Conservation Genetics

, Volume 11, Issue 4, pp 1389–1404

Barcoding bushmeat: molecular identification of Central African and South American harvested vertebrates

Authors

    • Department of Ecology and Evolutionary Biology, N122 RamaleyUniversity of Colorado
    • Patuxent Wildlife Research CenterUS Geological Survey
  • Greta L. Meyers
    • Department of Environmental ScienceBarnard College
  • Sergios-Orestis Kolokotronis
    • Sackler Institute for Comparative GenomicsAmerican Museum of Natural History
  • Matthew S. Leslie
    • Sackler Institute for Comparative GenomicsAmerican Museum of Natural History
  • Andrew P. Martin
    • Department of Ecology and Evolutionary Biology, N122 RamaleyUniversity of Colorado
  • George Amato
    • Sackler Institute for Comparative GenomicsAmerican Museum of Natural History
Research Article

DOI: 10.1007/s10592-009-9967-0

Cite this article as:
Eaton, M.J., Meyers, G.L., Kolokotronis, S. et al. Conserv Genet (2010) 11: 1389. doi:10.1007/s10592-009-9967-0

Abstract

The creation and use of a globally available database of DNA sequences from a standardized gene region has been proposed as a tool for species identification, assessing genetic diversity and monitoring the legal and illegal trade in wildlife species. Here, we contribute to the Barcode of Life Data System and test whether a short region of the mitochondrial cytochrome c oxidase subunit 1 (COX1) gene would reliably distinguish among a suite of commonly hunted African and South American mammal and reptile species. We used universal primers to generate reference barcode sequences of 645 bp for 23 species from five vertebrate families (Crocodilidae, Alligatoridae, Bovidae, Suidae and Cercopithecidae). Primer cocktails yielded high quality barcode sequences for 179 out of 204 samples (87.7%) from all species included in the study. For most taxa, we sequenced multiple individuals to estimate intraspecific sequence variability and document fixed diagnostic characters for species identification. Polymorphism in the COX1 fragment was generally low (mean = 0.24%), while differences between congeneric species averaged 9.77%. Both fixed character differences and tree-based maximum likelihood distance methods unambiguously identified unknown and misidentified samples with a high degree of certainty. Barcode sequences also differentiated among newly identified lineages of African crocodiles and identified unusually high levels of genetic diversity in one species of African duiker. DNA barcoding offers promise as an effective tool for monitoring poaching and commercial trade in endangered species, especially when investigating semi-processed or morphologically indistinguishable wildlife products. We discuss additional benefits of barcoding to ecology and conservation.

Keywords

Barcode of lifeCaimanCrocodilesCytochrome c oxidase subunit 1 (COX1)HuntingMolecular forensicsPrimatesWildlife monitoringMaximum likelihood phylogenyUngulates

Introduction

The hunting of tropical wildlife has historically been conducted for subsistence consumption and for local trade. But current trends in wildlife harvest from across the globe suggest that the volume of extraction of wild game, or “bushmeat”, has increased considerably, and many species are in sharp decline due to over exploitation (Albrechtsen et al. 2007; Bennett et al. 2007; Milner-Gulland et al. 2003; Redford 1992). Former locally-based subsistence economies have become global, and bushmeat is now a significant export product traded at regional and international scales. Of equal importance to the deleterious impacts on wild populations and ecosystems promoted by unregulated harvest are the potential human and agricultural heath impacts stemming from the increased potential for zoonotic disease transfer (Chomel et al. 2007; Milius 2005).

Considering the international trade in bushmeat and fisheries, estimates of its worth are in excess of US$60 billion per year, with wildlife and wildlife products contributing US$5–15 billion (Baker 2008). A significant portion of this trade is illegal (US$5–8 billion, in Baker 2008), involving species that are protected by national laws and international conventions governing the use of wildlife and wildlife products. Given the illicit nature of the trade, it is difficult to accurately assess and monitor the volumes and species involved (Chomel et al. 2007; Milius 2005) and thus fully understand existing and potential impacts on economies, wildlife populations and health.

Existing legislation and treaties governing the trade in wildlife, such as the Convention on the International Trade of Endangered Species (CITES) and the United States Endangered Species Act (ESA), are based on the recognition of distinct population or taxonomic units. At a minimum, enforcement of regulations depends upon an ability to identify suspected illegal products at the species level. Accurate identification is often impaired due to the types of products involved, which are typically processed and difficult to identify using morphological techniques. To improve our ability to detect, monitor and control the trade in wildlife and wildlife products, more accurate and efficient methods of species identification are required. Identification of fish and wildlife species targeted for commercial trade is considered among the most useful applications of molecular ecology (Baker 2008).

The ability to identify wildlife products, whether as processed meat, skins or whole animals, is being formalized by the development of DNA sequence databases using a standardized gene fragment (Ratnasingham and Hebert 2007; Ross et al. 2003). A database of single-gene “barcodes” has been proposed to classify the complete diversity of life (Hebert et al. 2003a; Ratnasingham and Hebert 2007) and proponents argue that such a tool could variously be employed for defining taxonomic units for conservation (Neigel et al. 2007; but see Rubinoff 2006), biological inventory (Janzen et al. 2005) and species discovery (Bickford et al. 2007; Hebert et al. 2004; Kaila and Stahls 2006; Witt et al. 2006). The 5′ Folmer region of the mitochondrial (mtDNA) cytochrome c oxidase I (COX1) gene has been recommend as a standard for DNA barcoding (Folmer et al. 1994; Hebert et al. 2003a, b; Ivanova et al. 2007). Although there has been considerable criticism of the philosophical and practical underpinnings of DNA barcoding (DeSalle 2006; Fitzhugh 2006; Rubinoff 2006; Rubinoff et al. 2006; Song et al. 2008), its application for species identification has largely been uncontested (e.g. Rubinoff et al. 2006).

The use of sequence data for investigations of endangered species collected in commercial markets is widely reported. While the COX1 region has not been used as a standard metric, these studies have employed mitochondrial markers to identify samples to the species level. Yan et al. (2005) used the cytochrome b (CYTB) gene to identify Chinese alligators (Alligator sinensis) from fresh and partially cooked meat found in Chinese markets. Baker and colleagues used mtDNA sequences and microsatellites to identify endangered whale species sampled from markets in Korea and Japan (Baker et al. 1996, 2002). Using diagnostic characters in CYTB, Birstein et al. (1998) identified three species of caviar-producing sturgeon and discovered that nearly a quarter of commercially available caviar lots sold in New York City were mislabeled. Marko et al. (2004) also used CYTB sequences to determine that 77% of fish sold in the US labeled as red snapper were actually other species. Martin (1991) and Moura et al. (2008) used mtDNA sequences, including COX1, to identify commercially fished shark species when morphological characters (e.g. fins, heads) were equivocal in discerning among and within genera. The identification of endangered wildlife products in each of these cases would not have been possible without molecular methods because vendors had a vested interest in concealing the identity of the species being sold.

Ours is the first study to examine the utility of universal COX1 primers as a standard metric to identify multiple species for monitoring the global trade in wildlife, with particular emphasis on species commonly traded in bushmeat markets. Our study includes species from five taxonomic families: bovids (duikers and spiral-horned antelope; genera: Cephalophus, Tragelaphus), suids (red river hog; genus: Potamochoerus), cercopithecoid primates (old world monkeys and mangabeys; genera: Cercopithecus, Lophocebus), alligators (genera: Caiman, Melanosuchus, Paleosuchus) and crocodiles (genera: Crocodylus, Osteolaemus, Mecistops) (Table 1). We obtained samples either from museum collections or from in situ captures. We use fixed diagnostic characters to compare interspecific and intergeneric levels of variation and bootstrap node support from a maximum likelihood (ML) tree-based approach to assess the monophyly of closely related species. Diagnostic characters and phylogenetic support were also used to assign a small number of unidentified samples to species and to examine the museum collection and field specimens for possible errors in species identification or labeling. The poor quality of some tissue used in this study reflects the suboptimal conditions encountered when working with material sampled from bushmeat markets, processed wildlife products, or reference museum samples and, therefore, provides a robust evaluation of the utility of DNA barcoding for wildlife monitoring and investigations.
Table 1

Sample list, sample size, number of samples that failed to sequence, tissue type, sample origin, and conservation status for 11 genera included in the COX1 barcoding study. Unknown samples are listed with their respective known family classification

Order

Family

Species

Museum

Wild

Failed

Storage method

Origin

Statusa

Crocodilia

Alligatoridae

Paleosuchus palpebrosus

4

2

Tissue (blood) in buffer

Brazil

Least concern

Paleosuchus trigonatus

3

Tissue (blood) in buffer

Brazil

Least concern

Caiman crocodilus chiapasius

4

Tissue (blood) in buffer, dried blood on gauze

Costa Rica

Least concern

Caiman yacareb

8

3

Dried blood on gauze

Bolivia; Brazil; Peru

Least concern

Caiman c. crocodilus × yacare

1

Tissue (blood) in buffer

Brazil

Least concern

Melanosuchus niger

4

Dried tissue (non-blood)

Peru

Least concern

Caiman latirostris

8

2

Tissue (blood) in buffer, dried blood on gauze

Bolivia; Brazil; Paraguay

Least concern

Unknown caiman

1

Tissue (blood) in buffer, dried blood on gauze

Bolivia

 

Crocodylidae

Crocodylus acutus

1

Dried tissue (non-blood)

Oaxaca, Mexico

Vulnerable

Mecistops cataphractus

10

Dried tissue (non-blood)

Gabon; Congo

Data deficient

Crocodylus niloticus

8

Dried tissue (non-blood)

Gabon

Least concern

Crocodylus niloticus

1

Dried tissue (non-blood)

Congo

Least concernc

Osteolaemus tetraspis

1

16

3

Dried tissue (non-blood)

Gabon; Cameroon

Vulnerable

Osteolaemus osborni

11

1

Dried tissue (non-blood), tissue in buffer

Congo

Vulnerable

Unknown African crocodiled

5

4

Dried tissue (processed leather)

Unknown

 

Artiodactyl

Bovidae

Tragelaphus eurycerus eurycerus

1

Dried blood on filter paper

Ndoki, Congo

Near threatened

Cephalophus nigrifrons

5

Tissue in buffer

Ndoki, Congo

Least concern

Cephalophus leucogaster

8

1

Tissue in buffer

Ndoki, Congo

Least concern

Cephalophus dorsalise

10

1

Tissue in buffer

Kabo, Congo

Least concern

Cephalophus callipygus

26

3

Dried blood on filter paper

Ndoki, Congo

Least concern

Cephalophus monticola

26

2

Dried blood on filter paper, tissue in buffer

Ndoki, Congo

Least concern

Unknown ungulate

5

1

Dried blood on filter paper

Congo

 

Suidae

Potamochoerus porcus

8

1

Tissue in buffer

Ndoki, Congo

Least concern

Primates

Cercopithecidae

Cercopithecus cephus

7

Dried blood on filter paper, tissue in buffer

Ndoki, Congo

Least concern

Cercopithecus nictitans

8

1

Dried blood on filter paper

Ndoki, Congo

Least concern

Cercopithecus pogonias

6

Dried blood on filter paper, tissue in buffer

Ndoki, Congo

Least concern

Lophocebus albigena

8

Dried blood on filter paper

Ndoki, Congo

Least concern

Total

  

40

164

25

   

aAs established by the International Union for the Conservation of Nature’s (IUCN) Red List of Threatened Species, www.iucnredlist.org

bIncludes one sample originally identified as M. niger

cThe IUCN has yet to recognize the new crocodile species and their status remains the same as the original taxa

dFrom US Fish & Wildlife Service, National Wildlife Property Repository (http://www.fws.gov/mountain-prairie/law/property_repository.html)

eIncludes one sample originally identified as C. callipygus

Materials and methods

Species identification and sample collection

Specimens used in this study were identified and collected under four different scenarios. The first involved harvested mammals surveyed during a prior study of bushmeat hunting in the Republic of Congo (Table 1; Eaton 2002). Species identifications were made by MJE and trained field assistants using African mammal guides (Estes 1991; Kingdon 1997) in conjunction with the knowledge of local hunters. Voucher photographs were taken of all species and later confirmed using Nowak (1999). All samples, with the exception of five unidentified ungulates, were collected from freshly killed, whole animals in which species identification was unambiguous. The five unidentified samples were collected from partially butchered animals that could only be confirmed as belonging to ungulates based on hair pattern and coloration. Samples of Osborn’s dwarf crocodiles (Osteolaemus osborni) and a single Nile crocodile (Crocodylus niloticus) were collected from intact, hunted animals during a subsequent study in the Republic of Congo (Eaton et al. 2009; Thorbjarnarson and Eaton 2004). The second sample collection scenario consisted of capture and identification of live African crocodiles for systematics and ecological research in the Republics of Congo and Gabon (Eaton 2006; Eaton and Barr 2005; Eaton et al. 2009; Thorbjarnarson and Eaton 2004). Although there have been recent revisions to the taxonomy of all African crocodiles (Eaton et al. 2009; Hekkala 2004; McAliley et al. 2006; Schmitz et al. 2003), identification to genus (and in most cases to species) was unambiguous. The third source of samples was voucher crocodilian specimens obtained from museum collections. These included dried blood and tissue from six species of South American caiman, the American crocodile (Crocodylus acutus) and the Central African dwarf crocodile (Osteolaemus tetraspis) provided by the Yale Peabody Museum (YPM) and the American Museum of Natural History (AMNH; see Table 2 for details and accession numbers). One sample was identified only as ‘caiman’ (YPM 15394) and analyzed as an unknown species. The final source of samples used in this study was wildlife products confiscated by the US Fish and Wildlife Service as illegal imports and donated to this project as research material. Aside from originating in Africa, the specimens (five crocodile skin handbags) contained no additional information on location or species. These unidentified samples were included in the study to test the ability of standard barcode primers to amplify highly degraded material. Field-collected tissue samples were stored in 10% buffered EDTA-DMSO and kept at room temperature for up to several months before being stored at −20°C; blood samples were applied directly to Whatman filter paper (Florham Park, NJ), then dried and stored at room temperature for nearly 10 years.
Table 2

Accession and locality information for museum and US Fish and Wildlife crocodilian specimens sequenced for COX1 barcoding

Study ID

Institution accession #a

Species

Country

Locality

Collection year

Sex

Caiman1

YPM 15394

Caiman

Bolivia

Tarija Dept

1986

 

C. c. chiapasius3

YPM 15709

Caiman crocodilus chiapasius

Costa Rica

Puntarenas Province

1993

M

C. c. chiapasius4

YPM 15713

Caiman crocodilus chiapasius

Costa Rica

Puntarenas Province

1993

M

C. c. chiapasius16

YPM 15741

Caiman crocodilus chiapasius

Costa Rica

Alajuela Province

1993

 

C. c. chiapasius17

YPM 15743

Caiman crocodilus chiapasius

Costa Rica

Alajuela Province

1993

 

C. yacare7

YPM 15402

Caiman crocodilus yacare

Bolivia

SantaCruz Dept

1989

 

C. yacare8

YPM 15401

Caiman crocodilus yacareb

Bolivia

El Beni Dept

1986

 

C. yacare15

YPM 15683

Caiman crocodilus yacare

Brazil

Mato Grosso State

1987

 

C. yacare32

YPM 15570

Caiman crocodilus yacare

Brazil

Mato Grosso Estate

1987

F

C. yacare33

YPM 15410

Caiman crocodilus yacare

Bolivia

Pando Dept

1986

 

C. yacare34

YPM 15775

Caiman crocodilus yacareb

Paraguay

Misiones Dept

1986–1987

 

C. yacare35

YPM 15669

Caiman crocodilus yacareb

Brazil

Mato Grosso State

1986–1987

 

C.c. crocodilus × yacare2

YPM 15547

C.c.crocodilus × yacare

Brazil

Amazonas State

1987

F

C. latirostris5

YPM 15754

Caiman latirostris

Paraguay

Presidente Hayes Dept

1986–1987

 

C. latirostris6

YPM 15755

Caiman latirostris

Paraguay

Neembucu Dept

1986–1987

 

C. latirostris18

YPM 15392

Caiman latirostris

Bolivia

Tarija Dept

1986

 

C. latirostris19

YPM 15554

Caiman latirostrisb

Brazil

Espirito Santo State

1989

 

C. latirostris20

YPM 15551

Caiman latirostrisb

Brazil

Espirito Santo State

1989

 

C. latirostris21

YPM 15548

Caiman latirostris

Brazil

Mato Grosso du Sol

1987

M

C. latirostris22

YPM 15754

Caiman latirostris

Paraguay

Presidente Hayes Dept

1986–1987

 

C. latirostris23

YPM 15393

Caiman latirostris

Bolivia

Tarija Dept

1986

 

M. niger9

YPM 15833

Melanosuchus niger

Peru

Madre de Dios Dept

1984

 

M. niger10

YPM 15834

Melanosuchus niger

Peru

Madre de Dios Dept

1984

 

M. niger24

YPM 15695

Melanosuchus niger

Peru

Madre de Dios Dept

1987

F

M. niger25

YPM 15832

Melanosuchus nigerc

Peru

Madre de Dios Dept

1984

 

M. niger26

YPM 15835

Melanosuchus niger

Peru

Madre de Dios Dept

1984

 

P. palpebrosus11

YPM 15703

Paleosuchus palpebrosus

Brazil

Mato Grosso State

1987

 

P. palpebrosus12

YPM 15702

Paleosuchus palpebrosusb

Brazil

Rondonia State

1987

 

P. palpebrosus28

YPM 15697

Paleosuchus palpebrosusb

Brazil

Rondonia State

1987

 

P. palpebrosus29

YPM 15704

Paleosuchus palpebrosus

Brazil

Mato Grosso State

1988

 

P. trigonatus13

YPM 15705

Paleosuchus trigonatus

Brazil

Rodinia State

1987

F

P. trigonatus31

YPM 15700

Paleosuchus trigonatus

Brazil

Rodinia State

1987

 

P. trigonatus14

YPM 15699

Paleosuchus trigonatus

Brazil

Rodinia State

1987

 

C. acutusM7

AMNH R100634

Crocodylus acutus

Mexico

Oaxaca

1967

 

O. tetraspisM11

AMNH R75421

Osteolaemus tetraspis

Cameroon

Unknown

Unknown

 

1USFWS

USFWS PB800

African crocodile1d

Africa

Unknown

Unknown

 

2USFWS

USFWS C5236

African crocodile2b

Africa

Unknown

Unknown

 

3USFWS

USFWS C7848

African crocodile3b

Africa

Unknown

Unknown

 

4USFWS

USFWS C6688

African crocodile4b

Africa

Unknown

Unknown

 

5USFWS

USFWS C3508

African crocodile5b

Africa

Unknown

Unknown

 

aYale Peabody Museum (YPM); American Museum of Natural History (AMNH); US Fish & Wildlife Service National Wildlife Property Repository (USFWS)

bUnsuccessfully sequenced

cDetermined to be C. yacare

dDetermined to be C. niloticus

DNA extraction, amplification and sequencing

Samples included in the study had been stored as preserved fresh tissue, dried tissue, blood dried in buffer, blood dried on filter paper, and processed skin products. DNA was extracted from tissue and blood using DNEasy kits (Qiagen) in a pre-PCR laboratory to prevent contamination. Extractions followed the manufacturer’s protocol for buffered animal tissues but were modified slightly for extraction of DNA from dried blood or older tissue to ensure maximum yield from low-quality samples. Modifications included incubating the lysed tissue at 65°C for 15 min after adding AL buffer and incubating again at 4°C for 1 h after adding ethanol. To maximize final yield of genomic DNA, 75 μl of AE buffer, preheated to 70°C, was added and left to incubate for 45 min before centrifuging and collecting flow-through. Because the processed crocodile leather products were the most difficult to extract, we assessed their DNA yield using a NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific).

Sequence data were generated using one of three universal COX1 primer “cocktails” (Table 3; Ivanova et al. 2007). Primer selection for each taxon was made through polymerase chain reaction (PCR) optimization of cocktails ‘COI-1,’ ‘COI-2’ and ‘COI-3’ and one individual primer (VF1d_t1/VR1d_t1) using representative samples from each genus of artiodactyl, primate and crocodilian included in the study, including positive control samples previously sequenced at other gene regions. Table 3 summarizes the taxon-specific primer combinations. PCR was performed in Mastercycler EP gradient S thermocyclers (Eppendorf) in a 25 μl reaction volume containing 1.0 μl genomic DNA (~25 ng/μl), 1× PCR Buffer (Fisher Scientific), 0.24 μM dNTPs, 15 ng BSA, 1 U Taq polymerase (Fisher Scientific), and 0.4 μM of each forward and reverse primer or primer cocktail (Integrated DNA Technologies). Failed amplifications were repeated under the same conditions with 2 μl of genomic DNA. Generally, the optimized PCR thermal cycling profile was a step-up protocol of 94°C for 3 min, 5 cycles of 94°C for 30 s, 51.1°C for 40 s and 72°C for 1 min, followed by 30 cycles of 94°C for 30 s, 56.9°C for 40 s and 72°C for 1 min, with a final extension at 72°C for 10 min. The COI-1 primer cocktail required a different annealing temperature, (54.2°C) for the first 5 cycles. PCR products were visualized by agarose gel electrophoresis, and the cocktail or primer pair yielding the brightest, thinnest band was chosen for each taxon. PCR products were purified with Ampure magnetic beads (Agencourt Bioscience) on a BioMek FX robotic platform (Beckman Coulter Inc.). Primer cocktails COI-2 and COI-3 included a modified forward and reverse universal M13 tag (Ivanova et al. 2007). A single primer pair, M13(−21) and M13(−27), was used for bidirectional sequencing reactions of PCR products generated from these two cocktails using BigDye v1.1 chemistry (Applied Biosystems Inc.) on a 3730xl DNA Analyzer (Applied Biosystems, Inc.). Forward and reverse sequences were assembled and edited in Sequencher 4.6 (Gene Codes Corp.) and verified by eye. Contig sequences were aligned using ClustalW (Thompson et al. 1994) as implemented in MEGA 4.0 (Tamura et al. 2007).
Table 3

Primers, primer cocktails and universal M13 tails used to sequence all samples included in this study. Table and primer information modified from Ivanova et al. (2007). Position of M13 tail indicated by [M13F] or [M13R]

Name

Ratio

Cocktail name/Primer sequence 5′-3′

COI-1 duikers (Cephalophus spp.), red river hog (Potamochoerus porcus), gray-cheeked mangabey (Lopohocebus albigena)

  FF2d

1

TTCTCCACCAACCACAARGAYATYGG

  FR1d

1

CACCTCAGGGTGTCCGAARAAYCARAA

COI-2 guenons (Cercopithecus spp.), in bold; bongo (Tragelaphus eurycerus)

  LepF1_t1

1

[M13F]ATTCAACCAATCATAAAGATATTGG

  VF1_t1

1

[M13F]TCTCAACCAACCACAAAGACATTGG

  VF1d_t1

1

[M13F]TCTCAACCAACCACAARGAYATYGG

  VF1i_t1

3

[M13F]TCTCAACCAACCAIAAIGAIATIGG

  LepRI_t1

1

[M13R]TAAACTTCTGGATGTCCAAAAAATCA

  VR1d_t1

1

[M13R]TAGACTTCTGGGTGGCCRAARAAYCA

  VR1_t1

1

[M13R]TAGACTTCTGGGTGGCCAAAGAATCA

  VR1i_t1

3

[M13R]TAGACTTCTGGGTGICCIAAIAAICA

COI-3 (crocodilians)

  VF2_t1

1

[M13F]CAACCAACCACAAAGACATTGGCAC

  FishF2_t1

1

[M13F]CGACTAATCATAAAGATATCGGCAC

  FishR2_t1

1

[M13R]ACTTCAGGGTGACCGAAGAATCAGAA

  FR1d_t1

1

[M13R]ACCTCAGGGTGTCCGAARAAYCARAA

  M13F (−21)

 

TGTAAAACGACGGCCAGT

  M13R (−27)

 

CAGGAAACAGCTATGAC

Sequence analysis

Unidentified crocodilian and ungulate samples were sequenced but then set apart from the remaining data set. All sequences obtained from species identified through museum collections or by independent morphological evaluation were grouped into their respective taxa and examined for diagnostic molecular characters that could be used as identifiers in future studies. Because species misidentifications are possible in field studies, as well as in the accession of voucher museum specimens, we examined both field and museum samples for obvious errors in species designation. Samples with possible mistaken identities were compared with all other known species to determine if a match could be made based on diagnostic characters. Because of small sample sizes for several taxa and the associated problems of overestimating diagnostic character sites (Brower 1999; Davis and Nixon 1992), we also confirmed misidentified specimens by means of their placement on phylogenetic trees. We used maximum likelihood (ML) phylogenetic inference as implemented in RAxML 7.0.4 (Stamatakis 2006) using the general time-reversible (GTR) substitution model (Lanave et al. 1984; Rodriguez et al. 1990) with rate heterogeneity parameters modeled by the Gamma (Γ) distribution and four rate categories (Yang 1994). Node support was evaluated with 100 rapid bootstrap replicates (Stamatakis et al. 2008). Outgroups included the American alligator (Alligator mississippiensis) for both Crocodylidae and Alligatoridae, the bongo (Tragelaphus eurycerus eurycerus) for the ungulate phylogeny, and the rhesus macaque (Macaca mulatta) for primates. Resulting trees were inspected for monophyletic groupings and for phylogenetic support values of nodes subtending possibly misidentified samples. Mistakes that could unambiguously be attributed to a species based on matching of diagnostic characters and phylogenetic placement were included in their respective taxonomic group for subsequent analyses.

Sequence variability in COX1 was evaluated at three hierarchical levels: among conspecifics, among congeneric species and among genera within each of the three orders included in this study. The bongo (T. eurycerus eurycerus) and the American crocodile (Crocodylus acutus), each represented by only one sample, were excluded from these analyses. Intraspecific comparisons quantified nucleotide diversity for each species based on average nucleotide substitutions per site (transitions + transversions) using a Tamura-Nei model with pairwise sequence comparisons (Tamura and Nei 1993). To compare congeneric species, we identified fixed nucleotide character differences for every species pair within each represented genus and calculated species divergence using the same Tamura-Nei substitution model. This same approach was used to compare average genus-level nucleotide divergence among orders. Positions that contained a fixed character state among all individuals within a species and an alternate but fixed state at the homologous site in the second species were considered diagnostic at the species level. We used MEGA for genetic diversity and divergence calculations and for visual examination of homologous character state positions among congeneric species. We assessed species monophyly and divergence among sister taxa by observing the resulting ML trees and node support values.

Treating the unknown ungulate, caiman and crocodile samples as a test case for species identification, we compared their sequences against diagnostic characters from our set of known species and included these samples in a reanalysis of ML phylograms. Unidentified samples were assigned to described species based on three criteria. First, an unidentified sample had to be included within a monophyletic group with bootstrap support exceeding 95%. Second, its sequence had to be at least 98% similar to the most common haplotype from a described taxon. Third, its inclusion into a particular monophyletic group had to preserve diagnostic character sites previously identified in that group. We considered these criteria sufficiently conservative to prevent type I errors (incorrectly attributing a sample to the wrong species).

Results

Primer selection and sequencing of degraded tissue and blood

The universal primers of Ivanova et al. (2007) reliably sequenced a 645 bp fragment of COX1 mtDNA for all mammal and reptile species included in this study. Based on gel visualization, the COI-I cocktail worked optimally for duikers (Cephalophus spp.), the red river hog (Potamochoerus porcus) and the gray-cheeked mangabey (Lophocebus albigena). COI-2 worked best for the bongo, while a single primer pair from this cocktail [V(F,R)1d_t1] was optimal for the guenons (Cercopithecus spp.). Cocktail COI-3 worked for all species of crocodilians (Table 3).

Because monitoring of the bushmeat trade using DNA barcoding identification will often require amplification of degraded tissue samples, we evaluated the success of universal COX1 primers on samples varying widely in age and curation method. The crocodile skin products proved difficult, with widely varying but generally low average DNA concentrations (13.5 ng/μl, SD = 13.9) and low purity (average 260:280 nm = 1.46, SD = 0.34). We were able to extract template DNA from two of the five leather products but only one of these produced a bidirectional COX1 sequence (DNA yield = 16.35 ng/μl). We were able to obtain high-quality sequence data from a total of 179 of 204 samples tested (87.7%; Table 1). Of 43 samples that originally failed to sequence, seven were successfully sequenced following a second round of PCR amplification and 11 others produced sequences after a re-extraction of genomic DNA. Of the 25 samples that ultimately failed to yield sequences, nine produced visible PCR bands of the approximate molecular weight of the COX1 fragment, but failed to sequence even after a second round of extraction and amplification. Final clean trace files of all sequenced samples (with the exception of six previously unidentified specimens, Table 1) were contributed to the BOLD database (Ratnasingham and Hebert 2007) at http://www.barcodinglife.org and sequences deposited on NCBI GenBank under accession numbers: GQ144467–GQ144639.

Correction of misidentified specimens

Based on diagnostic sites and branch placement on the ML tree, we discovered two discrepancies in species identification. One caiman sample (M. niger25) aligned unambiguously with Caiman yacare (Fig. 1a), but was labeled as belonging to Melanosuchus niger. It is unclear whether the specimen’s label was switched in the lab or the collections, misidentified in the field or incorrectly accessioned; the YPM reptile staff is investigating (G.J. Watkins-Colwell, pers. comm.). One ungulate sample collected in the field and labeled as Peter’s duiker (C. callipygusYF42) was identical to the most common haplotype of the bay duiker (Cephalophus dorsalis; Fig. 1b).
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-009-9967-0/MediaObjects/10592_2009_9967_Fig1_HTML.gif
Fig. 1

Maximum likelihood phylograms for a South American caimans, b African duikers, and c African crocodiles. Unknown, blind or misidentified samples are shown in bold type. Node support values are based on filtering the best maximum likelihood tree through 100 rapid bootstrap replicate trees. Log-likelihood and alpha shape parameter values, respectively, were a −1,949.953015 and 0.199779, b −1,888.9429 and 0.02, and c −1,965.181261 and 0.248764

Intraspecific sequence variation

Within-species nucleotide diversity (average rate of nucleotide substitutions per site) in the COX1 gene ranged from 0.0 to 1.92% (Table 5). The greatest amount of variability was observed in the Nile crocodile (Crocodylus niloticus). However, recent studies of this species using several mtDNA and nuDNA markers support a species-level division between northwestern and southeastern Africa (Hekkala 2004; Schmitz et al. 2003), though the geographic extent of each clade is not yet resolved. Eight of our Nile crocodile samples were captured in Gabon and, when we analyzed this geographic group separately, exhibited <0.01% nucleotide diversity (Table 5).

We observed the second greatest amount of intraspecific diversity (1.18%) in 26 specimens of Peter’s duiker (Cephalophus callipygus; Table 5). A subset of eight C. callipygus exhibited fixed differences at 13 out of 21 variable sites found within the larger group (Table 5) and formed a paraphyletic assemblage with moderate bootstrap support (<85%, Fig. 1b). When polymorphisms were evaluated separately for the two groups, diversity was 0.41% for the eight individuals and 0.05% in the remaining 18 individuals (Table 5). To reduce the likelihood that we had sequenced a pseudogene in any of the C. callipygus samples, we inspected the reading frames and found no premature stop codons. All substitutions were synonymous, thus not resulting in amino acid replacement.

Congeneric species differences

Based on the results of previous phylogenetic (Eaton et al. 2009) and morphological (Brochu 2007) studies, we treated African dwarf crocodiles (genus Osteolaemus) from Congo and Gabon as distinct species. We also considered the spectacled caiman (Caiman crocodilus chiapasius) and Yacare caiman (C. yacare) to be distinct taxa (Busack and Pandya 2001; Vasconcelos et al. 2006) and allowed for the possibility that our Nile crocodile (Crocodylus niloticus) samples represent distinct geographic lineages, based on the recent studies noted above.

The number of potentially diagnostic nucleotide characters between congeneric species ranged from 87 sites (13.5% of the sequenced fragment length) between mustached (Cercopithecus cephus) and crowned (C. pogonias) guenons, to only four characters (0.6% of the COX1 fragment) between Caiman yacare and Caiman crocodilus chiapasius (Table 4). Nucleotide divergence values for these species pairs were 0.164 and 0.013, respectively (Table 5). C. yacare showed some evidence of paraphyly with C. c. chiapasius (bootstrap support = 87%), based on one C. yacare sample (C. yacare7), while the C. c. chiapasius samples themselves grouped more strongly (96%; Fig. 1a).
Table 4

Nucleotide positions and character states of diagnostic sites in a 645 bp COX1 gene fragment. Twenty-one species from five tropical vertebrate mammals and reptiles are represented (see Table 1 for details on taxa and sample numbers and localities). Ambiguity codes denote that this position for a given species is not fixed and that the homologous position for another taxon is diagnostic

https://static-content.springer.com/image/art%3A10.1007%2Fs10592-009-9967-0/MediaObjects/10592_2009_9967_Tab4a_HTML.gif
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-009-9967-0/MediaObjects/10592_2009_9967_Tab4b_HTML.gif
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-009-9967-0/MediaObjects/10592_2009_9967_Tab4c_HTML.gif
Table 5

Inter- and intraspecific nucleotide differences in the COX1 gene for three orders of tropical mammals and reptiles. The lower triangular matrix quantifies the number of fixed, diagnostic nucleotide positions between species pairs within each order; values in the upper matrix represent pairwise nucleotide divergence (Tamura-Nei) across the 645 bp fragment of COX1. Diagonal values (in bold) are the average number of base substitutions per site (×100) for intraspecific comparisons

 

Crocodilia

1

2

3

4

5

6

7

8

9

10

11

1.

M. niger (4)

0.000

0.132

0.129

0.083

0.082

0.086

0.229

0.222

0.212

0.243

0.232

2.

P. palpebrosus (2)

73

0.310

0.066

0.120

0.110

0.118

0.229

0.214

0.213

0.230

0.254

3.

P. trigonatus (3)

74

37

0.100

0.133

0.110

0.119

0.217

0.212

0.208

0.235

0.256

4.

C. latirostris (6)

49

66

75

0.100

0.078

0.082

0.225

0.208

0.211

0.235

0.247

5.

C. yacare (5)

48

62

66

45

0.230

0.013

0.212

0.217

0.208

0.235

0.246

6.

C. c. chiapasius (4)

50

64

67

46

4

0.420

0.225

0.227

0.216

0.244

0.249

7.

M. cataphractus (10)

126

125

121

124

122

126

0.030

0.136

0.136

0.135

0.152

8.

C. niloticus-Gabon (8)a

123

118

119

116

123

128

79

0.070

0.053

0.173

0.182

9.

C. niloticus-Congo (2)a

114

113

112

114

113

118

76

31

0.160

0.178

0.205

10.

O. osborni (10)

131

125

128

128

129

134

77

96

95

0.000

0.098

11.

O. tetraspis (14)

131

139

111

138

136

139

85

100

106

57

0.280

 

Artiodactyla

12

13

14

15

16

17

18

19

   

12.

C. nigrifrons (5)

0.00

0.103

0.072

0.131

0.082

0.082

0.085

0.327

   

13.

C. dorsalis (9)

61

0.06

0.091

0.130

0.097

0.098

0.098

0.289

   

14.

C. leucogaster (7)

41

51

0.34

0.124

0.093

0.092

0.099

0.290

   

15.

C. monticola (24)

74

69

66

0.53

0.111

0.109

0.119

0.305

   

16.

C. callipygus (23)

44

46

46

54

1.18

0.264

   

17.

C. callipygus grp1 (15)

50

56

52

58

0.05

0.024

0.318

   

18.

C. callipygus grp2 (8)

49

55

54

59

13

0.41

0.328

   

19.

P. porcus (7)

140

128

125

120

124

132

134

0.13

   
 

Primates

20

21

22

23

       

20.

C. cephus (7)

0.150

0.111

0.164

0.240

       

21.

C. nictitans (7)

62

0.540

0.157

0.203

       

22.

C. pogonias (6)

87

72

0.340

0.194

       

23.

L. albigena (8)

118

99

102

0.110

       

aAnalyzed as a single species, intraspecific variability of C. niloticus was 1.92%

The average number of diagnostic sites for the 17 congeneric species pairs was 52.3 (8.1% of fragment length; SD = 18.4 sites), corresponding to an average genetic divergence of 0.098 (SD = 0.036). Position and character states of fixed nucleotide sites are presented in Table 4. Among the three orders compared, average congeneric nucleotide divergence was smallest in the crocodilians (0.065, SD = 0.03; excluding monotypic genera Melanosuchus and Mecistops) and highest among the three cercopithecoid primate species (0.144, SD = 0.029; excluding monotypic Lophocebus). Average pairwise divergence among the five duiker species (0.104, SD = 0.02) was intermediate between primates and crocodilians. Figure 2 demonstrates the range of intraspecific nucleotide diversity found in this study as compared to sequence divergence between congeneric species pairs.
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-009-9967-0/MediaObjects/10592_2009_9967_Fig2_HTML.gif
Fig. 2

Frequency histogram of intraspecific COX1 sequence diversity (“species”) compared to nucleotide divergence between pairs of congeneric sister species analyzed in this study. Only sister species Caiman yacare and C. crocodilus chiapasius had a divergence value overlapping the range of intraspecific nucleotide polymorphisms. The taxonomy of these two species is still being investigated

Higher-order comparisons

Comparing sequences between genera revealed average divergence values ranging from 0.084 (CaimanMelanosuchus) to 0.212 (CercopithecusLophocebus) within families, to a high of 0.295 between Suidae (Potamochoerus) and Bovidae (Cephalophus; Table 6). The greatest divergence between Alligatoridae and Crocodylidae was Paleosuchus to Osteolaemus (0.244; Table 6).
Table 6

Average nucleotide divergence (Tamura-Nei) between genera within orders for a 645 bp fragment of COX1 mtDNA

Order (Family)

Genera compared

Divergence

Artiodactyla

Cephalophus–Potamochoerus

0.295

Primates

Cercopithecus–Lophocebus

0.212

Crocodilia

Caiman–Crocodylus

0.214

Caiman–Mecistops

0.221

Caiman–Osteolaemus

0.243

Melanosuchus–Crocodylus

0.217

Melanosuchus–Mecistops

0.229

Melanosuchus–Osteolaemus

0.238

Paleosuchus–Crocodylus

0.212

Paleosuchus–Mecistops

0.223

Paleosuchus–Osteolaemus

0.244

(Alligatoridae)

Caiman–Melanosuchus

0.084

Caiman–Paleosuchus

0.118

Melanosuchus–Paleosuchus

0.131

(Crocodylidae)

Crocodylus–Osteolaeumus

0.184

Crocodylus–Mecistops

0.136

Osteolaemus–Mecistops

0.143

Diagnosis of unknown samples

We determined the species identity of the unknown crocodilian and ungulate samples by matching ostensibly fixed character sites and by placement and likelihood support in the phylogenetic trees. The sequence for the unknown caiman (Caiman1) matched the most common haplotype of Caiman latirostris and fell within this monophyletic group with 100% node support (Fig. 1a). The museum sample identified as a hybrid caiman (C. c. crocodilus × yacare) grouped with C. c. chiapasius + C. yacare7, but with low bootstrap support (34%; Fig. 1a, node value not shown). This sample, however, matched with C. yacare at all diagnostic sites distinguishing the two species (Table 4). The crocodile skin handbag (1USFWS) matched closely (99.8% identity) with the one Nile crocodile sampled in the Congo. Differentiation of these two samples from Nile crocodiles collected in Gabon received high bootstrap support (99%, Fig. 1c). Comparing these putative groupings revealed 31 potentially fixed character differences and a sequence divergence of 0.053 between Nile crocodile clades (Table 4, 5). One unidentified ungulate sample (Uniden2) matched the most common Cephalophus monticola haplotype, while the remaining three (Uniden10, 11, 15) matched the most common C. callipygus haplotype. Phylogenetic placement of all four unknown ungulate samples was unambiguous (Fig. 1b). No unknown or misidentified primate samples were detected and therefore the phylogeny is not shown.

Discussion

Methodological considerations

Universal primer cocktails compiled by Ivanova et al. (2007; Table 3) successfully amplified the targeted barcode region for all eleven genera included in this study. High quality sequence data were generated from older specimens and those stored under suboptimal conditions, including dried tissue and blood maintained at room temperature for more than 20 years. DNA extraction and sequencing of confiscated crocodile leather products proved difficult, however, with only one of five skins identified. The tanning process used to preserve the crocodile leather, although apparently minimal, likely contributed to DNA degradation. Following additional optimization of PCR thermal profiles and reaction volumes, and re-extraction of difficult samples, we were ultimately unable to produce sequences from 12.3% of tissue samples. Tissue stored in buffered solution required the least procedural modifications for amplification and sequencing. Specimens stored as dried tissue or dried blood on gauze stored longer than 15 years were the most unreliable in producing high quality sequences. Problems associated with amplification of degraded DNA are well known (Deagle et al. 2006; Shapiro et al. 2004; Vuissoz et al. 2007) and it is unlikely that recovering full-length COX1 barcodes (~650 bp) from processed animal products will be routinely successful. Hajibabaei et al. (2006) recognized this limitation and designed compatible primers to sequence short COX1 barcode fragments (221 and 134 bp) in Lepidoptera and Hymenoptera. Appreciating the importance of a varied molecular toolkit, we are using the taxa included in the current study to develop and test a suite of mini-barcode primers to be paired with the existing universal 5′ primer set of Ivanova et al. (2007) to sequence fragments in the range of 150–350 bp. Initial trials on samples with low probabilities of producing full barcode sequences (leather products and museum specimens) are promising and appear to capture sufficient numbers of diagnostic characters for species identification (data not shown). The primer sets and our results will be reported in a subsequent publication.

Intraspecific variation

Intraspecific variation in the barcode region was low and generally fell within reported ranges of within-species mtDNA divergence (rarely exceeding 2%, Avise 2000). In this study, we encountered two species with genetic distances approaching the expected upper range of intraspecific variation. We observed the greatest amount of nucleotide diversity (1.92%) in the Nile crocodile (C. niloticus). This crocodilian is now considered to be two genetically distinct lineages based on multiple lines of evidence (Hekkala 2004; Schmitz et al. 2003). Although the previous studies did not include COX1, we sequenced the 12S rRNA gene for Nile crocodiles from Congo and Gabon (Eaton, unpublished data) to compare with the lineages described by Schmitz et al. (2003) and Hekkala (2004). We found that Gabon crocodiles grouped strongly (ML bootstrap support = 96%) with the assemblage including South Africa, Madagascar, Kenya, Zimbabwe, Sudan and Egypt (GenBank accession numbers: AY195943, AY195945, AY195946, AY195950–AY195955). The Congo specimen and handbag aligned with samples from Senegal, Chad, the Gambia and Mauritania (AY195944, AY195947–AY195949, AY195956, AY195957) with similar bootstrap support (96%). Although it is not possible to compare Nile crocodile COX1 barcodes directly to clades produced by the 12S rRNA gene, we infer that the differentiation of barcode samples from Congo and Gabon corroborate the earlier studies and represent a geographic division between the former conspecific crocodiles.

Peter’s duiker (Cephalophus callipygus) was the second species with relatively high levels of intraspecific diversity. A subgroup of eight individuals displayed putative fixed characters at more than half of the variable sites for this species and formed a paraphyletic assemblage (Fig. 1b). Although NuMts (mitochondrial pseudogenes in the nucleus) are not easy to diagnose in cases of noncoding DNA (Kolokotronis et al. 2007), they often contain premature stop codons and frame-shift mutations (e.g. Chung and Steiper 2008; Lemos et al. 1999). Our sequence data contained neither, suggesting NuMts are unlikely to explain the observed sequence variation and phylogenetic pattern. In an extensive DNA barcode survey of birds, Kerr et al. (2007) were able to detect a low prevalence of pseudogenes due to their generally reduced size (100–200 bp) and, presumably, disrupted reading frames. Hybridization with another duiker species is also an unlikely explanation for these findings, as our sample set included all sympatric duikers found in the Republic of Congo with the exception of the yellow-backed duiker (C. silvicultor), a much larger species not likely to interbreed with C. callipygus. Our observation of cryptic genetic variation in Peter’s duiker is corroborated by at least one other molecular study which found two major clusters of C. callipygus in forests of central Gabon, although the authors offered no discussion of this observation (van Vliet et al. 2008).

Because inference on divergence was based on a single mitochondrial locus within a sympatric population of C. callipygus, we cannot reject the possibility that our observations may conflict with the genome tree or true phylogeny for this species due to introgressive hybridization or incomplete lineage sorting (Funk and Omland 2003). Such concerns are reduced in the case of the Nile crocodile, for which analysis of additional genes (including nuclear loci) corroborate those of COX1 in refuting an imperfect species taxonomy (Hekkala 2004; Schmitz et al. 2003). However, since our results are reported for allopatric populations of Nile crocodiles, introgression may in fact be an important consideration in future analyses if single-gene phylogenies are constructed from specimens sampled in putative hybrid zones.

Interspecific variation

While a character-based approach has been recommended as an improvement over distance-based thresholds for species identification (Rach et al. 2008; Rubinoff et al. 2006), the interspecific differences we present here are based on both distance metrics and diagnostic nucleotide characters. Mean congeneric nucleotide divergence in our study was 9.8% and the average number of fixed characters was 52.3. Variation between congeneric species pairs substantially exceeded levels of intraspecific polymorphism and fell within the range of genetic divergences reported for a wide range of vertebrate taxa. Hebert et al. (2003b) found that 93.8% of vertebrate congeners had a sequence divergence between 4 and 32% (mean = 9.6%) for the same region of COX1 while Johns and Avise (1998) reported average genetic distances for congeneric mammal and reptile species generally exceeded ~3% in the mitochondrial cytochrome b gene. One exception observed here was between Caiman yacare and C. crocodilus chiapasius, whose taxonomy is still under debate, but for which a small number of diagnostic characters (n = 4) may reliably identify individuals to species (Table 4).

Identification of unknown and misidentified samples

The primary intent of this work was to evaluate whether COX1 barcoding would serve as a reliable means to identify wildlife species sampled during bushmeat monitoring or while investigating trade in embargoed wildlife products. Matching homologous diagnostic sites and phylogenetic methods unequivocally assigned all unknown samples to the correct species, including caiman, crocodiles and ungulates. Barcode sequences also identified errors in identification or labeling that occurred during field collection or in museum accessioning.

Conclusions

Our results suggest that with minimal effort and simple refinements to DNA extraction and PCR protocols, accurate barcode sequence data can be obtained from most wildlife products encountered in bushmeat monitoring programs and wildlife investigations. Sequencing shorter barcode fragments should increase the success of working with degraded DNA samples and we have begun work to modify universal primers that will contribute to this effort for a wide range of taxa. Strong phylogenetic support and the high frequency of fixed character states between closely related taxa offer convincing evidence that COX1 barcoding gene will reliably diagnose many common African and neo-tropical bushmeat species. Bushmeat monitoring and investigations of wildlife commercialization and trade are likely to benefit from this molecular approach, especially when sampling from semi-processed products (e.g. from urban or import markets), when working with products that are difficult to identify (e.g. bird and reptile eggs, fish species, skins), or when focusing in areas containing morphologically cryptic species. In an effort to reduce inevitable errors while conducting field research and in museum curation, mtDNA barcoding offers a simple, low-cost and accurate method for verifying species identities. Although additional independent lines of evidence are needed to substantiate the levels of divergence observed in Peter’s duiker (C. callipygus), barcodes may have highlighted a novel evolutionary lineage worthy of further investigation. Finally, generating a database of barcode sequences for tropical wildlife will offer researchers, conservationists and managers an effective tool for more precisely delineating the extent, range and genetic diversity of species of concern.

Acknowledgments

This work was supported by the Alfred P. Sloan Foundation and the Richard Lounsbery Foundation. The National Science Foundation and the American Museum of Natural History’s Research Experience for Undergraduates Program supported the laboratory work of GLM. The AMNH sponsored Aritra Datta and Arlene Amador to extract and sequence the USFWS crocodile skin products and we thank them for their efforts. Field support was provided to MJE by the Wildlife Conservation Society’s Congo and Gabon programs, the National Geographic Society, the Rufford Foundation, Lincoln Park Zoo’s Asia & Africa Fund and the Mac-Arthur Program of the University of Minnesota. MJE thanks Paul and Sarah Elkan (WCS-Congo), Debora Pires, and Congo field assistants Yamba Flavien, Bienvenu Kimbembe and Rufin Lekana. New World crocodilian samples were collected by Peter Brazaitis, Carlos Yamashita (IBAMA, Brazil) and George Rebelo (INPA, Brazil) and provided by G. J. Watkins-Colwell of the Peabody Museum. Ellen Bean and three anonymous reviewers greatly improved the clarity and scope of this manuscript.

Supplementary material

10592_2009_9967_MOESM1_ESM.htm (22 kb)
Supplementary material

Copyright information

© Springer Science+Business Media B.V. 2009