First steps of bipedality in hominids: evidence from the atelid and proconsulid pelvis

Pelvic anatomy in Ardipithecus

There is mounting evidence that the LCA of panids and hominids lacked most of the specialized (derived) characters associable with suspension and vertical climbing seen in extant African apes [e.g., narrowed sacrum, elongated iliac isthmus, modification of femoral insertion characters of the hip musculature, reduction in the number of lumbar vertebrae to an average of 3.5, elongated sacra (by conversion of somite fates from lumbar to sacral) and elongated thorax (by conversion of somite fates to thoracic—especially advanced in the bonobo)] (Almeçija et al., 2013; Crompton, Vereecke & Thorpe, 2008; Kivell & Schmitt, 2009; Lovejoy & McCollum, 2010; McCollum et al., 2010; Reno, 2014; Thorpe, Holder & Crompton, 2007; Ward, 2015). It is unlikely that these characters were present in the LCA, since the metacarpus, carpus, limb proportions, foot, femur, humerus and ulna in ARA-VP-6/500 (“Ardi”) exhibit morphologies that lack definitive modifications for suspension (Almeçija, Smaers & Jungers, 2015; Lovejoy et al., 2009a; Lovejoy et al., 2009b). They are instead more consistent, based on the considerably likely substantial body mass of Ardipithecus, with deliberate quadrumanual climbing and bridging accompanied by ulnar withdrawal and posterolateral shoulder relocation with attendant invagination of the vertebral column (see below) (Lovejoy et al., 2009c; White et al., 2015). These were accompanied by substantial lateral enlargement of the iliac blade and reduction of the retroauricular region of the os coxa in Ar. ramidus indicating spinal invagination into the thorax (Lovejoy & McCollum, 2010; Lovejoy et al., 2009d).

Exceptions to the generalized primitive state in Ar. ramidus are its “upper” pelvis and lateral foot, which are consistent with a functional hip stabilization mechanism and rigid toe-off during upright walking. Its ilium is similar to those of later hominids and includes an anterior inferior iliac spine likely developed by a separate ossification center, a hominid apomorphy. Unfortunately, the specimen lacks both its lumbar column and a sufficient amount of its sacrum to permit direct demonstration of its capability for lordosis (Lovejoy et al., 2009a; Lovejoy et al., 2009b; Lovejoy et al., 2009c; Lovejoy et al., 2009d). This presents an interpretive conundrum. How can we then deduce how lordosis was likely achieved in the earliest phases of upright walking in the hominid clade?

The lumbosacral and pelvic anatomy of the LCA

The pelvis of the middle Miocene proconsulid, Ekembo nyanzae (McNulty et al., 2015), was generally similar to those of Old World monkeys (Ward, 1991; Ward, 1993; Ward et al., 1993). Its ilia were superoinferiorly long with a substantial gap between the sacral promontory and pubic symphysis, and the iliac fossa was quite narrow. The os coxa exhibits a relatively massive retroauricular portion (i.e., the iliac tuberosity) (Ward, 1991). It is now clear that proconsulids and other Miocene hominoids such as Nacholapithecus lacked tails (Nakatsukasa et al., 2003a; Nakatsukasa et al., 2003b; Nakatsukasa et al., 2004; Ward, Walker & Teaford, 1991), even though most were still largely above branch quadrupeds (Beard, Teaford & Walker, 1986; Morbeck, 1975; Napier & Davis, 1959; Ward, 1993; Ward, 2015). Tail loss was likely a hominoid synapomorphy by at least 17.9 mya (Nakatsukasa et al., 2004). The lumbar column was long, probably numbering six or seven ribless vertebrae (McCollum et al., 2010; Ward, 1991).

A dramatic shift in pelvic proportions is present in Ar. ramidus. The relative sizes of its pre- and retro-auricular portions are almost fully modern and very unlike those of Ekembo having been transformed by substantial vertebral column invagination and very likely migration of the lumbar transverse processes from mid-centrum to pedicle (Lovejoy et al., 2009c; White et al., 2009b). A pelvic fragment of the Miocene taxon Pierolapithecus catalaunicus (Hammond et al., 2013) and the os coxa of the late Miocene Oreopithecus bambolii have been described (Harrison, 1986; Harrison, 1991; Rook et al., 1999; Straus, 1963; Wood & Harrison, 2011), although the latter’s poor state of preservation appears not to have been fully appreciated (White et al., 2015). Arguments that its lumbar vertebral bodies show wedging to facilitate lordosis (Kohler & Moya-Sola, 1997) have been rigorously rejected (Russo & Shapiro, 2013). Moreover, the os coxa’s extreme compression during fossilization puts little confidence in claims that it exhibited a “true” anterior inferior iliac spine (White et al., 2015), as its general form is grossly inconsistent with origin by a secondary apophysis which characterizes the structure in all known hominids (contra Harrison, 1991; Harrison & Rook, 1997; Straus, 1963). Nevertheless, these Miocene fossils do provide a basis for reconstructing the LCA’s general pelvic form.

The iliac isthmus in late Miocene pelves must have still been superoinferiorly long with a substantial “promontory-symphysis vertical gap,” but likely with some lateral expansion (i.e., “flaring”) of the iliac fossa (Hammond et al., 2013). P. catalaunicus suggests some modifications of the limbs, thorax and pelvis for more competent arboreal clambering and deliberate climbing (including ulnar withdrawal and a more elliptical thorax than monkeys and proconsulids), but not to the degree seen in Ar. ramidus (Moya-Sola et al., 2004). This is especially true of the position of the lumbar transverse processes, whose origins were not yet fully pedicular as in hylobatids and other Miocene taxa such as Hispanopithecus and Morotopithecus (MacLatchy, 2004; Moya-Sola & Kohler, 1996). In the present paper we distinguish “deliberate climbing” as predominantly above branch cautious based on the relatively large body mass of several Miocene specimens that would render the leaping and acrobatic behavior seen in smaller primates hazardous because of substrate reactive elasticity. We distinguish “vertical climbing” (sensu stricto, as opposed to generally climbing upward) as that form of ascension of large trunks performed specifically by living African apes (see DeSilva, 2009; Fleagle et al., 1981 for illustration). These are not necessarily mutually exclusive within a given primate’s locomotor repertoire, however they do impose different functional demands and likely selective pressures on the skeleton depending on their relative frequency.

This pelvic form of Miocene apes such as Ekembo, when combined with a relatively long lumbar column, provides a basis for two key questions concerning the acquisition of lordosis: (1) how permissive were the vertebral column and pelvis of the LCA for lordosis, and (2) which of the two morphological shifts observed in Ardipithecus (reduction of iliac height and/or expansion of sacral breadth) was more likely to have occurred first in the evolution of earliest hominid bipedality?

Atelids as a model for the relationships between the pelvis and lumbar column

New World atelids (Alouatta, Ateles, Brachyteles, and Lagothrix) are unique because they frequently engage in caudal suspension (Cant, 1986; Iurck et al., 2013; Johnson & Shapiro, 1998; Lemelin, 1995; Mittermeier & Fleagle, 1976), which often causes their lower spines to enter into substantial sagittal recurvature similar to lordosis in hominids (Fig. 1A). The muriqui (Brachyteles) and spider monkey (Ateles) are particularly relevant as they have also undergone lumbar column reduction resulting in approximately 6 post-transitional vertebrae similar to early hominids (Williams, 2011). Examination of their pelvic and sacral morphology therefore may provide indirect evidence of the potential pathway toward lordosis in earliest hominids.

In order to address this, we have reviewed lumbar and pelvic form and function in these primates. First, we predict that the ability to generate lordosis should facilitate the attainment of more erect hind limb postures. To address this, we report observations of terrestrial behavior of northern muriquis (Brachyteles hypoxanthus) and utilize opportunistic observations of bipedal locomotion in Ateles. Second, we hypothesize that lordosis will be accompanied by features associated with caudal lumbar emancipation. As such, we investigate iliac height, lower lumbar ligamentous support, and sacral width in relevant extant and fossil anthropoids. Finally, we provide a possible scenario for the role of pelvic and sacral form in the evolution of bipedality.

Brachyteles bipedality

Observations of northern muriquis (B. hypoxanthus) were conducted at the Reserva Particular de Patrimônio Natural-Feliciano Miguel Abdala in Caratinga, Minas Gerais, Brazil. In recent years, as a consequence of substantial local demographic change, we have noted an increase in the frequency and duration of bouts of terrestrial behavior in one population of unprovisioned northern muriquis living at high density (Mourthé et al., 2007; Tabacow, Mendes & Strier, 2009). These enable terrestrial behavior to be monitored in a naturalistic setting. We photographed and video recorded bouts of terrestrial behavior to determine the general nature of their bipedal posture (Fig. 1 and Videos S1 & S2). Methods were approved by the University of Wisconsin Animal Care Committee (protocol L00104 through April 2011; then a Wildlife Waiver). The Brazilian government, CNPq, and the administration of the field site provided permission.

Given the likely kinematic variability in bipedal locomotion of non-human primates, we also surveyed video resources for further examples of bipedal behavior of atelids (youtube.com and arkive.com using a combination the search terms “spider monkey,” “Ateles,” “muriqui,” “woolly spider monkey,” or “Brachyteles” and “bipedal” or “walking”). As these involved opportunistic observations the descriptions we have provided are necessarily qualitative.

Iliac and sacral anatomy

In order to compare the ligamentous anatomy of primates (especially atelids and Old World monkeys) with free (potentially lordotic) versus constrained (restrictive of significant lordosis) last lumbar vertebrae, we conducted detailed dissections of spider (Ateles sp.), muriqui (Brachyteles sp.), howler (Allouata sp.), and langur (Presbytis sp.) monkeys as well as a single gibbon (Hylobates sp). The first three taxa allow a comparison of the shortened and longer backed atelids, and the langur was chosen because it should represent the more generalized catarrhine condition. The gibbon provides a comparison to hominoid lumbosacral and pelvic anatomy absent the extreme specialization observed in the large-bodied great apes. These specimens are housed in the collections of the Cleveland Museum of Natural History, Kent State University, Northeastern Ohio Medical University, and Case Western Reserve University.

We also collected metric and nonmetric data from the pelves of 150 skeletonized specimens in the Cleveland Museum of Natural History and Harvard Museum of Comparative Zoology (Table 1 and Table S1). Unfortunately, we did not obtain a suitable sample of measurable muriqui specimens, but we did obtain metrics from each of the three other atelid genera. We paid particular attention to the position of the sacral promontory in relation to the iliac crest and ischiopubic ramus (Fig. 2). From these data we calculated the relative iliac height above the sacral promontory within the pelvis (see Fig. 3 and its legend). For size normalization we used acetabular diameter (Plavcan, Hammond & Ward, 2014) to calculate the Iliac Height Ratio. To determine the relationship between alar breadth and centrum breadth, we performed an analysis of covariance (ANCOVA) between genus means of log alar breadth and log centrum breadth with a binary categorical variable (monkey versus ape) as a covariate. To account for potential autocorrelation in related taxa, we used a phylogenetic generalized least squares (pGLS) regression in the ‘caper’ package in R (Orne et al., 2011; Team , 2012). We obtained the phylogenetic tree and branch lengths from the 10kTrees Project (Fig. S1) (Arnold, Matthews & Nunn, 2010).

Proconsulid sacral reconstruction

The partial skeleton of Ekembo nyanzae (formerly Proconsul) from Mfangano Island, Kenya, includes a portion of the first sacral body (S-1:KNM-MW 13142-M) (Ward et al., 1993). We used a Kenya National Museum cast as the “core” of our reconstruction of the sacrum. The specimen’s inferior portion includes the rim of the first sacral foramen on the left side, thus indicating the craniocaudal height of the S-1 body, and its posterior portion indicates overall thickness of S-1 just lateral to the left articular process (see Fig. 2 in Ward et al. 1993). The superior surface of the S-1 centrum is essentially intact with only minor abrasion. We used “extra firm” oil-based modeling clay to build the left ala and the (missing) articular facets of S1, incorporating but not obscuring the fossil’s S-1 fragment. Elements 2-5 were added as well to create a reasonable facsimile of the likely entire structure of the original specimen, although these details have no bearing on the role of our reconstruction for the current report. Most importantly, for the auricular portion of the sacrum, clay was molded to exactly conform to the well-preserved auricular surface of the nearly complete os coxa of KNM-MW-13142-D (also a Kenya National Museum cast). This molded surface was then mated to the S-1 cast absent the addition of any material not needed merely to successfully fuse the clay auricular mold to the left side of the specimen’s S-1 plaster body as preserved in the cast. Ward et al. (1993) describe the left ala as being “preserved adjacent to the first sacral body, but [with] its ventral edges…eroded away” (p. 84). Our reconstruction thus minimized the effects of this erosion and produced essentially a 3-D version of Ward et al.’s drawing of the likely pelvic structure of the specimen (their Fig. 14), although their version appears to indicate the addition of more material than was actually required for our version (Fig. S2). Nevertheless, the latter produced a slightly broader ala than indicated in their scaled drawing (see below). Reconstruction was completed by mirror imaging the (missing) opposite side. The clay model was then molded and cast in plaster. Dimensions of the original specimen are available from Ward et al. (1993:85). They report that the “articular surface of the first sacral body is elliptical in outline, 19.6 dorsoventrally and 30.0 mediolaterally.”

Lordosis and upright posture in atelids

Atelids have previously been observed during bouts of terrestrial behavior only rarely (Campbell et al., 2005; Dib, Oliva & Strier, 1997; Mourthé et al., 2007). Systematic analyses of substrate use in Brachyteles showed that by 2005 muriquis of all age-sex classes in our study group occasionally engaged in terrestriality, and by 2007 adult males were spending 1.5% of their time on the ground, nearly a 50% increase from the 0.8% of time they spent on the ground in 2006 (Tabacow, Mendes & Strier, 2009). Their terrestrial activities had also diversified to include both essential ones (e.g., drinking, traveling across gaps in the forest, and feeding) and nonessential ones (e.g., socializing, including mating, resting, and traveling in areas where arboreal alternatives were available). Increasing terrestriality has continued to be present in this group and in the other three muriqui groups in our study population through July 2015 (K Strier, 2015, unpublished data). In 2011, with the use of strategically-deployed motion-sensitive camera traps, we were able to demonstrate for the first time that their terrestriality occurs in the absence of observers. Video images (Videos S1 and S2) show adult male muriquis descending to the ground to feed on fallen fruits in an open area within the forest, engaging in reassuring social contact as they move through the area with a combination of quadrupedal and bipedal postures. Thus, despite the overall rarity of terrestriality in Brachyteles, examination of posture in these naturalistic bouts makes it clear that individuals can readily achieve both a fully erect trunk and a near complete extension of the lower limb. Examples of these postures are shown in Figs. 1B, 1C and 1D. These observations indicate that the conformation of muriqui lumbar, pelvic and sacral anatomy is sufficient to produce fully extended hind limb postures, and that such behaviors are not dependent on habituation or training in captive animals. In particular, Figs. 1B and 1C indicate that hind limb extension can be achieved without the substantial lateral rotation of the limb observed in orangutans (see below) (Stern & Larson, 1993; Tuttle, 1979).

We also searched for other instances of videos depicting bipedal posture and locomotion in atelids to confirm the capacity of bipedal posture and locomotion in other species. As expected these tended to be highly variable using a variety of postures given their primary use of quadrupedal locomotion. However, opportunistic observations of spider monkeys (Ateles sp.) demonstrate that near or actually fully extended hip and knee postures can be achieved during bouts of bipedal locomotion in this species as well (Fig. 4 and Video S3).

Ligamentous support

We further explored if mechanisms associated with caudal lumbar mobility extended to also include soft tissues. Either direct articulation of the transverse processes or bilateral ligamentous attachment would limit mobility in vertebrae located directly between the iliac crests. Each of the atelids and the gibbon had reduced lumbar columns relative to those of the langur and the presumed catarrhine ancestral condition (McCollum et al., 2010; Pilbeam, 2004; Schultz & Straus, 1945). All of the dissected atelids exhibited a free last lumbar vertebra, while the most caudal lumbar in the langur and gibbon were instead located between the more dorsally extended ilia confirming previous observations (Figs. 2 and 3). Dissection revealed substantially less ligamentous restriction in the atelids than in the gibbon and langur, but ligamentous tissue was generally denser and more elaborate in the langur than in the gibbon (Fig. 5). Presumably this is related to the retention of a massive erector spinae in the cercopithecoids such as the langur (Benton, 1967), and loss of the tail (with some partial invagination of the spine) in the gibbon. It appears that any lordosis in the langur is accomplished mostly by the superoinferior length and number of its lumbar and post-transitional thoracic vertebrae (along with some presumed differential disc compression) as observed in macaque monkeys trained to habitually walk bipedally (Preuschoft, Hayama & Günther, 1988).

Sacral alar breadth

Another osteological feature that can impact lumbar mobility is the breadth of the sacrum and its effect on interauricular distance. To determine relative sacral breadth in atelids, we normalized alar breadth by the transverse breadth of the first sacral centrum (as a measure of body size) in anthropoids (Fig. 6). A comparison of the two regression lines representing monkeys and apes is of interest. The sacra of both New World and Old World monkeys, including atelids, are broader than those of suspensory apes. In contrast, Homo and Australopithecus individuals plot well above the combined monkey regression line. This shows that atelids possess a broad sacrum similar to those of other monkeys and that apes alone are distinguished by extreme sacral narrowing.