UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Genome-wide analysis of DNA methylation variance in healthy human subjects Jiang, Ruiwei 2015

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


24-ubc_2015_september_jiang_ruiwei.pdf [ 2.74MB ]
JSON: 24-1.0166261.json
JSON-LD: 24-1.0166261-ld.json
RDF/XML (Pretty): 24-1.0166261-rdf.xml
RDF/JSON: 24-1.0166261-rdf.json
Turtle: 24-1.0166261-turtle.txt
N-Triples: 24-1.0166261-rdf-ntriples.txt
Original Record: 24-1.0166261-source.json
Full Text

Full Text

GENOME-WIDE ANALYSIS OF DNA METHYLATION VARIANCE IN HEALTHY HUMAN SUBJECTS  by  Ruiwei Jiang  B.Sc., The University of Victoria, 2011  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF SCIENCE in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Genome Science and Technology)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  September 2015  © Ruiwei Jiang, 2015 ii  Abstract  DNA methylation is a type of epigenetic modification that modulates gene expression by acting as an intermediate between genes and environment; this in turn could trigger phenotypic changes with widespread implications in both disease and population models. Unlike DNA sequence, which is relatively stable and finite, DNA methylation presents itself differently in different tissues, and it is described as the sum of interactions affecting attachment of methyl groups to DNA mostly as a result of development and aging, with minor influences from stochastic variability, and environmental factors. Most studies involving DNA methylation focus on finding epigenetic changes related to pathogenicity or disease, as a result, there are certain foundational questions that remain unanswered. In order to translate the current knowledge into reliable insights, it is important to answer these questions, then standardize research methods and establish reference epigenomes. Here we begin to address this challenge through two avenues: epigenomic characterization and environmental interaction. To characterize the epigenome, we monitored the peripheral blood mononuclear cell DNA methylation levels from healthy subjects over a circadian day, a month, and under prolonged sample storage. We also investigated tissue specific variability in DNA methylation by comparing matched peripheral blood mononuclear and buccal epithelial cell samples from healthy subjects. Lastly, we analyzed the impact of diesel exhaust on the DNA methylation. We discovered that while overall DNA methylation was stable within a circadian day, certain loci demonstrated significant changes over the course of a month. Prolonged sample storage, on the other hand, had an even larger effect on DNA methylation. When we compared differences across tissues, we found that although both tissues showed extensive probe-wise variability, the specific regions and magnitude of that variability differed iii  strongly between tissues. Lastly, in light of environmental influences, we observed that DNA methylation was sensitive to even short-term exposure to diesel exhaust, and we identified associated CpG sites across the functional genome, as well as in Alu and LINE1 repetitive elements, with most of these exposure sensitive sites demonstrating loss of DNA methylation.  iv  Preface This thesis was written in plural since work described here contains contributions from other members of the Kobor lab, most involving organizing and conducting microarray experiments, and interpretation of data analysis results.  The Illumina GoldenGate and 450k methylation assays described in this thesis was conducted primarily by Sarah Neumann and Lucia Lam, with assistance from members of the Kobor lab. The subsequent data analysis was conducted by Ruiwei Jiang. Presentation of results, and writing of sections 1.2.2, 3.3, and 4.1.3 were conducted with contributions from both Ruiwei Jiang and Dr. Meaghan J Jones. Presentation of results, and writing of all other sections were conducted by Ruiwei Jiang with input from Dr. Meaghan J Jones.  A version of sections 1.2.2, 3.3, and 4.1.3 has been published. Jiang R, Jones MJ, Chen E, Neumann SM, Fraser HB, Miller GE, Kobor MS. (2015) Discordance of DNA methylation variance between two accessible human tissues. Scientific Reports. 5: 8257. I conducted all the analysis. Results interpretation and writing were done in collaboration with Dr. Meaghan Jones.  A version of sections 3.4, and 4.2 has been published. Jiang R, Jones MJ, Sava F, Kobor MS, Carlsten C. (2014) Short-term diesel exhaust inhalation in a controlled human crossover study is associated with changes in DNA methylation of circulating mononuclear cells in asthmatics. Particle and Fibre Toxicology. 11:71. I conducted all the analysis and wrote most of the manuscript.  v  All experimental protocols were approved by the University of British Columbia’s Research Ethics Board, and all methods were carried out in accordance with the approved guidelines. All subjects gave written consent before participating. Protocol for subject exposure to diesel exhaust was approved by the institutional review board for human studies at the University of British Columbia under H08-02288.  vi  Table of Contents  Abstract .......................................................................................................................................... ii Preface ........................................................................................................................................... iv Table of Contents ......................................................................................................................... vi List of Tables ................................................................................................................................ ix List of Figures ................................................................................................................................ x List of Abbreviations .................................................................................................................. xii Acknowledgements .................................................................................................................... xiii Dedication ................................................................................................................................... xiv Chapter 1: Introduction ......................................................................................................................... 1 1.1 DNA methylation........................................................................................................................ 1 1.1.1 Epigenetics and the central dogma ..................................................................................................... 1 1.1.2 The biology and functional importance of DNA methylation ............................................................ 3 1.1.3 The dynamic nature of DNA methylation is associated with diseases and phenotypes ..................... 5 1.1.4 Epigenetic epidemiology and current technologies ............................................................................ 6 1.2 Normal variations in DNA methylation...................................................................................... 8 1.2.1 Temporal stability of methylation mark over days and within a circadian day .................................. 8 1.2.2 Interindividual variability in DNA methylation in the context of tissue specificity ........................... 9 1.3 Environmentally associated variability of DNA methylation ................................................... 12 1.3.1 Clinical phenotypes of environment particulate matter pollution..................................................... 12 1.3.2 Impact of particulate matter pollution on human DNA methylation ................................................ 13 vii  1.4 Rationale and thesis objectives ................................................................................................. 16 Chapter 2: Material and Methods....................................................................................................... 19 2.1 Overview .................................................................................................................................. 19 2.2 Study design and data preprocessing ........................................................................................ 19 2.2.1 Illumina GoldenGate array experiments ........................................................................................... 19 2.2.2 Illumina Infinium HumanMethylation450 array experiment ........................................................... 20 2.2.3 DNA methylation temporal stability and experimental influences – study design ........................... 20 2.2.4 DNA methylation temporal stability and experimental influences – data preprocessing ................. 21 2.2.5 Interindividual variability in DNA methylation – study design ....................................................... 22 2.2.6 Interindividual variability in DNA methylation – data preprocessing .............................................. 22 2.2.7 Environmental effects on DNA methylation – exposure design ...................................................... 22 2.2.8 Environmental effects on DNA methylation – data preprocessing .................................................. 24 2.3 Statistical analysis ..................................................................................................................... 25 2.3.1 Differential and tissue specific DNA methylation ............................................................................ 25 2.3.2 Intraclass correlation ........................................................................................................................ 25 2.3.3 Correlation between tissues .............................................................................................................. 26 2.3.4 Concordance in probe variance across tissues .................................................................................. 26 2.3.5 DNA methylation and demographics ............................................................................................... 27 2.3.6 Bootstrapping and resampling .......................................................................................................... 27 2.3.7 Principal component analysis ........................................................................................................... 29 2.3.8 Identification of Alu, LINE1, and microRNA associated CpG sites ................................................ 30 2.3.9 Linear regression modeling .............................................................................................................. 30 2.3.10 Linear mixed effects modeling ....................................................................................................... 31 2.3.11 DAVID functional analysis ............................................................................................................ 31 Chapter 3: Results ................................................................................................................................ 32 3.1 Overview of analysis and main findings................................................................................... 32 3.2 Temporal- and processing- dependent variations in DNA methylation ................................... 32 viii  3.2.1 DNA methylation in PBMCs was largely stable throughout the circadian day ................................ 33 3.2.2 Fluctuation in DNA methylation during the course of a month ....................................................... 34 3.2.3 Processing delays led to significant change in DNA methylation levels .......................................... 34 3.2.4 DNA methylation stability was associated with genomic characteristics......................................... 35 3.3 Interindividual variability of BEC and PBMC ......................................................................... 44 3.3.1 PBMC and BEC had substantially different DNAm profiles ........................................................... 44 3.3.2 Variability of DNAm was largely tissue-specific ............................................................................. 46 3.3.3 Tissue-specific association between DNAm and demographic factors ............................................ 48 3.4 Genomic DNA methylation levels are sensitive to diesel exhaust-exposure............................ 57 3.4.1 Diesel exhaust was associated with changes in DNA methylation................................................... 57 3.4.2 Diesel exhaust-associated changes were found in genes relevant to allergic disease ....................... 60 3.4.3 Alu and LINE1 CpG sites showed methylation changes post diesel exhaust exposure ................... 61 3.4.4 MiR21 showed decrease in methylation upon diesel exhaust exposure ........................................... 62 Chapter 4: Discussion ........................................................................................................................... 72 4.1 Epigenomic characterization .................................................................................................... 73 4.1.1 DNA methylation is variable over the course of one month ............................................................ 73 4.1.2 DNA methylation is sensitive to delay in sample processing ........................................................... 74 4.1.3 PBMCs and BECs differ in certain key epigenetic features ............................................................. 76 4.2 Environmental characterization ................................................................................................ 80 Chapter 5: Conclusion and Future Directions ................................................................................... 86 Bibliography ................................................................................................................................ 88 Appendices ................................................................................................................................. 106  ix  List of Tables  Table 1 Demographic information  ............................................................................................. 106 Table 2 DAVID gene ontology enrichment for tissue-specific probes ....................................... 106 Table 3 Probes in PBMC found to display high confidence associations with gender and BMI 107 Table 4 Demographic information .............................................................................................. 110 Table 5 Top 4 DAVID functional annotation clusters in 2827 probes associated with diesel exhaust-exposure......................................................................................................................... 111 Table 6 CpG sites found to have significant decrease in methylation as a result of DE exposure through LME modeling............................................................................................................... 114  x  List of Figures Figure 1 Methylation measurements from the same individuals in the morning and evening showed no change in genomewide or probe-level methylation...................……………………37 Figure 2 Methylation measurements from the same individuals over the course of one month showed significant variability in specific CpG sites. .................................................................... 39 Figure 3 Genomewide methylation measurements changed in response to delayed sample processing, with particular CpG sites showing especially large fluctuations. .............................. 41 Figure 4 Methylation changes associated with the course of one month as well as processing time showed enrichment for hypomethylated and intermediate methylated regions. ........................... 43 Figure 5. BEC displayed overall lower DNAm than PBMC. ....................................................... 50 Figure 6 Correlation of DNAm was higher within than between tissues. .................................... 51 Figure 7 Tissue-specific DNAm was enriched in regions of low CpG density. ........................... 52 Figure 8 Variability in DNAm was tissue-specific. ...................................................................... 53 Figure 9 DNAm variability was tissue specific.. .......................................................................... 54 Figure 10. Relative levels of DNAm among individuals were not translatable across tissues. .... 55 Figure 11 Demographic factors were associated with DNAm changes in both PBMC and BEC.56 Figure 12 Percent variance within the dataset accounted for by the first 22 PCs. ........................ 63 Figure 13 Heatmap showing correlation among demographic variables and differential cell counts. ........................................................................................................................................... 64 Figure 14 PCs were associated with demographics and biological variables. .............................. 65 Figure 15 Diesel exhaust-associated exposure patterns were captured in PC 22 ......................... 66 Figure 16 Gene-specific methylation changes were found at CpG sites across the genome. ....... 68 xi  Figure 18 Mean beta value of exposure to FA at 0hr, 6hr, and 30hr for the 170 probes found to be significant for DE exposure but not for FA exposure. .................................................................. 69 Figure 19 DNA methylation of CpG sites overlapping with Alu and LINE1 repetitive elements was associated with diesel exhaust. .............................................................................................. 70 Figure 20 A CpG site residing in the miR-21 genomic locus changed in DNAm in response to DE. ................................................................................................................................................ 71    xii  List of Abbreviations  DNAm – DNA methylation EWAS – epigenome-wide association study PBMC – peripheral blood mononuclear cells BEC – bucal epithelial cells DE – diesel exhaust FA – filtered air CGI – CpG Islands DNMT – DNAm transferase enzyme HC – high CpG density IC – intermediate CpG density LC – low CpG density PM – particulate matter APEL – Air Pollution Exposure Laboratory ICC – intraclass correlation coefficient BP – biological process SD – standard deviation FDR – false discovery rate BMI – body mass index PC – principal components PCA – principal component analysis BAS – basophil LYM – lymphocytes  MON – monocytes Treg – regulatory T cell xiii  Acknowledgements  I offer my enduring gratitude to my program, my fellow students who have supported and accompanied me during my study. I owe particular thanks to Dr. Michael S. Kobor, whose scientific guidance and mentorship has made a significant impact on my education, and without whom this work would not have been possible. Thank you Dr. Christopher Carlsten for collaborative support and for guidance through my research on environmental impacts on DNA methylation. xiv  Dedication  I dedicate this work to all the schools, organizations and individuals that have guided me intellectually and morally, and supported me financially throughout my educational pursuit.1  Chapter 1: Introduction  1.1 DNA methylation 1.1.1 Epigenetics and the central dogma Epigenetics is the study of changes in gene expression during development and cellular differentiation that does not involve changes in DNA sequences. These changes are the sum of molecular-biochemical mechanisms that manifest as covalent attachment of chemical groups to DNA and histones (Shenderov & Midtvedt, 2014). The term epigenetics is broadly encompassing and refers to a variety of mechanisms that are important contributors to gene expression, including histone modifications, histone variants, RNA-based mechanisms and DNA methylation (DNAm) (Bonasio, Tu, & Reinberg, 2010). Throughout growth and development, epigenetic alterations are intrinsically programmed and occur at defined stages leading to tissue-specific expression of genes, or they could also occur sporadically due to environmental and lifestyle influences (Kota & Feil, 2010; Law & Jacobsen, 2010; Okano, Bell, Haber, & Li, 1999; Reik, 2007). In germ cells, the epigenome is reset to prepare for the next generation (Feil & Fraga, 2011; Shenderov & Midtvedt, 2014).  The two main forms of epigenetic regulation are histone modifications and DNAm. In the nucleus, chromatin is the higher order structure resulting from the tight packing of genetic information. The structural availability of the chromatin in combination with a host of covalently attached molecular marks is responsible for transcriptional control of the genome. Therefore, alterations to the structure or the covalent marks are key in changing the expression behavior of genes. It is known that post-translational modification of histone complexes are involved in 2  nucleosome arrangements, as well as the folding and unfolding of chromatin that is directly associated with the development of the organism (Bonasio, Tu, & Reinberg, 2010; Shenderov & Midtvedt, 2014). DNAm, on the other hand, is a mark that is inherited through cellular division as well as in the offsprings of the organisms. The methylation mark is established and maintained by the DNAm transferase enzymes (DNMTs), mainly DNMT1, DNMT3A, DNMT3B, and DNMT3L, the first of which has a high affinity for hemimethylated DNA, while the rest are involved in de novo methylation events (Denis, Ndlovu, & Fuks, 2011; Kota & Feil, 2010; Law & Jacobsen, 2010; Okano, Bell, Haber, & Li, 1999; Reik, 2007).  Advancements in epigenomics offers an alternative, albeit complex, way by which genetic activity is regulated. It is clear now that cellular homeostasis exists in complex feed-forward and feed-back cycles that involve programming and reprogramming of the epigenome during initial development and throughout life (Shenderov & Midtvedt, 2014). Factors that have been shown to impact the epigenome can be internal, such as microbiota, as well as external, such as lifestyle choices, upbringing and environmental stressors (Shenderov & Midtvedt, 2014). In the individual, epigenetic characteristics are different between tissues and developmental stages. While in the population, epigenetic characteristics differ among individuals and between different populations ( Fraser, Lam, Neumann, & Kobor, 2012; Heyn et al., 2013). Variations in DNAm have been observed between African, Asian and European, and this variability encompasses difference in drug response, susceptibility to pathogens, etc (Heyn, 2014).   Taken together, we now know that epigenetic programming could occur readily after fertilization, facilitating a crosstalk between genes and the environment. This knowledge about 3  epigenetics opens up new avenues for researchers to engage with human phenotypes, life experience, and as well as therapeutic interventions for disease onset and prognostics. Furthermore, it is important for researchers in disease and population demographics to understand the negative impact of controllable external conditions, so that they could influence policies and improve societal wellbeing.  1.1.2 The biology and functional importance of DNA methylation Among the many factors constituting the epigenome, the best studied is DNAm, which primarily refers to a methyl group covalently attached to the C5 positions on cytosine in the context of CpG dinucleotides (Illingworth & Bird, 2009).  DNAm is associated with gene expression, and carries out X-chromosome inactivation (Bjornsson, 2004). Given that there are 4 different bases in DNA and 16 possible pairwise combinations of these bases, CpGs should occur at a frequency of 6%, but in reality they are underrepresented and only occurs at 5-10% of the expected rate (Antequera & Bird, 1993). This underrepresentation is most likely due to the spontaneous deamination of methyl-cytosines to thymines. Nevertheless, CpGs are present on the majority of genomic elements, including gene promoters and bodies, and repetitive and transposable elements (Bird, 2002; Eckhardt et al., 2006; Weber et al., 2005). An important regulatory unit of DNAm is the CpG island (CGI); much of CpGs in the genome are present in that form. CGIs are around 0.5 to 5kb in length and are regions where there is an elevated C+G content (Das & Singal, 2004; Illingworth & Bird, 2009). CGIs can be found in the promoter of 60-70% of human genes including housekeeping and tissue specific genes, playing a role in tissue specificity and maintaining cellular homeostasis (Larsen, Gundersen, Lopez, & Prydz, 1992; Weber et al., 2005; J. Zhu, He, Hu, & Yu, 2008). Moreover, CGIs play a key role in the highly regulated event that 4  is gene transcription, and some experts believe that the co-localisation of CGIs to promoters may serve as landing signals for transcription factors; this conjecture came from the observation that despite the enrichment of transcription factor binding sites in the promoters, they are also ubiquitously present throughout the genome, thus presence of CGIs may serve as an additional signal for transcription factor to land on gene promoter binding sites (Illingworth & Bird, 2009). It is rare to have CGIs demonstrating high levels of methylation, and studies showed that only around 3-4% of CGI-promoters are highly methylated in healthy somatic tissues (Bird, 1995; Weber et al., 2005). Although overall findings seem to point to a role for CGI maintenance of gene expression such that higher methylation levels lead to gene repression, and lower methylation levels lead to expression, results from studies arrive at different conclusions. Some studies have found a low correlation between CGI methylation levels and actual gene expression levels, while other studies observed a general trend of association, where DNAm levels of expressed genes decrease at transcriptional start sites (Eckhardt et al., 2006; Fraser, Lam, Neumann, & Kobor, 2012 ; Illingworth & Bird, 2009; Lam et al., 2012; Y. Li et al., 2010; Volkmar et al., 2012).  With an understanding of the role DNAm plays in gene expression, it is clear that stability of DNAm is essential in maintaining normal cellular functions (Kulis & Esteller, 2010). Maintenance of the DNAm mark itself is attended to through addition and removal by many protein complexes (Feil & Fraga, 2011), including DNMTs 1, 1b, 1o, 1p, 2, 31, and 3b, as well as demethylases (Costello & Plass, 2001; Robertson, 2002). These protein complexes are responsible for the different levels of methylation across cell types, as well as the differences across individuals. 5   1.1.3 The dynamic nature of DNA methylation is associated with diseases and phenotypes Current approaches in investigations of biological roots of phenotypic traits appearing throughout the stages of life have heavily involved research in DNA sequence variants, and changes in gene expression (Bjornsson, 2004); this is especially true for studies on disease and cancer, in which malignancy is usually understood through mutation, gain, or loss of genetic information resulting in change in genetic activities and protein expression. However, we know that complex diseases often do not arise from a single gene mutation, thus it is wrong to assume that an understanding of genetics alone could explain the rise of complex diseases.  Compared to DNA sequence information, DNAm is incredibly malleable and it responds readily to internal changes and external stimuli (Mill & Heijmans, 2013). Specifically, there are three broad categories of influencers impacting epigenetic variations (Bjornsson, 2004). First there is parental influence, where events before and after conception could influence epigenetics (Cooney, Dave, & Wolff, 2002; Waterland & Jirtle, 2003). A second influencer is aging; studies have found age-related global hypomethylation of DNA and gene promoter silencing in most tissues, as well as age-related reactivation of the X chromosome (Bjornsson, 2004). Since the epigenome is critical in guarding the transcriptional status of genes, such age related attenuation of the epigenome could result in higher penetrance of diseases. Lastly there are environmental influencers. Environmental influences manifest as direct changes in response to stimuli, or as random stochastic events (Bjornsson, 2004). For example, toxin exposure experiments have shown DNAm to be sensitive to such exposures using both human and animal models 6  (Gluckman, Hanson, Buklijas, Low, & Beedle, 2009; Rosenfeld, 2010). As well, longitudinal cohorts study revealed that phenotypic differences between monozygotic twins are traceable to the DNAm changes that occurred after birth and throughout life (Wong et al., 2010). Other environmental chemicals and pollutants documented to affect DNAm include particulate air pollution, tobacco smoke, asbestos, ethal ions, silica, and benzene (Belinsky et al., 2002; Breitling, Yang, Korn, Burwinkel, & Brenner, 2011; Feil & Fraga, 2011).Such responsiveness of the epigenome to environmental cues hints at a complex method by which the static genome can by impacted, leading to gene expression changes and phenotypic outcomes.   By integrating DNAm into research and using it to complement mRNA expression findings, studies could seek to explain the late onset of diseases, the quantitative nature of complex traits, the effect environment has on disease progression, including modulating individual’s genetic predisposition to certain diseases, as well as how behavioral and social cues interact with the human body to elicit certain phenotypes (Feinberg 2004; Feinberg & Tycko 2004).  1.1.4 Epigenetic epidemiology and current technologies So far, it is established that to understand phenotypes as a function of genes and environment, we need to take into account both genetic and epigenetic variations. For example, the traditional model that investigates genetic variations alone could not explain why disease rate among monozygotic twins does not approach 100%. Thus an understanding of epigenetic variation, specifically when and how it changes, what variations are expected, and which could be a consequence of external influences is critical in gaining a thorough understanding of disease.  7  Rapid technological advancements allowing routine quantitative measurements of DNAm at multiple CpGs across the genome in a large number of subjects have facilitated the integration of DNAm in human population studies (Bock, 2012; Hatchwell & Greally, 2007). Although the only technology that could be currently considered to be “gold standard” for genomewide methylation assaying is the whole-genome bisulphite sequencing technique that interrogates around 28million CpG sites (Mill & Heijmans, 2013), this method is not efficient or affordable for epigenetic epidemiology studies that need to evaluate a large number of subjects at the same time. More readily adoptable options for human population studies are the single CpG site resolution Illumina platforms: 450k, 27k, and GoldenGate, each of which covers a unique number of selected CpG sites. The GoldenGate technology assays up to 1536 CpG sites from 807 genes, making it the platform with the least coverage out of the three. The 27k and 450k arrays allow targeting of 27,578 and 485,000 CpG loci, respectively, including those at high-value genetic regions and in methylation hotspots (Illumina Inc. 2010; Illumina Inc. 2010-2012; Illumina Inc. 2012).   As a result of this technological advancement, another aspect of epigenetics that is fast gaining relevancy is data quantity. Past research has taken a more targeted approach where selected loci are investigated for association with diseases. Currently, more studies are taking an alternative approach, where the Illumina technology is first used to maximize coverage, after which data analytics techniques are employed to identify sites or regions that are particularly sensitive or susceptible to influencers under investigation.  8  1.2 Normal variations in DNA methylation 1.2.1 Temporal stability of methylation mark over days and within a circadian day In the first part of the thesis we established that compared to DNA sequence information, DNAm is incredibly malleable and it responds readily to internal changes and external stimuli (Mill & Heijmans, 2013). In fact, it is this sensitive and dynamic nature of DNAm that makes it a key mechanism by which the static genome could be influenced, leading to changes in phenotypic traits. Because of this, DNAm has been the main focus of epigenetic investigations. However, the caveat is that DNAm could also be prone to stochastic variations that do not have any biological significance. Due to the nature of the covalent bond between a methyl group and cytosine, methylation studies are often conducted under the assumption that DNAm is stable. Therefore not enough knowledge exists on CpG loci that are stable versus those that may rapidly undergo changes in methylation levels. We believe that before conclusive findings could be gained from epigenetic studies, it is important to rigorously define the methodologies used in achieving the data. Furthermore, these considerations in combination with the high potential and growth of the field necessitates the existence of robust data resources similar to those that facilitate genetic studies. A critical component in building such data resources is characterizing the behavior of DNAm in healthy individuals over the course of multiple days, and even within a circadian day. Not much information currently exists on these fronts, thus we aim to take a step towards this direction of data standardization and building quality reference genomes. Another important component is method standardization, and in this thesis we aim to characterize the response of DNAm to delays in sample processing time.  9  1.2.2 Interindividual variability in DNA methylation in the context of tissue specificity Having established the importance of characterizing temporal variability, our next topic is interindividual variability in light of tissue specificity. Variation of DNAm between individuals is not as widely explored, yet it is of great importance for population epigenetics, as it is a prerequisite for any epigenetic association with either exposure or phenotype to be found.  It is reassuring that several studies have recently documented the existence of inter-individual DNAm differences within a given tissue (Bock, Walter, Paulsen, & Lengauer, 2008; Lam et al., 2012; Schneider et al., 2010; D. Zhang et al., 2010). Recently, the relevance of DNAm variability in particular to disease models has been demonstrated. One study showed that DNAm variability of uterine cervix cells differed between subjects who developed non-invasive cervical neoplasia versus those who were free of the disease (Teschendorff et al., 2012). Another study examined the relationship between DNAm variability in peripheral blood cells and obesity, and showed that at certain sites, DNAm exhibited higher variability in case than controls, and these sites enriched for genes associated with obesity and obesity related diseases (Xu et al., 2013). Finally, a study examining discordance in depression among monozygotic twins found differences in variance in affected versus unaffected twins (Byrne et al., 2013). Thus it appears that DNAm variability might be associated with disease risk and progression. Besides disease related variability, factors such as ethnicity, aging, environmental exposures, and genetic allelic variation all contribute to the epigenetic variation between individuals (Bjornsson et al., 2008; Feil & Fraga, 2011; Hatchwell & Greally, 2007; Heyn, 2014; Teschendorff, West, & Beck, 2013).  10  In part, tissue-specific DNAm and inter-individual variance is linked to distinct promoter structures and their epigenetic properties, particularly the presence of CGIs. CGIs are regions that harbor a higher than average density of CpG dinucleotides, which are otherwise underrepresented in the genome (Bjornsson et al., 2008; Hatchwell & Greally, 2007; Illingworth & Bird, 2009; P. A. Jones, 2012). Approximately 70% of human gene promoters are associated with CGIs, and methylation of CGIs has been shown to be highly correlated with gene expression levels (Bock, 2012; Laird, 2010; Larsen et al., 1992; Saxonov, Berg, & Brutlag, 2006). While several nuanced classifications for CGIs are in use, the one that provides the best enrichment discrimination separates CGIs into high CpG density islands (HCs), intermediate CpG density islands (ICs) and low CpG density islands (LCs) (Bjornsson, 2004; Hatchwell & Greally, 2007; Illingworth & Bird, 2009; Portela & Esteller, 2010; Rakyan, Down, Balding, & Beck, 2011; Weber et al., 2007). In somatic cells, DNAm patterns tend to be correlated with CpG density, as on average, HC regions have low levels of CpG methylation, while IC and LC regions have increasingly higher levels (Bjornsson et al., 2008; Eckhardt et al., 2006; Feil & Fraga, 2011; Hatchwell & Greally, 2007; Heyn & Esteller, 2012; Lam et al., 2012; Y. Liu et al., 2013; Mill & Heijmans, 2013; Weber et al., 2007). Epigenetic differences between tissues are primarily associated with DNAm differences in IC promoter regions, while reports are inconsistent as to which class is enriched for CpGs that vary between individuals (Byun et al., 2009; Das & Singal, 2004; Davies et al., 2012; Fernandez et al., 2012; Lam et al., 2012; D. Zhang et al., 2010; Ziller et al., 2013).   Currently, the vast majority of studies in the growing field of epigenetic epidemiology naturally rely on a limited number of easily accessible tissues such as blood and buccal epithelial cells 11  (BEC) due to their availability and non-invasive nature (Bock et al., 2008; Heijmans & Mill, 2012; Lam et al., 2012; Mill & Heijmans, 2013; Schneider et al., 2010; D. Zhang et al., 2010). The former often are further processed to obtain PBMC, which are an immunologically relevant fraction of lymphocytes that lacks multinucleated granulocytes. Importantly, PBMCs are derived from mesoderm whereas BEC are derived from ectoderm, thus representing two different germ layers and distinct developmental origins. PBMC is the tissue of choice for most of epigenetic epidemiology. Collection of blood is faster, less invasive, and more cost effective. Thus many studies are using PBMCs as a biomarker for association with pathogenicity and disease. There have been numerous documented association between blood DNAm and different types of cancers such as breast, bladder and colon (Cash et al., 2011; J. Y. Choi et al., 2009a; Pufulete et al., 2003; Terry, Delgado-Cruzata, Vin-Raviv, Wu, & Santella, 2011). PBMC DNAm has also been shown to be useful in experiments on neurological disorders such as Alzheimer’s Disease (Byun, Nordio, Coull, Tarantini, Hou, Bonzini, Apostoli, Bertazzi, & Baccarelli, 2012b; Chouliaras et al., 2010). Similarly, BECs is also widely accepted as a target tissue in the epigenetic community. Given the widespread use of either PBMCs or BECs, two fundamentally different tissues, for inferring population-level phenotypes, investigators have already assessed differential methylation at corresponding CpG sites with the two tissues (Lowe et al., 2013). It is now important to ask how these tissues compare in terms of variance, and to reconcile their DNAm variability in the context of tissue-specificity.   12  1.3 Environmentally associated variability of DNA methylation 1.3.1 Clinical phenotypes of environment particulate matter pollution Studies have shown that environmentally triggered phenotypes can have an epigenetic association, and this implication of epigenetics in complex phenotypes and disease is of great interest in the field. Environmental factors that trigger epigenetic changes include socio-economic aspects, pollution, stress, and personal habits such as smoking and diet. Although the precise mechanism by which environmentally elicited change in epigenetics is associated with phenotypic outcomes is difficult to establish, it is still important to investigate and document these relationships since humans are traversing within a sea of these chemicals every day.  One environmental issue that affect epigenetics with implications in global health is air pollution. Exposure to air pollutants is a serious public health concern that has been associated with cardiovascular and respiratory diseases (Brook et al., 2004; Peng et al., 2009; Zanobetti, Schwartz, & Dockery, 2000). Ambient air pollution has long been linked to health problems and disease exacerbations. There have been numerous recorded cases of disastrous air pollution throughout history, and one such example is the Long fog of 1952 from which led to thousands of deaths (Logan, 1953). Despite being composed of both gaseous and particulate components, it is thought that the latter is particularly detrimental to human health (Brook et al., 2004; Chow et al., 2006). Particulate matter (PM) present in air pollution is a heterogeneous mix of components varying in chemistry, origin, concentration and size (Brook et al., 2004; Chow et al., 2006). They contain organic chemicals, metals, soots, acids, soil or dust particles (Hou et al., 2014). It has been suggested that elemental components such as sulfur, calcium, aluminum and silicon contributes the most to PM toxicity. Because of the diversity of PM, they are categorized by 13  diameter in um, ranging from ultrafine (PM<0.1um), to fine (<2.5um), and to coarse (2.5-10um) (Brook, 2008).  Studies showed that PM is associated with both chronic and acute effects on health (Sava & Carlsten, 2012), including increased risk of lung cancer (Dockery et al., 1993; Lepeule, Laden, Dockery, & Schwartz, 2012; Vineis & Husgafvel-Pursiainen, 2005), exacerbation of asthma and its symptoms (Atkinson et al., 2001; Friedman, Powell, Hutwagner, Graham, & Teague, 2001; Guo et al., 1999; Klot et al., 2002), increased hospital admission rate for respiratory and cardiovascular diseases in both adults and children (Brook, 2008; Peng et al., 2009; Sava & Carlsten, 2012), as well as increased risk for mortality (Chow et al., 2006). Traffic-derived pollution appears particularly toxic, perhaps due to its abundance of PM (Hoffmann et al., 2007). Ambient PM is a heterogeneous mix of components varying in concentration, and chemistry (Sava & Carlsten, 2012). In urban environments, the major contributor to fine PM (diameter between 0.1μm and 2.5μm) is diesel exhaust (DE) (Laden, Neas, Dockery, & Schwartz, 2000). Due to its small size, fine PM can deposit deep in the lung, and its soluble components such as transition metals may cross the lung epithelium into systemic circulation and interact with internal organs (Brook et al., 2004).  1.3.2 Impact of particulate matter pollution on human DNA methylation Exposure to DE prompts the generation of reactive oxygen species, leading to oxidative stress and damage to cellular structures (Hiura, Kaszubowski, Li, & Nel, 1999; Kumagai et al., 1997; N. Li et al., 2002). On the molecular scale, DE has been found to change microRNA expression, increase production of allergic antibodies, up-regulate mRNA expression of pro-inflammatory 14  mediators and antioxidant enzymes, as well as decrease methylation of repetitive genomic elements (Al-Humadi et al., 2002; Baccarelli et al., 2009; Takizawa et al., 2000; Tsien, Diaz-Sanchez, Ma, & Saxon, 1997; Yamamoto et al., 2013). The hazardous effect of DE is also associated with asthma susceptibility and severity; for example, evidence links exposure to DE with decreased lung functions and increased airway resistance (Carlsten, MacNutt, Zhang, Sava, & Pui, 2014; Holloway et al., 2012). However, the precise mechanism by which pollution exacerbates asthma is not yet fully understood (Sava & Carlsten, 2012). It has also been hypothesized that exposure to DE may be partly responsible for the increase in allergic diseases in industrialized nations (Diaz-Sanchez, Proietti, & Polosa, 2003).  One possible mechanism through which air pollution impacts transcriptional pathways may be exposure-related epigenetic modifications. The effects of PM on DNAm have already been demonstrated in numerous studies. For example, it has been shown that exposure to PM ranging from 2.5-10um is associated with lowered Alu and LINE1 methylation, lowered methylation in proinflammatory genes (Baccarelli et al., 2009; Bollati et al., 2007; Madrigano et al., 2011; Tarantini et al., 2009), and decreased global methylation overall (Baccarelli et al., 2009; Bellavia et al., 2013; Bollati et al., 2007; Madrigano et al., 2011; Tarantini et al., 2009). Long terminal repeat elements are present in the genome as clusters of long and uninterrupted sequences (S. H. Choi et al., 2009b). LINE1 and Alu generally have higher methylation than the rest of the genome, and their methylation level is negatively correlated with mobility of retrotransposons (Y. Li et al., 2010). Increased mobility of these retrotransposons, especially at cancer-related genomic loci, is associated with mutation and higher tumorigenesis rates (Y. Li et al., 2010). In fact, it has been found that changes in cancer involved hypomethylation at repeated DNA 15  sequences (Ehrlich, 2002), leading to transcriptional activation of suppressed genes (Jürgens, Schmitz-Dräger, & Schulz, 1996). Thus higher DNAm at repetitive elements is associated with limiting retrotransposon mobility and stabilizing the genome. In fact, many of the recently studies surround how ambient PM change DNAm of LINE1 and Alu repetitive elements as well as that of pro-inflammatory and tumor suppressor genes (Bellavia et al., 2013; Hou et al., 2014; 2011; Tarantini et al., 2009). These studies have discovered that methylation at repeat elements is responsive to environmental influences: in individuals exposed to benzene, there is decrease in methylation at Alu and LINE1 (Bollati et al., 2007), while in individuals exposed to polycyclic aromatic hydrocarbons, there is increase in methylation at Alu and LINE1 (Pavanello et al., 2009). Changes in the methylation of Alu and LINE1 could affect chromosomal arrangements and gene expression that relates environment to disease.  Another epigenomic characteristic that would be insightful to explore is methylation of microRNAs. MicroRNAs are a family of small, non-coding RNAs, and they have important roles in regulation of gene expression for mRNA with matching sequences, controlling the mRNA copy number of certain genes and achieving post-transcriptional gene silencing (Ambros, 2003; Bartel, 2004; Lai, 2003). MiRNA regulates developmentally timed events, cellular differentiation, proliferation, and apoptosis (Ambros, 2003; Bartel, 2004; Carrington & Ambros, 2003; Johnston & Hobert, 2003; McManus, 2003; Yamamoto et al., 2013). And miRNA dysregulation has been implicated in numerous adverse phenotypes, such as lung diseases and inflammation (Holloway et al., 2012; Oglesby, McElvaney, & Greene, 2010; Small & Olson, 2011). Our interest in miRNA stemmed from the fact that they were also involved in airway inflammation (Lu, Munitz, & Rothenberg, 2009). In a previous study, we showed that DE-16  exposure changed the mRNA level of a few microRNAs. Thus it would be of interest to investigate whether methylation changes were also taking place behind that observation.  Given the significant impact on health that could be attributed to air pollution, it is important to understand the systemic impact of ambient pollutants on DNAm at sites across the genome and identify vulnerable genomic sites that are especially sensitive to exposures.  1.4 Rationale and thesis objectives Variations in DNAm is an emerging important topic in epigenetics; in fact, the International Human Epigenome Consortium is working on 1000 reference epigenomes to determine the degree of interindividual variations that exists in a host of different tissues (Abbott, 2010). But in order to thoroughly define epigenetic variations, we need to understand each source of variation, for it is only when this foundational knowledge is established, could be build robust epigenetic models on existing research evidence. In this thesis, we characterize DNAm under normal circumstances in healthy individuals by investigating temporal and individual variants, and we also investigate the response of DNAm to exposure to environmental compounds. We want to not only study the behavior of DNAm under these circumstances, but also to gain more information about the sensitive regions in the genome that are more susceptible to internal, external, and temporal factors. The results obtained would facilitate relating these variations in the epigenome to human population and disease models.   The work presented in this thesis is done on PBMC and BECs, both of which have been the tissues of choice for many epigenetic epidemiology studies due to easy collection, non-invasive 17  collection process, and cost effectiveness (Abouta et al., 2002; Bock et al., 2008; Heijman and Mill, 2012; Lam et al., 2012; Mill and Heijmann, 2013). These features combined make PBMC and BEC ideal for accommodating the large number of participants that is often seen in EWAS studies.   First we explore the stability of PBMC DNAm over the course of a circadian day and a month, as well as when subjected to delays in sample processing. Given that many population epigenetic studies repeatedly collect samples from the same individuals for longitudinal data, as well as carry out sample collection from large geographic areas, it is important to ensure that the observed methylation changes are directly or indirectly linked to the biological factors under study, instead of due to some spurious contribution from experimental methodologies, or the short- and long -term instability of DNAm. These insights would be valuable for they give investigators foresight in understanding which genes have naturally large variations over time, and would not be reliable indicators of change in longitudinal studies.   Another theme of variation characterized in this thesis is interindividual variance in DNAm of BECs and PBMCs tissues, both of which have been used extensively in epigenetic research (Baccarelli & Bollati, 2009; Baccarelli et al., 2010; Diaz-Sanchez et al., 2003; Kim et al., 2010; Mill et al., 2008; X. Zhu et al., 2011). We observe how these tissues compare in terms of variance, and we reconcile their DNAm variability with tissue specificity. Specifically, we compared matched PBMCs and BECs (obtained using cheek swabs) from a small community cohort of 25 healthy subjects at 998 CpG sites to examine how inter-individual variability differed across tissues.  18   Lastly, we investigate the sensitivity of DNAm to environmental PM pollution. Despite the numerous existing investigations regarding ambient PM and DNAm, no controlled investigation of PM on DNAm at sites distributed across the genome, an important starting point for unbiased mechanistic inquiry, has been reported. We were interested in understanding the systemic impact of air pollution from the perspective of DNAm, and thus focused on peripheral blood mononuclear cells (PBMCs). We hypothesized that short-term exposure to DE would lead to changes in DNAm status of PBMCs at CpG sites across the genome in asthmatic individuals, especially in genes relevant to the etiology of allergic diseases. Furthermore, we speculated that we would also observe changes in methylation of LINE1 and Alu repetitive elements, given that repetitive elements have been shown to be sensitive epigenetic indicators of environmental exposure (Baccarelli et al., 2009; Tarantini et al., 2009). Lastly, we investigated whether any methylation changes were linked to DE induced changes in microRNA expression that we have previously demonstrated in the same individuals tested here (Yamamoto et al., 2013).  19  Chapter 2: Material and Methods  2.1  Overview This thesis contains data from three separate studies. Both temporal and interindividual assays of DNAm variability involved cohorts of healthy subjects, while the environmental exposure assay involved asthmatic subjects and/or those with methacholine PC20 levels below 8mg/mL. The studies on temporal and interindividual variability were done using methylation assay technology Illumina GoldenGate, while the study on the effects of environmental exposure was done using the Illumina HumanMethylation450 BeadChip array (Illumina Inc., 2010-2012; Illumina Inc., 2012). All studies investigated DNAm changes by quantitative measurements of differentially methylated positions. Two distinct values of DNAm were calculated, beta-values and M-values. Beta-value has a range of 0 to 1 and approximately represents percent methylation. M-values are log transformation of beta-values, and are more statistically robust. Thus most statistical analyses were performed using M values while visualization and discussions were presented using beta-values.   2.2 Study design and data preprocessing 2.2.1 Illumina GoldenGate array experiments PBMC fractions were extracted from blood samples using density gradient centrifugation, after which standard techniques were used to extract genomic DNA as per previous studies. To measure DNAm, 750ng of DNA was treated using the EZ DNAm kit (Zymo Research, Orange, CA, US), which converts unmethylated cytosines to uracil. DNAm at 1,536 CpG loci was measured using Illumina GoldenGate Assay for DNAm profiling, as described in previous 20  studies (Bird, 1980; Yuen et al., 2009; 2011). Briefly, for each CpG site there were two allele-specific probes annealing to either the methylated or the unmethylated sequence. Annealed probes are then paired to locus-specific oligos, after which the ligated products are amplified using PCR. DNAm is evaluated as a value between 0 – 1, otherwise known as beta values, calculated from the intensity of the methylated and the unmethylated alleles (Lam et al., 2012; Weber et al., 2007; Yuen et al., 2011).  2.2.2 Illumina Infinium HumanMethylation450 array experiment All procedures conducted using commercially available kits were done so following the manufacturers’ protocol. Two μg of genomic DNA per sample was extracted from PBMCs using the DNeasy Blood & Tissue Kit (Qiagen, Valencia, CA, USA). One μg of the purified DNA was then bisulfite-converted using the EZ-DNA methylation kit (Zymo Research, Orange, CA, USA), which changed epigenetic data into sequence-based data by the selective conversion of unmethylated cytosines to thymidines. Bisulfite-converted DNA was assessed for concentration and quality using the NanoDrop, and 4μL of the conversion product was used for genome-wide DNAm evaluation at over 485,000 CpG sites using the Illumina Infinium HumanMethylation450 BeadChip array in house as described before (M. J. Jones et al., 2013).  2.2.3 DNA methylation temporal stability and experimental influences – study design In the first experiment, we looked at the temporal variability of PBMC DNAm among a cohort of 5 adult women in the Greater Vancouver Area. To test for short temporal effects, samples were collected from the same five subjects in the morning and evening of the same day, and then processed and analyzed immediately. To test for longer temporal effects, the samples were 21  collected from the same five subjects in the morning over a 4-week span at 0, 14, and 28 days, and processed and analyzed immediately. To test for effects of delayed processing time, samples were collected in the morning from the same five individuals, and then processed 0hr, 4hr, or 24hr after sample collection.  2.2.4 DNA methylation temporal stability and experimental influences – data preprocessing Assessment of DNAm using the GoldenGate array was conducted using the steps outlined in section 2.2.1. All samples were assayed in a single array run, providing confidence in their comparability. It is important to note that our analysis did not differentiate between DNAm and hydroxymethylation. DNA hydroxymethylation has been hypothesized to have a functional role outside of DNA demethylation, but as it is present at significant levels only in neural and pluripotent cells, this function is not likely to be important in non-neural tissue (Baccarelli et al., 2010; Lister et al., 2013; 2009).  Prior to analysis, DNA methylation data was subjected to stringent quality control. First, probes for SNPs sites were removed, along with polymorphic CpGs that could potentially interfere with analysis (Byun et al., 2009; Heyn & Esteller, 2012). Then, probes for which more than three samples had signals below background detection levels were also removed, leaving 855 CpGs that were used in further analysis. All analysis was performed using the R statistical computing software (R Core Team, 2012).  22  2.2.5 Interindividual variability in DNA methylation – study design The next experiment examined interindividual variability of DNAm among a cohort of 25 adults. Matching buccal and blood samples were collected from 25 adults in the Greater Vancouver Area. The PBMC fraction was extracted from blood using density gradient centrifugation, and for both BEC and PBMC, genomic DNA was extracted using standard techniques. Assessment of DNAm values was carried out as described in section 2.2.1.  2.2.6 Interindividual variability in DNA methylation – data preprocessing Data quality control was performed as follows. First, we removed all probes for SNPs and X-chromosome sites. Next, CpG sites where >23 subjects had beta values of zero, as well as sites where >3 subjects had signals below background detection were removed. Lastly, we removed probes with polymorphic CpGs that may interfere with the analysis (Byun et al., 2009; Heyn & Esteller, 2012). This left 998 CpG loci, and 25 subjects ranging from 26 to 45 years in age (10 male and 15 female). Methylation measurements of zero were replaced with the minimum measurement for either PBMC (0.033) or BEC (0.03). All analysis was performed using the R statistical computing software (http://www.r-project.org), and statistical tests were conducted with alpha level of 0.05. Data used in this study can be found in the Gene Expression Omnibus repository with the accession number GSE53396.  2.2.7 Environmental effects on DNA methylation – exposure design Sixteen participants were recruited at the Air Pollution Exposure Laboratory (APEL) in Vancouver, British Columbia, Canada. Written consent was obtained from all subjects, and the protocol was approved by the institutional review board for human studies at the University of 23  British Columbia (H08-02288). Participants were 19-to 35-year-old nonsmokers who had had physician-diagnosed asthma for at least 1 year and/ or a methacholine challenge with ≤ 8 mg/mL in terms of the PC20, a provocative concentration of methacholine that induces a 20% fall in forced expiratory volume in 1 sec (FEV1). All participants were stable in terms of asthma symptoms [assessed by the asthma control questionnaire] and were free of respiratory infections for 4 weeks prior to and during the study period (Juniper, O'Byrne, Guyatt, Ferrie, & King, 1999). The participants were free from current use of inhaled corticosteroids, regular use of bronchodilator, and use of vitamin A, C, or E supplements. Throughout the study, participants were asked to withhold long-acting β2-agonists 48 hr prior to spirometry, short-acting β2-agonists 8 hr prior to spirometry, and caffeine 4 hr prior to methacholine challenges. The participants maintained a stable diet, including intake of cruciferous vegetables, over the course of the study; thus, in the context of the crossover design, diet was not considered confounding. Details of the exposure were previously documented (Birger et al., 2011). This study followed a double-blind, crossover design in which each subject was subjected to either filtered air (FA) or DE (300ug/m3 PM2.5) for 2 hours on two separate occasions at least two weeks apart. The sequence of FA or DE exposure was randomized and counterbalanced, and exposures took place within the same season for a given subject. During exposures, subjects alternated between light exercise (15min) and rest (45min) on a stationary bicycle. The wattage of the stationary bike was calibrated in each individual to a minute ventilation of 15 L·min-1·m² body surface area. Blood was collected in EDTA and Vacutainer® CPT™ tubes (BD Biosciences, Franklin Lakes, NJ) immediately before exposure as well as 6 and 30 hours after exposure onset and processed within 4 hours. PBMC separation was performed following the protocol suggested by the manufacturer, after which PBMCs were stored at -80°C for analysis. Each subject was assessed 24  for GSTM1 gene deletion, GSTP1 rs1695 SNP and NFkB rs28362491 SNP genotypes by PCR-Restriction Fragment Length Polymorphism using DNA isolated from whole blood. Assessment of DNAm was carried out as described in Section 2.1.2.  2.2.8 Environmental effects on DNA methylation – data preprocessing Illumina GenomeStudio software was used to interpret array output, and signalA, signal, and probe intensity were exported into R for further processing and analysis. For each sample, probes with one or more samples that had undesirable detection p-values (p-value>0.01) or with one or more missing measurements were removed. Then probes residing on the X or Y chromosome were also removed to control for gender-derived differences in the array. Lastly, probes that cross-hybridize to somatic sites or to sites on the X or Y chromosomes, as well as probes that possibly reside at SNP sites (as defined by Illumina annotation or independent re-annotation) were removed, regardless of their allele frequency in the population, leaving 363340 probes and 96 samples from 16 subjects for analysis (Price et al., 2013). Such stringent filtration method was applied to ensure that methylation measurements of the CpG sites investigated in this study were most representative of the larger population.   Chip to chip color bias correction was performed in R using the built-in color-correction and background subtraction, and quantile normalization functions of the lumi package with default settings. Finally, we applied peak-based correction method which has been reported to improve data accuracy and reproducibility. Furthermore, peak based correction improves detection rates of differentially methylation at CpG sites that would otherwise have been missed 25  (Dedeurwaerder et al., 2011; Du, Kibbe, & Lin, 2008). The data obtained in this study has been deposited in the Gene Expression Omnibus repository under the accession number GSE56553.  2.3 Statistical analysis 2.3.1 Differential and tissue specific DNA methylation A probe-wise paired two-sample t-test (n = 25, two tailed) not assuming equal variance was used to compare the equality of mean probe methylation across BEC and PBMC. False discovery rate (FDR) was adjusted using Benjamini Hochberg correction, after which adjusted p-value of 0.001 was used as the significance level. Mean beta values were obtained for the probes in each tissue by averaging across the 25 individuals. Cross-tissue ∆β difference was calculated on a probe-wise basis by calculating the numerical difference between the average beta values.  Probes with significant differences in beta values between the two tissues were identified using a combination of t-test results and ∆β values. Tissue specific probes were defined as those with t-test p-value smaller than 0.001 and having an absolute ∆β of larger than 0.2. Highly tissue specific probes are defined as those with t-test p-value smaller than 0.001 and absolute ∆β larger than 0.4.  2.3.2 Intraclass correlation CpG sites were separated into three categories: 1. High CpG density (HC, CG content >55%, Obs/Exp CpG ratio >0.75 and length >500 bps) probes; 2. Intermediate CpG density (IC, CG content >50%, Obs/Exp CpG ratio >0.48 and length >200 bps) probes; and 3. Low CpG density (LC, non-islands or low-density CpG regions, non-HC/IC regions) probes (Weber et al., 2007). Probe-wise Intraclass correlation coefficient (ICC) was calculated using the one-way consistency 26  model in the irr R package. Calculation of ICC was done as described in equation (1). For probe i,  (1) Where MSwithin is the mean square value for within tissue difference, and MSacross is the mean square value for across tissue difference. As ICCi approaches +1, the within tissue difference is larger than the between-tissue difference for probe i; and as ICCi approaches -1, the between-tissue difference is larger than the within-tissue difference. Results from ICC offered insight into CpG regions that were more likely to display methylation difference across tissues.  2.3.3 Correlation between tissues Pearson’s two-tailed pairwise correlation was used to assess the comparability of inter-individual DNAm levels across BEC and PBMC. A Pearson’s correlation coefficient was calculated for each CpG site by correlating their methylation measurements for each of the individuals in BEC to their measurements in PBMC.  2.3.4 Concordance in probe variance across tissues DNAm range was visualized on a probewise basis using mean normalization. Range values were calculated as described in equation (2). For individual j of probe i,   (2) Due to truncation of variance at the extremes of DNAm, a beta value transformation was applied (Talens et al., 2010; Yuen et al., 2009; 2011). The transformation was achieved as described in equation (3). For individual j of probe i, ICCi = (MSwithin -MSacross )(MSwithin +MSacross )bnorm, j,i = b j,i -bi27   (3) After transformation, variance was also calculated for each probe and SD was derived by taking the square root value of variance. A levene’s test for equality of variance was conducted using the lawstat package in R using transformed BEC and PBMC DNAm values, and tissue as the grouping factor. P-values from the test were adjusted for FDR using Benjamini Hochberg correction, and an adjust p-value of 0.05 was used as the cut-off for significance.   2.3.5 DNA methylation and demographics Lastly, association between DNAm and demographic factors gender (15 male vs 10 female) and race (18 Caucasian versus 7 non-Caucasian) was investigated using the two-tailed Wilcoxon rank sum test for categorical variables. Multiple testing adjustments were done using the Storey’s q value method in the qvalue Bioconductor package. A q value of 0.05 was defined as the significance threshold. For risk factors BMI and alcohol consumption, association with DNAm was tested using a two-tailed Pearson’s correlation. Multiple testing adjustments were again carried out using the q value method, and probes with significant association to risk factors were defined as those with q value<0.05.  2.3.6 Bootstrapping and resampling In examining the temporal stability of DNAm, the bootstrap approach was consistently used to account for the small sample size, thus avoiding biased standard error estimations.  btransformed,i, j = tan-1bi, j1-bi, j( )éëêêùûúú2ìíïîïüýïþï28 Bootstrapped t-test A paired bootstrapped two-sample t-test was used to test for differential methylation between AM and PM measurements. For CpG locus i, the average difference between the means of AM and PM measurements was calculated. Then the data was randomly sampled n-times (n = 1000) with replacement, and the randomly sampled data was separated into AM versus PM groups. The p-value was calculated as the ratio of the number of resampled mean difference that is larger than or equal to the mean difference of the observed data divided by (n+1). Bootstrapped ANOVA A bootstrapped method similar to the bootstrapped t-test was used to test for the differences in methylation measurements over the course of a month. For CpG locus i, the ANOVA F-ratio was first obtained for the variation among the three days. Then the data was resampled n-times (n = 1000) with replacement, and the randomly sampled data was separated into three groups. The p-value was obtained from the ratio of the number of resampled F value that is greater than or equal to the F value of the observed data divided by (n+1). Linear mixed effects model bootstrap Linear mixed effects model was applied to the processing time delay data; this model incorporates fixed effects and random effects, the former represents changes in the population, while the latter represents changes to the individuals within the population. The bootstrap method reported here is a random effect coupled with residual bootstrap that has been previously documented (Thai, Mentré, Holford, Veyrat-Follet, & Comets, 2013). For each CpG locus, the mixed effects model was fitted onto the observed dataset, and then the random effect and 29  residuals were resampled from the distribution of the random effects and residuals of the simulated dataset. We drew n = 1000 bootstrapped datasets for each CpG locus, and the bootstrapped p-value is the number of resampled p-value that is smaller than or equal to the p-value of the observed data divided by (n+1).  2.3.7 Principal component analysis To examine impact of environmental exposure on DNAm, we made use of a principal component analysis (PCA) assisted initial filtering in which PCA was first used to identify the probes that were most likely to be impacted by DE-exposure. Mean normalization was conducted across samples, and the prcomp function from the R stats package was used to perform PCA on the dataset. The 0th principal component (PC) accounted for the difference in global probe intensity from one probe to the next and thus was not considered. To reduce the number of random variables under consideration, tests for association between experimental variables and PCs were carried out. Specifically ANOVA was used for nominal variables (time, subject, ethnicity, chip number, location on chip), spearman correlation was used for continuous variables (Height in cm, mass in kg, BMI, FEV1, age), and Wilcoxon-ranked sum test was used for dichotomous variables (FA vs. DE, DE hr6 vs. non-DEhr6, DEhr30 vs. non-DEhr30, DEhr6&30 vs. non-DEh6&30, sex, atopy, asthma, methylation, NFKB1 genotype, GSTP1 genotype, GSTM1 genotype). To control for type II error, Storey’s qvalue method was used with false discovery cutoff of 0.1 (high confidence) (Dabney, Storey, & Warnes).  30  2.3.8 Identification of Alu, LINE1, and microRNA associated CpG sites To determine if methylation of repetitive elements in the genome were sensitive to exposure to DE, we identified and analyzed probes overlapping with repetitive elements. To identify probes overlapping with Alu and LINE1 repetitive elements, the repeat element track was downloaded from the University of California Santa Cruz (UCSC) Genome Browser (http://genome.ucsc.edu), and probes having at least 15 base pairs (bp) overlap with Alu or LINE1 elements in the genome were identified; given that that the average Illumina probe is 50bp in length, this coverage would ensure that at least 30% of the probes resides within the repetitive elements. Furthermore, after testing multiple cutoffs, we found that a cutoff of 15bp gave the desirable number of Alu or LINE1 associated probes to represent coverage of Alu and LINE1 elements across the genome. A total of 7 probes overlapping with microRNAs miR-21, miR-30e, miR-215 and miR-144 were also identified and used in subsequent analysis.  2.3.9 Linear regression modeling The following linear regression model was applied to the 11378 lymphocyte and monocyte-count associated CpG sites identified by principal component analysis.  This model was used to identify sites that were significantly associated with blood cell counts. Yij represented the measured methylation value of individual i at CpG site j. β0 represented the overall intercept, β1 represented the effect of lymphocyte count on methylation value, β2 represented the effect of monocyte count, and εij represented the error term.     0 1 2β β  β İij ijY Lymphocyte Monocyte   31  2.3.10 Linear mixed effects modeling The following linear mixed effects model was applied to PCA identified DE-associated CpG sites, as well as sites overlapping with Alu, LINE1, and microRNA sites of interest using the lme function from the nlme package (Pinheiro et al., 2014). Since PC22 was associated with only diesel exposure, and it did not show confounding with any covariates, only exposure associated variables needed to be included in this model.  The model was used to test for change in methylation due to diesel exhaust exposure, assuming random intercept according to subject. Yijkl represented the measured methylation value of individual i, at CpG site j, exposure k and time l. β0 represented the overall intercept, μi represented the random intercept of individual i, β1 represented the main effects of exposure, β2 represented the main effects of time (pre-exposure 0 hr vs. post exposure 6&30 hr), and β3 represented the interaction effect between exposure and time, and lastly, εijkl represented the error term. This model was used to identify probes that changed in methylation as a result of exposure to DE but not FA.  2.3.11 DAVID functional analysis UCSC refgene accession-IDs corresponding to all probes tested in either the 450K or the GoldenGate platform involved in this analysis was used as background for DAVID GO analysis (http://david.abcc.ncifcrf.gov/) (Huang, Sherman, & Lempicki, 2009). CpG lists of interest were tested for enrichment of GO subcategories biological process and cellular component. Functional enrichment scores of larger than 1.3 were considered significant.       0 1 2 3β μ β  β β İijkl i ijklY Exposure Time Exposure Time      32  Chapter 3: Results 3.1 Overview of analysis and main findings This thesis characterizes temporal and interindividual DNAm variability, as well as variability in response to environmental compounds. When testing the temporal variability of DNAm in PBMC, and discovered that short-term variability over the course of one circadian day was almost non-existent; however, when that time period was stretched over one month, we found a number of CpG sites demonstrating sporadic changes in methylation. From the same experiment, we also found that delayed processing time directly impact the quality of DNAm, with decrease in PBMC methylation at many CpG sites detected after a 24 hour processing delay. The longer the incubation period, the larger the number of sites that showed a decrease in methylation. When we examined PBMC methylation in concert with BEC methylation, we found that besides tissue differential methylation, BEC and PBMC also differed vastly in terms of variability, and variability in one tissue did not predict variability in the other tissue. Furthermore, the level of inter-individual correlation between the two tissues was low. Lastly, investigation into environmental impacts on DNAm revealed that DNAm levels are sensitive to ambient PM pollution. There were changes in methylation at both LINE1 and Alu repetitive element sites, in response to DE exposure, as well as at CpG sites scattered across the genome. Furthermore, the majority of these changes were present as methylation decreases.  3.2 Temporal- and processing- dependent variations in DNA methylation DNAm of 1,536 independent CpG sites was quantitatively measured with the Illumina GoldenGate assay in PBMC DNA from five female volunteers.  We focused on females due to the natural hormonal changes happening over monthly timeframes.  After removing CpG sites 33  associated with single nucleotide polymorphisms, annealed to more than one gene locus, or had low detection p-values across the samples, DNAm levels of 855 CpG sites distributed over 565 genes were used as the basis for all subsequent analysis. For each subject, DNAm was analyzed for variation over the course of one day, one month, as well as changes associated with 24-hour processing delay.  3.2.1 DNA methylation in PBMCs was largely stable throughout the circadian day Blood samples were collected in the morning (AM) and evening (PM) from the same five subjects on the same day, after which the whole blood samples were processed immediately to extract PBMCs.  Comparing the resulting DNAm profiles, no significant changes were found in the total mean DNAm between AM and PM samples; Wilcoxon ranked sum test p-value>0.05 (Figure 1a). We next asked whether there were any differences across subjects between AM and PM at individual CpG loci.  This analysis showed that 90% of all CpG loci had absolute differences in beta value (∆β) of less than 0.096, and 70% had absolute differences of less than 0.048 (Figure 1b).  To determine if any CpG sites with relatively large ∆β could be substantiated statistically, a bootstrapped t-test was applied.  Only 2 specific CpG sites that differed significantly between morning and evening were identified (bootstrapped p-value<0.05, ∆β>0.05) (Figure 1c). The absence of skewing towards lower values in the bootstrapped p-value distribution further supported the lack of significant changes in DNAm between AM and PM samples (Figure 1d). These data suggest that DNAm remained largely stable during the day for the set of CpG sites tested.  34  3.2.2 Fluctuation in DNA methylation during the course of a month Given the relative stability of the DNA methylome throughout the day, we next examined whether DNAm fluctuated over the course of a month. Three morning draws of whole blood at two-week intervals on the same five individuals were carried out (day 0, 14, and 28).  ANOVA between the DNAm values for the three day-points revealed that mean DNAm levels did not fluctuate significantly over the course of one month (p-value > 0.05) (Figure 2a).  However, a large number of specific CpG sites displaying extensive variability among the three days were identified. From the recorded ∆β values for the three comparisons (d0 vs. d14, d0 vs. d28, day14 vs. d28), 138 CpG sites with ∆β changes >0.05 were found (Figure 2b). To identify CpG sites with statistically significant changes in methylation, a bootstrapped ANOVA was applied to the data. The balanced ANOVA p-value distribution confirmed that there were no temporal associated changes in DNAm on the genomic scale (Figure 2c). However, 58 CpG sites, equaling approximately 7% of all loci included in the study, had statistically significant changes in methylation over the course of one month, out of which 25 also had ∆β change of larger than 0.05 among one or more of the two-day comparisons (Figure 2d).  3.2.3 Processing delays led to significant change in DNA methylation levels Lastly, we assessed the effects of delays in blood processing. Blood samples were incubated undisturbed at room temperature for 0hr, 4hr, or 24hr before DNA for methylation analysis was purified. Although not substantiated by ANOVA (p-value=0.09), visual inspection of subject mean methylation profiles at each of the three time points suggested that there was a slight decrease in DNAm levels as processing was delayed (Figure 3a).  More strikingly, the overall distributions of observed changes in DNAm in relation to processing times demonstrated a shift 35  towards loss of methylation for the 4hr vs. 0hr comparison, which was even more pronounced in the 24hr vs. 0hr comparison.  More specifically, we found 121 and 173 probes with larger than 5% change in methylation from 0hr to 4hr, and from 0hr to 24hr respectively (Figure 3b). In order to identify the CpG sites driving this decrease in methylation, a probewise bootstrapped linear mixed effects model was applied to the data, from which the p-value distribution showed extensive skewing towards zero (Figure 3c). This deviation from a random distribution is indicative of association between DNAm and delayed processing.  Specifically, 76 CpG sites were identified with bootstrapped p-value less than 0.05, 52 of which also had methylation change of larger than 5% in one or more of the comparisons (0hr-4hr, 0hr-24hr) (Figure 3d). Most notably, 48 out of 52 sites had decreased methylation over the 24hr period, as illustrated by crowding at the left upper cell in the volcano blot (Figure 3d).   When comparing the overall magnitude of monthly variation versus blood processing, we noted the more pronounced skewing of the p value distribution associated with blood processing and also found a big difference in the number of statistically relevant CpGs with more than 10% methylation change.  3.2.4 DNA methylation stability was associated with genomic characteristics We next tested whether the CpG density of the region in which a given loci resided was predictive of genomic characteristics.  Taking into account the overall distribution of all probes on the array, surprisingly we did not find any enrichment for LC, IC, and HC regions among the variable probes.  However, CpG loci fluctuating during the month and those sensitive to processing delays were mostly hypo and intermediately methylated, in stark contrast to the overall bimodal distribution of DNAm of all 855 CpGs (Figure 4a and b).  This clear visual impression was statistically substantiated by bootstrapped kolmogorov-smirnov test p-values of 36  less than 0.05.  Lastly, we assessed the potential impact that weekly fluctuations and delays in blood processing might have on overall variability of DNAm in the larger cohort used for analysis of interindividual variability in the next step. The lists of day and processing CpG hits were compared to a list of most variable loci in PBMCs (based on GoldenGate array), as it was possible that the majority of variation observed did not reflect true biological variation but was rather an artifact of our collection and processing methods. We found that 1 of the day-point variable loci and 6 of the processing-time variable loci overlapped with the 25 most variable PBMC loci.                37                     Figure 1 Methylation measurements from the same individuals in the morning and evening showed no change in genomewide or probe-level methylation. (A) Subject mean methylation levels across 855 probes in the AM and PM. (B) Distribution of probewise Δβ values. For each probe, the mean methylation value across the five individuals was obtained for morning and evening measurements, after which the PM mean was subtracted from the AM mean to obtain the Δβ measurement. The dashed vertical line represents the mean Δβ for the distribution. (C) Volcano-plot of log10 p-values vs. Δβ measurements for all probes. P-values were obtained from 38  a bootstrapped t-test between AM and PM samples. The dashed lines represent significance cutoffs, with the horizontal line representing p-value=0.05, and the vertical lines representing Δβ ± 5%. (D) Distribution of p-values from the bootstrapped t-test. Skewing of the distribution to the right indicates lack of association between time of day and changes in DNAm.   39                       Figure 2 Methylation measurements from the same individuals over the course of one month showed significant variability in specific CpG sites. (A) Subject mean methylation levels across all probes for day 0, 14, and 28. The graph shows no visually discernible fluctuation in methylation value among the three day-points. (B) Distribution of Δβ values for all probes. The 40  mean methylation values across the five individuals were obtained for each probe, after which subtractions among the three day-points were performed to obtain the Δβ measurements. The dashed vertical lines represent mean Δβ for each distribution. (C) Distribution of p-values from the bootstrapped ANOVA. The lack of skewing within the distribution suggested an absence of association between DNAm levels and separate day-point measurements. (D) Volcano plot of log10 p-values vs. delta beta measurements for the comparisons among the three day-points. P-values were obtained from a bootstrapped ANOVA. The dashed lines represent significance cutoffs (p-value=0.05, Δβ ± 5%). A large number of probes passed these cutoffs.   41                        Figure 3 Genomewide methylation measurements changed in response to delayed sample processing, with particular CpG sites showing especially large fluctuations. (A) Subject mean methylation levels across all probes for samples immediately processed (0hr), and for samples kept at room temperature for 4hr or 24hr before processing. Overall decrease in methylation could be seen for each of the five subjects. (B) Distribution of Δβ values for all probes. The 42  mean methylation values across the five individuals were obtained for each probe, after which subtractions of the 4hr and 24hr means from the 0hr mean were performed to obtain Δβ measurements. The dashed vertical lines represent the mean Δβ for each comparison. The Δβ measurements from the 24hr vs. 0hr comparison were more negative than those from the 4hr vs. 0hr comparison. (C) Distribution of p-values from the bootstrapped mixed effects model. The left-skewing distribution demonstrates association between DNAm levels and delayed sample processing time. (D) Volcano plot of log10 p-values vs. delta beta measurements for two time-point comparisons. P-values were obtained from a bootstrapped mixed effects model allowing intercept to vary according to the individual. The dashed lines represent significance cutoffs (p-value=0.05, Δβ ± 5%). Both plots demonstrated high number of probes with negative Δβ values, with the 24hr vs. 0hr comparison showing a higher concentration of negative Δβ values.   43               Figure 4 Methylation changes associated with the course of one month as well as processing time showed enrichment for hypomethylated and intermediate methylated regions. (A) Density distribution of average beta values for 25 hit probes showing significant changes over the course of one month. The three day-points are separated and distinguished by color. (B) Density distribution of average beta values for the 25 day-hit probes versus that for all 855 probes. (C) Density distribution of average beta values for 52 hit probes showing significant changes in methylation in response to delayed processing time. The three time-points are separated and distinguished by color. (D) Density distribution of average beta values for the 52 processing-time hit probes versus that for all 855 probes.   44  3.3 Interindividual variability of BEC and PBMC To investigate interindividual variability, we recruited 25 healthy adults from the Greater Vancouver area to our study cohort, and collected BEC and PBMC from each individual (Table 1). This design allowed for a rigorous matched comparison between the two tissues, eliminating potential confounds due to inter-individual differences in environments or genetic background. DNAm of 1,506 promoter-associated CpGs biased towards cancer genes was quantitatively determined using an array-based approach (Bibikova, 2006; Byrne et al., 2013; Byun et al., 2009). After removing CpGs located on the sex chromosomes to avoid biases due to our cohort consisting of males and females and CpGs overlapping with SNPs or annealing to multiple regions in the genome, as well as 28 CpG sites where more than 3 samples had bad detection p-values, our final dataset consisted of 998 CpGs (Byrne et al., 2013; Byun et al., 2009)  3.3.1 PBMC and BEC had substantially different DNAm profiles We first interrogated broad tissue differences in DNAm across the population. Each of the 25 individuals had higher mean PBMC DNAm than BEC DNAm, and this cross-tissue difference was statistically significant by a paired Wilcoxon ranked sum test p-value of 3.16e-14 (n = 25, two tailed) (Figure 5A). While both tissues had the expected bimodal distribution of DNAm for each single CpG site across all subjects, there were noticeable differences in the specific fractions. Categorizing mean CpG methylation according to a published categorization schema, we found 62.4% of sites in BEC and 56.9% of sites in PBMC were hypomethylated (<20%), 16.7% in BEC and 20.3% in PBMC were heterogeneously methylated (20-80%), and 20.8% in BEC and 22.7% in PBMC were hypermethylated (>80%) (Figure 5B) (Y. Li et al., 2010).  45  To determine whether differences in DNAm between individuals or between tissues contribute more to probe-wise variability, we next compared between-individual and between-tissue correlations. Across the 998 CpG sites, the average two-tailed Pearson’s correlation coefficient of DNAm profiles among individuals was 0.980 for BEC and 0.987 for PBMC, and the coefficient between the two tissues was 0.773 (Figure 6A). Separating all probes into 212 LCs, 136 ICs, 503 HCs, and 147 uncategorized sites, we found that the relatively low between-tissue correlation was driven by LCs sites, which had a mean correlation of 0.514 between PBMCs and BEC; on the other hand, HC and IC had mean correlations of 0.748 and 0.757, respectively, and uncategorized sites had a higher correlation of 0.795 (Figure 6B). To explore tissue-specific DNAm in more detail, we identified probes that were differentially methylated between BEC and PBMC using a paired two sample t-test (n = 25, two tailed) followed by Benjamini Hochberg FDR correction (Benjamini & Hochberg, 1995). With this approach, we found that 537 CpGs (53.8%) were significantly different between PBMC and BEC with p-value < 0.001 (Figure 7A). Of these significant sites, 178 (17.8%) had a difference in average beta value (∆β) > 0.2 across the two tissues, and 121 of the 178 (12.1%) had ∆β > 0.4 (Figure 7A). Since the collection of 178 probes passed statistical threshold for significance, and was of reasonable size, we conducted functional enrichment analysis on the 178 highly tissue specific probes (located in 137 unique genes) using the online bioinformatics tool DAVID to determine if differentially methylated sites were representative of cellular processes specific to BEC or PBMCs (Huang et al., 2009; Weber et al., 2007). Using the recommended cutoff of enrichment score ≥ 1.3, two clusters were considerably enriched (Table 2). Cluster 1 contained biological process (BP) GO term immune system process (GO:0002376) accounting for 30 genes, as well as molecular function GO term cytokine activity (GO:0005125) accounting for 11 genes. Cluster 2 contained cellular component 46  GO term extracellular region (GO:0005576) accounting for 48 genes. Both cluster 1 and cluster 2 seemed to be enriching for PBMC and BEC specific processes, respectively, suggesting a reassuring level of association between DNAm and the unique functionalities of different cell types.  Stratifying tissue-specific DNAm loci according to CGI density, we found that LC regions, which constituted only 21% of all CpG sites on the array (212 out of 998 probes were located in LC regions), accounted for 31% of the 449 differentially methylated sites (hypergeometric p-value of 1.02e-12 indicating significant enrichment), consistent with this class having the lowest correlation between BEC and PBMCs. To further corroborate this finding, we calculated probewise ICC for each of the three LC, IC, and HC categories. ICC is a reliability index which, in this case, was used to measure the degree of between-tissue methylation difference: ICC values closer to -1 indicate larger difference between tissues than within tissues, while ICC values closer to +1 indicate larger within-tissue difference. Distribution of ICC values showed that the majority of HC and IC probes were clustered around ICC of 0, indicating equitable within and between tissues differences (Figure 7B). On the other hand, LC probes were clustered around ICCs of 0 and -1, with the latter displaying the most prominent peak, which means that the majority of LC probes harbored larger between than within tissue differences. These results thus suggested that tissue specific DNAm were enriched for regions of low CpG density.  3.3.2 Variability of DNAm was largely tissue-specific Having defined tissue-specific DNAm, we next turned to inter-individual variability within each tissue and, in particular, how this variability compared between the two tissues examined. For 47  each CpG site, we calculated the inter-individual range of beta values for both PBMC and BEC, and then visualized their overall distribution (Figure 8A). BEC had a larger overall DNAm range than PBMC, which was confirmed statistically by ANOVA on the BEC and PBMC distributions (p-value of 2.05e-8). Next we determined the degree of variation for each CpG by calculating the standard deviation (SD) of each CpG site for the two tissues. In the combined BEC and PBMC data, we compared the number of highly variable probes across tissues, using SD > 0.1 and SD > 0.3 as two levels of cutoffs for high variability. Using a cutoff of 0.1, 123 (12.3%) of PBMC and 150 (15.0%) of BEC loci were variable; using a SD cutoff of 0.3, 26 (2.6%) of PBMC and 55 (5.5%) of BEC loci were variable (Figure 8B). To determine the regions at which variable loci were most likely to be found, we performed hypergeometric tests on CpG density compositions, and found that variable loci (150 with SD > 0 .1) in BEC were enriched for LC regions (p-value of 2.39e-5, 51 LC loci) and IC regions (p-value of 2.19e-2, 28 IC loci), but depleted of HC regions (p-value of 3.51e-8, 45 HC loci). Highly variable BEC loci with SD>0.3 (55 loci) were enriched for LC regions (p-value of 0.01, 18 LC loci), but depleted of HC regions (p-value of 0.02, 20 HC loci). Similar trends were observed for variable loci in PBMCs (SD>0.1, 123 loci), which were enriched for LC regions (p-value of 1.57e-2, 35 LC loci) and IC regions (p-value of 9.38e-3, 25 IC loci), but depleted of HC regions (p-value of 4.59e-3, 48 HC loci), though highly variable PBMC loci (SD>0.3, 26 loci) were not enriched or depleted of any particular region.  Taking advantage of our matched study design, we compared variability of sites between BEC and PBMC directly by calculating the variance for each probe, and determining equality of variance using Levene’s test (Levene, 1960). This site-by-site evaluation of variance equality showed that 346 (34.7%) loci did not have equitable variance across the two tissues (Levene test p-value < 0.05). Of these loci, 144 showed larger BEC variance, and 202 showed larger PBMC 48  variance. This is in contrast to the earlier analysis where we found more sites with large SD in BEC when examining each tissue separately. Plotting the variance of PBMC versus the variance of BEC, we obtained a Pearson’s correlation coefficient of 0.26 (Figure 9). By setting variance in BEC as a reference, we searched for probes for which the PBMC variance deviates more than ±20% from the BEC variance. Using this method we found 584 probes residing within the ±20% interval, and 414 probes outside the interval (Figure 9). These observations indicated that variance was specific to the tissue; a probe that was highly variable in BEC would most likely not behave the same way in PBMC.   We next sought to find out how similar between-individual patterns of DNAm were across tissues. In other words, if individual a had higher DNAm than individual b at probe i in PBMC, would the same or opposite trend be observed in BEC. We carried out a probewise Pearson’s correlation of the DNAm values between BEC and PBMC and found that only 29 probes displayed relatively good correlation between the two tissues (absolute Pearson’s correlation coefficient > 0.5) (Figure 10). A total of 23 probes out of 29 survived FDR  correction with a medium confidence FDR of 0.25, and 13 probes survived using a high confidence FDR of 0.05. This indicated that in most cases, the relative individual level of DNAm at a specific probe was not translatable across tissues.  3.3.3 Tissue-specific association between DNAm and demographic factors We next performed an exploratory analysis to examine each tissue individually for associations with a small set of demographic factors that we had collected in our cohort. Given the small size of our cohort, these analyses explore how the differences in variability between BEC and PMBC 49  affect their association with demographic factors, rather than constituting a statistically significant determination of specific probes that are associated with these demographics. As a rough guide to the relative associations between DNAm and the demographics in our tissues, in addition to the number of associated probes we also examined the distributions of unadjusted p-values. As we have previously described, left skewing of p-value distributions is indicative of a higher likelihood of association (Lam et al., 2012). The variables examined included: age (26-45), gender (15 male versus 10 female), ethnicity (18 Caucasian versus 7 non-Caucasian), and body mass index (BMI, 18.18-47.95). We used a two tailed Wilcoxon rank sum test for gender and ethnicity, and a spearman two-sided correlation for age and BMI, and FDR was corrected with Storey’s qvalue method (Dabney, Storey, & Warnes). No specific probes were found to be associated with age, but skewed p-value distributions indicated a signature of age (Figure 11). Using a FDR ≤ 0.05 (high confidence), in BEC 0 probes were associated with gender while 4 probes were associated with BMI. The p-value distributions for the association of gender and BMI with methylation in BEC were quite different (Figure 11). Gender showed the left-skewing associated with higher confidence associations, while BMI showed the flat distribution that generally indicates low likelihood of association (Figure 11). In PBMCs 40 probes were associated with gender (10 with ∆β > 5%), and 51 probes were associated with BMI (31 with ∆β > 5% between subjects with highest and lowest BMI) using an FDR of 0.05 (Table 3). These results were both supported by their skewed p-value distributions (Figure 11).  50   Figure 5 BEC displayed overall lower DNAm than PBMC (A) Sample mean DNAm levels for BEC (grey) and PBMC (red). Dashed red line represents mean DNAm for each tissue. Each of the 25 individuals displayed higher average PBMC DNAm than BEC DNAm. (B) Distribution of PBMC (grey) and BEC (red) beta values. BEC had more hypomethylated CpG sites than PBMC, and PBMC had slightly more hypermethylated sites than BEC.   51                     Figure 6 Correlation of DNAm was higher within than between tissues. (a) Distribution of Pearson’s correlation coefficient for correlations within BEC, within PBMC, and between BEC and PBMC. In each instance, dashed lines represent the average correlation coefficient: within BEC=0.980, within PBMC=0.987, between tissue=0.773. (b) Distribution of Pearson’s correlation coefficient for BEC versus PBMC sample-wise correlation with CpG sites separated into four categories: low, intermediate, and HC, and uncategorized. Low CpG density islands displayed the lowest correlation coefficients.  52             Figure 7 Tissue-specific DNAm was enriched in regions of low CpG density. (a) Probewise tissue specific DNAm was evaluated by t-test p-value and ∆β. Categories of differentially methylated probes were identified by color. 12.1% of all CpG loci (red) had p-value<0.001 and ∆β>40%, and 17.8% of CpG loci (yellow) had  p-value<0.001 and ∆β>20%,; (b) Distribution of ICC for high, intermediate, and low CpG density regions. Probes residing in    53  regions of low CpG density had higher proportion of ICC values near -1.0.             Figure 8 Variability in DNAm was tissue-specific. (a) Distribution of DNAm range of 998 CpG sites for the two tissues. BEC had larger probe methylation range than PBMC. (b) Boxplot of SD for 26 sites with SD>0.3 in PBMC and the equivalent sites in BEC (c) Boxplot of SD for 55 sites with SD>0.3 in BEC and the equivalent sites in PBMC. High SD in one tissue did not predict high SD in the other tissue.   54         Figure 9 DNAm variability was tissue specific. 584 probes had PBMC variance that deviated <±20% from their corresponding BEC variance. 414 probes were found beyond the ±20% threshold, further demonstrating the variance discordance between BEC and PBMC.   55           Figure 10 Relative levels of DNAm among individuals were not translatable across tissues. Distribution of Pearson’s correlation coefficient (r) for probe-wise correlation between BEC and PBMC. Dashed line represent Pearson’s correlation coefficients of 0.5.    56              Figure 11 Demographic factors were associated with DNAm changes in both PBMC and BEC. Distribution of unadjusted p-values for association between gender (left), BMI (middle), and age (right) in BEC (top) and PBMC (bottom). Left-skewed p-value distributions showing an enrichment of CpG sites with small p-values suggested correlations of certain demographic factors with DNAm. This is in contrast to the relatively uniform p-value distribution of BMI and BEC, suggesting lack of such correlation.   57  3.4 Genomic DNA methylation levels are sensitive to diesel exhaust-exposure Having explored both temporal and interindividual-variability of DNAm, the last part of this thesis investigates how DNAm responds to exposure to DE. For the purpose of this experiment, sixteen subjects with physician-diagnosed asthma and/or baseline methacholine PC20 below 8 mg/mL were recruited (Table 4). A cross-over experimental design was implemented where subjects were exposed to either DE or FA, followed by a washout period. After the washout, the subjects were then exposed to the condition opposite to their initial exposure. By examining CpG sites across the genome, as well as those in LINE1, Alu, and miRNA regions, we found that short-term exposure to DE resulted in DNAm changes at CpG sites residing in genes involved in inflammation and oxidative stress response, repetitive elements, and microRNA. This provides plausibility for the role of DNAm in pathways by which airborne PM impacts gene expression.  3.4.1 Diesel exhaust was associated with changes in DNA methylation Using PCA, a dimensionality reduction technique, we decomposed the data to a set of 95 principal components (PCs), each of which explained a dominant and independent pattern of variation across the samples; we then assessed the relative contribution of each PC to the overall variance (Figure 12). PCA facilitated linear decomposition of the data, allowing identification of data dimension that significantly captured the association between DNAm change and exposure to DE. By focusing our analysis on the CpG sites most representative of that dimension, we were able to analyze only CpG sites with higher chance of being DE-associated. Surveying the pattern of methylation within the data, we found that demographic variables were associated with a number of PCs, and many of the associations were driven by correlations inherent among these variables (Figure 13; significant associations were captured after multiple testing correction by 58  the q-value method with FDR = 10%). In particular, age was a major factor driving data variability in PCs 1, 4, and 5, and BMI was strongly associated with PC4 (Figure 14). Since PBMCs are heterogeneous, we considered the contribution of blood cell composition to methylation change. To achieve this, we tested for association between PCs and white blood cell counts. We did not find association between PCs and any granulocyte cell types, as expected since they are removed during PBMC isolation. However, monocyte and lymphocyte counts were positively associated with PCs 7 and 14, and PCs 5, 11, 15, respectively. This indicated that inter-individual difference in blood cell composition was also a source of variance in the dataset. To account for blood composition difference and avoid confounding between diesel related methylation change and change resulting from blood cell population shifts, we identified CpG sites that were significantly associated with changes in blood count. First we calculated the probe-wise loading values of PCs 5, 7, 11, 14 and 15; for CpG site i with loading value m for PC n, the larger the magnitude of m, the closer the methylation pattern of CpG i approximates PC n. Using cutoffs of +/-3SD, we found 19250 sites that were potentially associated with blood cell counts. By regressing DNAm measurements at these sites against lymphocyte and monocyte cell counts, we identified 11378 sites that were significantly associated with blood cell counts (multiple testing correction by the q-value method; FDR =10%). All hits reported to be associated with DE in this study have been filtered to exclude these sites associated with blood cell composition.  A lingering concern with DNAm time-series data is that there would an influence of time, stochastic or biological, that affects measurements. We did not find association of time (0 hr to 59  30 hr) with any PC, indicating that DNAm did not vary as a result of measurements taken across the three time points.  Having controlled for potentially confounding variables, we next sought to identify CpG sites associated with DE exposure. We assessed three different scenarios of DE-induced changes assuming baseline measurements at FA-0 hr, FA-6 hr, FA-30 hr, and DE-0 hr (serving as pre-exposure baseline), and tested for association between these scenarios and principle component scores using Wilcoxon’s ranked sum test. Scenario-1 hypothesized that methylation changes are detectable 6 hr post DE-exposure, but diminish at 30 hr post-exposure (thus, it compared DE6hr against non-DE6hr). Scenario-2 hypothesized that changes are detectable 30 hr post-exposure (thus, it compared DE30hr against non-DE30hr). Scenario-3 hypothesized that changes are detectable at 6 hr and persist at 30 hr post-exposure (thus, it compared DE6hr&30 hr against non-DE6hr&30 hr). Scenario-3 was significantly correlated with PC22, which accounted for 0.6% of the total variance. Importantly, PC22 was not associated with any experimental confounds such as demographic factors and blood cell counts. To further examine the pattern of variation underlying PC22, we derived the PC22 scores, which could be used to interpret the correlation between that PC and variable Scenario-3. By ordering the PC22 scores into DE6hr&30 hr versus non-DE6hr&30 hr, we observed that DE6hr&30 hr samples exhibited mostly positive scores with a mean of 0.048, whereas baseline samples exhibited mostly negative scores with a mean of -0.024. This observation was statistically substantiated by a two-sample t-test (p-value = 1.6e-4) (Figure 15a). Next, we identified DE-associated CpG sites by calculating the probe-wise loading values of PC22 (Figure 15b). Using cutoffs of +/-3SD, we found 2827 such sites. Functional analysis of these probes using DAVID revealed enrichment of genes 60  involved in regulation of protein kinase and NFkB pathways (enrichment score of 3.01; Table 5) (Huang et al., 2009).  3.4.2 Diesel exhaust-associated changes were found in genes relevant to allergic disease The deviation from random of the raw p-value distributions showed that the post-DE (6 hr&30 hr) versus pre-DE (0 hr) comparison demonstrated high association with DNAm change, suggesting that they were more correlated than that expected by chance, while the same association was not found for post-FA versus pre-FA (Figure 16a). Consistent with the pronounced skewing of the p-value distributions, 170 differentially methylated positions (DMPs) were significant for DE exposure (after multiple testing correction using high confidence FDR of 10%), but not FA exposure. Out of these 170 sites, 25 were previously identified to be significantly associated with blood cell counts; these sites were removed from proceeding analysis, leaving 145 DMPs that demonstrated significant change in methylation as a result of exposure to DE. Specifically, 102 out of the 145 DMPs decreased in methylation (Table 6a), while 43 sites increased in methylation (Figure 16b and c, Table 6b). We calculated the average beta value change (Δβ) between post-exposure and pre-exposure for the DMPs, and we found 1 probe showing increase and 6 probes showing decrease in DNAm of greater than 5%. Interestingly, 1 out of 8 GSTP1 sites (cg09038676) were among the significantly correlated loci. For GSTP1, we genotyped subjects for the A- > G substitution at position 313 of the GSTP1 gene, and we used this information to stratify the samples in an attempt to determine whether the difference in magnitude of DNAm change was related to genotype. As expected, at baseline under the FA condition, DNAm levels at cg09038678 for subjects with the G allele were significantly different from that of subjects with the A allele (two-sample t-test p-value = 61  0.00024). Furthermore, individuals with the A allele showed significant change pre- and post- DE exposure, but not individuals with the G allele (Figure 17).  Lastly, we tested whether DE-related effects carried over to the subsequent FA exposure. Hypothetically, if carryover effects were present then probes found to have increased methylation would show higher mean FA methylation in subjects exposed to diesel first (vice versa for probes found to have decreased methylation). We compared FA mean methylation values in participants who were exposed to DE before FA by first separating the 145 DMPs into ones with decreased and increased methylation in response to DE, then separating the groups further according to exposure order (Figure 18). Neither visual inspection of the graphs nor Welch’s two-sample t-test supported any carryover effect (p-value > 0.05), consistent with our previous assessment of effective binding (Carlsten et al., 2013).  3.4.3 Alu and LINE1 CpG sites showed methylation changes post diesel exhaust exposure To investigate the impact of DE-exposure on the methylation of Alu and LINE1, we identified in silico all probe positions on the array that shared a larger than 15 base-pairs overlap with Alu or LINE1 repetitive elements in the genome. Applying LME modeling to these CpG sites, both Alu and LINE1 elements exhibited changes in DNAm as a result of DE exposure, and no changes as a result of FA exposure (Figure 19). 25/1271 (2%) and 31/1118 (3%) DMPs (CpG sites significantly associated with blood cell counts were not considered DMPs) were identified for Alu and LINE1, respectively (Table 6c and d). Out of the 25 Alu DMPs, 12 increased in methylation while 13 decreased in methylation after DE exposure; similarly for the 31 LINE1 DMPs, 13 increased while 18 decreased in DNAm after exposure. The significant loci found for 62  Alu and LINE1 elements did not overlap with the 145 PCA DMPs, since none of their PCA loading values made the ±3 SD cutoff.  3.4.4 MiR21 showed decrease in methylation upon diesel exhaust exposure Recently we identified changes in expression in peripheral blood of four microRNAs (miR-21, miR-30e, miR-215 and miR-144) in response to DE exposure in the same subjects that were involved in this study (Yamamoto et al., 2013). To determine whether differential methylation could be a mediator for the expression of these four microRNAs, we localized a total of 7 probes on the array that overlapped with the genomic positions of these microRNAs. A probe residing within miR-21 (cg07181702), which demonstrated increase in expression in response to DE, showed significant decrease in methylation in this study (and absence of significant change with FA): cg07181702 decreased in methylation by 3.9% in response to DE exposure (significant with 10% FDR) (Figure 20) (Yamamoto et al., 2013).   63           Figure 12 Percent variance within the dataset accounted for by the first 22 PCs.  (0th PC was disregarded since it is focused on probe offsets).   64             Figure 13 Heatmap showing correlation among demographic variables and differential cell counts. Red indicates negative Pearson’s correlation coefficient of -0.5. Green indicates positive Pearson’s correlation coefficient of 1.0. Abbreviations are as follows: basophils (BAS), lymphocytes (LYM), monocytes (MON).   65            Figure 14 PCs were associated with demographics and biological variables. For PCs 1-22, only those with associations were included. Colors correspond to p-values where darker colors indicate higher association, and lighter colors indicate lower association. Abbreviations are as follows: basophils (BAS), lymphocytes (LYM), monocytes (MON).   66               Figure 15 Diesel exhaust-associated exposure patterns were captured in PC 22. a) PC22 was associated with model-3 (DE6&30 hr vs. non-DE6&30 hr). Samples were sorted according to DE6&30 hr samples versus non-DE6&30 hr baseline samples and their PC22 associated PC scores were displayed. There was a visible pattern indicative of DE-exposure with the DE6&30 hr samples showing mostly positive scores, and the baseline samples showing mostly negative scores. b) For each probe in the dataset, its loading value in association with PC22 was calculated. The +/-3SD and +/-6SD positions are marked by vertical dashed lines. And the probes harboring the largest contribution to diesel exposure methylation pattern were selected for using the +/-3SD cut off, resulting in 2827 hits.   67             Figure 16 Gene-specific methylation changes were found at CpG sites across the genome. a) Unadjusted LME model p-value distributions for the effect of pre (0 hr) versus post (6-30 hr) exposure for FA, and that for DE. The uniform distribution expected by chance is indicated with horizontal dashed lines. Lack of positive skewing in the FA distribution indicated no association between FA exposure and differential methylation, whereas positive skewing in the DE plot indicated heavy association of DE exposure with differential methylation. b) Volcano plot of 2827 PCA hits. Negative log10 of fdr 10% adjusted p-values from LME modeling of the PCA hits were visualized against their corresponding Δβ values. Probes that were significant with a high confidence fdr of 10% were indicated in red. c) Distribution of ∆β obtained from subtracting pre-FA beta values from post-FA beta values, and from subtracting pre-DE beta values from post-DE beta values for the PCA identified probes that were significantly altered by DE exposure. ∆β from DE samples were much larger in magnitude than that from FA samples.   68            Figure 17 Changes in methylation of the GSTP1 probe cg09038676 in response to DE-exposure stratified between subjects with the A allele and those with the G allele.   69             Figure 18 Mean beta value of exposure to FA at 0hr, 6hr, and 30hr for the 170 probes found to be significant for DE exposure but not for FA exposure. The beta value differences were first stratified between probes with decreased and increased methylation in response to DE, and then further divided between subjects who were exposed to DE first and those who were exposed to FA first.   70             Figure 19 DNA methylation of CpG sites overlapping with Alu and LINE1 repetitive elements was associated with diesel exhaust. A) Distributions of unadjusted p-values from LME modeling applied to CpG sites overlapping with Alu and LINE1 elements. The uniform distribution expected by chance is indicated with horizontal dashed lines. Positive skewing of DE p-value distributions indicated association with differential methylation, whereas lack of skewing of FA p-value distributions indicated absence of association. B) Distributions of Δβ obtained from subtracting pre-FA beta values from post-FA beta values, and from subtracting pre-DE beta values from post- DE beta values for LINE1 and Alu probes that were significantly altered in response to DE exposure. Δβ from DE samples were much larger in magnitude than that from FA samples.   71           Figure 20  A CpG site residing in the miR-21 genomic locus changed in DNAm in response to DE. Post and pre methylation beta values of CpG site cg07181702 (residing within miR-21) found to have significantly decrease in methylation by 3.9% in response to DE-, but not FA-, exposure.    72  Chapter 4: Discussion The aim of this thesis was to characterize epigenetic variability as well as environmental influences on epigenetics in order to build foundational understandings for this field of study. We achieve this by exposing temporal and differential tissue variabilities present in healthy human subjects, as well as demonstrating how ambient PM could result in differential DNAm. The results reported here defined baselines for other studies, and it is only by defining what is normal could we define what is abnormal. The growing importance of DNAm variability is quite visibly within the epigenetic community. Study has shown that just like differential methylation, which uses at mean methylation as opposed to variance, differential variability of CpG site are also powerful indicator of obesity-associated methylation sites; at those methylation sites, obesity cases had more variable methylation than lean controls (Xu et al., 2013). On the note of environmental influences, humans are exposed to an abundance of chemicals throughout life. The key message here was that even short-term exposure to PM could impart strain on DNAm. On one hand, this emphasizes the need to establish an environmental database of epigenetic effects; on the other hand, it raises concerns for the social policies that impact global health. Lastly, concepts here were investigated in the two most accessible peripheral tissues PBMC and BECs. For disease-specific studies, primary tissues from the source of the disease are often required. For example, studies investigating the link between DNAm and disease in the brain would require tumour or autopsy samples. However, the goal of EWAS studies mostly differs from that of disease specific studies. First EWAS studies often do not examine specific diseases or particular conditions, instead, they investigate a wide range of epigenetic characteristics in association with effectors, such as diet, ethnicity, lifestyle, and environment, etc.; since these effectors would pose a systemic influence on the human body, it is better to choose a generic 73  representative tissue, such as PBMC or BEC. Also, these studies require larger cohorts; therefore, it is much more cost and time effective to collect the easily accessible tissues with the assumption that changes observed in these tissues reflect systemic changes.   4.1 Epigenomic characterization 4.1.1 DNA methylation is variable over the course of one month The first concept we established is the temporal variability of DNAm. Since methylation is a malleable mark responsive to even short term environmental cues, it is key to understand how it changes in over the circadian rhythm in healthy individuals. Specifically, we were interested in identifying CpG sites that vary stochastically over the period of a day, as well as those that vary over the period of a month. Results showed that methylation is likely stable over the course of a day. Interestingly, results from the DE exposure study corroborated with this finding. In that study we did not find any arbitrary variability attributable to sample collection time. Thus according to these results, statistical models detecting changes in methylation would not need to correct for staggers in sampling time within a circadian day. On the other hand, we discovered certain CpG sites to exhibit variability over the course of one month. A study done on the temporal variability of DNAm at 12 promoter regions over the course of 4 days also reported finding stochastic variability; according to the study, there are certain correlations between the variability of a CpG loci and its genomic location and methylation level (Byun, Nordio, Coull, Tarantini, Hou, Bonzini, Apostoli, Bertazzi, & Baccarelli, 2012b). Although we did not find associations between these CpG sites and the CpG density of the region in which they reside, we did find that most of the variable sites were hypomethylated, which corroborated with the results documented in the similar study that examined 12 promoter loci (Byun, Nordio, Coull, Tarantini, 74  Hou, Bonzini, Apostoli, Bertazzi, & Baccarelli, 2012b). Our results showed that sample collection over the course of a month would not cause random variability on the global scale. But because certain CpG loci demonstrated to be intrinsically more variable, thus studies should either account for staggers in sampling time or avoid drawing conclusions from CpG loci in high variable regions.  4.1.2 DNA methylation is sensitive to delay in sample processing Besides stochastic temporal variations, we also determined how sample processing delays could affect DNAm levels, leading to changes in methylation from ex vivo influences that could complicate examination of in vivo biological effects. Epigenome wide association studies involve collecting samples from a large number of subjects across a wide geographic region. Therefore temporal lag between the time of sample collection and that of sample processing is not uncommon. Since sample processing immediately after collection is often not possible, understanding how increased delay in processing time would affect changes in DNAm would be the next best strategy. Our results demonstrated that prolonged sample incubation indeed resulted in significant undesirable changes, reflecting an experimental confound rather than true biological findings. This observation was reminiscent of similar documented effects of delayed processing time on PBMC mRNA expression, where it was found that processing delays as short as a few hours led to both increase and decrease in expression of genes (Barnes, Grom, Griffin, Colbert, & Thompson, 2010; Duvigneau, Hartl, Teinfalt, & Gemeiner, 2003). In this case, while resting cells showed methylation changes in both directions, the dominant pattern of change was decrease in methylation; such pattern could be explained by decreased cell viability, which led to decreased methylation maintenance. Another potential scenario was change in cell composition. 75  PBMC consists of many different cell types, thus cell death could be concentrated in a specific type of cell, leading to amplification of DNAm patterns of other cell populations. Lastly during the collection and storage process, PBMC was exposed to external stresses such as shear stress, as well as to the foreign surface of apparatuses. Shear stress in particular could contribute to decrease in methylation, for it induces production of reactive oxygen species which mostly include hydroxyl radicals (Paravicini & Touyz, 2006; Smolarek et al., 2010). Radicals can react with the methyl group on the C5 position leading to demethylation and deamination (Smolarek et al., 2010). We found CpG loci within genes ALOX12, CDH13, EPHA2, KLF5, and STAT5, to have significant change in methylation due to prolonged storage (Diakiw et al., 2013; Kuang et al., 2010; Morales et al., 2012; Shivapurkar et al., 2004; Q. Zhang, Wang, Liu, & Wasik, 2007). This may be cause for concern, as the PBMC methylation levels of these genes are implicated in disease models. It would be ideal to process PBMC samples immediately after collection, but this is not feasible for most EWAS studies. One way to circumvent this problem is to remove the time sensitive methylation loci from analysis, the caveat of which would be removing sites that are important for the biological processes investigated. Thus, to account for prolonged storage, we suggest that process-time delay be either included as a variable in the statistical modeling or be regressed out of the dataset before analysis. Our group has documented a regression approach for differential cell counts; applying that method here, the dataset would be regressed against the recorded processing times or sample collection time, after which addition between a mean matrix and the regression residual matrix would rid the data of undesirable temporal effects (Lam et al., 2012).  76  4.1.3 PBMCs and BECs differ in certain key epigenetic features The tissue specificity of DNAm is a foundational knowledge of epigenetics. For research studies, DNAm difference between tissues is closely associated with tissue choice, as well as generalizing finding in one tissue to the systemic level. Recently, tissue variability has been gaining traction as a unit of measure as importance as tissue specificity. Thus a challenge in this thesis is to understand interindividual variability in the context of tissue specificity. We achieved this by taking advantage of matched tissue samples from a cohort of healthy adults to examine basic distribution of DNAm, its variation between individuals, and its broad association with selected demographic factors for 998 CpG methylation sites. Given the fundamental role of the epigenome in specifying tissue differences, it was not surprising that a large fraction of CpGs had substantially different DNAm levels between the two tissues. While the epigenetic differences between tissues greatly overshadowed inter-individual CpG methylation differences in the same tissues, the latter nevertheless encompassed a number of loci and revealed some important biological principles. Specifically, we showed that variation in DNAm can be tissue specific, and in addition, exploratory analysis showed that the association of DNAm with demographic variables was at least partly tissue-specific as well. As such, these results have practical implications for sample selection for population epigenetics studies.  With regards to tissue-specific DNAm marks, our results derived from matched samples largely confirmed principles from a large body of published data, most of which were derived from unmatched samples. We discovered that CpGs differing between tissues were not only significant in number but also large in the magnitude of absolute methylation difference. Reassuringly and consistent with the published literature, these CpGs resided primarily in LC 77  regions, and consequently, these also had by far the lowest pairwise correlation across tissues (Byun et al., 2009; Davies et al., 2012; Fernandez et al., 2012; Lam et al., 2012; Yuen et al., 2011; D. Zhang et al., 2010; Ziller et al., 2013). This implies that DNAm of PBMC and BEC is more different at regions of low CpG density than at medium or high density. We also replicated the finding that BECs have overall lower DNAm levels than PBMCs (Bjornsson, 2004; Hatchwell & Greally, 2007; Lowe et al., 2013). The issue of genetic variants that affect DNAm, known as methylation quantitative trait loci or mQTLs, is a concern in many epigenetic studies(Gutierrez-Arcelus et al., 2013; Moen et al., 2013; D. Zhang et al., 2010). Here, however, our matched design means that our comparisons across tissues are not confounded with respect to mQTLs, since we performed paired testing in samples from the same individual. When comparing across individuals, it is possible that mQTLs contribute to the variance in probes, however we believe this to be highly unlikely as the overall known percentage of CpG sites associated with mQTLs is very small, generally estimated between 1% and 7%  (Fraser, Lam, Neumann, & Kobor, 2012; Gutierrez-Arcelus et al., 2013; van Eijk et al., 2012).   Inter-individual variability across human tissues has been investigated in several recent studies (Bock et al., 2008; Lam et al., 2012; Schneider et al., 2010; D. Zhang et al., 2010). Our matched sample design allowed us to investigate inter-individual variability across the two most accessible human tissues. Highly variable probes were identified largely in low CpG density regions in both tissues, a finding which is consistent with published reports, though overall, studies are divided on which class of CGI is the most variable (Davies et al., 2012; Lam et al., 2012; Ziller et al., 2013). In our study, several important epigenetic features differed substantially between the two tissues. First, we found that when examining each tissue 78  separately, BEC had a larger variable range and contained more highly variable sites than PBMCs. Second, BEC had a stronger association between probe variability and CGI type compared to PBMC. Third, and perhaps most importantly, CpGs that had a variable range in one tissue did not necessarily have the same range in the other tissue. In fact, variance of DNAm for a given site differed at over a third (34.7%) of loci tested. Also, despite having fewer sites with high variance overall, PBMCs had more sites with higher variability within the loci that showed discordant methylation between the two tissues.  This observation was particularly important for epigenetic population studies that rely on these easily accessible tissues for indicators of DNAm variability; depending on which tissue is chosen for the study, the outcome of identified differentially variable CpG sites could be vastly different. This could then affect the tissue’s associations with the biological and deterministic factors that are often of interest in epigenetics studies.   Consistent with the differences in variability across tissues, we next showed that associations between DNAm and a small set of demographic factors were also different between the two tissues. In particular, while both tissues had CpG sites that were associated with gender, only PBMCs demonstrated a substantial number probes that were statistically significant. Similarly we identified more CpGs associated with BMI in PBMCs than in BEC, consistent with a recent report which also showed DNAm associations with BMI in PBMCs (PhD et al., 2014). Although we did find an age signature from the positive skewing of p-value for both tissues, we did not find significantly related CpG sites. The lack of significant age-related sites could be due to the limited age range (26-45) of this study compared to studies dedicated to examining age-related 79  changes in DNAm (generally <25 to >75), or to the relatively restricted distribution of the CpGs analyzed (Hannum et al., 2013; Horvath et al., 2012; McClay et al., 2014; Weidner et al., 2014).  Collectively, these results present somewhat of an interesting conundrum: BEC DNAm was more variable across individuals, yet PBMC DNAm was more strongly associated with the demographic variables tested here. In contrast, published work suggests that BECs are a more suitable tissue than peripheral blood for population epigenetic studies (Lowe et al., 2013). This was largely based on the finding that BECs containing more hypomethylated regions than PBMCs, which not only seem to overlap with hypomethylated regions in other tissues but also tend to cluster around disease associated SNPs (Lowe et al., 2013). However, given that PBMCs represent a circulating tissue directly involved in the immune and inflammatory systems and are thus functionally related to many major diseases, it is our opinion that in most cases, PBMCs would be the better choice (Hatchwell & Greally, 2007; Y. Li et al., 2010; Michels et al., 2013).  It is of course possible that for specific studies, selection of the appropriate tissue might be context-dependent, but our results show that for population epigenetics studies more broadly, the implications of tissue-specific variability need to be carefully considered when assessing DNAm, in particular when the primary tissue of interest is not accessible.  Lastly, both stability and variability of DNAm were based on methylation assessments using the Illumina GoldenGate technology. The GoldenGate array has limited CpG loci coverage, with an average of 1.5 sites per gene. As well, genes selected for the assessment in the array is biased towards genes involved in cancer, DNA repair, differentiation, apoptosis, etc. Thus, the observations obtained here must be interpreted in the context of these limitations. However, 80  given that we demonstrated value using the GoldenGate, it follows that studies involving better genomic coverage are needed to fully explore the questions posed here.  4.2 Environmental characterization In the last part of this thesis, we assess the sensitivity of DNAm to short exposure to ambient PM pollution. As well, we investigate the parallels between our epigenetically based findings and mRNA expression changes found in a previous study.  Inhaled air pollution including emissions from diesel engines have been associated with a host of cardiovascular and respiratory diseases, imparting a significant strain on public health (Brook, 2008; Peng et al., 2009). Data presented here on DNAm changes in response to short-term exposure to DE demonstrated a potential epigenetic mechanism for biological responses to DE exposure. PCA enabled detection of DE signals from subject-level noise generated by demographic variables such as age, ethnicity, and BMI, and focused the investigation onto a smaller list of pertinent CpG sites. Furthermore, the DE-associated methylation changes in LINE1 and Alu CpG sites corroborated with previous research demonstrating sensitivity of repetitive elements to environmental exposure (Baccarelli et al., 2009; Bellavia et al., 2013). Lastly, methylation change was also observed for miR-21, a microRNA associated with oxidative stress and allergic inflammation. Collectively, these results were consistent with our hypothesis that epigenetics servers as a potential avenue through which exposure to air pollution impacts biological systems.  81  Air pollution has long been linked to diseases including asthma and cardiovascular issues (Bredy et al., 2010; Lam et al., 2012; Peng et al., 2009; Zanobetti et al., 2000). What is still lacking in our understanding is defining the biological pathways that may be either a mediator or a consequence of the association. DE is a major source of fine particles that have been shown to impact gene expression, and change DNAm at inflammation associated sites (Baccarelli et al., 2009; Bellavia et al., 2013; Hou et al., 2011; 2014). Our results support previous studies that have found an association between PM exposure and decrease in DNAm of target promoter, candidate genes and repetitive elements (Baccarelli et al., 2009; Bellavia et al., 2013; Tarantini et al., 2009). We suggest that this could be explained through the effects of DE on the dynamics between DNA and DNA methyltransferases. First, ROS-induced oxidative damage to methyl-CpG binding protein recognition sequences could have inhibited the capacity for DNA methyltransferase to bind and methylate DNA (Valinluck et al., 2004). Secondly, decrease in saturation of the DNA methyltransferase enzymes could potentially come into play, since studies have shown a dependent relationship between exposure to particles and decrease in mRNA transcripts of the DNA methyltransferases Dnmt1, Dnmt3a, Dnmt3b in mouse macrophages (Miousse et al., 2014). Decreased mRNA transcripts could lower the cellular concentration of methyltransferases and as a result decease the methylation level of CpG sites. Both of these scenarios could be part of the mechanism leading to loss of methylation that partially explains why the majority of the gene-specific sites found here showed decrease in methylation.  DAVID functional enrichment analysis of the 2827 DE-associated probes identified by PCA revealed enrichment of NFkB-related functions. NFkB is a redox-related transcription factor that activates a pro-inflammatory response to ROS induced oxidative stress (Baeuerle & Henkel, 82  1994; Chan et al., 2010; Donaldson et al., 2005; Mazzoli-Rocha, Fernandes, Einicker-Lamas, & Zin, 2010). NFkB has been shown to activate in response to DE exposure, increasing the downstream expression of inflammatory cytokines (Takizawa et al., 1999). Besides NFkB, we also discovered enrichment of the MAPK protein kinase pathway. Previous studies have shown that MAP kinase is a stress-activated kinase pathway that can be induced by DE exposure (Hashimoto et al., 2000; Hiura et al., 1999). Given the proposed role of DNAm as a mediator between environment and gene expression, the observations made here further demonstrated that a tight circuitry exists between transcriptional pathways known to be involved in response to DE and DNAm changes elicited by DE exposure.  Results from PCA followed by regression modeling in this study revealed that in agreement with our hypothesis, DE was associated with changes in DNAm at genes that are known to be associated with inflammation and oxidative stress, most notably GSTP1 (Romieu et al., 2006; Zimniak et al., 1994). Studies investigating the effects of GSTP1 gene polymorphism found that subjects homozygous for the GSTP1 G allele have lower functional levels of the enzyme, and thus are at increased risk for oxidative stress, lung cancer, and asthma (Miller et al., 2003; Romieu et al., 2006; Tamer et al., 2004). Since the GSTP1 associated probe (cg09038676) was located in the body of the GSTP1 gene, then its increase in DNAm would likely to be associated with increase in GSTP1 transcription, resulting in higher levels of the enzyme to combat the effects of oxidative stress. Interestingly, this effect was most profound for subjects with at least one A allele, suggesting that these individuals are less susceptible to the negative effects of DE exposure than individuals homozygous for the G allele, consistent with the literature (MacIntyre et al., 2014). Non-asthmatic controls were not included in this study, thus we cannot conclude 83  that the changes observed are specific to asthmatics, but our finding bolsters mechanistic plausibility nonetheless. Furthermore, the significant baseline (FA) methylation difference between subject with and without the GSTP1 G allele substitution that cg09038676 could have been a methylation associated SNP; however, due to the fact that only 5 of 16 subjects had an A allele, we were not able to quantitatively assess this possibility.  Besides GSTP1, we also found decreased methylation in a probe residing within the body of HDAC9 (cg24458314), a class IIa histone deacetylase (de Zoeten, Wang, Sai, Dillmann, & Hancock, 2010). A study involving regulatory T (Treg) cells in mice showed that absence of HDAC9 enhances the suppressive ability of Treg cells, resulting in decreased immune responsiveness and inflammation (de Zoeten et al., 2010). Changes in DNAm of HDAC9 such as we observed could impact downstream gene expression that then modifies the allergic airway inflammation in response to DE exposure. Considering that Tregs constitute only a small fraction of PBMCs, we speculate this small change might indicate a profound effect on the Treg population; however, a more concrete conclusion could not be reached without independent examination of Treg cell-specific methylation.  Lastly, we discovered that cg05094429, which resides in the promoter region of CCR6 gene, decreased in methylation after exposure to DE. CCR6 is expressed by both Treg and Th17 cells, and it is a key regulator of the migration of these cells to sites of inflammation (Yamazaki et al., 2008). Lack of this protein in Th17 cells hampers the recruitment of both Th17 and Tregs (Yamazaki et al., 2008). Thus it is possible that decreased methylation of CCR6 in the promoter region could have resulted in increased expression of CCR6, eventually leading to increased 84  presence of Th17 and Treg cells responding to DE induced inflammation. It should be noted an effect of ambient PM on DNAm patterns has been previously documented in numerous studies (Baccarelli et al., 2009; Bellavia et al., 2013; Hou et al., 2011; 2014; Tarantini et al., 2009). Therefore, despite the fact that most changes found here were small in magnitude, the overall findings reported here demonstrated an effect of DE that is concordant with existing observations and is impactful on the genomic scale.  Repetitive elements comprise of half of the genome and, under normal conditions, harbor higher DNAm in comparison to the rest of the genome (A. S. Yang et al., 2004). Repetitive elements are activated during cellular stress, which in this case could be elicited by exposure to DE (Lucchinetti et al., 2006). In accordance with past research on effects of PM, we found both increase and decrease in Alu and LINE1 methylation after DE exposures (Miousse et al., 2014). In some cases, demethylation of Alu and LINE1 increases genomic instability, which could mean that DE exposure predisposes some cells to genomic rearrangements (Bennett et al., 2008; Cho et al., 2007; Gaudet et al., 2003). Furthermore, repeat elements also impact adjacent sequences; they can propagate the spread of DNAm to nearby sequences or serve as insulators (Lunyak et al., 2007; X. Wang et al., 2011). A recent study has shown that sequences close to a 3’ repeat element demonstrated better methylation stability (Byun, Nordio, Coull, Tarantini, Hou, Bonzini, Apostoli, Bertazzi, & Baccarelli, 2012a). Therefore, it is conceivable that disruption of repetitive element DNAm due to DE exposure could be associated with genomic rearrangement as well as affecting methylation stability of nearby sequences causing changes in gene expression.  85  In this study we also examined the effects of DE on methylation of microRNAs, which are a class of small noncoding RNAs with functions in post-translational regulation of expressed genes that are important mediators of cellular processes (L. He & Hannon, 2004). Pollutants can cause microRNA dysregulation, which could lead to lung diseases and inflammation (Yamamoto et al., 2013). In particular, miR-21 is involved in oxidative stress and allergic inflammation, and has been shown to be up-regulated in asthma (Bollati et al., 2010; Cheng et al., 2009; Lu et al., 2009; Yamamoto et al., 2013). Our results showing that DNAm was also altered after DE build upon a previous study on the same individuals that reported changes in miR-21 expression upon DE exposure (Yamamoto et al., 2013), though this observation was associative and we cannot conclude that the changes in methylation are directly responsible for the changes in expression.  Lastly we were able to demonstrate that DE-associated changes lasted at least 30 hr post-exposure. This observation was consistent with reports showing that methylation changes elicited by environmental exposure could persist for days in the absence of persistent triggers, though the precise dynamics of this in the context of air pollution requires further study. These results should be interpreted in light of the limited sample size, lack of another dataset sufficiently similar to attempt validation, and the systemic nature of the tissue investigated. Nevertheless, this investigation presents a novel approach to analyzing the association between PM and genome-wide DNAm, and although we did not observe residual effects in the subjects after 2 weeks, long-term or repeated exposures to DE may lead to accumulative effects. Future studies could build upon the approach presented here to investigate similar events in airway tissues.  86  Chapter 5: Conclusion and Future Directions In this thesis we began to carefully dissect genomic and environmental characterization of DNAm. For the former, we were successful in identifying sensitive regions of the genome that demonstrated stochastic temporal variability, and which were responsive to extrinsic delays in sample processing. We also showed that DNAm variability is a signal as visible as tissue-specific methylation. Despite the GoldenGate array offering limited coverage of the genome, our results substantiated our hypothesis that the knowledge currently existing within the field will remain of limited value until these issues are addressed and solutions are implemented in research methods. With environmental characterization, we found methylation changes across the genome in response to short-term DE exposure. Interestingly, many of the CpG sites were associated with expression and cellular pathways involved in stress response, and in order to gain mechanistic insights into such association, results from this methylation study would need to be interpreted in light of expression data. Without expression data, it is difficult to say whether the methylation changes found directly impacts stress response pathways, or whether methylation could just be relics of past transcription factor binding and merely reflects gene expression changes. Nevertheless, DNAm is the sum of experiences and extrinsic factors that crosses an individual’s path throughout life, one of which is environmental compounds that interacts with the human body on a daily basis. Here we demonstrated the extensive influence short-term exposure to DE could have, and many other studies have also investigated the effects of different compounds on DNAm. But the reality is that humans are never exposed to just one chemical at any particular time. Thus, the next key step would be to study the health consequences of exposure to a multiplicity of chemicals. 87  In conclusion incorporation of epigenetics in studies of disease and phenotypes is fast gaining importance. The infinite malleability of epigenetics compliments the finite possibilities of genetics with susceptibility to programming control. As this field of study gains recognition and momentum, it becomes important to characterize the epigenome in healthy individuals and define epigenetic variability under normal conditions and in response to external stressors. Such initiatives will facilitate the establishment of rigorous research method and quality reference epigenomic data, and support research efforts towards influencing epigenomic programming.     88  Bibliography Abbott, A. (2010, February 4). Project set to map marks on genome. Nature, pp. 596–597. Al-Humadi, N. H., Siegel, P. D., Lewis, D. M., Barger, M. W., Ma, J. Y. C., Weissman, D. N., & Ma, J. K. H. (2002). Alteration of intracellular cysteine and glutathione levels in alveolar macrophages and lymphocytes by diesel exhaust particle exposure. Environmental Health Perspectives, 110(4), 349–353. Ambros, V. (2003). MicroRNA pathways in flies and worms: growth, death, fat, stress, and timing. Cell, 113(6), 673–676. Antequera, F., Bird, A. (1993). CpG Islands. EXS, 64, 169–185. Atkinson, R. W., Anderson, H. R., Sunyer, J., Ayres, J., Baccini, M., Vonk, J. M., et al. (2001). Acute effects of particulate air pollution on respiratory admissions: results from APHEA 2 project. Air Pollution and Health: a European Approach. American Journal of Respiratory and Critical Care Medicine, 164(10 Pt 1), 1860–1866. doi:10.1164/ajrccm.164.10.2010138 Baccarelli, A., & Bollati, V. (2009). Epigenetics and environmental chemicals. Current Opinion in Pediatrics, 21(2), 243–251. doi:10.1097/MOP.0b013e32832925cc Baccarelli, A., Wright, R. O., Bollati, V., Tarantini, L., Litonjua, A. A., Suh, H. H., et al. (2009). Rapid DNA methylation changes after exposure to traffic particles. American Journal of Respiratory and Critical Care Medicine, 179(7), 572–578. doi:10.1164/rccm.200807-1097OC Baccarelli, A., Wright, R., Bollati, V., Litonjua, A., Zanobetti, A., Tarantini, L., et al. (2010). Ischemic Heart Disease and Stroke in Relation to Blood DNA Methylation. Epidemiology, 21(6), 819–828. doi:10.1097/EDE.0b013e3181f20457 Baeuerle, P. A., & Henkel, T. (1994). Function and activation of NF-kappaB in the immune system. Annual Review of Immunology, 12(1), 141–179. Barnes, M. G., Grom, A. A., Griffin, T. A., Colbert, R. A., & Thompson, S. D. (2010). Gene Expression Profiles from Peripheral Blood Mononuclear Cells Are Sensitive to Short Processing Delays. Biopreservation and Biobanking, 8(3), 153–162. doi:10.1089/bio.2010.0009 Bartel, D. P. (2004). MicroRNAs: genomics, biogenesis, mechanism, and function. Cell, 116(2), 281–297. Belinsky, S. A., Palmisano, W. A., Gilliland, F. D., Crooks, L. A., Divine, K. K., Winters, S. A., et al. (2002). 89  Aberrant promoter methylation in bronchial epithelium and sputum from current and former smokers. Cancer Research, 62(8), 2370–2377. Bellavia, A., Urch, B., Speck, M., Brook, R. D., Scott, J. A., Albetti, B., et al. (2013). DNA hypomethylation, ambient particulate matter, and increased blood pressure: findings from controlled human exposure experiments. Journal of the American Heart Association, 2(3), e000212. doi:10.1161/JAHA.113.000212 Benjamini, Y. & Hochberg, Y. (1995). Controlling the false discovery rate: a practical and powerful approach to multiple testing. J Roy Stat Soc B Met 57,  289-300. Bennett, E. A., Keller, H., Mills, R. E., Schmidt, S., Moran, J. V., Weichenrieder, O., & Devine, S. E. (2008). Active Alu retrotransposons in the human genome. Genome Research, 18(12), 1875–1883. doi:10.1101/gr.081737.108 Bibikova, M. (2006). High-throughput DNA methylation profiling using universal bead arrays. Genome Research, 16(3), 383–393. doi:10.1101/gr.4410706 Bird, A. (2002). DNA methylation patterns and epigenetic memory. Genes & Development, 16(1), 6–21. doi:10.1101/gad.947102 Bird, A. P. (1980). DNA methylation and the frequency of CpG in animal DNA. Nucleic Acids Research, 8(7), 1499–1504. Bird, A. P. (1995). Gene number, noise reduction and biological complexity. Trends in Genetics : TIG, 11(3), 94–100. doi:10.1016/S0168-9525(00)89009-5 Birger, N., Gould, T., Stewart, J., Miller, M. R., Larson, T., & Carlsten, C. (2011). The Air Pollution Exposure Laboratory (APEL) for controlled human exposure to diesel exhaust and other inhalants: characterization and comparison to existing facilities. Inhalation Toxicology, 23(4), 219–225. doi:10.3109/08958378.2011.562256 Bjornsson, H. (2004). An integrated epigenetic and genetic approach to common human disease. Trends in Genetics, 20(8), 350–358. doi:10.1016/j.tig.2004.06.009 Bjornsson, H. T., Sigurdsson, M. I., Fallin, M. D., Irizarry, R. A., Aspelund, T., Cui, H., et al. (2008). Intra-individual change over time in DNA methylation with familial clustering. JAMA : the Journal of the American Medical Association, 299(24), 2877–2883. doi:10.1001/jama.299.24.2877 Bock, C. (2012). Analysing and interpreting DNA methylation data. Nature Reviews: Genetics, 13(10), 705–719. doi:10.1038/nrg3273 90  Bock, C., Walter, J., Paulsen, M., & Lengauer, T. (2008). Inter-individual variation of DNA methylation and its implications for large-scale epigenome mapping. Nucleic Acids Research, 36(10), e55–e55. doi:10.1093/nar/gkn122 Bollati, V., Baccarelli, A., Hou, L., Bonzini, M., Fustinoni, S., Cavallo, D., et al. (2007). Changes in DNA Methylation Patterns in Subjects Exposed to Low-Dose Benzene. Cancer Research, 67(3), 876–880. doi:10.1158/0008-5472.CAN-06-2995 Bollati, V., Marinelli, B., Apostoli, P., Bonzini, M., Nordio, F., Hoxha, M., et al. (2010). Exposure to metal-rich particulate matter modifies the expression of candidate microRNAs in peripheral blood leukocytes. Environmental Health Perspectives, 118(6), 763–768. doi:10.1289/ehp.0901300 Bonasio, R., Tu, S., & Reinberg, D. (2010). Molecular Signals of Epigenetic States. Science, 330(6004), 612–616. doi:10.1126/science.1191078 Bredy, T. W., Sun, Y. E., & Kobor, M. S. (2010). How the epigenome contributes to the development of psychiatric disorders. Developmental Psychobiology, 52(4), 331–342. doi:10.1002/dev.20424 Breitling, L. P., Yang, R., Korn, B., Burwinkel, B., & Brenner, H. (2011). AR TICLETobacco-Smoking-Related Differential DNA Methylation: 27K Discovery and Replication. The American Journal of Human Genetics, 88(4), 450–457. doi:10.1016/j.ajhg.2011.03.003 Brook, R. D. (2008). Cardiovascular effects of air pollution. Clinical Science (London, England : 1979), 115(6), 175–187. doi:10.1042/CS20070444 Brook, R. D., Franklin, B., Cascio, W., Hong, Y., Howard, G., Lipsett, M., et al. (2004). Air pollution and cardiovascular disease: a statement for healthcare professionals from the Expert Panel on Population and Prevention Science of the American Heart Association. (Vol. 109, pp. 2655–2671). Presented at the Circulation. doi:10.1161/01.CIR.0000128587.30041.C8 Byrne, E. M., Carrillo-Roa, T., Henders, A. K., Bowdler, L., McRae, A. F., Heath, A. C., et al. (2013). Monozygotic twins affected with major depressive disorder have greater variance in methylation than their unaffected co-twin, 3(6), e269–6. doi:10.1038/tp.2013.45 Byun, H. M., Siegmund, K. D., Pan, F., Weisenberger, D. J., Kanel, G., Laird, P. W., & Yang, A. S. (2009). Epigenetic profiling of somatic tissues from human autopsy specimens identifies tissue- and individual-specific 91  DNA methylation patterns. Human Molecular Genetics, 18(24), 4808–4817. doi:10.1093/hmg/ddp445 Byun, H.-M., Nordio, F., Coull, B. A., Tarantini, L., Hou, L., Bonzini, M., Apostoli, P., Bertazzi, P. A., & Baccarelli, A. (2012a). Temporal stability of epigenetic markers: sequence characteristics and predictors of short-term DNA methylation variations. PLoS ONE, 7(6), e39220. doi:10.1371/journal.pone.0039220.s009 Carlsten, C., MacNutt, M. J., Zhang, Z., Sava, F., & Pui, M. M. (2014). Anti-Oxidant N-Acetylcysteine Diminishes Diesel Exhaust-Induced Increased Airway Responsiveness in Person with Airway Hyper-Reactivity. Toxicological Sciences, 139(2), 479–487. doi:10.1093/toxsci/kfu040 Carlsten, C., Oron, A. P., Curtiss, H., Jarvis, S., Daniell, W., & Kaufman, J. D. (2013). Symptoms in response to controlled diesel exhaust more closely reflect exposure perception than true exposure. PLoS ONE, 8(12), e83573. doi:10.1371/journal.pone.0083573 Carrington, J. C., & Ambros, V. (2003). Role of microRNAs in plant and animal development. Science, 301(5631), 336–338. Cash, H. L., Tao, L., Yuan, J.-M., Marsit, C. J., Houseman, E. A., Xiang, Y.-B., et al. (2011). LINE-1 hypomethylation is associated with bladder cancer risk among nonsmoking Chinese. International Journal of Cancer. Journal International Du Cancer, 130(5), 1151–1159. doi:10.1002/ijc.26098 Chan, K.-K., Shen, L., Au, W.-Y., Yuen, H.-F., Wong, K.-Y., Guo, T., et al. (2010). Interleukin-2 induces NF-kappaB activation through BCL10 and affects its subcellular localization in natural killer lymphoma cells. The Journal of Pathology, 221(2), 164–174. doi:10.1002/path.2699 Cheng, Y., Liu, X., Zhang, S., Lin, Y., Yang, J., & Zhang, C. (2009). MicroRNA-21 protects against the H(2)O(2)-induced injury on cardiac myocytes via its target gene PDCD4. Journal of Molecular and Cellular Cardiology, 47(1), 5–14. doi:10.1016/j.yjmcc.2009.01.008 Cho, N.-Y., Kim, B.-H., Choi, M., Yoo, E. J., Moon, K. C., Cho, Y.-M., et al. (2007). Hypermethylation of CpG island loci and hypomethylation of LINE-1 and Alu repeats in prostate adenocarcinoma and their relationship to clinicopathological features. The Journal of Pathology, 211(3), 269–277. doi:10.1002/path.2106 Choi, J. Y., James, S. R., Link, P. A., McCann, S. E., Hong, C. C., Davis, W., et al. (2009a). Association between global DNA hypomethylation in leukocytes and risk of breast cancer. Carcinogenesis, 30(11), 1889–1897. doi:10.1093/carcin/bgp143 92  Choi, S. H., Worswick, S., Byun, H.-M., Shear, T., Soussa, J. C., Wolff, E. M., et al. (2009b). Changes in DNA methylation of tandem DNA repeats are different from interspersed repeats in cancer. International Journal of Cancer. Journal International Du Cancer, 125(3), 723–729. doi:10.1002/ijc.24384 Chouliaras, L., Rutten, B. P. F., Kenis, G., Peerbooms, O., Visser, P. J., Verhey, F., et al. (2010). Progress in Neurobiology. Progress in Neurobiology, 90(4), 498–510. doi:10.1016/j.pneurobio.2010.01.002 Chow, J. C., Watson, J. G., Mauderly, J. L., Costa, D. L., Wyzga, R. E., Vedal, S., et al. (2006). Health effects of fine particulate air pollution: lines that connect. Journal of the Air & Waste Management Association (1995), 56(10), 1368–1380. Cooney, C. A., Dave, A. A., & Wolff, G. L. (2002). Maternal methyl supplements in mice affect epigenetic variation and DNA methylation of offspring. The Journal of Nutrition, 132(8 Suppl), 2393S–2400S. Costello, J. F., & Plass, C. (2001). Methylation matters. Journal of Medical Genetics, 38(5), 285–303. Das, P. M., & Singal, R. (2004). DNA methylation and cancer. Journal of Clinical Oncology : Official Journal of the American Society of Clinical Oncology, 22(22), 4632–4642. doi:10.1200/JCO.2004.07.151 Davies, M. N., Volta, M., Pidsley, R., Lunnon, K., Dixit, A., Lovestone, S., et al. (2012). Functional annotation of the human brainmethylome identifies tissue-specific epigeneticvariation across brain and blood. Genome Biology, 13(6), R43. doi:10.1186/gb-2012-13-6-r43 de Zoeten, E. F., Wang, L., Sai, H., Dillmann, W. H., & Hancock, W. W. (2010). Inhibition of HDAC9 increases T regulatory cell function and prevents colitis in mice. Gastroenterology, 138(2), 583–594. doi:10.1053/j.gastro.2009.10.037 Dabney A., Storey J.D., Warnes G.R.: qvalue: Q-value estimation for false discovery rate control. R package version 1.34.0. Denis, H., Ndlovu, '. N., & Fuks, F. (2011). Regulation of mammalian DNA methyltransferases: a route to new mechanisms. Nature Publishing Group, 12(7), 647–656. doi:10.1038/embor.2011.110 Diakiw, S. M., Perugini, M., Kok, C. H., Engler, G. A., Cummings, N., To, L. B., et al. (2013). Methylation of KLF5 contributes to reduced expression in acute myeloid leukaemia and is associated with poor overall survival. British Journal of Haematology, 161(6), 884–888. doi:10.1111/bjh.12295 Diaz-Sanchez, D., Proietti, L., & Polosa, R. (2003). Diesel fumes and the rising prevalence of atopy: an urban 93  legend? Current Allergy and Asthma Reports, 3(2), 146–152. Dockery, D. W., Pope, C. A., Xu, X., Spengler, J. D., Ware, J. H., Fay, M. E., et al. (1993). An association between air pollution and mortality in six U.S. cities. The New England Journal of Medicine, 329(24), 1753–1759. doi:10.1056/NEJM199312093292401 Donaldson, K., Tran, L., Jimenez, L. A., Duffin, R., Newby, D. E., Mills, N., et al. (2005). Combustion-derived nanoparticles: a review of their toxicology following inhalation exposure. Particle and Fibre Toxicology, 2, 10. doi:10.1186/1743-8977-2-10 Duvigneau, J. C., Hartl, R. T., Teinfalt, M., & Gemeiner, M. (2003). Delay in processing porcine whole blood affects cytokine expression. Journal of Immunological Methods, 272(1-2), 11–21. Eckhardt, F., Lewin, J., Cortese, R., Rakyan, V. K., Attwood, J., Burger, M., et al. (2006). DNA methylation profiling of human chromosomes 6, 20 and 22. Nature Genetics, 38(12), 1378–1385. doi:10.1038/ng1909 Ehrlich, M. (2002). DNA methylation in cancer: too much, but also too little. Oncogene, 21(35), 5400–5413. doi:10.1038/sj.onc Feil, R., & Fraga, M. F. (2011). Epigenetics and the environment: emerging patterns and implications. Nature Publishing Group, 13(2), 97–109. doi:10.1038/nrg3142 Feinberg, A. P. (2004). The epigenetics of cancer etiology. Seminars in Cancer Biology, 14(6), 427–432. doi:10.1016/j.semcancer.2004.06.005 Feinberg, A. P. (2007). Phenotypic plasticity and the epigenetics of human disease. Nature, 447(7143), 433–440. doi:10.1038/nature05919 Feinberg, A. P., & Tycko, B. (2004). The history of cancer epigenetics. Nature Reviews Cancer, 4(2), 143–153. Fernandez, A. F., Assenov, Y., Martin-Subero, J. I., Balint, B., Siebert, R., Taniguchi, H., et al. (2012). A DNA methylation fingerprint of 1628 human samples. Genome Research, 22(2), 407–419. doi:10.1101/gr.119867.110 Fraser, H. B., Lam, L. L., Neumann, S. M., & Kobor, M. S. (2012). Population-specificity of human DNA methylation. Genome Biology, 13(2), R8. doi:10.1186/gb-2012-13-2-r8 Friedman, M. S., Powell, K. E., Hutwagner, L., Graham, L. M., & Teague, W. G. (2001). Impact of changes in transportation and commuting behaviors during the 1996 Summer Olympic Games in Atlanta on air quality and 94  childhood asthma. JAMA : the Journal of the American Medical Association, 285(7), 897–905. Gaudet, F., Hodgson, J. G., Eden, A., Jackson-Grusby, L., Dausman, J., Gray, J. W., et al. (2003). Induction of tumors in mice by genomic hypomethylation. Science, 300(5618), 489–492. doi:10.1126/science.1083558 Gluckman, P. D., Hanson, M. A., Buklijas, T., Low, F. M., & Beedle, A. S. (2009). Epigenetic mechanisms that underpin metabolic and cardiovascular diseases. Nature Reviews. Endocrinology, 5(7), 401–408. doi:10.1038/nrendo.2009.102 Guo, Y. L., Lin, Y. C., Sung, F. C., Huang, S. L., Ko, Y. C., Lai, J. S., et al. (1999). Climate, traffic-related air pollutants, and asthma prevalence in middle-school children in taiwan. Environmental Health Perspectives, 107(12), 1001–1006. Gutierrez-Arcelus, M., Lappalainen, T., Montgomery, S. B., Buil, A., Ongen, H., Yurovsky, A., et al. (2013). Passive and active DNA methylation and the interplay with genetic variation in gene regulation. eLife, 2, e00523. doi:10.7554/eLife.00523 Hannum, G., Guinney, J., Zhao, L., Zhang, L., Hughes, G., Sadda, S., et al. (2013). Genome-wide Methylation Profiles Reveal Quantitative Views of Human Aging Rates. Molecular Cell, 49(2), 359–367. doi:10.1016/j.molcel.2012.10.016 Hashimoto, S., Gon, Y., Takeshita, I., Matsumoto, K., Jibiki, I., Takizawa, H., et al. (2000). Diesel exhaust particles activate p38 MAP kinase to produce interleukin 8 and RANTES by human bronchial epithelial cells and N-acetylcysteine attenuates p38 MAP kinase activation. American Journal of Respiratory and Critical Care Medicine, 161(1), 280–285. Hatchwell, E., & Greally, J. M. (2007). The potential role of epigenomic dysregulation in complex human disease. Trends in Genetics, 23(11), 588–595. doi:10.1016/j.tig.2007.08.010 He, L., & Hannon, G. J. (2004). MicroRNAs: small RNAs with a big role in gene regulation. Nature Reviews: Genetics, 5(7), 522–531. doi:10.1038/nrg1379 Heijmans, B. T., & Mill, J. (2012). Commentary: The seven plagues of epigenetic epidemiology. International Journal of Epidemiology, 41(1), 74–78. doi:10.1093/ije/dyr225 Heyn, H. (2014). A symbiotic liaison between the genetic and epigenetic code, 1–6. doi:10.3389/fgene.2014.00113/abstract 95  Heyn, H., & Esteller, M. (2012). DNA methylation profiling in the clinic: applications and challenges. Nature Reviews: Genetics, 13(10), 679–692. doi:10.1038/nrg3270 Heyn, H., Moran, S., Hernando-Herraez, I., Sayols, S., Gomez, A., Sandoval, J., et al. (2013). DNA methylation contributes to natural human variation. Genome Research, 23(9), 1363–1372. doi:10.1101/gr.154187.112 Hiura, T. S., Kaszubowski, M. P., Li, N., & Nel, A. E. (1999). Chemicals in diesel exhaust particles generate reactive oxygen radicals and induce apoptosis in macrophages. The Journal of Immunology, 163(10), 5582–5591. Hoffmann, B., Moebus, S., Mohlenkamp, S., Stang, A., Lehmann, N., Dragano, N., et al. (2007). Residential exposure to traffic is associated with coronary atherosclerosis. Circulation, 116(5), 489–496. doi:10.1161/CIRCULATIONAHA.107.693622 Holloway, J. W., Savarimuthu F.S., Fong, K. M., & Yang, I. A. (2012). Genomics and the respiratory effects of air pollution exposure. Respirology, 17(4), 590–600. doi:10.1111/j.1440-1843.2012.02164.x Horvath, S., Zhang, Y., Langfelder, P., Kahn, R. S., Boks, M. P., van Eijk, K., et al. (2012). Aging effects on DNA methylation modules in human brain and blood tissue. Genome Biology, 13(10), R97. doi:10.1186/gb-2012-13-10-r97 Hou, L., Zhang, X., Tarantini, L., Nordio, F., Bonzini, M., Angelici, L., et al. (2011). Ambient PM exposure and DNA methylation in tumor suppressor genes: a cross-sectional study. Particle and Fibre Toxicology, 8, 25. doi:10.1186/1743-8977-8-25 Hou, L., Zhang, X., Zheng, Y., Wang, S., Dou, C., Guo, L., et al. (2014). Altered methylation in tandem repeat element and elemental component levels in inhalable air particles. Environmental and Molecular Mutagenesis, 55(3), 256–265. doi:10.1002/em.21829 Huang, D. W., Sherman, B. T., & Lempicki, R. A. (2009). Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nature Protocols, 4(1), 44–57. doi:10.1038/nprot.2008.211 Illingworth, R. S., & Bird, A. P. (2009). CpG islands – “A rough guide.” FEBS Letters, 583(11), 1713–1720. doi:10.1016/j.febslet.2009.04.012 Illumin Inc. (2010). GoldenGate® Assay for Methylation and BeadArrayTM Technology. Illumina Technical Note. Illumin Inc. (2010-2012). DNA Methylation Analysis. Illumina Data Sheet: Epigenetics. 96  Illumin Inc. (2012). Infinium® HumanMethylation450 BeadChip. Illumina Data Sheet: Epigenetics. Johnston, R. J., & Hobert, O. (2003). A microRNA controlling left/right neuronal asymmetry in Caenorhabditis elegans. Nature, 426(6968), 845–849. doi:10.1038/nature02255 Jones, M. J., Farré, P., McEwen, L. M., Macisaac, J. L., Watt, K., Neumann, S. M., et al. (2013). Distinct DNA methylation patterns of cognitive impairment and trisomy 21 in Down syndrome. BMC Medical Genomics, 6, 58. doi:10.1186/1755-8794-6-58 Jones, P. A. (2012). Functions of DNA methylation: islands, start sites, gene bodies and beyond. Nature Reviews: Genetics, 13(7), 484–492. doi:10.1038/nrg3230 Juniper, E. F., O'Byrne, P. M., Guyatt, G. H., Ferrie, P. J., & King, D. R. (1999). Development and validation of a questionnaire to measure asthma control. The European Respiratory Journal, 14(4), 902–907. Jürgens, B., Schmitz-Dräger, B. J., & Schulz, W. A. (1996). Hypomethylation of L1 LINE sequences prevailing in human urothelial carcinoma. Cancer Research, 56(24), 5698–5703. Kim, M., Long, T. I., Arakawa, K., Wang, R., Yu, M. C., & Laird, P. W. (2010). DNA Methylation as a Biomarker for Cardiovascular Disease Risk. PLoS ONE, 5(3), e9692. doi:10.1371/journal.pone.0009692.s002 Klot, S. V., Wolke, G., Tuch, T., Heinrich, J., Dockery, D. W., Schwartz, J., et al. (2002). Increased asthma medication use in association with ambient fine and ultrafine particles. European Respiratory Journal, 20(3), 691–702. doi:10.1183/09031936.02.01402001 Kota, S. K., & Feil, R. (2010). Epigenetic Transitionsin Germ Cell Development and Meiosis. Developmental Cell, 19(5), 675–686. doi:10.1016/j.devcel.2010.10.009 Kuang, S. Q., Bai, H., Fang, Z. H., Lopez, G., Yang, H., Tong, W., et al. (2010). Aberrant DNA methylation and epigenetic inactivation of Eph receptor tyrosine kinases and ephrin ligands in acute lymphoblastic leukemia. Blood, 115(12), 2412–2419. doi:10.1182/blood-2009-05-222208 Kulis, M., & Esteller, M. (2010). DNA Methylation and Cancer. Epigenetics and Cancer, Part A (1st ed., Vol. 70, pp. 27–56). Elsevier Inc. doi:10.1016/B978-0-12-380866-0.60002-2 Kumagai, Y., Arimoto, T., Shinyashiki, M., Shimojo, N., Nakai, Y., Yoshikawa, T., & Sagai, M. (1997). Generation of reactive oxygen species during interaction of diesel exhaust particle components with NADPH-cytochrome P450 reductase and involvement of the bioactivation in the DNA damage. Free Radical Biology and Medicine, 97  22(3), 479–487. Laden, F., Neas, L. M., Dockery, D. W., & Schwartz, J. (2000). Association of fine particulate matter from different sources with daily mortality in six U.S. cities. Environmental Health Perspectives, 108(10), 941–947. Lai, E. C. (2003). microRNAs: Runts of the Genome Assert Themselves. Current Biology, 13(23), R925–R936. doi:10.1016/j.cub.2003.11.017 Laird, P. W. (2010). Principles and challenges of genome- wide DNA methylation analysis. Nature Reviews: Genetics, 11(3), 191–203. doi:10.1038/nrg2732 Lam, L. L., Emberly, E., Fraser, H. B., Neumann, S. M., Chen, E., Miller, G. E., & Kobor, M. S. (2012). Factors underlying variable DNA methylation in a human community cohort. Proceedings of the National Academy of Sciences of the United States of America, 109(Supplement 2), 17253–17260. doi:10.1073/pnas.1121249109/-/DCSupplemental Larsen, F., Gundersen, G., Lopez, R., & Prydz, H. (1992). CpG islands as gene markers in the human genome. Genomics, 13(4), 1095–1107. Law, J. A., & Jacobsen, S. E. (2010). Establishing, maintaining and modifying DNA methylation patterns in plants and animals. Nature Publishing Group, 11(3), 204–220. doi:10.1038/nrg2719 Lepeule, J., Laden, F., Dockery, D., & Schwartz, J. (2012). Chronic exposure to fine particles and mortality: an extended follow-up of the Harvard Six Cities study from 1974 to 2009. Environmental Health Perspectives, 120(7), 965–970. doi:10.1289/ehp.1104660 Leven, H. (1960) [Robust tests for equality of variances] Contributions to Probability and Statistics [Olkin, L., Ghurye, S.G., Hoeffiding, W., Madow, W.G. & Mann, H.B.] [278-292][Stanford University Press] Li, N., Kim, S., Wang, M., Froines, J., Sioutas, C., & Nel, A. (2002). Use of a stratified oxidative stress model to study the biological effects of ambient concentrated and diesel exhaust particulate matter. Inhalation Toxicology, 14(5), 459–486. doi:10.1080/089583701753678571 Li, Y., Zhu, J., Tian, G., Li, N., Li, Q., Ye, M., et al. (2010). The DNA methylome of human peripheral blood mononuclear cells. PLoS Biology, 8(11), e1000533. doi:10.1371/journal.pbio.1000533 Lister, R., Mukamel, E. A., Nery, J. R., Urich, M., Puddifoot, C. A., Johnson, N. D., et al. (2013). Global Epigenomic Reconfiguration During Mammalian Brain Development. Science, 341(6146), 1237905–1237905. 98  doi:10.1126/science.1237905 Lister, R., Pelizzola, M., Dowen, R. H., Hawkins, R. D., Hon, G., Tonti-Filippini, J., et al. (2009). Human DNA methylomes at base resolution show widespread epigenomic differences. Nature, 462(7271), 315–322. doi:10.1038/nature08514 Liu, Y., Aryee, M. J., Padyukov, L., Fallin, M. D., Hesselberg, E., Runarsson, A., et al. (2013). epigenome-wide association data implicate DNA methylation as an intermediary of genetic risk in rheumatoid arthritis. Nature Biotechnology, 31(2), 142–147. doi:10.1038/nbt.2487 Logan, W. P. D. (1953). Mortality in the London fog incident, 1952. Lancet, 1(6755), 336–338. Lowe, R., Gemma, C., Beyan, H., Hawa, M. I., Bazeos, A., Leslie, R. D., et al. (2013). Buccals are likely to be a more informative surrogate tissue than blood for epigenome-wide association studies. Epigenetics, 8(4), 445–454. doi:10.4161/epi.24362 Lu, T. X., Munitz, A., & Rothenberg, M. E. (2009). MicroRNA-21 is up-regulated in allergic airway inflammation and regulates IL-12p35 expression. Journal of Immunology (Baltimore, Md. : 1950), 182(8), 4994–5002. doi:10.4049/jimmunol.0803560 Lucchinetti, E., Feng, J., Silva, R. D., Tolstonog, G. V., Schaub, M. C., Schumann, G. G., & Zaugg, M. (2006). Inhibition of LINE-1 expression in the heart decreases ischemic damage by activation of Akt/PKB signaling. Physiological Genomics, 25(2), 314–324. doi:10.1152/physiolgenomics.00251.2005 Lunyak, V. V., Prefontaine, G. G., Núñez, E., Cramer, T., Ju, B.-G., Ohgi, K. A., et al. (2007). Developmentally regulated activation of a SINE B2 repeat as a domain boundary in organogenesis. Science, 317(5835), 248–251. doi:10.1126/science.1140871 MacIntyre, E. A., Brauer, M., Melén, E., Bauer, C. P., Bauer, M., Berdel, D., et al. (2014). GSTP1 and TNF Gene variants and associations between air pollution and incident childhood asthma: the traffic, asthma and genetics (TAG) study. Environmental Health Perspectives, 122(4), 418–424. doi:10.1289/ehp.1307459 Madrigano, J., Baccarelli, A., Mittleman, M. A., Wright, R. O., Sparrow, D., Vokonas, P. S., et al. (2011). Prolonged Exposure to Particulate Pollution, Genes Associated with Glutathione Pathways, and DNA Methylation in a Cohort of Older Men. Environmental Health Perspectives, 119(7), 977–982. doi:10.1289/ehp.1002773 Mazzoli-Rocha, F., Fernandes, S., Einicker-Lamas, M., & Zin, W. A. (2010). Roles of oxidative stress in signaling 99  and inflammation induced by particulate matter. Cell Biology and Toxicology, 26(5), 481–498. doi:10.1007/s10565-010-9158-2 McClay, J. L., Aberg, K. A., Clark, S. L., Nerella, S., Kumar, G., Xie, L. Y., et al. (2014). A methylome-wide study of aging using massively parallel sequencing of the methyl-CpG-enriched genomic fraction from blood in over 700 subjects. Human Molecular Genetics, 23(5), 1175–1185. doi:10.1093/hmg/ddt511 McManus, M. T. (2003). MicroRNAs and cancer. Seminars in Cancer Biology, 13(4), 253–258. doi:10.1016/S1044-579X(03)00038-5 Michels, K. B., Binder, A. M., Dedeurwaerder, S., Epstein, C. B., Greally, J. M., Gut, I., et al. (2013). Recommendations for the design and analysis of epigenome-wide association studies. Nature Methods, 10(10), 949–955. doi:10.1038/nmeth.2632 Mill, J., & Heijmans, B. T. (2013). From promise to practical strategies in epigenetic epidemiology. Nature Rev Genet, 14(8), 585–594. doi:10.1038/nrg3405 Mill, J., Tang, T., Kaminsky, Z., Khare, T., Yazdanpanah, S., Bouchard, L., et al. (2008). Epigenomic Profiling Reveals DNA-Methylation Changes Associated with Major Psychosis. The American Journal of Human Genetics, 82(3), 696–711. doi:10.1016/j.ajhg.2008.01.008 Miller, D. P., De Vivo, I., Neuberg, D., Wain, J. C., Lynch, T. J., Su, L., & Christiani, D. C. (2003). Association between self-reported environmental tobacco smoke exposure and lung cancer: modification by GSTP1 polymorphism. International Journal of Cancer. Journal International Du Cancer, 104(6), 758–763. doi:10.1002/ijc.10989 Miousse, I. R., Chalbot, M.-C. G., Aykin-Burns, N., Wang, X., Basnakian, A., Kavouras, I. G., & Koturbash, I. (2014). Epigenetic alterations induced by ambient particulate matter in mouse macrophages. Environmental and Molecular Mutagenesis, 55(5), 428–435. doi:10.1002/em.21855 Moen, E. L., Zhang, X., Mu, W., Delaney, S. M., Wing, C., McQuade, J., et al. (2013). Genome-wide variation of cytosine modifications between European and African populations and the implications for complex traits. Genetics, 194(4), 987–996. doi:10.1534/genetics.113.151381/-/DC1 Morales, E., Bustamante, M., Vilahur, N., Escaramis, G., Montfort, M., de Cid, R., et al. (2012). DNA Hypomethylation at ALOX12Is Associated with Persistent Wheezing in Childhood. American Journal of 100  Respiratory and Critical Care Medicine, 185(9), 937–943. doi:10.1164/rccm.201105-0870OC Oglesby, I. K., McElvaney, N. G., & Greene, C. M. (2010). MicroRNAs in inflammatory lung disease - master regulators or target practice? Respiratory Research, 11(1), 148. doi:10.1186/1465-9921-11-148 Okano, M., Bell, D. W., Haber, D. A., & Li, E. (1999). DNA methyltransferases Dnmt3a and Dnmt3b are essential for de novo methylation and mammalian development. Cell, 99(3), 247–257. Paravicini, T., & Touyz, R. (2006). Redox signaling in hypertension. Cardiovascular Research, 71(2), 247–258. doi:10.1016/j.cardiores.2006.05.001 Pavanello, S., Bollati, V., Pesatori, A. C., Kapka, L., Bolognesi, C., Bertazzi, P. A., & Baccarelli, A. (2009). Global and gene-specific promoter methylation changes are related to anti-B[ a]PDE-DNA adduct levels and influence micronuclei levels in polycyclic aromatic hydrocarbon-exposed individuals. International Journal of Cancer. Journal International Du Cancer, 125(7), 1692–1697. doi:10.1002/ijc.24492 Peng, R. D., Bell, M. L., Geyh, A. S., McDermott, A., Zeger, S. L., Samet, J. M., & Dominici, F. (2009). Emergency admissions for cardiovascular and respiratory diseases and the chemical composition of fine particle air pollution. Environmental Health Perspectives, 117(6), 957–963. doi:10.1289/ehp.0800185 PhD, K. J. D., PhD, C. P. N., PhD, L. T., PhD, J. K. S., MSc, D. A. S., MSc, S. W., et al. (2014). DNA methylation and body-mass index: a genome-wideanalysis. Lancet, 383(9933), 1990–1998. doi:10.1016/S0140-6736(13)62674-4 Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D. (2014): R Core Team: nlme: Linear and nonlinear mixed effects models. R package version 3.1-115, URL: http://CRAN.R-project.org/package=nlme Portela, A., & Esteller, M. (2010). Epigenetic modifications and human disease. Nature Biotechnology, 28(10), 1057–1068. doi:10.1038/nbt.1685 Price, M. E., Cotton, A. M., Lam, L. L., Farré, P., Emberly, E., Brown, C. J., et al. (2013). Additional annotation enhances potential for biologically-relevant analysis of the Illumina Infinium HumanMethylation450 BeadChip array. Epigenetics & Chromatin, 6(1), 4. doi:10.1186/1756-8935-6-4 Pufulete, M., Al-Ghnaniem, R., Leather, A. J. M., Appleby, P., Gout, S., Terry, C., et al. (2003). Folate status, genomic DNA hypomethylation, and risk of colorectal adenoma and cancer: a case control study. Gastroenterology, 124(5), 1240–1248. doi:10.1016/S0016-5085(03)00279-8 101  Rakyan, V. K., Down, T. A., Balding, D. J., & Beck, S. (2011). Epigenome-wide association studies for common human diseases. Nature Reviews: Genetics, 12(8), 529–541. doi:10.1038/nrg3000 Reik, W. (2007). Stability and flexibility of epigenetic gene regulation in mammalian development. Nature, 447(7143), 425–432. doi:10.1038/nature05918 Robertson, K. D. (2002). DNA methylation and chromatin - unraveling the tangled web. Oncogene, 21(35), 5361–5379. doi:10.1038/sj.onc.1205609 Romieu, I., Ramirez-Aguilar, M., Sienra-Monge, J. J., Moreno-Macias, H., del Rio-Navarro, B. E., David, G., et al. (2006). GSTM1 and GSTP1 and respiratory health in asthmatic children exposed to ozone. The European Respiratory Journal, 28(5), 953–959. doi:10.1183/09031936.06.00114905 Rosenfeld, C. S. (2010). Animal Models to Study Environmental Epigenetics. Biology of Reproduction, 82(3), 473–488. doi:10.1095/biolreprod.109.080952 R Core Team (2012). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org/ Sava, F., & Carlsten, C. (2012). Respiratory health effects of ambient air pollution: an update. Clinics in Chest Medicine, 33(4), 759–769. doi:10.1016/j.ccm.2012.07.003 Saxonov, S., Berg, P., & Brutlag, D. L. (2006). A genome-wide analysis of CpG dinucleotides in the human genome distinguishes two distinct classes of promoters. Proceedings of the National Academy of Sciences of the United States of America, 103(5), 1412–1417. doi:10.1073/pnas.0510310103 Schneider, E., Pliushch, G., Hajj, El, N., Galetzka, D., Puhl, A., Schorsch, M., et al. (2010). Spatial, temporal and interindividual epigenetic variation of functionally important DNA methylation patterns. Nucleic Acids Research, 38(12), 3880–3890. doi:10.1093/nar/gkq126 Shenderov, B. A., & Midtvedt, T. (2014). Epigenomic programing: a future way to health? Microbial Ecology in Health & Disease, 25(0), 1157. doi:10.1016/j.cell.2011.02.013 Shivapurkar, N., Takahashi, T., Reddy, J., Zheng, Y., Stastny, V., Collins, R., et al. (2004). Presence of simian virus 40 DNA sequences in human lymphoid and hematopoietic malignancies and their relationship to aberrant promoter methylation of multiple genes. Cancer Research, 64(11), 3757–3760. doi:10.1158/0008-5472.CAN-03-3307 102  Small, E. M., & Olson, E. N. (2011). Pervasive roles of microRNAs in cardiovascular biology. Nature, 469(7330), 336–342. doi:10.1038/nature09783 Smolarek, I., Wyszko, E., Barciszewska, A. M., Nowak, S., Gawronska, I., Jablecka, A., & Barciszewska, M. Z. (2010). Global DNA methylation changes in blood of patients with essential hypertension. Medical Science Monitor : International Medical Journal of Experimental and Clinical Research, 16(3), CR149–155. Takizawa, H., Ohtoshi, T., Kawasaki, S., Abe, S., Sugawara, I., Nakahara, K., et al. (2000). Diesel exhaust particles activate human bronchial epithelial cells to express inflammatory mediators in the airways: a review. Respirology, 5(2), 197–203. Takizawa, H., Ohtoshi, T., Kawasaki, S., Kohyama, T., Desaki, M., Kasama, T., et al. (1999). Diesel exhaust particles induce NF-κB activation in human bronchial epithelial cells in vitro: importance in cytokine transcription. The Journal of Immunology, 162(8), 4705–4711. Talens, R. P., Boomsma, D. I., Tobi, E. W., Kremer, D., Jukema, J. W., Willemsen, G., et al. (2010). Variation, patterns, and temporal stability of DNA methylation: considerations for epigenetic epidemiology. The FASEB Journal, 24(9), 3135–3144. doi:10.1096/fj.09-150490 Tamer, L., Calikoğlu, M., Ates, N. A., Yildirim, H., Ercan, B., Saritas, E., et al. (2004). Glutathione‐ S‐ transferase gene polymorphisms (GSTT1, GSTM1, GSTP1) as increased risk factors for asthma. Respirology, 9(4), 493–498. Tarantini, L., Bonzini, M., Apostoli, P., Pegoraro, V., Bollati, V., Marinelli, B., et al. (2009). Effects of particulate matter on genomic DNA methylation content and iNOS promoter methylation. Environmental Health Perspectives, 117(2), 217–222. doi:10.1289/ehp.11898 Terry, M. B., Delgado-Cruzata, L., Vin-Raviv, N., Wu, H. C., & Santella, R. M. (2011). DNA methylation in white blood cells: Association with risk factors in epidemiologic studies. Epigenetics, 6(7), 828–837. doi:10.4161/epi.6.7.16500 Teschendorff, A. E., Jones, A., Fiegl, H., Sargent, A., Zhuang, J. J., Kitchener, H. C., & Widschwendter, M. (2012). Epigenetic variability in cells of normal cytologyis associated with the risk of futuremorphological transformation. Genome Medicine, 4(3), 24. doi:10.1186/gm323 Teschendorff, A. E., West, J., & Beck, S. (2013). Age-associated epigenetic drift: implications, and a case of 103  epigenetic thrift? Human Molecular Genetics, 22(R1), R7–R15. doi:10.1093/hmg/ddt375 Thai, H.-T., Mentré, F., Holford, N. H. G., Veyrat-Follet, C., & Comets, E. (2013). A comparison of bootstrap approaches for estimating uncertainty of parameters in linear mixed-effects models. Pharmaceutical Statistics, 12(3), 129–140. doi:10.1002/pst.1561 Tsien, A., Diaz-Sanchez, D., Ma, J., & Saxon, A. (1997). The organic component of diesel exhaust particles and phenanthrene, a major polyaromatic hydrocarbon constituent, enhances IgE production by IgE-secreting EBV-transformed human B cells in vitro. Toxicology and Applied Pharmacology, 142(2), 256–263. doi:10.1006/taap.1996.8063 Valinluck, V., Tsai, H.-H., Rogstad, D. K., Burdzy, A., Bird, A., & Sowers, L. C. (2004). Oxidative damage to methyl-CpG sequences inhibits the binding of the methyl-CpG binding domain (MBD) of methyl-CpG binding protein 2 (MeCP2). Nucleic Acids Research, 32(14), 4100–4108. doi:10.1093/nar/gkh739 van Eijk, K. R., de Jong, S., Boks, M. P., Langeveld, T., Colas, F., Veldink, J. H., et al. (2012). Genetic analysis of DNA methylation and gene expression levels in whole blood of healthy human subjects. BMC Genomics, 13(1), 1–1. doi:10.1186/1471-2164-13-636 Vineis, P., & Husgafvel-Pursiainen, K. (2005). Air pollution and cancer: biomarker studies in human populations. Carcinogenesis, 26(11), 1846–1855. doi:10.1093/carcin/bgi216 Volkmar, M., Dedeurwaerder, S., Cunha, D.A., Ndlovu, M.N., Defrance, M., Deplus, R., Calonne, E., Volkmar, U., Esteve, M.I., Naamane, N., Guerra, S.D., Masini, M., Bugliani, M., Marchetti, P., Cnop, M., Eizirik, D.L., Fuks, F. (2012). DNA methylation profiling identifies epigenetic dysregulation in pancreatic islets from type 2 diabetic patients. EMBO J, 31(6), 1405-1426. doi:10.1038/emboj.2011.503 Wang, X., Fan, J., Liu, D., Fu, S., Ingvarsson, S., & Chen, H. (2011). Spreading of Alu methylation to the promoter of the MLH1 gene in gastrointestinal cancer. PLoS ONE, 6(10), e25913. doi:10.1371/journal.pone.0025913.s001 Waterland, R. A., & Jirtle, R. L. (2003). Transposable Elements: Targets for Early Nutritional Effects on Epigenetic Gene Regulation. Molecular and Cellular Biology, 23(15), 5293–5300. doi:10.1128/MCB.23.15.5293-5300.2003 Weber, M., Davies, J. J., Wittig, D., Oakeley, E. J., Haase, M., Lam, W. L., & Schübeler, D. (2005). Chromosome-104  wide and promoter-specific analyses identify sites of differential DNA methylation in normal and transformed human cells. Nature Genetics, 37(8), 853–862. doi:10.1038/ng1598 Weber, M., Hellmann, I., Stadler, M. B., Ramos, L., Pääbo, S., Rebhan, M., & Schübeler, D. (2007). Distribution, silencing potential and evolutionary impact of promoter DNA methylation in the human genome. Nature Genetics, 39(4), 457–466. doi:10.1038/ng1990 Weidner, C. I., Lin, Q., Koch, C. M., Eisele, L., Beier, F., Ziegler, P., et al. (2014). Aging of blood can be tracked by DNA methylation changes at just three CpG sites, 15(2), 1–11. doi:10.1186/gb-2014-15-2-r24 Wong, C. C., Caspi, A., Williams, B., Craig, I. W., Houts, R., Ambler, A., et al. (2010). A longitudinal study of epigenetic variation in twins. Epigenetics, 5(6), 516–526. doi:10.4161/epi.5.6.12226 Xu, X., Su, S., Barnes, V. A., De Miguel, C., Pollock, J., Ownby, D., et al. (2013). A genome-wide methylation study on obesity: Differential variability and differential methylation. Epigenetics, 8(5), 522–533. doi:10.4161/epi.24506 Yamamoto, M., Singh, A., Sava, F., Pui, M., Tebbutt, S. J., & Carlsten, C. (2013). MicroRNA expression in response to controlled exposure to diesel exhaust: attenuation by the antioxidant N-acetylcysteine in a randomized crossover study. Environmental Health Perspectives, 121(6), 670–675. doi:10.1289/ehp.1205963 Yamazaki, T., Yang, X. O., Chung, Y., Fukunaga, A., Nurieva, R., Pappu, B., et al. (2008). CCR6 regulates the migration of inflammatory and regulatory T cells. Journal of Immunology (Baltimore, Md. : 1950), 181(12), 8391–8401. Yang, A. S., Estécio, M. R. H., Doshi, K., Kondo, Y., Tajara, E. H., & Issa, J.-P. J. (2004). A simple method for estimating global DNA methylation using bisulfite PCR of repetitive DNA elements. Nucleic Acids Research, 32(3), e38. doi:10.1093/nar/gnh032 Yuen, R. K. C., Avila, L., Peñaherrera, M. S., Dadelszen, von, P., Lefebvre, L., Kobor, M. S., & Robinson, W. P. (2009). Human Placental-Specific Epipolymorphism and its Association with Adverse Pregnancy Outcomes. PLoS ONE, 4(10), e7389. doi:10.1371/journal.pone.0007389.s010 Yuen, R. K., Neumann, S. M., Fok, A. K., Peñaherrera, M. S., McFadden, D. E., Robinson, W. P., & Kobor, M. S. (2011). Extensive epigenetic reprogramming in human somatic tissues between fetus and adult. Epigenetics & Chromatin, 4(1), 7. doi:10.1186/1756-8935-4-7 105  Zanobetti, A., Schwartz, J., & Dockery, D. W. (2000). Airborne particles are a risk factor for hospital admissions for heart and lung disease. Environmental Health Perspectives, 108(11), 1071–1077. Zhang, D., Cheng, L., Badner, J. A., Chen, C., Chen, Q., Luo, W., et al. (2010). AR TICLEGenetic Control of Individual Differencesin Gene-Specific Methylation in Human Brain. The American Journal of Human Genetics, 86(3), 411–419. doi:10.1016/j.ajhg.2010.02.005 Zhang, Q., Wang, H. Y., Liu, X., & Wasik, M. A. (2007). STAT5A is epigenetically silenced by the tyrosine kinase NPM1-ALK and acts as a tumor suppressor by reciprocally inhibiting NPM1-ALK expression. Nature Medicine, 13(11), 1341–1348. doi:10.1038/nm1659 Zhu, J., He, F., Hu, S., & Yu, J. (2008). On the nature of human housekeeping genes. Trends in Genetics : TIG, 24(10), 481–484. doi:10.1016/j.tig.2008.08.004 Zhu, X., Liang, J., Li, F., Yang, Y., Xiang, L., & Xu, J. (2011). Analysis of associations between the patterns of global DNA hypomethylation and expression of DNA methyltransferase in patients with systemic lupus erythematosus. International Journal of Dermatology, 50(6), 697–704. doi:10.1111/j.1365-4632.2010.04804.x Ziller, M. J., Gu, H., Müller, F., Donaghey, J., Tsai, L. T.-Y., Kohlbacher, O., et al. (2013). Charting a dynamic DNA methylation landscape of the human genome. Nature, 500(7463), 477–481. doi:10.1038/nature12433 Zimniak, P., Nanduri, B., Pikuła, S., Bandorowicz-Pikuła, J., Singhal, S. S., Srivastava, S. K., et al. (1994). Naturally occurring human glutathione S-transferase GSTP1-1 isoforms with isoleucine and valine in position 104 differ in enzymic properties. European Journal of Biochemistry / FEBS, 224(3), 893–899.   106  Appendices Appendix A  Supplementary Tables Table 1 Demographic information for the 25 healthy subjects involved in the study. Variable Measurements Gender: Male 15 [60%]† Age, years 34.0 ± 5.8 [26.0 - 45.0]* Ethnicity   Caucasian 18 [72%]  Non-Caucasian 7 [28%] BMI, kg/m2 24.5 ± 5.5 [18.2 – 47.6] †count [percentage]. *mean ± standard deviation [min - max]   Table 2 DAVID gene ontology enrichment for tissue-specific probes using sub-ontologies biological process and cellular component.  Cluster Enrichment score GO term # of associated loci 1 1.63 Extracellular region (GO:0005576) 47 2 1.59 Immune system process (GO:0002376) 29 Cytokine activity (GO:0005125) 11         107  Table 3 Probes in PBMC found to display high confidence associations with gender and BMI (FDR≤0.05). Gender associated high confidence probes (40)  Probe ID CpG Island Gene Symbol Chromosome CpG Density 1 ABO_P312_F Y ABO 9 HC 2 ADCYAP1_P398_F Y ADCYAP1 18 HC 3 APBA1_P644_F Y APBA1 9 HC 4 COL4A3_E205_R Y COL4A3 2 HC 5 CYP1B1_P212_F Y CYP1B1 2 HC 6 DLC1_P695_F N DLC1 8 LC 7 DST_E31_F Y DST 6 HC 8 ETS1_E253_R Y ETS1 11 HC 9 HOXA5_P479_F Y HOXA5 7 HC 10 HPSE_P29_F Y HPSE 4 HC 11 IGF1_E394_F N IGF1 12 LC 12 IGFBP7_P371_F Y IGFBP7 4 HC 13 IHH_P529_F Y IHH 2 HC 14 IL16_P93_R N IL16 15 LC 15 IL2_P607_R N IL2 4 LC 16 JAG2_P264_F Y JAG2 14 HC 17 MAP3K9_E17_R Y MAP3K9 14 HC 18 MLH1_P381_F Y MLH1 3 HC 19 MLH3_P25_F Y MLH3 14 HC 20 MMP10_E136_R N MMP10 11 LC 21 MPO_P883_R N MPO 17 LC 22 MSH3_E3_F Y MSH3 5 HC 23 PDE1B_P263_R Y PDE1B 12 IC 24 PGF_E33_F Y PGF 14 HC 25 PLAGL1_P236_R Y PLAGL1 6 HC 26 PTGS1_P2_F Y PTGS1 9 IC 27 RASA1_E107_F Y RASA1 5 HC 28 SERPINA5_E69_F N SERPINA5 14 LC 29 SFTPB_P689_R N SFTPB 2 LC 30 SOX1_P1018_R Y SOX1 13 HC 31 STK11_P295_R Y STK11 19 HC 32 TCF7L2_P193_R Y TCF7L2 10 HC 33 TDGF1_P428_R N TDGF1 3 IC 34 TFDP1_P543_R Y TFDP1 13 HC 35 TMEFF1_E180_R Y TMEFF1 9 HC 108  Gender associated high confidence probes (40)  Probe ID CpG Island Gene Symbol Chromosome CpG Density 36 TP73_E155_F Y TP73 1 HC 37 TP73_P496_F Y TP73 1 HC 38 TRAF4_P372_F Y TRAF4 17 HC 39 WT1_E32_F Y WT1 11 HC 40 XPC_P226_R Y XPC 3 HC BMI associated high confidence probes (51)  Probe ID CpG Island Gene Symbol Chromosome CpG Density 1 AATK_P519_R Y AATK 17  2 ABCA1_P45_F Y ABCA1 9 HC 3 ABO_E110_F Y ABO 9 HC 4 ACVR1_P983_F N ACVR1 2 HC 5 ACVR1C_P115_R Y ACVR1C 2 HC 6 AREG_E25_F Y AREG 4 HC 7 CASP2_P192_F Y CASP2 7 HC 8 CCND2_P898_R Y CCND2 12 IC 9 CD81_P272_R Y CD81 11 HC 10 CDKN2B_E220_F Y CDKN2B 9 HC 11 COL18A1_P365_R Y COL18A1 21 LC 12 COL4A3_P545_F Y COL4A3 2 HC 13 CPNE1_P138_F Y CPNE1 20 - 14 EPHB6_E342_F Y EPHB6 7 HC 15 EVI2A_P94_R N EVI2A 17 LC 16 FGFR1_E317_F Y FGFR1 8 HC 17 FHIT_E19_R Y FHIT 3 IC 18 HDAC7A_P344_F N HDAC7A 12 - 19 HHIP_E94_F Y HHIP 4 HC 20 HLA-DPB1_E2_R N HLA-DPB1 6 LC 21 HOXA5_E187_F Y HOXA5 7 HC 22 HOXB2_P488_R N HOXB2 17 IC 23 IGSF4C_P533_R Y IGSF4C 19 HC 24 IHH_P529_F Y IHH 2 HC 25 INHA_P1144_R Y INHA 2 LC 26 IRAK3_P13_F Y IRAK3 12 HC 27 ITGA6_P718_R Y ITGA6 2 - 28 ITGB4_E144_F Y ITGB4 17 HC 29 MAGEL2_P170_R Y MAGEL2 15 IC 109  BMI associated high confidence probes (51)  Probe ID CpG Island Gene Symbol Chromosome CpG Density 30 MAP3K8_P1036_F Y MAP3K8 10 HC 31 MAPK14_P327_R Y MAPK14 6 HC 32 MAPK9_P1175_F N MAPK9 5 LC 33 MCM2_P241_R Y MCM2 3 HC 34 MOS_P27_R Y MOS 8 IC 35 MST1R_E42_R Y MST1R 3 HC 36 MUC1_P191_F Y MUC1 1 IC 37 MYOD1_P50_F Y MYOD1 11 HC 38 NBL1_P24_F N NBL1 1 LC 39 NPR2_P1093_F Y NPR2 9 HC 40 NTRK2_P10_F Y NTRK2 9 HC 41 NTSR1_E109_F Y NTSR1 20 HC 42 PRSS8_E134_R Y PRSS8 16 LC 43 PTGS1_P2_F Y PTGS1 9 IC 44 PTK2_P735_R Y PTK2 8 HC 45 PURA_P928_R Y PURA 5 HC 46 SKI_E465_R Y SKI 1 HC 47 SMARCA3_P109_R Y SMARCA3 3 HC 48 SNRPN_E14_F N SNRPN 15 IC 49 TFAP2C_P765_F Y TFAP2C 20 HC 50 TMPRSS4_E83_F N TMPRSS4 11 LC 51 ZIM2_E110_F Y ZIM2 19 -          110  Table 4 Subject demographics used with PC scores to determine main drivers of variation in data Variable Measurements Gender: Male 7 [44%]† Age, years 28.7 ± 6.7 [19.2-46.5]* Ethnicity   Asian 2 [13%]  Caucasian 12 [75%]  Middle Eastern 1 [6%]  South Asian 1 [6%] Height, cm 170.0 ± 12.2 [151.0-185.5] Mass, kg 71.9 ± 15.5 [53.0-104.6] BMI, kg/m2 24.8 ± 3.9 [19.8-34.7] FEV1, % predicted 90.3 ± 14.1 [66.0-120.0] Atopy: atopic 9 [56%] Asthma diagnosis: yes 12 [75%] Methacholine-responsive: yes 15 [94%] GSTP1: G allele 11 [69%] Lymphocytes, K/μL 1.9 ± 0.5 [1.2-2.6] Monocytes, K/μL 0.43 ± 0.13 [0.20-0.70] Basophils, K/μL 0.03 ± 0.05 [0.00-0.10] †count [percentage]. *mean ± standard deviation [min - max]          111  Table 5 Top 4 DAVID functional annotation clusters. Enrichment of cellular component, biological process, and molecular function in the 2827 PCA filtered probes was calculated using all probes involved in this analysis as the background.     Annotation cluster1 Enrichment Score: 3.426416727305296 GO Category GO Term Count Fold Enrichment Benjamini Hochberg adjusted p-value Molecular Function GO:0005088~Ras guanyl-nucleotide exchange factor activity 25 2.814 0.004 Biological Process GO:0051056~regulation of small GTPase mediated signal transduction 47 1.881 0.105 Molecular Function GO:0005085~guanyl-nucleotide exchange factor activity 33 2.167 0.020 Molecular Function GO:0005089~Rho guanyl-nucleotide exchange factor activity 20 2.701 0.020 Biological Process GO:0046578~regulation of Ras protein signal transduction 39 1.884 0.246 Molecular Function GO:0030695~GTPase regulator activity 63 1.584 0.035 Biological Process GO:0035023~regulation of Rho protein signal transduction 23 2.316 0.112 Molecular Function GO:0060589~nucleoside-triphosphatase regulator activity 64 1.572 0.035 Molecular Function GO:0005083~small GTPase regulator activity 43 1.620 0.129 Molecular Function GO:0005096~GTPase activator activity 28 1.298 0.918 Molecular Function GO:0008047~enzyme activator activity 41 1.223 0.920   112  Annotation cluster 2 Enrichment Score: 3.3598211037329535 GO Category GO Term Count Fold Enrichment Benjamini Hochberg adjusted p-value Molecular Function GO:0046872~metal ion binding 470 1.150 0.022 Molecular Function GO:0043167~ion binding 479 1.146 0.017 Molecular Function GO:0043169~cation binding 472 1.144 0.020 Molecular Function GO:0008270~zinc ion binding 267 1.182 0.114 Molecular Function GO:0046914~transition metal ion binding 303 1.111 0.542 Annotation cluster 3 Enrichment Score: 3.0207832225704467 GO Category GO Term Count Fold Enrichment Benjamini Hochberg adjusted p-value Biological Process GO:0010627~regulation of protein kinase cascade 44 1.732 0.096 Biological Process GO:0043123~positive regulation of I-kappaB kinase/NF-kappaB cascade 22 2.238 0.127 Biological Process GO:0043122~regulation of I-kappaB kinase/NF-kappaB cascade 23 2.115 0.171 Biological Process GO:0009967~positive regulation of signal transduction 48 1.605 0.179 Biological Process GO:0010740~positive regulation of protein kinase cascade 31 1.819 0.219 Biological Process GO:0010647~positive regulation of cell communication 51 1.524 0.278 Annotation cluster 4 Enrichment Score: 2.7298240154764244 GO Category GO Term Count Fold Enrichment Benjamini Hochberg adjusted p-value Biological Process GO:0010941~regulation of cell death 115 1.394 0.190 Biological Process GO:0043067~regulation of programmed cell death 114 1.387 0.189 Biological Process GO:0042981~regulation of apoptosis 113 1.389 0.154 113          Annotation cluster 4 Enrichment Score: 2.7298240154764244 GO Category GO Term Count Fold Enrichment Benjamini Hochberg adjusted p-value Biological Process GO:0010942~positive regulation of cell death 68 1.556 0.131 Biological Process GO:0006917~induction of apoptosis 53 1.662 0.126 Biological Process GO:0012502~induction of programmed cell death 53 1.657 0.108 Biological Process GO:0043065~positive regulation of apoptosis 67 1.552 0.098 Biological Process GO:0008624~induction of apoptosis by extracellular signals 24 2.229 0.105 Biological Process GO:0043068~positive regulation of programmed cell death 67 1.540 0.099 Biological Process GO:0012501~programmed cell death 84 1.383 0.252 Biological Process GO:0016265~death 94 1.309 0.363 Biological Process GO:0008219~cell death 93 1.304 0.391 Biological Process GO:0006915~apoptosis 80 1.335 0.392 Biological Process GO:0043069~negative regulation of programmed cell death 48 1.301 0.699 Biological Process GO:0060548~negative regulation of cell death 48 1.297 0.701 Biological Process GO:0043066~negative regulation of apoptosis 47 1.292 0.731 Biological Process GO:0006916~anti-apoptosis 25 1.184 0.960 114  Table 6 CpG sites found to have significant decrease in methylation as a result of DE exposure through LME modeling. For each probe, Δβ(FA6&30hr-FA0hr) represents the change in beta value from pre-FA exposure at 0hr and post-FA exposure at 6 and 30hr. Δβ(DE6&30hr-DE0hr) represents the change in beta value from pre-DE exposure to post-DE exposure. Closest TSS gene name indicates the gene name of the closest transcription start site. UCSC RefGene name indicates the name of the gene at which the probe is located. UCSC refgene group indicates the genomic region at which the probe is located. a) Sites with decrease in methylation were mined from the 2827 probes identified to be associated with DE exposure patterns. Sites were ordered from largest Δβ to smallest Δβ. b) Sites with increase in methylation were mined from the 2827 probes identified to be associated with DE exposure patterns. Sites were ordered from largest Δβ to smallest Δβ. c) Sites were mined from the 1118 probes with ≥15bp overlap with LINE1 elements in the genome. Sites were order from largest decrease in methylation to largest increase in methylation. d) Sites were mined from the 1271 probes with ≥15bp overlap with Alu elements in the genome. Sites were order from largest decrease in methylation to largest increase in methylation. a) Illumina probe ID Δβ(FA6&30hr-FA0hr) Δβ(DE6&30hr-DE0hr) Closest TSS gene name UCSC RefGene name UCSC refgene group cg27183818 -0.038 -0.069 BCL10   cg03917666 -0.025 -0.058 AK055803   cg20495370 -0.023 -0.058 SLC7A7 SLC7A7;SLC7A7;SLC7A7 Body;Body;Body cg14412134 -0.037 -0.057 AKAP5 MTHFD1 Body cg24414363 -0.019 -0.055 CENPM CENPM;CENPM;CENPM TSS200;Body;Body cg04211179 -0.032 -0.054 ZBTB17 ZBTB17 5'UTR cg16266809 -0.01 -0.049 ASXL1   cg21664636 -0.033 -0.048 BC015590   cg12078154 -0.01 -0.048 RPTOR RPTOR;RPTOR Body;Body cg27141509 -0.011 -0.047 TRNA_Val   cg01965380 -0.013 -0.046 LRRC8D LRRC8D;LRRC8D 5'UTR;5'UTR cg16462648 -0.016 -0.044 ANKRD55   cg23732024 -0.015 -0.044 LY96 LY96 Body cg02156723 -0.015 -0.043 MIIP   cg07462448 -0.03 -0.043 CASP7 CASP7;CASP7;CASP7;CASP7 5'UTR;5'UTR;5'UTR;5'UTR cg08314949 -0.013 -0.043 RPTOR RPTOR;RPTOR Body;Body cg23295629 -0.027 -0.042 PVT1 PVT1 Body cg26992245 -0.013 -0.042 MIR3148   cg19149314 -0.003 -0.042 STK17B HECW2 3'UTR cg23304023 -0.021 -0.042 TACC2 TACC2;TACC2 Body;Body 115  Illumina probe ID Δβ(FA6&30hr-FA0hr) Δβ(DE6&30hr-DE0hr) Closest TSS gene name UCSC RefGene name UCSC refgene group cg15006828 -0.022 -0.04 BC100777   cg11876705 -0.015 -0.04 RASGRP4 RASGRP4;RASGRP4;RASGRP4;RASGRP4;RASGRP4;RASGRP4;RASGRP4 TSS1500;TSS1500;TSS1500;TSS1500;TSS1500;TSS1500;TSS1500 cg03613942 -0.019 -0.039 BC034940   cg10063575 -0.021 -0.039 AMHR2 AMHR2;AMHR2;AMHR2 Body;Body;Body cg08131547 -0.024 -0.039 ZNF121 ZNF121 5'UTR cg01464186 -0.029 -0.039 SGMS1 SGMS1 5'UTR cg01894985 0.002 -0.039 MYLK MYLK;MYLK;MYLK;MYLK 5'UTR;5'UTR;5'UTR;5'UTR cg08243465 -0.015 -0.039 PMEPA1 PMEPA1;PMEPA1;PMEPA1;PMEPA1 5'UTR;5'UTR;Body;Body cg24458314 -0.021 -0.038 HDAC9 HDAC9;HDAC9;HDAC9;HDAC9;HDAC9 Body;Body;Body;Body;Body cg18986967 -0.009 -0.038 AGK   cg07000713 -0.027 -0.037 FAM19A2 FAM19A2 5'UTR cg02319972 -0.006 -0.037 MGC15885   cg19904265 -0.021 -0.036 MIR4251 PRDM16;PRDM16 Body;Body cg03655684 0.002 -0.036 CLCN7 CLCN7;CLCN7 Body;Body cg01117384 -0.01 -0.036 PMEPA1 PMEPA1;PMEPA1;PMEPA1;PMEPA1 5'UTR;5'UTR;Body;Body cg17212019 -0.025 -0.036 NEFL   cg27457191 -0.009 -0.036 PHTF2 PHTF2;PHTF2 5'UTR;5'UTR cg11465442 -0.017 -0.036 DEFB136 DEFB136 TSS200 cg11798406 -0.017 -0.036 LINC00114 NCRNA00114 Body cg08786003 -0.015 -0.035 FCRL3 FCRL3 TSS200 cg21585138 -0.017 -0.035 CISH CISH;CISH Body;Body cg10246903 -0.021 -0.035 TSC22D1   cg03999067 -0.009 -0.035 U6   cg23867673 -0.011 -0.035 CDH23 CDH23;CDH23 Body;Body cg03081173 -0.021 -0.035 HCG27 HCG27 Body cg25596287 -0.017 -0.034 RCC1 SNHG3-RCC1;RCC1 Body;TSS1500 116  Illumina probe ID Δβ(FA6&30hr-FA0hr) Δβ(DE6&30hr-DE0hr) Closest TSS gene name UCSC RefGene name UCSC refgene group cg22930549 0.009 -0.034 Metazoa_SRP RAD51L1;RAD51L1;RAD51L1 Body;Body;Body cg01613294 -0.004 -0.034 APOL3 APOL3;APOL3;APOL3;APOL3;APOL3;APOL3 5'UTR;TSS1500;TSS1500;TSS1500;TSS1500;5'UTR cg26216876 -0.012 -0.033 SLMO2 SLMO2 Body cg02857074 -0.008 -0.033 AK055932 CACNA2D1 Body cg05462446 -0.012 -0.033 ADAMTS18   cg13461554 -0.025 -0.033 LHCGR   cg13212186 -0.023 -0.032 HCG27 HCG27 Body cg25503381 -0.02 -0.032 THBS1 THBS1 TSS1500 cg11321190 -0.024 -0.032 LRRFIP2   cg19291696 -0.022 -0.032 USP12   cg21097090 -0.008 -0.031 TNFAIP8 TNFAIP8;TNFAIP8 Body;Body cg13541527 -0.012 -0.03 SNORD52 C6orf48;C6orf48;SNORD52 5'UTR;5'UTR;TSS1500 cg06190732 0.003 -0.03 SERPINA3 SERPINA3 TSS200 cg27104695 -0.013 -0.03 SH3BP5 SH3BP5;SH3BP5 Body;5'UTR cg07206827 -0.009 -0.03 LINGO4 LINGO4 TSS1500 cg17851868 -0.019 -0.03 FLJ30838   cg25812095 -0.013 -0.03 GTDC1 GTDC1;GTDC1 5'UTR;5'UTR cg19222784 -0.011 -0.029 NAV2 NAV2;NAV2;NAV2;NAV2 TSS1500;Body;Body;Body cg05094429 -0.006 -0.029 CCR6 CCR6;CCR6 5'UTR;TSS200 cg00465739 -0.017 -0.029 ATP8A2 ATP8A2 Body cg15464148 -0.02 -0.029 LPAR5 LPAR5 5'UTR cg20427318 -0.012 -0.029 LOC154092   cg15192146 -0.01 -0.028 AK098570   cg08025405 -0.008 -0.028 ASB3   cg15718287 -0.018 -0.028 DCLK1 DCLK1 5'UTR cg21536074 -0.015 -0.027 GLI3 GLI3 Body cg27305009 -0.017 -0.026 WIPF1 WIPF1;WIPF1 5'UTR;5'UTR cg01890712 -0.012 -0.026 OR9Q1 OR9Q1 TSS1500 cg05033369 -0.009 -0.025 FCRLA FCRLA TSS1500 cg00135497 -0.014 -0.025 LRAT   cg25298754 -0.002 -0.024 ZBED2 ZBED2;CD96;CD96 5'UTR;Body;Body cg07376029 -0.023 -0.024 GC GC TSS1500 117  Illumina probe ID Δβ(FA6&30hr-FA0hr) Δβ(DE6&30hr-DE0hr) Closest TSS gene name UCSC RefGene name UCSC refgene group cg04230397 -0.005 -0.024 MUC21 MUC21 Body cg12159992 -0.01 -0.023 MYBPC1 MYBPC1;MYBPC1;MYBPC1;MYBPC1 Body;Body;Body;Body cg02989940 -0.007 -0.023 AHSP AHSP;AHSP 5'UTR;1stExon cg13262467 -0.009 -0.022 LRRIQ4 LRRIQ4 TSS1500 cg08241528 -0.013 -0.022 RBPJ   cg25617519 -0.013 -0.022 KLHL29 KLHL29 5'UTR cg16455376 -0.007 -0.021 CARHSP1   cg05246530 -0.003 -0.021 ST3GAL2 ST3GAL2 5'UTR cg02046532 -0.008 -0.021 DEFB129 DEFB129 TSS200 cg25609393 -0.003 -0.021 HCN1 HCN1 Body cg11342453 -0.018 -0.02 HIST1H3F   cg23670794 0 -0.02 ZBED2 ZBED2;CD96;CD96 5'UTR;Body;Body cg26207423 0.003 -0.02 HMBOX1 HMBOX1;HMBOX1 5'UTR;5'UTR cg01919768 -0.007 -0.02 TFR2 TFR2 TSS1500 cg15150463 -0.01 -0.02 GJA10   cg17365725 -0.014 -0.02 D2HGDH D2HGDH Body cg17829936 -0.013 -0.02 TAAR5 TAAR5 1stExon cg06106484 -0.004 -0.019 TRNA_Pseudo  cg18447402 -0.009 -0.019 MRPS33 MRPS33;MRPS33 Body;Body cg07157117 -0.006 -0.019 AGXT2L1   cg13583523 -0.008 -0.018 LOC340107   cg26556196 0.001 -0.018 KHDRBS2 KHDRBS2 Body cg18837292 -0.005 -0.018 IRX1   cg15007959 0.008 -0.018 MYBPC2 SPIB Body cg27495728 -0.003 -0.016 TRNA_Trp   cg21071237 0.005 -0.014 ADAMTS15   cg02634628 -0.004 -0.013 UNCX   cg23652785 -0.01 -0.011 AF086258   cg02612650 0 -0.011 HIST1H3F   cg07229076 -0.003 -0.008 ARL4A ARL4A;ARL4A;ARL4A 5'UTR;TSS200;5'UTR cg04135246 0.001 -0.008 BC038465   cg12667031 -0.002 -0.007 SCYL3   118  Illumina probe ID Δβ(FA6&30hr-FA0hr) Δβ(DE6&30hr-DE0hr) Closest TSS gene name UCSC RefGene name UCSC refgene group cg04204002 0.002 -0.007 TMPRSS3 TMPRSS3;TMPRSS3 TSS1500;TSS1500 cg07221635 0 -0.005 PECR TMEM169;PECR;TMEM169;TMEM169;TMEM169 TSS200;1stExon;TSS200;TSS200;TSS200 cg11351779 0.003 -0.003 MPPED1 MPPED1 5'UTR b) Illumina probe ID Δβ(FA6&30hr-FA0hr) Δβ(DE6&30hr-DE0hr) Closest TSS gene name UCSC refgene name UCSC refgene group cg18944752 0.020 0.057 RGS12 RGS12;RGS12;RGS12 Body;Body;Body cg22331200 0.007 0.045 MPO MPO Body cg10049789 0.020 0.043 SH3TC1 SH3TC1 5'UTR cg03022891 0.021 0.040 TNNT3 TNNT3;TNNT3;TNNT3;TNNT3 Body;Body;Body;Body cg23756272 0.024 0.039 BCL2 BCL2 Body cg04309234 0.014 0.036 PRDM1   cg09729012 0.016 0.036 RYBP   cg07142009 0.016 0.033 XAF1   cg04109092 0.014 0.033 TTYH3 IQCE;IQCE Body;Body cg25437886 0.010 0.033 LIMD1 LIMD1 Body cg26060971 0.012 0.032 DNAH1 DNAH1 Body cg25150953 0.010 0.032 LIMCH1 LIMCH1;LIMCH1;LIMCH1 Body;Body;Body cg09597638 0.006 0.032 AB062083   cg06478504 0.017 0.031 HDAC4 HDAC4 Body cg13029400 0.005 0.029 ZBTB38 ZBTB38 5'UTR cg14440934 0.026 0.029 ZDHHC1 ZDHHC1 TSS1500 cg08564172 0.006 0.028 ENC1   cg23032129 0.011 0.026 SORT1 SORT1 TSS1500 cg17624536 0.015 0.025 PRDM1   cg16744531 0.007 0.025 B3GNT3 B3GNT3 TSS1500 cg01201279 0.012 0.024 AK094480 LEKR1 TSS1500 cg25158320 0.003 0.024 SERPINF1 SERPINF1;SERPINF1 1stExon;5'UTR cg11301250 0.001 0.023 VEPH1 VEPH1;VEPH1;VEPH1;VEPH1;VEPH1 TSS200;5'UTR;5'UTR;5'UTR;5'UTR 119  Illumina probe ID Δβ(FA6&30hr-FA0hr) Δβ(DE6&30hr-DE0hr) Closest TSS gene name UCSC RefGene name UCSC refgene group cg05084827 0.013 0.023 RPS27A C2orf63;C2orf63 Body;Body cg18551877 0.007 0.023 GLYATL2 GLYATL2 TSS200 cg16887334 0.005 0.021 OXT OXT TSS200 cg15600238 0.000 0.021 HOTTIP   cg06561886 0.009 0.020 SLC44A2 SLC44A2;SLC44A2;SLC44A2 5'UTR;1stExon;Body cg02331910 0.010 0.020 S100A13 S100A13;S100A13;S100A1;S100A13;S100A13;S100A13 5'UTR;5'UTR;TSS1500;TSS200;5'UTR;TSS1500 cg11343894 0.002 0.019 S100A13 S100A13;S100A13;S100A13;S100A13;S100A1;S100A13;S100A13 5'UTR;1stExon;5'UTR;TSS200;TSS1500;5'UTR;5'UTR cg09583957 0.003 0.018 GNAS GNAS;GNAS;GNAS;GNAS;GNAS 5'UTR;5'UTR;1stExon;1stExon;3'UTR cg27380292 0.004 0.018 MIR4499   cg19931644 0.005 0.018 LONRF1   cg16512163 -0.004 0.017 RASA3 RASA3 Body cg09322534 0.011 0.015 SLC2A1   cg09841842 0.005 0.015 U6 FRMD6 5'UTR cg19083407 0.003 0.015 DKFZp686E10196 PAX8;PAX8;PAX8;PAX8;PAX8;LOC440839;LOC654433 Body;Body;Body;Body;Body;Body;TSS1500 cg12502403 0.002 0.014 MARK2 MARK2;MARK2;MARK2;MARK2 1stExon;1stExon;1stExon;1stExon cg14192299 -0.001 0.014 NANOS2 NANOS2 1stExon cg07673080 -0.007 0.014 MIR4269 HDAC4 Body cg26018827 0.004 0.013 GEN1 GEN1;SMC6;GEN1;SMC6;GEN1 5'UTR;TSS1500;1stExon;TSS1500;5'UTR cg24291500 0.000 0.013 REST REST TSS1500 cg08260406 -0.008 0.012 OR2L13 OR2L13 TSS200 cg23899408 -0.003 0.010 HOOK2 HOOK2;HOOK2 Body;Body cg09038676 -0.001 0.010 GSTP1 GSTP1 Body cg16716750 0.000 0.010 RGS17 RGS17 5'UTR cg20291033 0.004 0.010 TNR TNR TSS200 120  Illumina probe ID Δβ(FA6&30hr-FA0hr) Δβ(DE6&30hr-DE0hr) Closest TSS gene name UCSC RefGene name UCSC refgene group cg14955916 0.001 0.010 DIP2C DIP2C Body cg20102877 -0.002 0.009 KRTCAP3 KRTCAP3;KRTCAP3 Body;Body cg06522412 0.003 0.007 MIR4493   cg17154022 0.003 0.007 HMGN2 HMGN2;HMGN2 1stExon;5'UTR cg11960393 0.003 0.007 LOX LOX TSS200 cg26802256 0.003 0.006 FOXK1   ch.13.32987270F 0.001 0.006 STARD13   ch.7.107350695F -0.001 0.004 DLD   cg11798043 -0.002 0.004 LMBRD1 LMBRD1 Body cg26056770 0.001 0.003 TMEM141 TMEM141 TSS1500 c) Illumina probe ID Δβ(FA6&30hr-FA0hr) Δβ(DE6&30hr-DE0hr) Closest TSS gene name UCSC refgene name UCSC refgene group cg11075561 -0.031 -0.043 MAPK13   cg05266321 -0.007 -0.028 CCR2 CCR2 3'UTR cg17729072 -0.012 -0.026 AK092451   cg04426653 -0.003 -0.025 BC033989   cg15210817 -0.016 -0.024 CSF3R CSF3R;CSF3R;CSF3R;CSF3R TSS1500;TSS1500;TSS1500;TSS1500 cg02224002 -0.008 -0.024 RNF166 RNF166 Body cg10136452 -0.011 -0.023 BC087858   cg25840780 -0.009 -0.022 NFATC2 NFATC2;NFATC2;NFATC2 Body;Body;Body cg22689597 -0.006 -0.022 LRRC10 LRRC10 TSS1500 cg14248598 -0.011 -0.021 C1S   cg25619459 -0.016 -0.020 CDK14 CDK14 Body cg05392244 -0.007 -0.019 FOXC1   cg21874404 -0.003 -0.017 SNORD112 MEG8 Body cg20728419 -0.006 -0.013 USP47 USP47 Body cg21750986 -0.006 -0.013 RAB12   cg12154976 -0.005 -0.012 RASGEF1C RASGEF1C 5'UTR cg01830883 -0.005 -0.012 WRAP73 WDR8 Body cg02713720 -0.001 -0.010 BC152379 FBLN2;FBLN2;FBLN2 Body;Body;Body 121  Illumina probe ID Δβ(FA6&30hr-FA0hr) Δβ(DE6&30hr-DE0hr) Closest TSS gene name UCSC RefGene name UCSC refgene group cg19587838 -0.002 0.009 SOX1   cg03993154 0.005 0.012 SLC13A1 SLC13A1 Body cg08995061 0.009 0.017 MLK7-AS1 ZAK Body cg09904296 0.001 0.019 GOLGA3 GOLGA3 TSS1500 cg20860806 0.004 0.019 SHC4   cg09500196 0.010 0.020 CARHSP1 CARHSP1 TSS1500 cg03997139 0.009 0.020 ROCK2 ROCK2 TSS1500 cg02215171 0.007 0.028 HERC5 HERC5 Body cg17554875 0.014 0.028 RNF130   cg12930304 0.015 0.028 SERPINB1   cg27220070 -0.001 0.031 EP400 EP400 TSS1500 cg12930392 0.005 0.032 PAK2 PAK2 5'UTR cg10225149 0.010 0.034 AKAP8 AKAP8L;AKAP8 Body;TSS1500 cg18437319 0.006 0.034 MCAT   cg13428066 0.013 0.044 KCNQ1OT1 KCNQ1;KCNQ1OT1;KCNQ1 Body;Body;Body d) Illumina probe ID Δβ(FA6&30hr-FA0hr) Δβ(DE6&30hr-DE0hr) Closest TSS gene name UCSC refgene name UCSC refgene group cg07128021 -0.018 -0.039 LOC283050 LOC283050;LOC283050;LOC283050 Body;Body;Body cg21593149 -0.017 -0.038 MAEA MAEA;MAEA Body;Body cg08343600 -0.015 -0.033 HNRNPF HNRNPF;HNRNPF;HNRNPF;HNRNPF;HNRNPF 5'UTR;TSS1500;5'UTR;5'UTR;5'UTR cg13502540 -0.021 -0.031 HCG27 HCG27 Body cg06493154 -0.019 -0.031 C6orf226 C6orf226 TSS1500 cg22553301 -0.004 -0.030 ZFR2 ZFR2 Body cg20130098 -0.017 -0.025 KALRN KALRN;KALRN Body;Body cg26785702 -0.020 -0.025 FOXR1 FOXR1 Body cg06482498 -0.010 -0.023 C6orf48 C6orf48;SNORD48;C6orf48 TSS1500;TSS1500;TSS1500 cg05669244 -0.009 -0.022 ZNF358   cg16747928 -0.009 -0.021 AK092451   cg02188939 -0.007 -0.020 SLC43A2 SLC43A2 Body cg08061463 -0.010 -0.020 TTLL13   cg02403031 -0.007 -0.016 Mir_633   122  Illumina probe ID Δβ(FA6&30hr-FA0hr) Δβ(DE6&30hr-DE0hr) Closest TSS gene name UCSC RefGene name UCSC refgene group cg22635673 0.006 0.010 BHLHA9   cg03148858 0.006 0.014 TLE2 TLE2 TSS1500 cg23482747 0.004 0.015 ARRDC2 ARRDC2 Body cg01801040 -0.001 0.016 TTC19 ZSWIM7;ZSWIM7 Body;Body cg00030420 0.011 0.019 SLC25A2 SLC25A2 TSS200 cg08812692 0.007 0.021 CLK4 CLK4 5'UTR cg00408567 0.003 0.028 FLJ38109 GALNT10;GALNT10 Body;Body cg13443575 0.009 0.033 SLFN13 SLFN13 TSS200 cg17653886 0.013 0.035 SLFN13 SLFN13 TSS1500 cg04354393 0.016 0.035 SLFN13 SLFN13 TSS200 cg00943851 0.015 0.036 AK092451   cg08866695 0.017 0.041 ACACB ACACB Body  


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items