Dietary change among hominins and cercopithecids in Ethiopia during the early Pliocene

Significance

Dietary change among hominins is a critical aspect of human evolution. Here we use carbon isotope data from fossil teeth of hominins, monkeys, and other mammals from Ethiopia to document C₄ food consumption by both hominins and the baboon, Theropithecus oswaldi, during the early Pliocene. The expansion of hominin diet and the appearance of the Theropithecus oswaldi lineage as early as 3.76 Ma mark a major ecological change within African primate communities. The ability to eat a range of C₃ and C₄ foods indicates that early Pliocene hominins were likely generalists who could thrive in different and perhaps varying environments.

Abstract

The incorporation of C₄ resources into hominin diet signifies increased dietary breadth within hominins and divergence from the dietary patterns of other great apes. Morphological evidence indicates that hominin diet became increasingly diverse by 4.2 million years ago but may not have included large proportions of C₄ foods until 800 thousand years later, given the available isotopic evidence. Here we use carbon isotope data from early to mid Pliocene hominin and cercopithecid fossils from Woranso-Mille (central Afar, Ethiopia) to constrain the timing of this dietary change and its ecological context. We show that both hominins and some papionins expanded their diets to include C₄ resources as early as 3.76 Ma. Among hominins, this dietary expansion postdates the major dentognathic morphological changes that distinguish Australopithecus from Ardipithecus, but it occurs amid a continuum of adaptations to diets of tougher, harder foods and to committed terrestrial bipedality. In contrast, carbon isotope data from cercopithecids indicate that C₄-dominated diets of the earliest members of the Theropithecus oswaldi lineage preceded the dental specialization for grazing but occurred after they were fully terrestrial. The combined data indicate that the inclusion of C₄ foods in hominin diet occurred as part of broader ecological changes in African primate communities.

The Pliocene is a critical time in human evolution when almost all early hominins became committed terrestrial bipeds and expanded their diets to include a wider range of resources than their ancestors. Recent stable isotope studies of Australopithecus afarensis teeth indicate that hominins had increased their dietary breadth to include significant amounts of C₄ or crassulacean acid metabolism (CAM) resources by 3.4 Ma (1), which is in contrast to its putative ancestor, Australopithecus anamensis, whose diet was limited to predominantly C₃ foods (2). The expansion in hominin diets to include significant amounts of C₄ and CAM resources indicates a transition toward more open-country foods, because C₃ plants include trees, shrubs, forbs, and cool-growing season grasses, whereas C₄ plants are primarily warm-growing season grasses and sedges, and CAM plants include cacti and succulents (3). This dietary expansion may have made it easier for hominins to survive in a greater range of environments or in environments that were more variable. However, we do not know when hominins started to include large proportions of C₄ resources in their diets, nor do we fully understand the ecological context of these changes (1, 4).

We use stable carbon isotope ratios of fossil hominin and cercopithecid teeth from Woranso-Mille in the western part of the central Afar Rift in Ethiopia (Fig. S1) to refine the timing of hominin dietary expansion to include C₄ resources. We also report isotope data from other mammals and soil carbonates to evaluate the environmental context and the diagenetic integrity of the isotope data. We use stable carbon isotope ratios, or δ¹³C values, of fossil teeth to evaluate the dietary proportion of plants that use the C₃ vs. C₄ photosynthetic pathway, based on the premise that C₃ and C₄ plants have distinct carbon isotope signatures and that the δ¹³C value of tooth enamel reflects the carbon isotope composition of an animal’s diet (5). Fossil teeth from the Woranso-Mille paleontological study area are well-suited to fill the temporal gap in the isotopic record of hominin diet because they are part of a record of Pliocene mammalian fossils that spans 3.76–3.2 Ma (6–11). The hominin fossils from Woranso-Mille include those that are morphologically intermediate between Au. anamensis and Au. afarensis, some that are definitively Au. afarensis, and others that represent additional species (7–9, 12). The cercopithecids include multiple species of colobines and at least two papionins (10). Theropithecus oswaldi cf. darti is the most common cercopithecid in the assemblages (>90% of identifiable cercopithecid specimens; 40% of the total identifiable mammal specimens) (6, 10); it represents the oldest and most primitive representative of the long-lasting T. oswaldi lineage whose morphology became increasingly specialized for grazing throughout the Pliocene and into the Pleistocene (13). The carbon isotope data from cercopithecids sympatric to hominins at Woranso-Mille make it possible to evaluate the ecological context for dietary changes in hominins.

Fig. S1.

Map with locations of sites discussed in text and listed in Datasets S2 and S5. Woranso-Mille, Middle Awash (Aramis, Asa Issie, Asa Koma), Dikika, Gona (Segala Noumou), and Hadar and are in the Afar region in Ethiopia. Lomekwi, Allia Bay, and Kanapoi are in the Omo-Turkana region of Kenya. Imagery is from NASA's Earth Observatory from August 2004 (72).

Results

We report carbon and oxygen stable isotope data from 152 fossil teeth of hominins, cercopithecids, bovids, giraffids, hippopotamids, suids, rhinoceratids, and proboscidea from three time intervals at Woranso-Mille (3.76–3.57 Ma, 3.57–3.47 Ma, and 3.47–3.2 Ma; see Supporting Information for additional context). Among these teeth, δ¹³C values range from −14.0 to +1.1‰ and represent the full spectrum of δ¹³C values expected for tooth enamel from mammals with diets of exclusively C₃ and exclusively C₄ resources (Fig. 1 and Dataset S1). Presumed browsers such as giraffids and colobines yield δ¹³C values at the lower end of this range, whereas δ¹³C values of presumed grazers including alcelaphini, Loxodonta, Anancus, and Notochoerus euilus are at the high end of this range. These distributions suggest that the carbon isotope data from the fossil teeth at Woranso-Mille preserve dietary information and are not products of diagenesis. There are no significant differences in δ¹³C values among any of the mammalian taxa from the different time intervals sampled at Woranso-Mille (P > 0.05), such that we consider the fossil δ¹³C data from all of the time intervals together.

Fig. 1.

Box and whisker plots of δ¹³C (A) and δ¹⁸O (B) values for the fossil tooth enamel data from Woranso-Mille. Median values are marked by a vertical line within the box, the edges of the boxes represent quartile values, the horizontal lines indicate the range, and outliers are plotted as circles.

Hominins.

Sampling was limited to fragmentary hominin teeth whose genus and species cannot be determined with certainty; the sample may include intermediate forms of the Au. anamensis–Au. afarensis lineage, Au. afarensis, Australopithecus deyiremeda, and a yet unnamed species represented by the BRT-VP-2/73 foot (9, 12). The Woranso-Mille hominin teeth yield a mean δ¹³C of −7.9 ± 2.5‰ (1σ) and range from −11.8 to −3.8‰ (n = 16, median −7.5‰). The δ¹³C values of hominin teeth from Woranso-Mille are indistinguishable from that of non-Theropithecus papionins, Tragelaphus, sivatheres, and hippopatimds (P > 0.1), but distinct from the δ¹³C values of all other taxonomic groups sampled (P < 0.001). Notably, δ¹³C values of hominins at Woranso-Mille are significantly higher than δ¹³C values of sympatric colobines and lower than δ¹³C values of sympatric Theropithecus individuals (Fig. 1). The range in δ¹³C values among hominins at Woranso-Mille indicates that some individuals had diets dominated by C₃ foods, whereas other individuals had a significant component of C₄ resources in their diet. The observed variation in δ¹³C values among hominins at Woranso-Mille may reflect dietary differences among hominin species; however, the fragmentary nature of the sampled teeth precludes us from connecting the stable isotope results to specific hominin taxa.

The range in δ¹³C values of the Woranso-Mille hominins overlaps with those from older hominins, such as Ardipithecus ramidus (median −10.3‰, range −11.2‰ to −8.5‰, n = 5) (14) and Au. anamensis (median −11.2‰, −12.0‰ to −9.5‰, n = 12) (2) (Fig. 2 and Dataset S2). The distribution of δ¹³C values of the Woranso-Mille hominins are significantly more positive than δ¹³C values of Au. anamensis (P = 0.002). The δ¹³C values of the Woranso-Mille hominins are indistinct from the majority of younger Pliocene hominins, including Au. afarensis (1), Australopithecus africanus (4, 15–18), and Kenyanthropus platyops (2), with the exception of two individuals of Australopithecus bahrelghazali that incorporated significantly more C₄ resources in their diet (19) and Australopithecus sediba, whose diet was restricted to C₃ foods (20). The Woranso-Mille hominins also have substantially more C₄ foods in their diets than the extant primates for which carbon isotope data from tooth enamel are available (Dataset S3), but we note that these data have mostly been generated from animals living in forests and do not represent the full spectrum of environments in Africa today. In summary, the δ¹³C values from the Woranso-Mille hominins indicate that C₄ foods were a significant component to hominin diets as early as 3.76 Ma.

Fig. 2.

The δ¹³C values of African fossil hominin (A) and cercopithecid (B) tooth enamel and (C) soil carbonate and leaf wax C₃₀-acid and (D) the percentage of grass pollen from eastern Africa. Published fossil hominin and cercopithecid δ¹³C data are from refs. 1, 2, 4, 14, 16–29, and 31–34. Soil carbonate δ¹³C values are plotted as means for 100-ky age bins to view long-term trends but are also plotted for individual soils from the Omo-Turkana, Afar, and Laetoli regions (9, 37, 41, 42, 56–65); ages are assigned as in original publication or using the assignment provided in ref. 66. Mean leaf wax δ¹³C values (C₃₀ n-alkanoic acids) for 100-ky bins and percentage of grass pollen from the Deep Sea Drilling Project Site 231 are from refs. 38 and 40. Fossil ages are plotted as midpoints if the age constraint is a range of dates. In A−C, more positive δ¹³C values indicate an increase in the contribution of C₄ vegetation. In C, error bars are 1σ on the mean.

Cercopithecids.

The δ¹³C values of the cercopithecids sampled from Woranso-Mille represent diets that ranged in proportions of C₃ and C₄ resource use (Figs. 1 and 2). The δ¹³C values of colobines average −11.9 ± 1.3‰ (n = 9) and represent diets of either pure C₃ vegetation or with minor amounts of C₄ plants and/or water-stressed C₃ vegetation. The δ¹³C values of T. oswaldi cf. darti sampled from Woranso Mille average −3.4 ± 1.8‰ (n = 39) and represent diets that are compatible with the inclusion of a large proportion of C₄ foods. The δ¹³C values from the other papionin samples, referred to as non-Theropithecus papionins henceforth, average −8.8 ± 3.9‰ and range from −12.0‰ to −2.6‰ (n = 9), reflective of diets that included little to no C₄ vegetation for some animals and a majority of C₄ vegetation for others.

Collectively, the T. oswaldi cf. darti teeth yield the highest δ¹³C values among all primates sampled from Woranso-Mille. They are distinct from δ¹³C values of the non-Theropithecus papionins, colobines, and hominins and from δ¹³C values of all other nonprimate taxa (P < 0.001), except for hippopotamids. The δ¹³C values of T. oswaldi cf. darti from Woranso-Mille are indistinguishable from the oldest Theropithecus sampled from the Turkana Basin, Theropithecus brumpti (21), but significantly lower than δ¹³C values from T. oswaldi that are younger than 2.0 Ma in Turkana, from Olorgesailie in Kenya, and from Swartkrans, Makapansgat, and Sterkfontein in South Africa (18, 21–24) (Fig. 2). All fossil Theropithecus δ¹³C values are significantly higher than the carbon isotope data available for extant primates in Africa (Dataset S3). Our data are consistent with studies from several sites in South Africa that show that Theropithecus consumed more C₄ vegetation than other cercopithecids and contemporaneous hominins (Fig. 2 and Fig. S2) (18, 22–24).

Fig. S2.

Body size vs. δ¹³C values of tooth enamel for fossil assemblages with isotopic data from multiple primate taxa for (A) Aramis, (B) Woranso-Mille 3.76−3.57 Ma, (C) Waranso-Mille 3.47−3.2 Ma, (D) Makapansgat, (E) Sterkfontein, and (F) Swartkrans. Size is plotted as the midpoint between the mean estimated body mass for females and males and is used here as a way to represent the range of physiologies among primates at a site. Error bars for body size represent the midpoint between male and female body mass. If error bars are missing, then data represent the only data available, which is either the mean value for males or females or an average estimate for the species. The body mass of the Woranso-Mille hominins is plotted both as the full estimated range for early Pliocene hominins and also as a more restricted range for Au. afarensis, both from ref. 73. The δ¹³C values are plotted as the means of the available data in reference to VPDB; error bars represent 1σ on the mean but are missing if n = 1. See Dataset S1 for the δ¹³C data from Woranso-Mille, Dataset S2 for the δ¹³C data from other sites, and Dataset S6 for body mass data.

The δ¹³C data from colobines come from at least two species at Woranso-Mille: Cercopithecoides cf. meaveae and one or more larger colobines. We do not have enough δ¹³C data from each taxon to detect differences among them, but they are all consistent with diets dominated by C₃ vegetation, with δ¹³C values that are indistinct from the non-Theropithecus papionins and Giraffa sp. at Woranso-Mille (P > 0.1) but significantly lower than δ¹³C values of Theropithecus, hominins, and the nonprimate mammals at Woranso-Mille (Fig. 1). The δ¹³C values from colobines at Woranso-Mille are indistinct from δ¹³C values published from fossil colobines at other localities in Ethiopia and South Africa (Fig. 2) (14, 22, 25, 26); however, they are significantly higher than carbon isotope data from extant colobines (Dataset S3), which might indicate either a small component of C₄ in their diet, a difference in digestive physiology, or that the C₃ resources they consumed had higher δ¹³C values than the plants consumed by extant primates in tropical evergreen and rainforests.

The δ¹³C values of non-Theropithecus papionins at Woranso-Mille overlap with δ¹³C values of hominins, colobines, giraffids, and Tragelaphus but are significantly lower than those of T. oswaldi cf. darti, the non-Tragelaphus bovids, hippopotamids, N. euilus, and Proboscidea. Among the nine teeth sampled, six yield δ¹³C values < −9‰ (−11.4 ± 1.0‰, n = 6) and overlap with the colobine δ¹³C data; however, three of the teeth (all from the 3.76–3.57 Ma interval) yield distinctively higher δ¹³C values (−3.7 ± 1.0‰), indicating diets with significant amounts of C₄ vegetation. There is nothing about the size, morphology, or anatomical position of these teeth that set them apart; as such, we interpret these data to indicate that either (i) some non-Theropithecus papionins overlapped in their dietary behavior with Theropithecus and consumed large amounts of C₄ vegetation or (ii) they are early Theropithecus that used C₄ resources but with tooth morphology that is not distinct enough to characterize them as Theropithecus. The δ¹³C values of the Woranso-Mille non-Theropithecus papionins are not significantly different from δ¹³C of other fossil non-Theropithecus papionins from Ethiopia, Tanzania, and South Africa (14, 18, 22–29) (Fig. 2).

Tooth Enamel Oxygen Isotope Data.

The δ¹⁸O values of teeth samples from Woranso-Mille range from −9.0‰ to +6.3‰ (Fig. 1) and exhibit fewer taxonomic distinctions than observed for δ¹³C. The majority of taxa have overlapping δ¹⁸O values, with the exception of the hippopotamid teeth, which are significantly lower than δ¹⁸O values of all other taxa sampled from the 3.76–3.57 Ma time interval (P < 0.02), except for Proboscidea, N. euilus, and sivatheres. We did not make this comparison for the other time intervals, as we only sampled hippopotamids from the 3.76–3.57 Ma time interval. However, among the samples from the youngest time interval (3.47–3.2 Ma), there are no distinctions in δ¹⁸O among taxa, except for δ¹⁸O of sivatheres, which are significantly higher than those of Theropithcus and N. euilus. There are not enough samples from the middle time interval (3.57–3.47 Ma) to make meaningful comparisons. Among taxa that were sampled from multiple time intervals (all of the primates, giraffids, Tragelaphus, and N. euilus), there are no temporal distinctions in the distributions of δ¹⁸O values, with the exception of Theropithecus and N. euilus, which exhibit an increase in mean δ¹⁸O values of 3.0‰ and 3.9‰, respectively, between the 3.76–3.57 Ma and 3.47–3.2 Ma sampling intervals. The total range in δ¹⁸O values among hominins (8.1‰) is similar to that for sympatric primates (6.5–7.8‰) and other mammals (5.0–11.0‰) at Woranso-Mille for which more than five teeth were sampled.

Absolute δ¹⁸O values of fossil teeth are difficult to compare between fossil localities (30); however, the low δ¹⁸O values of the hippopotamid teeth relative to most other taxa at Woranso-Mille are consistent with observations from other fossil sites in the Afar, including those with oxygen isotope data from hominins (1, 14). The offset between the giraffid and hippopotamid δ¹⁸O values in the 3.76–3.57 Ma aged sediments at Woranso-Mille is 3.3 ± 2.4‰, which can be compared with the δ¹⁸O giraffid−hippopotamid offset observed among fauna associated with Ar. ramidus at Aramis (9.0 ± 3.0‰) (14) and with Au. afarensis at Hadar (5.4 ± 4.0‰) (1). The oxygen isotope data may indicate a slightly wetter environment in Woranso-Mille at 3.76–3.57 Ma that at Hadar (ca. 3.4 Ma) or Aramis (4.4 Ma), assuming the relationship between aridity and the offset in δ¹⁸O values of giraffids and hippopotamids in eastern Africa today (30) applies to the Pliocene.

Soil Carbonates.

Soil carbonates from hominin fossil-bearing sequences at Woranso-Mille yield mean δ¹³C values of −6.1 ± 0.6‰ (n = 8) at Aralee Issie (3.76–3.57 Ma) and −9.2 ± 4.0‰ (n = 6) from Burtele and Nefuraytu (3.47–3.2 Ma) (Fig. S3 and Dataset S4). This distribution is similar to δ¹³C values of soil carbonates from Au. anamensis-, K. platyops-, and Au. afarensis-bearing sequences (Fig. S3 and Dataset S5). The δ¹⁸O values of soil carbonates at Woranso-Mille (−6.5 ± 0.9‰ at Aralee Issie and −3.7 ± 2.7‰ at Burtele and Nefuraytu) are within the range of δ¹⁸O values for soil carbonates from other Pliocene hominin-bearing sequences in eastern Africa (Fig. S3).

Fig. S3.

Distribution of δ¹³C (A) and δ¹⁸O (B) values of soil carbonates from fossil hominin-bearing strata at Worano-Mille and from other early to mid Pliocene sedimentary sequences that are associated with hominin fossils. Data from the late Miocene sediments bearing Ardipithecus kadabba fossils are also included for reference. See Datasets S4 and S5 for data and references. In these box plots, median values are marked by a vertical line within the box, the edges of the boxes represent quartile values, the horizontal lines indicate the range, and circles represent either single points (n = 1) or outlier values that are more than 1.5 times the interquartile distance.

Discussion

C₄ Resource Use Among Pliocene Hominins and Cercopithecids.

The carbon isotope data from Woranso-Mille indicate that hominins included significant amounts of C₄ resources in their diets as early as 3.76 Ma (Fig. 2). In this regard, δ¹³C values of the Woranso-Mille hominins indicate diets that are distinct from those of preceding taxa, Au. anamensis and Ar. ramidus, but similar to those of the majority of younger hominins, including Au. afarensis, K. platyops, and Au. africanus, although not to Au. bahrelghazali or Au. sediba (1, 2, 4, 14–20, 29, 31–34).

The expansion of hominin diets in the early Pliocene appears to be contemporaneous with the appearance of the first recognizable forms of Theropithecus (10, 35, 36). There are two well-known lineages of the genus, T. brumpti, largely restricted to the Turkana Basin, and the much more cosmopolitan T. oswaldi; both first appear ca. 3.8–3.6 Ma (10). The δ¹³C values for the earliest T. brumpti show that it incorporated a large component of C₄ resources in its diet (21). The δ¹³C values from Woranso-Mille extend this record of early C₄ consumption to the T. oswaldi lineage. The C₄-dominated diets of Theropithecus are unique among the cercopithecids; colobines maintain C₃-dominated diets through the Pliocene, and non-Theropithecus papionins always maintained a smaller component of C₄ in their diet than sympatric populations of Theropithecus (Fig. 2, Fig. S2, and Dataset S6). It is notable that δ¹³C values of hominin teeth are consistently higher than δ¹³C values of colobines throughout the Pliocene, indicating that (i) hominin and colobine diets were always distinct, (ii) hominins and colobines had different carbon isotope diet−tissue fractionation factors, reflecting differences in their physiologies, or (iii) hominins consumed some C₄ foods since 4.4 Ma.

Environmental Change.

Environmental context is an important part to understanding the evolutionary implications of the dietary expansion among hominins and Theropithecus to include C₄ resources; however, it has been difficult to establish definitive links between diet and environment for fossil primates (4, 21). The incorporation of significant amounts of C₄ resources in primate diets in the early Pliocene occurred amid a gradual trend of increasing C₄ vegetation in Africa that began in the late Miocene (37–39) (Fig. 2). This increase in C₄ vegetation is generally viewed as a decrease in woody cover (37), but pollen records indicate that C₃ grasses were present throughout the Pliocene and into the Pleistocene (38, 40). Pollen abundances and δ¹³C values of leaf waxes and soil carbonates indicate large proportions of grasses and C₄ plants in eastern Africa in the earliest Pliocene (5.0–4.0 Ma), with decreased amounts of C₄ vegetation and grasses by 4.0 Ma, followed by a small rise in C₄ vegetation between 3.9 and 3.7 Ma (37, 38, 40–42), which is coincident with the expansion of primate diets to include significant amounts of C₄ resources. Although the coincidence makes it tempting to link the C₄ dietary expansion among hominins to environmental factors, there are reasons to be cautious: (i) the soil carbonate records are sparse from this time interval, and (ii) when δ¹³C values of soil carbonates from hominin-bearing strata are compared directly (Fig. S3), there is no clear connection between environmental C₄ abundance and C₄-dominated diets among hominins. The oxygen isotope data from fossil giraffid and hippopotamid teeth may indicate wetter environments at Woranso-Mille than at both earlier and later hominin sites, but this could reflect local variation in water availability that is typical in eastern Africa today and does not necessitate any major climate change coincident with the inclusion of significant amounts of C₄ resources in hominin or Theropithecus diets. Given the collective data that are available, we do not see strong environmental trends that can be linked confidently to the C₄ dietary expansion among hominins and Theropithecus.

Dietary Breadth in Primate Communities.

The combination of the expanded breadth in hominin diet and the appearance of the T. oswaldi lineage as early as 3.76 Ma marks a major change in African primate communities. The appearance of Theropithecus is important to hominin paleoecology because Theropithecus was a large fraction of the identifiable mammalian fossils at Woranso-Mille (10) and was predominant in nearly all eastern African mammalian communities from 3.5 Ma to 1.0 Ma (36). The increased range in δ¹³C values of both hominins and cercopithecids after 3.76 Ma indicates that primate dietary breadth increased with the appearance of Theropithecus, which ultimately became a committed grazer (21). The use of both C₃ and C₄ foods by hominins likely reflects that they were generalists who could thrive in different environments, eating foods of varying quality and type (1, 4, 43). The inclusion of significant amounts of C₄ foods among hominins and papionins indicates niche broadening that appears unique among primates in the early Pliocene. We are not aware of significant dietary changes in other African mammalian orders at this time; Artiodactyla, Perissodactyla, and Probsocidea started to consume significant amounts of C₄ resources several million years earlier, in the late Miocene (39).

Relationship Between Diet and Morphology.

The craniodental features characterizing australopiths are usually interpreted as a character complex associated with adaptations for hard-object, tougher, and more abrasive food resources (44, 45), although other interpretations such as those based on dental microstructure may not corroborate the former interpretation (46). The first appearance of these features is interpreted as an adaptation to eating a broader array of resources than hominins did previously (47, 48). However, the niche expansion indicated by dentognathic morphology at 4.2 Ma is not reflected in carbon isotope data (2, 4, 14). Intake of large amounts of C₄ foods as early as 3.76 Ma may represent a subsequent dietary change, wherein hominins further increased their dietary breadth (Fig. 2). Increasing enamel thickness, megadontia, and mandibular robusticity among Pliocene hominins is consistent with the consumption of more C₄ food resources, which might have included underground storage organs (i.e., rhizomes, corms) (1, 49, 50). The long-term relationships between dentognathic morphology and C₄ dietary intake is exhibited by correlations between δ¹³C values and both postcanine and mandibular size for australopithecines (4); however, morphological changes are incremental and do not reflect the stepwise increase in C₄ resource consumption. We note that microwear data from Au. anamensis and Au. afarensis teeth are similar to one another and are not consistent with morphological changes that are often associated with increased specialization for hard-object feeding; instead, they are consistent with diets of soft and tough foods (51).

Among papionins, adaptations to terrestrial behavior were already apparent by the late Miocene (52), and all evidence suggests that the earliest Theropithecus was fully terrestrial (10, 36, 53, 54). The main morphological changes that occurred in Theropithecus 3.8–1.0 Ma include increasing body size and the dentognathic specialization for grazing (36, 53, 55), but the evidence for C₄-dominated diets among the earliest Theropithecus fossils indicates that the grazing behavior precedes most of the dental specializations for eating grass, although we recognize that Theropithecus’ initial use of C₄ grass may have focused on the nonblade part of grass.

Conclusions

Stable carbon isotope data from Woranso-Mille show that both hominins and Theropithecus included large proportions of C₄ foods in their diets as early as 3.76 Ma. This increased use of C₄ foods among both hominins and Theropithecus represents an expansion of dietary niche among primates to include both C₃ and C₄ resources that became typical in subsequent fossil primate communities for which there are isotopic data (Fig. S2). The C₄ dietary expansion was relatively late among primates, as other African mammalian orders had been consuming large amounts of C₄ foods since the late Miocene (39). For Theropithecus, the reliance on C₄ foods represents the initiation of its dietary specialization and its role as an abundant and widespread grazer within African mammalian communities (13). The evidence for C₄-dominated diets among the earliest identified specimens of Theropithecus precedes morphological specializations for grazing and provides a clear example of behavioral change preceding morphological adaptation. For hominins, the isotopic evidence for dietary expansion postdates the beginning of major morphological dentognathic changes among australopiths (47, 48) and, as such, it represents a subsequent phase of increasing dietary breadth. The unequivocal incorporation of C₄ resources by hominins and cercopithecids as early as 3.76 Ma demonstrates that primates were expanding the types of food resources and potentially the environments that they were able to exploit.

Methods

Fossil teeth were sampled in the National Museum of Ethiopia in Addis Ababa. Teeth were cleaned with gentle abrasion before sampling. Hominin teeth were sampled along previously broken surfaces. Enamel powders were treated with 10% (vol/vol) H₂O₂ and 0.1 M buffered acetic acid solution for 15 min each, rinsed with deionized water three times after each treatment, and dried at 60 °C overnight before analysis. Soil carbonates were sampled from distinct pedogenic carbonate zones ≥ 50 cm below the contact with the overlying unit. After digestion in phosphoric acid, ¹³C/¹²C and ¹⁸O/¹⁶O ratios of enamel and carbonate powders were measured on a Thermo MAT 253 isotope ratio mass spectrometer at Johns Hopkins University. External precision (1σ) of δ¹³C and δ¹⁸O values of the working fossil tooth enamel standards over the course of analysis was 0.3‰ and 0.2‰ for δ¹³C and δ¹⁸O, respectively. Isotope data are reported relative to the Vienna Pee Dee Belemnite standard. Statistics were performed in Kaleidograph using an unpaired Wilcoxon−Mann−Whitney Rank Sum Test.

SI Methods

Geologic Context and Age Constraints for Woranso-Mille Fossils.

Although individual fossils can be placed in more tightly constrained time intervals, for the purposes of this study, fossils were placed in three stratigraphic groupings. The oldest interval is bound by two radiometrically dated tephra, the Mille tuff (3.76 ± 0.02 Ma) and the Kilaytoli tuff (KT, 3.57 ± 0.014 Ma). The middle interval is bound by the KT at the base and the Burtele Tuff (3.469 ± 0.008 Ma) at its top (7, 9, 67). The youngest time interval is bound by the Burtele Tuff at its base and extends to 3.2 Ma (constrained by sedimentation rates) (9). Localities in the lower age bin (3.76–3.57 Ma) include ARI-VP-1, ARI-VP-2, ARI-VP-3, ARI-VP-5, KSD-VP-1, MKM-VP-1, MSD-VP-1, MSD-VP-2, MSD-VP-3, MSD-VP-5, MSD-VP-8, and MSD-VP-9. Localities included in the middle age bin (3.57–3.47 Ma) are BRT-VP-3 and KSD-VP-1. Localities included in the upper, younger age bin (3.47–3.2 Ma) are BKI-VP-2, BRT-VP-1, BRT-VP-2, LAD-VP-1, LAD-VP-2, LDD-VP-1, and NFR-VP-1. Note that specimens from KSD-VP-1 are included in either the middle or lower age bin depending on their position relative to a fault that cuts through this locality.

Sampling and Analytical Methods.

Fossil teeth and soil carbonates were sampled as described in Methods. To avoid sampling multiple teeth from one individual, we targeted the third molar where possible and did not sample teeth from the same taxonomic group if the fossils were recovered in close proximity to one another. Tooth enamel and soil carbonate powders were reacted with 100% H₃PO₄ under vacuum at 90 °C using a common acid bath device described by ref. 68. The resultant, purified CO₂ was then transferred to the microvolume of dual inlet system on a Thermo MAT 253 isotope ratio mass spectrometer at Johns Hopkins University for determination of ¹³C/¹²C and ¹⁸O/¹⁶O ratios. CO₂ was measured in reference to a working reference gas calibrated during each run to NBS-19 and calcite and bioapatite working reference materials. The CO₂ yields of enamel samples indicate that they contained 4–9% carbonate and did not contain large amounts of exogenous carbonate. An acid fractionation factor of 1.00725 (90 °C) was used for δ¹⁸O. All isotope data are reported relative to the Vienna Pee Dee Belemnite (VPDB) standard. Results are presented in Datasets S1 and S4.

Interpretation of δ¹³C Values.

We discuss carbon isotopic values using δ-notation in per mil (‰) units, where δ¹³C = (R_sample/R_reference – 1) × 1,000, and R represents the ¹³C/¹²C ratio of sample or reference material. For the interpretation of enamel δ¹³C values, we assign nominal preindustrial δ¹³C plant values of −26‰ and −12‰ to −10‰ for C₃ and C₄ plants in eastern Africa and assume that the isotope enrichment between diet and enamel for primates is ca. 14‰ (5), such that an animal with a pure C₃ diet would have an enamel δ¹³C value of ca. −12‰ and an animal with a pure C₄ diet would have an enamel δ¹³C value of ca. +2‰ to +4‰. We use these values only as rough guides to estimating the proportions of C₃ and C₄ resources consumed by an animal over the course of enamel mineralization, because there is a range in δ¹³C values of vegetation (diet) and there is uncertainty in the carbon isotope enrichment between diet and bioapatite for primates (5, 69). This is the approach used by ref. 2 to estimate the proportion of C₃ and C₄ plants in hominin diets. Wynn et al. in ref. 1 take a different approach and assign end-member δ¹³C values of enamel representing animals that had diets composed entirely of either C₃ or C₄ resources based on median values for giraffes and Alcelaphines, respectively. There were not sufficient numbers of fossil teeth representing Alcelaphines at the localities we sampled to take this approach with the Woranso-Mille data.

When we compare tooth enamel δ¹³C values from fossils to those of extant animals, all of the data from extant animals are corrected to δ¹³C values of CO₂ in A.D. 1750 values as done by ref. 2, using data from ref. 70 for pre-1980 δ¹³C_CO2-atm values and data from Mauna Loa, reported by ref. 71, for δ¹³C_CO2-atm after 1980, assuming that δ¹³C value of atmospheric CO₂ was −6.3‰ in 1750 A.D. These values are indicated in the text, Fig. 2, and Dataset S3 as δ¹³C₁₇₅₀. As noted by ref. 2, these values provide a maximum correction because tooth enamel formed before the death of the animal.