Speciation and DNA barcodes: testing the effects of dispersal on the formation of discrete sequence clusters

Abstract

Large-scale sequencing of short mtDNA fragments for biodiversity inventories ('DNA barcoding') indicates that sequence variation in animal mtDNA is highly structured and partitioned into discrete genetic clusters that correspond broadly to species-level entities. Here we explore how the migration rate, an important demographic parameter that is directly related to population isolation, might affect variation in the strength of mtDNA clustering among taxa. Patterns of mtDNA variation were investigated in two groups of beetles that both contain lineages occupying habitats predicted to select for different dispersal abilities: predacious diving beetles (Dytiscidae) in the genus Bidessus from lotic and lentic habitats across Europe and darkling beetles (Tenebrionidae) in the genus Eutagenia from sand and other soil types in the Aegean Islands. The degree of genetic clustering was determined using the recently developed 'mixed Yule coalescent' (MYC) model that detects the transition from between-species to within-population branching patterns. Lineages from presumed stable habitats, and therefore displaying lower dispersal ability and migration rates, showed greater levels of mtDNA clustering and geographical subdivision than their close relatives inhabiting ephemeral habitats. Simulations of expected patterns of mtDNA variation under island models showed that MYC clusters are only detected when the migration rates are much lower than the value of Nm=1 typically used to define the threshold for neutral genetic divergence. Therefore, discrete mtDNA clusters provide strong evidence for independently evolving populations or species, but their formation is suppressed even under very low levels of dispersal.

Figure 1

Sampling maps showing the collecting localities for (a) Bidessus diving beetles in western Europe and (b) Eutagenia darkling beetles in the Aegean archipelago in the eastern Mediterranean. Sampling sites for different species: filled squares, B. unistriatus; open squares, B. goudotii; stars, B. minutissimus; triangles, E. smyrnensis s.l.

Figure 2

Examples of simulated genealogies at four out of the eight migration rates (m) examined, increasing from Nm=(a) 0.0001, (b) 0.001, (c) 0.01 to (d) 0.1. Grey shading indicates branches allocated to the coalescent by the MYC model: (a) number of coalescent branching clusters (N_MYC)=16, lambda ratio (λ_D/λ_C)=0.272; Χ²p=0.000; (b) N_MYC=13, λ_D/λ_C=0.474, p<0.001; (c) N_MYC=4, λ_D/λ_C=0.001, p=0.016 and (d) N_MYC=1, p=1.000. Nodes marked with asterisks correspond to demes that were recovered as monophyletic.

Figure 3

Summary statistics of simulated datasets presented as mean values (error bars=1 s.d., n=100, 1 million generations; see text). (a) Proportion of starting (local) demes recovered as monophyletic on the simulated trees and the average nucleotide diversity within demes (π). Open bars, monophyly; hatched bars, nucleotide diversity (π). (b) Number of clusters detected by the MYC model of lineage branching and lambda ratio, identified based on the point of transition from species-level (λ_D) and coalescence (λ_C) branching rates. Open bars, no. of clusters; hatched bars, lambda ratio. Statistical significance was evaluated with a likelihood ratio test comparing the MYC model with one of uniform coalescent branching. The lambda ratio (λ_D/λ_C) provides a convenient measure of the degree to which long and short branches can be distinguished, but at high Nm the MYC model has poor support; thus high λ_D/λ_C results only from artefactual delimitation of MYC groups. ^*p<0.05, ^***p<0.001.

Figure 4

Bayesian trees for two empirical datasets. (a) Bidessus (Dytiscidae) including two lentic and one lotic species and (b) E. smyrnensis s.l. (Tenebrionidae) from sand and soil habitats. Grey shading indicates branches allocated to the coalescent by the MYC model. Numbers on the nodes correspond to posterior probabilities (only values above 0.90 are shown). These trees include all sequenced haplotypes, but in order to apply the MYC, only unique haplotypes were used.

Figure 5

Mean nucleotide diversity (π) when sequences are grouped into morphologically defined species (open bars), geographically defined populations (hatched bars) and MYC clusters (cross-hatched bars).