Genetic Determinants of Body Mass Index and Fasting Glucose Are Mediators of Grade 1 Diastolic Dysfunction.

Background Early (grade 1) cardiac left ventricular diastolic dysfunction (G1DD) increases the risk for heart failure with preserved ejection fraction and may improve with aggressive risk factor modification. Type 2 diabetes, obesity, hypertension, and coronary heart disease are associated with increased incidence of diastolic dysfunction. The genetic drivers of G1DD are not defined. Methods and Results We curated genotyped European ancestry G1DD cases (n=668) and controls with normal diastolic function (n=1772) from Vanderbilt's biobank. G1DD status was explored through (1) an additive model genome-wide association study, (2) shared polygenic risk through logistic regression, and (3) instrumental variable analysis using 2-sample Mendelian randomization (the inverse-variance weighted method, Mendelian randomization-Egger, and median) to determine potential modifiable risk factors. There were no common single nucleotide polymorphisms significantly associated with G1DD status. A polygenic risk score for BMI was significantly associated with increased G1DD risk (odds ratio [OR], 1.20 for 1-SD increase in BMI [95% CI, 1.08-1.32]; P=0.0003). The association was confirmed by the inverse-variance weighted method (OR, 1.89 [95% CI, 1.37-2.61]). Among the candidate mediators for BMI, only fasting glucose was significantly associated with G1DD status by the inverse-variance weighted method (OR, 4.14 for 1-SD increase in fasting glucose [95% CI, 1.55-11.02]; P=0.005). Multivariable Mendelian randomization showed a modest attenuation of the BMI association (OR, 1.84 [95% CI, 1.35-2.52]) when adjusting for fasting glucose. Conclusions These data suggest that a genetic predisposition to elevated BMI increases the risk for G1DD. Part of this effect may be mediated through altered glucose homeostasis.

diastolic dysfunction

fasting glucose

grade 1 diastolic dysfunction

Mendelian randomization

multivariable Mendelian randomization

polygenic risk score

type 2 diabetes

Clinical Perspective

What Is New?

Grade 1 left ventricular diastolic dysfunction (G1DD) is a modifiable risk factor associated with incident heart failure with preserved ejection fraction risk, but its genetic drivers are not known.

This study investigated the genetic drivers of G1DD measured by echocardiography in a European ancestry population aged <60 years.

Polygenic risk associated with body mass index variation is also associated with G1DD prevalence and is mediated, in part, by altered glucose homeostasis.

What Are the Clinical Implications?

Body mass index is an important modifiable risk factor for G1DD.

Prevention efforts should be directed to mitigate weight gain in early life to prevent adverse cardiac modeling that predisposes to G1DD and subsequent heart failure with preserved ejection fraction.

If G1DD is observed on echocardiogram, risk reduction strategies should include maintaining a healthy weight and controlling hyperglycemia.

Left ventricular diastolic dysfunction (DD) is a combination of 2 defects: (1) impaired myocardial relaxation ability and (2) reduced filling of the left ventricle in the absence of increased filling pressures. Early‐stage DD (grade 1 DD [G1DD]) is characterized by impaired relaxation, and even among young adults, early disease is associated with incident heart failure with preserved ejection fraction. , , Important modifiable risk factors for G1DD include obesity, diabetes, hypertension, and coronary heart disease. , , , , , , , , , Heart failure with preserved ejection fraction lacks life‐extending treatments, and delineating genetic drivers of diastolic dysfunction risk could help identify modifiable predisposing risk mechanisms to improve outcomes for patients. , , , , , , ,

The genetics of diastolic function are not well characterized. The ECHOGEN consortium examined diastolic function parameters in a large population and did not observe significant single nucleotide polymorphism (SNP) heritability estimates or SNPs associated with these parameters. However, structural traits related to diastolic function, such as left ventricular mass and other functional left ventricular measures, have been demonstrated to have a significant heritable component based on common SNPs. , This could suggest that diastolic dysfunction is a genetically heterogeneous phenotype representing the accumulated contributions from many risk mechanisms.

We hypothesized that examining early diastolic dysfunction in a clinical population would identify genetic risk mechanisms associated with this potentially reversible stage of cardiac remodeling. To address this hypothesis, we developed a large, genotyped population of individuals aged <60 years who had undergone transthoracic echocardiography studies as part of routine clinical care. We identified a genetic predisposition to elevated body mass index (BMI) and glycemic dysregulation associated with diastolic dysfunction. Addressing these risk mechanisms, especially obesity, may prevent or reverse the pathological cardiac changes associated with diastolic dysfunction.

Methods

An overview of the analytic approach to identify the underlying genetic risk mechanisms between genetically regulated comorbid traits and G1DD is presented in Figure 1. The genome‐wide association study (GWAS) summary statistics generated in these analyses are available from the corresponding author upon request.

Figure 1

Overview of the study design.

Grade 1 diastolic dysfunction cases and controls were selected from BioVU, Vanderbilt University Medical Center’s electronic health record (HER)–linked DNA biobank. A genome‐wide association study (GWAS) was performed on 2440 individuals. A polygenic risk score screen identified traits that shared genetic risk with grade 1 diastolic dysfunction. Associated traits were further investigated under a 2‐sample Mendelian randomization framework. Secondary analyses investigated for potential pleiotropic mediators.

Study Population

BioVU is Vanderbilt University Medical Center’s DNA biobank linked to a deidentified mirror of the electronic medical records of the Vanderbilt health system. Individuals seeking health care at Vanderbilt University Medical Center are consented to participate and may opt out at any time. Sample collection began in 2007 and is ongoing. The biobank comprises ≈260 000 individuals of European, African, and Asian ancestries. Access to the data in the biobank is overseen by the Vanderbilt Institutional Review Board. All participants provided written consent.

Individual‐level genotype data were obtained from BioVU. Approval for the present study was obtained from the Vanderbilt University Medical Center Institutional Review Board.

G1DD cases and controls were identified from transthoracic echocardiogram reports generated during routine clinical care, as previously described. Briefly, diastolic function stage was extracted from the first available transthoracic echocardiogram report and was assigned by a clinical echocardiographer at the time of collection of the transthoracic echocardiogram. Cases were individuals classified as impaired left ventricle relaxation (stage 1), whereas controls were individuals classified as having no dysfunction. Individuals with higher DD stages were excluded, as were individuals with a left ventricular ejection fraction <50%. Subjects who had echocardiograms collected during acute cardiac illnesses (eg, endocarditis), with exposure to chemotherapy, or with congenital diseases were also excluded. The analyses were restricted to individuals who had been previously genotyped on the Multi‐Ethnic Genotyping Array platform (described below) as part of a broad‐based institutional genotyping initiative. The analyses were further restricted to subjects aged <60 years and of European ancestry, as determined using HAPMAP reference populations in conjunction with genetic principal components.

Genetic Data

SNP genotyping of BioVU subjects was measured using the Illumina Infinium Multi‐Ethnic Genotyping Array platform. Quality control analyses used PLINK version 1.90β3 software. Before imputation, genetic data were filtered and standardized through the HRC‐1000G‐check tool version 4.2.5 (http://www.well.ox.ac.uk/~wrayner/tools/) and prephased using Eagle version 2.4.1. Principal components were calculated using the SNPRelate package. Data were imputed using the Michigan Imputation Server in conjunction with the 10/2014 release of the 1000 Genomes cosmopolitan reference haplotypes. Imputed data were filtered for a sample missingness rate <2%, a SNP missingness rate <4%, and SNP deviation from Hardy‐Weinberg P<10−06. After quality control, 7 585 258 SNPs were available for analysis. Genome‐wide study and polygenic risk scores were calculated using PLINK version 2.

Clinical Phenotypes

Age was defined as the age at the time of cardiac echocardiogram. Clinical diagnoses for the G1DD risk factors of obesity (278.10), type 2 diabetes (T2D) (250.20), ischemic heart disease (401.00), and hypertension (411.00) were defined using PheCodes, which are derived from International Classification of Diseases, Ninth and Tenth Revision (ICD‐9 and ICD‐10) codes.

Statistical Analysis

GWAS Summary Statistics

Summary statistics for BMI, systolic blood pressure, diastolic blood pressure, T2D, coronary artery disease, fasting glucose (FG), hemoglobin A1C (HbA1C), high‐density lipoprotein cholesterol, low‐density lipoprotein cholesterol, and triglycerides obtained from the publicly available large‐scale GWAS performed among individuals of European ancestry.

Baseline Characteristics

Mean age and the prevalence of comorbidities were calculated and stratified by G1DD case status. Differences in mean age and number of each sex, by G1DD status, was computed as a test of proportions between cases and controls using a Pearson χ2 test statistic. Significant differences in prevalence rates of comorbidities, by G1DD status, were assessed using logistic regression, adjusting for age and sex.

Genome‐Wide Association Study

The GWAS was used to identify SNPs associated with G1DD case‐control status (Data S1). The analyses used a logistic regression, assuming an additive genetic model, and was adjusted for age, sex and 5 principal components of ancestry using PLINK version 2. SNPs with an association of P<5×10−08 were deemed to be significant. We performed a power calculation based on our sample size and distribution of cases and controls (https://zzz.bwh.harvard.edu/gpc/cc2.html). We had >80% power to detect an association at genome‐wide significance for a SNP with an odds ratio (OR) >1.6, assuming a disease prevalence of 5% and a minor allele frequency >20%. Annotations for SNPs with an association of P<5×10−06 of GWAS results are presented in Data S2.

Polygenic Risk Score

To identify traits that share genetic risk with G1DD, polygenic risk scores (PRSs) for each trait were computed and then tested for an association with DD case‐control status, using PLINK version 2. An independent set of SNPs significantly associated with the respective trait (P<5×10−08) was selected using a pruning‐and‐thresholding algorithm that selected an Linkage Disequilibrium‐reduced (r 2<0.05) set of common SNPs with a minor allele frequency >5%. A genetically predicted trait score was then calculated for each individual in the G1DD cohort by summing the product of each SNP effect size and the SNP dosage (a value ranging from 0 to 2). The association with the polygenic risk score and DD status was tested using a logistic regression model that adjusted for sex, age, and 5 principal components as covariates. A Bonferroni‐adjusted P<0.01 (=0.05/5 phenotypes) was considered significant.

Mendelian Randomization

Mendelian randomization (MR) was used to further probe traits significantly associated with G1DD by PRS analysis. MR is an instrumental variable approach used to define causal relationships between exposures and outcomes. It uses SNPs associated with a chosen exposure as instrumental variables that define the direction and magnitude of associations between the exposure/risk factor and the chosen outcome, G1DD. To create genetic instruments for each risk factor, we used a pruning‐and‐thresholding algorithm that selected an Linkage Disequilibrium‐reduced set of SNPs with a minor allele frequency >5%. All MRs conducted were 2‐sample average inverse‐variance weighted method (IVWM) analyses, using the Mendelian Randomization R package. An association was considered significant for a P<0.05. Association measures represent the change in risk factor level versus the log odds of G1DD status. To ensure that significant associations were not caused by pleiotropy, sensitivity analyses were conducted using the pleiotropy‐robust MR‐Egger and weighted median methods to confirm the magnitude and direction of associations. IVWM results were considered reliable if they had a similar direction and magnitude of association as the other 2 methods.

In secondary analyses, we sought to identify candidate mediators of an observed genetic association between G1DD and BMI. MR association analyses were performed for FG, HbA1C, high‐density lipoprotein cholesterol, low‐density lipoprotein cholesterol, and triglycerides using the same methods as described above.

Multivariable Mendelian Randomization

Multivariable MR (MVMR) was used to determine whether an observed association between G1DD and BMI was independent of other modifiable risk factors. MVMR estimates the effects of multiple exposures on an outcome. MVMR analysis was performed that included BMI and each candidate mediator that significantly associated with G1DD by MR in the analyses described above. We identified those risk factors that decreased the primary mediator’s coefficient by >1.96 standard errors (P<0.05), as compared with the original coefficient after adjustment, and in the process, estimated the magnitude of the direct effect of the exposure and the indirect effect ascribed to the mediator.

Phenotypic Association of Genetic Risk Score

We examined if the PRS for BMI also associated with prevalent T2D, ischemic heart disease, and hypertension within this cohort. We ran a logistic regression for the outcome status in a model of the BMI PRS adjusted for age, sex, top 5 principal components, and G1DD status. We established a multiple comparisons threshold of <0.05/3 for significance.

Sensitivity Analyses to Examine Treatment Effects

Therapeutic interventions could lead to misclassification of case‐control status. Specifically, treatment of diabetes and hypertension by therapeutic agents could have prevented or reversed G1DD, resulting in misclassification of controls. To explore this possibility, we conducted sensitivity analyses by removing either controls treated for hypertension or for diabetes before the date of the echocardiogram. All primary analyses were repeated using these new control groups.

Results

The final study population comprised 2440 individuals, with 668 G1DD cases and 1772 controls (Table). There were 1457 (≈60%) women, and the mean age was 47.2 (SD, 10) years. Cases had significantly higher prevalence of diagnoses for ischemic heart disease (OR, 1.40 [95% CI, 1.15–1.71]; P=0.001), obesity (OR, 1.57 [95% CI, 1.28–1.92]; P=1.5×10−05), hypertension (OR, 2.14 [95% CI, 1.72–2.66]; P=2.61×10−11), and diabetes (OR, 1.88 [95% CI, 1.51–2.34]; P=2.4×10−09).

Table  

Demographic Profile of the Study Population

Characteristic*	All participants, n=2440	Cases, n=668	Controls, n=1772	P value
Age, y, mean (SD)	47.2 (10.0)	52.9 (5.9)	45.1 (10.3)	<2.2×10−16 †
Women	1457 (59.7%)	365 (54.6%)	1092 (61.6%)	0.002 †
Obesity	716 (29.3%)	248 (37.1%)	468 (26.4%)	1.5×1013 ‡
Ischemic heart disease	728 (29.8%)	286 (42.8%)	442 (24.9%)	0.001 ‡
Hypertension	1472 (60.3%)	527 (78.9%)	945 (53.3%)	2.6×10−11 ‡
Diabetes	654 (26.8%)	263 (39.4%)	391 (22.1%)	2.4×10−9 ‡

Values in the table represent counts and column percentages, except for age.

P value for the difference in proportions between cases and controls is based on the value of Pearson χ2 test statistic.

Association P value for the risk factor from a logistic regression model adjusting for age and sex.

There were no common SNPs associated with G1DD case status at genome‐wide significance by the GWAS (Figure S1A). Though there were no genome‐wide significant SNPs, annotations from SNPs with an association P<5×10−06 indicated that these SNPs were associated with cardiometabolic phenotypes (Figure S1B, Tables S1A and S1B), and with the left ventricle in HiC chromatin interaction. We tested for associations between G1DD status and PRS for systolic blood pressure, diastolic blood pressure, BMI, T2D, and coronary artery disease to determine whether genetic variation underlying any of these risk factors also associated with G1DD prevalence. Only BMI was significantly associated (OR, 1.20 for 1‐SD increase in BMI [95% CI, 1.08–1.32]; P=0.0004) (Figure 2, Table S1C). Given that BMI demonstrated shared genetics with G1DD, we asked if the relationship held under an instrumental variable framework. Genetically predicted BMI was significantly associated with G1DD by IVWM (OR, 1.84 for 1‐SD increase in BMI [95% CI, 1.35–2.52]; P=0.0002). Similar results were seen across other MR methods (Table S2, Figure S2).

Figure 2

Genetic determinants of body mass index (BMI) share a genetic risk with grade 1 diastolic dysfunction (G1DD).

Forest plot summarizing associations between polygenic risk score (PRS) for diastolic blood pressure (DBP), systolic blood pressure (SBP), ischemic heart disease (IHD)/coronary artery disease (CAD) , type 2 diabetes (T2D), and BMI and G1DD. Odds ratio (OR) represents the change in risk for G1DD per standard deviation increase in the PRS.

To determine whether the BMI association may be mediated through lipids (low‐density lipoprotein cholesterol, high‐density lipoprotein cholesterol, triglycerides) or glycemic factors (FG, HbA1C), we first ascertained whether genetic instruments for these factors were associated with G1DD by MR analyses. There was only a significant association with FG (OR, 4.14 for 1‐SD increase in FG [95% CI, 1.55–11.02]; P=0.005) (Figure 3, Table S3), but not HbA1C (OR, 2.23 for 1‐SD increase in FG [95% CI, 0.59–8.44]; P=0.24). A BMI predictor was associated with FG by IVWM, but an FG predictor did not associate with BMI, suggesting that FG may be a downstream effector of BMI (Table S4, Table S5).

Figure 3

Genetic determinants of fasting glucose are associated with grade 1 diastolic dysfunction (G1DD) in 2‐sample Mendelian randomization analysis.

Forest plot of inverse variance instrumental variable estimates for glucose, hemoglobin A1C (HbA1C), high‐density lipoprotein (HDL) cholesterol, low‐density lipoprotein (LDL) cholesterol, and triglycerides levels and G1DD status. Odds ratio (OR) represents the change in risk for G1DD per standard deviation increase in the respective mediator.

MVMR was used to ascertain independent associations between BMI and FG using the IVWM. After adjustment, BMI was still significantly associated with G1DD, but not FG. For BMI, there was a modest attenuation of the OR from 1.89 (95% CI, 1.37–2.61) to 1.84 (95% CI, 1.35–2.52) (Table S6, Table S7, Figure S3). This suggests that 5.5% for the BMI effect is attributable to FG (Table S8). In sum, these data suggest that genetic factors that predispose to elevated BMI may impact the development of G1DD directly and may mediate a modest fraction of their effects by modulating glucose homeostasis (Figure 4).

Figure 4

Genetic determinants of body mass index (BMI) associated with grade 1 diastolic dysfunction (G1DD).

Higher prevalence of comorbid phenotypic associations with G1DD are observed in the electronic health record (EHR). The underlying genetic basis is driven, in part, by a genetic predisposition to elevated BMI directly and mediated through fasting glucose. SNPs indicate single nucleotide polymorphisms.

Finally, we tested the association between PRS for BMI and G1DD‐associated comorbidities. Among the G1DD‐associated comorbidities, only T2D was significantly associated with a BMI PRS (OR, 1.23 for 1‐SD change in BMI PRS [95% CI, 1.12–1.36]; P=1.4×10−05) (Table S9, Figure S4).

The sensitivity analyses were conducted after removing controls receiving either hypertension medications or diabetes medications before the date of the echocardiogram. The exclusion of these individuals did not alter the overall findings for either the controls receiving antihypertensive medications (Figures S5 and S6, Tables S10 through S17) or glucose‐lowering medications (Figures S7 and S8, Tables S18 and S25). Similar results were seen across the other MR methods (Tables S5, S12, S13, S20 and S21, Figures S6 and S8). In addition, no new associations were observed in the post‐GWAS or MR analyses with these exclusions.

The precision and variance estimate from MR is strongly influenced by the number of SNPs used as genetic instruments. The larger variance estimates associated with the instrumental variables for glycemic traits (HbA1C, FG) reflect the smaller number of SNPs significantly associated with these variables, as compared with the other instruments. Exclusion of the treated controls did not substantially change these estimates. For the cohort where controls receiving hypertension medication were removed from the analyses (Table S13), there was only a significant association with FG (OR, 9.70 for 1‐SD increase in FG [95% CI, 2.71– 34.71]; P=0.0005), but not HbA1C (OR, 4.01 for 1‐SD increase in FG [95% CI, 0.72–22.39]; P=0.11). For the cohort where controls receiving diabetes medication were removed from the analyses (Table S21), there was only a significant association with FG (OR, 5.76 for 1‐SD increase in FG [95% CI, 2.03–16.33]; P=9.93×10−04), but not HbA1C (OR, 2.48 for 1‐SD increase in FG [95% CI, 0.61–10.14]; P=0.21). The distributions of PRS for BMI and FG by G1DD case‐control status and concordance statistics for models with and without these PRSs are presented in Figure S9 and Table S26.

Discussion

This study examined genetic determinants of early diastolic dysfunction in a clinical population. A GWAS did not identify common SNPs significantly associated with G1DD status, consistent with prior studies examining diastolic function phenotypes. , However, several of the top SNPs had appeared to have relevance to cardiometabolic phenotypes and the left ventricle. Thus, we examined associations with genetic predictors of established risk factors, which is a more powered approach to detect weak associations. Although diagnoses of obesity, ischemic heart disease, hypertension, and T2D were more prevalent among cases, only a genetic predictor for BMI positively associated with G1DD risk. MVMR models suggested that a small portion of this risk may be mediated by glycemic dysregulation, as measured by FG levels. In sum, these results suggest that a genetic predisposition to elevated BMI contributes to G1DD risk, acting directly and through a mediated fraction impacting glycemic regulation.

Obesity‐driven remodeling and expansion of adipose tissue, in addition to systemic inflammation from obesity, is increasingly recognized as a key driver of dysfunction in the left ventricle. Excess body mass is associated with several adverse physiologic alterations and cardiac structure changes that adversely affect the heart. These include concentric left ventricular remodeling, right ventricular dilatation, increased epicardial fat thickness, and elevated left ventricular filling pressures. The adverse effect of obesity on cardiac diastolic function are recapitulated in genetic animal models. For instance, leptin‐deficient (ob/ob) or leptin receptor‐deficient (db/db) mice, both isolated models of obesity, develop cardiac diastolic dysfunction.

In these analyses, we observed that a polygenic predisposition toward elevated BMI was associated with an increased risk for early clinical diastolic dysfunction. A polygenic association between an exposure and a phenotype could be caused by either shared genetic mechanisms that affect both phenotypes or a direct mediating effect of the exposure on the outcome. MR analyses confirmed the BMI‐G1DD association and did not demonstrate heterogeneity in the association among the underlying SNP instruments. These results suggest that BMI is a mediator of disease, and all underlying genetic mechanisms that increase BMI also drive disease risk. The implications of these findings are that avoiding or possibly reversing obesity through any mechanism could decrease the risk of early diastolic dysfunction. Furthermore, because G1DD is often reversible with attenuation of risk factors, weight loss could improve function.

The prevalence of established risk factors in addition to obesity was higher among the G1DD cases than controls in this study population. However, we did not observe significant associations between genetic predictors of coronary heart disease, hypertension, and T2D. Hypertension and insulin resistance are well‐established downstream risk factors of obesity. Thus, it is possible that the higher prevalence of these risk factors among G1DD cases is because of the other secondary factors of metabolic syndrome.

Although we did not see an association with a genetic predictor of T2D, there was a positive association with a predictor of FG and G1DD. Of note, the genetic architectures of these phenotypes differ. , In multivariable analyses, the association with FG was no longer significant, and there was a modest attenuation of the association statistics associated with BMI, suggesting altered FG levels may be secondary to the effects of elevated BMI, and may mediate some of the risk associated with BMI. Elevated FG levels have been observed to be associated with diastolic dysfunction, though the associations have not been consistent. , , , Higher fasting plasma glucose levels among individuals without diabetes was also found to be an independent risk factor for heart failure hospitalization. Furthermore, SGLT2 (Sodium‐glucose Cotransporter‐2) inhibitor use has been shown to reverse diastolic dysfunction among individuals with diabetes. Perhaps these findings align the development of G1DD as an antecedent to the current paradigm of the diabesity phenotype, the combined burden of obesity and diabetes on heart disease.

The current study has limitations. The outcome studied was a binary outcome based on an echocardiographer’s clinical assessment, which can result in loss of power caused by binning and misclassification. There was limited power to detect SNP associations with a magnitude of effect usually observed for common SNPs (ie, an OR of <1.3) associated with a complex phenotype, which can lead to false‐negative findings by the GWAS. The study was done in a European ancestry population; this limits insights into other ancestries. The study cohort was curated from electronic health records in a health system, and controls are not necessarily healthy. The modest number of SNPs available for use in constructing the genetic instruments for the glycemic predictors likely contributed to the low precision in effect size estimates associated with these instruments.

Future studies to dissect out the impact of genetic determinants on the development of G1DD could include looking at more continuous measures of diastolic function, taking multitrait GWAS/polygenic risk approaches, and recruiting more subjects representing non‐European ancestries. The genetic underpinnings of later stages of diastolic dysfunction should be studied to identify genetic drivers of risk of late‐stage cardiac diastolic remodeling.

In conclusion, among multiple risk factors epidemiologically associated with diastolic dysfunction, only a polygenic predictor of BMI was associated with G1DD, suggesting a predisposition to elevated BMI could be an important driver of risk and may also underlie the development of other risk factors, such as impaired glucose homeostasis. Obesity is driven by gene‐x‐environment interactions, and thus a genetic predisposition is not deterministic of an individual’s fate. Treatment and prevention strategies that reduce BMI are apt to mitigate an important genetic driver of early diastolic dysfunction. Based on these results, there might be a role for preventive echocardiograms to detect early G1DD and mitigate downstream complications, including heart failure, among subjects seen to be increasing BMI, of high BMI, or showing altered glycemic homeostasis.

Sources of Funding

American Heart Association 16FTF30130005 (J.D.M.) and National Institutes of Health/National Heart, Lung, and Blood Institute R01‐HL142856 (J.D.M.). Vanderbilt University Medical Center’s BioVU is supported by institutional funding, private agencies, and federal grants. These include the National Institutes of Health–funded Shared Instrumentation Grant S10RR025141; and Clinical and Translational Science Awards grants UL1TR002243, UL1TR000445, and UL1RR024975. Genomic data are also supported by investigator‐led projects that include U01HG004798, R01NS032830, RC2GM092618, P50GM115305, U01HG006378, U19HL065962, R01HD074711; and additional funding sources listed at https://victr.vumc.org/biovu‐funding/.

Disclosures

None.

Data S1–S2

Tables S1–S26

Figures S1–S9

References 48, 49, 50, 51, 52

Click here for additional data file.