Genetic evidence on the origins of Indian caste populations

Abstract

The origins and affinities of the approximately 1 billion people living on the subcontinent of India have long been contested. This is owing, in part, to the many different waves of immigrants that have influenced the genetic structure of India. In the most recent of these waves, Indo-European-speaking people from West Eurasia entered India from the Northwest and diffused throughout the subcontinent. They purportedly admixed with or displaced indigenous Dravidic-speaking populations. Subsequently they may have established the Hindu caste system and placed themselves primarily in castes of higher rank. To explore the impact of West Eurasians on contemporary Indian caste populations, we compared mtDNA (400 bp of hypervariable region 1 and 14 restriction site polymorphisms) and Y-chromosome (20 biallelic polymorphisms and 5 short tandem repeats) variation in approximately 265 males from eight castes of different rank to approximately 750 Africans, Asians, Europeans, and other Indians. For maternally inherited mtDNA, each caste is most similar to Asians. However, 20%-30% of Indian mtDNA haplotypes belong to West Eurasian haplogroups, and the frequency of these haplotypes is proportional to caste rank, the highest frequency of West Eurasian haplotypes being found in the upper castes. In contrast, for paternally inherited Y-chromosome variation each caste is more similar to Europeans than to Asians. Moreover, the affinity to Europeans is proportionate to caste rank, the upper castes being most similar to Europeans, particularly East Europeans. These findings are consistent with greater West Eurasian male admixture with castes of higher rank. Nevertheless, the mitochondrial genome and the Y chromosome each represents only a single haploid locus and is more susceptible to large stochastic variation, bottlenecks, and selective sweeps. Thus, to increase the power of our analysis, we assayed 40 independent, biparentally inherited autosomal loci (1 LINE-1 and 39 Alu elements) in all of the caste and continental populations (approximately 600 individuals). Analysis of these data demonstrated that the upper castes have a higher affinity to Europeans than to Asians, and the upper castes are significantly more similar to Europeans than are the lower castes. Collectively, all five datasets show a trend toward upper castes being more similar to Europeans, whereas lower castes are more similar to Asians. We conclude that Indian castes are most likely to be of proto-Asian origin with West Eurasian admixture resulting in rank-related and sex-specific differences in the genetic affinities of castes to Asians and Europeans.

Figure 1

Phylogeny of haplogroup M in India. Phylogenetic relationships between HVR1 haplotypes were estimated by constructing reduced median networks. The size of each node is porportional to the haplotype frequency. Reticulations indicate parallel mutational pathways or multiple mutations. The identities of HVR1 mutations (numbered according to the Cambridge reference sequence +16000; Anderson et al. 1981) that define major haplogroup subsets are depicted along selected internodes. The coalescence estimate of Indian haplogroup-M haplotypes is 48,000 ± 1500 yr, suggesting that Indian-specific mtDNA haplotypes split from a proto-Asian ancestor in the late Pleistocene.

Figure 2

Major subsets of haplogroup M. Phylogenetic relationships of HVR1 haplotypes assigned to haplogroup M were estimated for: (a) 343 Indians (Quintana-Murci et al. 1999a; this study); (b) 16 Turks and 78 Central Asians (Comas et al. 1998; this study); (c) 60 Mongolians (Kolman et al. 1996); (d) 25 Ethiopians (Quintana-Murci et al. 1999a); (e) 56 Chinese (Horai et al. 1996; this study); (f) 103 Japanese (Horai et al. 1996; Seo et al. 1998). The founding node of each network (M*) differs from the CRS (Anderson et al. 1981) by transitions at np 10398, 10400, and 16223. The frequency of each subset of haplogroup M is indicated. Each phylogenetic network was pruned by eliminating branches containing haplotypes summing to a frequency of <5% (these branches were binned with the founder haplotype, M*). The identities of HVR1 mutations (numbered according to the CRS −16,000; Anderson et al. 1981) that define major haplotype subsets are depicted along selected internodes.

Figure 3

Neighbor-joining network of genetic distances among caste communities estimated from 40 Alu polymorphisms. Distances between upper castes (U; Brahmin, Vysya, Kshatriya), middle castes (M; Yadava, Kapu), and lower castes (L; Mala, Madiga, Relli) are significantly correlated with social rank.