Samples

Skull samples and measures were obtained from specimens in museum collections, those collected in the field, or animals provided by indigenous hunters. Skull and tissue samples obtained from collected specimens in Brazil were approved by Ministério do Meio Ambiente, Sistema de Autorização e Informação em Biodiversidade—SISBIO (number: 21055-2; issue date: 11 December 2009). No permit was required for field sampling carried out in French Guiana. A sample of Tapirus indicus, Desmarest, 1819, was provided by the San Diego Zoo, San Diego, California. Sequence data for other specimens and species from different countries were retrieved from GenBank (Benson et al. 2005), or published elsewhere (de Thoisy et al. 2010).

Morphological studies

Canonical variate analysis (multivariate analysis of variance [MANOVA]–canonical variate analysis [CVA]) of 22 cranial measurements (Supporting Information S1, DOI:  10.1644/12-MAMM-A-169.S1) was performed with the software PAST version 2.14 (Hammer et al. 2001), using the Hotelling's Bonferroni-corrected option to test the discrimination of each of the species. Data were available for all living species: T. terrestris: n = 52, T. indicus: n = 3, T. bairdii: n = 4, T. pinchaque (Roulin, 1829): n = 3, Tapirus sp. nov.: n = 8 (for a list of the specimens see Supporting Information S2, DOI:  10.1644/12-MAMM-A-169.S2). Some fossil species were included, which represent all of the South American and several North American taxa known from skulls complete enough to take most, if not all, the measurements used. Although some skulls of Tapirus sp. nov. collected by the Karitiana Indians were partially damaged, so some measurements are missing, they were included in some of the analyses. Qualitative comparison of discrete characters was derived from available specimens, museum collections, and the literature. A figure including the measurements, their descriptions, and a list of specimens used in the analyses are provided in Supporting Information S1 and S2.

Morphological cladistic analysis was performed using TNT version 1.1 (Goloboff et al. 2008) to infer phylogenetic relationships of the tapirs using parsimony with the script aquickie.run, provided with the TNT package. The matrix (see Supporting Information S3, DOI:  10.1644/12-MAMM-A-169.S3) included 15 taxa and 60 cranial, dental, and postcranial characters, modified from Hulbert and Wallace (2005), with characters treated as unordered (the character list is shown in Supporting Information S4, DOI:  10.1644/12-MAMM-A-169.S4). The in-group included all known living species and 6 fossil Tapirus from the Miocene to the Pleistocene of North and South America. Out-groups included the extinct genera Plesiotapirus and Paratapirus, following previous propositions (Hulbert 1999).

Molecular phylogenetic analyses

The following total sample numbers, or sequences, or both, were included in the analyses: T. terrestris: n = 52, T. bairdii: n = 3, T. pinchaque: n = 5, T. indicus: n = 4, Tapirus sp. nov.: n = 4. Equus caballus or Rhinocerotidae, or both, were used as out-groups.

The DNA sequences from 3 mtDNA genes (cytochrome b [Cytb], cytochrome oxidase I [COI], and cytochrome oxidase II [COII]) from living Tapirus species were generated and compared to previously published data (Ashley et al. 1996; Norman and Ashley 2000; de Thoisy et al. 2010). We included a thorough analysis of the Cytb in a large geographic sampling in South America, such as has been shown to be useful in recovering mammalian phylogeny and exposing cryptic species (Baker and Bradley 2006; Redondo et al. 2008). Sequence alignments were performed using Clustal W 2.1 (Larkin et al. 2007) and edited or concatenated, or both, when necessary with the Alignment Explorer function implemented in MEGA 5.1 (Tamura et al. 2011). The resulting alignments were used in phylogenetic reconstructions by 2 methods. First, a maximum-likelihood search was performed in PhyML 3.0.1 (Guindon et al. 2010) using the GTR+Γ+inv model with parameters estimated from the data by the maximum-likelihood approach. The initial tree was a BioNJ tree and the search was performed using both NNI and SPR rearrangements in each interaction keeping only the tree with best likelihood score. After a full PhyML search on tree topology and parameters, the maximum-likelihood tree was used again as an initial tree for a new run and the search was repeated until no significant gain in likelihood was observed in a likelihood-ratio test. The confidence in the clades was assessed through the approximate likelihood-ratio (aLRT) test (Anisimova and Gascuel 2006; Anisimova et al. 2011). Second, we used a Bayesian inference method with MrBayes 3.2.1 (Ronquist et al. 2011) using a GTR+Γ+inv model of nucleotide evolution and assuming flat Dirichlet priors for parameter estimates and unconstrained branch lengths. The search was carried out using 2 independent runs of 4 Markov chains (1 cold and 3 heated) for 3 × 10⁶ generations, sampling every 300 generations with a burn-in of 25% of the samples for the estimates of topology and parameters. Stationarity of the chains was checked by plotting parameters against generation in Tracer 1.5 (Rambaut and Drummond 2007).

We also carried out phylogenetic reconstructions based on maximum-parsimony criteria for the Cytb data set using the software TNT. Additionally, we used the algorithm Median-Joining (Bandelt et al. 1999), as implemented in the software Network 4.6 (Fluxus Technology Ltd., Suffolk, England), to build a Cytb haplotype network to infer all the most-parsimonious relationships among haplotypes, which allowed the inclusion of multiple allele states and observation of homoplastic reticulations.

Divergence times

Two methods were used to estimate divergence times among the Tapirus lineages. In the 1st, we used the Linetree method of Takezaki–Rzhetsky–Nei (Tamura et al. 2007), which assumes a strict molecular clock. We assumed an evolutionary substitution rate of 2.5% per million years (Nabholz et al. 2008). Furthermore we used estimated dates from the fossil record to calibrate the molecular clock and set boundaries in some of the clade splits on the tree to refine the estimates (see Hulbert 1999; Spassov and Ginsburg 1999; Holanda and Cozzuol 2006; Ferrero and Noriega 2007; Holanda et al. 2011; and the morphological analysis below). These age boundaries determined by fossil dates were set in the split between the Asian and American tapirs (15 ± SD 5 million years ago [mya]), for the divergence between T. bairdii and the South American tapirs (7.5 ± 1 mya), and also the diversification of South American tapirs (0.13 ± 0.1 mya). In a 2nd method, we used a Bayesian estimation implemented in BEAST 1.5.3 (Drummond et al. 2007), with a relaxed molecular clock using the same fossil calibrations above to set the means and standard deviations of a normal distribution used as prior probabilities for the nodes' ages (Drummond and Rambaut 2006). We also used a strict clock model without constraints on the divergence times of the nodes, with the evolutionary rate described above (2.5%). All estimates were congruent.

We have deposited all newly generated sequences in GenBank and the COI sequences of all Tapirus species in the Barcodes of Life Database (Ratsinagan and Hebert 2007). GenBank accession numbers for sequences generated in this work are GU593658–GU593682 and GU737551–GU737565. Details can be found in Supporting Information S5 (DOI:  10.1644/12-MAMM-A-169.S5).

Etymology

Arabo kabomani signifies tapir in the Paumarí native language from southern Amazonas, Brazil, where the holotype was collected in December 2009.

Holotype

Universidade Federal de Minas Gerais (UFMG) 3177, a complete skeleton and parts of the skin of a young adult male, with complete fused epiphyses in long bones, but with 3rd molars unerupted (Figs. 1a–e). This specimen has the long-bone epiphyses and vertebral disks fused, indicating physical maturity, although tooth-eruption is incomplete. Tapirs with M1 erupted are already sexually mature and the skull and size subsequently changes little or not at all. The only significant change after M1 eruption is closure of the sutures (M. A. Cozzuol, pers. obs.).

Referred specimens

UFMG 3176, a skull, ribs, and a vertebra of a young adult male (2nd molars unerupted), hunted a few months before collection at 8°11′13.2″S, 65°41′41.2″W. Universidade Federal de Rondonia (UNIR-M21); a skull of a mature adult, sex unknown (3rd molars in use) with indication of being shot, from the right margin of the Madeira River, a few kilometers north of Porto Velho, Rondônia, Brazil (8°38′08.3″S, 63°53′00″W). Six partial skulls (UFMG 3178–3183), hunted by Karitiana Indians in their territory, Rondônia, Brazil, and donated by them to one of us (SN). American Museum of Natural History (AMNH) 36661, partial skull and skin of an adult young male, collected by Theodore Roosevelt in January 1912, in Porto Campo at Sepotuba River, Mato Grosso, Brazil.

Type locality

Southern Amazonas, Brazil, near BR 319 Highway, about 90 km north from Porto Velho, Rondônia, Brazil (8°07′45.73″S, 63°42′09.64″W; Fig. 2).

Distribution

The new species is present in Amazonas, Rondônia, and Mato Grosso states in Brazil and in Amazonas Department in Colombia. The habitats in the localities where the species was recorded so far are mosaics of forest and open savanna. Local people's knowledge and photographic documents also suggest that it may be present in the eastern Amazon along the Guiana Shield (Amapá in Brazil and southern French Guiana; Fig. 2).

Diagnosis

Tapirus kabomani is the smallest living tapir, with total length 130 cm, height at shoulder 90 cm, and body mass estimated at about 110 kg. Externally, it differs from the sympatric T. terrestris by darker hair, lower mane, broader forehead, and smaller size; cranially, it differs from T. indicus because the maxillary–premaxillary suture ends anterior to the canine; from T. bairdii by the absence of an ossified nasal septum and dorsal maxillary flanges; from T. terrestris by a lower sagittal crest, frontals broad and inflated behind the nasals, extending up to the frontal–parietal suture; from T. pinchaque by the gently inclined sagittal crest rising posteriorly instead of parallel to the toothrow; from T. bairdii and T. indicus by having a single narrow sagittal crest; from T. pinchaque, T. terrestris, T. bairdii, and T. indicus by a shallower and less dorsally extended meatal diverticulum fossa, and generally smaller size. It differs from all living species plus the extinct T. webbi, T. veroensis, T. johnsoni, and T. cristatellus by its relatively short limbs (femur length shorter than dentary length). Molecularly, the 4 samples of T. kabomani analyzed shared 7 unambiguous apomorphic characters in mtDNA Cytb (960 base pairs [bp]), COI (650 bp), and COII (642 bp) genes, compared with T. indicus, T. bairdii, T. pinchaque, and T. terrestris. The following autapomorphic characters define T. kabomani: position 620 in the Cytb gene (homoplasically present in an individual of T. terrestris as shown by the parsimony reconstruction), position 401 and 413 of the COI gene, and positions 254, 329, 389, and 395 of the COII gene. T. kabomani also is defined by absence of 3 synapomorphic positions (302, 401, and 497 of the Cytb gene) that unite the clade T. pinchaque–T. terrestris (see Supporting Information S5 for GenBank accession numbers).

Potential synonyms

Many names have been proposed for putative taxa of the genus Tapirus from South America (Hershkovitz 1954), and possibly 1 or more of those names refer to the species we are describing here.

Several early names were proposed as substitutes for the original Linnean name, Hippopotamus terrestris, or were independent descriptions of the same species, and thus are objective synonyms of T. terrestris (Linnaeus, 1758), type locality Pernambuco, Brazil. In this category are included Hydrochaerus tapir Erxleben, 1777, Tapir suillus Blumenbach, 1779, and Tapir americanus Gmelin, 1788, from Suriname; Tapir anta Zimmermann, 1780, from Pernambuco, Brazil; and T. tapirus Merriam, 1895. No types are available for these names.

Tapirus terrestris has no originally designated holotype and its type locality is Pernambuco, Brazil (Linnaeus, 1758).

Other names proposed for putative new species, including T. maypuri Roulin, 1829, from the Guianas, T. sabatyra Liais, 1872, from Pernambuco, Brazil, T. brasiliensis Liais, 1872, from Minas Gerais, Brazil, and T. anulipes Hermann, 1924, from Mato Grosso, Brazil, also have no designated holotypes and cannot be assigned. All of these should be considered as nomina dubia.

Tapirus aenigmaticus Gray, 1872, from Macas, eastern Ecuador, was considered by Hershkovitz (1954) to be a young T. terrestris, based on its immature skull, despite that its associated skin supposedly showed characteristics that appear closer to some juveniles of T. pinchaque. As Hershkovitz (1954:476) noted, Gray's (1872) revision of tapirs is a “… confusing and misleading source of information. The work is characterized by numerous typographical errors, misquotations of authors, contradictions, and assumptions derived from specimens mislabeled as to sex and locality and mismatched as regards skins and corresponding osteological material.” The skull illustrated by Gray (1872:491) is from a very young animal and the globular shape of the braincase is most likely due to this condition. Because no sign of a sagittal crest is visible in this specimen, and because in T. terrestris the crest is formed at the fetal stage, it seems most likely that the skull and skin may represent a single individual of T. pinchaque. In fact, the collector of this specimen noted that it was captured along with an adult female that did not separate from it. For obscure reasons, Gray (1872) doubted this assertion, and placed the young animal in a different species than the adult, which he called T. leucogenys. The latter specimen has some characteristic features of T. pinchaque, such as white upper and lower lips, but does not possess all characteristics. In any case, neither the skull nor pelt matches the observed specimens of T. kabomani. However, in our analysis some DNA sequences from the lower cordillera in Ecuador were closer to T. pinchaque than to other T. terrestris (see Genetic Evidence section). The possibility should be pursued that those samples may represent animals like the one described as T. leucogenys.

Tapirus ecuadorensis Gray, 1872, from Macas, eastern Ecuador, was based on a juvenile skin. Hershkovitz (1954:485) states that “Descriptions of species based on skins of striped juveniles (T. aenigmaticus, T. ecuadorensis, T. peruvianus) and young adults with persistent juvenile striping (T. anulipes) are trivial.” Our search of photographic material, zoo specimens, and consultation with colleagues uncovered no work available on intraspecific and interspecific variation of young pelts in tapirs. Gray used few specimens to make his observations, giving it an unjustified taxonomic value. The skins are now more than 150 years old, and curatorial procedures at that time included formalin and arsenic as preservatives, which preclude DNA extraction and analyses by current methods.

Tapirus rufus Fisher, 1814, probably from French Guiana, was based on a skin and a skull, but efforts to locate the specimen in the institution where it was deposited have been unsuccessful, and it is assumed lost.

The type of T. laurillardi Gray, 1867, without precise type locality, based on an adult skull, we ascribe to T. terrestris (Gray, 1867:882, figure A) because it has a high and very curved sagittal crest, with origin in the frontals.

Tapirus peruvianus Gray, 1872, from Peruvian Amazon, was based on a juvenile skin and skull. Despite the inutility of the skin as a character, the skull lacks any indication of a sagittal crest, which precludes assignment to T. terrestris. Juveniles of comparable age are still unknown for T. kabomani. The morphology of the upper deciduous premolar lacks the cinguloid shelf of the protocone, like T. pinchaque, but this feature is present in T. terrestris and in all the known specimens of T. kabomani (Hershkovitz 1954:487). We therefore identify this specimen as a T. pinchaque.

We examined the holotype skull of T. spegazzinii, Ameghino, 1909 (Museo Argentino de Ciencias Naturales, Bernardino Rivadavia, Buenos Aires, Argentina [MACN] 5.41), from Rio Pescado, Departamento de Orán, Salta, Argentina. It has the skull characters of T. terrestris, with a long and high sagittal crest starting in the frontals.

Tapirus terrestris colombianus Hershkovitz, 1954, from El Salado, eastern slope of Sierra Nevada, department of Magdalena, Colombia, is based on a young adult male skull that has all the characteristics of T. terrestris.

Consequently, we cannot link any previous names to our specimens, and thus we propose a new taxon name.

Description

External appearance. For external characters we use information from local hunters, who identified the animals in the camera-trap photos (Fig. 3) from the type locality as belonging to the new species. The hair is dark, from dark gray to dark brownish. The mane, as an external expression of the sagittal crest, is lower and starts posteriorly. The forehead behind the nasals is broader than in T. terrestris. From the photos, specimens of known sex show that females have a gray–white area that extends from the lower jaw to the cheeks and the base of the ears, and extends ventrally to the neck, similar to that of T. bairdii. Males do not show this patch and seem to be smaller (Figs. 3a and 3b; Supporting Information S6 and S7, DOI:  10.1644/12-MAMM-A-169.S6 and DOI:  10.1644/12-MAMM-A-169.S7). The ear tips have a white line as in all Tapirus species.

Skull

The sagittal crest is single and narrow, as in T. terrestris and adults of T. pinchaque, but lower than in the 1st and higher than in the 2nd. Because only adult specimens are currently known, we do not know if the sagittal crest occurs in newborns or if it develops later, as in T. pinchaque (Holbrook 2002). Unlike the latter, the sagittal crest is not horizontal, almost parallel to the toothrow, but it gradually rises posteriorly, as in some young specimens of T. terrestris. However, the sagittal crest is lower, shorter, and wider than in the latter species. The frontal bones are inflated to form a large triangular convex exposure, with a longitudinal medial depression, ending posteriorly at the frontal–parietal suture, where the sagittal crest begins. In T. terrestris the sagittal crest starts in the frontals, much anteriorly than in T. kabomani. These features are illustrated in Fig. 4, in comparison with a skull of T. terrestris. The meatal diverticulum fossae are shallower than in the other species, with reduced dorsal extension. The nasals are similar to those of T. terrestris in shape and do not project upward, as in T. bairdii and T. pinchaque. As in all Neotropical species, the maxillary–premaxillary suture is anterior to the canine. The anteromedial process of the maxilla is dorsal to premaxilla. The rostrum is not upturned as in most adults of T. terrestris.

Some features show intraspecific variability. The spiral grooves in the nasals are shallow in UNIR-M21, AMNH 36661, and UFMG 3177, but a little deeper in the holotype. Diastema and rostrum are short in UNIR-M21, but not in UFMG 3177 and in the holotype.

Morphometrics

A MANOVA–CVA, including all living species, plus the holotype of T. rondoniensis (Fig. 5A) and then T. kabomani, T. pinchaque, T. terrestris, and T. rondoniensis (Fig. 5B), was performed.

In both analyses, T. terrestris discriminates clearly from all the others with significant discriminant function values, but the small sample sizes of the other species preclude the calculation of a discriminant function among them (see Supporting Information S8, S9, and S10; DOI:  10.1644/12-MAMM-A-169.S8, DOI:  10.1644/12-MAMM-A-169.S9, and DOI:  10.1644/12-MAMM-A-169.S10). In the CVA scatter-plots (Fig. 5) all species appear clearly separated. The single specimen of T. rondoniensis falls between T. terrestris and T. kabomani, but out of the range of both of them.

Morphological cladistics

The morphological cladistic analysis, which includes all living and selected fossil North and South American species, results in a single most-parsimonious tree. The genus Tapirus exhibits monophyly. The middle to late Miocene species T. johnsoni and the late Miocene T. webbi, both from North America, appear as successive clades, the latter is the sister group of the 2 major clades, one containing all of the South American species (T. kabomani, T. terrestris, T. pinchaque, T. cristatellus, T. mesopotamicus, and T. rondoniensis) and the other containing the Asian, North American, and Central American species (T. indicus, T. bairdii, T. polkesnsis, T. haysii, and T. veroensis) supported by 2 synapomorphies. The South American clade is supported by 5 synapomorphies. T. mesopotamicus is the sister group of the remaining South American species. T. kabomani appears as the sister taxon of T. rondoniensis, from which it differs by possessing 3 autapomorphies, whereas T. rondoniensis has 1 autapomorphy. The character distribution in these taxa precludes considering these 2 species as synonyms. They are, in turn, the sister group of a clade including T. terrestris, T. pinchaque, and T. cristatellus (Fig. 6).

The grouping of T. indicus with T. bairdii and several North American fossil species is the major difference between phylogenetic results of morphological and DNA characters (see below). The reason for this is not clear. Because the several fossil species included in the morphological matrix cannot be assessed genetically, a likely alternative would be a long-branch attraction causing T. bairdii to appear as sister taxon of T. indicus in the morphology analysis. Furthermore, many other genetic markers have shown T. bairdii as sister group of the South American tapirs (Steiner and Ryder 2011), the same topology we recover in our mtDNA analyses. Because no fossil tapirs from Asia are well known enough to be included in the morphological matrix, the position of T. indicus may change in future morphological analyses.

Genetic evidence

A total of 960 bp of the Cytb gene, 617 bp of the COI gene, and 690 bp of COII gene were sequenced for the holotype (UFMG 3177) and 1 paratype (UFMG 3176), and compared with sequences of 64 tapir individuals. Between-taxa distances for COI, COII, and Cytb data were calculated using Kimura 2 parameter (Supporting Information S11, S12, and S13; DOI:  10.1644/12-MAMM-A-169.S11, DOI:  10.1644/12-MAMM-A-169.S12, and DOI:  10.1644/12-MAMM-A-169.S13), and reveal a high similarity because of a close relationship among the 3 South American species.

Bayesian inference and maximum-likelihood phylogenetic trees for the Cytb data set were mostly congruent, with small discrepancies only in node support (Fig. 7), both having the same topology, which also was replicated using the Cytb + COI + COII data set (Supporting Information S14, DOI:  10.1644/12-MAMM-A-169.S14). The last one used a smaller sample set. The proposed T. kabomani formed a clade with moderate (Cytb, aLRT = 0.55, posterior probability [PP] = 0.73) to strong (all genes combined, aLRT = 0.88, PP = 0.93) support, sister to T. terrestris and T. pinchaque in all phylogenetic reconstructions, involving combined data set (Cytb + COI + COII) or only Cytb with a larger sample size. Moreover, a similar topology is observed in the maximum-parsimony tree (figure not shown) and median-joining network analyses (Fig. 8). In all trees, the 2 DNA samples from southeastern Colombia and 1 on the Brazilian border also were shown to belong to the T. kabomani clade (Fig. 7).

Tapirus pinchaque presented a well-supported clade, but T. terrestris was shown to be paraphyletic with T. pinchaque nested inside it (Fig. 7). T. terrestris appears to be composed of at least 5 different clusters (or clades including T. pinchaque) in the Bayesian inference reconstructions, each supported by PP > 0.99 (aLRT > 0.91 for maximum-likelihood tree). The restricted and peripheral distribution range of T. pinchaque suggests it to be the result of a peripatric speciation process. Molecular dating methods (linearized tree and Bayesian estimation—see “Materials and Methods”) indicate a late Miocene to early Pliocene time of divergence of T. bairdii and the South American clade (3.16–7.55 mya), and a late Pleistocene divergence for the T. kabomani and T. terrestris–T. pinchaque complex (288–652 thousand years ago; Fig. 7).

The Cytb median-joining network analysis (Fig. 8) supported the basal position of T. kabomani, but also indicated that T. terrestris includes at least 2 main clusters, separated by 6 mutational steps. T. pinchaque is nested between these 2 clusters (Figs. 7 and 8), 5 steps distant from the larger T. terrestris cluster, and 9 from the smaller one.