The Genetic Structure of a Tribal Population,

The Yanomama Indians are a South American tribe distributed over an irregular area approximately 200 x 330 miles. The gene frequencies observed at 12 loci in 47 villages within this area have bpen analyzed for the occurrence of clines. Apparently significant clines are observed for alleles of the Rh, MNSs, Kidd, Gm, Inv and serum albumin system. Available data concerning recent tribal expansion and admixture permit a tentative analysis of the causes of these clines. Although the action of selection cannot be rigorously excluded, it seems unlikely to be the major cause. Admixture with surrounding tribes plays a role which can be quantified because of the fortuitous circumstance of two genetic markers for admixture. It is suggested that an important factor in the origin of these clines is the manner in which the tribe has recently expanded through successive village fissionings and a predominantly centrifugal pattern of village migration. I N the past decade it has been amply demonstrated that "natural" human populations display extensive subdivision and genetic microdiff erentiation. Given the extent of interdeme variability within these populations, it becomes pertinent to investigate the organization of this genetic heterogeneity in subdivided populations. Thus far in our own program, the bulk of inquiry has been directed to two types of investigation (apart from the continued demonstration of the extensive nature of genetic microdifferentiation). On the one hand, measures of pairwise genetic relationship have been derived and the set of relationships for a group of demes shown to be in accord with ethnohistorical data (viz., WARD and NEEL 1970; WARD 1972), or with other biological attributes of the populations (viz., SPIELMAN 1973; NEEL, ROTHHAMMER and LINGOES 1974). On the other hand (cf., NEEL and WARD 1972), the genetic heterogeneity has been partitioned following the concepts of population subdivision elaborated by WAHLUND (1928) and modified by WRIGHT ( 1943,1965). In general, the analyses of other workers on a diversity of populations have tended to adhere in principle to these two types of analysis--though the methods vary. These studles were supported in part by AEC NSF Contract AT(ll-1)-1552 with the University of Michigan The project of which they were a part was an Integrated Research Program of the U S International Biological Program Present address CD380, Child Development and Mental Retardatlon Center, WJ-IO, Universlty of Washington, Seattle, Washington 98195 Genetics 82: 103-191 January, 1976. 104 R. H. WARD AND J. V. NEEL In this paper we pursue a third approach to the study of allelic variability, by investigating the association between allele frequencies and the spatial location of the demes within the population. This approach most commonly takes the form of a search for geographic clines in allele frequencies. There are at least three well-defined factors that can cause a significant degree of geographic structure in the pattern of alleles (i.e., clines) : selection, differential movement of population subdivisions, and immigration of “foreign” alleles into the population. However, for the majority of na turd populations, the background of information and allele frequencies are such that there is no obvious way of partitioning the observed geographic structure into components. The Yanomama offer the potential for partitioning the association between geography and allelic frequency into comporient parts. A recent report from this laboratory described the discovery of a L‘private’’ polymorphism of serum albumin (Yanomama-2) among the Yanomama Indians of Southern Venezuela and Northern Brazil (TANIS et al. 1974a). An unusual feature of this polymorphism was a pronounced cline, gene frequencies ranging from approximately 0.30 in the northwest of the tribal distribution to 0.03 in the southeast. This finding stimulated an investigation of all our accumulated genetic data on this group for the occurrence of clines. In this paper we will report significant clines for one or more alleles in 6 of 12 genetic systems examined. I t will be argued that selection is unlikely to be the major cause of this phenomenon. On the other hand, admixture (intrusion of genes from surrounding tribes) and the recent centrifugal expansion of the tribe do appear to have significant roles. It will be suggested that the role of structured population expansion in the production of clines in natural populations has been inadequately appreciated in the past. MATERIALS AND METHODS Traditionally, and in this paper, geographic structure has been elucidated by using regression techniques to measure the association between allele frequencies and a geographic component (e.g., latitude, longitude, altitude, etc.). However, the demonstration that allelic frequencies are monotonically increasing functions of a geographic attribute depends on evaluating the regression. The usual test of statistical significance is based on the inherent assumption that the set of spatially defined gene frequencies results from independent samples, both within and between demes. Although this assumption is often overlooked (disregarded), SMITH (1969) and others have clearly indicated that this situation rarely if ever attains for most biological populations, and certainly does not hold for the Yanomama. Hence the relationship between statistical “significance” and biological relevance is tenuous at best. In addition, the geographic placement of Yanomama villages is temporally unstable and nonrandom, being a complex function of past migrational history, present-day political relationships, the topography of the environment, etc. Given the marked degree of genetic microdifferentiation that exists in such populations, which is also nonrandom in distribution due to the frequent propagation of villages by fissioning along lineal lines (cf. NEEL and SALZANO 1967), the demonstration of a cline by any method of fitting data to a line or surface should be interpreted with caution with respect to both statistical and biological significance. This is because the error of estimate associated with any cline will be appreciable and not readily definable in biological terms since the sampling error of a cline will be inflated by errors of estimate both for the allelic frequency for each deme (cf. WARD et al. 1975) and the exact CLINES I N A N I N D I A N TRIBE 105 geographic location of the villages. The extent to which these deviations from the inherent assumptions of the model affect the analysis is not wholly clear. The statistical problems of estimating clinical variability in the simple case where a population is distributed as a set of finite aggregates in a uniform environment are appreciable: the complex realities of the Yanomama present a formidable challenge. In our analysis, while we invite comparison with the indicated levels of "statistical significance", we present our results in terms of a n ad hoc procedure. The model we use for our analysis utilizes the logit transformation of village allele frequencies, regressed on latitude and longitude. In the two-allele case the regression equation is defined by