Interindividual Variation in Cardiorespiratory Fitness: A Candidate Gene Study in Han Chinese People

Cardiorespiratory fitness, as assessed through peak oxygen uptake (VO2peak), is a powerful health indicator. We aimed to evaluate the influence of several candidate causal genetic variants on VO2peak level in untrained Han Chinese people. A total of 1009 participants (566 women; age [mean ± SD] 40 ± 14 years, VO2peak 29.9 ± 7.1 mL/kg/min) performed a maximal incremental cycling test for VO2peak determination. Genomic DNA was extracted from peripheral whole blood, and genotyping analysis was performed on 125 gene variants. Using age, sex, and body mass as covariates, and setting a stringent threshold p-value of 0.0004, only one single nucleotide polymorphism (SNP), located in the gene encoding angiotensin-converting enzyme (rs4295), was associated with VO2peak (β = 0.87; p < 2.9 × 10−4). Stepwise multiple regression analysis identified a panel of three SNPs (rs4295 = 1.1%, angiotensin II receptor type 1 rs275652 = 0.6%, and myostatin rs7570532 = 0.5%) that together accounted for 2.2% (p = 0.0007) of the interindividual variance in VO2peak. Participants carrying six ‘favorable’ alleles had a higher VO2peak (32.3 ± 8.1 mL/kg/min) than those carrying only one favorable allele (24.6 ± 5.2 mL/kg/min, p < 0.0001). In summary, VO2peak at the pre-trained state is partly influenced by several polymorphic variations in candidate genes, but they represent a minor portion of the variance.


Introduction
Cardiorespiratory fitness (CRF) is positively associated with endurance exercise performance [1] and is a strong prognostic factor of morbidity and mortality from all causes and, particularly, from cardiovascular disease (CVD) [2,3]. While both physical activity (PA) and exercise training can modify CRF and are inversely associated with morbidity and mortality rates [4], CRF per se is a much stronger predictor of prognosis in CVD and metabolic disorders [5,6]. The measure of an individual's peak capacity to perform dynamic aerobic exercise is dependent on the synergistic action of pulmonary, cardiovascular and muscle tissue via a suite of physiological actions that effectively transport and Genes 2020, 11, 555 2 of 16 deliver oxygen from the atmosphere to mitochondria in working muscles [7,8]. Accordingly, CRF can be assessed by directly measuring the peak oxygen uptake (VO 2peak ) reached during a graded dynamic exercise test until exhaustion, involving large muscle masses (e.g., running or bicycling), or by indirectly estimating this variable from the peak workload achieved. Nevertheless, direct evaluation of VO 2peak is considered the gold standard measure of CRF and, indeed, the American Heart Association recently advocated for the routine assessment of this measure as a clinical vital sign [9]. VO 2peak is characterized by a high interindividual variability even in people of the same sex, age and with the same level of PA and exercise training. This variability is believed to be related, at least partly, to heredity. A seminal study by Claude Bouchard and colleagues found comparable VO 2peak values in brothers of the same sibship, and the similarities in VO 2peak were even greater in dizygotic and monozygotic twins [10]. The authors suggested that the genetic effect on VO 2peak reached~40%. In a similar study of 170 individuals and their offspring (n = 259), it was found that about 50% of the interindividual variance in VO 2peak corresponded to heritable factors after adjusting for age, sex, body mass, and body composition [11]. These findings have been replicated in subsequent studies with siblings and twins [12] and, to date, it is commonly accepted that VO 2peak is influenced by both genetic (~50-60%) and environmental factors. It has also been reported that twins with similar VO 2peak values present with comparable levels of a variety of PA indices [13], suggesting that part of the heritability of VO 2peak in twins might be due to the similarity of their PA levels. In fact, in a recent analysis of 123,545 single nucleotide polymorphisms (SNPs), only nine were associated with VO 2peak [14]. The authors of this study found that those individuals whose genotype was associated with a high VO 2peak value had a lower CVD risk (e.g., less visceral fat or lower total blood cholesterol), but they did not calculate the additive effect that the nine SNPs had on the interindividual variability of VO 2peak . There is therefore controversy on the influence of genetics on VO 2peak , which mostly likely stems from the discrepancies between studies conducted on siblings/twins vs those conducted on individuals with no familial connection. In this regard, determining the actual genetic contribution to the interindividual variability in VO 2peak would be of major importance to inform how environmental factors-including lifestyle-might contribute to heightened VO 2peak values. It is possible that if the influence of genetics on VO 2peak is low, exercise training might be a determining factor to enhance 'innate' VO 2peak even in those less genetically predisposed, with obvious subsequent benefits for cardiovascular health. Indeed, previous research has reported VO 2peak increases of up to 44% after strenuous training interventions, which would support a strong influence of environmental factors on CRF [15].
Aerobic/endurance exercise-based training appears to be the most effective way to augment VO 2peak . Exercise training increases rather than decreases the individual differences seen at baseline VO 2peak because the response to training itself shows large variation [16]. A genome-wide association study based on 324,611 SNPs found that only 21 SNPs could explain 48.6% of the change in VO 2peak induced by a 20-week exercise training program [17]. Among them, rs6552828, located in the acyl-CoA synthase long-chain member 1 (ACSL1) gene, accounted by itself for 6% of the training-induced enhancement in VO2 peak . In a recent meta-analysis of 35 articles on the genetic influence on VO 2peak trainability, a total of 97 genes were associated with this phenotype, although only 13 genetic variants were reproduced by more than two investigations [18].
Knowledge on the genetic influence on baseline VO 2peak (i.e., in isolation from training) is mainly based on studies conducted on siblings/twins or in individuals of Caucasian descent, and it remains to be determined whether the genetic variants that might be associated with baseline VO 2peak are similar or different in individuals of other ethnicities. Thus, the aim of the present study was to assess the influence of several candidate genetic variants in the interindividual variation of baseline CRF measured as VO 2peak, in Han Chinese individuals.

Participants
A total of 1047 participants (56% women) volunteered to participate in the study. The sample was recruited from five cities in China: Beijing, Xi'an, Guangzhou, Shenyang, and Tianjin. Inclusion criteria were the following: male/female aged 18-69 years; being of Chinese (Han) descent and unrelated to the other participants; having no CVD, diabetes or abnormal glucose tolerance, or any other acute or chronic disease; and being untrained (i.e., ≤ 2 sessions/week of ≤ 30min of regular physical exercise in the previous 12 months). One week before the start of the investigation, participants were fully informed of the experimental procedures and signed an informed written consent to participate in the investigation. The study protocol was approved by the Institutional Review Board of the China Institute of Sport Science.

Experimental Design
This is an observational cross-sectional study aimed at determining the genetic influence of target genes on the interindividual variability in VO 2peak values in untrained Han Chinese individuals. We selected untrained individuals to avoid any influence of exercise training or planned PA in the analysis.

Experimental Protocol
The day of the first experimental trial, participants underwent a medical examination (including medical history and other routine physical examinations) carried out by a licensed physician, to ensure the suitability of all participants to take part in the research protocols. On the same day, whole body dual-energy X-ray absorptiometry (GE Lunar DPX system, Madison, WI, USA) assessments were performed and used to calculate body fat and fat-free mass following previous recommendations [19]. VO 2peak (in mL/kg/min) was determined during a continuous incremental exercise test to volitional fatigue performed on a bicycle ergometer (Ergoselect 100, Ergoline GmbH, Bitz, Germany). Before tests, participants were familiarized with the ergometer and with the rating of perceived exertion (RPE), as measured by the Borg 6-20 scale [20]. Participants performed a standardized warm-up (5 min cycling at 20 W and 60 rpm), and the workload (starting at 20 W) was then increased by 25 W (men) or 20 W (women) every 2 minutes until volitional exhaustion. In participants >60 years of age, the workload was increased by 20 W (men) or 15 W (women) every 2 minutes. During the test, gas exchange data were measured 'breath-by-breath' with a metabolic cart (MetaMax 3B, Cortex Biophysik GmbH, Leipzig, Germany). Certified calibration gases (16.0% O 2 , 5.0% CO 2 , Cortex Biophysik) and a 3-L syringe were used to calibrate the gas analyzer and the flow meter, respectively, before each test. VO 2peak was defined as the highest VO 2 value (60-s average) obtained during the test. VO 2peak was considered valid when participants achieved at least two of the following criteria: (i) RPE >17, (ii) VO 2 difference between the last two consecutive loads <0.15 L/min, (iii) respiratory exchange ratio >1.1, and (iv) peak heart rate >85% of the age-adjusted estimate [21]. Heart rate was recorded with a chest strap transmitter (Polar RS400, Polar Electro, Kempele, Finland). The environmental temperature was similar in all measurement centers (~22 • C, 40% relative humidity).
On a different day during the week of testing, genomic DNA was extracted from peripheral whole blood samples using the Wizard Genomic DNA Purification Kit (Promega; Madison, WI, USA). Genotyping was performed at Shanghai Benegene Biotechnology, LTD (Shanghai, China). For analysis, a list of 125 SNPs (Tables A1 and A2, Tables A3 and A4) for the Han population of Beijing, China (CHB) was obtained from the International HapMap Project database. Haplotype-tag SNPs were selected using the following criteria: minor allele frequency ≥0.01 and measure of linkage disequilibrium (r 2 > 0.8). Initially, genes associated with cardiovascular responses to exercise were chosen, and genes associated with endurance performance, muscle performance, or body composition were then added as all of these factors might contribute to the value of VO 2peak (Table 1). Abbreviations: SNP, single nucleotide polymorphism. Abbreviations for gene names: ACE, angiotensin-converting enzyme; ACE2, angiotensin-converting enzyme 2; ACSL1, acyl-CoA synthase long-chain member 1; ACTN3, alpha-actinin-3; AGT, angiotensinogen; AGTR1, angiotensin II receptor type 1; AGTR2, angiotensin II receptor type 2; BDKRB2, bradykinin receptor B2; FGF21, Fibroblast growth factor 21; FGFR2, fibroblast growth factor receptor 2; FNDC5, fibronectin type III domain-containing protein 5; FST, follistatin; FTO, fat mass and obesity-associated protein (also known as 'alpha-ketoglutarate-dependent dioxygenase'; GDF8, growth differentiation factor 8 (also known as For high-throughput genotyping of SNPs, we used a matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) platform (Agena, San Diego, CA, USA). Primers for the polymerase chain reaction (PCR) and single-base extension were designed using the Assay Designer software package (Assay Design Suite V2.0, Agena, San Diego, CA, USA). Genotyping was performed as described elsewhere [50].

Statistical Analysis
All statistical analyses were performed using SAS 9.4 statistical package (SAS institute, Inc., Cary, NC, USA) and PLINK (v1.07). Hardy-Weinberg Equilibrium (HWE) was tested using χ 2 tests. Linear regression analyses were conducted to assess the association-expressed as standardized regression coefficients (β)-between each SNP and VO 2peak , with age, body mass and sex as covariates. The Bonferroni correction for multiple comparisons was applied to test for statistically significant associations between SNPs and VO 2peak , thereby setting the minimum level of significance at p < 0.0004 (i.e., 0.05 divided by the number of SNPs, i.e., 125). A multivariable regression analysis was then conducted to assess the overall contribution of the most significant SNPs to the interindividual variability of VO 2peak . All SNPs with p < 0.05 were included, and a regression model with backward elimination was used to filter-out redundant SNPs. By using a threshold of 5.0 points in the variance inflation factor, we avoided multicollinearity. SNPs that were retained in the final backward elimination model were then analyzed with a multivariate regression model using forward selection. The produced Genes 2020, 11, 555 5 of 16 regression equation was accepted at a significance level of p <0.01. The values of R 2 were adjusted for the number of cases and parameters in the analysis. The relative contribution (R 2 ) of each SNP in relation to the explained variance in VO 2peak was calculated as follows (Equation (1)): In the SNPs retained in multiple regression, VO 2peak values were compared among genotypes by using one-way analysis of variance (ANOVA). When the ANOVA showed a significant F value, pairwise differences were assessed using the Tukey post-hoc test. By using the SNPs retained in multiple regression analyses, we calculated a weighted genotype score to assess the combined influence of the SNPs on VO 2peak following the procedure of Williams and Folland [51]. First, each genotype was scored within each SNP by assigning 0 arbitrary units (a.u.) to homozygotes for the allele theoretically associated with low VO 2peak , 1 a.u. to heterozygotes, and 2 a.u. to homozygotes for the allele associated with high VO 2peak , following an additive model. Each SNP was then weighted by its β-coefficient (allele effect) based on the assumption that all SNPs of interest have independent effects and contribute in an additive manner to VO 2peak . Finally, the scores obtained for each SNP were summed to obtain a unique weighted genotype score for each participant (theoretical range: 0-6 a.u.). For clarity, we merged data of participants by using intervals of 1 a.u. Differences in VO 2peak between participants in the different groups were assessed by one-way analysis of variance and using the least significant difference post hoc test. Finally, the ability of weighted genotype score to distinguish individuals with low or intermediate CRF (i.e., below or above 28 mL/kg/min, as proposed by Kodama et al. [52]) was assessed using a receiver operating characteristic (ROC) curve and by determining the area under the ROC curve (AUC).

Results
Of the initial 1047 individuals recruited, valid VO 2peak measurements were obtained for 1009 individuals (566 women), and thus only these participants were included in the analyses. The main characteristics of the participants are shown in Table 2. Table 2. Main characteristics of the study participants (N = 1009) and their association with peak oxygen uptake.
In multiple regression analysis, and after excluding those SNPs with collinearity, only three were retained in the final model (ACE rs4295, AGTR1 rs275652, GDF8 rs7570532), which explained together 2.2% (p = 0.0007) of the variance in VO2peak (Figure 2a, statistical power = 0.987). The partial contribution of each SNP to the variance in VO2peak is shown in Table 3. The explained variance of VO2peak increased to 50.1% (p < 0.0001) when including covariates such as age, sex and weight in the model (Figure 2b, statistical power = 1.00).  In multiple regression analysis, and after excluding those SNPs with collinearity, only three were retained in the final model (ACE rs4295, AGTR1 rs275652, GDF8 rs7570532), which explained together 2.2% (p = 0.0007) of the variance in VO 2peak (Figure 2a, statistical power = 0.987). The partial contribution of each SNP to the variance in VO 2peak is shown in Table 3. The explained variance of VO 2peak increased to 50.1% (p < 0.0001) when including covariates such as age, sex and weight in the model (Figure 2b, statistical power = 1.00). Genotyping was successful (i.e., successful determinations for all SNPs) in 1006 of 1009 participants (99.7%). From the 125 SNPs analyzed, 10 were discarded because they deviated from HWE (Table B1), 10 because they had a MAF <5% (Table B2), and two because only one genotype was detected across the group of participants (Table B3). From the remaining pool of 103 SNPs, only rs4295, located in the angiotensin-converting enzyme (ACE) gene, was significantly associated with VO2peak (p < 2.9 × 10 −4 , β = 0.87; minor allele (G) frequency, 38.1%, heterozygosity frequency, 47.4%).
In multiple regression analysis, and after excluding those SNPs with collinearity, only three were retained in the final model (ACE rs4295, AGTR1 rs275652, GDF8 rs7570532), which explained together 2.2% (p = 0.0007) of the variance in VO2peak (Figure 2a, statistical power = 0.987). The partial contribution of each SNP to the variance in VO2peak is shown in Table 3. The explained variance of VO2peak increased to 50.1% (p < 0.0001) when including covariates such as age, sex and weight in the model (Figure 2b, statistical power = 1.00).  Individual VO 2peak values for each genotype of the ACE rs4295, AGTR1 rs275652, and GDF8 rs7570532 polymorphisms are shown in Figure 3. The one-way ANOVA revealed statistically significant differences in ACE rs4295 (F = 4.95, p = 0.007) and AGTR1 rs275652 (F = 3.90, p = 0.021) polymorphisms, while the ANOVA did not show differences for GDF8 rs7570532 (F = 1.64, p = 0.194) polymorphism. Specifically, GG homozygotes in ACE rs4295 had a mean VO 2peak of 31.1 ± 7.9 mL/kg/min, which was higher than that found in heterozygotes (GC, 29.8±6.9 mL/kg/min; p = 0.049) or in homozygotes for the common allele (CC, 28.9 ± 6.7 mL/kg/min; p = 0.013). In addition, AA homozygotes in AGTR1 rs275652 Genes 2020, 11, 555 7 of 16 had a mean VO 2peak of 30.0 ± 7.3 mL/kg/min, which was higher than that found in homozygotes for the minor allele (CC, 25.5 ± 5.3 mL/kg/min; p = 0.024). Individual VO2peak values for each genotype of the ACE rs4295, AGTR1 rs275652, and GDF8 rs7570532 polymorphisms are shown in Figure 3. The one-way ANOVA revealed statistically significant differences in ACE rs4295 (F = 4.95, p = 0.007) and AGTR1 rs275652 (F = 3.90, p = 0.021) polymorphisms, while the ANOVA did not show differences for GDF8 rs7570532 (F = 1.64, p = 0.194) polymorphism. Specifically, GG homozygotes in ACE rs4295 had a mean VO2peak of 31.1 ± 7.9 mL/kg/min, which was higher than that found in heterozygotes (GC, 29.8±6.9 mL/kg/min; p = 0.049) or in homozygotes for the common allele (CC, 28.9 ± 6.7 mL/kg/min; p = 0.013). In addition, AA homozygotes in AGTR1 rs275652 had a mean VO2peak of 30.0 ± 7.3 mL/kg/min, which was higher than that found in homozygotes for the minor allele (CC, 25.5 ± 5.3 mL/kg/min; p = 0.024). . Box-and-whisker plots showing peak oxygen uptake (VO2peak) in the study participants according to genetic variations in the genes for angiotensin-converting enzyme (ACE; rs4295), angiotensin II receptor type 1 (AGTR1; rs275652), and growth differentiation factor 8 (GDF8, also known as 'myostatin'; rs7570532). The lines in the box represent the first, second (median) and third quartiles, and the whiskers represent 1.5 × interquartile ranges. Each dot represents one individual within the specified genotype. (*) Depicts a statistically significant difference from CC genotype in ACE rs4295 polymorphism at p < 0.05. ( †) Depicts a statistically significant difference from CC genotype in AGTR1 rs275652 polymorphism at p < 0.05.
A weighted genotype score was constructed using the three SNPs shown in model 1 of genetic-only influence. Participants were categorized with a genotype score from 0 a.u., indicating Figure 3. Box-and-whisker plots showing peak oxygen uptake (VO 2peak ) in the study participants according to genetic variations in the genes for angiotensin-converting enzyme (ACE; rs4295), angiotensin II receptor type 1 (AGTR1; rs275652), and growth differentiation factor 8 (GDF8, also known as 'myostatin'; rs7570532). The lines in the box represent the first, second (median) and third quartiles, and the whiskers represent 1.5 × interquartile ranges. Each dot represents one individual within the specified genotype. (*) Depicts a statistically significant difference from CC genotype in ACE rs4295 polymorphism at p < 0.05. ( †) Depicts a statistically significant difference from CC genotype in AGTR1 rs275652 polymorphism at p < 0.05.
A weighted genotype score was constructed using the three SNPs shown in model 1 of genetic-only influence. Participants were categorized with a genotype score from 0 a.u., indicating the presence of homozygosity for all the alleles associated with a lower VO 2peak in ACE (rs4295), AGTR1 (rs275652) and GDF8 (rs7570532), to 6 a.u., indicating the presence of homozygosity for all the alleles associated with a higher VO 2peak in the aforementioned SNPs. A linear effect was found for genotype score on VO 2peak (Figure 4). Specifically, the individuals with 6 a.u. had a higher VO 2peak than those with scores up to 4.0 a.u. (p < 0.05). In addition, participants with scores >2 a.u. had a higher VO 2peak than those with scores <1.0 a.u. (p < 0.05). ROC analysis showed significant discriminatory accuracy of the weighted genotype score in the identification of individuals with low/intermediate CRF (AUC = 0.542) with a sensitivity of 0.733 and a specificity of 0.305. genotype score on VO2peak (Figure 4). Specifically, the individuals with 6 a.u. had a higher VO2peak than those with scores up to 4.0 a.u. (p < 0.05). In addition, participants with scores >2 a.u. had a higher VO2peak than those with scores <1.0 a.u. (p < 0.05). ROC analysis showed significant discriminatory accuracy of the weighted genotype score in the identification of individuals with low/intermediate CRF (AUC=0.542) with a sensitivity of 0.733 and a specificity of 0.305. . Peak oxygen uptake (VO2peak) levels in the study participants according to the genotype score (computed by using a weighted score of angiotensin-converting enzyme rs4295, angiotensin II receptor type 1 rs275652 and growth differentiation factor 8 rs7570532 genotypes). Abbreviations/symbols: a.u., arbitrary units; a, difference from 0-0.9 a.u. at p<0.05; b, difference from 1.0-1.9 a.u. at p < 0.05; c, difference from 2.0-2.9 a.u. at p < 0.05; d, difference from 3.0-3.9 a.u. at p < 0.05.

Discussion
CRF, particularly when objectively determined as VO2peak, is strongly associated with endurance performance and health outcomes. Indeed, VO2peak reflects the peak integrative ability of the organism to deliver oxygen from the atmosphere to the mitochondria of working muscles. The VO2peak is thus determined, among other factors, by peak cardiac output and pulmonary ventilation, lung diffusion capacity, blood and plasma volume, hemoglobin mass, and muscle capillary density and oxidative capacity [53]. Importantly, the mean values of VO2peak of our participants (29.9 ± 7.1 mL/kg/min or 8.5 metabolic equivalents, i.e., METs) were barely above the minimum healthy threshold for all-cause and CVD mortality in middle-aged men/women (i.e., 8 METs [52]). It is thus of medical importance to determine whether genetic factors (including specific gene variants) are associated with variability of CRF around (i.e., above vs below) the 8-MET cutoff. Previous research in siblings/twins suggests that 50%-60% of the variance of VO2peak is associated with heredity [10,12]. These values seem surprisingly high given the variety of physiological processes and body tissues involved in the uptake and utilization of oxygen in muscle mitochondria. Indeed, there is open debate about the limits of the evidence that support the relative influence of genetics on the variability and trainability of CRF [54,55]. . Peak oxygen uptake (VO 2peak ) levels in the study participants according to the genotype score (computed by using a weighted score of angiotensin-converting enzyme rs4295, angiotensin II receptor type 1 rs275652 and growth differentiation factor 8 rs7570532 genotypes). Abbreviations/symbols: a.u., arbitrary units; a, difference from 0-0.9 a.u. at p<0.05; b, difference from 1.0-1.9 a.u. at p < 0.05; c, difference from 2.0-2.9 a.u. at p < 0.05; d, difference from 3.0-3.9 a.u. at p < 0.05.

Discussion
CRF, particularly when objectively determined as VO 2peak , is strongly associated with endurance performance and health outcomes. Indeed, VO 2peak reflects the peak integrative ability of the organism to deliver oxygen from the atmosphere to the mitochondria of working muscles. The VO 2peak is thus determined, among other factors, by peak cardiac output and pulmonary ventilation, lung diffusion capacity, blood and plasma volume, hemoglobin mass, and muscle capillary density and oxidative capacity [53]. Importantly, the mean values of VO 2peak of our participants (29.9 ± 7.1 mL/kg/min or 8.5 metabolic equivalents, i.e., METs) were barely above the minimum healthy threshold for all-cause and CVD mortality in middle-aged men/women (i.e., 8 METs [52]). It is thus of medical importance to determine whether genetic factors (including specific gene variants) are associated with variability of CRF around (i.e., above vs below) the 8-MET cutoff. Previous research in siblings/twins suggests that 50%-60% of the variance of VO 2peak is associated with heredity [10,12]. These values seem surprisingly high given the variety of physiological processes and body tissues involved in the uptake and utilization of oxygen in muscle mitochondria. Indeed, there is open debate about the limits of the evidence that support the relative influence of genetics on the variability and trainability of CRF [54,55].
Our findings question the high heritability of VO 2peak , at least in Chinese individuals with no familial connection. From the 125 SNPs selected for our study, only one (ACE rs4295) was associated with VO 2peak . Also, the best model obtained through multiple regression analyses could only explaiñ 2.2% of the interindividual variance in VO 2peak . As in the study by Bye et al. [14], we created a polygenic score to determine whether those individuals with a higher number of alleles associated with VO 2peak did indeed present with higher values of this parameter. The only differences found between our findings and those of the Bye et al. study were the number of SNPs included in the polygenic score (7 vs 3, respectively) and the use of an intermediate genotype score for heterozygotes, which was not included by Bye et al. Interestingly, in both studies, participants with the theoretically lowest (or 'less favorable') genotype scores had the lowest VO 2peak (22-24 mL/kg/min), which was significantly lower than for those with the theoretically highest (or 'most favorable') genotype score (~32 mL/kg/min). These findings suggest that only a small number of SNPs are associated with the odds of having high VO 2peak values in untrained individuals. The variance of the interindividual variability in VO 2peak explained with these genotypes is low and the addition of favorable alleles might produce a change of 8-10 mL/kg/min. This genetic influence might be considerable in clinical terms because each 1-MET (or 3.5 mL/kg/min) increase in CRF has been shown to confer a 12% improvement in survival in Caucasian (North-American) men [6]. Moreover, as mentioned above, it is of clinical importance to surpass the 8-MET threshold, and in fact, adults with a CRF clearly above this level (>10 METs) have a remarkably reduced CVD risk [56]. In this regard, the probability of surpassing the 8-MET threshold (equivalent to 28 mL/kg/min) was doubled in those participants that carried the six 'favorable' alleles ( Figure 4).
Only three SNPs were included in the final multiple regression model. ACE rs4295 has not been previously associated with endurance performance, but it is located in the same linkage disequilibrium block as the widely studied ACE insert(I)/deletion(D) polymorphism (rs4340) [57]. The ACE gene encodes angiotensin-converting enzyme and the I allele might be associated with lower circulating levels of enzyme, and the II genotype potentially associated with performance in endurance athletes (odds ratio 1.35; 95% confidence interval, 1.17 to 1.55 [58]). However, several studies have found no association between the ACE I/D genotype and VO 2peak values in trained [59] and untrained [60,61] individuals. With regard to the ACE rs4295 variation found in the present study, although its influence on CRF needs to be replicated in other cohorts, our findings bolster the role of angiotensin-converting enzyme and its coding gene as predictors of CRF-related phenotypes. We also found that carriage of the C allele in the AGTR1 rs275652 polymorphism was negatively associated with VO 2peak values. This gene encodes the angiotensin II receptor 1 (AT 1 R), and polymorphisms in AGTR1 have been suggested to be involved in the physiological response to hypoxia [62]. AT 1 R is broadly expressed in different tissues and mediates most of the classical actions of angiotensin II, including vasoconstriction and vascular smooth muscle cell proliferation [63]. Thus, under hypoxic conditions, angiotensin II engages AT 1 R to modulate the pulmonary vasoconstrictive response [64]. Although speculative, it is possible that the C allele in AGTR1 rs275652 might exacerbate pulmonary vasoconstriction during exercise owing to a higher activation of AT 1 R for a given concentration of angiotensin II [65]. The last SNP included in the model explaining VO 2peak was rs7570532, a genetic variation in GDF8 encoding myostatin. This and other SNPs in GDF8 have been indirectly associated with a major cardiometabolic condition, obesity [39], but other authors have reported no association of rs7570532 with endurance performance [40]. Myostatin controls the differentiation and proliferation of skeletal muscle throughout embryonic development and regulates muscle growth during adulthood. Mutations in GDF8 that produce non-functional myostatin result in the increased growth of skeletal muscle, demonstrating the existence of a powerful mechanism to control muscle size in normal individuals through this protein [66]. Based on these findings and given the positive association between muscle mass and VO 2peak [66][67][68], it is possible that GDF8 rs7570532 confers a small but significant predisposition to higher VO 2peak values. Further research on these three SNPs is clearly warranted.
We acknowledge that the current investigation has some limitations. First, our study sample was heterogeneous in terms of age, sex, and anthropometric characteristics (Table 1). Although we used these variables as covariates in linear regression analyses, the high variability of these variables might have partially influenced our results. In fact, when they were included in multiple regression analyses (Figure 2b), the explained variance of VO 2peak increased up to 50.1%. Second, our study only included participants of Han Chinese descent and the results might therefore not be applicable to other ethnicities. Of note, the Han Chinese constitute the world's largest ethnic group (constituting 18% of the global population), but further studies in other large ethnic groups will be needed to confirm/discard the generalizability of these results. Lastly, we only analyzed 125 SNPs and thus it is plausible that other candidate genes might have an influence on VO 2peak .

Conclusions
The present study shows that in a cohort of untrained Han Chinese individuals, VO 2peak is influenced by a very few polymorphic variations in key genes even in isolation of training adaptations. The genetic influence accounted for~2.2% of the interindividual variance in VO 2peak , at least with the 125 SNPs included in this investigation. Although more research is needed, these data suggest that environment, probably more than genetics, is responsible for most of the interindividual variability in VO 2peak among healthy Han Chinese adults.