A new DRB1 allele, a variant of DRw14 called DRB1^*1408, has been found in Polynesians (Gao et al. 1992a); this allele equates with ‘DRw6P’ described in earlier RFLP studies. Among Austronesian speakers tested to date this allele is essentially confined to Polynesians in whom it occurs with moderate frequency (5-7 per cent), but it is absent from Madang, New Britain and Fiji. Four instances of the novel allele have been seen in Melanesians from New Caledonia, but this could represent recent admixture with Polynesians from the Wallis Islands. Closer dissection of the novel Polynesian-specific variant at the DNA level shows that it represents a single nucleotide substitution in an allele still found in contemporary Polynesians, so is almost certainly an example of recent mutation. In Kimberley region Aborigines, three novel HLA-DRB1 mutations, not seen elsewhere, account for about 50 per cent of the HLA-DRB1 allele frequency (Gao et al. 1992b).

Figure 1. Phylogenetic analysis of the distributions of HLA-DR,DQ haplotypes in 20 populations. In Asia-Oceania there were 82 haplotypes; additional haplotypes found in Caucasoids were pooled to generate a 20×83 matrix for genetic distance calculations.

The mutation giving rise to DRB1^*1408 in Polynesians is of functional significance in that it results in an amino acid change at position 57 in the DR beta molecule. The PCR-based protocol is capable of detecting silent mutations in hypervariable regions of the DRB1 gene, and no novel silent mutations have been found in Polynesians. This suggests that the new functional mutation may have survived and flourished due to some selective advantage. Other evidence in favour of this hypothesis is that the same mutation has occurred in Australian Aborigines on a different DRB3 haplotype, suggesting independent mutations and convergent selection. The role of natural selection in shaping HLA-DR profiles is subject to debate (Hughes and Nei 1989), but it is possible that rare HLA types have been advantaged in epidemics.

Phylogenetic analyses of HLA-DR,-DQ haplotype distributions in the populations listed in Table 1 and in Caucasoids (Fernandez-Viña et al. 1991) are given in Figures 1 and 2, based on Nei’s distance statistic (Nei 1973). Figure 1 shows the extraordinary power of this small segment of the human genome, clustering populations in a manner partially expected from linguistic, anthropological and archaeological evidence (Bellwood 1989). The complete separation of Javanese and Polynesians, who show virtually no overlap in HLA-DR, DQ haplotypes (Gao and Serjeantson 1991a), is unexpected on linguistic grounds. The eigenvector (Figure 2) makes better use of the genetic data, representing genetic distance in two dimensions. Java is well-isolated from the other populations due to a near absence of DR4- and DRw6-related haplotypes; these haplotypes are well-represented in the other populations and account for the majority of HLA-DR types seen in Polynesia. The northern and southern Chinese populations cluster to the left of the eigenvector, while Xinjiang, a minority group with known Caucasoid ancestry, is positioned midway between Caucasoid and northern Chinese populations. The Polynesian groups have a position intermediate between northern China and coastal Melanesia. In the phylogenetic analysis, Melanesians from the north Papua New Guinea coast (Madang) cannot be discriminated from the Tolai of New Britain and these groups have a DR, DQ profile similar to that in Melanesians from New Caledonia. The Fijian sample reflects some admixture with Polynesian elements and is equidistant to New Caledonia and Western Samoa in the eigenvector diagram. Micronesians from Nauru and Kiribati are well-separated from Polynesians due to a high frequency of DRB1^*1202, a DR5 allele that is commonly found in Javanese, less frequently in southern Chinese, and rarely elsewhere. Non-Austronesian-speaking Melanesians from the New Guinea Highlands show closer affinity with Australian Aborigines than with any of the Austronesian-speaking Melanesian groups. The Australian Aboriginal populations cluster together, even though about 50 per cent of HLA-DR alleles are unique to Kimberleys Aborigines.

Figure 2. Eigenvector representation of genetic distances between 20 populations based on HLA-DR, DQ haplotype frequency distributions.

This genetic distance analysis differs from that based on non-HLA markers (Kirk 1989) in that, in Kirk’s study, Australian Aborigines showed closer affinities with Asian populations than with New Guinea Highlanders in the non-HLA analysis. However, we note that Kirk’s Aboriginal sample was from central Australia whereas we have studied coastal people from Cape York and the Kimberleys; serological HLA profiles of central and northern Aboriginal populations are markedly different (Hay et al. 1986). This analysis based on HLA-DR, DQ haplotypes is more consistent with multivariate distances based on cranial measurements (Pietrusewsky 1984), in that Australian populations are well-separated from Austronesian-speaking groups and from Southeast Asia, and in that Java clusters with Southeast Asia rather than with Polynesia. The affinity between Fiji and Polynesia seen in the HLA-DR, DQ analyses was not seen by Pietrusewsky (1984) but was evident in the anthropometric analysis of Howells (1970).