Genetic Exchange Across a Hybrid Zone Within the Iberian Endemic

Molecular Ecology (2005) 14, 245–254 doi: 10.1111/j.1365-294X.2004.02390.x

BlackwellGenetic Publishing, Ltd. exchange across a hybrid zone within the Iberian endemic golden-striped salamander, Chioglossa lusitanica

F. SEQUEIRA,*† J. ALEXANDRINO,*§ S. ROCHA,* J. W. ARNTZEN*‡ and N. FERRAND*† *CIBIO/UP-Centro de Investigação em Biodiversidade e Recursos Genéticos, Campus Agrário de Vairão, Rua Padre Armando Quintas, 4485–661 Vairão, Portugal, †Departamento de Zoologia e Antropologia, Faculdade de Ciências, Universidade do Porto, Praça Gomes Teixeira, 4099-002 Porto, Portugal, ‡National Museum of Natural History, PO Box 9517, 2300 RA Leiden, the Netherlands, §Museum of Vertebrate Zoology, University of California, Berkeley, 3101 Valley Life Science Building # 3160, Berkeley, CA 94720– 3160, USA

Abstract The study of hybrid zones resulting from Pleistocene vicariance is central in examining the potential of genetically diverged evolutionary units either to introgress and merge or to proceed with further isolation. The hybrid zone between two mitochondrial lineages of Chioglossa lusitanica is located near the Mondego River in Central Portugal. We used mitochondrial and nuclear diagnostic markers to conduct a formal statistical analysis of the Chioglossa hybrid zone in the context of tension zone theory. Key results are: (i) cline centres are not coincident for all markers, with average widths of ca. 2–15 km; (ii) heterozygote deficit was not observed across loci near the transect centre; (iii) associations of parental allele combinations (‘linkage disequilibrium’ R) were not detected either across loci or across the transect. These observations suggest that the Chioglossa hybrid zone is not a tension zone with strong selection against hybrids but instead one shaped mostly by neutral mixing. The patterns uncovered suggest a complex history of populations over a small scale that may be common in southern Pleistocene refugia. Keywords: allozymes, asymmetric introgression, Chioglossa lusitanica, cline analysis, gene flow, hybrid zone, maximum likelihood, mtDNA Received 03 June 2004; revision received 30 September 2004; accepted 01 October 2004

refugial areas. This pattern of ‘refugia within refugia’ is Introduction becoming especially apparent in the Iberian Peninsula The climatic oscillations of the Quaternary had a profound after phylogeographical studies burgeoned in the region impact on many organisms, subdividing their ranges over (Gomez & Lunt 2004 and references therein). As distinct distinct Pleistocene glacial refugia with the consequent evolutionary units are uncovered in many organisms, it is genetic diversification of evolutionary units that often came important from both evolutionary and conservation per- to meet in hybrid zones (sensu Arnold 1997) after climatic spectives to examine their degree of genetic cohesiveness amelioration (Hewitt 1996, 1999, 2004; Taberlet et al. 1998). in zones of parapatry (i.e. gene flow across hybrid zones). Complete speciation appears not be a common outcome of Here we take advantage of one paradigm in the Iberian Pleistocene vicariance but stages of incipient speciation are refugial historical biogeography, the golden-striped sala- observed at hybrid zones with several degrees of genetic mander Chioglossa lusitanica (described below), to emphasize isolation (Avise et al. 1998; Hewitt 2000). Recent work in how the study of genetic cohesiveness between parapatric western Eurasia suggests that diversification of evolution- evolutionary units is central to make predictions on alter- ary units not only occurred between the three southern native evolutionary outcomes in Quaternary diversifying peninsulas (Iberian, Apennine, Balkans) and the Caucasus taxa (e.g. lineage amalgamation vs. further isolation). but also between smaller range fragments within these The genetic cohesiveness between evolutionary units can be examined in hybrid zones where genetically distinct Correspondence: João Alexandrino, Fax: 1510 6438238, E-mail: populations meet and produce hybrids (Barton & Hewitt [email protected] 1985; Harrison 1990; Arnold 1997). Selection may act on

246 F. SEQUEIRA ET AL. new combinations of alleles generated by recombination existence of a hybrid zone between the northern and southern (Harrison 1990, 1993). This kind of interaction often results forms with some mixing, but sampling was not sufficient in the formation of clines at different loci or phenotypic to describe the genetic structure of the zone in detail traits, the position and width of which are determined by (Alexandrino et al. 2000). a balance between dispersal into the zone and any counter The purpose of our work is to undertake a formal selection acting on hybrids. These cline shapes are rela- population genetic analysis of the hybrid zone between the tively independent of the type of selection acting against two evolutionary units of C. lusitanica. Here, we apply hybrids which is usually unknown (endogenous, exogen- maximum likelihood tension zone models (Szymura & ous or both; Kruuk et al. 1999). The ‘coincidence’ of the cline Barton 1986, 1991; Barton & Gale 1993; Phillips et al. in centres and the ‘concordance’ of cline widths are expected press) to quantify cline width, cline concordance/coincidence when a hybrid zone first forms from secondary contact. and genotypic disequilibria at five diagnostic loci (four After several generations of hybrid formation and back- allozymes and mtDNA) across a transect perpendicular crossing through continued dispersal into the zone, selec- to the Mondego River. We will use the above hybrid zone tion against hybrid genotypes, epistatic effects between properties not only to learn about the genetic differences loci and genome-wide linkage disequilibria will generate a between the populations involved but also to investigate if steep step in allele frequencies that will result in coincidence a barrier to gene exchange (i.e. hybrid counter selection) and concordance among clines for independent traits (i.e. between differentiated gene pools is in place. If selection tension zones; Barton 1983; Barton & Hewitt 1985; Barton is operating against hybrids, we expect to observe cline & Gale 1993; Gavrilets 1997; Barton & Shpak 2000). If selec- widths that are narrow relative to dispersal and significant tion against hybrids is strong, the centre of the zone will act genetic disequilibria in the centre of the zone. as an effective barrier to the introgression of both nega- tively selected and neutral alleles, but positively selected Materials and methods alleles could still cross such a barrier (Piálek & Barton 1997). A very different multilocus cline shape pattern can arise if Collection of samples selection is acting only weakly on loci, which will free clines to act independently of each other rather than as sin- Sampling was carried out in central Portugal within and gle nonrecombining unit. Each cline may then respond to around the putative contact zone between the southern independent forces such as locus-specific selection and/or and the northern population groups of Chioglossa lusitanica drift with the resulting pattern being noncoincidence (Alexandrino et al. 2000). This area (ca. 50 × 60 km) is located and nonconcordance of clines (Barton & Bengtsson 1986; in the southern part of the range and comprehends the Harrison 1990; Barton & Gale 1993; Searle 1993). A narrow mountain ranges of Buçaco (North), Lousã (West), Muradal hybrid zone with coincident and concordant clines and (South) and Estrela (East), and is crossed by two main rivers strong selection against hybrids is more likely to effectively — Mondego and Zêzere (Table 1 and Fig. 1). Salamanders maintain isolation and further promote divergence between were collected at 14 sites between 1998 and 2003 and diverged forms than a hybrid zone with wider clines asso- the tail end of each individual was removed and frozen at ciated with neutral mixing. −70 °C for genetic analysis. Sample size ranged from 15 to The golden-striped salamander, Chioglossa lusitanica 47, including samples reported upon earlier (sites 1, 3, 5, 7, (Salamandridae), is an Iberian endemic species in which at 11, 12 and 13; Alexandrino et al. 2000). least two local refugia have been identified from protein and mitochondrial (mt) DNA data. The salamander in- Markers habits the banks of small streams over mountainous areas characterized by high rainfall and mild summer and Allozyme electrophoresis. We conducted allozyme electro- winter temperatures (Arntzen 1981; Teixeira et al. 2001). phoresis of products encoded by the seven allozyme loci Groups of populations at either side of the Mondego River reported to be polymorphic in C. lusitanica by Alexandrino in central Portugal are genetically distinct (Alexandrino et al. (1997, 2000): alcohol dehydrogenase (Adh), phospho- et al. 2000, 2002) and have slightly different ecological pro- glucomutase (Pgm-1), phosphogluconate dehydrogenase files, with climatic conditions in the south that is harsher (Pgd) and four peptidases (Pep-A, Pep-B, Pep-C and Pep-D). than in the north (Arntzen & Alexandrino 2004). Divergence Only four of these (Adh, Pep-C, Pep-D and Pgm-1) were time estimates from genetic data (Alexandrino et al. 2000) diagnostic between northern and southern C. lusitanica in combination with the extrapolation of bioclimatic and were therefore useful for hybrid zone analysis. models (Teixeira et al. 2001; Teixeira & Arntzen 2002) suggest that the two evolutionary units may have been separated RFLP (restriction fragment length polymorphism) analysis. intermittently throughout the Pleistocene. The pattern of DNA extraction and PCR (polymerase chain reaction) ampli- genetic variation across the Mondego River suggests the fication of an 840 bp fragment of mitochondrial cytochrome b

GENETIC EXCHANGE ACROSS A SALAMANDER HYBRID ZONE 247

Fig. 1 Maps showing: (a) the distribution of Chioglossa lusitanica (shaded area, from Arntzen & Wallis 1999) and study area (square); (b) Sampling localities forming the transect (dotted line) with areas of medium (light shading) and high habitat suitability (dark shading) as inferred by bioclimatic modelling (Teixeira et al. 2001). Maximum and minimum altitudes are shown by triangles.

Table 1 Sampling localities over a south to north transect in Chioglossa lusitanica (see also Fig. 1)

Coordinates

Locality Code Elevation (m) Latitude (°N) Longitude (°W) UTM X UTM Y

Foz do Giraldo 1 700 40.00656 7.697167 611.2 4429.3 Alvôco da Serra 2 800 40.28744 7.678861 612.3 4460.5 Açor-Margaraça 3 600 40.22056 7.918722 592.0 4452.8 Barril do Alva 4 250 40.30564 7.959750 588.4 4462.2 Castanheira de Pêra 5 500 40.09131 8.201139 568.1 4438.2 Póvoa de Fiscal 6 300 40.11400 8.224333 566.1 4440.7 Vilarinho 7 300 40.11931 8.209028 567.4 4441.3 Riba de Cima 8 200 40.25914 8.236806 564.9 4456.8 Torres do Mondego 9 100 40.20764 8.360750 554.4 4451.0 Misarela 10 250 40.21844 8.358306 554.6 4452.2 Várzea 11 300 40.24825 8.375667 553.1 4455.5 Buçaco 12 350 40.36433 8.352194 555.0 4468.4 Saide 13 300 40.44617 8.324306 557.3 4477.5 Linhar de Pala 14 350 40.50425 8.244639 564.0 4484.0

(cytb) were undertaken as described in Alexandrino et al. group (haplotype A). Following the digestion of 5 µL

(2002). Two diagnostic restriction endonucleases (REs PvuII of PCR product with 10 mm Tris-HCl pH 7.5, 10 mm Mgcl2, and SfaNI, Fermentas) assays were designed to distinguish 50 mm NaCl, 0.1 mg/mL BSA and 1.5 units of RE enzyme between the two lineages of C. lusitanica using available for 3 h at 37 °C, RE fragment patterns were scored with mtDNA sequence data (Alexandrino et al. 2002). PvuII agarose gel electrophoresis. identifies two recognition sites (113 and 398) for the southern group (haplotype A) and one recognition site (398) Statistical analyses for the northern group (haplotype B). SfaNI identifies two recognition sites (231 and 531) for the northern group Allele frequencies were determined by gene counting. (haplotype B) and one recognition site (531) for the southern Private alleles were defined as those observed in a single

248 F. SEQUEIRA ET AL.

population. The genetic data analysis software (Lewis using four variables — pmin, pmax (the maximum and & Zaykin 2001) was used to calculate the average number minimum gene frequency values at the tail ends of a cline), of alleles (A) and average expected heterozygosity (HE). cline centre and cline width. Given the nature of the All markers were then collapsed to two allele systems by available data, we collapsed the sampled sites onto a one allocating all rare and low frequency alleles either to the dimensional transect for the purposes of analysis. Cline southern or the northern groups depending of their location fitting was undertaken along a best fit axis through the in the transect (see Table 2). Maximum likelihood clines sampled populations, equating to a compass bearing of were fitted independently for each locus to population 120°. This orientation approximately represents the real allele or lineage frequency data across the transect using orientation of all sites in the transect with the exception the compound tanh and exponential model of Szymura & of sites 2, 3 and 4 which were therefore excluded from Barton (1986) implemented in the ‘analyse’ application the analysis for purposes of cline fitting (see Discussion). (Barton & Hewitt 1985). We fitted sigmoid clines in ‘analyse’ For ease of representation, distance (km) along the fitted

Table 2 Allele frequency, sample size (N), average number of alleles (A) and average expected heterozygosity (HE) at eight polymorphic loci across the Chioglossa lusitanica hybrid zone

Sampling site Allele/ Locus lineage 1234567891011121314

N 21 19 21 25 23 24 27 25 15 31 24 31 26 32 mtDNA AS 1.00 0.68 1.00 0.92 1.00 1.00 1.00 1.00 1.00 1.00 1.00 0.77 — — BN — 0.32 — 0.08 ——————— 0.23 1.00 1.00 N 28 19 32 24 44 28 31 31 15 39 27 47 36 38 Adh 1 N ——— 0.54 0.43 0.48 0.31 1.00 1.00 1.00 1.00 0.99 1.00 1.00 2S 1.00 1.00 1.00 0.46 0.57 0.52 0.69 ——————— 3 N ——————————— 0.01 — — N 27 19 21 24 31 27 28 37 15 40 24 26 26 39 Pep-A 1 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 0.90 1.00 1.00 2 ——————————— 0.10 — — N 26 18 32 22 44 27 29 27 14 36 23 46 33 26 Pep-B 1 1.00 1.00 1.00 1.00 1.00 1.00 0.98 0.78 0.68 0.74 0.50 0.51 0.77 0.35 2 ——————— 0.22 0.21 0.26 0.48 0.49 0.22 0.65 3 ————————— 0.01 0.02 — — — 4 —————— 0.02 — 0.11 ————— 5 ———————————— 0.01 — N 27 19 14 12 29 21 23 18 11 9 17 19 20 14 Pep-C 1N ———— 0.17 0.12 0.36 0.33 1.00 0.95 1.00 0.97 1.00 1.00 1.00 2S 1.00 1.00 1.00 0.83 0.88 0.64 0.67 ——————— 3N ———————— 0.05 — 0.03 — — — N 28 19 32 24 44 27 31 37 15 40 27 47 46 39 Pep-D 1N 0.02 — 0.08 0.27 0.03 0.32 0.37 1.00 1.00 0.99 1.00 1.00 1.00 1.00 2S 0.98 1.00 0.92 0.73 0.97 0.68 0.63 ——————— 4N ————————— 0.01 ———— N 16 19 21 25 31 27 23 37 15 40 24 29 26 39 Pgd 1 0.87 0.97 0.90 1.00 0.95 0.94 0.93 1.00 1.00 1.00 1.00 1.00 1.00 1.00 2 0.13 0.03 0.10 — 0.05 0.06 0.07 ——————— N 27 19 32 19 44 27 30 37 15 39 27 46 46 37 Pgm-1 1FN ——— 0.37 — 0.06 0.08 0.49 0.66 0.60 0.83 0.84 0.80 0.89 1SN ——— 0.08 — — — 0.09 0.07 0.04 — 0.12 0.18 0.08 3FS 1.00 1.00 1.00 0.55 0.94 0.83 0.75 0.35 0.20 0.35 0.17 0.01 — — 3SS ———— 0.06 0.11 0.17 0.07 0.07 0.01 — — — 0.03 3VS ——————————— 0.02 — — 4N ——————————— 0.01 — — 5N ———————————— 0.01 — A 1.29 1.14 1.29 1.71 1.71 1.86 2.00 1.57 1.86 1.71 1.57 2.00 1.57 1.43

H E 0.036 0.019 0.045 0.246 0.138 0.257 0.270 0.139 0.154 0.133 0.122 0.140 0.096 0.093

S, N (subscript) — for alleles allocated either to the southern or the northern population groups, respectively (see text).

© 2004 Blackwell Publishing Ltd, Molecular Ecology, 14, 245–254 GENETIC EXCHANGE ACROSS A SALAMANDER HYBRID ZONE 249 transect was measured from the southernmost site in our Results sample (no. 1). Locality samples were weighted by effective sample size, a measure of the number of independent Analysis of the total genetic variation sampling events given observed genetic disequilibria: deviation of the population from Hardy–Weinberg equili- Allozymes. The number of alleles (electromorphs) ranged brium means the states of the two alleles sampled at a locus from two at Adh, Pep-A, Pep-C and Pep-D to seven at the are not statistically independent. Effective sample sizes Pgm-1 locus (Table 2): allele Adh*3 was found only at site 12 taking into account maximum likelihood estimates (MLEs) (frequency of 0.01); Pep-C*3 was found in two sites near the of FIS, were calculated for each sample at each locus as Mondego River (9 and 11) with frequencies of 0.05 and (Baird, SJE personal communication) 0.03, respectively; Pep-D*3 was found only at site 10 with frequency of 0.01. At the Pgm-1 locus, five low frequency 2N N = alleles (Pgm-1*1S, Pgm-1*3S, Pgm-1*3 V, Pgm-1*4, Pgm-1*5) e 1 + F IS were found with frequencies ranging from 0.01 to 0.11. The where N is the number of diploid individuals sampled, average number of alleles ranged from 1.14 at site 2 to 2.00 and Ne is the effective number of alleles sampled scaled at sites 7 and 12. Expected heterozygosity ranged from less such that complete relatedness results in effective sample than 2% at site 2 to ca. 25% at sites 4, 6 and 7. size ~1, while no relatedness but complete heterozygote For further analysis, rare and low frequency alleles were deficit results in effective allele sample size N. Cline allocated either to the northern group or the southern group coincidence and cline concordance between distinct loci depending on their location in the study transect (see and between different studies were evaluated combining Table 2). Substantial latitudinal variation was observed in maximum likelihood cline fitting procedures (Barton & pooled allele frequencies at the loci Adh, Pep-C, Pep-D and Hewitt 1985) with the approach recently described by Pgm-1 with fixed or nearly fixed differences between sites (Phillips et al. 2004) — for each model and locus, the located at the edges of the transect (1–2 vs. 12–14; Fig. 2). likelihood surface was explored stepwise along the axes Sites 4, 5, 6 and 7 showed intermediate allele frequencies at for both centre position c and width w with the other para- Adh, Pep-C, Pep-D. At Pgm-1 intermediate allele frequen- meters free to vary at each point. In this way, the likelihood cies were observed at sites 4 and 8–11 (Fig. 2). profiles (Hilborn & Mangel 1997) for both c and w can be constructed. For example, coincident vs. staggered cline mtDNA. Sites at the northern edge of the transect (13–14) centre hypotheses can be compared as follows: summing were fixed for lineage B haplotypes while sites south of log likelihood c profiles over a set L of loci results in the log Buçaco (site 12) were generally fixed for lineage A haplo- likelihood profile for the ML shared centre of the L loci. The types. Lineage A and lineage B haplotypes were observed centre coincidence ML can be compared with the sum of together at sites 2, 4 and 12 (Table 2 and Fig. 2). noncoincident c profile MLs using a likelihood ratio test, under the assumption that twice the difference in log e Hardy–Weinberg equilibrium and genotypic likelihood (G = −2∆LL) between two models asymptotically disequilibrium across the hybrid zone follows a χ2 distribution with the degrees of freedom being the difference in the number of parameters. Within the Analysis of genotypic frequencies produced estimates of contact zone Hardy–Weinberg equilibrium (HWE) was FIS that were not significantly different from zero across all assessed for each marker using MLE estimates of FIS in allozyme loci with the exception of Pep-C (Table 3). This ‘analyse’. Hardy–Weinberg equilibrium was also assessed result is because of the heterozygote deficit observed only at each site for each marker within the contact zone. ‘Scaled at site 5 (FIS = 0.513; 2 mL unit bounds 0.046–0.869) given average pairwise linkage disequilibrium’ (R = D/(pquv)1/2) that genotype frequencies at Pep-C were in equilibrium through the cline was assessed for diagnostic markers by at all the remaining sites (results not shown). Significant partitioning the variance in ‘hybrid index’ summed across deficiency or excess of heterozygotes was absent when diagnostic loci (Barton & Gale 1993). consensus FIS estimates were plotted for all four allozyme

Table 3 Maximum likelihood estimates of Locus F C (km) W (km) IS FIS, position of cline centre (c) and cline width (w) with the corresponding two log- − Adh 0.087 ( 0.089–0.263) 24.1 (23.1–25.1) 10.7 (8.2–14.2) likelihood unit confidence limits, for five Pep-C 0.245 (0.018–0.477) 24.5 (24.5–27.0) 6.3 (4.8–9.3) loci across the Chioglossa lusitanica hybrid Pep-D 0.000 (0.000–0.131) 26.2 (25.2–27.2) 8.2 (6.2–10.7) zone. Distances are measured along the − Pgm-1 0.028 ( 0.020–0.149) 34.5 (33.5–35.5) 14.5 (12.5–17.0) fitted transect, starting from the southernmost mtDNA not applicable 48.2 (47.2–49.7) 2.0 (1.0–3.5) study locality

Fig. 2 Pie diagrams representing the frequencies of southern (white shading) and northern (black shading) alleles at four nuclear loci (a: Adh, b: Pep-C, c: Pep-D and d: Pgm-1) and the mitochondrial DNA locus (e: Cyt-b) across the hybrid zone within Chioglossa lusitanica. loci against distance along the transect with the exception allele at the Pgm-1 locus in a sample of 37 otherwise northern- of site 14 (Fig. 3a). Heterozygote deficit at this site (FIS = like individuals, which suggests recent immigration. Across 1.000; 2 mL unit bounds 0.260–1.000) was the result of the all the diagnostic loci, significant ‘scaled average pairwise presence of one individual homozygous for the southern linkage disequilibrium’ (R) was not observed (Fig. 3b).

than at mtDNA (MLE = 1.98 km; 2 mL unit bounds 0.98– 3.48 km). Constraining all loci to a common cline width

(GW same-W diff. = 24.849, 4 d.f., P < < 0.05) resulted in significant decrease in likelihood but the comparison of likeli-

hood profiles for Adh, Pep-C and Pep-D (GW same-W diff. = 3.262, 2 d.f., P > 0.05) and Adh and Pgm-1 (GC same-C diff. = 2.000, 1 d.f., P > 0.05) suggested similar cline widths.

Discussion The hybrid zone between two evolutionary units of Chioglossa lusitanica is located near the Mondego River in central Portugal (Alexandrino et al. 2000, 2002, 2004). Presently, both field surveys and ecological modelling suggest that the contact between southern and northern population groups spearheads over two narrow stretches of favourable mountainous habitat (Teixeira et al. 2001), close to localities 8 and 9–11. Elsewhere in the Mondego River valley the habitat is mostly unfavourable. We conducted a formal statistical analysis of the C. lusitanica hybrid zone in the context of tension zone model theory with the following key results: (i) cline centres are not coincident for all markers, with average widths of c. 2– 15 km; (ii) heterozygote deficit was not observed across loci near the transect centre; (iii) associations of parental allele combinations (‘linkage disequilibrium’ R) were not detected both across loci and across the transect. These observations are not consistent with strong selection acting against hybrids suggesting that the Chioglossa hybrid zone is not a tension zone but instead one shaped by neutral mixing and a complex history of populations (see the following discussion). Fig. 3 Transect across the hybrid zone within Chioglossa lusitanica with (a) heterozygote deficit (FIS) (b) linkage disequilibrium (R), and (c) maximum likelihood fitted cline curves for diagnostic loci. Genetic structure of the zone: tension zone vs. neutral Distances are in km, starting from the southernmost study diffusion locality. Gray symbols represent sites 2, 3 and 4 excluded from Nuclear allozyme loci. Clines were coincident and concordant cline fitting (see text). between three diagnostic nuclear allozyme loci across the hybrid zone (Adh, Pep-C, and Pep-D). The cline for Pgm-1 was however, noncoincident with other nuclear markers Cline shape across the hybrid zone (displaced ca. 10 km to the north) and its cline width was Visual inspection of likelihood profiles (not shown) sug- concordant only with Adh. Because (i) the average width of gested coincidence of cline centres between the loci Adh, clines for allozyme loci appears to vary along a continuum ≈ Pep-C and Pep-D (MLEs 24–26 km) but not between these between ca. 6–14 km (with wPep-C < wPep-D < wAdh < wPgm-1) and Pgm-1 (C ≈ 34 km) and mtDNA (C ≈ 48 km; see also and (ii) because we cannot rule out that further sam- Table 3 and Fig. 3c). Constraining all markers to a common pling of the zone (in areas adjacent to Lousã, both north centre resulted in a significant decrease in likelihood and south) would reveal cline concordance for those markers,

(GC same-C diff. = 262.862, 4 d.f., P < < 0.05). Constraining only it is reasonable to suggest that clines for nuclear markers Adh, Pep-C and Pep-D to a common centre did not result in are in fact concordant with an overall average width of a significant decrease in likelihood (GC same-C diff. = 3.181, 2 10 km (conservatively bound between 5 and 17 km; d.f., P > 0.05), suggesting cline coincidence. Table 3). Cline widths were variable (MLEs = 1.98–14.50 km; Given that the contact between the two evolutionary Table 3) with wider clines at Adh, Pep-C, Pep-D and units of Chioglossa must have been possible since the Pgm-1 (MLEs at 6.34 < 8.20 < 10.70 < 14.50 km, respectively) Holocene, it is reasonable to assume that the hybrid zone

© 2004 Blackwell Publishing Ltd, Molecular Ecology, 14, 245–254 252 F. SEQUEIRA ET AL. formed at least several hundred generations ago (≈2500) populations is likely to explain the observed narrower and is now at equilibrium. At migration-selection equilibrium, clines for mtDNA than for allozymes. the effective selection acting against hybrids can, in Drift and a neutral diffusion cline could also account principle, be calculated from hybrid zone tension model for the noncoincidence of mtDNA clines with respect to theory using the formula s = 8σ 2/w2, where σ is the stand- nuclear markers perhaps associated with the movement of ard deviation of distance between parent and offspring the hybrid zone (e.g. as in Triturus sp.; Arntzen & Wallis per generation and w is the cline width (Bazykin 1969). 1991, 1999). In this respect, the observation of haplotypes of We can roughly estimate the dispersal parameter at σ both mtDNA lineages at sites located to the east of the main ≈ 60 m·gen−1/2 from a mark-recapture study carried out transect, ca. 20 km south of the mtDNA fitted cline centre, for one year along a stream in northern Portugal (Arntzen suggests either that (i) the mtDNA cline centre was located 1981). This may be an underestimate given (i) that the south of its present position in the past or (ii) there was study spanned only one fourth of the species generation mtDNA capture (e.g. as in the house mouse hybrid zone; time (ii) larval drift along streams (Thiesmeier 1994) and Gyllensten & Wilson 1987) as a signature of historical intro- juvenile dispersal (iii) long distance migration movements gressive hybridization. While these hypotheses are not along moist brook embankments (Arntzen 1994) and (iv) mutually exclusive, the further sequencing of those northern extinction/recolonization patterns in the zone are not taken haplotypes could help testing whether the observations into account. Assumptions made, to maintain the wide are the result of recent vs. historical hybridization events. clines (relative to dispersal) observed at nuclear markers We cannot rule out, however, that differential selection of would require selection against heterozygotes at a single alleles (e.g. in Caledia captiva; Shaw et al. 1993) acting across locus s = 8σ 2/w2 of ca. 0.03% (0.01–0.1%). These very low the neutral diffusion clines in Chioglossa could have pro- levels of selection, in combination with the absence of gen- duced the observed noncoincident patterns between both otypic disequilibria across nuclear markers and across the mtDNA and nuclear markers and between Pgm-1 and zone, suggest that recombination at nuclear markers is three other nuclear allozyme loci. not having a negative effect on hybrid genotypes and Detailed studies of hybrid zones have usually confirmed hybridization should proceed according to a neutral the expectation of narrow clinal variation for interactions diffusion process (Endler 1977; Hewitt 1988). between recognized species (e.g. Szymura & Barton 1986, 1991; Arntzen & Wallis 1991; Szymura et al. 2000; Babik Nuclear vs. mitochondrial markers. The cline for the mtDNA et al. 2003) and wider clines associated with nonconcordant locus is in sharp contrast with clines at all four nuclear patterns between markers for interactions between marker loci by being both noncoincident (mtDNA cline genetically divergent intraspecific groups (e.g. Ensatina sp., centre displaced 20 km to the north compared to clines Wake & Schneider 1998; Sceloporus sp., Marshall & Sites for Adh, Pep-C and Pep-D) and strikingly narrower 2001; Salamandra sp., Garcia-Paris et al. 2003). For example, (nonconcordant) with w ca. 2 km (1–3.5 km). Stronger selec- the narrow coincident and concordant clines observed at tion acting on the mitochondrial genome than allozymes hybrid zones between two species of Bombina (Szymura & may not explain the observed nonconcordance, given Barton 1986, 1991) contrast with wide and noncoincident that (i) mtDNA-specific selection against hybrids would clines at the zone of transition between two mtDNA line- still be very weak, at s = 8σ2/w2 ca. 0.7% (0.2–3%), and ages of Ensatina eschscholtzii platensis in the Sierra Nevada (ii) disequilibria between mitochondrial and nuclear loci of California that appear to be merging after Pleistocene were not observed. vicariance (Jackman & Wake 1994; Wake 1997; Wake & Other explanations for why the mitochondrial cline is Schneider 1998). In this respect, the absence of genetic narrower than allozyme clines in Chioglossa are associated barriers to the neutral diffusion of alleles observed between with the maternal inheritance, lack of recombination and two evolutionary units of Chioglossa suggests that genetic smaller effective population size of cytoplasmic compared cohesiveness is maintained by population-level processes to nuclear markers: (i) the increased effects of drift would common within good biological species, i.e. panmixis in cause mtDNA to cross contact zones slower and remain genetically admixed populations. more stationary whereas nuclear markers would spread faster trough the populations in contact (see Endler 1977; The hybrid zone in space and time Avise 1994); (ii) asymmetric assortative mating could pre- vent mitochondrial introgression; (iii) sex-specific differ- The width of a neutral diffusion cline will be a function ences associated with dispersal (e.g. as in Salamandra sp.; of the time as secondary contact (T) and the dispersal rate Garcia-Paris et al. 2003) or mating behaviour (e.g. as in Hyla (σ) — w = σ1.68T−1/2 (Endler 1977). We can then use average sp.; Lamb & Avise 1986). Because we have no evidence for w for nuclear loci (10 km) and a conservative value for T sex-specific differences in dispersal and mating in C. lusi- (2500 generations) to estimate the dispersal parameter tanica, we suggest that drift acting with neutral mixing of σ = 120 m·gen−1/2. Alternatively we can use w and our

© 2004 Blackwell Publishing Ltd, Molecular Ecology, 14, 245–254 GENETIC EXCHANGE ACROSS A SALAMANDER HYBRID ZONE 253 direct estimate of σ to obtain T = 9722 generations. So, References under a neutral diffusion model, and given cline width Alexandrino J, Arntzen JW, Ferrand N (2002) Nested clade ana- estimates for nuclear loci, we conclude that either the lysis and the genetic evidence for population expansion in the dispersal parameter is underestimated, as expected (see phylogeography of the golden-striped salamander, Chioglossa previous discussion), or the hybrid zone existence precedes lusitanica (Amphibia: Urodela). Heredity, 88, 66–74. the Holocene. If the two population groups have been Alexandrino J, Ferrand N, Arntzen JW (1997) Genetic variation in diverging since the early Pleistocene, as was hypothesized some populations of the golden-striped salamander, Chioglossa (Alexandrino et al. 2000), then it is likely that several instances lusitanica (Amphibia: Urodela), in Portugal. Biochemical Genetics, 35, 371–381. of isolation and secondary contact took place during this Alexandrino J, Froufe E, Arntzen JW, Ferrand N (2000) Genetic period as habitat conditions prevented or promoted con- subdivision, glacial refugia and postglacial recolonization in the nectivity across the Mondego valley, respectively. This golden-striped salamander, Chioglossa lusitanica (Amphibia: would explain the observed cline shapes at nuclear markers Urodela). Molecular Ecology, 9, 771–781. suggesting secondary contact before the Holocene. Repeated Alexandrino J, Teixeira J, Arntzen JW, Ferrand N (2004) Historical instances of genetic exchange throughout the Pleistocene biogeography and conservation of the golden-striped sala- would have been facilitated not only by climatic amelior- mander (Chioglossa lusitanica) in northwestern Iberia: integrating ecological, phenotypic and phylogeographic data. In: Phylogeo- ation but also by the historical persistence of diversifying graphy in Southern European Refugia: Evolutionary Perspectives on populations in areas adjacent to the Mondego valley, as the Origins and Conservation of European Biodiversity (eds Weiss S, inferred from mtDNA sequence data (Alexandrino et al. Ferrand N), p. (in press). Kluwer, Dordrecht, The Netherlands. 2002). The question of how often these close populations Arnold ML (1997) Natural Hybridization and Evolution. Oxford on either side of the Mondego valley have been connected University Press, New York. by gene flow cannot addressed with our current observa- Arntzen JW (1981) Ecological observations on Chioglossa lusitanica tions but could be examined in the future by further (Caudata, Salamandridae). Amphibia-Reptilia, 1, 187–204. Arntzen JW (1994) Speedy salamanders: sedentariness and migra- sequencing both mitochondrial and nuclear markers across tion of Chioglossa lusitanica. Revista Española de Herpetologia, 8, the hybrid zone. Contemporary gene flow is currently 81–86. being addressed using microsatellite markers. Both Arntzen JW, Alexandrino J (2004) Ecological modelling of genetic- historical and contemporary gene flow across the Mondego ally differentiated units of the Iberian endemic golden-striped valley are central to understanding how diversification has salamander, Chioglossa lusitanica. Herpetological Journal, in press. been taking place in Chioglossa, perhaps with historically Arntzen JW, Wallis GP (1991) Restricted gene flow in a moving persistent populations diverging over a small geograph- hybrid zone of the newts Triturus cristatus and T. marmoratus in western France. Evolution, 45, 805–826. ical scale. The hybrid zone formation in Chioglossa would Arntzen JW, Wallis GP (1999) Geographic variation and taxonomy repeatedly have resulted therefore from small scale of crested newts (Triturus cristatus superspecies): morphological recolonization of the habitat newly available after climatic and mitochondrial data. Contributions to Zoology, 68, 181–203. amelioration. This pattern recently uncovered for Chioglossa Avise JC (1994) Molecular Markers, Natural History and Evolution. (Alexandrino et al. 2002) and presently under study Chapman & Hall, New York. may be commonly found within refugia where small Avise JC, Walker D, Johns GC (1998) Speciation durations and scale spatial heterogeneity contributes to diversification Pleistocene effects on vertebrate phylogeography. Proceedings of the Royal Society of London, Biology Sciences, 265, 1707–1712. between populations. As phylogeographical patterns emerge Babik W, Szymura JM, Rafinski J (2003) Nuclear markers, mito- in Iberia, the study of spatial, ecological and genetic chondrial DNA and male secondary sexual traits variation in a connectivity at hybrid zones between diverged evolu- newt hybrid zone (Triturus vulgaris × T. montandoni). Molecular tionary units is a window open to understanding the Ecology, 12, 1913–1930. processes underlying biological diversification in southern Barton NH (1983) Multilocus clines. Evolution, 37, 454–471. European refugia. Barton N, Bengtsson BO (1986) The barrier to genetic exchange between hybridizing populations. Heredity, 57, 357–376. Barton NH, Gale KS (1993) Genetic analysis of hybrid zones. Acknowledgements In: Hybrid Zones and the Evolutionary Process (ed. Harrison RG), pp. 13–45. Oxford University Press, New York. We are grateful to Claudia Oliveira for field assistance, Marisa Barton NH, Hewitt GM (1985) Analysis of hybrid zones. Annual Azevedo, Catarina Pinho and Raquel Godinho for technical assistance, Review of Ecology and Systematics, 16, 113–148. and to Stuart Baird for discussions on hybrid zone theory. This Barton NH, Shpak M (2000) The effect of epistasis on the structure work was funded by the ‘Instituto da Conservação da Natureza’ of hybrid zones. Genetical Research, 75, 179–198. (ICN), the ‘LIFE’ program of the European Community and the Bazykin AD (1969) Hypothetical mechanism of speciation. Evolu- ‘Fundação para a Ciência e a Tecnologia’ (SFRH/BD/3365/ tion, 23, 685–687. 2000 PhD grant to FS, PRAXIS XXI/BD/5917/95 PhD grant to Endler JA (1977) Geographic Variation, Speciation and Clines. Princeton JA, SFRH/BPD/3597/2000 Postdoctoral grant to JA and project University Press, Princeton, NJ. POCTI/BSE/40987/2001). Collection permits were provided Garcia-Paris M, Alcobendas M, Buckley D, Wake DB (2003) Dis- by the ICN. persal of viviparity across contact zones in Iberian populations

of fire salamanders (Salamandra) inferred from discordance of Searle JB (1993) Chromosomal hybrid zones in eutherian mam- genetic and morphological traits. Evolution, 57, 129–143. mals. In: Hybrid Zones and the Evolutionary Process (ed Harrison Gavrilets S (1997) Hybrid zones with Dobzhansky-type epistatic RG), pp. 309–353. Oxford University Press, New York. selection. Evolution, 51, 1027–1035. Shaw DD, Marchant AD, Contreras N et al. (1993) Genomic and Gomez A, Lunt DH (2004) Refugia within refugia: patterns of environmental determinants of a narrow hybrid zone: cause phylogeographic concordance in the Iberian Peninsula. In: Phylo- or coincidence. In: Hybrid Zones and the Evolutionary Process geography in Southern European Refugia: Evolutionary Perspectives on (ed Harrison RG), pp. 165–195. Oxford University Press, New the Origins and Conservation of European Biodiversity (ed. Weiss S, York. Ferrand N), p. (In the press). Kluwer, Academic Publishers, Szymura JM, Barton NH (1986) Genetic analysis of a hybrid zone Dordrecht, The Netherlands. between the fire-bellied toads, Bombina bombina and B. variegata, Gyllensten UB, Wilson AC (1987) Interspecific mitochondrial near Cracow in southern Poland. Evolution, 40, 1141–1159. DNA transfer and the colonization of Scandinavia by mice. Szymura JM, Barton NH (1991) The genetic structure of the hybrid Genetical Research, 49, 25–29. zone between the fire-bellied toads Bombina bombina and Harrison RG (1990) Hybrid zones: windows on the evolutionary B. variegata: comparisons between transects and between loci. process. Oxford Surveys in Evolutionary Biology, 7, 69–128. Evolution, 45, 237–261. Harrison RG (1993) Hybrid Zones and the Evolutionary Process, Szymura JM, Uzzell T, Spolsky C (2000) Mitochondrial DNA p. 364. Oxford University Press, New York. variation in the hybridizing fire-bellied toads, Bombina bombina Hewitt GM (1988) Hybrid zones — natural laboratories for evolu- and B. variegata. Molecular Ecology, 9, 891–899. tionary studies. Trends in Ecology and Evolution, 3, 158–167. Taberlet P, Fumagalli L, Wust-Saucy AG, Cosson JF (1998) Com- Hewitt GM (1996) Some genetic consequences of ice ages, and parative phylogeography and postglacial colonization routes in their role in divergence and speciation. Biology Journal of the Europe. Molecular Ecology, 7, 453–464. Linnean Society, 58, 247–276. Teixeira J, Arntzen JW (2002) Potential impact of climate warming Hewitt GM (1999) Postglacial recolonization of European biota. on the distribution of the golden-striped salamander, Chioglossa Biology Journal of the Linnean Society, 68, 87–112. lusitanica, on the Iberian Peninsula. Biodiversity and Conservation, Hewitt G (2000) The genetic legacy of the quaternary Ice Ages. 11, 2167–2176. Nature, 405, 907–913. Teixeira J, Ferrand N, Arntzen JW (2001) Biogeography of the Hewitt GM (2004) Genetic consequences of climatic oscillations in golden-striped salamander Chioglossa lusitanica: a field survey the quaternary. Philosophical Transactions of the Royal Society of and spatial modelling approach. Ecography, 24, 618–624. London Series B Biology Sciences, 359, 183–195. Thiesmeier B (1994) Trophische Beziehungen und Habitatprä- Hilborn R, Mangel M (1997) The Ecological Detective: Confronting ferenzen sympatrisch lebender Salamandra salamandra Models with Data. Princeton University Press, Princeton. und Chioglossa lusitanica-Larven. Abhandlungen und Berichte für Jackman TR, Wake DB (1994) Evolutionary and historical analysis Naturkunde, 17, 119–126. of protein variation in the blotched forms of salamanders of the Wake DB (1997) Incipient species formation in salamanders of the Ensatina complex (Amphibia, Plethodontidae). Evolution, 48, Ensatina complex. Proceedings of the National Academy of Sciences 876–897. of the United States of America, 94, 7761–7767. Kruuk LEB, Baird SJE, Gale KS, Barton NH (1999) A comparison Wake DB, Schneider CJ (1998) Taxonomy of the plethodontid of multilocus clines maintained by environmental adaptation or salamander genus Ensatina. Herpetologica, 54, 279–298. by selection against hybrids. Genetics, 153, 1959–1971. Lamb T, Avise JC (1986) Directional introgression of mitochondrial DNA in a hybrid population of tree frogs — the influence of F. Sequeira is focusing his work on the population genetics mating behavior. Proceedings of the National Academy of Sciences of Chioglossa lusitanica, especially the fine-scaled evolutionary of the United States of America, 83, 2526–2530. processes occurring across a hybrid zone within the species. This Lewis PO, Zaykin D (2001) Genetic data analysis: computer work is part of his research towards a PhD degree. J. Alexandrino program for the analysis of allelic data, Version 1.0 (d16c). Free worked on the phylogeography and historical biogeography of program distributed by the authors over the internet from C. lusitanica for his PhD and started the work on the hybrid zone. http://lewis.eeb.uconn.edu/lewishome/software.html. He is currently finishing postdoctoral research on the ecological Marshall JC, Sites JW (2001) A comparison of nuclear and mito- modelling, phylogeography and hybridization within the chondrial dine shapes in a hybrid zone in the Sceloporus grammicus Plethodontid salamander Ensatina ring species complex. Sara complex (Squamata: Phrynosomatidae). Molecular Ecology, 10, Rocha is a postgraduate student interested on phylogeny and 435–449. phylogeography of amphibians and reptiles. J. W. Arntzen is Phillips BL, Baird SJE, Moritz C (2004) A narrow zone of secondary Curator of Vertebrates at the National Museum of Natural History contact between morphologically cryptic phylogeographic line- (Leiden, the Netherlands). He has his main interests in systematics ages of the rainforest skink, Carlia rubrigularis. Evolution, 58, and evolutionary biology of European amphibians. Nuno Ferrand 1536–1548. is Associate Professor and heads the Research Centre in Bio- Piálek J, Barton NH (1997) The spread of an advantageous allele diversity and Genetic Resources. He is interested in a variety of across a barrier: the effects of random drift and selection against questions in evolutionary and conservation genetics. heterozygotes. Genetics, 145, 493–504.