Epigenetic Profiling of Obesity and Smoking

Show full item record



Permalink

http://urn.fi/URN:ISBN:978-951-51-5807-9
Title: Epigenetic Profiling of Obesity and Smoking
Author: Bollepalli, Sailalitha Spurthy
Contributor: University of Helsinki, Faculty of Medicine
Doctoral Program in Population Health
Institute for Molecular Medicine Finland (FIMM)
Department of Public Health, University of Helsinki
Publisher: Helsingin yliopisto
Date: 2020-03-27
URI: http://urn.fi/URN:ISBN:978-951-51-5807-9
http://hdl.handle.net/10138/310395
Thesis level: Doctoral dissertation (article-based)
Abstract: Obesity and smoking are the two major preventable causes of global mortality associated with a multitude of comorbidities, inflicting greater public health and economic burden. Complex interactions between genetic and environmental factors influence susceptibility to obesity and smoking. Epigenetic modifications provide a mechanistic link between genetic and non-genetic factors causing complex diseases or traits. Epigenetic modifications also function as an additional layer of gene regulation by modifying the structure and accessibility of DNA and chromatin. The fundamental objective of this thesis is to elucidate the role of epigenetic and transcriptomic markers in obesity and smoking. Hence, this thesis focuses on identifying epigenetic and transcriptomic markers associated with weight loss and smoking behavior using different study designs and by applying computational and statistical approaches. Genome-wide transcriptome and methylome were assessed in an unbiased, hypothesis-free setting to identify weight-loss and smoking-associated signals in Study I and II, respectively. Validation of the main findings from the discovery analyses and integration of transcriptomic and methylation data were performed to assess the validity and biological significance of the identified markers. A machine learning approach was employed in Study III to develop a robust smoking status classifier based on DNA methylation profiles. The performance of the classifier was tested in three different test datasets and also in comparison with two other existing approaches. Therefore, this thesis encompasses both application and method development aspects to achieve the corresponding aims of the studies. In Study I, clinical parameters, genome-wide transcriptome, and methylome analyses were assessed longitudinally at three time points during a one-year weight loss intervention study, to understand the temporal changes in transcriptome and methylome of subcutaneous adipose tissue (SAT) in response to weight-loss. Results from the discovery analyses were validated using monozygotic (MZ) twin pairs discordant for acquired obesity, to examine whether weight loss and acquired obesity exhibit reciprocal transcriptome and methylome profiles. Gene expression and methylation profiles of the SAT at the three time points were also integrated to enhance our understanding of their interaction and thereby their contribution in weight loss. Based on the weight loss trajectory of the participants, three comparisons were performed: short-term (baseline to the fifth month), continuous (fifth to twelfth month), and long-term weight loss (baseline to twelfth month). Clinical parameters were improved with the weight loss (e.g. from baseline to fifth month, total and low-density lipoprotein cholesterol; triglycerides; and systolic blood pressure decreased and insulin sensitivity increased) and several significant transcriptome profiles were identified in response to weight loss at the three comparisons. No genome-wide significant methylation profiles were identified for the three comparisons. However, several significant correlations were observed between expression and methylation, indicating a potential regulatory role of DNA methylation in weight loss -associated transcriptome profiles. At the pathway level, short-term weight loss was implicated in lipoprotein metabolism and long-term weight loss associated with various pathways associated with multiple functions of the SAT. Furthermore, several weight loss -associated genes exhibited opposite direction of expression in acquired obesity in the validation cohort of MZ twins, validating the robustness of identified associations. In Study II, discovery analyses focused on understanding the widespread effects of smoking on SAT by simultaneous assessment of genome-wide transcriptome and methylome of SAT. Discovery analyses performed on the current (n=54) and never (n=291) smokers in the TwinsUK cohort identified 42 significantly differentially methylated signals and 42 significant differentially expressed genes (DEG) indicating a substantial impact of smoking on metabolically important SAT. Integration of these results revealed an overlap at five genes (AHRR, CYP1A1, CYP1B1, CYTL1, and F2RL3) comprising 14 CpG sites. To further characterize the widespread effects of smoking on metabolic disease risk three adiposity phenotypes (total fat mass [TFM], android-to-gynoid fat ratio [AGR] and visceral fat mass [VFM]) were assessed with regards to the identified smoking-associated methylation and expression signals. Three CpG sites in CYP1A1 showed significant associations with VFM and AGR, and an inverse association was identified between methylation levels of cg14120703 (NOTCH1) and AGR. To validate these associations, a subset of younger Finnish twins (n=69, 21 current smokers) was used as a replication cohort. The overall inverse association between cg10009577 (CYP1A1) and AGR was replicated and exhibited a similar direction for interaction effects between smoking status and AGR. However, this association did not reach the genome-wide significance level. Expression levels of F2RL3 showed a significant association with all three adiposity phenotypes. While OR51E1 expression levels were significantly associated with AGR and VFM. Our results show that smoking affects both the methylome and transcriptome of the SAT with overlapping signals. Furthermore, smoking-associated methylation and transcriptome profiles are also associated with adiposity phenotypes indicating a broader impact of smoking on human metabolic health. In Study III, I developed a methylation-based smoking status classifier using a machine learning approach to overcome the limitations of cotinine and carbon monoxide biomarkers (i.e. limited to measuring recent exposure to smoking due to their short half-lives in body fluids) and the existing DNA methylation score-based approaches and to advance the practical applicability of smoking-associated methylation signals. I considered three smoking status categories (current, former and never) and used multinomial LASSO regression coupled with internal cross-validation to build the classifier. I demonstrated the global applicability and robustness of our classifier by evaluation of its performance in three independent test datasets from different populations and also compared the performance with two existing approaches. Our classifier differs from the existing approaches by curtailing the need to compute a threshold value specific to each dataset to predict smoking status. Our classifier showed good discriminative ability in identifying current and never smokers compared to other approaches. I also performed an extensive phenotypic evaluation to understand the results of our classifier. Accurate classification of former smokers is challenging as their classification is affected by cessation time and smoking intensity prior to quitting. I provide the functionalities of our classifier including other the two methods as an R package EpiSmokEr (Epigenetic Smoking status Estimator), facilitating prediction of smoking status in future studies. In conclusion, this doctoral thesis (1) enhances our understanding of obesity and smoking by integrating methylation and transcriptome data and identifying several weight-loss and smoking-associated signals, (2) shows wide-spread impact of smoking on metabolic health risk by evaluating the associations between smoking-associated signals and adiposity measures, and (3) demonstrates the role of DNA methylation profiles as a robust biomarker to predict smoking status by developing a smoking-status classifier.
Subject: Public Health
Rights: This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.


Files in this item

Total number of downloads: Loading...

Files Size Format View
EPIGENET.pdf 2.558Mb PDF View/Open

This item appears in the following Collection(s)

Show full item record