Publications

BACKGROUND: Fibrinogen plays an essential role in blood coagulation and inflammation. Circulating fibrinogen levels may be determined based on interindividual differences in DNA methylation at cytosine-phosphate-guanine (CpG) sites and vice versa.

OBJECTIVES: To perform an EWAS to examine an association between blood DNA methylation levels and circulating fibrinogen levels to better understand its biological and pathophysiological actions.

METHODS: We performed an epigenome-wide association study of circulating fibrinogen levels in 18 037 White, Black, American Indian, and Hispanic participants, representing 14 studies from the Cohorts for Heart and Aging Research in Genomic Epidemiology consortium. Circulating leukocyte DNA methylation was measured using the Illumina 450K array in 12 904 participants and using the EPIC array in 5133 participants. In each study, an epigenome-wide association study of fibrinogen was performed using linear mixed models adjusted for potential confounders. Study-specific results were combined using array-specific meta-analysis, followed by cross-replication of epigenome-wide significant associations. We compared models with and without CRP adjustment to examine the role of inflammation.

RESULTS: We identified 208 and 87 significant CpG sites associated with fibrinogen levels from the 450K (p < 1.03 × 10-7) and EPIC arrays (p < 5.78 × 10-8), respectively. There were 78 associations from the 450K array that replicated in the EPIC array and 26 vice versa. After accounting for overlapping sites, there were 83 replicated CpG sites located in 61 loci, of which only 4 have been previously reported for fibrinogen. The examples of genes located near these CpG sites were SOCS3 and AIM2, which are involved in inflammatory pathways. The associations of all 83 replicated CpG sites were attenuated after CRP adjustment, although many remained significant.

CONCLUSION: We identified 83 CpG sites associated with circulating fibrinogen levels. These associations are partially driven by inflammatory pathways shared by both fibrinogen and CRP.

BACKGROUND. The confounder-corrected chemical shift-encoded MRI (CSE-MRI) sequence used to determine proton density fat fraction (PDFF) for hepatic fat quantification is not widely available. As an alternative, hepatic fat can be assessed by a two-point Dixon method to calculate signal fat fraction (FF) from conventional T1-weighted in- and opposed-phase (IOP) images, although signal FF is prone to biases, leading to inaccurate quantification. OBJECTIVE. The purpose of this study was to compare hepatic fat quantification by use of PDFF inferred from conventional T1-weighted IOP images and deep-learning convolutional neural networks (CNNs) with quantification by use of two-point Dixon signal FF with CSE-MRI PDFF as the reference standard. METHODS. This study entailed retrospective analysis of data from 292 participants (203 women, 89 men; mean age, 53.7 ± 12.0 [SD] years) enrolled at two sites from September 1, 2017, to December 18, 2019, in the Strong Heart Family Study (a prospective population-based study of American Indian communities). Participants underwent liver MRI (site A, 3 T; site B, 1.5 T) including T1-weighted IOP MRI and CSE-MRI (used to reconstruct CSE PDFF and CSE R2* maps). With CSE PDFF as reference, a CNN was trained in a random sample of 218 (75%) participants to infer voxel-by-voxel PDFF maps from T1-weighted IOP images; testing was performed in the other 74 (25%) participants. Parametric values from the entire liver were automatically extracted. Per-participant median CNN-inferred PDFF and median two-point Dixon signal FF were compared with reference median CSE-MRI PDFF by means of linear regression analysis, intraclass correlation coefficient (ICC), and Bland-Altman analysis. The code is publicly available at github.com/kang927/CNN-inference-of-PDFF-from-T1w-IOP-MR. RESULTS. In the 74 test-set participants, reference CSE PDFF ranged from 1% to 32% (mean, 11.3% ± 8.3% [SD]); reference CSE R2* ranged from 31 to 457 seconds-1 (mean, 62.4 ± 67.3 seconds-1 [SD]). Agreement metrics with reference to CSE PDFF for CNN-inferred PDFF were ICC = 0.99, bias = -0.19%, 95% limits of agreement (LoA) = (-2.80%, 2.71%) and for two-point Dixon signal FF were ICC = 0.93, bias = -1.11%, LoA = (-7.54%, 5.33%). CONCLUSION. Agreement with reference CSE PDFF was better for CNN-inferred PDFF from conventional T1-weighted IOP images than for two-point Dixon signal FF. Further investigation is needed in individuals with moderate-to-severe iron overload. CLINICAL IMPACT. Measurement of CNN-inferred PDFF from widely available T1-weighted IOP images may facilitate adoption of hepatic PDFF as a quantitative bio-marker for liver fat assessment, expanding opportunities to screen for hepatic steatosis and nonalcoholic fatty liver disease.

PURPOSE: Liver cancer incidence among American Indians/Alaska Natives has risen over the past 20 years. Peripheral blood DNA methylation may be associated with liver cancer and could be used as a biomarker for cancer risk. We evaluated the association of blood DNA methylation with risk of liver cancer.

METHODS: We conducted a prospective cohort study in 2324 American Indians, between age 45 and 75 years, from Arizona, Oklahoma, North Dakota and South Dakota who participated in the Strong Heart Study between 1989 and 1991. Liver cancer deaths (n = 21) were ascertained using death certificates obtained through 2017. The mean follow-up duration (SD) for non-cases was 25.1 (5.6) years and for cases, 11.0 (8.8) years. DNA methylation was assessed from blood samples collected at baseline using MethylationEPIC BeadChip 850 K arrays. We used Cox regression models adjusted for age, sex, center, body mass index, low-density lipoprotein cholesterol, smoking, alcohol consumption, and immune cell proportions to examine the associations.

RESULTS: We identified 9 CpG sites associated with liver cancer. cg16057201 annotated to MRFAP1) was hypermethylated among cases vs. non-cases (hazard ratio (HR) for one standard deviation increase in methylation was 1.25 (95% CI 1.14, 1.37). The other eight CpGs were hypomethylated and the corresponding HRs (95% CI) ranged from 0.58 (0.44, 0.75) for cg04967787 (annotated to PPRC1) to 0.77 (0.67, 0.88) for cg08550308. We also assessed 7 differentially methylated CpG sites associated with liver cancer in previous studies. The adjusted HR for cg15079934 (annotated to LPS1) was 1.93 (95% CI 1.10, 3.39).

CONCLUSIONS: Blood DNA methylation may be associated with liver cancer mortality and may be altered during the development of liver cancer.

Hemophilia-A (HA) is caused by heterogeneous loss-of-function factor (F)VIII gene (F8)-mutations and deficiencies in plasma-FVIII-activity that impair intrinsic-pathway-mediated coagulation-amplification. The standard-of-care for severe-HA-patients is regular infusions of therapeutic-FVIII-proteins (tFVIIIs) but  30% develop neutralizing-tFVIII-antibodies called "FVIII-inhibitors (FEIs)" and become refractory. We used the PATH study and ImmunoChip to scan immune-mediated-disease (IMD)-genes for novel and/or replicated genomic-sequence-variations associated with baseline-FEI-status while accounting for non-independence of data due to genetic-relatedness and F8-mutational-heterogeneity. The baseline-FEI-status of 450 North American PATH subjects-206 with black-African-ancestry and 244 with white-European-ancestry-was the dependent variable. The F8-mutation-data and a genetic-relatedness matrix were incorporated into a binary linear-mixed model of genetic association with baseline-FEI-status. We adopted a gene-centric-association-strategy to scan, as candidates, pleiotropic-IMD-genes implicated in the development of either ³2 autoimmune-/autoinflammatory-disorders (AADs) or ³1 AAD and FEIs. Baseline-FEI-status was significantly associated with SNPs assigned to NOS2A (rs117382854; p=3.2E-6) and B3GNT2 (rs10176009; p=5.1E-6), which have functions in anti-microbial-/-tumoral-immunity. Among IMD-genes implicated in FEI-risk previously, we identified strong associations with CTLA4 assigned SNPs (p=2.2E-5). The F8-mutation-effect underlies  15% of the total heritability for baseline-FEI-status. Additive genetic heritability and SNPs in IMD-genes account for >50% of the patient-specific variability in baseline-FEI-status. Race is a significant determinant independent of F8-mutation-effects and non-F8-genetics.

BACKGROUND: Inorganic arsenic (As) may increase the risk of cardiovascular disease (CVD) and all-cause mortality through accelerated aging, which can be estimated using epigenetic-based measures.

OBJECTIVES: We evaluated three DNA methylation-based aging measures (PhenoAge, GrimAge, DunedinPACE) (epigenetic aging measures) as potential mediators of the previously reported association of As exposure with CVD incidence, CVD mortality, and all-cause mortality in the Strong Heart Study (SHS), an epidemiological cohort of American Indian adults.

METHODS: Blood DNA methylation and urinary As levels were measured in 2,323 SHS participants (41.5% men, mean age of 55 years old). PhenoAge and GrimAge values were calculated using a residual-based method. We tested the association of urinary As with epigenetic aging measures using linear regression, the association of epigenetic aging measures with the three health outcomes using additive hazards models, and the mediation of As-related CVD incidence, CVD mortality, and all-cause mortality by epigenetic aging measures using the product of coefficients method.

RESULTS: SHS participants with higher vs. lower urinary As levels had similar PhenoAge age, older GrimAge age, and faster DunedinPACE. An interquartile range increase in urinary As was associated with higher of PhenoAge age acceleration [mean difference (95% confidence interval)=0.48 (0.17, 0.80) years], GrimAge age acceleration [0.80 (0.60, 1.00) years], and DunedinPACE [0.011 (0.005, 0.018)], after adjusting for age, sex, center location, genetic components, smoking status, and body mass index. Of the 347 incident CVD events per 100,000 person-years associated with a doubling in As exposure, 21.3% (9.1, 57.1) and 22.6% (9.5, 56.9), were attributable to differences in GrimAge and DunedinPACE, respectively.

DISCUSSION: Arsenic exposure was associated with older GrimAge and faster DunedinPACE measures of biological age. Furthermore, accelerated biological aging measured from DNA methylation accounted for a relevant fraction of As-associated risk for CVD, CVD mortality, and all-cause mortality in the SHS, supporting the role of As in accelerated aging. Research of the biological underpinnings can contribute to a better understanding of the role of aging in arsenic-related disease. https://doi.org/10.1289/EHP11981.

The common marmoset (Callithrix jacchus) is one of the most widely used nonhuman primate models of human disease. Owing to limitations in sequencing technology, early genome assemblies of this species using short-read sequencing suffered from gaps. In addition, the genetic diversity of the species has not yet been adequately explored. Using long-read genome sequencing and expert annotation, we generated a high-quality genome resource creating a 2.898 Gb marmoset genome in which most of the euchromatin portion is assembled contiguously (contig N50 = 25.23 Mbp, scaffold N50 = 98.2 Mbp). We then performed whole genome sequencing on 84 marmosets sampling the genetic diversity from several marmoset research centers. We identified a total of 19.1 million single nucleotide variants (SNVs), of which 11.9 million can be reliably mapped to orthologous locations in the human genome. We also observed 2.8 million small insertion/deletion variants. This dataset includes an average of 5.4 million SNVs per marmoset individual and a total of 74,088 missense variants in protein-coding genes. Of the 4956 variants orthologous to human ClinVar SNVs (present in the same annotated gene and with the same functional consequence in marmoset and human), 27 have a clinical significance of pathogenic and/or likely pathogenic. This important marmoset genomic resource will help guide genetic analyses of natural variation, the discovery of spontaneous functional variation relevant to human disease models, and the development of genetically engineered marmoset disease models.

OBJECTIVES: To examine the associations of dietary Mg intake with inflammatory biomarkers (C-reactive protein (CRP) and interleukin 6 (IL-6)), and the interaction of dietary Mg intake with single nucleotide polymorphism (SNP) rs3740393, a SNP related to Mg metabolism and transport, on CRP and IL-6 among American Indians (AIs).

METHODS: This cross-sectional study included AI participants (n = 1,924) from the Strong Heart Family Study (SHFS). Mg intake from foods and dietary supplements was ascertained using a 119-item Block food frequency questionnaire, CRP and IL-6 were measured from blood, and SNP rs3740393 was genotyped using MetaboChip. Generalized estimating equations were used to examine associations of Mg intake, and the interaction between rs3740393 and dietary Mg, with CRP and IL-6.

RESULTS: Reported Mg intake was not associated with CRP or IL-6, irrespective of genotype. A significant interaction (p-interaction = 0.018) was observed between Mg intake and rs3740393 on IL-6. Among participants with the C/C genotype, for every 1 SD higher in log-Mg, log-IL-6 was 0.04 (95% CI: -0.10 to 0.17) pg/mL higher. Among participants with the C/G genotype, for every 1 SD higher in log-Mg, log-IL-6 was 0.08 (95% CI: -0.21 to 0.05) pg/mL lower, and among participants with the G/G genotype, for every 1 SD higher in log-Mg, log-IL-6 was 0.19 (95% CI: -0.38 to -0.01) pg/mL lower.

CONCLUSIONS: Mg intake may be associated with lower IL-6 with increasing dosage of the G allele at rs3740393. Future research is necessary to replicate this finding and examine other Mg-related genes that influence associations of Mg intake with inflammation.

BACKGROUND: Identifying lipidomic markers of diet quality is needed to inform the development of biomarkers of diet, and to understand the mechanisms driving the diet- coronary heart disease (CHD) association.

OBJECTIVE: To identify lipidomic markers of diet quality and examine whether these lipids are associated with incident CHD.

METHODS: Using LC-MS, we measured 1,542 lipid species from 1,694 American Indian adults (aged 18-75, 62% female) in the Strong Heart Family Study. Participants were followed for development of CHD through 2020. Information on past year diet was collected using the Block Food Frequency Questionnaire, and diet quality was assessed using the Alternative Healthy Eating Index-2010 (AHEI). Mixed-effects linear regression was used to identify individual lipids cross-sectionally associated with AHEI. In prospective analysis, Cox frailty model was used to estimate the hazard ratio (HR) of each AHEI-related lipid for incident CHD. All models adjusted for age, sex, center, education, BMI, smoking, alcohol drinking, level of physical activity, energy intake, diabetes, hypertension, and use of lipid-lowering drugs. Multiple testing was controlled at false discovery rate <0.05.

RESULTS: Among 1,542 lipid species measured, 71 lipid species (23 known) including acylcarnitine, cholesterol esters, glycerophospholipids, sphingomyelins and triacylglycerols, were associated with AHEI. Most of the identified lipids were associated with consumption of ω-3 fatty acids. In total, 147 participants developed CHD during a mean follow-up of 17.8 years. Among the diet-related lipids, 10 lipids (5 known, including CE(22:5)B, PC(p-14:0/22:1)/PC(o-14:0/22:1), PC(p-38:3)/PC(o-38:4)B, PE(p-18:0/20:4)/PE(o-18:0/20:4), and SM(d36:2)A) were associated with incident CHD. On average, each standard deviation increase in the baseline level of these 5 lipids was associated with 17%-23% increased risk of CHD [HR (95%CI) ranged from 1.17 (1, 1.36) to 1.23 (1.05, 1.43)].

CONCLUSION: We identified lipidomic markers of diet quality in American Indian adults. Some diet-related lipids are associated with risk of CHD beyond established risk factors.

INTRODUCTION: Native American communities suffer disproportionately from elevated metal exposures and increased risk for cardiovascular diseases and diabetes. DNA methylation is a sensitive biomarker of aging-related processes and novel epigenetic-based "clocks" can be used to estimate accelerated biological aging that may underlie increased risk. Metals alter DNA methylation, yet little is known about their individual and combined impact on epigenetic age acceleration. Our objective was to investigate the associations of metals on several DNA methylation-based aging measures in the Strong Heart Study (SHS) cohort.

METHODS: Blood DNA methylation data from 2,301 SHS participants was used to calculate age acceleration of epigenetic clocks (PhenoAge, GrimAge, DunedinPACE, Hannum, Horvath). Urinary metals [arsenic (As), cadmium (Cd), tungsten (W), zinc (Zn), selenium (Se), molybdenum (Mo)] were creatinine-adjusted and categorized into quartiles. We examined associations of individual metals through linear regression models and used Bayesian Kernel Machine Regression (BKMR) for the impact of the total metal mixture on epigenetic age acceleration.

RESULTS: The mixture of nonessential metals (W, As, Cd) was associated with greater GrimAge acceleration and DunedinPACE, while the essential metal mixture (Se, Zn, Mo) was associated with lower epigenetic age acceleration. Cd was associated with increased epigenetic age acceleration across all clocks and BKMR analysis suggested nonlinear associations between Se and DunedinPACE, GrimAge, and PhenoAge acceleration. No interactions between individual metals were observed. The associations between Cd, Zn, and epigenetic age acceleration were greater in never smokers in comparison to current/former smokers.

CONCLUSION: Nonessential metals were positively associated with greater epigenetic age acceleration, with strongest associations observed between Cd and DunedinPACE and GrimAge acceleration. In contrast, essential metals were associated with lower epigenetic aging. Examining the influence of metal mixtures on epigenetic age acceleration can provide insight into metals and aging-related diseases.

Background: The confounder-corrected chemical-shift-encoded MRI (CSE-MRI) sequence used to determine proton-density fat fraction (PDFF) for hepatic fat quantification lacks wide availability. Hepatic fat can alternatively be assessed using a 2-point Dixon method to calculate signal fat fraction (FF) from conventional T1-weighted (T1W) in-and-opposed-phase (IOP) images, although signal FF is prone to biases leading to inaccurate quantification. Objective: To compare hepatic fat quantification between PDFF inferred from conventional T1W IOP images using deep-learning convolutional neural networks (CNNs) and 2-point Dixon signal FF, using CSE-MRI PDFF as reference standard. Methods: This study entailed retrospective analysis of 292 participants (mean age, 53.7±12.0 years; 89 men, 103 women) enrolled at two sites from September 1, 2017 to December 18, 2019 in the Strong Heart Family Study (a prospective population-based study of American Indian communities). Participants underwent liver MRI (site A: 3 T; site B: 1.5 T) including T1W IOP MRI and CSE-MRI (used to reconstruct CSE-PDFF and CSE-R2* maps). Using CSEPDFF as reference, a CNN was trained in 218 (75%) randomly selected participants to infer voxel-by-voxel PDFF maps from T1W IOP images; testing was performed in the remaining 74 (25%) participants. Parametric values from the entire liver were automatically extracted. Per-participant median CNN-inferred PDFF and median 2-point Dixon signal FF were compared with reference median CSE-MRI PDFF using linear regression analysis, intraclass correlation (ICC), and Bland-Altman analysis. Code is publicly available: https://github.com/kang927/CNN-inference-of-PDFF-from-T1w-IOP-MR. Results: In the 74 test-set participants, reference CSE-PDFF ranged from 1% to 32% (mean, 11.3%±8.3%); reference CSE-R2* ranged from 31 to 457 s-1 (mean, 62.4±67.3 s-1). Agreement metrics with reference CSE-PDFF for CNN-inferred PDFF were: ICC=0.99, bias=-0.19%, 95% limits of agreement (LoA)=[-2.80%, 2.71%], and for 2-point Dixon signal FF were: ICC=0.93, bias=-1.11%, LoA=[-7.54%, 5.33%]. Conclusion: Agreement with reference CSE-PDFF was better for CNN-inferred PDFF using conventional T1W IOP images than for 2-point Dixon signal FF. Further investigation is needed in individuals with moderate-to-severe iron overload. Clinical Impact: CNN-inferred PDFF using widely available T1W IOP images may facilitate adoption of hepatic PDFF as a quantitative biomarker for liver fat assessment, expanding opportunities to screen for hepatic steatosis and non-alcoholic fatty liver disease.