Clinal variation in the nuclear DNA of Europeans

LOUNPS CHIKHI,1'2 GIOVANNI DESTRO-BISOL,3 VINCENZO PASCALI,4 VANESSA BARAVELLI, MARINA DOBOSZ,6 AND GUIDO BARBUJANI 1'2

Abstract Allele frequencies are clinally distributed for many protein polymorphisms in Europe, suggesting that the current populations are derived from an ancestral group that expanded from the Near East. It is not yet fully established whether that expansion took place during the Neolithic or earlier or whether the detectable protein variation faithfully reflects the underlying molecular variation. In this study we address the latter question by describing geographic patterns of genetic diversity at seven highly polymorphic DNA markers. Two of these markers are minisatellites, four are microsatellites, and the seventh is a locus of the HLA system. By analyzing a database of 304 samples, with more than 130,000 chromosomes, we found evidence for a major clinal component of genetic variation. At most loci spatially close populations resemble each other genetically, and the degree of genetic similarity, as measured by spatial autocorrelation statistics, decreases at increasing distances. The observed patterns of molecular variation do not seem to differ qualitatively from those identified for protein polymorphisms. This suggests that low levels of population structuring, described in some mitochondrial DNA studies, may reflect different evolutionary histories for nuclear and maternally inherited markers or, alternatively, that spatial patterns of mitochondrial DNA variation may need more sensitive statistical methods to be recognized.

Clines can result from expansion events, that is, from a combination of demographic growth and range expansion (Endler 1977; Menozzi et al. 1978). Many allele frequencies show clinal distributions in Europe (Menozzi et al. 1978; Sokal et al. 1989; Cavalli-Sforza et al. 1993), with gradients extending from the Near East into the Iberian peninsula and northwestern Europe. Allele frequencies also correlate with the likely dates of origin of agriculture, as inferred from the archeological record (Menozzi et al. 1978; Sokal et al. 1991). That correlation suggests a possible historical cause of the clines, namely, the expansion of early agriculturalists from the Near East into Europe (Menozzi et al. 1978). Ammerman and Cavalli-Sforza (1984) called this expansion demic diffusion, and their modeling showed that the observed patterns could be explained by dispersal of populations from the Near East. To account for such broad clines, it is necessary to assume that the dispersing population kept growing in number as it came to occupy new regions while at the same time undergoing limited admixture with the populations encountered during the expansion. Although models involving total replacement of previously settled populations have been shown to fit the data as well (Barbujani, Sokal et al. 1995; Fix 1997), there is broad agreement that the continent-wide clines of allele frequencies in Europe reflect the farming expansion and hence originated during the Neolithic. Comparison with linguistic evidence agrees with this view (Renfrew 1987, 1992; Barbujani and Pilastro 1993), although it does not prove it (Sokal et al. 1992).

So far, large-scale studies of European populations with DNA markers have not supplied unquestionable evidence. Until recently, mitochondrial DNA (mtDNA) was the only marker with sufficient molecular data for a statistically robust analysis. Studies of the distribution of mtDNA in Europe showed extensive diversity within populations but a limited degree of diversity among them and little or no geographic structure (Pult et al. 1994; Bertranpetit et al. 1995; Sykes et al. 1996; Richards et al. 1996)-a somewhat unexpected finding that conflicts with previous results of protein studies. The estimated ancient age of most mtDNA alleles was interpreted as evidence of an ancient Paleolithic origin of the populations (Richards et al. 1996). This raised the suspicion that variation at the protein level may not faithfully mirror DNA diversity, with the obvious implication that previous evolutionary inferences based on allele frequency data should be reconsidered. There are alternatives, however. One is that perhaps mtDNA varies in a peculiar way because of its maternal mode of inheritance or because of other evolutionary constraints that cannot be specified at this time. Another alternative is that, to identify the signature of recent evolutionary processes on DNA data, whether mitochondrial or otherwise, we need more specific statistical tools than those leading to construction of gene trees.

In line with this second alternative, Semino et al. (1996) showed that two Y-chromosome restriction fragment length polymorphism (RFLP) alleles form clines across much of Europe, suggesting that some level of population structuring exists at the DNA level too. Here, we describe the overall patterns of geographic variation for seven hypervariable nuclear DNA markers. Using a spatial autocorrelation analysis of the distribution of DNA alleles, we provide evidence of broad clinal patterns, spanning most of the continent and often extending as far as the Near East. These results indicate that some processes that shaped allele frequency variation in Europe have also left a detectable mark at the molecular level for several nuclear DNA loci. Although our analysis refers to a short span of evolutionary time, it may help to reconcile molecular and nonmolecular evidence, possibly better than other, more complicated models.