Open Collections

UBC Faculty Research and Publications

Cophylogeny of the anther smut fungi and their caryophyllaceous hosts: Prevalence of host shifts and… Refrégier, Guislaine; Le Gac, Mickaël; Jabbour, Florian; Widmer, Alex; Shykoff, Jacqui A; Yockteng, Roxana; Hood, Michael E; Giraud, Tatiana Mar 27, 2008

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-12862_2007_Article_669.pdf [ 663.41kB ]
JSON: 52383-1.0223734.json
JSON-LD: 52383-1.0223734-ld.json
RDF/XML (Pretty): 52383-1.0223734-rdf.xml
RDF/JSON: 52383-1.0223734-rdf.json
Turtle: 52383-1.0223734-turtle.txt
N-Triples: 52383-1.0223734-rdf-ntriples.txt
Original Record: 52383-1.0223734-source.json
Full Text

Full Text

ralssBioMed CentBMC Evolutionary BiologyOpen AcceResearch articleCophylogeny of the anther smut fungi and their caryophyllaceous hosts: Prevalence of host shifts and importance of delimiting parasite species for inferring cospeciationGuislaine Refrégier1, Mickaël Le Gac1,2, Florian Jabbour1, Alex Widmer4, Jacqui A Shykoff1, Roxana Yockteng1,3, Michael E Hood5 and Tatiana Giraud*1Address: 1Ecologie, Systématique et Evolution, Bâtiment 360, Université Paris-Sud, F-91405 Orsay cedex, France ; CNRS F-91405 Orsay cedex, France, 2Department of Zoology, 6270 University Boulevard, Vancouver BC V6T 1Z4, Canada, 3MNHN UMR 5202, Unité Origine, structure et évolution de la biodiversité, Département Systématique et Evolution, 16 rue Buffon CP 39 75005, France, 4ETH Zurich, Institute of Integrative Biology, Plant Ecological Genetics, Universitätstr. 16, 8092 Zurich, Switzerland and 5Department of Biology, McGuire Life Sciences Building, Amherst College, Rts 9 & 116, Amherst, MA 01002-5000, USAEmail: Guislaine Refrégier -; Mickaël Le Gac -; Florian Jabbour -; Alex Widmer -; Jacqui A Shykoff -; Roxana Yockteng -; Michael E Hood -; Tatiana Giraud* -* Corresponding author    AbstractBackground: Using phylogenetic approaches, the expectation that parallel cladogenesis shouldoccur between parasites and hosts has been validated in some studies, but most others providedevidence for frequent host shifts. Here we examine the evolutionary history of the associationbetween Microbotryum fungi that cause anther smut disease and their Caryophyllaceous hosts. Weinvestigated the congruence between host and parasite phylogenies, inferred cospeciation eventsand host shifts, and assessed whether geography or plant ecology could have facilitated the putativehost shifts identified.For cophylogeny analyses on microorganisms, parasite strains isolated from different host speciesare generally considered to represent independent evolutionary lineages, often without checkingwhether some strains actually belong to the same generalist species. Such an approach may mistakeintraspecific nodes for speciation events and thus bias the results of cophylogeny analyses ifgeneralist species are found on closely related hosts. A second aim of this study was therefore toevaluate the impact of species delimitation on the inferences of cospeciation.Results: We inferred a multiple gene phylogeny of anther smut strains from 21 host plants fromseveral geographic origins, complementing a previous study on the delimitation of fungal speciesand their host specificities. We also inferred a multi-gene phylogeny of their host plants, and thetwo phylogenies were compared. A significant level of cospeciation was found when each hostspecies was considered to harbour a specific parasite strain, i.e. when generalist parasite specieswere not recognized as such. This approach overestimated the frequency of cocladogenesisPublished: 27 March 2008BMC Evolutionary Biology 2008, 8:100 doi:10.1186/1471-2148-8-100Received: 28 August 2007Accepted: 27 March 2008This article is available from:© 2008 Refrégier et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.Page 1 of 17(page number not for citation purposes)because individual parasite species capable of infecting multiple host species (i.e. generalists) werefound on closely related hosts. When generalist parasite species were appropriately delimited andBMC Evolutionary Biology 2008, 8:100 a single representative of each species was retained, cospeciation events were not morefrequent than expected under a random distribution, and many host shifts were inferred.Current geographic distributions of host species seemed to be of little relevance for understandingthe putative historical host shifts, because most fungal species had overlapping geographic ranges.We did detect some ecological similarities, including shared pollinators and habitat types, betweenhost species that were diseased by closely related anther smut species. Overall, genetic similarityunderlying the host-parasite interactions appeared to have the most important influence onspecialization and host-shifts: generalist multi-host parasite species were found on closely relatedplant species, and related species in the Microbotryum phylogeny were associated with members ofthe same host clade.Conclusion: We showed here that Microbotryum species have evolved through frequent hostshifts to moderately distant hosts, and we show further that accurate delimitation of parasitespecies is essential for interpreting cophylogeny studies.BackgroundHost-specific differentiation of parasites, also referred toas specialization, may arise as a consequence of limiteddispersal or adaptive constraints [1]. Some parasite spe-cies or lineages may indeed have evolved a restricted hostrange simply because they have not come in contact withother host species, for instance when these occur in allo-patry. Alternatively, host specificity may arise because ofadaptive specialization [2], where trade-offs or fitnesscosts of being generalist parasites can lead to host-specificdifferentiation even in sympatry [1-4].Host specificity is often expected to lead to cospeciation,i.e. parasite speciation tracking that of the host. For infer-ring whether cospeciation has occurred, one usually com-pares host and parasite phylogenies. Cospeciation yieldscongruent phylogenies (i.e. parallel cladogenesis), as hasbeen previously illustrated, for example in the cases ofanimal hosts and parasitic lice [5] and of some mutualistassociations [6,7]. In contrast, the colonization of newhosts, either followed by parasite speciation (host shift) ornot ('failure to speciate' [8]), generally decreases phyloge-netic congruence [9,10]. Additional processes are alsoexpected to reduce the congruence between host and par-asite phylogenies, such as species extinctions [5] andduplications (i.e. speciation of the parasite in absence ofhost speciation).Several recent studies comparing the phylogenies ofhighly specific parasites and their hosts revealed wide-spread incongruence [9,11-14], showing that apparentlystrict host specificity is not sufficient to impede host shiftsover the long term. In cases of host shifts, colonization isexpected to be most likely between geographically over-lapping hosts. The probability of the parasite being able todevelop on a new host may also be influenced by thetigating geographical and ecological similarities betweenhosts can therefore help reconstructing evolutionary his-tory of host shifts.In addition, if related parasites more easily infect hostswith similar ecologies and if chemical, physiological andecological characters in part covary with phylogeny, falseconclusions can arise from cophylogeny analyses. In thefirst place, host shifts will preferentially occur betweenphylogenetically closely related hosts [15], which can gen-erate similar degrees of congruence between parasite andhost phylogenies as would cospeciation [16]. Second,generalist parasites that can infect several host species arelikely to be found on closely related host species, eitherbecause these generalist species are the result of a lack ofparasite speciation following host speciation or of hostrange expansion among phylogenetically close hosts thatshare biochemical and ecological characteristics. If gener-alist species are not recognized as such, intraspecific nodesin parasite phylogenies may be mistaken for speciationevents and then misinterpreted as cospeciation events ifthe generalist parasites infect sister host species. This latterpotential pitfall in cophylogeny analyses has not beeninvestigated yet to our knowledge.The anther smut fungus Microbotryum violaceum (Pers.:Pers) Deml & Oberw. (= Ustilago violacea (Pers.) Fuckel)(Basidiomycota: Pucciniomycotina, Microbotryaceae) isan obligate parasite on many Caryophyllaceae. It has beenrecorded on 92 plant species in Europe and on 21 plantspecies in North America [17]. The Caryophyllaceae –Microbotryum pathosystem is a model in many fields ofevolutionary biology [18-20]. In diseased plants, diploidteliospores of M. violaceum replace pollen in the anthers,and are dispersed by insect visitors. Host specific diver-gence of M. violaceum has been of debate for about a cen-Page 2 of 17(page number not for citation purposes)degree to which the new host shares chemical, physiolog-ical and ecological characters with the original host. Inves-tury. Spore color [21], mating behavior [21],morphological differences [22], and cross-inoculationBMC Evolutionary Biology 2008, 8:100 [23,24], all suggested that M. violaceumstrains found on different host species were at least par-tially differentiated. More recently, genetic analyses of M.violaceum populations from many hosts have revealedstrong genetic differentiation [25-28]. Finally, anapproach using multiple gene phylogenies has firmlyestablished that most host races of M. violaceum representmultiple independent evolutionary lineages, highly spe-cialized on a single or a few host species [29]. Microbot-ryum violaceum is thus a complex of more than fifteen truesibling species, showing a strong post-zygotic isolation[30].The host family, Caryophyllaceae, has a global distribu-tion with highest diversity in the holarctic, but also highdiversity in the Mediterranean and Irano-Turanean region[31]. The majority of the approximately 2,200 species ofthe family are heliophytes occurring in dry, open habitats.Some species are restricted to mountainous regions andthe family is totally absent from lowland rain forests [31].Microbotryum violaceum commonly causes disease on spe-cies from two subfamilies of the Caryophyllaceae, theAlsinoideae and the Caryophylloideae, and is most preva-lent on perennials [17]. The systematics of the Caryophyl-laceae still mainly relies on morphological charactersalthough there have been recent efforts to reconstruct thephylogenetic relationship among genera based on molec-ular data [32-34]. This is unfortunately insufficient foraddressing questions on the Caryophyllaceae – Microbot-ryum association.The goals of this paper are to reconstruct the evolutionaryhistory of associations between the Microbotryum anthersmut fungi and their Caryophyllaceae hosts and to test theeffect of parasite species delimitation on cophylogenyanalyses. We addressed the following questions:1) What is the evidence for cospeciation and/or host shiftsin this system?2) Does delimitation of cryptic fungal species influencethe results of cophylogeny analyses?3) Do host shifts occur preferentially onto phylogeneti-cally close hosts?4) Does geography explain parasite similarity?5) Do ecological factors, such as pollinator spectra orplant habitat, influence host shifts?ResultsHost phylogenyThe topologies of the host phylogenetic trees based on theITS, intron trnL and spacer trnLF sequences showed no sig-nificant incongruence neither when assessed via anApproximately Unbiased (AU) tests (Table 1) nor wheninspected by eye (see Methods for details). A tree wastherefore constructed using the concatenated loci (Fig. 1).This tree adds support to the monophyly of the generaDianthus, Saponaria, and Lychnis, as previously reportedfor a smaller set of species [33,35]. Within Silene, we iden-tified two highly supported clades, that we named Silenetype I and Silene type II.Smut phylogenies and identification of phylogenetic speciesOur current findings lend additional support for the exist-ence of a number of cryptic Microbotryum species, eachspecific to one or a few host species [29]. As before, thespecies status of the various clades rests on the congruenceof the phylogenetic relationships of many strains inferredfrom the three fungal genes analyzed (Additional file 1, 2and 3; see Methods for details on the inference of phylo-genetic species). This congruence is only not met for theclade formed by the strains collected from S. acaulis asobserved previously [29]. For these strains the γ-tub phyl-ogeny differs from the two other single gene phylogenies(see AU tests in Table 2). The two placements of this line-age being basal, this incongruence is probably due to anancient introgression or hybridization [29]. Without theS. acaulis strains there was no remaining incongruence: allsupported nodes were identical in the three gene trees. Wetherefore concatenated the three genes in a dataset with-out the S. acaulis strains, yielding the same tree topologyTable 1: Results of the AU (Approximately Unbiased) tests for the plant dataset.Enforced topology (MP tree with bootstraps > 70)ITS trnL trnLFGene used ITS 0.612 0.326 NAtrnL 0.388 0.630 NAtrnLF 0.388 0.301 NAPage 3 of 17(page number not for citation purposes)P-values lower than 0.05 indicate that the likelihood of the topology obtained using one focal gene is significantly different from the likelihood of the enforced topologies (obtained using each other gene). NA indicates that no tests were possible because the phylogenetic relationships potentially incongruent between the alternative topologies were not resolved in the focal topology.BMC Evolutionary Biology 2008, 8:100 4 of 17(page number not for citation purposes)Bayesian 50% majority-rule consensus tree of the Caryophyllaceae hosts used in this study based on the concatenation of the ITS, intron trnL nd spacer trnLFFigure 1Bayesian 50% majority-rule consensus tree of the Caryophyllaceae hosts used in this study based on the con-catenation of the ITS, intron trnL and spacer trnLF. Statistical supports indicate Bayesian Posterior Probabilities (BPP)/Maximum Parsimony Bootstraps/Neighbor-Joining Bootstraps. Only nodes supported by more than two methods are indi-cated, the significant statistical supports being considered as higher than respectively 0.9/70/70. The tree is rooted based on previous studies (see text).StellariaGypsophilaSaponariaAtocionSilenetype ISilenetype IILychnisDianthus0.1Stellaria sp. A. rupestrisS. vulgarisS. dioicaS. latifolia1.00/99/98S. lemmoniiS. carolinianaS. virginica1.00/100/1001.00/99/1001.00/98/98S. nutansS. acaulisS. otites1.00/97/90L. flos-cuculiL. flos-jovis1.00/98/971.00/100/100G. repensSa. ocymoidesSa. officinalis1.00/100/1000.97/89/73D. carthusianorumD. monspessulanusD. gratianopolitanusD. superbusD. sylvestris1.00/100/1001.00/89/82BMC Evolutionary Biology 2008, 8:100 in the individual phylogenies but with higher supportfor some nodes (Fig. 2).In addition to the cryptic species identified previously[29], three new phylogenetic species could be inferredfrom the strains analyzed in this study, respectively on L.flos-jovis, on Saponaria officinalis, and on S. otites (seeFig. 2 for species nomenclature extended from ref. [29]).We had access to only a single strain from S. lemmoniiand Stellaria sp. plants, so we could not infer the existenceof specific fungal phylogenetic species on these hosts.However, because these strains were both strongly sepa-rated from other identified phylogenetic species in theirrespective clades and because they were isolated fromphylogenetically or geographically very distant hosts, wenevertheless considered them as separate lineages in thecomparison of host and fungal phylogenies below. ForStellaria, this choice is in agreement with an ITS phylog-eny supporting the monophyly of several strains collectedfrom this genus with the same phylogenetic placement asours [28].On the 21 host species screened in the end we identified atotal of 15 parasite species (Fig. 2). Most of them appearedspecific to a single host species. For both Atocion rupestris(previously Silene rupestris [36]) and Saponaria ocymoidesthe two strains analyzed belonged to different fungal spe-cies. Because these hosts are seldom infected (pers. obs.)we suspect that we picked up transient opportunisticinfections, so we did not consider these plant species astrue hosts of the corresponding fungal species. The fungalspecies identified on the Dianthus species and on Gyp-sophila repens appeared generalist, i.e. able to infect severalhost species. Speciation on these hosts may alternativelybe too recent to have allowed for sufficient moleculardivergence for us to detect.For some of the subsequent analyses (TreeMap, TreeFitter,Icong, see below and Methods), all nodes must beresolved. In that case, we used a previous study thatpolytomies, we considered several alternative topologies(see legend of Tables 3, 4 and 5 for details and nomencla-ture) including the branch of the fungal species on S. acau-lis for parasite phylogeny at the two alternativeplacements.Comparison of plant and fungal phylogeniesWe used two approaches for comparing host and parasitephylogenies: the more conventional one considered as aseparate taxon the parasite strains from different host spe-cies such that generalist species were represented by asmany terminal branches as host species on which theyoccurred (Fig. 3), while the second approach comparedspecies phylogenies (retaining a single representative perfungal species) (Fig. 4). In both cases, some broad-scalecongruence between the host and parasite phylogeniesappears by visual inspection (the fungal phylogenetic spe-cies infecting the Dianthus and the Saponaria are mono-phyletic, as are the plants) while fine-scale congruencebetween the host and parasite phylogenies is weak, in par-ticular among the parasites of Silene. However methodo-logical analyses reconstructed different evolutionaryhistories for Microbotryum-Caryophyllaceae associationdepending on which approach was chosen.When considering as a separate taxon the parasite strainsfrom different host species (i.e. retaining one strain perhost species and including opportunistic infections), allmethods used for comparing these host and parasite phy-logenies revealed a significant number of cospeciationevents or of congruence level as compared to randomassociations. TreeMap [37], which seeks to minimize hostshifts, inferred a significantly higher number of cospecia-tion events than expected from a random distribution butrequired many duplications and extinctions to achievethis (Table 3, upper half) and five distinct smut specieswere inferred on the ancestral host. The results were simi-lar regardless of which topology was chosen for the unre-solved nodes. Interestingly, TreeFitter [38] inferred amuch higher number of host shifts than cospeciation,Table 2: Results of the AU (Approximately Unbiased) tests for the Microbotryum violaceum dataset including strains from Silene acaulis.Topology enforced (MP tree with bootstraps > 70)β-tub γ-tub Ef1αGene used β-tub 0.995 0.002 0.346γ-tub 0.005 0.999 <0.001Ef1α <0.001 0.002 0.681P-values lower than 0.05 indicate that the likelihood of the topology obtained using one focal gene is significantly different from the likelihood of the enforced topologies (obtained using each other gene). When the Microbotryum strains parasitizing S. acaulis were removed in the fungal trees, all topologies were identical.Page 5 of 17(page number not for citation purposes)obtained high supports to resolve one of our polytomies[33] (see symbol * on Figs. 3 and 4). For the remainingduplication and extinction events, and this was trueregardless of the topology and the costs chosen (Table 4,BMC Evolutionary Biology 2008, 8:100 6 of 17(page number not for citation purposes)Bayesian 50% majority-rule consensus tree of the Microbotryum strains analyzed in this study based on the concatenation of β-tub, γ-tub and EF1α seq en es, and delimitation of the c rresponding speciesFigure 2Bayesian 50% majority-rule consensus tree of the Microbotryum strains analyzed in this study based on the con-catenation of β-tub, γ-tub and EF1α sequences, and delimitation of the corresponding species. Statistical supports indicate Bayesian Posterior Probabilities (BPP)/Maximum Parsimony Bootstraps/Neighbor-Joining Bootstraps. Only nodes sup-ported by more than two methods are indicated, the significant statistical supports being considered as higher than respectively 0.9/70/70. High support in individual gene trees was indicated: + for the γ-tub tree, × for the Ef1α tree, # for the β-tub tree. The tree is rooted based on previous studies (see text). Taxa labels correspond to the host plant on which fungal strains were collected. Brackets indicate clades and evolutionary units, i.e. cryptic fungal species, identified in the study using the three indi-vidual gene phylogenies (see text).0.1Svirginica 387Scaroliniana Sc2Svirginica TG117sv1.00/100/100Lflosjovis 410Lflosjovis M270.99/91/97Arupestris 383Svulgaris 30027Svulgaris 79130.91/79 /801.00/100/100Slemmonii SlemSotites SottLfloscuculi 9205Snutans 8742Snutans LbSnutans TG3031.00/100/100Svulgaris 7807Svulgaris 79031.00/99/981.00/85/Sdioica 7212Sdioica 7237Sdioica SdbArupestris 105021.00/99/100Slatifolia 4001Slatifolia 4106Slatifolia Sl21.00/95/871.00/100/1001.00/96/960.98/86/1.00/100/100Stellaria sp TG325Saofficinalis TG116Dcarthusianorum 7022Dsylvestris 9119Dsylvestris TG3171.00/99/100Dgratianopolitanus 332Dmonspessulanus TG328Dmonspessulanus 12920Dgratianopolitanus 330Dsuperbus 8718Dmonspessulanus 12919Dmonspessulanus TG3061.00 /92/930.99/ 86/751.00/100/100Dcarthusianorum TG66Grepens Gr137Dcarthusianorum 30901Dcarthusianorum 0902Dcarthusianorum 31602Dcarthusianorum TG651.00/100/1001.00/83/811.00/98/971.00/100/991.00/99/ 961.00/98/941.00/100/100 MvSspAMvLfjMvSv1MvSlemMvSotMvLfcMvStelMvSnMvSv2MvSdMvSlMvSoffMvDcMvDsp1MvDsp2Sacaulis 339Sacaulis 380Sacaulis Sa1Sacaulis 31401Sacaulis 31402MvSa+ x #+ x #x  x  + x #+ x + x ++ x + x + x +  #+  +  #+  x +  + x + x #+ x + x +  BMC Evolutionary Biology 2008, 8:100 half). The number of inferred cospeciation eventswas nonetheless higher than expected from a random dis-tribution in all cases. ParaFit [39] detected a significantcorrelation between the plant and fungal trees (ParaFit-Global = 0.0014, P < 0.0001 using the genetic distancesand ParaFitGlobal = 0.7643, P < 0.0001 using the patristicdistances). The Icong index [40] also indicated that thecongruence between plant and fungal trees was significantor marginally significant (Table 5, upper half). Further-more, we found a significant positive relationshipbetween the genetic distances between pairs of host plantsand the distance between the pairs of associated Microbot-ryum species using a Mantel test (a = 0.031, b = 0.125, P <0.001).The second approach, that we consider more correct,retains a single representative per fungal species, associ-ated to each of the multiple hosts for generalist species(Fig. 4), which leads to ignore the opportunistic infec-tions. This differs from the conventional approach in thatTable 3: Numbers of the different evolutionary events inferred in order to reconcile the plant and smut phylogenies by TreeMap (maximizing the number of cospeciation events).Approach Plant tree Fungal tree Cospeciation Duplication Host shift Sorting events Nb of events PKeeping one strain per host species Max A1 11 11 1 60 83 0.021*A2 12 11 0 58 81 0.041*Min A1 11 12 0 68 91 0.023*A2 11 12 0 67 90 0.019*Keeping a single representative per fungal speciesMinT B1 5 10 0 48 63 >0.5B2 5 10 0 46 61 >0.5MaxT B1 6 9 0 43 58 0.287B2 6 9 0 41 56 0.268Eight data sets were used. For the fungi, "A" trees contain one strain per smut species and per host (see Fig. 3) and "B" trees a single representative per smut species (see Fig. 4), and two possible topologies were used, "1" and "2", depending on the position of MvSa. For the plants, poorly supported nodes were resolved in order to maximize the congruence with the fungal trees in the "Max" trees and minimize it in the "Min" trees, and the "T" trees contain only host species on which fungal species are well established (i.e. ignoring the opportunistic infections). See also Figs. 3 and 4 for the correspondence between host and parasite phylogenies for the least congruent combinations.Table 4: Mean numbers of the different evolutionary events inferred in order to reconcile the plant and smut phylogenies by TreeFitter.Approach Plant tree Fungal tree Costs Cospeciation Duplication Host shift Sorting events Nb of events Total cost PKeeping one strain per host speciesMax A1 C 6 0.5 16.5 2.5 25.5 32 0.0048**D 1 0 22 0 23 0 0.0371*A2 C 7 0.5 15.5 3.5 26.5 31 0.0042**D 1 0 22 0 23 0 0.0399*Min A1 C 4 0 19 1 24 36 0.0430*D 1 0 22 0 23 0 0.0364*A2 C 5 0 18 2 25 35 0.0291*D 1 0 22 0 23 0 0.0434*Keeping a single representative per fungal speciesMinT B1 C 3 0 12 2 17 30 0.0652D 0.5 0 14.5 0 15 0 0.3209B2 C 3 0 12 2 17 23 0.0903D 0.5 0 14.5 0 15 0 0.3355MaxT B1 C 1 0 14 1.5 16.5 30 0.1937D 0 0 15 0 15 0 1B2 C 1 0 14 1.5 16.5 30 0.2331D 0 0 15 0 15 0 1Page 7 of 17(page number not for citation purposes)Eight data sets were used; see legend from Table 3 for their correspondence. Two sets of costs were used: C) maximizing the likelihood of cospeciation events: cospeciation = -1, duplication = 0, sorting events = 2, host shifts = 2; D) minimizing the number of events: cospeciation = 0, duplication = 0, sorting events = 2, host shifts = 0.BMC Evolutionary Biology 2008, 8:100 requires the proper delimitation of parasite species sincespecies are not defined by the host, but rather as a lineagewith an independent evolutionary history. In this caseneither TreeMap (Table 3, bottom half) nor TreeFitter(Table 4, bottom half) inferred significantly more cospe-ciation events than expected from a random distribution,Comparison of plant and fungal phylogenies using one strain per host speciesFigure 3Comparison of plant and fungal phylogenies using one strain per host species. Representation of the associations between Caryophyllaceae (left) and Microbotryum (right) with the a priori least congruent combinations between all possible resolved topologies for host and parasite trees, using one strain per host species (with 'Min' topology for the plant tree, and 'A1' topology for fungal tree, see Table 3). The symbol * highlights resolution of a polytomy in the plant tree according to a previous study [33]; The symbol ¤ highlights resolutions in the plant tree and in the fungal tree that differed from the other topology tested (Max versus Min in plants, and 2 versus 1 for fungi); The symbol - highlights resolutions that had no impact on Caryophyllaceae MicrobotryumStellariaSa. officinialisSa. ocymoidesG. repensD. carthusianorumD. monspessulanusD. sylvestrisD. gratianopolitanusA. rupestrisL. flos-cuculiL. flos-jovisS. nutansS. otitesS. acaulisS. dioicaS. latifoliaS. vulgarisS. lemmoniiS. carolinianaS. virginicaMvStellariaMvSaocymoidesTG96MvGrepensGr137MvDcarthusianorum30901MvDcarthusianorum7022MvSaofficinalisTG116MvDsylvestris9119MvDmonspessulanus12919MvDgratianopolitanus332MvLfloscuculi19205MvSnutansLbMvSvulgaris7807MvArupestris10502MvSdioica7212MvSaocymoidesOc1MvSlatifolia4001MvSottitesSottMvSlemmoniiSlemMvLflosjovis410MvArupestris3832MvSvulgaris30027MvSacaulis380MvScarolinianaSc2MvSvirginicaTG117sv*-¤¤¤¤¤--Table 5: Icong index and significance of the congruence level between Caryophyllaceae and Microbotryum trees.Approach Plant Tree Fungal tree Icong value PKeeping one strain per host species MaxD A1 1.400 0.007**A2 1.541 0.001****MinD A1 1.260 0.057A2 1.260 0.057Keeping a single representative per fungal species MaxDS B1 1.376 0.016*B2 1.549 0.001****MinDS B1 1.204 0.175B2 1.376 0.016*Eight data sets were used using fungal topologies described in Table 3. Plant trees had to be adjusted to be compatible with the fungal tree as the Icong program does not allow for multiple associations. In the name of the plant tree "D" indicates that hosts harbouring several fungal species were duplicated and "S" indicates that one host was chosen for generalist parasite species. These choices were made to maximize or minimize further the congruence with the fungal tree as compared to Max and Min trees (see legend Table 3). P-value < 0.05 indicated that the congruence between host and parasite trees was higher than that of two random trees.Page 8 of 17(page number not for citation purposes)congruence between the two phylogenies. Dots on the nodes indicate where cospeciation events were inferred by TreeMap.BMC Evolutionary Biology 2008, 8:100 of the chosen topologies for the unresolvednodes. For one of the two possible topologies of the plantphylogeny (MaxT, Table 4), TreeFitter even inferred onlyhost shifts when costs where set to minimize the totalnumber of events. Furthermore, the pairwise plant andMicrobotryum genetic distances were not significantly cor-related (Mantel test, a = 0.035, b = 0.064, P = 0.1173).Only the ParaFit and Icong analyses remained significant(ParaFitGlobal = 0.0014, P = 0.0002 using the genetic dis-tances and ParaFitGlobal = 0. 7492, P = 0.00010 using thepatristic distances; with Icong, P < 0.05 for three out offour of the combinations tested, Table 5, lower half), indi-cating that the parasite and host phylogenies were stillmore similar than expected by chance. The significantassociations in the ParaFit results included both associa-tions between Dianthus and Saponaria host species andtheir parasites, and some of the Silene species. The signifi-cance of these associations may be due to the symmetry ofthe two trees regarding the main clades: for instance, par-We conclude that the failure to appropriately delimit par-asite species and represent generalist parasite species asseveral separate taxa (one per host) had introduced a bias.Inspection of nodes for which reconciliation analyses sup-ported cospeciation in TreeMap showed that the strainsharboured by Dianthus spp. that belonged to the samegeneralist fungal species were considered as lineagesarisen by cospeciation in the TreeMap reconciliation. Inthe ParaFit analyses, this same group of strains contrib-uted almost half of the significant associations (data notshown). Thus, cospeciation events were inferred wherethere had been no speciation at all, but rather a failure tospeciate or a host range expansion towards closely relatedhosts. Thus failure to delimit parasite species inflated thesignificance of congruence between the host and parasitetrees.Geographic patternsMost European host plant species used in this study haveComparison of plant and fungal phylogenies using a single representative per fungal speciesFigure 4Comparison of plant and fungal phylogenies using a single representative per fungal species. Representation of the associations between Caryophyllaceae (left) and Microbotryum (right) with the a priori least congruent combinations between all possible resolved topologies for host and parasite trees, using a single representative per fungal species (with 'MinT' topology for the plant tree, and 'B1' topology for fungal tree, See Table 3). See Fig. 3 for symbol legend.Caryophyllaceae MicrobotryumStellariaSa. officinialisG. repensD. carthusianorumD. monspessulanusD. sylvestrisD. gratianopolitanusL. flos-cuculiL. flos-jovisS. nutansS. otitesS. acaulisS. dioicaS. latifoliaS. vulgarisS. lemmoniiS. carolinianaS. virginicaMvStelMvSaoffMvDcMvDsp1MvDsp2MvSotMvLfcMvSnMvSv2MvSdMvSlMvSlemMvLfjMvSv1MvSaMvSspA*-¤¤¤¤--Page 9 of 17(page number not for citation purposes)asites found on Dianthus and Saponaria are monophyleticas are the plants.overlapping ranges. For instance, all the phylogenetic spe-cies, except those of the genus Stellaria, include at leastBMC Evolutionary Biology 2008, 8:100 sample from the Western Alps and most of them atleast one sample from the French Pyrenees (Additionalfile 4). The current geographic distribution of host speciestherefore appears to contribute little to the phylogeneticrelationships among fungal species. In North America thehosts represent a clade whereas the smut sample collectedfrom S. lemmonii was not related to smuts from the twoother North American hosts (Fig. 2). Therefore we foundno geographic pattern explaining the phylogenetic rela-tionships of the fungal species.Relatedness among fungal species and host ecologyIn cases of host shifts, host ecology may play a role byfacilitating contact between a parasite and a new host thathas, for instance, similar pollinators or habitats as theoriginal host. We therefore investigated whether recenthost shifts, detected from non-congruence in the terminalbranches of the host and parasite phylogenies (see Fig. 5),occurred between hosts with similar ecologies such astype of habitats and pollinator guilds. We detected severalinteresting ecological similarities among hosts withrelated smut lineages (see boxes on Fig. 5): the pair S. vul-garis/L. flos-jovis grows on well drained soils, such as cal-careous meadows, at least in the Western Alps, whichcould have facilitated potential host shift 1. Sphingidae,Noctuidae (in particular Hadena bicruris), Apidae and Syr-phidae all visit Silene nutans, L. flos-cuculi, S. vulgaris, S. lat-ifolia and S. dioica and may thus have been the agents ofhost shifts 3 and 4. Regarding potential host shift 4, Lych-nis flos-cuculi, S. vulgaris, as well as S. nutans are oftenfound together on shady borders between fields andwoods. The Dianthus spp. and G. repens both grow onexposed rocky areas and are pollinated by Syrphidae,potentially favouring host shift 5. Intriguingly, S. otitesshares little similarities with other Silene from which itcould have inherited its parasites. It is mainly wind-polli-nated, only shares butterflies as insect pollinators with S.lemmonii which could have facilitated host shift 2, butthese two plant species have no current geographical over-lap to our knowledge. This suggests that the incongru-ences 1, 3, 4, and 5 (Fig. 5) between host and parasitephylogenies, if indeed the results of host shifts, may havebeen facilitated by ecological similarities of the plants.However, the lack of specificity of insect pollinators, thelarge overlap of the ecotypes of the different hosts and theprobable rapid evolution of ecological traits at the exam-ined scale render any such conclusions speculative.DiscussionCaryophylloideae phylogenyThe host phylogeny, here including 20 species within thesub-family Caryophylloideae provides interesting insightsin itself, because the relationships among Caryophyl-American hosts appear monophyletic, branching withinthe European species. This would be consistent with theirevolution by allopatric speciation following a single colo-nization event from Europe, after the separation of thesetwo continents. Fossil records suggest an origin of theCaryophyllaceae 70 Mya in Australia, the oldest fossils inEurope dating back only 30–50 Mya [35], by which timeNorth America was separated from Europe by the AtlanticOcean [41].Our dataset, with additional species compared to previousstudies [33,35], further supports the monophyly of thegenera Lychnis, Saponaria and Dianthus but unfortunatelyfails to resolve the monophyly of the genus Silene. The rel-ative position of species within the Dianthus clade was theleast resolved, suggesting recent radiation of Dianthus spe-cies. This may be particularly relevant for the study ofMicrobotryum-Caryophyllaceae association because thesmut lineages on Dianthus species also appear to be lesshost-specific than those on Silene species.Importance of delimiting parasite species in cophylogeny studiesWe detected a number of cryptic generalist species withmultiple closely related hosts. When these were dupli-cated in the phylogeny to have one representative per hostspecies, the congruence between the host and parasitephylogenies was artificially inflated. This shows that care-fully delimiting parasite species boundaries in cophylog-eny studies is important for assessing the degree ofcophylogenetic history, although it is rarely done formicro-organisms. Most often parasites are simply sam-pled from different host species and the resulting tree iscompared to the tree for host species [9,14]. Such anapproach implicitly assumes that each host species har-bours one distinct parasite species, an assumption that wehere show to be both unfounded and of potentially greatimportance in this kind of study. The Microbotryum-Cary-ophyllaceae system shows further that generalist parasitespecies are likely to infect a range of closely related hosts.Therefore parasite species boundaries must be carefullydelimited to avoid erroneous interpretations of the degreeof congruence between host and parasite phylogenies.Inference on Microbotryum evolutionary history: host shifts versus cospeciationAfter having carefully delimited fungal species, thenumber of cospeciation events inferred by TreeMap andTree Fitter when reconciling the Microbotryum phylogenyto that of their Caryophyllaceaous hosts was not higherthan expected from a random distribution, as suggestedby a previous study on a smaller sample [14] nor was therea significant correlation (Mantel test) between genetic dis-Page 10 of 17(page number not for citation purposes)laceae are still poorly known, in particular at the sub-genus level [32-34]. It is striking that the three Northtances of host plant and their respective anther smut spe-cies.BMC Evolutionary Biology 2008, 8:100 and ParaFit still detected significant congruencebetween the host and fungal phylogenies, the significantassociations in ParaFit being mostly due to the parasitespecies found on Dianthus and Saponaria. The significanceof the global ParaFit and Icong tests therefore most prob-ably stems, not from pervasive cospeciation events, butfrom the fact that the anther smut lineages infecting theDianthus and Saponaria plants were monophyletic, as werethe plants. This pattern is consistent with cospeciation orhost shifts between closely related members of theseclades, together with an absence of host shifts betweendistantly related hosts like Silene and Dianthus. Host shiftsappear to have occurred between hosts with limited phyl-ogenetic distances in the Microbotryum-Caryophyllaceaedisease system. Incipient host shifts by Microbotryum havebeen reported onto S. vulgaris from S. latifolia [18,42] andonto Gypsophila repens from the closely related Petrorhagiaimpeding shifts to hosts that are genetically too distant.Investigating the branch lengths and divergence times ofthe two phylogenies would be necessary to determinewhether cospeciation occurred at all [44]. Unfortunately,available fossil records do not allow calibration of thesequence divergence between the Microbotryum and itshost plant species [35,45].Overall, our results suggest that cospeciation is not therule in the Microbotryum-Caryophyllaceae system, thathost shifts were pervasive, but that fungal species couldnot shift to too distant host species (shift either associatedwith speciation or not). Comparing the results of reconcil-iation analyses to that of methods investigating simplecongruence was important for identifying these con-straints on host shifts imposed by host phylogeny.Ecology of the host plants of the different Microbotryum speciesFigure 5Ecology of the host plants of the different Microbotryum species. Phylogenetic tree of the Microbotryum species with indication, on the branches, of their plant host clade (see Fig. 1 for host clade delimitation). The numbers in brackets indicate the five terminal incongruences between the plant and fungal phylogenies. Here MvDsp1 and MvDsp2 were not distinguished (fused in MvDsp) because of poor resolution in this part of plant phylogeny and thus inability to identify any incongruence between the two phylogenies. Information about the ecology of the host plants is given: the most common clades of pollinators are indicated in black and type of habitat in grey (Fields from the U.S. were drawn apart to highlight the geographical barrier to host shift). Boxes indicate ecological similarities of host plants which parasite species are closely related.(1)(2)(3)(4)(5)				 !"	#		$%&		"Page 11 of 17(page number not for citation purposes)saxifraga [43]. The large-scale congruence detected byParaFit and Icong is thus likely to be due to constraintsSeveral factors may have limited the power of our analyseson cophylogeny. First, our collection was restricted to theBMC Evolutionary Biology 2008, 8:100 common Caryophyllaceae: we have sampled 21 hostspecies of the more than 100 known to be infected withanther smut [17]. Several additional independent fungallineages that may show more evidence of cospeciationthan the ones we have sampled are therefore likely toexist. However, larger sampling is more likely to decreaseand not increase the global congruence between hosts andparasites if host shifts occur between moderately distanthosts as shown for Microbotryum species on Silene. Andindeed larger sampling gave evidence for host shifts in thehighly specific ant-fungus growing parasites (compareconclusions of references [46] and [47]). Our conclusionthat Microbotryum has mainly evolved by host shifts istherefore highly likely to hold with a larger sampling. Sec-ond, the lineages that appear to infect more than one hostspecies could actually be host-specific but too recentlydivergent for our markers to detect differentiation. Thiswould increase the congruence of the Microbotryum andCaryophyllaceae phylogenies, but would not decrease thehigh number of host shifts required to reconcile the twophylogenies. It would thus not affect our conclusionsregarding M. violaceum evolutionary history: the Microbot-ryum-Caryophyllaceae system is another example of ahost-parasite association where cocladogenesis is not therule and highlights the importance of cross species diseasetransmission in the emergence of new parasites lineages[9,11-14].ConclusionAdaptive specialization that follows rare host shift eventsappears to be the major mechanism of speciation inMicrobotryum, as in many pathogenic fungi [4,48,49]. Thefrequency of host shifts in the Microbotryum-Caryophyl-laceae system, and in several other highly specific plant-parasite associations [9,13,14], illustrates that specializa-tion is far from an evolutionary dead end and that thediversification of specialist species can occur by other phe-nomena than cospeciation. In agreement with this idea,most previous cophylogeny studies on parasites able todisperse to novel hosts have reported evidence for hostshifts, even when the number of cospeciation events wasfound significant [47,50]. The possibility of host shift andthe degree of relatedness between ancestral and new hostswill a priori depend on several factors, such as geography,ecology, and the genetics of specialization. Current geo-graphic distribution of host species seems to be of littlerelevance for understanding the association betweenMicrobotryum and its hosts at a local or regional scale. Wedetected some interesting examples of shared ecologicaltraits that may have facilitated host shifts.MethodsTaxon Samplinghosts, we used data from 21 host species from Europe andNorth America for which we have observed Microbotryuminfections in natural populations. All host species ana-lyzed in this study belong to subfamily Caryophylloideae,except one Stellaria species, belonging to the subfamilyAlsinoideae [33]. To root host phylogenetic trees, the Stel-laria sp. was used as an outgroup based on a previousmolecular phylogenetic study [33]. To root anther smutphylogenetic trees, the strains from the North Americanhosts S. caroliniana and S. virginica were used because theywere previously shown to branch at the base of all the spe-cies analyzed in the present study [28]. Because the plantspecies are reasonably well established [31], a single plantsample was collected per host species. Some North Amer-ican Silene species are not monophyletic [51] but theywere not those included in our dataset. Several fungalsamples per host species were collected whenever possiblebecause the parasite taxonomy is currently being resolved(see above). The origins of anther smut and host samplesare given in the Additional files 4 and 5, respectively.Infected plants were detected by the violet, sporulatinganthers of open flowers and flower buds from these sameplants were collected and stored in individual paper orglacine envelopes on silica gel.Molecular MethodsHost plant DNA extraction and PCR amplification ofplant nuclear ITS and cpDNA (the intron within the trnLgene, hereafter trnL, and the intergenic region between thegenes trnL and trnF, hereafter trnLF) were performed asdescribed previously [52]. DNA from the fungal strainswas extracted from the cultures using a Chelex protocol[53]. Fungal DNA extraction and PCR amplification of theβ-tubulin (β-tub), γ-tubulin (γ-tub) and Elongation factor1 α (Ef1α) were amplified according to [29]. PCR frag-ments were purified and sequenced as described previ-ously [29,52].The programs Navigator PPC (Applied Biosystems) andBioedit 6.0.7 [54] were used to check sequence electroph-erograms. Multiple alignments based on consensussequences were carried out using BioEdit. Alignmentswere then checked and apparent alignment errors werecorrected by hand. Regions of ambiguous alignment andgaps were excluded from all analyses.Sequence dataThe sequences generated are available in GenBank (Acces-sion numbers plant ITS are EF407925–EF407945, forplant trnL EF407883–EF407903, for plant trnLFEF407904–EF407924, for fungal β-tub,DQ992076–DQ992113 and EF419304, γ-tub,Page 12 of 17(page number not for citation purposes)To test whether there is a phylogenetic associationbetween anther smut fungi and their CaryophyllaceaeDQ992114–DQ992147, and Ef1α,DQ992148–DQ992177 and EF419301–EF419303). TheBMC Evolutionary Biology 2008, 8:100 sequences generated previously [29] were also used,and are indicated in the Additional file 4.Phylogenetic AnalysisPhylogenetic trees were reconstructed by Bayesian infer-ence, maximum parsimony (MP) and Neighbor-Joining(NJ). MP and NJ analyses were performed using PAUP ver-sion 4.0b10 [55]. The following options were employedfor MP analyses in PAUP: heuristic search, charactersunordered with equal weight, starting tree obtained viastepwise addition option and constructed with randomsequence addition (10 replicates), branch swapping byTBR (tree bisection reconnection). A single MP tree wasrecovered for all datasets. NJ analyses were performedusing the molecular evolution models selected by AIC inModelTest 3.7 [56]. The models retained were TIM+G,gamma shape = 0.41 for the concatenated plant data set,HKY+G, Ti/Tv = 2.2, gamma shape = 1.12 for the fungal β-tub gene, HKY+I, invariable sites = 0.52 for the fungal γ-tub gene, GTR+G, gamma shape = 0.24 for the fungal Ef1αgene and GTR+G, gamma shape = 0.26 for the concate-nated fungal dataset. Bootstrap confidence values werecalculated for 1,000 pseudoreplicates. Bayesian analyseswere run using MrBayes version 3.0b5 [57]. Each run con-sisted of 4 incrementally heated Markov chains run simul-taneously, with heating value set to default (0.2). Priorswere constrained according to the results obtained by run-ning MrModeltest 2.2 [58]. Markov chains were initiatedfrom a random tree and run for increasing numbers ofgenerations, until the average standard deviationremained below 0.01, i.e. 1,000,000 generations for Ef1α,1,250,000 generations for γ-tub, 1,000,000 generationsfor β-tub, 500,000 generations for the fungal and plantconcatenated datasets. Trees were sampled every 50 gener-ations and the first 25% of trees were not taken intoaccount. We used a 50% majority rule consensus tree toobtain the Bayesian posterior probabilities (Bpp). Detailson the phylogenetic parameters used and output statisticsare available upon request. Data matrices and resultingtrees are available in TreeBase (submission ID numberSN3239; Journal Peer Reviewer's PIN number: 30601).We considered nodes as strongly supported by a givenmethod when they had values of Bayesian Posterior Prob-abilities/Maximum Parsimony Bootstraps/Neighbor-Join-ing Bootstraps at least equal to 0.9/70/70, respectively.Monophyly supported by at least two methods was con-sidered as significant.Congruence between individual phylogenies within fungal or plant systemsCongruence between individual phylogenies was esti-mated by Approximately Unbiased tests (AU) as imple-mented in CONSEL [59], by comparing for each gene theeach other gene) [60]. Likelihoods were obtained in PAUPusing the sequence evolution model selected, using AIC,according to the results of ModelTest v. 3.7 [56]. Theincongruence length difference test (ILD, [61]) was notused because several works have underlined that the ILDtest is a poor indicator of data set combinability [62].In absence of significant difference, we further checkedthe congruence of each node by visual inspection. Nodeswere considered as congruent in two gene phylogenieswhen supported by significant statistical values of at leasttwo of the three phylogenetic reconstruction methods ineach of the two phylogenies. Nodes were considered asincongruent between two gene phylogenies when signifi-cant statistical values of at least two of the three phyloge-netic reconstruction methods supported conflicting nodesbetween the two gene phylogenies.When we found no evidence for incongruence, the geneswere concatenated to perform combined analyses, simi-larly as described above. Consistency between the result-ing tree and individual gene trees was again checked byvisual inspection.Identification of fungal phylogenetic speciesTo detect phylogenetic species within M. violaceum, weused the criterion of phylogenetic congruence betweendifferent gene phylogenies [29]. We thus considered agroup of strains as an independent evolutionary lineagewhen 1) it was strongly supported as monophyletic bytwo of the three reconstruction methods in at least onegene phylogeny or in the concatenated phylogeny, and 2)this was not contradicted by the other gene phylogenies.Using three different methods of reconstruction allows usto be conservative in our species delimitation rule, and toavoid splitting the fungus into too many species based onsome artefact of one particular method. We consideredhere again nodes as strongly supported by a given methodwhen they had values of Bayesian Posterior Probabilities/Maximum Parsimony Bootstraps/Neighbor-Joining Boot-straps at least equal to 0.9/70/70, respectively.Comparison of host and fungal treesTo compare the plant and fungal phylogenies, we used thedata derived from the concatenated sequences. As wewanted to assess the impact of species delimitation, weretained successively: 1) one fungal strain per host species,and 2) one fungal strain per fungal species but linking itto all the hosts that this parasite species was found toinfect.For reconciliation analyses (TreeMap [37] and TreeFitter[38]) and Maximum Agreement Subtrees (Icong indexPage 13 of 17(page number not for citation purposes)likelihood of the MP topology obtained for this gene tothe likelihood of the enforced topologies (obtained with[40]), which do not accept polytomies, phylogenetic rela-tionships in our plant tree that were poorly supportedBMC Evolutionary Biology 2008, 8:100 resolved when possible according to previous studies(see symbol * on Figs. 3 and 4). For the remaining unre-solved nodes, the alternative placements were consideredas equally possible. In order to reduce the combinationsof plant and fungal tree comparisons, we used only twotopologies among all possible ones, for each of the plantsand fungi. One topology was chosen as a priori maximiz-ing the congruence with the other partner and the otherone minimizing it (see symbol ¤ on Fig. 3 and 4). For themethods using genetic distances (Mantel tests) or patristicdistances (ParaFit) original datasets were conserved butwe excluded S. acaulis, because of the incongruencebetween individual gene phylogenies, which may haveresulted from hybridization between two distant lineages.The history of association between hosts and fungi wasfirst investigated using reconciliation analysis as imple-mented in the programs TreeMap [37] and TreeFitter [38].TreeMap 1 uses a model to find optimal reconstructionsof the history of the association by maximizing cospecia-tion events [37]. In situations when host shifts are likelyto be common, as in our case [14], this methodology isless likely to find optimal solutions than when host shiftsare rare [63]. A later release of this program, TreeMap 2.0b[64], considers all potentially optimal solutions and offersa more appropriate means of dealing with host shifts.However, this program is currently limited in size andcomplexity of datasets that can be computed, and wecould not run the complete analysis with our dataset.Thus, in this study we used the program TreeMap 1 withthe heuristic search option to reconcile plant and fungaltrees. The program TreeFitter 1.0 uses a different algo-rithm and optimality criterion for reconciling the hostand fungal phylogenies, allowing reconstructions involv-ing many host shifts to be recovered. In both TreeMap andTreeFitter, one can test the null hypothesis that the twophylogenies are randomly related by comparing the scoresof optimal reconstructions (i.e. the number of cospecia-tion events for TreeMap and global cost for TreeFitter)with those of randomly obtained phylogenies throughpermutational procedures. We chose to randomize theparasite trees because in cophylogeny analyses, the hosttree is considered as given, and one wants to test whetherthe observed parasite tree is more congruent with the hosttree than are random parasite trees. Tests were performedbased on 3000 permutations for TreeMap and 10,000 per-mutations for TreeFitter. TreeFitter allows assignment ofdifferent costs to four types of events (i.e. cospeciation,duplication, extinction, and host shift). When using theoption seeking for costs with the highest likelihood, wefound a very large range of possible costs. We varied thesecosts to assess effects on the test results, and retained oneset of costs maximizing the likelihood of cospeciationAs a third approach we used the method ParaFit [39] thatuses matrices of principal coordinates, derived either frompatristic distances (summed branch lengths along a phyl-ogenetic tree; in that case, S. acaulis was excluded) orgenetic distances, and the matrix of presence/absence ofhost parasite associations. Using patristic distances allowstaking into account the most likely phylogenetic relation-ships in addition to genetic distances. We performed bothanalyses. The trace statistic is calculated by taking plant-fungus associations into account. The null hypothesis thatthe plant and fungal samples are randomly associated istested by a permutational procedure. Patristic distanceswere calculated from the unresolved trees derived fromthe concatenated datasets using PATRISTIC [65,66]. Phyl-ogenetic distances were calculated using the softwareMEGA [67]. Principal coordinates were then computedusing the software DistPCoA [68].The fourth method assessed whether the plant and fungaltopologies were more similar than expected by chanceusing the Icong index [40]. With this method, the topo-logical congruence of two trees is assessed through theirmaximum agreement subtree (MAST). A MAST is the larg-est possible tree compatible with two given trees [69] andis obtained by removing the minimum number of leaves(i.e. terminal branches) in both trees in order to obtainperfect congruence. Significant congruence is inferredwhen congruence between the two trees is higher thanthat of random trees with the same leaf number. As for theprevious methods, we reduced the number of topologiestested by choosing two topologies for the plant tree andtwo for the fungal tree. As this method requires that thetrees had the same number of leaves, hosts harbouringseveral fungal species were duplicated. When using a sin-gle strain per fungal species, we also had to select a singleone of the multiple hosts. The choice was made so that thecongruence with the fungal tree was maximized or mini-mized. We assessed the level of congruence for the fourresulting combinations. Finally, we correlated genetic dis-tances of host plants with the distances between associ-ated Microbotryum species. Genetic distances werecomputed as for ParaFit. As the data were not independ-ent, we tested the matrix correlation using permutations(Mantel test in the software Genepop [70,71]).Pollinator and ecological dataData on the common pollinators of the Caryophyllaceaespecies included in this study were collected from the lit-erature [27,72-77] and from personal observations andcommunications (J.A. Shykoff, A. Erhardt). Some individ-ual species, e.g. Hadena bicruris, were reported as pollina-tors, but most descriptions were made at highertaxonomic levels (e.g. genus or family) or common namePage 14 of 17(page number not for citation purposes)events inferred, and the other one minimizing the totalnumber of events.(e.g. moths or butterflies). This raises the concern thatplants pollinated by the same genus or family may not beBMC Evolutionary Biology 2008, 8:100 by exactly the same species or the same individu-als. For well studied plant species, however, only a few dif-ferent pollinators are reported and these pollinatorspecies are rather unspecialized, being found pollinatingseveral plant genera or families [27,72,74]. For instance,Macroglossum stellatarum (Sphingidae), Hadena bicruris,and Autographa gamma (Noctuidae) were all found on atleast three different host species [72]. Moreover, otherobservations showed that individual pollinators land suc-cessively on three sympatric species: Dianthus carthusiano-rum, Silene vulgaris and Silene nutans [27]. We thereforeconsidered that sharing a clade of pollinators impliedsharing at least one pollinator species, thus possiblyallowing M. violaceum spore transmission.To partition the plant species into ecological groups basedon their most common habitats, the French botanical website SOPHY [78], literature data [79], and personal obser-vations and communications (J.A. Shykoff, A. Erhardt, C.Bock) were used. Ecological similarity may lead to cross-species disease transmission even in the absence of sharedinsect pollinators because spores can also be disseminatedby wind, rain and phytophageous insects [27].Authors' contributionsTG, JAS and AW contributed to the conception and designof the study, to the acquisition and analysis of data, tocoordination of the study, and were involved in draftingthe manuscript. GR, FJ, MLG, MEH and RY participated inthe acquisition and analysis of data, and in drafting of themanuscript. All authors read and approved the final man-uscript.Additional materialAcknowledgementsWe thank E. Bucheli, B. Gautschi, M. Baltisberger, I. Till-Bottraud, M. Bar-toli, D. Galeuchet, S. Triki-Teurtroy and U. Carlsson-Graner for providing samples and the Parc National des Pyrénées for the permit to collect. We thank A. Erhardt and C. Bock for ecological information, respectively on pollinators and distribution of plant species. We also thank S. Widmer-Graf, O. Jonot, A. Gautier, O. Cudelou and B. Faivre for help in the lab, A. Gautier, Y. Brygoo and E. Fournier for access to sequencing facilities, and D. M. de Vienne for help in cophylogenetic analyses. Financial support for AW came from the Swiss National Science Foundation (Grant no. 3100AO-104114). TG acknowledges the grants ACI Jeunes Chercheurs (French Ministère de la Recherche), ANR 06-BLAN-0201 and ANR 07-BDIV-003, MEH a grant DEB 0346832 from the National Science Founda-tion and GR a post doctoral grant from the Fondation des Treilles.ReferencesAdditional file 1Bayesian 50% majority-rule consensus tree of the Microbotryum strains analyzed in this study based on the γ-tub gene. Statistical sup-ports indicate Bayesian Posterior Probabilities (BPP)/Maximum Parsi-mony Bootstraps/Neighbor-Joining Bootstraps. Only nodes supported by more than two methods are indicated, the significant statistical supports being considered as higher than respectively 0.9/70/70. The tree is rooted based on previous studies (see text). Taxa labels correspond to the host plant on which fungal strains were collected. Clades not supported in the individual tree are indicated in grey.Click here for file[]Additional file 2Bayesian 50% majority-rule consensus tree of the Microbotryum strains analyzed in this study based on the Ef1α gene. Statistical sup-ports indicate Bayesian Posterior Probabilities (BPP)/Maximum Parsi-mony Bootstraps/Neighbor-Joining Bootstraps. Only nodes supported by more than two methods are indicated, the significant statistical supports being considered as higher than respectively 0.9/70/70. The tree is rooted based on previous studies (see text). Taxa labels correspond to the host plant on which fungal strains were collected. Clades not supported in the individual tree are indicated in grey.Click here for file[]Additional file 3Bayesian 50% majority-rule consensus tree of the Microbotryum strains analyzed in this study based on the β-tub gene. Statistical sup-ports indicate Bayesian Posterior Probabilities (BPP)/Maximum Parsi-mony Bootstraps/Neighbor-Joining Bootstraps. Only nodes supported by more than two methods are indicated, the significant statistical supports being considered as higher than respectively 0.9/70/70. The tree is rooted based on previous studies (see text). Taxa labels correspond to the host plant on which fungal strains were collected. Clades not supported in the individual tree are indicated in grey.Click here for file[]Additional file 4Host species, name and sampling localities of the Microbotryum smut fungi analysed in this study.Click here for file[]Additional file 5Sampling localities of plant samples analysed in this study.Click here for file[]Page 15 of 17(page number not for citation purposes)1. Timms R, Read AF: What makes a specialist special?  TREE 1999,14(9):333-334.BMC Evolutionary Biology 2008, 8:100 van Tienderen PH: Evolution of generalists and specialists inspatially heterogeneous environments.  Evolution 1991,45(6):1317-1331.3. Kawecki TJ: Red Queen meets Santa Rosalia: arms races andthe evolution of host specialization in organisms with para-sitic lifestyles.  Am Nat 1998, 152:635-651.4. Giraud T: Selection against migrant pathogens: The immi-grant inviability barrier in pathogens.  Heredity 2006,93:316-318.5. Hafner MS, Page RDM: Molecular phylogenies and host-parasitecospeciation: gophers and lice as a model system.  Philos TransR Soc London 1995, 349:77-83.6. Downie DA, Gullan PJ: Phylogenetic congruence of mealybugsand their primary endosymbionts .  J Evol Biol 2005, 18:315-324.7. Kawakita A, Takimura A, Terachi T, Sota T, Kato M: Cospeciationanalysis of an obligate pollination mutualism: Have Glochid-ion trees (Euphorbiaceae) and pollinating Epicephala moths(Gracillariidae) diversified in parallel?  Evolution 2004,58(10):2201-2214.8. Johnson KP, Adams RJ, Page RDM, Clayton DH: When do parasitesfail to speciate in response to host speciation?  Systematic Biol-ogy 2003, 52(1):37-47.9. Roy BA: Patterns of association between crucifers and theirflower-mimic pathogens: host-jumps are more commonthan coevolution or cospeciation.  Evolution 2001, 55(1):41-53.10. Six DL: Phylogenetic comparison of ascomycete mycangialfungi and Dendroctonus bark beetles (Coleoptera: Scolyti-dae).  Ann Entomol Soc Am 1999, 92:159-166.11. Lopez-Vaamonde C, Charles H, Godfray J, Cook J: Evolutionarydynamics of host-plant use in a genus of leaf-mining moths .Evolution 2003, 57:1804-1821.12. Banks JC, Palma RL, Paterson AM: Cophylogenetic relationshipsbetween penguins and their chewing lice.  J Evol Biol 2006,19:156-166.13. Holst-Jensen A, Kohn LM, Jakobsen KS, Schumacher T: MolecularPhylogeny and evolution of Monilia (sclrotiniaceae) based oncoding and noncoding rDNA sequences.  Am J Bot 1997,84:686-701.14. Jackson AP: A reconciliation analysis of host switching in plant-fungal symbioses.  Evolution 2004, 58:1909–1923.15. Perlman SJ, Jaenike J: Infection success in novel hosts: an exper-imental and phylogenetic study of Drosophila-parasitic nem-atodes.  Evolution 2003, 57:544-557.16. de Vienne DM, Giraud T, Shykoff JA: Can host and parasite phyl-ogenies be congruent without cospeciation? A simulationapproach.  J Evol Biol 2007, 20:1428-1438.17. Thrall PH, Biere A, Antonovics J: Plant-life history and diseasesusceptibility - the occurrence of Ustilago violacea on differ-ent species within the Caryophyllaceae.  J Ecol 1993,81:489-498.18. Antonovics J, Hood M, Partain J, Heuhsen AM: The ecology andgenetics of a host-shift: Microbotryum as a model system.Am Nat 2002, 160:S40-S53.19. Antonovics J, Thrall PH, Jarosz AM, Stratton D: Ecological geneticsof metapopulations: the Silene-Ustilago plant-pathogen sys-tem.  In Ecological Genetics Edited by: Real LA. Princeton, New Jersey, Princeton University Press; 1994:146-170. 20. Garber ED, Ruddat M: Transmission Genetics of Microbotryumviolaceum (Ustilago violacea): a case study.  Adv Appl Microbiol2002, 51:107-127.21. Kneip H: Untersuchungen uber den Antherenbrand (Ustilagoviolacea Pers.): Ein Beitrag zum Sexualitätproblem.  ZeitschrtBot 1919, 11:275-284.22. Vanky K: The genus Microbotryum (smut fungi).  Mycotaxon1998, LXVII:33-60.23. Baker HG: Infection of species of Melandrium by Ustilago vio-lacea (Pers.) Fuckel and the transmission of the resultant dis-ease.  Annals of Botany 1947, 11(43):333-348.24. Liro IJ: Die Ustilagineen Finnlands.  Volume Vol. 1. Helsinki ; 1924. 25. Perlin MH: Pathovars or formae speciales of Microbotryumviolaceum differ in electrophoretic karyotype.  Int J Plant Sci1996, 157(4):447-452.26. Giraud T, Fournier E, Vautrin D, Solignac M, Shykoff JA: Isolation of44 polymorphic microsatellite loci in three host races of the27. Shykoff JA, Meyhöfer A, Bucheli E: Genetic isolation among hostraces of the anther smut fungus Microbotryum violaceum onthree host plant species.  Int J Plant Sci 1999, 160(5):907-916.28. Kemler M, Goker M, Oberwinkler F, Begerow D: Implications ofmolecular characters for the phylogeny of the Microbotry-aceae (Basidiomycota : Urediniomycetes).  BMC Evol Biol 2006,6:35.29. Le Gac M, Hood ME, Fournier E, Giraud T: Phylogenetic evidenceof host-specific cryptic species in the anther smut fungus.Evolution 2007, 61(1):15-26.30. Le Gac M, Hood ME, Giraud T: Evolution of reproductive isola-tion within a parasitic fungal species complex.  Evolution 2007,61(7):1781-1787.31. Bittrich V: Caryophyllaceae.  In The families and genera of vascularplants Volume 2, Magnoliid, Hamamelid, and Caryophyllid families. Editedby: Kubitzki K, Rohwer J, Bittrich V. Berlin, Germany , Springer Ver-lag; 1993:206-236. 32. Desfeux C, Maurice S, Henry JP, Lejeune B, Gouyon PH: Evolutionof reproductive systems in the genus Silene.  Proc R Soc Lond B1996, 263:409-414.33. Fior S, Karis PO, Casazza G, Minuto L, Sala F: Molecular phylogenyof the Caryophyllaceae (Caryophyllales) inferred from chlo-roplast matK and nuclear rDNA ITS sequences.  Am J Bot 2006,93(3):399-411.34. Popp M, Oxelman B: Evolution of a RNA polymerase gene fam-ily in Silene (Caryophyllaceae) - Incomplete concerted evo-lution and topological congruence among paralogues.  SystBiol 2004, 53(6):914-932.35. Jordan GJ, Macphail MK: A Middle-Late Eocene inflorescence ofCaryophyllaceae from Tasmania, Australia.  Am J Bot 2003,90(5):761-768.36. Oxelman B, Lidén M, Rabeler RK, Popp M: A revised classificationof the tribe Sileneae (Caryophyllaceae).  Nord J Bot 2001,20:743-748.37. Page RDM: Parallel phylogenies: reconstructing the history ofhost-parasite assemblages.  Cladistics 1994, 10:155-173.38. Ronquist F: Reconstructing the history of host-parasite associ-ations using generalised parsimony.  Cladistics 1995, 11:73-89.39. Legendre P, Desdevises Y, Bazin E: A statistical test for host-par-asite coevolution.  Syst Biol 2002, 51:217-234.40. de Vienne DM, Giraud T, Martin OC: A congruence index fortesting topological similarity between trees.  Bioinformatics2007, 23:3119-3124.41. Veevers JJ: Gondwanaland from 650-500 Ma assembly through320 Ma merger in Pangea to 185-100 Ma breakup: supercon-tinental tectonics via stratigraphy and radiometric dating.Earth-Science Reviews 2004, 68(1-2):1-132.42. Hood ME, Antonovics J, Heishman H: Karyotypic similarity iden-tifies multiple host-shifts of a pathogenic fungus in naturalpopulations.  Infect Genet Evol 2003, 2:167-172.43. Lopez-Villavicencio M, Enjalbert J, Hood ME, Shykoff JA, Raquin C,Giraud T: The anther smut disease on Gypsophila repens: acase of parasite sub-optimal performance following a recenthost shift?  J Evol Biol 2005, 18:1293-1303.44. Hirose S, Tanda S, Kiss L, Grigaluinaite B, Havrylenko M, TakamatsuS: Molecular phylogeny and evolution of the maple powderymildiew (Sawadaea, Erysiphaceae) inferred from nuclearrDNA sequences.  Mycol Res 2005, 109(8):912-922.45. Lepage BA, Currah RS, Stockey RA: The fossil fungi of the Princ-eton chert.  Int J Plant Sci 1994, 155:828-836.46. Currie CR, Wong B, Stuart AE, Schultz TR, Rehner SA, Mueller UG,Sung GH, Spatafora JW, Straus NA: Ancient tripartite coevolu-tion in the attine ant-microbe symbiosis.  Science 2003,299(5605):386-388.47. Gerardo NM, Mueller UG, Currie CR: Complex host-pathogencoevolution in the Apterostigma fungus-growing ant-microbe symbiosis.  BMC Evol Biol 2006, 6:88.48. Giraud T, Refregier G, de Vienne DM, Le Gac M, Hood ME: Specia-tion in fungi .  Fung Genet Biol 2008, in press:.49. Giraud T, Villaréal LMMA, Austerlitz F, Le Gac M, Lavigne C: Impor-tance of the life cycle in host race formation and sympatricspeciation in parasites.  Phytopathology 2006, 96:280-287.50. Huyse T, Volckaert AM: Comparing host and parasite phyloge-nies: Gyrodactylus flatworms jumping from Goby to Goby.Page 16 of 17(page number not for citation purposes)phytopathogenic fungus Microbotryum violaceum.  Mol EcolNotes 2002, 2(2):142-146.Syst Biol 2005, 54:710-718.Publish with BioMed Central   and  every scientist can read your work free of charge"BioMed Central will be the most significant development for disseminating the results of biomedical research in our lifetime."Sir Paul Nurse, Cancer Research UKYour research papers will be:available free of charge to the entire biomedical communitypeer reviewed and published immediately upon acceptancecited in PubMed and archived on PubMed Central BMC Evolutionary Biology 2008, 8:100 Popp M, Oxelman B: Origin and evolution of North Americanpolyploid Silene (Caryophyllaceae).  Am J Bot 2007, 94:330-349.52. Widmer A, Baltisberger M: Molecular evidence for allopolyploidspeciation and a single origin of the narrow endemic Drabaladina (Brassicaceae).  Am J Bot 1999, 86:1282-1289.53. Bucheli E, Gautschi B, Shykoff JA: Differences in population struc-ture of the anther smut fungus Microbotryum violaceum ontwo closely related host species, Silene latifolia and S. dioica.Mol Ecol 2001, 10:285-294.54. XX : Bioedit Program.   [].55. Swofford DL: PAUP*. Phylogenetic Analysis Using Parsimony(*and Other Methods). Version 4.   Edited by: . Sunderland, Mas-sachusetts , Sinauer Associates; 2003. 56. Posada D, Crandall KA: MODELTEST: testing the model ofDNA substitution.  Bioinformatics 1998, 14:817-818.57. Ronquist F, Huelsenbeck JP: MrBayes 3: Bayesian phylogeneticinference under mixed models.  Bioinformatics 2003,19:1572-1574.58. Nylander JAA: MrModeltest v2.   Program distributed by theauthor, Evolutionary Biology Centre, Uppsala University; 2004. 59. Shimodaira H, Hasegawa M: CONSEL: for assessing the confi-dence of phylogenetic tree selection.  Bioinformatics 2001,17:1572-1574.60. Shimodaira H: An approximately unbiased test of  phyloge-netic tree selection.  Syst Biol 2002, 51:492-508.61. Farris JS, Källersjö M, Kluge AG, Bult C: Testing significance ofincongruence.  Cladistics 1995, 10:315-319.62. Barker FK, Lutzoni FM: The utility of the incongruence lengthdifference test.  Systematic Biology 2002, 51(4):625-637.63. Page RDM, Charleston MA: Trees within trees: phylogeny andhistorical associations.  Trends Ecol Evol 1998, 13(9):356-359.64. Charleston MA, Page RDM: TreeMap for Macintosh. Ver. 2.0.2001 [].65. Patristic Program   []66. Fourment M, Gibbs MJ: PATRISTIC: a program for calculatingpatristic distances and graphically comparing the compo-nents of genetic change.  BMC Evol Biol 2006, 6:1.67. Kumar S, Tamura K, Jakobsen IB, Nei M: MEGA2: molecular evo-lutionary genetics analysis software.  Bioinformatics 2001,17:1244-1245.68. DistPCoA Program   []69. Finden CR, Gordon AD: Obtaining Common Pruned Trees.Journal of Classification 1985, 2(2-3):255-276.70. Genepop on the Web   []71. Raymond M, Rousset F: GENEPOP (v. 1.2): population geneticssoftware for exact tests and ecumenicism.  J Hered 1995,86:248-249.72. Kephart S, Reynolds RJ, Rutter MT, Fenster CB, Dudash MR: Polli-nation and seed predation by moths on Silene and alliedCaryophyllaceae: evaluating a model system to study theevolution of mutualisms .  New Phytol 2006, 169:667-680.73. Collin CL, Pennings PS, Rueffler C, Widmer A, Shykoff JA: Naturalennemies and sex: how seed predators and pathogens con-tribute to sex-differential reproductive success in a gynodio-ecious plant.  Oecologia 2002, 131:94-102.74. Bloch D, Werdenberg N, Erhardt A: Pollination crisis in the but-terfly-pollinated wild carnation Dianthus carthusianorum?New Phytol 2006, 169:699-706.75. Barthelmess EL, Richards CM, McCauley DE: Relative effects ofnocturnal vs diurnal pollinators and distance on gene flow insmall Silene alba populations .  New Phytol 2006, 169:689-690.76. Jürgens A, Witt T, Gottberger G: Reproduction and pollinationin central european populations of Silene and Saponaria spe-cies.  Bot Acta 1996, 109:316-324.77. Pettersson MW: Pollination by a guild of fluctuating moth pop-ulations: option for unspecialization in Silene vulgaris.  J Ecol1991, 79:591-604.78. SOPHY   []79. Bonnier G, De Layens G: Flore complète portative de la France,de la Suisse et de la Belgique.  Paris , Belin; 1986. yours — you keep the copyrightSubmit your manuscript here: 17 of 17(page number not for citation purposes)


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items