Complex spread of pastoralism to eastern Africa
Cluster 2 of the Kenyan samples on the PCA (
Fig. 2), with east African pastoralist-related ancestry, includes the newly reported groups/individuals from sites of the Savanna Pastoral Neolithic tradition in South Kenya: Kenya_LukenyaHill_3500BP, Kenya_HyraxHill_2300BP, and Kenya_MoloCave_1500BP, which fall into the beginning, middle, and end, respectively, of the Pastoral Neolithic period in Kenya, as well as a published ancient genome from Tanzania, Tanzania_Luxmanda_3100BP (
3), and other published Pastoral Neolithic genomes from eastern Africa (
4).
These samples show remarkable continuity of ancestry across a time span of 2000 years, presenting similar genetic profiles in PCA and clustering analysis (
Fig. 2 and fig. S1).
On the basis of previous models for Tanzania_Luxmanda_3100BP (
3), we first applied two-way ancestry models in qpAdm using Ethiopia_ 4500BP and a group of ancient Levantine individuals (
24), which we take as the closest available proxy for ancient northeastern African ancestry (
10,
11), as sources. Consistent with the findings of a previous aDNA study (
4),
we found this model to be insufficient (Fig. 3 and table S3) and demonstrate that an additional genetic component related to the present-day Dinka (a Nilotic-speaking group from South Sudan) is necessary to fit the data. In addition to qpAdm, we confirmed this affinity using a customized
f4 test (see Materials and Methods and fig. S3). In our final three-way model, which is qualitatively similar to the model proposed in (
4), we find 33 ± 11% and 24 ± 10% Dinka-related ancestry in Kenya_HyraxHill_2300BP and Kenya_LukenyaHill_3500BP, respectively, and lower proportions in Kenya_MoloCave_1500BP and Tanzania_Lxumanda_3100BP (
Fig. 3 and table S3).
While the estimated proportions of Levantine-related ancestry in all samples are rather constant (around 30 to 40%), we find that both the proportion of east African forager-related ancestry, as well as of Dinka-related ancestry, varies substantially across individuals. An earlier study (4) concluded that admixture between pioneering herders with Levantine-related ancestry and eastern African hunter-gatherers primarily occurred before their arrival in southern Kenya. However, our data suggest that periodic admixture between herders and hunter-gatherers, or populations predominantly carrying ancestry derived from them, may have continued into the PN. In particular, the newly reported 1500-BP individuals from Molo Cave carry 50% or more forager-related ancestry, and less Dinka-related ancestry, than observed in all other sequenced Pastoral Neolithic individuals (
4). A model of repeated interaction between foragers and herders is further supported by admixture date estimates using linkage disequilibrium decay, which suggest that admixture dates between ancestry related to Chalcolithic Levant (
24) and to Ethiopia_4500BP range from a few hundred to a few thousand years before the time of death of the individuals, with no clear correlation between admixture age and age of sample (fig. S4), inconsistent with a simple model of admixture, but suggesting either multiple events, or strong population structure preventing homogenization of ancestries over a long time period.
Despite only minimal archaeological evidence for the persistence of autochthonous hunter-gatherers in the Central Rift Valley this late in the PN (25), these genetic results suggest that communities with high or unadmixed hunter-gatherer–related ancestry continued to live alongside communities with high or unadmixed Pastoral-Neolithic related ancestry until nearly the Iron Age, leaving prominent genetic traces at Molo Cave. It is not yet clear from Molo Cave or other sites whether the timing and pace of admixture reflects adoption of herding by foragers, absorption of foragers into herding groups, or more complex intergroup social dynamics.
Combining evidence from both eastern African genetic clusters, we document very different patterns of interaction and admixture from sampled individuals along the eastern African and Lake Victoria shores relative to the patterns in the Central Rift. Near lake and ocean coasts, we see little evidence for pastoralist admixture into forager individuals [e.g., Kenya_Nyarindi_3500BP and two previously sampled individuals from Zanzibar (
3)].
Our analysis also demonstrates that the recently published individual from the cave site of Panga ya Saidi in coastal Kenya [Kenya_400BP (3)] similarly retains a predominantly eastern African forager ancestry, with only a small Levantine-related component. This is the exact opposite of the pattern observed in individuals around the Central Rift, where pastoralist-mediated, Levantine-related ancestry spread rapidly. It may be that delayed-return foragers in stable coastal and lacustrine environments were more demographically numerous and/or resistant to interactions with incoming food producers than other hunter-gatherers.
While our data support the three-component model for the Pastoral Neolithic (
4), our findings suggest greater complexity than initially proposed for the admixture of existing and incoming populations in this period. The fact that both Dinka-related ancestry and eastern African forager-related ancestry varies substantially in our samples and previously published samples suggests that the spread of herding either involved complex population structure maintained over a long time period or prevented homogenization of these ancestries, or multiple population movements with regionally distinct trajectories of interaction and admixture. This adds increasing resolution to proposed diversity of populations that contributed to the “moving frontier” model for herder dispersals in eastern Africa (
4,
26). Individuals from Molo Cave, Luxmanda, and Panga ya Saidi furthermore provide evidence that contact with eastern African foragers, who coexisted with food producing people until at least 400 BP (
Fig. 3A), was a continuous process, rather than one that occurred only during initial phases of contact.
The data also reveal that this interaction between herders and foragers was very imbalanced, with hunter-gatherer ancestry entering pastoralist populations, but little flow in the other direction. It is not clear what forms of social systems between herders and foragers may have resulted in this one-way admixture. In the past, it has been assumed that low herder population density and high risk of herd loss from epizootic disease would require herders to form closer relationships with local hunter-gatherers who had greater ecological knowledge of the landscape (
27,
28). This has been supported by evidence for herder-forager interactions at sites such as Crescent Island (
29) and Prolonged Drift (
30). Genetic evidence indicates that if these interactions occurred, then they were more structured and possibly more consistent with ethnographic client-patron relationships (
31), wherein individuals from hunter-gatherer communities may be slowly integrated into herder societies. It is possible that sex bias due to different social dynamics played a role in the observed asymmetric gene flow between the two groups. While we could not test this explicitly due to insufficient coverage on the X chromosomes, these dynamics have been previously described between foragers in central and southern Africa and Bantu-speaking farmers (
32).
advances.sciencemag.org
