Ardipithecus

Writing about the Sarmiento-White exchange [1] [2] a couple of weeks ago, I mentioned that I had three areas of comment. The molecular clock argument was the first, the pelvis the second. My trip to Europe got in the way of finishing up these notes, but after the molecular clock and the pelvis, I've come to the third area -- the canine-promolar complex.

Here is Sarmiento's exposition:

Fourteen of the 26 characters in table 1 in (1) common to Ardipithecus and Australopithecus are in the canine/premolar complex. However, reliance on the canine/premolar complex to diagnose hominids (in the classic sense) has misdiagnosed Miocene fossil apes (i.e., Oreopithecus and Ramapithecus) as early human ancestors (12, 13). Character polarity for this complex is not clear-cut, with many early hominoids, especially females, often showing a humanlike condition. The canine/premolar complex shows such a marked grade of character lability (e.g., conspecific males and females show the diagnostic character differences) that reversals in polarity could have occurred repeatedly over the evolutionary periods necessary for these fossil genera to differentiate (12). Approximation to the humanlike canine/premolar complex, therefore, does not indicate that Ardipithecus is a hominid or ancestral to Australopithecus any more than it indicates that Oreopithecus and the orangutan-like females of Sivapithecus, both of which also share a humanlike premolar/canine complex, are hominids or represent a descendant-ancestor continuum.

Sarmiento here accepts that the C/P₃ complex of Ardipithecus "approximates" the hominin condition. I would not go so far as this. I faced a big difficulty in understanding the description of the dentition by Suwa and colleagues [3]. As I pointed out last fall, they did not report standard measurements. Those have been hidden from everyone. Moreover, most of the metric comparisons presented graphically in the paper and its supplement are non-standard ratios.

Such comparisons have many disadvantages. Many lack any clear biological meaning (for example, why take the ratio of lower P₃height to M₁length?). Why use the maximum diameter of the P₃ without any reference to the angle of the tooth relative to the mesiodistal axis?

A few things are clear from the description and photos of upper canines and lower third premolars in figure S14 of Suwa et al. [3]. Many of the Ardipithecus upper canines are worn on the tip, and some of them are worn extensively. None of them have "honing" wear, which means wear that enhances the cutting function of their distal edges. The upper canines of Ardipithecus are like male bonobos in their crown diameter and intermediate between male and female bonobos in their labial height. There are some similarly large canines in Australopithecus afarensis, although the average in that species is significantly lower in crown diameter.

Australopithecus afarensis has a range of P₃ forms. Some are more apelike, asymmetrical crowns dominated by a single buccal cusp and angled relative to the mesiodistal axis. Others are more humanlike, with two more or less equal cusps and perpendicular to the mesiodistal axis. In Au. afarensis, the more apelike The Ardipithecus range of variation apparently never included any like these latter, humanlike P₃s from Au. afarensis. The P₃s of Ardipithecus also had substantially greater crown heights than typical of Australopithecus. The Ardipithecus P₃ is hominin-like in only two senses: It is smaller than the chimpanzee equivalent, and never has a facet for honing wear on the upper canine. This may be why Suwa and colleagues (2009) and White and colleagues [2] referred to the "C/P₃ complex" instead of simply the P₃. It may also explain why the research paper by Suwa and colleagues [3] did not illustrate any comparative statistics of the P₃. The fact is that the P₃ of Ardipithecus is by itself apelike.

White and colleagues respond to Sarmiento's points as follows:

The greatly expanded Ar. ramidus dental sample now further obviates Sarmiento’s assertions by establishing a metrically and morphologically refined Ar. kadabba-Ar. ramidus-Au. anamensis-Au. afarensis morphocline (5–7). It now seems clear that not all recovered Ar. ramidus canines can be female (7). Feminization of the male Ardipithecus C/P3 complex is robustly documented [detailed in the supporting online material in (7)]...

Wait a minute! "Robustly documented?!" There are no measurements!

...and is incompatible with Sarmiento’s argument that Ar. ramidus represents the stem taxon for both African apes and humans (1). If that were the case, a hominid-like C/P₃ complex with lack of honing would need to have evolved in Ar. ramidus, only to have independently reverted to the honing complexes in each African ape clade.

White and colleagues here link two assertions. First, they assert that the C/P₃ complex in Ardipithecus is actually like later hominins. On that assertion, the facts listed above are the pertinent ones. There are dental similarities, many of which are also shared with one or more lineages of Miocene apes. As Sarmiento claims, these similarities are arguably "shallow" -- mostly they reflect the fact that Ardipithecus has smaller canines than chimpanzees.

Second, White and colleagues assert that the canine morphology of chimpanzees and gorillas is unlikely to represent an evolutionary reversal. This second assertion is taken up later in the comment:

The character distributions we noted in the pelvis, C/P₃ complex, and basicranium are consistently indicative of a sister relationship of Ar. ramidus with Australopithecus (and later hominids). For Ar. ramidus to be a stem species of the African ape and human clade as Sarmiento advocates, its highly derived C/P₃ complex morphology, basicranial shortening, and iliac structure must have first emerged in some yet-unidentified Miocene ancestor before then reverting to an African ape–like condition. Such multiple, nonparsimonious character reversals are highly unlikely.

Shouldn't a sympathetic editor would have stopped them from including this passage? Lovejoy and colleagues' "Great Divides" paper [4], in the special issue of Science, was completely devoted to the argument that chimpanzees and gorillas derived most of their locomotor adaptations in parallel to each other. That includes a series of postcranial derived features ranging from the lumbar spine, wrist, hand and foot, limb proportions and the pelvis. No matter what we may think of Ardipithecus, the dentitions of chimpanzees and gorillas also evince substantial parallelism in enamel and crown morphology.

True or not, none of that is parsimonious.

I don't see any reason why the C/P₃ complex should be special evidence of relationship. Lovejoy and colleagues [4] had given some attention to the idea that morphological characters might be genetically correlated through changes to toolkit genes. For no system is such genetic correlation as likely as for the dentition. The serial homology among neighboring teeth reflects the action of precisely the same genes in different segments established early in dental development. Moreover, unlike the case of digits and limbs, the teeth may mechanically affect each other's development because they are adjacent to each other. That means that neither morphological reduction of the canines nor any change in their eruption schedule can be isolated from changes in the premolars.

Is it any surprise that the reduction of upper canines should result in a lack of honing wear on lower premolars? Honing is a consequence of a direct mechanical connection that becomes impossible with sufficient reduction of the canine.

Can we isolate a slight reduction in the lower third premolars from the significant reduction in the lower canines? As Suwa and colleagues (2009:95) point out, this lower canine evolution is not limited to hominins but also occurred in parallel in bonobos and some Miocene lineages:

A hominid-like incisiform LC morphology (high mesial shoulder, developed distal crest terminating at a distinct distal tubercle) is seen in some female apes (e.g., Ouranopithecus and P. paniscus), whereas the LCs of Ar. kadabba and Ar. ramidus tend to be conservative, exhibiting a strong distolingual ridge and faint distal crest, typical of the interlocking ape C/P₃ complex (4) (Fig. 1 and SOM text S1).

I just don't think that the Miocene ape record can sustain the argument that the C/P₃ "complex" carries much phylogenetic weight. A bonobo-sized canine is never going to be strong evidence linking a fossil to the hominins. That's part of why Ardipithecus remains such a doubtful case.

References

The other day, I started writing about the Sarmiento-White exchange on Ardipithecus, by describing how they disagree about the implications of the molecular clock.

What really prompted me to break up my discussion into three posts was that it takes quite a lot of space to explicate the features of the pelvis. I've taken care to reference the description by Lovejoy and colleagues (2009c), the general discussion of Ardi's locomotor anatomy in Lovejoy et al. (2009a, 2009b), and the discussion of early hominin pelvic evolution by Lovejoy and colleagues (1999).

I have a major hesitation that keeps me from writing anything about the Ardipithecus pelvis beyond those descriptions: Independent investigators at present cannot verify or replicate any comparisons made in Lovejoy and colleagues' analyses. Most of the measurements and many quantitative observations depend on a 3-d model. That model is not available for inspection, and the published description does not provide enough detail about the model to independently assess its accuracy. Worse, as I discussed last fall, the model appears to have been derived from the a priori expectations about pelvis evolution that Lovejoy and colleagues published in 1999.

As a result, I don't think any independent reader, including me, can tell how much of the model is real.

Given my problems understanding their pelvis 3-d model, I've decided to limit myself to the narrow points considered by Sarmiento's (2010) comment and White and colleagues' (2010) reply. Lovejoy and colleagues (2009b, 2009c) claimed that most of the pelvic anatomy of Ardipithecus is primitive for great apes, and that many of the pelvic features shared by chimpanzees and gorillas evolved in parallel in those two lineages. But they listed a few features that they considered to be derived in Ardipithecus and shared with Australopithecus. Sarmiento lists these, together with two features of the foot, and argues that they are not compelling evidence that Ardipithecus is a cladistic hominin:

Of the remaining characters listed as common to Ardipithecus and Australopithecus, none of the eight postcranial characters (sagittal iliac/isthmus orientation, slightly broadened iliac breadth, strong anterior inferior iliac spine formed by separate ossification center, robust second metatarsal base and shaft, dorsally domed second to fifth metatarsal heads, upwardly canted proximal foot phalanges, and short iliac isthmus and pubic symphysis outline), nor the other four craniodental characters [anterior basion position (14), advanced cranial flexion, and broad lower molars and mandibular corpus] are shown by systematic comparisons to be exclusive to humans or share-derived with humans. Nearly all are quantitative characters that appear in early hominoids (i.e., Oreopithecus and Dryopithecus) and have appeared independently in other primate lineages, and character simplicity is such that parallelisms or reversals in polarity cannot be demonstrated (12, 15).

I think Sarmiento's argument is entirely reasonable. Lovejoy and colleagues (2009a, 2009b) claimed a long series of parallelisms between chimpanzees and gorillas. Despite some reservations, I tend to agree -- Ardipithecus is primitive in its postcranial anatomy, and living apes are convergently derived. But take the argument to its logical end, and it becomes Sarmiento's. Ardi shares some postcranial features with hominins that living apes lack, but how do we know that any of them are derived? Or if they are derived, how do we know that they aren't trivially simple to evolve in parallel?

In their published reply to Sarmento, White and colleagues do not mention the long series of great ape postcranial features that they previously argued to be cases of parallel evolution (Lovejoy et al. 2009b, 2009c). Instead, they claim that three features of the pelvis are so convincingly like Australopithecus that Ardi must be a hominin:

Although isolated aspects of pelvic morphology of Oreopithecus may partially mimic those of Ar. ramidus [such as a projecting anterior inferior iliac spine (AIIS)], crucial postcranial elements of the latter (9, 10) are unambiguously derived toward the Australopithecus condition, to the exclusion of Oreopithecus. Some of these derivations probably stem from shared changes in pattern formation exhibited by both Ar. ramidus and Australopithecus. In the pelvis, these include (i) superoinferior approximation of the sacroiliac and acetabular joints by iliac isthmus shortening and (ii) a sagittally oriented and greatly broadened lower iliac isthmus accompanied by (iii) an exaggerated anterior margin, itself the product of a unique physis for the AIIS, shared only with phyletic hominids.

I find this reply very strange. The "shared changes in pattern formation" hypothesis actually supports Sarmiento's argument. If White and colleagues are correct about the morphogenetic basis of the Ardipithecus pelvic anatomy, that makes it more likely to have evolved convergently with Australopithecus, not less likely. Lovejoy and colleagues (1999) emphasized this point -- the pelvic features of hominins were likely to have evolved due to selection for a shorter pelvis, principally for biomechanical reasons, with other characters of the pelvis and femur changing entirely due to their genetic correlation with this major target of selection.

The reply omits the most persuasive of the derived features in hominins -- the short ilium -- which was at the center of Lovejoy and colleagues' (1999) account of hominin pelvic evolution. Here's a comparison of 3-d models:

Ardi looks very obviously like the human and Lucy, and very different from the chimpanzee, right? But I think that the chimpanzee model in this picture is larger than it should be, as the acetabulum looks much larger than Ardi even though Lovejoy and colleagues (2009c) report Ardi's acetabulum as right in the middle of the chimpanzee range. Maybe they chose a large chimpanzee, or built the Ardi 3-d model using the smaller end of their range of possible acetabular diameter. You see the problem of using a model instead of the actual fossil?

In any event, the differences between Ardi's os coxa and the chimpanzee's are obvious. Ardi has a much shorter ilium. The chimpanzee has an iliac blade that comes right out of the picture toward us, because it is oriented along a coronal axis. Ardi's angles forward, or anteriorly, like the hominins.

In fact, if we look at the model in superior view superimposed on Lucy's pelvis, you can see that Ardi's iliac blades angle even more anteriorly than Lucy's:

The three features White and colleagues (2010) list, as quoted above, are morphological side effects of the shorter, more sagitally angled ilia. Lovejoy and colleagues (1999) paper would likely have described these features as side effects of selection for a shorter pelvis with an anteriorly directed origin for the rectus femoris muscle.

The question is: How much of the functional similarity between Ardi and hominins is homology, and how much is convergence? Similarity may not reflect homology -- descent of the feature from the same ancestor.

That point is especially notable when White and colleagues (2010) discuss Oreopithecus -- an extinct ape whose pelvis shares some features with hominins, and other features with apes. Oreopithecus is not a hominin, but it may have had some adaptations to a bipedal stance. Yet it also shares features that Lovejoy and colleagues (2009b) have argued must have evolved convergently in orangutans, chimpanzees and gorillas. That seems like a real problem for the idea that Ardipithecus represents the primitive condition for such traits.

Here's the Oreopithecus paragraph from White et al. (2010), the first time that Ardipithecus and Oreopithecus pelvic features have been compared (other than here on the blog):

Indeed, Oreopithecus diverges from hominids remarkably in features ranging from limb proportions to dental anatomy. In the pelvis, it features bi-iliac entrapment of at least one lumbar vertebra and general immobilization of the lumbar column (including transformation of lumbar somites into its six-segment sacrum). Such changes stand in stark contrast to the six lumbar, four-segment sacrum of Au. afarensis, a character adumbrated by the precipitous reduction in iliac height (and extensive broadening) of the Ar. ramidus ilium (10). African apes have entirely rigidified lumbar columns that differ radically from those of hominids.

I think this comparison is very important. Oreopithecus is not a member of the orangutan clade, and Lovejoy and colleagues' (2009b) scenario implies that if Oreopithecus is a member of the African ape clade, it -- like chimpanzees and gorillas -- must have evolved these features convergently.

Can it be that orangutans, chimpanzees, gorillas, and Oreopithecus all acquired the distinctive "bi-iliac entrapment" of the lower lumbar vertebrae in four separate instances of evolutionary convergence? Put those together with the elongation of the arms, reduction in the length of the lumbar column, and sacralization of lumbar vertebrae. Far from a simple change, it a series of complicated, correlated changes. Lovejoy and colleagues (2009b) defended the hypothesis that these traits are parallelisms shared by all the lineages of living great apes. Now, White and colleagues (2010) are forced to posit a fourth independent evolution of many of these traits in Oreopithecus.

Despite those similarities to living great apes, Oreopithecus shares with hominins the development of a relatively prominent anterior inferior iliac spine. This implies an adaptation to hip flexion or knee extension with a more extended leg. Bipedal stance is one possible explanation for this anatomy, and is the explanation that Lovejoy and colleagues (2009c) offer for its presence in Ardipithecus. White and colleagues (2010) include this as their feature (iii), the "unique physis for the AIIS, shared only with phyletic hominids." But this description seems exaggerated, when we consider what Lovejoy and colleagues (2009c:71e3) actually wrote:

The form and size of the AIIS in ARA-VP- 6/500, as well as its projection anterior to the acetabular margin, indicate that this structure had already begun to appear and mature via a novel physis.

A "novel physis" refers to a separate growth plate for the anterior inferior iliac spine. Ardi was an adult, and her pelvis was fully developed. So there's no observing whether the anterior inferior iliac spine had its own growth plate. Lovejoy and colleagues (2009c, 2010) are just claiming there must have been one. What basis could there be for such a model, other than an allometric analysis of the anterior inferior iliac spine in humans and other primates where it is present -- such as Oreopithecus? Remember that Ardi is more than twice the body size of Oreopithecus, yet Rook and colleagues (1999) showed that the cancellous structure within the anterior inferior iliac spine of Oreopithecus is a close match to Homo. That anatomical similarity may imply a common developmental pathway in Oreopithecus and hominins.

Is the anterior inferior iliac spine homologous in Oreopithecus and Ardipithecus? If so, it is probably primitive for great apes, not derived in hominins. Does it have another functional role besides bipedal stance? If so, that functional role might well have occurred in Ardipithecus, another arboreal quadruped.

Could other features of Ardi's pelvis be consequences of arboreal quadrupedal locomotion in an ape with a long lumbar spine? The sagittal orientation of the iliac blades and isthmus is not like living great apes, but it is like living Old World monkeys. Ardi's ilia are shorter than monkey ilia, but the question deserves some serious allometric study. Also deserving of study is whether isthmus orientation in monkeys matches that of the iliac blades, and if not, why not? One hypothesis would be the morphogenetic effects of selection for a shorter ilium length, the scenario published by Lovejoy and colleagues (1999).

I don't think there's any question that the evolutionary scenario outlined by Lovejoy and colleagues (2009b) is highly non-parsimonious with respect to the postcrania. It requires the convergent evolution of a long suite of characters within all the living great apes in at least three separate evolutionary histories. Add in fossil apes -- at least Oreopithecus, and possibly Morotopithecus and Dryopithecus -- and the number of parallelisms is extreme. The chimpanzee-gorilla convergences go even further beyond those shared with orangutans to include the knuckle-walking features of the wrist and hand, and several dental characters.

White and colleagues (2010), as I'll describe in the next post, argue that the shared dental characters of Ardipithecus and Australopithecus necessitate their close relationship. Once this is assumed, the many postcranial convergences become necessary. In that perspective, it helps to "soften the blow" somewhat by identifying those postcranial features shared by Ardipithecus and the hominins.

From the perspective of the pelvis, I'll return to one feature of Ardipithecus that seems independent, shared with hominins, and lacking in Oreopithecus: the "precipitous reduction in iliac height," so obvious in the picture above. But Ardi's os coxa is badly crushed at the superior border of the ilium. My post from last fall includes photos of both Ardi's os coxa and the pelvis of Oreopithecus. Ardi's is relatively shorter, no question, and it lacks the great height on its medial aspect, that creates the "entrapment" of the last lumbar vertebra of Oreopithecus. But the crushing seems to obscure this anatomy, so that it's not possible to be sure from the photos.

I wish we had better than a cartoon model to compare. During the seven months since I first detailed what I see as weak points in the pelvic description, I've become less and less persuaded that the pelvic features reflect any hominin-like locomotor adaptations in Ardipithecus. There are many unresolved functional issues, which obscure the phylogenetic relations between living and fossil apes. Ardi makes every tree less parsimonious, no matter which branch we put her on. Shoe-horning her into the hominins doesn't solve many problems, and creates some intractable ones.

I find myself calling her an ape.

References:

   Abitbol MM. 1995. Reconstruction of the sts 14 (Australopithecus africanus) pelvis. Am J Phys Anthropol 96:143–158.

   Harrison T. 1986. A reassessment of the phylogenetic relationships of Oreopithecus bambolii. J Hum Evol 15:541–584.

   Harrison T. 1991. The implications of Oreopithecus bambolii for the origins of bipedalism. In: Coppens Y, Senut B, editors, Origine(s) de la bipédie chez les hominidés, Cahiers de Paléoanthropologie. Paris: Editions du CNRS. p 235–244.

   Köhler M, Moyà-Solà S. 1997. Ape-like or hominid-like? the positional behavior of Oreopithecus bambolii reconsidered. Proc Natl Acad Sci U S A 94:11,747–11,750.

   Lovejoy CO, Cohn MJ, White TD. 1999. Morphological analysis of the mammalian postcranium: A developmental perspective. Proc Natl Acad Sci U S A 96:13,247–13,252.

   Lovejoy CO, Simpson SW, White TD, Asfaw B, Suwa G. 2009a. Careful climbing in the Miocene: The forelimbs of Ardipithecus ramidus and humans are primitive. Science 326:70e1–70e7.

   Lovejoy CO, Suwa G, Simpson SW, Matternes JH, White TD. 2009b. The great divides: Ardipithecus ramidus reveals the postcrania of our last common ancestors with African apes. Science 326:100–106.

   Lovejoy CO, Suwa G, Spurlock L, Asfaw B, White TD. 2009c. The pelvis and femur of Ardipithecus ramidus: The emergence of upright walking. Science 326.

   Robinson JT. 1964. Adaptive radiation in the australopithecines and the origin of man. In: Howell FC, Bourlière F, editors, African ecology and human evolution. London: Methuen and Company, Limited. p 385–416.

   Rook L, Bondioli L, Köhler M, Moyà-Solà S, Macchiarelli R. 1999. Oreopithecus was a bipedal ape after all: Evidence from the iliac cancellous architecture. Proc Natl Acad Sci U S A 96:8795–8799.

Sarich VM. 1971. A molecular approach to the question of human origins. In (P. Dohlinow & V.M. Sarich, Eds.) Background for Man: Readings in Physical Anthropology, pp. 60‐81. Boston: Little, Brown.

Sarmiento EE. 2010. Comment on the paleobiology and classification of Ardipithecus ramidus. Science 328:1105. doi:10.1126/science.1184148

   White TD, Asfaw B, Beyene Y, Haile-Selassie Y, Lovejoy CO, Suwa G, WoldeGabriel G. 2009. Ardipithecus ramidus and the paleobiology of early hominids. Science 326:75–86.

White TD, Suwa G, Lovejoy CO. 2010. Response to Comment on the paleobiology and classification of Ardipithecus ramidus. Science 328:1105. doi:10.1126/science.1185462