Associations between gut microbiota and hand grip strength: a polygenetic scoring analysis and genome-wide environmental interaction study

Background We aim to explore the genetic association between hand grip strength (HGS) and gut microbiota (GM). The genome-wide association study (GWAS) data of GM was obtained from a recently published study, involving 2,646 individuals. Phenotypic data of left HGS (n = 375,229) and right HGS (n = 375,279) were derived from the UK Biobank. Firstly, PLINK 2.0 was used to calculate 114 GM related polygenetic risk score (PRS) of each subject based on GWAS result. Regression analysis was then conducted to evaluate the possible association of GM-PRS and HGS. Then genome-wide by environment interaction study (GWEIS) of GM related traits was then performed using a regression model adjusted for age, gender, BMI and 10 principal components. (P/A) and zero-truncated zero set as (AB) phenotype generated. RNT and HB (hurdle are two for AB) and 3


Abstract Background
We aim to explore the genetic association between hand grip strength (HGS) and gut microbiota (GM).

Methods
The genome-wide association study (GWAS) data of GM was obtained from a recently published study, involving 2,646 individuals.
Phenotypic data of left HGS (n = 375,229) and right HGS (n = 375,279) were derived from the UK Biobank. Firstly, PLINK 2.0 was used to calculate 114 GM related polygenetic risk score (PRS) of each subject based on GWAS result. Regression analysis was then conducted to evaluate the possible association of GM-PRS and HGS. Then genome-wide by environment interaction study (GWEIS) of GM related traits was then performed using a regression model adjusted for age, gender, BMI and 10 principal components.

Results
For PRS analysis, 13 candidate GM were identi ed to be associated with left HGS (P < 0.05), including G_Bi dobacterium (P = 8.36 ×

Conclusions
Our study holds great potential gaining a better understanding of the relationship between GM and HGS.

Background
Reduced muscle strength, as measured by hand grip strength (HGS), is considered a widely and accessible used proxy of muscle function. Therefore, grip strength serves as an established, simple and non-invasive marker of muscle tness. Meanwhile, HGS has been used as a signi cant predictor and biomarker of overall health of individual. In addition, various studies have shown a close correlation between HGS and many diseases related conditions, including long-term morbidity and mortality in adults [1], bone mineral density [2,3], fracture [4], depression [5], nutritional status [6], all-cause and cardiovascular mortality [7,8]. Previous literatures showed environmental exposures contributed a lot to the development of HGS, including age, gender, body size, low physical activity and smoking [9]. The heritability of HGS was demonstrated to be approximately 30%-50% [10]. Furthermore, multiple loci implicated in the genetic mechanisms of HGS have been found. Furthermore, Willems and colleagues identi ed 16 genome-wide signi cant loci associated with grip strength. And a number of these loci contain genes implicated in structure and function of skeletal muscle bres (ACTG1) [11].
Trillions of cells make up the gut microbiota (GM), including bacteria, viruses, and fungi, and they play a fundamental role in maintaining our health. The importance of GM in host physiology has received a great deal of attention in the last decade. According to published studies, changes in gut microbial diversity and composition have been demonstrated to be associated with several diseases and conditions [12], including neurologic conditions and psychiatric disorders [13], in ammation and immune function [14], metabolic diseases [15], skeletal and muscle function [16]. For example, a review study presented that gut microbes may contribute to the development of obesity and other associated chronic metabolic conditions [17]. Intriguingly, it was also found that GM in uences the skeletal homeostasis via affecting the host metabolism, immune function, hormone secretion, and the gut-brain axis, thus clarifying the relationship between GM and bone homeostasis [18].
Weaker muscle mass has been associated with reduced levels of physical performance [19]. In addition, reduced muscle mass and HGS re ected impaired nutritional status [20]. GM can in uence the host physiology by regulating multiple processes, but few studies have assessed the possible association between GM composition and muscle mass. Accumulating evidence has indicated that reduced muscle mass was associated with GM composition in mice [21]. For example, Lahiri et al. demonstrated the existence of a gut microbiota-skeletal muscle axis in germ-free mice, suggesting a role for GM in regulating skeletal muscle mass and function [22]. Nevertheless, limited evidence has targeted the association between GM and muscle mass in human studies. Given the reliable and accuracy of HGS as an indicator of general health and nutritional status, while the important role of GM in host physiology, we assumed there might be some functional interactions between GM and muscle strength in human.
Recent years, genome-wide association studies (GWAS) have showed that common complex disorders have a polygenic genetic architecture, which can be combined into a polygenic risk score that predicts an individual's susceptibility to diseases. Therefore, based on GWAS results, a powerful new method for exploring the genetic structure of complex traits and predicting disease risk is proposed, polygenic risk score (PRS) analysis [23]. PRSs are an estimate of disease risk conducted by the individual based on the risk alleles and the corresponding effect sizes obtained from the GWAS summary statistics. Interestingly, it can determine whether the pleiotropic effects of variations found in one disease lead to an increased risk of the other. Nowadays, PRS analysis has been widely applied for detecting shared genetic aetiology among traits [24].
Usually, the pathogenesis of complex diseases is thought to involve the interaction of genetic factors and environmental exposures, which makes it di cult to study separately [25]. Genome-wide by environment interaction study (GWEIS) is a simple but powerful method taking into account of the gene × environmental interaction(G × E)in the context of GWAS [26]. In addition, it is a complementation for GWAS analysis, for GWEIS focuses not only main effects, but also for marginal genetic effects, such as heterogeneity SNPs between subgroups de ned by some environmental exposure [27]. G × E studies for common complex phenotypes are needed and may be informative for identifying novel genomic loci and providing better insight into mechanisms for complex diseases. Nowadays, GWEIS has been used as a powerful tool to uncover newly susceptibility loci for complex diseases.
For example, a study suggested the APOE gene could be a possible candidate gene for the G × E interaction on grip strength trajectories [28].
In this study, using the GWAS datasets of GM and genotyping data of HGS from UK biobank, we rstly calculate the PRS of 114 GM related traits, and assessing the genetic correlation between GM related PRS and HGS. Then genome-wide gene × GM interaction analyses was performed to study the effect of GM on HGS. We hope that our study results provide novel clues for understanding the biological mechanism underlying the relationship between GM and HGS.

Ethic Statement
The signed consents were provided in the participants visit assessment. Ethical approval of UK Biobank was granted by the National Health Service National Research Ethics Service (reference 11/NW/0382).

Phenotype de nition of hand grip strength
Phenotypic data of HGS in this study was derived from the UK Biobank (https://www.ukbiobank.ac.uk/), which is a prospective study cohort consists of approximately 500,000 individuals aged between 49 and 60 years old [29]. Grip strength was assessed using a hydraulic hand dynamometer (Jamar J00105, Lafayette Instrument Company, Lafayette, Indiana), and the mean values of the right-hand and left-hand used in the analyses were expressed as kg per kg body weight [30].

Genotyping, imputation and quality control (QC)
Genotyping, imputation and QC were performed by the UK Biobank [29]. Brie y, SNP genotyping was performed using the Applied Biosystems UK BiLEVE Axiom Array by Affymetrix. Imputation was carried out with the IMPUTE4 program. The merged UK10K, 1000 Genomes phase 3 reference panels and the Haplotype Reference Consortium (HRC) data were used as the imputation reference panel, of which HRC data was de ned as the main imputation reference panel [31]. For the sample-based QC, the metrics of missing rate and heterozygosity were used to identify the poor-quality samples. For marker-based QC, batch effects, plate effects, departures from Hardy-Weinberg equilibrium, sex effects, array effects, and discordance across control replicates, these are all tested statistically here to identify poor quality markers, checking for consistency of genotype calling across experimental factors. The detailed information about the DNA extraction and genotyping are described in a published study [31].

GWAS data of gut microbiota
The GWAS data of 114 GM related traits was derived from a recently published study [32]. A total of 2,646 individuals from the Flanders region of Belgium were recruited into the Flemish Gut Flora Project (FGFP) and 2,688 stool samples were collected.
Informed consent was provided by volunteers using mail. DNA was extracted from frozen faecal samples using the PowerMicrobiome RNA Isolation Kit (MOBIO Laboratories). For microbial traits, a presence/absence (P/A) phenotype and a zerotruncated (all zero values set as missing) abundance (AB) phenotype were generated. RNT (rank-normal transformation) and HB (hurdle binary) are two microbial trait model for continuous (RNT | AB) and binary (P/A) traits [32].
Genotyping was conducted using the Human Core Exome v.1.0 array and the Human Core Exome v.1.1 array. Totally, 509,886 variants and 2,293 individuals were remained after marker-based QC and sample-based QC. FGFP genotype data were phased using SHAPEIT3 and imputed with IMPUTE4 using UK10K and all 1000 Genome Project phase 3 samples as the reference pane. Detailed information about study recruitment and sample collection, sequencing and microbiome data processing, genotyping, QC and imputation were described in the published study [32].
Polygenetic Risk Score analysis In this study, PLINK 2.0 was used to calculate polygenetic risk GM scores of each study subject based on GWAS result (http://www.cog-genomics.org/plink/2.0/) [33]. In brief, PRSs are the estimation of diseases risk, calculated by summing risk alleles, which are weighted by effect sizes derived from GWAS results. The GWAS summary statistics will be referred as the base and the dataset to be evaluated as the target. PRS analysis uses the beta coe cients from the GWAS as weight for each SNP allele in order to calculate an overall risk score for each individual subject in an independent target sample. Linear regression models were then used to test the association between the GM related PRS (predictor variable) and HGS variables (independent variables) using R (https://www.r-project.org/).

Genome-wide environmental interaction study (GWEIS)
Genome-wide gene-PRSs of GM interaction association analysis was performed here using a regression model adjusted for age, gender and principal components. PLINK 2.0 was used for the interaction analysis of the GWAS data (http://www.coggenomics.org/plink/2.0/) [33]. We used the following criteria for selecting SNPs. . SNPs with call rates 0.9 . Hardy-Weinberg equilibrium exact test P values 0.001 . minor allele frequencies (MAFs) 0.01. In addition, The unrelated subjects were generated with KING software, a rapid algorithm for relationship inference that allows the presence of unknown population substructure [34].
The quantile-quantile plot of interaction P values and the Manhattan plot of − log10 P of interaction were generated using"CMplot" package (https://github.com/YinLiLin/R-CMplot) in R platform. Signi cant SNPs for G × E effects were identi ed at a genome-wide signi cance threshold of P < 5 × 10 − 8 .

Discussion
Despite evidence suggesting that altered GM has been implicated in the pathogenesis of several complicated diseases, little work has been done to explore the genetic association between GM and HGS. In this study, using PRS and GWEIS analysis, several genetic associations between GM and HGS were identi ed, as well as signi cant loci interacted with GM effecting HGS. Polygenic risk scores (PRSs) analyses incorporating G × E effects provided evidence of shared aetiologies between HGS and GM, which might improve e ciency for candidate gene discovery.
Bi dobacteria was found to be associated with both left and right HGS in this study. Bi dobacteria are a group of bacteria that normally live in the intestines, most of which can be considered as a probiotic and taken by mouth as medicine. The role of Bi dobacteria in muscle function has been well studied recently. An experimental study investigated the effects of Bi dobacterium breve B-3 (B-3) on muscle function on mice [35]. And it turned out that the administration of B-3 on the mouse model could increase muscle mass and affect muscle metabolism, and the grip strength of heat-killed B-3 group was signi cantly higher than that of the control group [35]. Similarity, another study suggested that dietary supplementation with Bi dobacterium breve BR03 attenuates performance decrements and muscle tension in the days following muscle-damaging exercise [36]. Furthermore, Ni et al. and colleagues found Lactobacillus and Bi dobacterium supplementation enhanced muscle strength and function in aged mice by the GM regulation [37]. Also, the protecting role of Bi dobacterium in against obesity has been proposed [38]. Taken altogether, ndings from previous work we discussed above indicated the important role of Bi dobacterium in muscle mass and metabolism on mice. However, the number of evidences about this relationship in human studies remained scarce. In our analysis, we found Bi dobacteria was associated with both right and left HGS in human. Further supporting evidence is needed to prove our results.
G_Butyricicoccus is another common signi cant GM associated with both right and left HGS in our study. Interesting, the interaction between nutrients and the GM in the context of obesity and related metabolic diseases has been well studied [39].Among them, Butyricicoccus is one of the obesity related GM. For example, an experimental analysis found after transplantation of the GM from obese individuals to high-fat diet-fed mice, they responded differently to inulin supplementation, with Butyricimonas associated with the observed metabolic outcomes (decrease in adiposity and hepatic steatosis) in human-obesity mice [40]. Furthermore, Butyricicoccus drives the decrease of body mass index in response to inulin in obese individuals, suggesting the important role of characterizing the speci c consortia of bacteria in the context of obesity and metabolic diseases when improving in metabolic disorders by inulin [40]. Studies showed the potential role of Butyricicoccus in exercise physiology. For example, Tung et al. found a greater abundance and diversity of GM in intrinsic high exercise capacity mice than in low exercise capacity mice, indicating the potential effect of GM and functional proteins on intrinsic exercise capacity [41]. The important role of nutrients in muscle strength and HGS has been revealed in previous study [42], which is consistent with our results.
O_Lactobacillales is found to be associated with left HGS in our analysis. Lactobacillales is a well-known probiotic among the ingested-microorganism probiotics. Various biological function of Lactobacillales has been explored, including treating obesity [43], enhancing muscle strength and function [37], improving insulin sensitivity [44]. A study conducted on mouse model found forelimb grip strengths were 1.31 and 1.40 fold higher in the Lactobacillus plantarum TWK10-1X and Lactobacillus plantarum TWK10-5X groups than in the vehicle treatment group, indicating long-term supplementation with Lactobacillus plantarum TWK10 might increase in muscle mass and strength [45]. Similar conclusion about the role of Lactobacillus plantarum TWK10 in physiological homeostasis and health promotion was also drawn in human study [46]. In addition, another study evaluated the effect of Lactobacillus rhamnosus on GM, changes in permeability, and insulin sensitivity and signaling in high-fat diet and control animals, and they suggested the important role of probiotics to prevent and treat patients with obesity and insulin resistance [47].
G_Veillonella, G_Ruminococcus and P_Firmicutes were also identi ed to be related with HGS in our study. An analysis found Ruminococcus and Veillonella were signi cantly related to muscle improvement outcomes after soy-whey blended protein (SWP) treatment in hematopoietic stem cell transplantation patients, suggesting that intestinal microbiota might affect the regulation of muscle metabolism [48]. Interesting, gender and body mass index may in uence intestinal microbiota. It was found that Veillonella is signi cantly higher in fecal samples in men compared to women, and these differences may be in uenced by the grade of obesity [49]. The Firmicutes is one of two groups of bene cial bacteria that dominant in the human gut. A meta-analysis concluded that changes in Firmicutes and Bacteroidetes phyla/species levels might be signi cant indicators for childhood obesity [50]. Likewise, same conclusion was also drawn by other literature [51]. A high value of evidence indicated that muscle mass is an important criterion in nutrition status diagnosis, and HGS is used as an e cient tool for evaluating muscle functioning. Given the fact that the important role of HGS as an outcome predictor and marker of nutritional status [52], the relationship among microbiota, HGS and obesity deserves to be deeply explored. Our results were consistent with previous studies.
For GWEIS analysis, several signi cant SNPs were identi ed to be interacted with G_Butyricicoccus for HGS. Among them, rs41310432 is the most signi cant one (NASP, P = 3.86 × 10 − 8 ). Nuclear autoantigenic sperm protein (NASP) encodes a H1 histone binding protein and it is cell-cycle regulated [53]. It is expressed in all cells undergoing division. Functions of NASP mainly involve liver cancer [54], immune response [55], and cell proliferation [56]. For example, previous studies have shown that in lupus model mice, NASP gene mutation can change the proportion of immune cells in the spleen and aggravate the autoimmune response [55]. In addition, Yang and colleagues found somatic NASP negatively regulates TRAF6, and is a key regulator of innate immunity [57].
Interestingly, TRAF6 has been demonstrated to be a novel regulator of skeletal muscle atrophy [58]. Another study showed that overexpression of miR-125b-5p targeting TRAF6 may provide a promising therapeutic approach to treat muscle atrophy [59]. We found NASP is interacted with G_Butyricicoccus for HGS in our study. Given the potential role of in ammation [60] and muscle mass in HGS based on previous studies, the association between HGS and NASP needs to be studied further.
Another suggestive gene we identi ed in GWEIS analysis is CACNA1H. Calcium Voltage-Gated Channel Subunit Alpha1 H (CACNA1H), encodes a T-type member of the alpha-1 subunit family, a protein in the voltage-dependent calcium channel complex. Previous study indicated compound heterozygous CACNA1H mutations might lead to severe congenital amyotrophy [61]. Another experimental analysis suggested reduced expression of CACNA1H related to multiple vaginal delivery is associated with muscular atrophy, indicating the key regulator role of CACNA1H in skeletal muscle function [62]. In our study, we found CACNA1H is interacted with G_Butyricicoccus and might affect HGS in UK Biobank samples. Further analyses are warranted to validate our results.
In this study, we analyzed the association between HGS and GM using large sample from UK Biobank via utilizing PRS and GWEIS analysis. PRSs analyses incorporating G × E effects provided evidence of shared aetiologies between HGS and GM. As far as we known, this is the rst systemic study exploring the effect of GM as environmental factor on HGS. Besides, insight on interactions between GM and muscle mass from previous studies are mostly based on mouse studies rather than human studies. In addition, the large sample sizes guaranteed the accuracy of our results. Our study holds great potential for clarifying the functional relevance of GM with HGS.
Several issues in our study should be noted. Although several signi cant GM were identi ed to share common genetic etiology with HGS, as well as signi cant genes interacted with GM for HGS, further biological research should determine if these genes are true susceptibility loci for HGS in relation to GM. In addition, participants from our study are all European ancestry and so the conclusions that can be drawn are limited. Further analysis should include other ethnic groups.

Conclusions
In summary, we identi ed genetic associations between GM and HGS using PRS analysis and signi cant genes interacted with GM effecting HGS using GWEIS analysis. Our results have provided entry points to analyze the mechanisms underlying the relationship between GM and HGS and providing an insight on the pathogenesis and therapeutic strategies of these musculoskeletal diseases. Availability of data and materials

Abbreviations
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.