A new african fossil caprin and a combined molecular and morphological bayesian phylogenetic analysis of caprini (Mammalia: Bovidae)

Authors


Correspondence: Faysal Bibi, Museum für Naturkunde, Invalidenstr. 43, Berlin 10115, Germany. Tel.:+49 30 2093 9113; fax: +49 30 2093 8565; e-mail: faysal.bibi@mfn-berlin.de

Abstract

Given that most species that have ever existed on Earth are extinct, no evolutionary history can ever be complete without the inclusion of fossil taxa. Bovids (antelopes and relatives) are one of the most diverse clades of large mammals alive today, with over a hundred living species and hundreds of documented fossil species. With the advent of molecular phylogenetics, major advances have been made in the phylogeny of this clade; however, there has been little attempt to integrate the fossil record into the developing phylogenetic picture. We here describe a new large fossil caprin species from ca. 1.9-Ma deposits from the Middle Awash, Ethiopia. To place the new species phylogenetically, we perform a Bayesian analysis of a combined molecular (cytochrome b) and morphological (osteological) character supermatrix. We include all living species of Caprini, the new fossil species, a fossil takin from the Pliocene of Ethiopia (Budorcas churcheri), and the insular subfossil Myotragus balearicus. The combined analysis demonstrates successful incorporation of both living and fossil species within a single phylogeny based on both molecular and morphological evidence. Analysis of the combined supermatrix produces superior resolution than with either the molecular or morphological data sets considered alone. Parsimony and Bayesian analyses of the data set are also compared and shown to produce similar results. The combined phylogenetic analysis indicates that the new fossil species is nested within Capra, making it one of the earliest representatives of this clade, with implications for molecular clock calibration. Geographical optimization indicates no less than four independent dispersals into Africa by caprins since the Pliocene.

Introduction

Caprini – goats and relatives – comprises a taxonomically diverse and geographically widespread clade of ruminant herbivores. Ranging from the large high-arctic musk ox (Ovibos moschatus), to the small capricorn of Indonesian and Malaysian forests (Capricornis sumatraensis), to the emblematic Alpine ibex (Capra ibex), living caprins are tropical to boreal in distribution and are successfully established across most of Eurasia. They are also present in North America where they represent four of the five bovid species with natural ranges outside the Old World. Caprins are conspicuously rare in Africa, with only three species naturally occurring in restricted ranges on the continent today. They are also very rare in the African fossil record, but the material available still indicates a greater range in that continent throughout the late Neogene than the meagre distribution today. We here describe a new fossil caprin species from the early Pleistocene of the Middle Awash, Ethiopia. We perform a phylogenetic analysis of living and fossil caprin species using a combined molecular and morphological character data set. The analysis allows us to place the new species phylogenetically and within a biogeographical context, to investigate other parts of the caprin tree and to compare the results of molecular versus morphological approaches to phylogeny.

Caprin species are well adapted to rocky and mountainous terrain, possessing characteristically shortened metapodials, an adaptation for sure-footedness over speed. Otherwise, however, caprins vary widely in form, which has led to a constantly changing suprageneric taxonomy (Pilgrim, 1939; Simpson, 1945; McKenna & Bell, 2000). More recent phylogenetic work, spurred particularly by molecular analyses, has proposed new classificatory schemes, with traditionally recognized groups such as Saigini, Ovibovini and Rupicaprini now considered polyphyletic, and the position of the chiru (Pantholops hodgsonii) as the sister taxon to Caprini supported (Gentry, 1992; Groves & Shields, 1996; Gatesy et al., 1997; Hassanin et al., 1998, 2012; Vrba & Schaller, 2000; Ropiquet & Hassanin, 2005b). Ultimately, however, the evolutionary history of Caprini cannot be told without consideration of the fossil record. Although molecular phylogenetic analysis has significantly advanced systematics of living species, analyses hoping to discuss any aspect of the evolutionary history of this clade cannot afford to exclude fossil taxa. For example, it is clear from the fossil record that the modern diversity and geographical distribution of Caprini can be considered restricted and relict by comparison to times past. Caprines, viewed in light of their fossil record, therefore make a particularly good group for the study of Quaternary biodiversity loss, as their decline since Plio-Pleistocene times is no doubt at least partly a result of changing palaeoenvironmental factors. Numerous clades have benefitted from phylogenetic analyses that combine living and fossil taxa using molecular and morphological character ‘supermatrices’ (e.g. placentals, Asher, 2007; cetaceans, Geisler et al., 2011; squamates, Müller et al., 2011). Gatesy et al. (1997) and Gatesy & Arctander (2000) combined the morphological character matrix of Gentry (1992) with both mitochondrial and nuclear genes to investigate the relationships among living bovids, but included no fossil species. The analysis we present here for the first time combines molecular and morphological characters along with all living and three fossil caprine taxa in a single ‘supermatrix’. Analyses of the combined and separate data sets allow for an examination of agreement and conflict between the morphological and molecular evidence to the phylogeny of Caprini. Our phylogenetic framework also represents a first step towards combining neontological and palaeontological studies with the aim of producing a single phylogeny that includes all species of Bovidae.

The terms ‘Caprinae’ and ‘Caprini’ have been used interchangeably in the literature of the last decade. For the present paper, we use the term ‘Caprini’ (adj. caprin) for the clade including all goats and sheep and close relatives (musk oxen, takin, serows, etc.) but not Pantholops hodgsonii. Caprini can be diagnosed by shortened metacarpals and is synonymous with the traditional Caprinae (e.g. McKenna & Bell, 2000). Pantholops differs in numerous morphological and behavioural characteristics, and keeping it separate from Caprini would retain a much more ecologically useful taxonomy. Caprina is a large subclade of Caprini which includes goats (Capra spp.), sheep (Ovis spp.) and close relatives (cf. Caprini of Simpson, 1945; Caprina of Hassanin et al., 2009, 2012).

Systematic palaeontology

Bovidae

 Antilopinae Gray 1821

  Caprini Gray 1821

   Caprina Gray 1821

    Capra Linnaeus, 1758

Genus diagnosis

The crown clade, comprising all the descendants of the most recent common ancestor of all living species of Capra, is well supported by both morphological and molecular phylogenetic analyses. In terms of cranial morphology, species of the clade may be recognized by horns that are upright, not strongly divergent, long, curving backwards scimitar-like, bearing large nodosities on the sheaths (except in Capra falconeri in which horns are spiralled and smooth), deeply permeated by frontal sinus which is strutted into large chambers, and with a relatively long braincase bearing enlarged mastoids. Additional synapomorphies according to our analysis below (Fig. S2) are as follows: posteriormost nasal boundary located far posteriorly, reaching level of fronto-lachrymal suture; frontals in intercornual area expanded and clearly raised; parietal dorsal cranial surface often marked by pronounced swelling or boss; supraoccipital often marked by relief and set higher than parietal.

Capra wodaramoya sp. nov. Fig. 1

Holotype

GUN-VP-8/1 partial cranium

Type locality

Guneta Vertebrate Paleontology Locality 8 (GUN-VP-8), in the Middle Awash, Afar Region, Ethiopia.

Age

Estimated at 1.9–2.0Ma based on independent biochronological, stratigraphic and geochronological (Ar/Ar) grounds (T. White pers. comm.).

Etymology

From the Afar language words for goat (wodara) and head (moya)

Other referred specimens

None.

Diagnosis

A Capra of very large size differentiated from all other caprins by horn cores that are very large, antero-posteriorly elongated, very strongly mediolaterally compressed, lacking sharp keels and torsion, curving backwards continuously, and presumably very long when complete. In cross-section, horn cores are slightly wider centrally than anteriorly or posteriorly, with a long axis that is significantly rotated relative to the sagittal plane (c.25°). The braincase is very long and narrow and is very highly flexed on the face as indicated by a very obtuse angle between the dorsal cranium and the nuchal plane (145°).

Comparisons

Capra wodaramoya shares with other Capra species horn cores that are relatively large, strongly compressed, upright and deeply permeated by frontal sinuses that are divided into large chambers, and a braincase that is long and low, flexed on the face, smoothly curved between its lateral and dorso-lateral aspects, and with an enlarged mastoid region and a robust nuchal area. It differs from all species of Capra in the degree to which it has developed certain of these features, particularly the very large but very compressed and rotated horn cores, and the very long and highly flexed braincase.

The massive, strongly compressed horn cores of Capra wodaramoya are reminiscent of those of Tossunnoria pseudibex from the late Miocene of China (Bohlin, 1937). Tossunnoria differs in having a braincase that is shorter and much less flexed on face, frontal sinuses that do not penetrate deep into horn core, horn cores that are more divergent, curving laterally, probably shorter with rapid distal tapering and an associated rugose surface, and a frontal that is less expanded and less raised between the horncores.

Description

GUN-VP-8/1 is a partial cranium lacking the face, with only the proximal portions of both horn cores preserved. Measurements are given in Table 1. The fronto-parietal suture is very strongly anteriorly indented and the interfrontal and fronto-parietal sutural ridges are quite raised, more so than in any living Capra species, although these may be allometric effects of the sheer size of the horn core bases. The mastoids are large and protruding. The temporal squamosal suture curves upwards in its posterior half. The occipital nuchal plane is quite flat and not flexed about the median nuchal ridge, about as high as it is wide, with clear vertically oriented median and lateral nuchal ridges and pronounced rugosities (including two deep fovea at the supraoccipital) just to either side of the median plane. The nuchal surface is very robust and tuberous, deeply textured by large and deep foveae, and speaks to a skull supported by heavy nuchal musculature. The basioccipital is short with wide anterior tuberosities, and only a small isthmus of connection between the occipital condyles and the posterior tuberosities. A relatively deep fovea is located centrally between the post-tuberosities, occipital condyles and foramen magnum. A peculiar unique feature that might be subject to individual variation is that the parieto-temporal (squamosal) suture is curved upwards in its posterior part, resembling the condition in Budorcas taxicolor, and not straight like in most Caprini. Taken together, these features indicate a caprin species that, although bearing similarities to living Capra species, is in fact very derived in morphology and without any obvious relationship to any one of the living species.

Table 1. Measurements of the GUN-VP-8/1 partial cranium, holotype of Capra wodaramoya
Maximal antero-posterior length of the basal horn core (DAP)R: 105.1 mm L: 106.1 mm
Maximal transverse with of the basal horn core (DT)R: 49.9 mm L: 46.3 mm
Horn core length preservedR: 145 mm L: 200 mm
Transverse distance between the outer edges of the basal horn cores c.185 mm
Transverse distance between the inner edges of the basal horn cores c.16 mm
Rotation of the horn core DAP axis to the sagittal plane25°
Horn core divergence angle40°
Horn core inclination angle (measured against the dorsal braincase)85°
Minimum width of the braincase93.5 mm
Braincase width at the mastoids126.7 mm
Distance between bregma and the nuchal plane70.4 mm
Occipital height from the superior foramen magnum to the superior edge of the supraoccipital72.5 mm
Occipital height from basion to the superior edge of the supraoccipital93.4 mm
Outer distance between the posterior tuberosities of the basioccipital process41.7 mm
Outer distance between the anterior tuberosities of the basioccipital process38 mm (estimated)
Outer distance between the occipital condyles75.6 mm

Materials and methods

The cytochrome b genome (1140 base pairs, except Capra walie for which only 600 bp were available) was downloaded from GenBank for 38 species (Table 2). 52 osteological characters (51 from the skull and 1 postcranial, Supporting Information) were formulated and coded for 28 species using specimens at the American Museum of Natural History, the Muséum National d'Histoire Naturelle and the National Museum of Ethiopia (Table 2). The total sample comprises 40 bovid species, representing Caprini, Pantholops hodgsonii, species of Alcelaphini and Antilopini, Oreotragus oreotragus, and the boselaphin Tetracerus quadricornis.

Table 2. GenBank accession numbers and their references for sources of cytochrome b data, and museum skeletal specimens studied for morphological character scoring. Museum prefixes: AMNH, American Museum of Natural History (New York); MNHN, Muséum National d'Histoire Naturelle (Paris); GUN and AL, National Museum of Ethiopia (Addis Ababa)
TaxonGenBank Accession no.ReferenceSkeletal Specimen no.
Tetracerus quadricornis AF036274 Hassanin & Douzery (1999)MNHN 1975-838, 1988-223
Oreotragus oreotragus AF036288 Hassanin & Douzery (1999)MNHN 1967-913, 1932-33
Saiga tatarica AF064487 Hassanin et al. (1998)
Nanger granti AF034723 Hassanin et al. (1998)
Damaliscus pygargus AF036287 Hassanin & Douzery (1999)MNHN 1970-384, 1944-279
Addax nasomaculatus AF034722 Hassanin et al. (1998)
Pantholops hodgsoni AF034724 Hassanin et al. (1998)MNHN 1993-4237, 1877-31
Ammotragus lervia AF034731 Hassanin et al. (1998)AMNH 35056, 14518
Budorcas churcheri AL 136-5, AL 427-3A, AL 1666-1A, AL 1666-1B
Budorcas taxicolor AY669320 Ropiquet & Hassanin (2005a)AMNH 110475, 57016
Capra aegagrus FJ936175 Rezaei et al. (2010)AMNH 54612, 88697
Capra caucasica AF034738 Hassanin et al. (1998)
Capra cylindricornis AF034737 Hassanin et al. (1998)
Capra falconeri AF034736 Hassanin et al. (1998)AMNH 55821, 54610
Capra ibex AF034735 Hassanin et al. (1998)MNHN 2000-416, 1997-1303
Capra nubiana AF034740 Hassanin et al. (1998)AMNH 82264
Capra pyrenaica FJ207528 Hassanin et al. (2009)MNHN 1864-78, A12050
Capra sibirica AF034734 Hassanin et al. (1998)AMNH 60364, MNHN 1964-404
Capra walie EU368863 Gebremedhin et al. (2009)AMNH 188342, 54374
Capra wodaramoya GUN-VP-8/1
Capricornis crispus FJ207533 Hassanin et al. (2009)
Capricornis sumatraensis AY669321 Ropiquet & Hassanin (2005a)AMNH 110480, 45348
Hemitragus hylocrius AY846792 Ropiquet & Hassanin (2005b)AMNH 54755, 54757
Hemitragus jayakari AY846791 Ropiquet & Hassanin (2005b)
Hemitragus jemlahicus AF034733 Hassanin et al. (1998)AMNH 90152, 90085
Myotragus balearicus AY380560 Lalueza-Fox et al. (2005)AMNH 92321, 105252
Nemorhaedus caudatus U17861Groves & Shields (1996)
Naemorhedus griseus FJ207532 Hassanin et al. (2009)AMNH 28237, 110482
Oreamnos americanus AF190632 Hassanin & Douzery (2000)AMNH 35334, 70573
Ovibos moschatus AY669322 Ropiquet & Hassanin (2005a)AMNH 28072
Ovis ammon AF034727 Hassanin et al. (1998)AMNH 54884, 54931
Ovis aries AF034730 Hassanin et al. (1998)MNHN A12190, A12168
Ovis canadensis EU366066 Rezaei et al. (2010)MNHN 1929-417, 1926-171
Ovis dalli AF034728 Hassanin et al. (1998)
Ovis orientalis EU365973 Rezaei et al. (2010)
Ovis vignei AF034729 Hassanin et al. (1998)AMNH 88722
Pseudois nayaur AF034732 Hassanin et al. (1998)AMNH 117566, 117565
Pseudois schaeferi AF398355 Zhou et al. (2003)
Rupicapra pyrenaica AF034726 Hassanin et al. (1998)
Rupicapra rupicapra AF034725 Hassanin et al. (1998)AMNH 24193, 90296

The combined supermatrix was assembled in Mesquite 2.74 (Maddison & Maddison, 2010) and is archived in TreeBase (http://purl.org/phylo/treebase/phylows/study/TB2:S12447). Cytochrome b sequences were aligned using the CLUSTALW method in Jalview 2.7 (Waterhouse et al., 2009) and checked by eye. Bayesian analysis was performed in MrBayes 3.2.0 (Ronquist & Huelsenbeck, 2003) using two characters partitions (DNA and Standard). The DNA character partition (characters 1–1140) was run on the GTR+G+I model as determined by testing among 24 substitution models using the AIC criterion in jModeltest (Posada, 2008). The morphological character partition was run using the default settings for the Mk model, which assumes no constant characters are present (Mkv model, Lewis, 2001), and for which we allowed character rates to vary along a gamma distribution. Each analysis comprised 4 runs, each of which had 4 chains. All analyses were run for a total of 2.5 million generations (with the standard deviation of split frequencies reaching <0.01). Trees were sampled every 500 generations with the first 25% discarded as burn-in, and summarized using a 50% majority rule tree showing all compatible partitions.

The combined data set and the morphology-only data set were also analysed using parsimony in PAUP* 4.0b10 (Swofford, 2002). Heuristic searches were run for 1000 repetitions using random addition sequence and tree bisection reconnection, with all characters unordered and of equal weight (=1). Bootstrap was run for 1000 repetitions holding a maximum of 1000 trees each. Optimization of the morphological characters on the tree produced by the combined supermatrix run was performed in MacClade 4.08 (Maddison & Maddison, 2000) using the Trace All Changes option, with polytomies fixed as hard speciations, as well as in Mesquite 2.74 using the Trace Character History with the Mk likelihood option. In all analyses, Tetracerus quadricornis was set as the outgroup taxon.

Results

In the course of assembling the final matrix and running different iterations of the analysis, we observed that outgroup selection and the number of taxa (including taxon duplications) often had strong effects on certain parts of the tree topology. Without an in-depth investigation, it seemed that certain taxa were much more susceptible to topological shifts than others (among these Capra falconeri and the Budorcas clade).

Figures 2-4 present the trees resulting from the Bayesian phylogenetic analyses using both cytochrome b and morphology (Fig. 2), just cytochrome b (Fig. 3) and just morphology (Fig. 4). Figure 4 additionally compares the tree resulting from parsimony analysis of the morphological data. The combined Bayesian analysis of molecular plus morphological characters (Fig. 2) provides a phylogeny largely controlled by the molecular data but affected enough by the morphological data to reflect shared form and to allow for the integration of fossil taxa into the analysis. The combined analysis indicates that the monophyly of genus Capra is well supported, as is the placement of Capra wodaramoya within it. The cytochrome b analysis differs from the combined analysis mainly in failing to recover monophyletic Caprina and Capra clades. The morphological analyses differ from the other two mainly in failing to recover Caprini, and in the presence of monophyletic Budorcas+Ovibos and ibex clades. Differences between the trees produced by the parsimony and Bayesian analyses of the morphological data (Fig. 4) are reassuringly few (e.g. Müller & Reisz, 2006), limited mainly to resolution among the basal caprins and the chiru. The strict consensus resulting from parsimony analysis (Fig. S1) of the combined molecular plus morphological data set is effectively a less resolved version of the Bayesian tree (Fig. 2). The parsimony majority rule consensus reveals some relationships that are different from those in the Bayesian analysis, but these are aberrant with respect to all other studies (Fig. S1).

Figure 1.

GUN-VP-8/1 partial cranium, holotype of Capra wodaramoya. Views from left to right, top to bottom: right lateral, anterior, left lateral, ventral basioccipital, dorsal, infero-posterior, supero-anterior. Cross-section of left horn core base shown to scale oriented so anterior is to the top of the page. Scale bar = 5 cm total for the basioccipital view, 10 cm total for all others.

Figure 2.

Tree resulting from the Bayesian analysis of the supermatrix of cytochrome b plus 52 morphological characters. Numbers at nodes are posterior probabilities. Taxa with African distributions are coloured green. Caprina, the largest subclade of Caprini, is boxed in yellow.

Figure 3.

Tree resulting from the Bayesian analysis of just the cytochrome b data set. Numbers at the nodes represent posterior probabilities. Clades Caprina and Capra are not recovered here.

Figure 4.

Trees resulting from analyses of morphological data alone. (a) Strict consensus of 58 most parsimonious trees of 159 steps (consistency index excluding uninformative characters = 0.38). Bootstrap values are provided for nodes with ≥50% support. Branch lengths are proportional to the number of state changes (both unambiguous and ambiguous) along a branch. (b) The Bayesian tree with node posterior probabilities and relative branch lengths. There is high congruence between both trees, indicating that the choice of analytical method does not significantly influence topology.

Discussion

The new bovid species from the Middle Awash, at just older than 1.9 Ma, represents another exciting but poorly represented find of an African Plio-Pleistocene fossil caprin. Fortunately, the only known specimen is a partial cranium well-enough preserved to allow for some level of taxonomic diagnosis and phylogenetic analysis. The posterior cranium of Capra wodaramoya is slightly larger than those of a male Ethiopian ibex (C. walie), but the horns of the fossil species are simply massive, approaching the basal proportions seen in a male argali (Ovis ammon). The horns would have coursed upwards and backwards like in the living ibex, and were probably very impressive when complete. This large fossil caprin joins a range of other ‘supersized’ antelopes recorded from the African early Pleistocene, including Megalotragus kattwinkeli Hippotragus gigas, Tragelaphus strepsiceros gigas and Pelorovis oldowayensis. Leakey (1965) noted a period of early Pleistocene gigantism in particular among the fossil taxa of Bed II at Olduvai Gorge. This is a subject area ripe for further research. The general rarity of fossil caprins in the deposits of the eastern African Rift Valley is probably related to their preference for mountainous terrains far from the depositional axis. Their occasional presence, however, probably indicates some tenuous link with nearby mountainous terrain and, in the case of Capra wodaramoya, with the abutting Ethiopian highlands.

Phylogenetic analysis

The Bayesian analysis of the combined molecular and morphological supermatrix produces a tree (Fig. 2) that is well resolved and reflects a phylogeny more in line with the total evidence than those derived from either data set alone. Separate analyses of the morphological and molecular characters reveal several areas of conflict between the two data sets (Figs. 3, 4), whereas optimization of the morphological characters on the tree from the combined analysis (Fig. S2) reveals areas of consensus. Analyses using parsimony produce trees that are similar to those resulting from Bayesian analysis but often with less resolution or with relationships that are atypical (Fig. 4, Fig. S1).

A common problematic area among all the analyses is that of basal relationships within Caprini. While these appear resolved in the morphological parsimony tree (Fig. 4a), none of the nodes have bootstrap support and the nested position of Pantholops violates caprin monophyly. A lack of resolution at the base of Caprini has plagued most phylogenetic studies to date, leading to suggestions of rapid adaptive radiation shortly after the tribe's origins in the Miocene (Hassanin et al., 1998; Ropiquet & Hassanin, 2005b). We propose that incorporation of more fossil taxa into the phylogenetic framework – especially Miocene caprins (Gentry, 2000) – might help resolve basal caprin relationships through improved morphological character optimization for this part of the tree.

Our Bayesian analysis of the combined molecular and morphological data (Fig. 2) revalidates the sister-group relationship of Pantholops hodgsonii to Caprini, united by possession of a small central incisor, among other characters (Fig S2). Pantholops lacks the key character of shortened metacarpals, which we here take to be diagnostic of Caprini. The chiru has been shown to further differ from caprins in numerous phenotypic, including behavioural, characters (Vrba & Schaller, 2000), and so, while recognizing their sister-clade relationship, we prefer to keep Pantholops and Caprini separate in what is surely a more ecologically meaningful taxonomy.

Lalueza-Fox et al. (2005) found the extinct subfossil Myotragus balearicus to be closely related to Ovis, although with low node support (mostly <50%), and a phylogeny requiring polyphyly of the Caprina. Our Bayesian analyses only recognized clades with ≥50% support, and so we did not recover a similar result. In our analyses, the position of Myotragus is relatively unresolved, forming part of the polytomy at the base of Caprini. Long molecular and morphological branch lengths (Fig. 2) reflect the fact that Myotragus is very highly derived, and its closest relatives among living species may not be easy to recognize.

It is satisfying to see the placement of the Pliocene fossil takin Budorcas churcheri from Hadar (Gentry, 1996) as the sister taxon to the living takin. The presence of a takin – an animal today associated with high altitudes of the Himalayan Plateau – in the late Pliocene of Ethiopia remains curious. Budorcas churcheri indicates that the living takin is a geographical relict, the last remaining of a once more widespread Budorcas clade. The Budorcas clade itself here takes part in the large polytomy at the base of Caprini. Other molecular analyses, however, have sometimes placed the takin closer to Capra (Hassanin et al., 2009), closer to Ovis (Groves & Shields, 1997; Hassanin et al., 1998) or more towards the base of the Caprini (Ropiquet & Hassanin, 2005b).

The morphology-only analyses also recovered a clade uniting Budorcas spp. and Ovibos moschatus. This clade is supported by 2 unambiguous character state changes (29:0-1, horn cores arising behind orbit, and 40:0–1, horn sheath forming a basal boss) in characters that are otherwise homoplastic, and has bootstrap and posterior probability support values that are relatively high (Fig. 4). The concept of a monophyletic Budorcas–Ovibos clade has been effectively buried (Gentry, 1992; Groves & Shields, 1997; Hassanin & Douzery, 1999), and it is interesting to see it resurface here on the basis of morphological apomorphies. Hassanin et al. (2009) used the name Ovibovina for a clade comprising Ovibos+Capricornis+Naemorhedus, which is also recovered here in both the molecular and combined analyses (Figs 2, 3). Morphologically, Ovibovina is supported by a single unambiguously optimized character (lack of horn core compression, which reverts in Ovibos, Fig. S2).

Although its composition has changed, Caprina (traditionally ‘Caprini’) is one of the few traditional subclades of Caprini (traditionally ‘Caprinae’) to have survived recent systematic shakeups. Caprina is here recovered in the morphological and combined analyses, but not in that of the cytochrome b data alone. Morphological support for Caprina is high, with no less than eight unambiguously optimized synapomorphies (Fig. S2). Caprina is the largest resolved subclade of Caprini, comprising the living goats (Capra spp.), sheep (Ovis spp.) and close relatives (Ammotragus, Pseudois, Hemitragus). Hassanin et al. (2009) analysis found support for inclusion of Oreamnos, Rupicapra and Budorcas within Caprina as well.

Hemitragus is by all accounts polyphyletic (Ropiquet & Hassanin, 2005a). Although the three included species are morphologically similar, even our morphology-only analysis did not succeed in creating a Hemitragus clade. Our morphological and combined analyses place Hemitragus jemlahicus as the sister taxon to a monophyletic Capra clade, and not sister to Capra sibirica as found by the cytochrome b analysis. This replicates the results of other studies where mitochondrial phylogenies returned a polyphyletic Capra (Ropiquet & Hassanin, 2005a; Pidancier et al., 2006). Interestingly, these studies also investigated nuclear genes and found that nuclear DNA phylogeny supports monophyly of Capra, consistent with morphological observations.

Similarly, analysis of mitochondrial DNA indicates that ibexes (C. ibex, C. nubiana, C. pyrenaica, C. walie) form a paraphyletic grouping. Nuclear DNA, in contrast, supports monophyly of the ibexes including C. caucasica and C. cylindricornis (Pidancier et al., 2006), again in accordance with our morphological observations (Fig. 4). Our supermatrix analysis (Fig. 2) reflects cytochrome b ibex paraphyly, but is swayed enough by the morphological characters to separate C. sibirica and Hemitragus jemlahicus and recover a monophyletic Capra. Close mtDNA relationships are found between the two African ibexes and also between the two European species.

Monophyly of the genus Capra is well supported by ten unambiguous cranial characters and high bootstrap and posterior probabilities in all analyses. The new Ethiopian fossil species is nested deep within the Capra clade. The affinities of Capra wodaramoya appear to lie specifically with the Eurasian ‘bezoar-type’ goats and not the African or Eurasian ibexes, with biogeographical implications discussed below.

Molecular clock calibration and diversification times

At 1.76 Ma, Capra dalli from the Pleistocene of Georgia (Bukhsianidze & Vekua, 2006) already provided indications of the minimum age of the Capra clade. Now, Capra wodaramoya marginally extends this age further to 1.9 Ma, providing an updated minimum-age datum point that may be used for molecular clock calibration and divergence estimation. Similarly, Budorcas churcheri provides a minimum age of around 3 Ma for the takin lineage.

Lalueza-Fox et al. (2005) produced molecular clock–derived clade divergence estimates for Caprini. Their analysis was based on the mitochondrial cytochrome b and 12S genes and the nuclear 28S rDNA gene and was calibrated using the age of isolation of the Myotragus lineage on the Balearic islands, presumed to be with the end of the Messinian Salinity Crisis at 5.35 Ma. Their estimates provide dates of around 1.5Ma for the origin of Capra and 6.2 Ma for the origin of Caprini+Pantholops, which are probably too young. Numerous caprins older than 6.2 Ma are recorded from late Miocene sites in China, Turkey and Europe (Gentry et al., 1999). Some of these taxa are very derived morphologically, certainly more so than Pantholops or basal caprins such as Capricornis, making it likely they belong inside the crown group. One reason for Lalueza-Fox et al.'s results might be the use of the Myotragus calibration point as a hard date rather than as a minimum date: the insular isolation of Myotragus may well have taken place at 5.3 Ma, but the most recent common ancestor of Myotragus and its living sister taxon may have lived millions of years prior to that. By comparison, Ropiquet & Hassanin's (2005a) divergence estimates of 10.4–7.5 Ma for Caprini and 11.9–8.7 Ma for Caprini+ Pantholops accord better with the fossil record.

African biogeography

Caprins are today represented in Africa by only three species: the aoudad or barbary sheep (Ammotragus lervia) with a North African and Saharan montane distribution, the Nubian ibex (Capra nubiana) in parts of north-eastern Africa and Arabia, and the Ethiopian ibex (Capra walie), found only in the Simien Mountains of Ethiopia.

This representation is meagre compared to the current diversity and range of caprins across Eurasia. The Neogene fossil record, however, documents numerous Plio-Pleistocene caprin species in sub-Saharan Africa, from Ethiopia to South Africa (refs. in Vrba, 1995), up into the Holocene (Brink, 1999). Our phylogenetic analysis suggests that the two fossil species we considered, Capra wodaramoya and Budorcas churcheri, were each the product of an independent dispersal event into Africa, separate from those of the Ethiopian ibex or aoudad lineages. Along with the living species, this indicates caprins dispersed from Eurasia into Africa on no less than four separate occasions since the Pliocene (Fig. 2). What is not immediately clear is whether African fossil caprins such as these were part of long-lived African lineages, or were short-lived immigrants from Eurasia. Vrba (1995) noted wide morphological disparity among Plio-Pleistocene African caprins, suggesting no close relationships among the different species. Accordingly, she proposed that the record of African caprins represents numerous separate immigration events into the continent from Eurasia, particularly timed at 2.7–2.5 Ma and 1.9–1.7 Ma. The age and phylogenetic position of Capra wodaramoya provides an additional data point in support of Vrba's observations. Because caprins tend to inhabit terrain not amenable to fossilization processes (high altitude, highly erosional terrains), the rare appearance of caprins in the African fossil record might equally be a function of biased taphonomic and fossilization processes. Outside of a phylogenetic approach, it might be difficult to favour dispersalist versus taphonomic explanations for the spotty African caprin record. Addition of the remainder of African fossil caprin taxa into the phylogenetic framework we have developed can further test to what degree the African fossil caprin record is (i) the product of a few poorly sampled but endemically successful and long-lived lineages (null hypothesis) or (ii) the product of numerous short-lived immigration events in from Eurasia (Vrba's hypothesis).

Conclusions

The vast majority of species on Earth are extinct, and no phylogenetic approach can hope to capture a synthetic evolutionary history without the inclusion of fossil taxa. Bovidae is a wide and diverse clade of mammals whose evolutionary history covers most of the Old World Neogene and whose phylogenetic history stands to explain much about terrestrial large-mammal evolutionary patterns and processes of the last 20 million years. The rise of molecular phylogenetics has brought major advances and resolution to the phylogeny and systematics of Bovidae. Molecular phylogenetic analyses, however, do not normally have the option to investigate fossil taxa, and as such the fossil record has been largely ignored by phylogenetic investigation. We have here presented a combined molecular and morphological analysis sampling a subset of the Bovidae, demonstrating that living and fossil taxa may be analysed together in a combined molecular plus morphological character supermatrix. Subjected to Bayesian analysis, the combined data set produces good results that are more informative than analyses of either the molecular or morphological characters alone. Our analysis has also highlighted the importance of comparing different data sets separately, particularly as nuclear gene phylogenies have oftentimes been shown to conflict with mitochondrial phylogenies in favour of morphological observations. The examples of Capra and the ibexes are discussed above, and a further example can be found with Bison (Verkaar et al., 2004; Nijman et al., 2008).

Although only known from a single specimen, the new fossil caprin from Ethiopia differs from all other known caprin species, particularly in the great cranial flexion and the massive and highly compressed horn cores. Our phylogenetic analysis indicates that Capra wodaramoya belongs to the Capra clade, which is supported by numerous morphological apomorphies. At around 1.9 Ma, Capra wodaramoya provides a new minimum age for the origins of crown clade Capra, and the fossil's nested phylogenetic position makes it possible that the clade is significantly older still. Capra wodaramoya does not appear to be closely related to the existing Nubian or Ethiopian ibexes, thereby providing evidence for an independent immigration event of Caprini into Africa from Eurasia. Vrba (1995) discussed the significance of the numerous but fragmentary occurrences of African fossil caprins as correlated with global cooling events and associated faunal dispersal from Eurasia into Africa. Our analysis here provides further data in support of Vrba's hypothesis, but the whole picture will only come into view with inclusion of the remainder of the African fossil caprin record into a phylogenetic framework.

There is no reason not to quickly expand the taxonomic range of this study to include the entire fossil caprin record. In particular, the Eurasian Miocene caprin fossil record presents perhaps the best chance to resolve the unstable phylogenetic relationships at the base of Caprini (Gentry, 2000). The character set too can easily be increased, with many more morphological (skeletal and soft tissues), behavioural (e.g. Vrba & Schaller, 2000) and genomic characters (especially nuclear genes) added to the supermatrix. The end result should be a phylogeny that reflects the consilience of biological evidence at all scales.

Acknowledgments

We would like to thank Tim White and the Middle Awash Project; curators and collections managers at the National Museum of Ethiopia, Department of Mammalogy at the American Museum of Natural History (E. Westwig), and the collections of comparative anatomy at the Muséum National d'Histoire Naturelle (J. Lesur); K. Reed of the Hadar Project; A. Pinton, G. Métais and J. Müller for discussions; J. Tseng and X. Wang for providing images and information on Tossunnoria, and the editor and two anonymous reviewers for their comments. The lead author was supported by a National Science Foundation International Research Fellowship (grant 0852975), the Agence Nationale de la Recherche (ANR-09-BLAN-0238) and a Leibniz-DAAD Research Fellowship.

Ancillary