genetic analysis of chest dimensions in a high altitude tibetan population from upper chumik, nepal

AMERICAN JOURNAL OF HUMAN BIOLOGY 51719-724 (1993)

Genetic Analysis of Chest Dimensions in a High Altitude Tibetan Population From Upper Chumik, Nepal

S. WILLIAMS-BLANGERO', J. BLANGEROI, AVD C.M. BEALL' 'Department of Genetics, Southwest Foundation far Biomedical Research,, P.O. Box 28147, Sun Antonio, Texas 78228-0147, and "Case Western Reserve Uniuersity, Cleveland, Ohio 441 06

ABSTRACT Studies of high altitude Andean natives generally report large chest dimensions relative to sea level groups, while results from studies of Hima- layan populations are not consistent. One hypothesis is that this may represent different adaptive patterns in the two geographic areas. The purpose of this study is to explore the determinants of chest dimensions by assessing the genetic components of variat.ion in chest width and chest depth in a Tibetan population resident at 3250-3560 m in Upper Chumik, Nepal. Data were available for 608 individuals (298 males, 310 females) aged 2-79 years; 471 individuals could be assigned to 134 pedigrees containing between 2 and 10 people with data. The remaining 137 individuals were treated as independents. A maximum likelihood variance decomposition method was used to assess the genetic components of each trait. Growth in chest width and depth was modeled as a linear function of age until an asymptote, at which growth was considered to reach a plateau. Both chest width and chest depth were moderately heritable (h$idth = 0.50 * 0.10, h&,th = 0.49 * 0.08). There was a significant genetic correlation (0.43 +- 0.13) between chest width and depth, suggesting that some genes have pleiotropic effects on both traits. The observed significant genetic components of these morphological traits indicate that there is potential in this population for genetic adaptation of chest dimensions to the high altitude environment. c 1993 Wiley-Liss, Inc.

Studies of adaptation to the hypoxic high altitude environment have often focused on chest dimensions as a proxy measure of lung capacity (Frisancho, 1969; Frisancho et al., 1993; Hurtado, 1932; Mueller et al., 1978; Beall, 1982). The majority of these studies have been conducted in two geographic areas having populations with long histories of high altitude residence, the Andean region of South America and the Himalayan region of Asia. A general finding of the Andean studies is that high altitude natives have larger chest dimensions than low altitude natives (Frisancho et al., 1975; Beall et al., 1977; Mueller et al., 1978; Stinson, 19801, although Hoff (1974) did not detect this pattern. However, Himalayan high altitude natives do not consistently exhibit significantly larger chest dimensions than low altitude natives (Beall, 1982). It has been hypothesized that these divergent results may be due to different adaptive mecha-

nisms operating in the two geographic re- gions (Beall, 1982; Pawson, 1976).

The question of whether or not observed patterns in chest dimensions are due to developmental or genetic adaptation has been raised by a number of authors (Beall et al., 1977; Greksa, 1986, 1990; Frisancho and Greksa, 1989; Greksa and Beall, 1989). However, there has been limited consideration of the genetic influences on chest dimensions. For genetic adaptation to an environmental stress t o occur, the trait under consideration must have a genetic basis. This fundamental assumption, inherent to much of high altitude adaptation research, has rarely been tested explicitly.

Received May 5,1993: accepted July 5,1993. Address reprint requests to S. Williams-Blangero, Dept. of Ge-

netics, Southwest Foundation for Biomedical Research, P.0. Hox 28147, San Antonio, TX 78228-0147.

0 1993 Wiley-Liss, Inc

720 S. WILLIAMS-BLANGERO ET AL

TABLE I . Distribution of pedigree sizes

Pedigree Number of Number of size Dedimees individuals

1 2 3 4 5 6 7 8

10 Total

137 43 41 18 17 5 7 2 1

271

137 86

123 72

30 49 16 10

608

a5

Genetic epidemiological techniques are being applied to anthropological problems with increasing frequency (Williams-Blan- gero and Blangero, 1993), and the classical research questions in human adaptability are particularly well suited to this statistical genetic approach (Blangero, 1993; Beall et al., 1993). The purpose of this paper is to quantify the genetic components of two measures presumed to be associated with adaptive lung function, chest width and chest depth, in a high altitude Tibetan population residing in Upper Chumik, Nepal.

MATERIALS The data were collected in 1981 as part of

a study of high altitude adaptation in Ti- betan speaking Buddhists resident in Upper Chumik, Nepal (Beall and Reichsman, 1984). Upper Chumik is a network of six permanently inhabited villages at altitudes ranging between 3250 and 3560m in the Mustang District of north central Nepal (Schuler, 1987; Beall and Reichsman, 1984).

Two measures of chest size, chest width (transverse diameter) and chest depth (an- teroposterior diameter), were evaluated. The chest dimensions were measured by Cynthia Beall while subjects were standing at mid-expiration according to the protocols described by Cameron et al. (1981). Data on the two chest dimensions were available for 608 individuals, 298 males and 310 females. The subjects ranged in age from 2 to 79 years. Utilizing information on mother, fa- ther, and household, the data were assem- bled in pedigrees utilizing the pedigree based data management system PEDSYS (Dyke, 1989). Table 1 presents the distribution of pedigree size in the sample; 471 of the individuals belonged to a total of 134 pedigrees. These pedigrees contained between 2

and 10 individuals with data. There were 137 unrelated (and hence independent) individuals.

STATISTICAL METHODS Chest width and chest depth are corre-

lated and should not be analyzed as independent traits. Therefore, rather than using a traditional univariate variance decomposition approach to assess the genetic components of chest dimensions, a multivariate covariance decomposition approach is re- quired t o incorporate the pleiotropic effects of genes on both traits (Hopper and Mathews, 1982; Lange and Boehnke, 1983; Williams-Blangero and Blangero, 1992).

Multivariate genetic model A multivariate quantitative genetic anal-

ysis using the general method developed by Blangero and Konigsberg (1991) was performed. This technique allows for simulta- neous evaluation of genetic effects and the effects of covariates (such as age and sex) on a set of related traits by finding parameter estimates that maximize the likelihood of observing the data. For example, if we have information t quantitative traits, the phenotypic vector (y) of the j th individual given his covariates is assumed to take the following form:

(y,IxJ = I*. + gJ + P ' X j + €, (1)

where p is a t x 1 vector of means, g is a vector of additive genetic effects, x is a q x 1 vector of covariates, p is a q x t matrix of regression coefficients for the q covariates, and E is a vector of random environmental deviations. Assuming that bothg and E have means of zero, the expectation of yJ is

E(y,) = p + p'xj. (2)

Equation ( 2 ) defines the mean effects model. While a simple linear model for covariate effect was used here, any nonlinear model may also be used.

Complete specification of the statistical model requires that, in addition to the mean effects, the variance components be defined. For the multivariate polygenic model, the total phenotypic covariance matrix is

GENETIC ANALYSIS OF CHEST DIMENSIONS 721

where G is the additive genetic covariance matrix due to polygenes and E is the covariance matrix due to random environmental effects.

In order to estimate G , data from sets of related individuals organized into pedigrees which are assumed to be independent of one another are necessary. The likelihood function for a given pedigree takes the following multivariate normal form:

where vec( ) is a matrix operator that stacks the columns of a matrix underneath one another, Y is the n x t matrix of phenotypes, 1 is an n x 1 column vector on ones, and X is the n x q matrix of covariates. In the above likelihood function, the phenotypic covariance matrix for the pedigree is

Var[vec(Y)I = f1 = G @ 2 0 + E @ I n (5)

where @ is a Kronecker product operator, is the kinship matrix for the pedigree, and

I, is an n x n identity matrix, Evaluation of the likelihood in Eq. (4) re-

quires inversion of an nt x nt matrix which can be time-consuming for large pedigrees. A method that circumvents this problem by simplifying the multivariate likelihood via a transformation that simultaneously orthog- onalizes P, G, and E was used. This method is described fully by Blangero and Konigs- berg (1991). The transformation makes the traits genetically and environmentally un- correlated. Multivariate likelihoods then re- duce t o the product of t univariate likelihoods. Using this method, the likelihood computations are much faster than those based on inversion of the complete phenotypic covariance matrix.

Linear response and plateau model Age-dependent growth in chest dimen-

sions was explicitly incorporated in the genetic model. Based on preliminary explor- atory analyses of the chest dimension data available for the Upper Chumik population, sex-specific growth in chest width and chest depth was modeled as a linear function of age, until an asymptote a t which growth reached a plateau.

The expected phenotype for males is given

(6)

by

E(y) = pM + Nage,) age*

where age” = age - aM if age < aM = O if age 3 aM

In Eq. (6), pM is the expected mean trait value in males after growth as stopped at age aM. The parameter p(age,) is simply the regression coefficient on age for ages less than a,.

For females an additional sex effect lp(sex)l was incorporated, and the expected phenotype is given by

E(y) = pM + p(sex) + P(age,) age” (71

where age* = age - aF if age < aF = o if age 2 aF.

Such a linear response and plateau model has been used successfully in cross-sectional growth analyses (Konigsberg et al., 1990) and in previous genetic analyses (Konigs- berg and Cheverud, 1992; Williams-Blan- gero et al., 1992) of quantitative traits.

Parameter estimation The parameters of the model were esti-

mated by numerically maximizing the likelihood given in Eq. (4) over all pedigrees. The computer program PAP (Hasstedt, 19891, incorporating the subroutines developed by Blangero and Konigsberg (1991), was used to perform all genetic analyses. To simplify hypothesis testing, the variance components were reparameterized to include phenotypic standard deviations for chest width (upw) and chest depth (up& trait-specific heritabilities ( h t and h:), the additive genetic correlation between the two traits (pG), and the environmental correlation between chest dimensions (pE).

Hypothesis testing All hypothesis tests were performed using

likelihood ratio statistics. The likelihood ratio statistic (A) is calculated as twice the difference between the In likelihood of a fully parameterized general model (i.e., the most complex model considered where all previously defined parameters are estimated) and a constrained submodel repre- senting the hypothesized model for the trait. This statistic is distributed similarly t o a


M

?'ABLE 2. Maximum likelihood estimates of mean efects

Parameter Chest width Chest depth

FM 27.214 f 0.112 19.504 ? 0.108 p(sex) -2.139 i 0.145 -1.166 t 0.141

0.638 ? 0.019 0.305 IO.015 p(ageF) 0.599 ? 0.026 0.316 ? 0.018 a M 21.468 ? 0.367 25.763 2 0.761 a,.. 18.729 ? 0.453 22 382 t 0 723

chi-square with the degrees of freedom being determined as the difference in number of estimated parameters between the general model and the submodel.

RESULTS AND DISCUSSION Table 2 presents the maximum likelihood

estimates of parameters for the linear response and plateau model. These estimates were obtained simultaneously with those of the variance components in the genetic analyses. Expected chest width in adult males is 27.21 cm. The slope of response to age for chest width was 0.64 cdyear in males and growth in chest width ceased at 21.5 years. Adult females had an expected chest width 2.14 cm less than males (25.07 cm). They exhibited a very similar rate of growth to the males, the slope being 0.59 cdyear, but growth in this chest dimension ceased at 18.7 years, over 2 years earlier than in males. Expected chest depth was 19.50 cm for adult males and 18.33 cm in adult females. Chest depth demonstrated a slower rate of growth than chest width but was again similar for both sexes; the slope was 0.30 in males and 0.32 in females. Growth in chest depth plateaued at 25.6 years in males and 22.4 years in females.

Figure 1 shows the linear response and plateau functions obtained for the two measures of chest size for males and females. The plot is truncated at 30 years of age, although the plateaus reflect data through age 79 years.

Several hypotheses regarding the mean effects parameters were tested. Results are presented in Table 3. Adult females were significantly smaller than males for both chest width and chest depth as inferred from likelihood ratio tests where the submodels were constrained so that there were no sex effects fi.e., @(sex) = 01. The rate of growth was not significantly different for either trait in the two sexes. Finally, for growth in

25

20

15

I " ~

0 10 20 30

Age

Fig. 1. Plot of linear response and plateau functions for chest width and chest depth. Age is arbitrarily truncated a t 30 years.

TABLE 3. Likelthood ratio tests of mean effects

Chest width Chest depth Hypothesis A 1 P '4 1 P

p(sex) = 0 186.54 <0.0001 55.72 10.0001 P(ageM) = p(age,) 1.54 0.215 0.23 0.630 OLM = a p 21.03 <0.0001 4.66 0.031

TABLE 4. Maximum likelihood estimates of uariancelcouariance components

Parameter B * SE

1.359 2 0.041 0.500 ? 0.098 1.243 * 0.037 0.489 -t 0.084 0.430 t 0.131 0.114 2 0.117

both chest width and chest depth, females exhibited significantly earlier ages at plateau.

Table 4 presents the maximum likelihood estimates of variance/covariance components and Table 5 gives the corresponding statistical test results. Chest width, with a phenotypic standard deviation of 1.36 cm, is slightly more variable than chest depth (uPD = 1.24). The heritabilities reflect the proportion of variance among individuals that is attributable to additive genetic factors. Approximately 50% of the variation in

GENETIC ANALYSIS OF CHEST DIMENSIONS 723 TABLE 5. Likelihood ratio tests of

variancelcooariance comuonents

Hypothesis A, P

h $ = O 31.19 <0.0001 hg=O 35.31 ~0 .0001 Pc, = 1 19.02 ~0.0001 PG = 8.47 0.004 pE = 0 0.85 0.357

chest width is due to genetic factors, and about 49% of the variation in chest depth is due to genetic components. Both these heritabilities are significantly different from zero according to a likelihood ratio test, with corresponding P values of less than 0.0001. The relatively small standard errors for the heritabilities reflect the efficient use of all possible biological relationships in the estimation process, unlike most previous studies which are limited to nuclear families.

There was a significant genetic correlation of 0.43 i 0.13 between chest width and chest depth in the Upper Chumik sample. The hypothesis that there was complete pleiotropy, i.e., that the genetic correlation between the traits was 1, was tested using a likelihood ratio test. The hypothesis was rejected with a P value less than 0.0001. The hypothesis that there was no pleiotropy, i.e., that no shared genes influence the traits, was also tested, and this was also rejected with a P value of 0.004. These results indicate that there are some common genes which influence both chest width and chest depth. It is interesting to note that the genetic correlation was substantially higher than the phenotypic correlation, which was calculated as a function of the observed maximum likelihood variance/covariance component estimates as 0.27. Evolutionary dynamics are more dependent on genetic correlations than phenotypic correlations, suggesting that even traits with low phenotypic correlations may in fact be evolving as a closely related genetic complex.

The random environmental correlation between the two chest dimensions was not significantly different from zero (0.11 & 0.12, P = 0.36). This suggests that shared random environmental effects do not exert major influences on the covariation between chest width and depth.

Genetic variance for most traits is re- markably stable across populations. Finding differences in genetic variance for chest dimensions between populations may be in-

dicative of differences in selection histories between the groups for the traits in questions. Heritabilities may of course differ due to between-population differences in the environmental variance. The genetic standard deviations and heritabilities for chest width and chest depth obtained for the Upper Chu- mik population are shown in Table 4. These heritability estimates are fairly similar to those observed in the Phala nomads of Tibet living at 4870-5480 m (h$ = 0.64 * 0.14, h; = 0.38 f 0.13) (unpublished data). The values for the genetic standard deviations for residents of Upper Chumik (crGw = 0.96, uGD = 0.87) are virtually identical to those of Phala nomads (uGw = 1.07, crGD = 0.86) for chest width and chest depth. Unfortu- nately, the only available genetic analysis of these traits in a South American high altitude population, the Aymara (Kramer, 1992), does not provide enough information to generate an estimate of the genetic standard deviation in this population. Interest- ingly, the heritabilities obtained for chest dimensions in this Aymara population are smaller (h$ = 0.15 ? 0.10, hf!, = 0.28 2 0.101, which may reflect more recent intense directional selection in this population than in the Himalayan groups. However, this in- ference should be viewed with extreme cau- tion since simplified approaches which do not make full use of all available data were used to calculate the Aymara heritabilities.

Genetic variance is a prerequisite for natural selection resulting in genetic adaptation to the environment. In order to under- stand the evolution of traits in response to the high altitude environment, further studies which estimate directly comparable genetic variances for associated traits in both high and low altitude populations are re- quired. Once the presence of genetic variance for a quantitative trait is established, the genetic component of an individual's phenotype can be estimated. The relation- ship between these individual genetic values and the true currency of natural selection, fertility and mortality, may then be evaluated revealing the current status of se- lective pressures.

In summary, two aspects of chest morphology, chest width and chest depth, have significant moderate heritabilities. There was a significant genetic correlation between chest width and depth suggesting that some genes may influence both traits. These results indicate that there is potential


for genetic adaptation of chest dimensions to the high altitude environment in this Ti- betan population. Future analyses will focus on refining genetic models to determine if the effects of specific individual genes on chest dimensions can be detected.

ACKNOWLEDGMENTS The data for this paper were gathered

during fieldwork supported by the National Science Foundation under Grant BNS 80- 14317 awarded to CMB. The statistical methods and their computer implementa- tions were developed while J R was supported by NIH Grants HL28972, HL45522, GM31575, and DK44297, and SW-B was supported by NIH Grants HL45522 and RR08122. We thank H.M.G. of Nepal for granting permission to collect these data.

LITERATURE CITED Beall CM (1982) A comparison of chest morphology in

high altitude Asian and Andean populations. Hum. Biol. 54t14.5-163.

Beall CM, Baker PT, Baker TS, Haas J D (1977) The effects of high altitude on adolescent growth in south- ern Peruvian Amerindians. Hum. Biol. 49r109-124.

Beall CM, Klangero J, Williams-Blangero S, Goldstein MC (1993) A major gene for 0, saturation of arterial hemoglobin in Tibetan highlanders, Am. J. Phys. An- thropol., 941-966.

Beall CM, Reichsman AB (1984) Hemoglobin levels in a Himalayan high altitude population. Am. J. Phys. An- thropol. fi3;301-306.

Blangero J 1993) Statistical genetic approaches to human adaptability. Hum. Biol. 65, in press.

Blangero J, Konigsberg LW (1990) Multivariate segregation analysis using the mixed model. Genet. Epide- miol. 8: 299-3 16.

Cameron N. Hiernaux J , Jarman S, Marshall WA, Tan- ner JM, Whitehouse J H (1981) Anthropometry. In JS Weiner and JA Lourie (eds): Practical Human Biol- ogy. London: Academic Press, pp. 26-52.

Dyke B (1989) PEDSYS: A Pedigree Data Management System. User’s Manual. Population Genetics Labora- tory Technical Report No. 2, Southwest Foundation for Biomedical Research, San Antonio, Texas.

Frisancho AR (1969) Human growth and pulmonary function of a high-altitude Peruvian Quechua population. Hum. Biol. 41r365-379.

Frisancho AR. Borkan GA, Klayman J E (1975) Pattern of growth of lowland and highland Peruvian Quechua of similar genetic composition. Hum. Biol. 47~23% 243.

Frisancho AR, Frisancho HG, Albalak R, Villena M, Vargas E, Soria R (1993) Lung volumes of natives and Bolivians of non-native ancestry living at high altitude. Am. J . Phys. Anthropol., Supplement 1639 (abstract).

Frisancho AR, Greksa LP (19891 Developmental responses in the acquisition of functional adaptation to

high altitude. In MA Little and JD Haas (eds1: Human Population Biology: A Transdisciplinary Science. New York: Oxford University Press, pp. 203-221.

Greksa LP (1986) Chest morphology of young Bolivian high-altitude residents of European ancestry. Hum. Riol. 58r427-443.

Greksa LP (1990) Developmental responses to high-altitude hypoxia in Bolivian children of European ancestry: A test of the developmental adaptation hypothesis. Am. J. Hum. Biol. 2t603-612.

Greksa LP. Beall CM (1989) Development of chest size and lung function at high altitude. In MA Little and JD Haas (eds): Human Population Biology: A Trans- disciplinary Science. New York: Oxford University Press, pp. 222-238.

Hasstedt SJ (1989) Pedigree Analysis Package. Salt Lake City: Department of Human Genetics, Univer- sity of Utah.

Hoff C (1974) Altitudinal variations in the physical growth and development of Peruvian Quechua. Homo

Hopper JL, Mathews JD (1982) Extensions to multivariate normal models for pedigree analysis. Ann. Hum. Genet. 46t373-383.

Hurtado A (1932) Respiratory adaptation in thc Indian natives of the Peruvian Andes: Studies at high altitude. Am. J. Phys. Anthropol. 17t137-165.

Konigsberg LW, Cheverud J M (1992 1 Uncertain pater- nity in primate quantitative genetic studies. Am. J . Primatol. 27t133-143.

Konigsberg LW, Falk D, Hildebolt C, Vannier M, Chev- erud J , Helmkamp RC 11990) External brain morphology in rhesus macaques (Macnca mulattai. J. Hum. Evol. 19r269-284.

Kramer AA (1992) Heritability estimates of thoracic skeletal dimensions for a high-altitude Peruvian population. In RB Eckhardt and TW Melton (edsj: Popu- lation Studies on Human Adaptation and Evolution in the Peruvian Andes. Occasional Papers in Anthropol- ogy, No. 14. University Park, PA Matson Museum of Anthropology, Pennsylvania State University, pp. 25- 49.

Lange K, Boehnke M (1983) Extensions to pedigree analysis. IV. Covariance components models for multivariate traits. Am. J. Med. Genet. 14t513-524.

Mueller WH, Schull VN, Schull WJ, Soto P, Rothham- mer F (1978) A multinational Andean genetic and health program: Growth and development in an hypoxic environment. Ann. Hum. Biol. 51329-352.

Pawson IG (1976) Growth and development in high altitude populations: A review of Ethiopian, Peruvian, and Nepalese studies. Proc. Royal SOC. London, B 194r83-98.

Schuler SR (1987) The Other Side of Polyandry: Prop- erty, Stratification, and Nonmarriage in the Ncpal Himalayas. Boulder, CO: Westview.

Stinson S (1980) The physical growth of high altitude Bolivian Aymara children. Am. J. Phys. Anthropol. 52t377-385.

Williams-Blangero S, Blangero J (1992) Quantitative genetic analysis of skin reflectance: A multivariate approach. Hum. Biol. 64t35-49.

Williams-Blangero S , Rlangero J (1993) Genetic epidemiological methods in anthropological research. Hum. Biol. 65, 871-874.

Williams-Blangero S, Blangero J , Mahaney MC (1992) Segregation analysis of craniometric traits incorporating genotype-specific growth patterns. Am. J . Hum. Genet. 51tA163 (abstract).

24;87-99.

genetic analysis of chest dimensions in a high altitude tibetan population from upper chumik, nepal

Documents