University of Southern California Dissertations and Theses

Air pollution, smoking, and multigenerational DNA methylation Signatures: a study of two southern California cohorts

(USC Thesis Other)

Air pollution, smoking, and multigenerational DNA methylation Signatures: a study of two southern California cohorts

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Copyright 2023 Sahra Mohazzab-Hosseinian
Air Pollution, Smoking, and Multigenerational DNA methylation Signatures: A Study of Two
Southern California Cohorts
by
Sahra Mohazzab-Hosseinian
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(EPIDEMIOLOGY)
December 2023

ii
ACKNOWLEDGEMENTS
I would like to express my deep gratitude to the following people:
1. My advisor, Dr. Carrie Breton, for her guidance and commitment to my academic
journey. Her mentorship has been instrumental in shaping this work.
2. My committee members: Dr. Erika Garcia, Dr. Joseph Wiemels, Dr. Nicholas Mancuso,
and Dr. Marconett for their valued insight.
3. The participants, staff, and principal investigators associated with the cohorts’ data used
in this thesis to whom which this work would not have been possible.
4. My family who provided support, friendship, good food, funny jokes, perspective, and
hāl (حال (throughout this process.

iii
TABLE OF CONTENTS
Acknowledgements……………………………………………………………………….………ii
List of Tables…………………………………………………………………………….…….....iv
List of Figures…………………………………………………………….……………………....v
Abstract……..…………………………………………………………………………………...vii
Chapter 1: Introduction...…………………………..…………………………………………......1
Chapter 2: Overview of Quantile G-Computation………..…………………………………......12
Chapter 3: Literature Review on Smoking and DNA methylation..….……………………..…..14
Chapter 4: Literature Review on Air Pollution and DNA methylation..………………….…….17
Chapter 5: Gestational Air Pollution and Maternal DNA methylation Trajectories……….…...21
Chapter 6: Gestational Air Pollution Changes and Intergenerational DNA methylation............48
Chapter 7: Smoking, Air Pollution, and Multigenerational DNA methylation………………...65
Chapter 8: Conclusions…………………………………………………………………………87
Bibliography……………………………………………………………………………………89

iv
LIST OF TABLES
Study One Participant Characteristics…………………………………………………………...36
Study One Ambient Air Pollution Mixtures and GA Results………………….……….………..42
Study Two Participant Characteristics…………………………………………………………...60
Study Two Ambient Air Pollutant Influenced Maternal CpG…………………………………...61
Study Three Participant Characteristics…….…………………………………………………....79
Study Three Grandchild Affected Grandmaternal PTS CpGs in Paternal Lineage……………...82
Study Three Intergenerational FBRSL1…….……………………………………….…………...82
Study Three Intergenerational ADARB2…….…………………………………………………...82

v
LIST OF FIGURES
CHS Study Design Schematic…..….……………………………………………………………..8
CHS Recruitment Neighborhoods….……………………………………………………………..9
MADRES Study Design Schematic…..…………………………………………………………10
MADRES Recruitment Neighborhoods…....……………………………...…………………….10
Thesis Overview Schematic………..……………………………………………………………11
Study One GA Volcano Plot…….………………..……………………………………………..37
Study One GA Manhattan Plot…….……………..……………………………………………..38
Study One PM2.5 Volcano Plot…………………………………………………………………..39
Study One PM2.5 Influenced TRIM28 Methylation……………………………………………...40
Study One PM2.5 Influenced FBRSL1 Methylation……………………………………………...40
Study One PM2.5 Postnatal Persistence in CHS…………….…………………………………...41
Study One GA GO Pathways….……………..…………….…………………………………....42
Study One GA Distribution………………….....………….…………………………………….43
Study One Consort Diagram………………….....………….…………………………………...44
Study One PM2.5 distribution…………..…….....………….……………………………………45
Study One Boxplots of Replicated CpGs…….....………….……………………………………46
Study One Directed Acyclic Graph………….....………….………………………………..…..47

vi
Study Two Partial Plots…………….…………………………………………………….……...62
Study Two GA Distribution………….…………………………………………………….…….63
Study Two Consort Diagram……......…………………………………………………..……….64
Study Three Participant Characteristics..…………………………….………………………......80
Study Three Volcano Plot…………....……………………………….…………………….…....81
Study Three Cohort Diagram….. …....………..…………………………………………….…...83
Study Three Consort Diagram….. …....………..…………………………………………….…..84
Study Three logFC and SE plots. …....………..…….……………………………………….…..85

vii
THESIS ABSTRACT
Background
Epigenetic states, like DNA methylation (DNAm), may increase the risk of disease in
humans. The developmental period is a particularly sensitive window for the mother because of
the increase in energetic demands to support development, and for the fetus and associated germ
cells as well due to epigenetic programming. Consequently, the developmental window can impact
the health of three generations at once: the gestating mother (grandparental), developing fetus
(parental), and germ cell (grandchild). This can occur potentially through direct intergenerational
epigenetic effects. Environmental toxicants can differentially or consistently impact epigenetic
processes across these three generations and subsequently increase risk of disease. Less has been
studied on the effect of toxicants on the gestating mother, as the literature focuses on the more
explicit epigenetic changes occurring in the developing fetus. Little is also published on
intergenerational epigenetic profiles, mainly due to lack of available data. This thesis aims to
address these gaps in the literature using longitudinal maternal whole blood DNAm signatures
paired to cord blood DNAm from the Maternal and Developmental Risks from Environmental
Stressors (MADRES) cohort, and three generations of buccal cell DNAm data in the Children’s
Health Study (CHS). This study will focus on the effects of particulate matter and ambient air
pollutant mixtures as an exposure, along with grandparental prenatal tobacco smoking (PTS) in
the CHS. Both cohorts are based in Southern California, where air pollution contributes to
significant mortality and morbidity across the region and is therefore of public health significance.
Objectives

viii
The first study aims to address the dearth of studies on the effects of prenatal particulate
matter and ambient air pollution mixtures on maternal epigenetic changes. The second study seeks
to examine ambient air pollution mixture differences on both changes in maternal DNAm and cord
blood DNAm to add to the literature on intergenerational signatures. The potential impact of air
pollutant related maternal DNAm changes inducing cord blood DNAm signatures will also be
evaluated. The final study will test for evidence of a direct intergenerational effect of
grandmaternal PTS on parental and grandchild DNAm signatures and examine the potential
synergistic or antagonistic impact of parental air pollution exposure on grandchild DNAm among
PTS-affected CpGs.
Methods
The first study used adjusted linear models with an interaction between gestational age
(GA) and particulate matter and used prenatal DNA methylation signatures as the outcome.
Quantile g-computation was also used with the four ambient air pollutants as an exposure mixture
including the following pollutants of interest used in this thesis: NO2, O3, PM2.5, and PM10 on an
epigenome-wide scale. The second study evaluated change in ambient air pollution mixture, with
quantile g-computation, on differences in maternal DNAm and impacts on infant DNAm. Cord
blood epigenome-wide impacts of ambient air pollutant affected maternal DNAm signatures will
also be evaluated. The third study analyzed linear mixed models with a random effect for family
to investigate the potential change in methylation between the referent grandmaternal methylation
profile versus parental and grandchild DNAm signatures among families with a smoking
grandmother. DNAm signatures analyzed in the linear mixed model were chosen based
on overlap in generational and sex stratified epigenome-wide analyses. The potential synergistic
or antagonistic impact of parental PM2.5 exposure on overlapping DNAm signatures will also be

ix
evaluated using quantile g-computation. Multigenerational exposure to smoking will be
accounted for in the mixture model.
Results
Prenatal PM2.5 and ambient air pollution resulted in changes to 136 and 30 maternal CpGs,
respectively, between early and late pregnancy in the MADRES cohort. There was no overlap
between the mixtures and PM2.5 DNAm identified. There was some evidence that exposure to
PM2.5 in pregnancy induced CpG signatures that were persistent into the postpartum period (>2
years) in one CpG. Changes in ambient air pollution between early and late pregnancy also induced
change in 22 CpGs in pregnancy, but there was no evidence that change affected DNAm in cord
blood. Grandmaternal smoking was associated with DNAm in male and female offspring and all
paternal and maternal grandchildren. There was overlap in FBRSL1 and ADARB2 methylation
between male offspring and the paternal female grandchild that differed among families exposed
to grandmaternal smoke compared to those who were not. Parental ambient air pollution exposure
did not synergistically or antagonistically impact PTS-affected DNAm in the grandchild
generation.
Conclusion
Molecular states like DNA methylation can by impacted by environmental toxicants, and
during the prenatal period this has the potential to affect the grandparental, parental, and grandchild
generation. Study one and study two demonstrated some effect of cumulative levels as well as
changes in prenatal PM2.5 and ambient air pollution exposures, affecting maternal DNA
methylation profiles in pregnancy. There was no evidence for this in our cord blood DNAm, but
our sample size was modest. Finally, there was evidence of an intergenerational effect of

x
grandmaternal smoking in FBRSL1, with persistent decreased methylation signatures among
families exposed to smoking in the paternal parent and female grandchild lineage.

Copyright 2023 Sahra Mohazzab-Hosseinian
CHAPTER 1: INTRODUCTION
The Developmental Origins of Health and Disease (DOHAD) hypothesis suggests
environmental exposures during gestation can have impacts on the developing fetus’s adult health
outcomes [1]. Development of the fetus and its associated organs increases the sensitivity of this
period to environmental insult [1,2]. Rapid cellular division to support organogenesis is
coordinated by the epigenome, a physical structure that surrounds DNA and packs it into its
chromatin state for transcriptional efficiency [2,3]. The epigenome consists of several molecular
modifications on DNA that may have an influence on gene expression [4]. Epigenetic molecular
markers are an independent structure to the genome and are mitotically stable [4]. The epigenetic
landscape, unlike DNA, varies across cell types. Differences in the epigenetic profile across cell
type act to support the diverse needs of tissues. The epigenome consists of three main molecular
markers: DNA methylation, non-coding RNAs, and histones [4].
DNA methylation (DNAm) consists of the addition of a methyl group onto the fifth carbon
of DNA’s cytosine ring [4]. This addition occurs at the 5’ end of guanines, coined a CpG
dinucleotide (CpG) [4,5]. These CpGs are located across the genome sparsely but can also cluster
into a “CpG island” (CGIs), which are usually located close to promoter regions and are
unmethylated [5]. Methylation in these regions is usually inversely associated with transcription
or gene expression [5,6]. Methyl groups are covalently transferred to the C5 position of the
cytosine via S-adenosyl methionine (SAM) [5,6]. Transfers from SAM to DNA are catalyzed by
a group of enzymes called DNA methyltransferases (DNMT), and DNMT enzymes are categorized
based on their mutually exclusive functions [5].

2
DNMT1 functions in the mitotic replicative maintenance of DNA methylation marks [5]
and DNMT3 is involved in de novo methylation for mammalian development [6]. DNMT1 is then
essential for cell replication to occur without defect to cellular functions, given the unique
epigenetic landscape of each cell type [5,6]. During DNA replication, generation of the new
“naked” strand produces a hemimethylated site and DNMT1 is recruited to methylate this strand
[5]. As a result, symmetry is maintained across both strands and mitotic maintenance of epigenetic
marks is established [5]. Variability in mitotic cell maintenance of epigenetic marks can cause
defects in organ function, potentially causing disease [6,7].This variability can increase in
likelihood with age, captured with DNA methylation aging clocks, and can also be induced by
external factors like environmental toxicants [6,7]. Internal factors, like stress or disease, can also
incur discrete DNA methylation markers [7,8]. Such DNA methylation signatures may act as
mediators increasing risk of disease in later life [8].
DNMT3 functions in de novo methylation processes [9]. In development, paternal and
maternal epigenetic marks are demethylated from the fertilized zygote until the blastocyst stage,
at the end of the first week after conception before implantation of the embryo into the uterine
lining. Specifically, the paternal sperm undergoes rapid demethylation [9]. This demethylation is
accomplished through TET enzymes and is necessary for the embryo’s organogenesis [9,10].
However, some regions of this genome are resistant to this demethylation, including imprinting
control regions (ICRs) [10]. Then, rapid methylation occurs to orchestrate organogenesis via
DNMT3 [9,10]. The epiblast, or the cell mass with pluripotent capacity to create all organs, gives
rise to primordial germ cells (PGCs) [9,10]. Demethylation occurs at the PGCs and is proceeded
by sex-specific methylation markers by DNMT3 that will result in prospermatagonia in males and
oocytes in females (that will give rise to the next generation of offspring) [10]. For the male fetus’s

3
prospermatogonia, paternal imprints are re-established before birth and maintained to sperm
development at puberty [11]. For the female fetus’s germ cells, oocytes are re-established with
maternal imprints after birth [10].
Variability in de novo methylation patterns can have potential multigenerational
consequences [12]. Toxicants can disrupt fetal epigenetic programming across development,
thereby directly affecting fetal DNA methylation [12]. These effects can potentially affect the
fetus’s germ cell via epigenetic defects to PGCs and therefore the grandchild generation of the
gestating individual regardless of fetal sex [12]. Female offspring are born with a finite number of
oocytes that have been exposed to toxicants during gestation. Male offspring have
prospermatagonia, which serve as precursors to sperm, but are present in a finite amount [10-12].
There is evidence for epigenetic stability between PGCs and prospermatagonia, as well as discrete
modifications in male germlines caused by environmental toxicants inducing methylation profiles
that are stable in cauda epididymal sperm [11]. Therefore, the female and male germline both serve
as a mechanism for grandmaternal exposures to affect grandchild methylation and health outcomes
[10-12].
Given the potential intra- and multigenerational health impacts of the developmental
window on adult health, extensive research has been conducted on prenatal environmental
exposures and its association with DNA methylation in human populations [13]. Data collection
in human populations is often more expensive and intensive than model organism-based research.
Tissues collected from human populations can either represent single cell or “bulk” cell types.
Single cell consists of one cell type (i.e., neuron, epithelial) and therefore the epigenetic pattern
that is functionally relevant for that cell is represented [13]. Single cell data may be more
representative of expression profiles of organs, often used in model organism research [13].

4
Whereas bulk tissue represents a mixture of cell type like whole blood (a composition of
granulocytes, neutrophils, beta cells, etc.) will represent an average profile of methylation across
the cell types in that tissue [14]. Bulk tissue is more widely available in human epigenetic
epidemiology DNA methylation studies because it is easier to collect [13,14].
Epigenetic epidemiology studies are concerned with the contribution of methylation
variability to disease or the impact of environmental exposures on methylation profiles (that may
act as mediators of disease risk) [13,15,16]. Studying the nature of this molecular modification in
humans can be complicated for several reasons. Many studies are focused on an exposure at a
critical period and DNA methylation as an outcome [13,15,16]. If only one timepoint of DNA
methylation is used, the temporal dynamics of when the exposure induced the signature is unclear,
and whether the signature is truly a causal effect or just an association is unknown [17].
Longitudinal follow-up with multiple DNA methylation timepoints can allow for exploration of
the temporal nature of DNA methylation dynamics [17]. When examining DNAm in a case-control
analysis, risk factors versus disease-specific signatures cannot be distinguished and reverse
causation may confound results [15,17]. Genetic variation also has a massive impact on DNA
methylation variability, and delineating SNP-level variation cannot be accomplished without
genotyping data [15,17]. Therefore, human studies of DNA methylation without coupled genetic
data are confounded [15,17].
Epigenetic epidemiology studies focus on exposures during the developmental period due
to its potential to affect adult health outcomes in the offspring [13,15]. Maternal exposures and
habits are the focus of these studies and offspring cord blood or pediatric DNA methylation
signatures act as the outcome, reflecting aberrations to de novo methylation processes. Some
studies also extend into mediation analysis to test the effect of environmentally induced DNAm

5
signatures on an outcome, where DNAm is considered a mediator between a periconceptional
exposure and an outcome [16].
There is an abundance of research on maternal habits and exposures on offspring DNAm
[16], and conspicuously much less research has been conducted on environmental exposures
influencing maternal DNAm [18-20]. Pregnancy is a pivotal period for women with far-reaching
health consequences, and aberrations to mitotic maintenance as an effect of exposures during
gestation may increase risk of later life disease [20]. Simultaneously, there has also been a dearth
of research on multigenerational epigenetic effects from maternal factors. This leaves a critical
knowledge gap on the effects of maternal exposures on maternal epigenetic or multigenerational
epigenetic signatures [13,20]. Moreover, the potential for changes in maternal epigenetics affected
by exposures during gestation potentially affecting multigenerational DNA methylation signatures
is also understudied. Upstream biological changes in the maternal body, as an effect of toxicants,
may also have direct consequences on de novo methylation processes in the fetus – potentially
though differential response to hormonal cues in gestation [13,15,20,21]. This thesis aims to
address these gaps in the literature.
This thesis was conducted using data from two Southern California cohorts: the prospective
pediatric Children’s Health Study (CHS) [22] and the prospective pregnancy Maternal and
Developmental Risks from Environmental Stressors (MADRES) [23]. The CHS is a prospective
pediatric cohort that investigated the impact of air pollution on pediatric respiratory health that
began recruitment in 1993. A subset of CHS participants along with their partner and child,
conditional on the participant’s child having an age less than 7, were followed-up on in 2013
(Figure 1, Figure 2). MADRES is a prospective pregnancy cohort investigating the impact of
environmental and social stressors on mother-child pairs that began recruitment in 2015 (Figure 3,

6
Figure 4) [23]. Ambient air pollution, as well as self-reported grandmaternal prenatal smoking
exclusively in the final study, will be the exposure used in this thesis.
Ambient air pollution is a broad term used to categorize outdoor air pollution exposure. It
consists of all suspended and sustained compounds present in the air [24]. Six ambient air pollutant
compounds have been identified the EPA to pose significant harm across the United States
including particulate matter, photochemical oxidants, carbon monoxide, sulfur oxides, nitrogen
oxides, and lead [24]. This these will use 24-hour nitrogen dioxide (NO2), 8-hour maximum
ground-level ozone (O3), particulate matter with aerodynamic diameter less than 2.5 microns
(PM2.5), and particulate matter with aerodynamic diameter less than 10 microns (PM10).
Sources of particulate matter include dust, dirt, soot, smoke, or drops of liquids [25]. These
sources can either be primary or emitted directly from their source (e.g., forest fires), or secondary
sources that form particulate matter from atmospheric chemical reactions [25]. Particle size bins
PM compounds into either PM2.5 or PM10, the former of which can travel into the lower respiratory
tract due to its small size [25]. Nitrogen oxides are primarily produced via combustion of nitrogen
interacting with the oxygen in the atmosphere (gas emitted from vehicles, farming equipment,
boats), or can also be emitted from natural sources (e.g., lightning) [24]. Volatile organic
compounds (VOCs) and nitrogen oxides interact with sunlight at ground level to produce ozone.
Due to the impact of sunlight on ozone formation, this study focuses on the 8-hour peak of ozone
for each day [25].
Ambient air pollution in Southern California increases risk of some respiratory outcomes
[25]. Though concentration of ambient air pollutants has decreased in the region since 1994 with
introduction of the emission regulatory policies in Southern California, it has not decreased stably
for all groups [26]. Reduction in ambient air pollution concentration has also plateaued in recent

7
years [26]. Extremes in climate in Southern California have increased since the 1970s, including
duration and size of wildfires, which contributes to ambient air pollution concentrations [27,28].
High temperature extremes can also exacerbate the effects of ambient air pollutants, with their coexposure increasing risk of mortality greater than their independent effects combined [27,28].
This thesis is concerned with the effect of ambient air pollution concentrations on DNA
methylation. Measurement of ambient air pollutant concentration to assign outdoor air pollution
exposure may be prone to error. Ambient air pollution, in this thesis, was assigned using measures
from inverse distance squared weighted spatial interpolations [29]. Distance from Environmental
Protection Agency (EPA) monitors is used in CHS and MADRES, with additional community
monitors installed in the CHS. Local effects of outdoor ambient air pollutants may be inaccurate
(e.g., living near a major roadway). This approach also assumes that participants are constantly at
their homes, and therefore ignores the day-to-day nature of outside air pollution exposure
depending on the participant’s normal schedule [29,30]. Moreover, participant preferential aspects
(e.g., windows open versus closed) may also impact the concentration of outdoor ambient air
pollution someone is exposed to. We expect this measurement error to have a non-differential
impact on the association between air pollution and DNAm.
The first study of this thesis aims to address the effects of PM2.5 and ambient air pollutant
mixtures on maternal DNA methylation trajectories in pregnancy in MADRES. The persistence of
PM2.5 and ambient air pollutants into the postpartum period will be explored using postnatal
maternal DNA methylation signatures in the CHS. The second study aims to address the effect of
ambient air pollution mixtures on both maternal and neonatal infant DNA methylation profiles.
The effect of ambient air pollutant exposure on each generation will be conducted independently,
and then the potential for changes in maternal epigenetic signatures affecting infant DNA

8
methylation will also be conducted. The final study aims to explore evidence for a direct
intergenerational effect of a maternal gestational exposure on child and grandchild epigenetic
profiles. The potential for parental life course exposure to exert synergistic effects on child DNAm
affected by grandmaternal exposure will also be explored (Figure 5).
Figure 1. Schematic of CHS Data Collection Procedures
Created with BioRender.com
Figure 2. CHS Recruitment Locations

9
Image taken from “Emissions reduction policies and recent trends in Southern California's
ambient air quality”. J Air Waste Manag Assoc. 2015
Figure 3. Schematic of MADRES Data Collection Procedures

10
Created with BioRender.com
Figure 4.

11
Image taken from presentation by Rima Habre, PhD “MADRES Center for Health Disparities
External Advisory Committee Meeting” May 19,2021.
Figure 5. Schematic of Thesis Projects

12
CHAPTER 2: OVERVIEW OF QUANTILE G-COMPUTATION AND
ITS APPLICATIONS IN EPIGENETIC EPIDEMIOLOGY
In environmental epigenetic studies of model organisms, researchers investigate the effect
of a compound on an epigenetic state [15]. Environmental exposures in human populations rarely
occur in isolation, as humans are exposed to a mixture of compounds across the life course. The
effect of an exposure mixture of several components can have an entirely different impact than its
singular components independently [31]. These mixtures can either have a consistent effect across
components that demonstrates a synergistic effect, or components of mixtures have counteracting,
or antagonistic, effects on an outcome [31]. Quantile g-computation is a statistical method used to
investigate the impact of exposure mixtures, and its components, on a health outcome. Quantile gcomputation can estimate the average mixture effect, as well as the effect of the mixture’s
component, on a health outcome [31,32]. The average mixture effect estimates the risk of
increasing each of the mixture’s components together by one quantile on an outcome [31].
Quantile g-computation is an extension of g-computation, which estimates counterfactuals
using standardization via a joint marginal structural model [32]. Quantile g-computation uses the
joint marginal structural model for estimating counterfactuals in g-computation in conjunction
with a quantized exposure [30,33,34]. The estimation of counterfactuals in the joint marginal
structural model in quantile g-computation requires the fulfillment of causal assumptions:
consistency, no interference, positivity, and no unmeasured confounding [31,33,34]. Consistency
requires specific, well-defined treatment groups, no interference requires that the treatment of one
group does not affect the outcome in another, positivity assumes that the study population is

13
capable of being exposed, and no unmeasured confounding assumes that there is no mixing of
effects in your exposure-outcome relationship of interest [33,34].
Quantile g-computation first transforms each component of a mixture into its respective
quantile [31]. Then, a linear model is implemented with the quantized exposure on a health
outcome, considering the impact of each component’s weights and adjusting for relevant
confounders [31]. Each component’s weights are defined as the effect size for that specific
component divided by the sum of the effect sizes for all the components [31]. Weights can either
by negative or positive for each component, indicating that component’s relative effect compared
to the entire mixture [31].
Applications of quantile g-computation to DNA methylation are limited. Xu et al
demonstrated an antagonistic effect between maternal smoking and maternal plasma folate on
offspring CYP1A1 methylation [35]. On a genome-wide scale, phthalates, and bisphenols
(demonstrating a mixture of effects to endocrine disrupting compounds) was not associated with
DNA methylation in cord blood [36]. This thesis will be testing for potential synergistic or
antagonistic, effects of similar compounds (particulate matter and smoking) of related mixtures.
Therefore, this thesis will be some of the first research published using quantile g-computation on
an epigenome wide scale.

14
CHAPTER 3: LITERATURE REVIEW OF AIR POLLUTION’S
EFFECT ON DNA METHYLATION PROFILES
The following will be a short literature review aimed to highlight consistent DNA
methylation signatures affected by air pollution. The goal of this literature review is not to focus
on sources of bias across the literature, but rather to highlight consistent DNA methylation
signatures identified and their direction of effect. The review included all meta-analyses in
PubMed when searching for articles including text words with “DNA methylation” and “air
pollution” and restricting to meta-analysis in filters. A total of 3 meta-analyses were included.
A meta-analysis was conducted of 11 epigenome-wide association studies (EWAS) and 16
global methylation levels on the effect of air pollution and DNA methylation in adults. Most of
these studies were conducted in either Western European or East Asian populations. Most studies
were assessing effects of long-term air pollution on DNA Methylation in adults between 45-75
years old. Studies either investigated short-term or long-term air pollutant exposure. There were
loci displaying some evidence of differential methylation ranging between 13 and 1345 CpGs from
cohorts 100 to 2956 in sample size. However, there were no loci consistent across any of these
studies among the EWASs, and only some evidence of consistency in the NXN gene with increased
methylation after exposure across two studies. Meta-analysis of global methylation levels
indicated decreased methylation with PM2.5 exposure were not significant with a confidence
interval overlapping the null B = -0.39 [-0.97, 0.19]. KEGG pathways were enriched for pathways
in African trypanosomiasis and Malaria in long-term effects of air pollution, and antifolate
resistance in short-term exposure pathways [37].

15
Another meta-analysis examined the effect of prenatal particulate matter exposure on
neonatal DNA methylation signatures in the Pregnancy and Childhood Epigenetics (PACE)
consortium (n = 1,949 for PM10 and n = 1,551 for PM2.5) and replication in a sample of newborns
from ALSPAC (n = 688). Replication was also explored in other cohorts with DNA methylation
profiles from 7-9 years of age. Discovery included 9 Western European and mainly White
American populations. Four CpGs were significantly increased with increasing PM10 in GNB2L1,
SNORD96A, FAM13A, SRPRB, P4HA2, and two CpGs were decreasing in USP4 and NOTCH4.
There was decreased methylation with increasing PM2.5 in PLXNNA4, ZNF705A, C14orf2,
TMCOR3, SFRS8, NEUROG1, and MRI1 increased methylation in 3 CpGs including PSG5,
C7orf50, and FINP1. There were no consistencies between the PM2.5 and PM10 results, though
different cohorts contributed to those respective effect sizes. DMRs persisted for the PM2.5 effect
in C7orf50, ZNF705A, PLAT, PSG5, and MRI1. There was no replication into early childhood
among the PM10 significant sites in the ALSPAC cohort. There was evidence of persistence into
later childhood in the BAMSE and MEDALL samples (between 7 and 9 years old) in the NOTCH4
and FAM13A genes, but NOTCH4 was not in a direction consistent with the discovery analysis.
For PM2.5, there were replicated findings in the MRI1 gene for adolescents in the HELIX sample,
but the beta coefficient direction was not consistent in replication. However, both H19 and
MARCH11 were significantly identified by DMRcate across PACE and ALSPAC. There was also
some overlap in nominally significant CpGs (n = 359) in PACE and 6,073 FDR significant CpGs
related to maternal smoking in pregnancy in another meta-analysis [38].
PACE published another meta-analysis of prenatal nitrogen dioxide exposure and neonatal
DNA methylation signatures in a sample size of 1,508 for the discovery analysis. Look-up analyses
were conducted in cohorts with children aged 4 (n = 733) and 8 (n = 786). There were significant

16
associations with LONP1, cg24172570 (upstream of HIBADH), and SLC25A28 – with only
increased methylation associated with increasing exposure in SLC25A28. Importantly, adult
reference cell composition estimates were adjusted for in this study (not neonatal cell composition
estimates – which include nucleated red blood cells), and therefore bias may be induced by residual
confounding [39].
Very few of the studies included across these meta-analyses adjusted for single nucleotide
polymorphism (SNP) effects, and therefore are confounded by genetic variation. The effects in
these meta-analyses, when unadjusted for genetics, represents a mixture of both SNP-level and
potentially environmental effects. Risk of environmental hazards may impact marginalized
communities or geographic areas with racial and/or ethnic structure, in addition to the direct impact
of genetic variation on DNA methylation [40].
Some issues across the meta-analyses include lack of diversity in sample populations, lack
of consistency across replicated effects, lack of replication, and no exploration of mixture effects.
Moreover, inconsistent use of cell adjustment may be contributing to confounding. Consistency
across replicated effects may be affected by sources of particulate matter which may differ across
studies and geographic location – there was evidence of inconsistent effects of prenatal PM10 on
offspring methylation in NOTCH4 and FAM13A. Moreover, of all the adult populations tested,
none of them were conducted specifically in pregnant people. This thesis aims to specifically
address these gaps by investigating the effect of ambient air pollution mixtures and particulate
matter on maternal prenatal and postnatal profiles [33,34,37-39].

17
CHAPTER 4: LITERATURE REVIEW OF PRENATAL MATERNAL
SMOKING AND ITS EFFECT ON OFFSPRING DNA
METHYLATION
The goal of this literature review is to highlight consistencies across meta-analyses with
regards to offspring DNA methylation signatures affected by maternal smoking in pregnancy. A
total of three meta-analyses will be highlighted. The PACE consortium conducted a meta-analysis
of 6,685 newborn DNA methylation signatures and prenatal smoking. The prevalence of exposure
was between 13% sustained prenatal smoking at 25% any prenatal smoking during pregnancy.
Among the replication cohort with older children, only 8% and 13% reported persistent or ever
smoking during pregnancy. The discovery meta-analysis found 6,063 CpGs (FDR < 0.05) related
to persistent maternal prenatal tobacco smoke exposure with 52% exhibiting coefficients with
increased methylation. This effect persisted into later childhood with all CpGs displaying
nominally significant persistent associations, and 148 met FDR significance. Among these
persistent effects, 73% were consistent in the same direction in older childhood and 61% had an
attenuated effect estimate. Among the 148 persistent CpGs with FDR significance, all CpGs had
a consistent direction of effects with the discovery cohort – including MYO1G, AHRR, and
CYP1A1 which have all previously been associated with signatures of smoking. The most
significant association was in the AHRR gene, of which there were multiple CpGs with different
directions of association (increased or decreased methylation). In the any smoking meta-analysis,
there was a reduction in the number of findings (4,653 CpGs) [41].

18
A meta-analysis was conducted to assess the effect of prenatal smoking to DNA
methylation signatures in adolescence and adulthood in 2821 individuals from five cohorts [34].
There was evidence for differential methylation in 69 CpGs. These 69 CpGs were consistent in
direction of effect to a meta-analysis in newborns [33]. This included MYO1G (increased
methylation), CYP1A1 (increased methylation), CNTNAP2 (decreased methylation), FRMD4A
(increased methylation), GFI1 (decreased methylation), AHRR (increased methylation), NRP2
(decreased methylation), ARL4C (increased methylation), C11orf52 (increased methylation),
CACNB4 (increased methylation), RUNX1 (increased methylation), ETHE1 (increased
methylation), ZNF395 (decreased methylation) RHOA (decreased methylation), PRXX2 (increased
methylation), OLFM1 (increased methylation), FAM184B (increased methylation), ANXA4
(increased methylation), PKNOX2 (increased methylation), and C16orf70 (increased methylation).
Sensitivity analysis included restriction to those offspring who did not report regular smoking, and
the significant findings were still consistent. There was also evidence for a dose-response
association with maternal smoking and CYP1A1 methylation. Moreover, adjustments for maternal
smoking had no effect on the estimates. Finally, there were no longitudinal trends observed in
changes in DNA methylation in these CpGs – indicating a static and stable effect across childhood
[42].
Another meta-analysis was conducted in placental DNAm data with 443 CpGs identified
to be associated with persistent maternal smoking in pregnancy in a sample of 1,700 placentas.
Across cohorts, there was more consistency in results among the CpGs associated with persistent
versus any maternal smoking. The largest magnitude of effects was in the EDC3, WBP1L, and
KDM5B genes. Most of these CpGs were downregulated among smoking mothers. Findings also

19
included AHRR hypomethylation as an effect of maternal smoking, a direction inconsistent with
the neonatal DNA methylation literature [43].
There have also been two studies that examined the effect of grandmaternal smoking during
pregnancy and differential methylation in grandchildren in ALSPAC with 1226 individuals. One
study conducted in ALSPAC used two levels of stratification by sex: at the parental and grandchild
level producing a total of four sex specific EWASs for the maternal and paternal lineage effect.
There were no significant associations at the Bonferroni adjustment level when examining the
effect of grandmaternal smoking across all the EWASs. There was some evidence for the effect of
maternal grandmother on female grandchild methylation at 15-17 years old at cg19782749, with
increased methylation because of exposure. At birth, there was evidence for differential
methylation in cg19426678 because of maternal grandmother smoking among all grandchildren.
Among paternal grandmother, there was evidence for differential methylation among
grandchildren 15-17 years old in cg27456137 on the X chromosome and cg15068552 on
chromosome 7. There was also some evidence of a paternal grandmother effect on female
grandchildren in cg22682200 on chromosome 10 and cg26827966 on chromosome 5, both
associated with decreased methylation in smoking. None of these sites were consistent with
maternal DNA methylation smoking signatures [44].
In the Isle of Wight cohort, a similar analysis was applied but in a subset of 161
grandchildren. No sex-stratified analyses were conducted. Moreover, this analysis was restricted
to methylation heritable via the maternal lineage. Twenty-five CpGs sites were associated with
grandmaternal smoking including SIGLECL (increased methylation), IGFALS (increased
methylation), CACNA1C (decreased methylation), MYOM2 (decreased methylation), LY6GC
(decreased methylation), LRRC32 (decreased methylation), FLJ32810 (increased methylation),

20
ANKRD31 (increased methylation), UVRAG (increased methylation), TXNL4A (decreased
methylation), CRYBG3 (decreased methylation), IGF1R (decreased methylation), MYOM3
(decreased methylation), FOXO1 (decreased methylation), DCAF4 (decreased methylation),
VAMP8 (decreased methylation), SHISA7 (increased methylation), BRUNOL4 (decreased
methylation), PRDM6 (increased methylation), ANUBL1 (increased methylation), LOC728989
(decreased methylation), C6orf99 (decreased methylation), SDC1 (decreased methylation), STK24
(increased methylation). Notably, none of these overlapped with models of maternal smoking.
There was no evidence of DNA methylation differences in known smoking affected CpGs
including CYP1A1, AHRR, and GFI1 [45].
Overall, there is stronger consistency and evidence for persistence of smoking-related
DNA methylation signatures, especially compared to particulate matter. There are replicated
CpGs across cohorts including AHRR, CYP1A1, and GFI1. However, there is much less
consistency with regards to grandmaternal smoking effects, though this is much less data
available on this matter (as there were no meta-analyses available on grandmaternal effects) [41-
45,33,34].

21
CHAPTER 5: CHANGE IN MATERNAL DNA METHYLATION
ACROSS PREGNANCY IN RELATION TO PARTICULATE
MATTER AND AMBIENT AIR POLLUTION EXPOSURE
Abstract
Background
Pregnancy involves multifactorial biological changes to support the metabolic needs of
mother and fetus, some of which may have the potential to impact later maternal or child health.
Biological changes such as DNA methylation may be affected by environmental exposures
experienced in pregnancy, the consequences of which are largely unknown for downstream
maternal health.
Objectives
This study aimed to identify differential methylation between early and late pregnancy
whole blood prenatal samples (N = 128) in the MADRES pregnancy cohort. We assessed the
impact of gestational age (GA) as well as the effects of particulate matter with aerodynamic
diameters <2.5 μm (PM2.5) on change in DNA methylation across pregnancy (N = 127). The
persistence of prenatal PM2.5-driven DNA methylation signatures across tissue type and time was
evaluated in a separate sample of postnatal mothers (N = 31). The effect of ambient air pollutant
mixtures (PM2.5, PM10, nitrogen dioxide NO2, and ozone O3) on change in maternal DNA
methylation trajectories was evaluated as well in MADRES.
Methods

22
DNA methylation loci and regions that changed over time in pregnancy were identified
using adjusted linear models accounting for differences between participants. We then evaluated
whether a two-way interaction with gestational age (GA) to evaluate whether the effect of PM2.5
differed across gestation. Persistence of prenatal PM2.5 methylation changes was evaluated in a
sample of postnatal mothers in the Children’s Health Study (CHS) using adjusted linear models.
The effect of prenatal ambient air pollution mixtures on DNA methylation trajectories was
estimated using quantile g-computation. Overlapping significant results across the models were
reported. FDR adjusted p-values were used for significance.
Results
We identified 0.75% of CpGs (5,139) on the EPIC array having significantly differential
DNA methylation values between early and late pregnancy in adjusted models. PM2.5 was
associated with methylation differences in 136 CpGs and 33 DMRs, none of which overlapped
with the CpGs identified to differ between early and late pregnancy. There was some evidence for
persistence of prenatal- PM2.5 methylation changes among 1 CpG in CHS. 40 CpGs were identified
to differ between early and late pregnancy in response to prenatal ambient air pollution mixture
exposure.
Discussion
A subset of CpGs were identified that changed over time in pregnancy. Significant CpGs
identified to change in pregnancy were not modified by prenatal PM2.5 exposure. Instead, PM2.5
CpGs displayed GA-dependent effects in unique CpGs. The effect of PM2.5 differed across
gestation, but there was only some evidence that this effect persisted in a small postnatal sample.

23
Ambient air pollution also influenced some CpGs, which did not overlap with either the main
effect or PM2.5 analysis.
Introduction
Pregnancy is a multifactorial biological process that increases the energetic demands of the
maternal body to support fetal organ development, including changes to the respiratory system.
Increases in progesterone during pregnancy result in threshold decreases to the central chemoreflex
to carbon dioxide, reducing arterial carbon dioxide levels [46,47]. Uterine enlargement, in
response to fetal growth, shifts the diaphragm, reducing expiratory reserve volume [46,47]. As an
effect of these changes, minute ventilation is increased across pregnancy, increasing risk of
hypoxia and sensitivity of the respiratory system to environmental pollution [47].
Prevalence of pregnancy complications and more generally of high-risk pregnancies and
associated risk factors has increased with time [49]. Public health campaigns and research have
focused on the effect of environmental or psychosocial stressors on infant outcomes, but similar
efforts have not been made in maternal perinatal care [20,49].
Improvement in maternal care can reduce risk of adverse outcomes in pregnant people.
Investigating the biological changes associated with pregnancy, and environmental contaminants
that modify these changes, can aid in these efforts [20,49,50]. One molecular mechanism that can
capture change is DNA methylation (DNAm). DNAm is an epigenetic marker associated with the
addition of a methyl group on the 5-carbon cytosine ring of DNA, reducing expression within
promoter regions [50]. Prenatal molecular factors, like DNAm, may interact with environmental
factors to increase risk of health outcomes in the mother [20,49-51]. Environmental literature
surrounding prenatal health and DNAm focuses on adverse outcomes of the fetus and modifiable

24
maternal exposures and there are few longitudinal studies focused on biological changes in
pregnant people. Among these few cohorts, trends were observed for overall decreases in
methylation [51] and increases in expression [52] during gestation. Less is known about the impact
of environmental exposures on maternal gestational DNAm signatures [20], though meta-analyses
have been published in other adult [37], neonatal [38], or pediatric [39] populations. Effects of
ambient air pollutants, specifically particulate matter with aerodynamic diameters less than 2.5
(PM2.5), on neonatal DNAm has been of interest due to its hypothesized to disrupt fetal epigenetic
reprogramming (small size allows it to penetrate the placental barrier) [50]. However, this small
size can also aggravate the lungs of the pregnant person. PM2.5 can travel into the lower respiratory
tract due to its small size and induce oxidative stress in alveolar cells and potentially other cells
via diffusion into the bloodstream [20,53]. Pregnant people are at specific risk to PM2.5 and other
ambient air pollutants because of the increased ventilation and mechanical changes in lung
physiology across gestation [20,46-47].
This study aimed to estimate differences in prenatal DNA methylation between early and
late pregnancy. Then, PM2.5. induced DNA methylation changes were analyzed. A mixture model
was also used to evaluate the effect of ambient air pollution mixtures of DNA methylation changes.
Persistence of these PM2.5 signatures was additionally evaluated in a subset of postnatal mothers
in a replication study.
Methods
Discovery Study Sample and Recruitment
The Maternal and Developmental Risks from Environmental Stressors (MADRES) is a
prospective pregnancy cohort in Los Angeles, California which began in 2015. Pregnant people

25
were eligible for recruitment into the MADRES prospective pregnancy cohort if they were <30
weeks gestational age at cohort entry, HIV negative, and were incarcerated at enrollment.
Participants were interviewed at each trimester, at birth, and postnatally. Additional information
on variables collected and exclusive criteria can be found here [23].
MADRES Maternal DNA methylation
EDTA tubes (BD # 366643) were used to collect blood samples, which were transported
upright on ice to the lab at Norris Cancer Center and centrifuged with overlying plasma removed
within an hour of sample collection. 10 mL of peripheral blood samples were collected during
early and late pregnancy, and PBMCs were isolated from these samples (Figure S1). The AllPrep
DNA Kit (Qiagen) was used to extract DNA, and within one hour of sample collection plasma
aliquots were stored. The EZ DNAm Kit (Zymo Research) was used to perform bisulfite
conversion of DNA and was performed in batches with families plated together. Illumina Infinium
HumanMethylationEPIC (850K) assay was used to quantify DNAm according to the
manufacturer’s recommended protocol with no other modifications.
MADRES Quality Control
Data analysis was performed in RStudio v4.1.2 [103]. Minfi was used to conduct probe
and sample level quality control [54]. Probes displaying a signal intensity indistinguishable from
negative control probes were removed from the analysis [55], in addition to polymorphic and
cross-reactive probes [56,57], sex, or contained a SNP at the interrogation or single nucleotide
extension were dropped [54]. Samples with median probe intensities less than 10 were removed
[54]. Multidimensional scaling (MDS) plots were additionally used to screen for outliers [54]. No
outliers were removed based on the plot. Dye-bias correction [58] followed by quantile

26
normalization [59] was used for sample-level normalization. Samples displaying discrepancies
between reported sex at birth signal intensity generated from the X and Y chromosome-associated
probes on the array were dropped. This analysis was restricted to participants with both paired
samples surviving quality control. Figure S2 is a consort diagram.
CHS Post-Natal DNA methylation
Postnatal samples were used from the CHS, a prospective pediatric cohort study that
initially began recruiting California schoolchildren in 1993 [22]. Two decades after the initial
follow-up, families were invited to participate in a follow-up mail-based study and buccal cell
samples were collected from the adult participant, their partner, and child. A subset of these
participants (N = 37) were females who had reported giving birth with available coupled
gestational ambient air pollution data. Informed consent was provided (IRB #HS-17-00778). The
Oragene prepIT L2P kit was used to extract DNA locally at -20°
C. DNAm was quantified with the
EZ DNAm Kit (Zymo Research) and was performed in batches with no modifications in the
manufacturing protocol, like the discovery population. Five samples were removed to sex
discrepancy, 1 for missing covariates, and 1 due to restriction for the toothbrush collection method
for consistency.
Cell Type Estimation
The EPIC reference platform was used with the Houseman Method in Minfi using the
“FlowSorted.Blood.EPIC” data for whole blood [14]. Composite cell type for blood was used for
maternal data (Salas 2022). For buccal cell data in the CHS, immune cell type proportions were estimated
using HEpiDISH [59].
Genetic Factors

27
Ancestry based SNP-variation impacts DNA methylation and can bias results. We used
EPISTRUCTURE principal components produced from the quality-controlled DNA methylation
datasets to help address this issue. The top two EPISTRUCTURE principal components were used
as adjustment variables in analyses [60].
Measurement of Gestational Age at Sample Collection
Difference in gestational age in days between the infant’s date of birth and sample
collection was calculated. Then, this delta was subtracted from gestational age at birth, abstracted
from a combination of medical records and self-report to generate gestational age at sample
collection. This measure was analyzed continuously.
Measurement of PM2.5 in CHS and MADRES
Residential addresses were collected prospectively during prenatal visits. Inverse-distancesquared weighted spatial interpolation from ambient air quality monitoring data (with an average
of four monitoring stations within 8 to 14 km of each residential address) was used to generate
estimates of 24-hour particulate matter with aerodynamic diameter less than 2.5 μm (PM2.5 ).
Weekly gestational PM2.5 from the beginning of pregnancy (gestational week 0) to late pregnancy
sample collection was averaged for exposure assignment. Distribution of PM2.5 is in Figure S3.
Covariate Selection
Confounders were identified using a directed acyclic graph (Figure S6). Confounder
adjustment including pre-pregnancy body mass index (BMI), total weight gain across pregnancy,
binary diabetes status (gestational diabetes, chronic diabetes, glucose intolerance), binary
hypertensive status (preeclampsia, eclampsia, preeclampsia-eclampsia, chronic hypertension),

28
ethnicity, origin (White non-Hispanic non-foreign born, Black non-Hispanic non-foreign born,
Hispanic foreign born, Hispanic non-foreign born, and multiracial non foreign-born participants),
the top two EPISTRUCTURE principal components [60], maternal age at birth, parity status, and
maternal education (less than high school, high school or some college, college degree). Covariates
to control for technical artifact include cell type change across the paired timepoints and batch.
Additionally, two surrogate variables produced from surrogate variable analysis was used to adjust
for additional unmodeled technical variation [61]. SVA estimates unwanted sources of
heterogeneity [61]. Three participants were removed from the analysis from self-reported maternal
smoking in the first and third trimester, given the low sample size (N = 258). Additionally, three
participants were removed from the PM2.5 and ambient mixtures due to missing data (N = 254).
For the PM2.5 and ambient mixtures analysis, average temperature, and season of first sample
collection (Fall, Winter, Spring, Summer) were adjusted for.
Statistical Analysis
The first objective was to estimate differential methylation between early and late
pregnancy in our paired samples. Next, we evaluated evidence for differential methylation between
early and late pregnancy that changed in response to an increase in PM2.5. We also examined the
effect of prenatal ambient air pollution mixtures on DNAm changes between early and late
pregnancy (in a non-GA specific manner) using quantile g-computation in the qgcomp package
[31].
Linear models for change in DNA methylation by GA were implemented in limma [92]
and DMRcate [93], accounting for paired correlation within subjects (duplicateCorrelation) to
estimate differentially methylated positions (DMPs) in our analytic sample (n = 129, N = 258).
Identical models were implemented in DMRcate to generate differentially methylated regions

29
(DMRs) associated with GA. Interaction tests between continuous PM2.5 and gestational age at
sample collection were executed in limma and DMRcate (n = 127) [96,97]. Models were adjusted
for all covariates mentioned above. The FDR p-value was used to determine significance.
Quantile g-computation (n = 127) was implemented in the qgcomp package [31] and
adjusted for all covariates and confounders using normalized m-values as the outcome and
including a random effect for ID to adjust for baseline differences between participants.
We evaluated the potential for persistent methylation signatures postnatally in CHS
performing a look-up analysis among loci that were significant in the PM2.5 and GA interaction in
MADRES. Then, adjusted linear models for immune cell proportion, batch, maternal age, and
child age in addition to an SVA model were implemented in limma. Unadjusted p-value < 0.05
were reported.
Pathway Analysis
DMPs were tested for enrichment of Kyoto Encyclopedia of Genes and Genomes (KEGG)
and Gene Ontology (GO) terms using gometh [62,63]. This method accounted for the number of
probes per gene. Enriched terms with an FDR less than 0.05 were reported.
Results
Participant Characteristics
Table 1 examines participant characteristics by PM2.5 quartiles. Significant (p < 0.05)
differences existed between groups in pre-pregnancy BMI, ethnicity and place of birth, maternal
education, and parity status. Individuals in the lowest PM2.5 quartile were more likely to have a
lower BMI than quartile two or three and were also more likely to be multiparous. Hispanics born

30
in the United States were the more prevalent ethnic and place of birth category across all quartiles
of exposures.
Differential Methylation Across Pregnancy
We identified 5,139 differentially methylated CpGs (FDR < 0.05) between early and late
pregnancy in pregnancy participants, or 0.75% of CpGs on the EPIC array (Figure 1). 63% of these
CpGs decreased in methylation across the gestational period (Figure 2). 880 DMRs displayed an
FDR < 0.05 in DMRcate. DMRs with the strongest difference in average regional beta value
between early and late pregnancy were in the HOXB-AS3, HOXB3, and HOXB4 genic regions, all
of which were associated with decreased methylation with increasing gestational age.
Particulate Matter 2.5 Influences Prenatal DNA Methylation Trajectories
Prenatal PM2.5 was associated with change in methylation between early and late
pregnancy in 136 CpGs (Figure 3), with the largest magnitude of effect for a locus in the island
region of the FRG1 gene and was associated with a 0.002 log-fold increase in methylation in
pregnancy per IQR (IQR = 2.4 µg/m3
) increase in PM2.5. TRIM28 was associated with a 0.0004
log-fold increase in methylation across pregnancy per IQR increase in early and late gestational
averaged PM2.5 exposure (Figure 4). 33 DMRs were generated in DMRcate among the 136 CpGs.
The FBRSL1 DMR had a 0.0004 average regional beta value increase difference between early
and late pregnancy per IQR increase in PM2.5 (Figure 5).
Ambient Air Pollution Mixtures Influence Prenatal DNA Methylation Trajectories
23 CpGs were identified to significantly decrease (FDR < 0.05) in methylation across the
paired prenatal samples and 7 were increased in methylation in response to prenatal ambient air
pollution mixture exposure. The largest magnitude of effect was in DEFA3 with a 1.7-unit m-value

31
decrease in early versus late pregnancy per quartile increase in ambient air pollution mixture (Ψ =
-1.7, 95% CI [-2.2,-1.1], unadjusted p-value = 1.1 x 10-8). 10 of the 30 CpGs with the largest effect
size were reported in Table 2.
Persistence Replication in CHS
Among the 136 significant PM2.5 CpGs in MADRES, cg07465899 displayed persistent
DNA methylation signatures in a consistent increased direction in a sample of postnatal mothers
in CHS (p = 3.4 x 10-4) (Figure 6).
Pathway Analysis
DMPs were enriched (FDR < 0.05) in 6 GO processes (Figure 7) and 2 KEGG terms using
goMeth. Enriched KEGG terms included 192 CpGs enriched in the chemokine signaling pathway
(FDR = 0.036) and 188 CpGs enriched in transcriptional misregulation in cancer (FDR = 0.036).
The most enriched GO term was “response to stress”. There were no enriched KEGG or GO terms
in the PM2.5 analysis.
Discussion
In this study, we found evidence for changes in DNA methylation across pregnancy in
5,339 CpGs and 880 DMRs. These loci were enriched in GO pathways associated with response
to stress and the immune system process (Figure 6). Certain characteristics of exposures in
pregnancy could alter cross-pregnancy DNA methylation levels. Exposure to PM2.5 alone
influenced differential methylation across pregnancy in a subset of 136 CpGs that have been
previously associated with particulate matter in human populations. We also found evidence that
the mixture of ambient air pollution also contributes to prenatal methylation changes in 30 distinct
loci.

32
Only two other studies have evaluated DNA methylation at multiple time points in
pregnancy. Our study results replicated more than 75% of the 196 significant CpGs in the Isle of
Wight Study [51] (Figure S4) and were consistent in the direction of effect. The Isle of Wight
found most significant CpGs to be lower in late than early pregnancy (91%) in their analysis [51],
but among all significant CpGs in MADRES, only 63% were downregulated. The most frequent
gene in the significant GA CpGs was HOXB3, a placental homeobox gene, that was also frequent
(>3 CpGs) in the Isle of Wight study [51].
We observed that the effect of PM2.5 on DNA methylation differed across pregnancy in
136 CpGs. Some findings in the PM2.5 model overlapped with previous literature on ambient air
pollutants and expression changes, including TRIM28 in human lung epithelial cells [64], COQ6
in mouse models [65], and ACP1 in humans employed in trucking [66]. In our study, PM2.5 was
associated with increased methylation (Figure 2) across pregnancy in the island region of TRIM28.
Li et al reported that TRIM28 deficient pregnant mice had a higher incidence of embryos with
morphological abnormalities - potentially caused by a differential uterine response to hormones
[67]. TRIM28 is also an essential gene for maintaining genomic imprinting throughout pregnancy
[67,68]. Both maternal and zygotic TRIM28 bind to the methylated allele of all imprinting control
regions to protect from genome-wide demethylation in these regions. When maternal TRIM28 is
downregulated, there can be potential incomplete penetrance of imprinting effects in offspring
[67,68]. Defects to imprinting maintenance can modify developmental trajectories in the fetus,
subsequently increasing risk of offspring health outcomes.
Though no DMPs were consistent in differential methylation across pregnancy methylation
and PM2.5 affected CpGs, there were overlapping DMRs including ESYT2. In the main effect
model, ESYT2 were associated with lower methylation in late compared to early pregnancy. In the

33
PM2.5 models, ESTY2 are associated with higher methylation in late versus early pregnancy, or an
antagonistic difference between the impact of gestational age versus PM2.5 in paired pregnancy
samples. One CpG displayed persistence in gestational specific PM2.5 exposures in postnatal
mothers in the CHS. The results from our persistence analysis suggests that there may be a
lingering effect of environmental stressors on maternal health, potentially extending beyond two
years postpartum (Figure 4).
Ambient air pollution mixtures also affected maternal DNAm signatures between early and
late pregnancy. A majority of CpGs in the ambient air pollution decreased with increasing
gestational age in DEFA3, ATF6, MOG, and RERE. Increased ATF6 expression has been
associated with PM2.5 in murine models, in a direction consistent with the PM2.5 weight in this
study [69]. ATF6 is an endoplasmic reticulum stress sensor, and increased gene expression is
necessary in response to cellular stress [69,70]. MOG is associated with the damage of the myelin
sheath in oligodendrocytes and has been implicated in mediating the effect of traffic related air
pollution and neuroinflammation in murine models [71].
Our results add to existing literature on the many epigenomic changes occurring in
pregnancy and provide evidence that environmental exposures such as ambient air pollution
mixtures and specifically PM2.5 have the potential to alter these biological processes. We found
evidence for perturbed methylation differences from prenatal PM2.5 exposure in other populations
as well as in loci critical for genomic imprinting maintenance in pregnancy.
Some strengths of this study include the use of longitudinal maternal prenatal signatures.
Previous studies have focused on the effect of PM2.5 on cord blood or pediatric DNA methylation
signatures, but the addition of maternal signatures provides additional insights into complex
downstream biological processes involved in pregnancy which may confer risk to the offspring.

34
That these loci have replicated across multiple human cohorts provides further confidence in in
the robustness of results and suggests future studies should include maternal DNA methylation
signatures when investigating the effects of environmental prenatal exposures. Such evidence may
speak to biological changes that may affect downstream placental function, increasing risk of cord
blood DNA methylation changes. We also found evidence of persistence of discriminating
methylation signatures among gestational PM2.5. significant CpGs in postnatal maternal samples
in a separate cohort. Given extensive prenatal biological changes, pregnancy may represent a
sensitive period in mothers for PM2.5. to increase risk of enduring methylation signatures that may
act as a mediator for later life disease.
Limitations of this study include the lack of pre-conceptional and postnatal DNA
methylation signatures in the discovery and replication sample, which would provide a more
complete gestational trajectories analysis. For the ambient air pollution mixtures and PM2.5
analysis, there were no CpGs in genes that overlapped with results in the air pollution DNAm
neonatal, pediatric, or adult meta-analyses. This study sample also had a high prevalence (>20%)
of hypertensive and glucose dysregulation in pregnancy, which may be inducing collider bias in
some CpGs due to adjustment for these complications while for other CpGs these complications
may be acting as an effect modifier [33,34]. Another limitation was the relatively small sample
size (N = 127) subjects, which may be limiting our power in the PM2.5. and main effect model.
Non-differential measurement error in air pollution exposure assignment may also be
biasing our results. Exposure assignment is a function of distance from the monitor, and the
assumption that regional ambient air pollutant concentrations are consistent (and unaffected) by
other local factors, other than distance, affecting internal dose is a limitation of this analysis. Future
studies should examine the effect of personal exposure to ambient mixtures. Residual confounding

35
from other regional environmental effects that are correlated with ambient air pollution
concentrations may be affecting these analyses, though this study used a DAG to adjust for the
education, ancestry, and ethnicity [33,34].
Conclusion
In this study, we observed changes in numerous DNA methylation levels across pregnancy.
Prenatal exposure to PM2.5. further altered DNA methylation trajectories across pregnancy and
replicated genes that have been previously implicated in air pollution. We further provide evidence
that certain prenatal PM2.5. driven changes in prenatal methylation may persist in mothers
postnatally for several years, which we hypothesize may act as mediators of later life disease risk.
Table 1

36
Figure 1

37
Volcano Plot of 5,139 (FDR < 0.05) significant loci in linear models for GA, with the top 50 hits
labeled by gene annotation. Blue indicates lower methylation in late versus early pregnancy
[104].

38
Figure 2
Manhattan Plot of 5,335 (FDR < 0.05) significant loci in linear models for GA, with the 10
smallest p-values labeled by CpG. The red line is the Bonferroni adjustment level [105].

39
Figure 3
Volcano Plot of 136 (FDR < 0.05) significant loci in linear models for GA*PM2.5, with the top
50 hits labeled by gene annotation. Blue indicates lower methylation in late versus early
pregnancy [104].

40
Figure 4
Plot of TRIM28 m-value CpG methylation and Gestational Age at Sample Collection among
individuals a standard deviation above the PM2.5 mean [104].
Figure 5

41
DMR plot from DMRcate depicting FBRSL1 methylation on the y-axis and regional marks on the
x-axis. Purple indicates early pregnancy methylation levels among lowly exposed (PM2.5 < 12
micrograms/m3), and green is late pregnancy methylation levels among highly exposed PM2.5
participants [93].
Figure 6
Volcano plot depicting Beta Value (% methylation) on the y-axis and stratified by PM2.5
quartiles. The left-hand plot depicts the MADRES prenatal data across early and late pregnancy
samples, and the right-hand plot depicts this association in our postnatal mothers stratified by
prenatal PM2.5 quartiles [104]
Figure 7

42
GO enriched pathways for FDR significant DMPs identified in the Gestational Age model [104].
Table 2
The top 10 highest effect sizes among FDR significant CpGs in the ambient air pollution
mixtures analysis. Adjusted RR is the mixture coefficient.

43
Supplementary Figure 1. Gestational Age Distribution Across the First and Second Sample
Collection
Distribution of Gestational Age in the analytic sample [104]
Supplementary Figure 2. Consort Diagram of Maternal DNA methylation Data

44
Created with BioRender.com

45
Supplementary Figure 3
Distribution of PM2.5 across the Gestational Period [103].

46
Supplementary Figure 4. Boxplots of Significant Differentially Methylation Positions
Boxplots of six overlapping CpGs identified in the Born to Life and Isle of Wight Study [51].
Supplementary Figure 5
Supplementary Figure 6

47
Directed Acyclic Graph [106]

48
CHAPTER 6: CHANGES IN AMBIENT AIR POLLUTION
MIXTURES AND INTERGENERATIONAL DNA METHYLATION
SIGNATURES
Background
Ambient air pollutant changes to the gestating maternal environment can have an impact
on the offspring’s health outcomes. Molecular states, like DNA methylation (DNAm), may be
affected by environmental exposures in pregnancy. Aberrations to maternal DNA methylation
profiles caused by toxicants during pregnancy may have downstream consequences on placental
nutrient supply, and therefore fetal development or DNA methylation signatures. Small toxicants
can also have a direct effect on fetal development due to their ability to penetrate the placental
barrier.
Objectives
This study estimated the effect of changes in mixtures of ambient air pollution (particulate
matter with aerodynamic diameters <2.5 μm (PM2.5), particulate matter with aerodynamic
diameters < 10 μm (PM10), nitrogen dioxide (NO2), and ozone (O3)) mixtures between early and
late pregnancy on maternal prenatal DNA methylation (N = 128) changes. The impact of changes
in mixtures on cord blood DNA methylation signatures was also conducted (N = 120). Then, the
effect of maternal DNA methylation profiles affected by changes (N = 62 maternal-child pairs) in
ambient air pollutants on cord blood DNA methylation will be investigated.
Methods
Changes in ambient air pollutants between early and late pregnancy was estimated using
inverse distance squared weighted spatial interpolation. For the effect of mixtures on changes in

49
DNA methylation, adjusted epigenome-wide quantile g-computation was applied independently
in maternal and cord blood datasets. To test the impact of aberrations in maternal DNAm
associated with changes in ambient air pollution mixtures on cord blood DNAm signatures,
independent cord blood EWASs were applied to the exposure of maternal DNAm differences
among significant maternal CpGs identified in the mixture model. Significance was determined at
an FDR level.
Results
We identified 22 maternal CpGs differences during pregnancy because of changes in
ambient air pollution mixtures. 11 of the CpGs decreased in methylation across the gestating
period. Change in 22 maternal CpGs associated with change in ambient air pollution mixture did
not independently impact cord blood epigenome wide profiles.
Discussion
Ambient air pollution changes resulted in significant changes in maternal methylation in
22 CpGs between early and late pregnancy, but there was no evidence of change among cord blood
profiles in this small cohort. There was no evidence for changes in upstream maternal DNA
methylation affected by ambient air pollutant mixtures inducing specific cord blood DNAm
signatures in this small sample.
Introduction
Ambient air pollution is a plethora of chemicals that are inhaled in an outdoor environment
and include particulate matter with aerodynamic diameters less than 2.5 μm (PM2.5), less than 10
μm (PM10), nitrogen dioxide (NO2), and ozone (O3). Common sources for ambient air pollutants
in Southern California include traffic congestion and industrial output [26-27]. Ambient air
pollution exposure is associated with several adverse health outcomes, including mortality [28].

50
Reducing ambient air pollutant exposure is critical, especially during the periconceptional period,
as this vulnerability window can affect health outcomes in both the mother and developing fetus.
Physiological and mechanical changes during pregnancy can increase the maternal body’s
vulnerability to changes in ambient air pollutants, potentially affecting molecular states like DNA
methylation [20,46,47]. Epigenetic reprogramming orchestrates fetal organ development and
pollutants may interfere with this process via penetration of the placental barrier disrupting fetal
nutrient supply [74-76]. Studies on ambient air pollutant and DNA methylation focus on maternal
periconceptional exposures and neonatal cord blood or pediatric DNA methylation signatures [38].
Most studies focus on cumulative exposure, rather than examining how differences in ambient air
pollutants can also elicit change in DNA methylation profiles [37-40]. Moreover, fewer studies
examine the effect of periconceptional period on two generations of DNA methylation.
Considering upstream changes in maternal biology affected by the environment may also impact
cord blood DNA methylation profiles.
In this study, the effect of mixtures of ambient air pollutants on DNA methylation
signatures in maternal and cord blood profiles were analyzed. Variation in maternal DNA
methylation changes in pregnancy, affected by changes in mixtures of ambient air pollutants, were
also examined for their potential to affect infant cord blood DNA methylation.
Methods
Discovery Study Sample and Recruitment
MADRES is a prospective pregnancy cohort and began recruitment through four prenatal
providers in Los Angeles, California in 2015 contingent on participants being less than 30 weeks
gestational age at cohort entry and at least 18 years old. Informed consent was provided by

51
participants prior to cohort entry by the Institutional Review Board (IRB) at the University of
Southern California (USC) (IRB #HS-15-00498). Details on all variables collected and additional
exclusion criteria is available here [23].
MADRES Maternal DNA methylation
EDTA tubes carrying 10 mL of peripheral blood samples collected during early and late
pregnancy were transported upright on ice to the lab at Norris Cancer Center. At the lab, samples
were centrifuged and overlying plasma was removed within one hour of sample collection. PBMCs
were isolated from peripheral blood samples. The AllPrep DNA kit (Qiagen) was used to extract
DNA, and the EZ DNAm Kit (Zymo Research) was used for bisulfite conversion and was
performed in batches with families plated together. The Illumina Infinium
HumanMethylationEPIC (850K) assay using the manufacturer’s recommended protocol with no
other modifications was used to quantify DNAm.
MADRES Cord Blood DNA methylation
Umbilical cord blood was collected at birth and frozen within 24 hours of delivery. Hospital
staff collected 10 mL of cord blood and placed the samples in a cooler on ice, where it was then
transported from the hospital to the lab. Centrifugation was used to separate plasma, PBMCs, and
red blood cells and then samples were frozen. The EZ DNAm Kit (Zymo Research) was used for
bisulfite conversion, and the Illumina Infinium HumanMethylation EPIC was used to quantify
DNAm under the same protocols as the maternal arrays.
MADRES Quality Control
RStudio 4.1.0 was used to perform all data analysis [103]. The minfi Bioconductor package
performed all probe and sample level quality control (QC) [54]. Probe removal consisted of poor

52
detection p-value probes, polymorphic, and cross-reactive probes were also removed [54-57]. Any
evidence of poor detection p-value across more than 10% of probes, sex discrepancy between
reported sex at birth and sex generated from the array, and a median probe intensity less than 10
were removed from the analysis [54]. Background and dye-bias correction was performed [55]
followed by quantile normalization [56]. MDS plots were also used for outlier detection [54].
Normalized data was log-transformed to M-Values for quantile g-computation analysis [31]. A
consort diagram illustrating QC across maternal and infant probes is shown in Figure S2. 8% and
10% of maternal samples and probes were dropped due to quality control, and 10% and 23% of
infant probes and samples.
Cell Type Estimation
Cord blood and blood composite cell type was used to estimate cell counts in infant and
maternal data in MADRES using the Houseman method [14]. This was executed using EPIC
reference platform with the “FlowSorted.Blood.EPIC” data in minfi [54].
Genetic Factors
SNP-variation impacted by ancestry biases DNA methylation results in an unknown
direction. The top two EPISTRCUTURE principal components were used from DNA methylation
datasets to adjust for this confounding variation in maternal and infant datasets [60].
Measurement of Gestational Age at Sample Collection
Gestational age at sample collection was analyzed continuously. Gestational age at sample
collection was derived from gestational age at birth and gestational age differences between sample
collection and date of birth. Gestational age at birth was abstracted from a combination of medical
records and self-report.

53
Measurement of Ambient Air Pollutants
Residential addresses were collected prospectively during prenatal visits. Inverse-distancesquared weighted spatial interpolation from ambient air quality monitoring data (with an average
of four monitoring stations within 8 to 14 km of each residential address) was used to generate
estimates of 24-hour particulate matter with aerodynamic diameter less than 2.5 μm (PM2.5 ), 24-
hour particulate matter with aerodynamic diameter less than 10 μm (PM10 ), 24-hour nitrogen
dioxide (NO2 ), and 8-hour daily maximums in ozone (O3 ). Weekly gestational air pollutants were
averaged from one year pre-conception to early pregnancy sample collection and from early
pregnancy collection to late pregnancy sample collection for exposure assignment.
Covariate Selection
Confounders were identified using a directed acyclic graph from prior literature.
Confounder adjustment including pre-pregnancy body mass index (BMI), total weight gain across
pregnancy, binary diabetes status (gestational diabetes, chronic diabetes, glucose intolerance),
binary hypertensive status (preeclampsia, eclampsia, preeclampsia-eclampsia, chronic
hypertension), ethnicity and origin (White non-Hispanic non-foreign born, Black non-Hispanic
non-foreign born, Hispanic foreign born, Hispanic non-foreign born, and multiracial non foreignborn participants), maternal age at birth, parity status, and maternal education (less than high
school, high school or some college, college degree). Season of first sample collection
(Fall,Winter,Spring,Summer), average periconceptional temperature changes, and average
periconceptional ambient air pollution exposure to isolate effects of change. Covariates to control
for technical artifact included cell type change across the paired timepoints, batch, and the top two
EPISTRUCTURE [60] principal components. Participants who reported smoking (N = 3) were

54
removed from the analysis given the small sample size. Participants missing more than a week of
periconceptional exposure were removed from the analysis (N = 2).
Statistical Analysis
Quantile g-computation was used to estimate differences in ambient air pollutants mixtures
driving DNA methylation changes in pregnancy and cord blood DNA methylation signatures at
birth [31]. Log-transformed M-values were used as the outcome and the exposure mixture included
differences in PM2.5, PM10, NO2, and O3 between early and late pregnancy. Confounding variables
were also included. All probes surviving quality control were tested (Figure S2). FDR adjustments
were used for p-value significance. Overlap in significant findings between the maternal and infant
dataset were reported. Maternal CpGs from the mixtures analysis were also used as exposures in
independent cord blood adjusted EWASs to test for the effect of maternal change in methylation
driven by changes in ambient air pollution mixtures on epigenome wide cord blood signatures.
Changes in maternal DNAm were calculated as the beta value difference between the early and
late pregnancy samples in normalized datasets.
Pathway Analysis
DMPs were tested for enrichment of Kyoto Encyclopedia of Genes and Genomes (KEGG)
terms using gometh [62,63]. This method accounted for the number of probes per gene. Enriched
terms with an FDR less than 0.05 were reported.
Results
Participant Characteristics

55
Table 1 examines maternal and infant characteristics by quartiles of differences in PM2.5
exposure between early and late pregnancy. There was evidence of proportional differences in
quartile of PM2.5 exposure and maternal ethnicity and place of birth and age.
Ambient Air Pollution Mixtures and Maternal Prenatal DNA methylation Differences
22 CpGs were identified to significantly decrease in methylation across pregnancy in
response to changes in ambient air pollution mixtures (Table 2). The CpG with the smallest pvalue was cg17613707 in the open sea region of the ANKRD55 gene with a 0.65 m-value increase
in early versus late pregnancy per quartile increase in ambient air pollution mixtures (Ψ = 0.65,
95% CI [0.42, 0.88], p = 2.6 x 10-9). The largest magnitude of effect was in cg04395153 in the
open sea region of MYOG with a 1.47 m-value decrease in early versus late pregnancy per quartile
increase in ambient air pollution mixtures (Ψ = -1.51 95% CI [-2.03,-0.99], p -value = 3.4 x 10-8)
(Table 2). A partial plot was generated to examine the impact of weighted component mixture
effects of ambient air pollution mixtures for MYOG and indicated synergistic negative effects
across all four pollutant effects, with the strongest weight in ozone (Figure 1).
Ambient Air Pollution Mixtures and Neonatal DNA methylation Signatures
There were no neonatal CpGs that were significant at the FDR < 0.05 level.
Effect of Ambient Air Pollution Affected Maternal CpGs on Cord Blood Epigenome-Wide
Signatures
There were no significant (FDR < 0.05) neonatal cord blood CpGs associated with change
in maternal ambient air pollution affected CpGs across the 22 EWASs implemented.
Pathway Analysis

56
Pathway analysis was conducted among the 22 maternal CpGs that were significant in the
mixture model. There were no enriched KEGG or GO terms at an FDR < 0.05 level using gometh.
Discussion
Change in periconceptional ambient air pollution mixtures were associated with change in
22 maternal CpGs between early and late pregnancy, with fifteen of the CpGs decreasing over
time. There was no evidence for differences in CpGs in response to periconceptional ambient air
pollution changes in pregnancy in our small cohort of neonates. Moreover, there was no evidence
that variation in 22 maternal CpGs affected by ambient air pollution mixtures is associated with
cord blood DNAm signatures.
This study found evidence of dynamic epigenetic changes between early and late
pregnancy affected by air pollution mixture changes in 22 CpGs. MYOG decreased in methylation
across gestation per quartile increase in pregnancy, with the largest observed effect size among
significant CpGs. MYOG has been implicated in embryonic myofiber development and adult
muscle repair, and knockouts of the gene in murine models was associated with reduced transcript
levels across several key genes in adult muscle tissue [96]. Decreases in methylation of the south
shore region of this gene may be associated with dysregulation of muscle repair in response to
ambient air pollution mixtures. ANKRD55 had the smallest observed p-value, and methylation has
been associated with impaired immune function [97]. SNPs in ANKRD55 have been associated in
GWASs across 18 different autoimmune or inflammatory conditions [98].
CDH13 has been associated with gene-environment interaction between air pollution,
methylation, and hypertension in an adult Korean male population [77,78]. Short-term exposure
to PM10 in adult male and females is associated with decreased ETHE1 expression, and in our

57
cohort we found increases to ambient air pollution mixtures increases methylation in the north
shore region of this gene [79]. Increased ETHE1 methylation has also been associated with
sustained maternal smoking in meta-analyses, in a direction consistent with the effect in this cohort
[80]. For example, TSPAN9 has been associated with increased methylation in offspring in mothers
exposed to particles in pregnancy in murine models, a direction inconsistent with the decreased
maternal effect in this cohort [95].
Some strengths of this study include the use of paired, prenatal DNAm signatures that
allowed for the examination of dynamic DNAm changes. Many studies examining the effect of
pollution are restricted to cumulative pollutant effects and cross-sectional DNAm data, and
therefore the temporal nature of epigenetic dynamics become unknown. Another strength includes
the investigation of DNAm changes in the prenatal period in response to an environmental
toxicant, which is usually limited to neonates. Though there were no significant associations in our
neonatal group, this may be an effect of small sample size. This study adjusted for an average
effect of ambient air pollution mixtures, and therefore the epigenetics impacts of change
independent of average pollutant effects may be minimal. However, we are limited in our
interpretation due to low sample size. We also did not consider the potential effect modification
by maternal hypertensive or diabetes status, which may affect the impact of environmental factors
to the mom and fetus.
The potential persistence of the maternal CpGs affected by ambient air pollution mixture
changes beyond the prenatal period is unknown. Therefore, we are limited as well to assess the
effect beyond the prenatal period in this sample. Another strength of this study includes the use of
multigenerational DNAm signatures, which allows us to assess if the effect of change in ambient
air pollutant mixtures is associated with similar or different CpGs across generations. Though we

58
found evidence of change on the maternal level, we acknowledge that there are several limitations
to our cord blood analysis that may be restricting results. This includes the lack of longitudinal
signatures, small sample size with no duplicates or replicates to enhance power, and the inclusion
of 22 confounding variables in the model diminishing power to detect significant effects. It is also
important to keep in mind that this study examines the effect of change, adjusted for cumulative
exposure, which contrasts with the literature that focuses on cumulative prenatal pollutant effects
in cord blood methylation. Considering all the limitations of this dataset, the effect in these two
CpGs is not necessarily driven by bias or limitations in power. The fetal epigenetic environment
may be impacted by small changes in air pollutants in this cohort.
Another important limitation to interpretation in this study is the high prevalence of
hypertensive and glucose disorders in the study sample. Given the impact of hypertension and
glucose dysregulation on maternal biology in pregnancy, the effects of these pregnancy
complications and chronic conditions may act as a mediator for some CpGs, while acting as an
effect modifier for others. Genetic data was also unavailable in this study and therefore
interpretation of effects should be extremely limited without follow-up on the potential genetic
effects that may be driving these associations. Though we adjusted for EPISTRUCTURE principal
components, the method is applied to distinguish distinct populations, and may not be applicable
for this potentially admixed dataset.
Conclusion
In this study, we observed some evidence of change in DNA methylation levels across
pregnancy driven by ambient air pollution mixtures. There was no evidence of a mixture effect in
cord blood samples in this small cohort. There was no evidence of changes in ambient air pollution
driving maternal epigenetic changes that induced specific offspring de novo epigenetic signatures.

59
This study was extremely limited in its small sample size across mothers (N = 128) and cord blood
DNAm signatures (N = 120) to estimate effects, which may be contributing to Type II error.
However, this small sample highlights the potential detrimental effects of ambient air pollutants
on gestational biological profiles in a socioeconomically vulnerable subgroup.
Table 1

60

61
Table 2
Maternal and Neonatal CpG methylation among CpGs that were significant in maternal ambient
air pollution mixture model.
Figure 1

62
Partial plot of effects of component ambient air pollutants on MYOG methylation changes
between early and late pregnancy. Component effects were synergistic, with the largest weight in
the ozone component [31].
Figure 2
Figure 3

63
Supplementary Figure 1. Gestational Age Distribution Across the First and Second Sample
Collection
Distribution of Gestational Age [104].
Supplementary Figure 2. Consort Diagram of Maternal DNA methylation Data

64
Created with BioRender.com

65
CHAPTER 7: SMOKING, AIR POLLUTION, AND
MULTIGENERATIONAL DNA METHYLATION SIGNATURES
Introduction
Smoking reduces lung function in smokers and families of smokers [81]. There is some
evidence that grandmaternal smoking in pregnancy increases grandchild asthma risk [82,83].
Epigenetic states, including DNA methylation, can act as potential mediators of this increased risk
[84]. There is some evidence this may occur through direct intergenerational signatures [85] –
referring to the exposure of the developing fetus and potentially its associated germ cells [84-85].
Pluripotent germs cells, or PGCs, in the developing fetus migrate to the fetal genital ridge by the
end of gestational week six before experiencing epigenome-wide demethylation [85,86].
Therefore, persistent exposures across gestation may be associated with increased risk of
multigenerational health outcomes via independent changes in DNA methylation in the fetus and
PGCs [85,86]. Transgenerational epigenetic inheritance is the persistence of this signature to the
unexposed generation [21]. Murine research has demonstrated some evidence of intergenerational
and transgenerational epigenetic inheritance among mice that are exposed to gestational nicotine
[87,88]. Results suggested that perinatal nicotine exposure is associated with reduced lung function
in F1 and F2 dams, global methylation patterns increased in the testis and decreased in the ovaries
of F1 dams, increased protein expression of fibronectin in the lungs of F1 and F2 dams, and
decreased peroxisome proliferator activated receptor gamma (PPAR) expression in F1 and F2
dams [87,88].
Similar models have not been investigated in humans, mainly due to lack of available data.
There are also challenges to studying the effects of nicotine in humans. Given the sensitivity of
epigenetic patterns affected by nicotine in murine models to counteracting compounds (e.g.,

66
rosiglitazone) [89] and the many carcinogenic compounds that exist in cigarette smoke, it’s
representations of singular compound effects in animal models may not be a realistic depiction of
carcinogens human are exposed to in cigarette smoke [90]. Humans also have more complex
exposure histories than model organisms and grandmaternal (G0) prenatal tobacco smoking (PTS)
-affected, shared intergenerational DNAm signatures may be driven by cohabitation,
socioeconomic factors, or other confounding factors that covary with smoking behaviors [91].
Cigarette smoke and air pollution across the parental (G1) generation may have synergistic,
antagonistic, or discrete impacts on G0-PTS affected loci in the grandchild (G2) generation [91].
This study proposes to investigate the effect of G0 PTS on sex-stratified epigenome wide
signatures across the G1 and G2 generation in the Children’s Health Study (CHS) [22] using EPIC
array based buccal cell data. Shared, statistically significant (FDR < 0.05) signatures across
generations will be used in a linear mixed model investigating epigenetic trajectories among
families affected by G0 PTS. Then, the impact of the mixture [31] of G1 childhood ambient air
pollution exposure on PTS-affected G2 signatures will be evaluated.
Methods
Children’s Health Study Sample and Recruitment
The CHS is a prospective cohort study that recruited schoolchildren from the Southern
California region from the 1990s to 2000s [22]. The study’s goal was to determine the effects of
air pollution and other environmental exposures on respiratory health. Buccal cell samples were
collected from a subset of 500 recruited participants (G1) and their families (G0). Approximately
two decades after baseline, a convenience sample of adult-aged index participants (G1) were
invited to participate in a follow-up mail-based study. Buccal samples were collected from the
adult participant (G1) their partner, and their child (G2)(Supplementary Figure 1). Recruitment was

67
restricted to adult participants with a child under the age of 7. Informed consent was provided (IRB
#HS-17-00778).
Children’s Health Study DNA methylation Sample Collection
Pediatric buccal cells were collected with swab using the Oragene OC-175 kit or toothbrush
collection methods for each family trio in approximately 5mL of buffer. DNA was extracted with
the Oragene prepIT L2P kit. The extracted DNA was stored locally at -80°
C. The Zymo EZ DNAm
kit was used to perform bisulfite conversion, and the Illumina EPIC DNAm protocol was used to
generate the data.
Children’s Health Study DNAm Quality Control
All data analysis was performed in R (version v4.1.0, R Core Team 2021) [103]. Quality
control and normalization of data were performed separately at each generation. Sample and probe
level quality control were performed using standard protocols outlined by the minfi [54]
Bioconductor package. Briefly, poor detection p-values were computed across probes,
representing those probes with no significant difference in detection between background and
control probes, and were removed from the analysis. If a sample had more than 10% of poor
(p>0.01) detection p-value probes, it was removed from the analysis. Cross-reactive and
polymorphic probes were also removed [54-56]. Sex predicted from intensity of X and Y
chromosomes was used as a quality control check [54]. Noob background correction for dye-bias
[57] followed by quantile normalization [58] was used for normalization. SNP-associated probes
were removed from the analysis. This analysis was restricted to autosomes [54]. Log-transformed
beta-values were used in downstream regression analysis. Overall, 89% of probes and 92% of
samples were retained in the parent generation, and 89% of probes and 97% of samples in the
offspring generation (Supplementary Figure 2).

68
Cell Type Estimation
For buccal cell data in the CHS, immune cell type proportions were estimated using
HEpiDISH [59].
Genetic Factors
Ancestry based SNP-variation impacts DNA methylation and can bias results. We used
EPISTRUCTURE principal components produced from the quality-controlled DNA methylation
datasets to help address this issue in the G2 EWAS. The top two EPISTRUCTURE principal
components were used as adjustment variables in covariate-based sensitivity analyses [60].
Measurement of Air Pollution
Residential addresses were collected prospectively during CHS follow-up. Inversedistance-squared weighted spatial interpolation from ambient air quality monitoring data (with an
average of four ambient air pollutant monitoring stations within 0.1 to 5 km) was used to generate
estimates of 24-hour particulate matter with aerodynamic diameter less than 2.5 μm (PM2.5 ), 24-
hour particulate matter with aerodynamic diameter less than 10 μm (PM10), 24-hour nitrogen
dioxide (NO2), and 8-hour maximum ozone (O3). Yearly gestational PM2.5, PM10, NO2, and O3
from one year pre-conception to adulthood (age 18) sample collection was averaged for exposure
assignment.
Measurement of Smoking
Smoking exposure was collected via questionnaire. Questionnaire data was collected at
different timepoints in the grandparental and parental generation (Figure S1). At the grandparental
level, two questions were used: “Did the biological mother smoke during pregnancy?” and “Is
there any daily smoking in your home?” At the parental generation, several questions were used
to determine smoking status: “Did you or the mother smoke one month prior to or anytime during

69
your pregnancy with this child?”, “Did/does anyone in your household smoke cigarettes, cigars or
pipes inside on a daily basis while your child was at home since he or she has been born?” and
“Have you ever smoked cigarettes?”.
Measurement of Wheeze
Wheeze was measured at childhood in both the parental and grandchild generation. To
determine the outcome, questionnaires were distributed to the respective parents of each
generation. Questionnaires distributed to the grandparental generation to estimate parental wheeze
risk was “Has your child ever wheezed for 3 or more days out of the week for a month or longer?”
and questionnaires to the parental generation included “Has your child ever had wheezing or
whistling in the chest at any time in the past?”
Epigenome-Wide Association Studies
EWASs were implemented at the parental and grandchild level independently and stratified
by sex in limma [96], given the impact of sex on epigenetic heritability in animal models [87,88].
Six EWASs were then implemented at the G1 male (N = 72), G1 female (N = 148), G2 female from
paternal lineage (N = 43), G2 female from maternal lineage (N = 70), G2 male from paternal lineage
(N = 32), and G2 male from maternal lineage (N = 60). Surrogate Variable Analysis (SVA) [61]
was used for covariate adjustment of measured and unmeasured confounders. Briefly, SVA
removes variation attributable to interesting sources (grandmaternal smoking) and uses this
residual variation to generate surrogate variables, or variables associated with measured or
unmeasured confounders. All surrogate variables produced by SVA were used for adjustments in
limma [96]. Significance was based on FDR estimation in limma [96]. SVA was also compared to
a covariate model including age, collection method, batch, G0 wheeze status, G1 ever smoking, G1
PTS smoking, G1 secondhand smoke exposure (in adulthood or childhood), EPISTRUCTURE

70
principal components, G1 wheeze (in the G2 EWAS), G0 education and immune cell content [59].
Due to missing qfactors and genotype data in the G1, covariate models only included N = 40 G1
males and N = 35 G1 females. There was no adjustment for cohabitation in adulthood between the
G0 and G1 generation, or great-grandmaternal smoking because these factors were unmeasured in
our study population. The effect of G0 PTS was allowed to vary in linear models across wheeze
status at G1 or G2 using an interaction term. Effect modification was explored because PTS-affected
loci may increase risk for wheeze and/or PTS-affected loci may be confounded by wheeze status
due to reverse causation. The DMP with the smallest p-value and largest effect size in the main
effect and effect modification were reported at each generation from the SVA model. Overlap in
both the main effect and interaction model were reported. Persistence of the direction of
coefficients for these signatures in the covariate model was also reported.
Pathway Analysis
DMPs were tested for enrichment of Kyoto Encyclopedia of Genes and Genomes (KEGG)
terms using gometh [62,63]. This method accounted for the number of probes per gene. Enriched
terms with an FDR less than 0.05 were reported.
Linear Mixed Model
Linear mixed models with a random effect for family were implemented among significant
loci identified in the EWASs across overlapping across generations. Adjusted linear models
included covariates for individual collection method, G0 asthma, G1 wheeze, G1 ever smoking, G1
persistent smoking, G1 PTS smoking, G1 secondhand smoke exposure (in adulthood or childhood),
batch, G0 education, and immune cell content. Due to sample size restrictions (N = 211 participants
in linear mixed model), effect modification by wheeze status was only explored in the G2
generation. G0 female methylation was the reference group for estimation of generational-specific

71
methylation among families with a grandmother who smoked in pregnancy versus those who did
not. Confidence intervals and p-values were generated using Satterwaithe’s approximation for pvalues in the lmerTest package in RStudio [94]. P-Values less than 0.05 were considered
significant.
Quantile G-Computation
To estimate potential synergistic or antagonistic effects of G1 ambient air pollution
exposure on G2 PTS-affected methylation adjusted for G1 smoking exposure, we used quantile gcomputation [31] to assess the potential mixture effects of average parental childhood
O3,NO2,PM25, and PM10 exposure adjusted for smoking across the life course in the G1 (consisting
of either G0 use in utero, G1 secondhand smoke in childhood or adulthood, or personal use). Models
were also adjusted for G0 education and G2 age.
Results
G1 (Parental) PTS Epigenome-Wide Association Studies in Participants
278 DMPs were significantly (FDR < 0.05) different in PTS-exposed versus unexposed G1
males. 67% of significant DMPs were reduced in exposed compared to unexposed to participants.
G1 males exposed to PTS had 4.8 higher log m-value methylation in LINC01484, the largest effect
size, compared to unexposed participants. Among G1 males exposed to PTS with wheeze, 243
DMPs were significant compared to unexposed participants. 233 CpGs persistent among G1 males
with and without wheeze exposed to PTS, and no CpGs displayed a consistent direction of effect
(Figure 2).
1 DMP was significantly (FDR < 0.05) different in PTS-exposed versus unexposed G1
females with 0.24 logFC lower methylation in cg16495320 mapped to DYRK1A. Among G1

72
females exposed to PTS with wheeze, no DMPs were significant compared to unexposed
participants. Across the G1 group, there was no overlap in DMPs.
G2 (Grandchild) PTS Epigenome-Wide Association Studies in Participants in Maternal Lineage
Among G2 females exposed to grandmaternal PTS, 0 DMPs were significant compared to
unexposed participants with no wheeze. 81 CpGs were specific to associations in the PTS-affected
wheeze group. Methylation was increased in 65% of the PTS-affected wheeze group compared to
the unexposed group.
1 DMP was significantly (FDR < 0.05) different in PTS-exposed versus unexposed G2
males with 1.1 logFC lower methylation in cg02950812. This effect was persistent among G2 males
with and without wheeze exposed to PTS included, with 2.5 logFC higher methylation in those
exposed with wheeze and lower methylation in those exposed with wheeze compared to unexposed
participants. There was no shared overlap across G2 females and G2 males in the maternal lineage.
G2 (Grandchild) PTS Epigenome-Wide Association Studies in Participants in Paternal Lineage
7 DMPs was significantly different among G2 females exposed to grandmaternal PTS
compared to unexposed participants without wheeze. All significant DMPs were reduced in
exposed compared to unexposed to participants. G2 females exposed to PTS had 2.0 logFC lower
methylation in cg07441518, the largest effect size, compared to unexposed participants. Among
G2 females exposed to PTS with wheeze, 198 DMPs were significant compared to unexposed
participants. CpGs persistent among G2 females with and without wheeze exposed to PTS included
EIF2C2, with higher methylation in in those exposed with wheeze and lower methylation in those
exposed without wheeze compared to unexposed participants.

73
Among G2 males, there were also 7 DMPs associated with exposure to PTS compared to
unexposed participants. 3 out of the 7 DMPs had reduced methylation levels in exposed compared
to unexposed. The largest effect size was in cg20799941 in LMNB1 with a 1.8 logFC lower
methylation in exposed versus unexposed participants. Among G2 males exposed to PTS with
wheeze, 212 DMPs were significant compared to unexposed participants. 2 CpGs persisted among
G2 males with and without wheeze exposed to PTS in LMF1 and KCNJ5. In KCNJ5 and LMF1,
methylation was increased in those without wheeze and decreased in those with wheeze relative
to unexposed participants. There was no shared overlap across G2 females and G2 males in the
paternal lineage (Table 2).
Overlapping Epigenome-Wide Association Results across Maternal Lineage
There were no DMP CpGs overlapping across generations in the PTS group affected or
unaffected by wheeze.
Overlapping Epigenome-Wide Association Results across Paternal Lineage
There was 2 DMRs overlapping across generations and there were no DMP CpGs
overlapping across generations. There was some overlap in annotation to the UCSC Reference
Genome (GrCh38/hg19) in the ADARB2 gene between G1 males (cg10174867) and G2 females
(cg27230989) unaffected by wheeze (Figure 2). There was no overlap between G1 males and G2
males unaffected by wheeze. Among G1 males and G2 females with wheeze, affected by PTS, there
was overlap in 16 different genes, including in FBRSL1.
Linear Mixed Model
Linear mixed models were implemented for ADARB2 and FBRSL1 in paternal lineage for
the G1 and G2 generation, using G0 methylation as a reference. FBRSL1 (cg25977879) methylation

74
was lower in G1 males (B = - 0.05 95% [-0.1,-0.005], p = 0.04) and G2 females (B = - 0.07 95% [-
0.1,-0.02], p = 0.008) in families with a grandmother who smoked during pregnancy with a
wheezing grandchild compared to those not exposed to grandmaternal smoke among the group of
grandchildren without wheeze (Table 3). No differences were observed among the group of
grandchildren with wheeze. ADARB2 was not persistently differentially methylated in G1 males or
G2 females compared to G0 methylation in families in which the grandmother smoked when
evaluated in stratified models by grandchild wheeze. Grandmaternal PTS reduced ADARB2
methylation in G1 males (B = -0.17 95% CI [-0.97,0.62]) and increased in G2 females (B = 2.04
95% CI [1.0,3.1]) compared to unexposed G1 males and G2 females (Table 4) among the group
without wheeze. This effect was not significant (p>0.05) in G1 males. In families with G2 wheeze,
methylation was reduced in G1 males (B = -0.38 95% CI [-1.7,0.93] and G2 females (B = -0.26
95% CI [-1.4,0.94] but this effect was not significant (Table 3).
Quantile G-Computation
To assess if there were potential synergistic or antagonistic effects of ambient air pollution
across the parental life course on PTS-affected G2 CpGs, quantile g-computation was
implemented. There was no association (p>0.05) of the mixture effect on G2 CpG methylation in
cg27230989 or cg27230989.
Discussion
This study investigated the multigenerational epigenetic effects of grandmaternal (G0)
smoking in the parental (G1) and grandchild (G2) generation. PTS impacted methylation loci across
all EWAS analyses which were stratified by sex and generation. 7 DMPs were significantly

75
associated with grandmaternal PTS in the G2 paternal lineage with no wheeze. There were 278
significant DMPs in the G1 males but only 1 significant DMP in the G1 females.
In the paternal lineage, there was differences in methylation among G2 males and G2
females who were either affected or unaffected by wheeze. There were more significant DMPs in
the G2 females with wheeze compared to those without wheeze, but a similar number of DMPs in
G2 males with wheeze compared to those without wheeze. Disease states like wheeze may impact
DNAm, and therefore the results in our interaction model are impacted by reverse causation.
LMF1 methylation had the largest effect size in G2 males and has been associated with prenatal
smoking in meta-analysis, and in a positive direction consistent with the higher effect observed in
this cohort [38]. EIF2C2 methylation had the largest effect size in G2 females, in a negative
direction consistent with meta-analysis [38]. In the maternal lineage, there were no significant
DMPs among G2 females without wheeze, though there were DMPs among the G2 females with
wheeze. There were 1 significant DMP in the G2 males, that was persistent among the wheeze and
without wheeze group.
Results from the literature indicate global methylation increases in the testis of F1
nicotine exposed males and increased expression in fibronectin proteins [87]. In this study, there
were overall decreases in paternal parent and paternal granddaughter in FBRSL1 methylation in
families exposed to grandmaternal PTS among those without G2 wheeze.
FBRSL1, or fibrosin-like 1, methylation has been associated with personal signatures of
cigarette smoking [99] .Secreted fibrosin may induce fibroblast proliferation, which plays a
crucial role in maintaining the extracellular matrix which is responsible for wound healing [100].
Given that the results from the murine model in transgenerational effects of nicotine exposure
increases fibronectin expression [97], this may be consistent with the decreases in methylation

76
observed in a fibrosin-like gene in this study. There is some human evidence that de novo
truncating mutations in FBRSL1 is associated with reduced respiratory function, along with
developmental delay [101]. Interestingly, FBRSL1 is also a maternally expressing imprinting
gene [102].
One limitation of this analysis is the small sample size across exposure groups. Among G1
males, 8 were exposed to PTS and 4 had asthma. Among G1 females, there were 14 exposed to
PTS and 4 had asthma. The results from the G1 male analysis should be taken with extreme
limitation given the small sample sizes across the PTS and PTS and wheeze group. Among
participants included in this analysis that responded to the mail-based questionnaire, 70% were
female. Selection biases for the male population may be impacting results away from the null.
Recall bias, for example smoking history, may be driving results away from the null. Co-morbidity
of other conditions related to wheeze may also be impacting results away from the null, given the
impact of in utero smoking on other adult health conditions.
There were overlapping results in the G1 male and offspring G2 females in ADARB2. There
was some evidence of an intergenerational effect among families with a history of grandmaternal
PTS and reduced methylation in ADARB2 in the female grandchild unaffected by wheeze, but this
effect did not persist in the paternal group. Potential direct intergenerational effects can also be
possible through the germline of the male parent [11,12]. Prospermatagonia are present in a finite
amount in the developing fetus’s germline and act as precursors to sperm. Evidence for epigenetic
stability has been shown in vitro between PGCs and prospermatagonia [11]. F1 male dams directly
exposed to environmental toxicants during the gestational period also have methylation profiles
that are stable in cauda epididymal sperm [12]. Results from the model organism transgenerational
epigenetic literature indicate global methylation increases in the testis of F1 exposed males

77
exposed to nicotine [87,88]. In this study, G1 males had reduced methylation in over 70% of genes,
indicating potential increased expression in promoter regions – a direction inconsistent with this
effect [85].
Results from our mixture model indicated that there was no evidence for the joint impact
of average childhood ambient air pollution affecting variation in PTS-affected G2 ADARB2 or
FBRSL1 in this cohort, adjusted for G1 life course exposure. Effects of PTS in utero at the G0 level
may be impacting methylation independently of exposure to other smoking effects.
This study had several limitations that may be biasing results or restricting generalizability.
Issues of power due to low sample size may be inducing Type I or II error in this sample. Moreover,
collection method varied across generations and therefore may also be biasing our results in an
unknown direction, given that the entire parental generation used the toothbrush and the grandchild
used swab and both methods are prone to different types of human error. There was some
availability of duplicate samples within a generation that used both collection methods. There were
no apparent differences in FBRSL1 or ADARB2 methylation across collection method in each
generation among those with available samples. Visualizing the standard errors and logFC values
from the paternal parent in both the main effect and effect modification, there did not appear to be
a specific trend [Figure S3]. The biggest limitation to this study is the inability to delineate genetic
versus epigenetic effects (due to lack of genotyping data across generations) and this would bias
our results away from the null. Future studies or versions of this paper must incorporate the impact
of local or distal effects of the genome on FBRSL1 or ADARB2 methylation.
Conclusion

78
This study is one of the first to demonstrate an epigenetic effect at the parental and
grandchild generation affected by grandmaternal smoking during pregnancy and resulting in
wheeze. Evidence at the parental and grandchild level was consistent in some CpGs with results
from meta-analyses of PTS in newborns. Genotype, cohabitation, and reverse causation may be
biasing our results away from the null. Results from our mixture model indicate there was no joint
or compounding effect in PTS-affected CpGs and smoking or ambient air pollution exposure
across the lifespan.
Table 1

79

80
Figure 1
Diagram of sample characteristics. Created with BioRender.com

81
Figure 2
Volcano Plot of the Paternal Parent [107]. Right plot Created with BioRender.com
Table 2

82

83
Supplementary Figure 1
Created with BioRender.com
Supplementary Figure 2

84
Created with BioRender.com

85

86
Scatterplots for the logFC (x-axis) and standard error (y-axis) for both the main effect (top) and
effect modification (below) [103].

87
CHAPTER 8: CONCLUSIONS
Summary
Two small studies from Southern California were used to evaluate the effects of air
pollution on DNA methylation signatures across families in the region. We found some evidence
that developmental periods can act as periods of risk for maternal, offspring, and grandchild.
Maternal epigenetic effects are understudied, but there was evidence for differential change among
mothers exposed to both higher cumulative averages of air pollution (study one) and changes in
ambient air pollution concentrations (study two) in MADRES. There was no evidence the latter
induced specific cord blood DNAm signatures, though there was minimal change in ambient air
pollution concentration between early and late pregnancy in the MADRES cohort. There was some
evidence that paternal grandmaternal smoking displayed a direct intergenerational effect of
smoking on FBRSL1 and ADARB2 methylation in the paternal parent and female grandchild
generation. Parental particulate matter did not have any effects on FBRSL1 or ADARB2
methylation in grandchildren.
Conclusions and Implications
This thesis highlights the importance of studying the impacts of environmental exposures
in gestating mothers. It should be a goal of health researchers to improve the landscape surrounding
mothers to decrease mortality and morbidity risk from pregnancy. Moreover, though the
prevalence of maternal smoking during pregnancy has decreased over time, this thesis provided
evidence that there may be a biological legacy of the use of these extremely toxic substances. If
exposure to tobacco smoke can induce such multigenerational effects, then it is possible other
toxicants may too.

88
Future Directions
Future directions should include the exploration of joint cumulative and change in ambient
air pollution or particulate matter impacting intergenerational DNA methylation profiles.
Examining personal monitoring data is also an important step to isolate extremely specific
elements driving DNA methylation change. Disentangling the effect of a single element on the
change in DNA methylation can improve our understanding of the total mixture effect.

89
BIBLIOGRAPHY
1. Barker DJ. The origins of the developmental origins theory. J Intern
Med 2007;261(5):412–17.
2. Gluckman PD, Hanson MA, Cooper C et al.. Effect of in utero and early-life conditions on
adult health and disease. N Engl J Med 2008;359(1):61–73.
3. Moore LD, Le T, Fan G. DNA methylation and its basic function.
Neuropsychopharmacology. 2013 Jan;38(1):23-38. doi: 10.1038/npp.2012.112. Epub
2012 Jul 11. PMID: 22781841; PMCID: PMC3521964.
4. Hamilton JP. Epigenetics: principles and practice. Dig Dis. 2011;29(2):130-5. doi:
10.1159/000323874. Epub 2011 Jul 5. PMID: 21734376; PMCID: PMC3134032.
5. Jin B, Robertson KD. DNA methyltransferases, DNA damage repair, and cancer. Adv Exp
Med Biol. 2013;754:3-29. doi: 10.1007/978-1-4419-9967-2_1. PMID: 22956494; PMCID:
PMC3707278.
6. Almouzni G, Cedar H. Maintenance of Epigenetic Information. Cold Spring Harb Perspect
Biol. 2016 May 2;8(5):a019372. doi:
7. Bell, C.G., Lowe, R., Adams, P.D. et al. DNA methylation aging clocks: challenges and
recommendations. Genome Biol 20, 249 (2019). https://doi.org/10.1186/s13059-019-
1824-y
8. Jin Z, Liu Y. DNA methylation in human diseases. Genes Dis. 2018 Jan 31;5(1):1-8. doi:
10.1016/j.gendis.2018.01.002. PMID: 30258928; PMCID: PMC6147084.
9. Lee HJ, Hore TA, Reik W. Reprogramming the methylome: erasing memory and creating
diversity. Cell Stem Cell. 2014 Jun 5;14(6):710-9. doi: 10.1016/j.stem.2014.05.008.
PMID: 24905162; PMCID: PMC4051243.
10. Sasaki, H., Matsui, Y. Epigenetic events in mammalian germ-cell development:
reprogramming and beyond. Nat Rev Genet 9, 129–140 (2008).
https://doi.org/10.1038/nrg2295
11. Ben Maamar M, Beck D, Nilsson E, McCarrey JR, Skinner MK. Developmental alterations
in DNA methylation during gametogenesis from primordial germ cells to sperm. iScience.
2022 Jan 19;25(2):103786. doi: 10.1016/j.isci.2022.103786. PMID: 35146397; PMCID:
PMC8819394.

90
12. Anway MD, Cupp AS, Uzumcu M, Skinner MK. Epigenetic transgenerational actions of
endocrine disruptors and male fertility. Science. 2005 Jun 3;308(5727):1466-9. doi:
10.1126/science.1108190. Erratum in: Science. 2010 May 7;328(5979):690. PMID:
15933200.
13. Bakulski KM, Fallin MD. Epigenetic epidemiology: promises for public health research.
Environ Mol Mutagen. 2014 Apr;55(3):171-83. doi: 10.1002/em.21850. Epub 2014 Jan
21. PMID: 24449392; PMCID: PMC4011487.
14. Salas, L.A., Zhang, Z., Koestler, D.C. et al. Enhanced cell deconvolution of peripheral
blood using DNA methylation for high-resolution immune profiling. Nat Commun 13, 761
(2022). https://doi.org/10.1038/s41467-021-27864-7
15. Ladd-Acosta C, Fallin MD. The role of epigenetics in genetic and environmental
epidemiology. Epigenomics. 2016 Feb;8(2):271-83. doi: 10.2217/epi.15.102. Epub 2015
Oct 27. PMID: 26505319.
16. Yamada L, Chong S. Epigenetic studies in Developmental Origins of Health and Disease:
pitfalls and key considerations for study design and interpretation. J Dev Orig Health Dis.
2017 Feb;8(1):30-43. doi: 10.1017/S2040174416000507. Epub 2016 Sep 9. PMID:
27609592.
17. Fedak KM, Bernal A, Capshaw ZA, Gross S. Applying the Bradford Hill criteria in the
21st century: how data integration has changed causal inference in molecular
epidemiology. Emerg Themes Epidemiol. 2015 Sep 30;12:14. doi: 10.1186/s12982-015-
0037-4. PMID: 26425136; PMCID: PMC4589117.
18. Knight, A.K., Park, H.J., Hausman, D.B. et al. Association between one-carbon
metabolism indices and DNA methylation status in maternal and cord blood. Sci Rep 8,
16873 (2018). https://doi.org/10.1038/s41598-018-35111-1
19. Gruzieva O, Merid SK, Chen S, Mukherjee N, Hedman AM, Almqvist C, Andolf E, Jiang
Y, Kere J, Scheynius A, Söderhäll C, Ullemar V, Karmaus W, Melén E, Arshad SH,
Pershagen G. DNA Methylation Trajectories During Pregnancy. Epigenet Insights. 2019
Aug 13;12:2516865719867090. doi: 10.1177/2516865719867090. PMID: 31453433;
PMCID: PMC6696836.
20. Zuccarello D, Sorrentino U, Brasson V, Marin L, Piccolo C, Capalbo A, Andrisani A,
Cassina M. Epigenetics of pregnancy: looking beyond the DNA code. J Assist Reprod
Genet. 2022 Apr;39(4):801-816. doi: 10.1007/s10815-022-02451-x. Epub 2022 Mar 17.
PMID: 35301622; PMCID: PMC9050975.
21. Niu Z, Mohazzab-Hosseinian S, Breton CV. Transgenerational epigenetic inheritance:
Perspectives and challenges. J Allergy Clin Immunol. 2023 Jun;151(6):1474-1476. doi:
10.1016/j.jaci.2023.02.027. Epub 2023 Mar 8. PMID: 36893860.

91
22. Peters JM, Avol E, Navidi W, London SJ, Gauderman WJ, Lurmann F, Linn WS, Margolis
H, Rappaport E, Gong H, Thomas DC. A study of twelve Southern California communities
with differing levels and types of air pollution. I. Prevalence of respiratory morbidity. Am
J Respir Crit Care Med. 1999 Mar;159(3):760-7. doi: 10.1164/ajrccm.159.3.9804143.
PMID: 10051248.
23. Bastain TM, Chavez T, Habre R, Girguis MS, Grubbs B, Toledo-Corral C, Amadeus M,
Farzan SF, Al-Marayati L, Lerner D, Noya D, Quimby A, Twogood S, Wilson M, Chatzi
L, Cousineau M, Berhane K, Eckel SP, Lurmann F, Johnston J, Dunton GF, Gilliland F,
Breton C. Study Design, Protocol and Profile of the Maternal And Developmental Risks
from Environmental and Social Stressors (MADRES) Pregnancy Cohort: a Prospective
Cohort Study in Predominantly Low-Income Hispanic Women in Urban Los Angeles.
BMC Pregnancy Childbirth. 2019 May 30;19(1):189. doi: 10.1186/s12884-019-2330-7.
PMID: 31146718; PMCID: PMC6543670.
24. Solomon PA. Air pollution and health: bridging the gap from sources to health outcomes.
Environ Health Perspect. 2011 Apr;119(4):A156-7. doi: 10.1289/ehp.1103660. PMID:
21459697; PMCID: PMC3080947.
25. Manisalidis I, Stavropoulou E, Stavropoulos A, Bezirtzoglou E. Environmental and Health
Impacts of Air Pollution: A Review. Front Public Health. 2020 Feb 20;8:14. doi:
10.3389/fpubh.2020.00014. PMID: 32154200; PMCID: PMC7044178.
26. Lurmann F, Avol E, Gilliland F. Emissions reduction policies and recent trends in Southern
California's ambient air quality. J Air Waste Manag Assoc. 2015 Mar;65(3):324-35. doi:
10.1080/10962247.2014.991856. PMID: 25947128; PMCID: PMC5737709.
27. Nussbaumer CM, Cohen RC. Impact of OA on the Temperature Dependence of PM 2.5 in
the Los Angeles Basin. Environ Sci Technol. 2021 Mar 16;55(6):3549-3558. doi:
10.1021/acs.est.0c07144. Epub 2021 Mar 4. PMID: 33661623.
28. Rahman MM, McConnell R, Schlaerth H, Ko J, Silva S, Lurmann FW, Palinkas L,
Johnston J, Hurlburt M, Yin H, Ban-Weiss G, Garcia E. The Effects of Coexposure to
Extremes of Heat and Particulate Air Pollution on Mortality in California: Implications for
Climate Change. Am J Respir Crit Care Med. 2022 Nov 1;206(9):1117-1127. doi:
10.1164/rccm.202204-0657OC. PMID: 35727303; PMCID: PMC9704834.
29. Susanto F, de Souza P, He J. Spatiotemporal Interpolation for Environmental Modelling.
Sensors (Basel). 2016 Aug 6;16(8):1245. doi: 10.3390/s16081245. PMID: 27509497;
PMCID: PMC5017410.
30. Rivera-González LO, Zhang Z, Sánchez BN, Zhang K, Brown DG, Rojas-Bracho L,
Osornio-Vargas A, Vadillo-Ortega F, O'Neill MS. An assessment of air pollutant exposure
methods in Mexico City, Mexico. J Air Waste Manag Assoc. 2015 May;65(5):581-91. doi:
10.1080/10962247.2015.1020974. PMID: 25947316; PMCID: PMC4670782.

92
31. Keil AP, Buckley JP, O'Brien KM, Ferguson KK, Zhao S, White AJ. A Quantile-Based gComputation Approach to Addressing the Effects of Exposure Mixtures. Environ Health
Perspect. 2020 Apr;128(4):47004. doi: 10.1289/EHP5838. Epub 2020 Apr 7. PMID:
32255670; PMCID: PMC7228100.
32. Keil AP, Buckley JP, Kalkbrenner AE. Bayesian G-Computation for Estimating Impacts
of Interventions on Exposure Mixtures: Demonstration With Metals From Coal-Fired
Power Plants and Birth Weight. Am J Epidemiol. 2021 Dec 1;190(12):2647-2657. doi:
10.1093/aje/kwab053. PMID: 33751055; PMCID: PMC8796809.
33. Hernán MA, Robins JM (2020). Causal Inference: What If. Boca Raton: Chapman &
Hall/CRC.
34. Pearl J. An introduction to causal inference. Int J Biostat. 2010 Feb 26;6(2):Article 7. doi:
10.2202/1557-4679.1203. PMID: 20305706; PMCID: PMC2836213.
35. Xu R, Hong X, Ladd-Acosta C, Buckley JP, Choi G, Wang G, Hou W, Wang X, Liang L,
Ji H. Contrasting Association of Maternal Plasma Biomarkers of Smoking and 1-Carbon
Micronutrients with Offspring DNA Methylation: Evidence of Aryl Hydrocarbon Receptor
Repressor Gene-Smoking-Folate Interaction. J Nutr. 2023 Aug;153(8):2339-2351. doi:
10.1016/j.tjnut.2023.05.002. Epub 2023 May 6. PMID: 37156443; PMCID:
PMC10447613.
36. Sol CM, Gaylord A, Santos S, Jaddoe VWV, Felix JF, Trasande L. Fetal exposure to
phthalates and bisphenols and DNA methylation at birth: the Generation R Study. Clin
Epigenetics. 2022 Oct 10;14(1):125. doi: 10.1186/s13148-022-01345-0. PMID: 36217170;
PMCID: PMC9552446.
37. Wu Y, Qie R, Cheng M, Zeng Y, Huang S, Guo C, Zhou Q, Li Q, Tian G, Han M, Zhang
Y, Wu X, Li Y, Zhao Y, Yang X, Feng Y, Liu D, Qin P, Hu D, Hu F, Xu L, Zhang M. Air
pollution and DNA methylation in adults: A systematic review and meta-analysis of
observational studies. Environ Pollut. 2021 Sep 1;284:117152. doi:
10.1016/j.envpol.2021.117152. Epub 2021 Apr 15. PMID: 33895575.
38. Gruzieva O, Xu CJ, Yousefi P, Relton C, Merid SK, Breton CV, Gao L, Volk HE, Feinberg
JI, Ladd-Acosta C, Bakulski K, Auffray C, Lemonnier N, Plusquin M, Ghantous A, Herceg
Z, Nawrot TS, Pizzi C, Richiardi L, Rusconi F, Vineis P, Kogevinas M, Felix JF, Duijts L,
den Dekker HT, Jaddoe VWV, Ruiz JL, Bustamante M, Antó JM, Sunyer J, Vrijheid M,
Gutzkow KB, Grazuleviciene R, Hernandez-Ferrer C, Annesi-Maesano I, Lepeule J,
Bousquet J, Bergström A, Kull I, Söderhäll C, Kere J, Gehring U, Brunekreef B, Just AC,
Wright RJ, Peng C, Gold DR, Kloog I, DeMeo DL, Pershagen G, Koppelman GH, London
SJ, Baccarelli AA, Melén E. Prenatal Particulate Air Pollution and DNA Methylation in
Newborns: An Epigenome-Wide Meta-Analysis. Environ Health Perspect. 2019
May;127(5):57012. doi: 10.1289/EHP4522. Epub 2019 May 31. PMID: 31148503;
PMCID: PMC6792178.

93
39. Gruzieva O, Xu CJ, Breton CV, Annesi-Maesano I, Antó JM, Auffray C, Ballereau S,
Bellander T, Bousquet J, Bustamante M, Charles MA, de Kluizenaar Y, den Dekker HT,
Duijts L, Felix JF, Gehring U, Guxens M, Jaddoe VV, Jankipersadsing SA, Merid SK,
Kere J, Kumar A, Lemonnier N, Lepeule J, Nystad W, Page CM, Panasevich S, Postma D,
Slama R, Sunyer J, Söderhäll C, Yao J, London SJ, Pershagen G, Koppelman GH, Melén
E. Epigenome-Wide Meta-Analysis of Methylation in Children Related to Prenatal NO2
Air Pollution Exposure. Environ Health Perspect. 2017 Jan;125(1):104-110. doi:
10.1289/EHP36. Epub 2016 Jul 22. PMID: 27448387; PMCID: PMC5226705.
40. Villicaña S, Bell JT. Genetic impacts on DNA methylation: research findings and future
perspectives. Genome Biol. 2021 Apr 30;22(1):127. doi: 10.1186/s13059-021-02347-6.
PMID: 33931130; PMCID: PMC8086086.
41. Joubert BR, Felix JF, Yousefi P, Bakulski KM, Just AC, Breton C, Reese SE, Markunas
CA, Richmond RC, Xu CJ, Küpers LK, Oh SS, Hoyo C, Gruzieva O, Söderhäll C, Salas
LA, Baïz N, Zhang H, Lepeule J, Ruiz C, Ligthart S, Wang T, Taylor JA, Duijts L, Sharp
GC, Jankipersadsing SA, Nilsen RM, Vaez A, Fallin MD, Hu D, Litonjua AA, Fuemmeler
BF, Huen K, Kere J, Kull I, Munthe-Kaas MC, Gehring U, Bustamante M, SaurelCoubizolles MJ, Quraishi BM, Ren J, Tost J, Gonzalez JR, Peters MJ, Håberg SE, Xu Z,
van Meurs JB, Gaunt TR, Kerkhof M, Corpeleijn E, Feinberg AP, Eng C, Baccarelli AA,
Benjamin Neelon SE, Bradman A, Merid SK, Bergström A, Herceg Z, Hernandez-Vargas
H, Brunekreef B, Pinart M, Heude B, Ewart S, Yao J, Lemonnier N, Franco OH, Wu MC,
Hofman A, McArdle W, Van der Vlies P, Falahi F, Gillman MW, Barcellos LF, Kumar A,
Wickman M, Guerra S, Charles MA, Holloway J, Auffray C, Tiemeier HW, Smith GD,
Postma D, Hivert MF, Eskenazi B, Vrijheid M, Arshad H, Antó JM, Dehghan A, Karmaus
W, Annesi-Maesano I, Sunyer J, Ghantous A, Pershagen G, Holland N, Murphy SK,
DeMeo DL, Burchard EG, Ladd-Acosta C, Snieder H, Nystad W, Koppelman GH, Relton
CL, Jaddoe VW, Wilcox A, Melén E, London SJ. DNA Methylation in Newborns and
Maternal Smoking in Pregnancy: Genome-wide Consortium Meta-analysis. Am J Hum
Genet. 2016 Apr 7;98(4):680-96. doi: 10.1016/j.ajhg.2016.02.019. Epub 2016 Mar 31.
PMID: 27040690; PMCID: PMC4833289.
42. Wiklund, P., Karhunen, V., Richmond, R.C. et al. DNA methylation links prenatal
smoking exposure to later life health outcomes in offspring. Clin Epigenet 11, 97 (2019).
https://doi.org/10.1186/s13148-019-0683-4
43. Everson TM, Vives-Usano M, Seyve E, Cardenas A, Lacasaña M, Craig JM, Lesseur C,
Baker ER, Fernandez-Jimenez N, Heude B, Perron P, Gónzalez-Alzaga B, Halliday J,
Deyssenroth MA, Karagas MR, Íñiguez C, Bouchard L, Carmona-Sáez P, Loke YJ, Hao
K, Belmonte T, Charles MA, Martorell-Marugán J, Muggli E, Chen J, Fernández MF, Tost
J, Gómez-Martín A, London SJ, Sunyer J, Marsit CJ, Lepeule J, Hivert MF, Bustamante
M. Placental DNA methylation signatures of maternal smoking during pregnancy and
potential impacts on fetal growth. Nat Commun. 2021 Aug 24;12(1):5095. doi:
10.1038/s41467-021-24558-y. PMID: 34429407; PMCID: PMC8384884.

94
44. Watkins SH, Iles-Caven Y, Pembrey M, Golding J, Suderman M. Grandmaternal smoking
during pregnancy is associated with differential DNA methylation in peripheral blood of
their grandchildren. Eur J Hum Genet. 2022 Dec;30(12):1373-1379. doi: 10.1038/s41431-
022-01081-2. Epub 2022 Mar 28. PMID: 35347270; PMCID: PMC9712525.
45. Luo R, Zhang H, Mukherjee N, Karmaus W, Patil V, Arshad H, Mzayek F. Association of
grandmaternal smoking during pregnancy with DNA methylation of grandchildren: the Isle
of Wight study. Epigenomics. 2021 Sep;13(18):1473-1483. doi: 10.2217/epi-2020-0433.
Epub 2021 Oct 1. PMID: 34596434; PMCID: PMC8503801.
46. Pascual ZN, Langaker MD. Physiology, Pregnancy. 2022 May 19. In: StatPearls [Internet].
Treasure Island (FL): StatPearls Publishing; 2023 Jan–. PMID: 32644730.
47. Jensen D, Wolfe LA, Slatkovska L, Webb KA, Davies GA, O'Donnell DE. Effects of
human pregnancy on the ventilatory chemoreflex response to carbon dioxide. Am J Physiol
Regul Integr Comp Physiol. 2005 May;288(5):R1369-75. doi:
10.1152/ajpregu.00862.2004. Epub 2005 Jan 27. PMID: 15677521.
48. National Academies of Sciences, Engineering, and Medicine; Health and Medicine
Division; Division of Behavioral and Social Sciences and Education; Board on Children,
Youth, and Families; Committee on Assessing Health Outcomes by Birth Settings; Backes
EP, Scrimshaw SC, editors. Birth Settings in America: Outcomes, Quality, Access, and
Choice. Washington (DC): National Academies Press (US); 2020 Feb 6. 4, Systemic
Influences on Outcomes in Pregnancy and Childbirth. Available from:
https://www.ncbi.nlm.nih.gov/books/NBK555488/
49. Neggers YH. Trends in maternal mortality in the United States. Reprod Toxicol. 2016
Sep;64:72-6. doi: 10.1016/j.reprotox.2016.04.001. Epub 2016 Apr 7. PMID: 27063184.
50. Branco MR, King M, Perez-Garcia V, Bogutz AB, Caley M, Fineberg E, Lefebvre L, Cook
SJ, Dean W, Hemberger M, Reik W. Maternal DNA Methylation Regulates Early
Trophoblast Development. Dev Cell. 2016 Jan 25;36(2):152-63. doi:
10.1016/j.devcel.2015.12.027. PMID: 26812015; PMCID: PMC4729543.
51. Gruzieva O, Merid SK, Chen S, Mukherjee N, Hedman AM, Almqvist C, Andolf E, Jiang
Y, Kere J, Scheynius A, Söderhäll C, Ullemar V, Karmaus W, Melén E, Arshad SH,
Pershagen G. DNA Methylation Trajectories During Pregnancy. Epigenet Insights. 2019
Aug 13;12:2516865719867090. doi: 10.1177/2516865719867090. PMID: 31453433;
PMCID: PMC6696836.
52. Knight AK, Dunlop AL, Kilaru V, Cobb D, Corwin EJ, Conneely KN, Smith AK.
Characterization of gene expression changes over healthy term pregnancies. PLoS One.
2018 Oct 10;13(10):e0204228. doi: 10.1371/journal.pone.0204228. PMID: 30303981;
PMCID: PMC6179206.

95
53. Møller P, Danielsen PH, Karottki DG, Jantzen K, Roursgaard M, Klingberg H, Jensen DM,
Christophersen DV, Hemmingsen JG, Cao Y, Loft S. Oxidative stress and inflammation
generated DNA damage by exposure to air pollution particles. Mutat Res Rev Mutat Res.
2014 Oct-Dec;762:133-66. doi:
54. Aryee MJ, Jaffe AE, Corrada-Bravo H, Ladd-Acosta C, Feinberg AP, Hansen KD, Irizarry
RA (2014). “Minfi: A flexible and comprehensive Bioconductor package for the analysis
of Infinium DNA Methylation microarrays.” Bioinformatics, 30(10), 1363–
1369. doi:10.1093/bioinformatics/btu049.
55. Pidsley et al., Critical evaluation of the Illumina MethylationEPIC BeadChip microarray
for whole-genome DNA methylation profiling. Genome Biology (2016)
56. McCartney et al., Identification of polymorphic and off-target probe binding sites on the
Illumina Infinium MethylationEPIC BeadChip. Genomics Data (2016)
57. TJ Triche, DJ Weisenberger, D Van Den Berg, PW Laird and KD Siegmund Low-level
processing of Illumina Infinium DNA Methylation BeadArrays. Nucleic Acids Res (2013)
41, e90. doi:10.1093/nar/gkt090.
58. N Touleimat and J Tost. Complete pipeline for Infinium Human Methylation 450K
BeadChip data processing using subset quantile normalization for accurate DNA
methylation estimation. Epigenomics (2012) 4:325-341.
59. Zheng SC, Webster AP, Dong D, Feber A, Graham DG, Sullivan R, Jevons S, Lovat LB,
Beck S, Widschwendter M, Teschendorff AE A novel cell-type deconvolution algorithm
reveals substantial contamination by immune cells in saliva, buccal and
cervix. Epigenomics (2018) 10: 925-940. doi: 10.2217/epi-2018-0037.
60. Rahmani E, Shenhav L, Schweiger R, Yousefi P, Huen K, Eskenazi B, Eng C, Huntsman
S, Hu D, Galanter J, Oh SS, Waldenberger M, Strauch K, Grallert H, Meitinger T, Gieger
C, Holland N, Burchard EG, Zaitlen N, Halperin E. Genome-wide methylation data mirror
ancestry information. Epigenetics Chromatin. 2017 Jan 3;10:1. doi: 10.1186/s13072-016-
0108-y. PMID: 28149326; PMCID: PMC5267476.
61. Leek JT, Johnson WE, Parker HS, Jaffe AE, Storey JD. The sva package for removing
batch effects and other unwanted variation in high-throughput experiments.
Bioinformatics. 2012 Mar 15;28(6):882-3. doi: 10.1093/bioinformatics/bts034. Epub 2012
Jan 17. PMID: 22257669; PMCID: PMC3307112.
62. Phipson, B., Maksimovic, J., and Oshlack, A. (2016). missMethyl: an R package for
analysing methylation data from Illuminas HumanMethylation450
platform. Bioinformatics, 15;32(2), 286–8.
63. Geeleher, P., Hartnett, L., Egan, L. J., Golden, A., Ali, R. A. R., and Seoighe, C. (2013).
Gene-set analysis is severely biased when applied to genome-wide methylation
data. Bioinformatics, 29(15), 1851–1857.

96
64. Vuong NQ, Goegan P, Mohottalage S, Breznan D, Ariganello M, Williams A, Elisma F,
Karthikeyan S, Vincent R, Kumarathasan P. Proteomic changes in human lung epithelial
cells (A549) in response to carbon black and titanium dioxide exposures. J Proteomics.
2016 Oct 21;149:53-63. doi: 10.1016/j.jprot.2016.03.046. Epub 2016 Apr 12. PMID:
27084686.
65. Liu S, Zhang W, Zhang F, Roepstorff P, Yang F, Lu Z, Ding W. TMT-Based Quantitative
Proteomics Analysis Reveals Airborne PM2.5-Induced Pulmonary Fibrosis. Int J Environ
Res Public Health. 2018 Dec 31;16(1):98. doi: 10.3390/ijerph16010098. PMID: 30602677;
PMCID: PMC6339163.
66. Chu JH, Hart JE, Chhabra D, Garshick E, Raby BA, Laden F. Gene expression network
analyses in response to air pollution exposures in the trucking industry. Environ Health.
2016 Nov 3;15(1):101. doi: 10.1186/s12940-016-0187-z. PMID: 27809917; PMCID:
PMC5093980.
67. Li, R., Wang, T., Marquardt, R.M. et al. TRIM28 modulates nuclear receptor signaling to
regulate uterine function. Nat Commun 14, 4605 (2023). https://doi.org/10.1038/s41467-
023-40395-7
68. Alexander KA, Wang X, Shibata M, Clark AG, García-García MJ. TRIM28 Controls
Genomic Imprinting through Distinct Mechanisms during and after Early Genome-wide
Reprogramming. Cell Rep. 2015 Nov 10;13(6):1194-1205. doi:
10.1016/j.celrep.2015.09.078. Epub 2015 Oct 29. PMID: 26527006; PMCID:
PMC4644443.
69. Marín-Palma, D., Fernandez, G.J., Ruiz-Saenz, J. et al. Particulate matter impairs immune
system function by up-regulating inflammatory pathways and decreasing pathogen
response gene expression. Sci Rep 13, 12773 (2023). https://doi.org/10.1038/s41598-023-
39921-w
70. Laing S, Wang G, Briazova T, Zhang C, Wang A, Zheng Z, Gow A, Chen AF, Rajagopalan
S, Chen LC, Sun Q, Zhang K. Airborne particulate matterselectively activates endoplasmic
reticulum stress response in the lung and liver tissues. Am J Physiol Cell Physiol. 2010
Oct;299(4):C736-49. doi: 10.1152/ajpcell.00529.2009. Epub 2010 Jun 16. PMID:
20554909; PMCID: PMC2957267.
71. Onoda A, Kawasaki T, Tsukiyama K, Takeda K, Umezawa M. Carbon nanoparticles
induce endoplasmic reticulum stress around blood vessels with accumulation of misfolded
proteins in the developing brain of offspring. Sci Rep. 2020 Jun 22;10(1):10028. doi:
10.1038/s41598-020-66744-w. PMID: 32572058; PMCID: PMC7308306.
72. Adivi A, Lucero J, Simpson N, McDonald JD, Lund AK. Traffic-generated air pollution -
Exposure mediated expression of factors associated with demyelination in a female
apolipoprotein E-/- mouse model. Neurotoxicol Teratol. 2022 Mar-Apr;90:107071. doi:
10.1016/j.ntt.2022.107071. Epub 2022 Jan 10. PMID: 35016995; PMCID: PMC8904307.

97
73. Hooper LG, Kaufman JD. Ambient Air Pollution and Clinical Implications for Susceptible
Populations. Ann Am Thorac Soc. 2018 Apr;15(Suppl 2):S64-S68. doi:
10.1513/AnnalsATS.201707-574MG. PMID: 29676646; PMCID: PMC5955035.
74. Ghazi T, Naidoo P, Naidoo RN, Chuturgoon AA. Prenatal Air Pollution Exposure and
Placental DNA Methylation Changes: Implications on Fetal Development and Future
Disease Susceptibility. Cells. 2021 Nov 5;10(11):3025. doi: 10.3390/cells10113025.
PMID: 34831248; PMCID: PMC8616150.
75. Isaevska E, Moccia C, Asta F, Cibella F, Gagliardi L, Ronfani L, Rusconi F, Stazi MA,
Richiardi L. Exposure to ambient air pollution in the first 1000 days of life and alterations
in the DNA methylome and telomere length in children: A systematic review. Environ Res.
2021 Feb;193:110504. doi: 10.1016/j.envres.2020.110504. Epub 2020 Nov 20. PMID:
33221306.
76. Fedulov AV, Leme A, Yang Z, Dahl M, Lim R, Mariani TJ, Kobzik L. Pulmonary exposure
to particles during pregnancy causes increased neonatal asthma susceptibility. Am J Respir
Cell Mol Biol. 2008 Jan;38(1):57-67. doi: 10.1165/rcmb.2007-0124OC. Epub 2007 Jul 26.
PMID: 17656681; PMCID: PMC2176127.
77. Kim HJ, Min JY, Min KB, Seo YS, Sung J, Yun JM, Kwon H, Cho B, Park JH, Kim JI.
CDH13 gene-by-PM10 interaction effect on lung function decline in Korean men.
Chemosphere. 2017 Feb;168:583-589. doi: 10.1016/j.chemosphere.2016.11.022. Epub
2016 Nov 10. PMID: 27839881.
78. Kim HJ, Seo YS, Sung J, Son HY, Yun JM, Kwon H, Cho B, Kim JI, Park JH. Interactions
of CDH13 gene polymorphisms and ambient PM10 air pollution exposure with blood
pressure and hypertension in Korean men. Chemosphere. 2019 Mar;218:292-298. doi:
10.1016/j.chemosphere.2018.11.125. Epub 2018 Nov 20. PMID: 30476760.
79. Lee MK, Xu CJ, Carnes MU, Nichols CE, Ward JM; BIOS consortium; Kwon SO, Kim
SY, Kim WJ, London SJ. Genome-wide DNA methylation and long-term ambient air
pollution exposure in Korean adults. Clin Epigenetics. 2019 Feb 28;11(1):37. doi:
10.1186/s13148-019-0635-z. PMID: 30819252; PMCID: PMC6396524.
80. Wiklund, P., Karhunen, V., Richmond, R.C. et al. DNA methylation links prenatal
smoking exposure to later life health outcomes in offspring. Clin Epigenet 11, 97 (2019).
https://doi.org/10.1186/s13148-019-0683-4
81. Thacher JD, Schultz ES, Hallberg J, Hellberg U, Kull I, Thunqvist P, Pershagen G,
Gustafsson PM, Melén E, Bergström A. Tobacco smoke exposure in early life and
adolescence in relation to lung function. Eur Respir J. 2018 Jun 7;51(6):1702111. doi:
10.1183/13993003.02111-2017. PMID: 29748304; PMCID: PMC6003782.
82. Li YF, Langholz B, Salam MT, Gilliland FD. Maternal and grandmaternal smoking
patterns are associated with early childhood asthma. Chest. 2005 Apr;127(4):1232-41. doi:
10.1378/chest.127.4.1232. PMID: 15821200.

98
83. Magnus MC, Håberg SE, Karlstad Ø, Nafstad P, London SJ, Nystad W. Grandmother's
smoking when pregnant with the mother and asthma in the grandchild: the Norwegian
Mother and Child Cohort Study. Thorax. 2015 Mar;70(3):237-43. doi: 10.1136/thoraxjnl2014-206438. Epub 2015 Jan 8. PMID: 25572596; PMCID: PMC5034931.
84. Feinberg AP, Koldobskiy MA, Göndör A. Epigenetic modulators, modifiers and mediators
in cancer aetiology and progression. Nat Rev Genet. 2016 May;17(5):284-99. doi:
10.1038/nrg.2016.13. Epub 2016 Mar 14. PMID: 26972587; PMCID: PMC4888057.
85. Breton CV, Landon R, Kahn LG, Enlow MB, Peterson AK, Bastain T, Braun J, Comstock
SS, Duarte CS, Hipwell A, Ji H, LaSalle JM, Miller RL, Musci R, Posner J, Schmidt R,
Suglia SF, Tung I, Weisenberger D, Zhu Y, Fry R. Exploring the evidence for epigenetic
regulation of environmental influences on child health across generations. Commun Biol.
2021 Jun 22;4(1):769. doi: 10.1038/s42003-021-02316-6. PMID: 34158610; PMCID:
PMC8219763.
86. Soubry A, Hoyo C, Jirtle RL, Murphy SK. A paternal environmental legacy: evidence for
epigenetic inheritance through the male germ line. Bioessays. 2014 Apr;36(4):359-71. doi:
10.1002/bies.201300113. Epub 2014 Jan 16. PMID: 24431278; PMCID: PMC4047566.
87. Rehan VK, Liu J, Naeem E, Tian J, Sakurai Rlme, Kwong K, Akbari O, Torday JS.
Perinatal nicotine exposure induces asthma in second generation offspring. BMC Med.
2012 Oct 30;10:129. doi: 10.1186/1741-7015-10-129. PMID: 23106849; PMCID:
PMC3568737.
88. Rehan VK, Liu J, Sakurai R, Torday JS. Perinatal nicotine-induced transgenerational
asthma. Am J Physiol Lung Cell Mol Physiol. 2013 Oct 1;305(7):L501-7. doi:
10.1152/ajplung.00078.2013. Epub 2013 Aug 2. PMID: 23911437; PMCID:
PMC3798762.
89. Leslie FM. Multigenerational epigenetic effects of nicotine on lung function. BMC Med.
2013 Feb 4;11:27. doi: 10.1186/1741-7015-11-27. PMID: 23379695; PMCID:
PMC3585797.
90. Li Y, Hecht SS. Carcinogenic components of tobacco and tobacco smoke: A 2022 update.
Food Chem Toxicol. 2022 Jul;165:113179. doi: 10.1016/j.fct.2022.113179. Epub 2022
May 25. PMID: 35643228; PMCID: PMC9616535.
91. Baccarelli A, Kaufman JD. Ambient particulate air pollution, environmental tobacco
smoking, and childhood asthma: interactions and biological mechanisms. Am J Respir Crit
Care Med. 2011 Dec 15;184(12):1325-7. doi: 10.1164/rccm.201109-1695ED. PMID:
22174109; PMCID: PMC4354508.
92. Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, Smyth GK. limma powers
differential expression analyses for RNA-sequencing and microarray studies. Nucleic
Acids Res. 2015 Apr 20;43(7):e47. doi: 10.1093/nar/gkv007. Epub 2015 Jan 20. PMID:
25605792; PMCID: PMC4402510.

99
93. Peters TJ, Buckley MJ, Statham AL, Pidsley R, Samaras K, V Lord R, Clark SJ, Molloy
PL. De novo identification of differentially methylated regions in the human genome.
Epigenetics Chromatin. 2015 Jan 27;8:6. doi: 10.1186/1756-8935-8-6. PMID: 25972926;
PMCID: PMC4429355.
94. Kuznetsova A, Brockhoff PB, Christensen RHB (2017). “lmerTest Package: Tests in
Linear Mixed Effects Models.” Journal of Statistical Software, 82(13), 1–
26. doi:10.18637/jss.v082.i13.
95. Fedulov AV, Leme A, Yang Z, Dahl M, Lim R, Mariani TJ, Kobzik L. Pulmonary exposure
to particles during pregnancy causes increased neonatal asthma susceptibility. Am J Respir
Cell Mol Biol. 2008 Jan;38(1):57-67. doi: 10.1165/rcmb.2007-0124OC. Epub 2007 Jul 26.
PMID: 17656681; PMCID: PMC2176127.
96. Meadows E, Cho JH, Flynn JM, Klein WH. Myogenin regulates a distinct genetic program
in adult muscle stem cells. Dev Biol. 2008 Oct 15;322(2):406-14. doi:
10.1016/j.ydbio.2008.07.024. Epub 2008 Jul 30. PMID: 18721801.
97. Ugidos N, Mena J, Baquero S, Alloza I, Azkargorta M, Elortza F, Vandenbroeck K.
Interactome of the Autoimmune Risk Protein ANKRD55. Front Immunol. 2019 Sep
18;10:2067. doi: 10.3389/fimmu.2019.02067. PMID: 31620119; PMCID: PMC6759997.
98. Liu, L., Khan, A., Sanchez-Rodriguez, E. et al. Genetic regulation of serum IgA levels and
susceptibility to common immune, infectious, kidney, and cardio-metabolic traits. Nat
Commun 13, 6859 (2022). https://doi.org/10.1038/s41467-022-34456-6
99. Joehanes R, Just AC, Marioni RE, Pilling LC, Reynolds LM, Mandaviya PR, Guan W, Xu
T, Elks CE, Aslibekyan S, Moreno-Macias H, Smith JA, Brody JA, Dhingra R, Yousefi P,
Pankow JS, Kunze S, Shah SH, McRae AF, Lohman K, Sha J, Absher DM, Ferrucci L,
Zhao W, Demerath EW, Bressler J, Grove ML, Huan T, Liu C, Mendelson MM, Yao C,
Kiel DP, Peters A, Wang-Sattler R, Visscher PM, Wray NR, Starr JM, Ding J, Rodriguez
CJ, Wareham NJ, Irvin MR, Zhi D, Barrdahl M, Vineis P, Ambatipudi S, Uitterlinden AG,
Hofman A, Schwartz J, Colicino E, Hou L, Vokonas PS, Hernandez DG, Singleton AB,
Bandinelli S, Turner ST, Ware EB, Smith AK, Klengel T, Binder EB, Psaty BM, Taylor
KD, Gharib SA, Swenson BR, Liang L, DeMeo DL, O'Connor GT, Herceg Z, Ressler KJ,
Conneely KN, Sotoodehnia N, Kardia SL, Melzer D, Baccarelli AA, van Meurs JB,
Romieu I, Arnett DK, Ong KK, Liu Y, Waldenberger M, Deary IJ, Fornage M, Levy D,
London SJ. Epigenetic Signatures of Cigarette Smoking. Circ Cardiovasc Genet. 2016
Oct;9(5):436-447. doi: 10.1161/CIRCGENETICS.116.001506. Epub 2016 Sep 20. PMID:
27651444; PMCID: PMC5267325.
100. Kendall RT, Feghali-Bostwick CA. Fibroblasts in fibrosis: novel roles and mediators. Front
Pharmacol. 2014 May 27;5:123. doi: 10.3389/fphar.2014.00123. PMID: 24904424;
PMCID: PMC4034148.

100
101. Ufartes R, Berger H, Till K, Salinas G, Sturm M, Altmüller J, Nürnberg P, Thiele H, Funke
R, Apeshiotis N, Langen H, Wollnik B, Borchers A, Pauli S. De novo mutations in
FBRSL1 cause a novel recognizable malformation and intellectual disability syndrome.
Hum Genet. 2020 Nov;139(11):1363-1379. doi: 10.1007/s00439-020-02175-x. Epub 2020
May 18. PMID: 32424618; PMCID: PMC7519918.
102. Anisimov SV. A Prevalence of Imprinted Genes within the Total Transcriptomes of Human
Tissues and Cells. Mol Biol Int. 2012;2012:793506. doi: 10.1155/2012/793506. Epub 2012
Sep 11. PMID: 22997578; PMCID: PMC3446743.
103. RStudio Team (2020). RStudio: Integrated Development for R. RStudio, PBC, Boston,
MA URL http://www.rstudio.com/.
104. Wickham H (2016). ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag New
York. ISBN 978-3-319-24277-4, https://ggplot2.tidyverse.org.
105. Turner S (2018). “qqman: an R package for visualizing GWAS results using Q-Q and
manhattan plots.” The Journal of Open Source Software. doi:10.21105/joss.00731.
106. Textor J, van der Zander B, Gilthorpe MS, Liśkiewicz M, Ellison GT (2016). “Robust
causal inference using directed acyclic graphs: the R package 'dagitty'.” International
Journal of Epidemiology, 45(6), 1887–1894. doi:10.1093/ije/dyw341.
107. Blighe K, Rana S, Lewis M (2023). EnhancedVolcano: Publication-ready volcano plots
with enhanced colouring and labeling. R package version
1.18.0, https://github.com/kevinblighe/EnhancedVolcano.

Abstract (if available)

Abstract Background
Epigenetic states, like DNA methylation (DNAm), may increase the risk of disease in humans. The developmental period is a particularly sensitive window for the mother because of the increase in energetic demands to support development, and for the fetus and associated germ cells as well due to epigenetic programming. Consequently, the developmental window can impact the health of three generations at once: the gestating mother (grandparental), developing fetus (parental), and germ cell (grandchild). This can occur potentially through direct intergenerational epigenetic effects. Environmental toxicants can differentially or consistently impact epigenetic processes across these three generations and subsequently increase risk of disease. Less has been studied on the effect of toxicants on the gestating mother, as the literature focuses on the more explicit epigenetic changes occurring in the developing fetus. Little is also published on intergenerational epigenetic profiles, mainly due to lack of available data. This thesis aims to address these gaps in the literature using longitudinal maternal whole blood DNAm signatures paired to cord blood DNAm from the Maternal and Developmental Risks from Environmental Stressors (MADRES) cohort, and three generations of buccal cell DNAm data in the Children’s Health Study (CHS). This study will focus on the effects of particulate matter and ambient air pollutant mixtures as an exposure, along with grandparental prenatal tobacco smoking (PTS) in the CHS. Both cohorts are based in Southern California, where air pollution contributes to significant mortality and morbidity across the region and is therefore of public health significance.
Objectives
The first study aims to address the dearth of studies on the effects of prenatal particulate matter and ambient air pollution mixtures on maternal epigenetic changes. The second study seeks to examine ambient air pollution mixture differences on both changes in maternal DNAm and cord blood DNAm to add to the literature on intergenerational signatures. The potential impact of air pollutant related maternal DNAm changes inducing cord blood DNAm signatures will also be evaluated. The final study will test for evidence of a direct intergenerational effect of grandmaternal PTS on parental and grandchild DNAm signatures and examine the potential synergistic or antagonistic impact of parental air pollution exposure on grandchild DNAm among PTS-affected CpGs.
Methods
The first study used adjusted linear models with an interaction between gestational age (GA) and particulate matter and used prenatal DNA methylation signatures as the outcome. Quantile g-computation was also used with the four ambient air pollutants as an exposure mixture including the following pollutants of interest used in this thesis: NO2, O3, PM2.5, and PM10 on an epigenome-wide scale. The second study evaluated change in ambient air pollution mixture, with quantile g-computation, on differences in maternal DNAm and impacts on infant DNAm. Cord blood epigenome-wide impacts of ambient air pollutant affected maternal DNAm signatures will also be evaluated. The third study analyzed linear mixed models with a random effect for family to investigate the potential change in methylation between the referent grandmaternal methylation profile versus parental and grandchild DNAm signatures among families with a smoking grandmother. DNAm signatures analyzed in the linear mixed model were chosen based
on overlap in generational and sex stratified epigenome-wide analyses. The potential synergistic or antagonistic impact of parental ambient air pollution exposure on overlapping DNAm signatures will also be
evaluated using quantile g-computation. Multigenerational exposure to smoking will be accounted for in the mixture model.
Results
Prenatal PM2.5 and ambient air pollution resulted in changes to 136 and 30 maternal CpGs, respectively, between early and late pregnancy in the MADRES cohort. There was no overlap between the mixtures and PM2.5 DNAm identified. There was some evidence that exposure to PM2.5 in pregnancy induced CpG signatures that were persistent into the postpartum period (>2 years) in one CpG. Changes in ambient air pollution between early and late pregnancy also induced change in 22 CpGs in pregnancy, but there was no evidence that change affected DNAm in cord blood. Grandmaternal smoking was associated with DNAm in male and female offspring and all paternal and maternal grandchildren. There was overlap in FBRSL1 and ADARB2 methylation between male offspring and the paternal female grandchild that differed among families exposed to grandmaternal smoke compared to those who were not. Parental ambient air pollution exposure did not synergistically or antagonistically impact PTS-affected DNAm in the grandchild generation.
Conclusion
Molecular states like DNA methylation can by impacted by environmental toxicants, and during the prenatal period this has the potential to affect the grandparental, parental, and grandchild generation. Study one and study two demonstrated some effect of cumulative levels as well as changes in prenatal PM2.5 and ambient air pollution exposures, affecting maternal DNA methylation profiles in pregnancy. There was no evidence for this in our cord blood DNAm, but our sample size was modest. Finally, there was evidence of an intergenerational effect of grandmaternal smoking in FBRSL1, with persistent decreased methylation signatures among families exposed to smoking in the paternal parent and female grandchild lineage.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Prenatal air pollution exposure, newborn DNA methylation, and childhood respiratory health

PDF

Air pollution, mitochondrial function, and growth in children

PDF

Genetic epidemiological approaches in the study of risk factors for hematologic malignancies

PDF

A cohort study of air-pollution and childhood obesity incidence

PDF

Personal exposure to particulate matter PM2.5 sources during pregnancy and birthweight

PDF

Air pollution and breast cancer survival in California teachers: using address histories and individual-level data

PDF

Preprocessing and analysis of DNA methylation microarrays

PDF

Associations between ambient air pollution and hypertensive disorders of pregnancy

PDF

Examining exposure to extreme heat and air pollution and its effects on all-cause, cardiovascular, and respiratory mortality in California: effect modification by the social deprivation index

PDF

Ambient air pollution and lung function in children

PDF

Association of maternal and environmental factors with infant feeding behaviors in a birth cohort study

PDF

The interplay between tobacco exposure and polygenic risk score for growth on birthweight and childhood acute lymphoblastic leukemia

PDF

Prenatal environmental exposures and fetal growth in the MADRES cohort

PDF

Prenatal exposure to organophosphate esters and early neurodevelopment

PDF

A pilot study of a global approach to assessing air pollution exposure in port communities: passive air monitoring of nitrogen dioxide concentrations

PDF

Association of traffic-related air pollution and lens opacities in the Los Angeles Latino Eye Study

PDF

Evaluation of new methods for estimating exposure to traffic-related pollution and early health effects for large population epidemiological studies

PDF

The environmental and genetic determinants of cleft lip and palate in the global setting

PDF

Airway inflammation and respiratory health in the Southern California children's health study

PDF

Chronic eye disease epidemiology in the multiethnic ophthalmology cohorts of California study

Asset Metadata

Creator Mohazzab-Hosseinian, Sahra (author)

Core Title Air pollution, smoking, and multigenerational DNA methylation Signatures: a study of two southern California cohorts

School Keck School of Medicine

Degree Doctor of Philosophy

Degree Program Epidemiology

Degree Conferral Date 2023-12

Publication Date 10/30/2024

Defense Date 10/05/2023

Publisher Los Angeles, California (original), University of Southern California (original), University of Southern California. Libraries (digital)

Tag Air pollution,DNA methylation,epigenetics,family,longitudinal,multigenerational,OAI-PMH Harvest,prenatal,smoking

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Breton, Carrie (committee chair), Garcia, Erika (committee member), Mancuso, Nicholas (committee member), Marconett, Crystal (committee member), Wiemels, Joseph (committee member)

Creator Email mohazzab@usc.edu,smohosseinian@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC113763043

Unique identifier UC113763043

Identifier etd-MohazzabHo-12444.pdf (filename)

Legacy Identifier etd-MohazzabHo-12444

Document Type Dissertation

Format theses (aat)

Rights Mohazzab-Hosseinian, Sahra

Internet Media Type application/pdf

Type texts

Source 20231103-usctheses-batch-1104 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu