MALACOLOGIA. 2007, 49(2): 293-311

SPECIES DELIMITATION IN THE GENUS BYTHINELLA(MOLLUSCA: CAENOGASTROPODA: RISSOOIDEA):

A FIRST ATTEMPT COMBINING MOLECULAR AND MORPHOMETRICAL DATA

Jean-Michel Bichain'", Marie-Catherine Boisselier-Dubayle",Philippe Bouchet ' & Sarah Samadi'

ABSTRACT

Within the springsnail genus Bythinella, few dlscontlnuous morphological characters al­low to unambiguously delineate species-Ievel taxa . Opinions on the alpha-taxonomy ofthe group are divergent, with sorne authors recognizing every morphologically distinctlocal form as a species, whileethers interpreting such geographical forms as intraspecificvariation. Because the value of morphological characters was rarely contrasted with mo­leculardata, suchopinions remained untested. In thiswork, variation betweenpopulationswas studied through genetics isozymes , phylogeny (DNA), and morphometrics. Elevenpopulations represenling rive putative specles were sampled from the french Pyrenees, aregion where a high number of nominal species are classically recognized. Based ongenetic and phylogenetic analyses, the material clusters lnto three groups, of which oneconsists of several nominal species. Environmental factors , sexual dimorphism, and ge­netic factors contribute ta the significant morphological variation observed within the ge­neticgroups. Thus, the numberof species ofBythinella recognized inthe Pyreneesappearsprobably overestimated, andthecharacters traditionally usedfor species delimitation shouldbe re-evaluated.

Keywords: hydrobioid springsnails, ITS-1 nucleargene, isozyme,phylogeny, morphom­etry, multivariate analyses.

INTRODUCTION

Among gastropods, hydrobioid springsnailsare one of those taxa in which the lack of ob­viousdiscriminating morphological charactersmakes species delimitation diffieult, resultinghistorically in the establishment of nurnerousnominal species and currently in a variety ofspecialist opinion on the taxonomie validity ofthese taxa (Kabat & Hershler, 1993; Herschler& Ponder, 1998; Wilke et al., 2001; Szarowska& Wilke, 2004). A consequence is that differ­ent speeialists have differentevaluationsof themagnitude of biodiversity. The difficulty of re­solving sueh differing opinions is also ham­pered by the lack of a solid theoretical andmethodological framework, de facto render­ing many taxonomie opinions untestable hy­potheses. However, and although delimitating

species boundaries is a central aim of aipha­taxonomy, there ls no consensus coneerningthe meaning of the term "species" (as in­stances of the extensive literatureon thisques­tion , see, e.g, Howard & Berlocher, 1998 ;Wilson, 1999; Winston , 1999 : Wheeler &Meier, 2000; Hey, 2001a, b; Mallet, 2001; Noor,2002).

The objective of this paper is ta test the va­lidity of the traditional delimitation of somespecies belonging ta the genus BythinellaMoquin-Tandon , 1856, based upon a speciesconcept that permits testable hypothèses. Forthispurpose, we use the Hennigian inter-nodalspecies concept recently developed and for­malized by Samadi & Barberousse (2006).This species definition is close ta the Evolu­tionary Species Concept (Simpson, 1961 ,modified by Wiley, 1981; Wiley & Mayden ,

'Muséum National d'Histoire Naturelle, Département Systématique et Evolution, Unité Taxonomie et Collections (USM602­UMS2700), Case Postale 51, 55 Rue Buffon, F-75005 Paris CEDEX 05, France

' Systématique. Adaptation et Evolution, UM R 713 8 P6-IRD·MNHN-CNRS (UR IRD 148), Service de systématiquemoléculaire (CNRS, IFR 101 ), Département Systématique et Evolution (USM 603), Muséum National d'Histoire Naturelle ,Ca se Postale 26 , 57 Rue Cuvier. F-752 31 Paris CEDEX 05 , France*Correspondlng author: jean-michel.bichain@educagri.fr

293

294 BICHAIN ET AL.

2000), the Internodal Species Concept(Kornet, 1993) and the General Lineage Spe­cies Concept (De Dueiroz, 1998, 1999) . ForDe Dueiroz (1998), mast "species concepts"are ill-named, because they should rather beregarded as recognition criteria that, associ­ated with a formai definition of the "species"that is independent of the mechanism by whichspecies are recognized, permit delimitation ofspecles in practice . As suggested by theseauthors, the various techniques used in sys­tematics can be set in a coherent frameworkstarting tram species descriptions basad onmorphology and/or behavioural and ecologi­cal attributes . These descriptions are the pri­mary hypotheses for species delimitation thatcorrespond ta alpha-taxonomy. Second, thesehypotheses can be tested against criteria de­rived trom the varions so-called "species con­cepts", such as the Phylogenetic SpeciesConcept (PSC) and/or the Biological SpeciesConcept (BSC), corresponding respectively tathe approaches of phylogenetic systematics(sensu Hennig , 1966) and evolutionary sys­tematics (sensu Mayr, 1942). Finally, the re­sults of these tests permit the selection ofcharactersets corresponding to more robustprimary descriptions. Here we attempt ta de­Iimit sorne Bythinella species, in a restrictedgeographical area, in coherence with thetheory of evolution and ta justify why, as ar­gued for example by Sites & Marshall (2003,2004), we need an integrated approach in tax­onomy.

Bythinella species live in emerging ground­water springs and in the uppermost coursesof small streams, where they can form largepopulations (1,000-1 O,OOOs individuals).Sorne taxa live exclusively in groundwater(Giusti & Pezzoli, 1980; Falniowski, 1987;Bernasconi , 2000; Bertrand , 2004; Bichain etal., 2004). The smail (usually around 2.5 mm ,maximum 4 mm), nondescript shell is charac­terlzed by a blunt apex, with an overall ovoidto conical-elongate shape, most often smooth,but occaslonally with spiral keel(s) or one ormore axial varices on the last adult whorl. Spe­cies of Bythinelfa are gonochoristic. Males arecharacterized anatomically by a penis-flagel­lum complex; females have a bursa copulatrixand a seminal receptacle. France is an areaof high specifie richness for the genus. Of thespecies currently recognized in Europe, 62%occur in France, ofwhich 90% are endemic taFrance (Fauna Europaea, 2004).

ln Bythinelfa , as in the majority of hydrobioid(sensu Davis , 1979) genera , species are pri-

marily delimited based on non-discrete char­acters. A consequence of this continuousvariation of character states is a range of taxo­nomic opinions among the various taxonomieauthorities . For example, in Italy, based onshell morphology and reproductive anatomy,Alzona (1971) recognized elght species ofBythinelfa, whereas Giusti & Pezzoli (1980)regarded ail Italian populations as belongingto two species. Based on shell characters,Radoman (1976) recognized about 12 speciesin the Balkans and Asia Minor. Based on shellmorphology, soft part anatomy, and geographi­cal distributions, Falkner et al. (2002) recog­nized 42 valid species in France, with severalapplications of species names differing fromthose of Bernasconi (2000). Thirteen of the sespecies were restricted ta the Pyrenees andtheirfoothills . However, in Poland analyses ofenzyme polymorphism and morphometricstudies performed on shell and anatomicalcharacters (Falniowski et al. , 1998, 1999;Mazan, 2000 ; Mazan & Szarowska, 2000a , b)questioned the ability of the shell and anatomi­cal characters used in alpha-taxonomy ta de­lineate species of Bythinelfa . Consequently,because of these various opinions, whieh rec­ognize these taxonomie entities alternativelyas "good species" or not, we have no clearidea of the true diversity of this group.

Because oftheir narrow ranges and speeial­ized habitat, hydrobioid springsnails in gen­eral, and Bythinelfa species in particular, arevulnerable ta even small-scale habitat trans­formation , such as trampllng by cattle or artifi­cial diversion of springs. Several hydrobiidspecies are considered extinct , and many oth­ers are categorized as threatened (IUCN 2005,http://www.redllst.org). Of the species occur­ring in France, nine are nationally protected(statutory order October 7,1992). Legal pro­tection was bestowed upon them based on thestate of taxonomie knowledge in the 1980s,and it is now fitting ta evaluate whether thenames involved in the legal texts apply ta realunits of biodiversity. Of the putative speciesanalyzed in the present study, two are nation­ally protected .

Speeies delimitation in a number of nominalBytlJinella speeies from the Pyrenees wasexplored according ta criteria derived from thePSC and the BSC , and our purpose was tatest if sets of morphological characters (tradi­tian al or newly developed) correspond ta thesecriteria. For this purpose, we used populationsrepresenting nominal speeies sampled fromtheir type loealities or from the nearest pos-

SPECI ES DELIMITATION IN THE GENUS BYTHINELLA 295

sible area if the preci se type locality could notbe foun d or no longer exis led . We performedrnult lvar tate analyses on shel l parameters toevaluate whether the pattern of she ll variationwithin and amo ng populat ions permit s non­ambiguously att ributing each of the sampledpopulation ta a putative spe cies, and we testedspecies delimitation resulting tram Iraditionalmorphological approaches aga inst BSC andPSC criteria. The BSC crite rion am ounts tadelimitation of species as reproductively iso­lated groups that became genetically isolaledfrom other such groups. Isolation was evalu­ated indirecUy through analysis of gene flowamongpopulations usingelectrophoretic vari­ability of isozymes. In the mast comman viewof the PSC , spec ies are monophyletic groups.Monop hyly of groups was identi fied using thepattern of variab ility of the ITS-1 nuclea r gene .These three approaches were compared inarder ta evaluate the potential use of lhe mor­phological characters in the de limitation ofBythinella species .

MATERIAL AN D METHODS

Taxa and Populations Studied

The taxa selected for the prese nt study es­sentia ily orig inate from the Dé par tementAriège , in the cen tral/northeastern Py renees.The area has extensive karsts , with numer­DUS springs where BythinelJa species live indense populations. Based on shell and repro­ductive anatomy, Bernasconi (2000) had pro­posed an alph a-taxonomy of Byth inella fromthisarea, andwe followed his treatment in howwe applied specifie names ta each of the popu ­lations sampled .

Eleven populations were sampled that, basedon Bernasconi (2000) and Falkner et al. (2002),correspond ta five speci es (Table 1, Fig. 1).Ten of these populations, representing fourputative species, are from the Pyrenean foot­hills, but B. reyniesil (Dupuy, 1851) is a high erailitude species . Bythinella simoniana (Moquin­Tandon , 1856), a species relatively weil char­acterized by one or more axial varices on thebody wharf , was sampled from lhree popula­tions (Eng , Ca l and Sou). Fou r populal ions(Td11, Td12, Suz and Roq) are attributable taB. utricufus (Pa ladilhe, 1874) .

At Aud inac, we sam pled Bythinella from atherm al spring (ca . 18-20'C; Aud 1), from acoId sp ring (ca . 13' C; Aud2) just 30 metersaway, and from the co nfluence (Aud3) of the

two springs. The the rmal spring is the typelocalil y of B. rubiginosa (Boub èe, 1833), anominal speci es considered ta be restrictedta this locality; the applica tion of the speciesnam e is thus unam biguous. Th e populationsampled fro m Aud2 had been identified byBernasconi (2000) as B. eurystoma (Paladilhe ,1870). However, specimens do not match theoriginal descriptionofthis nominal speciesweil(type locality is St Jean-de-Fos, DépartementHé raull ); we co nsequently do not foll owBernasconi in app lying this name ta that popu­lation, which we will from herson des ignateas B. cf . eurystoma. The th ird po pu lalion(Aud3) was not inci uded in Bern asconi'sdataset, and we co uld not easi ly place it inone of the nominal species.

As a geographical out-group within Bytflinella,we used B. vir/dis (Po iret. 1801). the type spe­cies of the genus, whic h we samp led from thetype localily (Che) in the east of the Paris ba­sin .1l has an ovoid shell that cleariy sets it apartfrom ail the Pyrenean Bythinella species in­cluded in our study.

Specimenswere collected by washing smallpebbles, aquatic vegeta lio n and dead leavesove r Iwo sieves (2 mm and 450 pm mesh).ForanatomicaJand biometrica! studies, speci­mens were fixed in 70% ethanol; for molecu­lar studies, specimens we re Frazen alive al-80' C. Given the very sma ll size of the spec i­mens, each of the different analyse s (morpho­metric , isozyme and DNA seq uence analyses)could notbe carried out on the same individu­ais.

Isozyme Electrophoresis

Protein extraction was carried out on wholeanim aIs (including the shell), and followed theprotocol of Boissel ier-Du bayle & Gofas (1999) .El ectrophor esis was done usin g ve rtica lacryla mide gels on discontinuou s systems .The running buffe r was Iris-glycine (4.95 mM ,pH 8.3). Of lhe eight enzyme systems as­sayed, thre e gave scorable band ing patternsand are used in this stud y (pho spho glucomu­tase : Pgm, EC 5.4.2.2; glucose-6-phosph ateisomerase : Gpi, EC 5.3.1.9; aspartate ami­notransferase: Aat, EC 2.6.1.1 ). Because oflhe small size of the specimens, il was techni­cally not feas ible ta study ail three enzymesystems on the same individual. We therefo reextracted betwee n 30 and 40 indi viduals pe rpopu lation, giv ing approximately 20 for eac hloci. In talai, 439 individuals were used for theanalyses (Table 1).

'"<D

TABLE 1. Bythinella populations characleristics and number of individuals used for each analysls.Ail localilies were clte d by Bernasconi (2000) . Tdl1Cl

and Tdl2 are populations in two adjacent springs separated by 10 rneters. Both fall within the generalized type locality of Bythinella utricutus. N =number of individuals used for enzymatic activities, N.

I! = number of shells measured. oB

Population information ITS-1 sequencesNlIh and

Type Altilude (in DNA GenBank MNHNNominal species Code Localily Localilies Coordinates meters) N" isolate number number

B. viridis Che yes Chery-Chartreuve (02) 3"37'51"E 130 25 Che1 D0 318901 27Moulin de Veau 49"15'45"N Che3 D0 318902 Moll5959

B. rubiginosa Aud1 yes Audinac- les-Bains (09) 1"10'56"E 450 35 Aud11 D0318905 30thermal spring 43"00'22"N Aud12 D0318906 Mol15960

B. reyniesii Par no Boussenac (09) 1"27'7"E 1250 38 Por1 D0318899 30Col de Port 42"53'58"N Por2 D0 318900 Molt5961 ID

B. simoniana Alas (09) 1°01'23"E 30 CiCat no 510 36 Cat3 D0318903 I

Sainte-Catherine 42"57'19"N Molt5962 ~

Engomer (09) 1"04'17"E 30 zEng no 460 38 No sequenced m

Arguilla 42"56 '53"N Mol15963 -1

SouClermon t (09) 1"16'48"E

300 40 Sou2 D031890430 »

noLa Souleille 43"01'29"N Moll5964 r

B. cf. eurystorna Aud2 no Audinac- les-Bains (09) 1"10'56"E 450 38 Aud23 D0 318907 30no thermal spring 43"00'22"N Aud26 D0318908 Moll5965

Bytl line/la sp. Aud3 no Audinac- les-Bains (09) 1"10'56"E 450 34 Aud32 D0318909 30connuence of AUD1 and AUD2 43"00'22"N Aud31 D0318910 Moll5966

B. utticuïus Suz no La Basti de-de-Sérou (09) 1"25'10"E 410 34 Suz3 D0318914 30Coi de Suzan 43"01'16"N Suz5 D0 318915 Moll596 7

Tdl1La Bastide-de-Sé rou (09) 1"27'16"E

460 37Tdl15 D0318912 30

yesTour de Loly 43"00'56"N Tdl12 D0318913 Moll5968

Tdl2La Bastide-de-Sérou (09) 1"27'16"E

460 42 No sequenced 29yes

Tour de Loly 43"00'56"N Moll5969Roque fort les Cascades (09) 1°4S'44"E 59

Roq no Cascades des Turasses 42"57'22"N460 33 Roq2 D0318911 Moll5970

-

SPECIES DELIMITATION IN THE GENUS BYTHINELLA 297

Two polymorphie loci were interpreted forPgm, only one for Aar and Gpi. Multiple lociencoding the same enzyme were numberedin arder of decreasing mobility. The alleieswere numbered with the same system.

For the four putative loci , data were scoredin a matrix of individual genotypes. For eachpopulation , allele frequencies at each locus ,percentage of polymorphie loci at 0.95 (P),mean number of allèles per locus (A), meanobserved and expected heterozygosities (Hoand He), pairwise theta (Weir & Cockerham,1984) among populations (e) were estimatedusing Genetix version 4.02 (Belkhir et al. ,2001). Departures from Hardy Weinberg equi­iibrium (HWE) were tested in two steps: first aglobal test was performed using the exact HWtest of Haldane (1954), Weir (1990) , and Guo

42·40'

& Thompson (1992): second , heterozygotedeficit or excess was tested with U tests . Ge­notypic iinkage disequilibria were measuredwithin each population and for each pair oflocus, and exact tests of genotypic difleren­tiatlon were performed per locus and for eachpopulation pair.Ail these tests were performedwith Genepop version 3.4 (Raymond &Rousset , 1995). A sequential Bonferroni cor­rection (Holm, 1979; Rice, 1989) was usedwhen several statistical tests were performedsimultaneously (for HWE, heterozygote defi­cit or excess, linkage disequiiibrium).

A dendrogram of the 12 populations wasconstructed to analyse the distribution of ge­netic variation . The chord distance (Cavalli­Sforza & Edwards , 1967) was computedbetween ail pairs of populations and the popu-

2·

FIG. 1. Geographieal location of Bythinella populations sampled. For details of the cantiguous sta­tions trom Audinac (Aud) and La-Bastide-de-Sérou (TdL), see Table 1.

298 BICHAIN ET AL .

Jations we re c1ustered using the neighbour­jo ining (NJ) method . Th e chor d distance as­sumes that allele frequency changes resulttrom genetic drift atone: il does not assumethat population sizes have remained constantand eq ual in size (Felsenstein, 1993). Thisanalysis was con ducted us ing Populationsversion 1.2 .28 (Langella, 200 2).

DN A Techniques and Ph ylog enetic Analyses

The ribosom al interna i tran scribed spacers(IT8-1 and IT8 -2) are intergen ic DN A se ­qu ences that, because of a relatively fasl rateof evolution, are useful in resolving phyloge­netic relation ships of closely related taxa (see ,for exa mple, 8 chilthu ize n et al., 2004, for mol­luscs, or for other zoologica l groups see (Hilliset at., 1996 ; Colema n, 2003; C hen et al.,2004 ). We here use the variability of IT8 -1 thatlies betweenribosomal nuelear genes 185 and5.88 .

DNA was extr acted from whole individualsin cluding sh ell s , usin g the OIAGENDNea sy· 96 ki t. The entire IT8-1 region wa samplified using Iwo universal primers IT8-2 (5'­GCT GCC TIC TIC ATC GAT GC-3') and IT8 ­51 (5'-A GG TGA CCT GCG GAA GGA TCATI-3') (Hilli s & Dixon , 1991).

PCR reactions w ere per formed in a fin alvolume of 50 ~I , us ing approximate ly 5 ng oftem plate ONA, 2.5 mM MgCI ,.0.6 ~M of eachprimer, 0.26 mM of eac h nucleotide, 5% DMSOand 1.5 unit of Taq polym erase (Obioge ne).Amplification products we re generated by aninit ial denatura tion step of 4 min at 94°C fol­lowed by 40 cycles at 94 "C for 30 s,50°C for30 5 and 30 5 at 72·C, and a fina l extens ion at72 °C for 5 min. PCR products were pu rifiedus ing the OI Aqu ic k PC R pu ri fi cat ion k it(Qiagen) and sequenced on a Ceq2000 '" au­tom ated sequ encer (Beckman), in bo th direc­tions ta confirm accuracy of each sequence.We obt ained 17 seq ue nces corresponding toone or Iwo individuals ana lysed for each popu­lation (Table 1), aligned with the ClustalW al­gori thm (Thompson et al., 1994) implementedin the BioE dit 5.0.0 sequence alignment edi­tor softw are packag e (Ha ll, 1999).

The DNA fragment sequenced included sev­eral highly variable ragions with numerous in­sertions/deletions. These variable ragionswere removed from the data se t and manuallyali gned foll owing the proced ure of Barriel(1994) : num ber of inde ls, number of changesand number of modified sites are minimized

in lhisarderof priorities, a transition beingpre­fe rred ta a transversion in the ambiguous ar­eas. Each plausible align ment for each of thehighly va riable regions was lested with PAUp·4 .0b (8 wofford, 20 00). The criterion used tochoose an alignment among the different align­ments tested was to ob tain the most pars imo­nious tree.

After removing the first 5' ambiguo us vari­able region whic h wa s located belween theRNA188 ge ne and the first IT8-1 conservedregion, the final data matrix was composed of17 seq uences and 360 bp. A maximum parsi­mony analysis wa s performed usingPAU P· 4.0b. Gaps we re included as missingdata . The ana lysis was performe d using aheurlstic search wilh 100 random-addi tlon rep­lications, branch swapp ing by the Tree Bisée­tion and Reconn ection (TBR) algorithm . To testthe robustness of the resu lts, the Bremer in­dex (Bremer, 1994) was ca lculated and 100bootstrap replicates were carried out, using thesam e heuristic sea rch settings.

Morphometrica l A nalys is

Shell Measurements - Betwe en 30 and 59ad ult shells were used per population . 8 hellswere washed in a 5% Chlorox solution for 30min, then rinsed in distilled water, transferredto abs olute ethanoi and dried . The she lls wereplaced on an adhesive support in a stand ard­ized position (fo r more detai ls , see the legendof Fig. 2) and then digitized with a graduatedscale using a stereomicroscope associatedwit h a digital camera. An orthonormal frame(Ox, Oy) and the aperture center were over­laid on the digital pic ture, and 12 landmarkswe re then positioned us ing the TpsDig 1.23software package (Rohlf , 2001). The coordi­nates of thes e landmarks wi thin the orth ono r­mal frame (Ox, Oy ) wer e used to take 15measurements characterizing individual shells(Fig . 2).

Sex Determination - 8 exing the an imal re­qui red dissolving the shell for dissection. Afterdigitalization, spe cimens were removed fromthe adhesive support , placed for four min in asolution or 5% hydro chloric acid and 70% etha­nol for shell dis solu tion, and then rinsed in dis­tilled water. Disse ctions were done under adissecting stereomicroscope. Sax was deter­mined by the presence of the flagellum-peniscom plex for ma les and by the presence of thealbumen and capsule glands for females .

-SPE CIES DELIMITATION IN TH E GENUS BYTHINELLA 299

I ~

r I•

Wap2

"'--''-+- - - - -1- -''>.......- 'f--- -1-'-- l-l-t-

910

x

y

oWlw

Wsh

FIG. 2. Morphologieal parameters taken from shell of Bythinella in standard position: the collurnelaraxis and the aperture plane are parallel to the horizontal plane. Left - Landmarks locations. Or~

thonormal frame (Ox, Oy): the Oy axis Is paraliel ta Ihe collumelar axis and tangential ta the left edgeof the shell. The Ox axis is perpendicular to Oy axis and tangenlial to the lower edge of the aperture.The aperture center ls the center of the E ellipse which is adjusted with the aperture edge. Righl­Measurements taken, Shell height (L,,), aperlure length (L,,, =2'[x,-x )' +(y,-y,,),D, aperture heighl(L" ,) , aperture diameler (W,,, =2, [x,,-x,,)' +(y,,-y,,),J, aperture widlh (\lJ"1 =x,-x,,), shell widlh (W",),wldth of lasl whorl (W~ = x, ), widlh 01 penultimale whorl (W.. = ' [x, -x, '+(y,-y l'D, widlh of ante­penultimatewhorl (WlM' = ~[x :fXJ )2+(Y2-YJ)2]). Height measurementsof the spireandof each wholi weretaken on the left and on the riqht sidas of the shell: left and right spire heights (LSP1 =YI-Y2'~2 =YçYs).left and right helghls of penultimate whorl (L.... L..,), left and right heiqhts of last whori (~" ~,) .

Analysis of Sexual Oimorphism - Sexual di­morphism in Bythinella had been reported byFalniowski (1987) but never quanti fied. Weassess ed first how sheli dlmorphism might in­terfere with specie s delimitation wl thin popu­lations "For this. we analyzed sheli parametersin 59 individuals Irom Roq . The input data lnmiilim eters were transformed lnta Log-sh apeRatio (LSR) data in arder ta limit an eventualsize effe ct (Mosiman n, 1970 ). We first per­formed a Principal Component Analysis (PCA)ta explore how the range of sheli param eterswas dis tributed between males and females .Then a Discriminant Function Analysis (OFA)was used ta test stat istically the differe ncesbetween males and fema les and then ta iden-

tily which shen paramelers reflecl sexual dl­morphism. Wil ks' Lambda (WL) used in anANOVA (F) lest of mean differences was usedto test if the discriminant model as a wholewas slgnifica nt. Then , if the global F test wassignificant, each variable was tested usingWilks' lambda to determine which variable dif­fered significantly in mean between discrimi­nated groups.

Second , we evaluated the effecl of sexualdimorphism among population s, taxonomiegroups and genetic groups. For this purpose,an analysis of variance (ANOVA) with two fixedfactors (population and sex) was carried outon the parameters reveaied by the DFA as sig­nificantly contributing to shell dimorphism in

300 BICHAIN ET AL.

the population analysed. This analysis permit­ted comparison of the relat ive elfects of sexand population location on shell shape, andidentification of the interaction between thesetwo factors. Tukey's HSD post hoc test and itsassociated probability were used ta examinethe statistical significance of the differencesbetween aI l pairs of means.

Analysis of Shell Variation Between Popu­lations and Taxa - DFAs and assoclated Wilks'Lambda were performed on ail 15 measure­rnents afte r LSR transformation, and two dis­criminant factors we re tested: (i) the Taxondiscriminati ng factor: popu lations referred apriori to a nominal species based on the col­lection site were tested a posteriori ta checkwhether they couId be discriminated based onshell parameters (popu lation Aud3 was ex­ciudad tram the analysis because it was notassigned ta a nominal taxon); (ii) the Geneticdiscriminating factor: clusters of populationsresulting lrom the genet ic and phylogeneticanalyses were tested a posteriori ta checkwhether specimens grouplng together could

,--- - - - Che 1il. viridis -

also be discriminated based on thelr shell. Wethen comparedthereciassification scores of thediscriminating linearfunctions of the twogroup­ing methods (i.e., taxon or genetic groups).

Ali analyses were performed with the soft­ware package STATISTI CA 6.0 (Statsoft , 2001)and the level of significance use d was 5%.

RESULTS

Isozymes Polymorphism

The three enzyme systems involving lourputative loci revealed Irom three ta five allelesper locus (Table 2), and the mean number ofalleles per population ranged trom 1.0 ta 2.25(Table 3). The Aar-t locus was the least poiy­morphlc, with a total 01three alleles overall andwith each population being monomorphlc. Theone population (taxon) used as the out-group(Bythinella viridis) was monomorphic for ail fourloci, as was the Sou population. For the otherpopulations , the mean observed heterozygos­ity (Ho) ranged lrom 0.015 ta 0.241 . Genotype

Chc!-- - - - - - - - - - - ­Chc3- - - - - - - - - - - - -

I-- - - - Po, 1B. reynies iiPor~

100

15

Tdll Tdll .1

160

ruu.zrau B. mr iculusRoq 1

95 1--60 3

Roq SU,,~64

SU l. Suz5

Ion. . . . . . _-_._-- --- -_._-- -_.....

100Cut

cm3~ 12

Eng B. stmonùnm Soul_ 0._' _ 70

Audl .3 1

54Sou

Aud2.6 1 77

, AUd2 1B. cf. eurystoma A "dll~1

35Aud!.:!"t,dl 1B. rubiginosa Aud3.1

74 Alld] . Bythinclia sp . Aud3.2. .. .. . - .. .. . . . . __. .. -

FIG. 3. NJ tree constructed trom isozyme data on the 0 chard distance (Cavalli-Sforza & Edwards,1967) between populations of Bythinella (left) with the Bootstrap values and parsimony tree (strictconsensus) computed from ITS-1 sequences(right)with a heuristic search (Boatstrap values above,Bremer index below).

SPECIES DELIMITATION IN THE GENUS BYTHINELLA 301

TABLE 2 . Allele frequencies observ ed at 4 loci on 12 populations of Bythinella. N ;; number of indl-viduals ana lysed for each locus.

Aud1 Aud2 Aud3 Cat Eng Sou Tdl1 Tdl2 Roq Suz Par Che

Pgm-I (N) (24) (24) (23) (24) (22) (28 ) (25) (25) (24) (24) (22) (7)1 0.023 1.000

2 0.417 0.458 0.674 0.250 0.068 0.140 0.100 0.063

3 0.563 0.458 0.326 0.688 0.886 1.000 0.340 0.300 0.854 0.208

4 0.021 0.083 0.063 0.023 0.520 0.600 0.146 0.729

5 1.000

pgm-2 (N) (24) (24) (18) (24) (18) (28) (21) (20) (19) (19) (17) (9)1 1.000 1.000 1.000 0.646 0.889 1.000 0.262 0.225 0.029

2 0.354 0.111 0.738 0.775 1.000 1.000

3 0.9714 1.000

Gpi- l (N) (24) (25) (22) (27) (22) (30) (27) (30) (26) (26) (24) (19)1 1.000 1.000

2 0.958 1.000 0.955 0.981 0.932 1.000 1.000 1.000 1.000 1.000

3 0.042 0.045 0.019 0.068

Aat-I (N) (9) (8) (12) (12) (17) (16) (10) (14) (1) (1) (14) (10)1 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000

2 1.000 1.000

TABLE 3. Genetic variability computed on the four analysed loci on Bythinel/a populations. Niot ;;total number of individuals used for isozyme actlvlttes: n ;;mean number of individuals analysed perlocus; P ;; percentage of polymorphie loci; A ;; mean number of allele per locus; Ho ;; observedheterozygosity; He ;; expected helero zygosily; HWE ;; Hardy-Weinberg equilibrium tests (Fischer'smelhod for test of significance. · significant at 0.05 level).

Population Nlo t n P A Ho He HWE

Aud1 35 20.3 0.5 1.75 0.104 0.150 0.22Aud2 38 20.3 0.25 1.5 0.146 0.146 1

Aud3 34 18.8 0.5 1.5 0.164 0.135 0.712

Cal 36 21.8 0.75 2.0 0.134 0.244 0.003"

Eng 38 19.8 0.75 2.25 0.124 0.137 0.146

Sou 40 25.5 1.0

Tdl1 37 20.8 0.5 1.75 0.241 0.251 0.179

Tdl2 42 22.3 0.5 1.75 0.162 0.227 0.158

Roq 33 17.5 0.25 1.25 0.052 0.064 0.398

Pa r 38 19.3 0.25 1.25 0.015 0.015 1

Suz 34 17.5 0.25 1.5 0.073 0.107 0.089

Che 25 11.3 1.0

302 BICHAIN ET AL.

TABLE 4. F-statistics indices (Weir &Cockerham,1984) calculated for the Bythinefla populationsanalysed (excluding out-group and Par popula­tions. see text for exp ianations ).

Locus F,s FST FIT

Pgm-l 0.209 0.332 0.472Pgm-2 0.179 0.718 0.769Gpi-l -0.033 0.020 -0.012afl 0.189 0.481 0.579

frequencies generally matched Hardy Weinbergexpectationsaftar a sequenUal Bonferroni cor­rection, except for the Cat population,which hada heterozygosity deficil. No genotypic linkagedisequilibrium between pairs of lociwas round.

These four lociwere together diagnostic fartwo of the six presumptive taxa: (i) individuelsfrom the type-Iocality of B. viridis had diag­nostic alle les for the two loci Pgm-l and Pgm­2; and (ii) individuals from Por, attributed to B.reyn iesii, had a diagnostic allele for Pgm-2.Furthermore, individuals attributed to B. viridisand B. reyniesii together differed from ail otherByth inefla specimens by the same diagnosticalleles at the two loci Gpi- l and Aat -1.Ai l otherpopulations sampled from the Pyrenean foot­hills shared alleles with different frequ enciesat ail the loci scored. Thus, gene flow betweenany ofthese three groups and any other popu­lation from our sample can be excluded ,whereas gene flow among populations framIhe Pyrenean foothills cannot be excluded.

The NJ tree cons lructe d on the chord dis­tance illustrates Ihe genetic relationships be­tw een po pu lations (F ig . 3, left ). Tw opopulations, B. viridis and the Por population(attributed to B. reyniest î; were highly diver­gent from ail others . For other populations, thetopology of the tree revealed two main geneticgroups. First, the four popul ations Td11 , Td12,Suz, and Roq (attributed to B. utricu/us) elus­tered together, with values of 0 ranging from0.001 (Td11 vs. Td12) to 0.1 (Td11 vs. Roq).Second, the populations Aud1 (B. rubiginosa),Aud2 (B. cf. eurystoma),Aud3, Sou, Eng, andCat (the last three attributed ta B. simoniana)clustered together, wit h 0 ranging from 0.011(Audl vs. Aud2) to 0.139 (Aud3 vs. Sou ).

Because the patt ern of shared alleles indi­cated that gene flow cou ld not be excludedamong the Pyrenean foothill populations, weanalysed the genetic structure among themusing F statistics (Table 4) . The out-grou p and

the Par population of B. reyniesiiwere not in­ciudad in this analysis. The genetic varianceamon g populations (FST= 0.48 1) representedthe princ ipal component of the total gen eticvariance of the pop ulation s analysed (FIT =0.579); the genetic variation within populationswas relatively low (FIT = 0.189). Thus, theseBythinefla populations are highly differentiatedgenetically. The e values estimated betweenpairs of populations were significantly differ­ent trom zero, except between populationscollec ted in the same locality (Aud 1, Aud2,Aud3; Td1 1, Td12) or nearby (Td12 with Suz).The genotypic differentiations calculated foreach pair of populations and for ail loci revealsignificant P values , except between popula­tions located in the same locality (Aud1 to 3and Tdl1 and 2).

Phylogenetic Analysls

Of the 360 bp analysed , 76 posi tions werevariable and 48 were phylogenetically infor­mative. Figure 3 (right) displays the strict con­sensus tree based on 2,700 equi-parsimonioustrees resulting fram the heuristic search. Thetree lenglh was 96 steps with CI =0.96 and RI=0.96.

The topology revealed three distinct clades.The first clade (clade R) includ ed the individu­ais from Por referred to B. reyniesii. This cladewas supported by strong Bootstrap values (=100) and Bremer inde x (= 15). The secondclade (clade U) included the individuals frompopulations referred to B. utricu/us (TdI1, Suzand Roq). The thi rd clad e (clade S) includedthe individuals referred 10B. rubiginosa (Aud1),B. cf. eurys toma (Aud2) , B. simoniana (Catand Sou), and the Aud3 popuia tion. Clade Uand S constituted a monophyletic group, whichwas also supported by strong Bootstrap val­ues (; 100) and Bremer index (= 12).

The phylogenetic ana lysis corrobora ted theenzyme polymorphism studie s. Indeed , thethree monophyletic groups revealed corre­sponded ta the B. reyn ies! population and thetwo main groups revealed by the isozymeanalyses. Therefore, in the following morpho­logical analysis, we consider these groups asa hypothesis alternati ve to the classi cal tax­onomy, which can be formulated as follows:four popu lation groups of our dataset consti­tute four distinct species: Group V = Che (usedas out-group ); Group R = Por; Group S ; Aud1,Aud2,Aud3, Cat, Sou and Eng; Group U ; Tdl1,Td12; Suz and Roq.

SPECIES DELIMITATION IN THE GENUS BYTHINELLA 303

Morphometrical Analysis

Sexual Dimorphism - Overall 358 specimenswere successfully sexed, 55% males and45%females . Within a single population , 59 speci­mens from Roq included 36 (61%) males and23 (39%) females. PCA was carried out on the15 shell parameters LSR transformed. The first(PC1) and second (PC2) principal componentsrespectively account for 63.5% and 13.4% ofthe total variation (Fig . 4). Three groups ofstrongly correlated parameters were detected:(i) global shell size parameters (L"" W"" W,,,,L w1 ' Lpw2' L,wl' L'W2)' (ii) spire parameters (Wnw'

WfWlI L~P l andLsP2) and(Hi) aperture parameters(L~Pl ' L..P2' WflP1• WIlP2 ) · The firsttwo sets ofvari­ables, which are Iinked ta global shell size ,were primarily associated with PC1 . This firstaxis was thus interpretedas a size factor. Thethird set of parameters, related to shell aper­ture, were prlmarily assoclated with PC2. Theprojection of the individuals in the PC1 x PC2factorial plane revealed that, whereas femalesand males had similar distribution patternsalong PC1, they differed along PC2 (Fig. 4).

The DFA computed with sex as the discrimi­nate factor was slgnificant (WL = 0.4241. P =0.0002) . Scores ofWL with the probability as­sociated with each variable indicatedthat thetwo sexes differed significantly in one aper­ture parameter (W" , with WL =0.501 and p <0.008). The global reclassification score of thediscriminating linear functions was 86.44%(78.26% for females and 91.67% for males).These results suggest that females differ from

males by having a larger aperture width (W" ,),but not by shell size.

8etween populations , the same DFA analy­ses were then performed on ail populationsreferred , taxonomically and genetically, ta thegroup utriculus: 136 specimens were sexed,62% males and 38% females. This DFA wassignificant (WL =0.725, P =0.0004 ), but thetwo sexes differed significantly in the param­eter Lap1' another aperture parameter (LIIP1 withWL =0.755, P =0.028).

Finally , a DFA was performed with sex asthe discriminate factor on the global data setof 358 sexed specimens. The analyses wasglobally significant (WL =0.914 , P =0.011),indicating that the set of variables chosen al­lowsdiscrimination betweenthe sexes. Scoresof WL with the probability associated with eachvariable suggested that the two sexes differedsignificantly onJy by one aperture parameter(L'O, with WL =0.927 and p < 0.030).

ln order ta test for the respective effects ofsexualdimorphism and taxonomie differenees,two analyses of variance (ANOVA) were per­formed on the LSR global data set on the twoaperture parameters Wap2 and LIIP1 that signifi­eantly differed between males and females.First , we tested the effects of the two factorstaxonomie group and sax and their interaction ,and then the effect of the factors genetic groupand sax and their interaction. Thesa twoANOVAs revealed that bath effects of taxo­nomie group and genetic group were signifi­cant for the two parameterswhereas the effeetof sax was not significant in any analysis.

• . ; 0

• ~ •• •

l'C I fiJ.5%o

•

•

FIG. 4. Sexual dimorphism in Roq population of Bythinella: results of PCA. Left Correlation of the 15parameters measured on the shell to the first two principal components PC1 and PC2. Right Projec­tion of the individuals in the first factorial plan PC1 x PC2. Open circles are females andfilled eirelesare males.

304 BICHAIN ET AL.

TABLE 5. Correlations rnatrlx between the variables and the first two discriminate functions (F1 andF2) for the two alternative Bythine/la hypotheses (Taxon group and Genetic group).

Taxon group hypothesis Geneticgrouphypothesis

Shell parameters F1 F2 F1 F2

Lsh -0.027 0.189 -0.058 -0.243Wsh -00460 0.160 -00495 -0.216Lap2 0.001 -0.300 0.017 0.327Wap1 0.076 -0.217 0.103 0.265Wlw -0.504 0.084 -00407 -0.008Lpw1 -0.068 0.209 -0.102 -0.267Lpw2 -0.090 0.228 -0.125 -0.289L1w1 -0.108 0.200 -0.147 -0.267L1w2 -0.188 0.184 -0.237 -0.264Waw 0.263 -0.184 0.307 0.256Wpw 0.166 -0.069 0.207 0.145Lsp1 0.368 -0.179 00418 0.256Lsp2 00465 -0.147 0.523 0.230Eingenvalue 3.756 1.383 3.308 1.318

Cum. Eigenv. 53.8% 73.7% 63.3% 88.5%

For each of the two ape rture parame ters, thetaxonomie group x sex interaction was notsignificant whe reas the genetie group x sexinteraction was significant but only for the Wop,

parameter (L'P' F-test = 1.32 w ith P =0.267and W'P' F-test =4.226 wit h P = 0.006). Forthis parame ter, the post hoc test revealed thatthe geneticgroup x sex inte raction was caused

- 8 -4 4 - 8

- 4 _. .

- 4F2 25.2 %

4

FIG. 5. DFAtastingthe two alternative hypotheses Taxon groups vs Genetic groups within Bythinellaspecies. Left Taxon groups hypothesis. Filled circles: B. viridis (one population), open clrcles : B.utrieu/us (four populations), filledtriangles: B. reyniesii (one population), open triangles: B. simoniana(three populations), filled squares: B. cf. eurystoma (one population), open squares: B. rubiginosa(one population). Right Genetic groups hypotheses. Filled circles: group V (B. viridis, one popula­tion), fill ed triangles: group R (B. reyniesii, one population), open squares: group U (B. utrieu/us, fourpopulations), filled rhombus: group S (B. simoniana, B. rubiginosa, B. cf. eurystoma and Aud3 popu­lation, total six populations).

SPECIES DELIMITATION IN THE GENUS B YTHINELLA 305

TABLE 6. Global reclassification scores of the discriminating linear functions for the two Bythinellahypotheses. Each line gives the number of specimens of a given taxon classifie d as any of the taxa.and the percentage of each that are correctly classified.

Taxon group hypothesis

aythinefla % correct viridis reyniesii utriculus simoniana cf. eurystoma rubiginosa

viridis 88.9% 24 0 3 0 0 0reyniesii 90.0% 0 27 3 0 0 0utriculus 87.2% 0 5 129 13 1 0simoniana 82.2% 0 0 13 74 1 2cf. eurystoma 83.3% 0 0 1 0 25 4rubiginosa 83.3% 0 0 1 2 2 25

Total 85.6% 24 32 150 89 29 31

Genette group hypothesis

% correct Group V Group R Group U Group S

Group V 81.5% 22 0 5 0Group R 90.0% 0 27 3 0Group U 86.5% 1 5 128 14Group S 92.8% 0 0 13 167Talai 89.4 % 23 32 149 181

by Por (group R, attribuled ta B. reyniesiJ), lnwhich the sex ratio was 24% males and 76%females.

Tasummarize, the analysisshows thatvaria­tionofaperture parametersis mainly explainedby differences between "spec ies" (defined oneither the taxonomie or genetie dei imitationhypotheses ), and that sexua l dlmorphismwithin a population is manifest inaperture size.Ta minim ize the effect of the sexual dimor­phism, the Iwo aperture parameters were re­moved in the subsequent analyses.

Shefl Variation Among Pop ulations and Taxa- Ta explore the pattern of shell variation, aPCA was carried out on the whole LSR datasel(385 individuals of bath sexes). The first fac­torial plane accounted for 92.1% of the globalvariation with 80.3% by PC1 and 11 .8% byPC2 (graph not presented here). Three groupsof strongly correlated parameters were de­tected: (i) spire and aperture parameters nega­tively correlated with the PC1, (Ii) helght of thevarious whorls positlvely correlated wilh PC1,and (ili) shell width and width of the last whorlnegatively correlated with PC2. The projectionof the individuals in this faclo rial plane re­veaJed a strong overlap between the variouspopulati ons. On the PC2 axis, only individu-

ais of B. viridis had a distinct position. Thus,except for this taxon, the parameters used inthis analysis dld not reveai any differen cesamong groups of individuals when using ei­ther the taxonomlc or the genetic hypotheses.

Taxon Hypothesis - The DFAon the LSR glo­bal data set (Iess the taxonomically doubtfulAud3 population and L, 1 and W, ,) performedwith Taxon as discriminats factor was signifi­cant(WL ~ 0.023, P < 0.05). Scoresof WL wilhprobabili ty ass oc iated with each variableshowed that ail parameters significantly dis­criminate the taxonomie groups. Projection ofthe individuais on the first Iactorial plane F1 xF2 (each axis respectively accounting for 53.8%and 19.8% of the total variance) allowed dis­crimination, mainly on the F1 axis. of a clusterofindividuals attributed ta B. viridis. The F1 axiswas mainly correlated wilh spire parameters(W w' L 1 and L. ), whereas the F2 axis wasma'inly ëxplaine;rby a size effect (Table 5). In­deed, for this axis, ail variables conlributed thesame weight and were strongly correlaled (Fig.5). In this F1 x F2 faclorial plane, other taxe­nomic groups were poorly resolved with anextensive overlap among them. For example,individualsreferred toB. simoniana. B. cf. eury­stoma and B. rubiginosa were not separated.

306 BICHAIN ET AL.

The globa l reclassification score (Table 6)of the discrimina ting linear functions was85.6%. with speci fie scores ranging trom83.3% (i ndividuals attr ibu led 10 B. cf.eU/ystoma and B. rubiginosa ) to 90.O% (indi­viduals allribuled to B. reyniesil).

Genetic Hypofh esis - DFAs performed onthe LSR global data sel (Iess L. , and W, , )with Genetic groups as discrim~nate fact~rwe re also signi fic ant (WL ; 0.062, P < 0.05).Scores of WL with probability associated foreach variable showed that ail param eters sig­nifican tly discrim inate d the genetic groups.Th e F1 x F2 factorial plan accounted fo r88.5% of the global variance, wi th 63.3 % forF1 and 25 .2% for F2 (Table 5). The spire pa­rameters are positively co rrelated wi th F1,whereas W h and Wtw were negatively corre­lated with the sa,me axis. The. W aPI and L "P2

paramete rs conlnbule d ess entia lly to F2. Theprojec lion of individuats in lh is facto rial planerevealed three distin ct groups on lhe F1 axis,correspo ndi ng to individuals assigned togroups V, U and S defined beiow (Fig . 5). Onthe other hand , F2 allowed separation of groupR from ail other groups . Th e global reclassifi­catio n scores (Table 6) we re 89.3% under theGenetic hypot hesis , with reclassificalionscores ranging from 81.5% (group V) to 92.8%(group S).

From this result , we ca n ident if y fourmorphogroups basad mainly on spire param­eters : (i) group V with a sma ll spire and a largelast whorl that gives the she ll an ovoid outli ne;(ii) group S wi th a iarge overall size and a highspire that gives lhe shell an elo ngate outl ine;(iii) group U, inle rmediate between V and S,with anoulline that can be described as ovoid ­elongate ; (iv) group R wi th a spire heig ht equa lto the height of the last whorl and a large ap­erture , tha t gives the shell a pupoid outline.

DISCUSSION AND CONCLUSION

Our analyses confirmed that ail specimen satlributed to B. viridis (Fig. 6A) , which are usedas out-g roup , we re morp hologically and ge­netically distincttrom ailPyrenean specimens.

Within the Pyrenean sample, two speciesdelimitations resulting trom our analysis con­firmed earHer tradilional species deiimitatians(Bernasconi, 2000). Specimens att ributed toB. reyniesii (Fig. 6B) we re both morp holog i­cally and ge nelically distin ct from ail oth erspecimens included in our study. The morpho-

metric analysis describes the shell as havinga spire equal to the height of the last whorland a large aperture, resulting in a pupoidoverall shape. However, since a single popu­lation was included in the study, this descrip­tion lacks robustness andwe do not know howthe characters observed may be subject tophenolypic plas ticity or to genelic variability.

The second vatidated species was B. utricu­fus (Fig. 6F), in which specimens from ali popu­lat ions c1ustered both as a monophyletic group(DNA sequences) and as a gene tic group(isozymes) . We hypoth esize that the entireclade can be classified as a single species. Asseveral populatio ns are included within thisspecies, our morphometrie characterisationintegrates at least part of the phenolypic plas­ticity component of shell shape variability.

Ali other populations formed a single geneti cand phylogenetic group. Morphometric analy­sis describes the shell within this group ashaving a large ove rail size and a high spirethal gives il an elongate shape. This groupincluded the nominal B. simoniana (Fig . 6D) ,characterized in traditional taxonomy by axialvarices on the last wharl - a character notconfirmed in our analysis as of specifie diag­nostic value - and also included specimensIrorn the two springs complex at Audinac, theco le spring (Fig. 6C) and the therm al springtype iocality of B. rubiginosa (Fig. BE), and theircon fluence. Thus, this c1uster groups at leasttwo nominal taxa that shou ld be cons idered asingle specles , to be nam ed B. rubigin osabas ed on the Principle of Prior ity (ICZN CodeArt . 23.3). The high quantita tive (size) or quali­tative (varic es) shell variabil ity observed in thisgro up may be due either to phenotypic plas­licity or to intraspecific genetic variation.

However , althoug h DFA ana lyses allowseparation of these genetic groups, the PCAresu lts highligh ted overlaps among mosl popu­lations. indicating that a priorideterminationsat the species level rank in Bylhinella are stilldifl icull even when using the shell characlershere identifie d. Moreover, these shell charse­ters canno t be validated for alpha-ta xonomywithout a robust test at the scale of the genus.We show here that it is possible to give analternative speci es delimita tion hypoth esis tothe traditional view in an evolutionary frame­wo rk and that these newly defined ent itiesexh ib it distinct morphological features. Weagree with Wilke et al. (2002) about the ne­cessityto use molecular markersand morpho­metrical approaches in concert to study crypticor cJosely relaled species.

SPECIE S DELIMITATION IN THE GENUS BYTHfNE LLA 307

FIG. 6. Shells of lhe nominal species of By/hinel/a studied. A: B. viridis (Che): 8 : B. reyniesi;(Par): C: B. cf . eurystome (Aud2): D: B. simoniana (En9); E: B. rubiginosa (Aud1); F: B. utricu­lus (Roq) (Scale 1 mm).

Genetic studi es of Polish populal ions re­vealed weak genetic diffe renlialion primaritycorre laled wilh ge ogra phic di stance(Falniowski et al., 1998, 1999). The formersuggesled Ihal ", .. Ihe superspecies status ofIhe genus (Giusti & Pezzoli , 1977) seems themostjuslified", The lerm superspecies was firslused by Mayr (1931) and subsequenlly de­fined by Mayr & Ashlock (1991) as "a mono­phyletic group of closely related and enlirelyor largely allopatric species that are tao dis­tinct to be included in a single species or lhaldemonstrate their reproduction isolal ion in azone of contact. " Our data don't refute thisopinion within the Pyrenean area where amosaic of populations probably belonging to

different species more or less morphologicallydifferentiated seems ta coexist in allopatry.

Nevertheless, we show that sorne shell char­acters thal were lraditionally used 10 segre­gale species of Bythinella (see, for example,Bernasconi, 2000) do not necessarily havetaxonomiesignificance, a result agreeing withIhal of olh ers working on olher hydrobioidgroup s (Falniowsk i & Wilke, 2001 ; Wilke &Falniowski, 2001; Szarow ska & Wilke, 2004).The high inlraspeclfic shell variabilily may belinked 10biotic faclors, such as food availabil­ity, parasile-induced gigantism (Jourdane,1979; Gorbushin, 1997; Probsl & Kube, 1999),or sexua l dimorphism (Ponder el al., 1999;Kurala & Kikuchi, 2000; Velecka & Jüllner,

308 BICHAI N ET AL.

2000; Chiu et al., 2002). Our study carraba ­rates Faln iawski 's (1987 ) observation thatsexua l dimorphism is rather law in Bythinella,females differi ng from males by their gener­ally larger shell aperture. Hawever, as sex ra­tio appea rs nat lixed in po pulations 01Bythinella, superficia l marphal agical differ­ences among populations may also be a con­sequence 01dlfferent sex ratios. and cantributeta fuzzy specie s delimilatians. Shell varlabll­ity may alsa be linked ta abiatic pararneters,such as temperature (Brown & Richard son,1992). In the three Audinac populati ons of B.rubiginasa. specimens fram the thermal springare largest, spe cimens from the cold springare sma llest, and speci mens from theconnuence are intermediate.

Toconclude, the present work offers an ex­ample of a three-ste p integ rative taxanamystarting tram (i) primary hypotheses (currentspecies delimitatians based on phenetic char­acie rs), ta (ii) the test 01 these hypatheses bycriteria lhat are canceptually sound. Popula­tion geneties appraaches allaw ta idenlify re­pr aducti vel y isal ated gr oup s. Malecularphylageny allaws the recognition of manaphyl­etic groups and thus the detectian 01 groupsthal share ancestry. These Iwo criteria, derivedlrom the BSC and the PSC . permit new spe­cies delimitatian, and finally (Iii) a leedback taa new taxonomie hypothesis associated withnew phenetic descriptars. Therefare, the finaldescription corresponds ta the mast filt inghypathesis given the data at lhe time of thedescription.

A correct delim itatian of species has impar­lant consequences bath for a reliable estima­l ion of biad ive rsity but alsa when draw ingconservationpriorities. Two of the nominal spe­cies in the present study (B. viridis and B.reyniesil) are categarized as Vulnerable in theIUCN Red Lisl and are alsa protected underFrench law. Bath emerged fram the presentstudy as probabl y distinct evalutianary units.

Our study was restricted ta a small part ofthe geag raphical range of Bythinella includ­ing a subset of ils nominalspeciesand used alaw number of isazyme loci and a short geneIragment. Thu s, our results are of preliminarynature. In arder ta validate them, we need taexpand this approa ch ta a global revisian ofthe genus, including supplementary mitachan­drial and nuclear marksrs. Hawever, lhe chal­lenges to expanding thismodel are immense.T he Iiterature on European hydrobioids is re­piete with nam es of unkn own significance(Davis, 2004 ) that blur evaluatians of regianal

species richness , conservation priorities andevolut ionary history. Th is is because manynom inal species we re estab lished by 1911

' and20~ century authars that (i) worked on a localor regianal basis, and/or (ii) worked autside acontext of evo lutionary syste matics. Their ap­proach la alpha-tax anamy has alsa accasian­ally injected in the literature names based onfew, sometimes single. specimens only weaklydifferentiated canchalagicall y. Because thepresent study shows that geneti c or phylage­netic distinctiveness cannat be inferred fromIhe degree of shell distinctiveness, the chal­le nge to a co mp rehensive re vi s ion ofBythinella is that these nominal species haveta be re-evalu aled based on papula l ianssampled alive fram their type lacalily. At thismoment, we do nat knaw whether an integra­live revisian of Bythinella will result in just afew or seve re ! dozen spe cle s .

Haw do Ihe present results impact the sys­tematics of fassil Bythine/ia, and daes Ihe re­suit of the present study preclude naming anynew Bythinella species based on shell char­acters only? Despite the numerou s nominalspecies of Bythine/ia described by 19'" and 20·century autha rs , new exploratio n 01 littl e­known aqui fers, in partlcular hypogean envi­ronments in sauthern France, stilileads ta thediscavery 01 new marphatypes. Sametimes ,anly empty shells are knawn and living speci­men s escape collection despite inten sivesearches. We understand that il may be nec­essary, for communicatio n or conservationpurposes, ta name such taxa; however, it isessential thatthe primary description is basedon shell characters Ihat have been tested fortheir robustness regard ing spec ies delimita­tian.

Stabil ity in the alpha-taxanamy 01hydrobiaids will mast likely be reached thraughan inlegralive taxanamy approach (Sites &Marshall, 2003 , 2004), A possible appraachwauld be Ihe Hennigian inter-nodal specie sconcept derived from the Theary of Evolution(Samadi & Barberou sse, 2006) that hereproved an efficient taal in the study 01PyreneanBythine/ia.

ACKNOWLEDGM ENTS

The sequenci ng was done in the Service deSys lématique Molécu laire atthe Muséum na­tional d'Histoire naturelle of Par is with thetechn ical help of Annie Ti ll ier, JosieLabaurdiére and Céline Banilla. We are grate-

SPECIES DELIM ITATIO N IN THE GENUS 8 YTHINEL LA 309

fui ta Ahmed Abdau, Ala in Bertrand , BenoitFonta ine and Ol ivier Gargamin y for the ir co l­laboratio n during the malerial sampiing and taPhi lippe Deliat for his he lp in the marphamet­ric ana lyses. Special thanks ta Rob ert Cawieand Sim on Tillier for their hel pful camments.

L1TERATURE CITE D

ALZO NA, C ., 197 1, M al acofaun a ita lica .Catalogo e bibliografia dei molluschi viventi,ter restri e d'acqua dolce. Alti della Societaitaliana di Scienze Naturali e deI Musee Civicodi Storia Naturale di Milano, 111 : 1-433.

BARRIEL , V.• 1994. Phyloq énles molécu lairesel insertions-délétions de nucléotides.Comptes Rendus de l'Academie des Sciencesde Paris. Sciences de la Vie, 317: 693-701 .

BELKH IR , K., P. BORS A , 1. CH IKHI, N.RAUFASTE & F. BONHOMME, 200 1, GeneUx4.02, logiciel sous Windows TM pour lagéné tique de s population s. Mon tpellier,Laboratoire Génome, Populations, Interac­lions : CNRS UMR 500 . Unive rsit é deMontpellier Il.

BERNASCONI, R.. 2000 , Révision du genreByt hineUa (Moquin-Tan do n, 185 5) (Gas­tropoda , Prosobranch ia . Hyd robiidae ,Amnicolinae, Bythinellini). Documents Malaca·logiques, Hors -Série 1: 1-126.

BERTR AND , A., 2004, Atlas préliminai re derépartition géog rap hique des mollusquessouterrains de France (Mollusca: Rissoidea:Caenogastropoda ). Do cuments Malaco­logiques. Hors Série 2: 182. .

BICHAIN , J. M., C. BOUDSOCQ & V. PRIE ,2004, Les mollusques souterrains du réseaukarstique de Padirac (Lot, France) et micro­répartition de 8ythinella padiraci Locard, 1903(Mollusca , Caenogas tropoda, Rissooidea).Karst%gia, 43( 1): 9-18.

BOISSELIER-DUBAYLE, M. C. & S. GOFAS,1999, Ge netic relationships between marineand marginal-ma rine populations of Cerithiumspecies from the Mediterranean Sea . MarineBiology, 135: 671-682.

BREM ER, K., 1994, Branch support and tree sta­bilily. Cladistics , 10: 295-304.

BROWN. K. M. & T. D. RICHARDSON, 1992,Phenotypic plasticity in the life histories andproduction of two warm -temperate viviparldprosabranchs. The Veliger , 35(1): 1- 11.

CAVALLI -SFORZA, L. L. & A. W. F. EDWARDS,1967 , Phylogenetic analysls: models and esti­mation procedures. Evolution, 32: 55 0-570.

CHEN, C. A., C. C. CHANG, N. V. WEI. C. H.CHEN , Y. T. LEIN, H. E. LIN, C. F.DAI & C. C.WALL ACE, 2004, Secon dary struc ture andphylogenetic utility of the ribosomal internaitranscribed spacer 2 (ITS2) in scleractiniancorals. Zoo/ogical Studies, 43(4 ): 759-771 .

CHIU, Y. W, H. C. CHEN, S. C. LEE & C. A.CHEN, 2002 , Morpho metri c analysis of shelland operculum variations in the viviparid snail,

Cipangopaludina chinensis (Mollusca: Gas­tropoda ), in Taiwan . Zoo/ogieal Studies , 41(3):321-331 .

COLEMAN , A. W , 2003 , ITS2 is a double-edgedtool for eukaryote evolutionary. Trends in Ge­netics , 19: 370-375.

DAVIS, G M.• 1979, The orlqin and evolution ofthe gastropod family Pomatiopsidae , with em­phasis on the Mekong River Triculinae. Mono­graph of the Academy of Natural Sciences ofPhilade lphia , 20: 1- 120.

DE QUElROZ, K., 1998, The generai lineageconcept of species, species criteria, and theprocess of speclatlon: a conceptual unificationand terminological recommendations. Pp. 57 ­75, in: D. J. HOWARD & S. H. BERLOCHER, eds.,Endless forms. Species and specialïon. OxfordUniversity Press, Oxford. xii + 470 pp.

DE QUEIROZ. K. , 1999, The general Iineageconcept of species and the defining propertiesof the species ca lego ry. Pp. 49-89. in: R. A.WILSON, ed ., Species, new interdiscip/inaryessays. Massachuse tts: MI T Press, Cam­bridge. xxi + 325 pp.

FALKNER, G, T. E. J . RIPKEN & M. FALKNER.2002 , Mo/Jusques continentau x de la France :liste de référence annotée et bibliographie.Patrimoines naturels, 52, Paris. 350 pp.

FALNIOWSKI, A.. 1987, Hydrobioidea of Poland(Ga stropoda , Prosobranchia). Folia Malaco­logiea, 1: 1- 122.

FALNI OW SKI. A .• K . MAZAN & M. SZA­ROWSKA, 1999 , Homozygote exces s andgene flow in the sprin~ snail 8ythinella (Gas­trcpoda. Prosobra nchla). Journal of Zootoqi­cal Systema tic and Evolutionary Research, 37 :165-175.

FALNIOWSKI . A ., M . SZAROWSKA, W.FIALKOWSKt & K. MAZAN, 1998. Unusualgeographic pattern of interpopulation variationin a spring snail Bythine/la (Gastropoda,Prosobranchia). Journal of Natura l History, 32 :605-616.

FALNIOWSKI, A. & 1. WILKE, 2001 , The genusMarstoniopsis (Gastropode : Rissooidea): Intra­and Intergeneric phylogene tic relationships.Journal of Mollusean Studies, 67: 483-488.

FAUNA EUROPAEA 2004. Fauna Europa ea da­tabase. Available: http://www.fauna eur.orglAcce ssed: 10 Janua ry 2005

FELSEN STEIN, J .• 1993. PHYL/P. Phyloqenet icInference Package. version 3.Sc. University ofWashington, Seattle .

GIUSTI, F.& E. PEZZOLI. 1977, Primo cantributoalla revisione deI genere Bythinella in ltalia.Natura Bresciana, Brescia , 14: 3-80.

GIUSTI, F. & E. PEZZOLl , 1980, Gastropodi, 2(G astropoda : Prosobranchia: Hydrob ioidea,Pyrguloidea). Guide peril riconoscimento dellespecie animali della acque interne Italiana , 8 :67 pp.

GORBUSHIN, A. M., 1997, Field evidenc e oftrematode-induced gigantism in Hydrobia spp.(Gastropoda: Prosobranchia ). Journal of Ma­rine Biotoçy, 77: 785-800.

GUO , S. W. & E. A. THOMPSON, 1992, Per­formed the exact test of Hardy-Weinberg pro-

310 BICHAIN ET AL.

portion formultiple all èles. Biometries, 48: 361­372.

HALDANE, J. B. S., 1954, An exact test for ran­domness of mating. Journal of Genetics, 52 :631-635.

HALL, 1. A., 1999, BioEdit: a user-friendly bio­logical sequence alignment editer and analy­sis program for Windows 951gB/NT. Nue/sicAcids Symposium Series. 41: 95-98.

HENN IG, W., 1966, Phylogenetic systematics.University of lIIinois Press, Urbana. 263 pp.

HERSCHLER, R. & W. F. PONDER , 1998, A re­view of morphologicaJ characters ofhydrobioidsnails. Smitnsonten Contributions to Zo%gy,600: 1-55.

HEY, J., 2001 a, Genes, categories and species:the evolutionary and cognitive causes of thespecies problem. Oxford University Press,Oxford. xvii + 217 pp.

HEY, J., 2001 b, The mind of the species prob­lem. Trends in Eco/ogy and Evolution, 16(7):326-329.

HILLlS , D M & M. 1. DIXON, 1991, Ribosoma lDNA: molecuJar evolution and phyJogeneticinference, The Quaterly Review of Bioloçy,66(4): 411-451.

HILLlS, D. M., C. MORITZ & B. K. MABLE, 1996,Mo/ecu/ar systematics, 2nd ed. Sinuar Associ­ates, Sunderland. xvi + 655 pp.

HOLM, S., 1979, A simple sequentially rejectivemultiple test procedure. Scandinavian Journalof Statist/cs, 6: 65-70.

HOWARD, D. J. & S. H. BERLOCHER, 1998,Endless forms. species and speciation. OxfordUniversity Press. New York. xii + 470 pp.

JOURDAN E, J., 1979, Le cycle biologique deMaritrema pyrenaica Deblock et Combes,1965. Parasite de micromammifères insecti­vores pyrénéens. Annales de Parasitologie,54(4) : 449-456.

KABAT, A. & R. HERSHLER, 1993, The proso­branch snail family Hydrobiidae (Gastropoda,Rissoo idea): review of classification andsupraspecifictaxa. Smithsonian Contributionsto Zoology, 547: 1-94.

KORNET, D. J., 1993, Permanent splits as spe­ciation events: a formai reconstrustion of theintemodal species concept. Journal of Theo­retical Biology, 164: 407-435.

KURATA, K. & E. KIKUCHI, 2000, Comparisonsof life-history traits and sexual dimorphismbetween Assiminea japonica and An­gustassiminea castanea (Gastropoda:Assimineidae). Journal of Molluscan Studies ,66: 177-196.

LANGELLA, 0., 2002, Population 1.2.28. Logicielde génétique des populations. LaboratoirePopulations, Génétique er Evolution, CNRSUPR 9034, Gif-sur-Yvette.

MALLET, J., 2001, Specles, concept of. Vol. 5.Pp. 427-440, in: S. A.LEVIN, ed., Encyc/opediaof biodiversity. Academie Press, New York.

MAYR, E., 1931, Notes on Ha/cyon chloris andsorne of its subspecies. Museum Noviietes,469 : 1-10.

MAYR, E., 1942, Systematics and the origin ofspecies . ColumbiaUniversityPress, New York.xiv + 334 pp.

MAYR, E. & P. ASHLOCK, 1991, Principles ofsystematic zoology , 2nd ed. Mc Graw Hill, NewYork. xx + 475 pp.

MAZAN, K., 2000, Morphological and allozymicvariation within and between populations ofBythlnella Moquin -Tandon, 1855 (Gastropoda:Prosobranchia). 1. Morphological characters.Folia Ma/acologica, 8(2): 107- 139.

MAZAN, K. & M. SZAROWSKA, 2000a, Morpho­logical and allozymic variation within and be­tweenpopulations of BythinellaMoquin-Tandon ,1855 (Gastropoda: Prosobranchia). 11. Pheneticanalysis. Folia Malacologica, 8(3): 189-213.

MAZAN, K. & M. SZAROWSKA, 2000b, Morpho­logical and allozymic variation within and be­tweenpopulations ofBythinella Moquin-Tandon,1855 (Gastropoda: Prosobranchia). III. Phylo­geneticanalysis. Folia Malac%gica, 8(4): 257­269.

MOSIMANN, J. E., 1970, Size-allometry size andshape variableswithcharacterization of the lognormal and generalized gamma distributions.Journal of the American Statistica/Association,65: 930-945.

NOOR, M. A. F., 2002, Is the bio logical speciesconcept showingits age? Trends in Ec%gy &Evolut/on, 17: 153-154.

PONDER, W. F., S. CLARCK & A. C. MILLER,1999, A new genus and two new species ofHydrobiidae (Mollusca. Gastropoda, Caen 0­gastropoda) from south Western Australia.Journa/ of the Royal Society of Western Aus­tralia, 82: 109- 120.

PROBST, S. & J. KUBE, 1999, Hlstopathologi­cal effects of larval trematode infections inmudsnails and their impact on hast growth:what causes gigantism in Hydrobia ventrosa(Gastropoda: Prosobranchia). Journal of Ex­perimenlal Marine Biology and Ecology, 238:49-68.

RADOMAN, P., 1976, Speciation within the fam­ily Bythinellidae on the Balkans and Asia Mi­nor. Zeitlschrift fürZootcçiecne. Systematik undEvolutionsforschung, 14(2) : 130-152.

RAYMOND, M. & F. ROUSSE T, 1995, Genepop(version 1.2): population genetics softwareforexact tests and ecumenicism. Journal of He­redity, 86: 248-249.

RICE, W. R., 1989, Ana lyzing tables of statisti­cal tests. Evolution, 43: 223-225.

ROHLF, F. J., 2001, TPSdlg version 1.23. StateUniversity at Stony Brook, New York.

SAMADI, S. & A. BARBEROUSSE, 2006, Thetree, the network and the species. BiologicalJournal of the Linean Society, 89: 509-521.

SCHILTHUIZEN, M., E. GUTIELlNG, C. H. M.VAN MOORSEL, W.WELTER-SCHULTES, M.HAAS E & E. GITTENBERG ER, 2004,Phylogeography of the land snail Albinarial1ippolyti(Pulmonata, Clausiliidae) From Crete,inferred from ITS-1 sequences.Biological Jour­nal of the Linnean Society, 83: 317-326.

SP ECIES DELIMITATION IN TH E GENUS BYTHINELLA 311

SIMPSO N, G, G.. 1961 , Principles of animaltax­onomy. ColumbiaUniversity Press. New York.247 pp,

SITES , J, & J. MARSHALL, 2003, Delimitingspecies: a renaissance issue in systematic bi­ology. Trends in Ecology and Evolulion, 18(9):462-470.

SITES, J. & J. MARS HALL, 2004, Operationalcriteria for delimiling species. Annual reviewsof Ecology and Systemalic, 35: 199-227.

STATSOFT, 2001, Sial/sI/ca version 6 ,www.statsoft.com/french/wefcome.hlml.

SWOFFORD, D. L., 2000, PAUP·. PhylogenelicAnalysis Using Parsimony (*and Other Meth­ods). Version 4. Sinaue r Assoclates. Sunder­land, Massachusetts.

SZAROWSKA, M. & T, WILKE, 2004, Sadlerianapannonica (Frauenfeld, 1865): a lithoglyphid,hydrobiid or amnicolid taxon? Journal of Mo/­lusean Studies, 70: 49 -57,

TH OMPSON , J. D., D. G. HIGGINS & T. J.GIBSON, 1994, CLUSTAL W: improving thesensilivity of progressive multiple seq uencealignment through sequence weighting. posi­tion-specifie gap penalties and weight matrixchoice. NucleicAeids Reseercn. 22: 4673-4680.

VELECKA, 1. & L. JOTINER, 2000, Appli cationof McKendrick-von Foerster's model of popu­lation dynamics la freshwaler snails: specifi­cation of the madel using la boratoryexperiments. Acta Universitetis PalackianaeOlomucenis Facu/las Rerum Naturalium, 38:85-95.

WEI R, B. S., 1990, Genetlc dala analysis.Sinauer Assoclates, Sunderland, Massachu­setts. xii + 445 pp.

WEIR, B. S. & C. COCKERHAM, 1984 , Estimat ­ing F-staslistics for the analysis of populationstructure. Evolution, 38 : 1358- 1370.

WHEELER, a. D. & R. MEIER, 200 0, Speciesconcepts and phylogenelic ttieory. ColumbiaUniversity Press, New York. xii + 230 pp.

WILEY, E. O., 1981 , Phylogenetlcs: the theO/yand practice of Ihe phylogenetic systematics.Wiley and Sons, New York. xv + 439 pp.

WILEY, E. O. & R. L. MAYDEN , 2000 , The evo­lutionary species concept. Pp. 70-89, in: Q. D.WHEELER & R. MEIER, eds .. Species conceptsand phylogenetic theory: a debate. ColumbiaUniversity Press, New York. xii + 230 pp.

WILKE, T., G. M. DAVIS, A. FALN IOWSKI, F.GIUSTI, M. BODON & M. SZAROWSKA, 2001,Molecular systematicsofHydrobiidae(Mollusca:Gastropoda: Rissooidea): Testing monophylyand phylogeneticrelationships. Proceedings ofthe Academy of Natural Sciences of Philadel­phia, 151: 1-21.

WILKE, T. & A. FALNIOWSKI, 2001 , The genusAdriohydrobia (Hydrobiidae, Ga stropod a):polytypic species or polymorphie populations?Journal of Zoo/agical Systematic and Evolu­tion Researeh, 39: 227-234.

WILKE, T., M, PFEN NINGER & G. M. DAVIS,2002, Anatomical variation in cryptlc mudsnailspecies: statistical discrimination and evolu­tionary significance. Proceedings of theAcad­Emy ot Natural Sciences ot Philadelphia, 152:45-66.

WILSON, R. A., 199 9, Speeies: new interdisc i­plinary essays. Massachusetts: MIT Press,Cambridge. xxi + 325 pp.

WINSTON, J. E., 1999, Describing species:prsc­licaltaxonomieprocedure forbiolagists. Colum­bia University Press, New York. xx + 518 pp.

Revised ms. accepted 29 September 2006