Demographic and Selection Histories of Populations Across the Sahel/Savannah Belt

Abstract

The Sahel/Savannah belt harbors diverse populations with different demographic histories and different subsistence patterns. However, populations from this large African region are notably under-represented in genomic research. To investigate the population structure and adaptation history of populations from the Sahel/Savannah space, we generated dense genome-wide genotype data of 327 individuals-comprising 14 ethnolinguistic groups, including 10 previously unsampled populations. Our results highlight fine-scale population structure and complex patterns of admixture, particularly in Fulani groups and Arabic-speaking populations. Among all studied Sahelian populations, only the Rashaayda Arabic-speaking population from eastern Sudan shows a lack of gene flow from African groups, which is consistent with the short history of this population in the African continent. They are recent migrants from Saudi Arabia with evidence of strong genetic isolation during the last few generations and a strong demographic bottleneck. This population also presents a strong selection signal in a genomic region around the CNR1 gene associated with substance dependence and chronic stress. In Western Sahelian populations, signatures of selection were detected in several other genetic regions, including pathways associated with lactase persistence, immune response, and malaria resistance. Taken together, these findings refine our current knowledge of genetic diversity, population structure, migration, admixture and adaptation of human populations in the Sahel/Savannah belt and contribute to our understanding of human history and health.

Keywords: CNR1 gene; Africa; admixture; camel herders; pastoralists; population structure; selection.

Fig. 1.

Geographical locations and PCA for each assembled dataset. (A) Locations and PCA results for only Sahelian populations included in the present study. (B) Locations and PCA results of Sahelian populations across the Sahel/Savannah belt included in the present study and in previous studies (Triska et al. 2015; Haber et al. 2016; Vicente et al. 2019) (further details were included in supplementary fig. S2, Supplementary Material online). (C) PCA plots of worldwide populations included in the High-SNP density dataset (further details were included in supplementary fig. S4, Supplementary Material online). Populations were grouped according to their major linguistic affiliations (left plot) and geographical locations (right plot). Each PCA plot shows the first and second PC projections, and in parentheses the proportion of variance explained for each PC projection. Legend and further details about all the populations included in each dataset are detailed in supplementary fig. S3 and table S2, Supplementary Material online.

Fig. 2.

Patterns of population structure and migration rates across the Sahel/Savannah belt. Figure showing genetic cluster membership estimated for all the populations included in the High-SNP density database using unsupervised clustering ADMIXTURE analysis: (A) at K = 4 and (B) at K = 15 (further details were included in supplementary figs. S9–S13 and tables S4–S5, Supplementary Material online). The size of the pie charts is in relation to the sample size of each studied population. Results for the remaining European populations were included in supplementary fig. S12, Supplementary Material online. (C) To better visualize the ADMIXTURE results at K = 4, the values of the four components were plotted on a geographical map using the Kriging method, highlighting the West African ancestry in red, the Central African ancestry in purple, the Middle Eastern ancestry in light blue, and the European ancestry in dark blue. For each ancestry, cluster assignments below 30% were not included in the figure. (D) Effective migration rates estimated using FEEMS. Figure showing fitted parameters in log-scale with lower effective migration shown in orange and higher effective migration shown in blue. A topographic map of the Sahel/Savannah belt was also included (supplementary fig. S10E, Supplementary Material online).

Fig. 3.

Patterns of runs of homozygosity (ROH) and effective population sizes in studied Sahelian populations. (A) Linear regression between the total length of ROH and total sum of ROH longer than 1.5 Mb in each Sahelian population. (B) Linear regression between mean number of ROH and total sum of ROH longer than 1.5 Mb in each Sahelian population. In each figure, linear regressions were estimated using the generalized linear model (GLM) regression function in R. Values of each individual from each population are also shown in the background of each figure (and further details in supplementary fig. S17, Supplementary Material online). (C) Mean ROH length categories of all the populations included in the High-SNP density dataset. Figure showing the average for each category for: the Rashaayda Arab population (in black), European populations, Middle Eastern populations, North African populations, and African populations. Further details for each population were included in supplementary figs. S18–S22 and table S6, Supplementary Material online. (D) Effective population sizes (N_e) in Sahelian populations for the last 50 generations were estimated using IBDNe (supplementary fig. S25, Supplementary Material online). To better visualize the results, interactive plots were created for figures C and D (see Interactive_plot-fig_3C.html and Interactive_plot-fig_3D.html).

Fig. 4.

Genome-wide selection signals in Sahelian populations. XP-EHH-based scan for selection after comparing Rashaayda Arab population and African populations from: (A) the Western Sahelian region (1—Gambian in Gambia [GWD], 2—Bedik in Senegal, 3—Halpularen in Senegal, 4—Fulani in Guinea, and 5—Yoruba in Nigeria [YRI]) (supplementary fig. S35, Supplementary Material online); (B) the Central Sahelian region in Chad (1—Baggara Arab, 2—Dangaleat, 3—Daza, 4—Maba, and 5—Zaghawa) (supplementary fig. S36, Supplementary Material online); and (C) the Eastern Sahelian region in Sudan (1—Baggara Arab, 2—Kababish Arab, 3—Daju, 4—Nuba Koalib, and 5) Zaghawa (supplementary fig. S37, Supplementary Material online). Each figure highlights the location of the candidate region for selection in the CNR1 gene. Manhattan plots for this chromosome 6 and also a genomic region of 20 Mb around CNR1 gene were also included (supplementary figs. S38–S40, Supplementary Material online). (D) Lactase persistence associated haplotypes in studied Sahelian populations. Figure shows the EHH around the ancestral (in blue) and the derivate (in red) variants, rs182549 (−22018*A), which is in strong LD with rs4988235 (−13910*T), the European lactase prsistence associated variant.