Variation in GYS1 Interacts with Exercise and Gender to Predict Cardiovascular Mortality

Background The muscle glycogen synthase gene (GYS1) has been associated with type 2 diabetes (T2D), the metabolic syndrome (MetS), male myocardial infarction and a defective increase in muscle glycogen synthase protein in response to exercise. We addressed the questions whether polymorphism in GYS1 can predict cardiovascular (CV) mortality in a high-risk population, if this risk is influenced by gender or physical activity, and if the association is independent of genetic variation in nearby apolipoprotein E gene (APOE). Methodology/Principal Findings Polymorphisms in GYS1 (XbaIC>T) and APOE (-219G>T, ε2/ε3/ε4) were genotyped in 4,654 subjects participating in the Botnia T2D-family study and followed for a median of eight years. Mortality analyses were performed using Cox proportional-hazards regression. During the follow-up period, 749 individuals died, 409 due to CV causes. In males the GYS1 XbaI T-allele (hazard ratio (HR) 1.9 [1.2–2.9]), T2D (2.5 [1.7–3.8]), earlier CV events (1.7 [1.2–2.5]), physical inactivity (1.9 [1.2–2.9]) and smoking (1.5 [1.0–2.3]) predicted CV mortality. The GYS1 XbaI T-allele predicted CV mortality particularly in physically active males (HR 1.7 [1.3–2.0]). Association of GYS1 with CV mortality was independent of APOE (219TT/ε4), which by its own exerted an effect on CV mortality risk in females (2.9 [1.9–4.4]). Other independent predictors of CV mortality in females were fasting plasma glucose (1.2 [1.1–1.2]), high body mass index (BMI) (1.0 [1.0–1.1]), hypertension (1.9 [1.2–3.1]), earlier CV events (1.9 [1.3–2.8]) and physical inactivity (1.9 [1.2–2.8]). Conclusions/Significance Polymorphisms in GYS1 and APOE predict CV mortality in T2D families in a gender-specific fashion and independently of each other. Physical exercise seems to unmask the effect associated with the GYS1 polymorphism, rendering carriers of the variant allele less susceptible to the protective effect of exercise on the risk of CV death, which finding could be compatible with a previous demonstration of defective increase in the glycogen synthase protein in carriers of this polymorphism.

INTRODUCTION Cardiovascular (CV) disease (CVD), including coronary heart disease (CHD) and stroke, is the leading cause of death and disability in the Western world [1] and is thought to result from a complex interaction between genetic and environmental factors. Such risk factors are age, male gender, smoking, hypertension, diabetes, dyslipidemia [2] and physical inactivity. The genetic constitution of an individual usually determines how the individual responds to these risk factors. Therefore, it is necessary not only to identify which genetic variants increase susceptibility to the disease but also which environmental risk factors act in concert with these genes. In addition, the cellular environment in men and woman can be very different given known differences in hormonal milieu and gene expression [3]. Therefore, it is reasonable to consider the possibility that gender specific gene-environment interactions could modify the penetrance and expression of the trait.
Muscle glycogen synthase is the key enzyme in the synthesis of glycogen in skeletal muscle. A polymorphism (XbaI) in intron 14 of the glycogen synthase gene (GYS1) has been associated with lower glycogen synthase activity, T2D, features of the metabolic syndrome (MetS) and with myocardial infarction in males [4][5][6][7][8] but association to T2D has not been consistently replicated in all studies [9,10]. Interestingly, electrical stimulation of skeletal muscle to mimic physical exercise increased the amount of glycogen synthase in carriers of wild-type C-allele but not in carriers of the T-allele. As a consequence, carriers of the T-allele may benefit less from physical exercise than carriers of the normal allele [11]. GYS1 is located on chromosome 19q13.3, a region that has in several linkage studies been linked to MetS and T2D associated phenotypes [12][13][14][15][16][17]. Further, the GYS1 locus was in the HERITAGE family study linked to glucose effectiveness in response to endurance exercise [18].
GYS1 is separated only by 4.1 million base pairs from the gene coding for apolipoprotein E (APOE), which constitutes three common genetic isoforms in plasma and is known to play an important role in lipid metabolism [19]. The APOE4 isoform encoded by the e4 allele is associated with elevated serum totaland low density lipoprotein (LDL)-cholesterol concentrations [20,21] and with coronary heart disease (CHD) [22,23]. In addition, a -219 (G.T) polymorphism in the APOE promoter has in vitro been shown to decrease transcriptional activity of APOE [24] and has been reported to associate with severity of coronary artery disease [25] and increased risk for myocardial infarction [26].
Given the considerations above, we set out to test 1) whether the GYS1 polymorphism is associated with CV mortality in individuals from a large T2D family study from Finland, the Botnia Study. In particular we were interested in putative gender differences as the GYS1 polymorphism has earlier been associated with myocardial infarction only in males, 2) whether physical exercise would act as an environmental factor interacting with the effect associated with the GYS1 polymorphism as this has earlier been shown to be associated with defect in stimulation of glycogen synthase protein levels after muscle stimulation, and 3) to test if our results with the GYS1 polymorphism are independent of the adjacent APOE. As the endpoint we used CV mortality after a median follow up period of 8 years.

Study Population
The Botnia Study was initiated in 1990 and represents a large population-based family study in Finland and Sweden aiming at identification of genes increasing susceptibility to T2D, MetS and related disorders. Details of the study cohort, sampling strategy as well as anthropometric and metabolic measurements have been described in detail [27,28]. The study protocol was approved by the local ethics committees and an informed consent was obtained from each subject before participating in the study. The present study was restricted to the original Botnia cohort of 4654 subjects from 965 families (2142 males, 2512 females, age 58.2613.8 years) from Western Finland. At the baseline examination, a structured questionnaire was completed by specially trained nurses, covering information about diseases other than T2D (particularly hypertension, coronary heart disease, myocardial infarction and stroke) and data on smoking habits and physical activity during work and leisure time. Both previous and current smokers were recorded as smokers. Physical activity level during work was defined on a scale from 0 to 6 according to level of physical activity (0 coding for no work and 6 for highest level) while physical activity during leisure time was estimated by a scale from 1 to 3 (1 = almost no activity at all; 2 = sometimes, but not regular; 3 = regular physical activity). Information on work and leisure time physical activity was combined to obtain an estimate of total physical activity level and classified as: 1) no physical activity or low physical activity (work level of 0 to 2 in combination of leisure time level of 1); 2) normal to high physical activity (work activity level 0-2 in combination of leisure time activity of .1; or work activity level $3 in combination of any leisure time activity level). When division between high and normal physical activity was needed, normal physical activity was defined as work activity level $3 and leisure time of ,3 and high physical activity was defined as leisure time activity of 3 in combination with any work activity level. Glucose tolerance, assessed by an oral glucose tolerance test, and MetS were defined according to current World Health Organization (WHO) criteria [29]. Insulin resistance was estimated as the Homeostasis Model Assessment index (HOMA IR = fasting serum insulin*fasting plasma glucose/22.5).
Total and CV mortality was assessed with median follow up time of 7.9 years and mortality data were obtained from central death-certificate registry in Finland. CV mortality was classified using the 9 th revision of the International Classification of Diseases (CV diagnosis codes 390-459) before 1997 and the 10 th revision (codes 100-199) thereafter. Causes of death were classified as 1) CV death (CHD, cerebrovascular disease (including both thrombotic stroke and cerebral haemorrhage) or other CV events (including pulmonary embolism, abdominal aortic aneurysm, hypertensive complications, general atherosclerosis and peripheral artery disease with gangrene) or 2) other causes of death (neoplasma, violent or other).

Genotyping
A total of 4654 subjects were genotyped for the XbaI polymorphism in intron 14 (rs8103451) of GYS1 and for the APOE isoforms encoded by amino acid substitutions at residues 112 (rs429358) and 158 (rs7412), for the -219G.T promoter polymorphism (rs405509). The XbaI polymorphism in GYS1 was genotyped using single base pair extension on AB3100 (Applied Biosystems) and the APOE polymorphisms were genotyped using allelic discrimination on AB7900 at the SWEGENE DNA genotyping Laboratory. Before any analyses were performed, the expected risk-genotypes for GYS1 and APOE were defined as CT or TT (GYS1 XbaI), e3e4 or e4e4 (APOE codon 112 and 158 polymorphisms) and TT (APOE -219 polymorphism), respectively. Riskalleles were defined according to previous T2D and MetS association study results for GYS1 XbaI [4,7,8] and reports on APOE and risk of coronary disease [22,25,26]. To assure high quality of the produced genotypes, a random sample of 17.8% of all GYS1 XbaI genotypes were repeated using PCR and restriction fragment length polymorphism and the concordance rate was 99.9% [30].

Statistical Analysis
Allele-and genotype frequencies between groups were compared by the x 2 test or by Fisher's exact test whereas multiple regression was used to compare clinical variables between groups, adjusting for age, sex and BMI. Hardy-Weinberg equilibrium (HWE) was tested using exact test (http://ihg.gsf.de/cgi-bin/hw/hwa1.pl) with alpha level of ,0.05 for rejection. For the survival analyses the data were treated as left truncated and right censored, meaning that age was the basic time variable. Survival curves were obtained with the Kaplan-Meier estimator, and nonparametric two-sample tests for genetic effects were performed with the logrank test. Covariates from the baseline visit were used. Effects of genetic and clinical variables on survival time were analysed with uni-and multivariate Cox regression analyses, stratified for sex and using a robust variance estimate to adjust for within family dependence by treating each pedigree as an independent entity when calculating the variance. The univariate analyses were performed to obtain relevant set of variables for multivariate analyses, therefore these p-values were not corrected for multiple testing. The multivariate Cox models were obtained by stepwise forward inclusion of the covariates and statistical significance of the model was analysed using the Wald test. Individuals with missing data for any of the covariates were excluded from the analyses. Due to missing data on microalbuminuria (data missing for 35%) this variable was not included in the multivariate analysis.
Multiple tests were performed within the study (2 genes with 3 polymorphic sites and subanalyses in males and females). Concerning the XbaI polymorphism in males subanalyses were also performed according to physical activity level (low or normal to high). We did not correct for the number of analysed genes and polymorphisms as this study was designed to test the hypothesis that the T-allele of the GYS1 XbaI polymorphism could be associated with CV mortality and as the APOE markers were studied to test if the GYS1 results are independent of the adjacent APOE. For the gender-specific analyses and for the analyses in individuals with different physical activity levels we report both non-adjusted (p) and adjusted (p c ) p-values. The gender-specific analyses were multiplied with a factor of 3 (3 groups; all, males, females) and the physical activity analyses with a factor of 6 (3 groups with either low or normal to high physical activity).
All statistical analyses were performed using Number Crunching Statistical Systems version 2004 (NCSS; Kaysville, Utah, USA) or R (www.r-project.org). Two sided p-values of less than 0.05 were considered statistically significant. Estimates of linkage disequilibrium were calculated using the Haploview program [31]. Power calculations were performed using the normal distributions for the coefficient estimates in the Cox regression model [32].

Clinical and metabolic risk factors for CV mortality
During a median follow-up time of 7.9 years, 749 of the 4654 individuals (16.1%) had died and of them 409 (54.6%) due to CV causes (Table 1). Total mortality was slightly higher among males than among females (17.4 vs. 15.0%, p = 0.029), while frequency of CV mortality did not significantly differ between males and females (9.2 vs. 8.4%, p = 0.32). Subjects who died of CV causes had lower high density lipoprotein (HDL) cholesterol levels compared to both living subjects (p,0.0001) and individuals who died of other than CV causes (p = 0.0009). They also had higher triglyceride levels (p,0.0001) and higher frequency of T2D (p,0.0001), MetS (p = 0.0002), hypertension (p = 0.015), microalbuminuria (p,0.0001), earlier CV events (,0.0001) and lower physical activity level than subjects who were alive. CV death was associated with higher BMI (p = 0.046), total cholesterol levels (p = 0.0037), frequency of T2D (p,0.0001) and earlier CV events (p,0.0001) than death of other causes (Table 1).
Male gender, abdominal obesity, dyslipidaemia, T2D, hypertension, microalbuminuria, earlier CV events, smoking and low physical activity level were significant predictors of CV mortality among all individuals in univariate Cox regression analyses (Table 2). Gender specific univariate analyses identified low HDL cholesterol, T2D, hypertension, microalbuminuria, earlier CV events and physical inactivity as significant risk factors in both genders. Smoking was a significant risk factor only among male subjects while abdominal obesity and elevated triglyceride levels were significant predictors of CV death only in females (Table 2).
In multivariate analyses T2D, elevated fasting insulin concentration, earlier CV events, low physical activity and smoking were significant risk factors for CV mortality in males (model 1 in Table 3). In females T2D, high fasting glucose concentration, hypertension, earlier CV events, and physical inactivity were significant risk factors for CV mortality (model 1 in Table 4). Due to lack of data for a large part (35%) of the study subjects, microalbuminuria was not included in the multivariate model.

GYS1 XbaI as a genetic predictor for CV mortality
The frequency of the XbaI risk genotypes (CT or TT) did not significantly differ between patients who died of CV causes, other causes or survivors when all subjects were included in the analyses (13.3%, 10.5%, and 12.0%) ( Table 5). However, in gender-specific analyses, males with CV death had more often the CT/TT geno-  Figure 1A). The CT/TT genotypes did not predict mortality due to other than CV causes. Furthermore, when analysed together the GYS1 XbaI polymorphism, T2D, earlier CV events, low physical activity and smoking were significant risk factors for CV mortality in males (model 2 in table 3).
Does physical activity influence the effect associated with the genetic variation in GYS1 on CV mortality risk? CV mortality was significantly higher among individuals with low physical activity level compared to individuals with normal (29.9 vs. 7.1%, p,0.0001, corrected for sex and age) or high (29.9 vs. 5.7%, p,0.0001) physical activity level. The difference was not significant between groups reporting normal or high physical activity level (7.1 vs. 5.4%, p = 0.35) suggesting minimal or no protective effect above a normal level of physical activity on CV mortality risk. In a multivariate Cox regression analysis both physical activity (hazard ratio (HR) 3.2 [2.2-4.6], p,0.0001, p c ,0.0001) and the XbaI polymorphism (HR 2.6 [1.7-3.8], p,0.0001, p c ,0.0001) were strongly associated with CV mortality. While physical activity itself (normal or high) had a strong protective effect on CV mortality, this effect was attenuated in carriers of the CT/TT-genotypes of the XbaI polymorphism; physically active males with the CT/TT genotypes had a 2.7-times higher risk for CV mortality compared to CC-genotype carriers (HR 2.7 [1.8-4.1], p,0.0001, p c ,0.0001) (Figure 2).

APOE polymorphisms as genetic predictors for CV mortality
The frequency of the APOE e2/e3/e4 risk genotypes (e3e4 or e4e4), the APOE -219 risk genotype (TT) or the risk genotype combination of APOEe and APOE -219 (-219TT/e4) did not   significantly differ between individuals who died of CV causes and other subjects (Table 5). However, females in the CV mortality group had more often the APOE -219 TT-genotype, in particular -219TT/e4 compared to surviving females (26.9 vs. 19.8%, p = 0.019, p c = 0.057 and 17.5 vs. 9.4%, p = 0.00048, p c = 0.0014). No effect of the APOE variants was observed in males, this effect being restricted to females in whom both the APOE e4-allele, the -219 TT-genotype and their combination were significant predictors of CV mortality (Table 2 and Figure 1B), but not of non-CV mortality. When genetic and non-genetic factors were included in the analysis of risk of CV death, the APOE risk genotype combination, high fasting glucose level, high BMI, hypertension, earlier CV events and low physical activity were significant risk factors for CV mortality in females (model 2 in table 4).

Cholesterol levels and CV mortality according to APOE e3e4/e4e4 and -219 TT genotypes
The APOE e3e4/e4e4 and -219 TT genotypes (and their combination) were associated with increased total and LDL cholesterol levels in both males and females (after adjustment for age, T2D and BMI). Interestingly, although the APOE e3e4/e4e4 variants had a statistically weaker effect on total cholesterol in female carriers ( In males, the APOE e3e4/e4e4 genotypes affected total cholesterol in carriers (5.961.0 vs. 5.661.2 mmol/l, p = 0.031, for the e3e4/e4e4 genotype vs. other genotypes, respectively) but not after correcting for multiple testing (p c = 0.19) and not in non-carriers (5.761.0 vs. 5.661.1 mmol/l, p = 0.31) of the APOE -219 TT genotype. As in females, the APOE -219 TT genotype had no significant effect on cholesterol values neither in carriers nor in non-carriers of the APOE e3e4/e4e4 genotypes. In contrast to females, neither APOE e3e4/e4e4 nor APOE -219 TT predicted CV mortality in males.

Independency between GYS1 and APOE as risk factors for CV mortality
To investigate whether the 'at-risk' genotypes of GYS1 and APOE contributed independently to the CV mortality risk, we performed Cox regression analyses by entering both genes into the equation.

DISCUSSION
The key findings of the present study were that 1) the XbaI polymorphism in GYS1 was associated with CV mortality in males; 2) although physical activity markedly reduces risk of CV death, this protective effect was attenuated in male carriers of the XbaI polymorphism; and 3) despite the fact that GYS1 is adjacent to APOE on chromosome 19q13, the effect of the GYS1 polymorphism on CV mortality is independent of the effect of APOE, which exerts a strong effect on CV mortality risk by its own. Interestingly, this risk seems to be restricted to females and cannot fully be explained by the effect of the APOE alleles on cholesterol levels. Several studies performed in different ethnic populations have reported linkage to chromosome 19q13 for LDL-cholesterol- [33][34][35][36][37] or triglyceride levels [38], obesity [39,40], as well as for insulin resistance and T2D related phenotypes [12][13][14][15][16][17] but the underlying genetic variants have not been identified. In addition, glucose effectiveness in response to exercise training as well as significant sex specific differences in heritability models and sex interaction for HDL cholesterol have been mapped to the 19q13 region [18,41]. We therefore set out to study the contribution of two candidate genes in this region, GYS1 and APOE to CV mortality risk, focusing particularly on the role of putative interaction between GYS1 polymorphism and gender and/or physical activity level to affect the CV mortality rate.
The GYS1 XbaI polymorphism was significantly associated with increased risk for CV mortality in males, a result supported by our previous independent finding of an association between myocardial infarction and this particular polymorphism only in males in another study population [7].
As anticipated, a low physical activity level was a severe risk factor for CV mortality. The novel finding of our study was that the protective effect of physical exercise was attenuated in carriers of the XbaI polymorphism. This goes along with the hypothesis advanced by a Canadian study [11] that carriers of the risk Tallele have a defect in their ability to increase the glycogen synthase protein in response to neuromuscular electrical stimulation (as a proxy for physical exercise) [11]. An increase in glycogen synthase protein would promote glycogen formation which, in turn, could have a beneficial effect on exercise capacity. The downside of this message is that all individuals would not respond to physical exercise in the same way. The positive message is that the ''non-responder'' group is relatively small (frequency of CT/ TT genotypes in the population is only 12%) and in 88% of the population exercise exerts a highly beneficial and protective effect on risk of CVD.
We have no apparent explanation for why the effect of the XbaI polymorphism was restricted to males. One potential explanation could be that women have less muscle mass and muscular strength than men, but also a tendency to metabolise fat rather than carbohydrate during exercise [42]. Moreover, women seem less vulnerable to exercise-induced sudden death [42]. A potential explanation could be that both exercise training and oestrogen increase Akt phosphorylation and glycogen synthase kinase-3 inactivation leading to increased glycogen synthase activity. Interestingly, markedly higher myocardial Akt nuclear activity has been reported in females than in males as well as in precompared to post-menopausal woman [43]. If this also applies to skeletal muscle it could provide a potential explanation for the observed gender-specific effect.
Our results are in agreement with the earlier results reporting either the APOE e4-allele or the APOE -219 TT-genotype as risk factors for CVD and/or mortality [22,23,25,26] but, interestingly, we could observe this effect only in females. Caution is, however, warranted in the interpretation of the gender-specific effects of APOE as the study included a large number of patients with T2D. It is known, that men with T2D have an excess mortality compared with women with T2D resulting in a relative increase in the frequency of female T2D patients with aging. It is therefore still possible that the APOE polymorphisms had an effect in male T2D resulting in premature death. A sub-analysis of men and women divided by the median of age did, however, not support such an explanation. Also, power is reduced when the analysis is restricted to gender. The power in a Cox regression analysis depends among other things on the accrual time during which patients are recruited, mean time to failure and the expected effect sizes (http://biostat.mc.vanderbilt.edu/twiki/bin/view/ Main/PowerSampleSize) [44]. We used the standard Normal theory to calculate the power to detect a HR of a certain size [32]. Our study had a 73-96% power to detect hazard ratios of 1.5 for the analysed polymorphisms.
In conclusion, we demonstrate a protective effect of physical activity on CV mortality. However, in male subjects this effect was attenuated in carriers of the rare allele of the XbaI polymorphism in GYS1. This finding could be compatible with a previous demonstration of defective increase in the glycogen synthase protein in carriers of this polymorphism. We could exclude that the association between the GYS1 polymorphism and CV mortality was due to the adjacent APOE gene. Instead, we demonstrated that this gene exerted an increased risk of CV mortality in females. These findings re-emphasize the need to consider the effect of genetic variants in complex diseases in concert with their environmental triggers but also to evaluate whether females and males respond differently to genes and the environment.