Calcium isotopic ecology of Turkana Basin hominins

Diet is a major driver of hominin evolution, but most of the geochemical evidence relies on carbon isotopes (δ13C). Here, we report enamel stable calcium isotope (δ44/42Ca) values against δ13C values for several hominins and co-existing primates in the Turkana Basin area, circa 4 to 2 Ma. Australopithecus anamensis clusters with mammal browsers, Kenyanthropus platyops is distinct from A. anamensis in foraging into more open environments and the coexisting Theropithecus brumpti encompasses both the grazer and omnivore/carnivore domains. Early Homo is remarkable for its wide distribution in δ44/42Ca values, possibly reflecting omnivorous and opportunistic preferences. Paranthropus boisei is uniquely distributed in the δ13C versus δ44/42Ca iso-space being distinct from all other hominins from the Turkana Basin area as well as from the co-existing Theropithecus oswaldi. Several hypotheses are explored to discuss the unique δ44/42Ca values of Paranthropus boisei including significant differences observed with δ44/42Ca values recently reported for P. robustus from South Africa, questioning the monophyly of this genus.

N on-traditional stable isotopes are providing new avenues of research for exploring the ecology, physiology, and dietary preferences of extinct organisms, including our own lineage. Current evidence for early hominin diet relies on comparative osteology, dental tribology, and geochemistry [1][2][3][4][5][6][7][8][9][10] . But in the latter case, collagen nitrogen and its isotopes, a wellused trophic level proxy, are rarely preserved in such ancient (i.e., >1 Ma) contexts. While non-traditional isotopes emerge as a new toolkit to paleodietary inference [11][12][13][14][15] , a great deal of the geochemical evidence for early hominin diets relies on carbon isotopes [1][2][3][4][5][6][7][8][9] and to a lesser extent on trace element concentrations preserved in mineralized tissues such as fossil bone or teeth [16][17][18][19] . The carbon isotopic composition of bioapatite reflects the photosynthetic pathway (i.e., C 3 , C 4 , or CAM) of the ultimate plant source and has been useful for reconstructing past vegetation evolution 20 and exploring the feeding ecology of African hominins and associated faunas. African hominins cover the whole C 3 -C 4 spectrum. For example, in East Africa, early australopithecines derived most of their food from C 3 sources 1,4,6 whereas more recent forms such as Paranthropus boisei almost exclusively relied on C 4 sources 3,5 . However, significant proportions of C 4 sources are recorded in some early forms such as Australopithecus afarensis or Kenyanthropus platyops 1,2,7 . In Central Africa, Australopithecus bahrelghazali represents another example of an early hominin that was sourcing its food in a C 4 environment 8 . Another significant result of carbon isotope studies is the marked difference in δ 13 C values between megadont australopithecines, providing evidence that the South African Paranthropus robustus and the East African Paranthropus boisei were exploiting different resources 9,17 . Such results were independently corroborated with microwear texture 10,21 . Nonetheless, other proxies are needed to infer hominin diet beyond the ultimate plant source. Here, we investigate resource use in Turkana Basin hominins from Kenya using calcium and carbon isotopes and discuss their relevance for inferring the evolution of dietary habits in the hominin lineage Seminal studies on calcium isotopes in vertebrates 22 recognized a decrease in the 44 Ca/ 42 Ca ratio with increasing trophic level. Subsequent studies supported this trophic level effect, notably in the marine realm 23,24 , but also raised awareness on individual physiological variability influencing Ca isotope variability 25,26 or on the insensitivity of Ca isotope ratios to distinguish between primary and secondary consumers 27 . However, the constant δ 44/42 Ca offset of about −0.6‰ from dietary Ca to bone 22,28 supports the idea of a propagation-related calcium isotope fractionation from one trophic level to another. This pattern reflects a shared physiological feature of all vertebrates including terrestrial and aquatic mammals, as measured in six different species of mammals including foregut and hindgut digestive physiologies (as reviewed in 14 ). Moreover, this trophic level prediction recently found support as measured in a variety of modern and extinct terrestrial faunas 29,30 . Yet, disentangling dietary sources remains challenging because dietary items possess highly variable calcium concentrations and calcium isotopic compositions as exemplified by plants (as compiled in 14,29 ), bone 31 or milk 28,32 . Notably, this last source is 44 Ca-depleted 28 and analyzing tissues mineralizing during the pre-to postnatal transition has proved useful to study nursing in modern and fossil hominins 32,33 . Therefore, the isotopic composition of a consumer's tissue is controlled by the mass-weighted isotopic average of all the dietary sources and in our present case, we do not include data from early forming teeth (e.g., first molars). In other words, whether 44 Ca-enriched or 44 Ca-depleted, a narrow range of calcium isotopic composition in a consumer tissue may reflect a specialized diet (e.g., late forming teeth of Gorilla, see below and Supplementary Data 1) whereas a scattered distribution should reflect a more complex interplay between resource diversity, isotopic variation in resources, and differential consumption of resources by individuals within a group. As explained above, we favor the hypothesis of a Ca isotope variability primarily related to diet but other aspects such as speciesdependent physiologic parameters 25,26 will have to be further explored in future calcium isotope mammal studies. Therefore, in mammals, calcium isotopes can be used to monitor nursing in early formed teeth 28,32 ; later formed teeth can provide information on the intake of calcium (Ca) from adult dietary sources 29 .
Here, we report Ca isotope values for several early hominin species and associated nonhuman primates from the Turkana Basin area. We infer a unique trophic ecology and/or physiology for Paranthropus boisei according to its distinct isotopic distribution.

Results & discussion
Ca isotopic variability in modern Gorilla teeth. Teeth belonging to four modern western lowland gorilla (Gorilla gorilla gorilla) individuals from La Lopé National Park, Gabon, analyzed for Ca isotopes 33 , are presented here as a comparison with other hominin and non-hominin primates ( Fig. 1  teeth were formed early and include an incisor (G72) with a distinctly low δ 44/42 Ca value as well as a first molar (G39) with a surprisingly high δ 44/42 Ca value, being indistinct from the two other late forming gorilla teeth (second and third molars) available (Supplementary Data 1). The first molar starts mineralization earlier than the incisor (i.e., before birth) and completes mineralization before the third year (within the fourth year for the incisor) 34 . Nevertheless, it should be stressed that tooth enamel from that first molar was sampled for the entire height of the crown, thus mixing enamel formed in utero with enamel formed during or after completion of weaning. Because such teeth mineralize until 3 year of age after birth 35 , weaning and adult food items other than breast milk (in this case plant material) would have contributed to the high δ 44/42 Ca values observed here 36 . It should also be added that dental development is genetically controlled and the huge genetic diversity reported in western lowland gorillas 37 suggests that a wider spectrum of mineralization patterns exists beyond the published ages reconstructed from histological studies. Therefore, that the first molar (G72) did mineralize later than expected from the dental development models would not be surprising. The two late forming teeth (G40 and G41) display 44 Ca-enriched compositions and are interpreted as reflecting the post-weaning period of tooth formation during which the individuals were taking their Ca from adult food 33 . A detailed survey of Ca isotope variability in gorilla teeth and other modern primates will certainly bring interesting results to discuss physiological versus dietary influence on Ca isotopic fractionation processes, as currently investigated in humans 26,32,38 .
Ca isotopic variability in non-hominin fossil primates. Our stable calcium isotope values for various hominins and coexisting primates from the Turkana Basin area (n = 69) range between −0.69‰ and −1.88‰ (Figs. 1 and 2; Supplementary Data 1). Recently published calcium isotope data for coeval large mammals were shown to preserve their pristine composition, as discussed against elemental concentrations of diagenetic origin of the same samples 29 (see also Methods). This dataset presents a similar ordering between modern and fossil faunas but also some differences for certain taxa such as suids and saber-tooth cats, interpreted to have had changing dietary ecologies through time 29 .
Papionines have often been considered as analogues to understand dietary preferences of early hominins 1 . The 44 Cadepleted isotope values of modern Papio anubis (Fig. 1) are to be considered under the light of the broad feeding preferences in cercopithecines 39,40 . In comparison, the more positive Ca isotope values of Parapapio sp. invite to consider that extinct taxa may not have been similar in their feeding preferences, although one must take into account that modern Papio anubis has a wide distribution in Africa today and our data cannot reflect such spatial variations. Modern geladas do intensively feed on high altitude C 3 grasses 41 and our δ 44/42 Ca values (Supplementary Data 1) for two modern grass-eating Theropithecus gelada (−1.40‰) are consistent with a grass-eating ecology. Grazers are 44 Ca-depleted relative to leaf-eaters by about 0.3‰ and other modern grazers such as zebras, buffaloes, warthogs, and hippos also display comparably low calcium isotope values that overlap with carnivore values 29 (Fig. 2). Morphological evidence shows grass-eating adaptations in extinct Theropithecus 42 ; carbon isotopes confirm an early dietary adaptation to C 4 grasses as an important component of their diet 43  nearly pure-C 3 diet 1 similar to Ardipithecus ramidus 4 . Although early studies based on cranial morphology suggested tough/ abrasive food items in the diet of A. anamensis, more recent studies based on microwear and carbon isotopes proposed a diet of soft plant items from the C 3 environment 1,44 . Here, the distribution of calcium isotope values in A. anamensis (−1.18 ± 0.17‰, 1SD, n = 8) is not distinctly 44 Ca-depleted and overlaps much of the modern and fossil East African browser domains (Fig. 2) and indicates that it was foraging on a variety of C 3 resources, perhaps fruits or plants, as previously suggested from carbon isotopes 1  P. boisei possesses a unique tooth morphology with thick enamel and flatten occlusal surfaces when worn, and dental microwear studies do suggest a diet of soft plant items 21 . If leaves appear to be a good dietary candidate, C 4 leaves other than blades from grasses or sedges are not obviously identified in modern ecosystems. Theropithecus gelada cannot be considered a modern primate analogue for P. boisei, given their differences in δ 44/42 Ca values. A modern primate analogue for P. boisei that would feed on C 4 vegetation other than grass or sedge does not exist. Gorillas could be viewed as a remote plant-eating analogue. The two lateforming modern western lowland gorilla teeth from La Lopé in Gabon display calcium isotope values around −1.05‰ (Supplementary Data 1) 33 , comparable with the values of P. boisei. Although those gorillas feed on fruits, leaves, and ants, these resources arise from C 3 , not C 4 plants 11 . While dicot leaves and fruits are 44 Ca-enriched, the only dicot leaves that follow the C 4 photosynthetic pathway are represented by forbs, which would be found in relatively wet environments 46 and could be a dietary candidate for P. boisei according to oxygen isotope composition of P. boisei teeth suggesting water dependency 5 . Conducting a calcium isotope survey of such plants but also of C 4 sedges, especially in including plant parts that are known to fractionate differently 47 , would certainly add relevant data to better tackle this problem. On the other hand, we cannot exclude that both G. gorilla and P. boisei had physiological similarities that would lead to 44 Ca/ 42 Ca values that are enriched relative to other primates.
Whatever the exact dietary component or physiological specificity of P. boisei was, this species remains remarkable for exhibiting 44 Ca-enriched tooth enamel within a narrow range of isotope values, which most likely reflect a specialized diet on items with low isotopic variability 48 possibly associated to a physiology different than other primates. The range of calcium isotope values of P. boisei not only differs from other hominins of the Turkana Basin, it also differs from that of its South African congener P. robustus, which shows lower calcium isotope values 33 (Fig. 3). Interestingly, P. robustus and A. africanus do not differ in δ 13 C values 9 , nor do they differ in δ 44/42 Ca values 33 (Fig. 3). P. robustus was interpreted to have had a flexible diet together with the contemporaneous early Homo from South Africa 9,49 , a hypothesis that seems corroborated by the overlap of calcium isotope values of P. robustus with early Homo and A. africanus 33 . From our study as well as previous ones based on δ 13 C values 1 or tooth microwear 21 , the emerging picture could be that of P. boisei in East Africa with a specialized diet versus P. robustus in South Africa with a flexible diet 9,49 . These results confirm discrepancies between diets but apparently shared masticatory function and weaken the case for Paranthropus monophyly based solely on dentognathic features 50 .
In summary, calcium isotope ratios in a survey of hominin lineages in East Africa show that P. boisei has a unique 44 29 and Tacail et al. 33 . EA East Africa, SA South Africa. Values selected for representatives of the associated-fauna include only those values for organisms circa 2 Ma; for this reason, the browser category may appear underrepresented and the apparent difference may reflect specific dietary differences rather than differences in the substrate isotope baseline. Note that carnivore values in EA are dispersed as explained by a dataset that includes sabertooth felids. The boxes represent the first and third quartiles with the medians as horizontal lines. The lower and upper whiskers represent 1.5 * the interquartile range (numbers of biologically independent samples per group: n = 7 for South African carnivores; n = 10 for East African carnivores; n = 7 for South African browsers; n = 4 for East African browsers; n = 10 for South African grazers; n = 14 for East African grazers; n = 38 for Paranthropus robustus; n = 8 for Paranthropus boisei). δ 44/42 Ca values are expressed both against ICP Ca Lyon (left) and SRM915a (right). Source data are provided as a Source Data file. ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17427-7 ratio compared with other hominins including the South African P. robustus, all of which have similar ratios to each other but not to P. boisei. This calcium isotope difference could be due to differences in diet, or in physiology, and remains a unique character of P. boisei compared with other hominins.

Methods
Fossil samples. Fossil samples were available from the collections of the National Museums of Kenya and the Turkana Basin Institute. Those fossil samples are part of a paleoecology project reported in Cerling et al. 51 . Powdered enamel was collected using a low-speed dental drill along the broken surfaces of tooth enamel. Each sample represents an average of the crown height in order to provide enough material for the published carbon isotope analyses initially envisioned. 100-200 µg of powder per sample was used for Ca isotope purification. Sample chemistry. The chemical processing of the samples follows the method described in details in Martin et al. 29 and Tacail et al 33,52 . Briefly, enamel samples were dissolved in 300 μl of suprapure 1 M HCl acid and were subsequently processed through AG50X-W12 cation exchange resin in 1 M HCl medium to dispose of sample matrix. Ca and Sr fractions were collected in 6 M HCl medium. Ca fractions were then separated from Sr by loading samples onto columns filled with Sr-specific resin (Eichrom Sr-Spec) in suprapure 2 M HNO 3 medium. Blanks for the whole procedure did not exceed 100 ng Ca.
Isotopic measurements. Calcium isotope abundance ratios ( 44 Ca/ 42 Ca and 43 Ca/ 42 Ca) were measured using a multi-collector ICP-MS (MC-ICP-MS, Neptune Plus, Thermo). After purification, Ca samples were dissolved in ultrapure 0.05 M HNO 3 and Ca concentration was set at 2 ppm for all samples and standards. All Ca isotope compositions are expressed using the 'delta' notation defined as follows for the 44  Diagenesis. It is well known that diagenetic processes can alter the original isotopic composition of mineralized tissues. That Ca is a major constituent of bioapatite (at a concentration of about 35-40% in weight) represents a solid reason to investigate its isotopic composition in fossil bone and teeth as old as the Cretaceous 30,31,53 . Here, before chemical purification of Ca isotopes, all samples underwent the standard protocol that is applied for carbon isotopic analyses by removing potential secondary carbonates 51 . Measuring trace element concentrations requires more than half a milligram-when such elements are present-and therefore necessitates large quantities of sample powder at the cost of precious samples, such as the hominin teeth analyzed in the present study. In this study, our average sample weight uptake was about 100 µg, which is suitable for Ca isotope analysis but obviously not for trace element concentrations. Correlations between trace elements of diagenetic origin and Ca isotopic compositions were assessed in a previous study of fossil mammals from the same Turkana Basin localities 29 as those of the hominins analyzed in the present study. We concluded that the impact of diagenesis on our Ca isotopic measurements was minimal 29 and by extension, we consider such conclusions applicable to the present hominin dataset.
Comparing East African versus South African datasets. Because a substrate effect cannot be excluded on the fractionation of calcium isotopes, we used associated faunal calcium isotope compositions 29,33 to serve as a baseline. This issue has not been explored in detail yet 29 but it is worth repeating here that Ca isotopic composition in rocks is rather homogenous 60 and a substrate effect may not be pre-eminent in vertebrate tissues. The Turkana fossil assemblage circa 2 Ma includes six felids and four hyaenids calcium isotope values more variable than the range of values of the seven carnivores from South Africa that also includes felids and hyaenids. This dispersion in Ca isotope values is explained by the presence in the Turkana dataset of saber tooth felids, whose diet was specialized as previously discussed 29 . The fossil browsers from Turkana (-1.16‰, n = 4) versus browsers from South Africa (-1.38‰, n = 7) show different values. Browsers from South Africa consist of one genus only, Tragelaphus, which is known in modern South Africa to include about one-third of C 4 component in its diet 61 and might explain some of its values to be 44 Ca-depleted. The low number of browsers from East Africa that includes two giraffids and two bovids does not permit to satisfyingly compare both browser datasets. Finally, the fossil grazers consist of a larger dataset; taxa from Turkana (-1.25‰, n = 14) versus grazers from South Africa (-1.30‰, n = 10) gave consistent values indicating that beyond some variability that may arise from seasonal variation into food items (although this is restrained by the sampling protocol-see above), nursing processes or other unconstrained environmental processes, it is reasonable to compare both South African and East African palaeoenvironments.