Emergence and intensification of dairying in the Caucasus and Eurasian steppes

The study addresses when and how dairying emerged and intensified among Pontic-Caspian steppe populations and how animal management strategies influenced their mobility and later migrations. Dairying became a vital component of Holocene food systems, with milk providing nutrient-rich, storable products via fermentation. While ruminant dairying is first attested in Anatolia in the 7th–6th millennia BC and spread to Europe and Africa by the late 6th millennium BC, its initial dispersal into Asia and the steppe is less well known. The Pontic-Caspian steppe served as a major vector for the spread of mobile pastoralism and dairying and is linked to Afanasievo-related expansions into Mongolia and to Yamnaya-related expansions into Europe (Corded Ware and Bell Beaker). Yet, the timing of dairying adoption, the focal livestock species, and the economic underpinnings of Eneolithic and Early Bronze Age mobility in this source region remained unclear. Prior subsistence in the southern Russian plain and Caucasus was dominated by hunting; domesticates (sheep, goats, cattle, pigs) arrived by the 5th millennium BC from Anatolia. Subsequently, mobile pastoralism focused on sheep and cattle emerged on the steppe (Late Maykop, Yamnaya), with horses gaining importance in the late 3rd–early 2nd millennia BC. Climatic deterioration around the 4.2 ka event led to regional depopulation by 1700 BC. Despite genetic similarities among Bronze Age steppe groups, precise economic practices, especially dairying, were poorly resolved due to limited settlement evidence and the focus on mortuary contexts. This study leverages dental calculus proteomics to directly detect milk consumption and reconstruct dairying strategies across regions and periods.

Previous research shows early dairying in Anatolia (7th–6th millennia BC) and its spread to Europe and Africa by the late 6th millennium BC. The Eurasian steppe has been implicated in transmitting dairying eastward, with dairying in Mongolia introduced ca. 3000 BC alongside Afanasievo herders linked to Yamnaya. Steppe populations are associated with the westward spread of Corded Ware and Bell Beaker cultures. Archaeofaunal and isotopic studies across the Pontic-Caspian steppe indicate dominance of sheep, goat, and cattle in the 4th–2nd millennia BC and suggest a stronger contribution of caprine products to human diets; however, isotopic data could not directly identify dairy consumption and reflect complex ecological variability. Zooarchaeological studies in the North and South Caucasus debated the timing of dairying, with clear, systematic evidence for targeted milk production largely appearing in the later Bronze and Iron Ages. Settlement scarcity on the steppe and focus on kurgans hindered reconstruction of dairying’s extent. Proteomic analysis of dental calculus has previously provided direct evidence of milk consumption in other regions, motivating its application here to address gaps in the steppe and Caucasus.

Study design and sampling: Dental calculus from 45 individuals spanning the Neolithic to Greco-Roman periods (ca. 6000 BC–200 AD) was analysed from 29 sites across four regions: North Caucasus (n=27), South Caucasus (n=9), Oka-Volga-Don (n=7), and East Urals (n=2). Sampling took place in museums/archaeological institutions and at a dedicated ancient biomolecules laboratory, using nitrile gloves and sterile curettes. Approximately 5–13 mg of calculus per individual was used. Radiocarbon dating: 24 new AMS dates were generated at multiple laboratories (Mannheim; Helsinki; Oxford; Moscow), and 21 previously published dates were compiled, giving 38 directly dated individuals (45 total dates). Calibration used OxCal v4.4 with IntCal20. Proteomics: Proteins were extracted using a filter-aided sample preparation protocol optimized for ancient proteins. Calculus was demineralized (0.5 M EDTA); proteins were solubilized/reduced (4% SDS, 0.1 M DTT, 0.1 M Tris HCl), buffer-exchanged (8 M urea), alkylated (iodoacetamide), and digested overnight with trypsin in 0.05 M TEAB at 37 °C. Peptides were acidified and desalted (C18 stage tips). LC–MS/MS was performed on a Q-Exactive coupled to a Waters ACQUITY UPLC M-Class. MS1 (300–1700 m/z, 70,000 resolution at 200 m/z, AGC 3e6, 110 ms); DDA of top 12 ions with HCD (NCE 25), MS2 at 35,000 resolution; dynamic exclusion 30 s; singly charged/unassigned precursors excluded. Chromatographic separation used a trap (Symmetry C18) and analytical column (HSS T3 C18, 75 μm × 250 mm) at 50 °C with a 73-min gradient. Data processing: Spectra were converted with MSConvert and searched with Mascot (SwissProt 2019_08) using trypsin specificity, parent tolerance 10 ppm, fragment tolerance 0.050 Da, one missed cleavage; fixed carbamidomethyl (C); variable deamidation (N/Q) and oxidation (M/P). Validation in Scaffold v4.9.0 with PeptideProphet (>90% peptide probability) and ProteinProphet (>95% protein probability, ≥2 unique peptides). Proteins sharing evidence were grouped by parsimony. Non-template extraction controls monitored contamination; no dietary proteins were detected in controls. Identification of dairy proteins targeted beta-lactoglobulin (BLG), alpha-lactalbumin, and alpha-S1-casein; taxonomic assignment relied on diagnostic BLG peptides discriminating Ovis, Capra, Bovinae, and Equus.

• Milk proteins were identified in 34 of 45 individuals overall. BLG was detected in all 34 milk-positive individuals; alpha-lactalbumin in 2 and alpha-S1-casein in 2 individuals.
• North Caucasus: Milk proteins were present in 26 of 27 individuals (96%), with high peptide spectral matches per individual (mean 47 ± 27 PSMs). Earliest evidence (PG2001, 4338–4074 BC) shows dairying in the late 5th millennium BC. Eneolithic, Early and Late Maykop, Steppe Late Maykop, and early Yamnaya individuals consumed exclusively sheep (Ovis) milk. From ca. 2800 BC, dairying diversified to include goat (Capra) and cattle (Bos/Bovinae) alongside sheep in late Yamnaya, NCC, Catacomb, late NCC, and Lola/post-Catacomb individuals, with many consuming products from two or three species. Early Iron Age evidence includes horse (Equus) milk (individual MK5018).
• South Caucasus: Milk proteins detected from the 4th millennium BC onward. The earliest milk-positive individual (ALX002, 3776–3651 BC) shows cattle (Bovinae) milking nearly 1,000 years earlier than earliest cattle milk evidence in the North Caucasus (ca. 2700 BC). Across periods, cattle, goat, and sheep milk were identified; no horse milk detected.
• Oka-Volga-Don: Limited evidence of dairying; 6 of 7 individuals were negative for milk proteins across Eneolithic to Middle Bronze Age. Only RVK001 (late Catacomb, 2339–2148 BC) was positive for sheep, goat, and cattle milk.
• East Urals: Two Srubnaya-Alakul individuals (ca. 1900–1600 BC) yielded milk proteins: NEP008 (sheep and cattle) and NEP013 (non-specific bovid).

The findings directly resolve the antiquity and species focus of early dairying in the North Caucasus, demonstrating sheep-focused dairying among Eneolithic groups by the late 5th millennium BC, likely introduced with livestock from the south, and adopted by local transitional foragers regardless of ancestry. During the Maykop and Steppe Maykop periods, dairying remained specialized on sheep, while cattle were primarily used for traction and social display, explaining the disparity between archaeological visibility of cattle and proteomic evidence of sheep dairy consumption. Around the Yamnaya horizon and into the Middle Bronze Age, climatic aridification and environmental stress coincided with diversification of dairy herds to include goat and cattle, enabling exploitation of varied steppe environments and potentially driving increased mobility and broader cultural expansions (e.g., Corded Ware). In the Lola period, heightened drought likely reduced the dairying importance of water-demanding cattle. After a centuries-long depopulation of the steppe, Early Iron Age groups reintroduced a broad mobile dairy economy and adopted horse milking, aligning with historical accounts of steppe nomads and the ecological suitability of horses. Regionally, the South Caucasus shows earlier cattle milking than the North, reflecting environmental and cultural differences, while the Oka-Volga-Don region exhibits a later adoption of ruminant dairying consistent with its forager-based economies transitioning to agropastoralism in the Middle Bronze Age. Overall, the proteomic evidence clarifies the temporal and taxonomic patterning of dairying and links economic strategies to climate and mobility.

Proteomic analysis of dental calculus reveals that dairying was integral to pastoralist economies in the North Caucasus and adjacent regions from the Eneolithic through the Greco-Roman periods. Initial strategies emphasized sheep dairying with cattle used for traction; fully mobile pastoralism emerged with Yamnaya. As climate deteriorated during the Middle and Late Bronze Ages, herders diversified dairy livestock (sheep, goats, cattle) and expanded mobility, ultimately leading to steppe abandonment in the mid-second millennium BC. The steppe was later repopulated in the Early Iron Age with resumed ruminant dairying and the addition of horse milking. These shifts in dairying practices and herd composition illuminate the economic foundations behind third-millennium BC mobility and expansions across Eurasia. Future research integrating ancient livestock genomics and expanded proteomic datasets across the steppe will refine our understanding of the origins and dispersal of dairying breeds and technologies.

Reconstruction is constrained by the scarcity of settlements and reliance on mortuary contexts in the steppe, potentially biasing samples. Regional sample sizes are limited, especially for key periods (e.g., Kura-Araxes in the South Caucasus) and for the Oka-Volga-Don region, where only one individual was milk-positive. Proteomic detection reflects consumed proteins preserved in calculus and may underrepresent low-frequency or seasonally consumed milks. Some comparative evidence for early horse milking relies on sites with problematic dates (e.g., Kriviyansky IX), and direct evidence for certain cultural phases and regions remains sparse. Environmental and cultural variability across sites could introduce heterogeneity not fully captured by the dataset.