Cataloging the very large number of undescribed species of insects could be greatly accelerated by automated DNA based approaches, but procedures for large-scale species discovery from sequence data are currently lacking. Here, we use mitochondrial DNA variation to delimit species in a poorly known beetle radiation in the genus Rivacindela from arid Australia. Among 468 individuals sampled from 65 sites and multiple morphologically distinguishable types, sequence variation in three mtDNA genes (cytochrome oxidase subunit 1, cytochrome b, 16S ribosomal RNA) was strongly partitioned between 46 or 47 putative species identified with quantitative methods of species recognition based on fixed unique (“diagnostic”) characters. The boundaries between groups were also recognizable from a striking increase in branching rate in clock-constrained calibrated trees. Models of stochastic lineage growth (Yule models) were combined with coalescence theory to develop a new likelihood method that determines the point of transition from species-level (speciation and extinction) to population-level (coalescence) evolutionary processes. Fitting the location of the switches from speciation to coalescent nodes on the ultrametric tree of Rivacindela produced a transition in branching rate occurring at 0.43 Mya, leading to an estimate of 48 putative species (confidence interval for the threshold ranging from 47 to 51 clusters within 2 logL units). Entities delimited in this way exhibited biological properties of traditionally defined species, showing coherence of geographic ranges, broad congruence with morphologically recognized species, and levels of sequence divergence typical for closely related species of insects. The finding of discontinuous evolutionary groupings that are readily apparent in patterns of sequence variation permits largely automated species delineation from DNA surveys of local communities as a scaffold for taxonomy in this poorly known insect group.

Mitochondrial genome sequences are important markers for phylogenetics but taxon sampling remains sporadic because of the great effort and cost required to acquire full-length sequences. Here, we demonstrate a simple, cost-effective way to sequence the full complement of protein coding mitochondrial genes from pooled samples using the 454/Roche platform. Multiplexing was achieved without the need for expensive indexing tags (‘barcodes’). The method was trialled with a set of long-range polymerase chain reaction (PCR) fragments from 30 species of Coleoptera (beetles) sequenced in a 1/16th sector of a sequencing plate. Long contigs were produced from the pooled sequences with sequencing depths ranging from ∼10 to 100× per contig. Species identity of individual contigs was established via three ‘bait’ sequences matching disparate parts of the mitochondrial genome obtained by conventional PCR and Sanger sequencing. This proved that assembly of contigs from the sequencing pool was correct. Our study produced sequences for 21 nearly complete and seven partial sets of protein coding mitochondrial genes. Combined with existing sequences for 25 taxa, an improved estimate of basal relationships in Coleoptera was obtained. The procedure could be employed routinely for mitochondrial genome sequencing at the species level, to provide improved species ‘barcodes’ that currently use the cox1 gene only.

The demand for scientific biodiversity data is increasing, but taxonomic expertise is often limited or not available. DNA sequencing is a potential remedy to overcome this taxonomic impediment. Mitochondrial DNA is most commonly used, e.g., for species identification (“DNA barcoding”). Here, we present the first study in arthropods based on a near-complete species sampling of a family-level taxon from the entire Australian region. We aimed to assess how reliably mtDNA data can capture species diversity when many sister species pairs are included. Then, we contrasted phylogenetic subsampling with the hitherto more commonly applied geographical subsampling, where sister species are not necessarily captured. We sequenced 800 bp cox1 for 1,439 individuals including 260 Australian species (78% species coverage). We used clustering with thresholds of 1 to 10% and general mixed Yule Coalescent (GMYC) analysis for the estimation of species richness. The performance metrics used were taxonomic accuracy and agreement between the morphological and molecular species richness estimation. Clustering (at the 3% level) and GMYC reliably estimated species diversity for single or multiple geographic regions, with an error for larger clades of lower than 10%, thus outperforming parataxonomy. However, the rates of error were higher for some individual genera, with values of up to 45% when very recent species formed nonmonophyletic clusters. Taxonomic accuracy was always lower, with error rates above 20% and a larger variation at the genus level (0 to 70%). Sørensen similarity indices calculated for morphospecies, 3% clusters and GMYC entities for different pairs of localities was consistent among methods and showed expected decrease over distance. Cox1 sequence data are a powerful tool for large-scale species richness estimation, with a great potential for use in ecology and β-diversity studies and for setting conservation priorities. However, error rates can be high in individual lineages.

We investigate mitochondrial (COI, 16S rDNA) and nuclear (ITS2, 28S rDNA) genetic structure of North East Atlantic lineages of Terebellides, a genus of sedentary annelids mainly inhabiting continental shelf and slope sediments. We demonstrate the presence of more than 25 species of which only seven are formally described. Species boundaries are determined with molecular data using a broad range of analytical methods. Many of the new species are common and wide spread, and the majority of the species are found in sympatry with several other species in the complex. Being one of the most regularly encountered annelid taxa in the North East Atlantic, it is more likely to find an undescribed species of Terebellides than a described one.