DNA barcode and minibarcode identification of freshwater fishes from Cerrado headwater streams in Central Brazil

The extraordinary species diversity of the Neotropical freshwater fish fauna is world renown. Yet, despite rich species diversity, taxonomic and genetic resources for its Cerrado ichthyofauna remain poorly developed. We provide a reference library of 149 DNA barcodes for 39 species/lineages of Cerrado headwater stream fishes from the Brazilian Distrito Federal and nearby areas and test the utility of distance-based criteria, tree-based criteria and minibarcodes for specimen identification. Mean Kimura 2-parameter genetic distances within species to orders ranged 1 center dot 8-12 center dot 1%. However, mean intraspecific v. congeneric-interspecific distances (0 center dot 9-1 center dot 3%) overlapped extensively and distance-based barcoding failed to achieve correct identifications due to c. 4-12 center dot 1% error rates and 19 center dot 5% ambiguous identifications related to the presence of singletons. Overlap was reduced and best-match success rates improved drastically to 83 center dot 5% when Characidium barcodes representing potential misidentifications or undescribed species were removed. Tree-based monophyly criteria generally performed similarly to distance methods, correctly differentiating up to c. 85% of species/lineages despite neighbour-joining and Bayesian tree errors (random lineage-branching events, long-branch attraction). Five clusters (Ancistrus aguaboensis, Characidium spp., Eigenmannia trilineata, Hasemania hanseni and Hypostomus sp. 2) exhibited deep intraspecific divergences or para-/polyphyly and multiple Barcode Index Number assignments indicative of putative candidate species needing taxonomic re-examination. Sliding-window analyses also indicated that a 200 bp minibarcode region performed just as well at specimen identification as the entire barcode gene. Future DNA barcoding studies of Distrito Federal-Cerrado freshwater fishes will benefit from increased sampling coverage, as well as consideration of minibarcode targets for degraded samples and next-generation sequencing.