We sequenced 105 individuals for cytochrome b
) and 45 individuals for 16S, plus two outgroup taxa (Additional file 1
). Ingroup base frequencies are similar to those found in other frogs [29
]: cyt b
: A = 0.240, C = 0.313, G = 0.145, T = 0.301; 16S: A = 0.299, C = 0.251, G = 0.194, T = 0.255. Ingroup cyt b
sequences collapsed into 21 unique haplotypes and 16S into 15 unique haplotypes. For the 16S-cytb data combined (32 individuals; 20 haplotypes and two outgroup taxa; 1248 bp), 227 included characters were parsimony informative.
We found no conflicting phylogenetic signal between the cyt b and 16S data (partition homogeneity test, P = 1.0), justifying the use of the combined data partition in maximum parsimony (MP) analyses. Both MP and Bayesian analyses of the 16S-cytb data produced well-supported trees identical in all major topological details and with three major clades: an Upland Clade (Figure localities Hola Vida, Santa Clara, EBJS, Llanganates, Cando, and Chonta Yacu), a Lowland Clade (Yasuní, Puca Chicta, EBJS, Serena, Auca 14, La Selva, and Cuyabeno); and a Southeastern clade (Kapawi and Auca 14) (Figure ). The topology of the 16S-cytb trees is congruent with trees resulting from separate analyses of the full 16S and cyt b data (results not shown).
Figure 1 Map of sample localities. Map of sample localities across eastern Ecuador: Chonta Yacu (1), Reserva Cuyabeno (2), Serena (3), Cando (4), Jatun Sacha Biological Station (EBJS) and surrounding area (5), Puca Chicta (6), Auca 14 Road near Dayuma (7), Parque (more ...)
Figure 2 Phylogenetic tree of Ecuadorean E. ockendeni. Bayesian phylogenetic tree of E. ockendeni samples and two outgroup taxa. Numbers in brackets correspond to localities in Figure 1. The topology and support were congruent with the MP tree. Posterior probabilities (more ...)
Corrected p-distances between E. ockendeni
haplotypes were high, ranging up to 20% between some haplotypes (Additional file 2
). Mean net divergence ± standard error [30
] between the Southeastern Clade and the Upland Clade was 15.45% ± 1.83, between the Southeastern Clade and the Lowland Clade was 15.26% ± 1.90, and between Upland and Lowland Clades was 12.01% ± 1.72. Average sequence divergence was low within the Upland and Lowland clades, at 1.35 and 1.77% respectively, while the Southeastern Clade, which has fewer haplotypes and more geographically distant sampling, showed more intraclade diversity at 5.34%.
For the cyt b data, the identity of haplotypes varied among localities but seven of the 13 sites had only a single haplotype (Table ). Nucleotide diversities within localities range from 0 to 0.105 ± 0.079. However, high nucleotide diversity within localities EBJS (5) and Auca 14 (7) is an artefact of finding sympatric but genetically distinct clades at those localities. When calculations of locality nucleotide diversity are separated by clade, the values range from 0 to 0.0030 ± 0.003 (Table ). HKY+G was the best-fit model of nucleotide substitution for cyt b as chosen by hLRT.
Population cyt b diversities
The ratio of nonsynonymous to synonymous mutations between all pairs of clades as calculated with a McDonald-Kreitman test does not suggest that selection is responsible for the cyt b divergence among lineages (Upland versus Lowland, P = 1.0; Upland versus Kapawi, P = 1.0; Kapawi versus Lowland, P = 0.3). Overall non-significant value of Tajima's D (0.255; P > 0.1) and within clades (Upland Clade D = -1.707, P > 0.05, Lowland Clade D = -0.834, P > 0.1, insufficient sample size to test Southeastern Clade) also implies neutrality (Table ).
Historical population expansion by clade
A survey of five tetranucleotide microsatellite loci across individuals of the Upland and Lowland clades in the Napo River area (localities 3 – 11), including a locale where both mitochondrial clades are present, shows that microsatellite genotypes at all loci are very different between Upland and Lowland clades. This implies complete reproductive isolation. The microsatellite library was developed based on Lowland Clade frogs. Consequently, in individuals from the Upland Clade the microsatellite loci are either non-functional (not amplifying after repeated attempts) or allele sizes are non-overlapping or have very different range of sizes, with the Upland Clade always having the larger mean allele size (Table ). Sequences of large alleles from a subset of samples from three of the five loci (Eloc-Batman&Robin, Eloc-Laurel&Hardy, and Eloc-Thelma&Louise) demonstrate that very large allele sizes in Upland Clade individuals are due to an increase in the number of microsatellite repeats (unpubl. data).
Estimating divergence and expansion
We used three different methods to estimate divergence: the net divergence method to estimate species divergence time; a Bayesian MCMC method to estimate lineage divergence (time to most recent common ancestor, TMRCA); and, a coalescent MCMC approach to estimate migration and TMRCA. Of these, the net divergence method and the Bayesian MCMC method (as implemented in the program BEAST) gave similar temporal results. Divergence estimates from coalescent analyses (as implemented in the program MDIV) were typically less than half the age of the other two estimates.
By the net divergence method, time of divergence between the Upland and Lowland lineages using our slowest estimated rate of evolution (0.6%/MYR) was 20.02 ± 2.87 mya (early Miocene) while the faster rate (1.0%/MYR) estimates the same split at 12.01 ± 1.72 mya (mid-Miocene) (Table ). Southeastern Clade split from Upland and Lowland clades approximately 25 ± 3 mya (late Oligocene) at the slower substitution rate and 15 ± 2 mya (mid-Miocene) by the faster rate.
Time of divergence among E. ockendeni clades based on net divergence
From BEAST, assuming a constant molecular clock and rates of 0.6 and 1.0% substitutions per million years, we estimated the TMRCA of the entire ingroup to be 24.39 mya (late Oligocene) and 14.61 mya (mid-Miocene), respectively. For the TMRCA of the Upland and Lowland clades, the constant clock TMRCA estimates 15.19 mya (mid-Miocene) or 9.11 mya (late Miocene), respectively. The uncorrelated, relaxed clock estimates were not substantially different: assuming mean substitution rates of 0.6 and 1.0%, the TMRCA estimates for the entire ingroup were 27.01 and 15.10 mya, respectively, while those for the Lowland/Upland clade were 15.23 and 9.03 mya, respectively (Table ).
Time of divergence among E. ockendeni clades based on Bayesian coalescent estimation
Coalescent calculations (implemented in the program MDIV) of the Upland and Lowland clade divergence resulted in an average θ of 6.23 ± 0.13 and an average T of 8.78 ± 0.31 (Table ). Estimates of TMRCA are 7.74 mya assuming a substitution rate of 0.6%/MYR and 4.64 mya assuming the faster substitution rate of 1.0%/MYR. In models of Southeastern Clade versus Upland Clade divergence, θ averaged 8.43 ± 0.27 and T 7.73 ± 0.52, suggesting a TMRCA of 10.55 mya under the slower substitution rate and 6.33 mya, with the faster rate. Models of Southeastern versus Lowland clades identified an average θ of 9.45 ± 0.38 and an average T of 8.98 ± 0.69, or 9.99 mya or 7.77 mya TMRCA, depending on substitution rate. For all clade comparisons M modelled best as 0 suggesting there is no gene flow among clades. The estimated time since population divergence among all clades is less than the TMRCA. MDIV estimates of TMRCA and species divergence are non-overlapping with BEAST and the net divergence method.
Estimates of time since population divergence (Tpop) and time to most recent common ancestor (TMRCA), inferred using a coalescent MCMC approach to estimate migration and divergence
Estimating population expansion
We used four different methods to try and identify and estimate the timing of population expansion. Based on mismatch analyses, the sudden expansion model of population growth cannot be rejected for the Upland or the Lowland clades, although the SSD probability was marginal for the Lowland Clade (P = 0.056) (Table ). Using a mutation rate of 1.0%/MYR and the peak of the mismatch distribution, τ, to estimate the time of population expansion suggests an Upland Clade expansion began in the latter half of the Pleistocene (793 000 YBP), a Lowland Clade expansion (if it occurred) somewhat later (154 000 YBP). The model of sudden population expansion is rejected for the Southeastern Clade. Raggedness indices suggest population expansion (curves are not significantly different than smooth) in all three clades. Fu's F and Tajima's D are not different than would be found under a stable population size. Therefore, we have conflicting evidence from different methods but some suggestion of population expansion particularly in the Upland Clade.