Exploring Cuba's population structure and demographic history using genome-wide data

Abstract

Cuba is the most populated country in the Caribbean and has a rich and heterogeneous genetic heritage. Here, we take advantage of dense genomic data from 860 Cuban individuals to reconstruct the genetic structure and ancestral origins of this population. We found distinct admixture patterns between and within the Cuban provinces. Eastern provinces have higher African and Native American ancestry contributions (average 26% and 10%, respectively) than the rest of the Cuban provinces (average 17% and 5%, respectively). Furthermore, in the Eastern Cuban region, we identified more intense sex-specific admixture patterns, strongly biased towards European male and African/Native American female ancestries. Our subcontinental ancestry analyses in Cuba highlight the Iberian population as the best proxy European source population, South American and Mesoamerican populations as the closest Native American ancestral component, and populations from West Central and Central Africa as the best proxy sources of the African ancestral component. Finally, we found complex admixture processes involving two migration pulses from both Native American and African sources. Most of the inferred Native American admixture events happened early during the Cuban colonial period, whereas the African admixture took place during the slave trade and more recently as a probable result of large-scale migrations from Haiti.

Figure 1

Genetic diversity in Hispanic/Latino populations. MDS plots based on pairwise ASD matrix for the Cuban provinces and worldwide populations included in the called “Cuba-World” dataset (see further details in Supplementary Table S1), using (a) two-dimensional and (b) three-dimensional metric MDS projections for the first, second, and third axes. Most Cuban individuals are much closer to the European populations than the African populations. This is consistent with the Cuban population having experienced a three-way continental admixture between European, African, and, to a lesser extent, Native American source populations.

Figure 2

Genetic diversity across Cuba. (a) Bar-plot of four-way continental ancestry proportions in Cuban individuals and worldwide individuals estimated using unsupervised ADMIXTURE analysis at K = 4. Geographic distribution of (b) European (or green cluster), (c) African (or blue cluster), (d) Native American (or red cluster), and (e) East Asian (or purple cluster) ancestries across Cuban provinces estimated using unsupervised ADMIXTURE analysis at K = 4 (map on the top), and across Cuban municipalities estimated using RFMix analysis (EM = 2) (map on the bottom). Figure showing in grey the Cuban municipalities that were not included in the study (Supplementary Fig. S16a).

Figure 3

Sex-specific admixture patterns in Cuba. (a) Bar-plots of individual ancestry proportions across Cuban provinces estimated using RFMix (EM = 2) on the basis of autosomal (top; Auto.) and X-chromosome (bottom; X-chr.) data. Each vertical line represents one individual based on individual fractions for European (EUR), African (AFR), Native American (NAM), and East Asian (EAS) ancestry. (b) ΔAdmix ratios in all the fifteen Cuban provinces show different sex-specific admixture patterns across the island. Bar-plots for each ancestry were plotted with respective confidence intervals (95% CI).

Figure 4

Fine-scale genetic structure across non-masked reference populations and masked Cuban haploid genomes. Figure showing the ancestry-specific MDS obtained for each ancestry using the approach described in Browning et al. (MDS-based ASPCA). The placement of masked Cuban individuals against a backdrop of non-masked reference populations for (a) European, (b) African, and (c) Native American ancestries.

Figure 5

Time-frame of admixture events estimated in Cuba using TRACTS analysis. Figure showing the best-fitting model based on length distributions of ancestry tracts within each Cuban region, allowing for one European, two African, and two Native American migration events. The data points represent the observed distribution of ancestry tracts estimated using RFMix (EM = 2), while solid-coloured lines represent the distribution from the predicted model, and shaded areas indicating confidence intervals (68.3% CI) of the predicted model. The admixture timing of the best-fitting model is displayed to the right of each graph. Sizes of the dots represent inferred average proportions for each ancestry (see Supplementary Table S10). Best-fitting models were selected based on the strength of the log-likelihood of each Cuban region (−566.7 in Western, −553.2 in Central, and −505.8 in Eastern region).