The prevalence of type 2 diabetes (T2D) is expected to rise sharply for the next 40 yr to a level where one in three U.S. adults could be affected (8). Numerous prospective epidemiological studies indicate that regular physical activity is related to a 15%–60% reduction in risk of T2D (reviewed in ) and that behavioral intervention that promotes physical fitness can reduce progression from prediabetes to T2D by up to 58% (22). Cardiorespiratory fitness has been inversely associated with incident T2D (25) and cardiovascular events (23). Exercise programs designed to increase physical fitness are recommended to patients with established T2D. The benefit of exercise can be seen with improved insulin sensitivity as well as reduced adiposity and adipose tissue inflammation (3).
Studies in animal models demonstrate a significant role for genetic background in physical endurance (2). Similarly, cardiorespiratory fitness in humans was found to be heritable, with heritability estimates ranging between 25% and 65% (reviewed in ). Genome-wide association studies (GWAS) conducted in the Framingham Heart Study and HERITAGE Family Study using large arrays of single nucleotide polymorphisms (SNPs) identified no variants associated with pretraining levels or changes in heart rate or fitness in response to training at the genome-wide significance level (P value < 5 × 10−8) (7,39). Suggestive signals, however, were identified in the ryanodine receptor gene (RYR2) as well as in other genes that have a plausible role in fitness including ACE, ADRB1, AGT, AGTR1, KCNH8, and others. In another study, SNPs in three muscle-related genes (CNTF, AMPD1, and NR3C1) predicted whether a patient with coronary artery disease responded to a 3-month ambulatory supervised exercise training regimen (38). Finally, a study using a combination of transcriptomics and genomics demonstrated that about half of the variance of V˙O2max trainability was accounted for either by the abundance of 29 muscle transcripts or by 11 SNPs (20). While these studies demonstrate that genetic predictors of fitness are starting to emerge, there is currently insufficient evidence to implicate specific genes responsible for the interindividual variation in fitness. Newer gene-centric array-based genotyping technologies that permit improved coverage of the candidate genes, and potentially deep re-sequencing approaches, that capture genetic diversity across populations may prove more effective in identifying fitness genes. Here, we analyzed data from the ITMAT-Broad-CARe (IBC) chip (19), primarily aimed at assaying SNPs in candidate genes and pathways for cardiovascular, inflammatory, and metabolic phenotypes to better define the complex and poorly characterized role of genetics in human fitness.
The Action for Health in Diabetes (Look AHEAD) randomized clinical trial demonstrated that an intensive lifestyle intervention (ILI), including both caloric restriction and physical activity, produced significantly greater weight loss and improved measures of glucose control in participants with established T2D after 1 yr, compared with a control intervention of diabetes support and education (DSE) (29). The ILI was also effective in increasing cardiorespiratory fitness in Look AHEAD subjects (16).
Here, we hypothesized that genetic variants modify the fitness response to ILI compared with DSE in the presence of established T2D. To test this hypothesis, we analyzed whether SNPs within genes already implicated in physical fitness and present on the IBC chip were associated with changes in fitness in response to 1 yr of intervention in Look AHEAD. A differential response to intervention by genotype would help identify biological pathways involved in fitness.
MATERIALS AND METHODS
The design and methods of the Look AHEAD trial have been reported elsewhere (36), as have the baseline characteristics of the entire randomized cohort (9). Among 5145 ethnically diverse overweight and obese Look AHEAD subjects with T2D and age 45 to 76 yr at baseline, 4041 provided consent and DNA for genetic analysis. The Look AHEAD trial was approved by local institutional review boards, including genetic analyses.
Subjects were randomly assigned to DSE or ILI. DSE received standard care plus 3 education sessions during the 1-yr period. ILI included individual and group contact throughout the year focusing on caloric restriction and increased physical activity, with the goal of achieving 10% or greater weight loss. ILI participants were instructed initially to increase their physical activity to at least 50 min·wk−1, progressing to at least 175 min·wk−1 by week 26, with the intensity being moderate to vigorous (similar to brisk walking). Participants were also encouraged to increase lifestyle forms of physical activity (using stairs rather than elevators, walking rather than riding, and reducing use of labor saving devices).
Subject characteristics, including age, sex, medication use, and race/ethnicity, were collected via questionnaire at baseline. Weight at baseline and 1 yr postrandomization was measured using the standardized methods as described previously (21).
Cardiorespiratory fitness was assessed using a graded exercise test (GXT) on a calibrated motor-driven treadmill as previously described using a standardized protocol (16). A self-selected walking speed of 1.5, 2.0, 2.5, 3.0, 3.5, or 4.0 mph was used with the speed held constant throughout the test. The grade of the treadmill was initiated at 0% and increased by 1% each minute until test termination. During the last 10 s of each minute and at the point of test termination, the heart rate was measured from a 12-lead ECG, and RPE was measured using the Borg 15-category scale (scale ranges from 6 to 20). Blood pressure was assessed during the last 45 s of each even minute and at test termination. A maximal graded exercise to the point of volitional fatigue was conducted at baseline. The baseline GXT was considered valid provided that that subject achieved either 85% of age-predicted maximal heart rate (defined as 220 − age) computed as if not taking a medication that would affect the heart rate response to exercise or RPE >18 if the subject was taking a medication that would affect the heart rate response to exercise (e.g., beta blocker). This baseline test was used to exclude individuals for whom exercise may have been contraindicated before study randomization. Because of cost constraints associated with the need for physician’s presence for a maximal test regardless of health status, subjects completed a submaximal GXT at 1 yr using the same walking speed and grade increments as was used for the baseline test; however, the test was terminated at the point where the participant first exceeded 80% of age-predicted maximal heart rate if not on a beta-blocker at either baseline or year 1 or first exceeded RPE = 16 if on a beta-blocker at either baseline or year 1. The workload at test termination at 1 yr was compared with the workload from baseline where the same heart rate (80% age-predicted maximal heart rate) or RPE (RPE = 16) was met on the baseline GXT. These workloads were converted to estimated METs using the American College of Sports Medicine’s metabolic calculations for estimating energy expenditure (1), and the change in fitness was computed as the difference in METs at the same submaximal heart rate or RPE between the baseline and 1 yr GXT.
Genotyping and candidate gene selection
Genotyping was carried out on leukocyte DNA using the Illumina CARe iSelect chip (19), as previously described (28). Briefly, genomic DNA was extracted from whole blood (FlexiGene DNA Kit; Qiagen Inc., Valencia, CA) and genotyping was carried out at the Children’s Hospital of Philadelphia. SNPs were clustered into genotypes using the Illumina BeadStudio software and subjected to quality control filters. Samples were excluded for individual call rates <90%, sex mismatch, and duplicate discordance. SNPs were removed for call rates <95%. Because of the low power for capturing genetic effects of the many low-frequency variants included in the design, we filtered out SNPs of minor allele frequency (MAF) <5%. This left 32,561 SNPs on the IBC chip with MAF ≥5%, whose mean genotyping success rate was 99.8%. After excluding individuals that failed the IBC chip genotyping, had low call rate, or had discrepancy between self-reported and X-chromosome-determined sex, the study cohort consisted of 3899 individuals. We performed a detailed literature review of all available fitness genetic association studies. From our review, we selected studies with a substantial sample size (>470 participants) that identified a total of 158 candidate genes previously reported by candidate gene, genetic linkage and GWAS, and gene expression studies to be associated with fitness traits based on standardized exercise treadmill test traits (Table 1, Supplementary Digital Content, http://links.lww.com/MSS/A343); 63 of these genes were represented on the IBC. We then performed gene-level replication by prioritizing the analysis of 1317 SNPs within the 63 genes included on IBC.
We conducted a joint analysis of baseline and 1-yr METs measurements, using an unstructured covariance matrix. Longitudinal models evaluating the effects of time (baseline vs 1 yr), study arm (ILI vs DSE), and individual SNPs markers (0/1/2 minor allele copies) and their interactions on fitness outcomes were estimated with Splus 8.2 (Tibco Software, Inc., 2010) using restricted maximum likelihood. An additive genetic model was assumed for all genetic markers, with regression coefficients interpreted as the effect on METs of each additional copy of the corresponding minor allele.
After excluding SNPs in linkage disequilibrium (LD, r2 > 0.3), EIGENSTRAT was used to compute principal components for use as covariates to control for population admixture in the regression analyses (26).
Our models additionally adjusted for study site, sex, age, weight, use of beta-blockers, and the first two principal components to adjust for population admixture and controlled for the effect on fitness of both baseline values and of change in time-varying covariates allowing these effects to differ by study arm.
For candidate gene analyses, involving 1317 SNPs within 63 candidate genes previously reported to be associated with exercise treadmill test traits that were present on IBC (Table 1, Supplementary Digital Content, http://links.lww.com/MSS/A343), we determined the number of uncorrelated markers to be 687, after accounting for LD using the Li and Ji approach (24). Therefore, after adjustment for multiple hypothesis testing, a P value threshold for statistical significance was set at 7.4 × 10−5 when testing for 1-yr change in either the ILI or DSE arms. However, because these analyses attempt to replicate the associations with genetic markers previously implicated in METs and/or related treadmill-test traits, we also point out at least nominal (P value < 0.05) associations.
For chipwide analyses, we also calculated the effective number of uncorrelated markers among the 32,561 IBC SNPs under investigation and found it to equal 17,669 after LD correction (24). After controlling for multiple comparisons, this resulted in a chipwide significance threshold of P = 2.9 × 10−6. We used a false discovery rate (FDR) approach to guide our reporting of suggestive (FDR<20%) associations, operationalized via a rank ordering of the genetic markers according to their q-values. FDR controls the expected proportion of false-negative results among those deemed significant. The q-values are marker-specific quantities that recalibrate the rank ordering of P values by the probability that they represent a false discovery, calculated using the q-value package of Dabney et al. (10).
Given higher power needed to detect interaction effects, we did not explicitly test for ILI-DSE differences in genetic effects on METs change across the entire marker set. However, we do report these interactions for the subset of markers showing associations with METs change in either study arm, reducing the number of multiple comparisons. However, it may also cause us to miss interactions caused by genetic effects on both ILI and DSE change that are modest in size and of opposite sign.
In addition to the full-sample analyses, we conducted a sensitivity analysis excluding individuals receiving beta-blocker medications because their METs phenotype was calculated using different methodology than for the remaining study participants.
To identify biological associations between the top genes involved in cardiorespiratory fitness, the Gene Relationships Among Implicated Loci (GRAIL) was used (33). GRAIL scores association signals by evaluating whether observed genomic regions are nonrandomly linked to the other genes through word-similarity metrics in PubMed abstracts (33). We used the list of SNPs that showed at least nominal associations in the candidate genes and possible associations from the chip-wide analyses (q-value < 0.30) with 1-yr response to ILI to assess the degree of connectivity between the genes.
Look AHEAD genetic study
At baseline, 3899 Look AHEAD subjects who participated in this genetic study were evenly distributed between the ILI and DSE intervention arms with regard to age, sex, ethnicity, weight, and baseline fitness (Table 1). At 1 yr, METs levels of study participants that did not use beta-blockers at baseline increased, on average, by 1.02 U in individuals in ILI versus 0.23 U in DSE (Table 1). Comparable intervention effects were observed among individuals receiving beta-blockers at baseline, with METs levels increasing by 0.91 U in ILI versus 0.21 U in DSE. Beta-blocker use itself was stable across time, with only 5.3% of subjects switching regimens from baseline to follow-up.
Candidate gene analysis of treatment response
We analyzed the association of 1-yr change in METs for 1317 SNPs (Table 1, Supplementary Digital Content, http://links.lww.com/MSS/A343). After the adjustment for multiple hypothesis testing, one significant association was identified between RUNX1 rs9976623 (MAF = 0.24) and 1-yr change in METs in the ILI group (P = 1.9 × 10−5; Table 2; Figure 1A, Supplementary Digital Content, http://links.lww.com/MSS/A344. Carriers of the RUNX1 rs9976623 minor allele in the ILI group gained 0.19 ± 0.04 less METs per copy than noncarriers. Carriers in the DSE group showed no significant difference in 1-yr METs (P = 0.82). These ILI-DSE differences resulted in a nominally significant interaction (P = 4.8 × 10−3; Table 2) between rs9976623 minor allele status and treatment response. Two other SNP in moderate LD with rs9976623 showed nominal significance as indicated on the regional plot (see Figure 1A, Supplementary Digital Content, http://links.lww.com/MSS/A344). Minor alleles of multiple common COL4A1 SNP, in LD with each other, showed at least nominally significant associations with 0.15 ± 0.04 more METs gain per copy than noncarriers in the ILI arm (P < 2.6 × 10−4; Table 2; Figure 1B, Supplementary Digital Content, http://links.lww.com/MSS/A344. No difference was detected in the DSE group (P < 0.18), resulting in at least nominally significant treatment–genotype interactions (P < 1.1 × 10−3, Table 2). In addition, at least nominally significant associations for within-arm change were detected for ACE and AGT in the DSE group and for PRKAG2 in the ILI group (P < 5 × 10−4, Table 2).
Of the remaining 58 candidate genes on the IBC chip, at least nominally significant associations were observed between 1-yr change in response to ILI and DSE and multiple SNP within 29 genes, including RYR2, CASR, and ACE (see Table 1, Supplementary Digital Content, http://links.lww.com/MSS/A343). Moreover, the majority of these genes also showed at least nominal genotype–intervention interactions, indicating effect modification in response to ILI and DSE based on the genotype status (data not shown).
Chipwide association analysis of response to lifestyle intervention
No variants were associated with 1-yr METs change in response to ILI or DSE at a chip-wide level of significance. However, in addition to RUNX1, 13 SNP representing 11 independent loci showed genetic associations with ILI treatment response at q-value < 0.20 after the adjustment for covariates (Table 3). The majority also showed nominal genotype–treatment interaction (P < 0.05). These findings included 2 SNPs each at TBC1D1 and MTMR15, and one SNP each at GNAI2, PROP1, FKBP7, SMURF1, NRG3, PLA2G4B, C20orf75, and THBD. An intergenic SNP on chromosome 11p15.5 was not mapped to any known gene (Table 3).
Twenty-three percent of the Look AHEAD genetic sample (N = 893) received beta-blockers at the time of the treadmill test. As METs were estimated differently among participants receiving beta-blockers (see Materials and Methods section), we performed a sensitivity analysis excluding such subjects. In candidate gene analysis, multiple COL4A1 SNP were at least nominally associated with fitness response to intervention with rs11069830 passing the threshold for suggestive association (Table 4). In chipwide analyses, the elimination of subjects taking beta-blockers changed the order of some of the top hits in the ILI group, with FKBP7 rs17225700 now showing a stronger effect that passed the chipwide significance threshold (1-yr gain of 0.47 ± 0.09 less METs per allele copy; P = 5.3 × 10−7; Table 4). Specifically, Figure 1 demonstrates a significant effect of ILI on fitness change in FKBP7 rs17225700-A allele carriers that was significantly diminished in GG homozygotes for both men and women. Substantial genotype–treatment interaction was also detected (P = 5.3 × 10−5). The new signals with q-value < 0.20 that were not detected in the pooled analysis included DDN, RGS2, ALCAM, VAX2, and LOC652968 (Table 4).
To identify pathway connections between genes associated with fitness, we performed a GRAIL analysis. A strong connectivity between the genes by their enrichment in overlapping pathways was identified among 24 of the 33 previously reported fitness genes that were found to be at least nominally associated with change in METs in Look AHEAD and were in the GRAIL database, including those that influence vascular tone, muscle protein synthesis, and contraction as well as cardiac energy substrate dynamics (Figure 2, Supplementary Digital Content, http://links.lww.com/MSS/A345). Genes with the largest number of connections included AGT, CAV1, NOS3, ADRA1A, ADRA1B, COL4A1, COL4A2, and ACE (GRAIL P < 2 × 10−6, data not shown). Of the 35 genes associated with METs response to ILI in the chipwide analysis at q-value < 0.30, eight genes, GHSR, HCRTR2, PROP1, VWF, THBD, TGFB3, CORIN, and TIPM1, demonstrated significant connections, with GRAIL P < 0.05 (Fig. 2).
This is the largest study to date in a cohort of well-characterized overweight and obese individuals with T2D determining whether genetic variants located at or near genes previously associated with fitness traits or within ∼2100 genes associated with cardiovascular, inflammatory, and metabolic traits help explain variation in fitness response to a 1-yr lifestyle intervention.
The candidate gene analysis identified a significant signal in RUNX1, previously identified as a fitness gene from the aerobic training-responsive transcriptome (20) that in the present study was associated with 1-yr METs response in the ILI group and genotype–intervention interaction. RUNX1 is the runt-related transcription factor gene involved in erythropoesis. Thus, there is a plausible contribution to fitness via its effects on the red blood cell pool and the delivery of oxygen to tissues, including muscle, due to improved oxygenation during times of physical stress. Carriers of the RUNX1 rs9976623 minor allele increased their fitness by 0.19 U less per copy in response to ILI; there was no association observed in the DSE group. We also detected an at least nominally significant association with COL4A1 and PRKAG2 with METs change in ILI. COL4A1 is the major type IV alpha collagen chain of basement membranes that has been observed to be differentially expressed in human muscle of high responders when compared with low responders after 6 wk of aerobic exercise training (20). Variants in PRKAG2, an energy sensor that modulates glucose uptake and glycolysis leading to enhanced glycolytic capacity and protection against hypoxic injury in tissues such as the heart (35), were also identified by the GWAS of treadmill exercise responses in the Framingham Heart Study to be associated with heart rate during the recovery period after exercise (39).
Variants in angiotensinogen (AGT) and angiotensin converting enzyme (ACE) were suggestively associated with change in fitness, but only in the DSE group. The ACE gene insertion/deletion (I/D), known to affect serum enzyme levels (34), has been associated with exercise response (17) and muscle endurance by some studies, but not others (14). The closest associated coding SNP rs1800764 was located only 140 base pairs away from the I/D variant, and has been reportedly associated with diabetic kidney disease (13).
We also observed nominally significant associations of multiple RYR2 gene variants with fitness response. RYR2 SNPs were found to be associated with heart rate during treadmill test in the Framingham Heart Study (39) and implicated in V˙O2max training response to a standardized 20-wk exercise program in the HERITAGE study GWAS (7). The SNPs identified in our study were found to not be in LD with the RYR2 SNP previously reported. RYR2 is a cardiac-type ryanodine receptor that plays a key role in triggering cardiac muscle contraction (4). Defects in RYR2 are the cause of familial arrhythmogenic right ventricular cardiomyopathy 2 and of exercise-induced polymorphic ventricular tachyarrhythmias (30).
Next, in an analyses of all SNPs across the IBC chip, no chipwide significant associations were identified in treatment response that passed the correction for multiple hypothesis testing. The strongest association with regard to ILI response was detected for TBC1D1, with carriers of the minor allele being more likely to gain METs. TBC1D1 is an insulin-sensitive regulator of GLUT4 function in skeletal muscle, suggesting that variation in TBC1D1 may alter glucose uptake, which could have effects on physical fitness. Importantly, a nonsynonymous polymorphism in the TBC1D1 gene (R125W, rs35859249), located ∼34 kb upstream of our top TBC1D1 hit, has been associated with severe familial obesity (27). Interestingly, TBC1D1 variation was also found to be associated with 1-yr weight loss in the ILI group (26).
Secondary analysis of data after exclusion of subjects receiving beta blockers revealed a variant in FKBP7 that showed a significant association with ILI response and treatment interaction. Carriers of the minor allele randomized to ILI showed a 0.47 ± 0.09 less MET increase per copy, whereas no difference between the genotypes was detected in the DSE group. Kavanagh et al. (18) reported that each 0.3 MET increase in peak V˙O2 above a threshold of 3.7 METs associated with a marked benefit in prognosis of cardiovascular and all-cause mortality and conferred a 10% reduction in cardiac mortality in women with known coronary artery disease, indicating that the effect size observed in our study may have clinical implications. FKBP7 is a member of the FKBP-type peptidyl-prolyl cis/trans isomerase family that interacts with FK-506, which is the drug target of rapamycin known to influence muscle protein synthesis in response to exercise (12). The role for the molecular chaperone FKBP7 in cellular signaling is not defined; however, our studies raise the possibility that FKBP7 may modulate cellular signaling processes related to fitness.
We sought to identify connections between all significant genes and those with the association q-value < 0.20 that influence fitness by applying GRAIL, a program that uses abstracts from the entirety of the published scientific literature to look for relatedness among genes within associated regions that may represent key pathways (33). COL4A1, whose variants we found to be associated with behavioral treatment interaction and fitness response in the ILI group, was linked with integrin beta 1 (ITGB1), along with caveolins (CAV1 and CAV2). These genes were expressed differentially in human muscle from high and low responders to 6 wk of aerobic exercise training (20). GRAIL also identified connectivity with nitric oxide synthase 3 (NOS3), or endothelial NOS, which regulates vascular smooth muscle relaxation, as well as ADRA1A and ADRA1B, which are expressed in the heart and play a major role in smooth muscle contraction (Figure 2, Supplementary Digital Content, http://links.lww.com/MSS/A345. ADRA1B has been involved in the control of vascular tone linked to cardiomyopathy and heart failure (5). Pathway-based analysis also suggested the involvement of endogenous hormones in cardiorespiratory response to ILI through association signals in GHSR, a growth hormone secretagogue receptor, and PROP1, responsible for pituitary development and hormone expression (Fig. 2). Interestingly, these genes were also biologically related to the hypocretin receptor type 2 (HCRTR2) gene, which encodes a G-protein-coupled receptor involved in the regulation of feeding behavior.
The strengths of this study include its randomized assignment of a lifestyle intervention of documented health relevance and the objective measurement of fitness by treadmill testing in the largest sample size to date. Further, the randomized intervention reduces the effect of confounding factors in association studies based on cross-sectional and observational data. Finally, the use of GRAIL to identify subsets of genes involved in similar biological processes related to cardiorespiratory fitness increases the confidence in the biological plausibility of our findings, beyond that provided by statistical estimates of FDR.
Although the IBC chip is a strength of our study, permitting analysis of more than 32,000 SNP in relation to fitness response, we note that this genotyping array is focused on ∼2100 candidate genes previously associated with cardiovascular, inflammatory and metabolic phenotypes. Indeed, of 158 genes selected from the prior literature (Table 1, Supplementary Digital Content, http://links.lww.com/MSS/A343), we were able to represent at least one SNP in 63 of these genes. For example, no SNPs were available in PAPSS2, the region identified in a prior GWAS of physical activity participation (11). In several cases, the SNP available was not in close proximity or co-inherited with the previously identified marker. Therefore, although many new regions were queried in the present analysis, this array provided a limited window through which prior candidate gene and GWAS studies associated with fitness could be replicated, leaving the possibility that more direct replication attempts may prove fruitful.
Our intervention is also a strength, given its basis on the intervention deployed in the Diabetes Prevention Program that was successful in reducing diabetes incidence for 4 yr among individuals with impaired glucose tolerance (22). The Look AHEAD intervention also successfully increased cardiorespiratory fitness across more than 2500 overweight individuals with T2D. The Look AHEAD intervention, providing physical activity goals and behavioral strategies and counseling to support physical activity uptake and maintenance, nonetheless differs from many prior genetic studies of exercise involving directly supervised exercise. In particular, individuals exposed to the intervention may not have taken up exercise in the same frequency, intensity, and duration as would likely occur under supervised exercise training. We believe these approaches are complimentary, as genetic factors that influence uptake of exercise could be equally important as genetic factors that influence physiologic response to a standardize exercise training. Fitness changes resulting from flexible supervised moderate physical activity may be applicable to a larger segment of the population in the community and may elicit different physiologic changes when compared with those observed with standardized supervised exercise training.
We acknowledge additional limitations in our study, including the use of a submaximal test at year 1 and lack of replication cohort with comparable intervention and outcome measurements. While restricting our analysis to 1-yr follow-up may mitigate detectable genetic effects associated with the ability to improve cardiorespiratory fitness, we note that the largest change in weight and fitness in the Look AHEAD cohort occurred during the first year of ILI (40). Similar to other genetic association studies based on randomized clinical trials, our results may not be generalizable to the general population and our power to detect modest effects is limited. Because of the strong overlap between body size and fitness, weight change may influence change in physical fitness. To control for the effect of weight on fitness, we incorporated weight at baseline and 1-yr postintervention in our models. Finally, data on physical activity were not included in this study. Self-reported physical activity was available only in a subset of Look AHEAD participants, which would have considerably reduced our active sample size and power. Furthermore, our group has found that self-reported physical activity appears to overestimate exercise behaviors when compared with objectively-measured physical activity (6).
In summary, using a gene-centric genotype chip with ∼2100 genes implicated in cardiovascular, inflammatory, and metabolic traits, we identified genetic associations of RUNX1 and FKBP7, involved in erythropoesis and muscle protein synthesis, respectively, with change in cardiorespiratory fitness in response to lifestyle intervention. Despite the substantial sample size and state-of-the-art statistical approaches, we were able to detect a small number of significant associations. Our findings speak to the complex nature of the fitness phenotype, with multiple genes likely to be involved, each of modest effect size. We may have failed to identify important fitness genes or gene variants that were not included on the IBC chip. Future genome-wide association studies or genetic analyses that include rare variants (MAF<5%) may identify fitness gene variants not described here.
Looking forward, the replication of our findings in independent cohorts would allow for a targeted validation of the novel variants in the context of less stringent P values, such as those used in our study, due to a focused hypothesis testing approach. Similar to the analysis of other common traits, meta-analysis of fitness data from GWASs with larger sample sizes can identify novel variants with smaller effects. However, it is important to acknowledge that fitness tests are very labor intensive and may not be readily performed on large epidemiologic cohorts. Nonetheless, standardized measures of fitness and physical activity could be adopted by large studies to allow for more powerful joint analyses. For example, PhenX is a tool designed to build consensus for standard measures of phenotypes and exposures used in genetic studies (15) that could help standardize fitness measures suitable for large genetic studies. In addition, next generation sequencing of DNA from participants with “phenotypic extremes” (i.e., highly physically trained individuals or persons resistant to training) may identify genetic variants responsible for extreme fitness responses. Mendelian randomization studies aimed at determining whether the contribution of the variants in the candidate fitness genes is causal for the development of cardiovascular and metabolic outcomes are warranted. Lastly, it will ultimately be important to integrate multiple genetic variations in the DNA code (e.g., SNPs, copy number variants, and methylation patterns) and gene expression to further explore the role of the genome in fitness.
Funding and Support
This study is supported by the Department of Health and Human Services through the following cooperative agreements from the National Institutes of Health: DK57136, DK57149, DK56990, DK57177, DK57171, DK57151, DK57182, DK57131, DK57002, DK57078, DK57154, DK57178, DK57219, DK57008, DK57135, and DK56992. The following federal agencies have contributed support: National Institute of Diabetes and Digestive and Kidney Diseases; National Heart, Lung, and Blood Institute; National Institute of Nursing Research; National Center on Minority Health and Health Disparities; NIH Office of Research on Women’s Health; and Centers for Disease Control and Prevention. This research was supported in part by the Intramural Research Program of the National Institute of Diabetes and Digestive and Kidney Diseases. The Indian Health Service (I.H.S.) provided personnel, medical oversight, and use of facilities. The opinions expressed in this paper are those of the authors and do not necessarily reflect the views of the I.H.S. or other funding sources.
Additional support was received from The Johns Hopkins Medical Institutions Bayview General Clinical Research Center (M01RR02719); the Massachusetts General Hospital Mallinckrodt General Clinical Research Center (M01RR01066); the University of Colorado Health Sciences Center General Clinical Research Center (M01RR00051) and Clinical Nutrition Research Unit (P30 DK48520); the University of Tennessee at Memphis General Clinical Research Center (M01RR0021140); the University of Pittsburgh General Clinical Research Center (M01RR000056 44) and NIH grant (DK 046204); the VA Puget Sound Health Care System Medical Research Service, Department of Veterans Affairs; and the Frederic C. Bartter General Clinical Research Center (M01RR01346), and DK090043 to J.M.M.
The following organizations have committed to make major contributions to Look AHEAD: Federal Express; Health Management Resources; Johnson & Johnson, LifeScan Inc.; Optifast-Novartis Nutrition; Roche Pharmaceuticals; Ross Product Division of Abbott Laboratories; Slim-Fast Foods Company; and Unilever.
The results of the present study do not constitute endorsement by the American College of Sports Medicine.
Dr. Unick received an honorarium for a scientific presentation at the International Diabetes Interchange Symposium
National Institute of Diabetes and Digestive and Kidney Diseases: Barbara Harrison, M.S.; Van S. Hubbard, M.D., Ph.D.; Susan Z. Yanovski, M.D.
National Heart, Lung, and Blood Institute: Lawton S. Cooper, M.D., MPH; Jeffrey Cutler, M.D., MPH; Eva Obarzanek, Ph.D., M.P.H., R.D.
Centers for Disease Control and Prevention: Edward W. Gregg, Ph.D.; David F. Williamson, Ph.D.; Ping Zhang, Ph.D.
1. American College of Sports Medicine. Guidelines for Exercise Testing and prescription
. In: Franklin BA, editor. Philadelphia: Lippincott Williams & Wilkins; 2005. pp. 62–6.
2. Barbato JC, Koch LG, Darvish A, Cicila GT, Metting PJ, Britton SL. Spectrum of aerobic endurance running performance in eleven inbred strains of rats. J Appl Physiol
. 1998; 85 (2): 530–6.
3. Belalcazar LM, Reboussin DM, Haffner SM, et al. A 1-year lifestyle intervention for weight loss in individuals with type 2 diabetes reduces high C-reactive protein levels and identifies metabolic predictors of change: from the Look AHEAD (Action for Health in Diabetes) study. Diabetes Care
. 2010; 33 (11): 2297–303.
4. Benkusky NA, Farrell EF, Valdivia HH. Ryanodine receptor channelopathies. Biochem Biophys Res Commun
. 2004; 322 (4): 1280–5.
5. Benoit MJ, Rindt H, Allen BG. Cardiac-specific transgenic overexpression of alpha1B-adrenergic receptors induce chronic activation of ERK MAPK signalling. Biochem Cell Biol
. 2004; 82 (6): 719–27.
6. Bond DS, Jakicic JM, Unick JL, et al. Pre- to postoperative physical activity changes in bariatric surgery patients: self report vs. objective measures. Obesity (Silver Spring)
. 2010; 18 (12): 2395–7.
7. Bouchard C, Sarzynski MA, Rice TK, et al. Genomic predictors of the maximal O(2) uptake response to standardized exercise training programs. J Appl Physiol
. 2011; 110 (5): 1160–70.
8. Boyle JP, Thompson TJ, Gregg EW, Barker LE, Williamson DF. Projection of the year 2050 burden of diabetes in the US adult population: dynamic modeling of incidence, mortality, and prediabetes prevalence. Popul Health Metr
. 2010; 8: 29.
9. Bray G, Gregg E, Haffner S, et al. Baseline characteristics of the randomised cohort from the Look AHEAD (Action for Health in Diabetes) study. Diab Vasc Dis Res
. 2006; 3 (3): 202–15.
11. De Moor MH, Liu YJ, Boomsma DI, et al. Genome-wide association study of exercise behavior in Dutch and American adults. Med Sci Sports Exerc
. 2009; 41 (10): 1887–95.
12. Drummond MJ, Fry CS, Glynn EL, et al. Rapamycin administration in humans blocks the contraction-induced increase in skeletal muscle protein synthesis. J Physiol
. 2009; 587 (Pt 7): 1535–46.
13. Ezzidi I, Mtiraoui N, Kacem M, Chaieb M, Mahjoub T, Almawi WY. Identification of specific angiotensin-converting enzyme variants and haplotypes that confer risk and protection against type 2 diabetic nephropathy. Diabetes Metab Res Rev
. 2009; 25 (8): 717–24.
14. Frederiksen H, Gaist D, Bathum L, et al. Angiotensin I-converting enzyme (ACE) gene polymorphism in relation to physical performance, cognition and survival–a follow-up study of elderly Danish twins. Ann Epidemiol
. 2003; 13 (1): 57–65.
15. Hamilton CM, Strader LC, Pratt JG, et al. The PhenX Toolkit: get the most from your measures. Am J Epidemiol
. 2011; 174 (3): 253–60.
16. Jakicic JM, Jaramillo SA, Balasubramanyam A, et al. Effect of a lifestyle intervention on change in cardiorespiratory fitness in adults with type 2 diabetes: results from the Look AHEAD Study. Int J Obes (Lond)
. 2009; 33 (3): 305–16.
17. Jones A, Montgomery HE, Woods DR. Human performance: a role for the ACE genotype? Exerc Sport Sci Rev
. 2002; 30 (4): 184–90.
18. Kavanagh T, Mertens DJ, Hamm LF, et al. Peak oxygen intake and cardiac mortality in women referred for cardiac rehabilitation. J Am Coll Cardiol
. 2003; 42 (12): 2139–43.
19. Keating BJ, Tischfield S, Murray SS, et al. Concept, design and implementation of a cardiovascular gene-centric 50 k SNP array for large-scale genomic association studies. PLoS ONE
. 2008; 3 (10): e3583.
20. Keller P, Vollaard NB, Gustafsson T, et al. A transcriptional map of the impact of endurance exercise training on skeletal muscle phenotype. J Appl Physiol
. 2011; 110 (1): 46–59.
21. Kelley DE, Goodpaster BH. Effects of exercise on glucose homeostasis in Type 2 diabetes mellitus. Med Sci Sports Exerc
. 2001; 33 (6 Suppl): S495–501; discussion S28–9.
22. Knowler WC, Barrett-Connor E, Fowler SE, et al. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med
. 2002; 346 (6): 393–403.
23. Laukkanen JA, Kurl S, Salonen R, Rauramaa R, Salonen JT. The predictive value of cardiorespiratory fitness for cardiovascular events in men with various risk profiles: a prospective population-based cohort study. Eur Heart J
. 2004; 25 (16): 1428–37.
24. Li J, Ji L. Adjusting multiple testing in multilocus analyses using the eigenvalues of a correlation matrix. Heredity (Edinb)
. 2005; 95 (3): 221–7.
25. Lynch J, Helmrich SP, Lakka TA, et al. Moderately intense physical activities and high levels of cardiorespiratory fitness reduce the risk of non-insulin-dependent diabetes mellitus in middle-aged men. Arch Intern Med
. 1996; 156 (12): 1307–14.
26. McCaffery J, Papandonatos GD, Huggins G, et al. Human cardiovascular disease IBC chip-wide association with weight loss and weight regain in the Look AHEAD trial. Human Heredity
. 2013; 75 (2–4): 160–74.
27. Meyre D, Farge M, Lecoeur C, et al. R125W coding variant in TBC1D1 confers risk for familial obesity and contributes to linkage on chromosome 4p14 in the French population. Hum Mol Genet
. 2008; 17 (12): 1798–802.
28. Peter I, McCaffery JM, Kelley-Hedgepeth A, et al. Association of type 2 diabetes susceptibility loci with one-year weight loss in the look AHEAD clinical trial. Obesity (Silver Spring)
. 2012; 20 (8): 1675–82.
29. Pi-Sunyer X, Blackburn G, Brancati FL, et al. Reduction in weight and cardiovascular disease risk factors in individuals with type 2 diabetes: one-year results of the look AHEAD trial. Diabetes Care
. 2007; 30 (6): 1374–83.
30. Priori SG, Napolitano C, Tiso N, et al. Mutations in the cardiac ryanodine receptor gene (hRyR2) underlie catecholaminergic polymorphic ventricular tachycardia. Circulation
. 2001; 103 (2): 196–200.
31. Qi L, Hu FB, Hu G. Genes, environment, and interactions in prevention of type 2 diabetes: a focus on physical activity and lifestyle changes. Curr Mol Med
. 2008; 8 (6): 519–32.
32. Raychaudhuri S. VIZ-GRAIL: Visualizing Functional Connections Across Disease Loci. Bioinformatics
. 2011; 27: 1589–90.
33. Raychaudhuri S, Plenge RM, Rossin EJ, et al. Identifying relationships among genomic disease regions: predicting genes at pathogenic SNP associations and rare deletions. PLoS Genet
. 2009; 5 (6): e1000534.
34. Rigat B, Hubert C, Alhenc-Gelas F, Cambien F, Corvol P, Soubrier F. An insertion/deletion polymorphism in the angiotensin I-converting enzyme gene accounting for half the variance of serum enzyme levels. J Clin Invest
. 1990; 86 (4): 1343–6.
35. Russell RR 3rd, Li J, Coven DL, et al. AMP-activated protein kinase mediates ischemic glucose uptake and prevents postischemic cardiac dysfunction, apoptosis, and injury. J Clin Invest
. 2004; 114 (4): 495–503.
36. Ryan DH, Espeland MA, Foster GD, et al. Look AHEAD (Action for Health in Diabetes): design and methods for a clinical trial of weight loss for the prevention of cardiovascular disease in type 2 diabetes. Control Clin Trials
. 2003; 24 (5): 610–28.
37. Teran-Garcia M, Rankinen T, Bouchard C. Genes, exercise, growth, and the sedentary, obese child. J Appl Physiol
. 2008; 105 (3): 988–1001.
38. Thomaes T, Thomis M, Onkelinx S, et al. A genetic predisposition score for muscular endophenotypes predicts the increase in aerobic power after training: the CAREGENE study. BMC Genet
. 2011; 12: 84.
39. Vasan RS, Larson MG, Aragam J, et al. Genome-wide association of echocardiographic dimensions, brachial artery endothelial function and treadmill exercise responses in the Framingham Heart Study. BMC medical genetics
. 2007; 8 (1 Suppl): S2.
40. Wing RR. Long-term effects of a lifestyle intervention on weight and cardiovascular risk factors in individuals with type 2 diabetes mellitus: four-year results of the Look AHEAD trial. Arch Intern Med
. 2010; 170 (17): 1566–75.