Introduction

The endemic cyprinid Chondrostoma lemmingii (Steindachner 1866) shows one of the widest distribution ranges within the native freshwater fish fauna of the Iberian Peninsula (Doadrio et al., 1991). More precisely, its current distribution comprises, from south to north, the Guadalquivir, Odiel, Guadiana, Sado and Tagus rivers, as well as some tributaries of the south-western Spanish Duero. In contrast, the majority of the endemic freshwater fishes of the Iberian Peninsula present narrower distribution ranges, which are constrained in many cases to a single basin or a portion of it (Doadrio et al., 1991; Almaça, 1995). Because of its wide distribution and its rather old origin (early Pliocene, Zardoya & Doadrio, 1998) this species represents an excellent model to understand the fragmentation of the hydrographical basins in the Iberian Peninsula and the processes of divergence and speciation of the Iberian cyprinid fauna.

This species has received the attention of morphologists (Collares-Pereira, 1980, 1983; Elvira, 1987a,b; 1997; Casado, 1995) and geneticists (Coelho et al., 1997), who principally investigated its taxonomy in relation to other cyprinid species. Interestingly, a recent study which focused on the morphometric variability of several populations (Casado, 1995) suggested a significant differentiation in meristic characters of the Duero basin population compared to the rest. In the absence of appropriate genetic data, the morphological divergence detected in the northern population of C. lemmingii may be interpreted either as phenotypic plasticity or as a sign of undiscovered genetic differentiation.

To determine the extent of the genetic differentiation of C. lemmingii among hydrographical basins, to test whether the morphological divergence of the Duero basin population has a genetical basis, and to further our understanding of C. lemmingii phylogenetic relationships, we have analysed allozymes, as well as the complete nucleotide sequence of the mitochondrial cytochrome b gene. Allozymes are useful molecular markers in detecting the pattern of genetic variability and differentiation of populations. The cytochrome b gene has been recognized as particularly suitable for recovering phylogenetic relationships (Zardoya & Meyer, 1996), especially of cyprinids (Briolay et al., 1998; Zardoya & Doadrio, 1998, 1999).

Materials and methods

Allozymes

Nine C. lemmingii samples were studied from representative sites of the main Spanish hydrographical basin found throughout its distribution (Table 1, Fig. 1). Additionally, a population of the related species C. lusitanicum, from the Arade basin, and a population of C. arcasii, from the Miño basin, were included in the analyses, as well as L. carolitertii from the Duero basin that was used as outgroup. Taxonomic designations follow Zardoya & Doadrio (1998).

Table 1 Localities and sample size of the populations analysed
Full size table
Figure 1
figure 1

Distribution of sampling localities for Chondrostoma lemmingii and the rest of the Iberian species considered in the study. •, C. lemmingii; ◊, C. lusitanicum; □, C. arcasii; ▵, C. macrolepidotus; , C. willkommii; ⋆, L. carolitertii. 1, Huebra River; 2, Gavilanes River; 3, Turones River; 4, Tiétar River; 5, Aurela River; 6, Fresnedoso River; 7, Estenilla River; 8,Quejigares River; 9, Robledillo River; 10, Montemayor River; 11, Boina River; 12, Caboalles River; 13, Lozoya River; 14, Sobral River; 15, Jándula River and 16, águeda River.

Full size image

Skeletal muscle and liver samples were removed in the field, transported to the laboratory in liquid nitrogen and homogenized, the supernatant being stored at −70°C. Horizontal starch gel (12%) electrophoresis (Pasteur et al., 1987) was performed to analyse 19 enzyme systems encoded by 26 presumptive loci (Table 2). Loci and alleles were designated following the recommendations of Shaklee et al. (1990). Diagnostic loci will be used as one of the criteria to elucidate genetic differentiation. We consider a locus to be diagnostic for one population when all of the alleles seen at the locus in this population are never seen in any other population. Similarly, a unique allele for a population is an allele which is never seen in any other population, but which need not have a frequency of one, or even a high frequency in this population.

Table 2 List of enzymes and electrophoretic conditions used
Full size table

Genetic distances among samples were calculated using either Nei’s (1978) genetic distance and Cavalli-Sforza & Edwards’s (1967) chord distance. Nei’s (1978) distances were used to compare the present data with values reported in the literature, and Cavalli-Sforza & Edwards’s (1967) chord distances were calculated for their easily interpretable geometric basis (Swofford & Olsen, 1990). Phenetic relationships among samples were estimated based on both distances, using UPGMA clustering. All computations were performed with the BIOSYS-2 computer program (Swofford et al., 1997). To test the robustness of the results, a bootstrap analysis was performed with 500 replicates (Neighbor, PHYLIP 3.5.5, Felsenstein, 1993).

Wright’s (1965) FST values were calculated to estimate genetic divergence among populations, and their significance was tested using a nonparametric permutation approach (significance level P < 0.05). Computations were made using the analysis of molecular variance procedure (Excoffier et al., 1992) in the AMOVA option of ARLEQUIN v. 1.1 program (Schneader et al., 1997).

To test for correlation between genetic distance matrices obtained from allozymes and cytochrome b gene sequences, a Mantel nonparametric test was used (Sokal et al., 1986) with 500 permutations. The significance of correlations was assessed by computing t-values.

Cytochrome b nucleotide sequence

Cytochrome b sequence data were obtained from two individuals of almost all populations of C. lemmingii that were analysed with allozymes (Fig. 1, Table 1). Cytochrome b nucleotide sequences of the related Iberian species C. willkommii (AF045984), C. lusitanicum (AF045986), C. arcasii (AF045979) and C. macrolepidotus (AF045980) (Zardoya & Doadrio, 1998) and the outgroups Rutilus rutilus (Y10440) and Telestes souffia (Y10439) (Briolay et al., 1998) were also included in the phylogenetic analyses.

Total cellular DNA was extracted by phenol/chloroform extraction and ethanol precipitation (Towner, 1991). PCR amplification of the entire cytochrome b gene was carried out using the primers L14724 and H15915 (Schmidt & Gold, 1993). Reactions were performed in a total volume of 25 μL containing 67 mM Tris-HCl, pH 8.3, 1.5 mM MgCl2, 0.4 mM of each dNTP, 2.5 μM of each primer, template DNA (10–100 ng), and Taq DNA polymerase (1 U; Promega). Double-stranded products were amplified using the following cycling profile: denaturing at 92°C for 60 s, annealing at 48°C for 90 s and extending at 72°C for 180 s. A final extension for 240 s at 72°C was also employed.

PCR products were cloned using the pGEM-T vector (Promega) into E. coli JM109 and sequenced using the FS-Taq Dye Deoxy Terminator cycle-sequencing kit (Applied Biosystems Inc.) on an automated DNA sequencer (Applied Biosystems Inc.) following the manufacturer’s instructions. DNA sequences of both strands were obtained using M13 universal (forward and reverse) sequencing primers.

Sequences were aligned by hand by using as reference the published sequence data of C. lemmingii for cytochrome b (Zardoya & Doadrio, 1998; GenBank accession numbers AF045987–AF045989). Sequence data were analysed with maximum parsimony (MP) using PAUP* version 4.0b2a (Swofford, 1998). Heuristic searches were conducted with 10 random stepwise addition of taxa, followed by branch swapping using the TBR routine (MULPARS option in effect). Transversions (Tv) were given eight times the weight of transitions (Ts). Neighbor-joining (NJ) (Saitou & Nei, 1987) analyses based on HKY85 corrected distance matrices (with empirical Ts/Tv ratios and base frequencies) were performed with PAUP* version 4.0b2a. Maximum-likelihood (ML) analyses were performed using the Quartet Puzzling method (Strimmer & Von Haeseler, 1996) (implemented in PAUP* version 4.0b2a). Robustness of the inferred MP, NJ and ML trees was tested by bootstrap analysis (Felsenstein, 1985) with 100 pseudoreplications, respectively.

Results

Genetic differentiation

Within the genus Chondrostoma, the most differentiated samples were those of C. lemmingii from the Duero Basin (D=0.349–0.578). Interestingly, the rest of the C. lemmingii populations showed lower Nei distance values to C. arcasii (D=0.249–0.336) and C. lusitanicum (D=0.159–0.274) than to the populations of C. lemmingii from the Duero basin (D=0.349–0.475). The Guadalquivir sample, Robledillo, showed a higher genetic differentiation (Table 4).

Table 4 Above diagonal Wright’s (1965) FST values, below diagonal Nei’s (1978) genetic distances
Full size table

Using both Nei’s genetic distance and Cavalli-Sforza & Edwards’s chord distance and the UPGMA cluster method, two trees were obtained with exactly the same topology. Based on the bootstrap support, five groupings are discernible (Fig. 2a).

Figure 2
figure 2

(a)UPGMA cluster dendrogram based on Cavalli-Sforza & Edwards’s (1967) chord genetic distances among the samples analysed. Bootstrap values (500 replicates) are provided for clades found in more than 50% of the bootstrapped trees. (b) Neighbor-joining dendrogram based on HKY85 corrected distance matrices (with empirical Ts/Tv ratios and base frequencies) among the samples analysed of the genus Chondrostoma. The numbers at each node represent bootstrap values (100 pseudoreplications) for both NJ (above branch) and the MP (below branch) analyses.

Full size image

The first grouping includes the three samples of C. lemmingii that occur in the south-west area of the Spanish Duero basin. The second and third clades discriminate the C. arcasii and the C. lusitanicum samples, respectively, and the fourth grouping separates the sample of C. lemmingii from the Guadalquivir basin. Finally, the fifth grouping include all samples from the Tagus and Guadiana basins.

Four diagnostic loci (IDHP-2*, IDHP-3*, MDH-B* and PEP*) and one unique allele (MDH-A*95) were present in the Duero group with respect to the rest of C. lemmingii populations. Moreover, the allele CAH-1*100 was fixed, but also appeared in low frequency in other samples (Table 3). A lower number of exclusive characters were detected in C. arcasii (five unique alleles) and C. lusitanicum (three diagnostic loci).

Table 3 Allele frequency for 24 polymorphic loci examined in the samples of the genera Chondrostoma and Leuciscus
Full size table

The population of the Guadalquivir basin was found to carry a diagnostic locus (MDH-1*) with respect to other samples of C. lemmingii.

Significant FST values (Table 4) were obtained when different groups of samples were analysed. High FST mean values were found when all Chondrostoma samples (FST=0.748) or all C. lemmingii samples (FST=0.737) were surveyed. However, FST values decreased considerably when the samples from Duero basin were removed (FST=0.332). According to the AMOVA, nearly 63% of the variation detected in the populations of C. lemmingii was due to differences among hydrographical basins, whereas 7.5% was due to variation among populations within the same drainage.

The complete cytochrome b nucleotide sequence (1140 bp) was obtained for the nine C. lemmingii samples analysed. Sequence divergence of haplotypes supported a differentiation pattern among populations similar to that detected with Nei’s genetic distances. Low sequence divergence was found within the Tagus, Guadiana and the Montemayor River sample from Guadalquivir (range 0.08–1.3%). The Robledillo River sample (Guadalquivir basin) differed from the latter by a mean of 4.3%. Nevertheless, the highest differences were detected between the samples from the Duero basin and the rest of the basins, with a mean of 9.0% (range 7.5–9.6%). Lower pairwise sequence divergences were found between C. arcasii, C. lusitanicum and C. macrolepidotus than were found between these species and the samples from the Duero basin, for which the mean was 8.5% (see Table 5).

Table 5 Pairwise sequence divergence (uncorrected p distances) among populations of the genus Chondrostoma. For C. arcasii, the mean value of the two populations is showed
Full size table

The congruence between the differentiation patterns obtained from allozyme and from cytochrome b sequence data was supported significantly by a Mantel test (P=0.01) showing a good correlation between Nei’s genetic distances and sequence divergence (uncorrected p distances) among populations (r=0.89, t=4.28).

Phylogenetic relationships

A total of 210 positions of the C. lemmingii cytochrome b data set were phylogenetically informative sites under the parsimony criterion. No ambiguous alignments were found and no gaps were postulated. An overall Ts/Tv ratio of 7.94 was calculated for this data set. Sequence divergences were plotted against Ts/Tv ratios for every position and no saturation was detected (data not shown). Hence, all codon positions were included in the phylogenetic analyses.

The phylogenetic analyses of the Iberian Chondrostoma data set with MP and NJ, using Telestes souffia and Rutilus rutilus as outgroup, recovered identical trees (Fig. 2b). They are congruent to that obtained from the allozyme data, with the exception made by the position of C. arcasii. The robustness of these trees were confirmed by bootstrapping. In the NJ analysis, two main clades could be distinguished. The first clade included C. arcasii, C. macrolepidotus and C. willkommii and, as sister group, the samples of C. lemmingii from the Duero basin. In the second clade, C. lusitanicum showed a basal position to a group including the C. lemmingii samples from the Tagus, Guadiana and Guadalquivir.

Interestingly, in the second clade the populations of C. lemmingii were grouped according to hydrographical basins, and three groups could be distinguished: the first included the samples from rivers Tietar and Aurela from the Tagus basin, the second comprised the three samples from the Guadiana basin, and the third included the samples from the Guadalquivir. The Robledillo River population, from the Guadalquivir basin, appeared to be significantly differentiated and basal to the rest of the samples.

Similar and congruent topologies were recovered by the MP analysis (with a single most parsimonious tree of 1108 steps) when a 8:1 Tv:Ts weighting was assumed, and the same species were used as outgroups. The Maximum parsimony tree also recovered the population of C. lemmingii from Duero as the sister group of C. arcasii, C. macrolepidotus and C. willkommii, but this relationship was not supported by a high bootstrap value.

Discussion

High levels of genetic differentiation were found within populations of C. lemmingii. The differentiation pattern obtained was identical using both allozyme and DNA sequence data, and grouped the samples according to drainages. Our findings suggest that the configuration of the actual Iberian hydrographical basins was one of the main historical events that generated barriers and isolated populations.

Diagnostic loci, Nei’s genetic distances and FST values, as well as the percentage sequence divergence, indicate that the Duero basin population shows higher levels of genetic differentiation from all the other samples and recognized species analysed in this study. Therefore, this population should be considered as a new undescribed species located in the south-western Spanish Duero basin. A morphological description of this taxon will be the subject of further studies.

The presence of diagnostic loci can be considered as the strongest criterion for species recognition (Beerli et al., 1995; Doadrio & Carmona, 1998; Mendoza-Qijano et al., 1998) since, in a concordant geographical framework, and within reproductively connected populations, we can, in this way, distinguish populations that may be isolated only ephemerally from those that have been isolated long enough to differ at multiple diagnostic markers (Baum & Donoghue, 1995). Diagnostic loci have been used increasingly, recently, to evaluate the species differentiation within Iberian cyprinid genera. For instance, one diagnostic locus separates C. polylepis from C. willkommii, four separate C. lemmingii from C. polylepis and six separate C. lemmingii and C. willkommii (Coelho, 1992; Coelho et al., 1997). Likewise, three loci diagnosed C. lemmingii as distinct from C. lusitanicum and the same genetic distances and FST values distinguished C. lemmingii from C. arcasii, although no diagnostic loci were found between the latter species pair (this study).

A long-term interruption of gene flow between the upper Guadalquivir and Guadiana basins is also supported by the allozyme data. One diagnostic locus and evident differences in allele frequencies at some other loci (Table 3) were detected in the Robledillo River sample. In contrast, a low level of genetic divergence was detected between the Guadiana and Tagus populations using allozymes, despite the greater differentiation between them that was observed with mtDNA. This pattern can be interpreted as a heightened sensitivity of mtDNA as a marker of population structure (Taylor et al., 1997; Triantafyllidis et al., 1999).

Phylogenetic relationships analysed among the Chondrostoma species and the samples of C. lemmingii were congruent with those previously published using the cytochrome b gene (Zardoya & Doadrio, 1998). Several studies agree that the substitution rate of the cytochrome b gene is adequate for establishing the phylogenetic relationships at the genus and species level for the Cyprinidae family (Briolay et al., 1998; Zardoya & Doadrio,1998, 1999). Our results confirmed its utility also at the population level.

Allozymes and cytochrome b sequences were congruent in assessing the phylogenetic relationships within C. lemmingii populations and some related species. Our data support the redefinition of the species formerly known as Rutilus lemmingii to Chondrostoma lemmingii (Collares-Pereira, 1980; Coelho et al., 1997; Zardoya & Doadrio, 1998).

Phenetic and phylogenetic relationships among populations of C. lemmingii support a phylogeographical pattern in which differentiation is driven mainly by the isolation of hydrographical basins. This pattern is also supported by an AMOVA analysis, since almost two-thirds of the genetic variation detected within the samples of C. lemmingii was explained by differences among basins. Consequently, fragmentation of an ancestral widespread population in several regions, corresponding to hydrographical drainages, seems to be the main source of current genetic differentiation.

However, the high divergence detected between the Duero basin populations with respect to the rest contrasted with that found among the populations of the Tagus, Guadiana and Guadalquivir basins, indicating different levels of genetic differentiation among drainages. Assuming no significant different substitution rate (at the 5% level) within Chondrostoma populations (Zardoya & Doadrio, 1999), the high genetic divergence of the Duero population has to be explained by additional historical events, older than the hydrographical configuration.

The Duero River and the rest of the Iberian main rivers started to acquire their current configuration during the Pliocene (López-Martínez, 1989; De La Peña, 1995). However, palaeogeographical data indicate that the Duero basin split earlier, derived from an ancient endorrheic lagoon in the Miocene period. Moreover, during the Miocene period, the Duero drainage was composed of the former Duero endorrheic basin and the ancient Ciudad Rodrigo endorrheic basin located in the south-western part of the Spanish Duero basin (De La Peña, 1995; López-Martínez, 1989), which almost perfectly matches the current distribution range of the new putative species. Following a calibration of a molecular clock of 1.52% sequence divergence per million years for European cyprinids (Zardoya & Doadrio, 1998), the Duero basin population is predicted to have started its differentiation during the Messinian period (6 MYA). In this period, it is likely that the Duero and Ciudad Rodrigo endorrheic basins almost dried up and became isolated one from the other.

The population from the upper Guadalquivir must have started to diverge in the south of the Iberian Peninsula prior to the complete isolation of the Tagus and Guadiana basin. Although bootstrap values are not high enough to support statistically the differentiation of the Robledillo River sample, the presence in this population of one diagnostic locus indicates an incipient speciation that may have started in the middle Pliocene period (3 MYA).

Thus, molecular markers have demonstrated their usefulness in revealing the genetic variability among putative populations of Chondrostoma lemmingii. The high congruence between nuclear and mitochondrial DNA markers in assessing the genetic differentiation among the samples analysed strongly supports previous morphometric findings that suggest a clear differentiation of the Duero basin population. Our data demonstrate that this population has diverged enough to be considered a new species, and suggest that ancient endorrheism, together with the current hydrographical configuration, as the palaeogeographical events that started population differentiation and speciation of cyprinids in the Iberian Peninsula.