Blog

The springboard for a recent news@nature.com item by Hannah Hickey “Butterflies poke holes in DNA barcodes” is a report by Gompert et al in press in Mol. Ecology on genetic differences between two subspecies of Melissa blue butterfly, Lycaeides melissa melissa and L. m. samuelis. The latter subspecies is commonly known as “Karner blue” and is listed under the USA Endangered Species Act. Analysis of mitochondrial DNA revealed some populations of Karner blue have distinct COI sequences but those populations adjacent to the range of L. m. melissa subspecies do not. This result is not surprising. For one, DNA barcoding does not aim to separate subspecies. Subspecies are geographic variants within species whose differences shade into one another so it would be surprising if any single gene showed a sharp demarcation between populations. Most subspecies do not show diagnostic genetic differences, and when such differences are found, it has often led to proposals to elevate them to species status.

Regarding the utility of DNA barcoding, the findings with Melissa blues are unremarkable, as there are cases in all animal groups studied so far in which barcoding narrows identification to a few closely-related species, but no further. For example, see my earlier entry on comparing barcode performance. It may be helpful to point out that DNA barcoding is an instrument, not a theory. Cases of partial resolution do not “disprove” barcoding or invalidate its use. In fact, one application of DNA barcoding will be to quickly highlight such cases which may be biologically interesting as they likely represent recent speciation, ongoing hybridization, or synonymy.

A more relevant Nature article that Ms. Hickey might have cited is Als et al study of Maculinea large blues, a related group morphologically similar, taxonomically confusing, and highly endangered butterflies. Large blues have “extraordinary parasitic lifestyles…later instars live in ant nests where they either devour the brood (predators), or are fed mouth-to-mouth by adult ants (cuckoos)”. Genetic analysis using mitochondrial and nuclear DNA uncovered numerous cryptic species with unsuspected host specificity, thereby both multiplying the challenge and providing the key to conservation, the need to conserve both ant hosts and butterflies.

As highlighted by the large blue study, the larger and more exciting challenge for biodiversity science will be how to incorporate the enormous number of genetically and biologically distinct forms whose discovery is facilitated by large-scale barcoding.

Photo of Rebel’s large blue Maculinea rebeli and phylogeny showing cryptic species among predatory Maculinea from Nature article by Als et al.

Suggested metric, terminology, and standard graphic

How well do barcodes distinguish among species? A standardized, simple quantitative method and terminology for comparing barcoding performance among different data sets will be helpful.

In trying to answer this question, I aim to promote terminology that does not include “error”. In my view, it generally does not make sense to talk about the error rate of barcoding. Barcoding is an instrument akin to a telescope, except that it is designed to resolve species, not stars. A telescope that does not resolve a double star is not wrong, it simply lacks sufficient resolution. Also, the term error rate implies there is an accurate reference standard in species identification. As systematists emphasize, species definitions are hypotheses and frequently undergo revision. Thus in this view barcoding performance, effectiveness, and resolution are useful descriptive terms and are more informative than barcoding error rate.

What we want is an approach that quantitatively compares barcoding with current taxonomy. In the future, taxonomy may incorporate some of the groups discovered through barcoding as recognized species, perhaps will combine some of the recognized species with overlapping barcodes into single species, and additional sequence data may enable resolution of species with overlapping barcodes. To start, a 2 x 2 table comparing recognized species to distinct barcode groups:

Suggested terminology:

Barcode group (or cluster): the shallowest branch in a neighbor-joining tree that corresponds to one or more recognized species or potential split within a recognized species.

Distinct barcodes: a barcode group that corresponds to a recognized species or a potential split within a recognized species. This definition can incorporate whatever criteria are used for recognizing splits (such as criteria that have been used to define provisional species, ESUs).

Barcode resolution: #barcode groups/total #species, in which total #species includes recognized species plus provisional species/ESUs.

This definition of barcode resolution incorporates “partially-resolved” species, so that if, for example, 8 species are resolved into 4 barcode groups, then resolution for that set would be 4/8 = 50%. Alternatively, if idea of partial resolution is not helpful, resolution could be defined more simply as a + b (green + yellow)/total #species.

Suggested graphic: Applying this to recent barcode data sets:

Suggested color scheme: As in table, green (=good!) matches current taxonomy; yellow represents novel species/provisional species/ESUs (yellow like an early bud that lacks chlorophyll), and gray (as in a gray indeterminate zone) represents recognized species with overlapping barcodes. By definition, all potential splits/ESUs have distinct barcodes, so d) in the 2 x 2 table is blank. As barcode findings are incorporated into taxonomy, I expect that the proportion that is green will increase—the greening of barcoding and taxonomy!

Mark Stoeckle