Report

Genetic Evidence for High-Altitude Adaptation in Tibet

See allHide authors and affiliations

Science  02 Jul 2010:
Vol. 329, Issue 5987, pp. 72-75
DOI: 10.1126/science.1189406

Abstract

Tibetans have lived at very high altitudes for thousands of years, and they have a distinctive suite of physiological traits that enable them to tolerate environmental hypoxia. These phenotypes are clearly the result of adaptation to this environment, but their genetic basis remains unknown. We report genome-wide scans that reveal positive selection in several regions that contain genes whose products are likely involved in high-altitude adaptation. Positively selected haplotypes of EGLN1 and PPARA were significantly associated with the decreased hemoglobin phenotype that is unique to this highland population. Identification of these genes provides support for previously hypothesized mechanisms of high-altitude adaptation and illuminates the complexity of hypoxia-response pathways in humans.

The Tibetan highlands are one of the most extreme environments inhabited by humans. Many present-day Tibetan populations are thought to be descendants of people who have occupied the Tibetan Plateau since the mid-Holocene, between 7000 and 5000 years ago (1), and possibly since the late Pleistocene, ~21,000 years ago (2, 3). Compared with Andean populations living in similar high-altitude conditions, Tibetans exhibit a distinct suite of physiologic traits: decreased arterial oxygen content, increased resting ventilation, lack of hypoxic pulmonary vasoconstriction, lower incidence of reduced birth weight, and reduced hemoglobin (Hb) concentration (on average, 3.6 g/dl less for both males and females) (48). Neighboring Han Chinese individuals and other nonadapted lowland visitors to high-altitude regions develop increased Hb concentration to compensate for the hypoxic high-altitude environment (9), and this response is associated with adverse effects (10, 11).

High-altitude Tibetans maintain normal aerobic metabolism, despite profound arterial hypoxia (4), perhaps through the existence of changes in the oxygen-transport system. For example, elevated circulating NO levels increase vasodilation and blood flow (12), which, when combined with increased ventilation (13), may increase the availability of oxygen to cells (4). Collectively, these traits strongly suggest that Tibetans have adapted uniquely to extreme high-altitude conditions. The genetic basis of this adaptation, however, remains unknown.

We used two intersecting criteria to identify genes potentially involved in high-altitude adaptation: First, a priori candidates for adaptation to high-altitude hypoxia were chosen because of their known functions (14). Second, a genome-wide scan was conducted to identify regions that show strong evidence of local positive selection in high-altitude Tibetans (Fig. 1). To generate a set of a priori functional candidate loci, we constructed a list of Gene Ontology (GO) project categories (15) associated with the traits discussed above (Table 1). We merged genes from this list with those in the Panther-defined pathway “hypoxia response via activation of hypoxia-inducible factor (HIF)” (16), a major transcriptional regulator of oxygen homeostasis (17) that is probably associated with high-altitude adaptation. The resulting set of 247 functional candidate loci is listed in table S2.

Fig. 1

Gene regions responsible for adaptation to high-altitude hypoxia in Tibetans. (A) The strategy used to identify a list of genes related to high-altitude adaptation to hypoxia relies on three sets of genes. The set of functional candidates (yellow) consists of genes associated with physiological traits related to hypoxia (see Table 1 for categories). The XP-EHH (light blue) and iHS (dark blue) selection candidate sets include genes in the top 1% of the empirical distributions of XP-EHH and iHS results, respectively, excluding those with evidence of positive selection in neighboring populations (see SOM). The intersection of functional candidates with selection candidates (outlined in black) is enriched for regions containing genes that contribute to local adaptation to hypoxia in Tibetans. The genes in the intersection of functional candidates with iHS selection candidates still exhibit genetic variability in the population. (B to D) Comparison between Tibetan and CHB-JPT genomic regions identified in selection scans. The top and bottom halves of each figure represent chromosome regions in the Tibetan (number of chromosomes = 62) and CHB-JPT populations (62 randomly drawn chromosomes from 90 individuals), respectively, for the (B) EPAS1, (C) EGLN1, and (D) HMOX2 genes identified in XPEHH, both scans, and iHS, respectively. The three SNPs with the highest iHS and XP-EHH scores (indicated by an asterisk) were designated as the core haplotype for each genomic region. All haplotypes were sorted to the horizontal midline of each panel based on the length of uninterrupted matches to the reference sequence. See fig. S3 and table S5 for the remaining seven regions and details about these regions.

Table 1

Categories used to define the functional candidate a priori gene list. N, number of genes in each category. Total number of unique genes: 247.

View this table:

We next identified alleles subject to strong recent positive selection (a selective sweep) in a sample of 31 unrelated Tibetans who were genotyped for one million single-nucleotide polymorphisms (SNPs) using the Affymetrix Genome-Wide Human SNP 6.0 Array. These individuals showed no evidence of admixture with neighboring populations [see supporting online material (SOM), figs. S1 and S2]. To pinpoint loci under positive selection, we first used the cross-population extended haplotype homozygosity (XP-EHH) statistic (18) to make comparisons between the Tibetan highland population and the combined HapMap Chinese (CHB) and Japanese (JPT) lowland populations (19). The XP-EHH statistic assesses haplotype differences between two populations and is designed to detect alleles that have increased in frequency to the point of fixation or near-fixation in one of the populations (18, 19). A comparison with the CHB-JPT sample is appropriate because these populations have historically lived at low altitude and exhibit a small overall genetic distance to our Tibetan samples (figs. S1 and S2). We also identified partial selective sweeps (in which the adaptive variant is not yet near fixation) using the integrated haplotype score (iHS), a statistic based on the extent of decay of linkage disequilibrium surrounding a variant subjected to natural selection (20). The XP-EHH and iHS tests both have substantial statistical power to detect natural selection using genome-wide SNP genotypes, even when sample sizes are limited (19, 21).

Many selective events are likely to be shared among multiple populations, but we wish to focus only on those that are specific to high-altitude Tibetan populations. To do this, we divided the genome into consecutive, nonoverlapping 200-kb regions and excluded those that yielded significant values (P < 0.01) for the iHS test in neighboring Asian populations (see SOM). Seven of the 247 functional candidate loci were located in these excluded regions and were therefore eliminated from subsequent analyses. In our Tibetan sample, we then calculated an XP-EHH and iHS summary statistic (see SOM) for each genomic region, using an empirical significance level of 0.01 to identify a set of regions that exhibit evidence of local positive selection.

This set of regions contained 10 of the 240 genes from our functional candidate list (Fig. 1 and Table 2). Six of these genes were identified by the XP-EHH test, and five were identified by the iHS test. EGLN1 was identified by both, and a previous study suggests that this gene may also be under selection in Andeans (22). EGLN1and EPAS1 are both in the HIF pathway (23) and show elevated FST values (fig. S4). Three additional loci, EDNRA, PTEN, and PPARA, are also associated with HIF activity (Fig. 2) (24, 25). PPARA and ANGPTL4 are in the peroxisome proliferator–activated receptor (PPAR) lipid metabolism pathway, and CYP17A1 and CYP2E1 are cytochrome P450 genes (23). The two remaining genes are HMOX2, involved in HIF-independent oxygen sensing (26) and hypoxic endothelial cell survival (27), and CAMK2D, which mediates NO production in response to changes in intracellular calcium (28). Because the selection signals correspond to 200-kb sections of the genome rather than to individual genes, we performed additional analyses to localize the selection signal within each of the 10 regions (29). For each gene of interest, we observed multiple localization signals near the gene or bracketing the gene (fig. S4).

Table 2

List of putatively advantageous genes identified by the intersection of functional and selection candidate gene lists. An en dash (–) indicates nonsignificant P values (see tables S11 and S12 for all values). All gene descriptions are based on RefSeq (23) unless otherwise noted.

View this table:
Fig. 2

Selection candidates involved in the HIF pathway. Genes identified as selection candidates that are related to the HIF pathway (outlined in black boxes) are illustrated below. [Gene descriptions and regulation during hypoxic/normoxic conditions are provided in table S2. Note that HIF1AN was not included on the a priori functional candidate list but was identified by both XP-EHH and iHS analyses (see SOM, tables S2, S11, and S12).] Genes indicated in the gray boxes are provided for reference.

These results suggest that high-altitude adaptation in Tibetans has resulted from local positive selection on several distinct genes. However, even in the absence of selection, some loci will appear in the intersection of our functional candidate list and selection screens just by chance. To test the statistical significance of the observed pattern, we carried out a randomization test by resampling sets of 240 loci 1 million times from a list of all known autosomal genes (30). For each set of randomly chosen loci, we tabulated the number of loci that intersected the genomic regions identified by our selection scans. On average, the XP-EHH and iHS intersections contained 2.7 and 1.4 genes, respectively, significantly fewer than the six (P < 0.05) and five (P < 0.01) genes actually observed.

In contrast to the fixed or nearly fixed alleles detected by XP-EHH, the iHS test identifies regions affected by incomplete selective sweeps. Therefore, inter-individual genetic variation should be observable at these loci, allowing us to test for their association with Hb concentration, which is highly heritable and has distinctively low levels in high-altitude Tibetan populations (4). In each of the five 200-kb regions identified as targets of selection by iHS, we defined the selected core haplotype as the one containing the three SNP alleles that exhibit the most extreme iHS value (see SOM, table S6). We then used stepwise linear regression to test for correlations between Hb concentration and these five core haplotypes. The five independent predictor variables are the numbers of putatively advantageous haplotypes (0, 1, or 2) present in each 200-kb region. Because sex and age affect Hb concentration in Tibetans and Han Chinese, and the age effect differs between males and females (9), covariates are included in the regression analysis to allow an independent age effect in males and females (see SOM, tables S7 to S9).

The putatively advantageous haplotypes of EGLN1 and PPARA both show significant negative correlations with Hb concentration (P < 0.002 and 0.0009, respectively) (tables S8 and S9). A genome-wide regression analysis showed no excess of associations (398,020 SNPs with minor allele frequency ≥ 0.15 and no missing genotypes) (see SOM, figs. S1, S2, and S5). The phenotypic effects are substantial: Each additional copy of an advantageous haplotype at either locus decreases Hb concentration by ~1.7 g/dl on average (Fig. 3). This effect is greater than the well-established sex-related difference in high-altitude Tibetans (~1.1 g/dl) (9). The strong and significant association between Hb concentration and haplotype variation at EGLN1 and PPARA provides evidence of a genetic contribution to a form of high-altitude adaptation that appears to be unique to Tibetan populations.

Fig. 3

Genotype-phenotype association of the inferred adaptive EGLN1 and PPARA haplotypes with Hb concentration. Hb concentrations are plotted against the number of putatively advantageous haplotypes at (A) EGLN1 and (B) PPARA (from 0 to 2 haplotype copies) for 30 Tibetan individuals. A box-and-whisker plot is overlaid on the data points to show the median (red line), upper and lower quartiles (box ends), and most extreme measurements (whiskers). (C) Because the effects at the two loci are similar in magnitude, direction, and significance, they are combined here for purposes of illustration. Hb concentrations are plotted against the combined number of EGLN1 and PPARA haplotypes carried by an individual for both loci (from 0 to 4 copies) (table S9).

A potential explanation for the correlation between EGLN1 and decreased Hb concentration lies in the regulation of HIF and its target genes. EGLN1 targets two HIFα proteins for degradation under normoxic conditions, decreasing the transcription of HIF-regulated targets such as EPO, the erythropoietin gene whose product induces red blood cell (RBC) production (Fig. 2). Furthermore, mutations in EGLN1 prevent targeted degradation of HIF, leading to polycythemia (excessive production of RBCs) in mice and humans (31).

Although PPARA has not previously been considered as a candidate gene for high-altitude adaptation, it interacts with components of the HIF pathway. PPARA expression is inhibited by HIF1 during hypoxia in mice (Fig. 2) (25), and genes targeted by HIF are regulated by a HIF-independent mechanism involving PPARG coactivator-1α (32). In addition, a PPARA agonist, the antidiabetic agent tesaglitazar, resulted in decreased Hb levels during human clinical trials (33). This effect is consistent with the correlation between the putatively advantageous PPARA haplotype and Hb concentration found here.

It is plausible that the diminished Hb levels found in Tibetans offset complications associated with sustained high Hb levels (for instance, hyperviscosity) seen in non-Tibetans exposed to high-altitude conditions (10, 11). Alternatively, decreased Hb levels could be a side effect of other phenotypes that are the actual targets of natural selection. Functional analysis of genes such as EGLN1 and PPARA, which have undergone positive selection and are associated with Hb levels in this Tibetan sample, will further increase our understanding of genetic adaptation to high-altitude environments. In addition, we will better understand the human response to hypoxia, which has important implications for the prevention and treatment of mountain sickness, high-altitude pulmonary and cerebral edema, and other hypoxia-related diseases.

Supporting Online Material

www.sciencemag.org/cgi/content/full/science.1189406/DC1

Materials and Methods

Figs. S1 to S6

Tables S1 to S12

References

  • * The Research Center for High-Altitude Medicine initiated the research project and was primarily responsible for phenotyping and DNA collection.

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We thank all participants of this study. We are grateful to A. R. Rogers, J. Seger, D. J. Grunwald, and W. S. Watkins, who provided helpful comments for the manuscript, and S. R. Woodward and J. E. Gomez (from the Sorenson Molecular Genealogy Foundation) for providing samples used for comparative analyses. This work was supported by the National Basic Research Program of China number 2006CB504100, the National Natural Science Foundation of China number 30393133, and the University of Utah Research Foundation Seed Grant number 51003402. Funding was also provided by GM059290 (NIH) to L.B.J., HL50077 (NIH Heart, Lung, and Blood Institute) and 1P01CA108671-O1A2 (National Cancer Institute) Merit Review Award to J.T.P., and DK069513 (The Primary Children’s Medical Center Foundation National Institute of Diabetes and Digestive and Kidney Diseases) and The University of Luxembourg–Institute for Systems Biology Program to C.D.H. T.S.S. was supported by NIH Genetics Training Grant T32. All studies have been performed with informed consent approved by the Institutional Board of Qinghai Medical College of Qinghai University in Xining, Qinghai Province, People’s Republic of China. All SNP genoptypes are deposited in Gene Expression Omnibus, with accession code GSE21661. These data, as well as phenotype data, are also available on our laboratory Web site, http://jorde-lab.genetics.utah.edu. Please contact R.L.G. for access to DNA samples.
View Abstract

Navigate This Article