Present address: Department of Zoology, University of British Columbia, Vancouver, British Columbia, V6T1Z4, Canada.


Cryptic structure of species complexes confounds an accurate accounting of biological diversity in natural systems. Also, cryptic sibling species often become specialized to different ecological conditions, for instance, with host specialization by cryptic parasite species. The fungus Microbotryum violaceum causes anther smut disease in plants of Caryophyllaceae, and the degree of specialization and gene flow between strains on different hosts have been controversial in the literature. We conducted molecular phylogenetic analyses on M. violaceum from 23 host species and different geographic origins using three single-copy nuclear genes (β-tub, γ-tub, and Ef1α). Congruence between the phylogenies identified several lineages that evolved independently for a long time. The lineages had overlapping geographic ranges but were highly specialized on different hosts. These results thus suggest that M. violaceum is a complex of highly specialized sibling species. Two incongruencies between the individual gene phylogenies and one intragene recombination event were detected at basal nodes, suggesting ancient introgression events or speciation events via hybridizations. However, incongruencies and recombination were not detected among terminal branches, indicating that the potentials for cross-infection and experimental hybridization are often not sufficient to suggest that introgressions would likely persist in nature.

Species complexes are composed of genetically isolated lineages that are not distinguishable on the basis of purely morphological criteria. Such difficulties have been encountered in almost all taxonomic groups, even the most studied birds and mammals (e.g., Burbidge et al. 2003; Ravaoarimanana et al. 2004), but especially in groups with fewer obvious taxonomic criteria like fungi (e.g., Geiser et al. 1998; Koufopanou et al. 2001; Pringle et al. 2005). Detecting the cryptic structure of species complexes is essential for an accurate accounting of the biological diversity in natural systems, especially as the sibling species often have evolved different specializations. Investigating cryptic genetic structure in relation to specialization is particularly relevant in parasites. For example, many plant enemies considered as broad generalists have recently been recognized as complexes of sibling species, specialized on different hosts; see for instance reviews by Burnett (2003) for fungi and by Dres and Mallet (2002) for insects.

Several tree-based methods are used to delimit sibling species (Sites and Marshall 2004). Multiple-gene phylogenetic approaches have been particularly useful within the fungal kingdom (e.g., Koufopanou et al. 1997; Geiser et al. 1998; O'Donnell et al. 2000; 2001, Dettman et al. 2003a; Kasuga et al. 2003; Staats et al. 2005; Pringle et al. 2005; Fournier et al. 2005), which use the phylogenetic concordance of multiple unlinked genes to indicate a lack of genetic exchange and the evolutionary independence of lineages. Thus species boundaries can be determined in spite of a lack of other taxonomic characters or experimental crossing possibilities (Avise and Wollenberg 1997; Taylor et al. 2000; Dettman et al. 2003a,b). This approach is conservative, in that ancestral polymorphism may persist in species having recently diverged, preventing the detection of barriers to gene flow (Nichols 2001). In fungi, the species boundaries recognized by this approach have been found to be in good agreement with those identified by mating tests, although the use of mating compatibility can underestimate the number of species compared to the phylogenetic approach (Dettman et al. 2003a,b). A theoretical study has shown that this discrepancy can be caused by specificities in the life cycles of some fungi (Giraud et al. 2006).

Here we use multiple gene phylogenies to investigate the genetic structure within the fungal pathogen Microbotryum violaceum (Basidiomycota). The existence of “host races” within M. violaceum has been debated since the early 20th century, that is, groups of strains specialized on different hosts with restricted gene flow among them. Diverse approaches have been used to address this issue, including inoculation and hybridization experiments (Goldschmidt 1928; Biere and Honders 1996; Shykoff et al. 1999; Van Putten et al. 2003), morphological variation (Garber et al. 1978), as well as molecular variation for karyotypes (Perlin 1996; Perlin et al. 1997), microsatellite allele frequencies (Shykoff et al. 1999; Bucheli et al. 2000, 2001; Giraud et al. 2002; Van Putten et al. 2005), and DNA sequences (Freeman et al. 2002). Some studies provide evidence for host specialization of M. violaceum (Goldschmidt 1928) with limited gene flow among strains on different hosts even when in sympatry (Bucheli et al. 2000, 2001; Van Putten et al. 2005). In contrast, some experimental data highlighted the possibility of cross-infection and the production of viable hybrids among strains from different hosts (Biere and Honders 1996; Shykoff et al. 1999; Van Putten et al. 2005).

Using multiple gene phylogenies, we aimed to answer the following questions: (1) Are there genetically isolated lineages within M. violaceum that have not been distinguished based on morphological criteria? (2) Are these lineages specialized on single host species, or are they each found to parasitize several hosts? and (3) Is there a geographical segregation of the different evolutionary lineages?

Material and Methods


Microbotryum violaceum causes a sexually transmitted disease, anther smut, of plants in the Caryophyllaceae (Thrall et al. 1993). Diploid teliospores are produced in the anthers of infected plants, replacing the pollen, and female structures are aborted. Deposited by pollinators on a new host, teliospores germinate and undergo meiosis. Sexual conjugation between two haploid cells of opposite mating types is required prior to infection, but selfing is frequent (Delmotte et al. 1999; Giraud 2004; Giraud et al. 2005). Microbotryum violaceum has been studied as a model in several fields of genetics, ecology, and evolutionary biology (Garber and Ruddat 2002; Martinez-Espinoza et al. 2002).


For DNA sequencing, 56 strains of M. violaceum from 23 host species in Western Europe and America were used (Table 1). Strains were collected between 2000 and 2004 and stored at 4°C under desiccation. Teliospores were grown for ca. one week at 23°C on sterile GMB2 media, that is, GMB1 media (Thomas et al. 2003) modified as follows: instead of glucose 4 g/l and yeast extract 10 g/l, we added glucose 10 g/l, yeast extract 1.5 g/l, and malt extract 3 g/l. Resulting cultures of haploid sporidia were diluted with water and then replated, yielding colonies derived from separate haploid sporidia. These single-sporidial colonies were isolated and cultured for DNA extraction.

Table 1.  Isolates of Microbotryum violaceum used in this study: name (including the abbreviated name of the host species), host species, and geographic area of collection.
Sample nameHost speciesRegion
Dalpinus1Dianthus alpinusThe Alps, Switzerland
Dcarthusianorum_7022Dianthus carthusianorumThe Alps, Switzerland
Dcarthusianorum_7042Dianthus carthusianorumThe Alps, Switzerland
Dcarthusianorum_7515Dianthus carthusianorumThe Alps, Switzerland
Dcarthusianorum_30901Dianthus carthusianorumThe Pyrenees, France
Dcarthusianorum_30902Dianthus carthusianorumThe Pyrenees, France
Dcarthusianorum_31602Dianthus carthusianorumThe Alps, Switzerland
DgratianopolitanusDianthus gratianopolitanusJura, Switzerland
Dmonspessulanus_12919Dianthus monspessulanusThe Pyrenees, France
Dmonspessulanus_12920Dianthus monspessulanusThe Pyrenees, France
Dneglectus_1Dianthus neglectusThe Alps, Italy
Dneglectus_2Dianthus neglectusThe Alps, Italy
Dsuperbus_8718Dianthus superbusThe Alps, Switzerland
Dsylvestris_6733Dianthus sylvestrisThe Alps, Switzerland
Dsylvestris_6740Dianthus sylvestrisThe Alps, Switzerland
Dsylvestris_9111Dianthus sylvestrisThe Alps, Switzerland
Dsylvestris_9119Dianthus sylvestrisThe Alps, Switzerland
Dsylvestris_1Dianthus sylvestrisThe Alps, Switzerland
Dsylvestris_us1Dianthus sylvestrisThe Alps, Switzerland
Grepens_137Gypsophila repensThe Alps, Italy
Grepens_6Gypsophila repensThe Alps, Italy
Lflos-cuculi_9203Lychnis flos-cuculiThe Alps, Switzerland
Lflos-cuculi_9205Lychnis flos-cuculiThe Alps, Switzerland
Lflos-cuculi_10934Lychnis flos-cuculiThe Alps, Switzerland
Saccaulis_bSilene acaulisThe Alps, Switzerland
Saccaulis_31401Silene acaulisThe Pyrenees, France
Saccaulis_31402Silene acaulisThe Pyrenees, France
Saccaulis_1Silene acaulisThe Alps, Switzerland
Saccaulis_2Silene acaulisThe Alps, Switzerland
Scaroliniana_1Silene carolinianaUSA
Scaroliniana_2Silene carolinianaUSA
Sdioica_7212Silene dioicaThe Alps, Switzerland
Sdioica_7237Silene dioicaThe Alps, Switzerland
Sdioica_bSilene dioicaBrittany, France
Slatifolia_4001Silene latifoliaParis region, France
Slatifolia_4106Silene latifoliaParis region, France
Slatifolia_10002Silene latifoliaThe Alps, Switzerland
Slatifolia_10006Silene latifoliaThe Alps, Switzerland
Slatifolia_2Silene latifoliaUSA
SlemoniiSilene lemoniiUSA
Smaritima_1Silene maritimaSomerset, U.K.
Smaritima_3Silene maritimaSomerset, U.K.
Snutans_7901Silene nutansThe Alps, Switzerland
Snutans_8742Silene nutansThe Alps, Switzerland
Snutans_lbSilene nutansJura, France
SocymoidesSilene ocymoidesThe Alps, Switzerland
SofficinalisSilene officinalisThe Alps, Switzerland
SottitesSilene ottitesThe Alps, Switzerland
SparryiSilene parryiUSA
Srupestris_10502Silene rupestrisThe Alps, Switzerland
Svulgaris_7806Silene vulgarisThe Alps, Switzerland
Svulgaris_7807Silene vulgarisThe Alps, Switzerland
Svulgaris_7903Silene vulgarisThe Alps, Switzerland
Svulgaris_7913Silene vulgarisThe Alps, Switzerland
Svulgaris_30027Silene vulgarisThe Pyrenees, France
Svulgaris_30030Silene vulgarisThe Pyrenees, France
CcaespitosaCalandrinia caespitosaChile


DNA was extracted from single-sporidial colonies using the Chelex (Biorad, Marne-la-coquette, France) protocol (Bucheli et al. 2001). Approximately 600 bp of the β-tubulin (β-tub), 700 bp of the γ-tubulin (γ-tub), and 1000 bp of the Elongation factor 1 α (Ef1α) were amplified by polymerase chain reaction (PCR) using the primers Tub 5-IN/Tub 3-IN, T1/T2 (Freeman et al. 2002) and Ef1u1f (ACGGTCTGACGCATGTCAC)/Ef1u1r (CAAGAACATGATCACTGGTACCTC), respectively. For samples in which the Ef1α gene could not be amplified, internal primers Ef1m4f (GAAGGACTCGACGCACATC) and Ef1m4r (ACGCTCTCCTCGCCTTCA) were used, yielding 800 bp.

PCR amplifications were performed using standard techniques (available upon request). PCR products were purified using PEG (Sigma-Aldrich, Lyon, France) and sequenced on ABI 310 (Applied Biosystem, Courtabœuf, France) or CEQ 8000 (Beckman Coulter, Villepinte, France) using dye-terminator chemistry. The γ-tub and Ef1α genes were sequenced in both directions, whereas the β-tub was sequenced using only the Tub 5-IN primer. The sequences are available in GenBank (accession numbers DQ074479–DQ074635).


The three genes used can be considered as unlinked given that (1) they are located on different chromosomes in 2 Basidiomycetes having their genome sequenced (Ustilago maydis and Cryptococcus neoformans) and (2) the karyotype of M. violaceum has been shown to be highly polymorphic, even among strains from the same host (Hood 2002; Hood et al. 2003; Hood and Antonovics 2004), suggesting that few genes may have remained durably linked in M. violaceum.

Sequences were aligned in Bioedit v6.0.7 and corrected by hand when necessary. Regions with ambiguous alignments were excluded from all analyses. Phylogenetic trees were reconstructed by maximum parsimony (MP), neighbor joining (NJ), and Bayesian inference. MP and NJ analyses were performed using PAUP version 4.0b10 (Swofford 2003). MP analyses were performed using a heuristic search. For NJ analyses, the ModelTest v3.5 program (Posada and Crandall 1998) with Akaike information criterion (AIC) was used to select the models that best fit our data. Bootstrap confidence values were calculated for 1000 pseudoreplicates (Felsenstein 1985). Bayesian analyses were run using MrBayes version 3.0b5 (Ronquist and Huelsenbeck 2003). Each run consisted of four incrementally heated Markov chains run simultaneously, with heating value set to default (0.2). Priors were constrained according to the ModelTest results. Markov chains were initiated from a random tree and run for 500,000 generations. We used a 50% majority rule consensus tree to obtain the Bayesian posterior probabilities (BPP), considering the trees sampled after the likelihood scores had reached stationary. Details on the phylogenetic analyses are available upon request. Monophylies supported by bootstrap ≥ 75% and BPP ≥ 0.95 were considered as significant. A node was considered as strongly supported when significant using two of the reconstruction methods.


To detect intragene recombination we used two phylogenetic methods (RDP by Martin and Rybicki 2000 and Bootscanning by Salminen et al. 1995) and two substitution distribution methods (Maximum χ2 by Maynard Smith 1992 and Maximum mismatch χ2 by Posada and Crandall 2001). These four different methods were used as implemented in the software RDP2 (Martin et al. 2005), and details about the parameters used are available upon request. To minimize the risk of false positives, we only considered as reliable potential recombination events detected by more than one method (Posada 2002).


Sequences from different fungi were used to root the phylogenies. Sequences were chosen as the best hits to our sequences in GenBank. To root the β-tub gene tree, we used Monascus eremophilus (GenBank accession number AY498603), Rhodotorula glutinis (L47266), Cercophora sparsa (AY600253), and Platygloea pustulata (AY371532). To root the Ef1α gene tree, we used Agaricostilbum hyphaenes (AY879114), Sporobolomyces syzygii (AB127096), Bannoa hahajimensis (AB127093), Puccinia graminis (X73529), and Rhodotorula mucilaginosa (AF016239). The γ-tub gene of M. violaceum could not be aligned with any published fungal sequences, even those of U. maydis. To root the trees, we also sequenced a smut sample infecting a non-Caryophyllaceae host (Calandrinia caespitosa) from Chile. The β-tub (DQ074479) and Ef1α (DQ074536) sequences could be obtained in this smut sample but not γ-tub. Given the surprising position of the C. caespitosa sample in the trees, DNA extraction, amplifications, and sequencing were performed a second time independently to discard the possibility of a contamination.


Congruence between gene phylogenies was estimated using the incongruence length difference (ILD) test (Farris et al. 1994) as implemented in PAUP, the approximately unbiased (AU) test (Shimodaira 2002) as implemented in CONSEL (for assessing the confidence of phylogenetic free selection) (Shimodaira and Hasegawa 2001), and by visual inspection of topologies and statistical supports (Mason-Gamer and Kellogg 1996). The ILD test is known to be very conservative (Darlu and Lecointre 2002), so the null hypothesis of congruence was rejected only if P < 0.001. By visual inspections, incongruence between gene phylogenies was concluded when conflicting nodes were supported by significant statistical values; nodes were considered as congruent when identical and supported by significant statistical values in the phylogenies considered. AU tests were conducted for each incongruency detected by visual inspection, by comparing for each gene the likelihood of the MP topology obtained for the focal gene to the likelihood of the conflicting topology obtained with another gene. Likelihoods were obtained in PAUP using the sequence evolution model selected according to the ModelTest results.

A restricted dataset excluding the strains showing incongruent positions when comparing the single gene phylogenies was used to concatenate the three genes, in order to improve the power of detection of monophyly. Nodes that were neither significantly congruent nor incongruent when comparing the single gene phylogenies but that were found strongly supported in the concatenated analysis were considered as significantly congruent. Only statistical supports and topologies were considered in the combined analyses.


To detect independent evolutionary units within M. violaceum, we used the criterion of phylogenetic congruence between different gene phylogenies (Taylor et al. 2000; Dettman et al. 2003a). We thus considered a group of strains as an independent evolutionary lineage when (1) it was significantly supported as monophyletic in at least one gene phylogeny by two of the three reconstruction methods, and (2) this was not contradicted by the other gene phylogenies. In addition, we recorded whether the independent evolutionary lineages geographically overlapped, that is, whether they contained individuals geographically closer to individuals from other lineages than to individuals from its own lineage.



Information on datasets and phylogenetic trees is shown in Table 2. For each gene, Bayesian, MP, and NJ consensus trees revealed the same relationships between the significantly supported clades. Therefore, only the Bayesian consensus trees of the Ef1α, β-tub, and γ-tub genes are shown in Figures 1, 2, and 3, respectively.

Table 2.  For the different trees constructed, information on the sequence dataset (number of sequences, including the Calandrinia sample, number of aligned sites number of variable sites, and number of parsimony informative sites) and information on maximum parsimony (MP) trees (number of steps and retention index [RI]).
 Number of sequencesTotal number of sitesVariable sitesParsimony informative sitesNumber of tree stepsRI
Ef1α51 754163 (22%)111 (15%)2740.92
β-tub57 356 54 (15%) 45 (13%)1190.85
γ-tub48 636147 (23%)111 (17%)1070.96
Combined281742279 (16%)183 (11%)Not relevantNot relevant
Figure 1.

Bayesian 50% majority-rule consensus tree based on the Ef1α gene (mean lnL =−2968). Nodes not strongly supported are represented as unresolved. Statistical supports indicate Bayesian posterior probabilities (BPP)/maximum parsimony bootstraps/neighbor joining bootstraps. Only statistical supports higher than 0.9/70/70 are indicated. Numbered arrows indicate conflicting nodes between the different gene phylogenies or intragene recombination. Taxa labels correspond to the host plant on which fungal strains were collected. Brackets indicate clades and evolutionary units identified (see text). □ indicates strains from The Pyrenees, ▴ strains from The Alps, inline image strains from Paris region, ◊ strains from Brittany, ○ strains from Jura, ◆ strains from Somerset, and no symbol indicates strains from America.

Figure 2.

Bayesian 50% majority-rule consensus tree based on the β-tub gene (mean lnL =−1039). See Figure 1 for the legend of the bootstraps, arrows, and taxa labels.

Figure 3.

Bayesian 50% majority-rule consensus tree based on the γ-tub gene (mean lnL =−2238). See Figure 1 for the legend of the bootstraps, arrows, and taxa labels.

The ILD test indicated that the individual gene phylogenies were at the limit of the incongruence significance, with P= 0.001. In agreement, visual inspection of the nodes and supports showed that most of the phylogenetic relationships were congruent among the three genes, such that the same clades were significantly supported in the different gene trees, with a few exceptions outlined below. In particular, all phylogenies strongly supported four deep clades, named A, B, C, and D (Figs. 1, 2, and 3). However, statistical support was generally lower using the β-tub gene for all the reconstruction methods except Bayesian inference (Fig. 2), and with the γ-tub some nodes were better resolved than with the two other genes.

Visual inspection of topology and supports detected three conflicting nodes between the γ-tub phylogeny and the other gene phylogenies. First, the position of the clade containing fungal strains from Silene acaulis (clade D1) was different between the γ-tub phylogeny and the two other phylogenies (incongruency 1 on Figs. 1, 2, and 3). The AU tests were also significant for this node (P= 0.01, 0.003, and < 0.001 when enforcing the position of the clade D1 in the β-tub and Ef1α phylogenies as in the γ-tub phylogeny and in the γ-tub phylogeny as in the β-tub and Ef1α phylogenies, respectively). Second, within clade D, the γ-tub phylogeny strongly supported the Silene vulgaris and Lychnis floscuculi strains as monophyletic, whereas the Ef1α phylogeny strongly supported the S. vulgaris and Silene nutans strains as monophyletic (incongruency 2, Figs. 1 and 3). The AU tests were also significant for this node (P= 0.03 and < 0.001 when enforcing the monophyly of the S. vulgaris and S. nutans strains in the γ-tub phylogeny and of the S. vulgaris and L. floscuculi strains in the Ef1α phylogeny, respectively). Third, the Dianthus sylvestris strain 1 and D. sylvestris strain 6740 were strongly supported as monophyletic in the γ-tub phylogeny, whereas in the β-tub phylogeny the D. sylvestris strain 6740 was not in the same strongly supported clade as the D. sylvestris strain 1 (incongruency 3, Figs. 2 and 3). The AU tests were also significant for this node (P < 0.001 and P= 0.005 when enforcing the D. sylvestris strain 1 and D. sylvestris strain 6740 as monophyletic in the β-tub phylogeny and as belonging to different clades in the γ-tub phylogeny, respectively).

When removing these strains showing incongruencies among individual phylogenies, the three gene tree topologies appeared visually identical, but for the significance of statistical support for various nodes. The ILD test confirmed that the gene phylogenies are then more congruent (P= 0.06, NS). The phylogenetic relationships inferred from the concatenated dataset were the same as for the individual gene trees, but with higher statistical supports (Fig. 4).

Figure 4.

Topology of the tree inferred from the concatenated analyses using the restricted combinable dataset. See Figure 1 for the legend of the bootstraps, arrows, and taxa labels.

Phylogenetic analyses of the β-tub and Ef1α genes were first performed using several fungal sequences as outgroups; the γ-tub tree remained unrooted. The root positions found using the β-tub and Ef1α genes were identical (Figs. 1 and 2) and fully concordant with the results obtained by Freeman et al. (2002) using ITS, and can therefore be considered as reliable. Surprisingly, the three reconstruction methods strongly supported the branching of the strain from the non-Caryophyllaceae host C. caespitosa within M. violaceum clade D in both the β-tub and Ef1α phylogenies (Figs. 1 and 2). The other outgroups strongly supported the monophyly of all the M. violaceum strains plus the strain infecting C. caespitosa for both the β-tub and Ef1α trees. Outgroup sequences were removed from the datasets for subsequent analyses, except those of the C. caespitosa sample that did not appear actually to be an outgroup.


Maximum χ2 and Maximum mismatch χ2 methods each indicated multiple recombination events in Ef1α, but only one recombination event was detected by both methods and was therefore considered as significant. The Ef1α alleles of the strains involved belonged to clade C and resulted from recombination between a haplotype similar to the Silene lemonii strain and a haplotype similar to the S. acaulis strains. The entire clade C may therefore have experienced recombination, causing the position of this clade in the Ef1α phylogeny to remain uncertain (incongruency 4, Figs. 1, 2, and 3). Estimated beginning breakpoints for the insertion of the S. acaulis-like haplotype in the S. lemonii-like haplotype varied between the positions 357 and 507 of the aligned datafile, depending on the method and on the focal sequence. Estimated ending breakpoint for this insertion was the position 695 of the aligned datafile. None of the four tests of recombination (RDP, Bootscanning, Maximum χ2, and Maximum mismatch χ2) detected evidence of recombination within β-tub nor γ-tub.


Using the criterion of congruence between individual gene phylogenies, 11 independent evolutionary lineages were identified, labeled on the figures by the initials of their main host species (e.g., MvSa for M. violaceum evolutionary unit infecting S. acaulis, or MvSspA for M. violaceum evolutionary unit infecting Silene species native from America). Figure 5 summarizes the supported phylogenetic relationships among these 11 independent evolutionary lineages, as inferred from the different trees constructed. The evolutionary lineages each included mainly strains infecting the same host species, and all the evolutionary lineages geographically overlapped with several other lineages. Five individual strains (from S. lemonii, Silene ottites, Silene officinalis, Dianthus alpinus, and Gypsophila repens) were each strongly supported as evolutionarily independent of the 11 lineages identified above, and were therefore not included in any of these lineages. We, however, did not name specific lineages for them because these strains each branched alone in the trees.

Figure 5.

Summary of the phylogenetic relationships established between the 11 independent evolutionary lineages identified. The two lineages displaying incongruencies for which the position in the tree could not be resolved are not connected to it.



The multiple gene phylogenies approach in M. violaceum provided evidence of several independent evolutionary lineages evolving without gene flow for enough time for the ancestral polymorphism to be either lost or fixed in the descendant populations. Several groups were strongly supported as monophyletic by the three independent gene phylogenies. Moreover, the phylogenetic relationships supported by these three nuclear genes are also fully congruent with the results obtained with ITS sequence data from an independent set of strains from the same range of hosts (M. E. Hood, M. Le Gac, and T. Giraud, unpubl. ms.). We detected 11 independent evolutionary lineages, MvSspA, MvSa, MvSv1, MvSl, MvSn, MvSd, MvSv2, MvLf, MvDsp1, MvDsp2, and MvDc, that had largely overlapping geographic ranges. However, it remains to be determined whether regions or host species where sampling was limited may contain some unresolved cryptic species, and whether lineages that appear to infect more than one host species may actually be very host-specific but too recently divergent for detection using our markers.

Our results show that there is not frequent introgressions among the different lineages despite the possibility of cross-infections observed in experimental inoculations between some host plants and even in nature (Biere and Honders 1996; Shykoff et al. 1999; Antonovics et al. 2002; Hood et al. 2003; Van Putten et al. 2005) and the possibility of hybridization between strains from different hosts (Goldschmidt 1928; Biere and Honders 1996; Shykoff et al. 1999; Van Putten et al. 2003). The few incongruencies detected between individual gene phylogenies and the intragene recombination event generally correspond to ancient events. This is in agreement with the idea that in vitro interfertility, and even the presence of hybrids in nature, do not suggest that frequent introgression is likely (Coyne and Orr 2004, p. 42).


Interestingly, the host species S. vulgaris appears to be parasitized by two independent divergent lineages of M. violaceum, as previously suggested based on microsatellites (Bucheli et al. 2000). Here, we showed that the species Dianthus carthusianorum also appears to be parasitized by two distinct M. violaceum lineages. In both cases, the two lineages infecting the same host plant occur in the same mountains and even in the same host population for S. vulgaris strains (Svulgaris_7903 and Svulgaris_7913), raising questions of the conditions that allow their coexistence. Further investigations should be conducted to address why different lineages on the same host do not fuse or why one lineage is not outcompeted by the other.


The lineage MvSspA includes most of the strains from native U.S. hosts and therefore corresponds to the native North American hosts strains clade described by Freeman et al. (2002). These authors reported a clear distinction between strains infecting native North American hosts and European hosts, suggesting a single and ancient divergence event. Their study was, however, based on a single gene, ITS, included collections from few European host species, and was restricted to east-coast populations in North America. In the present study, a strain from an additional native host in the western regions of North America, S. lemonii, was included and did not branch within the established North American clade A, but at the base of a clade containing European strains. This suggests a more complex geographic colonization and diversification history of M. violaceum than was previously known (Freeman et al. 2002).


Four incongruencies were detected among the individual gene phylogenies. Because the third incongruence (incongruency 3) is located within a lineage, MvDsp1, and even within a host race, it probably only reflects usual intraspecific recombination. The three remaining incongruencies, strongly supported by comparing topologies and supports of the three phylogenies and AU tests (incongruencies 1 and 2) or by testing for intragene recombination (incongruency 4), may be due to genetic exchange between lineages, to the maintenance of ancestral polymorphism, or to convergence. Convergence appears highly improbable because apomorphies yielding incongruencies were numerous and were due to synonymous or intronic substitutions (data not shown). Ancestral polymorphism and ancestral recombination are not likely either given the great divergence between the different DNA fragments detected as incongruent and the fact that long stretches of DNA have the same evolutionary history. For instance γ-tub and Ef1α from S. acaulis strains branch in two very distant positions in the Microbotryum tree, but within each gene no site contradicts their respective position. The incongruencies 1, 2, and 4 therefore suggest the existence of genetic exchange events between divergent lineages of M. violaceum, that is, ancient introgressions or speciation by hybridization. These incongruencies indeed involved all the strains from the given lineages. This indicates an ancient recombination event that may have occurred during speciation, and may thus reflect a speciation by hybridization, or genetic exchange just after speciation, when introgression was still possible. Signatures of ancient introgressions have been reported in other fungal species complexes, for instance, in Epichloë spp. (Schardl et al. 1997) and Botrytis spp. (Staats et al. 2005).


Samples of M. violaceum collected from the same host plants but in different geographic regions were genetically much closer than smuts collected on different host species in the same fields. This suggests a major role of specialization in the origin and maintenance of the M. violaceum lineages. However, this also makes the role of geographic isolation during speciation much more difficult to assess. Speciation events may have occurred in allopatry, and the current geographic distribution of species is the result of postspeciation dispersal events. It is indeed theoretically difficult to explain how two lineages can emerge in sympatry without divergent adaptations. It is therefore likely that two lineages infecting the same host species, as it is the case on S. vulgaris or D. carthusianorum, have first diverged in allopatry.


The degree of specialization and differentiation within M. violaceum has been debated for many decades. Genetic data suggested that this fungus was divided in several host-specific and genetically differentiated groups (Perlin 1996; Perlin et al. 1997; Shykoff et al. 1999; Bucheli et al. 2000, 2001; Giraud et al. 2002; Van Putten et al. 2005), although it was not clear whether these represented host races or true species. In contrast, cross-inoculation studies suggested that there was no strong specialization in the anther smut (Biere and Honders 1996; Shykoff et al. 1999; Van Putten et al. 2005), and the viability of hybrids both in vitro and in natura suggested that gene flow could be frequent among strains of different host species (Van Putten et al. 2005). The present study settles this issue, clearly showing the existence of independent evolutionary lineages, not exchanging gene for a long time, even when in sympatry, and establishing clear genetic boundaries for these lineages. Furthermore, this study suggests for the first time the existence of ancient introgression or hybridization events and large-scale species dispersal.

Associate Editor: H. Hollocher


The authors thank Y. Brygoo, B. Faivre, and A. Gautier for help with sequencing; A. Thomas, M. Bartoli, I. Till-Bottraud, M. Arroyo, A. Widmer, J. Antonovics, S. Triki-Teurtroy, S. Ribstein, and T. Darras for help in spore collection; J. A. Shykoff and R. Yockteng for critically reading the manuscript; H. Hollocher and three anonymous referees for constructive comments, and the French Pyrenees National Park for collection permit. TG acknowledges a grant “ACI Jeunes Chercheurs” from the French Ministry of Research and MEH. acknowledges award DEB-0346832 from the National Science Foundation. An ECOS grant (FONDECYT Cooperative Grant 7020956) allowed the spore collection in Chile. MLG was supported by a 2004 Systematics Research Fund award from the Linnean Society and the Systematics Association.