Search

Introduction: Associations of nonalcoholic fatty liver disease (NAFLD) with dementia are controversial. The full spectrum of NAFLD from simple steatosis to fibrosis has been less investigated. Hypothesis: The severity of NAFLD, especially with the stage of liver fibrosis, is associated with dementia. Methods: The associations between midlife (1990-92, N=9283, mean age 57, female 55%) and late-life (2011-13, N=5138, mean age 75, female 58%) NAFLD and all cause dementia were quantified by Cox regression. NAFLD was characterized using the fatty liver index and liver fibrosis assessment model (FIB-4). Dementia was adjudicated and ascertained through Dec. 31, 2019 from neuropsychological assessments, annual participant or informant contact, and medical record surveillance. We used Cox models with adjustment for demographics, APOE ε4, alcohol, and kidney function. Additional adjustments were made for metabolic factors to explore the independent contribution of NAFLD to dementia. Results: During a median follow-up of 24.5 years after midlife NAFLD assessment, 1854 participants developed dementia. During a median of 6.3 years in late-life, there were 893 dementia cases. A graded increase in dementia risk across the spectrum of NAFLD was observed at midlife (p-trend Figure ). In fact, NAFLD in late-life tended to be protective against dementia. After adjusting for the metabolic factors, the associations remained statistically significant in some categories. Conclusions: Remodeling of hepatic architecture and dysregulation in hepatocellular function in NAFLD at midlife plausibly induce peripheral promoters of neurodegeneration and may expose a large segment of the population to increased risk of dementia. The inverse association of NAFLD with dementia at late-life requires further investigation, possibly reflecting the depletion of susceptibles, the reversibility of NAFLD due to weight loss, and the lack of accuracy in identifying NAFLD using clinical prediction models at older age.

Introduction: Analyzing plasma proteins is a novel approach to uncover molecular mechanisms by which psychosocial stress may lead to chronic illness, including cardiometabolic disease. Proteomic analyses of cardiometabolic disease have been reported; however, proteomic studies of stress and depressive symptoms are limited. Hypothesis: We hypothesized that plasma protein concentrations are both positively and negatively associated with greater depressive symptoms and perceived global stress, respectively, in African American adults from the Jackson Heart Study (JHS). Methods: Participants from exam 1 (2000-2004) with SomaScan 1.3k platform proteomics data were utilized (n=2143, mean age=55.3). Depressive symptoms and global stress scores were measured via the 20-item Centers for Epidemiological Studies (CES-D) Scale and an 8-item global perceived stress scale adapted for the JHS, respectively. Multivariable linear models were used to test the association between depressive symptoms and global stress scores and proteins, controlling for age, sex, and proteomics batch. Bonferroni correction was used to adjust the p-value threshold for multiple testing (3.78e-5, for 1,322 tested proteins). Results: Six (6) proteins were found to be significantly associated with either depressive symptoms or perceived global stress. Angiopoietin-2 (β=0.014, SE=0.003, p =1.33e-5), growth/differentiation factor 15 or MIC-1 (β=0.014, SE=0.003, p=7.08e-8), and kynureninase (β=0.014, SE=0.003, p =1.55e-5) were positively associated with depressive symptoms. Contactin-5 (β=-0.017, SE=0.003, p =1.40e-7) and neural cell adhesion molecule 1 (120 kDa isoform) (β=-0.016, SE=0.003, p =6.74e-7) were negatively associated with depressive symptoms. Leukotriene A-4 hydrolase or LKHA4 (β=0.024, SE=0.005, p =2.72e-6) was positively associated with perceived stress. We note that several of these proteins which are associated with higher levels of stress or depression symptoms have also been associated with cardiovascular disease and related phenotypes; for example, LKHA4 was the lead association with left ventricular mass in previous JHS analyses. Conclusions: JHS proteomic profiles revealed differences in protein concentrations by psychosocial measures. We will next explore the extent to which identified proteins may mediate relationships between stress and depressive symptoms and cardiovascular endpoints. Future investigations may identify potentially targetable proteomic mechanisms by which psychosocial stressors manifest into disease.

Introduction: Clonal hematopoiesis of indeterminate potential (CHIP) is an aging-related accumulation of somatic mutations in hematopoietic stem cells, leading to clonal expansion. CHIP presence has been implicated in elevated risks for coronary heart disease (CHD) and death, but its association with incident type 2 diabetes (T2D) is unknown. Hypothesis: We hypothesized that CHIP is associated with elevated risk of incident T2D. Methods: CHIP was derived from whole genome sequencing of blood DNA in NHLBI Trans-omics for Precision Medicine (TOPMed) cohorts. We analyzed 17,637 participants without prior T2D, cardiovascular disease, or cancer at blood draw, with prospective follow-up for incident T2D. We evaluated baseline prevalence of CHIP vs. no CHIP with incident T2D risk using Cox regression. We also investigated CHIP variants previously related to CHD: DNMT3A , TET2 , ASXL1 , JAK2 , and TP53 . We estimated multivariable-adjusted hazard ratios and 95% confidence intervals (HR [CI]) adjusted for age, sex, body mass index, smoking, alcohol, and education. We combined cohort estimates via fixed effects meta-analysis. Results: On average, participants were age 63.4 years (SD=11.5) and 76% female. Prevalence of CHIP was 6.0% (1,055) at baseline. There were 2,467 incident T2D cases over mean=9.8 years follow-up. Compared to those without a mutation, having CHIP was associated with a 23% higher T2D risk, both overall (combined HR=1.23; 95% CI=1.04, 1.45), and among those with CHD-related CHIP mutations (87% of total CHIP): HR=1.23 (1.03, 1.46). Although those with CHIP mutations of TET2 (HR=1.48; 1.05, 2.08) and ASXL1 (HR=1.76; 1.03, 2.99) had larger elevations in T2D risk, and DNMT3A was suggestive of increased T2D risk (HR=1.15; 0.93, 1.43), statistical power was limited for JAK2 and TP53 mutation analyses. Conclusions: CHIP was associated with higher incidence of T2D. CHIP mutations located on loci previously implicated in aging and CHD were also related to T2D, suggesting shared pathology of atherosclerosis and T2D.

Introduction: The aromatic amino acids tyrosine and phenylalanine may increase type 2 diabetes risk, but few studies have examined whether these critical molecular precursors exert additional phenotypic effects and whether these effects are independent of insulin resistance. We evaluated associations between two aromatic amino acids - tyrosine and phenylalanine - with a broad range of phenotypic categories in the UK Biobank using polygenic risk score (PRS) instrumental variables that are robust to confounding and reverse causation and tested for mediation by glycated hemoglobin (HbA1c) as a measure of insulin resistance. Hypothesis: We hypothesized that tyrosine and phenylalanine would show broad phenotypic effects and that many of these effects would be mediated by HbA1c. Methods: We constructed PRS (Crosspred) using all nominally significant and common (minor allele frequency >5%) variants from genome-wide association studies (SAIGE) of unrelated European ancestry participants in the UK Biobank. We evaluated associations with 273 curated phenotypes spanning 20 categories, including endocrine, circulatory, and cancer related traits. We corrected for multiple comparisons using false discovery rate Results: A total of 108,554 participants had tyrosine or phenylalanine measured at baseline (mean age=57 years; 54% (58,880/108,554) female). PRS were predictive of tyrosine (R 2 =0.28) and phenylalanine (R 2 =0.26). Tyrosine and phenylalanine PRS showed significant associations with 121/273 (44%) and 124/273 (45%) of phenotypes, respectively, including chronic kidney disease, serum testosterone, lymphocyte count, and myocardial infarction (MI; n=3,349 cases, mean years of follow-up=11). For incident MI, every one standard deviation increase in the tyrosine PRS increased the odds of MI by 4.4% (total effect odds ratio (OR) = 1.044 (95% confidence interval (CI): 1.002, 1.089). The direct effect of the tyrosine PRS on incident MI was slightly decreased (OR= 1.040 (95% CI: 0.997, 1.084) when extending the statistical model to examine mediation by HbA1c, with an estimated indirect effect of 1.004 (95% CI: 1.003, 1.006) and an estimated percent mediated by HbA1c of 9.95% (95% CI: -0.04%, 19.90%). Findings were consistent for phenylalanine, where the percent of the total effect mediated by HbA1c was 8.94% (95% CI: -1.93, 19.80%). Conclusions: The effect of tyrosine and phenylalanine on incident MI may primarily operate through pathways other than glucose dysregulation.

Introduction: Although thiazide diuretics are common therapies for the treatment of hypertension, evidence suggests that they are also associated with elevated blood lipid concentrations after their initiation. Pharmacogenomics studies could help identify potential biological pathways underlying this phenomenon, but few well-powered studies have been conducted. Methods: In participants of European ancestry from the UK Biobank, we conducted a genome-wide association study (GWAS) to examine whether common variants (minor allele frequency [MAF] >1%) modified the effect of thiazide diuretic use on three lipid measures: low density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), and triglyceride (TG) concentrations (mmol/L). TG concentrations were natural log transformed, and LDL-C concentrations were adjusted for statin use. Given the conflicting evidence about the impact of loop diuretics on blood lipid concentrations, unexposed participants were defined as those taking neither thiazides nor loop diuretics. GWAS of UK10K plus 1000-Genomes Phase 3 imputed data were performed using SAIGE, adjusting for age, sex, fasting status, body mass index, study center, and the first 10 ancestral principal components. Results: Approximately 7.1% (n=27,971) of participants (n=395,058; mean age=57 years; 54% female [n=214,334]) used thiazides at study baseline. The thiazide-LDL-C GWAS (n=9,680,686 variants) identified one chromosome 19 locus harboring several genome-wide significant variants that modified the effect of thiazide diuretic use on LDL-C. The lead variant rs2199576 (MAF = 0.18) had an effect estimate (standard error) of 0.159 (0.013) among participants exposed to thiazides, and an effect estimate of 0.083 (0.003) among participants unexposed to any diuretic, yielding a significant variant-by-thiazide interaction effect estimate of 0.076 (0.013) ( p = 1.62 x 10 -9 ). These effect estimates are comparable to previously identified main effects of blood lipids loci. The rs2199576 variant resides near the PVRL2 locus, a cholesterol-responsive gene that is located within 100 kb of TOMM40 and the APO cluster, which have both been associated with blood lipid concentrations and Alzheimer’s disease. Conclusion: Our results may provide insights into pathways that influence the adverse effects of thiazide diuretic use and highlight the large sample sizes needed to detect modest pharmacogenetic effects.

Introduction: Despite decades of research linking obesity with cardiovascular disease (CVD) and its risk factors, major research gaps remain, with few studies evaluating the often-described but poorly understood heterogeneity in CVD risk observed within obese populations. One plausible, but largely unexplored source of heterogeneity in obesity-associated CVD risk is the impact of body fat distribution (central versus overall obesity), obesity duration and severity. Longitudinally assessed anthropometric measures and GWAS derived from these exposures may improve understanding of functional properties of loci associated with obesity and downstream influences on CVD. Methods: We respond to these research gaps by using 25 years of Atherosclerosis Risk in Communities (N=14,514, mean baseline age=54 years; 55% Female; 25.73% African-American) data and latent class mixed models to derive four longitudinal measures of obesity distribution, duration, and severity, operationalized as obesity subtypes. Genome wide association studies on the subclasses were conducted using logistic multi-variate regression models to test for associations between variants and BMI. Meta-analyses across race and sex were conducted in METAL. Functional properties of single nucleotide polymorphisms (SNP) were assessed in FUMA. Results: The four obesity subclasses were: decline (4.1%), stable/slow decline (67.8%) [referent], moderate increase (24.6%), and rapid increase (3.6%). The lead SNP from the decline group was known, on chromosome 7 (rs1196508, beta (SE) = -0.44 (0.089), p = 6.84x10 -7 , minor allele frequency (MAF) = 0.36), and nearest to a tyrosine phosphatase receptor, PTPRZ1 , in the central nervous system. We combined moderate and rapid increase for analyses; they shared a common top, unknown, SNP on chromosome 5 (rs76956550, beta (SE) = -0.37 (0.074), MAF=0.12), nearest to PGAM5P1 . The top SNP within the rapid increase group was unknown, on chromosome 14 (rs580888, beta(SE) = -0.57 (0.11), MAF=0.35), and nearest to the OTX2 gene, a transcription factor involved in production of dopaminergic neurons. Mutations of this gene are associated with pituitary hormone deficiency. Conclusion: Our GWAS of obesity subclasses demonstrated unique genetic and functional properties between groups, supporting the contention that CVD risk may differ across obese individuals. We intend to replicate our novel association of OTX2 with BMI, waist-to-hip ratio, and waist circumference in people who rapidly gained weight in other large longitudinal cohorts.

Introduction: Experimental and observational evidence suggests that aberrant branched chain amino acid (BCAA; leucine, isoleucine, and valine) metabolism promotes insulin resistance and downstream disease, prompting interest in BCAA catabolism as a treatment target. Most published research has evaluated the three BCAAs in combination, although each BCAA may harbor distinct phenotypic effects, as valine is glucogenic, leucine is ketogenic, and isoleucine is glucogenic and ketogenic. Methods: Because studies examining individual phenotypic effects of BCAAs that account for the effects of other BCAAs have been limited by high correlation between BCAAs and strong confounding by dietary patterns and shared genetic loci, we applied a novel causal inference method developed to address these limitations: multivariable Mendelian randomization (MVMR). In a study of UK Biobank European ancestry participants, we examined 1,468 phenotypes that included metabolic, cardiovascular, cancer, cognitive, hematopoietic, and musculoskeletal domains. Polygenic risk score (PRS) instrumental variables were constructed using genotypic and circulating BCAA phenotypic data in European ancestry UK Biobank participants and the Crosspred method was used to correct for overfitting. Significant independent causal effects were defined as estimates with P- values< 1.1x10 -5 [0.05/(1,468 phenotypes*3 BCAAs)]. Results: The n=97,469 participants with measured BCAAs were on average middle aged (mean age=57), female (54%), and overweight [mean body mass index (BMI)=27]. Each BCAA PRS strongly predicted its circulating concentrations ( P- value-100 ). MVMR methods did not identify any phenotypes for which all three BCAAs showed evidence of a directionally consistent and significant independent causal effect. However, there were numerous instances where BCAAs harbored independent and directionally inconsistent statistically significant effects. For example, significant positive causal effects for valine independent of isoleucine and leucine and significant negative causal effects for leucine independent of valine and isoleucine were identified for BMI, glycated hemoglobin, and type 2 diabetes; no significant independent causal effects for isoleucine with BMI, glycated hemoglobin, and type 2 diabetes were identified. In contrast, models without control for the effects of other BCAAs suggested that isoleucine, leucine and valine each had significant positive causal effects on BMI, glycated hemoglobin, and type 2 diabetes. When examining other phenotypes using MVMR, heterogeneity in the effects of individual BCAAs was detected for cardiovascular and cognitive phenotypes, among others. Conclusion: BCAAs may harbor distinct and opposing effects on metabolism, health, and disease, motivating further studies that examine the independent effects of these essential amino acids.

Background: Heart failure (HF) risk and the underlying risk factors vary by race. Traditional models for HF risk prediction treat race as a covariate in risk prediction and do not account for significant parameters such as cardiac biomarkers. Machine learning (ML) may offer advantages over traditional modeling techniques to develop race-specific HF risk prediction models and to elucidate important contributors of HF development across races. Methods: We performed a retrospective analysis of 4 large, community cohort studies (ARIC [Atherosclerosis Risk in Communities], DHS [Dallas Heart Study], JHS [Jackson Heart Study], and MESA [Multi-Ethnic Study of Atherosclerosis]) with adjudicated HF events. The study included participants who were >40 years of age and free of HF at baseline. Race-specific ML models for HF risk prediction were developed in the JHS cohort (for Black race–specific model) and White adults from ARIC (for White race–specific model). The models included 39 candidate variables across demographic, anthropometric, medical history, laboratory, and electrocardiographic domains. The ML models were externally validated and compared with prior established traditional and non–race-specific ML models in race-specific subgroups of the pooled MESA/DHS cohort and Black participants of ARIC. The Harrell C-index and Greenwood-Nam-D’Agostino χ 2 tests were used to assess discrimination and calibration, respectively. Results: The ML models had excellent discrimination in the derivation cohorts for Black (n=4141 in JHS, C-index=0.88) and White (n=7858 in ARIC, C-index=0.89) participants. In the external validation cohorts, the race-specific ML model demonstrated adequate calibration and superior discrimination (Black individuals, C-index=0.80–0.83; White individuals, C-index=0.82) compared with established HF risk models or with non–race-specific ML models derived with race included as a covariate. Among the risk factors, natriuretic peptide levels were the most important predictor of HF risk across both races, followed by troponin levels in Black and ECG-based Cornell voltage in White individuals. Other key predictors of HF risk among Black individuals were glycemic parameters and socioeconomic factors. In contrast, prevalent cardiovascular disease and traditional cardiovascular risk factors were stronger predictors of HF risk in White adults. Conclusions: Race-specific and ML-based HF risk models that integrate clinical, laboratory, and biomarker data demonstrated superior performance compared with traditional HF risk and non–race-specific ML models. This approach identifies distinct race-specific contributors of HF.

Cardiovascular disease (CVD) is the leading cause of death in the U.S. and disproportionately affects African Americans (AAs). Routinely measured circulating red blood cell traits, which are highly heritable and differ by ethnicity, are independent predictors for CVD-related traits including hypertension, stroke, coronary heart disease, and CVD mortality. Many genetic loci associated with red blood cell count (RBC), hemoglobin (HGB), hematocrit (HCT), mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC), and red cell distribution width (RDW) have been identified. However, identified genetic variants do not fully explain the heritability of these traits. Epigenetic alterations in DNA methylation likely also explain a portion of red cell trait variance, and detecting methylation quantitative trait loci (meQTLs) can provide critical insight into the development of CVD and CVD health disparities. We performed an epigenome-wide association analysis to identify CpG sites at which DNA methylation levels are associated with red blood cell traits in 1753 participants from the Jackson Heart Study, a population-based cohort of AAs. DNA methylation was measured by the Illumina MethylationEPIC array, interrogating approximately 850,000 CpG sites in peripheral blood mononuclear cells at the baseline exam. Associations were assessed using linear mixed models adjusted for age, sex, cell proportions, genetic ancestry, and experimental batch effects. A Bonferroni-corrected p-value of 9x10 -8 was used to assess statistical significance. We identified many significant differentially methylated CpG sites associated with red blood cell traits. Analysis of RBC, HGB, MCV, MCH, and MCHC revealed a novel highly significant association with a CpG annotated to a non-coding RNA (cg11703701; RBC P = 5.19x10 -22 , HGB P = 1.19x10 -11 , MCV P = 4.69x10 -59 , MCH P = 2.68x10 -67 , MCHC P = 2.32x10 -31 ), as well as a strong signal for a reprogramming-specific differentially methylated region (cg04321267; RBC P = 1.67x10 -16 , HGB P = 3.58x10 -12 , MCV P = 2.12x10 -49 , MCH P = 1.74x10 -57 , MCHC P = 5.91x10 -30 ). Multiple CpGs annotated to genes HBA1 and HBA2 were associated with RBC, MCV, MCH, and MCHC, while ITPKB (cg23740281; HGB P = 4.88x10 -10 , HCT P = 6.20x10 -11 ) and ALDH2 (cg17969951; HGB P = 3.03x10 -11 , HCT P = 8.31x10 -9 ) were significantly associated with both HGB and HCT. Additional CpG sites were significantly associated with HCT ( PLXND1 cg22902177; P = 7.54x10 -11 and ARL1 cg23903357; P = 9.86x10 -11 ) and RDW ( CPNE2 cg09018739; P = 3.03x10 -22 ). These findings shed light on potential hematologic and CVD mechanisms in understudied populations. Future work will explore the role of neighboring SNPs in mediating observed methylation-trait associations, and replicate results in an additional multi-ethnic cohort.

Platelets are critical in inflammation, wound healing, and thrombosis. Platelet count (PLT) and mean platelet volume (MPV) are heritable, frequently assessed measures for which hundreds of genetic associations have been identified, predominantly in European populations. However, causal variants are unknown at most loci, and additional genetic determinants in diverse ancestral populations remain to be discovered. We performed a multi-ancestry whole genome sequencing analysis for PLT in three Phase 1 TOPMed cohorts, the European American Framingham Heart (FHS, n=2259) and Old Order Amish (OOA, n=1095) studies and the African American Jackson Heart Study (JHS, n=3523). MPV was additionally analyzed in JHS and FHS. Heritability in JHS and FHS was 45-73%. Single variant and aggregate burden and sequence kernel association (SKAT) tests [5 and 50 kilobase (kb) sliding windows] were performed using a mixed model approach in GENESIS, with aggregate tests for variants with a minor allele frequency (MAF) < 1% only. Inverse normalized residuals were adjusted for sex, age, study, and an empirical genetic relationship matrix, allowing for heterogeneous variance by study. For PLT, several associations had the same index SNP ( ARHGEF3 , TPM4 ) or were in moderate LD ( TAOK1 , r 2 =0.46) with a previously reported signal; two associations were not in LD (r 2 -8 , MAF=16%), an intergenic variant upstream of telomerase reverse transcriptase ( TERT ), is common only in African ancestry populations (~36%, 1% in Europeans) and was significant only in JHS (p=4.7 x 10 -9 , p=0.91 FHS, absent OOA). This variant and an LD proxy (rs10052815) overlap a GATA1 binding site in erythroblasts, the sister lineage of platelet producing megakaryocytes. An intronic TERT variant (rs2736100) 17 kb away (r 2 =0.025, D’=0.418 in included TOPMed cohorts) has been associated with PLT in Europeans, as well as with red blood cell traits, telomere length, pulmonary fibrosis, and multiple cancer subtypes. The second locus (rs73437176, p=1.5 x 10 -8 , MAF=12%), an intronic variant in progestin and adipoQ receptor family member V ( PAQR5 ), is also more frequent in African ancestry populations (~25%, 4% in Europeans) and significant only in JHS (p=2.5 x 10 -8 , FHS p=0.47, OOA p=0.12); the nearest previously identified signal is ~4 Mb away and not in significant LD (D’DNM3 , ARHGEF3 , CCDC71L , JMJD1C , RHOF , TAOK1 , TPM4 ), with either the same lead variant as previously reported or a variant in close LD (r 2 >0.75). Aggregated sliding window tests did not identify any additional signals. These results show suggestive evidence for association with PLT for two novel variants common only in African ancestry populations, but larger sample sizes of diverse ancestry will be necessary for replication and further discovery of novel loci.

Plasma D-dimer, a fibrin degradation product, is a biomarker of both coagulation and fibrinolysis and a risk factor for both arterial and venous thrombosis. African Americans (AAs) have increased risk of thrombosis and higher levels of D-dimer as compared to European populations, but little is known about the genetic architecture of D-dimer in AAs. D-dimer was assessed in 4163 AAs from the Jackson Heart Study (mean age 55 years, 62.1% female). We hypothesized that D-dimer levels would associate with incident cardiovascular disease events and with both coding and noncoding genetic variants. Median (interquartile range) D-dimer in the cohort was 0.40 (0.26-0.64) μg/ml. D-dimer was associated with increased risk of coronary heart disease (hazard ratio (HR) per standard deviation (SD)=1.26, p=0.012), stroke (HR=1.27, p=0.018), heart failure (HR=1.33, p=0.001), and mortality (HR=1.25, pF3 (top variant rs2022030, p=5.71x10 -12 , minor allele frequency=10.6%, imputation r 2 =1). F3 codes for tissue factor, a necessary co-factor for factor VIIa in the initiation of the extrinsic blood coagulation pathway. The F3 locus had been previously associated with D-dimer levels in European populations but has not been reported for AAs. Data were also available from the Illumina HumanExome Beadchip, version 1.1, in 2502 individuals; single variant analyses and gene-based tests for nonsynonymous, splicing, and nonsense variants with a minor allele frequency GPRASP1 (G protein-coupled receptor associated sorting protein 1, rs145805077, p= 1.84x10 -7 , minor allele frequency=0.3%) was significantly associated with decreased D-dimer. No gene-based test result met the strict multiple comparison significance threshold; the top gene-based result with the sequence kernel association test (SKAT) was for six variants in PPM1M (protein phosphatase, Mg2+/Mn2+ dependent, 1M, p=6.09x10 -6 ). These results highlight D-dimer as an important predictor of cardiovascular disease risk in AAs and that levels of this biomarker are associated with the F3 locus.