Population genetics of the Chilean frog Batrachyla leptopus ( Leptodactylidae )

A variacao eletroforetica de proteinas codificadas por 14 loci foi analisada em oito populacoes (5 continentais e 3 insulares) da ra leptodactilidea chilena Batrachyla leptopus. A proporcao geral de loci polimorficos foi estimada como sendo de 18,7% e o numero medio de alelos por loco, 1,2, enquanto que as heterozigosidades observada e esperada foram 1,7 e 5,1%, respectivamente. O coeficiente esperado de identidade genetica foi 0,940; o numero correspondente para a distância genetica foi 0,063. A analise estatistica F mostrou um coeficiente de endogamia total (Fit) de 0,855 e altos niveis de subdivisao genetica (Fst = 0,596), assim como de endogamia dentro das populacoes (Fis = 0,640). Contudo, houve apenas um nivel moderado de diferenciacao genetica (Fst = 0,181) entre o grupo insular de populacoes e o grupo continental.

The temperate Nothofagus forests of southern Chile and Argentina are characterized by endemism of their reduced anuran fauna (Formas, 1979).Some members of this fauna (Caudiverbera and Eupsophus) are known from the Tertiary period (Schaeffer, 1949).One species of this small group of frogs, Batrachyla leptopus Bell, is quite small (30-35-mm snout vent length) and has a curious biology, with the eggs being deposited on the ground, near water (Busse, 1971).The main distribution of this species in Chile is continental, with some isolated populations on Chiloé Island, separated from the mainland by the Chacao channel (2.3-6 km wide, 50-100 m deep).The present distribution area of B. leptopus was intensively glaciated (Heusser and Flint, Statistical analysis All measures of genetic variability and population differentiation were calculated using BIOSYS-1 (Swofford and Selander, 1989).To determine whether heterozygosity levels were significantly different from those expected under panmixia, the observed and expected proportions of heterozygotes were compared by the Wilcoxon matched pairs signed rank test using the SPSS/pc + 5.0 computer package (Norusis, 1992).Levene's correction for small sample sizes was applied to the data (Levene, 1949).A locus was considered polymorphic when the frequency of the major allele was 95% or less.Heterogeneity of allele frequencies among the populations was investigated for each locus with contingency χ 2 analysis using the method of Workman and Niswander (1970).
The genetic relationships among populations were depicted by a Wagner tree (Swofford and Selander, 1989), since its algorithm does not assume a uniform rate of protein change.A population of B. taeniata from La Saval (Valdivia Province) was used as the outgroup.The intersample genetic variation, using Rogers' genetic distance (Rogers, 1972), was examined through multidimensional scaling, using NTSYS (Rohlf, 1992).The correlation be-tween genetic distance (Rogers, 1972) and geographic distance was examined by the Mantel test (Mantel, 1967), as implemented in NTSYS (Rohlf, 1992).Isolation by distance model (Wright, 1943) was analyzed with regression of F st /(1 -F st ) estimates for pairs of populations on the logarithm of the geographic distance for populations (Rousset, 1997).The GENEPOP program (ISOLDE option) (Raymond and Rousset, 1997) was used to test the hypothesis of isolation by distance.F-statistics (F st , F is , F it ) were calculated according to Wright (1965), as implemented in FSTAT (Goudet, 1995), and the statistical significance per locus and among loci was tested according to Excoffier et al. (1992) and Hudson et al. (1992).
The gene flow was estimated with Slatkin's formula (1985) rare allele method, which uses the formula ln p (1) = -0.505ln (Nm) + (-2.44),where p (1) is the average frequency of the private alleles and Nm is the product of the population size and immigration rate.Because sample sizes differed among populations, the correction recommended by Slatkin (1985) was applied, in which Nm is divided by the ratio of the average sample size to 25.Thus, Nm c = Nm (N/25) -1 , where N = (Σ Ni/n), the average sample size.Ni is the number of specimens in a given sample and n is the number of samples studied.For our average sample size of N = 12.2, the correction factor was 0.819.The program GENEPOP 3.1 (Rohlf, 1992) was used in estimating the level of gene flow.The method of Nei (1975) was used to estimate the maximum possible value of m (migration rate) using the formula I = m/(m + v), where I is the Nei normalized coefficient of genetic identity and v is the mutation rate.The value used for the mutation rate was 2 x 10 -6 per locus per generation.A minimum estimate of the average effective population size was derived by dividing Nm (derived from F st ) by m (obtained from Nei's genetic identity).The insular and continental populations were grouped into two clusters, and different populational calculations were carried out separately for both groups in order to make appropriate comparisons.
The average number of alleles per locus was 1.2.In simple regressions, percentage of polymorphic loci correlated strongly with the mean number of alleles per locus (r = 0.897).No significant relationships were found between sample size and percentage of polymorphic loci (r = 0.319), and sample size and mean number of alleles per locus (r = 0.194).
D-statistics indicated a deficit of heterozygotes in the loci for PGD (six populations), GPI (three populations), GPD-2 (one population) and GOT (one population), and an excess in the MDH-1 locus (one population; Table I).The heterozygote deficiency was 11.1% and the heterozygote excess 1.4%, suggesting that some of the populations were influenced by non-random mating (Gallardo and Köhler, 1992).The mean expected and observed heterozygosities, calculated for the populations, were 0.017 and 0.051, respectively (Wilcoxon z value = -3.516,P < 0.001).The following loci were fixed for the same allele in all populations: LDH-1, LDH-2, GPD-1, NP, and GOT-1.

Genetic population structure, gene flow, and effective population size
The heterogeneity of allele frequencies among the populations, investigated for each locus with contingency χ 2 analysis (Workman and Niswander, 1970), revealed significant heterogeneity in allele frequencies among populations (χ 2 = 870, d.f.= 91, P < 0.001).The high values of heterogeneity among populations indicate similarly high levels of inbreeding of further population subdivision as suggested by the positive F is (0.640) values (Chesser, 1983).
The standardized variance of allele frequencies indicated a high level of interpopulational genetic differentiation (F st = 0.596, P < 0.005).On average, 60% of the total variance of allele frequencies was attributable to genetic differentiation among the populations whereas 40% was found within populations.
The high positive value for F it (0.855, P < 0.005) indicated a greater number of homozygous individuals relative to that expected, when the data for all populations were pooled.The high positive F is value (0.640, P < 0.005), due to excess homozygotes within populations, indicated high levels of inbreeding in these populations of B. leptopus.
The estimate of gene flow, using the rare alleles method (Table I; Slatkin, 1985), was one migrant per 1.6 generations, which is not strong enough to counteract the effects of drift.The estimate of the minimum effective population size was 5600 individuals (the Nm and m values were 0.17 and 3 x 10 -5 , respectively).Since the Nm was very low, we concluded that the populations of B. leptopus were highly differentiated (Hartl and Clark, 1989).

Genetic and geographical correlation
The average genetic identity was 0.940 (Table II).The hierarchical analysis, based on Rogers' genetic distance (Rogers, 1972), produced clusters of populations that did not conform to their north-south distributional patterns (see Figures 1 and 2).The populations of La Picada and Chepu, 157 km apart, were grouped in the same cluster

B. leptopus populations
Distance from root 0.00 0.25 0.08 0.17 since they shared an equal degree of genetic similarity (Table II).In contrast, the Puyehue and La Picada populations, 50 km apart, were allocated separately.The Mantel test for a correlation between matrices of Rogers' (1972) genetic distance and geographic distance was not significant (r = -0.082, P = 0.330), indicating that genetic similarity does not correspond to geographic proximity.The isolation-by-distance test indicated a lack of significant correlation (r = 0.651) between F st /(1 -F st ) and the logarithm of geographical distance (Rousset, 1997).This shows that the genetic differentiation is not related to the geographical distance.On the other hand, it suggests that this species may have only recently invaded this area since it is not in equilibrium (Slatkin, 1993).The lack of association between genetic differentiation and geographical distances suggests differentiation by a founder effect, mutation, and drift (Mayr, 1968;Chesser, 1983).
The southwestern part of South America was affected by three to four pleistocenic glaciations which reached the eastern slopes of the coastal range (Figure 1) (Paskoff, 1977).The last of these glaciations occurred between 20,500 and 14,000 years BP (Mercer, 1972).Consequently, most animal and plant species were displaced to the western slopes, where they remained in refugial areas (Formas, 1979;Ashworth and Hoganson, 1993;Villagrán et al., 1996).After retreat of the glaciers, B. leptopus populations may have recolonized their primitive areas of distribution.Based on this scenario, the lack of correlation between genetic and geographical distances is interpreted as the result of a relatively recent recolonization from the west.The allozymic data are consistent with the notion that a founder effect and genetic drift were associated with these post-glacial events.

Insular and continental populations
The continental group of populations was significantly more polymorphic (28.6%) than the insular one (21.4%),and the mean number of alleles per locus was also higher in the continental (1.8) than in the insular group (1.5).The mean heterozygosity was lower in the mainland populations (0.7% direct-count, 11.8% expected) than in the Chiloé Island populations (3.5% direct-count, 5.8% expected).The F st value (0.181) and the Rogers' genetic distance (0.017) indicated a low degree of differentiation between these two groups of populations.
The hierarchical analysis (Wagner tree) built on Rogers' genetic distance produced a topology not totally consistent with the populations' geography.Using Rogers' genetic distances, interpopulational genetic variation was further examined.Non-hierarchical, multidimensional scaling allowed two slightly different sets of populations to be recognized: an insular group (Puntra, Chepu, Dalcahue) and a continental one (La Saval, Pucatrihue, Puyehue, La Picada and Lenca) (Figure 3).
The calculated gene flow, using the rare alleles method (Table I ;Slatkin, 1985), between the insular and continental populations was one immigrant per three generations, and does not agree with the known situation.Since these animals have very little tolerance of salt water, oceans are effective barriers to dispersal (Porter, 1972).The estimated gene flow between the Chiloé Island and mainland populations was interpreted as resulting from genetic interchanges that took place when both areas were connected during the last glacial maximum (18,000 years BP) by a wide land bridge (Climap, 1981;Moreno et al., 1994).The low values for Rogers' genetic distance (0.017) and genetic differentiation (F st = 0.181) between the insular and continental populations of B. leptopus are consistent with this scenario.
The level of gene flow and the genetic distances between insular and continental populations of B. leptopus suggest that the Chacao channel has acted as a recent barrier to gene flow between these groups of populations.

Figure 1 -
Figure 1 -Map of the locations where populations of Batrachyla leptopus were collected in southern Chile.Inset: Distribution of B. leptopus in southern Chile and Argentina.

Figure 2 -
Figure 2 -Wagner tree showing the allozymic relationships among Batrachyla leptopus populations.The tree was rooted using the outgroup method.Cophenetic correlation = 0.999.

Figure 3 -
Figure 3 -Bi-dimensional ordination of the eight populations of Batrachyla leptopus using the multidimensional scaling method.

Table I -
Allelic frequencies, polymorphism (%), heterozygosity (observed and expected), and mean number of alleles per locus at nine variable loci for eight populations of Batrachyla leptopus.