Aliso: A Journal of Systematic and Floristic Botany Aliso: A Journal of Systematic and Floristic Botany

Microsatellite alleles were used to delimit the genetic boundaries and divergence of the two relictual endemic Pyrenean taxa Borde rea chouardii and B. pyrenaica (Dioscoreaceae ), and to infer the different life histories followed by each species. Our study was conducted on the same populations previously analyzed with allozymes and RAPD markers. The three studied data sets were congruent in the inference of a single evolutionary scenario for the split of the two Borderea taxa from a common Tertiary ancestor in the Prepyrenees, thus supporting their taxonomic treatment as separate species. However, the more variable SSR and RAPD data provided better resolution for a stepping-stone model of local colonization of B. pyrenaica populations from southern Prepyrenean refugia to the northern Pyrenees. SSR markers proved to be more robust than RAPD markers in assessing the genetic structure of recently diverged populations of B. pyrenaica and thus qualified as the best molecular markers for fine-scale evolutionary investigations of Dioscoreaceae. Furthermore, microsatellites rendered unique clues to decipher the mechanisms involved in the origin of these relictual species and their genetic background. Borderea was shown to be a tetraploid genus of hybrid origin with a chromosome base number of x = 6. Phylogenetic data, karyological evidence, and our present knowledge based on microsatellite analyses allowed us to speculate that the Pyrenean endemic genus Borderea and its sister taxon, the Mediterranean genus Tamus, represent some of the oldest paleopolyploid lineages of the mostly pantropical


INTRODUCTION
The genus Borderea Miegev., endemic to the central Pyrenean and Prepyrenean mountain ranges, has been considered to be a Tertiary relictual lineage of Dioscoreaceae (Gaussen 1965) based on the fact that the vast majority of its family members (more than 600 spp.) show a present pantropical distribution with only a few taxa growing outside of that range (Knuth 1924;Burkill 1960;Dahlgren et al. 1985).
The species of this genus have been subjected to several taxonomic rearrangements throughout history.The two currently accepted taxa (Borderea pyrenaica and B. chouardii;cf. Heywood 1980;Villar et al. 2001) were first described within the large pantropical genus Dioscorea L. as D. pyrenaica Bubani & Bordere ex Gren.and D. chouardii Gaussen, respectively (Gaussen 1952(Gaussen , 1965)).The name Borderea arose in the mid-nineteenth century (Miegeville 1866) to describe specimens from the Pyrenees (B.pyrenaica Bubani ex Miegev.), which differed from Dioscorea mainly in their dwarf habit and wingless seeds.Karyological analyses of the two species demonstrated that both taxa shared not only these remarkable morphological attributes, but also a chromosome number of 2n = 24, which was assumed to represent a chromosome base number of x = 12, distinct from that shown for most Dioscorea taxa (x = 10).In turn, this chromosome value was used as a further argument to classify both species under the same genus Borderea (Heslot 1953).On the basis of morphological similarities with Bor-derea, other taxa native to geographical regions apart from the Pyrenean range were also transferred to this genus.Thus, the Chilean endemic Dioscorea humilis Bert.ex Colla was renamed as B. humilis (Bert.ex Colla) Pax.However, this and two other Andean dwarf endemics show certain distinctive morphological features within Dioscoreaceae (i.e., the possession of prominent pistillodes in the male flowers, round capsules, and strongly emarginate leaves) that favored their separate treatment as members of the independent genus Epipetrum Phil.(Knuth 1924).Milne-Redhead (1963) described from Kenya Dioscorea gilletii Milne-Redh., a taxon potentially close to Borderea based on its wingless seeds and other less consistent traits.The extreme taxonomic importance given to the wingless seed morphological character moved Huber (1998) to classify the east African D. gilletii and the two Pyrenean endemics as the disjunct members of the small section Borderea within the large genus Dioscorea.A careful examination of the type material for D. gilletii (K! H3644/83, H3645/83) revealed several morphologically distinctive features with respect to those exhibited by Borderea (Segarra-Moragues and Catal<in 2005, unpubl.data), suggesting that D. gilletii was more closely related to other Dioscorea species than to Borderea.Recent phylogenetic studies of Dioscoreaceae (Caddick et al. 2002a) based on rbcL, atpB, 18S ribosomal DNA (rDNA) sequences, and morphological characters demonstrated that Dioscorea s.l.(Huber 1998) is paraphyletic and that the dioecious Dioscoreaceae (i.e., Borderea, Tamus L.) are embedded within a cl ade of mo noecious spec ies.These findin gs, supported by evide nce that the putatively distinctive traits of these genera are not as unique as prev io usly thoug ht, prompted the inc lusion of Borderea and its Mediterranean sister genus Tamus, w ithin Dioscorea (Scho ls et aJ . 2001 ;Caddi ck et al. 2002b).
Regardless of its taxono mic attributio n, Borderea represents an evolutionary split from an old Diosco reaceae lineage that successfull y adapted to and co lo ni zed the centraJ region of the Py renees and, as c urrentl y circ umscribed, o nl y includes two taxa, B. py renaica Miegev.a nd B. chouardii (Gaussen) H es lot (Fig. 1).These two spec ies are e nde mic of the centra l Pyrenean and Prepyrenean mo untain ranges that present some of the longest Life spans re ported for herbaceous plants (Garcfa and Anto r 1995a.;G arcfa 1997;Garcfa et al. 2002), inc luding some indi viduaJs over 300 years o ld.Bo th taxa are di oecio us geophytes and apparentl y o nl y reproduce sexua ll y (G arcia and Antor 1995b;Garcfa et al. 1995Garcfa et al. , 2002)).Borderea chouardii is a chas mophyti c species that has been classified as " in danger of extincti on" in the A nnex II of the Habitats Directi ve of the E uropean U ni on and as "criticall y endangered " in the Spanish Red List of Endangered Nati o naJ Plants (G arcfa 1996;Varios Auto res 2000;Moreno-S aiz et aJ. 2003).It is onl y kn own from a single po pulation of approx imate ly 2000 indi vidu als (Garcfa et al. 2002), located in one of the southernmost Spani sh Prepyrenean mounta in ranges (Sope ira, Huesca province: Fig. 2) growing on limesto ne c liffs at lower elevati o ns (ca.800 meters above sea level [m .a.s.J. ]).The extre mely limited effec ti ve popul ati o n size and reduced area of occupancy (less than 1000 m 2 ) of B. chouardii, coupled with its reduced capability to coloni ze new habitats caused by its limited seed di spersal system (postcarpotropi sm), could dri ve thi s pl ant into extincti on because of either biological stochasti c events or anthropogeni c acti on.Borde rea pyrenaica, tho ugh mo re widespread than its congener B. chouardii, is confined to a narrow geographic area of 160 km 2 in the centraJ Pyrenean and Prepyre nean regio n (Fig. 2), w here it inhabi ts mo bile calcareo us screes above 1800 m. a.s.l.Its popul ati ons are di stributed in three main mo unta in " islands," the largest one expa nding aro und th e Mo nte Perdi do massif in the Pyre nean axial divide and two mo re reduced populati o n cores located in the Pre pyre nean Coti e ll a and Turb6 n massifs, res pecti vely (Fig. 2).The geographical map di stances among the po pulations of B. pyrenaica growing a lo ng the Pyre nean range are short (less th an 15 km), however, they are separated by some of the hi ghest peaks of thi s mountain chai n.These peaks constitute naturaJ barriers between the more abundant Spani sh popul atio ns on the southe rn side of the Pyre nees (around the Ordesa and Pineta valleys) and the more sporadi c po pulati o ns that grow to the north in F rance aro und the G avarni e Valley.By contrast, the Prepyrenean po pul ati o ns of B. py renaica are located further away, in two isolated Spani sh mo untain mass ifs that are separated by deep valleys and by more than 30 km (Coti e ll a) and 50 km (Tur-b6n) from the Pyrenean core, re pecti vely.T hese geographic fea tures c urrentl y prevent any gene fl o w between the Pyrenean and the Prepyre nean po pulati o n cores.On the other hand, the southernmost Prepyrenean B. pyrenaica popul ati o n of 1\.1rb6n is located at a lmost the sam e geographi c di stance from its conspecific Prepyre nean core at C oti e ll a (20 km), as fro m its congener B. chouardii at Sopeira (25 km ) (Fi g. 2).Po pul ati ons of B. pyrenaica are somewhat larger-some comprising mo re than 10,000 reproducti ve indi v iduals that inhabit wide, a lmost pri stine hi gh-mountai n areas.Yet, despite its restricted geographic di stribution, popul ations of B. pyrenaica are less threate ned fro m intrinsic o r extrin sic factors than B. chouardii.
The two species are di vergent for several morpho logical characters related to the size and shape of the fruit and the seeds, the thickness and colo r of the leaf, and th e shape of the leaf apex (G aussen 1952) (Fig. 1).They are al so geographicall y separated (Fig. 2) a nd show distinct eco logicaJ prefe rences (G aussen 1952, 1965 ).However, they share a close morphology (Fig .1) that moved some authors to specul ate about the taxonomic di stinctness of B. chouardii fro m its congener B. pyrenaica, suggesting that B. chouardii could be a subspecies of B. pyrenaica (Burkill 1960).However, no formal proposals for taxonomic change were suggested.The scarcity of available material for B. chouardii for comparative studies (i.e., collection is prevented by Spanish laws) contributes to their uncertain taxonomic status, although a monographic study of the genus is currently under way (Segarra-Moragues and Catalan 2005, unpubl. data).
Previous molecular studies based on allozymes conducted on six populations of B. pyrenaica and on the only known population of B. chouardii detected very low levels of genetic variability in these taxa (Segarra-Moragues and Catalan 2002).Nonetheless, the greatest genetic distances were those between B. chouardii and all the B. pyrenaica populations, but the relationships among B. pyrenaica populations could not be ascertained with confidence due to the low levels of polymorphism detected by these markers.To address this issue, a further population genetic analysis of B. chouardii and B. pyrenaica was conducted using highly variable random amplified polymorphic DNA (RAPD) markers (Segarra-Moragues and Catalan 2003).This study revealed a strong molecular distinctness for the two taxa and allowed a molecular characterization of most studied individuals.However, in spite of the larger amounts of genetic diversity detected within B. pyrenaica, very few population-unique bands were detected, resulting in an intermingled hierarchy of RAPD phenotypes.These results were interpreted as being the consequence of historical events, supporting a recent post-glacial expansion evolutionary scenario for the B. pyrenaica populations, rather than the homogenizing effect resulting from present-day gene flow among populations, which we presumed should be low in view of the large geographical distances, natural barriers that separate the three main population cores (Fig. 2), and related biological factors such as the type of pollination vectors in B. pyrenaica (mainly ants; Garcia et al. 1995) and the limited seed dispersal capability.
Simple sequence repeats (SSRs; Tautz 1989) are codominant markers of the nuclear genome that are useful for population genetic studies (Degen et al. 1999;Naito et al. 1999;Perera et al. 2000;Sun et al. 2001;Al-Rabab'ah and Williams 2002), molecular identification of closely related taxa or populations (Bruschi et al. 2000;Macaranas et al. 2001 ), and estimation of dates of origin of hybrid species (Welch and Rieseberg 2002).Different types of microsatellite alleles have been broadly used in plant genomic analyses (Morgante and Olivieri 1993;Wang et al. 1994), even for the assessment of genetic relationships between wild relatives and derived cultivars of species of agronomic interest (Anthony et al. 2002;Hormaza 2002;Palombi and Damiano 2002).The use of microsatellites, however, is still limited among Dioscoreaceae where they have only been used in the characterization of individuals of the wild yam species Dioscorea tokoro Makino (Terauchi and Konuma 1994) and for the characterization of germplasm stocks of the white yam (D. rotundata Poir.) (Mignouna et al. 2003).Comparative studies between RAPD and SSRs demonstrated better performance of the latter markers in detecting the genetic structure of populations and in providing a higher number of polymorphisms able to characterize close species, infraspecific taxa, populations, individuals, and even clonal plant sports (Bech-er et al. 2000;Mengoni et al. 2000;Staub et al. 2000;Palombi and Damiano 2002;Mignouna et al. 2003).
In order to gather further information on the genome divergence and past evolutionary histories of these two species, we assessed the genetic differentiation and population structure of the palaeoendemic Borderea taxa through SSR analysis and compared these results to the data previously obtained from allozymes and RAPD markers.Since conservation resources are often limited, the identification of B. chouardii as an independent taxonomic entity and evolutionary lineage from B. pyrenaica was imperative.The highly variable codominant single-locus SSR alleles could also help to unravel other biological features of Borderea that have passed undetected in our previous molecular surveys.Because of the potential risk that the use of a single molecular marker could result in misleading data and the benefits derived from the performance of combined studies with congruent molecular markers, we conducted a further combined analysis of RAPD and microsatellite markers in the Borderea populations with the intention of obtaining a better picture on the genomic characteristics of the studied taxa.A further goal of our investigation was to evaluate the reliability of the SSR markers in resolving the population structure of B. pyrenaica compared to that obtained from RAPDs, in order to perform a large-scale population genetic study of this taxon.

Population Sampling, DNA Extraction
The present study was conducted on the same populations and individuals previously analyzed for allozyme and RAPD markers (Segarra-Moragues andCatalan 2002, 2003).A total of 407 individuals collected from seven populations of Borderea were included in the survey (Fig. 2).The ratio of male to female of 1: 1 was kept in the original sampling scheme.Sampling included the only known population of B. chouardii (BcOl: Sopeira, Huesca, Spain, n = 47), and six populations of B. pyrenaica (60 individuals each) distributed along its geographical range.Four of the B. pyrenaica populations are located in the Pyrenean axial divide; two of them occur on the northern face of the Monte Perdido massif (BpOl: La Planette, Gavarnie, France; Bp02: Les Rochers Blancs, Gavarnie, France), and the other two grow on the southern face of this mountain range (Bp03: Pineta, Huesca, Spain; Bp04: Ordesa, Huesca, Spain).The remaining B. pyrenaica populations inhabit the more distant Prepyrenean massifs (Bp05: Cotiella [La Vasa Mora], Huesca, Spain; Bp06: Turb6n, Huesca, Spain).
Fresh leaves from all sampled individuals were dried in silica gel and used for DNA isolation.DNA was extracted following the hexadecyltrimethylammonium bromide (CTAB) protocol of Doyle and Doyle ( 1987) adapted for miniprep extractions.DNA concentration was calculated by comparison to marker VII (Roche, Barcelona, Spain) concentration on agarose gel.Samples were diluted to a final concentration of ca. 5 ng/f.Ll in 0.1 X TE buffer and used for further DNA amplifications.

RAPD and SSR Amplifications
The RAPD analyses corresponded to that described in Segarra-Moragues and Catalan (2003).In brief, 12 RAPD primers out of 40 assayed (Operon Technologies, Alameda, California, USA, kits A and B) in a previous pilot study were selected for the screening of all studied individuals.Amplifications were carried out in 20 f.d total volume containing 1 X buffer (Ecogen, Machynlleth, Powyes, UK), 2.5 mM MgC1 2 , 0.4 mM of each dNTP, 4 pmoles of primer, 1.0 unit of Taq DNA polymerase (Ecogen), and 2 ng template DNA.The amplification program consisted of an initial step of DNA melting of 4 min at 94°C, followed by 40 cycles at 94°C for 1 min, 39°C for 1 min, and 72°C for 1.5 min, followed by an elongation step of 72°C for 7 min.The amplified products were resolved in 2% agarose gels stained with ethidium bromide; electrophoresis was set at lOOV during 4 hr in 0.5X TBE buffer.RAPD bands were visualized with UV transmitted light and captured with Gel Doc 1000 (BioRad, Hercules, California, USA).RAPD amplifications were repeated at least twice in order to check the reproducibility of the banding profiles.Ten individuals of B. pyrenaica failed to produce RAPD amplicons and the number of samples was reduced accordingly for these markers as follows (BpOl, n =58; Bp02, n =56; Bp03, n =58 and Bp05, n = 58).
The SSR analyses were based on the previous loci-characterization surveys conducted by Segarra-Moragues et al. (2003, 2004).Enriched genomic libraries in trinucleotide (CTT) motifs were separately constructed for B. chouardii (Segarra-Moragues et al. 2003) andB. pyrenaica (Segarra-Moragues et al. 2004).A total of 10 and 7 primer-pairs were designed to amplify the corresponding microsatellite regions in B. chouardii and B. pyrenaica, respectively.Transferability tests under multiplexed conditions were then assayed for the 17 microsatellite loci in both species resulting in successful cross amplifications for all 407 studied individuals (Catalan et al., unpubl. data).PCRs (polymerase chain reactions) were performed in 20 f.d reactions containing 3-5 pmoles each of the fluorescein labeled forward and unlabelled reverse primers, 1 X Taq buffer (Promega, Barcelona, Spain), 2 mM MgCl 2 , 0.4 mM of each dNTP, and 1 unit of Taq DNA polymerase (Promega), and approximately 5-8 ng DNA.The PCR program for the transferred loci consisted of an initial melting step (94°C, 4 min) followed by 30 cycles (94°C, 45 sec; annealing temperature [55-60°C], 45 sec; and 72°C, 1 min-1min 20 sec) and a final extension step (72°C, 7 min).PCR conditions for the loci developed and amplified in each separate source species are described in Segarra-Moragues et al. (2003) for B. chouardii and in Segarra-Moragues et al. (2004) for B. pyrenaica.Products were run on an ABI 310 automated DNA sequencer (Applied Biosystems, Madrid, Spain).Fragment lengths were assigned with GENESCAN and GENOTYPER software (Applied Biosystems) using ROX-500 as the internal lane standard.

Data Analysis
Borderea was recently discovered to be a tetraploid genus on the basis of molecular microsatellite data (Segarra-Moragues et al. 2003, 2004) and the present study (Fig. 3; see also Discussion); this is in contrast to previous chro-mosome counts by Heslot (1953) that suggested Borderea could be a diploid taxon with 2n = 24.Allelic inheritance analyses are currently underway (Catalan et al. unpubl. data) in order to determine whether disomic (amphidiploid) or tetrasomic (autotetraploid) inheritance occur in Borderea; thus, information on genotypes of individuals is not yet available.Microsatellite alleles have been coded as binary data in polyploid plant species (Mengoni et al. 2000) b~sed on t~e irr:possibility of distinguishing duplex and simplex d1allehc combinations and different triallelic combinations in some cases.The linear combination of presence/absence of codominant SSR bands generates phenotypic binary microsatellite patterns for each individual that can be compared with similar binary patterns obtained from the dominant RAPD markers (Mengoni et al. 2000;Staub et al. 2000;Hormaza 2002;Palombi and Damiano 2002).We applied the binary code system ( 1/0) to both the distinct RAPD phenotyp_es obtained from the 12 selected RAPD primers and the different SSR phenotypes resulting from the scoring of alleles at the 17 microsatellite loci across all investigated individuals.SSR bands that presented the same electrophoretic mobility were assumed to be homologous based on the success of all attempted cross-amplifications carried out between the two closely related Borderea species and the conserved trinucleotide changes observed between alleles of the same loci.Both separate (SSR, RAPD) and combined (SSR + RAPD) data matrices were constructed and used for further genetic analyses using different computer programs.
Genetic distances between phenotypes were calculated through several metric distances.The simple matching (SM) metric based on shared presences and absences of bands was used to compute distances between SSR phenotypes.Dice's (D) and Jaccard's (J) similarity coefficients both excluding shared absences of bands and the pairwise difference (PD) distance (Excoffier et al. 1992) were used to compute distances between SSR and RAPD phenotypes in the separate data matrices and in the combined one.SM, D, and J coefficients were calculated with NTSYSpc vers.2.11a (Rohlf 2002) and the PD distance was computed with ARLEQUIN vers.2.000 (Schneider et al. 2000).Correlation between these metrics was calculated through a Mantel test with 1000 replicates (Mantel 1967) using NTSYSpc.Genetic distances between RAPD phenotypes based on D, J, and PD showed significant high correlations among them (Segarra-Moragues and Catalan 2003).These metrics and the SM metric were also highly correlated when analyzing the SSR phenotypes (PD/J r = -0.992P < 0.001; PD/D r = -0.981P < 0.001; PD/SM r = -0.993P < 0.001; J/D r = 0.992 P < 0.001; J/SM r = 0.987 P < 0.001; D/SM r = 0.991 P < 0.001) and the combined data matrix of RAPD + SSR phenotypes in B. pyrenaica (PD/J r = -0.980P < 0.001; PD/D r = -0.973P < 0.001; J/D r = 0.998 P < 0.001).As all coefficients were highly correlated with each other the pairwise difference distance was chosen for subsequent analyses.

Population Structure
The genetic structure of the taxa and populations of Borderea was first studied through the analysis of the molecular variance (AMOVA; Excoffier et al. 1992) using ARLE-QUIN.Although AMOV A was originally designed for re-  striction fragment length polymorphism (RFLP) haplotypes it has been widely used to analyze binary coded phenotypes (i.e., RAPDs, cf.Steward and Excoffier 1996;Gabrielsen et al. 1997;Martinet al. 1997;Palacios and Gonzalez-Candelas 1997;Wolff et al. 1997;AFLP, cf. Palacios et al. 1999;and SSRs, cf. Bruschi et al. 2000).AMOV A analysis was performed at different hierarchical levels within Borderea: (i) all samples considered as belonging to the same species (Borderea s.l.; cf.Burkill 1960) The relationships among all SSR phenotypes were visualized by a neighbor-joining (NJ) tree constructed with MEGA vers.2.0 (Kumar et al. 2001) in which statistical robustness of the groupings was assessed by a 1000-replicates bootstrap analysis (Felsenstein 1985) using PAUP* vers.4.0 beta 10 (Swofford 2002), and by multivariate prin-cipal coordinate analyses (PCO) conducted with NTSYSpc (Rohlf 2002).Three different approaches were performed in the PCO analyses: (i) with the whole set of samples, to visualize the multidimensional relationships of the SSR phenotypes of both taxa; (ii) with a subset of the matrix, containing only the SSR phenotypes of B. pyrenaica, to search for differences in the molecular spatial distribution of phenotypes among populations of this taxon; and, (iii) with the combined SSR + RAPD data matrix to assess the consistency of the spatial distribution of populations depicted by the two sorts of molecular markers in B. pyrenaica.The results rendered by these analyses were compared with those reported in the previous RAPD survey (Segarra-Moragues and Catalan 2003).Genetic distances based on pairwise FsT statistics between populations were used to construct unweighted pair-group method with arithmethic averaging (UPGMA) phenograms using NTSYSpc and bootstrapped with POPU-LATIONS vers.1.2.28 (Langella 2000).Correlations between genetic and geographic distances between populations were assessed by means of a 1000 replicates Mantel test using NTSYSpc.

Relationships Between Borderea Phenotypes
As stated in Segarra-Moragues and Catalan ( 2003), the 12 RAPD primers generated 112 bands, of which only four Table I.Alleles found in B. chouardii and B. pyrenaica for the 17 SSR loci studied.For each species and locus, the number of alleles (NA) and the allele sizes in Bp.In bold, the alleles common to both species.
The microsatellite genetic study of Borderea s.l.conducted here for the first time, detected similar levels of genetic variability to that provided by the RAPD markers; however, the resolving power of these SSR markers was notably higher than that shown by the less reliable RAPD markers.The 17 reproducible SSR loci detected a total of 172 bands (alleles) across the 407 studied individuals.Thirty out of 65 bands present in B. chouardii were exclusive to this taxon (46.15%) and, of these, three (10%) were fixed and diagnostic, separating it from its congener, whereas 107 out of 142 bands were exclusive to B. pyrenaica (75.35%), and three of these (2.83%) were diagnostic for this species.Thirty-five (20.35%) out of 172 total bands were shared between the two taxa.A summary of the SSR alleles detected in Borderea s.I. and in each of the independent species B. chouardii and B. pyrenaica is shown in Table 1.SSR markers were even more precise than RAPD markers in identifying each of the 407 individuals studied by their own SSR phenotypes.
Principal coordinate analysis of the whole SSR data matrix (Borderea s.l.) showed a complete differentiation between the B. chouardii and the B. pyrenaica phenotypes that clustered separately on the space delimited by the first two axes that accumulated 32.92% of the variance (Fig. 4), a result similar to that obtained from RAPD analysis (Segarra-Moragues and Catalan 2003).However, in contrast to the poor genetic population structuring shown by RAPDs ( cf.Segarra-Moragues and Catalan 2003), subsequent PCO analysis of SSR phenotypes restricted to the B. pyrenaica data set distinguished a clear-cut clustering among phenotypes belonging to the five geographical regions (Fig. 5).The 3D projection of the SSR phenotypes in the space defined by the first three axes that accumulated 30.94% of the variance separated the more isolated southernmost Prepyrenean population of Turb6n (Bp06) from the rest along the negative extreme of axis 1 and the positive extreme of axis 3. Other clusters corresponding to the Prepyrenean populations of Cotiella (Bp05) and the Pyrenean Spanish populations of Ordesa (Bp04) and Pineta (Bp03) are located in intermediate positions on axis 1, but separate along positive to negative positions on axis 2. Finally, a mixed cluster of phenotypes from the French Pyrenean populations of Gavarnie (Bp01 and Bp02) differentiated along the positive extreme of axis 1 (Fig. 5).Phenotypes of the two French northern populations of the Pyrenees partially intermixed in their common cluster due to the close proximity of these populations, which is less than 3 km apart.The spatial pattern shown by the B. pyrenaica population clusters in this molecular PCO space is in agreement with their geographical distribution (Fig. 2).
The unrooted NJ tree constructed from pairwise distances between the 407 SSR phenotypes also revealed the differentiation of two main clusters (Fig. 6) that corresponded to phenotypes of B. chouardii and B. pyrenaica, respectively, with branch divergence showing 100% bootstrap support.In contrast to previous results based on RAPD analysis (Segarra-Moragues and Catalan 2003), hierarchy of SSR phenotypes across the six studied populations of B. pyrenaica was resolved in this NJ tree (Fig. 6).Most of the phenotypes from each of the studied populations of B. pyrenaica joined in separate clusters indicating a certain degree of genetic isolation although their respective branches were not supported.Several phenotypes of the French populations from the north side of the Pyrenees (Bp01 and Bp02) appeared intermingled in a less differentiated cluster, a direct consequence of their close geographical proximity and the likely existence of present day gene flow between them.On the other hand, some phenotypes from the Ordesa population (Bp04), from the south side of the Pyrenees, clustered together with those from the French north side (Bp01 and Bp02), whereas the vast majority of the remaining phenotypes clustered with the other Spanish population of Pineta (Bp03) on the south side of the mountains.Phenotype clusters corresponding to the Prepyrenean populations of B. pyrenaica at Cotiella (Bp05) and Turb6n (Bp06) shared less genetic affinities to the Pyrenean population cores (BpO I, Bp02, Bp03, and Bp04), paralleling their geographically isolated distribution (Fig. 2, 5).The southernmost Prepyrenean population of Turb6n (Bp06) showed a basal diverging clustering within the B. pyrenaica group and represented the population of this taxon most similar to that of the congener B. chouardii.
Multivariate PCO analysis was also conducted on the combined SSR + RAPD data matrix (results not shown); the tridimensional plotting of phenotypes in the space defined by the first main axes was less hierarchically structured than that obtained from SSRs.The same spatial differentiation pattern was observed for the clusters of phenotypes corresponding to the Pyrenean and Prepyrenean populations of B. pyrenaica in that projection as in the one obtained from the SSR markers (Fig. 5).However, in the latter, the first three axes accumulated a lower percentage of variance (23.81%) and the separation among clusters was not as neat.This lack of resolution was caused by the inclusion of a poorly resolved set of RAPD phenotypes (Segarra-Moragues and Catalan 2003) into the combined data matrix.A similar loss of hierarchical resolution was observed in the NJ tree based on the combined data set (results not shown) indicating that the RAPD markers are less valuable in differentiating the genetic structure of recently diverged populations in contrast to the powerful discriminating value demonstrated by microsatellites.

Population Genetic Structure
Partitioning of genetic variance within Borderea was obtained through AMOV A analysis (Table 2).The genetic differentiation between B. chouardii and B. pyrenaica previously detected by RAPD markers was corroborated by the statistical analysis of the SSR phenotypes.The first distribution analysis attributed 48.52% of the variance to differences among populations of Borderea s.l.when all samples were considered to be one species, indicating a strong heterogeneity in that group.This was further confirmed when the samplings were treated as separate species in the second analysis (B. chouardii vs. B. pyrenaica), the differences among taxa accumulating 48.99% of the total variation, whereas differences among populations and within populations were only of 18.73 and 32.28%, respectively.The FsT values for the RAPD and SSR differentiation between the two species were highly significant (0.79 and 0.68, respectively; P < 0.001) in both cases.
AMOV A analyses conducted at different hierarchical levels within B. pyrenaica always revealed higher genetic diversity within populations than either between populations or between geographical regions regardless of the data matrix used for comparison (Table 2; SSRs, RAPDs).However, the SSR markers always detected lower levels of genetic variability within populations, but higher levels of genetic differentiation among populations and among regions than the RAPD markers in all assayed cases, indicating a more conserved nature and pointing toward their suitability for a better assessment of the genetic relationships among recently diverged populations.
The lowest value for the percentage of genetic variation accumulated among populations from the same area (15 .14, 8.80, and 12.63%, for the SSRs, RAPDs, and combined analyses, respectively) was that obtained when the populations were divided into four geographical ranges (northern Pyrenees [BpOl,Bp02] vs. southern Pyrenees [Bp03, Bp04] vs. Prepyrenean Cotiella [Bp05] vs. Prepyrenean Turb6n [Bp06]), suggesting a close genetic relationship and homogeneity of populations from the same region.At the same time, the highest percentage of partitioning of variance among regions also was obtained for this subdivision of populations, both for the SSRs (23%) and the combined analyses (17.28% ).Whereas for RAPDs, the highest value of divergence among regions (8.59%) was obtained when the southernmost Prepyrenean population of Turb6n (Bp06) was considered separate from the rest.These results indicated that SSRs are more precise in depicting genetic relationships among closely related populations.Nonetheless, the two molecular markers are coincident in showing the populations of the southern Prepyrenean ranges as the most genetically distant and those from the northern Pyrenean range as the most recently derived.

Genetic and Geographical Distances between Taxa and Populations
Genetic distances among populations of Borderea were based on FsT values, the analogue of <PsT values, calculated from the Euclidean Distance.The FsT coefficients were used  to construct UPGMA phenograms based on both SSR and RAPD distance matrices.The two types of molecular markers were concordant in depicting the same linkage among populations (Fig. 7); however, patristic distances and levels of bootstrap support were higher in the SSR phenogram than in the RAPD phenogram.In both cases, the greatest distances were those observed between the single population of B. chouardii and all six populations of B. pyrenaica (ca.0.7 for SSRs and >0.8, RAPDs).These results were supported in both clusterings by full bootstrap percentages ( 100% ).Distances between populations of B. pyrenaica ranged from moderate (>0.4,SSR) to low (<0.2,RAPD) values.The two closest populations were those from the northern Pyrenees at Gavarnie (BpOl, Bp02), which are also geographically close and show strong bootstrap support (I 00% SSR and 96% RAPD).Another tight cluster was composed of the southern Pyrenean populations (Bp03, Bp04), which are well supported in the SSR clustering (71%) and, in tum, linked to the northernmost Prepyrenean population of Cotiella (Bp05).The greatest distance (ca.0.42, SSRs; ca.0.20, RAPDs) was that between the southernmost Prepyrenean population of Turb6n (Bp06) and the rest of the B. pyrenaica populations.The genetic affinities among populations of B. pyrenaica are mostly concordant with geographic distances.Nevertheless, populations from the French side of the Pyrenees (Bp01 and Bp02) appear to be less closely related to those of the Spanish side (Bp03 and Bp04) although the map distance between them is shorter than that of the latter populations to Prepyrenean population Bp05.This suggests a certain degree of isolation and reduced gene flow between the two sides of the Pyrenean axial divide due to geography.The Mantel correlation test between genetic and geographic distances computed for B. pyrenaica populations showed significant values of 0.71 (P < 0.01) and 0.68 (P < 0.05) for SSR and RAPD markers, respectively, indicating that some populations located in closer geographic proximity are genetically less closely related than other populations that are geographically further apart, but that show higher genetic affinities.These results point toward past climatic oscillatory changes and geography as the main factors intervening in the postglacial colonization pathways followed by the populations of B. pyrenaica and in the maintenance of their present genetic relationships.

Microsatellite Markers and the Attributes and Relationships of Genus Borderea with Respect to other Dioscoreaceae
Comparative studies from a large pool of different molecular markers are the most accurate way to test the validity of potential evolutionary scenarios for any group of living organisms (A vise 1994 ).Of the three molecular marker systems assayed in Borderea (allozymes, RAPDs, SSRs; cf.Segarra-Moragues andCatalan 2002, 2003;Segarra-Moragues et al. 2003, 2004, and the present study) microsatellitesdue to their codominant nature and their capacity to detect high, but stable levels of polymorphism-have provided the most robust data set to investigate the past evolutionary history of the Borderea taxa and populations.These properties qualify SSRs as the best molecular tools for detailed finescale evolutionary and taxonomic investigations of little-differentiated populations of Dioscoreaceae.Nonetheless, microsatellite data are mostly congruent with allozyme and RAPD data in depicting a similar evolutionary scenario for the ancestral Pyrenean yams.SSR markers have also rendered unique clues to decipher the mechanisms involved in the origin of these relictual species and their genetic background.
In two initial SSR assays conducted on single populations of B. chouardii (Segarra-Moragues et al. 2003) and of B. pyrenaica (Segarra-Moragues et al. 2004) each separate set of loci showed individuals with up to four alleles.Cross amplifications of these 17 loci along all 407 studied individuals have confirmed the previous findings (Fig. 3), thus reaffirming the tetraploidy of the two Borderea taxa.This is the first record concerning the polyploid nature of this relict Pyrenean genus.The two species of Borderea were considered to be diploid by Heslot (1953) who counted 2n = 24 and n = 12 chromosomes in B. pyrenaica and 2n = 24 chromosomes in B. chouardii.We assumed Borderea was diploid and had a chromosome base number of x = 12, close to that presented by most Dioscorea taxa (x = 1 0), and used this karyological character as a further criterion to distinguish Borderea from Dioscorea.No other cytogenetic studies have been conducted in Borderea since those of Heslot (1953), and all later authors have accepted x = 12 as the chromosome base number of the Pyrenean yams (Burkill 1960;cf. Gaussen 1965;Huber 1998).On the other hand, the Mediterranean genus Tamus, with a chromosome number of 2n = 48, was believed to be a tetraploid taxon that shared with Borderea the chromosome base number x = 12 indicative of a close evolutionary relationship between them (Burkill 1960;Huber 1998).Other karyological surveys of Dioscoreaceae have shown that the most common chromosome base numbers in the family are x = 10, present in nine paleotropical sections of Dioscorea and in the holarctic section Macropoda Uline, and x = 9, present in four out of nine tropical sections of Dioscorea and in Rajania L., whereas the American D. mexicana Scheidw., shows x = 8 (Huber, 1998).
An immediate conclusion from our microsatellite survey is that if Borderea is a tetraploid genus with 2n = 24 chromosomes, then its chromosome base number is not x = 12, but x = 6.Up to now x = 6 is the smallest chromosome base number recorded in the mostly pantropical Dioscoreaceae (Burkilll960; Dahlgren eta!. 1985;Huber 1998).However, Borderea may well represent a case of a secondary base number (x = 12) of polyploid derivation (Stebbins 1971) where the original base number x = 6 was doubled by tetraploidization giving rise to the gametic number n = 12 observed by Heslot (1953) and interpreted as a functionally x = 12.
Further insights into the genomic inheritance of nuclear chromosome markers in Borderea have also been obtained from the analysis of microsatellite alleles.A more detailed statistical study on SSR inheritance patterns in the Borderea species is presently underway.However, for most of the studied SSR loci both B. pyrenaica and B. chouardii show predominant duplicate disomic inheritance (Catalan et a!.unpubl.data) reinforcing the hypothesis of a hybrid origin of these polyploid taxa (Segarra-Moragues and Catalan 2002).Fixed heterozygous microsatellite profiles as well as variable, but cosegregating allelic patterns in SSR loci are concordant with previous findings based on fixed heterozygous patterns for some allozyme loci (PGI-2, IDH) in the likely existence of a past hybridization event that resulted in the present known genus Borderea.Amphipolyploidy is recognized as the more common polyploidization mechanism in flowering plants (Stebbins 1950(Stebbins , 1956(Stebbins , 1971;;Stace 1987;Soltis andSoltis 1993, 1999) and is of special relevance in relictual lineages of many families of angiosperms (Stebbins 1971;Soltis and Soltis 1993).Borde rea fits well within an archaic amphidiploid scenario whereas other polyploid Dioscorea taxa (with several multiples of 2n = 10) seem to have had a more recent origin-especially those concerned with the highly polyploid cultivated yams (Huber 1998).
The phylogenetic studies of Caddick et a!.(2002a) based on analysis of rbcL sequences demonstrated that polyploid Mediterranean Tamus was the closest relative of the Pyrenean endemic Borderea.According to these results, and based on our present knowledge about ploidy levels and inheritance patterns in Borderea, we speculate that these two sister genera could also have a common amphidiploid origin derived from common ancestors with x = 6.Thus, tetraploid Borderea (2n = 24) and octoploid Tamus (2n = 48) could constitute some of the oldest extant paleopolyploid lineages of Dioscoreaceae.
Borderea and Tamus are sympatric in the Pyrenees; these two ancient genera belong to the same biogeographical Mediterranean region, although Tamus is widespread in the pan-Mediterranean area, whereas Borderea is restricted to the central Pyrenean zone.Chloroplast sequence data, karyological evidence, and our present microsatellite data reinforce the hypothesis of a common hybrid origin of these two genera from Tertiary Dioscoreaceae ancestors that likely became extinct with the advent of glaciation, but whose descendants survived as newly arisen polyploids in Mediterranean refugia during the coldest periods of the Quaternary and successfully adapted to the newly deserted areas during the postglacial warming era.
Whereas the close relationship of Borderea and Tamus is undisputedly supported by various data sources in spite of some apparent morphological differences-the twining habit and berry fruit shown by Tamus in contrast to the non-twining dwarf habit and capsule fruit present in Borderea-the purported close affinity of Borderea to other taxa of Dioscoreaceae seems to be less reliable.Thus, the attributed affinity of the Borderea taxa to the east African Dioscorea gilletii, classified under the same sect.Borderea of a large genus Dioscorea s.l. by Burkill ( 1960) and Huber (1998), and their putative past ancestry from a pan-Thetyan tropical Dioscoreaceae lineage (Burkill 1960) are likely spurious.New phylogenetic analyses based on chloroplast sequence data (rbcL, matK) (P.Wilkins and P. Scholz pers.comm.)indicate that Borderea and Tamus are sister group to an isolated Mediterranean clade whereas D. gilletii is nested within an independent African Dioscorea clade.Similarly, the presumed closeness of Borderea to the Andean endemic orophyte genus Epipetrum and their hypothesized origin from a common pan-Atlantic Dioscoreaceae ancestor (Braun-Blanquet 1948) are not sustainable based on current biogeographical knowledge (even with the lack of molecular data for Epipetrum).Thus, the crucial shared morphological characters that most frequently have been used for classification purposes-such as, the possession of unwinged seeds, a feature shared by Borderea, D. gilletti, and Epipetrum, as well as by Tamus and Nanarepenta Matuda (cf.Caddick et a!. 2002b), and the acquisition of a dwarf mountain habit, common in Borderea and Epipetrum and, to a lesser extent, in D. gilletii as well as in other North American Dioscorea species (cf.Burkill 1960)-appear to be the consequence of parallel convergent evolutionary processes leading to independent Dioscoreaceae lineages in different geographical regions, and probably, at different evolutionary times.

The Species of Borderea and Their Evolutionary History
Microsatellite data also support the genetic distinctness of the two Borderea taxa.In concordance with previous molecular studies based on allozyme and RAPD markers (Segarra-Moragues andCatalan 2002, 2003), the SSR analyse\ have provided three new private alleles for each of B. chouardii and B. pyrenaica that account for the characterization and the separation of the two species.These markers, together with a private allozyme allele (PGMl-2) detected for B. chouardii (Segarra-Moragues and Catalan 2002) and 20 and 1 private RAPD markers detected for B. chouardii and B. pyrenaica, respectively, (Segarra-Moragues and Cat- ahin 2003) constitute a large mo lecular set of ide nti fiers fo r th ese Pyrenean endemics.Reconstructi o n of phenotypic relati onshi ps a mong popul ati ons of bo th species and di stancebased methods conducted with allozy mes, RAPD, and SSR markers always discriminate B. chouardii fro m B. py renaica as two di stinct geneti c entities.Thus, mo lecul ar evide nce based o n three different data sets indi cates an o ld di vergence of the two Borderea taxa and favo rs the ir taxo no mi c recogni ti o n as separate species.In contrast to the taxono mic treatment proposed by Burkill ( 1960) and in agreement w ith that defended by Gaussen ( 1952Gaussen ( , 1965)), Bo rde rea choua rdii should be considered a di ffere nt species f rom B. pyrenaica and not a me re subspecies of it.
The molecul ar data corre late well with the morpho logica l traits in the disti ncti on of the two taxa.Apart from traditi ona l features used to d ifferenti ate B. chouardii fro m B. pyrenaica (Gaussen 1952; Fig. 1), a careful examinati o n of new materials all o wed us to find new di stincti ve tra its for these species including th e coat, shape, and co lo r of the seeds.Hence, whereas B. chouardii bears brow ni sh fus ifo rm seeds with an apical caruncle, th ose of B. pyrenaica possess dark hori zontally-compressed seeds covered by an exte nded thin-layered carunc le (Segarra-Moragues and Catal::'i.n2005 , unpubl.data) .Thus, the new morpho logical characters contribute to the identi fica ti on of B. chouardii as a singular species and posit further arguments fo r its separate c lass ificatio n fro m B. pyrenaica.
Reconstructio n of the life hi stor y of Bo rderea was first atte mpted w ith th e more variabl e RAPD markers (Segarra-Moragues and Catalan 2003) due to the low variability detected w ith a ll ozy mes (Segarra-M orag ues and Cata la n 2002); however, the two markers agree o n the old di vergence of the two Borderea spec ies and o n a recent postglaci al diversifi cati o n of the present B. pyrenaica po pulations (Segarra-Moragues a nd Catalan 2003).The mic rosatellite data a lso support thi s evoluti onary scenario fo r Borderea and add more precise data fo r a coloni zing postg lacial mi gratory route of B. pyrenaica fro m southern P repyrenean refugia towards the northern Pyre nean ranges (Fig. 8).
Patte rns a nd mode ls of pla nt evolution during the oscillatory c limati c changes of the late Tertiary and Qu ate rn ar y ages in E uro pe a nd in the Mediterranean basin are di verse (H ewitt 1996; 2000).T he re is a genera l acceptance that the no rthern territori es of Europe were mostl y covered by ice during the g laciati o ns without any trace of pote nti a l refugia (nunataks) fo r plants and that postglacia l co lo ni zati o n fo llowed a general tre nd fro m south to no rth (Gabrie lsen et al. 1997;Tollefsrud et a l. 1998).Concomi tant with the retreat and advances of the ice sheets (Hew itt 1996; Co mes and Kadereit 1998; Gutie rrez-Larena et a l.2002), the she lterin g mounta in s of southe rn E urope are be lieved to have hosted intermi tte nt phases of expansio n and contracti on of pl ant popul atio ns derived fro m locall y ascending and descending migrato ry moveme nts on the mountains.Under such c ircum-stances further scenarios have been postulated for different plant species ranging from long-term isolation processes in \ ~eographically separate refugia to recent colonization events '-t_rom single or limited shelters (Bauert et al. 1998;Taberlet et al. 1998;Zhang et al. 2001).Borderea represents a typical example of the latter case and this scenario is also supported by the more accurate microsatellite data (Fig. 8).Despite concordance of phylogeographic patterns exhibited by several angiosperm groups in the southern European mountains, molecular studies have shown that each plant group has followed its own evolutionary history.The present day distribution of species and populations could be due, in part, to factors as different as the ecological affinities of the original founders; their ability to successfully invade newly vacated areas or zones already occupied by other pioneers; and, their differential rates of extinction in sheltered areas (Comes and Kadereit 1998).Hybridization likely played a crucial role in the conquest of newly vacated areas at higher altitudes-land made available by glacial retreat-as it increased the capability to invade more inhospitable environments (Soltis and Soltis 1999;Ellstrand and Schierenbeck 2000;Zhang et al. 2001).The case of Borderea well illustrates the dramatic differences in adaptive fitness shown by two closely related species of common hybrid origin.In spite of the potential advantages acquired through hybridization, the less successful B. chouardii-due to its inefficient seed dispersal mechanisms (postcarpotropism) described by Segarra-Moragues et al. ( 2005)-was nearly rendered extinct during the glacial periods and failed to colonize new habitats at postglacial times.The SSR data also confirm the strong genetic bottlenecking experienced by the only surviving population of B. chouardii where levels of genetic variability were considerably lower than those of its congener B. pyrenaica.By contrast, the aggressive hybrid B. pyrenaica evolved into a postglacial subalpine species that successfully colonized the barren scree of the central Prepyrenees and Pyrenees where competition with other plants was less stringent and seed dispersal could be mediated by grazing animals.
The SSR data favor a stepping-stone colonization hypothesis (Hewitt 2000) for B. pyrenaica, previously envisaged through RAPD analyses (Segarra-Moragues and Catalan 2003), and add further evidence for a fine-scale reconstruction of a saltatory migration route from southern Prepyrenean to northern Pyrenean massifs (Fig. 8).The relatively low levels of genetic differentiation found among populations of this taxon, in contrast to the higher levels of genetic variation detected within populations indicate (in the absence of present gene flow between the three main Prepyrenean and Pyrenean mountain ranges) that the present B. pyrenaica populations are of recent origin and that they were likely derived from a pre-Quaternary lineage subjected to severe population declines during the coolest glacial phases.The short time that has elapsed since the beginning of the postglacial colonization and expansion of B. pyrenaica until today (ca.10,000 yr) has prevented the genetic isolation of these recently arisen populations which do not show any private SSR alleles.Previous hypotheses, based on RAPD data, pinpoint the Prepyrenees as the likely place of speciation and divergence of the two Borderea taxa and as the starting point for the B. pyrenaica migratory way to the north (Segarra-Moragues and Catalan 2003) have been confirmed and reassessed by more solid data provided by microsatellites.
From our analyses of SSR alleles (Fig. 2, 6-8) we derived a more robust scenario for the successive advances and divergences of the B. pyrenaica populations.Our research corroborates that populations from an initial stock, isolated at the southernmost Prepyrenean massif of Turb6n, diverged to form the northernmost Prepyrenean populations at the Cotiella massif, and eventually, were distributed within the Pyrenean range.Moreover, microsatellites have confirmed the presence of three relictual SSR alleles, shared with its ancestral congener B. chouardii, in the populations of the two Prepyrenean massifs and their absence in those distributed in the Pyrenean core (Table 1).The ancillary data strongly support the single south-to-north migratory cline followed by B. pyrenaica from older and first established populations in the warmer and lower latitude refugia of the Prepyrenees toward the younger and more recently settled populations that colonized the cooler and higher mountain ranges of the Pyrenees concomitantly with the gradual retreat of the glaciers.The more discrete genetic structure found in the Prepyrenean populations (Fig. 6) indicates their relative genetic isolation from those of the Pyrenees.As a consequence of the postglacial warming trends, the Prepyrenean populations were left behind during the up-slope migration toward the higher northern massifs of the Pyrenees.These populations ended geographically isolated at the highest points of the southern Prepyrenean in mountain "isles" surrounded by foothills containing new montane vegetation (Fig. 8).
According to the microsatellite data, a small-scale colonization process can be proposed for the B. pyrenaica populations that reached the main Pyrenean core from the south.The expansion along the southern side of the central Pyrenees was favored by the availability of suitable habitats, whereas northward ascents were probably impeded by the higher altitudes of the Monte Perdido massif, which presently maintains some of the few relictual glaciers of the Pyrenean chain.However, this natural barrier was likely circumvented through lower-altitudinal mountain passes that facilitated the ultimate colonization of the northern side of the Pyrenees at the Gavarnie Valley (Fig. 2, 8).The close correspondence between some SSR phenotypes of the southern-Pyrenean populations of Ordesa and those of the northern-Pyrenean populations of Gavarnie confirms this connecting migratory pathway, whereas the limited differentiation observed among the northern-Pyrenean populations indicates a recent divergence and the likely existence of present gene flow among these populations.
Park of the Pyrenees for facilitating sample collections; and D. Najar and E. Perez-Collazos for their help in processing Fig. 1 and 8.This work has been supported by an Aragon Government (DGA) research grant Pl05/99-AV toP.Catalan and by a DGA doctorate fellowship to J. G. Segarra-Moragues.

Fig
Fig. I.-Habit o f Borderea taxa : (a) B. choua rdii, male plant and deta il of fruiting branch o f fe ma le pl ant; (b) B. py renaica, ma le pl ant and deta il o f fruitin g branch o f fe male pl ant.Scale bar = I em.Drawings reproduced by courtesy of M. Saul e ( 199 1 ).

Fig. 6 .
Fig. 6.-Neighbor-joining (NJ) tree of the 407 SSR phenotypes observed in Borderea.BpOl e and Bp02 0, correspond to French populations of B. pyrenaica; Bp03 • and Bp04 D to populations in the axial ranges on the Spanish side of the Pyrenees; Bp05 .A. and Bp06 6 to populations in the Prepyrenean ranges; BcOl + corresponds to B. chouardii.Support for the grouping is indicated on the branch by the bootstrap value.

Fig. 8 .
Fig. 8.-Hypothesized Borde rea split fro m a Tertiary ancestor that gave rise to the presentl y known Borde rea taxa and pote ntial postglac ia l colo ni zation route fo llowed by B. pyrenaica from the southern Prepy renean ranges to the northern Pyrenees.Circled areas ind icate current areas occ upied by B. chou.ardii(so li d line) and B. pyrenaica (dashed li ne), respectively.