Journal Logo


Advances in Exercise, Fitness, and Performance Genomics in 2015


Author Information
Medicine & Science in Sports & Exercise: October 2016 - Volume 48 - Issue 10 - p 1906-1916
doi: 10.1249/MSS.0000000000000982
  • Free


This publication is the latest installment of an annual publication on exercise genetics and genomics. It summarizes the scientific literature reported in the calendar year 2015. The review focuses on the strongest publications as defined by study design, sample size, analytical strategy, novelty, and relevance of phenotypes with potential implications for exercise science and sports medicine. The review is not a comprehensive summary of all published articles on genetics and genomics relative to exercise, physical activity, fitness, and performance, and standards for the selection of articles have been explained in prior yearly installments of the publication (e.g., see Loos et al. [28]). Given the dearth of studies meeting these strict requirements for inclusion, the authors of the present annual review are evaluating several options for future years. One option would be to publish the review biennially. Another option is to terminate the series. A third one would be to morph the current effort into a review with a broader focus. These and other potential avenues will be pondered by the present team in the coming months.

The 2015 review is organized around the following topics: a) physical activity behavior, b) muscular strength and power, c) cardiorespiratory fitness and endurance performance, d) body weight and adiposity, e) insulin and glucose metabolism, f) lipid and lipoprotein metabolism, and g) hemodynamic traits. The article ends with a summary of the noteworthy findings from 2015, as well as a discussion of direct-to-consumer (DTC) products, a conceptual framework for exercise genomics studies, and highlights the importance of the recently launched the National Institutes of Health (NIH) Common Fund program titled “Molecular Transducers of Physical Activity in Humans.”


In 2015, one interesting study related to the molecular genetics of physical activity behavior was published. Kostrzewa et al. (25) used a combination of animal models and human studies to identify potential chromosomal regions for spontaneous physical activity level. First, the investigators quantified running wheel activity in multiple mouse chromosome substitution strains (CSS) as well as in their progenitor lines, A/J and C57BL/B6 mice. The initial analyses suggested that the CSS2 line would be the most informative (significantly lower running wheel activity compared with the B6 mice) and that a quantitative trait locus (QTL) for wheel running activity would be located on mouse chromosome 2. The QTL was fine mapped using female mice from a CSS2 × C57BL/B6 F2 population and narrowed down to a 3.3-Mb region on chromosome 2. The QTL explained 16.6% of the variance in running wheel activity in the F2 population.

Next, the investigators identified a syntenic region on human chromosome 20 that mapped between 51.2 and 55.2 Mb (25). The region was tested for associations between questionnaire-based physical activity phenotypes and genome-wide single nucleotide polymorphisms (SNPs) in two large human studies: the TwinsUK sample and the ALSPAC cohort. In females of the TwinsUK sample, one of the tested SNPs (rs459465) remained statistically significant after correcting for multiple testing. Replication analysis of the same chromosomal region in adolescent females of the ALSPAC cohort revealed two SNPs (rs6022999 and rs6092090) that remained significantly associated with a dichotomous questionnaire–derived activity trait (exerciser = individual exercised 4 h·wk−1 or more; nonexerciser = individual exercised less than 4 h·wk−1). However, none of the SNPs was associated with objective measures (accelerometer) of physical activity. Thus, the results suggest that both human studies identified SNPs in the target region that remained significant even after multiple testing was accounted for. However, it should be noted that there was no overlap between the studies in terms of the specific SNPs associated with physical activity traits. The authors suggested MC3R, CYP24A1, and GRM8 as potential candidate genes in this region based on bioinformatics analysis.

The study by Kostrzewa et al. (25) provides an example of how animal studies could be used to identify target genomic regions for physical activity traits in humans. However, there are some caveats in their study design that need to be kept in mind. For example, the activity traits in animal and human studies focused on different ends of the physical activity level distribution. The mouse QTL reflected reduced running activity, whereas the human studies focused on self-reported phenotypes that address participation in “excessive” physical activity. Both the animal QTL analyses as well as human studies were conducted only in females, and it remains unclear if the same QTL and genetic associations would be evident in males as well. Finally, it remains uncertain if the SNPs detected in human cohorts are biologically and behaviorally meaningful. Nonetheless, there is potentially robustness in the associations as they were observed despite all the limitations highlighted previously. Further studies are warranted to verify if the proposed genes or any other gene(s) or regions on chromosome 20 associate with activity traits in a systematic and repeatable fashion in both genders, in different age-groups and across different ethnicities.


In 2015, few articles were identified in the area of genomics in muscular strength and power; two were selected for presentation here based on our standards for review, and both are focused on the ACTN3 R577X nonsense polymorphism. In the first study, Broos et al. (8) examined the influence of ACTN3 R577X on muscle strength and endurance characteristics in 226 untrained young men (mean 21 yr of age). Consistent with expected genotype distribution patterns for this polymorphism in people of European descent (30), 20% of subjects exhibited the XX genotype and 31% the RR genotype. Several measures of muscle strength and fatigue were obtained. The results showed squat jump (−5%), countermovement jump (−5%), and handgrip strength (−6%) all to be significantly lower in the XX genotype carriers compared with RR carriers (all P < 0.05), as were concentric isokinetic knee torques (corrected for maximal isometric force) at both 100 and 200°·s−1 (both P < 0.05). Similarly, fatigue index was 4% lower in the XX compared with RR genotype group (P < 0.05), indicating greater fatigue resistance in the XX genotype carriers. No correction for multiple statistical testing was conducted, but the findings do match expectations from the performance literature showing lower levels of strength- and power-related performance in ACTN3 XX compared with RR genotype carriers (13). That said, ACTN3 association studies with direct measures of muscle function such as those used by Broos et al. have tended to show more mixed results (2,18,31).

In the second study, Kikuchi et al. (22) examined the association of ACTN3 R577X genotype with muscle function in 1227 Japanese men and women age 25–85 yr, 31% of whom were XX genotype carriers, which is typical of east Asian populations (30). A variety of muscle function tests, including handgrip strength, chair stand test, and 8-ft walking test, were administered. Analyses were performed within younger (<55 yr) and older (≥55 yr) age-groups and adjusted for age and exercise habits. The researchers observed significantly poorer chair stand performance in the older male XX genotype carriers compared with RX and RR genotypes (P = 0.036), with ACTN3 genotype contributing 2.5% of the observed variance. No significant genotype differences were noted in younger males or in women of any age for any measure. What is most striking is the general similarity of muscle function results across the groups; even the statistically significant difference in chair stand performance in older XX men is quite subtle and translates into a difference of approximately one stand in 30 s compared with the other genotype groups. These findings add to the corpus of mixed results observed in association studies of ACTN3 R577X genotype and muscle-related traits in untrained and older individuals (2,18,31).


In 2015, the number of articles on genetics and cardiorespiratory fitness or endurance performance phenotypes continued to decrease. One article using a genome-wide association study (GWAS) approach deserved to be retained. It was based on a combination of measured physiological phenotypes and replications using several case–control cohorts (1). Although the sample size is extremely small for a GWAS, we have chosen to highlight the article not because of its findings, as they are most likely to be false positive, but because of the research strategy used by the authors. Ahmetov et al. (1) first examined the association between 1,140,419 SNPs (HumanOmni1-Quad BeadChips; Illumina Inc., San Diego, CA) and V˙O2max in 80 elite endurance athletes (46 males and 34 females) from Russia, and they found 15 suggestive “endurance alleles.” Thereafter, they claimed that they were able to “replicate” six SNPs in at least one female and one male subgroup when comparing the frequencies of the six most significant SNPs (P < 10−5) between 218 endurance athletes and control cohorts (192 Russian controls, 1367 European controls, and 230 Russian power athletes). They reported that three SNPs (NFIA-AS2 rs1572312, TSHR rs7144481, and RBFOX1 rs7191721) were associated with endurance athlete status. The C allele of the most significant SNP, rs1572312, was associated with higher mean values of V˙O2max in all athletes (males: AC genotype = 55.3 ± 7.5, CC genotype = 64.4 ± 6.4 mL·kg−1·min−1, P = 0.0051; females: AA/AC genotype = 51.5 ± 4.5, CC genotype = 58.5 ± 4.7 mL·kg−1·min−1, P = 0.0005). Furthermore, the frequency (%) of the rs1572312 C allele was significantly higher in elite endurance athletes (95.5%) in comparison with nonelite endurance athletes (89.8%, P = 0.026), Russian (88.8%, P = 0.007) and European (90.6%, P = 0.02) controls, and power athletes (86.2%, P = 0.0005) (1). The rs1572312 SNP is located in the nuclear factor IA antisense RNA 2 (NFIA-AS2) gene locus. This gene regulates the expression of the NFIA gene, which is involved in hematopoiesis encoding transcription factors activating erythropoiesis and depressing granulopoiesis (1).

There are some serious limitations to this study, with the low sample size in the GWAS and the small number of cases (athletes) for the replication analysis being the most obvious ones. In addition, one can ask whether a classification of the athletes as elite based on their V˙O2max (mean ± SD for long endurance athletes: male = 70.4 ± 4.6 mL·kg−1·min−1, female = 58.9 ± 3.8 mL·kg−1·min−1; middle and short endurance athletes: male = 60.1 ± 5.4 mL·kg−1·min−1, female = 54.5 ± 6.3 mL·kg−1·min−1) is justifiable. Moreover, their measured V˙O2max values seem to be rather low for world caliber endurance athletes. It could be speculated that V˙O2max values in the range of 60–70 mL·kg−1·min−1 in males or 50–60 mL·kg−1·min−1 in females could be reached by an appropriate exercise training regimen without a strong genetic predisposition. Nonetheless, the overall approach of the study is very interesting, and it suggests that it could generate useful information if it could be expanded to multi-center and perhaps international efforts.

One article with potential clinical relevance was published by Banting et al. (4). They investigated whether elite endurance athletes also have a genetic profile favoring a healthy metabolic risk factor profile. Five genetic markers (IGF2BP2 rs4402960, LPL rs320, LPL rs328, KCJN rs5219, and MTHFR rs1801133) known to be associated with obesity and its co-morbidities were compared between male endurance athletes (N = 254), power athletes (N = 207), and sedentary controls (N = 544). They found that male elite endurance athletes are less likely to carry metabolic risk alleles, specifically at IGF2BP2 rs4402960 and LPL rs320. Sprint/power athletes were twice as likely than the endurance athletes to have the IGF2BP2 rs4402960 TT genotype (associated with increased risk for diabetes and obesity) compared with the GT or GG genotypes (odds ratio = 2.00, 95% confidence interval [CI] = 1.01–3.95, P = 0.045), whereas sedentary controls were 022% more likely than endurance athletes to carry the LPL rs320 risk-related G-allele compared with the TT genotype (odds ratio = 1.22, 95% CI = 1.03–1.46, P = 0.024) (4). The authors claim that some genetic variants may have pleiotropic effects, predisposing to endurance performance but also to a reduced risk of developing metabolic morbidities. This would be compatible with the favorable longevity outcome observed in former endurance athletes compared with other athletes as well as compared with the general population (15). However, we are dealing here with extremely complex genotype–phenotype relationships. Moreover, as is well known, complex metabolic traits and their response to training are polygenic traits affected by large numbers of genes and DNA variants, which are unlikely to have been adequately covered by the five candidate genes tested by the authors.

In summary, the exercise genomics of cardiorespiratory endurance phenotypes did not experience much progress in 2015. Innovative approaches and more powerful and comprehensive efforts are needed.


In 2015, two studies that examined the interaction between physical activity and genetic susceptibility to obesity stood out (16,24). In the first study, Klimentidis et al. (24) examined whether physical inactivity influences the association between variation in FTO and BMI in 12,074 individuals from the Framingham Heart Study (FHS) (NOffspring = 3430; NThird Generation = 3888 all of European ancestry) and from the Women’s Health Initiative (WHI) (N = 1542 of Hispanic ancestry [HA]; N = 3214 of African ancestry [AA]). Analyses did not include WHI participants of European ancestry. Physical inactivity, here assessed by “time spent sitting” (TSS), has been proposed as a risk factor for cardiometabolic disease, independent from physical activity (37). The authors chose to study the FTO rs9939609 variant, which shows robust associations with BMI and other obesity traits in individuals of European ancestry, whereas associations for this variant in AA populations have been less consistent (27). Associations were significant between FTO rs9939609 and BMI in the FHS and also in WHI participants of HA, but not of AA. Interaction analyses showed that in the FHS, the association between FTO rs9939609 and BMI (Pinteraction = 0.0034) was greater in participants with more TSS compared with those with less TSS (Table 1). However, in the AA and HA participants of the WHI, the interaction tended to be in the opposite direction (Pinteraction = 0.02); i.e., the association between FTO rs9939609 and BMI was attenuated in participants with more TSS (Table 1).

Association of FTO rs9939609 with BMI in three categories of TSS.

Thus, increased physical inactivity assessed by TSS exacerbates the BMI-increasing effect of FTO rs9939609 in the FHS, in which all participants are of European ancestry. This is consistent with a previous study in which “television watching” was used as a proxy for physical inactivity (32). These observations on physical inactivity, combined with those previously reported on other lifestyle factors (23,33), such as physical activity and dietary intake, provide further support for the interaction between variation in FTO and lifestyle on obesity risk. Specifically, an unhealthy lifestyle seems to exacerbate the effect of FTO variation on obesity risk, at least in individuals from European ancestry. The reason why associations and interactions were weaker and even in opposite direction in the AA and HA populations may be due to the fact that FTO rs9939609 has been reported to show inconsistent results in AA populations (27). Other variants in FTO (e.g., rs1781794) have shown more convincing and consistent association with obesity traits in individuals of AA (27). It should also be noted that the BMI-increasing allele of FTO rs9939609 was associated with increased TSS (P < 0.001) in the FHS, and a similar trend was observed in the HA and AA of the WHI (P = 0.08). The authors performed mediation analysis and found that the association of rs9939609 with BMI is partially mediated by TSS, but that BMI also partially mediates the association between rs9939609 and TSS. Thus, disentangling the causal nature of these associations is a difficult task. This is a concern as interaction results may be biased when an SNP is associated with both the outcome (obesity risk) and the environment (TSS) (12).

Dudbridge and Fletcher describe how gene–environment (e.g., behavior) dependence can result in spurious gene–environment interactions (12). Specifically, they describe how dependence between gene and environment can arise through mediation, pleiotropy, and confounding (Fig. 1). The association of a gene with an environmental risk factor is often thought to imply that the environmental factor mediates the effect of the gene on the outcome of interest (Fig. 1A). However, the gene may have pleiotropic effects on both the environmental factor and the outcome, but the environmental factor may not cause the outcome (Fig. 1B). Lastly, gene–environment dependence can occur through confounding (Fig. 1C). Thus, a statistical interaction can be present between an SNP and an environmental marker (e.g., TSS) even if there is no biological interaction between the SNP and the environment (12). Thus, closely examining how genes and environmental factors relate to each other is crucial when analyzing and interpreting gene–environment interactions.

Gene–environment dependence by (A) mediation, (B) pleiotropy, and (C) confounding. E, environmental factor; C, unmeasured confounding factor; Common causes, causes of both environmental factor and outcome of interest. A, The environmental factor mediates the effect of the SNP on the outcome. B, The SNP is associated with both the environmental factor and the outcome, but the environmental factor is not directly associated with the outcome. Adapted from Dudbridge and Fletcher (12).

In a second study, Guo et al. (16) examined how the heritability of BMI differs according to physical activity in 8738 individuals from the FHS. Heritability was assessed using a mixed-model approach that assesses the contribution of genome-wide variation to BMI (39). The genomic influence on BMI among the 21- to 50-yr-old subgroup was significantly more pronounced in physically inactive compared with active individuals (Fig. 2), whereas no difference in genomic contribution was observed in the >50-yr-old subgroup (16).

Two normal curves representing the estimated random genomic effects of SNPs on BMI between physically active and inactive adults age 21–50 yr. A flatter curve indicates that more random effects are much larger than zero. The random effects of SNPs on BMI are substantially larger among those not engaged in physical activity than those engaged in physical activity (F value = 11.3, P < 0.00001). From Guo et al. (16) with permission.

Although there is growing evidence that physical activity is beneficial even in those who are genetically susceptible to obesity, so far, the FTO locus is the only one showing convincing interactions. Large-scale genome-wide interaction analyses will be needed to reveal additional loci that are sensitive to environmental influences.


In 2015, only two candidate gene studies have addressed the issue of gene–physical activity interactions for insulin and glucose metabolism phenotypes, and we consider that only one met our standards for this review (40).

The modifying effects of the PPARG Pro12Ala polymorphism on the association of moderate-to-vigorous–intensity physical activity (MVPA) and sedentary time behavior with markers of insulin sensitivity was investigated in 541 subjects at increased risk of type 2 diabetes (40). Subjects underwent an oral glucose tolerance test, and fasting and 2-h postchallenge glucose and insulin levels were used to calculate two measures of insulin sensitivity commonly used in epidemiological studies: the HOMA-IS and the Matsuda-ISI. The subjects were also asked to wear a triaxial accelerometer for a minimum of seven consecutive days during waking hours, and the number of counts per 15 s was used to categorize time spent in sedentary activities (<25 counts) and in MVPA (≥505 counts). Significant genotype–sedentary time interactions were not observed for either measure of insulin sensitivity. However, a significant genotype–MVPA interaction was observed with Matsuda-ISI (P = 0.005), but not for HOMA-IS. MVPA was strongly associated with insulin sensitivity in Ala12 carriers, every additional 30 min spent in MVPA being associated with a 69% higher Matsuda-ISI (P = 0.007), whereas in Pro12 allele homozygotes, the association was not significant. This candidate gene study, although based on a moderate sample size, is of particular interest because it is based on objective measurements of physical activity and on common insulin sensitivity measures derived from an oral glucose tolerance test.


In the 2015 calendar year, two interesting studies were published related to gene–exercise or gene–physical activity interactions on lipid and lipoprotein phenotypes. The first by Sarzynski et al. (34) integrated genomic and transcriptomic data to create a composite SNP-based predictor of the response of triglycerides (TG) to exercise training (△TG). The authors performed global RNA profiling of skeletal muscle taken at baseline from a subset of subjects from the HERITAGE Family Study (N = 49). They found that an 11-gene model predicted 27% of the variance in △TG, which was validated in an independent study. Next, the association of almost 500 SNPs in or near the 11 genes with △TG was tested in white HERITAGE subjects (N = 481), and SNPs in four genes were found to be nominally associated with △TG.

Sarzynski et al. (34) also performed a GWAS of △TG in HERITAGE whites (N = 478). The authors found 39 SNPs associated at P < 1 × 10−4, but none reached genome-wide significance (Bonferroni-corrected threshold of 1.8 × 10−7). In conditional heritability analyses, four SNPs were able to statistically account for all of the heritability of △TG, with these four SNPs each explaining 3.7% to 5.5% of the variance in a multivariate regression model. The authors selected these four SNPs to be used in an SNP summary score.

Thus, the authors created a combined SNP summary score from eight SNPs, four SNPs from the transcriptomic analysis (NSA2 rs1043968, SLC4A2 rs3793336, MACROD1 rs594461, and EEF2K rs11646610) and four SNPs from the GWAS analysis (CYYR1 rs222158, GLT8D2 rs2722171, RBFOX1 rs1906058, and ZNF385D rs2593324). Subjects with seven or more favorable alleles (e.g., allele associated with a decrease in TG with training) across the eight SNPs showed an adjusted mean decrease in TG with exercise training, whereas those with six or less favorable alleles had an adjusted mean increase in TG (Fig. 3). In regression models that included more than 50 baseline predictor variables, the SNP summary score was the strongest predictor of △TG, explaining 14% of the variance (approximately 8% from four GWAS SNPs, approximately 6% from four RNA predictor SNPs), whereas baseline TG explained 7.4% of the variance (34).

Adjusted mean response to exercise training triglycerides (ΔTG) across eight SNP summary score categories in HERITAGE Caucasian subjects. Values were adjusted for age, sex, baseline BMI, and baseline TG level. The number of participants within each SNP score category is indicated inside each histogram bar. From Sarzynski et al. (34) with permission.

This study (34) highlights how integrating data from multiple “omics” technologies may lead to the identification of more robust predictors of trait response to regular exercise. Although multiple genes and variants were associated with TG response to exercise training, the authors did not find any associations with the existing population-based TG GWAS loci. This suggests that the genes involved in modifying TG response to regular exercise are likely different from the genes contributing to variation in population TG levels (34). The study is limited by the fact that the SNP summary score was tested in the same cohort used to create the score. Thus, the results are likely overestimating the true strength of the associations. The individual variants and SNP scores now need to be tested in independent studies to examine whether these associations can be replicated and to potentially identify other predictors of TG response to exercise. Furthermore, replication studies are needed to examine whether these results apply to other modes and doses of endurance exercise.

In the second study, Justesen et al. (20) examined whether lifestyle factors modified the association of lipid-based genetic risk scores (GRS) with serum lipid traits in two population-based studies: the Inter99 study (N = 5961) serving as the discovery cohort and the Health2006 study (N = 2565) as the replication cohort. Specifically, the authors created weighted GRS from the SNPs reaching genome-wide significance (P < 5 × 10−8) in the latest GWAS meta-analysis for the following four traits: total cholesterol (N = 74 SNPs), low-density lipoprotein cholesterol (LDL-C, N = 58 SNPs), HDL-C (N = 71 SNPs), and TG (N = 39 SNPs). Each GRS was strongly associated with its corresponding trait in both the discovery and replication cohorts (P ≤ 5 × 10−25). However, only the HDL-C GRS showed a significant interaction with physical activity level in the Inter99 cohort (P = 0.00023), with the genetic effect attenuated in more active individuals (interaction effect, β = −0.0015 (0.0004)). Furthermore, the overall effect size of physical activity was small, and the interaction between HDL-C GRS and physical activity was not replicated in the Health2006 cohort (P = 0.86). The study by Justesen et al. (20) benefits from its large sample size and the inclusion of the latest population-based lipid GWAS SNPs. However, the study is limited by its cross-sectional and observational nature and the use of self-reported measures of physical activity. Larger, well-powered studies, including objective measures of physical activity and perhaps even genome-wide tests for interactions, may provide more insight into the modifying effects of physical activity on genetic predisposition for elevated serum lipid levels (20).


Few genetic association studies were published in 2015 on the relationship between acute or chronic exercise and hemodynamic traits or cardiovascular outcomes, and most used small sample sizes (N < 100). Despite the lack of adequately powered studies, some are worth mentioning because they can help understand the genetic basis of i) important cardiovascular outcomes and ii) human variability in the response of these outcomes to exercise interventions.

A recent meta-analysis of candidate gene association studies of the blood pressure response to acute and chronic exercise included data from 12 aerobic exercise training studies examining up to 11 candidate gene polymorphisms (9). Participants (total N = 3444; mean ± SD) were non-Hispanic middle-age adults (44 ± 11 yr) of both genders with prehypertension (systolic blood pressure = 134.7 ± 11.8 mm Hg; diastolic blood pressure [DBP] = 78.5 ± 9.6 mm Hg) after a “standard” moderately intense (~60% V˙O2peak) aerobic training program (three to four sessions per week of ~40 min per session, for an average of ~4 months). Of the polymorphisms examined (ACE rs4340, AGT rs699, CHRM2 rs324640 and rs8191992, CYBA rs4670 and rs1049255, GNB3 rs5443, IL6R rs2228145, LPL rs328, and TGFβ1 rs1800470), only the angiotensinogen (AGT) Met235Thr (rs699) variant was associated with the change in DBP (post- vs pretraining, P = 0.05) and explained only <1% of the variance in the DBP response to exercise (9). Reductions in DBP were greater among people with the AGT MM genotype (n = 174) (standardized mean effect size −0.56 [95% CI = −0.770 to −0.347], −6.2 mm Hg) than among people with the AGT TT genotype (n = 179) (effect size −0.30 [95% CI = −0.515 to −0.076], −3.9 mm Hg) (P = 0.04), but this association did not remain statistically significant after adjustment for multiple comparisons with a corrected alpha level of P = 0.02 (9). Of note, ∼41%–46% of the variance in the training adaptation of the postacute exercise hypotension response was not explained by factors related to clinical features, which may be explained by genetic factors or by factors not related to clinical features, which certainly remain to be determined. This is an important finding because BP is typically reduced (by 5–7 mm Hg) up to 24 h after exposure to acute aerobic exercise (the so-called postexercise hypotension) in healthy humans, and reductions of the same magnitude are often observed after training in the resting BP of hypertensive individuals. Such BP reductions are comparable with those elicited by common antihypertensive drugs. In adverse responders (6), regular exercise is unlikely to have such beneficial effects on elevated BP, but this remains to be investigated. Well-powered studies are needed to unveil the genetic factors that predict the individual response of BP and of other hemodynamic traits to an exercise intervention.

Identifying the genetic factors associated with potentially “adverse response” to sustained strenuous endurance exercise, even in people without known cardiovascular disease risk factors, is also important, especially when considering i) the growing number of people participating in strenuous endurance events (e.g., marathons) and ii) the increasing controversy regarding the potential “cardiotoxic” effects of long-term strenuous endurance exercise even in previously healthy hearts, mainly unsuitable cardiac remodeling, and increased risk of atrial fibrillation. In this regard, the provocative concept of “excessive endurance exercise” has been introduced (26). A matter of major concern is potential strenuous endurance exercise-induced dysfunction of the right ventricle (RV), as indicated by the presence of fibrotic patches with cardiac magnetic resonance imaging in long-term exercisers. Although underpowered, a recent study with 126 elite Hungarian athletes of mixed disciplines who underwent cardiac magnetic resonance imaging revealed that all showed the typical feature of the athlete’s heart, that is, increased stroke volume with eccentric hypertrophy (36). They also observed that carriers of the Asp allele of the Glu298Asp (rs1799983) polymorphism in the nitric oxide synthase 3 (NOS3) gene had higher values of RV mass (32 ± 6 vs 27 ± 6 g, P < 0.01) and RV stroke volume index (71 ± 10 vs 64 ± 10 mL, P < 0.01) compared with the Glu/Glu genotype. This association was not found in untrained controls (36). Further studies are clearly warranted to identify all DNA variants that may predispose to adverse RV remodeling.

Animal model studies can help gain mechanistic insight into the biological pathways associated with strenuous endurance exercise-induced “cardiotoxicity.” A recent mouse study (3) showed that 6 wk of endurance training increased vulnerability to atrial fibrillation in association with inflammation and fibrosis, with microarray results suggesting the involvement of a major inflammatory cytokine, tumor necrosis factor α (TNF-α), in exercise-induced atrial remodeling. Exercise training induced the TNF-α–dependent activation of both nuclear factor kappa light-chain enhancer of activated B cells (NFκB) and p38 mitogen-activated protein kinase, whereas the TNF-α inhibition (e.g., with TNFA gene ablation) prevented training-induced atrial structural remodeling and atrial fibrillation vulnerability, without affecting the beneficial physiological changes (3). Thus, future human candidate gene studies addressing the association between atrial fibrillation risk and long-term strenuous endurance events might target gene variants related to TNF-α, nuclear factor kappa light-chain enhancer of activated B cells, or mitogen-activated protein kinase pathways.

Although pulmonary arterial hypertension (PAH) is not frequently studied in exercise sciences, it represents an interesting model to understand the interplay between exercise and hemodynamic traits. Patients typically experience dyspnea and marked exercise intolerance because of RV failure and limited cardiorespiratory fitness, with V˙O2max or other cardiopulmonary exercise testing–derived variables being predictors of survival. Several randomized controlled trials have shown the feasibility and benefits of moderate-intensity aerobic exercise for improving the V˙O2max of these patients (10). There are also serum biomarkers of disease severity/mortality such as the N-terminal pro-brain natriuretic peptide or endostatin, a 20-kDa C-terminal fragment of type XVIII collagen, which is one of the most potent inhibitors of angiogenesis. Importantly, a loss-of-function, missense variant in the gene encoding endostatin (Col18a1) is linked to lower circulating endostatin levels and independently associated with reduced mortality in PAH patients (11). Thus, both the Col18a1 gene and its product are potential new therapeutic targets in PAH and possibly in other cardiopulmonary conditions associated with poor physical capacity. In fact, lower levels of endostatin might be a sign of increased angiogenesis and decreased inflammation and calcifying activity in physically active patients with cardiovascular disease compared with their inactive peers, and it might contribute to the damaging effect of physical inactivity in the human cardiovascular system (35). Whether DNA sequence variation contributes to the benefits of regular exercise in PAH patients is not known and represents an understudied area.

Exercise reduces cardiovascular-related mortality after acute myocardial infarction (AMI) and improves LV regeneration (19). Although several mechanisms have been postulated to support the anti-remodeling benefits of exercise training after AMI (improved endothelial function, myocardial contractility and autonomic balance, or lower systolic/diastolic wall stress), more research is needed. A major challenge is to unveil the molecular mechanisms that distinguish pathological from physiological cardiac hypertrophy. In this regard, the role of micro-RNAs (miRNAs) on the exercise benefits after AMI remains to be elucidated. miRNAs are small RNAs encoded in the genome that play key roles in a myriad of cellular processes and body functions, including cardiac function under physiological and pathological conditions. A recent review compiled the recent discoveries on miRNAs and the cardioprotective effects of endurance exercise (14). Studies in the heart have identified several anti- and prohypertrophic miRNAs, four of which are recognized as cardiac specific (and thus called “myomiRs”), being either antihypertrophic (miRNA-1 and miRNA-133a/b) or prohypertrophic (miRNA-208a/b and iRNA-499). A recent study evaluated the effect of exercise training on cardiac miRNA-1 and miRNA-214 expression after MI in rats (29). The results indicated that exercise training restored miRNA-1 and miRNA-214 expression levels, which was associated with the normalization of calcium handling and LV compliance in infarcted hearts after training. Thus, it would be of medical relevance to corroborate whether these miRNAs also have a positive effect on tissue regeneration and recovery of calcium handling capacity in patients post-MI and whether exercise restores the expression of these miRNAs in humans. In addition, it would be important to explore human DNA sequence variation in the coding frames of these miRNAs and the genomic elements to which they functionally bind.


Although only a small number of articles related to exercise genomics published in 2015 met our standards for inclusion, they include several articles that investigated gene–physical activity interactions, as well as studies that used creative study designs. In general, the gene–physical activity interaction papers showed that the genomic influences on cardiometabolic traits were more pronounced in inactive individuals and attenuated in more active people. For example, one study showed that physical inactivity exacerbated the BMI-increasing effect of an FTO variant in individuals of European ancestry, whereas another study showed that HDL-C GWAS loci exerted a smaller effect on plasma HDL-C in active individuals. Among the noteworthy reports was a study by Kostrzewa et al. (25), whereby a QTL of wheel running in mice led to the identification of SNPs in chromosome 20 associated with physical activity levels in humans. This shows how animal models could be used to identify candidate genomic regions for physical activity traits in humans. One of us (MAS) led a study (34) combining genome-wide and skeletal muscle transcriptome-wide profiling to identify predictors of TG response to exercise training. A global gene expression signature and a key GWAS SNPs were used to create an 8-SNP summary score that explained 14% of the variance in TG response to training. This study demonstrates how integrating data from omics technologies may improve our ability to identify predictors of exercise response.

DTC genetic testing

The ability (or lack thereof) to identify predictors of exercise or sports performance is an important issue related to exercise genomics, particularly with the proliferation of DTC genetic testing products over the past several years. DTC tests marketed toward sport or exercise performance are aimed at individuals, coaches, parents, athletes, sports teams, and sports medicine professionals and propose to guide talent development and screen people for specific sports and athletic events, particularly children and adolescents. Furthermore, several of these DTC products claim they can help personalize the design of training programs around an individual’s genetic profile to maximize results and “overcome” their genetic predispositions or limitations. For example, despite uncertainties in the large body of related literature, one of the most commonly included gene variants in exercise- or sport-related DTC tests is ACTN3 R577X, with a recent review even boldly outlining “effective utilization” of ACTN3 genotype information for physical training (21). The problem of the validity of DTC genetic tests is compounded by the fact that several companies are not even disclosing what variants are being tested. The entire area of DTC genetic testing for exercise and sports applications is plagued by a score of scientific, ethical, and legal issues, as recently summarized in a consensus report on DTC genetic testing for predicting sports performance and talent identification (38). The consensus paper contends that the scientific evidence for the predictive value of such tests, and therefore their usefulness in the context of training responses or talent identification in sport is virtually zero (38). Moreover, the results of the DTC genetic tests may have unintended consequences such as undesirable early sport specialization and reduced opportunities for athletic pursuits of youth, not to mention their potential psychosocial and financial implications. Recently, experts in exercise genomics have weighed in on this important topic and provide clear arguments against the current use of genetic testing for athletic performance, especially in children (17,38).

Conceptual framework

The strategy of combining omics with physiological and/or behavioral data seems like a very reasonable approach to increase our understanding of the genes, pathways, and networks contributing to human responsiveness to changes in physical activity behavior and exercise training. To better conceptualize this global approach, we have developed the conceptual framework shown in Figure 4. It aims at providing a general overview of aspects of designs, requirements, and technologies that could enhance exercise genomics studies along the research continuum from discovery to translation. The conceptual framework provides general indications on basic sample size requirements and technologies for discovery and replication phases based on whether rare, common, or candidate variants are being examined. Regardless of the variant type, the validation phase requires at a minimum broad bioinformatics analyses along with in silico, cell, and animal model experiments (Fig. 4). The translation of the findings is obviously a difficult step, but the framework calls mainly for strong randomized controlled trials using Mendelian randomization, with the goal of eventually applying the results in personalized exercise medicine situations. We emphasize that the framework represents a series of suggestions and is not meant to be comprehensive, as it does not capture all of the various study designs and technologies emerging in the field.

Simplified conceptual framework for exercise genomics studies. KO, knockout; med, medicine; RCT, randomized controlled trial.

The concepts represented within the framework as they relate to exercise genomics studies are not new. However, several barriers, as previously detailed in reports from our group and others (5,7,28), have prevented the exercise genomics field from using this integrative approach on a large scale. To improve the quality of science in exercise genomics, it has been suggested that the field no longer rely heavily on observational study designs and poorly justified candidate genes with small sample sizes (7). Instead, it was suggested that exercise genomics research be mostly composed of experimental and mechanistic studies, with replication studies being a critical component. To this end, collaborative research and data sharing would be necessary to produce the large resources needed for such an undertaking. However, lack of substantial funding may be the biggest barrier to such an enterprise, as randomized controlled exercise trials with standardized study designs and measures along with high-throughput technologies and robust computational and bioinformatics tools are costly and beyond the means of any one research site/team.


Recently, the U.S. NIH Common Fund Program titled “Molecular Transducers of Physical Activity in Humans” (MoTrPAC; was launched. This ambitious program will award at least US$170 million in grants for 6 yr to extensively catalog the biological molecules affected by acute and chronic exposures to physical activity (endurance and resistance) in humans, to identify some of the key molecules that underlie the systemic effects of physical activity, and to characterize the function of these key molecules. MoTrPAC will also develop a publically available database that any researcher can access to develop hypotheses related to the molecular mechanisms whereby physical activity improves or preserves health. MoTrPAC will enroll approximately 3000 participants in various physical activity studies; will collect blood, adipose, and muscle tissue samples under a variety of activity conditions; will support genomic, epigenomic, transcriptomic, proteomic, and metabolomic profiling in blood, adipose, and muscle samples under acute and chronic endurance or resistance physical activity exposures; will fund animal model studies of activity to obtain the same information on tissues and organs not accessible in humans; will incorporate additional animal model studies aimed at mechanistic questions; will provide comprehensive biostatistics and bioinformatics expertise; and will accumulate, verify, and store in a central repository all relevant data to be shared with other investigators interested in pursuing questions not directly addressed by the MoTrPAC teams of scientists.

The implications of MoTrPAC for exercise genomics are nothing but extraordinary. The research program will create a wealth of data kept in a central repository to which the international scientific community will have access for years and decades. MoTrPAC offers an opportunity to better understand at the molecular level the mechanisms responsible for the physiological and metabolic changes that are observed in response to acute and chronic exposure to endurance exercise and resistance exercise regimens. It will generate a wealth of information that could be used to investigate the molecular basis for human variability in responsiveness to exercise regimens in pediatric and adult populations, men and women, while taking account of ethnicity as well. There is little doubt that the program represents one large leap forward toward the goal of personalized exercise medicine.

In summary, the exercise genomics field is poised for substantial progress in the next decade, particularly with the advent of the NIH Common Fund MoTrPAC program and the continuing development and decreasing cost of high-throughput technologies. To keep pace with the growing field and advancing technologies and be able to take full advantage of the new opportunities, more computational biology and bioinformatics expertise is needed in the exercise science and sports medicine enterprise.

The authors thank Dr. James M. Hagberg for his contributions to the Human Gene Map for Performance and Health-related Fitness Phenotypes and subsequently to the Advances in Exercise, Fitness, and Performance Genomics series for the past decade. CB is partially funded by the John W. Barton Sr. Chair in Genetics and Nutrition. MAS is a consultant to Genetic Direction.

The others do not have conflicts of interest to disclose.

The results of the review do not constitute endorsement by the American College of Sports Medicine.


1. Ahmetov I, Kulemin N, Popov D, et al. Genome-wide association study identifies three novel genetic markers associated with elite endurance performance. Biol Sport. 2015;32:3–9.
2. Alfred T, Ben-Shlomo Y, Cooper R, et al. ACTN3 genotype, athletic status, and life course physical capability: meta-analysis of the published literature and findings from nine studies. Hum Mutat. 2011;32:1008–18.
3. Aschar-Sobbi R, Izaddoustdar F, Korogyi AS, et al. Increased atrial arrhythmia susceptibility induced by intense endurance exercise in mice requires TNFalpha. Nat Commun. 2015;6:6018.
4. Banting LK, Pushkarev VP, Cieszczyk P, et al. Elite athletes’ genetic predisposition for altered risk of complex metabolic traits. BMC Genomics. 2015;16:25.
5. Bouchard C. Overcoming barriers to progress in exercise genomics. Exerc Sports Sci Rev. 2011;39(4):212–7.
6. Bouchard C, Blair SN, Church TS, et al. Adverse metabolic response to regular exercise: is it a rare or common occurrence? PLoS One. 2012;7:e37887.
7. Bouchard C. Exercise genomics-a paradigm shift is needed: a commentary. Br J Sports Med. 2015;49:1492–96.
8. Broos S, Van Leemputte M, Deldicque L, Thomis MA. History-dependent force, angular velocity and muscular endurance in ACTN3 genotypes. Eur J Appl Physiol. 2015;115:1637–43.
9. Bruneau ML Jr, Johnson BT, Huedo-Medina TB, Larson KA, Ash GI, Pescatello LS. The blood pressure response to acute and chronic aerobic exercise: a meta-analysis of candidate gene association studies. J Sci Med Sport. 2015;19(5):424–31.
10. Buys R, Avila A, Cornelissen VA. Exercise training improves physical fitness in patients with pulmonary arterial hypertension: a systematic review and meta-analysis of controlled trials. BMC Pulm Med. 2015;15:40.
11. Damico R, Kolb TM, Valera L, et al. Serum endostatin is a genetically determined predictor of survival in pulmonary arterial hypertension. Am J Respir Crit Care Med. 2015;191:208–18.
12. Dudbridge F, Fletcher O. Gene–environment dependence creates spurious gene–environment interaction. Am J Hum Genet. 2014;95:301–7.
13. Eynon N, Hanson ED, Lucia A, et al. Genes for elite power and sprint performance: ACTN3 leads the way. Sports Med. 2013;43:803–17.
14. Fernandes T, Barauna VG, Negrao CE, Phillips MI, Oliveira EM. Aerobic exercise training promotes physiological cardiac remodeling involving a set of microRNAs. Am J Physiol Heart Circ Physiol. 2015;309:H543–52.
15. Garatachea N, Santos-Lozano A, Sanchis-Gomar F, et al. Elite athletes live longer than the general population: a meta-analysis. Mayo Clin Proc. 2014;89:1195–1200.
16. Guo G, Liu H, Wang L, Shen H, Hu W. The genome-wide influence on human BMI depends on physical activity, life course, and historical period. Demography. 2015;52:1651–70.
17. Guth LM, Roth SM. Genetic influence on athletic performance. Curr Opin Pediatr. 2013;25:653–8.
18. Hagberg JM, Rankinen T, Loos RJ, et al. Advances in exercise, fitness, and performance genomics in 2010. Med Sci Sports Exerc. 2011;43(5):743–52.
19. Haykowsky M, Scott J, Esch B, et al. A meta-analysis of the effects of exercise training on left ventricular remodeling following myocardial infarction: start early and go longer for greatest exercise benefits on remodeling. Trials. 2011;12:92.
20. Justesen JM, Allin KH, Sandholt CH, et al. Interactions of lipid genetic risk scores with estimates of metabolic health in a Danish population. Circ Cardiovasc Genet. 2015;8:465–72.
21. Kikuchi N, Nakazato K. Effective utilization of genetic information for athletes and coaches: focus on ACTN3 R577X polymorphism. J Exerc Nutrition Biochem. 2015;19:157–64.
22. Kikuchi N, Yoshida S, Min SK, et al. The ACTN3 R577X genotype is associated with muscle function in a Japanese population. Appl Physiol Nutr Metab. 2015;40:316–22.
23. Kilpelainen TO. Physical activity attenuates the influence of FTO variants on obesity risk: a meta-analysis of 218,166 adults and 19,268 children. PLoS Med. 2011;8:e1001116.
24. Klimentidis YC, Arora A, Chougule A, Zhou J, Raichlen DA. FTO association and interaction with time spent sitting. Int J Obes (Lond) [Internet]. 2015. Available from: doi:10.1038/ijo.2015.190.
25. Kostrzewa E, Brandys MK, van Lith HA, Kas MJ. A candidate syntenic genetic locus is associated with voluntary exercise levels in mice and humans. Behav Brain Res. 2015;276:8–16.
26. Lavie CJ, O’Keefe JH, Sallis RE. Exercise and the heart—the harm of too little and too much. Curr Sports Med Rep. 2015;14:104–9.
27. Loos RJ, Yeo GS. The bigger picture of FTO—the first GWAS-identified obesity gene. Nat Rev Endocrinol. 2014;10:51–61.
28. Loos RJ, Hagberg JM, Perusse L, et al. Advances in exercise, fitness, and performance genomics in 2014. Med Sci Sports Exerc. 2015;47(6):1105–12.
29. Melo SF, Barauna VG, Neves VJ, et al. Exercise training restores the cardiac microRNA-1 and -214 levels regulating Ca(2+) handling after myocardial infarction. BMC Cardiovasc Disord. 2015;15:166.
30. Mills MA, Yang N, Weinberger RP, et al. Differential expression of the actin-binding proteins, a-actinin-2 and -3, in different species: implications for the evolution of functional redundancy. Hum Mol Genet. 2001;10:1335–46.
31. Perusse L, Rankinen T, Hagberg JM, et al. Advances in exercise, fitness, and performance genomics in 2012. Med Sci Sports Exerc. 2013;45(5):824–31.
32. Qi Q, Li Y, Chomistek AK, et al. Television watching, leisure time physical activity, and the genetic predisposition in relation to body mass index in women and men. Circulation. 2012;126:1821–7.
33. Qi Q, Kilpelainen TO, Downer MK, et al. FTO genetic variants, dietary intake and body mass index: insights from 177 330 individuals. Hum Mol Genet. 2014;23:6961–72.
34. Sarzynski MA, Davidsen PK, Sung YJ, et al. Genomic and transcriptomic predictors of triglyceride response to regular exercise. Br J Sports Med. 2015;49:1524–31.
35. Sponder M, Fritzer-Szekeres M, Marculescu R, Litschauer B, Strametz-Juranek J. Physical inactivity increases endostatin and osteopontin in patients with coronary artery disease: Heart Vessels. 2015. Epub ahead of print.
36. Szelid Z, Lux A, Kolossvary M, et al. Right ventricular adaptation is associated with the Glu298Asp variant of the NOS3 gene in elite athletes. PLoS One. 2015;10:e0141680.
37. Thorp AA, Owen N, Neuhaus M, Dunstan DW. Sedentary behaviors and subsequent health outcomes in adults a systematic review of longitudinal studies, 1996–2011. Am J Prev Med. 2011;41:207–15.
38. Webborn N, Williams A, McNamee M, et al. Direct-to-consumer genetic testing for predicting sports performance and talent identification: consensus statement. Br J Sports Med. 2015;49:1486–91.
39. Yang J, Benyamin B, McEvoy BP, et al. Common SNPs explain a large proportion of the heritability for human height. Nat Genet. 2010;42:565–9.
40. Yates T, Davies MJ, Henson J, et al. Effect of the PPARG2 Pro12Ala polymorphism on associations of physical activity and sedentary time with markers of insulin sensitivity in those with an elevated risk of type 2 diabetes. PLoS One. 2015;10:e0124062.


© 2016 American College of Sports Medicine