Exome-Wide Association Study of Competitive Performance in Elite Athletes

im of the study was to identify genetic variants associated with personal best scores in Turkish track and eld athletes and to compare allelic frequencies between sprint/power and endurance athletes and controls using a whole-exome sequencing (WES) approach, followed by replication studies in independent cohorts. The discovery phase involved 60 elite Turkish athletes (31 sprint/power and 29 endurance) and 20 ethnically matched controls. The replication phase involved 1132 individuals (115 elite Russian sprinters, 373 elite Russian endurance athletes (of which 75 athletes were with VO 2maxmeasurements), 209 controls, 148 Russian and 287 Finnish individuals with muscle ber composition and cross-sectional area (CSA) data). None of the single nucleotide polymorphisms (SNPs) reached an exome-wide signi cance level (p< 2.3 10 7 ) in genotype–phenotype and case–control studies of Turkish athletes. However, of the 53 nominally (p< 0.05) associated SNPs, four functional variants were replicated. TheSIRT1rs41299232 G allele was signi cantly over-represented in Turkish (p= 0.047) and Russian (p= 0.018) endurance athletes compared to sprint/power athletes and was associated with increased VO 2max(p= 0.037) and a greater proportion of slow-twitch muscle bers (p= 0.035). TheNUP210rs2280084 A allele was signi cantly over-represented in Turkish (p= 0.044) and Russian (p= 0.012) endurance athletes compared to sprint/power athletes. TheTRPM2rs1785440 G allele was signi cantly over-represented in Turkish endurance athletes compared to sprint/power athletes (p= 0.034) and was associated with increased VO 2max(p= 0.008). TheAGRNrs4074992 C allele was signi

a greater proportion of slow-twitch muscle bers (p= 0.035). TheNUP210rs2280084 A allele was signi cantly over-represented in Turkish (p= 0.044) and Russian (p= 0.012) endurance athletes compared to sprint/power athletes. TheTRPM2rs1785440 G allele was signi cantly over-represented in Turkish endurance athletes compared to sprint/power athletes (p= 0.034) and was associated with increased VO 2max(p= 0.008). TheAGRNrs4074992 C allele was signi cantly over-represented in Turkish sprint/power athletes compared to endurance athletes (p= 0.037) and was associated with a greater CSA of fast-twitch muscle bers (p= 0.024). In conclusion, we present the rst WES study of athletes showing that this approach can be used to identify novel genetic markers associated with exercise- and sport-related phenotypes. Genes2023,14, 660.

Genes2023,14, 660 2 of 11 Keywords: athletic performance; athletics; track and eld; athletes; aerobic capacity; muscle hyper- trophy; sports genetics; WES; EWAS; GWAS 1. Introduction Whether pure talent or long-term experiences promotes athletic performance is one of the questionable issues [1]. Progression in the sport sciences has underlined that athletic performance was a phenomenon affected by lots of factors including physiology and environment [2]. Recent studies have also gured out the possible association of the genetic background of the athletes in their high personal performances, resulting in the rise of a novel scienti c branch, called sport genetics [3,4]. Sport genetics could be de ned as the investigation of the genes and their molecular mechanisms affecting athletic performance and the determination of the possible associ- ation of the variants, especially single nucleotide polymorphisms (SNPs), with diverse athletic parameters including branch or personal performances [5]. According to the studies on sport genetics, 66% of athletic performance has been linked to the genetic background [6]. Moreover, physical parameters were also associated with the genetic background. For instance, 44–68% of endurance and 49–56% of muscular force were shown to be affected by genetic variations [7,8]. Thus, both genetics and the environment, which would in uence each other, have key roles in athletic performance [9,10]. For example, training periods to reach a performance level were proved to be linked to the genetic background of the athletes [9]. Recently, identi cation of candidate genes and/or variants associated with sports parameters has greatly attracted scientists. Until now, more than 235 genetic variants have been linked to the athletic parameters [10,11]. However, the results of the single-gene and/or variant approach may mislead, owing to the ignorance of the other related genes and/or variants. Consequently, it was realized that the results for the associations of each gene and/or variant were controversial [12]. Hence, multigenetic factors should be targeted to totally explore the possible associations. In parallel, several genome-wide association studies (GWAS) have been conducted on sports genetics. GWAS is a powerful technique to cover all known or unknown SNPs [13–15]. GWAS has proposed novel associated genes and/or SNPs

that the results for the associations of each gene and/or variant were controversial [12]. Hence, multigenetic factors should be targeted to totally explore the possible associations. In parallel, several genome-wide association studies (GWAS) have been conducted on sports genetics. GWAS is a powerful technique to cover all known or unknown SNPs [13–15]. GWAS has proposed novel associated genes and/or SNPs for the athletic parameters such as endurance, aerobic capacity, metabolism, and muscle ber composition [16,17]. However, the complexity and cost of GWAS limit such studies, and pilot experiments are suggested [18]. Exome-wide association studies (EWAS) could be an alternative to overcome the problems with GWAS. EWAS has also been previously chosen as a strategy to nd the possible associations in the sports genetics [19]. The aim of the present study was to identify genetic variants associated with personal best scores in Turkish track and eld athletes and to compare allelic frequencies between sprint/power and endurance athletes and controls using a whole-exome sequencing ap- proach, followed by replication studies in independent cohorts of athletes and controls. 2. Materials and Methods 2.1. Ethical Approval The study was carried out in accordance with the Declaration of Helsinki, and ap- proval was obtained from the Gazi University Non-Interventional Clinical Research Ethics Committee (with the decision dated 5 April 2021 and numbered 09) and from the Ethics Committee of the Federal Research and Clinical Center of Physical-Chemical Medicine of the Federal Medical and Biological Agency of Russia (Approval number 2017/04). 2.2. Participants 2.2.1. The Turkish Cohorts The Turkish study involved 60 elite athletes (sprint/power: 11 females (35.5%) and 20 males (64.5%); endurance: 10 females (34.5%) and 19 males (65.5%); mean age SD:

Genes2023,14, 660 3 of 11 25.1 4.8; height (cm): 174.97 7.9; body weight (kg) 72.5 22.4; sport experience (year) = 9.4 4.8; personal best (PB) = 1005.63 94.55) licensed in different clubs and af liated with the Turkish Athletics Federation. The number of controls (non-athletes) was 20 (6 females (30.0%) and 14 males (70.0%); mean age SD: 23.5 7.1), and they were healthy unrelated citizens of Turkish descent without any competitive sports experience. The athletes were categorized as either sprint/power or endurance athletes as deter- mined by the distance, duration, and energy requirements of their events. All athletes were nationally ranked in the top ten in their sports discipline and had participated in international competitions such as the Olympic Games, European Championships, Uni- versiade, Mediterranean Games, and Balkan Championship. The sprint/power group included sprint and power athletes whose events demand predominantly anaerobic energy production. The athletes in this group (n= 31) were 100–400 m runners (n= 9), jumpers (n= 3), and throwers (n= 19). The endurance athlete group (n= 29) included athletes competing in long-distance events demanding predominantly aerobic energy production. This group included 3000 m (n= 12), 5000 m (n= 5), 10,000 m (n= 4), and marathon (n= 8) runners. The informed voluntary consent and demographic information forms were obtained from the participants before the measurements. The International Association of Athletics Federations (IAAF) score scale was used to determine the performance levels of the athletes, depending on their personal best/competitive performance [20]. For instance, the IAAF score scale of a male athlete who runs 100 m in 10.05 sec is 1189, while that of a marathon runner who completes the race in 2 h 20 min 11 sec is 997. Thus, the performance scale of the marathon runner is less than that of the 100 m runner. The IAAF scales are useful for the determination of performances of athletes from diverse athletics events and genders. 2.2.2. The Russian Cohorts The Russian case–control study involved 488 elite athletes (293 males and 195 females), of whom 115 were elite sprint/power athletes (29 100–400 m runners, 38 500–1000 m speed skaters, 22

less than that of the 100 m runner. The IAAF scales are useful for the determination of performances of athletes from diverse athletics events and genders. 2.2.2. The Russian Cohorts The Russian case–control study involved 488 elite athletes (293 males and 195 females), of whom 115 were elite sprint/power athletes (29 100–400 m runners, 38 500–1000 m speed skaters, 22 sprint cyclists, 26 50 m swimmers), and 373 were elite endurance athletes (52 rowers, 32 biathletes, 7 long-distance cyclists, 30 kayakers and canoers, 37 middle- and long-distance speed skaters, 92 cross-country skiers, 63 middle- and long-distance runners, 31 middle- and long-distance swimmers, 8 race walkers, and 21 triathletes). The athletes were Russian national team members (participants and prize winners in international competitions) who had never tested positive for doping. Of 373 endurance athletes, 46 male endurance athletes (rowers, kayakers, speed skaters, biathletes, and cross-country skiers) and 29 female endurance athletes (rowers, kayakers, speed skaters, biathletes, and cross- country skiers) participated in the study of aerobic performance. Controls were 209 healthy and unrelated citizens of Russia without any competitive sport experience. The Russian muscle biopsy study involved 148 physically active participants of Rus- sian origin (99 males: mean age SD: 30.4 7.9 years; 49 females: mean age SD: 27.1 7.3 years). 2.2.3. The Finnish Cohort The Finnish muscle biopsy study (replication phase) involved 287 individuals (167 males, age 59.5 8.1 years; 120 females, age 60.7 7.4 years) from the FUSION study as previously described [21]. 2.3. Evaluation of Muscle Fiber Composition by Immunohistochemistry 2.3.1. Russian Study Vastus lateralis samples were obtained from the left legs of the participants using the modi ed Bergström needle procedure with aspiration under local anesthesia using 2% lidocaine solution. Serial cross-sections (7 m) were obtained from frozen samples. The sections were then incubated at RT in primary antibodies against slow or fast isoforms of the myosin heavy chains, as previously described [17,22].

Genes2023,14, 660 4 of 11 2.3.2. Finnish Study Muscle ber composition in 287 Finnish individuals was estimated based on the ex- pression of the myosin heavy chain 1 (MYH1), myosin heavy chain 2 (MYH2), myosin heavy chain 7 (MYH7), Ca 2+ ATPase A1, and Ca 2+ ATPase A2 genes, as previously described [21]. 2.4. VO2maxMeasurement Maximal oxygen consumption rate (VO2max) in rowers, kayakers, speed skaters and biathletes was determined using an incremental test to exhaustion on speci c ergometers. VO2maxwas determined breath-by-breath using a MetaLyzer II (Cortex Bio-physik, Leipzig, Germany), MetaMax 3B (Cortex Biophysik, Leipzig, Germany) or MetaMax 3B-R2 gas analysis systems (Cortex Biophysik, Leipzig, Germany), as previously described [23]. 2.5. Whole-Exome Sequencing (WES) The peripheral blood obtained from the participants was processed to isolate total DNA by DNeasy Blood and Tissue Kit (Qiagen, Hilden, Germany) according to the manu- facturer's instructions. Next, qualities of isolated DNA were checked by 1% agarose gel, and the concentrations were determined by a NanoDrop (NanoDrop 1000 Spectropho- tometer V3.8; Thermo Scienti c, Waltham, MA, USA). WES was performed after library preparation by the Twist Human Comprehensive Exome Panel (Twist Biosciences, San Francisco, CA, USA) according to the supplier's instructions. Brie y, enzymatic DNA fragmentation was performed, and Twist Hybridization probes and Dynabeads™MyOne™ Streptavidin T1 (Invitrogen, Carlsbad, CA, USA) were used for the hybridization. After the steps of library enrichment and determination of the library sizes, the samples were uploaded to the ow cells and the run was performed by Illumina NextSeq500 (Illumina Inc., San Diego, CA, USA). Average read depth was aimed as minimum 200 . Raw data were processed to by the Genome Analysis Toolkit (GATK)'s [24]. The HaplotypeCaller program was used to obtain Binary Alignment Map (BAM) les and subsequently produce an output Variant Call Format (VCF) le via the GRCh38/hg38 reference genome. Finally, variants were annotated by ANNOVAR [25]. 2.6. Data Extraction As the primary evaluation of the data, the VCF les were combined, and 511,061 vari- ants were detected. Only SNPs were analyzed in the context of the present study. The vari- ants with a minor allele frequency (MAF)

an output Variant Call Format (VCF) le via the GRCh38/hg38 reference genome. Finally, variants were annotated by ANNOVAR [25]. 2.6. Data Extraction As the primary evaluation of the data, the VCF les were combined, and 511,061 vari- ants were detected. Only SNPs were analyzed in the context of the present study. The vari- ants with a minor allele frequency (MAF) < 0.01, incorrectly annotated, and non-autosomal were eliminated, and 219,232 SNPs were further evaluated. 2.7. Genotyping DNA samples from Russian individuals were obtained from leukocytes (venous blood). DNA extraction and puri cation from blood samples were performed using commercial kits (Techno-sorb), according to the manufacturer's instructions (Techno-clon, Moscow, Russia). Genotyping of the candidate SNPs from the discovery phase was performed using microarray technology [26]. DNA samples from Finnish individuals were extracted from the blood, and the poly- morphisms were genotyped using the HumanOmni2.5–4v1_H BeadChip array (Illumina, San Diego, CA, USA), as previously described [21]. 2.8. Statistical Analyses Association analyses of Turkish data were performed by a Chi-square test using thet R program [27]. During the EWAS, the uni ed mixed-model method [27] was used. y = X + S + e (1) where y is the phenotypic observation; X is the xed effect; and S is the SNP effect [27]. The statistical signi cance probabilities of the SNP effects were converted to log10p. The

Genes2023,14, 660 5 of 11 results of EWAS analyses were presented as a Manhattan Plot. The exome-wide signi cance level was set atp< 2.3 10 7 (i.e., 0.05/219,232 SNPs). Statistical analyses of Russian and Finnish data were conducted using GraphPad InStat Version 3.05 (GraphPad Software, Inc., San Diego, CA, USA) software. The PLINK 1.9 program (National Institutes of Health, Bethesda, MD, USA) was used to perform genetic data quality control, and PLINK 2.0 was used to perform principal component analysis and association testing via generalized linear models. Bcftools was used for vcf le conversion. The phasing and imputation of genotypes were completed using the shapeit2 and impute2 programs. Differences in phenotypes between groups were analyzed using regression analysis adjusted for covariates. The chi-square test ( 2) was used to test for the presence of the Hardy–Weinberg equilibrium (HWE). Thereafter, the frequencies of genotypes or alleles were compared between sprint/power and endurance athletes and controls using Fisher's exact test. All data are presented as means (SD). Thep-values < 0.05 were considered statistically signi cant. 3. Results 3.1. Discovery Phase None of the SNPs reached an exome-wide signi cance level (p< 2.3 10 7 ) in genotype–phenotype and case–control studies of Turkish athletes (Figure). The only SNP that was close to the threshold (p= 1.0 10 5 ) was rs8037843 in the Pyroglutamyl-Peptidase I Like (PGPEP1L) gene (Figure). Although rs8037843 correlated with personal bests in athletes, there were no allelic differences between the Turkish and Russian endurance and sprint/power athletes and controls with respect to this SNP (p> 0.05).Genes 2023, 14, x FOR PEER REVIEW 6 of 12 Figure 1. Manhattan plot showing associations between SNPs across chromosomes and personal bests in elite Turkish athletes. Light and dark blue: the illustration separating the consecutive chro‐ mosomes for comprehensibility. The genotypic differences between the groups were evaluated by principal compo‐ nent analysis on an SNP matrix (PCA). PCA of the genotyping data pointed out no signif‐ icant influence of sport disciplines (Figure 2) on genotype distributions. Figure 2. Principal component analysis on the SNP matrix showing genotype distributions across different groups in EV1

the consecutive chro‐ mosomes for comprehensibility. The genotypic differences between the groups were evaluated by principal compo‐ nent analysis on an SNP matrix (PCA). PCA of the genotyping data pointed out no signif‐ icant influence of sport disciplines (Figure 2) on genotype distributions. Figure 2. Principal component analysis on the SNP matrix showing genotype distributions across different groups in EV1 −EV2 (A) and EV3 −EV4 (B) planes. Comparisons of allelic frequencies between three groups (endurance vs. sprint/power athletes; endurance athletes vs. controls; sprint/power athletes vs. controls) showed 53 SNPs whose frequencies were significantly differentiated between the sprint/power and endurance group (but not in the separate sub‐groups of female and male athletes due to low sample sizes) (Supplementary Table S1). The genes in which these SNPs were located were further analyzed by the String database (v.11.5; https://string‐ Figure 1. Manhattan plot showing associations between SNPs across chromosomes and personal bests in elite Turkish athletes. Light and dark blue: the illustration separating the consecutive chromosomes for comprehensibility. The genotypic differences between the groups were evaluated by principal component analysis on an SNP matrix (PCA). PCA of the genotyping data pointed out no signi cant in uence of sport disciplines (Figure) on genotype distributions.

Genes2023,14, 660 6 of 11Genes 2023, 14, x FOR PEER REVIEW 6 of 12 Figure 1. Manhattan plot showing associations between SNPs across chromosomes and personal bests in elite Turkish athletes. Light and dark blue: the illustration separating the consecutive chro‐ mosomes for comprehensibility. The genotypic differences between the groups were evaluated by principal compo‐ nent analysis on an SNP matrix (PCA). PCA of the genotyping data pointed out no signif‐ icant influence of sport disciplines (Figure 2) on genotype distributions. Figure 2. Principal component analysis on the SNP matrix showing genotype distributions across different groups in EV1 −EV2 (A) and EV3 −EV4 (B) planes. Comparisons of allelic frequencies between three groups (endurance vs. sprint/power athletes; endurance athletes vs. controls; sprint/power athletes vs. controls) showed 53 SNPs whose frequencies were significantly differentiated between the sprint/power and endurance group (but not in the separate sub‐groups of female and male athletes due to low sample sizes) (Supplementary Table S1). The genes in which these SNPs were located were further analyzed by the String database (v.11.5; https://string‐ Figure 2. Principal component analysis on the SNP matrix showing genotype distributions across different groups in EV1 EV2 (A) and EV3 EV4 (B) planes. Comparisons of allelic frequencies between three groups (endurance vs. sprint/power athletes; endurance athletes vs. controls; sprint/power athletes vs. controls) showed 53 SNPs whose frequencies were signi cantly differentiated between the sprint/power and endurance group (but not in the separate sub-groups of female and male athletes due to low sample sizes) (Supplementary Table S1). The genes in which these SNPs were located were further analyzed by the String database (v.11.5;, accessed on 10 December 2022) for the functional interaction and pathway analyses. The results showed minimal interactions between the proteins, and the Markov Cluster Algorithm (MCL) option in the database demonstrated ve clusters (Figure).Genes 2023, 14, x FOR PEER REVIEW 7 of 12 db.org/, accessed on 10 December 2022) for the functional interaction and pathway anal‐ yses. The results showed minimal interactions between the proteins, and the Markov Cluster Algorithm (MCL) option in the database demonstrated five clusters (Figure 3). Figure 3. Interaction

This research explores genetic factors influencing athletic performance in elite Turkish athletes.