Downloaded from rspb.royalsocietypublishing.

org on October 1, 2010

Barcoding animal life: cytochrome c oxidase subunit 1

divergences among closely related species
Paul D.N. Hebert, Sujeevan Ratnasingham and Jeremy R. de Waard
Proc. R. Soc. Lond. B 2003 270, S96-S99
doi: 10.1098/rsbl.2003.0025

Supplementary data "Data Supplement"

http://rspb.royalsocietypublishing.org/content/suppl/2009/01/23/270.Suppl_1.S
96.DC1.html

References Article cited in:

http://rspb.royalsocietypublishing.org/content/270/Suppl_1/S96#related-urls

Email alerting service Receive free email alerts when new articles cite this article - sign up in the box at the
top right-hand corner of the article or click here

To subscribe to Proc. R. Soc. Lond. B go to: http://rspb.royalsocietypublishing.org/subscriptions

This journal is © 2003 The Royal Society

 Downloaded from rspb.royalsocietypublishing.org on October 1, 2010

species diversity. As such, these insects provided a chal-

lenging test for the ability of COI diversity to resolve spec-
ies boundaries.
Although Hebert et al. (2003) argued that a COI-based
identification system could be developed for all animals,
Barcoding animal life: scepticism has been expressed (Mallet & Willmot 2003).
Primary objections have focused on the concern that DNA
cytochrome c oxidase sequence differences among closely allied species will
subunit 1 divergences often be too small to allow their discrimination. Although
this issue has never been tested comprehensively, Johns &
among closely related Avise (1998) demonstrated that closely related species of
vertebrates regularly show more than 2% divergence at
species another mitochondrial gene, cytochrome b. The present
study addresses this issue further by examining the extent
of sequence diversity at COI among congeneric taxa in
Paul D. N. Hebert*, Sujeevan Ratnasingham
the major animal phyla. The most intensive analysis
and Jeremy R. deWaard
focuses on the arthropods because sequence information
Department of Zoology, University of Guelph, Guelph, for these organisms is particularly detailed owing to their
Ontario N1G 2W1, Canada
*
Author for correspondence (phebert@uoguelph.ca). high taxonomic diversity. However, COI divergences are
also examined among closely related species in all animal
Recd 09.03.03; Accptd 28.03.03; Online 15.05.03 phyla where data are available. In total, sequence diver-
gences are examined in more than 13 000 congeneric pairs
With millions of species and their life-stage trans- including representatives from 11 phyla. These results
formations, the animal kingdom provides a chal- support, with the exception of a single phylum, the con-
lenging target for taxonomy. Recent work has clusion that species-level diagnoses can routinely be
suggested that a DNA-based identification system, obtained through COI analysis.
founded on the mitochondrial gene, cytochrome c
oxidase subunit 1 (COI), can aid the resolution of
this diversity. While past work has validated the 2. MATERIAL AND METHODS
Sequences were extracted from GenBank between November 2002
ability of COI sequences to diagnose species in cer- and February 2003 for all congeneric species pairs of animals pos-
tain taxonomic groups, the present study extends sessing at least 400 bp of COI sequence from homologous sites. In
these analyses across the animal kingdom. The total, 2238 species met this criterion. Because their COI sequences
results indicate that sequence divergences at COI were acquired using varied primers, they derive from different sec-
tions of the gene. In practice, most comparisons involved either a
regularly enable the discrimination of closely allied sequence block that extended from near the 59 end of the gene to its
species in all animal phyla except the Cnidaria. This middle or a block extending from the midst of the gene to its 39 end.
success in species diagnosis reflects both the high Rates of sequence change in the 59 and 39 halves of COI were com-
pared by examining divergences in these regions for all pairwise com-
rates of sequence change at COI in most animal
parisons of the 260 animal species with full mitochondrial genomes
groups and constraints on intraspecific mitochon- in GenBank release 131.0. This analysis (results not shown) indicated
drial DNA divergence arising, at least in part, that sequence divergences in the halves were closely similar (mean
through selective sweeps mediated via interactions sequence divergence in COI-59 was 97.7% of that in the 39 region,
s.d. 6.2%). Because of this congruence, the measures of sequence
with the nuclear genome. divergence for other species pairs are analysed without reference to
their source region in the gene.
Keywords: molecular taxonomy; DNA barcode; Following the extraction of COI sequences from GenBank, p-
cytochrome c oxidase subunit 1; DNA; mitochondrial distances ( p = n d/n t where nd is the number of different nucleotides
between the two sequences and nt is the total number of nucleotides
examined) were determined for each congeneric species pair. A single
1. INTRODUCTION divergence value was obtained for all genera represented by two spec-
A final taxonomic system for the animal kingdom will ies, while n(n 2 1)/2 values were obtained for genera with three or
probably include at least 10 million species partitioned more species. The large number of sequences available for congeneric
species pairs in the Hexapoda permitted separate analysis for the
among more than a million genera. Given such high diver- four dominant orders (Coleoptera, Diptera, Lepidoptera and
sity, there is a growing realization that it is critical to seek Hymenoptera), while the remaining taxa were pooled as ‘other hexa-
technological assistance for its initial description and its pods’. Data for the chelicerates and crustaceans were more limited,
subsequent recognition (Godfray 2002; Blaxter 2003). so divergence values for these arthropods were not partitioned to the
ordinal level. Because less information was available for the Annelida,
Recent investigations have suggested the feasibility of cre- Chordata, Cnidaria, Echinodermata, Mollusca, Nematoda and Pla-
ating identification systems reliant on the analysis of tyhelminthes, these data were aggregated at a phylum level. Finally,
sequence diversity in small segments of DNA (Tautz et results for three small phyla (Bryozoa, Brachiopoda and
Onychophora) were pooled. An alignment for all of the animal COI
al. 2003). Hebert et al. (2003) focused this discussion by sequences in GenBank is available at www.uoguelph.ca/~phebert. As
proposing that a DNA barcoding system for animal life well, a list of the 447 genera examined in the present analysis,
could be based upon sequence diversity in cytochrome c together with sequence divergences (mean, maximum, minimum) for
oxidase subunit 1 (COI). They established that diversity the species pairs in each genus, is included in electronic Appendix
A, available on The Royal Society’s Publications Web site.
in the amino acid sequences coded by the 59 section of
this mitochondrial gene was sufficient to reliably place
species into higher taxonomic categories (from phyla to 3. RESULTS
orders). They also found that diversity in nucleotide COI divergences among the 13 320 species pairs ranged
sequences of the same gene region regularly permitted the from a low of 0.0% to a high of 53.7% (figure 1). While
discrimination of closely allied species of lepidopterans, a most pairs (79%) showed greater than 8% sequence diver-
group with modest rates of molecular evolution and high gence, levels did vary among higher taxonomic groups

Proc. R. Soc. Lond. B (Suppl.) 270, S96–S99 (2003) S96 Ó 2003 The Royal Society
DOI 10.1098/rsbl.2003.0025
 Downloaded from rspb.royalsocietypublishing.org on October 1, 2010
DNA divergences P. D. N. Hebert and others S97

Table 1. Mean and standard deviation of the percentage sequence divergences at COI for 13 320 congeneric species pairs in 11
animal phyla.
(The percentage of sequence divergence estimates falling in a particular range is also shown. n indicates the number of congeneric
pairs examined in each group.)

COI sequence divergence (%)

phylum n mean s.d. 0–1 1–2 2–4 4–8 8–16 16–32 321

Annelida 128 15.7 6.2 6.3 1.6 — 3.9 18.0 70.3 —

Arthropoda
Chelicerata 1249 14.4 3.6 — 0.2 0.2 2.0 50.8 46.8 —
Crustacea 1781 15.4 6.6 0.1 0.3 4.3 13.4 18.0 63.8 0.1
Coleoptera 891 11.2 3.8 2.2 1.6 3.0 8.0 74.2 11.0 —
Diptera 1429 9.3 3.5 0.9 2.1 4.1 14.0 76.2 2.7 —
Hymenoptera 2993 11.5 3.8 0.2 — 0.3 3.3 93.0 3.2 —
Lepidoptera 882 6.6 2.2 1.0 2.8 7.3 60.4 28.5 — —
other orders 1458 10.1 4.9 0.5 1.6 8.4 35.5 41.8 12.1 —
Chordata 964 9.6 3.8 1.2 0.7 4.3 19.2 61.7 12.9 —
Cnidaria 17 1.0 1.2 88.2 5.9 5.9 — — — —
Echinodermata 86 10.9 4.9 1.2 1.2 5.8 39.5 44.2 8.1 —
Mollusca 1155 11.1 5.1 1.2 1.9 4.0 15.0 67.5 10.0 0.4
Nematoda 49 11.0 2.9 — 2.0 — 22.4 73.5 2.0 —
Platyhelminthes 84 14.4 5.4 8.3 — — 4.8 44.0 42.9 —
minor phyla 154 13.3 9.7 0.6 1.3 2.6 39.6 38.3 16.9 0.7
overall 13 320 11.3 5.3 0.9 1.0 3.4 16.2 59.4 19.0 0.1

10 000 in their COI genes. In fact, more than 98% of species pairs
no. of congeneric species pairs

7918 showed greater than 2% sequence divergence. The few

2532 cnidarians included in the analysis showed far less diver-
1000 2155
gence than that typical for other animal phyla. This result
456 is not surprising as rates of mitochondrial evolution are
100 136 exceptionally low in these organisms (France & Hoover
76 2002; Shearer et al. 2002). Their stasis seems linked, at
39
10 least in part, to the presence of an excision repair system
8 absent in other animal mitochondria. Low rates of mito-
chondrial divergence also appear typical of the plant king-
1 dom (Palmer 1992). By contrast, levels of COI divergence
0.5 1 2 4 8 16 32 64
in the Fungi and Protista exceed those in animals (Ingaki
sequence divergence (%)
et al. 1998), suggesting that the prospects for extending a
Figure 1. COI sequence divergences for 13 320 congeneric COI-based identification system to these kingdoms are
pairs of animal species belonging to 11 phyla. high.
Except for the cnidarians, most congeneric species pairs
of animals showed enough sequence divergence to ensure
(table 1). The Cnidaria showed far lower COI divergences
their easy diagnosis. In fact, the mean divergence value of
than any other phylum with 94.1% of the cnidarian pairs
11.3% indicates that most pairs were separated by more
possessing less than 2% sequence divergence, while just
than 50 diagnostic substitutions in every 500 bp of their
1.9% of the species pairs in other taxonomic groups
COI gene. About 1.9% of the congeneric pairs in groups
showed such limited differentiation. This difference was
with normal rates of mitochondrial evolution showed less
highly significant (G = 489.55, p , 0.0001).
than 2% divergence, probably often reflecting short histor-
Levels of divergence varied among the four major insect
ies of reproductive isolation. Such pairs may be separable,
orders with lepidopterans showing a lower average
even in samples from natural populations where intraspec-
sequence divergence than the other three (table 1). Diver-
ific variation might be expected to cloud decisions. For
gences in the insects, as a group, were less than those in
example, lepidopteran species pairs with similarly low lev-
the other two dominant groups of arthropods, the che-
els of COI divergence were invariably resolved (Hebert et
licerates and crustaceans. Mean divergences in the other
al. 2003). Some additional cases of low divergence may
major phyla were comparable to the overall mean diver-
simply be artefacts generated by flawed identifications, but
gence value for the arthropods. Sample sizes were small
other cases of congruence will undoubtedly reflect mito-
for the minor phyla, but their mean level of divergence
chondrial introgression. In the latter situations, the analy-
appeared similar to those in the large phyla.
sis of a rapidly evolving nuclear sequence, such as the
internal transcribed spacer region of the ribosomal repeat,
4. DISCUSSION will aid taxonomic resolution.
This study has established that congeneric species of Although more work needs to be done to quantify levels
animals regularly possess substantial sequence divergence of COI divergence within animal species, a good sense of

Proc. R. Soc. Lond. B (Suppl.)

 Downloaded from rspb.royalsocietypublishing.org on October 1, 2010
S98 P. D. N. Hebert and others DNA divergences

the usual limits of intraspecific divergence in mitochon- proteins (Schmidt et al. 2001). Comparative studies of the
drial genes derives from phylogeographic analyses. Two patterns of genetic diversity in mitochondrial and nuclear
results of these studies are particularly important. First, genomes have additionally provided evidence for the
intraspecific divergences are rarely greater than 2% and depletion of mtDNA diversity via selective sweeps, per-
most are less than 1% (Avise 2000). Second, when higher haps mediated by the rise of mitochondrial variants with
divergences are observed, these variants ordinarily occur more effective nuclear interactions (Ballard 2000; Gerber
as geographical isolates, reflecting their origin in past epi- et al. 2001). These sweeps do not need to occur with high
sodes of gene pool fragmentation. Moreover, many of frequency to ensure the clear delineation of sister species.
these high divergences involve cases of taxonomic uncer- Given a mean species longevity of 5 million years, the
tainty where lineages share a species epithet, but their occurrence of sweeps at average intervals of 500 000 years
actual status is unclear (Avise & Walker 1999). Collec- would eliminate ancestral polymorphisms while allowing
tively, these phylogeographic studies do establish that the accumulation of large amounts of shallow diversity, a
intraspecific divergences are ordinarily well below those pattern congruent with that seen in nature. Interactions
that separate congeneric species pairs. Hence, COI diver- between mitochondrial and nuclear genomes should also
gences can serve as an effective tool in species recognition. ensure that horizontal transfers are ordinarily fatal. Of
Moreover, the fact that some ‘species’ show higher diver- course, cases of mitochondrial transfer have been detected
gences does not compromise the use of COI sequences between some closely allied animal species (e.g. Glemet et
for their identification. In fact, just the opposite, it allows al. 1998), but these may reflect situations in which key
delineation of the regional lineages that comprise them. nuclear genes were also introgressed, restoring func-
Concern has been expressed that efforts to base identifi- tionality to the misplaced mitochondria.
cation systems on mtDNA markers would fail because of The present results indicate that an identification sys-
the prevalence of horizontal transfers of mitochondria tem for animal life based on the COI gene will be highly
between divergent lineages and because closely allied effective. Although COI divergences appear too low to
species would regularly share mitochondrial polymor- regularly enable species diagnosis within the cnidarians,
phisms that were millions of years old (Mallet & Willmot generic-level identifications in these organisms remain a
2003). The present study suggests that such complications prospect. More importantly, the mitochondrial genomes
are rare. In fact, the clear delineation of most congeneric of closely allied species in other phyla, those that comprise
species pairs indicates a surprising ferocity of lineage prun- the bulk of animal diversity, regularly show sufficient
ing. The restricted levels of intraspecific mitochondrial sequence diversity to enable their discrimination. Even if
divergence that result from this pruning are critical to COI analysis simply generated robust generic assign-
taxon diagnosis and they may have a simple explanation. ments, its application would winnow the 10 million animal
Horizontal transfers of mtDNA and the persistence of species down to a generic assemblage averaging less than
ancestral polymorphisms require the autonomy of mito- 10 species, delivering a resolution of 99.9999% of animal
chondrial genomes. But, there is growing evidence that diversity in the process. As the present study has estab-
mitochondrial and nuclear genomes are linked in a pas de lished that species-level identifications are ordinarily
deux of surprising intimacy, best evidenced by studies on achieved, COI analysis actually provides a taxonomic sys-
cybrids. For example, despite the close genetic similarity tem that is chasing the last digit of animal diversity.
of their nuclear genomes, orang-utan mitochondria
experience a total collapse in respiratory capacity when Acknowledgements
placed in a human cell background (Barrientos et al. This work was supported by grants from NSERC and the Canada
2000). Even chimpanzee and gorilla mitochondria show Research Chairs programme to P.D.N.H. We thank Teri Crease,
Jinzhong Fu, Bob Ward and Jonathan Witt for their thoughtful com-
20% reductions in their oxidative capacity in a human ments on the manuscript.
cytogenetic setting (Barrientos et al. 1998). These effects
have been tied to interactions between gene products Avise, J. C. 2000 Phylogeography. The history and formation of species.
encoded by the nucleus and the mitochondrion; the three Cambridge, MA: Harvard University Press.
mitochondrial enzymes in the cytochrome c oxidase sys- Avise, J. C. & Walker, D. 1999 Species realities and numbers in sex-
ual vertebrates: perspectives from an asexually transmitted gen-
tem are decorated by 10 nuclear proteins critical to their ome. Proc. Natl Acad. Sci. USA 96, 992–995.
functionality (Wu et al. 2000). Similar nuclear–cytoplasmic Ballard, J. W. O. 2000 Comparative genomics of mitochondrial DNA
incompatibilities have been detected in studies that in Drosophila simulans. J. Mol. Evol. 51, 64–75.
employed recurrent backcrosses to transfer mitochondria Barrientos, A., Kenyon, L. & Moraes, C. T. 1998 Human xenomito-
chondrial cybrids. Cellular models of mitochondrial complex I
into a ‘foreign’ nuclear background. For example, deficiency. J. Biol. Chem. 273, 14 210–14 217.
backcrosses between regional groups of the copepod Barrientos, A., Muller, S., Dey, R., Weinberg, J. & Moraes, C. T.
Tigriopus californicus lead to offspring whose respiratory 2000 Cytochrome c oxidase assembly in primates is sensitive to
small evolutionary variations in amino acid sequence. Mol. Biol.
capacity is compromised (Willett & Burton 2001).
Evol. 17, 1508–1519.
Detailed molecular studies have revealed that selection Blaxter, M. 2003 Counting angels with DNA. Nature 421, 122–124.
acts particularly strongly to reconfigure mitochondrial France, S. C. & Hoover, I. L. 2002 DNA sequences of the mitochon-
gene products, apparently to optimize their interactions drial COI gene have low levels of divergence among deep-sea octo-
corals (Cnidaria: Anthozoa). Hydrobiol. 471, 149–155.
with nuclear ‘accessory’ proteins whose amino acid com- Gerber, A. S., Loggins, R., Kumar, S. & Dowling, T. E. 2001 Does
position is far more static. Moreover, the mitochondrial non-neutral evolution shape observed patterns of DNA variation
substitutions are tightly targeted. All three cytochrome c in animal mitochondrial genomes? A. Rev. Genet. 35, 539–566.
oxidase genes in primate mitochondria show elevated lev- Glemet, H., Blier, P. & Bernatchez, L. 1998 Geographical extent of
arctic char (Salvelinus alpinus) mtDNA introgression in brook char
els of non-synonymous substitutions at amino acid pos- populations (Salvelinus fontinalis) from eastern Quebec. Mol. Ecol.
itions that are closely juxtaposed to nuclear-encoded 7, 1655–1662.

Proc. R. Soc. Lond. B (Suppl.)

 Downloaded from rspb.royalsocietypublishing.org on October 1, 2010
DNA divergences P. D. N. Hebert and others S99

Godfray, H. C. J. 2002 Challenges for taxonomy. Nature 417, 17–19. Evolution of nuclear and mitochondrial encoded subunit interac-
Hebert, P. D. N., Cywinska, A., Ball, S. L. & deWaard, J. R. 2003 tion in cytochrome c oxidase. Mol. Biol. Evol. 18, 563–569.
Biological identifications through DNA barcodes. Proc. R. Soc. Shearer, T. L., Van Oppen, M. J. H., Romano, S. L. & Worheide,
Lond. B 270, 313–322. (DOI 10.1098/rspb.2002.2218.) G. 2002 Slow mitochondrial DNA sequence evolution in the
Ingaki, Y., Ehara, M., Watanabe, K. I., Hayashi-Ishimaru, Y. & Anthozoa (Cnidaria). Mol. Ecol. 11, 2475–2487.
Ohama, T. 1998 Directionally evolving genetic code: the UGA Tautz, D., Arctander, P., Minelli, A., Thomas, R. H. & Vogler, A. P.
codon from stop to tryptophan in mitochondria. J. Mol. Evol. 47, 2003 A plea for DNA taxonomy. Trends Ecol. Evol. 18, 70–74.
378–384. Willett, C. S. & Burton, R. S. 2001 Viability of cytochrome c geno-
Johns, G. C. & Avise, J. C. 1998 A comparative summary of genetic types depends on cytoplasmic backgrounds in Tigriopus californicus.
distances in the vertebrates from the mitochondrial cytochrome b Evolution 55, 1592–1599.
gene. Mol. Biol. Evol. 15, 1481–1490. Wu, W., Schmidt, T. R., Goodman, M. & Grossman, L. I. 2000
Mallet, J. & Willmot, K. 2003 Taxonomy: renaissance or Tower of Molecular evolution of cytochrome c oxidase subunit 1 in primates:
Babel? Trends Ecol. Evol. 18, 57–59. is there coevolution between mitochondrial and nuclear genomes.
Palmer, J. D. 1992 Mitochondrial DNA in plant systematics: appli- Mol. Phylogenet. Evol. 17, 294–304.
cations and limitations. In Molecular systematics of plants (ed. P. S.
Soltis, D. E. Soltis & J. J. Doyle), pp. 36–49. New York: Chap-
man & Hall. Visit http://www.pubs.royalsoc.ac.uk to see an electronic appendix to
Schmidt, T. R., Wu, W., Goodman, M. & Grossman, L. I. 2001 this paper.

Proc. R. Soc. Lond. B (Suppl.)