Genome-wide data from the Bubi of Bioko Island clarifies the Atlantic fringe of the Bantu dispersal

Abstract

Background: Bioko is one of the few islands that exist around Africa, the most genetically diverse continent on the planet. The native Bantu-speaking inhabitants of Bioko, the Bubi, are believed to have colonized the island about 2000 years ago. Here, we sequenced the genome of thirteen Bubi individuals at high coverage and analysed their sequences in comparison to mainland populations from the Gulf of Guinea.

Results: We found that, genetically, the closest mainland population to the Bubi are Bantu-speaking groups from Angola instead the geographically closer groups from Cameroon. The Bubi possess a lower proportion of rainforest hunter-gatherer (RHG) ancestry than most other Bantu-speaking groups. However, their RHG component most likely came from the same source and could have reached them by gene flow from the mainland after island settlement. By studying identity by descent (IBD) genomic blocks and runs of homozygosity (ROHs), we found evidence for a significant level of genetic isolation among the Bubi, isolation that can be attributed to the island effect. Additionally, as this population is known to have one of the highest malaria incidence rates in the world we analysed their genome for malaria-resistant alleles. However, we were unable to detect any specific selective sweeps related to this disease.

Conclusions: By describing their dispersal to the Atlantic islands, the genomic characterization of the Bubi contributes to the understanding of the margins of the massive Bantu migration that shaped all Sub-Saharan African populations.

Keywords: Bantu-speaking groups; Bioko Island; Bubi; Isolation; Population genetics.

Fig. 1

a Map of the Gulf of Guinea showing the location of Bioko Island and the neighbouring countries. b Map of the Bioko Island. c Genotype-based principal components analysis (PCA) plot obtained with EIGENSOFT smartpca. A dataset of 168 individuals from 14 populations and 581,224 SNPs are included. The percentage of variance explained is written along the axes. The maps have been created with R package

Fig. 2

Graphical representation of shared genetic components performed using the ADMIXTURE (K = 4) software: 169 individuals from 13 populations and 456,095 SNPs have been plotted. Although the cross-validation errors show the lowest value at K = 2, we have chosen K = 4 because it is the first plot where the crucial RHG, Bantu and Western African components are clearly identified

Fig. 3

Statistic f₃values obtained with popstats when taking San as outgroup. The statistic and one standard error deviation are presented for each combination test. Blue dots represent Bantu-speaking populations, Green dots represent Western-African populations, orange dots Represent Eastern Hunter-Gatherer populations and purple dots indicate Western Hunter-gatherer populations

Fig. 4

Matrix of pairwise F_st calculations. The F_st values have been calculated using plink and with a dataset of 592,395 SNPs

Fig. 5

Matrix of shared counts of haplotypes obtained using fineSTRUCTURE. The dataset includes 491,203 SNPs and 169 individuals. The structure of the matrix has been adapted using the fineSTRUCTURE dendrogram. Green label represents the Bubi individuals, red label is used to represent the Gabon-Cameroon Bantu-speaking individuals, Blue label represents the Angolan Bantu-speaking individuals, Yellow label represents the Western-African individuals, Light blue label represents the Eastern Hunter-Gatherer individuals and Brown label represents the and Western Hunter-gatherer populations

Fig. 6

a Average of the genome in homozygosity (in kb) for the Bubi and mainland Bantu populations. b Average of the shared IBD genomic blocks between the Bubi and mainland Bantu populations (IBDs > 2 cM). The maps have been created with R package