What did Hadropithecus eat, and why should paleoanthropologists care?

INTRODUCTION

Hadropithecus stenognathus, a recently extinct giant lemur (Primates, Archaeolemuridae) from Madagascar, bears remarkable resemblance in its cranial morphology and dental proportions to basal hominins (especially robust australopiths such as Paranthropus boisei and P. robustus) [Ryan et al., 2008] and in its molar occlusal morphology to grazing mammals such as geladas (Theropithecus), capybaras (Hydrochoerus), hippopotamuses (Hippopotamus), and kangaroos (Macropus) (Fig. 1). Like basal hominins, Hadropithecus has a short face, small and orthally implanted upper and lower incisors, small canines, early fusion of the mandibular symphysis, thick mandibular corpus, flaring zygoma, molariform posterior premolars and large and buccolingually expanded molars. Its mandibular ascending ramus is tall so that the temporomandibular joint is raised high above the mandibular occlusal plane. In contrast, its molar occlusal morphology looks nothing like that of a basal hominin. Its relatively tall cusps and crests form complex flat ribbons of enamel that retain sharp edges as they wear. Its enamel is only moderately thick (as in Theropithecus) and its enamel prisms are only weakly decussated [Godfrey et al., 2005]. Understanding how an animal with molars so similar to those of geladas and other grazers might also converge so strongly in cranial architecture, incisor and canine reduction, P4 molarization, and molar hypertrophication with basal hominins is clearly of paleoanthropological significance.

At least four hypotheses have been proposed to account for the morphological convergences of Hadropithecus with hominins, geladas, or both. Most have focused on diet, some more on the metric properties of food items, and others more on their material properties.

Testing the above hypotheses requires understanding more precisely what Hadropithecus ate. Stable isotope analysis can help us do exactly this. On the basis of its high stable carbon (δ¹³C) and nitrogen (δ¹⁵N) isotope values, Crowley & Godfrey [2013] posited a diet for Hadropithecus stenognathus rich in Didiereoideae (a subfamily of spiny succulent plants that is endemic to Madagascar). In southwestern Madagascar, these succulent plants, which rely on Crassulacean Acid Metabolism (CAM), have high δ¹³C values [Crowley & Godfrey, 2013; Kluge et al., 1991; 1995]. Additionally, Didiereoideae consumption fits the isotopic signal of Hadropithecus better than C₄ plants in southern Madagascar. Whereas Hadropithecus and Alluaudia (a genus of Didiereoideae) both have high δ¹⁵N values, grasses have low δ¹⁵N values [Crowley & Godfrey, 2013]. However, in the Central Highlands of Madagascar (hereafter CH) where Didiereoideae do not exist, isotope values for Hadropithecus are strikingly different from those from sites in the dry south—the Spiny Thicket (ST) and Succulent Woodland (SW) ecoregions.

Here we expand our sample of isotope data for giant lemurs and other subfossil species from the CH, and we present new comparative analyses. We do so with two primary objectives. First, we seek to better understand how Hadropithecus differs from other species in its stable isotope values within and across ecoregions. We focus particularly on how individuals from the CH differ from individuals in the ST and SW. We then pool our samples for each genus across ecoregions, and compare Hadropithecus to (1) other extinct lemurs, and (2) extinct hippopotamuses and elephant birds. To better understand isotopic variability, we also compare Hadropithecus to Microcebus (mouse lemurs), which is a well-studied extant lemur genus with a large available stable isotope database [Crowley et al., 2011b, 2013, 2014]. Our second primary objective is to explain geographic variation in the δ¹³C values of Hadropithecus. We expect δ¹³C values in bone to be influenced by diet, and thus to be strongly correlated with the isotope values of consumed plants; if this is the case, we can use the pattern of correlation between plant and Hadropithecus δ¹³C values to develop a list of candidate plants for consumption by this species. We also expect plant carbon isotope values to be influenced by soil salinity, light, moisture, and temperature [Amundson et al., 2003; Crowley et al., 2011b; van der Merwe & Medina, 1991], all of which may be correlated with elevation and latitude. Our goal is to determine which environmental variables best explain geographical variation in the carbon isotope values for Hadropithecus, and to identify candidate food plants that are (1) eaten by living lemurs, and (2) have δ¹³C values that vary in a manner that is consistent with geographic isotopic variation for Hadropithecus.

METHODS

We compiled previously published collagen δ¹³C and δ¹⁵N isotope data for bones from giant lemurs and hippopotamuses, and collagen δ¹³C values for bones and eggshells from elephant birds from the ST and SW ecoregions [Berger et al., 1975; Burney, 1999; Burney et al., 2004; Clarke et al., 2006; Crowley et al., 2012; Crowley & Godfrey, 2013; MacPhee, 1986]. To these data we added new isotope data from subfossil lemurs, hippopotamuses (Hippopotamidae), and elephant birds (Aepyornithidae), concentrating on specimens from the CH (Table I). We also assembled published stable isotope data for extant Microcebus fur from the ST, SW, and CH [Crowley et al., 2011b, 2013, 2014].

We included in our study all extinct lemur genera represented by 10 or more individuals as well as other extinct Madagascan megafauna with comparable samples. Our final dataset comprises δ¹³C and associated δ¹⁵N values (when available) for Hadropithecus (n = 11), Archaeolemur (n = 31), Megaladapis (n = 25), Pachylemur (n = 27), Palaeopropithecus (n = 35), hippopotamuses (n = 20), elephant birds (n = 10), and Microcebus (n = 463). We excluded Mesopropithecus spp., Daubentonia robusta, and Archaeoindris fontoynontii because of small sample sizes. Our Hadropithecus sample comes from five subfossil sites: Ampasambazimba (CH), Ankilibehandry (SW), Tsirave (SW), Anavoha (ST), and Andrahomana (ST) (see Fig. 2). Because the isotope data for Microcebus were collected from fur samples, we converted the fur values to look like subfossil collagen before comparing them to values of subfossil lemurs by (1) adding +0.9‰ to carbon and +0.8‰ to nitrogen isotope values (to allow for the average apparent enrichment between these two tissues in primates [Crowley et al., 2010]); and then (2) adding an additional +1.2‰ to fur carbon isotope values (to allow for the isotopic changes in atmospheric CO₂ resulting from the burning of fossil fuels in the Southern Hemisphere [Keeling et al., 2010]).

New subfossil isotope data (Table I) were generated following Crowley et al. [2011a]. We isolated collagen from the subfossil bone samples by decalcifying approximately 200 mg of each fragmented sample in 0.5 M EDTA for 10 days at 4 °C. We replaced the EDTA, sonicated the samples, and allowed them to sit an additional 10+ days at room temperature. We then rinsed the samples ten times using ultrapure water. We gelatinized the samples in 0.01 N HCl at 70 °C for 15 hr, filtered them using Whatman 1.5 μm glass fiber filters, and dried them under vacuum. Following sample treatment, we weighed 0.7 mg of collagenous residue into tin boats, combusted them, and analyzed their δ¹³C and δ¹⁵N values on a ThermoElectron (Finnigan) Delta-XP continuous flow system connected to an Elemental Analyzer at the University of California, Santa Cruz Stable Isotope Laboratory. Collagen preservation was evaluated using sample yield, isotope values, and elemental ratios [Ambrose, 1990].

We had previously compiled from the literature and from Google Earth a database that included latitude, longitude, and elevation for subfossil sites in Madagascar [Muldoon & Godfrey, 2013]. Using these data, we were able to extract estimated monthly site-specific rainfall and temperature data for the past 100 years from the online WorldClim database [Hijmans et al., 2005]. We selected January and July values for temperature and rainfall to represent austral summer and winter extremes, respectively. We also coded each site for soil salinity on the basis of its proximity to the coast (coastal sites are more saline than inland sites) using a simple binary code (coastal vs. inland). Finally, we compiled δ¹³C values from the literature for both C₃ and CAM plants from localities in the vicinity of Hadropithecus subfossil sites (Table II).

Plant isotope data were derived from multiple plant parts (leaves, fruits, and seeds). All plant δ¹³C values were adjusted for fractionation between herbivore collagen and diet (+5‰) [Ambrose and Norr, 1993; Crowley et al., 2011b; Kellner & Schoeninger, 2007; Koch et al., 1991; Sullivan & Krueger, 1981; van der Merwe, 1989; Vogel, 1978] and for changes in atmospheric CO₂ (+1.2‰) [Keeling et al., 2010]. Geographic distributions of candidate plants that may have been consumed by Hadropithecus were assessed for overlap with Hadropithecus sites using specimen geographic searches in www.tropicos.org.

To analyze data, we applied standard statistical tests (t-tests and ANOVAs with post hoc tests of honestly significant differences) using SPSS 22. We used correlation and regression analysis to examine the relationships between Hadropithecus bone δ¹³C values and the following characteristics of the specimen's geographic location: latitude, longitude, elevation, proximity to the coast, mean January and mean July rainfall and temperature, and regional δ¹³C values for C₃ and CAM plants.

This research adhered to the American Society of Primatologists' principles for the ethical treatment of primates.

RESULTS

Table III compares δ¹³C and δ¹⁵N values among giant lemur taxa living in different ecoregions. We pooled isotope values for the SW and the ST because they are not significantly different; thus, we compare values for the CH to a larger sample from arid and semi-arid habitats. Both δ¹³C and δ¹⁵N values are lower in the CH than in the SW/ST, almost always significantly so. This applies to individual genera, taken alone, and to a pooled sample of all giant lemurs.

Looking only at extinct lemurs from the CH, neither average δ¹³C values nor average δ¹⁵N values differ significantly among genera (Table III). Hadropithecus lies at the low end of the spectra for both δ¹³C and δ¹⁵N. In contrast, in the SW/ST, there are strongly significant differences in both δ¹³C values and δ¹⁵N values among lemur genera, and Hadropithecus has the highest values for both isotopes. Its δ¹³C values differ significantly from those of each of the other giant lemurs, and its δ¹⁵N values differ significantly from all other lemurs except Palaeopropithecus (Table III). The variance for δ¹³C values for Hadropithecus is high in the CH (standard deviation, or SD = 4.1‰) but unexceptional in the SW/ST (SD = 1.5‰).

Table IV compares carbon and nitrogen isotope values for lemur genera without regard to ecoregion. ANOVAs demonstrate significant differences among genera for both δ¹³C and δ¹⁵N values. Once again, Hadropithecus is unusual not merely for its high mean δ¹³C value, but also for its strikingly high δ¹³C variance (SD = 6.2‰). This exceptionally high variance is a consequence of dramatic inter-regional differences, coupled with high variance within the CH itself. Variance for δ¹⁵N is also high for Hadropithecus (SD = 4.8‰), though to a lesser degree.

Isotopic variance is exceptional in Hadropithecus despite the fact that our samples for each of the other giant lemur genera comprise multiple recognized species, while only one species is recognized for Hadropithecus. Data derived from a well-studied extant lemur genus, Microcebus, underscore the unusual nature of isotopic variation in Hadropithecus. In contrast to Hadropithecus, standard deviations of mouse lemur δ¹³C and δ¹⁵N from the CH, ST, and SW are much closer to those of other extinct lemurs (1.5‰ and 2.3‰ respectively).

Table V compares Hadropithecus to hippopotamuses and elephant birds from the ST, SW and CH. Mean δ¹³C and δ¹⁵N values for Hadropithecus are significantly higher than for either of the other two taxa. Elevated δ¹³C values for Hadropithecus suggest that this species' diet differed not merely from those of other giant lemurs but also from those of hippopotamuses and elephant birds. Higher δ¹⁵N values for Hadropithecus suggest that this lemur tended to occupy drier habitats than hippopotamuses (i.e., feeding further from bodies of water); nitrogen isotope data for elephant birds are insufficient to allow any conclusions about the habitats they occupied.

Combined, our results suggest that Hadropithecus had a highly unusual diet, and that in the CH it consumed plants that differed dramatically in their isotopic composition from those eaten in the ST or SW. Its isotopic signal can be summarized simply: Hadropithecus displays elevated δ¹³C and δ¹⁵N in arid areas (but not in the much wetter CH) and highly variable isotope values in the CH.

Several (but not all) of the environmental variables that we compiled to explore factors potentially influencing this pattern of variation (i.e., latitude, longitude, elevation, proximity to the coast, January and July mean rainfall and temperature, and regional δ¹³C values for C₃ and CAM plants) are themselves strongly collinear. Across Hadropithecus sites, localities at higher elevation are also significantly closer to the equator. Rainfall is highest during the rainy season (in January, austral summer) and at high elevations. Temperature also changes with season, but is not significantly correlated with elevation, likely because Hadropithecus sites that are closer to the equator are also higher in elevation. Temperature is significantly correlated with longitude (warmer on the west coast than further east, in the CH), both for January and July. The δ¹³C values of CAM plants are negatively correlated with elevation (CAM plants are more enriched in ¹³C at low elevations); there is no relationship between δ¹³C values of C₃ plants and elevation.

Two of our environmental variables are excellent predictors of δ¹³C values for Hadropithecus bone: elevation (r = −0.957, P < 0.001, n = 11) and δ¹³C values for CAM plants (r = 0.923, P < 0.001, n = 11). Two additional variables are significantly (but more weakly) correlated with δ¹³C values for Hadropithecus bone: January rainfall (r = −0.733, P = 0.01, n = 11) and latitude (r = −0.668, P = 0.025, n = 11). No other variables, including δ¹³C values for C₃ plants, are significantly correlated with δ¹³C values for this extinct lemur. The regression with the highest adjusted R² value (0.90) uses elevation and δ¹³C values for CAM plants to predict δ¹³C values for Hadropithecus bone. Its multiple R value is 0.961.

CAM plants that were consumed by Hadropithecus should (1) have δ¹³C values that vary in a manner consistent with geographic isotopic variation for Hadropithecus, (2) exist in the regions formerly occupied by Hadropithecus, and (3) be eaten, albeit perhaps in small quantities, by living lemurs. Table VI lists plant genera that appear to satisfy the first two criteria. Figure 1 demonstrates how dominant consumption of facultative CAM plants such as Kalanchoë spp. could account for the isotopic variation exhibited by Hadropithecus, even in the CH. The δ¹³C values for Kalanchoë in the CH show tremendous variation. Table VII provides observations that support the third criterion. In most cases, the parts eaten are leaves.

DISCUSSION

Isotope data provide strong support for the inference that Hadropithecus specialized on CAM (and not C₃) plants. The difference between CAM and C₃ plants in accounting for variation in the carbon isotope signature of Hadropithecus is impressive, and comprises strong evidence that Hadropithecus preferred CAM plants, even in the CH. CAM plants typically have elevated δ¹³C values in arid regions and can exhibit highly variable δ¹³C values in humid areas [Kluge et al., 1991, 1995]. The variability of δ¹³C values of Hadropithecus is truly unusual, not merely in comparison to other fossil lemurs, but in comparison to a very well measured living lemur, Microcebus. No other lemur, living or extinct, shows the combination of exceptionally high δ¹³C values in arid regions and extreme variability of δ¹³C values in humid regions, as is consistent with dedicated CAM plant consumption.

While our stable isotope values for Hadropithecus do not reveal plant part or parts eaten, there is good reason to believe that the succulent leaves of CAM plants were the dietary staples for Hadropithecus in all ecoregions. Leaves account for the vast majority of CAM consumption by extant lemurs (Table VII). Endemic CAM plants of Madagascar tend to use wind or water to disperse their seeds. Alluaudia and Kalanchoë are prime examples of this, with very tiny seeds adapted for wind dispersal and not animal consumption. Some endemic CAM plants, including all members of the Didiereoidea, have spines to protect their leaves, which strongly suggests that these plants are (or were) subjected to significant leaf predation.

There are also reasons to exclude a predominantly C₄ diet for Hadropithecus. Unlike C₄ plants, succulent CAM plants are abundant in the southern parts of the island and can provide food and water during the prolonged dry season, when grass leaves become desiccated and of little nutritional value. In arid places where Hadropithecus was abundant, isotope values in C₄ plants are less affected by aridity than are C₃ or CAM plants. Consequently, C₄ plants typically have lower δ¹⁵N values than C₃ or CAM plants in arid settings, including southern Madagascar [Crowley & Godfrey, 2013; Koch et al., 1991; Swap et al., 2004]. Therefore, grass consumption cannot explain the very high δ¹⁵N values for Hadropithecus in the arid south.

Minimal reliance by other endemic taxa on C₄ plants in Madagascar is also supported by their δ¹³C values. Neither Madagascan hippopotamuses nor elephant birds have isotope values indicative of a diet rich in C₄ grasses [Clarke et al., 2006; Crowley & Godfrey, 2013]. Bond & Silander [2007] note the existence (albeit scant) of enriched ¹³C for some hippopotamuses reported by Burney et al. [2004]. However, the only hippo δ¹³C values that match dedicated C₄ grass consumption derive from a single individual Hippopotamus whose skull and mandible may have been brought from continental Africa to the Académie Malgache in the 1800s for comparative analysis.

A final argument against C₄ grass consumption being critical to Hadropithecus is that C₄ grasses likely arrived too late to explain the derived suite of traits that distinguish Hadropithecus from its sister taxon, Archaeolemur. Recent phylogenetic research based on complete or nearly complete mitochondrial genome sequences for extinct and extant lemurs suggests that the archaeolemurids diverged from a palaeopropithecid/indriid clade roughly 24 Ma [Kistler et al., 2015]. This establishes an upper limit for the divergence timing of the two archaeolemurid genera. Closely related lemur genera for which divergence estimates are available fall between 10 and 15 million years (∼10.9 Ma for Pachylemur and Varecia, a bit more for Eulemur-Lemur and for Propithecus-Avahi) [Kistler et al., 2015]. A divergence date for Archaeolemur and Hadropithecus between 10 and 20 million years seems reasonable. This is exactly when the common ancestor of the Didiereoideae likely arrived on Madagascar. Estimates for the timing of the origin of the family to which the Didiereoideae belongs (i.e., the Didiereaceae, including the continental African Calyptrothecoideae and Portulacarioideae as well as the Madagascan Didiereoideae; see Bruyns et al., 2014]) vary from 15 to ∼30 Ma [Arakaki et al., 2011; Ocampo & Columbus, 2010]. Arakaki et al. [2011] place the initial diversification of the Madagascan subclade at ∼17 Ma. Ocampo & Columbus [2010] place the divergence of African and Madagascan clades at 12.1 Ma. But even that late estimate substantially predates the global radiation of C₄ grasses [Cerling et al., 1997] and likely post-Miocene spread of C₄ grasses to Madagascar [Bond et al., 2008], which is believed to be related to a decrease in concentrations of atmospheric CO₂ below a threshold that favored C₃ photosynthesis [Cerling et al., 1997].

There is substantial evidence that C₄ grasses expanded into tropical and subtropical latitudes of continental Africa during the Late Neogene [Strömberg, 2011; Uno et al., 2011]. Furthermore, that evidence supports a latitudinal gradient, with C₄ grasses expanding 4–5 million years later in southern Africa than in eastern and central Africa [Segalen et al., 2007; Strömberg, 2011]. If the same latitudinal gradient applies beyond continental Africa, one can expect a relatively late spread of C₄ grasses to Madagascar.

The timing of the diversification of the Madagascan Didiereoideae is critical because spines protecting the leaves from predation likely evolved in the ancestral lineage soon after initial colonization. Spines occur on all extant species of Didiereoideae but not their continental African relatives. Only lemurs and elephant birds were present on Madagascar when the ancestral didiereoid arrived on the island; hippopotamuses would have arrived later. Bond & Silander [2007] ruled out elephant birds on the basis of their feeding anatomy, which makes climbing lemurs the likely candidates. Exploitation by climbing lemurs would explain the evolution of spines on the tallest branches of didiereoid trees such as Alluaudia [Crowley & Godfrey, 2013], and the fact that spine length matches or slightly exceeds the leaves beneath them. Differential exploitation of this new resource might also help to explain the divergence of Archaeolemur and Hadropithecus.

Recent discoveries of some previously unknown postcranial bones of Hadropithecus are fascinating in this light [Godfrey et al., 1997, 2006; Lemelin et al., 2008], as they suggest that Hadropithecus, despite being highly terrestrial, was maladapted for speed but adept at climbing. The femur was slightly shorter than the humerus, but very robust. Its shaft was anteroposteriorly compressed as in slow climbers; its femoral condyles show an asymmetry similar to that of chimpanzees or gorillas, likely reflecting differential weight bearing during femoral rotation on the tibia under abducted femoral excursion. The patellar groove was shallow and wide. This animal was neither a leaper nor suspensory; it was also not cursorial. Polk et al. [2010] measured the density of subchrondral bone in the distal femur to reconstruct knee postures. They found a broad range, from highly flexed to highly extended. That Hadropithecus was slower and more deliberate than Archaeolemur is corroborated by data on the radius of curvature of its semicircular canals [Walker et al., 2008].

Curiously, Hadropithecus also had an unusual manus [Lemelin et al., 2008]. Its short digits suggest cercopithecoid-like terrestriality but differences from terrestrial cercopithecoids are also striking. Papionins in general, and Theropithecus in particular, have a long and mobile pollex and a very short index finger, which help them in securing small objects such as the seeds of grasses, as well as underground storage organs such as rhizomes of grasses or corms of sedges, which become essential resources when grasses are dry [Etter, 1973; Jablonski et al., 2002]. Hadropithecus, in contrast, had an exceptionally short thumb, which suggests poor manipulative skills. The few hand bones known for Hadropithecus include the fifth metacarpal and the wrist bone that articulates with it, the hamate. When fitted together, they form a peculiar angle; this carpometacarpal joint displays a hyperextended set [Lemelin et al., 2008]. This is a very odd adaptation with no known analogue in the world of primates or other mammals. One might imagine that Hadropithecus would have benefited from being able to remove leaves manually if it indeed specialized on the succulent leaves protected externally by long spines. At present we do not know how it would have done this. More discoveries of bones of its hand may help us to understand whether or not Hadropithecus' manual peculiarities relate to a unique feeding adaptation.

With our new interpretation of the diet of Hadropithecus, we can now better evaluate the hypotheses that have been proposed to explain this animal's dental convergences with geladas and largely cranial convergences with basal hominins. Grass blades, rhizomes, and small seeds of grasses are unlikely, despite being mechanically challenging, because the isotope signature does not fit, and because Hadropithecus and Archaeolemur likely diverged before C₄ grasses became widespread in Madagascar [H₁ and, in part, H₂]. There is no evidence that hard foods or large seeds were significant components of the diet of Hadropithecus [H₃]. Large seeds or fruits with hard protective coats cannot have been staples, because they derive virtually exclusively from C₃ plants. Heavy microwear pitting on the molars of Hadropithecus is likely to reflect high grit levels on the foods rather than the consumption of hard objects, as many endemic CAM plants of Madagascar have tiny seeds that are wind dispersed. In arid or relatively open habits, grit levels on leaves are likely to be high, whether the leaves are cropped from plants on the ground or from trees [Green & Kalthoff, 2015; Ungar et al., 1995].

Elements of Jolly's [1970] small object hypothesis [H₂] can be defended in that succulent leaves may require little incisal preparation and not more than a small gape, and may have been consumed in large quantities. Repetitive chewing may have been required. Furthermore, succulent leaves may well have been conveyed by hand to the mouth. It is not clear that the leaves of these succulents are mechanically challenging (they have yet to be systematically tested), but having molar teeth that function in the manner of a grazer is not surprising for an animal masticating large quantities of succulent leaves. High-bulk feeding [H₄] cannot be discounted; this is, in effect, a component of Jolly's small object hypothesis [H₂]. However, while some of our candidate plants do not grow tall, others are trees. Hadropithecus likely accessed its foods via climbing as well as on the ground. Here, Jolly's [1970] gelada analogy fails.

Finally, we stress that determining the diet of Hadropithecus is only the first step of several that must be taken if we are to understand its hominin and papionin likenesses. While our research here contributes importantly to this task, we recognize that more research is needed to trim our long list of candidate plants and eliminate some of them. Nitrogen isotope data, in particular, are scant.

Ultimately, we would like to be able to address the question: What ecological or dietary shift in the ancestral archaeolemurid could have produced the evolutionary shift in gnathic and dental structure that we observe in the Hadropithecus lineage? And why do some (but not all) of its features parallel apomorphies of basal hominins and others of Theropithecus? To address these questions, we must know the diets of the fossils, but that is only the beginning. We must then collect appropriate data on the mechanical properties of likely resources, the mechanical efficiency of the molars, and processes by which foods break down during mastication. In our view, the cranial and dental convergences of Hadropithecus to basal hominins and to grass-eating papionins are too striking to ignore, and understanding their significance could help us to better understand the adaptive significance of the craniodental anatomy of basal hominins and of grass-eating papionins themselves.