Population structure and history in East Asia

Recent investigations by using molecular markers to study patterns of genetic diversity in East Asia have tried to address two main questions. First, are northern and southern East Asian populations genetically distinct? Second, are northern and southern East Asian populations descendants of the same ancestral population, or are they descended from different populations?

Chu et al. inferred a distinction between southern and northern Chinese populations and a southern origin for northerners by analyzing phylogenetic trees of populations constructed by using microsatellite data (10). Among trees constructed by using the neighbor-joining method with bootstrapping, Chu et al. identified a “clear distinction between southern and northern Chinese populations” on the basis of the presence of a paraphyletic northern group and an almost monophyletic southern clade. Su et al. inferred support for regional clusters on the basis of an analysis of principal components of Y-chromosomal diversity and the observation of substantial lineage sharing among regions, also suggesting the possibility of a southern origin (11). In a map of the first two principal components of variance, northern populations clustered together, and southern populations clustered together. In contrast, we find support for neither a strong regional distinction nor a southern origin of Northeast Asian populations.

Four sections in Fig. 2 show the results from our four marker systems. In these maps, the first principal component accounted for 11, 34, 33, and 32% of the total dispersal in observed in mtDNA, STR, Y chromosome, and JC virus, respectively. The second principal component accounted for 10, 19, 23, and 14% of the variance in observed in mtDNA, STR, Y chromosome, and JC virus, respectively. These maps suggest an alternative explanation for diversity patterns in East Asia. Three features of the principal components maps in Fig. 2 are most important.

First, the STRs, the Y polymorphisms, and the JC virus polymorphisms are geographically structured so that the principal components give a good portrayal of the underlying genetic distances. There is, on the other hand, almost no structure in the mtDNA differences among regions. The genetic distances among populations from mtDNA describe something like a high-dimension sphere. Even though it is possible to generate an mtDNA phylogeny from the data, any such phylogeny could not reveal much of interest about population history here.

Second, although northern and southern populations generally fall into different regions of the principal components maps, the clusters are not distinct. For example, in the map of diversity in mtDNA, some southern populations such as Dai are much more similar to other northern populations than they are to other southerners, such as the Vietnamese or Malay Aborigines. The putative northern and southern clusters appear to blend across a cline; there is no abrupt change.

Third, populations sampled from adjacent geographical areas tend to be near each other on the principal components maps. This finding is consistent with previously published evidence of genetic isolation by distance in China (20) and explains the lack of clustering in the principal components maps. Because populations are isolated by distance, and because they were sampled on a predominantly north/south axis, the principal components maps can be segregated into northern and southern groups delineated by latitude. This feature explains the gradual, rather than sharp, divide among the arbitrarily chosen north/south divide.

The geographical organization of the principal components maps raises questions about the informativeness of human population “phylogenies.” If populations that are isolated by distance are sampled along an axis, is artificial phylogenetic signal introduced? One possible explanation for the repeated identification of northern and southern clades in East Asia (9, 10) is that oversampling along a north/south axis has spuriously influenced phylogenetic inferences. Such a bias would be consistent with the low bootstrap values supporting northern and southern clades observed by Chu et al. (10).

It is of some interest that the clearest north/south distinction among the four principal components maps in Fig. 2 is observed in the JC virus. The JC virus is a nonpathogenic urinary tract virus that is thought to be largely vertically transmitted, and as such it might be a replicate of mtDNA. However, the true extent of horizontal transmission in JC virus is unknown. Diversity patterns in the JC virus have been interpreted as evidence for a north/south distinction previously (21), and that the virus displays a stronger regional distinctiveness than the human genes suggests appreciable horizontal transmission may be present. The level of isolation by distance is more different among viral subpopulations than among their human hosts.

The lack of regional clusters brings inferences about directional migration into question as well. Su et al. suggest that the presence of every northern lineage in at least one southern population implies northward movement, but it is unreasonable to conclude that the northern population derived from the southern populations based on this evidence. One attractive alternative explanation for the asymmetry in lineage sharing is that northern and southern East Asian groups have had a long history of separation, but many lineages have migrated from north to south recently. Such asymmetric migration could easily generate regional differences in genetic diversity. However, this explanation would predict a clear genetic difference between northern and southern groups. Another explanation for the asymmetry in lineage sharing is suggested by regional differences in demography. Whereas southern populations reside mostly in high-density areas, northern areas are sparsely populated (22). Between-region migration, accompanied by high rates of genetic drift and lineage loss in northern groups, could account for an asymmetry in lineage composition without causing appreciable between-region divergence.

The potentially important role of Central Asia in questions about the genetic composition of East Asia is emphasized by patterns of mtDNA diversity in the region. In a comparison of mtDNA sequences from Europe, Central Asia, and East Asia, Comas et al. (23) found a closer affiliation between Mongols and Talas Kirghiz populations than between the Talas Kirghiz and Sarytash Kirghiz. No southern East Asians were included in Comas et al.'s study, and their relationships to Central Asians are unknown. However, the similarity of some northern East Asian populations to Central Asians indicates that the large migrations associated with trade along the Silk Road and during later times may have had an influence on diversity in the Far East. The inclusion of Central Asian samples in future studies including both northern and southern East Asians will be important in answering more detailed questions about East Asian origins.

The existence of a genetic distinction between northern and southern East Asia is not well supported. Patterns of genetic diversity in the area are more consistent with the notion that local gene flow since the end of the Pleistocene era has erased old human population differences over much of the world at neutral marker loci (24), and that much of the differentiation in the region is attributable to simple isolation by distance (20). Such erasure is expected to happen even if migration rates are relatively low (25). The lack of pattern in East Asia suggests that many of the anthropological trends previously held to define pervasive regional distinction are strictly cultural phenomena with no implications for genetic differentiation. This finding is itself interesting—regional cultural trends in East Asia seem to have persisted for lengthy time periods despite evident genetic continuity.

Article published online before print: Proc. Natl. Acad. Sci. USA, 10.1073/pnas.240441297.

Article and publication date are at www.pnas.org/cgi/doi/10.1073/pnas.240441297

We are indebted to Dr. Antonio Torroni for his detailed explanation of both experimental procedures and statistical methods. We thank Dr. Wen Wang and Long Nie for their intellectual contribution and support in sample collection. We also thank Xue-Mei Lu and Jing Luo for their comments on an early version of this paper. This work was supported by the Chinese Academy of Sciences, Natural Science Foundation of Yunnan Province, and National Sciences Foundation of China.