We analyzed a total of 2,275 individuals from 38 East Asian populations, especially those not included in the previous reports (northern Han Chinese, Tibetan and southern ethnic populations in China). Table 1 lists the frequencies of ADH1B*47His in the 38 populations. In general, the distribution pattern is consistent with the previous reports1417, and most of the populations (31/38) have frequencies higher than 50%. In Han Chinese, the highest frequency is detected in Zhejiang province of south-eastern China (98.5%), and those in the west have relatively low frequencies (60-70%). The same pattern is also observed for the other ethnic populations from China and Southeast Asia (Cambodia and Thailand) except for Tibetan (14.1% on average), Bulang (1.7%, an ethnic population from south-western China) and Cambodian (20.6%). All the five Tibetan populations from different geographic regions have low frequencies (13-21%). We created a contour map based on the data from the 38 populations and those published before (Figure. 1). The distribution of the frequencies of ADH1B*47His confirms its prevalence in East Asia and a clear east-to-west cline is observed.

To detect the molecular signature of recent selection on the ADH1B*47 polymorphism, we applied the LRH method and the iHS statistics using the genotype data from the HapMap project. The obtained iHS value for the core SNP (rs1229984) is -2.189 (the empirical p-value is 0.0269), an indication of selection. We then define the core region of ADH1B on the basis of five SNPs (rs4147536, rs1229984, rs1353621, rs1159918 and rs6810842) which determine the East Asian-dominant haplotype. We also select the flanking SNPs, extending both upstream and downstream to 250 Kb, to study the decay of LD from the core haplotype. We plot the haplotype-bifurcation diagrams18 for the two East Asian populations (Figure. 2) from HapMap (JPT: Japanese in Tokyo, Japan; CHB: Han Chinese in Beijing, China). At a minimum threshold of 9%, we define two core-region haplotypes in the JPT+CHB population. The haplotype CTTCG, which covers the derived variant of ADH1B*47His has an extended predominance by showing a thick branch in the haplotype-bifurcation diagram, clearly suggesting a long-range LD.

The EHH and REHH of the major core haplotypes (≥9%) are plotted against the distance away from the core for the JPT+CHB population (Figure. 2). The EHH of the CTTCG core haplotype decays more slowly than that of the other core haplotype (containing the ancestral variant ADH1B*47Arg) does. In addition, the upstream REHH value of the CTTCG is 17.329 (P = 0.01, by using 1-NORMSDIST). Again, this result is highly consistent with the previous studies, in which the molecular signature of selection was suggested in a wider genomic region containing the ADH1B locus among East Asian populations1517. The selection on ADH1B was also reported previously when the global populations were screened16. Additionally, a strong signature of positive selection was detected for the ADH gene cluster in a genome-wide analysis19. Collectively, the distribution of ADH1B*47His allele frequency in the populations studied cannot be explained by random genetic drift, and recent selection needs to be invoked.

Previous studies suggested a culture-related selection on the ADH1B*47His17. To test this, we superimposed the unearthed culture relic sites of rice domestication in East Asia and we observed a significant correlation of the ADH1B*47His allele frequencies with the ages of rice domestication (r = 0.769, p < 0.01, two-tailed t test; Figure 3; see Additional file 1). The origin of rice domestication occurred along the Yangtze River of southern China about 10,000 years ago2021. Based on the culture relics, the earliest rice sites are located in southern and south-eastern China (8,000-12,000 YBP), and then expanded to the central parts of China about 3,000-6,000 years ago, reaching Korea and Japan less than 3,000 years ago 2223. The spread of rice domestication agrees well with the distribution of ADH1B*47His, implying that rice domestication is likely the force driving up the frequency and expansion of ADH1B*47His in East Asia during the past 10,000 years. To see if the initial increase of ADH1B*47His in East Asia occurred during the same period as the emergence of rice domestication in early Neolithic time, we conducted molecular dating24 by typing the nearest STR loci (a CATA repeat STR located about 14 Kb upstream to the ADH1B locus, and a ATTC repeat STR located about 35 Kb downstream to the ADH1B locus) in 598 individuals randomly selected from the 38 populations. For phase reconstruction, only homozygous individuals with the ADH1B*47His alleles are included (see Additional file 2). The estimated ages based on the STRs are 5,525 YBP (CATA repeat), and 9,200 YBP (ATTC repeat). Considering that the two STR loci are still far away from the ADH1B locus, we also estimate the age of the ADH1B*47His based on the phased SNP haplotypes from the HapMap dataset (see Additional file 3). With the fine-scale genetic map, we selected 19 contiguous polymorphic SNPs to estimate the age (Table 2). Surprisingly, the estimated ages are extremely different between the upstream SNPs (114,693-208,919 yrs, 95% confidence interval) and the downstream SNPs (7,338-9,948 yrs, 95% confidence interval), which is due to the dramatic change of recombination rates in the studied genomic region. As suggested, the method based on the moments estimator24 is not suitable for the region of low average recombination rates. The previous genomic study based on the HapMap SNPs also excluded the regions with low average recombination rate25. Therefore, the age estimated based on the downstream SNPs seems to reflect the real age of ADH1B*47His allele, which is also consistent with the ages estimated from the STR variations. Taken together, the age of the derived allele at the ADH1B locus falls in the range of 10,000-7,000 years before present.

In this study, a total of 2,275 unrelated samples were collected from 38 populations (Table 1 and Figure 1). The Han Chinese samples were collected from individuals in 14 provincial areas whose geographic origins were assigned according to the birthplaces of their four grandparents, covering the major geographic regions in China. The other ethnic populations were sampled from south-western China and Qinghai-Tibet plateau where about 80% Chinese ethnic populations live with inhabited histories longer than 3,000 years34. All the samples were collected with informed consent. The protocol of this study was approved by the institutional review board of Kunming Institute of Zoology, Chinese Academy of Sciences.

Initially all the samples were genotyped for the ADH1BArg47His polymorphism (rs1229984). For genotyping by PCR-RFLP, we designed primers (forward primer, 5'FAM-GATTAGTAGCAAAACCCTCAAATAC-3'; reverse primer, 5'-CTAACCATGTGGTCATCTGCG-3') to cover this region. The restriction endonuclease used is Hin61 (Fermentas Life Sciences). The two contiguous microsatellites (CATA repeats and ATTC repeats) are from the UCSC Genome Database, located on Chr: 4 100471833-1889 and Chr: 4 100423654-3678. Both the microsatellite and the SNP genotyping were carried out by using an ABI 3130 sequencer, and the data was analyzed by using Genemapper software version 3.1 (Applied Biosystems). For haplotype analysis, we used the data from HapMap (Phase2.1)35.

We used the iHS statistics19 to detect recent positive selection. The obtained iHS value for the candidate SNP(rs1229984) and the empirical P-value for ADH1B were calculated by using Haplotter19. The extended haplotype homozygosity (EHH) and the relative EHH (REHH)18 were examined by the Sweep program, using the phased haplotype data set (CHB+JPT) from the HapMap project.

Allele age calculations are conducted by the standard methods published previously242536. In brief: t = [1/ln(1-c)]ln [(x(t)-y)/(1-y)], where t = allele age (in generations), c = recombination rate, x(t) = frequency in generation t, and y = frequency on ancestral chromosomes. We assume the origin of the ADH1B*47His allele is on the background of the ancestral allele haplotype, and the calculation utilizes the value of c, determined from the HapMap project recombination rate database http://hapmap.ncbi.nlm.nih.gov/downloads/recombination/latest/rates/. It should be noted that the East Asian samples from the HapMap are non-family data, therefore, limiting the estimation of recombination rates in these populations. Two types of polymorphism data have been used in our analysis. The phased haplotypes (CHB) containing a 45 Kb fragment, which includes 32 SNPs and covers the ADH1BArg47His polymorphism site, is obtained from the HapMap project website http://www.hapmap.org, and the microsatellites data is collected from the randomly selected 598 individuals. For conversion of time in generations, t, into time in years, a generation time of 25 years is assumed.