Isotopic analyses suggest mammoth and plant in the diet of the oldest anatomically modern humans from far southeast Europe

Relatively high 15N abundances in bone collagen of early anatomically modern humans in Europe have often been interpreted as a specific consumption of freshwater resources, even if mammoth is an alternative high 15N prey. At Buran-Kaya III, access to associated fauna in a secured archaeological context and application of recently developed isotopic analyses of individuals amino acids offer the opportunity to further examine this hypothesis. The site of Buran-Kaya III is located in south Crimea and has provided a rich archaeological sequence including two Upper Palaeolithic layers, from which human fossils were retrieved and directly dated as from 37.8 to 33.1 ka cal BP. Results from bulk collagen of three human remains suggests the consumption of a high 15N prey besides the contribution of saiga, red deer, horse and hare, whose butchered remains were present at the site. In contrast to bulk collagen, phenylalanine and glutamic acid 15N abundances reflect not only animal but also plant protein contributions to omnivorous diet, and allow disentangling aquatic from terrestrial resource consumption. The inferred human trophic position values point to terrestrial-based diet, meaning a significant contribution of mammoth meat, in addition to a clear intake of plant protein.

runoff into the shelter. In the upper unit (from silt clay level to , seasonal contrasts are less pronounced and eolian deposits are dominant; the climatic conditions become significantly drier than below. Layer 4 is more organic and coarser than below, which are characteristics of contrasting conditions (frost shattering, pedobiological activities).
Mammoth remains are only represented by a few processed artefacts made of ivory (bead and engraved plate). Other osseous artefacts are composed of 40 bone pointed tool fragments, with debris from in situ production, 1 perforated fox canine and 35 perforated mollusc shells.
Taphonomical studies show a limited modification of bone surfaces by carnivores, whose tooth marks are only observed on 1% of the NISP. Almost all skeletal elements of saiga antelope, fox and hare are represented but axial and girdle parts are relatively scarce. The bone material is very fragmented.
A high proportion of burned bones were found (25% of the total number of remains), resulting from intense combustion activities produced by humans. Cut marks and fracture features due to human activities were observed on 7% of the identified specimens, belonging to six taxa: saiga antelope, hare, bovine, horse, fox and wolf. Human modifications of the saiga bone surfaces (on cranial and postcranial remains) mainly reflect disarticulation and defleshing activities, breakage and marrow extraction. Consumption of hares is also evidenced by taphonomic modifications. Foxes also exhibit marks of disarticulation and defleshing, and also traces of fur removal on metapodials and tarsals.
Indices of seasonality tend to indicate a summer mortality of the saiga antelopes, which represent the main game of the hunter-gatherers in Buran-Kaya III. The subsistence activities in the studied area of the site, i.e. the rear part of the rock-shelter, seem relatively modest and do not exhibit the complete butchery chaîne opératoire. Only skinning, filleting and marrow extraction activities were identified. The first stages of butchery probably took place elsewhere, in another location or even another site. Furthermore, there is a high usage of bone as fuel within the site, which could be explained by a very cold environmental context with rare plant resources and/or by activities of preparation of meat and fat broth.
These zooarchaeological results from the Gravettian s.l. layers suggest short-term and recurrent settlements. The site of Buran-Kaya III would have been used as a typical temporary hunting camp, with more intense and/or long activities in layer 6-1 than 6-2.

Supplementary Data 3
The anatomically modern humans of Buran-Kaya III More than two hundred human remains were discovered in three well-documented Upper Palaeolithic layers (6-2, 6-1, 5-2), attributed to the Gravettian sensu lato technocomplex. Layer 6-1 has yielded the richest assemblage in terms of number of remains, minimum number of individuals and anatomical parts. These human remains are highly fragmented and consist mostly of cranial parts and teeth (90% of the remains), whereas the postcranial skeleton is barely represented (hand phalanx fragments). Based on the age determination from the dental remains, at least 5 individuals belonging to three developmental age groups (juvenile, subadult and adult) can be identified in this layer. The health of these individuals was good 4 , with no caries. Enamel dental hypoplasia was observed only in the youngest individuals at the beginning of their childhood (~3 years). Based on the combination of morphological features on the occipital and on the dental remains, the human remains from Buran-Kaya III are distinguished from Neanderthal and are attributed to Anatomically modern humans 5 .
Among the fragmented human remains (n=172) from layer 6-1, 8 % exhibits human modifications such as cut marks. Their morphology and their position in respect to muscular insertions suggest that scalping and disarticulation processes occurred. Three possible anthropogenic actions on human bodies can be hypothesized to explain the occurrence of these cut marks on several human remains: mortuary practice, dietary and non-dietary cannibalism. Comparative taphonomical analyses (e.g. spatial repartition of the remains, skeletal representation and surface bone modifications) between the human remains and the saiga antelope, which is the main game according to the archaeozoological analyses, show that these two taxa were not processed in a similar way. Furthermore human skulls were intentional selected. Consequently, the hypothesis of a dietary cannibalism 6 is not supported by these data. Specific mortuary practices (such as postmortem disarticulation processes of corpses for secondary disposal) or ritual cannibalism are proposed as alternative scenarios 5,7 .

Results of the SIAR reconstruction for the animal carnivores based on bulk isotopic ratios
The food categories to be tested through the SIAR model were defined as follow: deer and horse altogether (Deer&Horse), saiga antelope (Saiga), woolly mammoth (Mammoth) and hare (Hare).
For each category, we considered the average and standard-deviation of the δ 13 C coll and δ 15 N coll values of the specimens without distinction between level 6-1 and 6-2 (Supplementary Table 2). In the case of a single individual, as for mammoth and hare, we attributed a standard-deviation value that was calculated from other datasets measured on archaeological context (Supplementary Tables  3 and 4). In both cases, we selected the highest values obtained, which were to ± 0.4‰ and ±0.9‰ for mammoth (Maisières-Canal dataset) and ±0.5‰ and ±0.8‰ for hare (Combe-Saunière 1 dataset), corresponding to the standard-deviation of the δ 13 C coll and δ 15 N coll value respectively.

Supplementary Table 2.
Results of stable isotope analyses of collagen (δ 13 C coll , δ 15 N coll ) of the herbivores of level 6-1 and 6-2 of Buran-Kaya III: The carbon and nitrogen composition of the collagen is given through elemental composition (C coll , N coll ) and atomic ratio (C:N coll ). For each predator specimen (carnivore and human), the possible contribution of each food source was evaluated using SIAR and considering the combined δ 13

Isotope analyses of individual amino acids
Instrumental analysis was performed according to previous methods, with a few modifications of the equipment settings 20 . The amino acid derivatives were injected into the GC column using the  T P t e r r e s t r i a l = 2 T P t e r r e s t r i a l = 3 T P a q u a t i c = 2 T P a q u a t i c = 3 T P a q u a t i c = 4 thickness; Agilent Technology). The GC oven temperature was programmed as follows: isothermal hold at 40°C for 3 min; temperature ramp to 110°C at the rate of 15°C min −1 ; ramp to 150°C at the rate of 3°C min −1 ; ramp to 220°C at the rate of 6°C min −1 ; and subsequent holding isothermally at 220°C for 13 min. Carrier gas (He) flow rate through the GC column was 1.4 ml min −1 . CO 2 generated in the combustion furnace was eliminated by a liquid nitrogen trap. Standard mixtures of nine amino acids with known δ 15 N values were injected into the GC/C/IRMS every five runs to confirm the reproducibility of the isotope measurements. The mean accuracy and precision of the reference mixtures were 0.0‰ and 0.4-0.7‰ (mean of 1σ), respectively. Nitrogen isotopic composition of the following 10 amino acids were determined: alanine (Ala), glycine (Gly), valine (Val), leucine (Leu), isoleucine (Ile), proline (Pro), serine (Ser), glutamic acid (Glu), phenylalanine (Phe) and hydroxyproline (Hyp). All reported δ 15 N values for glutamic acid include a contribution from the α-amino group of glutamine, as glutamine is converted to glutamic acid during acid hydrolysis. The isotopic measurement of collagen amino acids has been performed by GC/C/IRMS after optimization of instrumental settings (temperature of the reaction furnaces, carrier gas flow rate, and sample amount) 21 . Isotopic difference between proline and hydroxyproline, it might relate to possible different factors such as site taphonomy or animal phylogeny. Nevertheless, these two amino acids are not involved in the TP calculation.