Identiﬁcation and Characterization of Genomic Predictors of Sarcopenia and Sarcopenic Obesity Using UK Biobank Data

: The substantial decline in skeletal muscle mass, strength, and gait speed is a sign of severe sarcopenia, which may partly depend on genetic risk factors. So far, hundreds of genome-wide signiﬁcant single nucleotide polymorphisms (SNPs) associated with handgrip strength, lean mass and walking pace have been identiﬁed in the UK Biobank cohort; however, their pleiotropic effects on all three phenotypes have not been investigated. By combining summary statistics of genome-wide association studies (GWAS) of handgrip strength, lean mass and walking pace, we have identiﬁed 78 independent SNPs (from 73 loci) associated with all three traits with consistent effect directions. Of the 78 SNPs, 55 polymorphisms were also associated with body fat percentage and 25 polymorphisms with type 2 diabetes (T2D), indicating that sarcopenia, obesity and T2D share many common risk alleles. Follow-up bioinformatic analysis revealed that sarcopenia risk alleles were associated with tiredness, falls in the last year, neuroticism, alcohol intake frequency, smoking, time spent watching television, higher salt, white bread, and processed meat intake; whereas protective alleles were positively associated with bone mineral density, serum testosterone, IGF1, and 25-hydroxyvitamin D levels, height, intelligence, cognitive performance, educational attainment, income, physical activity, ground coffee drinking and healthier diet (muesli, cereal, wholemeal or wholegrain bread, potassium, magnesium, cheese, oily ﬁsh, protein, water, fruit, and vegetable intake). Furthermore, the literature data suggest that single-bout resistance exercise may induce signiﬁcant changes in the expression of 26 of the 73 implicated genes in m. vastus lateralis, which may partly explain beneﬁcial effects of strength training in the prevention and treatment of sarcopenia. In conclusion, we have identiﬁed and characterized 78 SNPs associated with sarcopenia and 55 SNPs with sarcopenic obesity in European-ancestry individuals from the UK Biobank.


Introduction
Sarcopenia is an age-associated condition characterized by the loss of skeletal muscle strength and muscle mass, and in severe cases, followed by reduced physical performance (e.g., slower gait speed) [1]. In most cases (53-84%), sarcopenia co-exists with obesity [2,3]. Older adults are usually identified as having sarcopenic obesity if low muscle mass and strength, as well as increased adiposity are present [4][5][6]. Furthermore, individuals with sarcopenic obesity can be stratified into stage I (absence of clinical complications) or stage II (presence of clinical complications) [7].
The prevalence of sarcopenia and sarcopenic obesity depends on the diagnostic criteria used to describe these conditions, with the range from 10-27% for sarcopenia [6], and from 10-23% for sarcopenic obesity [8]. Both sarcopenia and sarcopenic obesity are related to negative health-related outcomes, such as increased risk of falls, disability, frailty, osteoporosis, type 2 diabetes (T2D), metabolic syndrome, poor glycaemic profiles (i.e., hyperglycaemia, high HbA1c, insulin resistance, etc.) cardiovascular diseases, dyslipidaemia, poor neurocognitive functioning and quality of life, decreased health span and mortality [9][10][11][12][13][14]. Importantly, while declines in lean mass could contribute to further gains in fat mass, high fat mass may also lead to accelerated loss of lean mass [10].
There is a wide range of individual variability in skeletal muscle quantity and quality even under the same biological (age, sex) and environmental (level and type of physical activity, macro and micronutrient intake, etc.) conditions. This variability partly depends on genetic factors, with high heritability for muscle strength (49-56%) [15] and fat-free mass (45-76%) [16]. So far, hundreds of genome-wide significant (p < 5 × 10 −8 ) single nucleotide polymorphisms (SNPs) associated with sarcopenia-related traits, such as handgrip strength (170 SNPs) [17][18][19], appendicular lean mass (1059 SNPs) [20] and walking pace (70 SNPs) [21] have been individually identified in the UK Biobank cohort. However, their pleiotropic effects on all three traits have not been investigated on a systemic level.
Since muscle strength has been shown to be positively correlated with muscle mass and walking speed [22,23], we hypothesized that these phenotypes might be positively associated at a genetic level as well. One might also suggest that alleles associated with all three sarcopenia-related traits (i.e., low muscle strength, low lean mass and slow walking pace) can be considered as the most robust predictors of sarcopenia. On the other hand, risk alleles for both sarcopenia and increased adiposity can be considered as genomic predictors of sarcopenic obesity.
The aims of the present study were threefold: (1) to identify SNPs with pleiotropic effects on handgrip strength, appendicular lean mass, usual walking pace and fat percentage using summary statistics from the UK Biobank cohort; (2) to identify the potential mechanism of action of each SNP on sarcopenia-related traits by searching for intermediate phenotypes and using mice knockout models; and (3) to investigate the effect of resistance exercise on the expression of sarcopenia-related genes using bioinformatic tools.

UK Biobank Study
The UK Biobank is an open-access large prospective study with phenotypic and genotypic data from more than 500,000 participants (>90% of participants are of white ethnicity) with an age range for inclusion of 40-69 years when recruited in 2006-2010 [24]. UK Biobank has approval from the North West Multi-centre Research Ethics Committee (MREC) as a Research Tissue Bank (RTB) approval (reference 11/NW/0382). Full written informed consent was obtained from all participants prior to the study.

Identification of Genomic Predictors of Sarcopenia and Sarcopenic Obesity Using UK Biobank Data
In the first stage, we used publicly available summary statistics from five published genome-wide association studies (GWASes) on handgrip strength [17][18][19], appendicular lean mass [20] and walking speed [21], making an initial list of 1299 genome-wide significant (p < 5 × 10 −8 ) SNPs (Table 1). To identify correspondences between phenotypes (for example, to find if a specific SNP associated with handgrip strength is also associated with appendicular lean mass and walking pace), we used publicly available summary statistics from GWASes on appendicular lean mass [20], appendicular lean mass in older adults [25], handgrip strength (left) [26], handgrip strength (right) [26], handgrip strength in older adults (weakness) [19] and usual walking pace [26] with less stringent p value threshold (p < 0.005) ( Table 1). SNPs associated with all three phenotypes (i.e., with appendicular lean mass, handgrip strength and walking pace) with consistent effect directions were considered as potential genomic predictors of sarcopenia. In the second stage, to test the hypothesis that sarcopenia-related SNPs are also associated with other health-related traits, we used summary statistics from GWASes on body fat percentage [26], type 2 diabetes [26], heel bone mineral density [27], frequency of tiredness [26], self-reported tiredness [26], recent feelings of tiredness or low energy [26], and falls in the last year [26] (Table 1). Risk alleles for both sarcopenia and increased body fat percentage were considered as genomic predictors of sarcopenic obesity. Furthermore, risk alleles for both sarcopenic obesity and type 2 diabetes were considered as genomic predictors of sarcopenic diabesity. We also used other traits (biochemical, anthropometric, physiological, behavioral) to identify shared genetic architecture between sarcopeniarelated traits and lifestyle exposures (Table 1).

Analysis of Sarcopenia-Related Polygenic Profiles in European Populations
Raw genetic data of sarcopenia-related SNPs of 503 anonymized individuals of European origin from the 1000 Genomes project (Phase 3) [35] were used to calculate a genetic sum score of risk alleles for each individual. This cohort was composed of five subgroups (British from England and Scotland (n = 91), Finnish in Finland (n = 99), Toscani in Italia (n = 107), Iberian populations in Spain (n = 107), Utah residents (CEPH) with Northern and Western European ancestry (n = 99). Unweighted polygenic risk scores (coded as 0, 1 and 2 for homozygous non-risk genotype, heterozygous genotype and homozygous risk genotype, respectively) were developed for the prediction of sarcopenia, sarcopenic obesity and sarcopenic diabesity in European populations. Individuals were evenly divided into 5 groups (20% each) with high (high number of risk alleles), above average, average, below average and low (low number of risk alleles) risks for sarcopenia, sarcopenic obesity and sarcopenic diabesity.

Analysis of Association of Sarcopenia-Related SNPs with Gene Expression
The Genotype-Tissue Expression (GTEx) portal [36] was used to analyze the association between sarcopenia-related SNPs and expression of genes in different tissues with the focus on skeletal muscle tissue and nervous system (p < 0.05). The GTEx project is an ongoing effort to build a comprehensive public resource to study tissue-specific gene expression and regulation. Samples were collected from 49 tissue sites across >800 individuals, primarily for molecular assays including whole genome sequencing (WGS), whole exome sequencing (WES), and RNA-Seq [37]. SNPs that were significantly (p < 0.05) correlated with expression of genes (levels of mRNAs) were considered as expression quantitative trait loci (eQTLs).

Analysis of Effects of Knockouts of Implicated Genes on Sarcopenia-Related Traits in Mice
Data from the International Mouse Phenotyping Consortium (IMPC) database [38] were used to assess the effects (p < 0.05) of genes knockout on lean mass, fat mass and grip strength in mice. The IMPC web portal makes available curated, integrated and analyzed knockout mouse phenotyping data from 9000 mouse lines [39].

Analysis of Effects of Strength Training on the Expression of Sarcopenia-Related Genes
Publicly available human skeletal muscle transcriptome dataset was used to check the significant effect (p < 0.05) of a single-bout resistance exercise on the mRNA of the sarcopenia-related genes in m. vastus lateralis of seven young men (age 23.3 ± 0.6 years) at 2.5 h and 5 h timepoints compared to baseline [40].

Potential Genomic Predictors of Sarcopenia and Sarcopenic Obesity
A flow diagram displaying the study design and the main findings is shown in Figure 1. First, by combining data from five published GWASes on handgrip strength [17][18][19], appendicular lean mass [20] and walking speed [21], we made a list of 1299 genome-wide significant SNPs. The 78 out of those SNPs (from 73 loci) were independently (i.e., all SNP with linkage disequilibrium (LD) threshold r 2 > 0.2 were excluded) associated (p < 0.005) with all three traits-appendicular lean mass, handgrip strength and walking pace-with consistent effect directions ( Table 2). These 78 SNPs can be considered as potential genomic predictors of sarcopenia and can be used as instruments for Mendelian randomization analysis to study and uncover causal relationships between sarcopenia and other traits.
Next, by using summary statistics from GWASs on other health-related traits we found that 55 of the 78 SNPs were associated with body fat percentage with consistent effect directions (i.e., the same allele is a risk variant for both sarcopenia and adiposity) and can be regarded as potential genomic predictors of sarcopenic obesity. Of these 55 SNPs, the 21 SNPs were also associated with the risk of T2D with consistent effect directions (potential genomic predictors of sarcopenic diabesity) ( Table 2).

Polygenic Analysis of Sarcopenia, Sarcopenic Obesity and Sarcopenic Diabesity
A genetic sum score of sarcopenia risk alleles composed of 78 SNPs was calculated for each of 503 individuals of European origin from the 1000 Genomes project. Individuals were then evenly (by~20%) divided into five groups. Carriers of 58-68 risk alleles had the lowest risk of sarcopenia, whereas carriers of 81-95 risk alleles had the highest risk. The distribution of risk alleles in each subgroup for sarcopenia, sarcopenic obesity and sarcopenic diabesity is shown in Table 3 and may be used to improve prediction of these disease states when incorporated into existing clinical risk tools in individuals of European origin.
By comparing human (GWAS, GTEx) and mice genes knockout data, we identified eight genes with the same direction of association. More specifically, while protective alleles in ADCY3 (rs10203386 T), BCKDHB (rs9350850 C), CEP192 (rs1786263 G), H1FX (rs4073154 G), and POLD3 (rs72977282 T) genes were associated with the increased expression of these genes in human tissues, the knockout of the corresponding genes in mice (Adcy3, Bckdhb, Cep192, H1fx, and Pold3) led to the decrease in lean mass and strength (with increase in fat mass).
On the other hand, while protective alleles in BTRC (rs10883618 A), LCORL (rs1472852 C), MTCH2 (rs11039324 G) genes were associated with a decreased expression of these genes in human tissues, the knockout of the corresponding genes in mice (Btrc, Lcorl, and Mtch2) led to the increase in lean mass and strength (with decrease in fat mass) (Supplementary Table S1).

Discussion
In this study, we identified and characterized 78 pleiotropic genomic predictors of sarcopenia based on previously discovered genome-wide significant SNPs associated with handgrip strength, appendicular lean mass and walking pace. Of the 78 SNPs, 55 polymorphisms were also associated with body fat percentage and 25 polymorphisms with type 2 diabetes (T2D), indicating that sarcopenia, obesity and T2D share many common risk alleles. It is, therefore, unsurprising that according to the Data from Health and Nutrition Examination Survey (NHANES), 83.6% of women and 79.3% of men (aged 60 and older) with sarcopenia also have obesity (i.e., sarcopenic obesity) [2].
The  Table S1). Interestingly, of the 73 genes, knockouts of 27 genes in mouse models led to functional consequences such as changes in the lean mass, fat mass and grip strength.
Of the 78 SNPs, 58 were identified as eQTL SNPs that correlated with expression of genes in various tissues including skeletal muscle and nervous system, indicating that they are likely to be functional and may influence multiple traits. Indeed, we found that risk alleles were also associated with other intermediate phenotypes of sarcopenia, namely tiredness, falls in the last year, low physical activity, and low bone mineral density. Furthermore, risk alleles were associated with neuroticism, time spent watching television, alcohol intake, smoking and poor diet (higher salt, white bread, and processed meat intake), whereas protective alleles were positively associated with serum testosterone, IGF1, and 25-hydroxyvitamin D levels, height, intelligence, cognitive performance, educational attainment, income, ground coffee drinking and healthier diet (muesli, cereal, wholemeal or wholegrain bread, potassium, magnesium, cheese, oily fish, protein, water, fruit, and vegetable intake). This is in line with the previous studies stating that low educational attainment [41], neuroticism [42], low testosterone levels [43], short stature [44], high alcohol [45], processed meat [46] and salt [47] intake, sedentary behavior (such as watching television) [48], smoking and physical inactivity [49] are associated with an increased risk of sarcopenia or low muscle strength, whereas coffee, magnesium, potassium, protein, vitamin D, water, oil fish, fruits and vegetables intake [46,[50][51][52][53][54] have protective effects against sarcopenia.
Future research is needed to test interventional strategies focusing on all these factors to evaluate improvement in muscle quality and quantity. Given that resistance exercise induces significant changes in the expression of 26 genes (out of 73 implicated genes) in human skeletal muscle compared to the pre-training state, our findings also partly explain beneficial effects of strength training in the prevention and treatment of sarcopenia [55].
The link of 78 SNPs with muscle strength and lean mass indicates that these markers may be important not only in the general population, but also in athletes. Indeed, of this panel of markers, some of protective alleles have been reported to be over-represented in elite sprinters (E2F3 rs4134943 T, FHL2 rs55680124 C, GDF5 rs143384 G, SLC39A8 rs13107325 C, and ZNF568 rs1667369 A) [56] and elite strength athletes (ADCY3 rs10203386 T, ADPGK rs4776614 C, MMS22L rs9320823 T and ZKSCAN5 rs3843540 C) [57] compared to controls. Furthermore, GBF1 rs2273555 G, MLN rs12055409 G, and MMS22L rs9320823 T alleles (all protective) were found to be positively associated with weightlifting performance [58,59]. One of markers (rs3734254) is located in the PPARD gene which is in high linkage disequilibrium (LD) with the PPARD rs2016520 SNP, which has been previously associated with endurance athlete status [60].
Our study presents novel data of sarcopenia-related genetic markers. However, there are also limitations. Firstly, our findings were based on summary statistics of three different phenotypes which were discovered in the whole sample of the UK Biobank. To confirm the association between identified SNPs, sarcopenia and sarcopenic obesity, a case-control study (individuals with confirmed sarcopenia vs. individuals with normal muscle quality and quantity) is needed in independent studies. Second, our results were obtained using genomic data of European-ancestry individuals from the UK Biobank. Therefore, the set of 78 SNPs should be analyzed for association with sarcopenia in other populations as well before implementation in practice. We also recognise small sample size (n = 7) in the study of transcriptomic responses to a single-bout resistance exercise and encourage independent replication in larger cohorts.

Conclusions
In conclusion, we have identified and characterized 78 SNPs associated with sarcopenia and 55 SNPs with sarcopenic obesity that highlight shared genetic architecture between sarcopenia-related traits and lifestyle exposures.
We strongly suspect that many additional common polymorphisms, and probably rare mutations as well, will be shown to be associated with sarcopenia-related traits in due course. Thus, we suspect that the 78 polymorphisms we have identified constitute only a small fraction of the genetic factors that influence muscle strength, muscle mass and walking pace. However, looking to the future, when thousands of polymorphisms will be discovered that contribute to the variability in sarcopenia-related traits, the power of such information (in conjunction with standard measurement data) as a practical tool for clinicians will emerge.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/nu15030758/s1, Table S1: Association of top sarcopenia-related SNPs and implicated genes with health-related traits, gene expression and responses to resistance training. Informed Consent Statement: Informed consent was obtained from all subjects involved in the UK Biobank study.

Data Availability Statement:
The data presented in this study are publicly available online at https://genetics.opentargets.org (accessed on 27 December 2022).