Reply to Whitfield

To the Editor:

Dr. John B. Whitfield (Whitfield 2002 [in this issue]) writes to call attention to the variation in alcoholdependence risk as a function of both the ADH1B Arg47His polymorphism and specific populations. We do not disagree with his result: our study (Osier et al. 2002) involved, primarily, normal individuals from multiple populations and showed considerable variation in allele frequencies at that site among the populations. In that report, we commented that the different risks of alcoholism associated with alleles at this site could be explained by other relevant variation on the specific haplotype at high frequency in eastern Asia. However, we did not show that haplotyping of the ADH Class I polymorphisms resulted in an unusually high F_st but that some individual sites in the gene cluster individually had unusually high F_st values. One particular haplotype does have a very large range of variation, but we do not have an appropriate empiric distribution for F_st values in multiallelic haplotype systems to show that it is unusually large. We showed that the “protective” allele, ADH1B*47His, occurs primarily on a specific haplotype in the Mediterranean and European populations studied but occurs on a different haplotype in eastern Asia. We concluded that the ADH1B*47His allele is likely to be old, a conclusion Whitfield reiterates.

We do disagree with some of the conclusions Whitfield reaches. He concludes that the effects of the ADH1B*47His allele are not additive. However, because of the strong linkage disequilibrium (LD) across the Class I gene cluster, Whitfield’s analysis showing nonadditive allelic effects uses the ADH1B*47His allele as a surrogate for the entire haplotype. Those analyses do not provide sufficient evidence to limit the effect to just that allele; some other variant on that haplotype could be relevant. Taken together, the evidence that the risk difference associated with this polymorphism is not the same in Europe as it is in eastern Asia and our demonstration that the haplotype containing the ADH1B*47His allele is different in Europe from the one in eastern Asia require us to focus on the entire haplotype and not just on this one site. The nonadditive effects cannot be attributed to the ADH1B*47His allele exclusively, as Whitfield himself notes earlier in his introductory paragraph.

Whitfield repeats a common error when he says LD will decrease over time, without noting all of the assumptions involved in that deterministic result. In regions of high disequilibrium caused by very low levels of recombination, the effects of random genetic drift can easily outweigh the deterministic expectation, as we have demonstrated (Calafell et al. 2001). Modern humans have existed outside of Africa for a relatively short time and had small population sizes during much of that time. Thus, drift associated with the expansion out of Africa and diversification around the world can swamp any factors like recombination that tend to reduce LD. Since these ADH cluster genes are involved in alcohol metabolism, we of course have the additional complication of determining to what extent natural selection may have played a role in altering the frequency of particular haplotypes in different geographical regions.

Finally, Whitfield implies it is important to determine whether social factors or LD are responsible for the differences in effects of ADH1B*47His in eastern Asia and in Europe. Social factors can indeed be very important in modifying risk associated with different genotypes. For example, it will be difficult to determine the actual risk associated with ADH variation in northern Africa and the Middle East, since most populations in those regions are Muslim and consumption of alcohol is proscribed by their religion. However, we note that the relevant “genetic” component is not strictly differences in LD but differences in what haplotypes are present. A site in the ADH cluster but not in LD with the ADH1B Arg47His site could have an epistatic effect such that only those chromosomes with particular alleles in coupling account for the protective effect. Moreover, background genotype clearly differs at the population level between eastern Asian populations and European populations (e.g., Calafell et al. 1998) adding yet another level of confounding on the path to understanding the role of the ADH1B Arg47His polymorphism in risk of alcoholism.