Lower Ilium Evolution in Apes and Hominins

Transcription

1 THE ANATOMICAL RECORD 300: (2017) Lower Ilium Evolution in Apes and Hominins ASHLEY S. HAMMOND 1 * AND SERGIO ALM ECIJA 1,2 1 Center for Advanced Study of Human Paleobiology, Department of Anthropology, George Washington University, Washington, District of Columbia Institut Catala de Paleontologia Miquel Crusafont, Universitat Autonoma de Barcelona, Barcelona, Spain ABSTRACT Elucidating the pelvic morphology of the Pan-Homo last common ancestor (LCA) is crucial for understanding ape and human evolution. The pelvis of Ardipithecus ramidus has been the basis of controversial interpretations of the LCA pelvis. In particular, it was proposed that the lower ilium became elongate independently in the orangutan and chimpanzee clades, making these taxa poor analogues for the pelvis of the LCA. This study examines the variation in relative lower ilium height between and within living and fossil hominoid species (and other anthropoids), and models its evolution using available fossil hominoids as calibration points. We find nuanced differences in relative lower ilium height among living hominoids, particularly in regards to gorillas, which do not have elongate lower ilia (because they are likely to represent the plesiomorphic hominoid condition for this trait). We also show that differences in relative lower ilium height among hominoid taxa are not readily explained by differences in size between species. Our maximum likelihood ancestral state reconstructions support inferences that chimpanzees (Pan troglodytes in particular) and orangutans evolved their elongate lower ilia independently. We also find that the predicted lower ilium height of the Pan-Homo LCA is shorter than all great apes except gorillas. This study adds to a growing body of evidence that finds different regions of the body show different evolutionary histories in different hominoids, and underscores that the unique combinations of morphologies of each modern and fossil hominoid species should be considered when reconstructing the mosaic nature of the Pan-Homo LCA. Anat Rec, 300: , VC 2017 Wiley Periodicals, Inc. Key words: pelvis evolution; ancestral state reconstruction; last common ancestor (LCA); Ardipithecus; Sivapithecus Additional Supporting Information may be found in the online version of this article. Grant sponsor: National Science Foundation; Grant number: BCS ASH; Grant sponsor: Spanish Ministerio de Economıa y Competividad ; Grant number: CGL P; Grant sponsor: CERCA Programme (Generalitat de Catalunya); Grant sponsor: Wenner-Gren Foundation; Grant sponsor: LSB Leakey Foundation. *Correspondence to: Ashley S. Hammond, Center for Advanced Study of Human Paleobiology, Department of Anthropology, George Washington University, Science and Engineering Hall, nd Street NW, Suite 6000, Washington, DC ashleyshammond@gwu.edu Received 2 April 2016; Accepted 9 October DOI /ar Published online in Wiley Online Library (wileyonlinelibrary. com). VC 2017 WILEY PERIODICALS, INC.

2 LOWER ILIUM EVOLUTION IN APES AND HOMININS 829 Elucidating the pelvic morphology of the Pan-Homo last common ancestor 1 (LCA) is crucial for understanding ape and human evolution. Not only does the hominoid pelvis display major transformations in morphology with adaptive shifts, it confers information about locomotor behavior (e.g., Waterman, 1929; Schultz, 1950; Robinson, 1972; Ward, 1991; Lewton, 2010, 2015), trunk morphology (e.g., Ruff, 1991; Simpson et al., 2008; Ruff, 2010; Betti et al., 2014), and obstetrical and biomechanical requirements (e.g., Ruff, 1995; Rosenberg and Trevathan, 2002; Brown, 2011; Grabowski et al., 2011). Presumed hominin (i.e., the human clade) synapomorphies of the ilium (i.e., the upper pelvis) have become central to debates about hominin origins (Lovejoy et al., 2009b,c; Wood and Harrison, 2011; White et al., 2015) and provide compelling evidence for facultative bipedalism in Ardipithecus ramidus (ARA-VP-6/500). The short, sagittally-directed, and laterally-flaring ilia of Australopithecus and Homo are widely thought to relate to weight-balance and gluteal abductor mechanisms that facilitate habitual bipedalism (e.g., Lovejoy, 2005). Ardipithecus shares these features, as well as a projecting anterior inferior iliac spine (AIIS) and distinct sciatic notch on the ilium (Lovejoy et al., 2009c), with later hominins. In comparison, the ilium of extant great apes is almost universally described as broad and long, and it is thought to reflect selection for great ape forelimbdominated climbing behaviors by promoting lumbar rigidity (e.g., via entrapment ) and providing a larger area for origin of key muscles (e.g., latissimus dorsi) (Ward, 1991; Lovejoy, 2005). Pan has been the most widely used referential model for the behavior and morphology of the Pan-Homo LCA in recent history (e.g., Zihlman et al., 1978; Pilbeam, 1996; Wrangham and Pilbeam, 2001; see also recent review in Sayers et al., 2012). This model implies that the LCA engaged in forelimb-dominated behaviors and knuckle walking like a chimpanzee or bonobo. As a result, it is generally assumed that the pelvis of the Pan-Homo LCA looked like that of modern panins and that the hominin pelvis has been entirely re-configured for bipedal locomotion. More recently, the reconstructed pelvis of Ardipithecus ramidus (ARA-VP-6/500) as well as other regions of the skeleton was used to support the counter argument that humans did not evolve from a chimpanzee-like ancestor (Lovejoy et al., 2009b). The pelvic morphology of the Pan-Homo LCA is therefore central to our understanding of both hominin and panin origins (by defining the starting point ). More broadly, inferring ancestral state morphologies will also help establish the evolutionary history of phenotypic evolution in extant great apes. Based on the Ardipithecus ramidus pelvis (ARA-VP-6/ 500), Lovejoy et al. (2009b,c) put forward a model of pelvic evolution with specific predictions regarding the timing of the emergence of certain pelvic morphologies in apes and hominins (Fig. 1). Lovejoy and colleagues (see also Machnicki et al., 2016) hypothesized that the Pan, Gorilla, and Pongo lineages all independently witnessed an increase in cranial iliac height and reduction in sacral alar breadth. Lower ilium height, traditionally defined as the length between the sacroiliac and acetabular joints, received particular attention. Lovejoy et al. (2009b, 2009c) hypothesized that the lower ilium, particularly the iliac isthmus (the mediolaterally narrowest part of the ilium), became elongate independently in the Pongo and Pan clades. In the hominin clade, on the other hand, the lower ilium was hypothesized to have become shorter beginning in the late Miocene (Lovejoy et al., 2009b,c), resulting in the hip and sacroiliac joints being relatively closer together in Plio-Pleistocene hominins. The reduced distance between these joints in hominins should decrease rotational torques between the sacroiliac (from weight of the torso) and hip joints (loads passing upwards from hindlimb loading) (Steudel, 1984), and should therefore be a major adaptation to bipedality in the bony pelvis. The Ardipithecus ramidus pelvic morphology was described as intermediate between Miocene ape Ekembo nyanzae 2 and bipedally-adapted Australopithecus in certain regards (Lovejoy et al., 2009c), which has implications for the pelvic morphology of the Pan-Homo LCA. Ekembo nyanzae (17.9 Myr) is a stem hominoid (but see Harrison, 2010) whose pelvic morphology is known from a partial skeleton (KNM-MW 13142) from Mfangano Island, Kenya. The Ekembo nyanzae pelvis (KNM-MW 13142) consists of a nearly complete hipbone 3 and partial sacrum, which, until recently, has provided the basis for most of what we know about Miocene hominoid pelvic morphology. The Ekembo nyanzae acetabulum has an expansive acetabular fossa and is shallow like Asian apes and atelines (Ward et al., 1993), but most other aspects of the hipbone especially the ilium are more similar to cercopithecoids or platyrrhines (Ward, 1991, 1993; Ward et al., 1993). The maximum and minimum estimates of lower ilium height (LIH; the distance between the hip and sacroiliac joints) for Ardipithecus ramidus (ARA-VP-6/500) are similar to Ekembo nyanzae (KNM-MW 13142) (Lovejoy et al., 2009c). Thus, Lovejoy et al. (2009c) inferred that the lower ilium height in Ardipithecus ramidus reflects the maintenance or slight shortening of the lower ilium height from the last common ancestor of African apes and hominins (Lovejoy s et al. GLCA ). Late Miocene fossil great apes the most crucial pieces of evidence to test hypotheses about the Pan- Homo LCA are almost entirely unknown during the time period when hominins first appeared (5-7 Ma). Pelvic material is known from the insular hominoid Oreopithecus bambolii (6-7 Ma), although efforts to integrate Oreopithecus into evolutionary scenarios are impeded by the uncertain phylogenetic position of this taxon (see historical review in Delson, 1986). The pelvic material for Oreopithecus includes a pubic symphysis 1 We do not follow the same naming conventions as Lovejoy et al. (2009a,b135) in this paper. In this paper, hominid refers to family Hominidae (i.e., the great ape and human clade) and hominin specifically refers to the human clade. We also use Pan-Homo LCA for the panin-hominin LCA, which is the CLCA abbreviation used by Lovejoy and colleagues. 2 Historically referred to as Proconsul nyanzae, this species was recently placed within new genus Ekembo by McNulty et al. (2015). 3 We use hipbone, rather than innominate bone, here because os coxae and hipbone are the anatomical terms preferred by the Terminologia Anatomica (e.g., see Tuttle 1988).

3 830 HAMMOND AND ALM ECIJA Fig. 1. Predictions about pelvic evolution in the Hominidae made by Lovejoy et al. (2009b,c). Figures of hipbones are scaled to the same approximate acetabular diameter. The Ardipithecus (ARA-VP-6/500) figure is modified from Lovejoy et al. (2009c) and reprinted with permission from AAAS. (BA71) and a nearly complete but very distorted pelvis (IGF11778 skeleton). Interpretations of the Oreopithecus postcranial morphology are controversial, with most interpretations oscillating between either suspensory (Jungers, 1987; Wunderlich et al., 1999) or bipedal adaptations (K ohler and Moya-Sola, 1997; Rook et al., 1999; Moya-Sola et al., 2008). Differing conclusions have almost certainly been exacerbated by the taphonomic deformation of the IGF11778 pelvis (as well as most of the associated skeleton), which is compressed into an almost two-dimensional structure (e.g., see White et al., 2015). Although a number of authors have argued that the Oreopithecus lower ilium is indeed short like hominins (Harrison, 1991; K ohler and Moya-Sola, 1997; Sarmiento, 2010), iliac length in Oreopithecus is particularly difficult to ascertain because of both the crushing and the displacement of at least one lumbar vertebrae (White et al., 2015). In recent years, additional fossil great ape pelvic remains have been reported (Ward et al., 2008; Ward et al., 2010; Morgan et al., 2011; Hammond et al., 2013; Morgan et al., 2015), which may help clarify the polarity of pelvic changes in the hominid clade. One such recently published hipbone is that of Sivapithecus indicus (12.3 Myr), a nearly complete ilium and ischium which most closely resembles early fossil hominoid Ekembo or platyrrhine Lagothrix (Morgan et al., 2015). The hipbone is described as having cercopithecoid-like iliac sacral surface and tuberosity widths, a robust lower ilium, a deep and steep-sided acetabulum with a craniallyexpanded lunate surface, and a small ischial tuberosity (Morgan et al., 2015). Importantly, the Sivapithecus facial morphology suggests a close phylogenetic relationship with Pongo, yet the postcrania looks more like Ekembo or an arboreal monkey (see discussions of the Sivapithecus Dilemma in Larson, 1998; Pilbeam and Young, 2001). Regardless of the exact phylogenetic placement of Sivapithecus, this fossil provides a much needed calibration point that makes it possible to more reliably make inferences about the expected pelvic morphology of the great ape-human and/or Pan-Homo LCAs. In this study we consider alternative phylogenetic hypotheses of this taxon to test how they affect our evolutionary results (see Phylogenetic Tree Construction section below). Fragmentary pelvic remains for great ape Pierolapithecus catalaunicus (11.9 Myr) have also been recently described (Hammond et al., 2013). Although much more limited inferences can be made from this partial ilium and ischium, Pierolapithecus is intermediate between primitive hominoid Ekembo nyanzae (KNM-MW 13142) and the extant apes in the projected width of the ilium.

4 LOWER ILIUM EVOLUTION IN APES AND HOMININS 831 However, most other aspects of the fragmentary ilium (e.g., gluteal surface concavity, sacroiliac joint shape, linea arcuata morphology on the lower ilium) are more similar to Ekembo than any other taxa (Hammond et al., 2013). Although lower ilium height cannot be directly inferred from the Pierolapithecus ilium fragment, these noted similarities between the Pierolapithecus and Ekembo ilia hints at the possibility that the Pierolapithecus lower ilium may have only been moderately long like in Ekembo. Throughout the trunk, Pierolapithecus displays features associated with orthograde behaviors (e.g., vertebral transverse process originating on the pedicle, highly curved rib bodies, phalangeal curvature), although it lacks adaptations in the phalanges for suspensory behaviors sensu stricto (Moya-Sola et al., 2004; Almecija et al., 2009; Susanna et al., 2014; but see Begun and Ward, 2005). Craniodentally, Pierolapithecus aligns with all extant great apes in basic structure, although preserving a primitive hominoid sagittal profile (Moya-Sola et al., 2004). Other authors even claim that it aligns with living African apes (review in Begun et al., 2012). Irrespective of their precise phylogenetic affinities, when considered together, the Sivapithecus and Pierolapithecus materials provide evidence that some primitive pelvic morphologies (e.g., wide iliac tuberosity, gluteal surface concavity, sacroiliac joint shape, linea arcuata morphology on the lower ilium) persisted until at least 12.3 Myr and 11.9 Myr in some members of the great ape and human crown group (see also Middleton et al., (this issue)). The objectives of this study are: First, to inspect patterns of variation/covariation in lower ilium height (relative to acetabular size) in a large anthropoid sample with particular attention paid to great apes and humans. Second, to generate ancestral state reconstructions of the relative lower ilium height for the Pan- Homo LCA, as well as the hominid and hominoid LCAs. These ancestral state reconstructions are then compared to the timing and overall conclusions about lower ilium length evolution in the Hominidae made by Lovejoy et al. (2009b,c). MATERIALS AND METHODS Sample This study sampled humans (n 5 58), non-human hominids (n 5 112), and hylobatids (n 5 61). In addition to hominoids, our sample incorporated a selection of relevant anthropoid outgroups representing different locomotor repertoires (see Table 1): cercopithecoids (n 5 95), and platyrrhine monkeys (n 5 33). Because we focus on estimating the lower ilium height in the Pan-Homo LCA (see below), and apes and humans are more influential in estimating the LCA values (i.e., because they are more closely related to the LCA than monkeys), a larger number of humans and apes were sampled than for anthropoid monkeys. This expansive sample of apes and humans allowed us to better understand the pattern of the inter/intra-specific variation in lower ilium height and more accurately represent hominoid species averages. Extant anthropoids were measured from material housed at the American Museum of Natural History (AMNH), Cleveland Museum of Natural History (CMNH), United States National Museum (USNM), Naturalis Biodiversity Center (NBD), Field Museum of Natural History (FMNH), Bavarian State Zoology Collections (ZSM), Royal Museum for Central Africa (RMCA), and Harvard s Museum of Comparative Zoology (MCZ). The human sample was selected in order to account for sexual dimorphism and ecogeographic variation in the pelvis. Human populations that were sampled include American blacks and whites (Terry Collection from the USNM), I~nupiak arctic peoples (AMNH), Zuni Pueblo Native Americans (USNM), small-bodied Andaman Islanders (AMNH), and narrow-bodied Americans from the Hamann-Todd Collection (CMNH). The narrowbodied individuals from the Hamann-Todd Collection are those with a pelvic bispinous breadth 185mm, and all of these individuals were black Americans. The fossil sample includes all undistorted, relatively complete hipbones that were available for study. Many published fossil pelves were necessarily excluded from this study because they are too fragmentary or too deformed 4. The fossils sampled here include KNM-MW (Ekembo nyanzae), YGSP (Sivapithecus indicus), KNM-RU 7694 (small Miocene hominoid of uncertain affinity, c.f. Limnopithecus legetet), A.L (Australopithecus afarensis), SK 3155 (Pleistocene hominin), Sts 14 (Australopithecus africanus), KNM-ER 3228 (Pleistocene hominin), OH 28 (Pleistocene hominin), and Kebara 2 (Homo neanderthalensis). The fossils that we consider the most influential for our evolutionary analyses are the fossil apes (i.e., KNM-MW 13142, YGSP 41216) and early hominins (i.e., A.L ), which help calibrate the ancestral state reconstruction for the Pan- Homo LCA. Lower ilium height was measured on the original fossils for A.L , KNM-ER 3228, and KNM- MW 13142, and all other fossil data were collected from high quality casts. Relative Lower Ilium Height Here we measure lower ilium height from the center of the acetabular fossa to the point where the linea arcuata meets the auricular surface. Lower ilium height has traditionally been measured either from the point of iliopubic junction to the auricular surface (as in Straus, 1929), from the center of the center of the acetabulum to the most caudal auricular surface (as in Ward, 1991), and from the posterior inferior iliac spine to the acetabular margin just below the anterior inferior iliac spine (as in Steudel, 1984). Although our measure differs slightly from past workers, we found that we were able to most consistently identify the junction of the linea arcuata and the auricular surface across taxa. A single study has found the sacroiliac joint to be positioned more superiorly in male common chimpanzees (Li, 2002), but this is the only evidence that there might be biases in measuring the lower ilium height in this manner. All measurements were taken by a single observer (ASH). Lower ilium height was inspected in relation to the superoinferior acetabulum diameter (i.e., lower ilium height/acetabular diameter ratio). Analyzing ratio data 4 Incomplete specimens necessarily excluded from this study include the Malapa hominins, Dinaledi Chamber hominins, Pierolapithecus catalaunicus, and KNM-ER 5881 (Ward et al., 2015a). Oreopithecus bambolii was considered too deformed for inclusion in this study.

5 832 HAMMOND AND ALM ECIJA TABLE 1. Extant sample composition and provenience Population Number (M/F) Institution Hylobates agilis 4/2 USNM Hylobates klossii 4/1 USNM, AMNH Hylobates moloch 4/3 ZSM, USNM Hylobates muelleri 2/3 MCZ, USNM Hylobates lar 11/5 MCZ, USNM, ZSM, CMNH Symphalangus 15/7 ZSM, USNM, NBD syndactylus Pongo pygmaeus 5/9 USNM, MCZ, ZMA, CMNH Pongo abelii 6/12 USNM, MCZ, CMNH Gorilla gorilla 15/16 USNM, CMNH Gorilla beringei 4/4 USNM Pan troglodytes 15/16 CMNH, USNM, CMNH Pan paniscus 4/6 RMCA Homo sapiens 5/5 USNM (TC) (Black Americans) Homo sapiens 5/5 USNM (TC) (White Americans) Homo sapiens 5/5 CMNH (HT) (narrow-bodied Black Americans) Homo sapiens 9/4 AMNH (I~nupiak arctic natives) Homo sapiens (small-bodied Andaman Islanders) 2/3 AMNH Homo sapiens 5/5 USNM (Native American Zuni) Cercopithecus mitis 3/3 USNM Theropithecus gelada 2/2 USNM Papio anubis 1/1 USNM Papio hamadryas 6/3 USNM, AMNH Papio cynocephalus 1/2 USNM, CMNH Macaca fascicularis 6/2 USNM, MCZ Colobus guereza 10/8 USNM Nasalis larvatus 8/6 MCZ, ZSM Pygathrix nemaeus 5/2 USNM, AMNH, FMNH Rhinopithecus roxellana 4/5 USNM Trachypithecus cristatus 10/5 MCZ, USNM, ZSM, NBD Alouatta palliata 4/1 USNM Alouatta caraya 3/3 AMNH Ateles fusciceps 3/2 USNM, FMNH Cebus apella 4/5 USNM, FMNH, NBD Saimiri sciureus 4/4 USNM Museum codes: American Museum of Natural History (AMNH), Cleveland Museum of Natural History (CMNH), Field Museum of Natural History (FMNH), Harvard s Museum of Comparative Zoology (MCZ), Royal Museum for Central Africa (RMCA), Naturalis Biodiversity Center (NBD), United States National Museum (USNM), Bavarian State Zoology Collections (ZSM). TC5 Terry Collection, HT5 Hamann-Todd. Gorilla beringei is the mountain gorilla, and does not include subspecies Gorilla beringei graueri. produces specimen-specific dimensionless shape data (Jungers et al., 1995), alleviating some of the issues with absolute size differences among taxa. The dimensions of the hip joint are widely used in size adjustment in anthropoids (e.g., McHenry, 1972; Ruff, 1988; Swartz, 1989; Jungers, 1991; Ruff, 1991; McHenry, 1992; Rafferty and Ruff, 1994; Auerbach and Ruff, 2004; Ruff, 2010; Plavcan et al., 2014) because there is a very close relationship between the hip joint size and body mass. The ranges of variation for all extant taxa were examined via boxplot alongside the individual fossil data, using the ggplot2 functions (Wickham, 2009) in R (R Development Core Team, 2012). Of species sampled here, only hylobatids, Ateles, and A.L are known to have slightly large acetabulae for their femoral head size (Ward, 1991; Plavcan et al., 2014) and/or body size (Schultz, 1969a,). Furthermore, the relative lower ilium height values for all hominid species and a subset of hylobatids (Symphalangus syndactylus and Hylobates lar, selected for their larger sample sizes) were statistically compared with a one-way analysis of variance (ANOVA) and post hoc Tukey s Honest Significance Difference (HSD) using built-in functions in R. The Tukey HSD test controls for Type I error across multiple comparisons. Significance among taxa was assessed at P 0.05 in all statistical tests. To ensure that there were not biases in the human data due to population or sexual dimorphism of the bony pelvis, relative lower ilium height was also examined by sex and population via boxplots and ANOVA. Scaling of the Lower Ilium Height To test if shape differences in relative lower ilium height in our modern hominoid sample could be caused by undetected size-related (i.e., allometric) shape changes, we enlisted species-specific pairwise comparisons of ordinary least squares regressions (OLS; via logtransformed variables using natural logarithms). We first tested whether all species share a same allometric direction (i.e., homogeneity of slopes) via analysis of the covariance (ANCOVA) of the two log-transformed variables: lower ilium height as dependent, acetabulum diameter as independent, and using each species as the interaction factor in the linear model. Subsequently, the scaled-adjusted means (i.e., regression intercepts) of each species for which homogeneity of slopes could not be discarded (i.e., taxa sharing an equivalent allometric direction) were compared via ANOVA and Tukey s HSD post-hoc multiple comparisons to detect grade shifts (i.e., using a multiple regression approach were the factor species was included as the second independent variable). These analyses were performed using the R package multcomp (Hothorn et al., 2008). Ancestral State Reconstructions and Evolution of the Lower Ilium Height Finally, we used ancestral state reconstruction to estimate the relative lower ilium height (lower ilium height/ acetabulum diameter) for all nodes in the hominoid tree and reconstructed the evolution of this trait along the anthropoid phylogenetic tree. This was accomplished by reconstructing the position of the internal nodes (i.e., ancestral states) using a maximum likelihood (ML) method for continuous characters (Felsenstein, 1988). For an evolutionary model based on normally distributed Brownian motion (Cavalli-Sforza and Edwards, 1967; Felsenstein, 1973, 1985), the ML approach yields identical ancestral state estimates to the squared-change

6 LOWER ILIUM EVOLUTION IN APES AND HOMININS 833 parsimony method accounting for branch length, which minimizes the total amount of shape change along all the branches of the tree (Maddison, 1991; Rohlf, 2002). Next, all edges along the tree are fractionated, and state estimates are computed at the midpoint of each fraction via interpolation using equation [3] of Felsenstein (1985). This approach creates the visual appearance of continuous color change along the edges of the tree. This method of visualizing trait evolution in a tree is explained in detail elsewhere (Revell, 2013). Ancestral state visualizations were computed using the R package phytools (Revell, 2012). 95% confidence intervals (95% CIs) for each hominoid ancestral node were computed using the fastanc function implemented in phytools, and are based on equation [6] of Rohlf (2001), that computes the variance on the ancestral states estimates. Once these variances are known, 95% CIs on the estimates can be computed as the estimates the square root of the variances. Phylogenetic Tree Construction The time-scaled phylogeny used in this study is based on a consensus chronometric tree of extant anthropoid taxa and Neandertals downloaded from 10kTrees Website (ver. 3; The 10kTrees Website provides phylogenies sampled from a Bayesian phylogenetic analysis of eleven mitochondrial and six autosomal genes available in GenBank (Arnold et al., 2010). For the case of the Pan-Homo LCA, the consensus tree produced an estimated age of Myr before the present. Branches for dated fossils (excluding Neandertals) were introduced to the consensus tree post hoc, as in other studies (e.g., Almecija et al., 2013, 2015). The published chronological age of the fossil is the tip date in the tree. The lineage represents the length of time between the expected divergence date and the tip date. Although the lineage lengths were selected on a case-by-case basis (see below), we standardized by 1 Myr lineage lengths when possible. Our trees and species mean data used in the ancestral state reconstructions are available in the supplemental data files. Australopithecus afarensis (A.L ) was positioned within the hominin clade in the consensus tree because it shares unequivocal synapomorphies with hominins. The tip date assigned to Australopithecus afarensis was 3.2 Myr, following the established Hadar geochronology (Walter, 1994). The divergence time of Australopithecus afarensis is unknown, and so the lineage length was estimated to be 1 Myr. Adding in ghost lineages to trees is a common practice in ancestral state reconstructions because it eliminates the spontaneous appearance of features and taxa on the tree (Steiper and Seiffert, 2012). The resulting divergence time (4.2 Myr) for Australopithecus afarensis coincides with the earliest known Australopithecus anamensis (White et al., 2006), which is hypothesized to be ancestral to Australopithecus afarensis (Kimbel et al., 2006). Ekembo nyanzae (KNM-MW 13142) was positioned as a stem hominoid. Ekembo (including Proconsul sensu stricto) is almost universally found to be a stem hominoid (Begun et al., 1997a,b; Ward et al., 1997; Alba et al., 2015; McNulty et al., 2015), although others interpret early Miocene taxa be either stem catarrhines or sister taxa to hominoids (Rossie et al., 2002; Harrison, 2010). The Hiwegi Formation from which the KNM-MW skeleton derives is dated to 17.9 Myr (Drake et al., 1988), and so this was selected as the tip date. The crown hominoid node was estimated to be at 19.6 Myr. Thus, Ekembo nyanzae was assigned a divergence date of 20.6 Ma, a divergence date that precedes the crown group node by 1 Myr. This divergence date is consistent with the timing of the earliest known Proconsul (sensu stricto), which is dated to at least 20 Myr (Harrison and Andrews, 2009). Two alternate phylogenetic trees were created to consider the two most likely phylogenetic hypotheses for Sivapithecus indicus (YGSP 41216). Sivapithecus is widely considered a pongine given its facial profile (Raza et al., 1983; Begun et al., 1997a; Ward, 1997; Kelley, 2002). However, even though it has clear hominid (i.e., great ape) facial affinities, its exhibits postcranial morphologies resembling proconsuloids (Pilbeam et al., 1990; Pilbeam and Young, 2001; Madar et al., 2002; Morgan et al., 2015). Thus, we also considered the hypothesis that Sivapithecus could be a stem hominid in order to test how this phylogenetic position affects our evolutionary analyses. The two trees constructed here differ only in the location and branch length of Sivapithecus indicus. Both trees have a 12.3 Myr tip date following Morgan et al. (2015). When positioned as a stem hominid (Tree 1), Sivapithecus indicus was assigned a divergence date that is 1 Myr prior to the crown hominid node. This selection results in a 16.1 Myr divergence date, with a 2.8 Myr branch length leading to the 12.3 Myr tip date. When positioned as a pongine (Tree 2), Sivapithecus indicus was assigned a 13.3 Myr divergence date, with a 1 Myr branch length leading to a 12.3 Myr tip date. There is not general consensus for date and/or phylogenetic placement of several fossil specimens that are analyzed in this study by boxplots (i.e., KNM-RU 7694, Sts 14, SK 3155, KNM-ER 3228, OH 28), and so these specimens were excluded from the ancestral state reconstructions. RESULTS Relative Lower Ilium Height An interesting result that can be appreciated after inspecting our boxplots (Fig. 2) is the heterogeneity in the ranges of variation of our extant sample. The boxplots demonstrate that all of the extant hylobatids, both orangutan species, and Pan troglodytes have long relative lower ilium heights (as captured by medians >2.5). On the other hand, Gorilla beringei, and to a lesser extent Gorilla gorilla and Pan paniscus, are more moderate (median values <2.5) and are overlapping with the fossil hominoids (i.e., KNM-MW 13142, YGSP 41216, KNM-RU 7694). Humans have the shortest species medians of any extant taxa, and all fossil hominins fall within the modern human range. It is worth noting that, among extant taxa, the upper range of modern humans only overlaps with the lowest range of Gorilla gorilla (a single individual), Gorilla beringei, Nasalis larvatus (proboscis monkeys), and Pygathrix nemaeus (doucs). The monkey sample is variable, with some colobines having a relatively short lower ilium, but with most monkeys having fairly elongate lower ilia. The species with the relatively longest lower ilium height in the

7 834 HAMMOND AND ALM ECIJA Fig. 2. Relative lower iliac height is compared by boxplot among anthropoid taxa. Hominins have the shortest relative lower iliac height, but the hominin range does technically overlap with Gorilla, Nasalis, and Pygathrix. The fossil hominoids are smaller than the large values seen in hylobatids-pongo-pan troglodytes, and are most similar to the moderate values seen in Gorilla and Pan paniscus. Box represents the interquartile range, centerline is the median, whiskers represent nonoutlier range, and dots are outliers. entire anthropoid sample is Cebus apella (capuchin monkeys), followed by Hylobates moloch (as captured by their medians and maximum values). There is substantial overlap between extant apes and monkeys in relative lower ilium length. The analysis of the variance (ANOVA) and the pairwise comparisons reveal that the relative lower ilium height is not equivalent among all extant hominoids (Table 2). Pairwise comparisons of means found that the relative lower ilium heights of siamangs, gibbons,

8 LOWER ILIUM EVOLUTION IN APES AND HOMININS 835 TABLE 2. Pairwise mean comparisons (Tukey s HSD) between extant hominids and a subset of hylobatid taxa for the relative lower ilium height (i.e., ratio to the acetabulum; A) and allometrically scaled (i.e., linear model; B) lower ilium height A H. lar S. syndactylus P. pygmaeus P. abelii G. gorilla G. beringei P. troglodytes P. paniscus S. syndactylus P. pygmaeus P. abelii G. gorilla < < < < G. beringei < < < < P. troglodytes < < P. paniscus H. sapiens < < < < < < < < B H. lar S. syndactylus P. pygmaeus P. abelii G. gorilla G. beringei P. troglodytes P. paniscus S. syndactylus P. pygmaeus < < P. abelii < < G. gorilla < < G. beringei P. troglodytes < < P. paniscus < H. sapiens < < < < < < In B, Gorilla beringei was not compared with the other taxa because it exhibited an allometric slope statistically different from other taxa in the sample. Adjusted P values reported, and significant values are marked in bold. orangutans, chimpanzees and bonobos are statistically equivalent, except in two comparisons (Pan paniscus Hylobates lar, P , and Pan paniscus Pan troglodytes, P < 0.01). Gorilla beringei and Gorilla gorilla are not significantly different from each other (P ). However, gorillas are significantly different from all other taxa except bonobos (Pan paniscus Gorilla gorilla, P 5 0.9, and Pan paniscus Gorilla beringei, P 5 0.2). As expected, modern humans are significantly different from all other taxa (P < 0.01). The relative lower ilium height in humans does not demonstrate a clear relationship with sex or population (Fig. 3). Although female humans tend to have relatively longer lower ilium height in certain populations (i.e., American whites, I~nupiak arctic peoples, narrow-bodied American blacks), this pattern is not present in all populations. Pooled human female (mean ) and male (mean ) relative lower ilium height values are not significantly different (one-way ANOVA p ). There is not an obvious pattern for how human lower ilium height might vary among human populations (one-way ANOVA p ), and the range of variation seen in humans overall is similar to the ranges observed in other anthropoids (Fig. 2). Finally, there is not a significant interaction between sex and population in human relative lower ilium height (two-way ANOVA p ). Scaling of the Lower Ilium Height Allometric relationships between lower ilium height and acetabulum diameter were examined in our sample of extant hominids and selected hylobatids (Fig. 4). The species-specific regression lines are provided (Table 3). From visual inspection, Gorilla beringei has a steeper slope than the other taxa, although isometry cannot be ruled out at this time because of the large confidence interval for the slope (95% CI: ). Isometry cannot be ruled out at this time either for Pan paniscus, Pongo abelii and Symphalangus syndactylus. Hylobates lar is the only taxon to exhibit a negative slope. However, statistically speaking, both Hylobates lar and Pan troglodytes exhibit slopes that cannot be ruled out as different from zero (Table 3). If this is the case, it would indicate that relative lower ilium height does not scale with acetabulum diameter (and is therefore unlikely to scale with overall body size) in these two species. Furthermore, Gorilla beringei is the species in which the allometric relationship accounts for most of the variance in lower ilium height (70% of the variance), followed by Pongo pygmaeus (63% of the variance). In the rest of the species, the relationship between lower ilium height and acetabulum diameter accounts for a smaller proportion of the variance (2% 37%). As mentioned, our individually-fitted regressions indicate that some species scale with negative allometry, while for others isometry cannot be discarded. However, these results do not consider the interaction with each of the regressions of the remaining species for assessing whether their slopes are homogeneous or not. This is especially difficult to evaluate in our data because the 95% CIs are very large. To circumvent this problem, we relied on an analysis of covariance (ANCOVA), the results of which demonstrate that there are not significant differences in the slopes (i.e., there is no interaction) between the species in our sample (F ; P ) once Gorilla beringei is removed from the analysis (ANCOVA results with Gorilla beringei: F ; P < 0.01). Thus, we compared vis-a-vis the shape ratios with their allometrically-adjusted means (i.e., intercepts) for the species in which homogeneity of slopes could not be rejected (i.e., all hominoid species excluding Gorilla beringei; compare Table 2A with 2B). ANOVA and Tukey s HSD multiple comparisons of these scaled means do not add support to the shape ratio differences between Gorilla gorilla Pongo spp. and Gorilla gorilla

9 836 HAMMOND AND ALM ECIJA Pan troglodytes. Additional differences were detected by the scaling analyses (but not the shape ratios) between Gorilla gorilla and Pan paniscus, among others. These scaling results for Gorilla beringei support the differences revealed by the shape ratios (i.e., Gorilla beringei is different from other great apes in its scaling pattern). Nevertheless, it seems that Gorilla gorilla falls in the same allomeric trajectory as the other great ape species (but not hylobatids; Fig. 4). Thus, the scaling results raise the possibility that the relatively shorter lower ilium height of Gorilla gorilla could relate to some, still to be explained, size-related shape change. Other interspecific differences revealed by the shape ratio are Fig. 3. Sex-specific relative lower iliac height is compared by boxplot. Box represents the interquartile range, centerline is the median, whiskers represent non-outlier range, and dots are outliers. supported by allometrically scaled means (Table 3), implying that the differences observed in the relative lower ilium height are not merely caused by size differences alone. Ancestral State Reconstructions and Evolution of the Lower Ilium Height The ancestral state reconstructions at key nodes in the Hominoidea are nearly identical in both trees (Fig. 5, Table 4). In other words, the hypothetical phylogenetic position of Sivapithecus in these two trees does not substantially change the inferences made about the Pan-Homo node, nor does it unduly influence any part of the trees. Both trees show a progressive and independent increase in relative lower ilium height in some but not all hominoid lineages (i.e., except Gorilla species and, to a lesser degree, Pan paniscus). In other words, the long lower ilia observed in some extant hominoids appears to have evolved after the hominoid LCA. The relative lower ilium height in the hylobatid LCA is inferred to have already been fairly long, and this proportion becomes even longer in gibbons after the last common ancestor shared with siamangs (Fig. 5). Within the great apes, our analysis reveals independent elongation in Pan troglodytes and Pongo. However, it should be noted that the proportions seen in gorillas (both species) and to a lesser extent in Pan paniscus are estimated to be plesiomorphic for hominids (although note that there is overlap in the ranges of shape variation; Fig. 2). Within monkeys, Cebus, Macaca, Cercopithecus and Ateles are also estimated to have had an independent elongation of the lower ilium. The most rapid transformation in the proportions of the lower ilium is in hominins (Fig. 5), which suggests that this region was under intense selection in this group. Fig. 4. Allometric relationships of lower ilium height and acetabulum diameter in hominoids. Regression lines are fitted independently to each of the extant species listed in the legend (right; see Table 3 for regression results). Fossils were plotted to facilitate visual comparisons, but they were not included in this analysis.

10 LOWER ILIUM EVOLUTION IN APES AND HOMININS 837 TABLE 3. Allometric regressions of lower ilium height and acetabulum diameter in modern hominoids N R 2 SEE P Slope 95% CI Intercept 95% CI H. sapiens P. paniscus P. troglodytes G. beringei G. gorilla P. abelii P. pygmaeus S. syndactylus H. lar Variables were transformed using natural logarithm prior to the analyses. Abbreviations: N, sample size; R 2, adjusted squared-r; SEE, standard error of the estimate; P, P-value, significant slopes (i.e., statistically different from zero) are marked in bold; CI, confidence interval. We find that the predicted relative lower ilium height of the Pan-Homo LCA (Tree , CI ; Tree , CI ) is shorter than Pan troglodytes (range ). The Pan-Homo LCA confidence intervals overlap with the lower range for gorillas (range Gorilla beringei ; Gorilla gorilla ) and some monkeys. The Pan paniscus range ( ) is just above the estimated range of the Pan-Homo LCA. In regards to fossil apes, the Pan-Homo LCA confidence intervals overlap with Ekembo nyanzae (1.979; provided in Supporting Information) but do not overlap with Sivapithecus indicus (2.154). DISCUSSION The ancestral state reconstructions 5 predict that the relative lower ilium height of the Pan-Homo LCA would have been similar to gorillas and Ekembo nyanzae. However, the predicted lower ilium height of the Pan-Homo LCA is shorter than Pan troglodytes, which has historically been the referential model of the LCA. If the Pan- Homo LCA did have a moderate lower ilium length (as indicated by our results, Fig. 5), the reduction in length that occurred in the earliest hominins is less profound than if they had evolved from a long, chimpanzee-like ilium. The ilium itself appears to be a stand-alone module of the pelvis (Lewton, 2012) that develops relatively independently from the lower pelvis (Young and Capellini, 2015), suggesting that iliac proportions may be more evolvable than more highly-integrated regions. To this end, there is evidence based on overall pelvic shape that a chimpanzee-like pattern of pelvic integration would take 1.7 times longer to evolve from the Pan- Homo LCA than a pelvis with a more human-like integration pattern (Grabowski, 2012), which underscores that the evolution of the Pan troglodytes lower ilium length is interesting in its own right. The ancestral state reconstructions indicate that Pan troglodytes has increased the relative length of its lower ilium since the last common ancestor that it shared with Pan paniscus, whereas Pan paniscus displays a somewhat more plesiomorphic relative lower ilium height (i.e., plesiomorphic for panin clade). 5 The position of Sivapithecus in the tree does not impact the results for the Pan-Homo LCA estimate, so we discuss the ancestral state results here as being equivalent. The boxplots demonstrate that there is substantial variation in and among hominoid taxa in relative lower ilium length, and the pairwise comparisons revealed significant differences between hominoids often assumed to be equivalent. Of particular note are the comparisons for Pan paniscus and Gorilla both gorilla species are significantly shorter than all other hominoids except Pan paniscus, and Pan paniscus is significantly shorter than Pan troglodytes. However, it is important to acknowledge that there is substantial overlap between the ranges of variation (e.g., most, but not all, of the Pan paniscus variation falls within their lowermost range of Pan troglodytes variation). The scaling relationship of the lower ilium to the acetabulum also differs in Gorilla beringei relative to other hominoids. It shows a rapid increase in lower ilium height with an increase in the body size surrogate (i.e., acetabulum diameter; Fig. 4 and our ANCOVA results). The unusual scaling pattern of Gorilla beringei is unlikely to merely represent a consequence of its large body size among primates, because Gorilla gorilla lower ilium height scales with negative allometry here (as Lewton, 2015), and follows a pattern closer to the other extant apes (at least those apes showing statistically significant allometric slopes of lower ilium height and acetabular diameter). However, Gorilla gorilla s lower ilium height has been reported to scale isometrically by Lovejoy et al. (2009c), but showing positive allometry for ilium length overall (Ward, 1991), which might relate to having a long cranial ilium specifically (e.g., Machnicki et al., 2016). Although only based on observational data in Figure 4, it seems that males in Gorilla and Pan species have similar scaling patterns, with the females driving our observed differences. The sample sizes used here are too small to test this observation, but these observations regarding scaling of the female lower ilium would be an interesting avenue for future exploration. In general, a specific study of hominoid pelvic allometry incorporating large samples of each species/subspecies/sex will be necessary to test some of these ideas. However, the shape differences (lower ilium height:acetabulum diameter ratio) detected between Gorilla gorilla and Pan troglodytes and Pongo species is not supported by our allometrically-scaled means (see Table 2), which makes interpreting some of our evolutionary results difficult. For example, our evolutionary modeling (based on shape ratios) indicates that both species of Gorilla, with a moderate relative lower ilium height, retain the

11 838 HAMMOND AND ALM ECIJA Fig. 5. Evolution of lower ilium height (relative to acetabulum diameter) in two phylogenetic scenarios. Tree 1 (above) and Tree 2 (below) are identical except for the position of Sivapithecus: in Tree 1 Sivapithecus is positioned as a stem hominid (i.e., great ape and human clade), whereas in Tree 2 it is positioned as a pongine (i.e., Pongo lineage). The term Homininae in this study refers to the African ape and human clade. The method used here estimates internal node values, and then visualizes evolutionary change using a continuous color interpolation along the edges of the tree. The reconstructed values for key nodes are shown for both trees. The ancestral states estimated for all of these nodes on both trees are nearly identical. Confidence intervals and dates for the ancestral states are reported in Table 4.

12 LCA node LOWER ILIUM EVOLUTION IN APES AND HOMININS 839 TABLE 4. Ancestral state estimates for relative lower ilium height (LIH: acetabulum diameter) Date of node (Myr) Tree 1 estimate Tree 1, 95% CI Tree 2 estimate Tree 2, 95% CI Homo Australopithecus , , P. troglodytes P. paniscus , , G. gorilla G. beringei , , P. pygmaeus P. abelii , , Symphalangus Hylobates , , Pan Homo , , crown Homininae , , Pongine Homininae , , Hominidae Hylobatidae , , Hominoidea Cercopithecoidea , , plesiomorphic relative lower ilium length for hominoids and hominids (Fig. 5). Likewise, the apparent similarities between the position of Miocene apes, hylobatids and some gorillas in the log-log biplot (Fig. 4) could be interpreted as the hominoid/hominid plesiomorphic condition from which Pan and Pongo have independently departed by elongating their lower ilium. However, our allometric regressions (i.e., with Gorilla gorilla falling on the same allometric line as most hominoids; Fig. 5) could also be interpreted as suggesting that the relatively longer lower ilium height of Pan and Pongo represents the hominid plesiomorphic condition and that Gorilla gorilla has departed from this because of sizerelated shape changes. If that is the case, the departure of Gorilla beringei from the general great ape trend must be explained by other biological phenomena (i.e., it is different from other great apes in both the shape ratios and its size-related shape changes). The hylobatid results are revealing on this matter because, even though the range of their shape ratios falls within the uppermost range of great ape variation (Fig. 2), the hylobatid allometric regressions fall instead in the lowermost range (Fig. 4). Given that shape ratios represent individual shape values whereas scaling results are sample dependent, and also given that hylobatids are dramatically smaller than hominids and also more distantly related, we favor the view that in this case the shape ratios more closely represent the true lower ilium relative height in our hominoid sample (e.g., Jungers et al., 1995). In any case, these results demonstrate that there is not a single great ape or African ape lower ilium length, and emphasize the importance of each living and fossil hominoid lineage for understanding hominoid pelvic evolution. We consider this finding not very surprising, as other recent studies have found that chimpanzees and orangutans often appear the most derived among the extant great apes, with gorillas appearing to retain primitive hand proportions (Almecija et al., 2015) and scapular shape (Green et al., 2016). Specific predictions about the sequence of lower ilium height evolution in gorillas were largely excluded from the framework established by Lovejoy and colleagues (Lovejoy et al., 2009b). This study fills some gaps in our knowledge about pelvic evolution by showing that gorillas have a moderate relative lower ilium height that is similar to what is predicted for basal great apes, which adds support to the hypothesis that Pan troglodytes and orangutans independently elongated their lower ilia (i.e., a scenario where chimpanzees and orangutans evolved a longer lower ilium becomes as parsimonious as a scenario where gorillas and hominins secondarily shortened theirs). Furthermore, it should be noted that the differences in relative lower ilium height observed between some of the taxa, such as hylobatids and great apes or Gorilla beringei and the remaining hominids, cannot be readily explained by differences in body size (i.e., the scaling results). Functional Hypotheses A long lower ilium has been hypothesized to be an adaptation to vertical climbing and/or suspension (Ward, 1991; Lovejoy et al., 2009c; Lewton, 2015). Lower ilium height is reported to decrease in large taxa or those thought to experience larger locomotor loads (Lewton, 2015). Our data also find that, within hominoids, taxa specifically using arm-swinging behaviors (rather than vertical climbing per se) have the most elongate lower ilium. Pongo and Pan use forelimb suspension more frequently than gorillas (especially Gorilla beringei) (Cant, 1987; Doran, 1996, 1997; Doran and McNeilage, 1998; Remis, 1998; Thorpe and Crompton, 2005). The more suspensory gibbons have the longest lower ilium height among hominoids, with gibbons even showing a more elongate lower ilium height than siamangs. This follows known differences among the lesser apes, with gibbons using high-speed brachiation more frequently (Fleagle, 1976), preferring a more fine-branched discontinuous canopy than used by siamangs (Fleagle, 1976; Caldecott, 1980; Gittins, 1983; Cannon and Leighton, 1994), and having a body mass distribution and proportions that are better suited for brachiation than found in siamangs (Zihlman et al., 2011). There are several possibilities for why a long lower ilium might be adaptive for forelimb-dominated behaviors. The long bony pelvis of extant hominoids occupies a larger percentage of the trunk (gibbons 47%; siamangs 49%; orangutans 51%; common chimpanzees 57%) than in Old World monkeys (37%) and, especially, in humans (42%) (Schultz, 1969a,b,), and so trunk rigidity should typically be higher (Thompson et al., 2015) in extant apes because a larger proportion of the torso is occupied by the bony pelvis (Smith and Savage, 1956; Kummer, 1975; Ward, 1991; Lovejoy et al., 2009c; Hunt, 2016). This explanation falls short on at least two accounts. To begin with, there are some generalized arboreal quadrupeds (i.e., Cebus, Macaca, Cercopithecus) demonstrating