GENETIC STRUCTURE OF HUMAN POPULATIONS III. DIFFERENTIATION OF ABO BLOOD GROUP GENE FREQUENCIES IN SMALL AREAS OF JAPAN* MASATOSHI NEI and YOKO IMAIZUMI Division of Genetics, National Institute of Radiological Sciences, Chiba, Japan Receivedii .xii.65 1.INTRODUCTION INa previous paper (Nei and Imaizumi, 1966a) it was shown that the local differentiation of ABO and MitT blood group gene frequencies in Japan has occurred largely by genetic random drift. In that investi- gation the population of Japan was divided into 45 different sub- populations or Prefectures (administrative units of Japan) the sizes of which were mostly one to two millions. However, the size of mating groups of neighbourhoods in Wright's (1946) sense appears to be much smaller than the sub-populations employed in the paper. It is, there- fore, expected that the gene frequencies are locally differentiated even within these sub-populations. We thus examined the degree of differ- entiation of ABO blood group gene frequencies in several small areas of Japan. 2.DIFFERENTIATION WITHIN PREFECTURES Thereare two Prefectures, in which the local variation of ABO blood group gene frequencies can be analysed. One is Kagoshima Prefecture in Kyushu Island and the other Tokushima Prefecture in Shikoku Island (cf fig. i in Nei and Imaizumi, 1966a). The data on the former were collected by Ono and Takagi (1942) and Makisumi (1958), and those on the latter by Katayama (i) and Minamibuchi (i). The areas of Kagoshima and Tokushima Prefectures are 9,104 km2 and 4,143 km2 respectively, both including several neighbouring small islands. The population sizes of these two Prefectures at the time of 1960 census are 1,962,998 and 847,279 respectively. Blood group typing was made on 6,926 individuals at i6 different locations in Kagoshima Prefecture and on 3,467 individuals at idifferent loca- tions in Tokushima Prefecture. The locations at which blood group typing was made are given in figs. i and 2. These locations represent mura, machi, or city, administrative units smaller than Prefecture in Japan. The frequencies of the gene JA, j8, and J0 controlling the ABO blood groups were estimated by Berenstein's method (cf. Nei and Imaizumi, 1966a). The results obtained are given in table i, in which * This investigation was supported in part by a grant from the Scientific Research Fund of the Ministry of Education, Japan. 2G2 46! 462 MASATOSHI NEt ANDYOKOIMAIZUMI the frequencies of 1B, and 1 are denoted by p,q,and rrespectively. In Kagoshima Prefecture, the 2-value for the departure from the Hardy-Weinberg equilibrium is significant at the 5 per cent, level in one of the r 6 sub-populations, whereas, in Tokushima Prefecture, it is significant at the same level in one of the 14 sub-populations. Thus, the 0 Makizono 0 0 inshima Mizobe FIG. s.—Ceographical distribution of the locations for blood group typing in Kagoshima Prefecture. observed phenotype frequencies agree fairly well with the theoretical expectations. The x2forthe heterogeneity of gene frequencies among the sub-populations is 79'2 with 30 degrees of freedom in Kagoshima Prefecture and 88'o with 26 degrees of freedom, both of which are significant at the o i per cent. level. The degree of differentiation of gene frequencies may be measured by the inbreeding coefficient of sub-populations relative to the total GENETIC STRUCTURE OF HUMAN POPULATIONS 463 population. This inbreeding coefficient can be estimated from the means and genetic variances and covariances, as described by Nei and Emaizumi (i g66a). The estimates of the inbreeding coefficient thus obtained are set out in table 2. The inbreeding coefficients averaged over all alleles and allele combinations for Kagoshima and Tokushima are both larger than the value previously obtained for the whole of Japan, i.e. 00007 fNeiand Imaizumi, 1966a). This may be due 0 .-0 lchiu rurumlya 0 Sakashukjfo 0. IoI 0 0 tamikito Nakakito FIG. .—Geographical distribution of the locations for blood group typing in Tokushirna Prefecture. to the fact that the sizes of sub-populations employed for the whole of Japan were much larger than those for the present investigation, as will be discussed later. Table 2 also indicates that the inbreeding coefficient orf is more variable in Tokushima than in Kagoshima according to the allele or allele combination concerned. This suggests that the differentiation of gene frequencies has not occurred at random, but the possibility of sampling accident cannot also he ruled out. To test the significance of individual values offs from o, an approximate x2was computed by the following formula = 464 MASATOSHI NEI ANDYOKOIMAIZUMI where SS stands for the weighted sum of squares of the gene frequency concerned and for the mean sampling variance of gene frequency estimate (cf.Neiand Imaizumi, i g66a). The number of degrees of freedom of this x2isequal to the number of sub-populations minus one. Note that the x2isbased on the assumption of independent distribution TABLEi ABO blood group frequencies and estimates of gene frequencies in sub-populations of Kagoshima and To/cushima Prefectures (i) Kagoshima Prefecture Blood group frequency Gene frequency Sub- population 0 A B AB p q r 2 Kaseda 452 3214P4 59-4 7-1 o283 o144 0573 P0 Ata 5oo 6o40-6184 -00264 0126 o6,o Kasasa 377 39.5 379183 43 0241 o•121 0-638 2-4 Boodomari 505 327467 143 63 0316 0-109 0575 0-4 Chiran 517 300 404226 70 0277 0-162 o561 40 Izaku 316 269440 19-6 9503I9 0158 o523 02 Beppu 327 32-4 45-0 156 7-0 0307 0121 0572 01 Kawanabe 304 30.9 497 135 590335 0103 0562 07 Ibusuki 318 252 50.9 157 82 o•363 0-128 0509 o'8 Aira 379 290 40-4 208 g8 o294 0-167 0539 0-0 Mizobe 565 297 43-6 57-7 90 0-311 0-144 0545 00 Makizono 721 277 465 z6- 93 0-335 o-i8 0527 00 Kirishima 789 29-7 42-6i8-6 91 0305 0150 0-545 00 Takasu 312 31'! 39'8 20-8 8-3 0-280 0-159 o-6i 0-2 Shibushi 297 30.3 404 i8'g 10-4 0'298 o'i8 0544 0-5 Sueyoshi 247 27'! 441207 8i 0311 O'157 0-532 P2 Total5 6,926 30'7 43.3 18-2 7-8 0-302 0'141 0-557 55 (ii) Tokushuna Prefecture Tokushima 467 223 450 22-0 10-7 0-337 o-,8i 0-482 i-6 Kamojima 242 302 388 203 10.7 0289 oi68 0-543 0-4 Kawata 137 241 3942P9146 0319 0-202 0-479 o-6 Miyama 77 29-9 40-2 18-2 1P7 0-305 o'i6s 0534 05 Shishikui 132 22'O 42-4 220 13-6 0-337 0,597 0-466 0-0 Orono 333 24-3 42'4222 II-! 0-319 0-184 0-497 02 Aioi 221 32'6 4P6 ,g'o 68 0-283 0-139 0578 0-5 Fukuhara 411 31-6 49-6 12-7 6-s 0-335 0-099 o-66 0-3 Sakashukito 196 30-6 48-0 i-8 6 0'321 0114 0'565 31 Nakakito 89 32-6 483 11-2 7.9 0-336 0-100 0564 0-3 Kamikito 87 29'g49.4 126 8-, 0347 0-109 0-544 00 Ichiu i8i 282 50.3 138 7.7 0353 0-114 0533 0-0 Furumiya 158 17-7 393 34.8 8-2 O'283 0-253 0-464 8-i Matsushima 736 315 390 203 9-2 0-280 o-i6o 0-560 0•1 Totals 3,467 280 433 195 9-2 0'312 0-157 0-531 P4 * Genefrequencies inthe row oft' Total" refer to the weighted means. =samplesize. GENETIC STRUCTURE OF HUMAN POPULATIONS 465 of different gene frequencies, which is true only for a large number of multiple alleles. At any rate, the x2thusobtained indicates that, in Tokushima, one of the three estimates offs, i.e.thatobtained from the frequency of JA, is not statistically significant, whereas, in Kagoshima, all of the three estimates are significant (table ). TABLE2 Means,genetic variances, and degrees of djffrentiation of gene frequencies in Kagos/iima and Tokushima Prefectures Kagoshima Tokushima Alleles — concerned Mean Mean cu frequency JST frequency fsr 11 O'3023 O'000474 O'oo225 03124 o'000128 o'ooo6o I O'1404 O'000187 O'00155 O'1567 o'ooii86 O'00897 JO O'5573 0000605 O'00245 05309 O'000836 o'00336 JA,Jfl —O'000025 o'ooo6o —O'000239 O'00488 14.10 —0'000443 0'00263 ooooi ii —o00067 1&I0 —O'000158 000202 —O'000947 O'01138 Mean oooigu 000475 genetic variance or covariance. fsr inbreeding coefficient. TABLE 3 x'-values for testing the sign j/icance of individual values of f51 x* Area D.F. P q Kagoshima 15 41.7*** 35.9** 43.8*** Tokushima 13 17'4 71.I*** 32.7** Kami-ina 5 I9'8 3'8 25.75** Chiisagata 8 7.0 24 7.9 Sado 6 27.2*** 9.7 22.7*** Tsushima ii 38'6 49.2*** 37.1*** Oki 4 6'o 29.9*** 114* Kuroshima 7 22'4 31.8*** 27.O*** * Significantat the 5 per cent, level. *Significantat the z per cent, level. ***Significantat the o' i per cent, level. The randomness of differentiation of gene frequencies can be tested by comparing the observed with the expected correlations of gene frequencies (Nei, 5965). The expected and observed correlations, obtained by the method given by Nei and Imaizumi (xg66a), are presented in table 4.