by (Vos et al., 1995) was carried out using an AFLP Core Reagent kit (Invitrogen). The protocol provided by the supplier was followed. The EcoRI and MseI restriction enzymes were used to restrict approximately 150 ng DNA. After adaptor ligation a 1:10 dilution was performed using TE buffer. The dilutions were used in the pre-selective amplification with EcoRI (E-A or E-G) and MseI (M-A or M-G) primers with one extra base. Each pre-selective amplification reaction contained 5 μL diluted restricted-ligated DNA, 0·5 U GoTaq® Flexi DNA Polymerase (Promega Corporation), 4 μL 5X reaction buffer, 1·5 mm MgCl2, 200 μm dNTPs and 0·25 μmEcoRI and MseI primers with one selective nucleotide, in a total volume of 20 μL. The following pre-selective amplification cycling conditions were used: 20 cycles of 94°C for 30 s, 56°C for 1 min and 72°C for 1 min. These amplified fragments were then subjected to selective amplification using primers which had two extra bases (E-AA with M-AA, M-AG, M-AT and M-AC; and E-GC with M-GC, M-GA, M-GT and M-GG). The EcoRI primers were HEX-labelled. The selective PCR contained 2 μL pre-selective amplified DNA, 0·5 U GoTaq® Flexi DNA Polymerase, 3 μL 5X reaction buffer, 1·5 mm MgCl2, 200 μm dNTPs and 0·25 μmEcoRI and MseI primers with two selective nucleotides, in a total volume of 15 μL. The selective amplification cycling conditions were: 12 cycles of 94°C for 30 s, 65°C for 30 s and 72°C for 1 min, followed by 23 cycles of 94°C for 30 s, 56°C for 30 s and 72°C for 1 min. Four microlitres of 100% formamide loading buffer were added to the amplified samples. The samples were denatured for 4 min at 95°C and were visualized on 6% polyacrylamide gels using a Gel-Scan 2000™ DNA fragment analyser (Corbett Life Sciences). Gels were run for 90 min at 2500 V.

Scoring and data analysis

Both monomorphic and polymorphic bands were scored and used in the data analysis. Bands were scored independently by two people and bands which showed large differences in intensities or could not be scored accurately by both people were removed from further analysis. Amplicons produced by each primer combination were scored as binary data.

Distance-based clustering analyses

A similarity matrix was constructed using the dice coefficient (Dice, 1945) in the qualitative data program of the ntsys-pc version 2.20f software package. Cluster analysis of the matrix values was performed by employing the unweighted pair-group method with arithmetic mean (upgma; Sneath & Sokal, 1973) provided in the sahn program of ntsys-pc and a dendrogram was produced using tree plot. Clade support was assessed through a 300-replicate bootstrap test in winboot (http://www.irri.org/science/software/winboot.asp) to define confidence intervals (Felsenstein, 1985). Nodes were considered as unsupported when bootstrap values were less than 70%.

Analysis of molecular variance (amova)

transformer-3 (Caujapé-Castells & Baccarani-Rosas, 2005) was used to convert the format of the AFLP data from the Microsoft Excel file format to the arlequin file format. Analysis of molecular variance was computed using the software arlequin version 2.0 (Excoffier et al., 1992; Schneider et al., 2000) with 1000 permutations. Genotypic data was partitioned into groups in order to test for genetic variation between PTT and PTM isolates overall and among (between) countries. Subsequently amova was performed for each P. teres form to test for genetic variation among and within sampling locations. As the isolate sampling area in South Africa was much smaller than that in Australia, the towns listed in Table 1 were used as the sampling locations for the SA samples, whereas the regions indicated in Table 2 were used as the sampling locations for the AUS samples.

Model-based clustering analyses

structure version 2.2 (Pritchard et al., 2000) was used to infer the genetic structure and the number of clusters or populations (K) in the dataset. Initially, the analysis was carried out on the whole AFLP dataset, consisting of 10 independent runs performed for each value of K, K varying from one to 11. The default settings of the program were used, i.e. admixture ancestry model, uncorrelated allele frequencies between populations and the degree of admixture, alpha, inferred from the data (Pritchard & Wen, 2003; Falush et al., 2007). Each run was set to a burn-in period of 10 000 iterations followed by 100 000 Monte Carlo Markov Chain (MCMC) iterations. As the outlier samples (B. sorokiniana, P. tritici-repentis and E. rostratum) interfered with the resolution of the analyses, they were omitted from further analysis. Final structure analysis was performed on the separate PTT and PTM AFLP datasets (PTM and PTT; as established by the form-specific markers and upgma cluster analysis). Ten independent runs of K, set between one and six, were performed for each dataset and to determine the consistency of the results the burn-in was increased to 50 000 followed by 500 000 MCMC iterations.

To estimate the most likely number of clusters the logarithmized probability of data [Pr(X∣K] or L(K) for each value of K (Pritchard et al., 2000) was used. This was compared to the statistic delta KK), the second-order rate of change of the likelihood function with respect to K (Evanno et al., 2005). In brief, the mean difference between successive likelihood values, L′(K), was calculated, after which the absolute values of the difference between successive values of L′(K), i.e. L″(K), were averaged over 10 runs and divided by the standard deviation.

clumpp version 1.1.1 (Jakobsson & Rosenberg, 2007) was used to permute one output from the 10 independent cluster outputs produced by structure. Graphs were constructed in Microsoft Excel 2007.


The classification of isolates as PTT or PTM according to their field identifications was verified prior to AFLP analysis by amplification with the PTT- and PTM-specific primers of Williams et al. (2001) and seven diagnostic microsatellite markers produced by Keiper et al. (2008). A further two diagnostic microsatellite markers from that study did not produce clear bands (hSPT2_24agac and hSPT2_5agac) and were discarded. Amplification with the selected markers indicated that four isolates (NB143, NB150, NB154 and NB160) originally identified as PTM were actually PTT (Table 2). Thirteen SA samples had been classified from leaf symptoms as B. sorokiniana, but microscopic studies of spore morphology indicated that they were Drechslera teres (anamorph of P. teres) (Sivanesan, 1987). The PTM markers amplified on DNA from these accessions and they were therefore reassigned as PTM samples (Table 1). This was confirmed by AFLP analysis (see next section). The PTT- and PTM-specific markers did not amplify on samples SB230 and SB170 from Australia and AFLP analysis confirmed that these samples were not P. teres isolates as they clustered with the B. sorokiniana isolates (see next section).

AFLP analysis

AFLP analysis was conducted on DNA of 23 SA and 37 AUS PTT isolates, 37 SA (including two controls discussed below) and 29 AUS isolates identified as PTM in the field, six B. sorokiniana isolates, one P. tritici-repentis and one E. rostratum isolate (Tables 1 and 2). Eight AFLP primer combinations were used and on average 50 loci were amplified with each primer combination. In total, 400 loci could be accurately scored across all samples and 168 of these loci (42%) were polymorphic in the P. teres samples. Independent DNA preparations (samples PTT37#1 and PTT37#2) and independent polymerase chain reaction samples (sample PTT21) produced the same banding patterns.

Distance-based clustering analysis

The distance-based clustering analysis subdivided the 134 isolates into seven groups (Fig. 1). The two main groups, I and III, contained 60 PTT and 59 PTM isolates (plus two controls), respectively. Group II consisted of one PTM isolate, SNB172, which clustered away from the main PTM group and only showed 74% similarity with the other PTM isolates. Two SA isolates (PTM28#1 and PTM63#2) formed a cluster on their own (group IV) and their identity could not be determined. The outliers, i.e. the P. tritici-repentis and E. rostratum isolates formed groups V and VI, respectively, and the six B. sorokiniana isolates together with the two AUS isolates originally classified as PTM (SB170 and SB230) formed group VII. Based on the coefficient of similarity, the P. teres isolates were only 14, 12 and 9% similar to these three outlying groups, respectively.

Figure 1.

 Dendogram produced using upgma cluster analysis based on DICE’s similarity coefficient calculated from 400 AFLP fragments using the following isolates: 23 South African Pyrenophora teres f. teres (PTT), 37 Australian PTT, 35 South African P. teres f. maculata (PTM), 29 Australian PTM, six Bipolaris sorokiniana, one Pyrenophora tritici-repentis and one Exserohilum rostratum.

Within the PTT and PTM groups the similarity between individuals was very high, with a minimum value of 90% (Fig. 1). However, distinct clusters containing only SA and AUS isolates were present. All but two of the SA PTM isolates clustered together in group III. PTM32 clustered with the AUS PTM isolates and PTM21 formed a cluster on its own. The AUS PTM isolates clustered together in group III, with the exception of two isolates (SNB167 and SNB222), which formed a separate cluster. The separation of the SA and AUS isolates within the PTT group was not as clear, with a number of SA isolates clustering amongst the AUS isolates.

Analysis of molecular variance (amova)

amova revealed highly significant differences among the two forms of P. teres, PTT and PTM, contributing 36·0% (< 0·0001) of the total genetic variation (Table 3). The smallest proportion, 17·72% (< 0·0001), was ascribed to the country of origin (SA or AUS), whilst 46·28% (< 0·0001) of variation existed within the individual P. teres populations.

Table 3.   Analysis of molecular variance (amova) of South African (SA) and Australian (AUS) Pyrenophora teres populations
Source of variationDegrees of freedomSum of squaresVariance componentsVariation (%)
  1. *< 0·001; **< 0·0001.

  2. ns, not significant.

  3. PTT, Pyrenophora teres f. teres; PTM, P. teres f. maculata.

Among groups (PTT vs. PTM)1683·828·7936·00**
Among populations within groups (SA vs. AUS)2277·124·3317·72**
Within populations1181332·9111·2946·28**
Among sampling locations (towns) for SA PTT444·841·7932·81**
Within sampling locations (towns) for SA PTT1761·303·6167·19*
Among sampling locations (regions) for AUS PTT222·850·518·89*
Within sampling locations (regions) for AUS PTT34177·235·2191·11**
Among sampling locations (towns) for SA PTM319·860·539·45ns
Within sampling locations (towns) for SA PTM29146·325·0690·55*
Among sampling locations (regions) for AUS PTM215·490·142·15ns
Within sampling locations (regions) for AUS PTM24155·956·5097·85*

In the inter-population amova, highly significant variation (32·81%; < 0·0001) was observed among SA PTT samples depending on sampling location, whereas no significant variation was observed with the SA PTM samples (Table 3). Although the variation observed among sampling locations with the AUS PTT samples was lower than with the SA PTT samples, it was still significant (8·89%; < 0·001), whilst no significant variation was observed with the AUS PTM samples among sampling locations. Variations within sampling locations were highly significant (< 0·001) for PTT and PTM isolates from both South Africa and Australia (Table 3).

Model-based clustering analysis

For the model-based clustering in structure the AFLP data were analysed separately for the PTT and PTM groups. Averaged over replicates, the log-likelihood values (as described in Pritchard et al., 2000) divided both the PTT and PTM isolates into three clusters, as the highest values [L(K)] were observed at = 3 (Fig. 2a). When the rate of variation in likelihood values between successive K values (ΔK statistic of Evanno et al., 2005) was examined, the number of clusters was subsequently reduced to two for the PTT isolates, whilst the ad-hoc statistic also indicated that the PTM isolates were assigned to three clusters (Fig. 2b). Even when the burn-in iterations were increased to 100 000 followed by 1 000 000 MCMC iterations for the PTT isolates, the highest value for ΔK was observed at = 2. The ad-hoc statistic method was chosen to determine the final value of K, as this is the recommended method to determine the number of clusters (Evanno et al., 2005; Basset et al., 2006; Stajner et al., 2008).

Figure 2.

structure analyses: most likely number of clusters (K) for Pyrenophora teres f. teres (PTT) and P. teres f. maculata (PTM) according to (a) mean log (ln) probability values [L(K) estimated over 10 independent runs for each value of K] and (b) values of ΔK calculated for each K.

Using the model-based cluster analysis, samples were then assigned into a specific group based upon the highest percentage of membership or co-ancestry (Fig. 3). Most isolates could be assigned to a specific group as they shared more than 80% common ancestry. A minority of isolates shared less than 80% similarity and were considered to be of mixed origin, i.e. they were representative of more than one group. The first group of PTT isolates consisted solely of 24 AUS isolates, whereas the second group consisted of seven AUS and 22 SA isolates. Six AUS (NBHRS08119, NB085, NB330, NB308, NB327 and NB321) isolates and one SA isolate (PTT51#1) were distinctly of mixed origin (Fig. 3a). In the PTM group one population consisted only of AUS isolates, the second consisted only of SA isolates and the third group consisted of only one AUS isolate (SNB172). Five AUS isolates (SNB164, SNB264, SNBHRS07033, SNB167 and SNB222) and six SA isolates (PTM66#1, PTM25, PTM67#2, PTM39#2, PTM32 and PTM21) could not be assigned to a population as they were of mixed origin (Fig. 3b).

Figure 3.

Figure 3.

 Population structure and upgma clustering of 120 South African and Australian Pyrenophora teres f. teres (PTT) (a) and P. teres f. maculata (PTM) (b) isolates. The estimated population structure is indicated on the left. Each individual is represented by a horizontal line; shading represent inferred populations. Clusters produced by upgma are indicated on the right. Samples of mixed origin are boxed.

Figure 3.

Figure 3.

 Population structure and upgma clustering of 120 South African and Australian Pyrenophora teres f. teres (PTT) (a) and P. teres f. maculata (PTM) (b) isolates. The estimated population structure is indicated on the left. Each individual is represented by a horizontal line; shading represent inferred populations. Clusters produced by upgma are indicated on the right. Samples of mixed origin are boxed.

The seven PTT isolates of mixed origin (NBHRS08119, NB085, NB330, NB308, NB327, NB321 and PTT51#1) observed with the model-based (structure) analysis were clustered amongst the other samples in the distance-based (ntsys) analysis and could therefore not be distinguished from the rest. Four of the 11 PTM isolates (PTM67#2, SNB167, SNB222, PTM21) of mixed origin also formed a cluster to one side of the main group in the distance-based analysis (Fig. 3). The other seven PTM isolates of mixed origin (SNB164, SNB264, SNBHRS07033, PTM66#1, PTM25, PTM39#2 and PTM32) clustered amongst the main group in the dendogram.


In this study, 58 SA and 66 AUS P. teres monoconidial isolates were investigated through AFLP analysis to establish genetic differences among and within these fungal populations. Previous studies have investigated the genetic variation of P. teres isolates collected from different regions all over the world (Jonsson et al., 2000; Campbell et al., 2002; Rau et al., 2003; Leisova et al., 2005; Bakonyi & Justesen, 2007; Serenius et al., 2007). However, most of these studies employed RAPD markers and did not use a large number of molecular markers in the cluster analysis. By contrast, the present study employed a significant number of AFLP markers and then used a model-based cluster analysis to determine the distribution of P. teres isolates. Apparently, only one other study has so far employed a model-based clustering approach to determine population structure in a fungus (Bayon et al., 2009).

Diagnostic markers (Williams et al., 2001; Keiper et al., 2008) were used to verify that the initial field classification of the isolates into the PTT and PTM groups was correct. This indicated that four Australian isolates had been misclassified when sampled. Furthermore the diagnostic markers amplified on 13 SA isolates which had been tentatively classified as SB (B. sorokiniana) indicated these accessions were all PTM. The difficulty with field identification based on visible symptoms has been recognized in other studies (Rau et al., 2003; Leisova et al., 2005). Furthermore the two forms of P. teres are difficult to discriminate based on spore morphology (Crous et al., 1995). This suggests that diagnostic markers should be used more frequently to classify barley foliar diseases, in particular NFNB, SFNB and SB, which are difficult to distinguish and can easily be mistaken for other spot-like symptoms (e.g. boron toxicity and genetic necrosis) on the leaves (Campbell et al., 2002).

Distance-based cluster analysis using AFLP markers separated the SA and AUS P. teres isolates into two distinct groups consistently identified as PTT and PTM by the diagnostic markers. A clear differentiation was observed between these groups and the other leaf pathogens used as outliers (P. tritici-repentis, E. rostratum and B. sorokiniana). The 14% genetic similarity between the P. teres and P. tritici-repentis isolates was similar to the 19% observed by Singh & Hughes (2006), who conducted a cluster analysis on 33 P. tritici-repentis isolates using two P. teres isolates as outliers. The similarity observed between the PTT and PTM clusters was 65%, thus clearly separating the two forms. Within each cluster the variation was very low (minimum similarity of about 90%). Similar results were observed in a number of other studies. For example, minimum similarities of 88% and 91% were observed within the PTT and PTM groups, respectively, in a study using RAPD analysis to determine the genetic relationship between 32 P. teres isolates from geographically diverse areas (Bakonyi & Justesen, 2007). In this study, the similarity between the PTT and PTM groups was also high (84%). In another study, nineteen reproducible RAPD loci were used to determine the genetic diversity of two Swedish net blotch populations each consisting of 64 monoconidial isolates. A mean similarity of 90% based on the genetic distance coefficients among all subpopulations was observed (Jonsson et al., 2000). By contrast, a high level of variation was observed in a study using AFLPs and the upgma cluster method with 37 PTT and 30 PTM isolates collected mainly from the Czech Republic and Slovakia (Leisova et al., 2005). The clear distinction between the two forms of P. teres was further confirmed by amova, with highly significant differences observed among the PTT and PTM groups. This suggests that sexual crossing between the two forms does not occur in these two countries and is in agreement with other studies indicating that recombination is rare between the two forms (Rau et al., 2003; Serenius et al., 2007). It is, however, in disagreement with the findings of Campbell et al. (2002), which indicated that sexual reproduction between the two forms is likely within SA barley-growing regions. Campbell et al. (2002), using RAPD analysis, identified unique net- and spot-type DNA bands in one isolate and therefore concluded that sexual recombination may occur between the two forms.

Significant variation was observed among the SA and AUS populations within groups (PTT and PTM), indicating that for each form of P. teres, South Africa and Australia harbour genetically distinct lineages, thus confirming the initial hypothesis. However, it is unlikely that these AFLP analyses can identify variation at the pathotype level, as several groupings at or above the 94% level of similarity contained isolates of distinctly different virulence spectra (data not shown). Further studies are planned to compare SA and AUS isolates using a common set of differential lines to determine the pathotype variation in SA net blotch populations.

A significant difference was also observed amongst sampling locations in which SA PTT and AUS PTT samples had been collected. By contrast, in the SA and AUS PTM samples the percentage of variation within sampling locations was much greater than among sampling locations and no significant variation was observed among either SA or AUS sampling locations. A similar partitioning of genetic variation was found by Serenius et al. (2007), who examined 116 AUS isolates of PTT and PTM using two AFLP primer combinations (87 unique genotypes). They also found higher genetic variation within than among sampling locations (fields). The findings of the present study also agree with those of Rau et al. (2003), who concluded from their study of Sardinian P. teres isolates that genetic divergence among PTT populations is higher than among PTM populations.

Pairwise genetic similarity values are calculated as a proportion of loci with shared alleles in distance-based approaches used in programs such as ntsys. With this approach the number of groups identified is based on a subjective cut-off made by the user (Pritchard et al., 2000; Lu et al., 2005). The model-based clustering algorithm used in programs such as structure identifies subgroups of accessions with distinct allele frequencies within the samples tested (Maccaferri et al., 2005). Whereas samples are not overlapping in the distance-based analysis, in the model-based analysis each sample is allowed to have membership of several different subgroups, with membership coefficients totalling one (Maccaferri et al., 2005). It was indicated previously that the ability of structure to converge to a robust solution is reduced when using systems with a complex structure (Stajner et al., 2008; Kiær et al., 2009). It was also found that by excluding the outliers in the structure analysis and by subdividing the P. teres isolates into the two forms, the ability of the program to distinguish differences within the PTT and PTM groups was increased. The distance-based model is therefore useful to first identify subgroups, which subsequently can be analysed in structure.

The high level of genetic relatedness observed within the PTT and PTM groups in the model-based cluster analyses suggests that reproduction in these fungi is, in the main, asexual. However, a small number of isolates of mixed origin were identified for both forms and therefore sexual reproduction cannot be entirely excluded. Of the SA isolates, six PTM were of mixed origin versus only one PTT sample. This suggests that in South Africa, sexual crosses between PTM isolates may be more frequent than for PTT. The occurrence of sexual recombination in P. teres has been suggested by several studies in different environments (Campbell et al., 2002; Wu et al., 2003), whilst Rau et al. (2003) proposed that the relative contribution of sexual and asexual reproduction varies among different environments. The AUS isolate SNB172 was distinct from other PTM isolates both in the distance- and model-based cluster analyses and needs to be further investigated.

Unfortunately P. teres samples collected in South Africa in the past (Campbell et al., 2002) have not been preserved (P. W. Crous, Centraalbureau voor Schimmelcultures, Netherlands, personal communication). Samples collected for this study will form the basis of an on-going collection available for future studies.

In conclusion, the AFLP analysis of PTT and PTM isolates indicated high genetic variation among the two forms of P. teres, as well as among the SA and AUS isolates. amova analysis indicated that genetic variation was considerably higher among sampling locations for PTT than for PTM isolates, whilst genetic variation was high within sampling locations for both. Overall, these results suggest that sexual reproduction/recombination between the two forms is unlikely; isolates are probably specific to a geographical region they occur in; and reproduction within the PTT and PTM groups occurs mainly asexually.


The authors would like to thank Dr Hugh Wallwork (SARDI) and Dr Sanjiv Gupta (Murdoch University) for the isolate samples provided by them. We also would like to thank Debbie Snyman, Denise Liebenberg and Lizaan Rademeyer for their technical help in the laboratory. This project was funded by the South African Winter Cereal Trust.