A Revision of Malagasy Species of Anochetus Mayr and Odontomachus Latreille (Hymenoptera: Formicidae)

Species inventories are essential for documenting global diversity and generating necessary material for taxonomic study and conservation planning. However, for inventories to be immediately relevant, the taxonomic process must reduce the time to describe and identify specimens. To address these concerns for the inventory of arthropods across the Malagasy region, we present here a collaborative approach to taxonomy where collectors, morphologists and DNA barcoders using cytochrome c oxidase 1 (CO1) participate collectively in a team-driven taxonomic process. We evaluate the role of DNA barcoding as a tool to accelerate species identification and description. This revision is primarily based on arthropod surveys throughout the Malagasy region from 1992 to 2006. The revision is based on morphological and CO1 DNA barcode analysis of 500 individuals. In the region, five species of Anochetus (A. boltoni sp. nov., A. goodmani sp. nov., A. grandidieri, and A. madagascarensis from Madagascar, and A. pattersoni sp. nov. from Seychelles) and three species of Odontomachus (O. coquereli, O. troglodytes and O. simillimus) are recognized. DNA barcoding (using cytochrome c oxidase 1 (CO1)) facilitated caste association and type designation, and highlighted population structure associated with reproductive strategy, biogeographic and evolutionary patterns for future exploration. This study provides an example of collaborative taxonomy, where morphology is combined with DNA barcoding. We demonstrate that CO1 DNA barcoding is a practical tool that allows formalized alpha-taxonomy at a speed, detail, precision, and scale unattainable by employing morphology alone.


Introduction
Anochetus and Odontomachus were treated globally by Brown [1,2]. This paper revises the genera for the Island of Madagascar and also includes new records from the Seychelles and Comoro Islands. The revision is based on morphological and CO1 sequence analysis of 500 individuals. We evaluate the role of DNA barcoding as a tool to accelerate species identification and description.
Anochetus and Odontomachus are closely related genera [1,3,4] characterized by long and straight mandibles inserted just on either side of the cephalic midline and with two or three large teeth near tip arranged in a vertical series (Figure 1a,b). The single tooth or spine at the apex of the petiole separates Odontomachus from the closely related genus Anochetus (which has two teeth or rounded margin). Odontomachus and Anochetus can also be easily distinguished by the characters on the back of the head. With head viewed from back near neck of pronotum, Odontomachus has dark, inverted Vshaped apophyseal lines that converge to form a distinct, sometimes shallow groove or ridge on upper back of head. In Anochetus, the V-shaped apophyseal lines are absent. In the same region of the back of head, however, nuchal carinae in Anochetus form an uninterrupted, inverted U-shaped ridge. In the field, small members of Anochetus might also be mistaken for Strumigenys, from which they may be distinguished by their one-segmented waist (vs. two segments in Strumigenys).
The utility of a standardized single gene for species recognition (but not phylogenetics) has been tested in an increasing swath of life. Here we tested how well a cytochrome c oxidase 1 (CO1) DNA barcode resolved species within Malagasy Anochetus and Odontomachus. In Madagascar, these ponerine genera are known to include species with independent colony formation by ergatoid (wingless) queens -and therefore are expected be a challenge for DNA barcoding using a single mitochondrial marker -but also include cases where prior taxonomy has not linked males with females and workers, nor has resolved obvious worker dimorphism as either caste variation or provisional species.
Species level taxonomy in these genera can be quite difficult. Brown [2] noted that males provide a useful source for species level delimitation. Males, however, are rarely associated with the worker castes. Brown [2:553] states: ''Unfortunately, males found associated in the nest with the female castes are known only for a minority of the species. Additional kinds of males are known from collections at light or by Malaise trap, but it has not yet been possible to link any of these securely to worker-based species.'' In this study, we used CO1 barcode sequences to associate worker, queen and male castes. We conclude that DNA barcoding will enable species delimitation, linking a greater range of the morphological diversity in ants (castes and sex), and further will provide a set of molecular characters that improve species delimitation and identification while making these hypotheses transparent and reproducible.

Methods
This revision is primarily based on arthropod surveys in Madagascar that included over 6,000 leaf litter samples, 4,000 pitfall traps, and 8,000 additional hand collecting events throughout Madagascar from 1992 through 2006 [5]. Also included are specimens from museums in Genoa, Geneva, Paris, London, Berlin, Tervuren, and Basel and the extensive collection of Gary D. Alpert located at the MCZC. Overall, this revision included the study of approximately 1,700 specimens of Anochetus and Odontomachus from 1014 recorded collecting events from throughout Madagascar with additional samples from Comoros and Seychelles. Roy Snelling (LACM) provided the records of O. simillimus from his work on the ants of Seychelles. Samples were selected for CO1 sequencing throughout the geographic range of each species. In total, 501 specimens were sequenced. Specimens examined from Madagascar are listed by increasing latitude within provinces.
All species and type material examined in this study have been imaged and are available on AntWeb (www.antweb.org). Material was deposited at the California Academy of Sciences, San Francisco (CASC); British Museum of Natural History, London (BMNH); and Museum of Comparative Zoology, Harvard University, Cambridge, Massachusetts (MCZC). All sequences, oligonucleotides and electropherograms are deposited in BOLD (www.barcodinglife.org), and sequence data has been deposited on Genbank.
In accordance with section 8.6 of the ICZN's International Code of Zoological Nomenclature, we have deposited copies of this article at the following five publicly accessible libraries: Natural History Museum, London, UK; American Museum of Natural History, New York, USA; Museum National d'Histoire Naturelle, Paris, France; Russian Academy of Sciences, Moscow, Russia; Academia Sinica, Taipei, Taiwan. The three new species names established herein have been prospectively registered in ZooBank [6][7][8], the official online registration system for the ICZN. The ZooBank publication LSID (Life Science Identifier) for the new species described herein can be viewed through any standard web browser by appending the LSID to the prefix ''http://zoobank.org/''.
New specific names in this work are attributive genitive nouns and thus invariant. Each specimen discussed below is uniquely identified with a specimen-level code (e.g. CASENT0003099) affixed to each pin. In addition, each specimen may include a collection code, which is a field number that uniquely identifies collecting events (e.g. BLF01652). Collection codes, when available, are associated with a collector and follow the collector's name.
Digital color images were created using a JVC KY-F75 digital camera and Syncroscopy Auto-Montage (v 5.0) software. All measurements were taken at 806 power with a Leica MZ APO microscope using an orthogonal pair of micrometers, recorded to the nearest 0.001 mm, and rounded to two decimal places for presentation. When more than one specimen was measured, minimum and maximum measurements and indices are presented. Measurements follow those used by Brown [1,2]. Abdominal segments are noted by ''A'' and the segment number, such as A2 for the petiole and A3 for the first gastral segment.

CO1 methods
Specimens were preserved in 95% ethanol in Madagascar and upon return to California were loaded into ScrewTop TrakMatesH boxes (Matrix Technologies) and shipped to the University of Guelph. There, DNA was extracted from tissues rich in mitochondria (e.g. legs), employing primers with high universality, and amplifying a PCR product approximately 600 bp in length. Total genomic DNA extracts were prepared from small pieces (#1 mm) of tissue using the NucleoSpinH 96 Tissue kit (Macherey-Nagel Duren, Germany), following the manufacturer's protocols. Extracts were resuspended in 30 ml of dH 2 O, and a 650base-pair (bp) region near the 59 terminus of the CO1 gene was amplified following standard protocol [9]. Briefly, full length sequences were amplified using primers (LF1-ATTCAACCAATCATAAAGATATTGG and LR1-TGATTTTTTGGACATCCAGAAGTTTA [10]). In cases where a 650 bp product was not successfully generated, internal primer pairs (LF1-ANTMR1-(see Table 1)) and (MLF1 -GCTTTCCC-ACGAATAAATAATA [11] -LR) were employed to generate shorter overlapping sequences that allowed the creation of a composite sequence (contig). PCR reactions were carried out in 96 well plates in 12.5 ml reaction volumes containing: 2.5 mM MgCl 2 , 5 pmol of each primer, 20 mM dNTPs, 10 mM Tris HCl (pH 8.3), 50 mM KCl, 10-20 ng (1-2 ml) of genomic DNA, and 1 unit of TaqDNA polymerase (PlatinumH Taq DNA Polymerase -Invitrogen) using a thermocycling profile of one cycle of 2 min at 94uC, five cycles of 40 sec at 94uC, 40 sec at 45uC, and 1 min at 72uC, followed by 36 cycles of 40 sec at 94uC, 40 sec at 51uC, and 1 min at 72uC, with a final step of 5 min at 72uC. Products were visualized on a 2% agarose E-GelH 96-well system (Invitrogen) and samples containing clean single bands were bidirectionally sequenced using BigDye v3.1 on an ABI 3730xl DNA Analyzer (Applied Biosystems). Contigs were made using Sequencher v4.0.5 (Gene Codes) and were subsequently aligned by eye in Bioedit [12]. Sequence divergences were calculated using the K2P distance model [13] and a NJ tree of distances [14] was created to provide a graphic representation of the patterning among-species divergences using MEGA3 [15], and BOLD [16]. Sequence neutrality  and rates of substitution were calculated with DnaSP v.3 [18]. Sequences and other specimen information are available in the project file ''Revision of Malagasy Anochetus and Odontomachus'' in the Published Projects section of the Barcode of Life website (www.barcodinglife.org) with complete collection information for each specimen deposited at www.antweb.org. All sequences from the barcode region have been deposited in Genebank (CO1: EF610629: EF611041, EF999925-EF999945).
A composite representation of variation within the CO1 DNA barcode for each of the eight species revised here is presented in Figures 15 and 16. We used the online program Fingerprint [19 -http://evol.mcmaster.ca/fingerprint] to illustrate the heterogeneity at a specific site within the barcode region as a percentage of the vertical line drawn to represent each base pair.
Diagnostic base pairs (or combination of base pairs) for each species within the Malagasy region are presented. This more cladistic interpretation of the DNA barcode data is very sensitive to the number of specimens analyzed -and the fewer specimens incorporated, the greater the likelihood that a rare haplotype is not reflected in the data. We present this method of analysis not to that our coverage of each species is sufficient to reflect the variation within a species, but rather to demonstrate that such an analysis is possible within a group of taxa or region, when there is good representation of the variability within a species. The nucleotide position given refers to the barcode region, and can be compared to their full mitochondrial position by adding 48 (as aligned to the Bos taurus complete mitochondrial genome sequence Genbank ref AY676873). The standard IUPAC ambiguity codes are used to denote intra-specific variation.
Complementary genetic analyses. In addition to the CO1 barcode, for some specimens we amplified portions of the rRNA gene regions: 18S, 28S (D2) and ITS1. Within the variable D2 region of 28S, the forward primer corresponds to positions 3549-3568 in Drosophila melanogaster reference sequence (Genbank M21017). Within the 18S sequence, the forward primer corresponds to positions 375-406 in Drosophila melanogaster reference sequence while the ITS1 sequence was generated using primers where the forward primer corresponds to positions 1822-  [20]. Wolbachia are obligate intracellular endosymbiotic bacteria that cause reproductive incompatibility between infected and uninfected lineages, resulting in an increased proportion of infected maternal lineages that cannot reproduce. Worker Diagnosis: Blade of mandible with five teeth and denticles located along distal two thirds of blade's length. Propodeum with short teeth (Fig. 5a). Dorsolateral margin of petiole with long spine (Fig. 5a). In frontal view, petiolar margin deeply U-shaped. Pilosity, sculpture as in Figures 2a,b.

Taxonomic synopsis
Male caste: Dorsolateral margin of petiole with acute spine. The species is most similar to A. goodmani, but can be easily distinguished by its petiole node with a pair of large apical spines.
Distribution and biology. The distribution is limited to collections made between 450 m and 750 m in rainforest in Parc National de Marojejy and 240 m from Ambanitaza near Antalaha Discussion: Specimens from Ambanitaza differ notably in shape of propodeal spines and length of spines on petiole from those of the type locality. Though these localities are quite close (40 km apart), these character differences are noticeable, and correspond to significant differences in CO1 (34 base pairs) and ITS1. While specimens from each location were invariant within 18S, there is a 7 bp insertion within ITS1 that is characteristic of the Ambanitaza population which is missing from all specimens from Marojejy. Ultimately, more collections need to be made and evaluated in order to test the hypothesis that these populations represent separate species. One important factor to consider in the testing of that hypothesis is reproductive strategy, which is, to our  knowledge, through fission. Though the queen caste is not known, based on overall similarity of workers with A. goodmani, it is likely that the queen of boltoni is wingless. Queens have never been collected during the 12 month malaise trap sampling even though males were collected. Species that reproduce by fission may show greater geographic differences in morphology and CO1.
Additional material examined for Anochetus boltoni: In addition to the type material, specimens from 4 additional collecting events from the following three localities were examined in this study. MADAGASCAR Worker Diagnosis: Blade of mandible with five teeth and denticles located at the distal half of the blade length. Petiole dorsal margin without spines. In front view, the dorsal petiolar margin flat with lateral margin rounded (Fig. 6b). Pilosity, sculpture as in The species is most similar to A. boltoni but can be easily distinguished by its petiole node without apical spines. No winged queens are known. Ergatoid queens were collected at six localities. In four of the collections, three ergatoid queens were collected in the same locality. They are very similar in size and shape to workers (Figs 2g,h), and have no ocelli (Fig. 2g). Males are not known.
Distribution and biology. A. goodmani is endemic to Madagascar and is widespread in northern and western parts of the island. It has been collected in dry forest and rainforest as low as 30 m in altitude and also in montane rainforest at the altitude 960 m on Montagne d'Ambre ( Fig. 6a), most frequently under stones (12 collections) and sifted litter (7), but also at light (1), beating low vegetation (3), rot pocket (1), in rotten log (6), ground foragers (1), ground nest (9), Malaise trap (1), on low vegetation (1), and pitfall traps (4). Discussion. Anochetus goodmani is characterized by extreme divergence within the barcode region. To date, sequencing complementary nuclear markers has provided some degree of support for the deepest CO1 divergences (between the north and south-west of Madagascar) as being separate species. Importantly however, ITS1 sequences as divergent have been produced from the same individual (Appendix S1 and Table 3). Although CO1 supports more than one operational unit within A. goodmani the hypothesis of cryptic species in relatively isolated environments requires further evidence with less ambiguity.
Additional material examined for Anochetus goodmani: In addition to the type material, specimens from 56 additional collecting events from the following 18 Table 2. Anochetus goodmani within-species pair-wise partitioning of genetic variance for the CO1 DNA barcode. K2P distances are beneath the diagonal and the number of substitutions are above the diagonal. Worker diagnosis: Inner blade of mandible without teeth and denticles; apical end of inner blade without a notched semicircular concavity (Fig. 2a). Eyes small (0.05-0.11 mm), projecting dorsolaterally. In full face view, antennal scape usually not reaching, and not surpassing posterior margin of occipital lobe. Dorsal surface of head with numerous short setae. Pilosity and sculpture as in Figures 3 a-d.
Queens alate: Very similar to workers, only slightly larger than respective size class (Figs 3e-h). Ergatoid queens not recorded.
Within a single locality, two size classes of workers, queens and males are present in this species, but the differences within a site do not hold up when variation across all sites is included. These differences suggest that two reproductive and developmental pathways can occur in this species. Further work is needed to explore the biotic or abiotic factors that trigger the development of small and large castes.
The species is most similar to A. madagascarensis but can be easily distinguished by its small eyes and scape that does not surpass the occipital lobe. A. madagascarensis has large eyes (0.24-0.26 mm), and scapes that surpass occipital lobes.
Distribution and biology. A. grandidieri is endemic to Madagascar and is widespread throughout Madagascar in forest and shrubland habitats below 1,550 m elevation (Fig 4b). It has been collected in gallery, dry, littoral, lowland, and montane forest, in desert spiny bush thicket in the southwest, and Uapaca woodland in the central plateau. As in many soil dwelling ants, A. grandidieri has reduced eyes (EL/HW 0.11-0.13) and short Worker Diagnosis: Inner blade of mandible without teeth and denticles; apical end of inner blade with notched semicircular concavity (Fig. 1a). Eyes large (0.24-0.26 mm), projecting dorsally. In full face view, antennal scape extends beyond posterior margin of occipital lobe. Dorsal surface of head asetose. Pilosity and sculpture as in Figures 4a,b. Queens alate: Very similar to worker and only slightly larger (Figs 4c,d). Queens of only one size. Ergatoid queens not recorded.
Males: Males light yellowish brown in color and with large projecting ocelli on vertex (Figs 4e,f). Males have been collected in Malaise traps in every month of the year and males have been noted to swarm and fly at dusk and early evening.
The species is most similar to A. grandidieri but can be easily distinguished by its large eyes (0.24-0.26 mm), and scapes that surpass occipital lobes.
Distribution and biology. A. madagascarensis is widespread throughout Madagascar in forest or shrubland habitats below 1100 m elevation and is also known from the Comoros. Forel's (1912:159) record of a male ''Anochetus sp.? africanus var. madagascariensis Forel'' from Seychelles, Mahé, has not been seen and confirmed. This record most likely refers to pattersoni. In Madagascar, madagascarensis is widespread and has been collected in gallery, dry, littoral, lowland, and montane forests, and in desert spiny bush thicket in the southwest of Madagascar. The longer scapes and larger eyes of A. madagascarensis compared to A. grandidieri, correlate with nesting and foraging above the soil layer. The species was most often recorded nesting in rotten logs (99 collection records) followed by sifted litter (41). In addition, it was collected from dead twigs above ground (1), rot pockets (2), ground foragers (20), ground nests (6), Malaise trap (14), on low vegetation (2), and pitfall traps (4).
CO1. Shallow intraspecific and deep interspecific divergences between A. madagascarensis and the other species. Average within species sequence divergence of 1.67% (SE = 0.055) (Fig. 16).  Worker Diagnosis: Dorsal margin of petiole node concave medially (not visible in figures of the workers but easily seen in the queen in Figure 7f.) Anterior portion of pronotal dorsum lightly sculptured compared to posterior portion of pronotum. Propodeal dorsum and angle transversely coarsely rugose, declivitous face below angle with transverse striae, with sculpture thinning near base of face; propodeum angulate in lateral view. Petiole scale broad; anterior half of first gastral tergum smooth and shiny with only fine punctures at base of setae. This species is most similar to the graeffei a widespread species across the Indo-Pacific, but differs from the latter by the pattern of sculpture on the mesosomal dorsum, shape the petiole (concave), broader petiole node as seen in lateral view, and its much larger size (HL+ML 1.99-2.12 mm in pattersoni, HL+ML,1.75 mm in graeffei).
Distribution and biology. This species is limited to the Aldabra group islands with most collections from Isle Picard. References and records to Anochetus madagascarensis [e.g.   Figures: worker 1b, 10a,b, 13c; queen 10c,d; male 11a, Worker Diagnosis: Workers of this species can be easily distinguished from troglodytes by their larger size, mandible with long, acute apical and preapical teeth and lack of extraocular furrows and temporal ridges on vertex. Brown [2] provides a description and additional references.

Check-List of Malagasy Odontomachus Species
Distribution and biology. O. coquereli is endemic to Madagascar and is restricted to eastern and northern montane rainforest, lowland rainforest, and littoral forest from 10 to 1325 m (Fig. 10a). It is most abundant at mid-elevations in the northeast such as in Marojejy National Park. Nests of O. coquereli are most commonly found in rotten logs and consist of small colonies. Queens of coquereli are wingless and very similar to workers; colonies reproduce by fission [28]. Males are collected in Malaise traps and yellow pan traps. Workers forage on the ground day and night. A few times BLF has seen solitary foragers high up on trunks and branches of large trees. It is not clear if they are foraging for plant or insect liquids up in the canopy.
There is notable geographic variation in shape of petiole, sculpture and number of preapical teeth. Preapical teeth and denticles range from 7-12. Occasionally, adjacent teeth may be fused at base to form a single bidententate tooth. However, there is no consistent concordant pattern to this variation. Molecular data are also extremely variable -suggesting that these isolated populations have long been separated. Rather than describing these populations as distinct species, we leave them here as a single species -a hypothesis that can be tested in the future with subsequent experiments in both the field and lab.
CO1: The barcode region is extremely variable (Fig. 16) -there is evident isolation by distance which is largely concordant with the biogeographic regions proposed by Wilme et al. [29].
Discussion: Odontomachus coquereli from Madagascar, the only species in the genus where winged queens have never been found.
Molet et al. [28] investigated the Marojejy population of O. coquereli, and based on demography, morphometry, allometry and ovarian dissections demonstrated that the winged queen caste has been replaced by a wingless reproductive caste and that the strategy of colonial reproduction is fission. A single wingless reproductive (ergatoid) was found in each colony. The smallest colonies consisted of at least 5 workers and the largest colonies never exceeded 40 workers, indicating a threshold size at which a colony divides in two daughter colonies. In contrast, O. troglodytes reproduces by non-claustral independent foundation and colonies can reach 1300 workers [30]. As in A. goodmani and A. boltoni, the other species without winged queens -there are deep CO1 divergences between different collection localities.
Specimens examined for Odontomachus coquereli: Odontomachus troglodytes Santschi Figures: worker 10e,f, 13a;      Because of its preference of secondary habitats, it is possible that troglodytes in Madagascar is a recent colonist from Africa, possibly introduced by humans. This is in contrast to coquereli which is most closely related to Melanesian species in the tyrannicus group.
Our collections in Madagascar were focused primarily on less disturbed habitats, thus the distribution map (Fig. 10b) probably does not reflect the full extent of its range. O. troglodytes was most often recorded nesting in rotten logs (30 collection records) followed by sifted litter (15). Males were collected at light, malaise traps, and yellow pan traps.
A lab colony was kept for a number of months and thrived on a diet of crickets, producing numerous larvae, brood, and males. The trap jaw behavior is very similar to that of O. bauri [39,Fisher unpublished]. When disturbed, the specimen use trap jaw propulsion to ''jump'' away.

Complementary analyses to CO1
In some instances we chose to amplify independent nuclear markers to help interpret CO1 divergences involving populations where specimens were morphologically cryptic. Because of their high copy number and relatively conserved primer regions, we selected three ribosomal regions to amplify: 18S, 28S and ITS1. We had high expectations for the utility of these markers to complement the mtDNA barcode analysis based on our own experiences with other taxa [40,41], the utility of these markers in other taxonomic groups where, for instance, ITS1 functions as a barcode [42], and, for 28S, based on predictions of others for the utility of this region as an alternative barcode region [43]. Unfortunately, we found that, while the CO1 data from species with exclusively (putatively) ergatoid queens had large phylogeographic signal, when compared to the three rRNA regions we utilized it was markedly simpler to generate, interpret and analyze. The rRNA markers utilized here, particularly 18S and 28S, can be useful for identifying interspecific (species as revised here) hybridization [see 40,41,43].

The role of CO1 barcoding in taxonomic revision
In traditional morphology-based taxonomy, morphologically discrete forms are tentatively recognized and hypothesized to be species. Taxonomists search for consistent phenotypic discontinuities that may indicate the occurrence of reproductive isolation. Many ant species, however, show considerable geographical variation in morphological characters. An additional complication for morphology-based taxonomy is the difference between castes within the same species, e.g. males, major and minor workers, and queens. Sequence data provide an alternative set of characters to assist in inferring species boundaries. In addition, like morphological data, hypotheses can be evaluated in light of additional data on specimen distribution, biology, and behavior. In the example of Anochetus of Madagascar, sequence data impacted the taxonomic process at the following steps: Caste association. Caste association, including male/female association, is a powerful contribution to taxonomic studies, especially for ants, which vary tremendously in morphology between sexes and castes. In this study, CO1 divergence was the principal source of data for revealing that small and large workers and queens are the same species. Though no morphological distinction in addition to size between the forms was noted, it remained unclear whether they belonged to the same species since no colony collection contained both size classes, even though they are often collected at the same site. One explanation is that small workers are produced by small queens. Small queens may represent an alternative reproductive strategy and may be only rarely produced by large queens. Further research will explore the reproductive biology of this species. The sequence data also confirmed the association of males collected in Malaise traps with the worker caste.
Type designation. The identities of many valid names are in question in Madagascar because insufficient geographic and morphological information was provided in their original descriptions, or type specimens are of uninformative minor worker castes or are damaged. For Anochetus, description of new species included the DNA barcode of a specimen from the paratype colony series to provide an additional tool for associating the name with type specimens. This facilitates linking the name to the type specimen if the identity of the type is called into question.

Evolutionary questions and biogeographic patterns.
Sequencing revealed patterns of geographic coherence and divergence that were not revealed in morphological analysis. A good example of this is the deep divergence in isolated populations of A. goodmani, and O. coquereli (see Species as hypotheses below). These results will direct future morphological and evolutionary studies on these divergent populations.
Identification. In-depth morphological study, a more timeconsuming process than the DNA analysis undertaken in this study, was applied to outliers identified by the DNA analysis. For example, in the inventory described in part I, 22 collections of Anochetus from three species were included. All were correctly identified using sequence data. Specimens within the same species that showed high sequence divergence, however, were culled for morphological scrutiny (e.g. A. madagascarensis from Amato and Binara).
Biogeography. This combination of traditional taxonomy and DNA barcoding has produced a wealth of biogeographic hypotheses to be tested. Do more basal lineages have more restricted or wider distributions, compared to younger taxa? Are evident patterns of genetic isolation by distance within the ergatoid ponerines examined here shared by all those with wingless queens? Are the mechanisms of isolation the same? Do the phylogeographic groupings correspond with the Wilme et al. [29] biogeographic regions hypothesized largely as related to primates? Taxonomy has always had this style of iterative hypothesis testing, but adding an explicit molecular component as with DNA barcoding -allows these hypotheses to be more transparent.

Species as hypotheses
The existence of any species is a hypothesis to be tested, and the transparency of species delimitation is one of the major additions that DNA barcoding brings to systematics. In our analysis, the deep sequence divergences within A. goodmani suggests that populations from the north and south of western Madagascar have a long history of isolation, and could in fact be separate species (Fig. 6). However, there are alternate hypotheses. This species has wingless queens. Species of ants that lack winged queens, reproduce by fission and have reduced dispersal ability, particularly when measured using a maternally inherited genetic marker. Thus, we might expect that those populations now restricted to isolated relict pockets of moist habitat in the dry west would show deep divergence [for example -44-46], and represent distinct, evolutionarily significant units [47], if not distinct species. By contrast in A. madagascarensis and A. grandidieri, where only winged queens have been observed, within-species sequence divergences are much lower. We are currently testing the hypothesis that female-limited dispersal has caused the extremely site-specific phylogeographic signal by assaying nuclear genes. It is possible that these populations, separated at such a large spatial scale, will show strong genetic differentiation for both nuclear and mtDNA markers between localities [44]. The CO1 analysis does not unequivocally indicate that A. goodmani is more than one species, but it does suggest future hypotheses of species membership to be tested.
Molecular approaches to species identification have been criticized for potentially overestimating [48,49], and/or underestimating biodiversity. Species diversity will be underestimated when collections include quickly evolving species-pairs [9] where interspecific divergences are less than or equal to intraspecific variation. Our data set contains one potential example of this phenomenon: individuals of Anochetus goodmani collected from Binara on the north east coast and Parc National de Kirindy Mite on the south west coast. Individuals from these populations are separated by, on average, 6.0% sequence divergence. Are these populations operating as separate species? Are these populations members of the same species but highly divergent? Our data alone cannot answer this question. But, of critical import, our data have identified a surprising level of within-species divergence and lays bare these differences to further study. A standard arthropod molecular clock for CO1 is 1.2-1.5% per million years [50][51][52]. In hymenopterans the rate for this gene is accelerated [53], and therefore average estimates should be interpreted with caution. However, the higher rates suggest that populations have been isolated for several hundred thousand years. The opportunity now exists to employ a suite of approaches (behavioral observations, tests of interbreeding, and phylogeographic resolution of more quickly evolving genetic markers) to test species membership.

CO1 and complementary genetic analyses
Of all the molecular data used here, the CO1 data was by far the easiest to generate and interpret. While an inter-gene/genomic comparison of utility was not the intent of this research, we feel it important to comment on these differences here, while presenting a full multigene phylogeographic analysis of the covariance of genetic diversity and geographic separation in another manuscript.
Within the context of a species description or revision, the relevant information here is that: 1) The CO1 barcode was very easy to generate. While the majority of specimens analyzed here are between 1-2 years old, we did generate full length barcodes (.600 bp) for specimens up to 14 years old. Barcodes were generated with the same primers and reaction conditions. Alternatively, rRNA data, variable at a species level, was often challenging to generate (i.e. sequence) due to long regions of t-repeats and uncharacterized intra-individual variation (in ITS1).

2)
We found no evidence for Numts [54] or other misplaced nuclear markers that would introduce conflict into our analysis if not spotted.

3)
CO1 sequences never showed intra-individual variation as did some of the rRNA markers. 4) Although the species described here (especially O. coquereli, A. goodmani and A. boltoni) contain large CO1 divergences, such variation is always geographically segregated, as one might expect from a species where the queens (when known) are ergatoid.
In the worst-case scenario, by describing species containing large intra-specific CO1 divergences, we have missed morphologically cryptic diversity within these species. However, the DNA data, collection records, measurements and photo-digital accessions are all preserved in publicly accessible databases, facilitating the testing (and potential refutation) of our one-species hypothesis in the traditional, iterative, process of alpha-taxonomy.

Collaborative Taxonomy
Species inventories are essential for documenting global diversity and generating necessary material for taxonomic study. However, for inventories to be relevant in the short term, the taxonomic process must reduce the bottlenecks in describing and identifying specimens. The shear diversity of arthropods can easily overwhelm an inventory system with too many specimens, the bulk of which are outside the focal expertise of the taxonomists. As an example, the NSF-funded Arthropod Inventory of Madagascar has shipped over a third of million specimens to over 150 participating taxonomic collaborators [5]. Major taxonomic products from these inventories, which will take decades to produce, represent only a fraction of the diversity collected, and provide no short-term return of biodiversity data to Madagascar.
The development of ''collaborative taxonomy'' would permit researchers to participate collectively in an accelerated teamdriven taxonomic process. Key participants in collaborative taxonomy are (i) inventory teams led by conservationists, ecologists, and taxonomists, (ii) traditional morphology-based taxonomists equipped with imaging tools, and (iii) geneticists. Under this plan, inventory teams would generate specimens and sequence data in collaboration with geneticists. Geneticists, in turn, would work directly with the taxonomist who identifies the need for additional sequencing of specimens. Taxonomists would then combine extensive sequencing data with their morphological and ecological analysis, assisted by new technologies in digital imaging and web-based delivery (e.g. www.antweb.org and www. barcodinglife.org), to infer species limits and frame evolutionary context for species.
Nothing can replace the countless hours of careful observation necessary to understand variation and to delimit species boundaries. However, the addition of sequence data provides a means to create short-term results from inventories and at the same time generate data helpful to taxonomists. For taxonomists, sequencing highlights the specimens most deserving of focused study. We tested this collaborative model by revising the ant genera Anochetus and Odontomachus of Madagascar using a combination of morphological and genetic character sets based on inventories in Madagascar.

Future
This study demonstrates how sequence data, combined with morphological analysis and innovations in imaging and web delivery, have set the stage for accelerated discovery and documentation of global species diversity. The combination of DNA sequence data with inventory and traditional taxonomy is a model that can be applied across disciplines and will allow analytical needs to scale to the enormity of the biodiversity crisis [55]. It will help in the identification and conservation of the evolutionary processes that generate and preserve biodiversity.
Little time remains to document and protect global biodiversity. Taxonomists, equipped with modern tools and collaborations, have a chance to move systematics to the forefront of conservation and the public's attention. With increased taxonomic output and improved public access and visibility, public support for the discovery of life on this planet will follow.

Supporting Information
Appendix S1 Accessions and collection information for all sequences created in this study. Appendix S2 This is the current article provided as a TaxonX XML document. TaxonX models taxonomic treatments and allows semantic enhancement so machines can understand, mine and extract the content (see http://plazi.org). Found at: doi:10.1371/journal.pone.0001787.s002 (0.11 MB RTF)