Commentary Genetic isolates: Separate but equal?

The tiny volcanic island of Tristan da Cunha is the most remote inhabited place on Earth. It lies in the South Atlantic, 1,700 miles and a week’s journey by boat from Cape Town, South Africa. The island’s population of 300 is descended from a handful of founders, mostly shipwreck survivors, who settled there in the 19th century. The small number of founders is reflected in the fact that only seven surnames and five mitochondrial lineages (1) are found among the inhabitants. At first glance, this population couldn’t be more different from that of Europe, inhabited for tens of thousands of years and home to more than half a billion people and dozens of ethnic groups. So it may come as a surprise that a study in this issue of the Proceedings (2) reports that these two populations, along with a wide range of others, are remarkably similar in one important respect—the level of linkage disequilibrium between nearby genetic loci. Taken at face value, this study has significant implications for human geneticists who seek to use increased linkage disequilibrium in isolated populations as a tool for unraveling the genetics of common diseases. Linkage disequilibrium refers to nonrandom association of alleles at two loci. To illustrate, consider locus A, with two equally frequent alleles A1 and A2, and locus B, with two equally frequent alleles B1 and B2. If the distributions of alleles at the two loci were independent of each other, the four haplotypes A1B1, A1B2, A2B1, and A2B2 would occur with equal frequency. A deviation of haplotype frequencies from these proportions defines linkage disequilibrium. For example, allele A1 may occur predominantly on chromosomes that carry allele B1. This might be the case if the mutation that gave rise to allele A1 at locus A originally occurred on a chromosome that carried allele B1 at locus B, and if the two loci are sufficiently closely linked that recombination hasn’t had time to randomize the haplotypes. The case of interest in human genetics is when locus A is a previously uncharacterized disease susceptibility locus, and locus B is a known polymorphic marker. If allele A1 confers increased risk of disease, then the frequency of A1 will be higher among patients—as will the frequency of B1 if there is linkage disequilibrium between the loci. The observed increase in frequency of B1 among patients then can be used to infer the presence of a nearby susceptibility locus. How closely linked must a marker be to show linkage disequilibrium with a susceptibility locus? The answer depends on the number of generations since the susceptibility allele first arose, because recombination tends to reduce linkage disequilibrium with each generation. In the case of common diseases observed worldwide, it has been hypothesized that many of the underlying susceptibility alleles are common gene variants that experience little selective pressure (3–5). Such alleles likely date back to before the ‘‘out of Africa’’ expansion of modern humans '100,000 years ago. Over the thousands of generations since that time recombination will have erased linkage disequilibrium except for markers very near the susceptibility locus. Sufficient marker density for detecting such susceptibility alleles by linkage disequilibrium thus may be difficult to achieve in practice, at least in the immediate future (unpublished work). One approach for getting around this problem is to take advantage of special populations that might show increased linkage disequilibrium. Two types of populations have been proposed for this purpose: those that are genetically isolated by geography or culture (6, 7), and those with a history of recent admixture between different ethnic groups (8, 9). In such populations, small numbers of founders, population bottlenecks caused by various catastrophic events, genetic drift, as well as interbreeding between groups with different allele distributions may alter haplotype frequencies and thereby create linkage disequilibrium anew, in effect resetting the clock. Does this, in fact, occur? It is certainly true that increased linkage disequilibrium is seen in genetic isolates around rare disease mutations (10), but the situation is far less clear for common variants. For example, even a narrow bottleneck may admit a large enough sample of chromosomes carrying a common variant that haplotype proportions are not significantly altered (unpublished work). Whether linkage disequilibrium around common variants is increased in special populations is currently an open question, as the demographic history of most populations is not known in sufficient detail for precise theoretical predictions, and few experimental studies have compared linkage disequilibrium across populations. Lonjou et al. (2) set out to address the question empirically. They cull data from the literature (11–13) on haplotype frequencies at two genomic regions in a wide range of populations and compare levels of linkage disequilibrium. The regions are the MNS blood group system on chromosome 4 and the Rh blood group system on chromosome 1. [Lonjou et al. also include data on the CD4 gene on chromosome 12 (see ref. 14), but because no information about haplotype frequencies is available for special populations, these data cannot be used to examine the question of interest and will not be considered further.] Linkage disequilibrium is computed for four pairs of loci: MN-Ss from the MNS system and C-D, C-E, and D-E from the Rh system. The populations examined include a number of subpopulations from each of eight major geographic regions: Europe, Near East, India and Pakistan, Far East, sub-Saharan Africa, the Americas, Oceania, and North Africa. Some of the subpopulations have histories of recent admixture between Europeans and another ethnic group. Also examined are six populations Lonjou et al. consider to be genetic isolates: Jews, Basques, Lapps (Saami), Eskimos, the Ainu of Japan, and the islanders of Tristan da Cunha. Lonjou et al. (2) examine the pattern of average linkage disequilibrium at the blood group loci and make the following observations. First, sub-Saharan Africa has consistently lower levels of linkage disequilibrium than all other populations. This observation is consistent with previous studies (14) and most likely reflects relatively recent common descent of nonAfrican modern humans from an ancestral African population.