HIPPOPOTAMIDAE AND

Boisserie, Jean-Renaud, Lihoreau, Fabrice, Orliac, Maeva, Fisher, Rebecca E., Weston, Eleanor M. & Ducrocq, Stéphane, 2010, Morphology and phylogenetic relationships of the earliest known hippopotamids (Cetartiodactyla, Hippopotamidae, Kenyapotaminae), Zoological Journal of the Linnean Society 158 (2), pp. 325-366 : 349-351

HIPPOPOTAMIDAE AND ANTHRACOTHERIIDAE

The selenodonty of the Anthracotheriidae , particularly advanced in Bothriodontinae , defines in a general dental morphology that is quite dissimilar from that observed in Hippopotamidae ( Fig. 1 and Supporting Information). This could be viewed as a major obstacle to accepting the nesting of the Hippopotamidae within the Anthracotheriidae . Two important contributions in favour of anthracotheriid– hippopotamid close relationships have dealt with this issue. First, Colbert (1935), basing his conclusions essentially on craniomandibular features, suggested two hypothetical scenarios leading to hippopotamid dentition. Evolution from a Merycopotamus -like anthracotheriid dentition was judged to be more straightforward than from a suid dentition. Secondly, the cladistic analysis of the whole artiodactyl phylogeny by Gentry & Hooker (1988) was largely based on cheek tooth characters. The preferred cladogram was not built following the parsimony rules of the initially used algorithm, but considered ‘realistic’ state reversals. A close relationship of hippopotamids with dentally archaic anthracotheriines was favoured.

The reassessment of kenyapotamine affinities was a particularly good opportunity to test hippopotamid relationships within artiodactyls on a wide array of cheek tooth characters and by the means of a formal cladistic analysis. A deep nesting of hippopotamids within bothriodontines was also obtained in previous works essentially based on craniomandibular features ( Boisserie et al., 2005a, b).

This and previous analyses ( Boisserie et al., 2005a, b) included Libycosaurus within the sister group of Hippopotamidae , alone or with other bothriodontines. A close phylogenetic relationship between Libycosaurus and Hippopotamidae has been recently criticized (Pickford, 2006: 270): ‘The anteriormost “premolar” (Pacc) in Libycosaurus represents a new acquisition, unique to Libycosaurus among artiodactyls, which primitively possess only four upper premolars. Incidentally, this strange dental acquisition provides strong evidence that Hippopotamidae have no close phylogenetic relationship to Libycosaurus , because in hippos there are only four (or three) premolariform teeth anterior to the M1/, and not five as in Libycosaurus .’ The author is correct in depicting this structure as quite unique amongst artiodactyls – it is indeed an autapomorphic trait of Libycosaurus ( Lihoreau, 2003; Lihoreau et al., 2006). However, the reasoning behind his phylogenetic argument is difficult to understand. The author would be correct if he had criticized a hypothesis presenting Libycosaurus as the forerunner of Hippopotamidae , but to our knowledge, no past or recent publications have suggested this. If his argument was put forward to refute the presence of Libycosaurus within the sister group of Hippopotamidae , this would similarly mean that Libycosaurus could not figure within the sister group of either anthracotheriids or artiodactyls, or of any known placental mammal younger than the late Cretaceous ( Novacek, 1986), a suggestion probably not intended by Pickford (2006).

More generally, the same author provided a list of differences and similarities amongst Hippopotamidae , palaeochoerids, and Anthracotheriidae ( Pickford, 2007b: table 8) in support of the following view: ‘the supposed synapomorphies between hippos and anthracotheres cited by Boisserie et al. (2005b) are either incorrectly reported or are more likely to be parallelisms’ ( Pickford, 2007b: 103). These features and their derived or primitive conditions were unfortunately not discussed, nor illustrated. This is highly regrettable, as we did discuss and illustrate many of them in a contribution ( Boisserie et al., 2005a) apparently ignored by Pickford (2007b), despite the fact that the manuscript was cited in the publication he criticized ( Boisserie et al., 2005b).

Another paper (Pickford, 2006), providing a detailed review of the anatomy of Libycosaurus anisae , and summarized by Pickford (2007b: table 7), was also cited against the phylogenetic hypothesis presented by Boisserie et al. (2005b). We would simply note that: (1) previous works by Boisserie et al. (2005a, b) did not discuss the affinities of Libycosaurus anisae , but those of Libycosaurus petrocchii ; (2) unpolarized character differences observed between a single species of anthracotheriid and Hippopotamidae (for which the comparative sample was not mentioned by Pickford, 2006) are not sufficient to invalidate an emergence of Hippopotamidae within Anthracotheriidae – a result obtained even when Libycosaurus was removed from the present data matrix and from those of Boisserie et al. (2005a, b). Similarly, we note that the recently suggested systematic changes within Libycosaurus ( Pickford, 2008) , not followed here, would not have any impact on this phylogenetic hypothesis.

Other studies, addressing the more general relationships between cetaceans and artiodactyls, did not recognize exclusive relationships between Anthracotheriidae and Hippopotamidae (O’Leary, 1999, 2001; O’Leary & Geisler, 1999; Geisler, 2001; Geisler & Uhen, 2003, 2005; Theodor & Foss, 2005; Geisler et al., 2007). However, these studies did not consider the full diversity of taxa characterizing Anthracotheriidae , and represented them mostly by northern American taxa excluding Eurasian anthracotheriines and bothriodontines – an issue acknowledged by Geisler et al. (2007).

Three Eurasian anthracotheriids were incorporated by Thewissen et al. (2007). None of them belong to Eurasian bothriodontines, and two of them, Anthracokeryx and Microbunodon , belong to the Microbunodontinae , a subfamily distinct from other anthracotheriids in displaying evolutionary trends analogous to those of moschids and tragulids ( Lihoreau & Ducrocq, 2007). The resulting phylogeny did not support a particularly close affinity between anthracotheriids and hippopotamids.

In contrast, O’Leary & Gatesy (2007) followed Boisserie et al. (2005a, b) in incorporating a number of Eurasian anthracotheriids, including Eurasian bothriodontines ( Merycopotamus , Libycosaurus , Bothriogenys ). Their results were broadly similar to ours, supporting close affinities between hippopotamids and most anthracotheriids, and excluding suoids from this relationship. However, some of their results appear inconsistent with ours. Notably, the position of Libycosaurus was most often not found in a clade with other anthracotheriids and Hippopotamidae , suggesting a polyphyletic Anthracotheriidae . This situation constitutes a major discrepancy with our results and other works nesting Libycosaurus within Bothriodontinae ( Pickford, 1991; Lihoreau & Ducrocq, 2007). This issue illustrates more generally the incongruence between the relationships within Anthracotheriidae obtained by O’Leary & Gatesy (2007) and ‘classical’ views on the phylogeny of the family ( Black, 1978; Pickford, 1991; Lihoreau & Ducrocq, 2007). In our opinion, the former should not be validated until further substantiation based on exhaustive reassessment focusing on material relevant to the family Anthracotheriidae .

Furthermore, we are convinced that the proper resolution of hippopotamid relationships requires the consideration of hippopotamid evidence beyond that included in the analyses cited above, i.e. solely the two extant species. Hippopotamus and Choeropsis may appear different enough to broadly represent the family Hippopotamidae but they are, nevertheless, representatives of lineages with peculiar adaptations and evolutionary histories ( Boisserie, 2007; Weston & Boisserie, in press). They should not be viewed as approximations of earlier species from the family, which in our opinion, provide more reliable data for resolving hippopotamid origins.

The results reported here and those of Boisserie et al. (2005b) independently offer support to a close relationship between crown bothriodontines (including Merycopotamus and Libycosaurus ) and Hippopotamidae , as one analysis of the present data matrix, performed after exclusion of characters shared with the data matrix of Boisserie et al. (2005b) yet still led to a similar result (66 characters remaining, see Supporting Information). Furthermore, cheek tooth characters alone support an emergence of hippopotamids within bothriodontines, a situation which may have been regarded as counterintuitive ( Boisserie & Lihoreau, 2006), but was foreseen by Colbert (1935). Of course, a scenario explaining the evolution of hippopotamid dentition from a bothriodontine dentition would depend on which bothriodontines are included within the sister group of Hippopotamidae .

Crown bothriodontines as sister group of Hippopotamidae

The complete data matrix analysis indicated that the sister group of Hippopotamidae included solely Libycosaurus and Merycopotamus , i.e. the most advanced Old World bothriodontines (or crown bothriodontines), to the exclusion of the other anthracotheriids considered in this work ( Fig. 9 ). This topology is based mainly on craniomandibular and frontal dentition character changes: width of mandibular symphysis (2:1); about constant corpus height (4:1); cylindrical roots of lower incisors (7:0, to be lost in Merycopotamus ); and large lower canines, at least in males, reaching below P 3 –P 4 and enamel extensively covering the crown (11:1, 12:1, 13:1). In previous contributions ( Boisserie et al., 2005a, b), a close relationship of Hippopotamidae with crown anthracotheriids was largely established on features from the same morphocomplexes. Notably, a reversal of the cylindrical lower incisor roots in Merycopotamus does not appear likely, and we favour the parallel evolution of this trait in Libycosaurus and Hippopotamidae .

In addition, dental features that support the clade ( Hippopotamidae + crown bothriodontines) are: the presence of one to several postparaconules on the P 3 (23:1); the partial fusion of the lingual and the distolabial roots on the P 4 (31:1) – an ambiguous synapomorphy not verifiable for middle Miocene Kenyapotamus and A. lothagamensis ; the great reduction or loss of the upper molar paraconule (44:2); and the great reduction or loss of the premetacristid on M 1-2 (69:1). Two reversals would have therefore occurred in Kenyapotamus : the complete redevelopments of the paraconule and of the premetacristid. Once again, this seems an unlikely scenario perhaps better interpreted as parallel losses.

This close relationship with crown bothriodontines fits scenarios proposed by Gaziry (1987) and Boisserie & Lihoreau (2006). They suggested that the origin of Hippopotamidae may be looked for within early to middle Miocene African bothriodontines including Afromeryx and Sivameryx . First, this hypothesis involves a certain degree of parallelism in craniomandibular structures (e.g. high orbits) between hippopotamids and crown bothriodontines, in particular Libycosaurus (see Boisserie et al., 2005a, b). Second, it also relies on a relatively complex modification of the cheek tooth pattern as discussed above and by Boisserie & Lihoreau (2006). In this regard, examin- ing the phylogenetic signal using only cheek tooth characters was an important test.

Advanced bothriodontines as sister group of Hippopotamidae

This test resulted in two conflicting phylogenetic hypotheses (see Supporting Information). Half of 14 resulting trees displayed the same topology as that shown in Figure 9 . In contrast, the other half exhibits a sister group of Hippopotamidae that includes crown bothriodontines and Elomeryx (‘advanced bothriodontines’; Fig. 10 ). This would indicate a possible emergence of Hippopotamidae within more basal bothriodontines, relatively close to Bothriogenys . The clade grouping Hippopotamidae and their sister group would be defined by the loss of the upper molar ectometacristule (48:2); the occurrence of a metaconid on P 3 (54:1); and the shallow ectoprotofossid on M 1-2 (66:1). The occurrence of one or several P 3 postparaconules (23:1) and the upper molar paraconule reduction (44:1) would be retained as ambiguous clade support.

This topology would not require a redevelopment of labial cingulum structures on hippopotamid upper molars. In fact, as discussed in Appendix 2 (character 34), this structure may not have disappeared in advanced bothriodontines, but simply merged with the extensions of the postparacristae and premetacristae. This situation differs from the simple cingular structures observed in Bothriogenys and Hippopotamidae . Similarly, this topology involves: no reappearance of the postprotofossid on P 4 in Hippopotamidae , but its presence as a plesiomorphic state found in anthracotheriines and basal bothriodontines, lost in more advanced bothriodontines; and no reversal in Kenyapotaminae of the paraconule reduction and of the important trigonid modifications evolving in parallel within the Hippopotaminae and crown bothriodontines. Most probably, the paraconule loss occurred independently on multiple occasions within Anthracotheriidae ( Lihoreau & Ducrocq, 2007) . This reduction may also have followed parallel pathways in basal hippopotamines and within crown bothriodontines, and a similar scenario may have applied to the evolution of the trigonid. The fusion of the P 4 roots would also have occurred independently within crown bothriodontines and Hippopotamidae .

If we extrapolate for noncheek-tooth characters, this topology ( Fig. 10 ) implies that the development of the mandibular symphysis and lower canines could have been reversed in Elomeryx , or been acquired in parallel in Hippopotamidae and crown bothriodontines. Similar parallelism would be found for: the cylindrical lower incisor root (developed in Hippopotamidae and Libycosaurus only); the constant mandibular corpus height; the extension of the enamel band to the whole length of the lower canine.

An emergence of Hippopotamidae within basal bothriodontines ( Fig. 10 ) seems ‘dentally more realistic’ than the scenario derived from the complete data matrix analysis. A dental transition could be relatively easy to imagine, especially if it involved cheek tooth enamel thickening ( Boisserie & Lihoreau, 2006). However, this scenario implies a parallelism between hippopotamines and crown bothriodontines greater than previously suggested ( Boisserie et al., 2005a, b), affecting more cranial and dental structures. This parallelism would appear to be particularly extreme between Libycosaurus petrocchii and the most advanced Hippopotaminae , such as Hippopotamus ( Boisserie et al., 2005a) . This situation could have resulted from the occupation of the same niche – in this case, that of semiaquatic large herbivores – by species with closely related genomes and sharing close social organizations (i.e. structured herds, with some form of segregation between males and females).

Figures

Figure 1. Occlusal views of upper dentition in Anthracotheriidae, Hippopotaminae, and palaeochoerids. A, Merycopotamus nanus, alveoli of right I1–C, and right P1–M3 (specimen number: Y47189, repository: Geological Survey of Pakistan, Museum of Natural History in Islamabad, Pakistan – GSP). B, Merycopotamus nanus, left* M2–M3 (Y47189, GSP). C, Anthracotherium sp., right P3–M3 (University of Montpellier 2 cast of ACQ 6608, Ecomusée des Phosphatières, Bach, France). D, Hexaprotodon garyam, right M3 (TM059-01-012, Centre National d’Appui à la recherche, Ndjamena, Chad – CNAR). E, Archaepotamus harvardi, left* P2–M3 (KNM-LP 8731, National Museums of Kenya, Nairobi – NMK). F, Ar. harvardi, left* M1–M2 (KNM-LT 23838, NMK). G, Palaeochoerus pusillus, right P4–M3 (Qu 15, Muséum National d’Histoire Naturelle, Paris, France – MNHN). H, Schizochoerus vallesensis, right P3–M3 (TRQ 1018, MNHN). I, Doliochoerus quercyi, left* M1? (Qu 7092, MNHN). * indicates reversed images for comparison purposes. Localities: Y, Potwar plateau, Pakistan; TM, Toros-Menalla, Chad; TRQ, Yassiören, Turkey; LP, Loperot, Kenya; LT, Lothagam, Kenya; Qu and ACQ, Quercy, France. Scale bars = 1 cm.

Figure 9. Strict consensus tree of the 14 most parsimonious phylogenetic trees (289 steps; consistency index = 0.474; retention index = 0.724) obtained by cladistic analysis (heuristic search) of the complete matrix (24 taxa and 87 characters; see Appendices and Supporting Information). Numbers at nodes are: Bremer index/ bootstrap index (1000 replicates,> 50% occurrences). CB, crown bothriodontines; Hid, Hippopotamidae; Hoid, Hippopotamoidea; Hin, Hippopotaminae; K, Kenyapotaminae; MMK, middle Miocene kenyapotamines; LMK, late Miocene kenyapotamines.

Figure 10. One of the eight most parsimonious phylogenetic trees (230 steps; consistency index = 0.470; retention index = 0.724) obtained by cladistic analysis (heuristic search) of the matrix including only cheek tooth characters (23 taxa and 66 characters – character 16 to character 81; see Appendices and online Supporting Information). Three other trees presented a similar topology within Hippopotamoidea, the four others displaying a topology compatible with Figure 9. Numbers at nodes are: Bremer index/ bootstrap index (1000 replicates,> 50% occurrences). AB, advanced bothriodontines; Hid, Hippopotamidae; Hoid, Hippopotamoidea; K, Kenyapotaminae; MMK, middle Miocene kenyapotamines; LMK, late Miocene kenyapotamines.