Molecular Systematics of the Old World Monkey Tribe Papionini

Home , Dinopithecus, Gorgopithecus, Papionini, Parapapio

Eugene E. Harris* Molecular systematics of the Old World Department of Anthropology, monkey tribe Papionini: analysis of the 25 Waverly Place, New York total available genetic sequences University, New York, NY 10003, U.S.A. The phylogenetic relationships among the genera of the tribe Received 23 March 1998 Papionini are inferred using a taxonomic congruence approach in Revision received which gene trees derived for eight unlinked genetic sequence datasets 20 January 1999 and are compared. Population genetics theory predicts that species accepted 10 April 1999 relationships will be revealed with greater probability when the topology of gene trees from many unlinked loci are found to be Keywords: Old World congruent. The theory underlying this approach is described. monkeys, Papionini, Monophyly of the mangabeys is not supported by any of the gene molecular phylogeny, trees; instead, they are polyphyletic with Cercocebus found to be the systematics, mangabeys, sister taxon to Mandrillus in ﬁve gene trees (with no conﬂicting Papio, Theropithecus, trees), and Lophocebus found to be closely related to Papio and/or Mandrillus, Lophocebus, Theropithecus in all trees. Theropithecus and Papio are not strongly Cercocebus, Macaca. supported as sister taxa (present in one or two trees only); Lophocebus and Papio are supported as sister taxa in the majority of trees. A close relationship between Mandrillus and Papio is not supported in any of the trees. The relationships among Papio, Lophocebus, and Theropithecus cannot be resolved by congruence, probably due to the short time interval estimated between their divergences. The mtDNA COII sequences are used to estimate divergence dates within the papionins. The internode between the divergences of these species is estimated to be between 290 ka and 370 ka. Lastly, the evolution of morphological features such as long faces, suborbital facial fossae, and terrestrial skeletal adaptations is discussed. 2000 Academic Press

Journal of Human Evolution (2000) 38, 235–256 doi: 10.1006/jhev.1999.0318 Available online at http://www.idealibrary.com on

Introduction Lophocebus), as opposed to the largely Asian distributed macaque genus (Macaca) The Papionini comprise a group of six (Strasser & Delson, 1987; Disotell, 1992; genera of Old World monkeys which are 1994; Disotell et al., 1992; Morales & geographically widespread and ecologically Melnick, 1998). As a group, papionins are diverse. They can be subdivided into two characterized by facial lengthening, usually groups, the exclusively African papionins, some development of facial fossae, increased including geladas (Theropithecus), baboons 1 use of terrestrial substrates, and a diploid (Papio), mandrills and drills (Mandrillus), karyotype of 42. All known systematic and the mangabeys (Cercocebus and studies of the Old World monkeys, based *Present address and address for correspondence: on either genetic or morphological data, Eugene E. Harris, Department of Genetics, Nelson Biological Labs, 604 Allison Road, Busch Campus, have found the papionin tribe to be a Rutgers University, Piscataway, NJ 08855-1059, monophyletic group. U.S.A.. Tel: (732)-445-2681; Fax: (732)-445-5870; The phylogenetic relationships among E-mail: [email protected] 1A population of Papio hamadryas is known to inhabit genera have been studied using essentially coastal regions of Yemen and Saudi Arabia. two classes of evidence, morphological and

0047–2484/00/020235+22$35.00/0 2000 Academic Press 236 . .  Macaca Theropithecus Cercocebus Lophocebus Mandrillus Papio Macaca Cercocebus Lophocebus Theropithecus Mandrillus Papio

a b Figure 1. Two morphologically-based trees of papionin phylogeny. Tree (a) was advocated by Kuhn (1967), Szalay and Delson (1979),andStrasser and Delson (1987); tree (b) is after Delson and Dean (1993). molecular. The various hypotheses for their molecular trees based on immunology and relationships based on morphology, how- mtDNA encoded COII sequences (Sarich, ever, strongly disagree with most molecu- 1970; Cronin & Meikle, 1979, 1982; lar phylogenies. This discordance among Disotell, 1992, 1994; Disotell et al., 1992)? hypotheses, while strongly apparent between There now exist nine datasets of DNA molecular and morphological hypotheses, and amino acid sequences for papionin is not peculiar to this dichotomy and is genera, which can be analyzed together to also found, although to a lesser extent, examine these three questions. These among molecular hypotheses (as this study sequences derive from genomic regions describes). belonging to different chromosomes or dif- There are two central systematic ferent genomes (i.e., nuclear and mito- questions. First, are mangabeys mono- chondrial) and therefore can be subdivided phyletic as posited by most morphological into separate linkage groups. A linkage trees (see Figure 1; Jolly, 1967; Kuhn, 1967; group represents a unit of the genome which Hill, 1974; Szalay & Delson, 1979; Strasser segregates independently and recombines & Delson, 1987) or polyphyletic as pro- freely with respect to other such groups. posed in molecular and chromosomal The tree estimated based on each gene studies (Cronin & Sarich, 1976; Hewett- region is properly conceived of as a gene tree Emmett et al., 1976; Hewett-Emmett & in contrast to a species trees (see Hey, Cook, 1978; Dutrillaux et al., 1979; 1994). The terminal branches of gene trees Disotell, 1992, 1994; Disotell et al., 1992; bear homologous gene sequences and not Harris & Disotell, 1998) and some morpho- organisms or ‘‘species’’. Likewise, the nodes logical studies (Groves, 1978; Fleagle & of gene trees represent ancestral DNA McGraw, 1989)? Furthermore, if they are sequences and not ancestral organisms or polyphyletic how is each mangabey genus ‘‘species’’. Proceeding back in time from the related within the papionin group? Second, terminal DNA sequences to the ancestral do the genera Papio and Mandrillus share a DNA sequence represents a process known close relationship within the papionins as as coalescence. proposed by all morphologically based It is often assumed that gene trees hypotheses (see Figure 1), or is Theropithecus accurately estimate the species tree. How- most closely related to Papio as proposed by ever, there are several reasons why this       237 assumption may be incorrect. First, gene when species divergences have occurred trees may be incorrectly estimated because relatively close in time (Nei, 1987; Pamilo & of sampling bias when trees are estimated Nei, 1988). from relatively short sequences. Tradition- A solution to the problem of gene tree/ ally, the length of homologous DNA species tree mismatch lies in the search for sequences compared across species is on the congruence among the branching patterns order of several hundred base pairs. Saitou of gene trees that are derived from separate and Nei (1986) show that this problem can linkage groups (Saitou & Nei, 1986; Pamilo be diminished if amply long DNA sequences & Nei, 1988; Wu, 1991). This is because are compared. Second, natural selection population level processes like speciation are could have constrained variation among expected to produce similar effects across lineages resulting in a tree that conflicts with many loci. In contrast, processes that can actual phylogenetic relationships. Third, obfuscate species relationships like random assuming that gene trees have been correctly lineage sorting and natural selection are estimated, incongruence between gene trees expected to have locus specificeffects (Hey, and the species tree can be due to random 1994). Following this reasoning, when a lineage sorting. The theoretical basis of majority of gene regions are concordant in random lineage sorting has been discussed supporting trees showing the same branch- extensively in the population genetics litera- ing pattern, the most likely explanation is ture (see Hudson, 1983; Nei, 1987; Pamilo & that this pattern was shaped by the actual Nei, 1988; Rogers, 1993; Hey, 1994; Moore, species divergences. Congruence among 1995, 1997; Avise & Wollenberg, 1997; gene trees derived from unlinked loci is Hoelzer, 1997; Doyle, 1997; Maddison, therefore a powerful tool for inferring 1997) and is briefly described here. species trees. This general approach has The coalescence time of homologous gene been described as taxonomic congruence sequences sampled from two sister species (Mickevich, 1978)—the agreement among will predate the divergence time of the the supported topologies of different data species. When the coalescence of homolo- sets (Miyamoto & Fitch, 1995). gous gene sequences occurs in the most In this paper, a robust phylogenetic tree recent common ancestral population of two for all papionin genera is developed using a species, the topology of the gene tree will be taxonomic congruence approach in which the same as the species tree. But if the two the total available genetic sequences for homologous DNA sequences fail to coalesce papionin genera are analyzed. The central in the ancestral population of these two phylogenetic questions presented above are species, instead coalescing in the common the focus of the study. The pattern of dis- ancestral population that these species share cordance among gene trees is interpreted in with a third species, then the gene tree may terms of population genetics theory and its not reflect the actual order of divergences significance for understanding the evolu- among the species. This is because the DNA tionary history of papionins is discussed. lineages found in the ancestral population Finally, the evolution of morphological fea- have randomly sorted into the three tures within the papionin group is discussed descendant species lineages. The mismatch in light of the molecular phylogeny. between gene trees (from multiple unlinked loci) and the species tree will occur with a Background probability that can be modeled by population genetics theory and may be consider- Harris & Disotell (1998) obtained gene trees able under certain conditions, for example for five nuclear gene regions including CD4, 238 . . 

Cercopithecus aethiops 5 Cercopithecus mitis 1 Macaca mulatta 1 d1 [51] Cercocebus galeritus 1 Cercocebus torq. lun. 1 d1 d1 Cercocebus torq. atys 1 d5 [79] Mandrillus sphinx [97] d1 [64] Mandrillus leucophaeus Papio anubis 1 d1 Theropithecus gelada 1 [71] d1 Lophocebus aterrimus 2 Lophocebus albigena 2 [78] d2 [81]

CD4 (Chr. 12)

Cercopithecus aethiops 3 Cercopithecus aethiops 4 d8 Cercopithecus mitis 1 [100] Cercopithecus mitis 2 Macaca mulatta 2 Cercocebus galeritus 1 Cercocebus torq. lun. 1 Cercocebus torq. lun. 2 d2 Cercocebus galeritus 2 d2 [81] d4 Mandrillus leucophaeus [90] [97] Mandrillus sphinx Cercocebus torq. atys 1 Cercocebus torq. atys 2 Theropithecus gelada 2 d2 Papio hamadryas 1 d2 [88] Papio hamadryas 2 [88] Lophocebus aterrimus 2 d3 Lophocebus albigena 2 [96] Lophocebus albigena 1 TSPY (Y) Figure 2. Eight gene trees obtained using maximum parsimony. The CD4, TSPY, 1,3 GT, and hemoglobin trees are all strict consensus trees. The Papio hamadryas 1 sequence in the TSPY tree is from Kim & Takenaka (1996). The --globin intergenic and COII & 12S rRNA combined mtDNA trees were the single MP trees estimated for these regions. The IRBP tree is the single MP tree obtained using an a priori weighting scheme in which transversions were given twice the weight as transitions. The numbers following taxa indicate diﬀerent individuals sequenced. In the hemoglobin tree, xa, xb, yc indicate diﬀerent alleles. The prion protein tree is a 50% majority rule tree. Bootstrap percentages (in brackets) and decay indices (denoted by ‘‘d’’) are given alongside the branches. A decay index of zero (d0) means this clade was not present in some but not all most parsimonious reconstructions (see Bremer, 1994).

TSPY, the --globin intergenic region, trees for these regions are used in the present 1,3 galactosyltransferase ( 1,3 GT), and analysis they are brieﬂy reviewed here [see IRBP (see Figure 2). These regions range in Harris & Disotell (1998) for additional length from 514 to 717 bp and consist of details]. The phylogenetic methods used noncoding DNA sequences comprising to obtain these trees are identical to the either intron segments, intergenic regions, methods employed here (see Materials and or pseudogene sequences. Since the gene methods).       239

Figure 2. (Continued—legend on facing p. 238.) 240 . . 

All three maximum parsimony (MP) trees Four additional genetic sequence datasets (length=42; see Figure 2) for CD4 are are included in the analyses presented here. identical and support mangabey polyphyly These include reanalyses of DNA sequence in which Cercocebus forms a unique clade datasets from the mitochondrially encoded with Mandrillus, and Lophocebus forms a cytochrome oxidase subunit II (COII) gene clade with Theropithecus and Papio. Lopho- (Disotell et al., 1992; Disotell, 1992, 1994); cebus mangabeys are supported as the the mitochondrially encoded 12S rRNA exclusive sister taxon to Theropithecus, with gene (Van der Kuyl et al., 1995a); the Papio being the sister taxon to this pair. The nuclear encoded prion protein gene (Schatzl position of Macaca is left unresolved in et al., 1995; Van der Kuyl et al., unpublished relation to the two African papionin clades. manuscript; Krakauer et al., 1996); as well A priori weighting did not change the as reanalysis of an amino acid sequence results. dataset from the nuclear encoded The six MP trees (length=51; Figure 2) hemoglobin gene (Hewett-Emmett et al., for TSPY are essentially identical to the tree 1976). for CD4, except that the relationships among Theropithecus, Papio, and Lophocebus are left unresolved in all trees. A priori Materials and methods weighting did not aﬀect the results. The single MP tree for the --globin In total, eight separate genetic loci are intergenic region, found in both uniform examined in the present paper. The and a priori weighted analyses (length=43; papionin taxa represented by these genetic Figure 2), is essentially the same as the datasets and used in the current analysis are previous trees with the exception that listed in Table 1. These sequences were Papio is supported as the sister taxon to either collected by the author (see Harris & Lophocebus. Disotell, 1998) or were downloaded from The strict consensus tree of ﬁve MP trees the GenBank database at the National (length=33; Figure 2) found for the 1,3 Center for Biotechnology Information. The GT region is relatively poorly resolved. The outgroup taxa in all analyses consisted of only clade supported is a Lophocebus–Papio Cercopithecus aethiops and/or C. mitis (and C. clade. The cladistic positions of all other mona in the analysis of the prion protein genera are unresolved at the base of the sequences), belonging to the tribe Cerco- papionin tree. A priori weighting did not pithecini, the nearest sister group to the change the results. Papionini (Strasser & Delson, 1987). Analysis of IRBP employing uniform Table 2 reports the length (in base pairs) weighting resulted in two MP trees and inferred chromosomal position of the (length=30; Figure 2). These trees both gene regions analyzed here along with support mangabey polyphyly with Cercocebus references to the studies in which these in a clade with Mandrillus, and Lophocebus in sequences were initially analyzed. The prion a clade with Papio. Theropithecus is either protein sequences, comprising approxi- placed in an unresolved position at the base mately 744 bp region of the second exon of of the African papionin tree (uniform the prion protein gene, were collected by weighting) or in a clade with Papio, and Schatzl et al. (1995) and Van der Kuyl et al. Lophocebus (in the single a priori weighted (unpublished manuscript). The second tree). However, Macaca is anomalously exon, encoding the entire prion protein, supported as the sister taxon to the is expressed at high levels in brain tissue Papio–Lophocebus clade. and is believed to be involved in several Table 1 Genetic sequences available for species within the tribe Papionini and for several outgroup species of Cercopithecus*      

--globin intergenic Prion hemoglobin Species name CD4 TSPY region a1,3GT IRBP protein amino acid COII 12SrRNA

Cercopithecus aethiops ++ + + + + + + + Cercopithecus mitis ++ + + + Cercopithecus mona + Cercocebus galeritus chrysogaster ++ + + + C. galeritus sp. ++ C. torquatus lunulatus ++ + + + C. torquatus atys ++ + + + + + + Lophocebus aterrimus ++ + + + + + + L. albigena albigena ++ + + + L. albigena sp. + Macaca mulatta ++ + + + + + + + Papio cynocephalus + Papio anubis ++ + + + + + + + Papio hamadryas + ++ Theropithecus gelada ++ + + + + + + + Mandrillus sphinx ++ + + + + + + Mandrillus leucophaeus ++ + + + + +

*This list is not meant to be comprehensive. It serves to outline which species are represented in the current study. 241 242 . . 

Table 2 Gene regions analyzed in this study subdivided into separate linkage groups

Presumed Size Human papionin Linkage (bp/amino chromosomal chromosomal Gene region group* acids) location† location‡ References§

Nuclear genome CD4 1 717 bp 12p12–pter 12 1 TSPY 2 692 bp Ypter–p11·2Y1 intergenic 3 672 bp Approx. 11p15·4–15·5111 1,3 galactosyltransferase 4 514 bp 9q33–q34 14 1 IRBP 5 685 bp 10p12 10 1 Prion protein 6 744 bp 20p12–pter 20 2 -chain hemoglobin 7 14 a.a. 16 16 3 Mitochondrial genome COII 8 697 bp ——4 12S rRNA 8 394 bp ——5

*The numbers assigned to linkage groups are arbitrary. All gene regions belong to diﬀerent linkage groups except the COII and 12S rRNA genes which belong to a single linkage group. †Human chromosomal locations were obtained from either literature sources or from any of several computer databases including The Human Genome Database, The Online Inheritance in Man data base, GenBank, or the Human Genome Map. p=short arm; q=long arm; ter=terminal. ‡Locations in papionins were inferred from chromosomal homology studies between humans and the papionin species Macaca fuscata (Weinberg et al., 1992). §1, Harris & Disotell (1998);2,Van der Kuyl et al. (unpublished manuscript);3,Hewett-Emmett et al. (1976); 4, Disotell et al. (1992);5,VanderKuylet al. (1994).

degenerative neurological diseases in mam- in respiration, and 12S rRNA codes for a mals, including primates. The hemoglobin ribosomal RNA molecule. sequences comprise 141 amino acid residues Genes falling on different chromosomes collected and analyzed initially by Hewett- necessarily have a recombination fraction of Emmett et al. (1976) and Hewett-Emmett & 50% and therefore segregate independently Cook (1978). with respect to each other in the formation The nuclear gene regions, including those of gametes. Such genetic regions constitute initially analyzed in Harris & Disotell (1998) distinct linkage groups. Since each nuclear are each found on different chromosomes in gene region analyzed for the papionins is humans and probably in papionins as well inferred to fall on a different chromosome, (see Table 2). Their inferred chromosomal there are a total of seven nuclear linkage location in papionins was based on the groups. The two mtDNA sequence datasets, chromosomal synteny study of Weinberg however, form a single additional linkage et al. (1992) in which humans were com- group since the mitochondrial genome is pared with the papionin species Macaca inherited independently of the nuclear fuscata. genome and undergoes no recombination The mitochondrial gene sequences among its constituent genes. include both the COII (Disotell et al., 1992) and 12S rRNA (Van der Kuyl et al., 1995a) Phylogenetic methods coding regions. The COII gene codes for Phylogenetic analyses were performed on cytochrome oxidase subunit II, a mitochon- each of the genetic data sets separately fol- drial electron-transport enzyme functioning lowing a taxonomic congruence approach       243

(Mickevich, 1978; Miyamoto and Fitch, reflecting empirically observed asymmetries 1995). An analysis was also done on a single in nucleotide substitutions (see Kimura, combined dataset in an analysis of ‘‘total’’ 1980; Nei, 1987; Ruvolo, 1997; Harris & evidence (see Kluge, 1989). As the number Disotell, 1998). A priori weighting of the of taxa and particular taxa sequenced for mtDNA sequences used the average of each gene region differed, a combined data- the empirically determined transition/ set was developed for only those taxa repre- transversion ratios (Tr/Tv) found for the sented in all gene regions. In the cases of COII and 12S rRNA sequences, respect- Lophocebus and Mandrillus, sequences from ively. Tr/Tvs were determined by counting different species within these genera needed the number of transition and transversion to be combined in order to form the ‘‘com- changes appearing on the MP trees found bined’’ sequence for these genera. Although for these genes. In the analysis of the this author believes that a combined analysis hemoglobin amino acid sequences all is not well justified theoretically, at least for substitutions were weighted equally. molecular data drawn from distinct genomic regions, it has been included because the approach has been strongly advocated by Results some traditional systematists (Kluge, 1989; Ernisse & Kluge, 1993; Kluge & Wolf, The results of all phylogenetic analyses 1993) and because the results of a combined including the trees obtained in Harris & analysis in the current study can be used to Disotell (1998) are summarized in Table 3. point out the shortcomings of this approach. Results from the reanalysis of the previously Phylogenetic analyses employed maxi- published genetic sequences from the COII, mum parsimony (MP) using the computer 12S rRNA, prion protein, and hemoglobin program PAUP version 3.1 (Swofford, genes are presented below. Additionally, 1993). All MP analyses employed PAUP’s both a consensus tree of the separate MP branch and bound search option. Deletions gene trees is developed, and a combined tree were coded as gaps and treated as a fifth is derived in an analysis of the genetic character state. Multiple base deletions were sequences from each different region pooled treated as a single event. Bootstrap confi- as a single dataset. The COII gene dence levels were based on 500 random sequences are used to estimate divergence samplings of the data. Decay indices (DI), dates for papionin genera. representing the difference in tree length between the most parsimonious tree pos- hemoglobin sessing a particular clade and the closest The MP trees originally obtained by suboptimal tree(s) in which this clade is not Hewett-Emmett et al. (1976; Figure 3) for present (Bremer, 1994), were calculated the hemoglobin amino acid sequences using AutoDecay version 2.9.6 (Eriksson, were similar to the strict consensus tree (of 1996). MP tree lengths and homoplasy seven most parsimonious trees, length=18, indices (HI) are described for all trees. HI=0·250; Figure 2) obtained in the re- Two different character weighting analysis of their sequences. The consensus schemes were used in the maximum tree is largely similar to the previous gene parsimony analyses. In the uniform weight- trees described, however, Lophocebus is sup- ing scheme, all nucleotide substitutions were ported as the sister taxon to Papio (as in the weighted equally. In the a priori weighting --globin intergenic region, 1,3 GT, scheme, nuclear sequences were weighted and IRBP trees) with Theropithecus as their with a 2:1 transition–transversion ratio, sister taxon. 244

Table 3 Papionin phylogenetic hypotheses and their support in the gene trees derived from unlinked gene regions as well as for all the gene regions combined

COII/12s chain Prion Phylogenetic CD4 intergenic 1,3 GT IRBP rRNA hemoglobin protein Combined hypotheses (12)* TSPY (Y) (11) (14) (10) (mtDNA)¶ (16) (20) sequences

Mangabey polyphyly ✔† ✔✔UR ✔✔ ✔ ✔ ✔ 

Mangabey monophyly ✗‡ ✗✗✗✗✗✗✗✗. 

Theropithecus–Papio–Lophocebus ✔✔ ✔ UR§ ✗✔ ✔ ✗ ✔. Cercocebus–Mandrillus ✔✔ ✔ UR ✔✔ UR UR ✔  Lophocebus–Papio ✗ UR ✔✔✔✔✔✗✔ Lophocebus–Theropithecus ✔ UR ✗✗✗✗✗✔✗ Theropithecus–Papio ✗ UR ✗✗✗✗✗✗✗ Mandrillus–Papio ✗✗ ✗ ✗✗✗ ✗ ✗ ✗

*Inferred chromosomal location in papionins. †A check-mark indicates that this relationship is supported by maximum parsimony. ‡An ✗ indicates that this relationship is not supported by maximum parsimony. §UR indicates that this relationship is unresolved. ¶All mtDNA trees referred to are those in which the 12S rRNA sequences for Theropithecus were excluded, and in which an a priori weighting scheme was employed (see text for details).       245

Prion protein Cercopithecus A 50% majority rule consensus tree of four most parsimonious trees (length=16; Macaca HI=0·333; Figure 2) for the prion protein Theropithecus gene region is largely anomalous compared 71% with the other gene trees. The genes for 100% Lophocebus Cercopithecus mona and Macaca group with Papio 71% Cercocebus, Mandrillus and Papio. The genes 57% of Lophocebus and Theropithecus form a Cercocebus related group. The a priori weighted MP tree a 71% Mandrillus is unchanged. Schätzl et al. (1995), Van der Kuyl et al. (unpublished manuscript), and Krakauer et al. (1996) similarly found that their analyses of prion protein sequences yielded a number of anomalous groupings Cercopithecus within the primates not found in other gene Macaca trees for this group. d23 Mandrillus [100] Cercocebus Combined COII and 12S mtDNA sequences d7 [100] Theropithecus The combined mtDNA sequences from the d31 COII and 12S rRNA genes (see Disotell [100] d2 Lophocebus [56] et al., 1992 and Van der Kuyl et al., 1995a, b Papio respectively) comprise a dataset of 1078 bp. Figure 3. (a) 50% majority rule consensus tree of the Since analyses of the 12S rRNA gene by seven separate gene trees obtained using maximum itself places the Theropithecus sequence parsimony. Percentages indicate the fraction of the total anomalously outside the papionins (perhaps gene trees supporting a particular clade; (b) single MP tree found in an analysis of a single dataset of all seven being a nuclear copy of the mitochondrial nuclear gene regions and the two mitochondrial gene gene; see Van der Kuyl et al., 1995b),2 regions combined. analyses were done with the sequence for T. gelada excluded. In the a priori weighted analyses, transversions were weighted six sister taxon to Mandrillus. Macaca is sup- times that of transitions using the mean of ported as the sister taxon to the African the Tr/Tv ratios calculated for both the papionins. However, the a priori weighted COII and 12S rRNA genes based on MP tree supports Lophocebus rather than their separate MP trees (7·9/1 and 4·8/1, Theropithecus as the sister taxon to Papio. respectively; mean=6·35/1). The uniformly weighted MP (length= Consensus tree. A 50% majority rule 386, HI=0·377; Figure 2) tree is fully consensus tree indicating congruent clades resolved and is largely similar to the previous amongst seven MP gene trees is given in gene trees (except prion protein). Lophoce- Figure 3. The consensus tree was derived bus, Papio, and Theropithecus form a clade, as based on the five gene trees from Harris & in most of the other gene trees, with Thero- Disotell (1998) and two of the gene trees pithecus as the sister taxon to Papio, exclusive derived in the current study. The actual of Lophocebus. Cercocebus is supported as the trees used in the consensus analysis were pruned versions of the trees in Figure 2,in 2Van der Kuyl et al. (1995b) found nuclear counterparts of the mtDNA encoded 12S rRNA gene in several which only a single sequence was selected to Old World monkey species. represent each papionin genus. The prion 246 . .  protein gene results were excluded from Lophocebus clade, are strongly supported by this analysis because its trees are notably decay indices as well as bootstrap measures. anomalous. Therefore, the tree represents the consensus of seven gene trees. Estimation of divergence dates. Divergence It is the case that all the gene trees reject dates were estimated based on the mtDNA monophyly of the mangabeys. Five gene COII DNA sequences using distances calcu- trees (71%) support mangabeys (Lophocebus lated in MEGA 1·01 (Kumar et al., 1993). and Cercocebus) as falling into two separate Distances are based on third codon pos- clades (CD4, TSPY, --globin intergenic itions only, because most substitutions at region, IRBP, combined mtDNA). In these these positions are synonymous (i.e., cause trees, Cercocebus is supported as the sister no amino acid change), and were corrected taxon to Mandrillus and Lophocebus in a using Kimura’s (1980) two-parameter clade with Theropithecus and Papio. Within method since considerably transition– the clade consisting of Lophocebus, Thero- transversion bias occurs. Estimations were pithecus, and Papio, five trees show Papio restricted to the COII gene since its esti- and Lophocebus as sister taxa exclusive to mated distances have the smallest associated Theropithecus, and two trees show either standard errors (e.g., the mean standard Lophocebus as the exclusive sister taxon to error of the distances is 14%). In contrast, Theropithecus (CD4), or these three taxa as the Kimura corrected distances for the an unresolved trichotomy (TSPY). The two nuclear DNA regions (TSPY, the -- trees that do not place each mangabey genus globin intergenic region, CD4, 1,3 GT, into different clades leave the relationships and IRBP) have associated standard errors of at least one of these genera (Cercocebus) that range from 28·29% (TSPY) to as great unresolved at the base of the African as 47·49% ( 1,3 GT) presumably due to papionin tree. These unresolved trees are relatively low levels of sequence divergence not incongruent with mangabey polyphyly. (see Harris and Disotell, 1998), thus giving Macaca is supported as the sister taxon to the COII distances relatively the highest the African papionins in three gene trees level of confidence. Divergence dates were --globin intergenic region, mtDNA, not estimated for the combined dataset of all hemoglobin). Only a single tree (IRBP) gene regions. Doing this ignores the signifi- conflicts with this position; the remaining cant problem caused by combining regions trees leave the relationship of this genus having disparate levels of diversity reflecting unresolved at the base of the papionin tree. different rates of evolution. It also ignores the widely different standard errors associ- A combined analysis of the ‘‘total’’ sequences ated with the distances estimated for these All seven nuclear datasets (including prion regions. protein) were combined with the mtDNA Separate estimates of the times of dataset (excluding the 12S rRNA sequence species divergences were made using two for T. gelada as above) to form a single large different calibration points: a divergence dataset consisting of 5252 bp of sequence between humans and orang-utans at around for papionins. The single uniformly 13 million years, and a divergence of Thero- weighted tree (length=590, HI=0·188, pithecus from the African papionins at Figure 3) supports a clade comprised of around 4 million years. These dates are the Papio, Lophocebus, and Theropithecus with earliest times at which we find fossils these first two genera linked as sister taxa; attributable to Sivapithecus, an early deriva- Cercocebus shares a unique clade with tive of the orang-utan lineage (Kappelman Mandrillus. All clades, except for the Papio– et al., 1991) and Theropithecus (Eck &       247

Table 4 Relative rate tests using Cercopithecus aethiops as the outgroup species based on the Kimura (1980) two-parameter corrected distances for third codon positions

Diﬀerence between ingroup branch distances Ingroup 1 Ingroup 2 (S.E. of diﬀerence) P value

Cercocebus galeritus Papio anubis 0·049 (0·085) 0·2810 Theropithecus gelada 0·066 (0·087) 0·2236 Lophocebus albigena 0·071 (0·083) 0·1949 Mandrillus leucophaeus 0·017 (0·064) 0·3936 Macaca mulatta 0·124 (0·092) 0·0901 Mandrillus leucophaeus Papio anubis 0·066 (0·089) 0·2266 Theropithecus gelada 0·084 (0·092) 0·1814 Lophocebus albigena 0·088 (0·086) 0·1515 Macaca mulatta 0·141 (0·093) 0·0643 Theropithecus gelada Papio anubis 0·017 (0·080) 0·4129 Lophocebus 0·005 (0·080) 0·4761 Macaca mulatta 0·058 (0·101) 0·2843 Macaca Papio anubis 0·075 (0·102) 0·2327 Lophocebus albigena 0·053 (0·098) 0·2946

Jablonski, 1984; Szalay & Delson, 1979), divergence of these genera from each other though Delson & Dean (1993) note that the may tend to be underestimated. When the earliest Theropithecus may be slightly more divergence date for the Macaca lineage recent, and they provide minimum estimates was estimated, in which the distances to of the divergence times for these species. Mandrillus and Cercocebus were averaged Dates were calculated assuming a uniform with the distances to the remaining genera, substitution rate and were based on the it was found to differ only slightly from the mean of the successive distances between estimated date when the distances to these a taxon and each in-group taxon. For genera were excluded (Table 5). example, for the divergence between Additionally, since one of the calibration Macaca and the African papionins, the points falls outside the papionins (and distances between the genus and Papio, within the hominoids), it is important to Theropithecus, Lophocebus, Cercocebus and calculate the relative rates between this Mandrillus were averaged and used in the group and the papionins. With respect to date calculations. this, Adkins et al. (1996) found that the rates Relative rate tests of lineages within the in these two groups are homogeneous when Papionini, using Cercopithecus aethiops as an comparing synonymous substitutions but outgroup, did not show any of the rates to be heterogeneous for nonsynonymous substi- significantly different (Table 4). However, tutions. This justifies the use here of dis- observation of the differences between distances restricted to third codon positions, tances does indicate a considerable rate sites at which substitutions are largely decrease between Cercocebus and Mandrillus synonymous. when compared with Macaca,afinding The estimated dates are given in Table 5. made previously by Disotell (1992). There- It is important to bear in mind that since the fore, the estimated time of divergence of a fossil calibration points represent minimum Cercocebus/Mandrillus lineage as well as the dates of divergence, estimated dates based 248 . . 

Table 5 Estimated divergence dates (in millions of years) within the Papionini based on Kimura (1980) two-parameter corrected distances for the mtDNA COII sequences using third codon positions only

Estimated date Estimated date (calibrated on (calibrated on Pongo/Homo Theropithecus Divergence divergence)* Range divergence)† Range

Macaca vs. African papionins 11·88 (12·49)‡ 10·49–14·68 9·27 (9·74) 8·18–11·45 Mandrillus/Cercocebus vs. other papionins 11·48 8·47–14·33 8·95 6·61–11·2 Mandrillus/Cercocebus divergence 4·12 — 3·21 — Theropithecus/Papio vs. Lophocebus 4·76 4·55–4·97 3·71 3·55–3·88 Theropithecus/Papio divergence 5·13 — 4·0§—

*The Pongo/Homo divergence was inferred to be 13·0 m.y.a. based on the earliest occurrence of fossils attributable to Sivapithecus (after Kappelman et al., 1991). †The Theropithecus divergence was inferred to be 4·0 m.y.a. based on the earliest occurrence of fossils attributable to this genus (after Eck & Jablonski, 1984; Szalay & Delson, 1979). ‡Dates in parentheses are estimations in which the distances to Cercocebus and Mandrillus were excluded because the branches leading to these genera show a rate slowdown relative to other papionin branches. §This data is the calibration point within the papionin group. on these calibrations are likely to be 290 ka to 370 ka, and that between Macaca underestimated. Equally important to and the African papionins is estimated to be remember, however, is that these dates are in the range of 320 ka to 400 ka. The short based on the divergences between DNA times estimated between these divergences sequences and not the divergence between (i.e., their internodal branches) are also indi- species. Since gene copies have coalescences cated by the maximum likelihood estimated that predate the species divergences, these branch lengths. In the maximum likelihood dates tend to be overestimates of the species tree for the combined mtDNA dataset (tree divergence, though presumably these differ- not shown), in which all branch lengths ent factors will act to balance each other to are significantly positive as measured by some extent. Because it is difficult to take Felsenstein’s (1993) criteria, these two these variables into account, the dates of internodes are approximately 31% and 33% divergence presented here are only intended of the average branch lengths in the tree, to be rough estimates. Nevertheless, the respectively. dates based on either calibration point are generally consistent with the inferred ages of Discussion papionin lineages for which a good fossil record is known (see Szalay & Delson, Gene trees, species trees, and population 1979). In light of an incomplete fossil record genetics theory these estimated dates provide important As discussed earlier, several factors may information concerning the relative times contribute to discordances among gene trees between species divergences within the and species trees. First, gene trees may be papionins. incorrectly estimated because of random Considering the times between species sampling error (Saitou & Nei, 1986), or divergences, two are found to be relatively because high substitution rates cause homo- short. The time between the successive plasy to bias the tree. Second, natural selec- divergences of Theropithecus, Lophocebus, tion can produce conflicting gene trees. and Papio is estimated to be in the range of Among the eight gene trees for the       249 papionins, only the prion protein gene tree, actual species tree) is the random sorting of which is notably anomalous, is suspected to ancestral DNA sequences into descendant be incorrectly estimated, likely as a result of lineages. For example, considering just three selection pressures on the prion protein. For species, the probability that a gene tree has instance, Van der Kuyl et al. (unpublished the same topology as the species tree is given manuscript) suggest that the relatively low by rates of substitutions at nonsynonymous P=12/3e T/2Ne sites (i.e., amino acid changing sites) compared to the much higher rates at synony- where T is the time between successive ff mous sites (i.e., that cause no amino acid species divergences, and Ne is the e ective change) indicate strong functional con- population size (of gene copies) (Pamilo & straint possibly acting to prevent incorrect Nei, 1988). For any particular gene tree, folding of the PrP protein known to cause two factors will tend to reduce the prob- disease (see Pan et al., 1993). ability that it matches the species tree, when

Error due to high substitution rates is T is small, and/or when Ne is large. This is probably unlikely, since most gene regions because random genetic drift is less efficient showed relatively low diversity (less than at fixing polymorphic DNA sequences 10% variability across papionin species for within a population over relatively short nuclear regions sequenced by Harris & periods of time or when effective population Disotell (1998). Nucleotide variability is sig- sizes are large. Under these circumstances nificantly greater in the COII and 12S rRNA the random sorting of ancestral DNA line- genes (Disotell et al., 1992; Van der Kuyl ages into descendant species lineages can et al., 1995a). However, error is minimized produce a gene tree different from the in the analysis of these data by employing a species tree. Examining the influence solely weighting scheme in which the relatively fast of time, Pamilo & Nei (1988) found changing transitions were downweighted that when the interval between successive compared to the more conservative trans- divergences among species is only several version changes. The error due to random hundred thousand years, or roughly 20,000 sampling which occurs in the analysis of to 30,000 generations (assuming a gener- relatively short regions of DNA is always ation of ten years), the probability of mis- difficult to rule out. Saitou & Nei (1986) match between gene and species tree may be have estimated that relatively longer DNA considerable. sequences (than those usually appearing Under these circumstances, increasing the in phylogenetic analyses) are needed to number of unlinked loci can considerably estimate the correct gene tree with 95% increase the probability of obtaining the probability. While this concern needs to be actual species tree (Saitou & Nei, 1986; balanced in light of the practical limitations Pamilo & Nei, 1988). Congruence in the to collecting large datasets in the laboratory, pattern of species relationships across mul- these same authors (Saitou & Nei, 1986) tiple separate gene trees can greatly increase point out that the same high probability the probability that real phylogenetic associ- can be achieved by sequencing multiple ations are identified (above the level at independent genes and searching for con- which these associations will occur by gruence across their separately estimated chance alone). Using the eight unlinked loci gene trees. analyzed here, this approach has been As discussed earlier, the third phenom- applied to the investigation of the central enon that can cause discordances among questions in papionin systematics outlined gene trees (and between gene trees and the earlier. 250 . . 

Systematic questions area of incongruence among the gene trees Are mangabeys monophyletic? Mangabey concerns the order of divergence of Lopho- monophyly is not supported by any of the cebus, Papio and Theropithecus. For three eight gene trees. The question is, then, to taxa, there are three possible alternative which papionin genera is each mangabey bifurcating rooted trees, [(Lophocebus, genus most closely related? For Cercocebus,a Papio), Theropithecus], [(Lophocebus, Thero- sister group relationship with Mandrillus is pithecus), Papio], and [Lophocebus, (Papio, well supported by five gene trees (CD4, Theropithecus)], all of which have been found TSPY, --globin intergenic region, IRBP, in the gene trees analyzed. The distribution mtDNA) while the remaining gene trees ( of gene trees supporting the alternative pair- 1,3 GT and hemoglobin) are not incon- ings of these taxa can be obtained. TSPY gruent with such a clade. The genus yields an unresolved tree for these taxa and Lophocebus is well supported as closely is excluded. As discussed earlier, prion pro- related to Papio and/or Theropithecus in all of tein is excluded since its gene trees are the gene trees, making the close relationship anomalous. The distribution is: 4 ( 1,3 among these three genera strongly corrob- GT, --globin intergenic region, IRBP, orated. However, resolving the detailed hemoglobin), 1 (CD4), and 1 (mtDNA MP relationships among these taxa is more tree; uniform weighting), respectively. The problematic since there is considerable con- distribution would change slightly to 5, 1, flict among the gene trees for these taxa (see and 0 if the mtDNA MP a priori weighted discussion below). A recent analysis of DNA tree is adopted which supports a Lophocebus/ sequences for two different genes (gamma 1 Papio clade. and 2) within the -globin cluster (chromo- A likelihood ratio test for multiple loci some 11) supports the close relationship developed by Wu (1991) can be used to between Papio and Theropithecus to the evaluate whether such a distribution permits exclusion of Lophocebus (Page et al., 1999). the selection of one particular tree as the actual species tree. In the case of six gene Are Papio and Mandrillus sister taxa? The trees, Wu’s test requires that all of these hypothesized sister group relationship be- trees support the same branching pattern in tween Papio and Mandrillus (in morphological order to reject the null hypothesis, that the studies) is not supported in any of the eight three genera form a trichotomy, or that one gene trees, thus strongly rejecting it. In fact, of the less frequent alternative trees may in at least five gene trees place these genera in fact represent the actual species tree (see two distinct clades. Again, recent analyses of Wu, 1991). For the present distributions, two genes within the -globin cluster (gamma statistical rejection of the null hypothesis is 1 and 2) also do not support a close relation- not possible. The acceptance of any single ship between these genera (Page et al., 1999). pairing among these three taxa as represen- These sequences may in fact be linked with tative of the actual species tree (on the basis the --globin intergenic sequences ana- of Wu’s 1991 test) may require further data lyzed in this paper. Therefore, they may not from unlinked genetic loci. Indeed, Ruvolo’s provide independent evidence concerning the (1997) recent application of the multi-locus relationships of these genera. However, they test towards resolving the relationships do increase the reliability of the gene tree among the three ‘‘African’’ hominoids based on this genetic locus. was able to find statistical support for a Pan–Homo pairing on the basis of a total of How are Lophocebus, Papio, and Thero- 14 unlinked loci. Moreover, Moore (1995) pithecus related? It is clear that the major has estimated that as many as 16 nuclear loci       251 may be required to reach statistical confi- low a ‘‘total’’ evidence approach and adopt dence in a species tree, particularly when the combined tree as the actual phylogenetic species’ divergences have occurred close in relationships of papionin genera. The goal of time. Despite such findings, there may be this approach is to maximize the informa- reason to suggest that a sister taxon relation- tiveness and explanatory power of the data ship between Lophocebus and Papio is the (Kluge, 1989; Ernisse & Kluge, 1993; Kluge favored phylogenetic hypothesis at present & Wolf, 1993). However, in the analysis of (if we insist on a bifurcating tree), since this unlinked genetic regions such an approach relationship is most frequently supported makes little theoretical sense. Unlinked gen- amongst the available gene trees. etic regions can have different biological Nevertheless, the discordances among properties (e.g., in gene function, levels of gene trees for these three genera presumably diversity, base composition, transition/ indicate that their divergences with respect transversion ratios) as well as different to each other took place over a relatively evolutionary properties (e.g., in gene co- short time interval. This is supported by the alescence, biparental versus uniparental relatively short time between their diver- inheritance, variation shaped by natural gences estimated based on the COII selection) which make it unnatural to com- sequences (ranging from 290 ka to 370 ka), bine them in a single dataset (Miyamoto and as well as the relatively short internodal Fitch, 1995). In the present study, the com- branch lengths (estimated by maximum like- bined analysis yielded a single bifurcating lihood) between their divergences. A parallel tree in which Lophocebus is supported as the example of this phenomenon is the case of sister taxon to Papio to the exclusion of the hominoid ‘‘trichotomy’’ where various Theropithecus. However, adopting this tree gene trees support alternative pairings would limit insight into the evolutionary amongst Homo, Pan and Gorilla with the history of the papionins especially the rapid- time between their divergences estimated to fire divergence of Lophocebus, Papio, and be perhaps as short as several hundred Theropithecus gained by examination of the thousand years (Rogers, 1993; Ruvolo, pattern of incongruence among separate 1997), though some estimates are consider- gene trees. ably greater (see Ruvolo, 1997). Given the similar situation for the Implications for morphological evolution within papionin genera, we might ask the following the Papionins. The strong support for questions. Why did Papio, Theropithecus, and mangabey polyphyly found here is in dis- Lophocebus speciate so relatively close in agreement with traditional phylogenies of time around 5 m.y.a. or so? Are there com- the mangabeys based on morphological evi- mon causal factors underlying the rough dence which suggest their monophyletic coincidence in the divergences among the grouping. The overall close morphological papionin genera and the slightly earlier resemblance between the two mangabey divergences among the large ‘‘African’’ genera therefore requires explanation. In hominoids? What were the speciation pro- contrast to the other papionin genera, both cesses involved in the divergence of these mangabey genera share moderately prog- species? nathic faces, deep suborbital maxillary fossae, a medium body size, and arboreal to Separate versus combined analysis of multiple semiterrestrial skeletal adaptations. At least gene regions two interpretations of the evolution of these One approach to resolving the region of features are possible under the present view incongruence in the papionin tree is to fol- of their relationships. Either they evolved 252 . .  in parallel in the two separate mangabey share a deep scapula; a humerus with a very lineages, or they are retained characters broad deltoid plane, a proximally extended from the common ancestor of the African supinator crest, a broad brachialis flange, papionins. For the cluster of characters, they and a narrow olecranon fossa with a deep indeed may represent a mixture of primitive lateral ridge; an ulna with a narrow coronoid and independently evolved characters. process and a relatively large radial notch; a Outgroup comparisons with Macaca radius and ulna with marked interosseous suggest that many of these characters are lines; a robust ilium, reduced gluteal primitive for the African papionins, such as tuberosity on the femur, subequal and sharp moderate prognathism, arboreal to semi- borders of the patellar groove, and a tibia terrestrial skeletal adaptations, and medium with a more rounded midshaft; and in the body size, all of which are variably found in dentition, relatively large and rounded P4s macaque species. However, the fact that in the upper and lower jaws. This suite of Macaca lacks even shallow suborbital fossae characters is interpreted to be functionally (Szalay & Delson, 1979) may suggest that associated with a general ecological special- the fossae are independently evolved in each ization in the Cercocebus/Mandrillus clade for mangabey genus. foraging for insects, hard nuts, and seeds on Despite an overall resemblance between the forest floor (Fleagle & McGraw, 1989), mangabey skulls, Groves (1978) docu- a lifestyle which they suggest may be mented a series of minor morphological dif- primitive for the papionin tribe. If the late ferences in the crania between Lophocebus Miocene to Pliocene genus Parapapio, often and Cercocebus mangabeys in cranial dimen- suggested to be an ‘‘archetypal’’ ancestor of sions, nasal bones, zygomatic arches, sub- either the entire tribe or just the African orbital fossae, tubular auditory meati, papionins (Simons & Delson, 1978; Leakey mandibular morphology, cranial vault & Delson, 1987), is found to display such a sutures, molar wear, central incisor mor- suite of characters this idea would be further phology, number and position of the malar supported. However, the frequent fossil and mental foramina, and the morphology presence of Parapapio at relatively dry of the lachrymal fossae. More recently, savannah or wooded savannah sites in East differences in the postcranial skeletons of Africa such as at Laetolil, Tanzania (Leakey these two genera have been documented & Delson, 1987) and the relatively dry and (Nakatsukasa, 1994, 1996) which are func- open habitats at South African sites such as tionally associated with a basic arboreal/ Sterkfontein, Makapansgat, Bolt’s Farm and semiterrestrial–terrestrial ecological division Taung (Simons & Delson, 1978; Delson, between them. Lophocebus is reported to be 1975) (does not seem to support) this sug- strictly arboreal, preferring the main canopy gestion. It is possible, however, that if the of the forest; whereas Cercocebus is semi- ancestral papionin species was found in terrestrial to terrestrial and frequently more densely forested habitats like those inhabits the forest’s understory and floor inhabited by species of Cercocebus and (Waser, 1984; Jones & Sabater Pi, 1968; Mandrillus today (in western and central Homewood, 1978). Africa) than taphonomic conditions may Fleagle & McGraw (1989) found a not have been favorable for its fossil number of osteological characters which preservation. support the genetic findings of mangabey The strongly supported hypothesis that polyphyly and which link Cercocebus with Papio and Mandrillus belong to two un- Mandrillus in a clade distinct from a Lopho- related lineages within the papionins cebus, Papio, and Theropithecus clade. They calls for re-evaluation of morphological       253 characters traditionally posited as linking lature. Obviously, these explanations are not them as sister taxa. The long faces shared exclusive, and the causal relationship by these genera apparently either evolved between a decrease in body size (and facial independently in these two genera, or are length) can be examined in comparative retained from their common African pap- ontogenetic and morphological studies of ionin ancestor. This first interpretation facial morphology in mangabeys and other would seem to be supported by the general papionin genera. Nevertheless, the hypoth- allometric trend exhibited by the large- esis that a large body size and long face bodied papionins in which disproportionate characterized the last common ancestor of lengthening of the face is correlated with the African papionins might be better sup- increasing body size (see Freedman, 1962; ported if additional mangabey features, Jolly, 1970). Lengthening of the face is common to both genera and functionally associated with large body size in a number uncorrelated with the face (e.g., features of of living and extinct Old World monkeys the postcranium such as size of articular including Macaca nigra, Gorgopithecus, surface areas, and diaphyseal cross sectional Dinopithecus, Paracolobus, and Rhinocolobus areas), were shown to reveal a phyletic (Szalay & Delson, 1979). It is probable that decrease in their overall size. the trend towards disproportionate facial Another finding of the current study is lengthening (and development of large that Theropithecus may not be the sister canine teeth) within this group is also related taxon to Papio, as suggested in immunologi- to a general social system in which there is cal and mtDNA COII studies (Sarich, 1970; strong sexual dimorphism with intense Disotell et al., 1992). This may indicate that intermale competition. the morphological features shared by these However, the alternative interpretation, genera—terrestrially adapted postcrania, that long faces may be primitive for the large body size, and long (although differ- African Papionini, is also tenable. This view ently shaped) faces—may also be indepen- has been suggested by Groves (1978) and dent acquisitions. Of course, as discussed more recently by Kingdon (1997) in light of before, the alternative explanation that they the emerging findings from molecular sys- are primitive for the group is also possible. tematics. According to such an interpret- Clearly, further investigation of the ation, the close resemblance in the long polarity of morphological evolution in faces of Papio and Mandrillus is a result of papionins is needed. Particular focus might their retaining the putative long face of the be on comparing the ontogeny of faces African papionin common ancestor. Thus, within the papionins (e.g. see Delson & the suborbital maxillary fossae possessed by Dean, 1993; Shah & Leigh, 1995), as well as mangabeys are explained as evolving inde- investigating the structural and functional pendently in each genus as a result of attributes of facial fosae, specifically the independent shortening of their faces as they deep suborbital fossae in the mangabeys. A became phyletically smaller from the puta- well supported molecular phylogeny will tively long-faced African papionin ancestor provide a useful template upon which (see Kingdon, 1997). morphological contrasts between genera and Kingdon (1997:42) suggests a structural species can be studied. basis for the development of the fossae as ‘‘the result of buckling in the plane of junc- Acknowledgements tion [between face and cranium]’’ whereas Groves (1978) emphasizes their functional I thank Todd R. Disotell, Terry Harrison, basis in preserving a complex facial muscu- Clifford J. Jolly, Eric Delson, David H. A. 254 . . 

Fitch, Jody Hey, and Rajasekhar Bandaru, Doyle, J. J. (1997). Trees within trees: genes and all of whom gave valuable suggestions on species, molecules and morphology. Syst. Biol. 46, 537–553. various aspects of this research. Eric Delson Dutrillaux, B., Fosse, A.-M. & Chauvier, G. (1979). gave extensive and careful suggestions which Étude cytogénétique de six espèces ou sous-espèces resulted in a much improved manuscript. de mangabeys (Papiinae sic, Cercopithecidae). Ann. Génét. 22, 88–92. Additionally, two anonymous reviewers gave Eck, G. G. & Jablonski, N. G. (1984). A reassessment very useful suggestions on the paper. This of the taxonomic status and phyletic relationships of research was supported by an NSF Doctoral Papio baringensis and Papio quadratirostris (Primates: Cercopithecidae). Am. J. phys. Anthrop. 65, 109– Dissertation Improvement Grant to T. R. 134. Disotell for E.E.H. and an NSF Career Eriksson, T. (1996) AutoDecay Version 2.9.6 (Hyper- Development Grant (SBR 9506892) to card stack distributed by the author). Botaniska Institutionen, Stockholm University, Stockholm. T. R. Disotell. Additional support came Ernisse, D. J. & Kluge, A. J. (1993). Taxonomic from Jody Hey at Rutgers University while congruence versus total evidence, and amniote the author was a postdoctoral fellow in his phylogeny inferred from fossils, molecules, and morphology. Mol. Biol. Evol. 10, 1170–1195. laboratory. Felsenstein, J. (1993) PHYLIP (Phylogenetic Inference Package), Version 3.5c. Department of Genetics, References University of Washington, Seattle. Adkins, R. M., Honeycutt, R. L. & Disotell, T. R. Fleagle, J. G. & McGraw, W. S. (1989). Skeletal and (1996). Evolution of eutherian cytochrome c oxidase dental morphology supports diphyletic origin of subunit II: heterogeneous rates of protein evolution baboons and mandrills. Proc. natn. Acad. Sci. USA and altered interaction with cytochrome c. Mol. Biol. 96, 1157–1161. Evol. 13, 1393–1404. Freedman, L. (1962). Growth of muzzle length Avise, J. C. & Wollenberg, K. (1997). Phylogenetics relative to calvaria length in Papio. Growth 26, and the origin of species. Proc. natn. Acad. Sci. USA 117–128. 94, 7748–7755. Groves, C. P. (1978). Phylogenetic and population Bremer, K. (1994). Branch support and tree stability. systematics of the mangabeys (Primates: Cercopithe- Cladistics 10, 295–304. coidae). Primates 9, 1–34. Cronin, J. E. & Meikle, W. E. (1979). The phyletic Harris, E. E. & Disotell, T. R. (1998). Nuclear gene position of Theropithecus: congruence among molecu- trees and the phylogenetic relationships of the lar, morphological and paleontological evidence. mangabeys (Primates: Papionini). Mol. Biol. Evol. Syst. Zool. 28, 259–269. 15, 892–900. Cronin, J. E. & Meikle, W. E. (1982). Hominid and Hewett-Emmett, D. & Cook, C. N. (1978). Atypical gelada baboon evolution: agreement between evolution of papionin -haemoglobins and indication molecular and fossil time scales. Int. J. Primatol. 3, that Cercocebus may not be a monophyletic genus. In 469–482. (D. J. Chivers & K. A. Joysey, Eds) Recent Advances Cronin, J. E. & Sarich, V. M. (1976). Molecular in Primatology, pp. 291–294. London: Academic evidence for the dual origin of the mangabeys among Press. Old World monkeys. Nature 260, 700–702. Hewett-Emmett, D., Cook, C. N. & Barnicot, N. A. Delson, E. (1975). Paleoecology and zoogeography of (1976). Old World monkey hemoglobins: decipher- the Old World monkeys. In (R. Tuttle, Ed.) Primate ing phylogeny from complex patterns of molecular Functional Morphology and Evolution, pp. 37–64. The evolution. In (M. Goodman & R. E. Tashian, Eds) Hague: Mouton. Molecular Anthropology, pp. 257–275. New York: Delson, E. & Dean, D. (1993). Are Papio baringensis R. Plenum Press. Leakey, 1969, and P. quadratirostris Iwamoto, 1982, Hey, J. (1994). Bridging phylogenetics and population species of Papio or Theropithecus? In (N. G. Jablonski, genetics with gene tree models. In (B. Schierwater, Ed.) Theropithecus: the Rise and Fall of a Primate B. Streit, G. P. Wagner & R. DeSalle, Eds) Molecular Genus, pp. 125–156. Cambridge: Cambridge Univer- Ecology and Evolution: Approaches and Applications, sity Press. pp. 435–449. Switzerland: Birkhauser Verlag Basel. Disotell, T. R. (1992). Molecular evolution of the Hill, W. C. O. (1974). Primates: Comparative Anatomy papionini (Primates: Cercopithecinae). Ph.D. Dis- and Taxonomy. Volume VII: Cynopithecinae. sertation. Harvard University. Edinburgh: The University Press. Disotell, T. R. (1994). Generic level relationships of Hoelzer, G. A. (1997). Inferring phylogenies from the Papionini (Cercopithecoidea). Am. J. phys. mtDNA variation: mitochondrial-gene trees versus Anthrop. 94, 47–57. nuclear gene-trees revisited. Evolution 51, 622–626. Disotell, T. R., Honeycutt, R. L. & Ruvolo, M. (1992). Homewood, K. M. (1978). Feeding strategy of the Mitochondrial DNA phylogeny of the Old-World Tana mangabey (Cercocebus galeritus galeritus) monkey tribe Papionini. Molec. Biol. Evol. 9, 1–13. (Mammalia: Primates). J. Zool., Lond. 186, 375–391.       255

Hudson, R. R. (1983). Testing the constant-rate Moore, W. S. (1997). Mitochondrial gene-trees versus neutral allele model with protein sequence data. nuclear-gene trees: a reply to Hoelzer. Evolution 51, Evolution 37, 203–217. 627–629. Jolly, C. J. (1967). The evolution of baboons. In (H. Morales, J. C. & Melnick, D. J. (1998). Phylogenetic Vagtborg, Ed.) The Baboon in Medical Research II, relationships of the macaques (Cercopithecidae: pp. 23–50. Austin: University of Texas Press. Macaca), as revealed by high resolution restriction Jolly, C. J. (1970). The large African monkeys as an site mapping of mitochondrial ribosomal genes. adaptive array. In (J. R. Napier & P. H. Napier, Eds) J. hum. Evol. 34, 1–23. Old World Monkeys, pp. 227–262. New York: Nakatsukasa, M. (1994). Morphology of the humerus Academic Press. and femur in African mangabeys and guenons: func- Kappelmann, J., Kelly, J., Pilbeam, D., Sheikh, K. A., tional adaptation and implications for the evolution Ward, S., Anwar, M., Barry, J. C., Brown, B., Hake, of positional behavior. African Study Monographs, P., Johnson, N. M., Raza, S. M. & Shah, S. M. I. Suppl. 21, 1–61. (1991). The earliest occurrence of Sivapithecus from Nakatsukasa, M. (1996). Locomotor diﬀerentiation the middle Miocene Chinji Formation of Pakistan. and diﬀerent skeletal morphologies in mangabeys J. hum. Evol. 21, 61–73. (Lophocebus and Cercocebus). Folia Primatol. 66, Kim, H.-S. & Takenaka, O. (1996). A comparison of 15–24. TSPY genes from Y-chromosomal DNA of the Nei, M. (1987). Molecular Evolutionary Genetics.New great apes and humans: sequence, evolution, and York: Columbia University Press. phylogeny. Am. J. phys. Anthrop. 100, 301–309. Page, S. L., Chiu, C. & Goodman, M. (1999). Kimura, M. (1980). A simple method for estimating Molecular phylogeny of Old World monkeys (Cerco- evolutionary rate of base substitutions through com- pithecidae) as inferred from -Globin DNA parative studies of nucleotide sequences. J. Molec. Sequences. Mol. Phyl. Evol. 13, 348–359. Evol. 16, 111–120. Pamilo, P. & Nei, M. (1988). Relationships between Kingdon, J. (1997). The Kingdon Field Guide to African trees and species trees. Molec. Biol. Evol. 5, 568–583. Mammals. San Diego: Academic Press. Pan, K.-M., Baldwin, M., Nguyen, J., Gasset, M., Kishino, H. & Hasegawa, M. (1989). Evaluation of Serban, A., Groth, D., Mehlhorn, I., Huang, Z., the maximum likelihood estimate of the evolutionary Fletterick, R. J., Cohen, F. E. & Prusiner, S. P. tree topologies from DNA sequence data, and (1993). Conversion of -helices into -sheets features branching order in Hominoidea. J. Mol. Evol. 29, in the formation of the scrapie prion proteins. Proc. 170–179. natn. Acad. Sci. USA 90, 10,962–10,966. Kluge, A. G. (1989). A concern for evidence and a Rogers, J. (1993). The phylogenetic relationships phylogenetic hypothesis of relationships among among Homo, Pan,andGorilla: a population Epicrates (Boidae, Serpentes). Syst. Zool. 38, 7–25. genetics perspective. J. hum. Evol. 25, 201–215. Kluge, A. G. & Wolf, A. J. (1993). Cladistics: what’sin Ruvolo, M. (1997). Molecular phylogeny of the a word? Cladistics 9, 183–199. hominoids: inferences from multiple independent Krakauer, D. C., Pagel, M., Southwood, T. R. E. & DNA sequence data sets. Mol. Biol. Evol. 14, 248– Zanotto, P. M. de A. (1996). Phylogenesis of prion 265. protein. Nature 380, 675. Jones, C. & Sabater Pi, J. (1968). Comparative ecology Kuhn, H.-J. (1967). Zur systematik der Cerco- of Cercocebus albigena (Gray) and Cercocebus torquatus pithecidae. In (D. Starck, R. Schneider & H.-J. (Kerr) in Rio Muni, West Africa. Folia primatol. 9, Kuhn, Eds) Neue Ergebnisse se Primatologie, 99–113. pp. 25–46. Stuttgart: G. Fischer. Saitou, N. & Nei, M. (1986). The number of nucleo- Kumar, S., Tamura, K. & Nei, M. (1993) MEGA: tides required to determine the branching order of Molecular evolutionary genetics analysis, version three species, with special reference to the human– 1.01. The Pennsylvania State University, University chimpanzee–gorilla divergence. J. Mol. Evol. 24, Park, PA 16802. 189–204. Leakey, M. G. & Delson, E. (1987). Fossil Cerco- Sarich, V. M. (1970). Primate systematics with special pithecidae from the Laetolil beds. In (M. D. Leakey reference to Old World Monkeys: a protein per- & J. M. Harris, Eds) Laetoli: a Pliocene site in North- spective. In (J. R. Napier & P. H. Napier, Eds) Old ern Tanzania, pp. 91–106. Oxford: Clarendon World Monkeys, pp. 175–226. New York: Academic Press. Press. Maddison, W. P. (1997). Gene trees in species trees. Schätzl, H. M., Da Costa, M., Taylor, L., Cohen, F. E. Syst. Biol. 46, 523–536. & Prusiner, S. B. (1995). Prion protein gene vari- Mickevich, M. F. (1978). Taxonomic congruence. ation among primates. J. Mol. Biol. 245, 362–374. Syst. Zool. 27, 143–158. Shah, N. & Leigh, S. (1995). Cranial ontogeny in three Miyamoto, M. M. & Fitch, W. M. (1995). Testing papionin genera. Am. J. phys. Anthrop. Suppl. 20, species phylogenies and phylogenetic methods with 404 (abstract). congruence. Syst. Biol. 44, 64–76. Simons, E. L. & Delson, E. (1978). Cercopithecidae Moore, W. S. (1995). Inferring phylogenies from and Parapithecidae. In (V. J. Maglio & H. B. S. mtDNA variation: mitochondrial-gene trees versus Cooke, Eds) Evolution of African Mammals, pp. 100– nuclear gene-trees. Evolution 49, 718–726. 119. Cambridge: Harvard University Press. 256 . . 

Strasser, E. & Delson, E. (1987). Cladistic analysis of Nuclear counterparts of the cytoplasmic mito- cercopithecid relationships. J. hum. Evol. 16, 81–99. chondrial 12S rRNA gene: a problem of ancient Swofford, D. L. (1993) PAUP: Phylogenetic Analysis DNA and molecular phylogeneies. J. Mol. Evol. 40, Using Parsimony, Version 3.1.1. Program and Docu- 652–657. mentation. Champaign, ILL, Center for Biodiversity, Waser, P. M. (1984). Ecological differences and behav- Illinois Natural History Survey. ioral contrasts between two mangabey species. In Szalay, F. & Delson, E. (1979). Evolutionary History of (P. S. Rodman & J. G. H. Cant, Eds) Adaptations for the Primates. New York: Academic Press. Foraging in Nonhuman Primates, pp. 195–214. New Van der Kuyl, A. C., Dekker, J. T. & Goudsmit, J. York: Columbia University Press. (unpublished manuscript). Evidence for an increased Weinberg, J., Stanyon, R., Jauch, A. & Cremer, T. rate of the hominoid prion gene during the period of (1992). Homologies in human and Macaca brain expansion. fuscata chromosomes revealed by in situ sup- Van der Kuyl, A. C., Kuiken, C. L., Dekker, J. T. & pression hybridization with human chromosome Goudsmit, J. (1995a). Phylogeny of African monkeys specific DNA libraries. Chromosoma 10, 265– based upon mitochondrial 12S rRNA sequences. 270. J. molec. Evol. 40, 173–180. Wu, C.-I. (1991). Inferences of species phylogeny in Van der Kuyl, A. C., Kuiken, C. L., Dekker, J. T., relation to segregation of ancient polymorphisms. Perizonius, W. R. K. & Goudsmit, J. (1996b). Genetics 127, 429–435.