Quechua in the Andes may be genetically adapted to altitude and able to resist decrements in maximal O2 consumption in hypoxia (ΔV̇o2 max). This hypothesis was tested via repeated measures of V̇o2 max (sea level vs. 4,338 m) in 30 men of mixed Spanish and Quechua origins. Individual genetic admixture level (%Spanish ancestry) was estimated by using ancestry-informative DNA markers. Genetic admixture explained a significant proportion of the variability in ΔV̇o2 max after control for covariate effects, including sea level V̇o2 max and the decrement in arterial O2 saturation measured at V̇o2 max (ΔSpO2 max) (R2 for admixture and covariate effects ∼0.80). The genetic effect reflected a main effect of admixture on ΔV̇o2 max (P = 0.041) and an interaction between admixture and ΔSpO2 max (P = 0.018). Admixture predicted ΔV̇o2 max only in subjects with a large ΔSpO2 max (P = 0.031). In such subjects, ΔV̇o2 max was 12–18% larger in a subgroup of subjects with high vs. low Spanish ancestry, with least squares mean values (±SE) of 739 ± 71 vs. 606 ± 68 ml/min, respectively. A trend for interaction (P = 0.095) was also noted between admixture and the decrease in ventilatory threshold at 4,338 m. As previously, admixture predicted ΔV̇o2 max only in subjects with a large decrease in ventilatory threshold. These findings suggest that the genetic effect on ΔV̇o2 max depends on a subject's aerobic fitness. Genetic effects may be more important (or easier to detect) in athletic subjects who are more likely to show gas-exchange impairment during exercise. The results of this study are consistent with the evolutionary hypothesis and point to a better gas-exchange system in Quechua.
- deoxyribonucleic acid
- genetic markers
- aerobic performance
native south american Quechua, and a few other native Andean ethnic groups, are thought to be descendant from groups who first reached the highland Andes ∼10,000 years ago (31). Given the extremes of altitude that are now permanently inhabited in the region [up to 5,200 m (50)], it seems reasonable to hypothesize that these populations derive in part from groups who experienced natural selection in the past favoring superior oxygen transport phenotypes. If so, one functional consequence of genetic adaptation might be an ability (in current populations) to limit normal impairments in oxygen uptake and/or utilization that occurs during strenuous exercise in hypoxia. Indeed, many previous studies have made this argument, pointing to the relatively high maximal oxygen consumption (V̇o2; V̇o2 max in ml · min-1 · kg-1) in hypoxia of Andean study groups (3, 11, 14, 29, 34, 35), or suggesting that such groups, compared with lowland groups, experience only a small decrement in sea level V̇o2 max [change (Δ) in V̇o2 max] when exposed to hypoxia (2, 11, 14, 23, 45, 47, 49). Regarding the latter, only a few studies have directly measured the ΔV̇o2 max in such groups using a repeated-measures design (2, 23, 47), and these report V̇o2 max decrements in Andean natives that are between ∼30 and 80% of the decrement seen in lowland comparison groups. In this regard, the smaller V̇o2 max decrement in Quechua may reflect an integrated functional response to hypobaric hypoxia that involves multiple physiological and biochemical systems (22). However, the reported differences in the magnitude of the V̇o2 max decrement may be exaggerated. In two of the studies cited above, the lowland comparison groups were trained athletes, and aerobic fitness has a well-known positive effect on the V̇o2 max decrement (12, 17, 28, 30, 32, 41, 42, 44).
In the present study, we assessed V̇o2 max decrement via a repeated-measures design in a large group (n = 30) of young Peruvian men of mixed Quechua and Spanish ancestry who were born and raised in Lima, Peru (sea level). These subjects were first measured in Lima and then were transported to Cerro de Pasco, Peru (4,338 m) for measurement after ∼12 h of exposure to hypobaric hypoxia. The main objective of our study was to assess the influence of Quechua vs. Spanish genetic admixture on the V̇o2 max decrement within this study group, controlling for variation in aerobic fitness that may impact the magnitude of decrement experienced. Our focus on individual genetic admixture level as a study-independent variable represents a new research strategy to detect the effects of genetic adaptation on physiological phenotypes. The approach is made possible by new molecular genetic techniques that give admixture estimates for individuals descendant from two or more parental populations. Admixture in the Andes is historically well documented and began ∼500 yr ago with the contact of Andean Native American, Spanish, and, to a lesser extent, West African groups. Thus the approach represents an alternative to the more typical study comparing an Andean native group with a lowland “control” group. Such studies ignore the reality of Andean population history and may be confounded by many factors, including unknown admixture levels in the so-called native group.
To estimate the extent of Spanish ancestry (admixture) for each individual within the sample, we used a panel of 22 ancestry-informative DNA genetic markers. This revealed an admixture range within the sample from <1 to 64% Spanish ancestry. From this, we hypothesized a positive relationship between the extent of Spanish ancestry and the magnitude of the V̇o2 max decrement, as would be expected if natural selection had favored hypoxia-tolerant phenotypes in past Quechua populations. Importantly, for the approach to be useful, the individual genetic markers need not be associated (i.e., linked) with whatever genes determine exercise capacity in Quechua. The markers are simply used to produce a probability estimate of the proportionate ancestry of an individual. As a construct that reflects ancestry (and not specific genes or markers), the admixture estimate may be associated with a specific physiological phenotype, even if the individual genetic markers are not. In this regard, the approach may be seen as a first step to identify physiological phenotypes for further genetic study.
MATERIALS AND METHODS
Subject Selection Criteria and Study Populations
The subjects for this study were young men (18–35 yr), who gave written, informed consent according to guidelines approved by the Institutional Review Boards at the University at Albany, State University of New York, and the Universidad Cayetano Heredia, Lima, Peru. Subjects were identified as nonsmokers and screened via a brief clinical history and medical examination for conditions contraindicating participation in the study protocols, including chronic obstructive respiratory diseases, cardiovascular disease, and renal disease. At screening, a venous blood sample was drawn from the antecubital vein, and Hb concentration (g/dl) was immediately determined by a Hemocue blood Hb analyzer (Angelholm, Sweden). Subjects with Hb <13.4 g/dl were considered anemic and were excluded from the study.
The subjects were recruited from within a specific district of Lima, Peru (Barrios Altos district, population ∼150,000). In this district, ∼10% of inhabitants are recent down-migrants from highland Peru. Individuals accepted into the study were born and raised in Lima or near sea level, and both sets of their parents and grandparents were born at an altitude >3,000 m. Thus all study subjects were first- or second-generation down-migrants. No specific attempt was made to recruit subjects based on surname composition or skin reflectance measures, both of which have been used in the past to assess ancestry. The majority of subjects described themselves as “Peruvians,” but acknowledged both their Quechua and Spanish origins.
Subjects were recruited into the study during the last 2 wk of July 2001 in Lima, Peru. Each subject completed baseline studies in Lima requiring ∼4 h of participation. These subjects were studied again within 2 wk at 4,338 m in the town of Cerro de Pasco, Peru. Cerro de Pasco is a 6- to 10-h bus ride from Lima on paved road. The first ∼4–6 h of the trip involve a steady gain in altitude to a high mountain pass (∼4,800 m). The road then descends to the Peruvian Altiplano (3,600–4,300 m) for the next 3–4 h of the trip. Subjects arrived in Cerro de Pasco from Lima in groups of four to five per day over a 2-wk period and rested in the laboratory for 2–4 h before studies were initiated. Thus these subjects were studied after 10–12 h of acute exposure to hypobaric hypoxia. Of 32 subjects measured in Lima, two were not measured in Cerro de Pasco. One was unable to make the trip for personal reasons, and the other was diagnosed with acute mountain sickness on arrival to Cerro de Pasco.
Anthropometry and Pulmonary Function
Standard anthropometry was performed on each subject by the same investigator. Measurements included height, weight, and skinfolds at subscapular, suprailiac, biceps, and triceps sites. Body density was calculated according to age and sex-specific equations given by Durnin and Womersley (10). Kashiwazaki et al. (26) have tested the validity of a number of reference equations against doubly labeled water measures of body composition in Bolivian Aymara and concluded that the reference equations above are the best available for use in Andean native populations. Percent body fat was calculated from the Siri equation. Pulmonary function was assessed on each subject in Cerro de Pasco by using a VS400 volumetric spirometer (Puritan-Bennett, Mallinckrodt, Hazelwood, MO), calibrated daily with a 3-liter calibration syringe. Each subject performed a maximal inspiration, followed immediately by a forced maximal expiration while in a standing position. From this procedure, the forced vital capacity (FVC) and forced expiratory volume made in 1 s were determined based on the best of at least two efforts. FVC and forced expiratory volume in 1 s measures were corrected for btps.
Genetic markers. Individual Spanish admixture proportion was estimated by using a panel of 22 informative genetic markers (MID-575, TSC1102055, WI-11153, MID-52, SGC30610, WI-17163, WI-9231, WI-4019, WI-11909, D11S429, TYR-192, DRD2 TaqD, DRD2 BclI, WI-14319, CYP19, PV92, WI-7423, CKM, MID-161, MID-93, FY, and F13B). The first 20 markers were selected because they show high-frequency differences (>30%) between Native American and Spanish populations, which are a priori the main parental populations in this sample. We also included in the panel two markers (FY and F13B) providing information on African ancestry. Details of these markers, including allele frequencies in all parental populations, DNA sequences, exact positions of single-nucleotide polymorphisms (SNPs), and the PCR primers and amplification conditions used are available from the dbSNP database (www.ncbi.nlm.nih.gov/SNP) under the submitter handle PSU-ANTH.
Genotyping. SNPs and insertion and deletion markers were scored by a melting-curve assay, in which the target sequence containing the SNP is amplified by PCR by using a mismatched primer where necessary to create an artificial restriction site polymorphism. PCR products were digested with a restriction enzyme, and the resulting restriction fragment-length polymorphisms are scored by their melting curves in a Hybaid DASH machine (ThermoHybaid). More details about this genotyping method can be found in Akey et al. (1). PCR products of the PV92 Alu insertion polymorphism were scored by conventional 2% agarose gel electrophoresis.
Admixture estimation. Usually, in samples from admixed populations, admixture is estimated by using the frequencies of samples representing the contributing parental populations as a contrasting reference. In this case, although there is information on the allele frequencies for these markers in several European populations, no such information is available for the Quechua parental population. Thus, to estimate admixture, we have used a strategy in which information on all parental frequencies is not required. We have used the program STRUCTURE, developed by Jonathan Pritchard, to infer admixture proportions in the samples from Lima and Cerro de Pasco. This program has been designed to infer the presence of genetic structure and to estimate the admixture proportions in samples with unknown genetic structure. This program can be used to estimate the number of subpopulations present in a sample and to assign individuals to each of those subpopulations, including estimates of individual admixture from each subpopulation. To estimate admixture in our sample, we prepared an input file, which included the genotype data for the 32 samples from Lima, 39 samples of highland-born subjects from Cerro de Pasco (not reported here), and 72 additional samples from Spain. We then ran the STRUCTURE program with K = 2 as the predefined setting for the number of populations, using 30,000 iterations for the burn-in period and 70,000 additional iterations to obtain parameter estimates. The output file provides an estimate of the European and Native American ancestry for each individual in the sample. We ran the program several times, with consistent results. More information about the program STRUCTURE can be found in Pritchard et al. (38). Using STRUCTURE in this manner has been shown to result in individual admixture estimates that are highly correlated with estimates made by using maximum likelihood and parental allele frequencies from both parental populations.
The estimates of individual admixture range from 0 (Quechua) to 1 (Spanish) and reflect the proportionate contribution of two different population histories to the genetic makeup of an individual.
Identical protocols to measure V̇o2 max (l/min) were administered in Lima and Cerro de Pasco. To begin, V̇o2 was measured at rest (5 min) with the subject seated. After resting measurements, V̇o2 max was measured on a mechanically braked Monarch 818e research ergometer. Subjects started with a workload of 1.0-kg resistance at 60 rpm, and resistance was incremented by 0.5 kg at constant rpm every 3 min until subject volitional fatigue. Subjects were given verbal encouragement, and V̇o2 max was defined as the highest level of V̇o2 averaged over the final minute of the test, concomitant with at least one of the following: a nonlinear increase in exercise ventilation, resulting in a respiratory exchange ratio (RER) >1.10, a plateau in the V̇o2-work rate relationship, or a maximal heart rate (HR) within 10% of the age-predicted maximum.
During V̇o2 testing, subjects breathed through a low-resistance breathing valve, and expired ventilation (V̇e, l/min btps), as well as the fractional concentrations of O2 and CO2 in expired air, was processed by a Parvo-medics True Max metabolic measuring system (Sandy, UT) to produce 1-min-interval calculations of V̇o2, carbon dioxide production (V̇co2), the RER, and the ventilatory equivalents for oxygen and carbon dioxide (V̇e/V̇o2 and V̇e/V̇co2, respectively). Gas analyzers were calibrated with standard gases before each exercise test. The pneumotach used to measure ventilatory flow was also calibrated before each test with a 3-liter calibration syringe. HR was continuously monitored via telemetry (Polar Electric Oy, Sweden) interfaced with the metabolic measuring system. Arterial oxygen saturation by pulse oximetry (SpO2) was continuously monitored by an Ohmeda 5740 pulse oximeter by using a finger-tip sensor (subjects were instructed not to grip with that finger). The pulse oximetry signal was acquired by an REM/400M data-acquisition system (CB Sciences) and recorded every 15 s during V̇o2 measurements. Ventilatory threshold (V̇ethresh) was determined graphically from the V̇e, V̇e/V̇o2, and V̇e/V̇co2, according to the method of Caiozzo et al. (7).
All variables were evaluated for normality by using the Kolmogorov-Smirnov test against a standard normal distribution by using the Lilliefors two-tail probability. The admixture variable was transformed by the natural logarithm [log(e) admixture] to achieve normality for statistical testing purposes. Differences between exercise response variables in Lima and Cerro de Pasco were evaluated by paired Student's t-test, whereas differences between subgroups established within the data set were evaluated by t-test for independent samples. The general linear model procedure from SYTAT version 5.1a (Macintosh) or version 9.0 (personal computer) was used to construct univariate and multivariate models predicting the ΔV̇o2 max (decrement) from sea level to 4,338 m. Such analyses are potentially problematic because change (Δ) depends on the initial or baseline value. Baseline, in this regard, is the value achieved at sea level for a given measure. To address this issue, we have adopted the statistical approach recommended by a number of authors (21, 25). That is, when analyzing a Δ-dependent variable, the baseline value must be entered as a covariate, despite the obvious mathematical relationship between dependent and independent variables. Similarly, a baseline value must be entered as a covariate (control) when analyzing a Δ-independent variable, despite the obvious mathematical collinearity between the baseline and Δ-independent variables. This procedure results in a nonbiased removal of baseline effects. In the present context, it allowed the evaluation of the effect of admixture on V̇o2 max decrement, independent of the level of V̇o2 max achieved at sea level.
Our analytic strategy to detect genetic variance in ΔV̇o2 max was as follows. First, multivariate models were constructed that controlled for baseline and other important covariate effects on ΔV̇o2 max. Then, genetic main effects were evaluated by the addition of the log(e) admixture variable. Last, interaction effects between log(e) admixture and relevant covariates were evaluated.
Values are expressed as means ± SD, unless otherwise indicated. Statistical significance criteria was P ≤ 0.05 for all tests.
Subject characteristics are given in Table 1. Compared with US National Health and Nutrition Examination Surveys reference standards, these subjects fell below the 5th percentile for stature by age, and near the 15th and 50th percentiles for weight by age and stature by weight, respectively. In this respect, they are typical of both highland- and lowland-born healthy Andean native men. Compared with a highland-born control population from Cerro de Pasco, they were slightly taller and fatter and had 15% smaller FVC, as expected (comparative data not shown). The average admixture rate was ∼10% Spanish ancestry and ranged from <1 to 64%.
Maximal exercise response data are given in Table 2. These subjects were well motivated to perform the exercise test, and all achieved the criteria for a true V̇o2 max during exercise at sea level. Even in Cerro de Pasco, where exercise was subjectively unpleasant, most subjects showed nonlinear increases in V̇e near the end of the V̇o2 max test, resulting in a high RER, and a majority (21 out of 30 subjects) had maximal HR levels within 10% of their age-predicted maximum.
The V̇o2 max decreased from sea level to 4,338 m by ∼19% (P < 0.001). Maximal HR was significantly lower at 4,338 m compared with sea level. V̇e btps was unchanged, but the V̇e btps/V̇o2 was significantly higher at 4,338 m. The RER was unchanged from sea level to 4,338 m. The mean V̇ethresh, expressed in liters per minute of V̇o2, was significantly lower at 4,338 m compared with sea level, i.e., 2.27 vs. 1.64 l/min (P < 0.01). V̇ethresh was also significantly lower at altitude when expressed as a percentage of V̇o2 max, i.e., 74.1 vs. 66.1% of V̇o2 max at sea level and 4,338 m, respectively.
Table 3 gives the matrix of correlation coefficients between variables that were used in subsequent multivariate analyses to model the effect of genetic admixture on ΔV̇o2 max. Variables that were not significantly related to ΔV̇o2 max, and thus not included in multivariate analyses, included the fat-free mass and Hb concentration. Sea level V̇o2 max, sea level V̇ethresh, ΔSpO2 measured at V̇o2 max (ΔSpO2 max), and ΔV̇ethresh were all strongly correlated to ΔV̇o2 max, i.e., correlation coefficients between 0.59 and 0.81. Figures 1 and 2 show the relationship between ΔV̇o2 max and sea level V̇o2 max and ΔSpO2 max, respectively, as both relationships are central to the multivariate analyses that follow. Figure 1 reveals minimal V̇o2 max decrement for individuals with <2.5 l/min sea level V̇o2 max, but substantially larger decrements (∼25%) for individuals approaching 4.0 l/min sea level V̇o2 max. Similarly, Fig. 2 reveals only minimal V̇o2 max decrements for individuals with <10% decrease in SpO2 max and larger V̇o2 max decrements for individuals decreasing SpO2 max by >20%.
Table 4 gives results of multivariate analyses modeling the decrease in V̇o2 max from sea level to 4,338 m. Model 1 (Table 4) is the simplest model and again demonstrates that the majority of the variability in ΔV̇o2 max is explained by sea level V̇o2 max (β = 0.516, R2 = 0.669, P < 0.001). Model building continued with the addition of the ΔSpO2 max variable, controlling for the sea level value of SpO2 max (model 2, Table 4). This model reveals a significant independent association of ΔSpO2 max on ΔV̇o2 max (β = 0.022, P = 0.033), increasing the R2 over model 1 by 0.058. The addition of log(e) admixture (model 3, Table 4) reveals a significant main effect of log(e) admixture (β = -0.205, P = 0.041) and a significant interaction of log(e) admixture with ΔSpO2 max (β = 0.014, P = 0.018).
The nature of this interaction is best understood by viewing plots that show the ΔV̇o2 max as a function of admixture and ΔSpO2 max in various study subgroups (Figs. 3 and 4). Subgroups were created by splitting the data sample above and below the median of admixture and ΔSpO2 max. The ΔV̇o2 max residuals were used as the dependent variable in these plots from the regression of ΔV̇o2 max on both sea level V̇o2 max and sea level SpO2 max. This allows a nonbiased assessment of the individual deviation from the overall group mean ΔV̇o2 max, adjusting for baseline (i.e., sea level) effects. Thus positive and negative values specify individuals with larger and smaller than average decrements in V̇o2 max, respectively, independent of the sea level V̇o2 max and the baseline SpO2 max. In fact, control for baseline SpO2 max (sea level SpO2 max) makes little difference in this regard, because sea level saturation was uniformly high and showed little variation relative to the SpO2 max observed in Cerro de Pasco. Nevertheless, as described in materials and methods, control for baseline is the correct approach to view the effect of admixture or ΔSpO2 max on ΔV̇o2 max.
Figure 3 shows the relationship between the V̇o2 max residuals and log(e) admixture in subgroups with the smallest (Fig. 3A) and largest (Fig. 3B) decreases in SpO2 max. Log(e) admixture was not related to ΔV̇o2 max in the subgroup with the smallest ΔSpO2 max (mean decrease 14.7 percentage points), but significantly correlated to ΔV̇o2 max in the subgroup with the largest ΔSpO2 max (mean decrease 20.8 percentage points) (R = 0.557, P = 0.031). Whereas subgroup analysis allows the visualization of this interaction effect, the transformation of a continuous variable (ΔSpO2 max) into a dichotomous variable results in a loss of information. Thus the reader should refer back to model 3 (Table 2) when gauging the true strength of association between admixture and ΔV̇o2 max.
Alternately, the interaction between admixture and ΔSpO2 max may be seen in Fig. 4, which shows the relationship between the ΔV̇o2 max residuals and ΔSpO2 max in subgroups with the lowest (Fig. 4A) and highest (Fig. 4B) levels of genetic admixture. ΔSpO2 max was not related to ΔV̇o2 max in the subgroup with the lowest genetic admixture (mean 1.1% European genetic influence), but was highly significantly correlated to ΔV̇o2 max in the high-admixture subgroup (mean 18.4% European genetic influence) (R = 0.652, P = 0.008).
To address the potential for confounding or spurious correlation, it is important to establish the general similarity between subgroups used in the analyses above. Comparative data in Table 5 show no significant differences for potential confounding variables between subgroups, with the exception of a significantly higher relative V̇o2 max (ml · min-1 · kg-1) at sea level in the low- vs. high-admixture subgroups. However, this difference does not explain (as a positive confounder) the interaction described above. That is, model 3 (Table 4) explicitly controls for sea level V̇o2 max when testing for the main and interaction effects of admixture on the V̇o2 max decrement. A model substituting relative V̇o2 max for absolute V̇o2 max yields the same qualitative result.
Models 4 and 5 (Table 4) test for the effects of ΔV̇ethresh, log(e) admixture, and the interaction of these two variables on ΔV̇o2 max. Model 4 reveals a significant positive association between ΔV̇o2 max and ΔV̇ethresh, controlling for sea level values of both (β = 0.297, P = 0.006). However, model 5 (Table 4) shows no main effect of log(e) admixture on ΔV̇o2 max (P = 0.319) and only a trend for interaction between log(e) admixture and ΔV̇ethresh (P = 0.095). It is worth noting that the interpretation of this interaction, if significant, would be the same as that described above between admixture and ΔSpO2 max. That is, the trend suggests an effect of admixture that is evident only in individuals with a large decrease in V̇ethresh. Similarly, the positive association between ΔV̇o2 max and ΔV̇ethresh is only evident in individuals with high Spanish admixture, but not evident in individuals with low Spanish admixture.
Additional multivariate models were run, controlling for ΔSpO2 max and ΔV̇ethresh simultaneously. These models do not change the basic results presented in models 3 and 5 (Table 4). That is, the interaction between log(e) admixture and ΔSpO2 max remained significant after control for ΔV̇ethresh (β = 0.011, P = 0.037). Similarly, the interaction between log(e) admixture and ΔV̇ethresh remained nonsignificant after control for ΔSpO2 max, although the P value once again reached a relatively low level (β = 0.086, P = 0.114). A final model testing for the three-way interaction among log(e) admixture, ΔV̇ethresh, and ΔSpO2 max revealed a marginally significant interaction in this regard (β = 0.004, P = 0.06). Interpretation of this interaction is similar to the interpretation given above for two-way interaction. That is, the admixture effect is evident in individuals with large decrements in SpO2 max and V̇ethresh, but absent in individuals showing only small decreases in SpO2 max and V̇ethresh from sea level to 4,338 m.
This study shows an association between Spanish admixture level and the V̇o2 max decrement in a subset of subjects showing larger than average altitude-related decreases in SpO2 max. In these subjects, high Spanish ancestry was associated with larger ΔV̇o2 max, consistent with the general hypothesis that Quechua natives of the highland Andes are adapted to high altitude based on population-specific genetic factors that have arisen as a consequence of natural selection. The strength of this evolutionary inference depends on the novel research design and on the choice of ΔV̇o2 max as the study-dependent variable. A repeated-measures design, which exploits intragroup variability in genetic admixture, has clear advantages over the traditional comparative approach i.e., Andean native vs. lowland “control.” Comparisons between groups may be confounded by many factors, including unknown levels of admixture in the study populations (4). Regarding the ΔV̇o2 max, this phenotype has long been considered an important marker of hypoxia sensitivity or tolerance (2, 23, 47).
The study approach depends on the well-documented (and different) population history of exposure to hypobaric hypoxia between Quechua and Spanish. In essence, the admixture estimate is a construct describing the proportionate contribution of alternate population histories to the genetic makeup of an individual. This is an important point because the association between individual admixture and ΔV̇o2 max is to be interpreted relative to population history, not relative to the specific genetic markers that were used to derive the estimates. In other words, for the approach to work, it is not necessary that there be a direct linkage between the genetic markers and the ΔV̇o2 max phenotype, especially as the markers used represent only a small fraction of the total genome. Our laboratory has previously reported that admixture can show a significant correlation with specific phenotypes, even if the markers informative for admixture are not physically linked to the trait in question (36, 43). The process of admixture does produce allelic associations between unlinked and linked loci, as a function of the level of frequency differences between the parental populations and admixture rate (8). However, it is important to note that, several generations after the admixture event, the association between linked markers will be much higher than the background association between unlinked markers. This could be exploited to map the genetic factors that contribute to hypoxia tolerance in the Quechua population. Presently, the genes involved are unknown.
The Association Between Admixture and ΔV̇O2 max
By itself, genetic admixture level was not related to variability in ΔV̇o2 max. However, after control for both sea level V̇o2 max and ΔSpO2 max, strong associations between genetic admixture and ΔV̇o2 max were revealed. Covariate control is an important step because, in this and previous studies (12, 17, 28, 30, 32, 41, 42, 44), the majority of the variability in the ΔV̇o2 max was explained by the sea level V̇o2 max. This effect is impressive as aerobically fit subjects typically lose two to four times more of their sea level V̇o2 max and show decrements in aerobic performance at even modest altitudes, i.e., <600 m (17). Similarly, this and other studies (12, 32, 41, 42, 44) demonstrate that much of the variability in the ΔV̇o2 max is explained by the ΔSpO2 max. Sea level V̇o2 max and ΔSpO2 max together explained nearly 73% of the variability in the ΔV̇o2 max within our study sample. Of the remaining variability, another 7% was explained by genetic factors assessed by the admixture variable (P = 0.041) and an admixture-by-ΔSpO2 max interaction effect (P = 0.018, model 3, Table 4). Thus the effect of admixture on ΔV̇o2 max depends on the magnitude of ΔSpO2 max. Admixture effects were large between subjects with large ΔSpO2 max (Fig. 3B), but not detectable between individuals who showed only modest decreases in SpO2 max (Fig. 3A). In the subset of individuals showing larger than average ΔSpO2 max, adjusted mean values for ΔV̇o2 max (ml/min) were 18% larger in the highest vs. lowest subgroups of Spanish ancestry, i.e., least squares mean values, adjusted for sea level V̇o2 max, were ∼739 and ∼606 ml/min, respectively.
One interpretation of this finding is that genetic effects are more important (i.e., easier to detect) across the range of admixture when subjects are aerobically fit. This makes sense because aerobically fit individuals are, a priori, more hypoxia sensitive, given their larger ΔV̇o2 max (see references cited above). In addition, athletes vs. nonathletes show a wider alveolar-arterial partial pressure difference during exercise and show impairments in gas exchange, even in normoxia (9, 40). These differences suggest a pulmonary limitation to exercise in athletes via ventilation-perfusion mismatch and/or diffusion limitation, both of which increase as problems with increasing exercise intensity and/or hypoxia (15, 19). Thus better pulmonary function in Quechua may be the basis for part of the admixture effect described, particularly as Andean natives are characterized by large lungs (6, 13, 18, 24) and high-pulmonary diffusion capacities (39, 46). One previous study is consistent with this hypothesis, showing higher arterial oxygen saturation during exercise in Andean natives compared with a highlandborn control group of lowland ancestry (5).
This interpretation is further supported by the V̇ethresh data. In this and previous studies (28, 41), ΔV̇ethresh explained a large proportion of the variability in ΔV̇o2 max. We also noted a trend for interaction (admixture-by-ΔV̇ethresh) that paralleled the admixture-by-ΔSpO2 max interaction (P = 0.095). That is, the admixture effect tended to be evident only in individuals who showed a large decrease in V̇ethresh. This may be a reflection of pulmonary limitation and a decrease in arterial O2 saturation in the most fit subjects, as experimental studies demonstrate an increase in blood lactate concentration when arterial saturation is lowered (27). In fact, ΔV̇ethresh and ΔSpO2 max were strongly correlated (R = 0.522, P < 0.01, Table 3).
Alternately, the relationship between ΔV̇o2 max and ΔSpO2 max may also be seen to depend on the genetic admixture level (Fig. 4). Theoretically, ΔV̇o2 max and ΔSpO2 max should show a positive linear relationship (as in Fig. 4B) because of the shape of the Hb-O2 equilibrium curve (OEC). This has been demonstrated by Ferretti et al. (12), who show nonlinear relationships between the ΔV̇o2 max and the fractional concentration of inspired oxygen from 0.1 (hypoxia) to 0.3 (hyperoxia) and also between the arterial saturation measured at V̇o2 max and fractional concentration of inspired oxygen. The nonlinearity in both relationships mirrors the OEC, and, consequently, a plot of ΔV̇o2 max vs. arterial saturation (or ΔSpO2 max) yields a linear plot. Interestingly, both aerobically fit and sedentary subjects fall on the same line (12), although aerobically fit subjects have both larger ΔV̇o2 max and ΔSpO2 max. This suggests that athletes vs. nonathletes operate on the steep part of the OEC at maximal exercise and, as a consequence, have little margin to increase oxygen flow conductance when oxygen levels decrease (12, 17, 30, 44). The absence of a positive relationship between ΔV̇o2 max and ΔSpO2 max in low-admixture subjects is thus unexpected (Fig. 4A).
In Quechua, it may be that variability in ΔV̇o2 max is not dependent on the OEC or that arterial desaturation is not a major factor driving the ΔV̇o2 max. If so, then other factors independent of O2 content may be implicated, including the cardiac output, the peripheral O2 diffusion (capillary to mitochondria), or the mitochondrial oxidative capacity itself. However, this study was not designed to address these specific possibilities. Also, V̇o2 max has been described as an integrated functional response variable, not defined (or limited) by a single factor, but rather set by the interaction of multiple O2 transport conductances in the lungs, circulatory system, and skeletal muscle (48). In this sense, it might be more realistic to consider the possibility that genetic adaptation in Quechua is the result of an integrated evolutionary response involving multiple interacting components of the O2 transport chain.
Limitations of the Present Study
The estimates of individual admixture were based on 22 genetic markers. These markers were selected to estimate admixture because they differed greatly in the frequency of specific alleles between Native American and European populations. They have been validated by genotyping diverse samples of Native American populations (Mayan, Southwestern Native Americans, Native Mexican, and Aymara) and European populations (Europe, Germany, and European Americans). Although this panel of markers is sufficient to obtain a precise estimate of group mean admixture (typically, with a standard error <3%), the application of these markers does not provide as precise an estimate of admixture at the individual level. The issue of quantifying the exact error around these estimates is a complicated one, and standard errors are not constant across the range of admixture. For example, standard errors are larger for individuals with high Spanish admixture. We estimate that, to obtain a precision of <3% in the individual admixture estimates, at least 80–100 ancestry-informative markers would be required. Thus it is important to mention that the estimates of individual admixture for each person show a wide confidence interval. A partial solution to this problem is to ensure a large sample size with a wide range of admixture across the study sample, as we have done here.
One weakness of this study is that arterial oxygen saturation was measured via pulse oximetry rather than from blood sampled during exercise. Fortunately, the study partially compensates for error in the SpO2 measure via a relatively large sample size. A large sample ensures accurate group mean values, assuming that there is no bias in the measurement. While a number of studies demonstrate the general validity and accuracy of SpO2 measures at rest and during exercise vs. blood-gas measures (20, 33, 41, 51), some studies suggest bias, especially during maximal exercise (37, 52). This external bias, or the idea that the “true” saturation cannot be measured via pulse oximetry at maximal exercise, is a minor problem in the present context. A larger potential problem would be the presence of internal bias, or the idea that measurement validity depends on conditions internal to the study. For example, it would be problematic if SpO2 validity depended on altitude or fitness level. There is no indication that this is the case, and, moreover, the repeated-measures approach minimizes some of the potential problems in this regard. Nevertheless, blood-gas studies in Andean natives will be necessary to confirm the results of this study and to further explore the specific mechanisms that explain hypoxia tolerance in Quechua natives.
The results of this study are consistent with the hypothesis that Quechua are genetically adapted to hypobaric hypoxia. We have employed a novel research strategy that is based on estimating the individual genetic admixture level within a specific study group using a panel of ancestry-informative genetic markers. We demonstrate an association between the extent of Spanish ancestry and an important functional phenotype (ΔV̇o2 max) that may be considered to define hypoxia tolerance. Whereas all comparative studies have some inherent problems that limit evolutionary inference (4, 16), the admixture approach applied here holds great promise. This is particularly true because the library of ancestry-informative genetic markers is expected to grow in the future, leading to increased precision in admixture estimation.
This work was supported in part by National Science Foundation Grant BCS-0129377 (to T. D. Brutsaert) and National Human Genome Research Institute Grant HG-002154 (to M. D. Shriver).
We thank the volunteers who gave their time, sweat, and DNA for this research.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2003 the American Physiological Society