UCLA

Introduction: Metabolic syndrome (MetS) has been identified as a significant contributor to increased risk of cardiovascular diseases and numerous adverse health outcomes, placing a considerable financial burden on healthcare systems and societal well-being. With the elderly population experiencing a notable and ongoing rise in MetS prevalence, it becomes crucial to grasp its pathogenesis and explore avenues for mitigating its progression. By understanding and addressing factors that exacerbate MetS, we may potentially hinder the development of chronic diseases and curb the associated escalation in healthcare costs. Recent advancements in aging biomarkers, such as epigenetic clocks, offer promising insights into various health conditions. However, existing research has primarily adopted a cross-sectional approach. Therefore, this dissertation aims to illuminate the prospective association between epigenetic markers and metabolic dysfunction within cohorts comprising racially diverse older adults. Specifically, there are three key objectives of this dissertation: 1) assess the effect of changes in metabolic dysfunction over time on clock accelerations; 2) construct regression models to predict changes in MetS components from initial epigenetic clock accelerations. Additionally, we evaluate the predictive capability of these epigenetic clock accelerations for the onset of MetS. And 3) we investigate the association between baseline DNAm level at CpG sites and subsequent changes in MetS components. Following this, we construct predictive models for incident MetS using baseline DNAm levels at CpG sites through two distinct methods: an epigenome-wide association study approach and a candidate CpG site approach.Methods: To address the first study objective, we used data from 867 participants in the Health and Retirement Study (HRS) who were assessed for demographic characteristics and metabolic biomarkers measured at baseline in 2008. In 2016, using newly collected blood samples, HRS estimated five epigenetic clocks for continuing respondents, including Horvath, Hannum, PhenoAge, GrimAge, and DundinPoAm. Epigenetic clock acceleration, defined as the difference between chronological age and epigenetic age, served as an age-adjusted biomarker for aging. Metabolic dysfunction level was assessed at both baseline and follow-up, which was defined as having three or more of the following abnormalities: hypertension, glycosylated hemoglobin A1c (HbA1c), total cholesterol, high-density lipoprotein cholesterol (HDL), and waist circumference. Linear regression models were conducted to assess each clock acceleration at follow-up in relation to changes in metabolic dysfunction level, while adjusting for baseline covariates and follow-up years. Separate linear regression analyses were employed to regress each clock acceleration against changes in each metabolic dysfunction component, controlling for baseline covariates and follow-up years. Finally, potential effect modification by sex was evaluated in the relationship between changes in metabolic dysfunction level and clocks. Following this, stratified analyses were performed using separate linear regression models for each sex to evaluate the effect of metabolic dysfunction level on clock accelerations. For the second objective, a total of 537 participants from the Women’s Health Initiative were included in this analysis. Demographic characteristics, health behaviors, and epigenetic clock data were collected at baseline, while measurements of MetS components were obtained at baseline and follow-up. Epigenetic age was calculated using five established algorithms: Horvath, Hannum, PhenoAge, GrimAge, and DunedinPACE. Logistic regression models were employed to assess the association between epigenetic clock acceleration and changes in each MetS component, controlling for baseline covariates, follow-up duration, and baseline MetS measures. Similar models were applied to participants at risk of MetS at baseline (n = 281) to explore the association between epigenetic clock accelerations and incident MetS. For the third objective, we analyzed the identical participant cohort as in aim 2, using data on the same covariates pertaining to demographic characteristics and health behaviors. DNAm levels were assessed in blood samples using the Illumina 450k array. Following quality control, our analysis incorporated 421,172 CpG sites. Logistic regression models were employed to assess the impact of DNAm levels at CpG sites on changes in individual MetS components, while adjusting for baseline covariates, follow-up duration, and baseline MetS measures. Incident MetS analysis was carried out in participants at risk of MetS at baseline (n = 281). Two approaches were employed, both using similar models as in the longitudinal change analysis. Model 1 utilized an EWAS approach, while Model 2 focused on a group of candidate CpG sites selected from the longitudinal change analysis. Results: Our findings for the first study objective are as follows. Our study participants in HRS cohort were predominantly female (61%), with a mean age of 66 years, and a mean follow-up time of 8 years. When investigating how changes in indicators of metabolic dysfunction over time affect subsequent clock accelerations, participants who transitioned from a high to a low risk of metabolic dysfunction experienced a slower PhenoAge (β = -1.61, 95% CI -3.11, -0.12) compared to those who remained consistently low risk. Regarding effects of individual metabolic components, participants who improved from abnormal to normal HDL levels exhibited accelerated PhenoAge (β = 3.05, 95% CI 1.59, 4.51) and DunedinPoAm (β = 0.03, 95% CI 0.01, 0.05) compared to those who maintained healthy HDL levels. Similarly, those who developed new abnormalities in HDL levels showed acceleration in PhenoAge (β = 1.89, 95% CI 0.50, 3.29) and DunedinPoAm (β = 0.02, 95% CI 0.01, 0.04) compared to those consistently within the healthy range. Moreover, individuals who returned their total cholesterol levels to normal experienced acceleration in Hannum (β = 2.96, 95% CI 0.91, 5.01), GrimAge (β = 3.19, 95% CI 0.33, 6.04), and DunedinPoAm (β = 0.06, 95% CI 0.02, 0.10), while those experiencing worsening levels over time exhibited acceleration in Hannum (β = 2.45, 95% CI 1.00, 3.90) but deceleration in DunedinPoAm (β = -0.04, 95% CI -0.06, -0.01). Participants with improved HbA1c levels also showed accelerated GrimAge (β = 1.83, 95% CI 0.59, 3.07) compared to those within the normal blood sugar range. Lastly, no heterogeneity by sex was observed in the relationship between metabolic dysfunction level and clocks. Regarding the second study objective, here are our findings. At baseline, WHI participants had a mean age of 63.6 years, with a mean follow-up duration of 15.4 years. Among the 537 participants, changes in glucose levels were significantly associated with accelerated Horvath (OR = 1.51, 95% CI 1.22, 1.87), PhenoAge (OR = 1.33, 95% CI 1.10, 1.61), and GrimAge (OR = 1.61, 95% CI 1.10, 2.35). Changes in waist circumference were associated with accelerated Hannum (OR = 1.46, 95% CI 1.14, 1.87) and DunedinPACE (OR = 1.33, 95% CI 1.06, 1.66) clocks, while changes in HDL levels were associated with accelerated DunedinPACE (OR = 1.43, 95% CI 1.15, 1.76). Similar trends were observed among participants at risk of MetS. Additionally, GrimAge was associated with changes in HDL (OR = 1.67, 95% CI 1.03, 2.72) and waist circumference (OR = 1.73, 95% CI 1.06, 2.85), while DunedinPACE was associated with changes in triglycerides (OR = 1.46, 95% CI 1.09, 1.95) and glucose levels (OR = 1.47, 95% CI 1.06, 2.02). In the incident MetS analysis, GrimAge (OR = 1.83, 95% CI 1.17, 2.87) and DunedinPACE (OR = 1.33, 95% CI 1.02, 1.75) were associated with the onset of MetS. The findings pertaining to the third study objective are outlined below. Among the 537 participants, we identified 92 CpG sites associated with changes in MetS components, annotated to 69 genes. Some of these genes have demonstrated functional relevance to MetS previously, such as CXCR7 (cg15066967) and KDR (cg14323109) for hypertension, PPT2 (cg24283914) for HDL levels, ZFHX3 (cg05918327), NEU1 (cg10849498), and BCKDHB (cg19242448) for triglycerides, and PLIN1 (cg08749443) for waist circumference. In the incident MetS analyses, Model 1 identified 19 CpG sites, while Model 2 found 14 CpG sites linked to MetS onset. Some of these sites have been annotated to genes relevant to MetS, including PRKCZ (cg00095688), GPR133 (cg14499928), IGFL-1 (cg25246031), and ESCRT (cg10035272). Both approaches produced ROC curves with good discriminability of MetS compared to a model considering only current MetS conditions and without DNAm markers. Conclusion: In sum, this dissertation reported the bi-directional relationship between epigenetic clocks and changes in metabolic dysfunction indicators, as well as several CpG sites that may be implicated in the progression of MetS in the elderly population. These findings indicate that metabolic dysfunction accelerates epigenetic clocks, which in turn predicts MetS incidence. Consequently, epigenetic clocks may serve as a valuable tool for categorizing risk among older adults, identifying those at high risk of metabolic dysfunction. Furthermore, the CpG sites associated with MetS progression could provide directions for further investigation into their biological functions and their contributions to MetS development.