Molecular and isotopic evidence for milk, meat, and plants in prehistoric eastern African herder food systems

Katherine M. Grilloa,1,2, Julie Dunneb,1, Fiona Marshallc, Mary E. Prendergastd, Emmanuelle Casanovab, Agness O. Gidnae, Anneke Janzenf, Karega-Muneneg, Jennifer Keuteh, Audax Z. P. Mabullai, Peter Robertshawj, Toby Gillardb, Caitlin Walton-Doyleb, Helen L. Wheltonb, Kathleen Ryank,3, and Richard P. Evershedb

aDepartment of Anthropology, University of Florida, Gainesville, FL 32611; bOrganic Geochemistry Unit, School of Chemistry, University of Bristol, BS8 1TS Bristol, United Kingdom; cDepartment of Anthropology, Washington University in St. Louis, St. Louis, MO 63130; dDepartment of Sociology & Anthropology, Saint Louis University, 28003 Madrid, Spain; eNational Museums of Tanzania, Dar es Salaam, Tanzania; fDepartment of Anthropology, University of Tennessee, Knoxville, TN 37996; gUnited States International University, Nairobi, Kenya; hSchool of Archaeology, University of Oxford, Oxford OX1 3TG, United Kingdom; iDepartment of Archaeology and Heritage Studies, University of Dar es Salaam, Tanzania; jDepartment of Anthropology, California State University, San Bernardino, CA 92407; and kAfrican Section, University of Pennsylvania Museum of Archaeology and Anthropology, Philadelphia, PA 19104

Edited by Melinda Zeder, National Museum of Natural History, Santa Fe, NM, and approved February 21, 2020 (received for review November 18, 2019) The development of pastoralism transformed human diets and selected for in pastoral populations through gene–culture co- societies in grasslands worldwide. The long-term success of cattle evolution, but the roles of cuisine, gut microbiomes, and pop- herding in Africa has been sustained by dynamic food systems, ulation interactions are poorly understood, given highly variable consumption of a broad range of primary and secondary livestock selection contexts worldwide (e.g., ref. 4). In southwestern Asia, products, and the evolution of lactase persistence (LP), which archaeological data suggest early dairying was based on sheep/goat allows digestion of lactose into adulthood and enables the milk- pastoralism; however, faunal remains and organic residue evi- based, high-protein, low-calorie diets characteristic of contempo- dence from pottery from the 9th millennium in NW Anatolia in- rary pastoralists. Despite the presence of multiple alleles associated dicate intensification of milk use centered on cattle (5). African with LP in ancient and present-day eastern African populations, the herding systems, beginning ∼8000 BP in what is now the Sahara, contexts for selection for LP and the long-term development of were characterized by greater mobility, later limited cultivation, ANTHROPOLOGY pastoralist foodways in this region remain unclear. Pastoral Neo- and dietary foci on cattle. In eastern Africa today, milk and milk c lithic ( . 5000 to 1200 BP) faunas indicate that herders relied on products from cattle (or camels in more arid regions), sheep, and cattle, sheep, and goats and some hunting, but direct information goats provide as much as 60% to 90% of total calories (6–8). Milk on milk consumption, plant use, and broader culinary patterns is shortages during droughts or dry seasons increase vulnerabilities rare. Combined chemical and isotopic analysis of ceramic sherds to malnutrition (9–11) and result in increased consumption of (n = 125) from Pastoral Neolithic archaeological contexts in Kenya meat and within-bone nutrients. Milk and meat are also com- and Tanzania, using compound-specific δ13CandΔ13Cvaluesofthe plemented by seasonally available wild plant foods and, today, major fatty acids, provides chemical evidence for milk, meat, and introduced staples (e.g., maize and beans). Tishkoff and col- plant processing by ancient herding societies in eastern Africa. These data provide the earliest direct evidence for milk product leagues (12, 13) suggest that strong selection resulted in the appearance of the C-14010 genetic variant for the LP trait in consumption and reveal a history of reliance on animal products ∼ and other nutrients, likely extracted through soups or stews, and eastern African populations an estimated 6,800 to 2,700 y ago. plant foods. They document a 5,000-y temporal framework for east- Evidence for early milk consumption is lacking, but discovery of ern Africa pastoralist cuisines and cultural contexts for selection for an individual buried in northern Tanzania, dating to 2200 to alleles distinctive of LP in eastern Africa. 2150 BP, with the C-14010 allele provides a pastoral context for

archaeology | ceramics | lipid residue analysis | pastoralism | Significance lactase persistence Lipid residue analysis of archaeological ceramics provides the astoralism is a dominant form of food production in grass- earliest direct chemical evidence for milk, meat, and plant Pland ecosystems globally (1), and genetic histories of pasto- consumption by pastoralist societies in eastern Africa. Data ralist populations reflect complex interplays among human for milk in specialized pastoral systems (c. 5000 to 1200 BP) migrations, interactions, and culinary practices (2). In eastern reveal changing selective pressures for lactase persistence and – Africa, mobile herding communities in marginal dryland envi- provide support for models of gene culture coevolution ronments have historically relied on milk-based, high-protein, among pastoral populations. low-calorie diets (3), facilitated by multiple genetic alleles for Author contributions: K.M.G., J.D., K.-M., K.R., and R.P.E. designed research; K.M.G., J.D., lactase persistence (LP) and digestion of milk into adulthood. M.E.P., A.O.G., J.K., A.Z.P.M., P.R., T.G., C.W.-D., and H.L.W. performed research; E.C. Although the earliest evidence for cattle, sheep, and goats dates contributed new reagents/analytic tools; J.D., E.C., T.G., C.W.-D., and H.L.W. analyzed to the Pastoral Neolithic (PN; c. 5000 to 1200 BP), when herders data; and K.M.G., J.D., F.M., M.E.P., A.J., P.R., and R.P.E. wrote the paper. spread into eastern Africa, little is known about the emergence of The authors declare no competing interest. milk as a food, settings for selection for eastern African alleles for This article is a PNAS Direct Submission. LP, or the long-term development of highly specialized pastoralist Published under the PNAS license. foodways. Direct evidence for the consumption of milk, for ex- 1K.M.G. and J.D. contributed equally to this work. ample, is absent in PN and later archaeological records. 2To whom correspondence may be addressed. Email: [email protected]. The genetic basis for LP facilitates the ability to survive on 3Deceased January 17, 2018. livestock milk, as regulatory mutations that affect production of This article contains supporting information online at https://www.pnas.org/lookup/suppl/ the enzyme lactase and its postweaning retention enable digestion doi:10.1073/pnas.1920309117/-/DCSupplemental. of milk sugars (lactose). It has long been hypothesized that LP is First published April 13, 2020.

www.pnas.org/cgi/doi/10.1073/pnas.1920309117 PNAS | May 5, 2020 | vol. 117 | no. 18 | 9793–9799 Downloaded by guest on September 30, 2021 selection (14). The existence of multiple other genetic variants first appear in the Turkana Basin of northwestern Kenya during associated with the LP phenotypic trait in contemporary eastern a period of aridity and social change c. 5000 to 4500 cal BP (3), African populations (12, 13) distinguishes this region from most when herder-fishers built a series of megalithic pillar sites. Some of of Eurasia, where a single allele for LP exists (2). All LP variants these are communal cemeteries in which people deposited highly are present at relatively low frequencies (<45%) (13) in eastern decorated ceramic vessels (15, 16). After c.3300calBP,there Africa, and it is likely this was also the case in the past (14). It is was a florescence of specialized herding economies throughout frequently argued that in the absence of LP, herder dependence south-central Kenya and into Tanzania. Herders moved into on milk can be facilitated by the preparation of fresh milk into wetter, topographically diverse, and more productive pastoral more easily digested fermented milk products (2, 13). grazing lands, where cattle herds can provide milk for human Although it is clear from the human genetic data that the consumption almost year-round (17–19). ability for herders to digest milk must have been strongly adap- The sites and samples chosen for this study are broadly repre- tive in eastern Africa, lack of direct information on when and sentative of the archaeological diversity of the PN. They include where milk was drunk, how ancient herders prepared their food, ceremonial and habitation sites, span the Turkana Basin in Kenya and how herders buffered milk shortages has constrained un- to the north-central Tanzanian highlands (Fig. 1), and cover the derstanding of prehistoric selective contexts for LP. Integrated ar- three major PN material culture traditions (see Table 1 for ref- chaeological and molecular evidence can provide a deeper erences). Jarigole (GbJji1, c. 5000 to 4500 cal BP) is a megalithic understanding of when, where, and why LP emerged in eastern pillar site on the east side of Lake Turkana, and Dongodien Africa, in addition to shedding light on broader dietary patterns. (GaJi4, c. 5000 to 4500 cal BP) is a roughly contemporaneous Here, we report lipid residue analyses of 125 ceramic vessels from open-air habitation site on the eastsideofLakeTurkana.Ngamuriak PN archaeological contexts in Kenya and Tanzania. These data c provide the earliest direct evidence for meat, milk, and plant pro- (GuJf6, . 2000 cal BP) is a settlement site in the Loita-Mara c. cessing or storage by ancient eastern African herders, and offer a region of southwestern Kenya. Luxmanda ( 3000 cal BP) is an- globally relevant perspective on the origins of pastoralist dietary other large open-air settlement site, located on the Mbulu Plateau adaptations and the evolution of milk-centered husbandry systems. in north-central Tanzania. A description of sampling strategies and a list of lipid-yielding sherds with contextual information can Pastoral Neolithic Herders be found in SI Appendix,TableS1; sherd photographs are in Fig. During the PN, communities herding cattle, sheep, goat, and 2. Previously published radiocarbon dates for each site, along with donkey spread throughout the eastern African Rift Valley and newly generated compound-specific radiocarbon dates on extracted beyond, interacting with indigenous hunter-gatherers. Livestock lipids (20), are in SI Appendix,TableS2.

Fig. 1. Density and distribution of excavated Pastoral Neolithic sites in eastern Africa: major sites with known coordinates are marked with black dots. Sites sampled for this study are marked with red triangles. Base map source: Natural Earth.

9794 | www.pnas.org/cgi/doi/10.1073/pnas.1920309117 Grillo et al. Downloaded by guest on September 30, 2021 Table 1. Summary table of archaeological information for sampled sites Date Archaeological Site range tradition Site description Ceramics Other material culture Fauna References

Jarigole c. 5000 to Nderit Megalithic Nderit ceramics: Diverse personal Only fauna 40 (GbJj1) 4500 cal communal vessels include ornamentation reported BP cemetery on the a variety of including stone, are fish eastern side of finely made forms bone, and ostrich vertebrae Lake Turkana and decorative eggshell beads; and small styles such as a relatively small fragments Classic Nderit lithic assemblage of bovid tooth basket-like (predominantly enamel bowls, often with obsidian); and internal scoring; zoomorphic other vessels have ceramic figurines red burnished ripple of wild fauna designs or incised and probable geometric patterns domestic cattle Dongodien c. 5000 to Nderit Open-air Nderit pastes Microlithic, Generalized 35, 41 (GaJi4) 4500 cal settlement and decorative predominantly pastoralism BP site on the techniques appear obsidian assemblage; with small eastern similar to those one unidentified stock, cattle, side of Lake seen at Jarigole incised bovid and wild fauna Turkana and Lothagam North bone including fish Ngamuriak c. 2000 Elmenteitan Open-air settlement Elmenteitan Distinctive Highly specialized 18, 42 (GuJf6) cal BP site in the ceramics: vessels blade-based pastoralism; Loita-Mara are typically obsidian focus on cattle, region of undecorated assemblage sheep, and

southwestern but contain large goats ANTHROPOLOGY Kenya quantities of sparkling mica temper; forms are varied but open lugged vessels are common Luxmanda c. 3000 SPN Highland open-air Narosura ceramics: Diverse chert, quartz, Highly specialized 26, 37, cal BP settlement vessels are and obsidian pastoralism; 43, 44 site on the commonly black SPN lithics; bone and focus on sheep, Mbulu Plateau, and exhibit ivory tools and goat, and north-central decorative motifs ornamentation; cattle Tanzania; largest- made by groundstone known and most comb-stamping artifacts southern of all PN or incising, found in settlement sites bands beneath rims; forms are typically open bowls in a range of sizes

− − Lipid Residue Results 11.5 mg·g 1, LUX019; Ngamuriak, 29.4 mg·g 1, NMK018). Recovery Rates and Concentrations. Lipid analysis and interpreta- Several sherds from both later sites contain high concentra- − − tions for PN ceramics were performed using established protocols tions of lipids (e.g., LUX003, 4.6 mg·g 1; LUX019, 11.5 mg·g 1; − − − (e.g., refs. 21 and 22). A total of 125 ceramic sherds were sam- LUX038, 5.9 mg·g 1; LUX043, 2.7 mg·g 1; LUX052, 2.0 mg·g 1; − − − pled from Jarigole (n = 20), Dongodien (n = 20), Ngamuriak NMK004, 2.3 mg·g 1; NMK005, 6.2 mg·g 1; NMK006, 13.2 mg·g 1; − − (n = 22), and Luxmanda (n = 63). Lipid recovery rates varied, NMK8,9,10, 3.4 mg·g 1; and NMK018, 29.4 mg·g 1). These vessels with rates at the two earlier sites, Jarigole and Dongodien, being were likely subjected to sustained use in the processing of high lower (25%, n = 5 at each site) than at Luxmanda (38%, n = 24), fat-yielding resources. and especially at Ngamuriak, where lipid recovery was very high (82%, n = 18). Lower lipid recovery rates may be taphonomic, Meat and Milk. Analysis of the total lipid extracts (TLEs; n = 125) but more likely suggest different patterns of vessel use. The from all sites, using gas chromatography (GC) and GC-mass mean lipid concentrations of the Jarigole and Dongodien sherds spectrometry (GC-MS), demonstrated that 52 sherds contained − − are 0.7 and 0.08 mg·g 1, respectively (Table 1), with maximum sufficient concentrations (>5 μg·g 1) of lipids to allow reliable − − lipid concentrations of 1.9 mg·g 1 (GAJ003) and 0.2 mg·g 1 interpretation (23). These extracts comprised lipid profiles (GBJ014), whereas the mean lipid concentrations for Luxmanda that demonstrated that palmitic (C16) and stearic (C18) fatty − − (1.3 mg·g 1) and Ngamuriak (3.6 mg·g 1) and the maximum lipid acids, typical of a degraded animal fat (Fig. 3 B–D), were the concentrations at these sites are considerably higher (Luxmanda, most abundant components (e.g., ref. 24). Further analyses by

Grillo et al. PNAS | May 5, 2020 | vol. 117 | no. 18 | 9795 Downloaded by guest on September 30, 2021 Fig. 2. Examples of sherds from each site containing lipid residues. Sherds A–G are considered Nderit pottery (see ref. 45) from the Turkana Basin; exterior surface is shown on the left, interior surface is shown on the right. Jarigole (A, GBJ001; B, GBJ014; C, GBJ016; D, GBJ018; E, GBJ007): A–D are Classic Nderit basketry-like vessels with unidentified forms; for A and B, note internal scoring, for C, note internal shallow impressions. Sherd E has incised grooves in a geometric pattern, common for shallow bowls. Dongodien (F, GAJ003; G, GAJ012): F is Classic Nderit, G has wide grooves, and three undecorated sherds with lipid residues are not pictured. Luxmanda (H, LUX019; I, LUX025; J, LUX065; K, LUX059): sherds are Narosura pottery, common to Savanna Pastoral Neolithic sites. Forms are open-mouthed bowls or globular vessels with comb-stamping or incising in bands under rims. Ngamuriak (L, NMK006; M, NMK8/9/10): sherds are Elmenteitan pottery, undecorated with forms including straight-sided or globular bowls/pots.

GC-combustion-isotope ratio MS (GC-C-IRMS) characterized further vessel (LUX025, −3.1‰), plotting at the boundary of the source of these animal products through determination of the range. Similarly, at Ngamuriak, vessel NMK020 plots close 13 Δ13 − ‰ the δ C values of the major fatty acids, C16:0 and C18:0 (Fig. 4), to the top of the range ( Cof 3.4 ), with vessel NMK8/ which allows differentiation between carcass fats from ruminant 9/10 plotting close to the boundary between meat and milk and nonruminant animals and between ruminant dairy and car- products. cass fats (21, 24). At Jarigole (n = 5), one lipid residue plots within the non- Plant Processing. At Luxmanda, two vessels, LUX032 and LUX057, Δ13 ruminant (or, possibly, plant) range, with the remaining four plot within the nonruminant/plant region with enriched Cvalues ‰ ‰ plotting within the ruminant adipose region. Ruminant dairy fats of 2.9 and 3.7 , respectively. Interestingly, both vessels have are differentiated from adipose fats when they display Δ13C low lipid concentrations, often indicative of plant processing, and δ13 – δ13 − ‰ the lipid profile of LUX032 includes a series of odd-over-even ( C18:0 C16:0) values of less than 3.1 (25). Thus, at n A Δ13 − -alkanes (Fig. 3 ), dominated by C31, likely originating from Jarigole one residue with a C value of 3.0 plots at the border – between adipose and milk fats. At Dongodien (n = 5), one vessel C3 or C4 wild grasses or lake-margin plants, such as sedges (28 31). plots in the ruminant dairy range, with two vessels plotting in the A further vessel, LUX030, is dominated by high concentrations of ruminant adipose range and a further two plotting in the non- even-over-odd numbered long-chain fatty acids, maximizing at C26, again strongly indicative either of an origin in leaf or stem epicu- ruminant/plant range. Nonruminant mammals potentially pre- ticular waxes (28). sent in the area during the PN include equids (zebra and donkey), suids (warthog and bushpig), and small mammals such Freshwater Resources. Fatty acid methyl esters (FAMEs) from all δ13 as hare and hyrax. At Jarigole and Dongodien, C16:0 values lipid residues were analyzed by GC-MS in SIM mode to check − ‰ − ‰ (range, 26.3 to 16.8 ) suggest that animals were sub- for the presence of aquatic biomarkers; namely, ω-(o-alkyl- sisting on both C3 and C4 plants, as confirmed by other datasets phenyl) alkanoic acids, vicinal dihydroxy fatty acids, and iso- SI Appendix at these sites (ref. 17 and , Fig. S2). prenoid fatty acids. These are routinely used to detect both The majority of the lipid-bearing residues at the two later PN marine and freshwater animal product processing (32–34); sites, Luxmanda (83%, n = 20) and Ngamuriak (94%, n = 17), however, no freshwater biomarkers were detected here. De- plot within the ruminant carcass range, confirming the dominant spite the prevalence of fish bones at Dongodien (35), fresh- δ13 use of ceramic vessels for meat processing. C16:0 values water resources such as fish may not have been prepared by (range, −22.1‰ to −12.0‰) for these sites suggest that animals boiling in pots and could have been cooked in other ways such were subsisting mainly on C4 plants, which aligns with stable as roasting, salting, or smoking. carbon isotope data for livestock tooth enamel across south- ern Kenya and northern Tanzania (refs. 17, 26, and 27 and SI Discussion Appendix, Fig. S2). One vessel at Luxmanda, LUX004, with a Lipid residue analysis of archaeological ceramics provides the Δ13Cof−5.4‰, plots firmly in the ruminant dairy region, with a earliest direct chemical evidence for milk, meat, and plant

9796 | www.pnas.org/cgi/doi/10.1073/pnas.1920309117 Grillo et al. Downloaded by guest on September 30, 2021 ANTHROPOLOGY Fig. 3. Partial gas chromatograms of trimethylsilylated lipid extracts of pottery. (A) Lipid distributions in LUX032 are indicative of leafy plant processing and

(B–D) showing typical degraded animal fat chromatograms, dominated by C16:0 and C18:0 fatty acids, from Dongodien, Luxmanda, and Ngamuriak, re- spectively. Circles, n-alkanoic acids (fatty acids, FA); triangles, n-alkanes; IS, internal standard, n-tetratriacontane.

consumption by pastoralist societies in eastern Africa. Our re- pastoralists in northern Kenya today revealed a broadly similar sults demonstrate that the earliest herders in the region, at pattern (38). In Samburu, this pattern reflects container choices, Dongodien, c. 5,000 y ago, predominately processed meat rather than degree of reliance on milk in the diet (9). Meat and products in their pottery vessels, while processing milk or milk bones are cooked in pots, while milk, for both ceremonial and products at low frequencies. The presence of milk at Dongodien daily use, is generally collected, stored, and processed in wooden provides evidence for an extended chronology for selection for vessels and gourds, which rarely preserve at archaeological sites. the distinctive eastern African basis for LP (C-14010), and raise Lipid assemblages at Ngamuriak and Luxmanda are domi- questions about whether these herders may have carried an allele nated by ruminant carcass product processing, which corre- for LP (e.g., G-13907) common in Sudan or Ethiopia today (see sponds well with highly fragmented cattle, sheep, and goat bone refs. 12 and 13). Meat, fat, and/or bones from livestock and wild assemblages (18, 26). Extremely high lipid concentrations from fauna (nonruminants) were the predominant products cooked or nearly all Ngamuriak vessels and many of the Luxmanda vessels served in ceramic pots found at the Dongodien settlement and the indicate that ceramic technologies throughout the PN played a Jarigole mortuary site. These food residues are consistent with critical role in the repeated preparation of meat, bones, marrow, faunal data from Dongodien; however, it is notable that few re- and/or fat, likely through breakage and boiling of bones. Fur- mains of nonruminant mammals (hyrax and hare) were identified thermore, lipid profiles at Luxmanda provide evidence for plant in the faunal assemblage (35). The organic residue data reveal that processing in the PN. Given otherwise good preservation of at least some highly decorated Nderit ceramics found at both early lipids at Luxmanda, the large number of sherds with no detect- PN mortuary contexts and habitation sites were used in the stor- able lipid residues may come from pots that served other func- age, preparation, or serving of food. All five vessels with preserved tions such as serving or storage of low lipid-yielding foods. lipids at Jarigole, and three from Dongodien, display low lipid At a larger scale, our findings demonstrate that as herders concentrations, suggesting they were only used sporadically for spread throughout eastern Africa following Saharan expansion, cooking or for serving foodstuffs. At Dongodien, two additional food-producing economies incorporated carcass processing in vessels (GAJ003 and GAJ013) yielded high lipid concentrations, pots and the use of milk from the earliest PN, and plant pro- indicating their possible use as cooking pots, although no use-wear cessing in pots by the later PN. Lipid residue data are relatively such as sooting on any Nderit pottery indicates use over fire. consistent across a wide geographic area, different site types, and In the wetter southern highlands and richer grasslands, faunal different material traditions: Throughout the PN, herders used remains show that later herders with different mortuary and pots to process predominantly ruminant carcass products and, ceramic traditions c. 3000 to 2000 BP kept large cattle herds and infrequently, milk or milk-based products. Notwithstanding the depended on livestock, with little to no reliance on wild fauna low frequency of milk processing, the evidence is robust. Hence, (36, 37). Lipid residues demonstrate that herders at the Lux- this time frame provides a relatively long period in which se- manda and Ngamuriak settlement sites used their pottery pre- lection for the LP trait may have occurred within multiple groups dominantly for processing ruminant carcass products, and to a in eastern Africa, which supports genetic estimations. Settings lesser extent for processing milk. Comparison with lipid residues for selection differed with social and economic organization and extracted from ceramics used by heavily milk-reliant Samburu environment. Diversified reliance on livestock, fishing, and hunting

Grillo et al. PNAS | May 5, 2020 | vol. 117 | no. 18 | 9797 Downloaded by guest on September 30, 2021 13 13 13 Fig. 4. Graphs showing the Δ C(δ C18:0 – δ C16:0) values from the potsherds at each of the sampled sites: Jarigole, Dongodien, Luxmanda, and Ngamuriak. The ranges shown here represent the mean ± 1SDoftheΔ13C values for a global database comprising published modern reference animal fats from Africa

(25), UK (animals raised on a pure C3 diet) (21), Kazakhstan (46), Switzerland (47), and the Near East (48).

at Lake Turkana 5,000 y ago provided herders in this arid region and an internal standard, typically 40 μg, was added to enable quantification of the with alternatives to reliance on milk or slaughtering of livestock lipid extract (n-tetratriacontane; Sigma Aldrich Company Ltd). Following the addi- δ13 during times of seasonal shortage or drought. Conversely, lipid tion of 5 mL H2SO4/MeOH2%to4%vol/vol( C of MeOH measured), the culture residue and faunal data reveal that pastoralists who lived at Nga- tubes were placed on a heating block for 1 h at 70 °C, mixing every 10 min. Once cooled, the methanolic acid was transferred to test tubes and centrifuged at muriak and Luxmanda in the rich grazing lands of southwestern 2,500 rpm for 10 min. The supernatant was then decanted into another furnace Kenya and northern Tanzania relied almost exclusively on their culture tube (II) and 2 mL dichloromethane (DCM) extracted double-distilled water livestock, and possibly plants. Given pastoralist specialization, se- was added. To recover any lipids not fully solubilized by the methanol solution, 2 × lection for LP in east Africa (the C-14010 allele) may have been 3mLn-hexane was added to the extracted potsherds contained in the original stronger among specialized pastoral groups after 3100 BP in more culture tubes, mixed well, and transferred to culture tube II. The extraction was southerly regions than earlier at Lake Turkana in the north. transferred to a clean, furnaced 3.5-mL vial and blown down to dryness. Following × The earliest herders in northern Kenya 5,000 y ago drank milk this, 2 2mLn-hexane was added directly to the H2SO4/MeOH solution in culture or ate milk-based products, but ceramic vessels with preserved tube II and whirlimixed to extract the remaining residues. This was transferred lipid residues, found at both ceremonial and settlement sites, were to the 3.5-mL vials and blown down until a full vial of solvent remained. Aliquots of the lipid extracts were derivatized using N,O-bis(trimethylsilyl)trifluoroacetamide predominantly used for processing or serving carcass products (BSTFA) containing 1% vol/vol trimethylchlorosilane (TMCS; Sigma Aldrich Company from livestock. Later specialized herders in southwest Kenya and Ltd.; 20 μL; 70 °C, 1 h). Excess BSTFA was removed under nitrogen, and the north-central Tanzania c. 3000 to 2000 BP intensively processed derivatized FAME was dissolved in hexane before GC, GC-MS, and GC-C-IRMS. livestock carcass fats in ceramic vessels, similar to how Samburu All FAMEs initially underwent high-temperature GC using a gas chro- herders use ceramics today. These results reveal the changing matograph fitted with a high-temperature nonpolar column (DB1-HT; 100% significance of livestock, and their milk, to herder diets and have dimethylpolysiloxane, 15 m × 0.32 mm i.d., 0.1 μm film thickness). The carrier broader relevance for understanding the biological histories of gas was helium and the temperature program comprised a 50 °C isothermal −1 herders in eastern Africa. These histories are likely the result of hold followed by an increase to 350 °C at a rate of 10 °C·min followed by a gene-culture coevolution involving complex interplays among mi- 10-min isothermal hold. A procedural blank (no sample) was prepared and analyzed alongside every batch of samples. Further compound identification grations, diverse livestock management strategies, local ecologies, was accomplished using GC-MS. FAMEs were then introduced by autosam- and dietary and culinary adaptations spanning thousands of years. pler onto a GC-MS fitted with a nonpolar column (HP-1; 100% dimethyl polysiloxane stationary phase; 60 m × 0.25 mm i.d., 0.1 μm film thickness). Materials and Methods The instrument was a ThermoFinnigan single quadrupole TraceMS run in EI Lipid analysis and interpretations were performed using established protocols de- mode (electron energy 70 eV, scan time of 0.6 s). Samples were run in full scribed in detail in earlier publications (e.g., refs. 21 and 22). All solvents used were scan mode (m/z 50 to 650), and the temperature program comprised an − HPLC grade (Rathburn), and the reagents were analytical grade (typically >98% of isothermal hold at 50 °C for 2 min, ramping to 300 °C at 10° C·min 1,fol- purity). Briefly, ∼2 g of potsherd were sampled and surfaces cleaned with a mod- lowed by an isothermal hold at 300 °C (15 min). Data acquisition and pro- eling drill to remove any exogenous lipids. The sherds were ground to a powder cessing were carried out using HP Chemstation software (Rev. B.03.02 [341],

9798 | www.pnas.org/cgi/doi/10.1073/pnas.1920309117 Grillo et al. Downloaded by guest on September 30, 2021 Agilent Technologies) and Xcalibur software (version 3.0). Peaks were scholarship on cattle pastoralism in eastern Africa continues to inspire. K.M.G. identified on the basis of their mass spectra and GC retention times by and M.E.P. thank the Tanzanian Commission for Science and Technology and comparison with the NIST mass spectral library (version 2.0). the Division of Antiquities for permission to excavate at Luxmanda and to Carbon isotope analyses by GC-C-IRMS were also carried out using a GC export ceramic sherds. Export permits were also issued by the National Museums of Kenya to K.M.G. for export of material from Jarigole and Dongo- Agilent Technologies 7890A coupled to an Isoprime 100 (EI, 70eV, three Faraday dien. Permits for export of Ngamuriak sherds were issued to K.R. Analyses of cup collectors m/z44, 45, and 46) via an IsoprimeGC5 combustion interface with a Dongodien and Jarigole ceramics were funded by NSF Doctoral Dissertation CuO and silver wool reactor maintained at 850 °C. Instrument accuracy was Improvement Grant #0752042 to K.M.G. and F.M. Fieldwork at Luxmanda was determined using an external FAME standard mixture (C11,C13,C16,C21,andC23) funded by a University of Wisconsin–La Crosse Faculty Research Grant to K.M.G of known isotopic composition. Samples were run in duplicate and an average in 2015, and by a Wenner-Gren Foundation grant to K.M.G. and a National taken. The δ13C values are the ratios 13C/12C and expressed relative to the Vienna Geographic Society grant (NGS-196R-18) to M.E.P. in 2018. J.K. received a UWL

Pee Dee Belemnite, calibrated against a CO2 reference gas of known isotopic Undergraduate Research Grant. Additional thanks to Natural Environment composition. Instrument error was ±0.3‰. Data processing was carried out using Research Council for partial funding of the National Environmental Isotope Ion Vantage software (version 1.6.1.0, Elementar). These methods have pre- Facility (Bristol laboratory), the Leverhulme Trust (RPG-2016-115) for a research grant to R.P.E. supporting a research fellowship to J.D., the European Research viously been detailed in, for example, ref. 39. All data discussed in this article will Council (H2020 ERC-2018-PoC/812917) for a research grant to R.P.E. supporting be made available to readers. a research fellowship to E.C., Helen Grant of the National Environmental Iso- tope Facility (Lancaster laboratory) for stable isotopic characterization of ref- ACKNOWLEDGMENTS. We would like to express our deep gratitude to erence standards and derivatizing agents, and Ian Bull and Alison Kuhl for Kathleen Ryan, who initially spearheaded this project. Her ethnographic technical help.

1. R. S. Reid, K. A. Galvin, R. S. Kruska, “Global significance of extensive grazing lands 25. J. Dunne et al., First dairying in green Saharan Africa in the fifth millennium BC. and pastoral societies: An introduction” in Fragmentation in Semi-Arid and Arid Nature 486, 390–394 (2012). Landscapes: Consequences for Human and Natural Systems, K. A. Galvin, R. S. Reid, 26. M. E. Prendergast, A. Janzen, M. Buckley, K. M. Grillo, Sorting the sheep from the R. H. B. Jr, N. T. Hobbs, Eds. (Springer Netherlands, 2008), pp. 1–24. goats in the Pastoral Neolithic: Morphological and biomolecular approaches at Luxmanda, 2. L. Ségurel, C. Bon, On the evolution of lactase persistence in humans. Annu. Rev. Tanzania. Archaeol. Anthropological Sci. 11,3047–3062 (2019). Genomics Hum. Genet. 18, 297–319 (2017). 27. A. Janzen, M. Balasse, S. H. Ambrose, Early pastoral mobility and seasonality in Kenya 3. F. Marshall, K. Grillo, L. Arco, “Prehistoric pastoralists and social responses to climatic assessed through stable isotope analysis. J. Archaeol. Sci., in press. risk in East Africa” in Sustainable Lifeways: Cultural Persistence in an Ever-Changing 28. P. E. Kolattukudy, Plant waxes. Lipids 5, 259–275 (1970). Environment, N. F. Miller, K. M. Moore, K. Ryan, Eds. (University of Pennsylvania 29. M. Maffei, Chemotaxonomic significance of leaf wax alkanes in the gramineae. Biochem. Museum, 2011), pp. 39–74. Syst. Ecol. 24,53–64 (1996). 4. C. Jeong et al., Bronze Age population dynamics and the rise of dairy pastoralism on 30. F. Rommerskirchen, A. Plader, G. Eglinton, Y. Chikaraishi, J. Rullkötter, Chemotaxo- the eastern Eurasian steppe. Proc. Natl. Acad. Sci. U.S.A. 115, E11248–E11255 (2018). nomic significance of distribution and stable carbon isotopic composition of long- 5. R. P. Evershed et al., Earliest date for milk use in the Near East and southeastern Europe chain alkanes and alkan-1-ols in C4 grass waxes. Org. Geochem. 37, 1303–1332 (2006). linked to cattle herding. Nature 455,528–531 (2008). 31. J. Dunne, A. M. Mercuri, R. P. Evershed, S. Bruni, S. di Lernia, Earliest direct evidence of ANTHROPOLOGY 6. K. A. Galvin, Food Procurement, Diet, Activities and Nutrition of Ngisonyoka, Turkana plant processing in prehistoric Saharan pottery. Nat. Plants 3, 16194 (2016). Pastoralists in an Ecological and Social Context (State University of New York at 32. F. A. Hansel, M. S. Copley, L. A. S. Madureira, R. P. Evershed, Thermally produced ω-(o- Binghamton, 1985). alkylphenyl)alkanoic acids provide evidence for the processing of marine products in 7. K. A. Galvin, Nutritional ecology of pastoralists in dry tropical Africa. Am. J. Hum. Biol. archaeological pottery vessels. Tetrahedron Lett. 45, 2999–3002 (2004). – 4, 209 221 (1992). 33. F. A. Hansel, R. P. Evershed, Formation of dihydroxy acids from Z-monounsaturated 8. J. Holtzman, Uncertain Tastes: Memory, Ambivalence, and the Politics of Eating in alkenoic acids and their use as biomarkers for the processing of marine commodities Samburu, Northern Kenya (University of California Press, 2009). in archaeological pottery vessels. Tetrahedron Lett. 50, 5562–5564 (2009). 9. K. M. Grillo, Pastoralism and pottery use: An ethnoarchaeological study in Samburu, 34. L. J. E. Cramp et al., Immediate replacement of fishing with dairying by the earliest Kenya. Afr. Archaeol. Rev. 31, 105–130 (2014). farmers of the Northeast Atlantic archipelagos. Proc. Biol. Sci. 281, 20132372 (2014). 10. M. Fujita, E. A. Roth, M. A. Nathan, E. Fratkin, Sedentism, seasonality, and economic 35. F. B. Marshall, J. W. Barthelme, K. Stewart, Early domestic stock at Dongodien in status: A multivariate analysis of maternal dietary and health statuses between pas- northern Kenya. Azania 19, 120–127 (1984). toral and agricultural Ariaal and Rendille communities in northern Kenya. Am. J. Phys. 36. F. Marshall, “Cattle herds and caprine flocks” in Early Pastoralists of South-Western Anthropol. 123, 277–291 (2004). Kenya, P. Robertshaw, Ed. (British Institute in Eastern Africa, 1990), pp. 205–260. 11. K. Sadler, C. Kerven, M. Calo, M. Manske, A. Catley, The fat and the lean: Review of 37. K. M. Grillo et al., Pastoral Neolithic settlement at Luxmanda, Tanzania. J. Field Archaeol. production and use of milk by pastoralists. Pastoralism 1, 291–324 (2010). 43,102–120 (2018). 12. S. A. Tishkoff et al., Convergent adaptation of human lactase persistence in Africa and 38. J. Dunne, K. M. Grillo, E. Casanova, H. L. Whelton, R. P. Evershed, Pastoralist foodways Europe. Nat. Genet. 39,31–40 (2007). recorded in organic residues from pottery vessels of modern communities in Samburu, 13. A. Ranciaro et al., Genetic origins of lactase persistence and the spread of pastoralism Kenya. J. Archaeol. Method Theory (2018). in Africa. Am. J. Hum. Genet. 94, 496–510 (2014). 39. J. Dunne et al., Pots, plants and animals: Broad-spectrum subsistence strategies in the 14. M. E. Prendergast et al., Ancient DNA reveals a multistep spread of the first herders Early Neolithic of the Moroccan Rif region. Quat. Int., in press. into sub-Saharan Africa. Science 365, eaaw6275 (2019). 40. C. M. Nelson, The work of the Koobi Fora Field School at the Jarigole Pillar Site. Kenya 15. E. A. Hildebrand et al., A monumental cemetery built by eastern Africa’s first herders Past Present 927,4 –63 (1995). near Lake Turkana, Kenya. Proc. Natl. Acad. Sci. U.S.A. 115, 8942–8947 (2018). 41. J. Barthelme, Fisher-hunters and Neolithic Pastoralists in East Turkana, Kenya (Oxford, 16. K. M. Grillo, E. A. Hildebrand, The context of early megalithic architecture in eastern 1985). Africa: The Turkana Basin c. 5000-4000 BP. Azania 48, 193–217 (2013). 42. P. Robertshaw, Early Pastoralists of South-Western Kenya (British Institute in Eastern 17. K. L. Chritz et al., Climate, ecology, and the spread of herding in eastern Africa. Quat. Sci. Rev. 204, 119–132 (2019). Africa, 1990). 18. F. Marshall, Origins of specialized pastoral production in East Africa. Am. Anthropol. 43. M. E. Prendergast et al., Pastoral Neolithic sites on the southern Mbulu Plateau, – 92, 873–894 (1990). Tanzania. Azania 48, 498 520 (2013). 19. F. Marshall, “The land of milk and honey: Food, identity and resilience in ancient East 44. M. C. Langley, M. E. Prendergast, K. M. Grillo, Organic technology in the Pastoral Africa” in The Origin and Evolution of Food Production and its Impact on Con- Neolithic: Osseous and eggshell artefacts from Luxmanda, Tanzania. Archaeol. Anthropol. – sumption Patterns, N. Sanz, Ed. (UNESCO, 2018), pp. 100–111. Sci. 11,114 (2019). “ 20. E. Casanova, T. D. J. Knowles, C. Williams, M. P. Crump, R. P. Evershed, Practical 45. S. Wandibba, The application of attribute analysis to the study of Later Stone Age/ ” considerations in high precision compound-specific radiocarbon dating: Eliminating Neolithic pottery ceramics in Kenya (summary) in Proceedings of the 8th Panafrican the effects of solvent and sample cross-contamination on accuracy and precision, Congress of Prehistory and Quaternary Studies, Nairobi, 5 to 10 September 1977, Anal. Chem. 90, 11025–11032 (2018). Leakey, Ogot, Eds. (International Louis Leakey Memorial Institute for African Pre- 21. S. N. Dudd, R. P. Evershed, Direct demonstration of milk as an element of archaeo- history, 1980), pp. 283–285. logical economies. Science 282, 1478–1481 (1998). 46. A. K. Outram et al., The earliest horse harnessing and milking. Science 323, 1332–1335 22. M. Correa-Ascencio, R. P. Evershed, High throughput screening of organic residues in (2009). archaeological potsherds using direct acidified methanol extraction. Anal. Methods 6, 47. J. E. Spangenberg, S. Jacomet, J. Schibler, Chemical analyses of organic residues in 1330 (2014). archaeological pottery from Arbon Bleiche 3, Switzerland – evidence for dairying in 23. R. P. Evershed, Experimental approaches to the interpretation of absorbed organic the late Neolithic. J. Archaeol. Sci. 33,1–13 (2006). residues in archaeological ceramics. World Archaeol. 40,26–47 (2008). 48. M. W. Gregg, E. B. Banning, K. Gibbs, G. F. Slater, Subsistence practices and pottery 24. R. P. Evershed et al., Chemistry of archaeological animal fats. Acc. Chem. Res. 35, 660– use in Neolithic Jordan: Molecular and isotopic evidence. J. Archaeol. Sci. 36, 937– 668 (2002). 946 (2009).

Grillo et al. PNAS | May 5, 2020 | vol. 117 | no. 18 | 9799 Downloaded by guest on September 30, 2021