Host-related genetic differentiation in the anther smut fungus Microbotryum violaceum in sympatric, parapatric and allopatric populations of two host species Silene latifolia and S. dioica

Authors

  • W. F. Van Putten,

    1. Department of Plant Population Biology, Netherlands Institute of Ecology, NIOO-KNAW, NL-6666 ZG Heteren, The Netherlands
    2. Department of Evolutionary Genetics, University of Groningen, Kerklaan, The Netherlands
    Search for more papers by this author
  • A. Biere,

    1. Department of Plant Population Biology, Netherlands Institute of Ecology, NIOO-KNAW, NL-6666 ZG Heteren, The Netherlands
    Search for more papers by this author
  • J. M. M. Van Damme

    1. Department of Plant Population Biology, Netherlands Institute of Ecology, NIOO-KNAW, NL-6666 ZG Heteren, The Netherlands
    Search for more papers by this author

W. F. Van Putten, Department of Evolutionary Genetics, University of Groningen, Kerklaan 30, NL-9751 NN Haren, The Netherlands.
Tel.: +31 50 363 2119; fax: +31 50 363 2348;
e-mail: w.f.van.putten@biol.rug.nl

Abstract

We investigated genetic diversity in West European populations of the fungal pathogen Microbotryum violaceum in sympatric, parapatric and allopatric populations of the host species Silene latifolia and S. dioica, using four polymorphic microsatellite loci. In allopatric host populations, the fungus was highly differentiated by host species, exhibiting high values of FST and RST, and revealed clear and distinct host races. In sympatric and parapatric populations we found significant population differentiation as well, except for one sympatric population in which the two host species grew truly intermingled. The mean number of alleles per locus for isolates from each of the host species was significantly higher in sympatric/parapatric than in allopatric populations. This suggests that either gene flow between host races in sympatry, or in case of less neutral loci, selection in a more heterogeneous host environment can increase the level of genetic variation in each of the demes. The observed pattern of host-related genetic differentiation among these geographically spread populations suggest a long-term divergence between these host races. In sympatric host populations, both host races presumably come in secondary contact, and host-specific alleles are exchanged depending on the amount of fungal gene flow.

Introduction

Plant parasites can often exploit more than one host species, and show intraspecific variation in host use. Different host species can represent different ecological niches to which the parasite can adapt by natural selection. When different host species occur in allopatry their parasite populations are isolated as well, and subject to random processes such as genetic drift especially when population sizes are small (Wright, 1931; Kimura, 1955). The process of genetic divergence is strongly enhanced by disruptive selection on habitat preference (Rice & Salt, 1990) or on fitness related aspects of the specific combination of host and parasite (e.g. in spider mites: Gotoh et al., 1993) where host and mate selection are correlated. As correlations between host (habitat) preference and assortative mating develop (e.g. offspring returning to the parental habitat to mate) host race formation can occur.

For plant-pathogenic fungi, several recent studies have used molecular markers to investigate the existence and genetic basis of host races, e.g. in Alternaria (brown spot fungus of citrus) and Macrophomina (charcoal rot fungus infecting root tissue of some crop species), showing genetic differentiation between host races and host preferences (Peever et al., 1999; Peever et al., 2000; Su et al., 2001) in allopatric host populations. In sympatric populations of hosts however, gene flow between fungal isolates may be common if prezygotic isolation mechanisms are weak or absent. In such cases gene flow between host races can occur either by host races infecting the ‘alien’ or non-native host species (individual movement) or, in species with sexual recombination, by alleles being exchanged between both host races (allelic gene flow), resulting in hybridization and introgression.

In this study we focus on the anther smut fungus Microbotryum violaceum, a well studied pathogen of the Caryophyllaceae (Pinks). Strains of this fungal pathogen that were isolated from a range of different allopatric populations of host species show varying degrees of host differentiation and specialization (Zillig, 1921; Biere & Honders, 1996a) and genetic differentiation based on karyotype (Perlin, 1996; Perlin et al., 1997), microsatellite data (Shykoff et al., 1999; Bucheli et al., 2000, 2001) and sequence variation at various genes (Freeman et al., 2002). In sympatry however, the effects of fungal gene flow on the extent of differentiation among fungal isolates, and whether fungal isolates from sympatric populations of hosts show host related genetic differentiation, is yet unclear. Two of the fungus’ host species, Silene latifolia and S. dioica, common dioecious herbs in Western Europe frequently occur in adjacent habitats (parapatry) despite their quite different habitat preferences. Interspecific hybrids between these plant species, reported to constitute more than 6% of a sympatric population (Biere & Honders, 1996b), clearly indicate extensive gene flow between these plant species. As this fungus is transmitted by pollinators of the host plants, this implies that there should be opportunities for fungal gene flow as well. Cross inoculation experiments have shown adaptations of fungal strains to their native host species for spore production although not for infection success (Biere & Honders, 1996a). Thus, cross-infection between strains could occur, depending on the behaviour of vectors, and gene flow could follow if fungal strains from the two host species can cross. Here, we investigate the extent of genetic differentiation and population sub-structuring of anther smuts in sympatric, parapatric and allopatric host populations. As no constraints on transfer of fungal strains between the two host species are expected in sympatric populations given the frequent hybridization between S. latifolia and S. dioica (Baker, 1947, 1948; Goulson & Jerrim, 1997), genetic population differentiation of the fungus with respect to host species could provide insight into whether host related differentiation is maintained despite the potential for gene flow. In addition, we study allelic diversity within fungal demes on each of the two host species in sympatry and allopatry. If fungal strains in sympatric populations indeed experience enhanced gene flow and/or a more heterogeneous (host) selection environment, we might expect that fungal demes on each of the host species could show higher levels of genetic diversity in sympatric host populations than fungal demes on the corresponding host species in allopatry.

Specific questions that will be addressed in this paper are: (i) to what extent are populations of M. violaceum genetically differentiated by host species, or by geographic distance? (ii) If M. violaceum shows overall genetic differentiation by host species, is such differentiation only observed in allopatric host populations, or also in sympatric/parapatric host populations where opportunities for fungal gene flow that can potentially disrupt host-specific differentiation are much larger? (iii) Is allelic diversity among fungal isolates from each of the host species higher when hosts occur in sympatry/parapatry than when they occur in allopatry?

Materials and methods

The species

The anther smut fungus Microbotryum violaceum (Pers.: Pers) Deml & Oberw. (=Ustilago violacea [Pers.] Fuckel) (Ustilaginaceae) (Deml & Oberwinkler, 1982) is a heterobasidiomycete that is obligatorily parasitic on susceptible members of the Caryophyllaceae, which it sterilizes to complete its life cycle (Baker, 1947). The fungus utilizes the anthers of its host to produce purple-brownish, diploid teliospores. In female hosts of dioecious species, the development of female reproductive tissue is halted (Audran & Batcho, 1981) and the expression of ‘male specific’ genes is induced (Scutt et al., 1997) leading to the development of anthers that produce fungal spores. Flowers of male hosts also contain teliospores in their anthers instead of pollen. Teliospores are diploid thick-walled cells that undergo meiosis when they germinate. After germination haploid sporidia of two mating types are produced that proliferate asexually by yeast-like growth. Sporidia of opposite mating type can conjugate to produce a dikaryotic infection hypha that enters the host. Spores are transmitted by the natural pollinators of their hosts, which also serve as vectors of this disease (Jennersten, 1983).

Silene latifolia Poiret (=Silene alba [Miller] Krause), the white campion, is a short-lived perennial weed that grows in open, disturbed habitats. S. dioica (L.) Clairv., the red campion, is a closely related perennial weed that mainly occurs at the edges of forests and in open woodland. Both species are dioecious and in areas where their habitats overlap hybridization frequently occurs (Baker, 1947, 1948; Goulson & Jerrim, 1997). Although both species are common in Western Europe, truly mixed sympatric populations are scarce or even absent because of differential habitat preferences.

Collection sites

We sampled eight populations of anther smut in Western Europe (see Fig. 1 for their geographical distribution), four from sympatric/parapatric populations of hosts, two from allopatric S. latifolia populations, and two from allopatric S. dioica populations. All populations contained at least a few hundred host plants at the time of sampling. The sympatric/parapatric populations are all patchy with respect to host species. At the Norg (Ng, 53° 06′ N, 6° 30′ E) sampling site (which has been studied extensively by Biere & Honders, 1996b, 1998), patches with predominantly S. latifolia and patches with predominantly S. dioica are close together, down to less than a meter. At the Abbertbos site (Ab, 53° 26′ N, 5° 49′ E), patches of both host species are further apart, down to a few tens of meters at closest, and at the Oxford site (Ox, 51° 41′ N, 1° 23′ W) patches with different host species are even more separated, down to a few hundred meters at closest from each other. The latter two populations could also be considered parapatric instead of sympatric. In the Kings Worthy population (Kw, 51° 05′ N, 1° 17′ W), S. dioica was present but no infected individuals were found, hence we lack fungal isolates from this host species at this site. The level of sympatry of this population was comparable with that of the Ab population. In all four sympatric/parapatric populations (Ng, Ab, Ox and Kw) hybrid hosts occurred within S. latifolia patches, rather than within S. dioica patches. Interspecific S. latifolia × S. dioica hybrids could be distinguished from the pure species forms by their intermediate morphology (Goulson & Jerrim, 1997), which is expressed in gradients of leaf shapes, flower colours and hairiness of stems (Baker, 1951). Hybrids may include both F1 and backcrosses. In populations that were classified allopatric in this study; Lac Vert (Lv, 48° 06′ N, 7° 07′ E), Meyendel (Md, 52° 10′ N, 4° 30′ E), Millingerwaard (Mw, 51° 52′ N, 6° 03′ E) and Wolfheze (Wh, 52° 00′ N, 5° 48′ E), the other host species was not observed within a range of a few kilometres.

Figure 1.

Geographical locations of the sampled smut populations of Microbotryum violaceum in Western Europe. Ng, Ab, Ox and Kw are sympatric/parapatric host populations containing Silene latifolia, S. dioica and interspecific hybrids, Wh and Mw are allopatric S. latifolia host populations, and Md and Lv are allopatric S. dioica host populations.

Sampling teliospores and microsatellite analysis

Teliospores were collected from as many infected male and female host plants as could be found per population. Closed flower buds were taken to avoid contamination where possible. While gently opening the flower buds, spores were collected in 1.5 mL microcentrifuge tubes. Although host plants can be multiply infected (Hood, 2003; Van Putten et al., 2003; Van der Wal et al., unpublished results), usually only one dikaryon succeeds in growing into a flower bud and in infecting that shoot (Day, 1980). Thus, teliospores isolated from single flower buds can be regarded as identical. Diploid teliospores from single infections were plated on standard medium (Cummins & Day, 1977). Haploid sporidia, produced after teliospore germination and meiosis, were grown for 1 week. From the agar medium, cells from many colonies were scraped off the plate, put into a microcentrifuge tube, and freeze dried for 24 h. Freeze dried samples were stored in a dry environment containing silica gel until DNA isolation. DNA was isolated using the PureGene Genomic DNA isolation kit for yeast (Gentra systems, Minneapolis, MN, USA). DNA was dissolved in a hydration solution (Gentra Systems) and stored at −20 °C until polymerase chain reaction (PCR) amplification. DNA from anther smut that is isolated following this ‘pooled culture’ procedure contains all genetic material that is present in the original diploid teliospore-parent, thus creating ‘pseudo-diploid’ DNA samples, provided that all meiotic products are produced and proliferate on the plate. We used four microsatellite loci that were developed by Bucheli et al., 1998.

The DNA from anther smuts isolated following the above-mentioned ‘pooled culture’ procedure only represents the genetic material present in the original diploid teliospore-parent if all meiotic products are produced and proliferate on the plate following teliospore germination. This might not always be the case. Some populations harbour strains that produce mating type biased offspring, because teliospores do not produce the full meiotic tetrad upon germination because of the presence of mating-type linked lethal alleles (Kaltz & Shykoff, 1997; Oudemans et al., 1998; Hood & Antonovics, 2000). This creates the potential problem of erroneously scoring isolates as homozygous at a locus when in fact one of the alleles is missing because half of the meiotic product has not proliferated because of mating-type linked lethals. In the studied populations, this problem appeared to be minimal. Isolates that were scored as homozygous at a particular locus generally showed higher peaks than heterozygotes that contained the same allele. Although we did not calibrate this, this may indicate that in the isolates that were scored as homozygous at the particular locus these alleles were indeed present in more than one copy. We are therefore confident that the pooled-culture DNA used in this study truly reflects the parental genotype.

Data analysis

To analyse genetic sub-structuring of fungal populations from the two host species we calculated observed (HO) and expected (HS) heterozygosities (Nei, 1987) using BIOSYS-1 (Swofford & Selander, 1987). To analyse genetic differentiation between fungal populations from the two host species we estimated FST (Weir & Cockerham, 1984) and RST (an F-statistic analogue that also takes into account differences in allele sizes; Slatkin, 1995), using FSTAT-2.9.3 (Goudet, 1995). Bootstrapped confidence intervals (RstCalc 2.2; Goodman, 1997 and GDA 1.1; Lewis & Zaykin, 2001) were used to test whether values were significantly different from zero (i) for isolates from the two host species that originated from the allopatric populations and (ii) similarly for isolates from the two host species that originated from each of the sympatric/parapatric populations. We used one-way anova to test whether FST and RST for the four sympatric/parapatric populations were significantly different from values for the four allopatric across-host species combinations.

To estimate gene diversity in fungal populations we calculated the mean numbers of alleles per locus (NA). To test whether NA within fungal demes collected from a particular host species was affected by the sympatry status of the population (sympatric/parapatric vs. allopatric), we performed ancovas with sympatry status of the population, host species, and the interaction between sympatry status and host species as dependent variables and sample size as a covariate.

Results

Population structure

Significant deviations from Hardy–Weinberg equilibrium (HW) were observed for all samples from S. dioica populations (1000 permutations; all P < 0.01), and for populations from allopatric S. latifolia (1000 permutations; all P < 0.05) (results not shown). Observed heterozygosities were lower than expected heterozygosities at three of the four loci (Table 1) and significant heterozygote deficit was observed in four of the populations (Table 2). Observed heterozygosities were significantly lower for samples from S. dioica than for samples from S. latifolia and hybrids (Table 2, contrast analysis procedure GENMOD in SASv8; inline image = 57.6, P < 0.001). All four loci contributed to significant genetic sub-structuring, both among populations and among host species, as evidenced by high values of FST and RST (Table 1).

Table 1.  Observed and expected heterozygosities (HO and HS), and estimates of genetic differentiation, FST an RST, for each of four microsatellite loci among isolates of the fungal pathogen Microbotryum violaceum from different populations and from the two host species Silene latifolia and S. dioica. Values for all four loci combined and 95% confidence limits for FST an RST are also indicated.
StructureLocusHOHSFSTRST
Among populations60.1380.3190.5980.808
110.6420.4030.2090.177
140.1130.4980.4590.386
180.1990.5340.3470.692
All0.2730.4390.4190.708
95% CI   0.299–0.5300.673–0.778
Among host species60.1270.5510.3790.767
110.5910.4030.2260.273
140.1340.6430.2730.315
180.2240.8440.1110.470
All0.2690.6100.2460.542
95% CI   0.149–0.3530.465–0.625
Table 2.  Molecular diversity in populations of the fungal pathogen Microbotryum violaceum, expressed as the mean number of alleles per locus (NA) and observed and expected heterozygosities (HO and HS) over four microsatellite loci. Sample sizes (n) in brackets.
PopulationSympatry status†Mean number of alleles per locus‡Heterozygosity
AllS.l.S.d.Hyb.HO ± SE§HS ± SE
  1. †S, sympatric; P, parapatric; A, allopatric.

  2. ‡Separate values are given for isolates from the hosts S. latifolia (S.l.), S. dioica (S.d.) and hybrids (Hyb.).

  3. §Significant heterozygote deficiency (P < 0.05) is denoted by an asterisk (*) in the column of HO.

Ng (38)S4.54.0 (21)3.0 (7)3.0 (10)0.31 ± 0.16*0.51 ± 0.13
Ab (39)S/P5.02.8 (17)3.8 (12)3.5 (10)0.41 ± 0.10*0.69 ± 0.05
Kw (14)S/P2.52.5 (12)1.5 (2)0.27 ± 0.220.37 ± 0.09
Ox (119)P5.54.5 (77)3.5 (36)2.5 (6)0.23 ± 0.12*0.60 ± 0.08
Lv (10)A2.02.0 (10)0.16 ± 0.110.34 ± 0.13
Md (15)A1.81.8 (15)0.00 ± 0.00*0.16 ± 0.10
Mw (15)A2.52.5 (15)0.27 ± 0.200.36 ± 0.16
Wh (9)A2.02.0 (9)0.29 ± 0.160.46 ± 0.05
Among populations (259) 8.3   0.27 ± 0.13*0.44 ± 0.09
Among host species (259) 8.3   0.27 ± 0.12*0.61 ± 0.10
 S. latifolia (151) 6.5   0.32 ± 0.18*0.58 ± 0.06
 S. dioica (80) 6.5   0.11 ± 0.02*0.58 ± 0.17
 Hybrids (28) 5.5   0.38 ± 0.16*0.67 ± 0.08

Host-related differentiation

Overall, there was significant genetic differentiation between the fungal isolates from the two host species (Table 1, FST = 0.246). For locus 6 (and to a lesser degree locus 18) we found host-specific alleles in allopatric populations of hosts (Fig. 2). Significant host-related differentiation was not only observed among strains from allopatric populations (Fig. 3), but also among strains from two of the sympatric/parapatric populations (Ab, Ox), showing significant values for both FST and RST (Fig. 3). Only in the Ng population no significant host-related differentiation was observed (Fig. 3). In this population, the alleles found among fungal isolates from S. dioica appeared to be a subset of the alleles observed among fungal isolates from S. latifolia from the same population, except for one allele at locus 6. FST values overall decreased from allopatric populations, where alternative hosts did not occur within 1000 m, to sympatric populations, where hosts co-occurred at scales down to 100 (Ox), 10 (Ab) and 1 (Ng) meter (Fig. 3). Values were significantly lower for the sympatric/parapatric populations than for the four allopatric across-host species combinations (F1,5 = 8.47, P < 0.05).

Figure 2.

Distribution of allelele sizes at locus 6 (A,B) and locus 18 (C,D) in allopatric (top) and sympatric populations (bottom) of the fungal pathogen Microbotryum violaceum sampled from the host species Silene latifolia (grey) and S. dioica (black) (n = 259). Note the bimodal distribution at both loci.

Figure 3.

Fixation indices FST and RST between anther smuts (Microbotryum violaceum) from the two host species Silene latifolia and S. dioica in sympatric, parapatric and allopatric host populations. Patches with different host species in the sympatric Norg host population grow truly intermingled (±1–10 m), in Abbertbos they are further apart (±10–100 m) and in Oxford they are more or less parapatric (±100–1000 m). Error bars represent 95% confidence intervals. Allopatric populations are Wh and Mw (smut from S. latifolia) and Md and Lv (smut from S. dioica).

The alleles at locus 6 that appeared to be host-specific in allopatric populations were frequently observed in heterozygous state in isolates from the sympatric/parapatric populations. If this were the result of hybridization and introgression upon (secondary) contact, the fraction of fungal isolates that are heterozygous with a mixed origin, i.e. having both host-specific alleles at locus 6, could be used as an upper estimate of the amount of effective fungal gene flow in such sympatric/parapatric populations. These values were 13.2% for the Ng (n = 38), 15.4% for the Ab (n = 39) and 1.7% for the Ox population (n = 119), indicating that the maximum effective gene flow in the sympatric/parapatric populations would be around 15%.

For the allopatric populations and the parapatric Ox population, RST values were larger than FST values (Fig. 3). Accordingly, analyses of molecular variance (amova, Michalakis & Excoffier, 1996) based on differences in allele frequencies only (assuming the infinite allele mutation model, FST), indicate that 25% of the observed molecular variance is because of differences between host species. However, when allele size is incorporated in the analyses as well (assuming the stepwise mutation model, RST), 57% of the observed variance could be attributed to host species. The difference between these amovas was primarily accounted for by large differences in the frequencies of the larger (S. dioica) and the smaller (S. latifolia and hybrids) alleles at loci 6 and 18 (Fig. 2). As these results may have been biased by the large sample size of the Ox population (46% of the samples) that showed the largest difference between FST and RST (Fig. 3), we repeated the analysis excluding this population. This yielded basically the same result. The proportion of the observed variance that could be explained by host differences was 20% for the FST based amova and 38% based on the RST based amova. Samples from S. dioica also showed higher frequencies of the larger alleles at locus 11, but had higher frequencies of the smaller alleles at locus 14.

Gene diversity in allopatric vs. nonallopatric host populations

We found 8, 3, 8 and 14 different alleles for locus 6, 11, 14 and 18, respectively, in the sampled anther smut populations. Gene diversity in samples from each of the two host species, expressed as the mean number of alleles per locus, was significantly lower in samples from allopatric host populations than in samples from sympatric/parapatric host populations (Fig. 4; ancovaF1,5 = 13.4, P < 0.05). The magnitude of the difference was similar for samples collected from S. latifolia and from S. dioica (no host species × sympatry status interaction, F1,5 = 0.45, n.s.). Similarly, allelic richness (El Mousadik & Petit, 1996), a measure of allelic diversity that is independent of sample size, was significantly different between samples from allopatric (RS = 1.93) and sympatric (RS = 2.81) populations of hosts (1000 permutations; P < 0.01). The pattern held true for fungal populations in both the Netherlands and the UK, strengthening the suggestion that higher levels of variation in the pathogen can be maintained in sympatric host populations.

Figure 4.

Allelic diversity (mean number of alleles per locus) at four microsatellite loci in the fungal pathogen Microbotryum violaceum, sampled from allopatric vs. sympatric/parapatric populations of its hosts Silene latifolia and S. dioica. Bars indicate 1 SE of the mean.

Discussion

Population structure

Significant substructuring of populations of M. violaceum was observed, both with respect to populations and host species. We found no evidence for isolation by distance, suggesting that differentiation within this fungus is host-related rather than related to the geographic distance between populations. In agreement with earlier studies of M. violaceum from S. latifolia and S. dioica (Bucheli et al., 2000, 2001) we observed significant heterozygote deficits (HD) in most of the studied populations. Values of HD (and derived statistics like FST) should be interpreted with caution in M. violaceum because intra-tetrad mating, which seems to be common in this species (Hood & Antonovics, 1998, 2000, 2003), can strongly affect levels of heterozygosity throughout the genome. As there is complete centromere linkage of the mating type locus, intra-tetrad mating (between cells of the two opposite mating types) will always bring together nuclei that were separated at the first meiotic division and hence will maintain heterozygosity at all loci tightly linked to their centromeres (Hood & Antonovics, 2000). Consequently, observed levels of heterozygosity are not necessarily interpretable through traditional population genetic models. Conversely, like selfing, intra-tetrad mating will result in rapid loss of heterozygosity at loci that are not closely linked to the centromere. In the absence of knowledge on linkage relationships of the studied loci, we cannot confirm whether the one locus (locus 11) showing a surplus of heterozygotes in our study is indeed more tightly linked to the centromere than the other three loci that show strong HD. Several alternative hypotheses could explain (part of) the observed heterozygote deficit, including the presence of null alleles, the Wahlund effect and high levels of (inter-tetrad) selfing. The presence of null alleles (Pemberton et al., 1995), which may cause erroneous scoring of samples in favour of homozygotes, does not appear to be a likely explanation as only few of our samples did not amplify during the PCR. The Wahlund effect, i.e. mixing of sub-populations that will lead to HD until one generation of random mating, might play a role. If sub-populations, in our case smuts sampled from different host species, do not mate randomly, HD will remain. Finally, selfing rates in this fungus seem to be quite high (Baird & Garber, 1979; Hood & Antonovics, 1998; Kaltz & Shykoff, 1999). In experimental inoculation studies using a mixed inoculum of teliospores carrying sporidial colony colour markers, Baird & Garber (1979) estimated that 93.4% of the teliospores produced after successful infection had resulted from inter- or intra-tetrad selfing. Similarly, Kaltz & Shykoff (1999) showed that conjugation rates between haploid sporidia derived from the same teliospore (selfing) were considerably higher than conjugation rates between sporidia from different teliospores (outcrossing).

Host related genetic differentiation

We observed strong host-related genetic differentiation among fungal isolates from the allopatric and two of the sympatric/parapatric host populations, confirming the existence of separate host races of M. violaceum on S. latifolia and S. dioica, as proposed by previous authors (Zillig, 1921; Biere & Honders, 1996a; Bucheli et al., 2001). Allele frequency differences at all four microsatellite loci contributed to the host related genetic differentiation. In agreement with earlier studies in Swiss populations (Bucheli et al., 2001), we found differences in the frequencies of larger and smaller alleles between samples from S. dioica and S. latifolia. The allele frequency distributions at two loci were clearly bimodal (Fig. 2), showing gaps of seven repeats (locus 6) and nine repeats (locus 18). When these fungal isolates are considered as one species, a more continuous distribution is expected under the stepwise mutation model (Kimura & Ohta, 1978). As multi-repeat jumps are considered to be rare (cf. Amos, 1999) and intermediate allele sizes have been reported at both loci in smut originating from other Caryophyllaceous host species (Shykoff et al., 1999; Bucheli et al., 2000), the gaps can indicate a long-term divergence between these two separate host races. By assuming a stepwise mutation model, large allele size differences indicate long-term divergence between the host races, at least for anther smut populations in Western Europe. One of the possible scenarios in that case would be that the two host races shared a common host species with a certain intermediate allele size at each of these loci. In the period subsequent to divergence from this common host, the mean allele length at the individual loci could have evolved independently, and in different directions.

Host-specific genetic differentiation between fungal isolates from S. dioica and S. latifolia was not only observed for strains from allopatric populations, but also for strains from two of the sympatric/parapatric populations. However, the fixation indices were significantly lower for the two most sympatric (Ng and Ab) populations than for the allopatric populations (Fig. 3), suggesting that the extent of differentiation is lower in more mixed populations where opportunities for fungal gene flow are increased. When the Ng population is considered in isolation, the allele frequency distribution of fungal isolates from the two host species might suggest a recent host shift from S. latifolia to S. dioica in this population, as alleles on the latter were largely a subset of the former. However, given that host-specific alleles of similar sizes were found in allopatric populations throughout Western Europe, the general picture that emerges is one of long-time divergence between strains from S. latifolia and S. dioica, and occasional secondary contact in the relatively rare parapatric or sympatric host populations. In such populations, the lower extent of host specific differentiation may then be the result of introgression. If so, the high fraction of strains in these populations that are heterozygous at loci showing strictly host-specific alleles in allopatry might reflect realized gene flow in such populations. However, it should again be noted that heterozygosity at loci closely linked to the centromeres may be easily maintained by intra-tetrad mating; the high values of such estimates (15% in the sympatric Ab population) should therefore be viewed as an upper estimate and interpreted with caution.

The reduced extent of differentiation in sympatric populations raises the question whether fungal strains from the two host species collected from sympatric populations are on average less well adapted to their respective hosts than fungal strains collected from allopatric populations. The limited data available (Biere & Honders, 1996a) do not seem to support this contention. Strains from S. latifolia and S. dioica, collected from the sympatric Ng population and six allopatric populations including Mw and Wh, on average show only slightly higher infection success on their native host species than on the alternative host (inline image = 2.81, P = 0.09) and the extent of host adaptation does not differ between population types (no interaction between allopatric vs. sympatric origin and native vs. alternative host species, (inline image = 1.60, P = n.s.).

Gene diversity of anther smuts in allopatric vs. sympatric host populations

Previous studies of genetic diversity within allopatric populations of the anther smut fungus M. violaceum have shown little variation in allozymes (Antonovics et al., 1996) in North America, and in microsatellite loci (Bucheli et al., 2001) in Switzerland. In allopatric populations of S. latifolia and S. dioica hosts, we find levels of variation – when variation is expressed in mean numbers of alleles per locus – that are comparable with these studies. Interestingly however, we find significantly higher levels of variation among fungal isolates from each of the host species when they originate from sympatric or parapatric populations than when they originate from allopatric host populations. This leads to the suggestion that higher levels of variation in the pathogen can be maintained in the presence of another host species and/or pathogen host race, even if the levels of fungal gene flow between fungi from one host species to another are low. In a scenario where the two host races have experienced long-term divergence and then come into secondary contact in parapatric or sympatric host populations, even low levels of gene flow will cause the mutual exchange of alleles and enlarge the variation in both host races, as was observed in a hybrid zone of two chromosome races of the common shrew (Wyttenbach et al., 1999). Selectively neutral variation may be maintained for long periods of time but will eventually get lost by drift. New variation can enter the subpopulation by gene flow, or by mutation. The higher levels of genetic variation in sympatric and parapatric populations may be maintained because fungal gene flow between sub-populations of different host species will always be higher in these populations than between allopatric populations of hosts, thus resulting in a different equilibrium of gene flow and drift. Alternatively, in sympatric host populations, the fungus experiences selection in a more heterogeneous environment through infection and reproduction on the two host species and their hybrids. A recurring change of host environment may thus slow down the fixation process, resulting in the maintenance of greater variation in the fungus. However, care should be taken in drawing conclusions from this dataset only. Populations are considered independent replicates in studying gene diversity in sympatric/parapatric vs. allopatric populations. This may not necessarily be the case.

Host-specific differentiation in sympatry: different markers, different answers

The results from this microsatellite study contrast with results of a survey of the allelic distribution at the yellow locus of the sporidial colony colour loci (SCC) in the Ng population (Van Putten, 2002) that showed clear and significant differentiation between the two host races at this locus, although to a significantly lower degree than what was observed between smut from allopatric host populations of these host species. This suggests that the yellow locus is subject to gene flow but that natural selection counteracts further convergence at this locus. Smuts from allopatric populations of S. latifolia are almost fixed for the yellow+ allele (sensuGarber et al., 1975) for sporidial colony colour (SCC) (Biere & Honders, 1996a; Van Putten, 2002), producing pink sporidial colonies when growing on standard medium because of high levels of cytochrome c (Will et al., 1984). Anther smut from allopatric S. dioica are almost fixed for the yellowy mutant allele, producing yellow sporidial colonies when growing on standard medium (Biere & Honders, 1996a; Van Putten, 2002). The yellow locus is one of several loci that affect sporidial colony colour (Garber et al., 1975). It specifies a polypeptide of the cyclase that catalyses the formation of β-ionone rings, resulting in β-carotene (Garber & Owens, 1980) causing the yellow colour of strains from S. dioica carrying this allele. Therefore, the selective neutrality of this SCC locus is unclear and there may be host-specific selection at the responsible locus. This would be in line with the observation that the degree of genetic differentiation as measured by coding genes (quantitative traits, expressed in QST) is typically larger than for neutral marker genes (Merilä & Crnokrak, 2001). Assuming that the Ng population of anther smut is in equilibrium, the selectively neutral markers may have converged, while variation at selectively less neutral markers may have maintained some of the differentiation.

Concluding remarks

Populations of anther smut sampled from allopatric and parapatric host populations of S. latifolia and S. dioica show significant differentiation. The genetic differentiation observed in this study is host related, rather than being related to the geographical distance between populations. The observation that host-specific microsatellite alleles that are separated by seven to nine repeats between fungal isolates occur in populations throughout Western Europe, may indicate a long-term divergence into two separate host races for S. latifolia and S. dioica. In (true) sympatry however, the extent of differentiation is much lower. Furthermore, the level of variation, expressed in mean number of alleles per locus, is higher anther smut from mixed (parapatric and sympatric) populations of hosts than in allopatry. Both suggest that there is interracial fungal gene flow that may cause both host races to converge when they come into secondary contact in sympatry.

Acknowledgments

The authors wish to thank Dave Goulson for help with finding and sampling the sympatric and parapatric populations of anther smut in the UK, Erika Bucheli for help with methodological issues, Jerome Goudet for useful discussions on analysing the data set, and Michael Hood, Peter van Dijk and two anonymous referees for their constructive comments on earlier versions of the manuscript. This study was financially supported by the Earth and Life Science Foundation (ALW) of the Netherlands Organization for Scientific Research (NWO), grant 805-36-391.

Ancillary