Blog

In 9 September 2008 Proc Natl Acad Sci USA researchers from Brigham Young University and University of South Carolina report that nuclear pseudogenes, if not excluded from analysis, can confuse COI DNA barcoding studies.  To my reading, this study re-iterates a well-understood hazard and proposes remedies that are already standard in most phylogenetic DNA work including DNA barcoding. 

Pseudogenes, first described by Jacq, Miller, and Brownlee in 1977, are non-functional genes that presumably arose from ancient duplication events and subsequent loss of function through accumulation of mutations. In sequencing studies, pseudogenes of protein coding genes are usually easily distinguished from their functional counterparts as they harbor insertions, deletions, and/or point mutations that interrupt the reading frame.

Pseudogenes derived from mitochondrial DNA, often called numts (nuclear copies of mtDNA) were first reported by Gellissen et al in 1983. A search of NCBI PubMed for “mitochondrial pseudogenes” shows 282 articles and 12 review articles over the past 25 years.

Song and colleagues analyzed mitochondrial COI sequences in grasshoppers (single individuals of four species representing different Acrididae subfamilies) and cave crayfish (119 individuals of four species in genus Orconectes collected at 56 localities in southeastern US).  Most of the analyses involved sequencing of cloned PCR products, which adds a level of complexity and is unlike any DNA barcoding study I am aware of. To skip to the conclusion, the authors emphasize that if numts generated by PCR amplification of mtCOI are NOT excluded, then it will confuse DNA barcoding or other phylogenetic studies. Since most of the numts generated in this study were easily recognized I do not understand why they did so much work (in all they sequenced 125 grasshopper clones and 560 crayfish clones) to reach this sensible but obvious conclusion.

First, grasshoppers. The authors amplified a subsegment of the COI barcode region (439 vs 648 bp in full-length barcode region; shorter amplicons are more likely to represent pseudogenes). The amplified products from the four individual grasshoppers were cloned, and 30 clones/species were sequenced, generating an average of 15 unique haplotypes per species. Of these, 97.3% had stop codons, meaning they could be immediately excluded as not representing true mtCOI sequences.  A full-length barcode sequence was amplified from 1 species, and cloned products yielded 19 paralogues (ie obvious pseudogenes).

Second, crayfish. The researchers amplified the full-length COI barcode region from 172 individuals using Folmer primers. “For 93 individuals, we were able to obtain clean COI sequences; however, 79 individuals from southern populations of O. australis and O. barri yielded ambiguous sequences.” To my reading, the next step would be to stop there and find different primers or PCR conditions that did not generate ambiguous sequences (indicating that more than one COI-like template was being amplified). Instead the authors proceeded to clone products from individuals that yielded ambiguous results and also from those with clean sequences “to determine whether numts were present but not being detected without cloning.” Not surprisingly, they found probable numts in all 4 species of crayfish, and interestingly some of the clones did NOT contain stop codons (ie might be mistaken for functional COI sequences). These apparent numts, which might be easily overlooked, came from the 2 species with ambiguous results on sequencing of uncloned products, which I take as further evidence that it would have been better to develop a different COI amplification protocol, assuming the goal is to accurately determine the barcode sequence.

Among other quality control standards in Barcode of Life Database (BOLD), COI sequences with stop codons, such as found in most pseudogenes in this study, are automatically flagged, signalling the researcher to re-check the data.  

Finally, it may be that some of what the authors call numts instead reflect heteroplasmy, ie differences among individual mitochondrial DNAs. Like static noise generated when you turn the volume up all the way, cloning is likely to reveal various mutations in some of the 10^17 or so mitochondrial genomes present in eukaryotic organisms. Looking ahead, it seems to me that the authors have missed an opportunity to contribute protocols or sequences that could be applied by other researchers to DNA barcoding of grasshoppers or crayfish.

In Pacific Fishing September 2008 (issue available on newstands, not yet on web) two New York City teenagers, Kate Stoeckle (my daughter!), 19, and classmate Louisa Strauss, 18, apply DNA-based identification to fish sold in their Manhattan neighborhood. The girls purchased 60 items from 14 establishments and sent samples to University of Guelph where graduate student Eugene Wong performed DNA barcode analysis. 14 (25%) of 56 samples with recoverable DNA were mislabeled, in all cases as more expensive or more desirable fish. Mislabeled items were sold at 2 of 4 restaurants and 6 of 10 grocery stores/fish markets. 

The frequency of mislabeling and the ease of high-schoolers obtaining DNA-based species identifications captured public interest. Their study was featured on page 1 in New York Times on August 22, one of the “quotes of the day” on CNN/Time website, a live segment on CBS TV Early Show, an interview on national public radio, and has appeared in over 350 print and news items and blogs from 34 countries in 10 languages so far, with particularly heavy coverage in China, Korea, and Japan, presumably related to dietary importance of fish in general and sushi in particular. 

This response demonstrates how powerful the FishBOL database is already (30,665 barcodes from 5,463 species so far, which represents about 20% of world fish), and hints at enormous uses that DNA barcoding will have as technology gets smaller and cheaper.

I will be away from Blog until mid-August.

COI barcodes aim to enable identification of species, assigning unknown specimens to known species, and helping flag genetically divergent organisms that may represent new species. Might barcodes also help understand relationships among species?

Here I look at one example from birds, comparing differences among COI barcodes to a recently revised phylogeny of terns (subfamily Sternini). According to American Ornithologists’ Union Check-list of North American Birds Supplement 47 (2006), “the data show that the genus Sterna as currently defined…is paraphyletic.”…”[W]e follow the recommendation of Bridge et al 2005 to resurrect four generic names currently placed in synonomy with Sterna.” The figure at left, taken from the 2005 paper by Bridge, Jones, and Baker, shows phylogenetic relationships based on 2800 bp of mtDNA from 33 species of terns (Bayesian tree with ML distances and ML boostrap support indices), and is juxtaposed to an NJ tree of COI barcodes from 29 of the same 33 taxa. The figure is colored according to the revised generic assignments (AOU 2006).

The topology of the COI NJ tree is similar to the larger data set tree, including that all currently recognized genera are reciprocally monophyletic, and most show similarly high boostrap values as in the Bayesian/ML analysis based on the larger data set. 

Of course mitochondrial DNA is widely used in analyzing relationships among animal species, including birds. Most of these studies are focused on relatively small groups of species, such as the tern study cited here. With growing DNA barcode libraries it will be increasingly possible to get at least a preliminary look at genetic relationships for large numbers of species (so far 2,393 avian species (24% of world birds) have barcode records in BOLD). This could be exciting!