Ancient DNA from the Green Sahara reveals ancestral North African lineage

Main

Following the last glacial period, a climatic transformation in the Sahara desert led to the AHP, which peaked around 11,000 to 5,000 years ago^3,4. During this period of increased humidity, the region transformed into a ‘Green Sahara’ with savanna-like landscapes, varying tree cover, permanent lakes and extensive river systems⁵ (Fig. 1). Evidence from ancient lake deposits, pollen samples and archaeological artifacts confirm human presence, hunting, herding and resource gathering in the currently arid desert region^1,6,7. However, despite this rich history, much about the genetic history of the human population of the Green Sahara remains unclear due to limited DNA preservation under the current climatic conditions. Ancient DNA data from northwestern Africa points to a stable and isolated genetic population from at least 15,000 to 7,500 years ago^2,8. This stability was disrupted by the arrival of early farming groups from southwestern Europe between 7,500 and 5,700 years ago who marked the beginning of the Neolithic in the Maghreb by introducing farming practices to the local foragers⁹. The earliest herders with their livestock entered Africa probably along the Sinai and the Red Sea routes, after which they rapidly spread into northeastern Africa and reached the Central Sahara around 8,300 years ago¹⁰. By 6,400 years ago, further gene flow occurred with the appearance of ancestry associated with Neolithic groups from the Levant, whose archaeological signatures are visible in the Eastern Sahara^9,11,12. A previous study¹³ analysed mitochondrial DNA (mtDNA) from individuals recovered from the Takarkori rock shelter in the Tadrart Acacus Mountains of southwestern Libya—the same individuals examined in this study—providing the first ancient DNA from pastoralists of the Green Sahara. However, non-recombining and therefore effectively single genetic loci like mtDNA have much less statistical power to reveal population dynamics than genome-wide autosomal data. Their origins and whether the arrival of pastoralism into the Green Sahara was linked to the movement of peoples from the Levant or rather cultural diffusion remain a matter of debate^10,14.

Here we present first genome-wide data obtained from the same two approximately 7,000-year-old Saharan herders, recovered from the Takarkori rock shelter in the Central Sahara (Supplementary Figs. 1.2–1.5), a site that has yielded an exceptional wealth of data and material remains^15–19. Our findings show that these individuals predominantly carry a previously unknown ancestral North African lineage that lacks the Neanderthal admixture typically found outside Africa and appears to have remained largely isolated, with the notable exception of small traces of Levantine admixture. These results support that pastoralism in the Sahara was established through cultural diffusion¹⁰ rather than significant human gene flow. Furthermore, the Takarkori individuals exhibit a close genetic affinity to Northwestern African foragers but no substantial ties with sub-Saharan African lineages, implying no detectable genetic exchange across the Green Sahara during the AHP from sub-Saharan to northern Africa. For a non-peer-reviewed Arabic summary of the article, see Supplementary Note 3.

Takarkori: a window into the Green Sahara

The Takarkori rock shelter—situated in southwestern Libya’s Tadrart Acacus Mountains—offers a remarkable glimpse into the Sahara’s greener past^15,20. The excavations at this archaeological site revealed a timeline of human settlement from Late Acacus hunter-gatherer-fishers from around 10,200 calibrated years before the present (cal. bp) to a long Pastoral Neolithic occupation, dated from approximately 8,300 to 4,200  cal. bp^10,21. Data from these latter periods traces the sociocultural trajectories of Neolithic herding societies in Central Sahara, from early livestock introduction to the development of a full pastoral economy characterized by transhumance and the use of secondary product^22,23 (Supplementary Note 1). Within the deepest recess of the rock shelter, 15 human burials were unearthed, pertaining to a timeframe between approximately 8,900 and 4,800 cal. bp²², with the majority dating back to the Early (8,300–7,300 cal. bp) and Middle (7,100–5,600 cal. bp) Pastoral period²². Strontium isotope analysis on the remains, primarily of women of reproductive age, children and juveniles, indicated a local geographical origin²⁴. Two naturally mummified adult female individuals, attributed to the Middle Pastoral Period, were selected for our DNA analysis. These individuals were directly radiocarbon dated to 7,158–6,796 and 6,555–6,281 cal. bp (95.4% probability), respectively¹³ (Supplementary Note 1).

Genomic analyses

We extracted DNA from the powdered tooth root from individual TKH001 and from two fibula fragments from TKH009. Given their extremely low endogenous human DNA content (0.085–1.363%; Supplementary Table 2.1), we opted for a DNA capture approach to cost-effectively retrieve informative single-nucleotide polymorphisms (SNPs) for autosomal analysis. Sampling of other skeletal elements in the future may provide DNA extracts with higher percentage of endogenous DNA that allow for whole-genome shotgun sequencing. Using dedicated ancient DNA protocols (Methods), we prepared DNA libraries and enriched them through a DNA hybridization approach with the Twist Ancient DNA panel²⁵ targeting 1.4 million SNPs for TKH001 and the 1240K panel²⁶ targeting 1.2 million SNPs for TKH009. Despite challenging preservation conditions, sequencing yielded 881,765 SNPs for TKH001 and 23,317 SNPs for TKH009 (Supplementary Table 2.4). For TKH001, we conducted additional enrichment for 1.7 million SNPs, targeting sites informative about Neanderthal and Denisovan admixture (Archaic Admixture SNP capture panel)²⁷. DNA sequences from both Takarkori individuals had a post-mortem degradation pattern typical of ancient DNA and low contamination estimates (Supplementary Figs. 2.1, 2.2 and Supplementary Table 2.3).

To visualize the variation in the genetic ancestry of the Takarkori individuals, we performed a principal component analysis (PCA) using genome-wide data from 795 present-day individuals from the whole African continent, the Near East and Southern Europe, all genotyped on the Human Origins SNP panel^28–38 (Supplementary Data 1 and 2). We then projected the Takarkori individuals and 117 relevant published ancient genomes^{2,8,9,35,36,39–44} (Supplementary Data 2) onto the first two principal components (PCs). The Takarkori individuals plot broadly intermediate between West African and Near Eastern groups on PC1, albeit closer to West Africans (Extended Data Fig. 1, Supplementary Fig. 2.6 and Supplementary Data 3). To obtain a finer-scale view, we restricted the African populations to West Africa, the Sahel and East Africa, while retaining the Near Eastern and Southern European populations (Supplementary Data 3). This PCA captured the geographical distributions of these populations, whereby PC1 separates African from non-African populations and PC2 differentiates within Africa, particularly separating Sahel/West African from East African populations (Extended Data Fig. 2). Both Takarkori individuals maintained a distinct position, intermediate between Sahel/West and East African populations (Fig. 2a), and formed a tight cluster with overlapping 95% confidence interval (CI) ellipses (Supplementary Fig. 2.4). The PCA projections broadly mirror geography, including the placement of Takarkori. For subsequent population genetic analyses, we merged the low-coverage data from the TKH009 individual with the high-coverage data from the TKH001 individual. We note that many of the signals in this group analysis are probably driven by TKH001 owing to its much higher coverage.

To probe for shared genetic drift of Takarkori with other ancient and modern genomes, we computed outgroup f₃ statistics³⁰ of the form f₃(Takarkori, X; South Africa 2,000 cal. bp), with X representing worldwide ancient and present-day test populations and South Africa 2,000 cal. bp as the deepest modern human outgroup lineage⁴⁴. We found that the Takarkori individuals share the most genetic drift with Moroccan Palaeolithic and Early Neolithic individuals. Specifically, we observed the highest level with an Epipalaeolithic individual from Ifri Ouberrid and similarly elevated shared drift with the 15,000-year-old foragers from Taforalt and the Early Neolithic individuals from Ifri n’Amr o’Moussa (Fig. 3a, Extended Data Fig. 3 and Supplementary Fig. 2.12). Both Epipalaeolithic and Early Neolithic groups have been previously shown to maintain high genetic continuity with the much older Taforalt group^8,9, explaining the similarly elevated shared drift statistics between all three groups and the Takarkori individuals.

When restricting the comparative groups to present-day African and Near Eastern populations, we found that a group of individuals defined as FulaniA in a previous study³⁸, which includes individuals from relatively less admixed Fulani herders from eight Sahelian countries, shows an increased genetic affinity to Takarkori, although less so to Taforalt (Extended Data Fig. 4 and Supplementary Fig. 2.16). This finding aligns with ref. ³⁸, which observed a non-sub-Saharan ancestry component in FulaniA similar to that found in the Taforalt and Late Neolithic Moroccans. To further probe for Takarkori-like ancestry in FulaniA, we conducted an f₄ analysis f₄(chimpanzee, X; Masai/Datog/Iraqw, Takarkori), using Masai/Datog/Iraqw as baseline references owing to their similar amount of out-of-Africa (OoA) ancestry to Takarkori (Supplementary Fig. 2.20.1–2.20.3). The results indicated that the FulaniA have an increased affinity to Takarkori-like ancestry, as do other Sahelian and West African groups. These findings are consistent with the archaeological evidence of the southward expansion of Pastoral Neolithic groups from Central Sahara. Rock art, ceramic production and funerary practices provide detailed indications of the spread of these herders at the end of the Middle Holocene, probably driven by the progressive aridification of the Central Saharan regions^45,46.

Given the high amount of shared genetic drift between the Takarkori and the ancestry first appearing in the 15,000-year-old Taforalt individuals in the outgroup f₃ statistics, we subsequently investigated whether the Takarkori genomes share more alleles with other human groups compared to the Taforalt genome. For this analysis, we computed the statistics f₄(chimpanzee, X; Takarkori, Taforalt), where X represents ancient and present-day groups from Africa and Eurasia. We obtained significantly positive values for Eurasian and Eurasian-admixed African groups, suggesting that the Taforalt group shares more alleles with these groups than Takarkori does. Conversely, none of the ancient or modern groups tested showed a significantly negative signal, indicating no detected closer affinity with Takarkori than Taforalt. Notably, both ancient and modern sub-Saharan groups, who are mostly unadmixed with Eurasian groups, yielded no significant value (|Z| < 3), suggesting that these groups are equally distant from both the Takarkori and Taforalt groups (Fig. 3b and Supplementary Data 4). Moreover, when running statistics f₄(Chimp, Takarkori; X, Taforalt), we found significantly positive values for every population X, indicating that Takarkori shares more alleles with Taforalt than any other group (Extended Data Fig. 5 and Supplementary Data 4).

At the mtDNA level, both Takarkori individuals belong to a basal branch of haplogroup N, representing one of the deepest mtDNA lineages outside sub-Saharan Africa and predating present-day N-derived mtDNAs¹³. Using BEAST analysis for mtDNA dating, which included additional sequences from Upper Palaeolithic individuals and the dataset described previously¹³, we corroborate the previous findings of ref. ¹³ that the Takarkori individuals carried a basal N haplogroup lineage¹³, and refined the molecular split date estimate to 61,343 years old (95% highest posterior density (HPD) = 54,408–69,046) (Extended Data Fig. 6). Notably, the mtDNA lineage of the Oase 1 individual falls more basal to haplogroup N, suggesting an earlier split from the OoA lineage before the divergence of the Takarkori lineage. However, owing to incomplete lineage sorting and mtDNA representing a single lineage, the exact timing of the underlying population splits remains uncertain.

Previous work modelled the Taforalt group’s ancestry as a two-way admixture of approximately 63.5% Natufian (ancient Levantine foragers) and 36.5% sub-Saharan African ancestries². However, this model using the software qpAdm² could not pinpoint the origin of Taforalt’s African ancestry, resulting in unknown ghost ancestry only broadly linked to South, East and Central African groups². Here we included Takarkori as a possible source of the African ancestry in Taforalt in comparison to several potential sources (namely Yoruba, Dinka, Mota, Cameroon Shum Laka, Botswana Xaro Early Iron Age (EIA) and Tanzania Zanzibar 1,300 cal. bp) through rotation-based qpAdm. We found that Saharan Takarkori provides a much better fit as an African proxy for Taforalt than the sub-Saharan groups, attaining a P value of >0.05, indicative of a much better model fit compared to the other sources (P < 2.84 × 10⁻³⁴) (Extended Data Figs. 7, 8 and Supplementary Tables 2.6, 2.7). With this revised model, we estimated that the Taforalt ancestry retains a comparable 60.8% (±1.8%) contribution from Natufians, with the remaining 39.2% (±1.8%) derived from Takarkori.

We next explored the direct genetic affinity between Takarkori and the OoA ancestry found in all non-African modern humans, in contrast to African groups. For this, we computed f₄ statistics of the form f₄(chimpanzee, Zlatý kůň; African, Takarkori). Zlatý kůň, a 45,000-year-old Upper Palaeolithic individual from Czechia, was used as a proxy for OoA ancestry as it is probably the oldest modern human sequenced to date and represents the deepest known human lineage after the OoA lineage splits from the African lineages⁴⁷. Our results showed positive values for Takarkori, indicating that it is genetically closer to Zlatý kůň than to sub-Saharan Africans, including Mota, a 4,500-year-old genome from East Africa. Nevertheless, various African populations with substantial OoA admixture were still genetically closer to Zlatý kůň than to Takarkori (Extended Data Fig. 9 and Supplementary Data 5).

These results raise the question of whether Takarkori’s ancestors are closely related to OoA groups but remained in Africa, or whether they received later gene flow from OoA groups. If Takarkori experienced such later gene flow, it would carry Neanderthal admixture that is found in all OoA groups. To explore this signal, we used the data generated using the Archaic Admixture SNP panel for Takarkori and the software admixfrog⁴⁸ that detects Neanderthal segments and included other ancient African and Eurasian groups as well as modern sub-Saharan African populations for comparison (Supplementary Note 2). We detected a total of 12 Neanderthal fragments that surpass 0.05 cM in length (approximately 50 kb), with the longest fragment located on chromosome 1 (Extended Data Fig. 10 and Supplementary Fig. 2.32), translating to a low level of Neanderthal ancestry in the Takarkori genome of approximately 0.15% (Fig. 4a). This percentage is less than a quarter of the Neanderthal ancestry in segments longer than 0.05 cM found in Taforalt and Neolithic Morocco individuals (0.6–0.9%), and about tenfold lower than in most OoA groups (1.4–2.36%), yet significantly higher than that in other ancient and contemporary sub-Saharan African genomes, where Neanderthal ancestry was completely absent. This pattern suggests that the Takarkori individuals have received a small amount of ancestry from OoA groups. However, the estimation of admixture dating of this OoA ancestry with DATES⁴⁹ based on linkage disequilibrium (Supplementary Figs. 2.26.1–2.26.3) indicated very ancient admixture events with substantial uncertainty.

Finally, we used the find_graphs() function from the ADMIXTOOLS 2⁵⁰ package to model Takarkori’s ancestral relationship with other populations. We used a function for automated graph exploration with a model that includes Mota, Iran Neolithic, Natufian, Taforalt, Takarkori and the outgroup chimpanzee. The model fits with small f-statistic residuals (max |f_4,expected − f_4,observed| = 0.27 s.e.). The fitted graph suggests that Takarkori traces most of its ancestry (93%) to a hitherto unknown North African population, in agreement with the isolation signature obtained from the f₄ statistics mentioned above (Fig. 3b). This unknown North African population is closely related to OoA populations and branches off the lineage leading to OoA later than the ancient individual from Mota Cave, Ethiopia, who represents the sub-Saharan African lineage most closely related to OoA groups identified so far. In our model, the remainder of the Takarkori individuals’ ancestry (7%) is derived from a deeply ancient Levantine source. The Levantine gene flow also accounts for the Neanderthal ancestry found in Takarkori as the Levantine Neolithic genomes carry around 1.86% Neanderthal ancestry, aligning with our estimates using admixfrog. The graph also models Taforalt as a mixture of a 40% contribution from a Takarkori-related branch and 60% from a Natufian-related branch, consistent with our qpAdm results (Fig. 4c and Supplementary Data 6).

Discussion

Our study introduces ancient genome-wide data from humans that inhabited the Green Sahara, providing unique insights into the genomic ancestry of populations in this region. The individuals from the Takarkori rock shelter predominantly carry an ancestral African lineage, representing an ancestry profile that has not been previously described. They share the most genetic drift with Epipalaeolithic and Early Neolithic individuals from Ifri Ouberrid and Ifri n’Amr o’Moussa, as well as the 15,000-year-old foragers from the Taforalt Cave in Morocco, suggesting a long-standing and stable population in North Africa before the AHP (14,500–5,000 bp). Plausibly, this ancestry was present in large parts of Northern Africa after the OoA event, and the Takarkori individuals inherited it from the group that inhabited the area during the final period of the Late Acacus (10,200–8,000 cal. bp). In Southwestern Libya, this period preceded the arrival of domesticates and is characterized by cultural advancements within those hunter–gatherer groups¹⁰. This included an increase in sedentism⁵¹, and the use of sophisticated material cultures such as pottery, basketry, and bone and wooden tools^52–54.

We found that individuals from Takarkori exhibit only a marginal amount of genetic admixture from Levantine groups, suggesting that the emergence of pastoralism in the Sahara was primarily driven by the dissemination of cultural practices rather than through large-scale human migration, as suggested on archaeological basis¹⁰. Material culture at the onset of the earliest pastoral period shows both continuity and change, reflecting possibly complex assimilation dynamics during socioeconomic transitions^10,19. These patterns may further suggest gradual cultural transformations rather than abrupt population replacement. This interpretation is further supported by the comparatively lower levels of Neanderthal genetic admixture found in the Takarkori individuals. Our admixture dating analysis points to events far back in time, suggesting a more heterogeneous spread of pastoralism and food production in the Sahara compared to Morocco and East Africa, where there was a noticeable increase in recent Levantine genetic admixture^9,35,39,43. In addition to these findings, a run of homozygosity (ROH) analysis for TKH001 revealed no ROH segments larger than 12 cM, indicating no close-kin inbreeding (Supplementary Fig. 2.28). Several shorter ROH segments over 4 cM suggest an effective population size (N_e) of around 1,000 individuals, reflecting a moderately sized population.

Our research offers insights into the ancestry of the previously published Taforalt hunter-gatherers. While a previous study² could not precisely ascribe the ‘sub-Saharan’ component in the Taforalt genome, we now identify this ancestry as a deep North African lineage, with higher proportions found in the Saharan Takarkori individuals. This refines the earlier model, which proposed a dual admixture of Natufian and broadly sub-Saharan African ancestries. Our updated model suggests that the Taforalt ancestry is composed of a 60% contribution from a Natufian-like Levantine population, with the remaining 40% derived from a Takarkori-like ancestral North African population (Extended Data Fig. 8). Notably, both the late Pleistocene Taforalt and the mid-Holocene Takarkori individuals demonstrate equally distant relationships with sub-Saharan African lineages. This pattern suggests that no substantial genetic exchanges across the Green Sahara occurred during the AHP or other humid periods preceding the Later Pleistocene. The Sahara, spanning around 9 million km² and housing diverse biomes, such as grasslands, wetlands, woodlands, lakes, mountains and savannas^55,56, probably saw fragmented habitats impacting human gene flow. These ecological barriers, combined with social and cultural barriers, spatial structuring of populations, and the selective adoption of specific practices, may have facilitated the widespread dissemination of similar archaeological features^57,58, while limiting extensive genetic admixture. This genetic discontinuity is consistent with modern data, which show substantial genetic differentiation across the Sahara beyond just a geographical gap⁵⁹. Our findings suggest that sporadic Green Sahara events, particularly before pastoralism, were insufficient to allow for considerable genetic exchange, mirroring the Sahara’s persistent role in limiting human genetic flow, as reflected in both ancient and modern population structures.

Our findings represent an important initial step, and future genetic studies could reveal more refined insights into human migration and gene flow across the Sahara. As sequencing costs continue to fall, whole-genome sequencing could enable more unbiased estimates regarding OoA events and other key aspects of human evolution.

Methods

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41586-025-08793-7.

Supplementary information

Author contributions

J.K. and S.d.L. designed the research (Conceptualization), with N.S. leading the formal analysis assisted by M.S.v.d.L., A.P.S., A.H., H.R., B.P. and K.P.; S.d.L., D.C., M.A.T. and G.M. provided the archaeological materials and conducted excavations (resources and investigation). Laboratory work (investigation) was headed by S.V. and R.A.B. with help from R.R., M.L. and A.M. Data curation and visualization were primarily managed by N.S., with M.S.v.d.L. contributing. M.F.M.A.-F. and M.T. managed project administration and handled permit acquisitions (project administration), and A.B. contributed to embedding the findings into the broader archaeological framework (integration into archaeological framework). N.S. drafted the manuscript, with all of the authors reviewing and editing. K.P., H.R., J.K. and S.d.L. provided supervision, and J.K. secured funding. Contributions are recognized using the CRediT Taxonomy labels.

Peer review

Funding

Open access funding provided by Max Planck Society.

Data availability

All newly reported ancient nuclear DNA data are archived in the European Nucleotide Archive (PRJEB84057).

Competing interests

The authors declare no competing interests.

Extended data

is available for this paper at 10.1038/s41586-025-08793-7.

Supplementary information

The online version contains supplementary material available at 10.1038/s41586-025-08793-7.