Genetic Scores of eNOS, ACE and VEGFA Genes Are Predictive of Endothelial Dysfunction Associated Osteoporosis in Postmenopausal Women

The present study aimed to examine the participation and contribution of endothelial nitric oxide synthase (eNOS), angiotensin converting enzyme (ACE) and vascular endothelial growth factor (VEGFA) genes for the risk of endothelial dysfunction (ED)-associated osteoporosis risk in postmenopausal women of Punjab, India. Women with ED were categorized into women with osteoporosis (n = 346) and women without osteoporosis (n = 330). They were examined for selected SNPs within eNOS, ACE and VEGFA genes. Linear regression analysis revealed a positive association of ED with bone mineral densities (BMDs) at femoral neck (r2 = 0.78, p < 0.001) and lumbar spine (r2 = 0.24, p = 0.001) after Bonferroni correction. Three susceptibility haplotypes were exposed within eNOS (CTAAAT), ACE (ACDG) and VEGFA (GATA) genes. Bearers of CTAAAT (OR 2.43, p = 0.007), ACDG (OR 2.50, p = 0.002) and GATA (OR 2.10, p = 0.009) had substantial impact for osteoporosis after correcting the effects with traditional risk factors (TRD).With uncertainty measure (R2h) and Akaike information criterion (AIC), best fit models showed that CTAAAT manifested in multiplicative mode (β ± SE: 2.19 ± 0.86, p < 0.001), whereas ACDG (β ± SE: 1.73 ± 0.54, p = 0.001) and GATA (β ± SE: 3.07 ± 0.81, p < 0.001) expressed in dominant modes. Area under receiver operating characteristic curve using weighted risk scores (effect estimates) showed substantial strength for model comprising TRD + GATA (AUC = 0.8, p < 0.001) whereas, model comprising TRD + GATA + CTAAAT exhibited excellent ability to predict osteoporosis (AUC = 0.824, p < 0.001)


Introduction
Osteoporosis is a complicated skeletal disorder, which is influenced by multiple factors and influences multiple elements of health, especially in women after menopause owing to endocrinological, physiological and psychological upheaval. Primarily, it is confirmed by low bone mineral density (BMD) that leads to brittle bones and fractures. The scientific literature reveals that during this phase of life, reduced hormone levels severely affect the vascular endothelium, leading to endothelial dysfunction [1][2][3]. Endothelial dysfunction is an abnormal state, whereby the endothelial cell walls fail to balance between vasorelaxation and vasoconstriction, due to reduced synthesis and availability of nitric oxide (NO), which largely mediates, manages and maintains it. Post-menopause-associated low estrogen levels downregulate NO production, resulting in impaired vascular tone and reactivity [4,5].
phenotypes [24]. The major allele G of a functional SNP rs2010963 (−1154G/C) of VEGFA gene has been observed to be associated with reduced VEGF transcription [25].
Many reports have analysed independently the role and contribution of these three pertinent candidates-eNOS, ACE and VEGFA genes-in vascular function [17,21,25]. These genes regulate endothelial function in various systems of the body through different mechanisms, especially in cardiovascular pathology, but their participation in bone vasculature remains elusive. Apropos to this, the present study aims to investigate the genetic contribution of some important SNPs of these genes through their gene-gene, gene-environment and haplotype specific interactions as genetic determinants and predictors of endothelial dysfunction inflicted osteoporosis in postmenopausal women.

Subjects
In the present study, preliminary examination of 2167 postmenopausal women who visited orthopedic wards of prominent hospitals (Rajindra Hospital, Aggarwal Orthopedic Hospital and Amrit Sagar Hospital) of Punjab, India was conducted to establish their baseline characteristics. After following the exclusion and inclusion criteria, 919 women were enrolled ( Figure 1). Their endothelial damage was tested and reactive hyperemia index (RHI), an indicator of endothelial dysfunction, was assessed (EndoPAT 2000 device by Itamar Medical Technology Ltd., Caesarea, Israel). After exclusion of 243 women who had normal vascular function, 676 postmenopausal women finally participated. Their BMD was examined (dual energy x ray absorptiometry: DXA) at L1 to L4 vertebrae of lumbar area (BMD_LS) and at neck of the femur (BMD_FN). Based on criteria of T scores given by WHO, 346 women had scores ≤ 2.5 and were categorized as women with osteoporosis. The remaining 330 women had T scores ≥ 1 and hence were categorized as women without osteoporosis. Considering aim of the study which demands clear pathologies and minimum stratification, so as to deduce exclusive effect of endothelial dysfunction, women having T-score between −1.0 and −2.5 (osteopenia) and women having fractures were not included. It is reasonable to investigate effect of endothelial dysfunction on confirmed low BMD (osteoporosis) whereas, osteopenia is midway phenotype where the effects of low bone mass has just started to express. Similarly, fractures may confound the analysis because these can be the outcomes of bone cysts, cancers or excessive use of glucocorticoids and bisphosphonates, irrespective of low BMD.
Only those subjects who had given their informed written consent were allowed to participate in the study. In order to avoid any bias, the case control status was coded and blinded to the researchers. The protocol of the study was approved by institutional ethical review board (Reference no. IEC2017/05, dated 20 January 2017) and conformed strictly to ethics for medical research involving human subjects (Helsinki Declaration).

Description of Variables
Body mass index (BMI) was calculated by Quetelet's index, which is weight in kilograms over height in meters squared (kg/m 2 ). Detailed menopause status for recording age and years since menopause (YSM) were verified from their medical records or through personal interviews. Systolic (SBP) and diastolic blood pressure (DBP) measurements were conducted, two times on resting subjects (at least for 10 min) after an interval of 3 min and their mean values were recorded. Lipid values of triglycerides (TG), high density lipoproteins (HDL) and total cholesterol (TC) were analysed by using assay kits (Erba Mannheim, London, UK) with one step enzymatic methods on Lisa scan plate reader (Erba Mannheim). It can identify a minimum of 0.1 mg/L of component. Low density lipoproteins (LDL) levels were calculated with the Friedewald equation. The inter-assay and intra-assay coefficients of variation (CVs) were 6.2 and 6.7, respectively.

Description of Variables
Body mass index (BMI) was calculated by Quetelet's index, which is weight in kilograms over height in meters squared (kg/m 2 ). Detailed menopause status for recording age and years since menopause (YSM) were verified from their medical records or through personal interviews. Systolic (SBP) and diastolic blood pressure (DBP) measurements were conducted, two times on resting subjects (at least for 10 min) after an interval of 3 min and their mean values were recorded. Lipid values of triglycerides (TG), high density lipoproteins (HDL) and total cholesterol (TC) were analysed by using assay kits (Erba Mannheim, London, UK) with one step enzymatic methods on Lisa scan plate reader (Erba Mannheim). It can identify a minimum of 0.1 mg/L of component. Low

Examination of Endothelial Function
A non-invasive technique using an EndoPAT 2000 device was conducted to check reactive hyperemia index (RHI), an indicator of endothelial function. According to the manufacturer's guidelines patients were counseled to not take any food items or beverages containing methylxanthine (caffeine, tea, chocolate, etc.,) at least 5 h before examination. After taking mean blood pressure of two measurements, finger probes were positioned, which measured endothelium dependent change in vascular tone. First of all baseline pulse variation was noticed and recorded for about six minutes and then occlusion to brachial artery was done up to at least 250 mmHg of systolic blood pressure with sphygmomanometer cuff. Non-endothelium-dependent alterations of the contra-lateral arm were compared to assess vascular tone. Now occlusion was released after 5 min which causes flow mediated dilation (FMD) and EndoPAT recorded RHI values as increase in pulse amplitude tone (PAT). Subjects having RHI < 1.67 were confirmed to have endothelial dysfunction. These measurements were operator independent being accomplished by in-built algorithm based dedicated software.

BMD Evaluation
DXA was employed to test BMD by using Hologic QDR 4500 system (Hologic Inc. Waltham, MA, USA). T scores were inferred based on their comparisons with peak bone mass of average 30 years old young individual of the same gender. On the basis of these T scores, those women were confirmed to have osteoporosis who had T score ≤ 2.5 and women without osteoporosis; who had T scores ≥ 1. The QDR system was standardized according to the guidelines of the manufacturer before testing. Inter-and intra-assay CVs for the checking of the BMD_FN and BMD_LS were 5.0 and 5.8, respectively.

Selection of the SNPs and Genotyping
In the clinical arena, it is well understood that functional manifestations of these three genes (eNOS, ACE and VEGFA) strongly impact endothelial function by inducing NO bioavailability in the endothelium, generation of super oxide anions that degrade NO and induce angiogenesis [17,[21][22][23][24][25]. Functional SNPs along with other significant candidates for each gene were selected on the basis of three points; (i) these SNPs should have been earlier verified by submitted information on the reference SNP collection databank at dbSNP (http://www.ncbi.nlm.nih.gov/snp), (ii) these SNPs should have been formerly identified as having an association with endothelial dysfunction (iii) all the SNPs should have been polymorphic with minor allele frequency > 0.05. Following these criteria, six SNPs of eNOS gene i.e., rs2070744, rs1799983, rs1800780, rs3918181, rs891512 and rs1808593, four SNPs of ACE i.e., rs4459609, rs1800764, rs1799752 and rs4343 and four SNPs of VEGFA i.e., rs2010963, rs699947, rs833061 and rs1570360 were selected. After extraction of deoxyribose nucleic acid (DNA) from whole blood with a normal salting-out procedure, PCR was used to amplify DNA with reaction mixture of 25 µL. High conformity restriction enzymes (NEBS, Hertfordshire, UK) were used to digest amplicons. Depending upon the product size, genotypes were assessed and typed on two percent to three percent agarose gels. The confidentiality of the subjects was maintained and all the experimental work was blinded to the clinical and case-control status in order to avoid any subjective bias. To verify reproducibility of genotyping, 15 percent of the respective samples were re-analyzed.

Population Stratification Analysis and Statistical Power of Genetic Association
Case-control design of the study may assume false positive inferences because of underlying population stratification (PS), which shows misleading differences in allele frequencies due to different ancestries, otherwise considering subpopulations of the population belonging to same ancestry. Using the software Arlequin ver. 3.0 [26], population comparisons were performed to compute pairwise fixation index (F st ), an indicator of how populations differ genetically. F st for within population differentiation was observed to be 0.035 ± 0.019, which showed that no considerable PS existed between groups in this study population that may confound the genetic analysis. Genetic association versus sample size was examined with 676 subjects (346 cases and 330 controls) by using the Power for Genetic Association Analysis (PGA) Package [27], which deduced appropriate sample size and minimum detectable relative risk (MDRR) using SNPs and haplotype effects under various models. A preliminary analysis indicated that this sample size (n = 676) would deliver more than 90 percent power to differentiate minimum genotype relative risk (MGRR) of 2.0 with apportioned value of minor allele frequency of at least 0.21 (rs1799883 in present sample) at significance of 0.05. Analysis with haplotype effect module suggested that sample size of 676 would deliver MDRR of 1.5 with substantial power (>90%). Further, power of genetic association was checked at more stringent significance levels (0.01, 0.001), which suggested that this sample size is sufficient to discriminate between genotype relative risk of 2.0 under all the genetic models (additive, dominant and recessive) with more than 80 percent statistical power.

Statistical Analysis
To examine differences between proportions or categorical data of the study groups, a chi-square test was applied, whereas a t test (Student's t) or Mann-Whitney-Wilcoxon rank-sum test was used for continuous data. Gene counting was done for calculating minor allele frequencies and Hardy-Weinberg equilibrium was evaluated with Fischer's exact test. Inter-relationship of BMD (target variable) at both sites and endothelial dysfunction (predictor variable) was assessed with linear regression analysis and summarized in whisker plots. Variance inflation factor (VIF) was calculated to check collinearity between both the explanatory variables (BMD_FN and BMD_LS). Extent of association between all the variables with BMD was examined with univariate regression analysis (GLM procedure). Those variables which showed significance in univariate model were involved in binary logistic regression analysis (backward step method) to identify their independent relationship with the dependent variable along with interaction analyses between them. Full logistic regression analysis was applied to calculate association of designated alleles with the risk of endothelial dysfunction associated osteoporosis in codominant, dominant, recessive, and multiplicative genetic models. Assuming genetic disease penetrance of 1, r and r 2 for genotypes AA, AB and BB respectively, codominant model specifies that risk of osteoporosis is increased by r-fold for genotype AB and r 2 -fold for genotype BB. For dominant model, either one or two copies of allele B are required for r-fold increased risk, recessive model demonstrates that two copies of allele B are required for r-fold increased risk and multiplicative model indicates that the osteoporosis risk is increased by r-fold with each additional B allele. Two locus epistasis effect between SNPs versus risk variables (gene-environment relationship) were analysed using the epiSNP software [28]. Haplotypes were generated using genotype data with the software Arlequin ver. 3.0 [26]. In order to understand risky haplotypes, multivariable regression analysis was used to compute odds ratios in unadjusted model and a model with adjusted values for risk variables after taking most prevalent haplotype as referent. Risky haplotypes were checked more stringently to understand their best modes of impact in different genetic models (Dominant, Recessive, Multiplicative and general). Best fit model was identified with Wald test and Akaike information criterion (AIC = −2 log-likelihood + 2 × number of parameters). Degree of haplotype uncertainty (R 2 h) was also investigated by the method of Stram et al. [29].

Receiver Operating Characteristic Curve Analysis
For the evaluation of discriminatory ability and predictive accuracy of collective effects of alleles within a gene (haplotype) and/or traditional risk factors (BMI, SBP and TG), area under receiver operating characteristic curve (AUROC) was modeled with risk scores. Risk scores for traditional risk factors (TRD) were taken as respective β coefficients (unweighted scores) obtained from logistic regression analysis. These values were further standardized by multiplying lowest absolute value of the coefficient with a number to become value of 1 and all β values were multiplied by that number to round them to the closest integer (weighted scores). For calculating genetic risk scores of haplotypes, individual scores were calculated based on the carriage of risky alleles in the susceptibility haplotype of each gene. Genotypes of risk (R) and non-risk (N) alleles were apportioned scores of 0 (NN), 1 (RN) and 2 (RR) for each SNP participating in respective susceptibility haplotype. Final risk score was deduced by summing up all SNP wise risk score of every individual. The values of area under curve (AUC), from 0.6-to 0.7, 0.7 to 0.8 and 0.8 to 0.9 were considered weak, acceptable and excellent, respectively.

Analysis of Variables at Baseline and Genetic Correlates
Baseline variables of postmenopausal women having endothelial dysfunction categorized according to osteoporosis status are summarized in Table 1. No evidence of statistically significant differences between the two groups for age, YSM, DBP, LDL, HDL, and TC (p > 0.05) was observed. Values of BMI, SBP and TG were considerable higher in women having osteoporosis than women without osteoporosis and these differences were statistically significant (p < 0.001) between both the groups. Albeit, all the postmenopausal women participated in the present study had confirmed endothelial dysfunction (RHI < 1.67), nonetheless, its effect was considerably enlarged in women having osteoporosis and differences between the groups were statistically dissimilar (p < 0.001). Similarly, it is evident that adjusted values of BMD at femoral neck and lumbar spine were markedly low in women having osteoporosis than women without it (p < 0.00). Genotype frequencies of all the SNPs of three genes (eNOS, ACE and VEGFA) were within range (p > 0.05) of Hardy-Weinberg equilibrium. Minor allele frequencies (MAF) of rs1800780, rs3918181 and rs1808593 of eNOS gene were similar between both groups (p > 0.05), however, for rs2070744, rs1799983, and rs891512, MAF differed significantly (p < 0.05) between osteoporotic and non-osteoporotic women. MAF of three SNPs of ACE gene i.e., rs1800764, rs1799752 and rs4343 were observed to be significantly dissimilar (p < 0.05), however MAF of rs4459609 was similar (p = 7.86) between women having osteoporosis and women without it. MAFs of SNP rs833061 within VEGFA gene were detected to be non-significant (p = 0.793) between both the groups of women, whereas MAFs of other three SNPs; rs2010963, rs699947 and rs1570360 were observed to be considerably different (p < 0.05) between them.

Identification of Independent Risk Variables
Univariate analysis of risk variables were analysed for assessing their impact on osteoporosis (Table 2), which showed that BMI, YSM, SBP, TG and BMDs at both the sites of femoral neck and lumbar spine were significant risk variables. Further, their analysis in binary logistic regression analysis showed that BMI (≥30 kg/m 2 ), YSM (>5 years), SBP (>120 mmHg), TG (>150 mg/dL) and BMDs at femoral neck (<0.7 g/cm 2 ) and lumbar spine (<0.8 g/cm 2 ) were independent risk predictors (p < 0.05) for osteoporosis risk, whereas, DBP, TC, LDL and HDL did not influence osteoporosis risk (p > 0.05).

Genotype Specific Implications of Genes through Different Genetic Models
Role and relevance of individual SNPs of eNOS gene were analysed by comparing carriage of risky allele between women with osteoporosis and women without osteoporosis by taking major allele as referent (Table 3). Genotype specific codominant, dominant, recessive and multiplicative models after adjusting the effect of variables revealed that minor allele C of rs2070744 were associated in codominant in heterozygous (   Assuming genetic disease penetrance of 1, r and r 2 for genotypes AA, AB and BB respectively, codominant model specifies that risk of osteoporosis is increased by r-fold for genotype AB and r 2 -fold for genotype BB. For dominant model, either one or two copies of allele Bare required for r-fold increased risk, recessive model demonstrates that two copies of allele B are required for r-fold increased risk and multiplicative model indicates that the osteoporosis risk is increased by r-fold with each additional B allele. Significant values are shown in bold face. Assuming genetic disease penetrance of 1, r and r 2 for genotypes AA, AB and BB respectively, codominant model specifies that risk of osteoporosis is increased by r-fold for genotype AB and r 2 -fold for genotype BB. For dominant model, either one or two copies of allele Bare required for r-fold increased risk, recessive model demonstrates that two copies of allele B are required for r-fold increased risk and multiplicative model indicates that the osteoporosis risk is increased by r-fold with each additional B allele. Significant values are shown in bold face.

Analysis of Linear Relationship of RHI with BMD
Linear regression analysis ( Figure 2) displayed that RHI predicted the values of BMD linearly at both femoral neck (r 2 = 0.78, p < 0.001) and lumbar spine (r 2 = 0.24, p = 0.001) after Bonferroni correction. Results in the analysis indicated that RHI and BMD were positively correlated and that with gradual increase of RHI (normal endothelial function, otherwise lesser the RHI < 1.67, more is the severity of endothelial dysfunction), BMDs at both sites also increased. and BMD, which reassured that RHI positively associated with BMD at femoral neck (p < 0.001) and lumbar spine (p = 0.009).

SNP-SNP Cross Talks, Risky Traits and Their Modes of Association
Several genotype specific single marker effects within SNPs of eNOS, ACE and VEGFA genes on risk covariates were deduced. All significant single marker effects (p < 0.05, r > 0.04) were further analysed for pair wise epistatic effects with Bonferroni corrections (Table 6). It was revealed that 10 SNP pair associations with risky traits were evident of osteoporosis risk. Functional SNP of eNOS gene; rs2070744 showed epistatic re- In multicollinearity analysis, VIF was observed to be 1.3, which suggested that BMDs at femoral neck and lumbar spine were not linearly dependent on each other. Whisker plots were generated after omitting outliers to clarify this relationship of RHI and BMD, which reassured that RHI positively associated with BMD at femoral neck (p < 0.001) and lumbar spine (p = 0.009).

SNP-SNP Cross Talks, Risky Traits and Their Modes of Association
Several genotype specific single marker effects within SNPs of eNOS, ACE and VEGFA genes on risk covariates were deduced. All significant single marker effects (p < 0.05, r > 0.04) were further analysed for pair wise epistatic effects with Bonferroni corrections ( Table 6). It was revealed that 10 SNP pair associations with risky traits were evident of osteoporosis risk. Functional SNP of eNOS gene; rs2070744 showed epistatic relationship with SNPs rs4343, rs1799983, rs1800764 and rs891512 influencing RHI (p = 0.001), LDL (p = 0.005), TG (0.001) and SBP (p = 0.003) through interactive (I), additive x additive (AA), dominant × additive (DA) and additive × dominant (AD) modes respectively. Interestingly, SNP pairs rs2070744-rs1800764 impacted TG through DA mode in control subjects (p = 0.041) also. Another SNP pairs which showed impact on the risk of osteoporosis through AA, dominant × dominant (DD) and I mode were rs4343-rs2010963 (p = 0.009), rs1800764-rs1799752 (p = 0.029) and rs1799983-rs891512 (p = 0.032) influencing RHI, TC and RHI respectively. Similarly, SNP rs1799983 coupled with rs699947 (p = 0.001), rs699947 with rs1799752 (p = 0.003) and rs891512 with rs1799752 (p = 0.022) to influence BMI, RHI and LDL through DD, I and DA modes, respectively. SNP pair rs1799983-rs699947 was observed to afflict DD influence in women without osteoporosis also (p = 0.041). Two locus SNP-SNP epistatic links without risk variables have been shown in figure embedded in the Table 6, to have quicker glance without some perplex interactions.

Haplotype Analysis, Their Contribution and Best Mode of Impact for Osteoporosis Risk
SNPs within eNOS gene (in the order of rs2070744, rs1799983, rs1800780, rs391881, rs891512 and rs1808593) developed into 64 possible haplotypes and out of them 29 haplotypes were visible. Of these, 21 haplotypes had frequencies less than 5 percent, therefore, excluded from the analysis. The remaining eight haplotypes captured 85percent of genetic variance of women having normal bone mass and 88 percent of women having osteoporosis ( Table 7).The major alleles at position 1, 2, 5 and 6 and minor alleles at 3 and 4 of studied eNOS SNPs appeared in the form of haplotype TGAAGT was having highest frequency, hence served as referent for the analysis. Minor alleles of all the SNPs except at position 6in the form of CTAAAT appeared to be the risky haplotype for osteoporosis risk (OR 2.80, 95%CI: 1.53-5.13, p = 0.001). Inter group comparisons of this haplotype after Bonferroni corrections exhibited significant differences approaching GWAS p values (p = 1 × 10−8). It was observed to be susceptibility haplotype for the risk of osteoporosis (OR 2.43, 95%CI: 1.22-4.71, p = 0.007), when its influence was examined after adjusting the effects of confounders (BMI, YSM, SBP and TG).

Haplotype Analysis, Their Contribution and Best Mode of Impact for Osteoporosis Risk
SNPs within eNOS gene (in the order of rs2070744, rs1799983, rs1800780, rs391881, rs891512 and rs1808593) developed into 64 possible haplotypes and out of them 29 haplotypes were visible. Of these, 21 haplotypes had frequencies less than 5 percent, therefore, excluded from the analysis. The remaining eight haplotypes captured 85percent of genetic variance of women having normal bone mass and 88 percent of women having osteoporosis ( Table 7).The major alleles at position 1, 2, 5 and 6 and minor alleles at 3 and 4 of studied eNOS SNPs appeared in the form of haplotype TGAAGT was having highest frequency, hence served as referent for the analysis. Minor alleles of all the SNPs except at position 6in the form of CTAAAT appeared to be the risky haplotype for osteoporosis risk (OR 2.80, 95%CI: 1.53-5.13, p = 0.001). Inter group comparisons of this haplotype after Bonferroni corrections exhibited significant differences approaching GWAS p values (p = 1 × 10 −8 ). It was observed to be susceptibility haplotype for the risk of osteoporosis (OR 2.43, 95%CI: 1.22-4.71, p = 0.007), when its influence was examined after adjusting the effects of confounders (BMI, YSM, SBP and TG). Haplotype analysis of SNPs in the order of rs4459609, rs1800764, rs1799752 and rs4343 of ACE gene exhibited 11 visible haplotypes but six haplotypes were excluded as these had lower frequencies (<5 percent) and were not amenable to be used for authentic results. The remaining five haplotypes showed 88-90 of genetic variability in both the groups of women. All the major alleles representing ATIA haplotype emerged to be the most common haplotype in both the groups, so it was taken as referent. All minor alleles of ACE gene SNPs, except at position 1 in the form of haplotype ACDG appeared to be risky (OR 3.03, 95%CI: 1.86-4.88, p < 0.001) and it was confirmed that it conferred 2.5 fold higher risk of developing osteoporosis in postmenopausal women having endothelial dysfunction after correcting the effect of risk predictors (OR 2.50, 95%CI: 1.28-3.96, p = 0.002).
Although the results implied that those postmenopausal women who possessed these susceptibility haplotypes were at higher risk of developing osteoporosis than those women who did not have it, in which best possible way these haplotypes inflicted their maximum effects needed to be identified (Table 8). Their functional effects on BMD were modeled and tested with Wald statistics under dominant, recessive, multiplicative and general modes of inheritance and selection of the best fit model was identified with AIC and R 2 h (Table 8). Analysis clarified that susceptibility haplotype CTAAAT of eNOS gene influenced osteoporosis risk in multiplicative mode (β ± SE: 2.19 ± 0.86, p < 0.001), haplotypes ACDG of ACE gene (β ± SE: 1.73 ± 0.54, p = 0.001) and haplotype GATA of VEGFA gene (β ± SE: 3.07 ± 0.81, p < 0.001) influenced bone mass in postmenopausal women in dominant modes. Models showing values after adjustment for risk covariates; body mass index, years since menopause, systolic blood pressure, triglyceride levels and BMDs at femoral neck and lumbar spine. a Estimated haplotype effect, P-asymptotic value, R 2 h-haplotype uncertainty measure, AIC-Akaike information criterion. Values in bold face show highest R 2 h values and lowest AIC. Dominant effect (women having one copy is at same risk as those women having two copies), Recessive (women having one copy is at the same risk as women having no copy), Multiplicative effect (women having one copy of the haplotype are at intermediate risk than women having no copy or two copies). Significant values are shown in bold face.

Discussion
The results obtained in the present study have highlighted that endothelial dysfunction impacts bone mass through genetic participation of eNOS, ACE and VEGFA genes. Furthermore, this research also illustrates that collaborative effects of genetic variants within these genes along with traditional risk factors are predictive of endothelial dysfunction-affiliated osteoporosis in postmenopausal women. Nonetheless, it is well understood in the clinical arena that hormonal insufficiency in postmenopause phase of life is not only detrimental to bone health, but also influences skeletal vasculature [1][2][3][4] but, whether such connections have genetic connotations, has been largely undefined, primarily because endothelial dysfunction has been considered as a relevant surrogate marker for cardiovascular and cerebrovascular diseases whereas, BMD is used for osteoporosis. In the present analysis, some variants within three noteworthy candidates, which impinge upon endothelial function (eNOS, ACE and VEGFA), have shown significant impact on the risk of osteoporosis through different genetic models. In individual studies, minor alleles; G, D (deletion) and major allele G of the SNPs of eNOS; rs1799983, ACE: rs1799752 and VEGFA: rs2010963 respectively has been associated with lower BMDs in Chinese Han, Turkish and Caucasian women [30][31][32], which is consistent with the findings of this study.
Whether the influence of these alleles capture the overall genetic variance for the risk of osteoporosis is questionable, as their individual effect may fluctuate when other SNPs also contribute, especially when they are non-randomly linked to each other. Only a few studies have investigated the gene-gene-, gene-environmental-and haplotype-specific effects for osteoporosis risk by involving endothelial function-oriented genes. In a previous report by our laboratory with a lesser sample size (n = 456), a hap-

Discussion
The results obtained in the present study have highlighted that endothelial dysfunction impacts bone mass through genetic participation of eNOS, ACE and VEGFA genes. Furthermore, this research also illustrates that collaborative effects of genetic variants within these genes along with traditional risk factors are predictive of endothelial dysfunction-affiliated osteoporosis in postmenopausal women. Nonetheless, it is well understood in the clinical arena that hormonal insufficiency in postmenopause phase of life is not only detrimental to bone health, but also influences skeletal vasculature [1][2][3][4] but, whether such connections have genetic connotations, has been largely undefined, primarily because endothelial dysfunction has been considered as a relevant surrogate marker for cardiovascular and cerebrovascular diseases whereas, BMD is used for osteoporosis. In the present analysis, some variants within three noteworthy candidates, which impinge upon endothelial function (eNOS, ACE and VEGFA), have shown significant impact on the risk of osteoporosis through different genetic models. In individual studies, minor alleles; G, D (deletion) and major allele G of the SNPs of eNOS; rs1799983, ACE: rs1799752 and VEGFA: rs2010963 respectively has been associated with lower BMDs in Chinese Han, Turkish and Caucasian women [30][31][32], which is consistent with the findings of this study.
Whether the influence of these alleles capture the overall genetic variance for the risk of osteoporosis is questionable, as their individual effect may fluctuate when other SNPs also contribute, especially when they are non-randomly linked to each other. Only a few studies have investigated the gene-gene-, gene-environmental-and haplotype-specific effects for osteoporosis risk by involving endothelial function-oriented genes. In a previous report by tive and dominant mode for inflicting endothelial-associated osteoporosis risk in postmenopausal women. A model comprising risk scores of traditional risk factors in addition to that of susceptibility haplotype CTAAAT (eNOS gene) and GATA (VEGFA gene) is capable of being excellent predictor of osteoporosis, likelihood of which prompts future genetic studies to probe it.  Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to ethical issues.