DNA barcoding is a recent technique that employs one or a few short, universal DNA regions to place sampled individuals into named species and to identify individuals as belonging to putatively undescribed species (http://www.barcodeoflife.org/). DNA-based identification promises a range of applications, including identification of organisms at cryptic life stages (e.g., seeds, seedlings, larvae), source identification of plant or animal parts (e.g., plant foodstuffs, herbal medicines, meats and furs from CITES-protected species), forensics, and surveys of poorly known biological communities e.g., tropical rainforests, deep-sea communities, microbial communities .
Although DNA barcoding of animals using mitochondrial genes has been done with high success , plants have proven to be somewhat recalcitrant to DNA barcoding. The low sequence variation in the plant mitochondrial genome has led to a search for alternative universal DNA barcodes for plants, which has proven difficult . Most genes tested as universal plant DNA barcodes are within the plastid genome, and a small number of them are becoming increasingly popular . A single gene is unlikely to provide enough resolution to differentiate all plant species , yet six plastid genes in combination still fail to discriminate all species within the genus Crocus. Other candidates include nuclear ribosomal genes ; however, their utility as DNA barcodes may be limited by incomplete concerted evolution, fungal contamination, and amplification failure . Although DNA barcoding using multiple genes has proven successful with high resolution for phylogenetically diverse communities, e.g., Panamian trees (98% species discrimination, ), Mesoamerican orchids and Kruger National Park trees (> 90% species identification, ), but see Gonzalez et al. (< 70% species identification,  for Amazonian trees), barcoding studies of single clades have had limited success, e.g., 43.5% species discriminated in Bromeliaceae , and 32% in Fraxinus. Alternative candidates for DNA barcodes are low-copy nuclear genes, which have received little attention (i.e., ). Problems expected with such genes include the design of universal primers, gene duplications, recombination, allopolyploidy and heterozygosity .
The native flora of Hawai‘i boasts extreme endemism (89% for angiosperms ) and offers a unique opportunity to evaluate DNA barcoding on species and communities of different ages. The Hawaiian flora is a rich but young assemblage, with the majority of lineages originating on the main islands within the past five million years . Hawaii’s main islands are part of a broader volcanic chain and span a natural age gradient from 0.5 to 5 my . Although many of Hawaii’s endemic plant lineages span the main islands, most species are restricted to a single island , and their maximum ages can be set to the age of the island on which they occur. As such, the Hawaiian flora allows estimation of a species-age or a community-age threshold below which DNA barcoding fails to delineate species.
To our knowledge, DNA barcoding has not been attempted on the Hawaiian flora, and, plastid genes (the most popular DNA barcodes) have comparatively been little used for phylogenetic studies of Hawaiian radiations, presumably due to a near absence of variation in these genes. Instead, most studies have used ribosomal genes [17–21] or in a few cases low-copy nuclear genes [22, 23], finding moderate levels of variation at these genes. One of the notable exceptions is the phylogeographic study of Metrosideros by Percy et al.  based on 10 plastid genes; nonetheless, these genes failed to fully resolve the evolutionary history of the genus within the Hawaiian Islands. The availability of DNA barcodes for the Hawaiian flora is particularly desirable as many native species are difficult to distinguish through vegetative characters alone, many are threatened by extinction, and hybridization appears to be common.
Our study focused on two plant genera of the Hawaiian Islands: Clermontia (Campanulaceae) and Cyrtandra (Gesneriaceae). Clermontia is an endemic genus of rainforest shrubs that are either epiphytic or terrestrial with bird-pollinated flowers that show great variation in flower morphology among species. The genus belongs to the Hawaiian lobeliads, the largest Hawaiian radiation , and comprises 22 species , most of which are found on the youngest islands of Maui and Hawai‘i (Big Island). Cyrtandra is a genus of understorey shrubs with somewhat uniform, white and probably insect-pollinated flowers and great variation in vegetative characters, although the adaptive significance of the latter is generally not clear. The genus comprises 53 species , all endemic, more or less evenly distributed among the main islands, with O‘ahu having the greatest diversity. In both genera, most species are restricted to a single island, and numerous cases of hybridization have been reported. Furthermore, circumscription of species with multiple-island distributions is often disputed. The estimated crown ages for Clermontia and Hawaiian Cyrtandra are 3.2 My  and 5.2 My , respectively.
In this study, we evaluate the potential of low-copy nuclear genes as DNA barcodes in Clermontia and Cyrtandra, and discuss some of their advantages and disadvantages compared to frequently used plastid genes. We did not attempt to identify universal barcodes, but rather conducted a pilot study to see how such markers would be informative.