The rise and transformation of Bronze Age pastoralists in the Caucasus

The study investigates how Bronze Age steppe pastoralist societies formed, expanded, and eventually transformed around the Caucasus, a key interface between Europe and Asia. The region’s complex ecologies and role as a corridor and barrier fostered diverse cultural and genetic processes from the Mesolithic through the Late Bronze Age. Despite the recognized impact of Caucasus-steppe populations on Eurasian genetics, languages, and cultures, their origins from local hunter-gatherers, interactions with early farmers from the Fertile Crescent, and the processes underlying their decline remain insufficiently resolved. This work aims to reconstruct population structure and dynamics across ~6,000 years, clarifying the emergence of steppe ancestry, the role of Maykop and Kura-Araxes networks, and the nature of mountain–steppe interactions leading to the consolidation of pastoral economies and later absorption into highland populations.

Prior research identified major Holocene population turnovers and admixture in Eurasia, including the role of steppe populations in Europe and Asia and the presence of Caucasus Hunter-Gatherer (CHG) ancestry. Earlier data from the Caucasus suggested eco-geographic structuring and contributions from Anatolian and Levantine Neolithic groups, as well as Eastern Hunter-Gatherers (EHG). Studies on Yamnaya expansions, dairying, and wheeled transport framed steppe mobility and influence. However, sampling gaps and coarse temporal resolution limited understanding of early Eneolithic formation of steppe ancestry, the extent of Maykop-related interactions, and the Late Bronze Age transformations. This study builds on and refines these findings by tripling available genomic data for the region, adding radiocarbon control, and applying fine-scale modeling (qpAdm, f-statistics, DATES, ancIBD, ROH).

• Sampling: Generated new genome-wide data for 131 ancient individuals from 38 archaeological sites across and around the Caucasus (piedmont, steppe, and highlands), spanning ~6,000 years (Mesolithic/Neolithic to Late Bronze Age). Obtained 84 new radiocarbon dates. After quality control and removal of related or insufficient-quality individuals (n=26 excluded), 102 unrelated individuals were retained for population-genetic analyses and merged with published ancient and present-day datasets.
• Chronological coverage: Mesolithic/Neolithic (7th–6th millennia BC, n=7), Eneolithic (5th millennium BC, n=11), Late Eneolithic/Early BA (4th millennium BC, n=20), Early and Middle BA (3rd millennium BC, n=51), final MBA and LBA (2nd millennium BC, n=42).
• Population-genetic analyses: Principal component analysis (PCA) projecting ancient individuals onto modern reference space; ADMIXTURE for ancestry components; outgroup f- and D-statistics to test affinities; qpAdm to model admixture using distal and proximal sources; DATES to estimate admixture times; ancIBD to detect up to 6th-degree relatedness; runs of homozygosity (ROH) to infer parental relatedness and effective population sizes.
• Modeling sources: Distal sources included Anatolia_PPN, Levant_PPN, CHG-Iran_N; proximal sources included Georgia_Neolithic, CHG, Caucasus_Eneolithic, Steppe_Eneolithic, Maykop (and subgroups), Armenia_C, Iran_C, WSHG and Botai-related groups, Ukraine_N/Meso, CTC-GAC, Yamnaya groups, Kura-Araxes, Catacomb, NCC, Srubnaya, Sintashta, Kazakhstan Kumsay EBA, among others.
• Site- and mound-level analyses: Relatedness within multiphase kurgans; evaluation of sex ratios; assessment of genetic continuity or turnover across cultural horizons within specific mounds.
• Interpretation integrated with archaeological context (material culture, mobility technologies, subsistence such as dairying).

• Early structure: Strong genetic differentiation across the Greater Caucasus during the Mesolithic: north characterized by EHG ancestry (e.g., SJG001) and south by CHG-related ancestry which later admixed with Anatolian Neolithic ancestry.
• Neolithic cline south of the mountains: Georgia_Neolithic and Armenia Neolithic groups fall on a cline between CHG and central Anatolian Neolithic (e.g., Çatalhöyük). Distal modeling supports mixtures of Anatolia_PPN, Levant_PPN, and CHG-Iran_N; Georgia_Neolithic can be modeled as a two-way mixture CHG + Çatalhöyük_N or Tell Kurdu.
• Formation of steppe ancestry: Earliest Steppe_Eneolithic in the North Caucasus formed by admixture of CHG-like and EHG-like ancestries, modeled at ~55% CHG-like and ~45% EHG-like. DATES estimates: Anatolia Neolithic–CHG/Iran_N cline formed ~6300–6000 BC; EHG–CHG cline formed ~5800–5300 BC.
• Eneolithic intermediates: Nalchik individuals (4930–4359 cal BC) fall between Steppe and Caucasus clusters indicating early fifth-millennium BC gene flow. KHB003 (4318–4057 cal BC) shows increased WHG/Anatolia_N affinity and can be modeled as CHG + Ukraine_Neolithic.
• Maykop-period structure (4th millennium BC): Three previously known clusters (Maykop_main, Steppe_Maykop, Steppe_Maykop_outlier1) confirmed; three new groups identified (Late_Steppe_Eneolithic, Late_Steppe_Eneolithic_outlier, Steppe_Maykop_outlier2). Maykop_main shows continuity with Caucasus_Eneolithic but requires an added Iran_C-related source in qpAdm (fit improves with Iran_C as third source). Steppe_Maykop carries substantial WSHG/ANE-related ancestry (up to 48%), unlike other contemporaneous groups. Steppe_Maykop_outlier2 modeled as ~62% Steppe_Maykop + ~38% Maykop_Novosvobodnaya. Late_Steppe_Eneolithic modeled at ~52% SJG001-like + ~48% CHG; Late_Steppe_Eneolithic_outlier modeled ~55% Caucasus_Eneolithic + ~45% Steppe_Eneolithic.
• Yamnaya and MBA continuity (3rd millennium BC): Ten new Yamnaya_North_Caucasus (Yamnaya_NC) individuals form a tight cluster with other Yamnaya; two-way models with CTC-GAC or Maykop_main only fit for western Ukraine_Yamnaya, not Yamnaya_NC/Samara. Three-way models (Steppe_Eneolithic + Maykop + Ukraine_Neolithic/Mesolithic) fit Yamnaya_NC. North Caucasus Culture (NCC) and Catacomb individuals cluster with Yamnaya; Catacomb modeled with Yamnaya_NC as a single local source; NCC modeled with Ukraine_Yamnaya. Admixture date estimates for Steppe groups cluster around 4800–4000 BC.
• Kura-Araxes and Caucasus MBA: Georgian Kura-Araxes individuals show continuity with Maykop/Caucasus ancestry; well-fitted models with Maykop (often Maykop_Novosvobodnaya as best single source), though some sites require additional Armenia_C or Iran_C ancestry. Iran_BA individuals near Tepe Hissar modeled with Iran_TepeHissar_C + Kura-Araxes.
• Final MBA and LBA transformations (2nd millennium BC): Steppe cluster space largely empty; most post-Catacomb individuals shift toward Caucasus, modeled as ~79% Catacomb + ~21% Kura-Araxes. Lola individuals (n=9) split into Lola_1 and Lola_2 by ANE levels; Steppe_Maykop-only continuity rejected; supported two-way models with Steppe_Maykop + NCC/Catacomb or local MBA Steppe + a Central Asian source (Kazakhstan Kumsay EBA). One individual (KVO013) aligns with Srubnaya; modeled as a mixture of BA Sintashta + Steppe. Prescythian_steppe individuals modeled as Srubnaya + Lola_1.
• Highland gene flow: MBA/LBA Caucasus highlanders shift toward Steppe in PCA; two-way models with Kura-Araxes (Caucasus proxy) + BA Steppe sources fit, with geographic cline (west showing more Srubnaya/Maykop affinity; east more Kura-Araxes). Armenian LBA (Lernakert) shows local continuity modeled with Armenia_MBA.
• Social structure and demography: Across 21 multi-burial kurgans (105 individuals), only 15 first-/second-degree relationships detected among 5,460 pairs (0.27%); even within-site (5.5%) or chrono-cultural overlap (1.3%) pairs largely unrelated, indicating kurgans generally not lineage-based burial grounds. Significant male burial bias in Steppe cluster (P=0.035) but not Caucasus (P=0.850). Steppe individuals show more short ROH (4–12 cM), indicating smaller effective population sizes; five Maykop_main cases of consanguinity (including first-cousin and sibling unions) identified.
• Macro-process: After the 4th millennium BC, Maykop innovations and contacts expanded but with limited intermarriage between piedmont Maykop_main and Steppe groups. Third-millennium BC sees homogenized Steppe ancestry (Yamnaya, NCC, Catacomb) and consolidation of mobile pastoralism. Second millennium BC shows steppe decline and abandonment (~1700 BC), re-emergent gene flow from Central Asia (ANE-bearing) and northwestern forest-steppe (Srubnaya), and absorption of Steppe ancestry into highland Caucasus populations.

Findings reveal two long-standing, genetically distinct populations separated by the Greater Caucasus: EHG-related groups to the north and CHG-related groups to the south, with early Neolithic admixture producing a CHG–Anatolia Neolithic cline in the south. Steppe ancestry emerged via EHG–CHG admixture in the mid-6th millennium BC, predating the Darkveti-Meshoko agropastoral expansion, and early interactions are evidenced by Eneolithic intermediates (e.g., Nalchik) and western-leaning outliers (KHB003). The Maykop phenomenon brought intense cultural exchange and technological innovations (wheeled transport, dairying, horse management) but limited gene flow from piedmont Maykop_main into Steppe groups; some Steppe groups acquired ANE-related ancestry from the northeast (Steppe_Maykop). During the third millennium BC, Steppe ancestry homogenized (Yamnaya, NCC, Catacomb) while Caucasus Kura-Araxes groups maintained continuity with Maykop-related ancestry and expanded south/east, sometimes incorporating Iran_C-like ancestry. The LBA marks renewed, multidirectional gene flow: Steppe groups shift toward Caucasus ancestry (post-Catacomb modeled with Kura-Araxes), while highlanders acquire Steppe ancestry, with regional structure along east–west gradients. Socially, kurgans were not primarily lineage-based; Steppe groups show male-biased burial and small effective sizes with regulated exogamy, whereas Maykop_main includes cases of close-kin unions, suggesting differing kinship structures. Overall, the results clarify the formation, consolidation, and eventual absorption of steppe pastoralist ancestry into mountain populations, reframing earlier hypotheses that emphasized mountain expansions into the steppe by highlighting instead the integration of Steppe groups into highland communities as steppe habitats deteriorated.

This study triples the genomic dataset for the Caucasus region, producing a 6,000-year transect that: (1) documents early Mesolithic divergence north/south of the Caucasus; (2) dates the formation of Steppe ancestry (EHG+CHG) to the mid-6th millennium BC; (3) disentangles Maykop-associated genetic continuity versus cultural transmission; (4) traces the homogenization and expansion of Yamnaya-related Steppe ancestry through the MBA; and (5) reveals LBA transformations with multi-source gene flow and the absorption of Steppe ancestry into highland Caucasus populations. The integrated archaeological-genetic framework explains how technological innovations and pastoral strategies underpinned mobility and interactions that reshaped Eurasia. Future research should refine radiocarbon chronology addressing reservoir effects, expand sampling in undersampled epochs and regions (especially for intermediate individuals during the EBA in both mountain and steppe zones), and include genomic studies of domesticated animals to track the spread of pastoral technologies and livestock lineages.

• Chronological uncertainty due to potential radiocarbon reservoir effects in Eneolithic and Bronze Age pastoralist contexts, complicating fine-grained sequencing of events. - Geographic and temporal sampling gaps, particularly for contemporaneous mountain and steppe populations during critical transitional periods (e.g., early Maykop/EBA), limit detection of genetically intermediate individuals. - Some ancestry inferences rely on proxy populations (e.g., Maykop_Novosvobodnaya, Kura-Araxes, WSHG/Botai), which may not perfectly represent true sources. - Limited power to reconstruct detailed kinship networks beyond detected degrees and to resolve subtle substructure within homogenized Steppe groups.