North Africans traveling north

Humans have always loved to move and to mate. In doing so, we generate complex patterns of diversity in language, custom, and culture across and within geographic regions. However, the richest and most durable signals of migration and admixture are evident in patterns of DNA sequence variation (1, 2). The study by Botigué et al. (3) reported in PNAS analyzes DNA sequence variation in Africa, the Near East, and Europe to advance our understanding of the peopling of Europe with a focus on the relatively recent contribution of North African ancestry to the contemporary population genetic structure of southern Europe.

The population history of Europe during the last ∼50,000 y corresponds to two major geologic time periods: the Pleistocene before ∼11,500 y ago, transitioning into the Holocene (Fig. 1). The major events of the Pleistocene include the initial pioneering colonization by humans during the Upper Paleolithic, as part of the overall dispersal of modern humans out of Africa (Fig. 1A). Subsequently, during the most recent major Ice Age (Last Glacial Maximum) extending from 25,000 to 19,500 y ago, people found refuge in a number of well-defined geographic areas, including southwest Europe (Franco-Cantabrian/ Iberian), along the Mediterranean coast including the Italian peninsula, the Balkans, the Levant, and on the east European steppes (Fig. 1B) (4, 5). The subsequent transition from Pleistocene to Holocene (disrupted briefly by a less severe cooling lasting about a millennium; Fig. 1C) also ushered in the dispersal to Europe of Near East migrants, bringing with them Neolithic skill sets and practices, including different forms of agriculture and animal domestication (Fig. 1D) (6). Both the recolonization from Ice Age population refugia and the Neolithic migration generated a backdrop for a reducing south to north European cline of DNA sequence diversity. Ever since, numerous migrations have followed along multiple intercontinental routes from the Copper Age to this day, including trans-Mediterranean movements (Fig. 1D), which are directly relevant to understanding population genetic structure in southern Europe (7).

Fig. 1.

Major stages in the peopling of Europe. A–E are illustrative of five major demographic stages that may have contributed to the current complex human population genetic structure of Europe. Time scales, expansion arrows, and specificity of geographic regions should not be regarded as pointing to the detailed migratory routes and dates.

Much of the genetic trace of the more remote and recent migratory events has been fruitfully elucidated by examining the distribution of ever more finely resolved phylogenies or haplogroups at the nonrecombining uniparental regions of the genome: namely, the Y chromosome and mitochondrial DNA, as has been recently summarized for Europe (5). However, population genetics studies corresponding to the Holocene, which initiated 10 millennia of relative climatic stability to this day, has particularly benefited from combining such uniparental lineage analysis with genomewide approaches (2). The rapid transition during the last decade to analysis of genomewide variation has enabled levels of resolution of population genetic structure, which eluded single locus analysis (8–11). This is because each independent locus has its own coalescent history and continues to add information until the point where the spacing of loci falls below the scale of linkage disequilibrium (the nonrandom association of neighboring markers) in the populations of interest. For more than a decade, it has been possible, and now straightforward and affordable using DNA BeadChip technology, to determine for any individual genome, the biparental allelic states (genotype) for a subset of almost all known single nucleotide variants (SNVs), or better still, the phased combination of genotypes for a group of SNVs at any and all given genomic regions of interest (haplotype). The technical ability to interrogate genomewide variation was accompanied by the development of analytic approaches based on either allele frequency or haplotype comparisons. The various categories of such analytic approaches, their strengths, and limitations have been summarized recently (2).

Using these approaches, multiple studies examining autosomal biparental variation among Europeans have yielded a set of consistent conclusions, which generate a multilayered pattern, to which Botigué et al. (3) validate an important North African layer (Fig. 1E). Within a global context, it is apparent that European populations share a distinctive background genetic layer that is readily separable from components dominating other continents, enabling quite reliable individual assignment of European ancestry, most evident in the Basque and Sardinian populations (1, 8–12). A next layer is evident as the consistent and reproducible distinction between “northern” and “southern” European population groups, with a clinal distribution of genetic variation consistent with a south to north expansion and/or a larger effective population size in southern than in northern Europe (10, 12, 13). Despite comparatively low average levels of genetic differentiation among Europeans, the decay of genetic similarity as a function of geographic distance within Europe gives rise to a rather amazing ability to overlay a geographical map of Europe with allele frequency-based measures of genomewide variation (9, 12). In certain cases, it is even possible to assign the ethnic affiliation of an individual with a high degree of fidelity for population isolates such as Basque, Sardinian, Finns, and Ashkenazi Jews (12, 14, 15), a genetic structure within a given geographic region (16), or even a village of origin (17).

In the European context, the question of the trans-Mediterranean gene flow has also been investigated. The maternal component of African lineages to the contemporary gene pool of Europeans could be readily estimated due to the clear partitioning of Sub-Saharan and European mtDNA lineages. Frequencies of such African lineages vary widely within Europe, ranging from 3% in southern Europe to 0.7% in central Europe and 0.5% in northern Europe (18, 19), with evidence that as much as 35% of African lineages form European-specific subclades, pointing to gene flow from Sub-Saharan Africa to Europe as early as 11,000 y ago (20). Similarly, the sharp discontinuities between northwestern Africa and the Iberian Peninsula at the level of the Y chromosome were followed by evidence of North African to south European migration, particularly to the Iberian Peninsula (21, 22). At the genomewide level, Auton et al. (23) reported the highest genomewide haplotype diversity of European samples in the Iberian Peninsula, together with the highest sharing of haplotypes with the Yoruba population of any European population examined. A more recent study reported that southern Europeans have inherited 1–3% African ancestry, with an average admixture timing of some 55 generations ago, coinciding with the end of the Roman Empire and subsequent Arab migrations (24).

This latter set of findings begs the question of whether this level of Sub-Saharan admixture is simply the carryover of Sub-Saharan African admixture embedded within a much larger recent North African contribution. It is this question that Botigué and colleagues successfully resolve (3). With the appropriate choice of sample sets, multiple independent analytic approaches converged on the conclusion that a relatively recent North African–specific, rather than Sub-Saharan, admixture has made a significant contribution to the population genomic structure of Europe, with a striking clinal pattern from prominence of such admixture in southwest Europe to vanishing in north and east Europe. With the reassuring exception of the Basque population isolate, the Iberian Peninsula showed the greatest imprint. Specifically, southwestern European populations averaged between 4% and 20% of their genomes assigned to a North African ancestral cluster, whereas this value did not exceed 2% in southeastern European populations. The study included an isolation-by-distance analysis, including a western Mediterranean waypoint. Isolation-by-distance resulting from a geographic barrier such as the Mediterranean can confound allele frequency based measures of admixture (25). The addition of an Eastern Nile Corridor waypoint might have provided an even more comprehensive picture, as would have a larger number of representative sources for Near East populations. The study also highlights the importance of combining allele frequency with haplotype-based approaches. Thus, the genomewide examination of long identical-by-descent haplotypes showed that the component of haplotype sharing between Europe and Sub-Saharan Africa appeared to be driven by the carryover of Sub-Saharan to northern Africa.

Botigué et al. (3) motivate further studies to unravel the complexity of trans-Mediterranean gene flow. First, it is important to clarify whether the strong North African signal results from multiple waves or continuous gene flow to Europe or rather can be mostly attributed to the strong hold of North African Moors in Iberia lasting for almost a millennium. Moreover, multiple complex scenarios distinguish direct gene flow from North Africa to Europe from indirect gene flow via a third region such as the Levant and/or by the combined movements of groups such as Jews and Phoenicians. Levant and Near Eastern gene flow to southern Europe could also have been either direct or via a North Africa route, much as that for the Arabian introgression first to North Africa and then to Iberia. Sorting this out is required as a next necessary step with careful comparison of adequately representative Near East with North African population sample sets.

Finally, Botigué et al. (3) make a valiant attempt to relate population genetic findings to patterns of common disease. While acknowledging the serious limitations of extrapolating a common disease risk genetic signature based on SNVs that are presumed to tag true risk causative mutations in one population (in this case a set of 53 tagging markers for multiple sclerosis in Europeans) to another population (North Africans) whose patterns of allele frequencies and linkage disequilibrium are markedly different. Nevertheless, the study goes on to report a higher than expected frequency of this very signature in North Africans. Given the disparity of actual multiple sclerosis disease favoring the populations of the temperate climates of northern Europe, the authors reach two major conclusions—both of which are valuable in their own right—but not likely a sequitur for the current study results. The first is that appropriate genomewide mapping of SNP signatures predicting predisposition to common disease requires delineation of the population genetic architecture of the population of interest. The second is that a true disease risk genomic signature does not equal a higher frequency of disease without the additional influence of gene-by-environment interactions.