SEARCH

SEARCH BY CITATION

Keywords:

  • coalescent theory;
  • colonization history;
  • host switch;
  • migration rates;
  • NIP1;
  • pathogenic fungi

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

The origins of pathogens and their past and present migration patterns are often unknown. We used phylogenetic haplotype clustering in conjunction with model-based coalescent approaches to reconstruct the genetic history of the barley leaf pathogen Rhynchosporium secalis using the avirulence gene NIP1 and its flanking regions. Our results falsify the hypothesis that R. secalis emerged in association with its host during the domestication of barley 10 000 to 15 000 years ago in the Fertile Crescent and was introduced into Europe through the migration of Neolithic farmers. Estimates of time since most recent common ancestor (2500–5000 BP) placed the emergence of R. secalis clearly after the domestication of barley. We propose that modern populations of R. secalis originated in northern Europe following a host switch, most probably from a wild grass onto cultivated barley shortly after barley was introduced into northern Europe. R. secalis subsequently spread southwards into already established European barley-growing areas.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Colonization and introduction are associated with founder effects that alter patterns of genetic variation of taxa, for example by reducing variability. A gradient of genetic variability is, for example, a common observation in biota of Northern Europe, and convincingly explained by postglacial (re-) colonization from the south (Hewitt, 1999, 2000). Because these genetic signatures may persist for thousands of years, they provide a powerful tool for describing historical changes in a species’ distribution and demography. The field of genetic epidemiology of infectious disease uses this signature to trace origins and lines of descent for pathogen populations. This can provide critical information on the ‘how and when’ of pathogen emergence, and indicate how pathogens spread among and within their host populations. Most studies in this area have focused on human pathogens, with the malaria-causing Plasmodium species being the most prominent (Escalante et al., 2005; Mu et al., 2005).

Considerably less information is available for plant pathogens. Notable exceptions, however, provided unexpected novel insight into the range of possible population genetic structures for plant pathogens. One dramatic finding was that the fungal-like protist Phytophthora infestans, causal agent of the potato late blight disease that caused the Irish potato famine, existed as a single globally distributed clone (Goodwin et al., 1994) that originated in Mexico on wild Solanum species. At the other end of the spectrum of genetic structures, the wheat leaf blotch pathogen Mycosphaerella graminicola was found to undergo regular sexual reproduction and appeared to be globally panmictic, though no clones existed beyond spatial scales of 1 or 2 m (Zhan et al., 2003). For M. graminicola it has been hypothesized that the center of origin is in the Middle East, and that humans dispersed the pathogen globally during the expansion of wheat cultivation (McDonald et al., 1999; Banke et al., 2004; Banke & McDonald, 2005).

It is now generally accepted that the cradle of agriculture (also called the ‘Fertile Crescent’) was located around the eastern edge of the Mediterranean Sea, including today's Middle East countries of Israel, Syria and Jordan. Archeological remains indicate that barley (Hordeum vulgare) was the first crop domesticated in this region. Radiocarbon dating of the oldest barley grains found in Syria place the beginning of agriculture to 10 000–15 000 years BP (Badr et al., 2000). Following the path of human migrations in the Neolithic age (Cavalli-Sforza et al. 1994), agriculture and barley spread from the Fertile Crescent via Anatolia all over Europe and reached northern Europe about 8000 BP (Haak et al., 2005).

Rhynchosporium secalis is a haploid fungal pathogen causing leaf scald of barley. Rhynchosporium secalis is carried from season to season on infected crop residues or in the tissues of infected seed. Spores may also be spread onto the seedling by rain-splash from crop residues or neighbouring plants. Surprisingly, an earlier study using RFLP data from a global sample of pathogen strains showed that R. secalis displayed the highest genetic diversity in Scandinavia followed by Switzerland (Zaffarano et al., 2006). This led to the hypothesis that the pathogen did not originate at the center of origin of barley, the Fertile Crescent, nor in a secondary center of diversity of barley, Ethiopia.

Host–parasite associations can be ideal to infer colonization histories because the population genetic structure of the pathogen is coupled to, and reflects the evolutionary history of the host (e.g. Fisher et al., 2001; Falush et al., 2003; Wirth et al., 2005). In plant–pathogen systems, the co-evolutionary histories are particularly tightly coupled in the case of gene-for-gene interactions (e.g. Thrall & Burdon, 2003). In the classic gene-for-gene model (Flor, 1955), resistance genes recognize pathogen elicitors encoded by avirulence genes that activate the plant's defence system. An avirulence gene (NIP1) was previously identified in R. secalis that specifically interacts with its corresponding resistance gene (Rrs1) in barley (Rohe et al., 1995).

Here we used nucleotide sequences of the NIP1 avirulence gene to infer the geographical origin and colonization history of R. secalis. More specifically, we tested the hypothesis that R. secalis on barley originated in the Middle East during the domestication of its host plant and subsequently was introduced into Europe with the introduction of barley agriculture by early farmers. We compared our findings with previous studies in another cereal pathogen, M. graminicola.

Classic approaches such as indirect estimates of migration rates derived from FST values (Slatkin & Barton, 1989), assume that populations have reached an equilibrium between drift and migration, are of equal size and follow a Wright-island model of migration (Templeton, 1998; Neigel, 2002). However, the historical time frame we are investigating is likely too short for haplotypes drawn from populations to have sorted into distinct lineages (Avise, 1994; Avise & Wollenberg, 1997), and hence, these populations are probably not in equilibrium. Thus, we applied model-based coalescent methods to nucleotide sequences of the NIP1 gene to infer the origin and colonization history of R. secalis. These methods provide a biologically more realistic approach because they do not rely on the assumption of populations in equilibrium or equal population size, and also allow the determination of the direction of gene flow (Beerli & Felsenstein, 2001; Wilson & Rannala, 2003).

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Geographical locations of sampling sites and details of the samples collected are provided in Fig. 1 and Table 1 respectively. All R. secalis populations were collected from naturally infected barley fields using standardized sampling strategies (McDonald et al., 1999). Some of the samples have previously been included in population genetic studies on more local scales (references in Table 1) using isozymes (McDermott et al., 1989), and RFLPs (McDonald et al., 1999; Salamati et al., 2000). Nucleotide variation at the NIP1 locus was investigated in a previous study that differentiated two virulence mechanisms for this fungal avirulence gene (Schürch et al., 2004). A detailed description of DNA extraction, PCR and sequencing protocols was provided in the same publication. A subset of these samples (161 isolates, 1455-bp excluding indels and ambiguous sites) was used in this study.

image

Figure 1.  Geographical locations of the 17 sampled Rhynchosporium secalis populations. Also shown are the proportions of the 37 detected haplotypes pooled according to genetically distinct regions (see Tables 1 and 2). Shared haplotypes among regions are indicated by colours and grey segments represent unique haplotypes.

Download figure to PowerPoint

Table 1.   Country of origin for the 161 Rhynchosporium secalis isolates used in the present study grouped into geographical regions.
CountryCodeSampling siteHostnReference
AustraliaAUS  49 
 New South Wales1Wagga-WaggaBarley7McDonald et al. (1999)
 New South Wales2MethulBarley/barleygrass16McDonald et al. (1999)
 New South Wales3RannockBarley/barleygrass16McDonald et al. (1999)
 Victoria4HorshamBarley7McDonald et al. (1999)
 Western Australia5QuindanningBarley3McDonald et al. (1999)
USACAL  12 
 California6DavisBarley12McDermott et al. (1989)
GermanyGER  14 
7VariousBarley14Zaffarano et al. (2006)
Great BritainGBT  9 
8VariousBarley9Zaffarano et al. (2006)
SwitzerlandSWI  6 
9EschikonBarley6Linde et al. (2003)
JordanJOR  22 
10Al RabaBarley8Zaffarano et al. (2006)
11Al RabaBarley14Zaffarano et al. (2006)
SyriaSYR  12 
12BatajikBarley4Zaffarano et al. (2006)
13Al Jurn Al AswadBarley4Zaffarano et al. (2006)
14Aim IsaaBarley4Zaffarano et al. (2006)
FinlandFIN  7 
15JokioinenBarley7Salamati et al. (2000)
NorwayNOR  30 
16BuskerudBarley22Salamati et al. (2000)
17StjordalBarley8Salamati et al. (2000)

Population genetic parameters such as nucleotide diversity (π), haplotype diversity (h), and number of segregating sites (S) were calculated with arlequin v3.01 (Excoffier et al., 2005). The same program was used to assess Wright's F-statistics FST for population differentiation for all population pairs. Significance of differentiation was estimated with the implemented Markov chain bootstrap method. An amova (also implemented in arlequin) was used to test the hierarchical genetic structure of the populations. Components of genetic variance were computed at three levels. Among groups, populations were a priori pooled into groups according to major geographical regions (see Table 2). To test if the evolution of the gene followed a neutral model, Tajima's D statistic was calculated in DnaSP v4.10 (Rozas et al., 2003).

Table 2.   Results of hierarchical analysis of genetic variance (amova).
Source of variationVariance componentsPercentage variationFixation indicesProbabilities
  1. Populations were a priori assigned into five geographical groups: Australia (containing populations 1–5), California (6), Western Europe (7,8), Central Europe (9), Middle East (10–14), and Northern Europe (15–17).

  2. V, variance; F, haplotypic correlation at corresponding levels; P, probability of a more extreme variance component than that observed by chance alone.

  3. *FCT is defined as the correlation of random haplotypes within a group of populations, relative to that of random pairs of haplotypes drawn from the whole sample.

  4. FSC is the correlation of random haplotypes within populations, relative to that of random pairs of haplotypes from the group.

  5. FST is the correlation of random haplotypes within populations, relative to that of random pairs of haplotypes drawn from the whole sample (Excoffier et al., 1992).

Among groups0.854 Va19.83FCT: 0.198*P < 0.001
Among populations within groups0.759 Vb17.62FSC: 0.220†P < 0.001
Within populations2.693 Vc62.54FST: 0.375‡P < 0.001

Excluding the R. secalis populations from Australia and California, which have only recently been introduced by humans, our sample sites approximate the south-to-north migration route of Neolithic farmers from Syria (where the oldest remains of barley were detected) to northern Europe. Under a scenario of sequential range expansion from a single population source we would expect a decline of nucleotide and haplotype diversity along the migration route due to founder effects (Austerlitz et al., 1997). Thus, we tested the hypothesis that R. secalis was introduced together with its barley host by plotting haplotype diversity against geographical distance from Syria. To reduce the effect of different sample sizes, haplotype diversity was adjusted to the smallest sample size (SWI, n = 6) using a Monte Carlo re-sampling routine (1000 simulated data sets per population) implemented in PopTools v2.6 (Hood, 2005). Significance of genetic and geographical correlation was assessed with a Mantel test.

Recombination violates the assumption in traditional molecular phylogenetics that there is only one evolutionary history in a sequence data set. Thus, it is important to detect recombination prior to inferring phylogenetic relationship. Using the Φ statistics (phi-test) by Bruen et al. (2006) as implemented in the program SplitsTree v 4.6 (Huson & Bryant, 2006), we found no significant evidence for recombination (P = 0.136). To test the south-to-north migration hypothesis, genealogical relationships among R. secalis haplotypes were inferred by constructing a statistical parsimony based haplotype network using tcs v1.21 (Clement et al., 2000). This program established a 95% connection limit between haplotypes at 16 mutational steps. Phylogenetic relationships among all haplotypes were also reconstructed with the genetic algorithm for maximum-likelihood (gaml) program (Lewis, 1998). We used the priors recommended by Lewis and stopped the program after 30 000 generations because the search failed to find a more likely tree. We used this program because it allows for a fast estimation of phylogenetic trees based on the ML approach compared to most other programs when many haplotypes are sampled. Furthermore, gaml implements the Hasegawa et al. (1985) nucleotide substitution model. Preliminary analysis based on likelihood ratio tests (LRTs) implemented in modeltest v3.7 (Posada & Crandall, 1998) in conjunction with paup* suggested that the HKY substitution model with a gamma distribution of rate variation among sites provided the best fit to the R. secalis data set. Because of the lack of an appropriate outgroup and the failure to amplify the NIP1 region in related taxa, the resulting tree was a posteriori rooted in paup* with the implemented method of Lundberg using an hypothetical ancestor.

Migration among regions was estimated with migrate version 1.7.3 (Beerli & Felsenstein, 2001). migrate is based on the coalescent theory and uses a maximum likelihood approach to estimate effective population sizes and a migration matrix. It allows for asymmetrical gene flow among subpopulations and does not assume equal population sizes. Following the authors’ recommendations, we made a first migrate run with the default values using FST to find the start parameters. Parameters obtained from this initial run were used to re-run the program with the following Markov chain settings; short chains 10, long chains 5, averaging over 5 replicates and a 4-chain heating scheme.

We applied a gene-tree population-tree approach (Knowles & Maddison, 2002) to test alternative hypotheses about the colonization history of R. secalis using mesquite v1.06 (Maddison & Maddison, 2005). Briefly, the program uses coalescent simulations to evaluate the discordance between a population tree (i.e. our hypotheses) and a contained gene tree derived from the data set. The amount of discordance caused by incomplete lineage sorting (called ‘deep-coalescences’; Maddison, 1997) is measured as additional parsimony steps necessary to fit a gene tree within a population tree.

We used the Bayesian coalescent framework implemented in beast v1.3 (Drummond & Rambaut, 2003) to estimate demographic parameters and the time to the most recent common ancestor (TMRCA) for R. secalis. There are only few published studies on molecular mutation rates in fungi, perhaps also because of the scarcity of the fossil record in this group. However, Kasuga et al. (2002) provided a robust estimate of average mutation rates in protein coding genes of Eurotiomycetes. This estimate of 0.9 × 10−9 to 16.7 × 10−9 substitutions per site per year was included in the coalescent analysis to calibrate trees and thus to obtain absolute values of TMRCA and population size. These two parameters were assessed with different models of population growth available in beast; constant size, exponential growth, expansion growth, and logistic growth. Three Monte Carlo-Markov Chain (MCMC) searches were performed for each growth model to adjust priors and to ascertain convergence. The best model fit was assessed by comparing ‘posterior’ statistics, which is the combined log likelihood distribution of the tree and the coalescent model.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Polymerase chain reaction amplification and direct sequencing yielded a 1524-bp long DNA fragment including the NIP1 gene (GenBank accession numbers AY507678AY507845). In the amino acid sequence coding for the mature NIP1, 19 sites were variable with a total of 23 mutations. One mutation was synonymous, 17 were nonsynonymous and the remaining six mutations were all located at a single codon. Schürch et al. (2004) provided more detailed analyses and information on the variation both at the nucleotide and amino acid level of NIP1.

The geographical distribution and relative frequencies of the 37 distinct haplotypes are shown in Fig. 1. There was substantial geographical structuring indicated by grey segments representing unique haplotypes. The most prominent example is Norway with 11 out of 12 haplotypes unique to this region. The other extreme is Australia, which shared haplotypes with high frequency with European regions. One exception was the globally distributed haplotype 3 that was detected in Norway, Finland and Great Britain, and which apparently was also recently introduced into California and Australia. In sharp contrast, haplotypes found in the Middle East (H-31 to H-37) were shared only among populations within this restricted region. This observation is contradictory to the expected distribution if the Middle East was the origin of R. secalis.

Analyses of pairwise haplotype frequencies (FST) among all 17 populations revealed no significant genetic differentiation among populations within the same major geographical regions as a priori defined in Table 1. Consequently, we pooled these undifferentiated populations for all subsequent analyses into the nine major groups depicted in Fig. 1.

amova (Table 2) corroborated these findings and attributed only 17% of total genetic variation among populations within groups. Nineteen per cent of total variation explained differentiation among groups, suggesting moderate, yet significant structuring on larger geographical scales. However, with 62% the highest variance was clearly found within populations, which is in accordance with previous findings that most of the genetic diversity in R. secalis can be found over small spatial scales. For example, Salamati et al. (2000) showed that 75% of the total genetic diversity in R. secalis across a continent was reflected within a sampling plot of only 1 m2.

Standard diversity indices are summarized in Table 3. At face value, whereas nucleotide diversity was relatively equal among regions, haplotype diversity appeared to be lowest in the Near East and highest in more northern groups. A notable exception to this trend is the nucleotide diversity found in Jordan and the number of segregating sites from the same region. This was caused by the single highly divergent haplotype H-33 (see also Figs 3 and 4). Tajima's D-test for neutrality was not significant for any region or across all regions.

Table 3.   Genetic diversity indices (excluding indels) for the Rhynchosporium secalis NIP1 gene.
PopulationSampling sitesMSπ (SD)h (SD)Tajima's DSignificance
  1. M, number of haplotypes; S, number of segregating sites; π, nucleotide diversity; h, haplotypic diversity and Tajima's D neutrality test; ns, not significant.

Australia1–57110.002 (±0.001)0.673 (±0.060)−0.202ns
California6470.002 (±0.001)0.712 (±0.105)0.461ns
Germany77190.005 (±0.002)0.846 (±0.074)0.911ns
Great Britain85130.003 (±0.001)0.833 (±0.098)−0.227ns
Switzerland92130.004 (±0.001)0.500 (±0.265)0.006ns
Jordan10, 115470.013 (±0.006)0.693 (±0.076)1.234ns
Syria12–14480.002 (±0.001)0.758 (±0.080)0.986ns
Finland15580.003 (±0.001)0.857 (±0.137)0.722ns
Norway16, 1713180.002 (±0.001)0.899 (±0.030)−1.501ns
All1–1737820.005 (±0.002)0.947 (±0.008)−1.728ns
image

Figure 3.  Haplotype network for the NIP1 locus based on 1455-bp sequences from Rhynchosporium secalis isolates originating from seven geographical regions; red, Norway; blue, Finland; green, Switzerland; white, Germany; grey, Great Britain; yellow, Jordan; orange, Syria. Haplotype numbers correspond to Fig. 1. Size of circles is proportional to haplotype frequency. Each line between haplotypes indicates one mutational change, or step. Black dots are hypothetical haplotypes not detected in the data set.

Download figure to PowerPoint

image

Figure 4.  Maximum likelihood tree of all 37 distinct haplotypes generated using gaml. The localities the haplotypes were sampled from are listed to the right (see also Fig. 1). The search for more likely trees was stopped after about 30 000 generations as the program failed to find a more likely tree at a ln(likelihood score) = −2792.5591. Because of the lack of an appropriate outgroup taxa, the tree was a posteriori rooted using the method of Lundberg implemented in paup*. For clarity, bootstrap support values (> 50%; 1000 replicates) are indicated on branches for major clades only.

Download figure to PowerPoint

Haplotype diversity (h) plotted against geographical distance corroborated the amova findings of group differentiation and resulted in a highly significant correlation (r = 0.645; P = 0.009), with populations from the Middle East showing the lowest and northern Europe showing the highest diversity (Fig. 2). Thus, the distribution of R. secalis haplotype diversity did not reflect the expected decline from south to north associated with a hypothesized colonization in this direction. Instead the opposite colonization scenario was supported. This gradient of increasing diversity towards northern Europe did not change after adjusting for sample size. None of the simulated values deviated significantly from the empirical values, indicating that the data from populations with smaller sample sizes contained enough information to reliably reflect population genetic parameters.

image

Figure 2.  Empirical (white dots) and simulated (black) haplotype diversities (h) and standard deviations as a function of geographical distance from Syria, a known centre of origin for barley.

Download figure to PowerPoint

The R. secalis haplotype network reflected the distribution of Fig. 1 and showed considerable geographical clustering among major regions (Fig. 3). Haplotype 11, found only in Norway, had the highest outgroup probability based on coalescence theory (Donnelly & Tavaré, 1986; Castelloe & Templeton, 1994). The likelihood was calculated as a function of the position of the haplotype in the network, its frequency and its number of connections with neighbouring haplotypes. Haplotype 11 formed the centre of a star-like structure, indicating a population expansion or migration. This star in turn formed the center of the entire network. Two notable observations not revealed in Fig. 1 are (i) the highly divergent haplotype H-33 sampled five times in Jordan. This haplotype formed a distinct clade that required 40 mutational steps, which is significantly greater than the maximum of 16 steps (95%) to be included in the network. Neutral theory suggests that there is a relatively high probability of having a very divergent haplotype such as H-33 (Tajima, 1989). We included H-33 in all analyses except when using BEAST to estimate TMRCA. This program requires samples to be monophyletic, which could not be justified for H-33 because of the lack of suitable outgroups (see above). And (ii) although the network is generally well-resolved as indicated by the few hypothetical steps necessary to connect most haplotypes, the genealogy of samples from Switzerland is poorly resolved. This indicates the need to increase both sample size and sampled populations for this region in future studies.

The phylogenetic tree derived with gaml was fully congruent with the findings from the previous analyses in showing that all of the populations are paraphyletic for at least one lineage. Furthermore, it identified haplotypes from northern Europe as being basal for the entire tree (H-16, H-19) as well as basal for subclades within this tree (i.e. H-17, and supporting the hypothetical root of the haplotype network, H-11).

The coalescent simulations strongly suggest that the observed phylogeographic pattern of haplotype diversity reflects a colonization scenario for R. secalis from north-to-south (Fig. 5). The deep-coalescent value from the empirical gene tree was significantly lower than those values obtained from the simulated gene trees under the south-to-north hypothesis (Fig. 5a). Hence, the south-to-north colonization hypothesis can be rejected in favour of the alternative. In contrast, the deep-coalescent value from the empirical gene tree falls clearly within the values obtained from the simulated gene trees under the north-to-south hypothesis (Fig. 5b). Thus, it is highly unlikely that the colonization of Europe by R. secalis originated in the Middle East and moved northwards.

image

Figure 5.  Comparison of deep-coalescents from the empirical gene tree to the expected distributions from gene trees simulated by neutral coalescence for the two alternative hypotheses: (a) The introduction of R. secalis was from south to north and (b) R. secalis was introduced from the north to the south.

Download figure to PowerPoint

Migration rates among major geographical regions were generally low, and in 23 out of 30 pairwise comparisons were less than one migrant per generation (Table 4). The two exceptions were high migration rates into Australia (5.0) and into California (12.9). These two regions have only very recently (e.g. 200 years BP for Australia) been colonized by Europeans, and these elevated rates may reflect recent and repeated introductions through exchange of infected seed (Fig. 6).

Table 4.   Pairwise likelihood estimates of directional migration rates expressed as the number of migrants per generation (2Nm) between major geographical regions (95% confidence intervals in parenthesis). Donor populations are shown on the left and receiving populations are given along the top.
 AUSCALGER-GBTSWIJOR-SYRFIN-NOR
AUS0.49 (0.10–1.80)0.66 (0.22–1.65)0.00 (0.00–0.24)0.09 (0.01–0.35)0.33 (0.08–1.58)
CAL0.23 (0.08–1.07)0.00 (0.00–0.00)0.00 (0.00–0.48)0.18 (0.04–0.47)0.76 (0.24–1.70)
GER-GBT4.98 (1.53–9.28)1.13 (0.33–2.35)0.00 (0.00–0.00)1.17 (0.40–2.31)0.00 (0.00–0.42)
SWI0.22 (0.08–0.68)0.34 (0.14–1.18)0.87 (0.33–1.80)1.35 (0.59–2.51)0.00 (0.00–0.00)
JOR-SYR0.05 (0.01–0.20)0.00 (0.00–0.00)0.22 (0.01–0.63)0.08 (0.00–0.28)0.00 (0.00–0.00)
FIN-NOR0.42 (0.10–1.94)12.87 (6.92–20.68)1.22 (0.51–2.54)1.41 (0.67–2.62)0.38 (0.13–1.29)
image

Figure 6.  Summary of results placing the origin of R. secalis in northern Europe. Black arrows indicate migration rates between regions. Only number of migrants per generation > 1 are shown. Grey arrows are migration routes of Neolithic farmers into Europe (modified from Diamond & Bellwood, 2003 and Falush et al., 2003).

Download figure to PowerPoint

We tested four demographic models for inclusion in the Bayesian analysis (Table 5). Although the likelihood distributions of the models partially overlapped, expansion and logistic models showed an improvement over the simpler constant size and exponential growth models. This suggests that our data set contains enough power to estimate the additional parameters in these more complex models. We used a conservative approach and applied the lower end (0.9 × 10−9 substitutions per site per year) of the range of known mutation rates for fungi. For all four models estimates of the TMRCA for R. secalis gave a time of coalescence later than the first known cultivation of barley, which occurred 10 000–15 000 BP. For the logistic model, which returned the highest likelihood, TMRCA was between 2500 and 5000 years ago. Applying a higher mutation rate would result in an even more recent coalescence time.

Table 5.   Bayesian estimates of population size in Rhynchosporium secalis and TMRCA comparing four different demographic models.
Demographic ModelTMRCA (thousand years)Population size, N0 (thousand)Posterior (log-likelihood)
  1. Lower and upper 95% highest posterior density interval in parenthesis.

Constant4.79 (2.99 to 6.92)10.44 (6.26 to 15.12)−2779.31 (−2793.31 to −2766.46)
Expansion4.91 (2.67 to 7.38)65.77 (8.26 to 171.33)−2778.57 (−2791.87 to −2764.13)
Exponential3.44 (2.24 to 4.78)78.42 (28.76 to 123.10)−2775.41 (−2789.57 to −2762.41)
Logistic3.67 (2.45 to 5.06)257.29 (45.60 to 554.24)−2768.86 (−2782.96 to −2755.71)

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Here we described the global population structure of the barley scald pathogen R. secalis using nucleotide sequence variation of the NIP1 avirulence gene and its flanking regions. The distribution of R. secalis haplotype diversity did not reflect the expected decline from south to north associated with a hypothesized colonization in this direction. Instead the opposite colonization scenario was supported by an increasing diversity towards northern Europe. This north-to-south colonization scenario was further corroborated by (i) hierarchic phylogenetic clustering and haplotype networks that both identified haplotypes from Scandinavian populations as being more basal (more ancient), and the root, respectively, (ii) contrasting the two hypothesis using coalescent based simulations and (iii) directional estimates of gene flow. Comparing these results with the well-documented colonization history of Neolithic farmers and their cultivated cereal crops provided an opportunity to infer the geographical origin of this fungus and the role that early human migration may have played in spreading this pathogen.

We found consistent and compelling evidence that R. secalis originated in northern Europe and moved southward long after the introduction of barley. This proposed northern origin is supported by the known ecological and climatic preferences of the fungus, which causes its greatest damage in cool and moist areas. Hence, the evolutionary history of R. secalis does not mirror the colonization history of Neolithic farmers and cultivated barley. We conclude that R. secalis did not originate in the Fertile Crescent through a co-evolutionary process during the domestication of its barley host. This conclusion stands in sharp contrast with the M. graminicola-wheat pathosystem. In a recent study, Stukenbrock et al. (2007) showed by comparing M. graminicola populations from cultivated wheat and wild grasses that this fungus indeed emerged in the Fertile Crescent via host specialization during the domestication of wheat (10 000 ybp). Unfortunately, we were not able to test the secondary centre of diversity of barley, Ethiopia, as a possible origin because we were not successful in amplifying the NIP1 gene in those samples.

We hypothesize that two mechanisms contributed to the north-to-south migration of the pathogen. The first mechanism is human-mediated distribution of infected seed. Rhynchosporium secalis is known to be seed-borne, and we propose that Neolithic farmers moved infected seed from north to south through trade following the host switch that occurred between 2500 and 5000 years ago. The second mechanism is movement via wind-dispersed ascospores. There is accumulating evidence for the rare development of a sexual stage and associated wind-dispersed ascospores in R. secalis (Zaffarano et al., 2006).

Although this work represents a significant addition to our earlier understanding, the colonization history and evolution of R. secalis is far from being completely understood. Additional sampling and additional loci would allow us to obtain a more comprehensive picture of the genetic epidemiology of this important plant pathogen. Although all our analyses provided consistent results indicating the usefulness of the NIP1 locus, we recognize that it is risky to rely upon only one DNA region, which is under positive diversifying selection (Schürch et al., 2004) to infer the history of R. secalis. Many population genetic and phylogenetic applications assume (explicitly or otherwise) that the DNA sequence variation under consideration is selectively neutral. Interestingly, the basis for supporting this assumption as well as the consequences for inferring the correct phylogeny has never been thoroughly addressed. Although we acknowledge that NIP1 is likely under selection, our interpretation of the results is convincingly supported by other independent observations. Most importantly, published and ongoing studies using neutral RFLP and SSR genetic markers are in full concordance with this study both with regard to the genetic patterns observed and the interpretation of the colonization history of R. secalis. For example, global analyses of microsatellites (C.C. Linde & B.A. McDonald, personal communication) and RFLPs (Zaffarano et al., 2006) both showed a significantly higher genetic diversity in R. secalis populations from northern Europe compared with populations from the Middle East. The latter analysis found no significant differences between allelic richness and gene diversity in the Middle East, where barley has been cultivated for thousands of years, compared with R. secalis populations from Australia, South Africa or California, where barley was introduced as a cereal crop only during the last few hundred years.

Combining the findings described in this paper with earlier results, we hypothesize that the barley scald pathogen emerged in northern Europe approximately 2500–5000 years ago, in a host shift that occurred after barley was introduced as a crop into the region. After the new pathogen became established on the barley host, it migrated to southern Europe on infected seed via the grain trade and the movement of Neolithic farmers until it reached the Fertile Crescent. During the colonization of the New World by Europeans, the pathogen was moved to the Americas and Australia.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

We thank the many collaborators involved in collecting infected leaves, including David Moody (Australia), Robert Loughman (Australia), Barbara Read (Australia), Marja Jalli (Finland), Buskerud Extension Office (Norway), John Speakman (western Europe), Saideh Salamati (Scandinavia) and Vincent Michel (Switzerland). This work was supported by the Swiss National Science Foundation (Grant 31-56874.99).

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References