History of the invasion of the anther smut pathogen on Silene latifolia in North America


  • Michael C. Fontaine,

    Corresponding author
    1. CNRS, UMR 8079, Orsay Cedex, France
    2. Department of Biological Sciences, University of Notre Dame, Notre Dame, IN, USA
    • Université Paris-Sud, Laboratoire Ecologie, Systématique et Evolution, Orsay Cedex, France
    Search for more papers by this author
  • Pierre Gladieux,

    1. Université Paris-Sud, Laboratoire Ecologie, Systématique et Evolution, Orsay Cedex, France
    2. CNRS, UMR 8079, Orsay Cedex, France
    3. Plant and Microbial Biology, University of California, Berkeley, CA, USA
    Search for more papers by this author
  • Michael E. Hood,

    1. Department of Biology, Amherst College, Amherst, MA, USA
    Search for more papers by this author
  • Tatiana Giraud

    1. Université Paris-Sud, Laboratoire Ecologie, Systématique et Evolution, Orsay Cedex, France
    2. CNRS, UMR 8079, Orsay Cedex, France
    Search for more papers by this author

Author for correspondence:

Michael C. Fontaine

Tel: +1 574 631 3904

Email: mikafontaine@gmail.com


  • Understanding the routes of pathogen introduction contributes greatly to efforts to protect against future disease emergence.

  • Here, we investigated the history of the invasion in North America by the fungal pathogen Microbotryum lychnidis-dioicae, which causes the anther smut disease on the white campion Silene latifolia. This system is a well-studied model in evolutionary biology and ecology of infectious disease in natural systems.
  • Analyses based on microsatellite markers show that the introduced American M. lychnidis-dioicae probably came from Scotland, from a single population, and thus suffered from a drastic bottleneck compared with genetic diversity in the native European range. The pattern in M. lychnidis-dioicae contrasts with that found by previous studies in its host plant species S. latifolia, also introduced in North America. In the plant, several European lineages have been introduced from across Europe. The smaller number of introductions for M. lychnidis-dioicae probably relates to its life history traits, as it is an obligate, specialized pathogen that is neither transmitted by the seeds nor persistent in the environment.
  • The results show that even a nonagricultural, biotrophic, and insect-vectored pathogen suffering from a very strong bottleneck can successfully establish populations on its introduced host.


Biological invasions have had dramatic ecological and economic impacts (Pimentel et al., 2001). In particular, biological invasions by fungal or fungal-like (i.e. Oomycetes) pathogens have caused many emerging and devastating diseases (Anderson et al., 2004; Desprez-Loustau et al., 2007). Among plant pathogens, striking examples include the introduced diseases that have affected the American chestnut (Castanea dentate), a formerly dominant tree that is now confined to being an understory shrub. The Ink disease caused by the oomycete Phytophthora cinnamomi first caused the demise of most chestnut trees in the southern part of their American distribution (Crandall, 1950). The chestnut blight fungus (Cryphonectria parasitica) then eliminated nearly all remaining native chestnut trees throughout eastern American forests during the 20th Century (Anagnostakis, 1987). Another example of devastating emerging fungal plant diseases is Dutch elm disease, first caused by Ophiostoma ulmi, which led to the destruction of many American and European elms (Ulmus Americana) (Brasier, 1991). Its sibling species Ophiostoma novo-ulmi, introduced in the 1960s and more aggressive than O. ulmi, then eliminated most remaining mature elms (Ulmus glabra, Ulmus procera and Ulmus minor) across North America and Europe (Brasier, 1991). Phytophthora cinnamomi is probably the most notorious invasive tree pathogen, known to attack and kill c. 5000 woody plant species in the world, causing highly destructive epidemics at ecosystem and landscape scales world-wide (Zentmyer, 1980; Shearer & Tippet, 1989; Shearer et al., 2004; Cahill et al., 2008).

Our primary food production is also at risk as a result of emerging crop diseases caused by fungi or oomycetes (Strange & Scott, 2005), the most dramatic historical example being the Irish Potato Famine caused by Phytophthora infestans on cultivated potato (Solanum tuberosum) at the beginning of the 1840s (Fry, 2008). More recent examples include the blast disease of wheat (Triticum aestivum) that appeared in Brazil in the 1980s and then spread to other South American countries (Urashima et al., 1993), as well as the Ug99 fungal pathotype causing stem rust disease of wheat, first identified in 1998 in Uganda and now threatening North Africa, the Middle East and Asia (Singh et al., 2008).

Retracing the routes of introduction of fungal pathogens and understanding the evolutionary processes during introduction and spread can contribute to invasion prevention and management programs (Facon et al., 2006). It is important to assess, for instance, whether multiple introductions are required for successful invasions and whether strong bottlenecks reduce the genetic variability of pathogens in introduced ranges (Dlugosch & Parker, 2008). In the case of an obligate and specialized pathogen, where invasion can occur only after introduction of the host, another relevant question is whether the pathogen came from the same geographic areas as its host, either because they took the same routes of introduction or because of local adaptation (i.e. when sympatric pathogens perform the best on their local hosts compared with other pathogen populations; Kaltz & Shykoff, 1998).

Here, we studied the history of the invasion of the anther smut pathogen of the white campion Silene latifolia in North America. Microbotryum violaceum sensu lato is a species complex of basidiomycete fungi responsible for the sterilizing and persistent anther smut disease in many species, mostly in the Caryophyllaceae. Infected plants contain fungal teliospores in place of the pollen and female structures do not mature; female plants in dioecious species also develop spore-bearing anthers. Teliospores are transmitted from diseased to healthy plants mostly by insects that normally serve as pollinators. Microbotryum fungi are obligate pathogens. The fungus is not transmitted in the seeds (Baker, 1947a,b), nor do the spores appear capable of persisting in the environment as the disease is restricted to perennial hosts that allow for transmission between living plants (Hood et al., 2010). Moreover, the sibling species encompassed in M. violaceum sensu lato (Kemler et al., 2006; Le Gac et al., 2007; Denchev et al., 2009) show strong host specificity (de Vienne et al., 2009a) and inter-sterility (de Vienne et al., 2009b). The most widely studied species is Microbotryum lychnidis-dioicae (Denchev et al., 2009) (called MvSl in Le Gac et al. (2007)), parasitizing S. latifolia, and this is a well-studied plant model in evolutionary biology and ecology (Bernasconi et al., 2009). Silene latifolia (syn. S. alba) is a dioecious herbaceous perennial with a history of human association, commonly found in disturbed areas such as roadsides, railroad embankments, cultivated fields, and abandoned lots (Baker, 1948). Microbotryum lychnidis-dioicae has a highly selfing mating system (Delmotte et al., 1999; Giraud et al., 2005, 2008a; Vercken et al., 2010), with mostly intratetrad matings (Hood & Antonovics, 2000, 2004). Sex is obligate in the life cycle of Microbotryum, involving the production of dikaryons that are the infectious structures.

Both the plant S. latifolia and its anther smut pathogen have been introduced in modern times to North America from Europe, and the invasion of S. latifolia is now well studied (Wolfe, 2002; Taylor & Keller, 2007; Keller et al., 2009, 2012). The plant species arrived in North America from Europe in the mid-1800s (Baker, 1947a,b) and has become a mildly problematic weed of cultivated fields in southern Canada and the northern USA (McNeil, 1977). An herbarium study suggested that the first introduced plant populations occurred probably in the north-eastern USA, from where plants spread predominantly southwards and became common only since the 1920s (Antonovics et al., 2003). Introduced populations of S. latifolia have been shown to originate from Europe, with independent introductions into eastern and western North America, and with a rather severe but short-lived genetic bottleneck (Taylor & Keller, 2007; Keller et al., 2012). Lineages of S. latifolia from both eastern and western Europe have been introduced in both eastern and western North America, offering opportunities for admixture among previously isolated lineages (Taylor & Keller, 2007; Keller et al., 2012). There are genetically based differences in life-history traits between S. latifolia populations from the native European range and the introduced North American range, in particular increased susceptibility to natural enemies in North America (Wolfe et al., 2004).

In contrast to the invasion by S. latifolia, the invasion history of its associated anther smut pathogen in North America has not been studied. The disease introduction is probably very recent, as an herbarium study did not find any diseased S. latifolia in plant collections from the USA (Antonovics et al., 2003), even though diseased plants are usually represented without much bias in herbaria (Antonovics et al., 2003; Hood et al., 2010). The first published record, to our knowledge, of the disease on S. latifolia in North America describes its occurrence in the central Appalachian Mountains of Virginia (Alexander & Antonovics, 1988). Sporadic observations have also been made in recent decades in Illinois (E. Garber, pers. comm.), Michigan (J. Antonovics, pers. comm.), Pennsylvania (E. Lyons & A. Jarosz, pers. comm.), and Massachusetts (T. Meagher, pers. comm.). Two specimens in mycological herbaria from the early 1960s place the pathogen on S. latifolia in Bronx and Dutchess counties of New York State (in the Gray and Kew mycological herbaria, respectively; M. E. Hood, pers. obs.). Previous studies on limited numbers of strains and based on electrophoretic data or internal transcribed spacer (ITS) sequences suggested that North American populations on S. latifolia were genetically more similar to English populations on S. latifolia than to North American populations on other Silene host species (Antonovics et al., 1996; Freeman et al., 2002), showing that these North American populations originated through an introduction rather than as a host shift from a host species native to North America. Few sample localities were included in these previous studies, so no comparisons could be made among European populations to identify the pathogen source(s) more specifically. Also, the possibility of a host shift as the source of the disease on North American S. latifolia from another European host species, particularly Silene dioica, cannot be excluded. Host shifts indeed occur in Microbotryum between closely related hosts (Antonovics et al., 2002; Lopez-Villavicencio et al., 2005; Refrégier et al., 2008; Gladieux et al., 2011), and natural and artificial cross-species disease transmission is successful between S. latifolia and S. dioica (Van Putten et al., 2003; de Vienne et al., 2009a).

Our aim here was therefore to reconstruct the invasion history of Microbotryum on S. latifolia by comparing the population genetics of North America and European using multilocus microsatellite genotyping of M. lychnidis-dioicae from S. latifolia and M. silenes-dioicae parasitizing S. dioica (Vercken et al., 2010; Gladieux et al., 2011). Concurrently, the mating system in introduced pathogen populations was assessed in comparison to its native range, as examples of striking changes in other introduced organisms include the loss of sexual reproduction in introduced fungal pathogens (Ali et al., 2010; Duan et al., 2010; Saleh et al., 2012) and changes in selfing rates in plants (Amsellem et al., 2001; Bossdorf et al., 2005; Colautti et al., 2010). We therefore analyzed North American populations of Microbotryum on S. latifolia in North America, representing good coverage of the extant populations, using genetic data to address the following specific questions: What are the European sources from which the North American populations originated, in terms of both species and populations? Can we detect footprints of a bottleneck, and if so how strong was it? Were invasive populations introduced from multiple sources? If so, did admixture occur? And/or did hybridization occur with native Microbotryum species? Has the mating system of the fungus changed compared with the populations of origin? How similar are the invasion histories of the plant S. latifolia and of its pathogen?

Materials and Methods

Teliospore collection and populations

The individuals of Microbotryum analyzed in this study were collected as diploid teliospores from 183 localities on Silene latifolia Poiret (= 773) across Europe and eastern North America (Fig. 1) and stored in silica gel (for a detailed description of the sampling see Supporting Information Table S1 for North America and Vercken et al. (2010) for Europe). DNA extracted from teliospores as described in Giraud (2004) from one flower per diseased plant was used for genetic analyses. Multiple infections by different genotypes are frequent in the SileneMicrobotryum system, but teliospores within a single flower originate from a single diploid individual (Lopez-Villavicencio et al., 2007).

Figure 1.

Map of sampled localities for Microbotryum lychnidis-dioicae in the native European range (MvSl, n = 701; left panel) and the introduced area in the USA (= 72; right panels). The sizes of the red dots are proportional to the sample size.

Microsatellite genotyping

Populations from Europe had been genotyped and analyzed in a previous study (Vercken et al., 2010). For populations from North America, teliospores were genotyped using 11 microsatellite markers following the protocol of Giraud (2004) (Table 1). Among the 11 microsatellite loci used, E14, E17 and E18 were described in Bucheli et al. (1998), SL8, SL9, SL12, SL19, SVG5, SVG8 and SVG14 were described in Giraud et al. (2008b), and SL5 was described in Refrégier et al. (2010).

Table 1. Genetic polymorphism at each locus in European and American samples of Microbotryum lychnidis-dioicae
N A H e F IS N A H e F IS
  1. N, sample size; A, number of alleles; He, expected heterozygosity; FIS value: ns, not statistically different from 0; *, < 0.05; **, P < 0.01.

E14670110.710.96**7020.07−0.03 ns
SL1268070.620.91**6820.04−0.02 ns
SVG1468370.310.90**6720.03−0.01 ns
± SD58.

Data analyses

Descriptive statistics

Polymorphism at each locus was quantified using number of alleles (A), allelic richness (Ar), allelic richness private to a specific grouping (pAr), unbiased expected heterozygosity (He), and fixation index (FIS); the four latter statistics control for number of samples. These statistics were calculated using fstat (Goudet, 2001) and adze (Szpiech et al., 2008). Departures from Hardy–Weinberg expectations were tested using exact tests implemented in genepop 4.0 (Raymond & Rousset, 1995; Rousset, 2008).

Population structure

We investigated population structure using two complementary approaches: principal component analyses (PCAs) and Bayesian model-based clustering. PCAs display genotypes in a multivariate space described by the principal component (PC) (Patterson et al., 2006; Jombart et al., 2009). Although the method does not rely on any population genetic model, it is useful to represent genetic relationships among individuals (McVean, 2009). By contrast, individual-based Bayesian clustering algorithms, such as structure, partition multilocus genotypes into clusters based on allele frequency and linkage equilibrium among loci, and estimate the admixture proportions to each cluster (Pritchard et al., 2000; Falush et al., 2003; Hubisz et al., 2009).

In order to check that the anther smut fungus infecting S. latifolia in North America originated from European strains also infecting S. latifolia, and not from another European Microbotryum species such as Microbotryum silenes-dioicae infecting S. dioica, we conducted PCAs on the European genotypic data from Vercken et al. (2010), which included both Microbotryum lychnidis-dioicae and M. silenes-dioicae, together with the North American samples genotyped for the present study. We then ran PCA on the M. lychnidis-dioicae data set from Europe and North America to provide a better resolution of the genetic structure. PCA was conducted on the microsatellite allele frequencies using the ADEgenet package (Jombart, 2010) in the R environment (R Development Core Team, 2010).

For the clustering analysis, the structure 2.3.3 program was used (Pritchard et al., 2000; Falush et al., 2003; Hubisz et al., 2009) with a haploid setting because Microbotryum is almost completely homozygous (Table 1 and 2). Run conditions for structure analyses were as follows: a series of independent runs were conducted with different proposals for the number of clusters (K), testing all values from 1 to 10. Each run used 1 000 000 iterations after a burn-in of 100 000 iterations, using a model allowing for admixture and correlated allele frequencies. To ensure convergence of the Markov Chain Monte Carlo (MCMC), we performed 10 independent replicates for each value of K and checked the consistency of results visually and using the procedure implemented in the program clumpp 1.1.1 (Jakobsson & Rosenberg, 2007). We used clumpp 1.1.1 to account for label switching and to identify potential distinct solutions among the results of independent replicate runs for each K. For that purpose, we computed with the greedy algorithm a symmetric similarity coefficient index between pairs of runs (100 random input sequences, G′ statistic).

Table 2. Pairwise differentiation among clusters expressed as FST values among clusters of Microbotryum lychnidis-dioicae (P < 0.001 for all values)
F ST BalkanNWestItalianSWestEastern

Differences in allelic frequencies between groups were assessed using exact tests implemented in genepop 4.0 (Raymond & Rousset, 1995; Rousset, 2008) and quantified using Weir and Cockerham's FST estimator (Weir & Cockerham, 1984). Differences in genetic polymorphism between groups were assessed using a Wilcoxon signed-rank test. The selfing rate within the North American population was estimated using the instruct program (Gao et al., 2007) using the same run settings as in Vercken et al. (2010) and we compared selfing rates with values previously estimated for European clusters (Vercken et al., 2010).


Population study

North American Microbotryum strains originate from European M. lychnidis-dioicae populations

The PCA on M. lychnidis-dioicae and M. silenes-dioicae microsatellite data confirms that all strains collected in North American are related to European M. lychnidis-dioicae populations on S. latifolia rather than being a host-shift from the pathogen on S. dioica (Fig. S1). We focused thus only on material of M. lychnidis-dioicae for subsequent analyses.

Low genetic diversity in North American populations

The genetic diversity in North American populations was much lower than those observed in Europe (Table 1). Most loci displayed a strong deficit in heterozygosity, clearly departing from Hardy–Weinberg expectations.

Genetic structure

The population structure of M. lychnidis-dioicae in Europe was very strong, as described in Vercken et al. (2010); increasing K consistently identified new distinct clusters in structure analyses. The main clusters in Europe at K = 3 (Fig. 2) were: the Western group (blue at K = 3), which split at higher K values into north Western (NWest) and south Western (SWest) clusters (respectively in blue and yellow at K ≥ 4); the Italian cluster (in green); and the Eastern cluster (in red at K = 3), which split at higher Ks into the Balkan and Eastern clusters (respectively in purple and red at K ≥ 5). The American populations remained together with the NWest cluster regardless of the number of clusters implemented (see K = 3–7; higher K values not shown). This indicates that American populations are closely related to the populations from northwestern Europe, and in particular to the populations from the UK. This was also indicated by the PCA (Fig. 3a), on which American genotypes clustered together with northwestern European populations; the first three PCs accounted for 35% of the total variance.

Figure 2.

Genetic clustering analyses of Microbotryum lychnidis-dioicae microsatellite genotypes. Barplots display estimated individual admixture proportion for the global data set (= 773), including samples from both Europe and USA, assuming K = 3 to K = 7 clusters (a). Each individual is represented by a thin vertical line divided into K colored segments that represent the individual's estimated admixture fractions in the putative number of K clusters assumed. Black lines separate individuals from different geographic areas. In each case, only the most frequent solution among the 10 replicates is displayed. (b, c) Maps show the admixture proportions estimated from the structure analysis at K = 7 averaged over localities.

Figure 3.

Principal component analysis (PCA) on Microbotryum lychnidis-dioicae microsatellite genotypes. (a) Genotypes from Europe and the USA are plotted in the three-dimensional space defined by the first three principal components (PCs). Each genotype has been colored according to the assignment values based on the clustering analysis in Vercken et al. (2010). (b) PCA conducted on all genotypes that displayed score values for the three first PCs overlapping with those from the USA. Countries in which these genotypes occurred are coded with different colors and symbols. The dotted lines split individual genotypes into three distinct groups consistent with the structure clustering analysis.

Even with high K values we could not obtain finer resolution for the origin of the North American populations using the full data set (Fig. 2a). However, the clustering analyses on a reduced data set (= 149) focusing on American samples and the European samples that belonged to the same cluster at K = 7 showed that all strains from North America clustered with a few strains collected in Scotland (Fig. 4). Estimations of the number of clusters for this reduced data set, based on either the posterior probability calculation (P(K|X), where X is the data) (Pritchard et al., 2000) or the method advocated by (Evanno et al., 2005), indicated that K = 3 was the most likely solution (P(K = 3|X) = 0.99 and maximum ΔK obtained for K = 3; Fig. S2). Indeed, the Scottish samples were most closely related to American samples, having a multi-locus genotype almost identical to some strains found in North America, differing only at one locus, E18, which was the most polymorphic locus, and for which the American population had an allele that was not shared by any European population of M. lychnidis-dioicae.

Figure 4.

Genetic clustering analyses of Microbotryum lychnidis-dioicae microsatellite genotypes on a reduced data set (= 147) consisting of USA samples and European samples which grouped with the USA samples at K = 7 in the global population structure analysis. structure barplots for K = 2 to K = 4 and maps displaying results at K = 4 in the native and introduced areas are provided in (a), (b) and (c), respectively.

PCA analyses provided results consistent with the clustering analyses. Indeed, when focusing on the European genotypes that displayed score values on the first three PCs that overlapped with those of the North American samples, the introduced genotypes again appeared most closeley related to the Scottish samples (Figs 3b, 4, S3).

The level of divergence in allelic frequencies (FST) between North American populations and the five European clusters identified in Vercken et al. (2010) was significant for all pair-wise comparisons and ranged between 0.31 and 0.64 (Table 2). The lowest value was observed between the North American populations and the NWest cluster, in agreement with the close genetic relationship estimated from the PCA and structure analysis. The North American populations displayed a significantly lower level of genetic diversity than the European NWest cluster (Table 3), as estimated either using the allelic richness (Ar) or the genetic diversity (He) (Wilcoxon signed rank test, P < 0.001 for both Ar and He statistics). North American populations also had a very low private allelic richness (pAr) compared with the European values, meaning that all the alleles found in the North American samples were also detected in European populations. The selfing rate within the North American population estimated using the instruct program was high, as expected given the known mating system in Microbotryum species. The estimated selfing rate was significantly lower than in European populations (Tables 1, 3), but the difference was small and the variance in FIS values across loci was large (Table 1).

Table 3. Descriptive statistics within each of the Microbotryum lychnidis-dioicae European clusters and in the USA
  1. N, within-cluster sample size; Ar ± SD allelic richness; pAr ± SD private allelic richness; He ± SD expected heterozygosity; FIS value; S, selfing rate inferred from instruct program.

N 131159143938472
A r 4.14 ± 0.634.44 ± 0.675.67 ± 0.563.70 ± 0.524.95 ± 0.631.65 ± 0.16
pA r 0.44 ± 0.220.74 ± 0.211.14 ± 0.290.46 ± 0.190.84 ± 0.380.03 ± 0.03
H e 0.46 ± 0.260.46 ± 0.330.65 ± 0.120.38 ± 0.270.43 ± 0.280.10 ± 0.17
F IS 0.850.910.950.930.960.84
S 0.930.950.940.910.920.88


This study confirms that the introduced M. lychnidis-dioicae populations on S. latifolia in North America originated from an introduction of European populations of M. lychnidis-dioicae specialized on S. latifolia, and did not result from a host shift. Yet, among the documented cases of recent disease emergence as a result of introductions of fungal or oomycete pathogens into new continents, infection of novel hosts (i.e. host-shifts, host-range expansion, or pathogen spillover) is often involved (Parker & Gilbert, 2004; Slippers et al., 2005). Examples include the introduction of C. parasitica in the USA, the most probable source of which was Japanese chestnut trees (Castanea crenata) that were imported and planted throughout the country (Milgroom et al., 1996; Anagnostakis, 2001).

The strong genetic structure of M. lychnidis-dioicae in Europe allowed us to determine that Scotland is the most probable source of introduction for the North American populations. No footprint of admixture or hybridization with native Microbotryum species was found, which probably relates to the high selfing rate in M. lychnidis-dioicae (> 85%; Hood & Antonovics, 2004; Giraud et al., 2005; Gladieux et al., 2011), as well as to the strong host specificity and post-zygotic isolation among Microbotryum species (de Vienne et al., 2009a,b). The high level of homozygosity found in the North American populations also indicated a high selfing rate of the introduced populations. The fungus probably has thus kept the same mating system as in its native area. This is in contrast to other fungal or oomycete pathogens that do not undergo sexual reproduction in their introduced ranges, such as the rice blast fungus Magnaporthe oryzae (Saleh et al., 2012), the yellow rust fungus Puccinia striiformis f.sp tritici (Ali et al., 2010; Duan et al., 2010), or the potato late blight oomycete P. infestans (Day et al., 2004). In the latter case, only one mating type – A1 – had long been present in the introduced populations in Europe, preventing the occurrence of sex (Day et al., 2004). The introduction of the A2 mating type from Mexico in the 1970s, c. 130 yr after the A1 mating type, caused dramatic changes in the population structure and aggressiveness of the European P. infestans population as a consequence of the resulting sexual recombination (Goodwin, 1997; Cooke et al., 2012). Similarly, Phytophthora ramorum and P. cinnamomi reproduce only clonally in their world-wide introduced ranges, as only their respective A2 mating types occur in their invasive natural populations (Zentmyer, 1980; Old et al., 1984; Arentz & Simpson, 1986; Shearer & Tippet, 1989; Grünwald et al., 2012). In Microbotryum, the failure to evolve asexual reproduction is not surprising, as the formation of a dikaryon by sexual fusion of haploid cells is required for plant infection. The lack of change in the mating system also contrasts to some invasive plants, which have evolved higher selfing rates (Amsellem et al., 2001; Bossdorf et al., 2005; Colautti et al., 2010); however, this may not be surprising either given the already very high selfing rates of Microbotryum in its native range (Vercken et al., 2010).

The introduction of M. lychnidis-dioicae on S. latifolia in North America appears to have been more recent, and with a much stronger bottleneck, than the introduction of the host plant species (Taylor & Keller, 2007). Substantially fewer propagules were successfully introduced for the fungus than for the plant, and only in a small range on the eastern coast of North America, while several introductions occurred for S. latifolia, and independently on the west and east coasts (Taylor & Keller, 2007). Even in the area where M. lychnidis-dioicae is present in North America, several European lineages of S. latifolia have been detected, originating from different European regions (Taylor & Keller, 2007). Local adaptation is therefore not likely to be involved in the low number of M. lychnidis-dioicae populations introduced. However, we have not been able to integrate samples of the pathogen from the northeastern USA where the fungus has also been observed in a few places (J. Antonovics, pers. comm.), and the possibility remains that additional allelic variation may exist in M. lychnidis-dioicae in North America that has not been captured by the current study. The situation in M. lychnidis-dioicae and S. latifolia thus contrasts with plant pathogens such as Mycosphaerella graminicola (Banke & McDonald, 2005; Stukenbrock et al., 2007), Ustilago scitaminea (Raboin et al., 2007), Magnaporthe oryza (Couch et al., 2005a), Venturia inaequalis (Gladieux et al., 2008) and P. infestans (Gomez-Alpizar et al., 2007), which seem to share the same geographic origins as their respective host plant, that is, wheat, sugarcane (Saccharum officinarum), rice (Oryza sativa), apple (Malus domestica) and potato, respectively. This does not seem to be the case, however, for the barley (Hordeum vulgare) pathogen Rhynchosporium secalis (Brunner et al., 2007; Zaffarano et al., 2008) or for the oilseed rape (Brassica napus) pathogen Leptosphaeria maculans (Dilmaghani et al., 2012).

It is quite remarkable that there has been trans-continental introduction of the anther smut disease at all, as there is no seed transmission, dispersal of Microbotryum occurs by insects, and the fungus appears unable to persist as an environmental contaminant (Hood et al., 2010). One may speculate that, as probably occurred for the seeds of the host, diseased flowers of S. latifolia could have been transported along with forage materials aboard ships, as this plant is common in hay fields. Entry of the pathogen may have occurred through seedlings germinating among the spores where the forage material was off-cast upon arrival to North America. Such a scenario may be consistent with the source populations of introduced populations being from the UK, as human and commercial exchange are known to be especially frequent between these countries, and especially with the east coast. The successful invasion by an insect-borne biotrophic fungus shows that even pathogens presenting life history traits that are a priori less favorable for invasion (Philibert et al., 2011) can still cause biological invasions. An alternative explanation, that a living, diseased plant was transported from Europe to North America, seems particularly unlikely as S. latifolia is a short-lived perennial, easily propagated by seeds, and not of particular interest among gardeners. However, it should be noted that such an event is not entirely unprecedented, as the garden favorite moss campion, Silene acaulis, appears to have been introduced along with its anther smut to Hawaii (Farr et al., 1989), where this extremely long-lived species is more likely to be collected and transported as living specimens. Also, the anther smut disease of Dianthus caryophyllus was introduced by unknown means to Massachusetts in 1948, but the disease did not become established in this nonnative host and was present only in agricultural settings (Spencer & White, 1951). Another example of a priori unlikely invasion that nevertheless occurred is the case of the forest pathogen Heterobasidion annosum introduced to Italy by US troops during World War II (Gonthier et al., 2007).

The low genetic diversity in introduced populations is an important factor that is expected to limit invasion success. Multiple introductions can increase genetic diversity in introduced ranges (Genton et al., 2005), which has been argued to facilitate biological invasions. Here again, however, the M. lychnidis-dioicae case shows that, with only a few genotypes, a population of a fungal plant pathogen can become established in a new geographic range.

Biological invasions of whole continents by one or a few clonal lineages in fact appear to be frequent in fungal and oomycete pathogens (Goodwin et al., 1994; Couch et al., 2005b; Enjalbert et al., 2005; Raboin et al., 2007; Singh et al., 2008; Fisher et al., 2009; James et al., 2009; Mboup et al., 2009), such as P. infestans on potato (Goodwin et al., 1994), P. cinnamomi on trees (Old et al., 1984; Arentz & Simpson, 1986; Dobrowolski et al., 2003), U. scitaminea on sugarcane (Raboin et al., 2007), Batrachochytrium dendrobatidis on amphibians (James et al., 2009), Seiridium cardinale on cypress (Cupressus macrocarpa) (Della Rocca et al., 2011), or Puccinia striiformis on wheat (Mboup et al., 2009). Other invasions nevertheless resulted from introductions of a few divergent lineages from distinct sources, such as Plasmopara halstedii on sunflower (Helianthus annuus) in France (Delmotte et al., 2008; Ahmed et al., 2012), C. parasitica on chestnut in France (Dutech et al., 2010), the grape (Vitis vinfera) powdery mildew fungus Erysiphe necator (Brewer & Milgroom, 2010), the apple scab fungus Venturia inaequalis (Gladieux et al., 2008), or the oilseed rape (Brassica napus) pathogen Leptosphaeria maculans (Dilmaghani et al., 2012). Admixture might then occur between the lineages and influence their evolutionary dynamics. Little admixture has, however, been found between divergent introduced lineages in C. parasitica (Dutech et al., 2010, 2012) and P. ramorum (Ivors et al., 2006; Mascheretti et al., 2008, 2009), probably because of the predominantly asexual mode of reproduction of those pathogens. By contrast, it has been posited that hybridization between distant lineages created the hypervirulent strain in the chytrid fungus threatening amphibians world-wide (Farrer et al., 2011) and similarly led to new pathogenic abilities on resistant cultivars in the downy mildew pathogen of sunflowers (Ahmed et al., 2012).

The most devastating epidemics on plants caused by invasive fungal and oomycete pathogens most often involve novel host species that are closely related to the host in the pathogens’ area of origin. The present study nevertheless adds to a number of cases of invasions by pathogenic fungi in plants following the introduction of its host plant and shows that, even where life-history traits are unfavorable for invasions (Philibert et al., 2011), such invasions may occur. Investigations of the evolution of introduced populations are also interesting for the understanding and management of biological invasions. The elucidation of the origin of introduced M. lychnidis-dioicae populations on S. latifolia in North America will make it possible to perform more directed cross-inoculations to investigate coevolution with the North American plant genotypes, and whether there has been local adaptation, as shown in European S. latifolia for resistance against M. lychnidis-dioicae (Kaltz et al., 1999).


T.G. acknowledges receipt of grant ANR 07-BDIV-003 (Emerfundis project) and M.C.F. a postdoctoral grant from the Ile de France Région. M.E.H. acknowledges receipt of grants NSF-DEB 0747222 and 1115765. We thank Janis Antonovics for helping with collections and comments on the manuscript, and Doug Taylor for samples.