Full text



Paige Annalise Bommarito

A dissertation submitted to the faculty at the University of North Carolina at Chapel Hill in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department

of Environmental Sciences and Engineering in the Gillings School of Global Public Health.

Chapel Hill 2020

Approved by: Rebecca C. Fry Kelly Ferguson Kun Lu


© 2020



Paige Annalise Bommarito: Associations Between Toxic Metals and Preeclampsia: An Interdisciplinary Approach to Understanding Disease Etiology

(Under the direction of Rebecca C. Fry)



I would like to thank my advisor, Rebecca Fry, for her guidance and support, as well as the rest of my dissertation committee: Kelly Ferguson, Kun Lu, David Richardson, and Jill Stewart. I would also like to acknowledge my funding sources, the National Institues of Health (R01 ES019315, P42 ES005948, 5U01DE014577, and ZIA103321), the National Institue of Environmental Health Sciences (T32ES007018 and R01ES018872), and the Children’s Health Exposure Analysis Research (CHEAR) Program (U2CES026555), and the Institute for







The Placenta and its Formation ...1

Preeclampsia ...4

Risk Factors for Preeclampsia: Demographic and Behavioral ...5

Risk Factors for Preeclampsia: Genetic ...7

Risk Factors for Preeclampsia: Clinical...7

Aberrant Placentation in Preeclampsia ...7

Model Organisms for Preeclampsia: In Vitro Studies ... 10

Model Organisms for Preeclampsia: In Vivo Studies ... 12

Preeclampsia Etiology ... 13

Toxic Metals and Preeclampsia... 14

Cadmium and Preeclampsia ... 15

The Potentially Protective Effect of Essential Metals ... 16

Research GapsiIn Toxic Metals and Preeclampsia ... 18

Project Approach ... 18



1.3 Materials and Methods ... 23

1.3.1 Subject Recruitment ... 23

1.3.2 Urinary Trace Metals Analysis ... 25

1.3.3 Plasma Biomarkers of Angiogenesis ... 25

1.3.4 Statistical Analysis ... 26

1.4 Results ... 28

1.4.1 Subject Demographics and Urinary Metals... 28

1.4.2 Urinary Trace Metals and Preeclampsia Risk ... 32

1.4.3 Urinary Trace Metals and Circulating Maternal Angiogenic Biomarkers ... 33

1.4.4 Trace Metals Mixtures, Preeclampsia Risk, and Angiogenic Biomarkers ... 36

1.5 Discussion ... 37

1.6 Conclusions ... 43


2.1 Overview ... 44

2.2 Objective ... 46

2.3 Materials and Methods ... 46

2.3.1 Study Recruitment ... 46

2.3.2 Sample Collection and Metals Analysis ... 47

2.3.3 Placental RNA Extraction and Gene Expression Analysis ... 48

2.3.4 Statistical Analysis ... 50

2.4 Results ... 52

2.4.1 Sample Description ... 52


2.4.3 Cadmium and Placental Gene Expression ... 57

2.4.4 Differentially Expressed Genes and Preeclampsia ... 60

2.5 Discussion ... 60

2.6 Conclusions ... 66


3.1 Introduction ... 67

3.2 Objective ... 68

3.3 Methods ... 69

3.3.1 Cell Culture and Treatment ... 69

3.3.2 Cytotoxicity Assay ... 69

3.3.3 Accumulation Assay ... 70

3.3.4 Glutathione Depletion Assay ... 71

3.3.5 RNA Extraction and mRNA Expression ... 71

3.3.6 Measurement of Protein Secretion ... 72

3.3.7 Trophoblast Migration ... 72

3.3.8 Statistical Analysis ... 73

3.4 Results ... 74

3.4.1 Cd Accumulates in HTR-8/SVneo Cells and Induces MT Expression ... 74

3.4.2 Cd Exposure Alters TGFB Pathway Expression ... 75

3.4.3 Cd Exposure Inhibits Trophoblast Migration ... 77

3.4.4 Cd-Induced GSH Depletion is Prevented by Zn Pre-treatment ... 77


3.4.6 Zn Pre-Treatment Prevents Cd-Induced Changes in TGFB

Target Genes ... 80

3.4.7 Zn Pre-Treatment Prevents Cd-Induced Inhibition of Trophoblast Migration ... 82

3.5 Discussion ... 82

3.6 Conclusions ... 87


4.1 Project Themes ... 89

4.1.1 Integration of Scientific Approaches Provides Mechanistic Support of Population-Level Observations ... 89

4.1.2 Essential Metals May Protect Against the Effects of Toxic Metals Within the Placenta ... 91

4.2 Public Health Relevance ... 92

4.3 Future Research Directions ... 93

4.3.1 Using Toxicological Evidence to Inform Bayesian Analyses of Epidemiologic Data ... 93

4.3.2 Examining the Effects of Cd as an Endocrine Disruptor Within the Placenta ... 98

4.3.3 Understanding Whether Cd Affects Trophoblast Differentiation and Placental Cellular Composition ... 100

4.4 Conclusions ... 103


A 1.1 Flow-Through Diagram Displaying Study Design. ... 104

A 1.2 Simplified Directed Acyclic Graph (DAG) Used to Identify Potential Confounders and/or Colliders. ... 105

A 1.3 Unadjusted and Adjusted Relationship Between Urinary Metals and the HR (95% CI) of Preeclampsia. ... 106


A 1.5 Standardized and Rotated Loading Factors and Communalities for

Each Variable Within Each Principal Component. ... 109 A 1.6 Interactions Between Individual Trace Metals Loading onto PC1

(Cu, Se, and Zn) and PC2 (Cd, Pb, and Mn)... 110 APPENDIX 2: SUPPLEMENTAL MATERIAL FOR CHAPTER 2 ... 111

A 2.1 Relative Expression of Genes Measured in the Placenta of

Motor-CC (N = 166). ... 111 A 2.2 Simplified DAG of the Association Between Placental Cd Levels

and Gene Expresion. ... 114 A 2.3 Estimates (95% CI) of Demogrpahic Variables in Relation to

Placental Metals in Motor-CC... 115 A 2.4 Relative Expression of Genes Measured in the Placenta of

Motor-CC (N=166). ... 117 A 2.5 Association Between Placental Cd and Gene Expression Stratified

by Low (< Median) vs. High (>Median) Levels of Placental Se Where

There Was Significance in One or Both Strata. ... 119 A 2.6 Association Between Placental Cd and Gene Expression Stratified

by (< Median) vs. High (>Median) Levels of Placental Zn Where There

Was Significance in One or Both Strata. ... 120 APPENDIX 3: SUPPLEMENTAL MATERIAL FOR CHAPTER 3 ... 121

A 3.1. Cytotoxicity for CdCl2, With or Without ZnCl2 Pre-Treatment, in

Htr-8/Svneo Cells. ... 121 A 3.2 Line Plot Displaying Observed AUC of Trophoblast Migration

Between (A) Controls and 10 µM CdCl2, and (B) Controls and 10 µM

CdCl2, With or Without ZnCl2 Pre-Treatment, in Htr-8/Svneo Cells. ... 122

A 3.3 Secreted Concentrations of (A) TIMP1 and (B) MMP9 and (C) the TIMP1/MMP9 Ratio in Htr-8/Svneo Cells Treated With Control Media or

10 µM CdCl2. ... 123

A 3.4 Secreted Concentrtaions of (A) TIMP1 and (B) MMP9 and the (C) TIMP1/MMP9 Ratio in Htr-8/Svneo Treated With Control Media or and

10 µM CdCl2, With or Without ZnCl2 Pre-Treatment... 124


A 4.1 Prior Distributions for Estimates Used in Bayesian Regression Models Examining Placental Cd Concentrations in Relation to Gene

Expression Within Strata of Placental Zn. ... 125 A 4.2 Comparison of Estimates from MLE and Bayesian Analyses of the

Association Between Placental Cd Levels and Placental Gene Expression



Table 1. Weighted LIFECODES demographic characteristics overall and by

preeclampsia status: crude N (weighted %) or weighted median (weighted IQR). ... 30 Table 2. Weighted distribution of specific gravity-adjusted trace metals from 3rd

study visit urine samples (µg/L) by preeclampsia case status (N=383). ... 32 Table 3. Adjusted association between urinary metals and the HR (95% CI) of

preeclampsia. ... 33 Table 4. Adjusted relationship between urinary trace metals and the percent

change (95% CI) in circulating maternal angiogenic biomarkers. ... 35 Table 5. Demographics of MOTOR-CC participants included in analysis. ... 53 Table 6. Distribution of metals in placenta of women in MOTOR-CC (n = 166)... 54 Table 7. Distribution of metals across demographic characteristics in the weighted

MOTOR-CC. ... 56 Table 8. Top 10 highest and lowest expressed genes measured in the placenta. ... 58 Table 9. Mean difference (95% CI) in gene expression associated with placental

Cd levels. ... 59 Table 10. Mean difference (95% CI) in gene expression associated with placental

Cd levels when co-adjusting for placental essential metals. ... 60 Table 11. Odds ratio (95% CI) of preeclampsia associated with placental gene

expression. ... 60 Table 12. Intracellular Zn and Cd concentrations in HTR-8/SVneo cells treated

with 0 µM, 0.044 µM, 1.0 µM, 10.0 µM CdCl2. ... 74

Table 13. Intracellular Zn and Cd concentrations in HTR-8/SVneo cells under the

ZnCl2 pre-treatment paradigm. ... 79

Table 14. Prior distributions for estimates used in main effects Bayesian regression models examining placental Cd concentrations in relation to gene

expression. ... 94 Table 15. Comparison of estimates from MLE and Bayesian analyses of the

association between placental Cd levels and placental gene expression. ... 95 Table 16. Prior distributions for estimates used in Bayesian regression models

examining placental Cd concentrations in relation to gene expression within strata


Table 17. Comparison of estimates from MLE and Bayesian analyses of the association between placental Cd levels and placental gene expression among



Figure 1. Anatomical diagram of the human placenta. A. cross-section of the

placenta. B. structure of chorionic villi. ...2 Figure 2. Two-stage conceptualization of preeclampsia (adapted from Steegers,

2010)...8 Figure 3. Overarching hypothesis for dissertation work. ... 19 Figure 4. Adjusted HR (95% CI) of preeclampsia with PCs loaded by toxic metals

(PC2) and seafood-associated metals (PC3) within strata of above vs. below

median levels of essential metals (PC1). ... 36 Figure 5. DAG displaying preterm birth as a (A) a mediator or (B) a collider in

the association between trace metals exposure and preeclampsia. ... 42 Figure 6. MT expression after 24-hour treatment with 0.044, 1.0, and 10.0 µM

CdCl2. A. MT1F, B. MT1X, and C. MT2A expression was increased in

HTR-8/SVneo cells treated with 1.0 and 10.0 µM CdCl2. Key: ** p < 0.05 in one-way

ANOVA (N = 4). ... 75 Figure 7. Expression of the TGFB Pathway. A. Heat-Map displaying fold-changes

for TGFB pathway genes after CdCl2 treatment. B. TIMP1 and C. MMP9

expression after CdCl2 treatment. Key ** p < 0.05 in one-way ANOVA (N = 4). ... 76

Figure 8. Trophoblast migration over 24-hours on scratch test. A. Representative figure displaying migration over 24-hours after the scratch for control and 10.0 µM CdCl2 treatment groups. B. Compiled trophoblast migration for CdCl2

treatments relative to control and 10.0 µM CdCl2 treatment groups. B. Compiled

trophoblast migration for CdCl2 treatments relative to controls across all

replicates. Key: ** p < 0.05 in linear regression (N = 5). ... 77 Figure 9. GSH depletion in HTR-8/SVneo cells following 24- or 48-hour

treatment with CdCl2. A. Intracellular levels of GSH following 24-hour CdCl2

treatment, B. GSH levels under a 2-hour ZnCl2 pre-treatment paradigm.

Intracellular levels of GSH following 48-Hour treatment with C. CdCl2 alone, or

D. 2-hour pre-treatment with ZnCl2 followed by CdCl2 treatment. Key: ** p <

0.05 in one-way ANOVA; * p< 0.05 compared to controls & # p < 0.05 compared

to 100 µM ZnCl2 pre-treated cells in two-way ANOVA (N = 3 - 6). ... 78

Figure 10. MT induction in HTR-8/SVneo cells following a 10 µM CdCl2

treatment, with and without a two-hour 100 µM ZnCl2 pre-treatment. A. MT1F,

B. MT1X, and C. MT2A expression is induced in HTR-8/SVneo cells after treatment with 10 µM CdCl2, regardless of ZnCl2 pre-treatment. Key: ** p < 0.05


Figure 11. TGFB Pathway expression in HTR-8/SVneo cells after 10 µM CdCl2

treatment, with and without a two-hour ZnCl2 pre-treatment. A. Heat map

displaying fold changes in TGFB pathway genes shows that patterns of expression are similar regardless of the ZnCl2 pre-treatment. B. TIMP1 and C.

MMP9 expression is not induced in the presence of a 100 µM ZnCl2

pre-treatment. Key: * p < 0.05 compared to controls & # p < 0.05 compared to ZnCl2

pre-treated cells in two-way ANOVA (N = 4). ... 81 Figure 12. Trophoblast migration, measured using a scratch test, in

HTR-8/SVneos treated with 10 µM CdCl2, with or without a 100 µM ZnCl2

pre-treatment. A. Representative graph from one replicate displaying trophoblast migration, under all treatment conditions, over 24-hours. B. Compiled migration across all replicates relative to a control. Key: * p < 0.05 compared to controls & # p < 0.05 compared to 100 µM ZnCl2 pre-treated cells in linear regression (N =

4). ... 82 Figure 13. hCG secretion in the JEG-3 choriocarcinoma cell line following

48-hour treatment with CdCl2. Key: ** p < 0.05 in one-way ANOVA with Dunnett’s

p-value correction. ... 99 Figure 14. Expression of syncytialization markers following CdCl2 treatment in

unstimulated BeWo cells. A. hCGα, B. hCGβ, C. SYN-1, D. SYN-2, and E.


LIST OF ABBREVIATIONS 17BHSD Hydroxysteroid 17-beta dehydrogenase 1

ACOG American College of Obstetricians and Gynecologists ADAM12 ADAM metallopeptidase domain 12

AMH Anti-mullerian hormone

ARNT Aryl-hydrocarbon nuclear translocator ART Assisted reproductive technologies

As Arsenic

Ba Barium

Be Beryllium

BMI Body mass index

BMP1 Bone morphogenic protein 1 BWH Brigham and Women’s Hospital

Cd Cadmium

CDH2 N-cadherin

CHEAR Children’s Health Exposure Analysis Research CI Confidence interval

Cr Chromium

Cr Cycle threshold

Cu Copper

CYP19A1 Cytochrome p450 family 11, subfamily A, member 1 DAG Directed acyclic graph


FDR False discovery rate

GAPDH Glyceraldehyde 3-phosphate dehydrogenase GCM1 Glial Cells Missing Transcription Factor 1 hCG Human chorionic gonadotropin

Hg Mercury

HLA-G Major histocompatibility complex, class 1, G

HR Hazard ratio

ICP-MS Inductively coupled plasma-mass spectrometry IGF Insulin growth factor

IL6 Interleukin 6 IL10 Interleukin 10 IQR Interquartile range LOD Limit of detection LRT Likelihood ratio test MMP Matrix metalloproteinase

Mn Manganese

Mo Molybdenum

MOTOR Maternal Oral Therpay to Reduce Obstetric Risk

MT Metallothionein

Ni Nickel

NIEHS National Institute of Environmental Health Sciences NODAL Nodal growth differentiation factor

NOS2 Nitric oxide synthase 2


Pb Lead

PC Principal component

PCA Principal components analysis PLGF Placental growth factor

PR Prevalence Ratio

PRL Prolactin

Se Selenium

sFlt-1 Soluble fms-like tyrosine kinase-1 SG Specific gravity

Sn Tin

SYN Syncitin

TBP TATA-box binding protein TGFB Transforming growth factor beta TIMP Tissue inhibitor of metalloproteinase

Tl Thallium

U Uranium

VEGF Vascular endothelial growth factor

VEGFR Vascular endothelial growth factor receptor

W Tungsten

YWHAZ Tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein zeta


INTRODUCTION The Placenta and its Formation

The placenta is a transient organ that comprises the maternal-fetal interface. While it is often disposed of after delivery, the placenta plays a crucially important role in maintaining pregnancy [1]. To do so, the placenta carries out a variety of different functions throughout pregnancy. First, the placenta is responsible for the transfer of nutrients, gases, and waste products between mother and fetus [1]. Second, the placenta is a major endocrine organ and is responsible for the production of several hormones integral to the maintenance of pregnancy and fetal development, such as human chorionic gonadotropin and progesterone [2]. Lastly, the placenta serves as a protective barrier against xenobiotics during pregnancy and can act to limit the transfer of harmful compounds to the fetus [3]. Notably, the placenta does not provide a perfect barrier and many harmful environmental contaminants are still transported across the placenta to the fetus [3]. Nevertheless, while often overlooked, the placenta is as an important factor contributing to the health of the offspring [4].


the intervillous space, where they are bathed in maternal blood. The intervillous space is the site of gas, nutrient, and waste exchange occurs [4]. Lastly, the basal decidua is a mixture of maternal uterine and fetally-derived cells [5]. Specifically, cells invade from the developing embryo into the uterine wall to anchor the fetus to the mother and adapt it for pregnancy [3]. In addition, maternal uterine arteries pass through the basal decidua and release blood into the intervillous space [5].

The placenta is derived from the outer layer of the blastocyst, known as the

trophectoderm, and begins formation when implantation occurs [1, 3, 6]. Cells within this layer, known as trophoblasts, are highly proliferative and are the first cells in the embryo to

differentiate [6]. Trophoblasts first differentiate into cytotrophoblasts, which serve as the ‘trophoblast stem cell’ [7]. These cells then differentiate along one of two distinct pathways – a villous or extravillous pathway [6]. When differentiating along the villous pathway, trophoblasts


form large, multi-nucleated syncytiotrophoblasts. Syncytiotrophoblasts are formed and maintained by the continuing fusion of an underlying population of cytotrophoblasts and

comprise the placental barrier [8]. Specifically, they line chorionic villi, separating maternal and fetal blood pools from one another in the intervillous space (Figure 1b) [8]. Given their location, they are integral to the transport of nutrients, waste products, and gases between mother and fetus [8]. In addition, syncytiotrophoblasts are major hormone producers within the placenta and serve as the main source of human chorionic gonadotropin, progesterone, and estrogens, among other hormones involved in placentation, fetal growth, and delivery [2].

On the other hand, a subset of cytotrophoblasts near the anchoring villi proliferate and undergo an epithelial to mesenchymal transition, becoming invasive extravillous trophoblasts [9]. These extravillous trophoblasts invade into the maternal decidua, partly enabled by secreting factors that digest the extracellular matrix, including matrix metalloproteinases, and urokinase-type plasminogen activator, among others [9]. As a result, these cells colonize the decidua and are involved in the lateral growth of the placenta from the chorionic plate, anchoring the organ to the mother’s uterus [6, 9]. In addition, some of extravillous trophoblasts invade the maternal spiral arteries and take on an endovascular phenotype. Endovascular extravillous trophoblasts invade into and are responsible for remodeling maternal spiral arteries from high-resistance/low capacitance to low-resistance/high-capacitance vessels needed to accommodate increased blood flow required later during pregnancy [10]. Notably, during this process of deep invasion and angiogenesis, extravillous trophoblasts invade beyond the boundary of the basal decidua and into the inner third of the maternal endometrium [11].


differentiation and function [11]. One example of such a disease is preeclampsia, which is a major contributor to maternal mortality worldwide [12]. While little may be known about what triggers the onset of preeclampsia, it is characterized by defects in trophoblast invasion and adaptation of maternal spiral arteries [13-19]. Moreover, the only cure for preeclampsia is delivery of the placenta [19]. The difficulty in preventing and treating preeclampsia, and other disorders of placentation, highlights the importance of the placenta during pregnancy and beyond.


Preeclampsia is a hypertensive disorder of pregnancy, classically defined as de novo hypertension (systolic blood pressure >140 mmHg or diastolic blood pressure > 90 mmHg) and proteinuria (> 300 mg/dL in a 24-hour urine collection or a protein/creatinine ratio > 0.3 mg/dL) after the 20th week of gestation [20]. More recently, however, the diagnostic criteria have been expanded to allow for the diagnosis of preeclampsia in the absence of proteinuria, but the presence of other markers of end organ dysfunction (i.e. thrombocytopenia, renal insufficiency, impaired liver function, etc.) [20]. Globally, preeclampsia affects approximately 5% of

pregnancies and may account for up to 15% of maternal deaths in some regions [21, 22]. Within the United States, preeclampsia is estimated to impact approximately 3% of pregnancies [23] and its incidence has increased approximately 25% over the past several decades [23, 24]. This increase may, in part, be driven by changes in the prevalence of risk and/or protective factors (discussed below, in section a), such as smoking, advanced maternal age, and obesity, as well as changes in diagnostic criteria.


preeclampsia as preeclampsia with severe hypertension (systolic blood pressure > 160 mmHg or diastolic > 110 mmHg), or other complications, such as thrombocytopenia, impaired liver function, renal insufficiency, pulmonary edema, new onset headaches, and visual disturbances [20]. Others have defined severe preeclampsia based on the timing of delivery (< 34 weeks gestation) [25]. Preeclampsia has also been sub-classified as early-onset vs. late-onset

preeclampsia, with early-onset preeclampsia being considered higher risk for adverse maternal-fetal outcomes [26, 27]. However, widespread agreement on the gestational week that defines early-onset vs. late-onset preeclampsia is lacking [26, 27]. In addition, other distinct types of preeclampsia exist. These include superimposed preeclampsia (i.e. proteinuria occurring in women with pre-existing chronic hypertension) and postpartum preeclampsia (i.e. de novo hypertension and proteinuria occurring after delivery) [28, 29]. There is also substantial

heterogeneity among these preeclampsia subtypes, suggesting potentially distinct etiologies for various forms of the disease [30-34].

Risk Factors for Preeclampsia: Demographic and Behavioral


adverse reproductive and developmental outcomes, it appears to be protective of preeclampsia [39]. Mechanisms underlying this observation are unknown, but may be related to the presence of carbon monoxide in cigarette smoke, which is believed to promote trophoblast migration, act as a vasodilator, suppress antiangiogenic factors that promote aberrant placentation, or activate the antioxidant system within the placenta [40, 41]. On the other hand, researchers employing counterfactual frameworks have also suggested that the observation may be due to a combination of selection bias, left truncation bias, and/or collider stratification bias in epidemiologic studies. For instance, many studies examining preeclampsia enroll only prevalent cases at birth,

neglecting to consider individuals that were unable to develop preeclampsia because their use of cigarettes also predisposed them to preterm birth and/or miscarriage [42]. Such effects may help explain the paradoxically inverse association between smoking and preeclampsia.


Risk Factors for Preeclampsia: Genetic

There is strong evidence for a genetic component to preeclampsia. Both maternal and fetal factors have been implicated in preeclampsia. Pregnancies in which either the mother or the father were exposed to preeclampsia in utero also have an increased risk of preeclampsia,

suggesting an inherited predisposition to the disorder [53]. Moreover, genotypes of many loci, including those related to trophoblast invasion (e.g. ENG, TGFB1 [54], and MMP1/3 [55]), angiogenesis (e.g. Flt1 [56], VEGF, and VEGFR [57, 58]), immune system/inflammation (e.g. HLA-G [59], and IL10 [60]), among others, have been associated with preeclampsia risk.

Risk Factors for Preeclampsia: Clinical

Many clinical risk factors have also been defined based on preexisting or comorbid conditions. Preeclampsia risk is higher among women with pre-existing diabetes, chronic hypertension, kidney disorders, autoimmune disorders, or polycystic ovarian syndrome, among other conditions [35, 61].

Women who conceive using assisted reproductive technologies (ART) are also at

increased risk of preeclampsia. Within ART, certain methodologies may confer greater risk than others. For instance, pregnancies conceived following egg donation or in-vitro fertilization may be at greater risk than others [62, 63]. In addition, women who are nulliparous, or who have had long intervals between pregnancies are also at higher risk of developing preeclampsia [25, 64].

Aberrant Placentation in Preeclampsia

Preeclampsia is believed to originate in the placenta. Classically, preeclampsia is


as deeply as in normal pregnancy. It is unknown what triggers this impaired placentation, but may originate with poor adaptation of the maternal decidua or impairments in trophoblast invasion and spiral artery adaptation [12]. As a result of this inadequate placentation, the maternal spiral arteries are unable to

accommodate the blood flow required to support the growing fetus [6]. Without adequate blood flow, the placenta may experience reduced or irregular perfusion, ultimately leading to oxidative stress within the organ [6].

During the second stage of preeclampsia, oxidative stress levels within the placenta cause apoptosis and necrosis of syncytiotrophoblasts. Dying syncytiotrophoblasts release their cellular components, including inflammatory molecules and anti-angiogenic factors, into maternal circulation [6]. These factors promote systemic maternal inflammation and endothelial dysfunction, giving rise to the characteristics that are used to diagnose the disorder [6]. Several biological pathways have received significant attention with respect to preeclampsia development. First, biological pathways that control trophoblast invasion have


been implicated in the pathogenesis of preeclampsia, a significant one being the transforming growth factor beta (TGFB) pathway [9, 66]. Specifically, the TGFB pathway inhibits trophoblast invasion by upregulating the expression of tissue inhibitor of metalloproteinases (TIMPs) [67, 68]. TIMPs inhibit trophoblast invasion by binding to matrix metalloproteinases (MMPs), which are enzymes secreted by invasive trophoblasts to digest the surrounding extracellular matrix [66-68]. In preeclampsia, placental expression of the TGFB pathway and TIMPs are higher, and MMPs are lower, indicating reduced invasive capacity of trophoblast cells [66, 69, 70]. Notably, when trophoblast invasion is impaired, invasion into and remodeling of the maternal spiral arteries is also reduced [11]. These changes are consistent with the concept of preeclampsia as being driven by shallow placentation [11].


Ultimately, the placenta’s endocrine function is an important factor in controlling placentation. Hormones, including hCG produced by trophoblasts themselves, control trophoblast differentiation into the extravillous and villous phenotypes in an autocrine and paracrine manner [72]. In addition, hCG produced by the placenta controls trophoblast invasion, angiogenesis, and overall placental growth. Specifically, hCG increases trophoblast invasion by decreasing the expression of TIMPs and increasing the expression of MMPs [73]. In addition, hCG acts as a vasodilator and enhances the production of VEGF to promote angiogenesis [74, 75]. Progesterone, which is also produced by syncytiotrophoblasts, modulates hCG production [76] and inhibits trophoblast invasion by upregulating TIMPs and downregulating MMPs [77]. However, other studies have also demonstrated that progesterone may promote trophoblast migration by increasing insulin growth factor (IGF) signaling [78], indicating potential timing and dose-dependent effects of progesterone on placentation. Other hormones, including estrogen, pregnancy-associated plasma proteins, inhibins and activins also play roles in controlling this tightly regulated process [2]. Thus, given that a major role of the placenta is as an endocrine organ, its hormone production is crucial to controlling placentation and has been implicated in the pathogenesis of preeclampsia [2].

Model Organisms for Preeclampsia: In Vitro Studies

In vitro models for the study of preeclampsia focus primarily on trophoblast cells because of their role in establishing the placenta and carrying out its functions throughout pregnancy. In other words, these cells are responsible for carrying out many of the processes that are

dysregulated during preeclampsia.


the lab [79]. Primary cells are believed to more closely represent placental cells as they behave in vivo. However, primary cells have major limitations [80]. For example, they may dedifferentiate when they lose cell-cell interaction and are removed from their normal milieu [81]. In addition, their proliferation in culture is short and limits their usefulness in toxiciological studies, where toxicant exposure periods may be several days [82].

Immortalized cell lines are most commonly used to study trophoblasts and include both malignant cells that have spontaneously immortalized and normal cells that have been

immortalized via viral transfection. Three of the most common cell lines all originate from the same choriocarcinoma and are known as the BeWo, JEG-3, and Jar cell lines [82]. While these cell lines originate from the same choriocarcinoma, they have features that represent distinct trophoblastic lineages. Currently, BeWos are estimated to be the most widely used placental cell line for the study of preeclampsia [82]. Despite this, non-malignant cell lines are typically preferred to malignant lines, which may have acquired genomic and phenotypic changes that do not reflect normal placental cells. One widely used non-malignant cell line is the HTR-8/SVneo cell line. The cell line is derived from HTR-8 cells, which were isolated from a first trimester placental tissue and transfected with the SVneo virus for immortalization [83]. HTR-8/SVneo cells retain phenotypic characteristics of extravillous trophoblasts [83]. While JEG-3 cells are good models of first trimester cytotrophoblasts with robust hormone production [84], and BeWos of syncytiotrophoblasts [82], HTR-8/SVneos provide a model of invasive extravillous


Model Organisms for Preeclampsia: In Vivo Studies

Given the ethical complications in studying the placenta before birth, placental development cannot easily be studied in humans. Explants and primary cells, as mentioned above, be isolated and cultured from delivered placenta for short periods of time [79]. Yet, the placenta can only be sampled at the time of delivery, limiting our ability to study the organ’s development. To compensate, a considerable amount of effort has also been put into identifying non-invasive biomarkers of preeclampsia risk that can be measured throughout pregnancy. A number of potential biomarkers have been identified, but two have been repeatedly shown to predict the onset of preeclampsia later in pregnancy. Specifically, maternal circulating sFlt-1 and PLGF (and the sFlt-1/PLGF ratio) have been demonstrated to accurately predict preeclampsia in pregnant women [85-87].

For more mechanistic studies of preeclampsia, rodents are the most common in vivo models utilized. For instance, mouse models of preeclampsia have been developed using genetic alterations or pharmacological agents to manipulate angiogenic pathways, oxidative stress conditions, or the maternal immune system to create preeclampsia-like symptoms (i.e.


been proposed. These include guinea pigs, who do spontaneously develop preeclampsia; sheep, which are common models of human reproduction; and non-human primates, several of which do share many similarities with humans [90]. However, the use of these larger organisms in laboratory studies are ethically more complicated, have substantially longer gestational periods, and are more expensive to maintain [90].

Preeclampsia Etiology

While several pathways involved in placental development have been implicated in preeclampsia, the underlying etiology is still largely unknown [12]. Notably, as previously mentioned, the incidence of preeclampsia has increased approximately 25% over the past few decades [23, 24]. Changes in the prevalence of risk factors, such as obesity, have been implicated in this increase. However, these risk ractors are unable to fully account for the observed increase in preeclampsia incidence [23]. Moreover, recent changes to preeclampsia diagnosis have made the disease criteria more stringent and would be expected to produce a decrease in preeclampsia incidence rather than an increase [24]. As a result, there is increasing interest in investigating exposure to environmental contaminants as a potential cause of preeclampsia. Environmental exposures may cause preeclampsia by (1) altering trophoblast differentiation, (2) interfering with trophoblast invasion and/or placental angiogenesis, or (3) producing oxidative stress within the placenta. Despite the interest in environmental contaminants, few have been linked to


Toxic Metals and Preeclampsia

Toxic metals are ubiquitous contaminants with no known role in the human body. They are linked to a variety of reproductive disorders, including miscarriage, intrauterine growth restriction, and preterm birth [92-94]. Toxic metals have been shown to cross the placenta and can be measured in placental tissue, indicating that placental cells are directly exposed to toxic metals during pregnancy [95]. They are also known to promote oxidative stress [96], alter processes of cell migration and invasion [97], and may affect trophoblast differentiation [98]. Therefore, toxic metals may have direct effects on the placenta and may contribute to setting an individual along a path of disease development (Figure 2).

In general, the excretion of toxic metals have been shown to be higher in women

experiencing preeclampsia compared to those with normotensive pregnancies [99]. Beyond this, several toxic metals, such as cadmium (Cd) and lead (Pb), have been linked with preeclampsia [91, 100, 101]. However, there is little evidence for an association between other toxic metals and preeclampsia. For instance, arsenic (As), which has known effects on the placenta, has not been associated with an increased risk of preeclampsia [102]. On the other hand, mercury (Hg) may be associated with preeclampsia, but has only been investigated among women with occupational Hg exposure [103]. Thus, more research is needed to more clearly investigate potential associations between toxic metals exposure and preeclampsia among


Cadmium and Preeclampsia

Early studies on prenatal metals exposure revealed that the placenta is a target organ for Cd. Specifically, while the placenta limits the transfer of Cd to the fetus, it selectively

accumulates within the organ due to highly inducible expression of metallothioneins (MTs) [95]. MTs are low molecular weight proteins that bind and sequester Cd within cells [95]. Therefore, the cells of the placenta may experience exposure to Cd that is up to several orders of magnitude higher than those found elsewhere in maternal or fetal tissues [104]. This means that placental cells, including trophoblasts, may be more susceptible to the effects of Cd exposure during pregnancy than others.


There is also toxicologic evidence to support links between maternal Cd exposure and preeclampsia. In rats,Cd exposure has been successfully used to create models of preeclampsia during pregnancy [112, 113]. Cd-exposed dams develop both hypertension and proteinuria and their placenta appear shallower than those in unexposed dams [98, 112, 113]. Mechanistic studies suggest that changes to placental oxidative stress, glucocorticoid signaling, and immunologic signaling within the uterus may, in part, mediate association between Cd and preeclampsia [112-114]. Studies in placental cells, including both JEG-3 and HTR-8/SVneo cell lines, have also demonstrated that Cd exposure inhibits hormone production [115, 116],

trophoblast migration [117, 118] , and alters the expression of angiogenic pathways [119], which may also promote the development of preeclampsia. Thus, the current toxicologic evidence suggests Cd may affect placentation via multiple pathways. In particular, previous research in our laboratory suggests that Cd may alter patterns of trophoblast invasion by inducing expression of the TGFB pathway [117, 118]. However, it is worth noting, that these studies were conducted in the JEG-3 cell line, which has been previously reported to lack TGFB-induced inhibition of trophoblast invasion [120]. Thus, more work is needed to elucidate the mechanisms by which Cd impairs trophoblast migration or invasion and contributes to shallow placentation and,

ultimately, preeclampsia.

The Potentially Protective Effect of Essential Metals


[121]. For instance, evidence exists to suggest Se, and Zn are protective of preeclampsia development based on meta-analyses of essential metal supplementation clinical trials [122, 123]. However, it is worth noting that essential metals function as antioxidants within a specific range, above which they may become pro-oxidants and contribute to oxidative stress

conditions [96]. The range within which essential metals function as antioxidants within the placenta is currently unknown.

Essential metals also protect against the effects of toxic metals [124]. Many essential metals, such as Zn or Ca, have many chemical similarities to toxic metals. As a result, there are numerous toxicokinetic and toxicodynamic interactions between toxic and essential metals. In fact, many of the effects of toxic metals exposure are derived from their disruption of essential metal homeostasis [124]. This is perhaps best characterized with respect to Cd and Zn. For instance, in perfused human placentae, Cd has been shown to affect Zn storage and inhibit its transport across the placenta [125, 126]. In other tissues, Zn has been shown to protect against the effects of toxic metals. Specifically, Zn can prevent Cd uptake, Cd-induced oxidative stress, and directly competes with Cd-binding throughout the human body [124]. Yet, despite this rich history of toxic-essential metals interactions, there is little research investigating these


Research GapsiIn Toxic Metals and Preeclampsia

There are several major gaps in our current understanding of how toxic metals may contribute to preeclampsia. First, much of the existing literature has focused on either maternal Pb or Cd exposure, with little to no research on other toxic metals [91]. Second, while essential metals are suspected to prevent preeclampsia, research on mixtures of toxic and essential metals is also lacking. Currently, research on chemical mixtures represents a major research objective of the National Institutes of Environmental Health Sciences [127] and is an important, unexplored research avenue in preeclampsia. Importantly, research into chemical mixtures includes

examining both mixtures of toxic exposures and mixtures of toxic and essential exposures, which may act antagonistically to one another. Third, beyond population-level studies on metals

exposure and preeclampsia incidence, few researchers have looked more in depth at the effects of metals on the placenta itself. This includes research using both in vitro models and human cohorts. Thus, little is known about what mechanisms may be driving observed relationships between toxic metals and preeclampsia. In particular, research on how toxic metals may affect placentation are critical to understanding potential links between exposure and preeclampsia, given that the origins of preeclampsia likely lie in early pregnancy during placental formation [12].

Project Approach


toxic metals and prevent adverse effects within the placenta. This project uses an

interdisciplinary approach, spanning from environmental epidemiology to in vitro toxicology, to examine these hypotheses. Moreover, this work is notable because it is the first to: (1) examine a broad range of trace toxic and essential metals, alone and in combination, in association with preeclampsia in a prospective study, (2) estimate the association between Cd accumulation and gene expression within the placenta, and (3) examine the effects of Cd on trophoblast migration, a key process underlying placentation, and test potential interactions between Cd and essential metals within the placenta using an in vitro approach.

This dissertation is organized into three chapters. Chapters 1 and 2 utilize population-based epidemiologic and molecular approaches, while Chapter 3 employs an in vitro toxicological

approach. Specifically, Chapter 1 examines 17 urinary trace metals, including both toxic and essential metals, in association with preeclampsia in a prospective birth cohort using both single-contaminant and mixtures-based approaches. Chapter 2 examines the association between placental Cd concentrations and the expression of gene pathways implicated in placentation in a case-control study of preeclampsia. Lastly, Chapter 3 investigates the effect of Cd exposure on trophoblast migration in the HTR-8/SVneo cell line, reflecting one potential mechanism by which Cd affects placentation. Taken together, this dissertation demonstrates that toxic metals


are associated with biomarkers of preeclampsia risk and further suggests that Cd can dysregulate gene expression and alter trophoblast behavior., providing evidence that Cd can alter



1.1 Background

Preeclampsia is a reproductive disorder that is diagnosed by de novo hypertension and proteinuria after 20 weeks gestation [128]. In the United States, the incidence of preeclampsia has increased over the past several decades and is estimated to affect approximately 5% of all pregnancies [12, 24]. Importantly, the underlying etiology of preeclampsia remains unknown. The origins of preeclampsia are hypothesized to lie within the placenta. Preeclamptic

pregnancies are characterized by a placenta with shallow invasion into the maternal decidua and incomplete remodeling of maternal spiral arteries [12]. As a consequence, the preeclamptic placenta is poorly perfused and hypoxic, giving rise to the systemic maternal endothelial dysfunction that characterizes the disorder [10, 12].

Given that preeclampsia is characterized by impaired spiral artery remodeling, angiogenic biomarkers may serve as predictors of preeclampsia risk [129, 130]. These biomarkers include soluble fms-like tyrosine kinase-1 (sFlt-1) and placental growth factor (PlGF). Both sFlt-1 and PlGF are members of the vascular endothelial growth factor (VEGF) family and play important roles in adapting maternal spiral arteries during placentation. More specifically, higher levels of maternal circulating sFlt-1 indicate anti-angiogenic activity, while lower levels of circulating PlGF reflect reduced placental vascularization [131]. Both biomarkers

1 A variation of this chapter previously appeared as an article in Environmental Health. The original citation is as

follows: Bommarito, P.A., Kim, S.S., Meeker, J.D., Fry, R.C., Cantonwine, D.E., McElrath, T.F., Ferguson, K.K. “Urinary trace metals, maternal circulating angiogenic biomarkers, and preeclampsia: a single-contaminant and


have been shown to predict preeclampsia [132]. In addition, the sFlt-1/PlGF ratio (i.e., a ratio of anti-angiogenic to angiogenic activity) may more accurately predict preeclampsia than either biomarker alone [133].

Prenatal exposure to metals may contribute to preeclampsia risk. Previous studies have reported associations between toxic metals, such as cadmium (Cd) and lead (Pb), and an

increased risk of preeclampsia [99, 109, 110, 134-136]. Alternatively, higher levels of essential metals, such as selenium (Se) and zinc (Zn), have been routinely associated with a decreased risk of preeclampsia [122, 137]. Possible mechanisms for these associations include that toxic metals may impair trophoblast invasiveness [98, 117, 118, 138], generate placental oxidative stress [112], or lead to maternal immunologic abnormalities [113], whereas essential metals have antioxidant properties that may reduce the effect of toxic exposures and promote normal placentation [123]. In addition, there are known toxicokinetic and toxicodynamic interactions between toxic and essential metals [139-141]. At an epidemiologic level, this has been observed with respect to both placental functioning and preeclampsia, suggesting that essential metals may mitigate the relationship between toxic metals and preeclampsia [109, 142, 143].

1.2 Study Objectives

While there is evidence for the role of metals in the development of preeclampsia, many of these studies come from populations with exceptionally high levels of toxic metals exposure [99, 134, 135]. Evidence is more limited in populations with lower exposures to toxic metals. In addition, while studies have examined relationships between individual trace metals and

preeclampsia, very few studies have investigated metals in combination with one another [99]. In the present study, we performed an exploratory analysis of the relationship between 17


in the LIFECODES birth cohort. In addition, we investigated the impact of metals mixtures using principal components analysis (PCA) to examine the relationship between groups of urinary metals, preeclampsia, and levels of circulating angiogenic biomarkers. We selected PCA, rather than other statistical methods for analyzing chemical mixtures as a method of dimension

reduction that would enable us to examine groups of metals that may share important exposure sources or toxicokinetics. Lastly, potential interactions between principal components

characterized by toxic and essential metals were also investigated.

1.3 Materials and Methods 1.3.1 Subject Recruitment

LIFECODES is an ongoing prospective birth cohort that began recruitment in 2006 at Brigham and Women’s Hospital (BWH) in Boston, MA. The study has few exclusion criteria. Specifically, women are eligible for enrollment into the LIFECODES cohort if they are at least 18 years of age, are seeking prenatal care before 15 weeks gestation, and intend on delivery at BWH [144]. At the first visit (median, 10 weeks gestation), women completed detailed

questionnaires of demographic information and medical history and provided informed consent. Gestational age was defined according to the guidelines of the American College of

Obstetricians and Gynecologists (ACOG), with the last menstrual period verified by first trimester ultrasound scanning [145]. Additional questionnaire data and urine samples are

collected at three subsequent study visits (median, 18, 26, and 35 weeks gestation) and stored at -80° C until analysis [146]. All research was approved by the Institutional Review Board at BWH and was deemed exempt by the University of Michigan and the National Institute of


The present analysis used subjects from a case-control study of preterm birth recruited from 2006 – 2008, which is nested in the parent LIFECODES cohort (Appendix 1.1). This case-control study included almost all cases of singleton preterm birth during this period and

unmatched singleton controls were included in a 1:3 ratio [147]. The primary purpose of this case-control study was to examine the relationship between prenatal environmental exposures and preterm birth [148]. Participants in the case-control study were included in this analysis if they had urine samples from the 3rd study visit available for metals analysis (n = 390).

Importantly, demographic characteristics in the subset of women with urine samples at the 3rd study visit is highly similar to both the larger case-control study of preterm birth [148] and the overall LIFECODES cohort [147]. To account for the disproportionate number of preterm births in our analysis, we applied inverse probability weighting based on preterm birth case status to all statistical analyses [149]. Thus, results are generalizable to what would have been observed in the parent LIFECODES birth cohort.

After delivery, preeclampsia diagnosis and diagnosis date were abstracted from medical records. Preeclampsia was defined according to ACOG guidelines between 2006 – 2008: elevated maternal blood pressure (> 140 mmHg systolic and/or > 90 mmHg diastolic) and proteinuria (> 300 mg/24 hours or a protein/creatinine ratio > 0.20) after 20 weeks gestation [128]. All cases of preeclampsia were verified by two maternal-fetal medicine specialists and diagnosis date was used to determine the gestational age of preeclampsia onset. In the case of conflict, a third specialist reviewed the case. Among the 390 participants, 28 women developed preeclampsia. Among these 28 preeclampsia cases, one woman developed preeclampsia

postpartum. For the present analysis, we excluded non-preeclamptics with gestational


against women who did not develop any form of pregnancy-induced hypertension, resulting in a final sample size of 383. Given the exclusion of individuals with pregnancy-induced

hypertension, the contrast in this study is the risk of preeclampsia in the LIFECODES birth cohort among individuals without any pregnancy induced hypertension where trace metals exposure was increased by an interquartile range compared to preeclampsia risk if trace metals exposure remains constant.

1.3.2 Urinary Trace Metals Analysis

Urine samples from the 3rd study visit were sent to NSF International (Ann Arbor, MI, USA) for trace metals analysis, in collaboration with the Children’s Health Exposure Analysis Research (CHEAR) Program. Trace metals were analyzed using inductively coupled plasma-mass spectrometry (ICP-MS), with methods described in further detail elsewhere [146]. The metals analyzed included arsenic (As), barium (Ba), beryllium (Be), Cd, copper (Cu), chromium (Cr), mercury (Hg), manganese (Mn), molybdenum (Mo), nickel (Ni), Pb, Se, tin (Sn), thallium (Tl), uranium (U), tungsten (W), and Zn.

When metal concentrations were measured below the limit of detection (LOD) by ICP-MS, the machine-read values were used in analysis [146]. If the reported value was below zero or blank, the value was replaced with LOD/√2 [150, 151]. Metals with very low detection rates (< 30%) were represented as detect versus non-detect in all subsequent analyses.

1.3.3 Plasma Biomarkers of Angiogenesis

Measurement of circulating maternal sFlt-1 and PlGF were measured in plasma samples collected at the 3rd study visit [129, 152]. Both sFlt-1 and PlGF were measured using


concentrations from 0.10 – 150 ng/mL were measured. The ratio of sFlt-1 to PlGF was also calculated for analysis [133]. The combined intra- and interassay coefficients of variation were < 7% for both PlGF and sFlt-1 [129].

1.3.4 Statistical Analysis

First, differences in demographic and clinical characteristics based on preeclampsia status were examined using t-tests, chi-square tests, or Fisher’s exact tests, where appropriate. Second, we analyzed trace metal distributions by calculating median and interquartile ranges (IQR) by preeclampsia status.

Third, the relationship between urinary trace metals and the hazard ratio (HR) of preeclampsia were estimated using Cox proportional hazard models with normalized sampling weights. Both unadjusted and adjusted Cox models were used to estimate the relationship between an IQR increase in urinary trace metals and the HR (95% Confidence Interval [CI]) of preeclampsia. The proportional hazards assumption was assessed by creating time-dependent variables for each predictor in the model. Tests of these time-dependent variables indicated that the proportional hazards assumption was satisfied. Trace urinary metals were log-transformed in order to improve model fit (i.e. lower AIC). Crude models were created with adjustment for gestational age at sample collection and, for metals that were modeled continuously, urinary specific gravity (SG). Potential confounders for adjusted analyses were identified using a

directed acyclic graph (DAG) and included age (years), body mass index (BMI; kg/m2), maternal


diabetes (yes/no), use of assisted reproductive technology (ART; yes/no), chronic hypertension (yes/no), self-reported use of multivitamins (yes/no), calcium supplements (yes/no) and iron supplements (yes/no) during pregnancy, and infant sex (male/female) (Appendix 1.2). Factors were retained in adjusted analyses if their inclusion influenced effect estimates by > 10%. As a secondary analysis, we recreated the same statistical models excluding BMI as a covariate because of its potential role as a collider [153].

Fourth, the relationships between urinary trace metals and maternal angiogenic biomarkers were assessed using linear regression models. Trace metal species were

transformed to improve model fit (i.e. lower AIC) and maternal angiogenic biomarkers were log-transformed to approximate a normal distribution. To ensure comparability between analyses, crude and adjusted models were constructed using the same confounders as the Cox models. Given that both the exposure and the outcome variables were log-transformed, beta estimates and standard errors from these regression models were converted to percent change (95% CI)

associated with an IQR increase in trace metals concentration for interpretability. Lastly, in addition to using single-contaminant models, urinary trace metals were

investigated as a mixture using PCA. PCA was applied to the set of 13 urinary trace metals with > 30% detection to produce principal components (PCs). Urine metals measures were

log-transformed and adjusted for SG prior to performing PCA to ensure normality and take the effect of hydration on the correlation of urinary metals into account. SG-correction was used to account for urine dilution using the following formula: MetalSG = Metal[(1.015-1)/(SG-1)], where

MetalSG is the specific gravity-corrected metal concentration, Metal is the raw urinary trace


the Scree test was used to determine a meaningful set of components. Additionally, PCs were only retained if at least three trace metals loaded into them with a factor loading > 0.40. Varimax rotation was used to maximize the variance in the factor pattern matrix [154]. While 13 urinary metals were initially included in PCA, Ba and Mo loaded onto multiple components, indicating that they were complex items [154]. Therefore, PCA was re-created excluding these metals. These PCA scores were then fit to (1) unadjusted and adjusted Cox proportional hazard models to estimate the relationship between a 1-unit change in PC score and the HR of preeclampsia, and (2) unadjusted and adjusted linear regression models to estimate the relationship between a 1-unit change in PC score and the percent change in circulating angiogenic biomarkers. Given that PCA revealed components loaded by toxic and essential metals, we examined interactions between groups of toxic and essential metals using nested interaction and main effects models. We additionally explored interactions between the individual metals that loaded onto PCs loaded by toxic and essential metals. Significant interactions were identified using a likelihood ratio test (LRT). Interactions were considered significant if the LRT p-value < 0.10. For these models, the PC loaded by essential metals was dichotomized at the median (< median vs. > median) in order to report the association between toxic metals and preeclampsia among individuals with low vs. high levels of essential metals.

All analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA). Unless otherwise stated, significance was defined as p-value < 0.05.

1.4 Results

1.4.1 Subject Demographics and Urinary Metals


Table 1. Weighted LIFECODES demographic characteristics overall and by preeclampsia status: crude N (weighted %) or weighted median (weighted IQR).

Overall (N=383; ∑weights = 946.7) Preeclamptic (N=28; ∑weights = 53.0)


(N=355; ∑weights = 893.7)

p* Maternal Age (years) 32.7 (29.1, 35.7) 33.0 (29.0, 35.1) 32.7 (29.1, 35.7) 0.62 Pre-pregnancy BMI


23.9 (21.4, 27.9) 29.8 (24.5, 36.7) 23.8 (21.3, 27.4) <0.01

Maternal Race

White 231 (60.3%) 21 (75.0%) 210 (59.2%)


African American 57 (14.9%) 6 (19.4%) 51 (14.8%)

Other 95 (24.8%) 1 (5.6%) 94 (26.1%)

Maternal Education†

<High School 53 (13.6%) 7 (21.5%) 46 (13.1%)


Technical College 57 (14.9%) 3 (9.7%) 54 (15.2%)

Junior College or Some College

115 (31.3%) 11 (40.3%) 104 (30.7%) > College 148 (40.3%) 7 (28.5%) 141 (41.0%) Maternal Health


Private/HMO/Self-Pay 311 (82.7%) 25 (90.3%) 286 (82.2%)


Public 63 (17.3%) 3 (9.7%) 60 (17.8%)

Maternal Smoking During Pregnancy§

No 353 (93.9%) 24 (91.7%) 329 (94.0%)


Yes 25 (6.1%) 4 (8.3%) 21 (6.0%)

Maternal Alcohol Use During Pregnancy‡

No 359 (95.2%) 28 (100.0%) 331 (94.9%)

Yes 15 (4.8%) 0 (0.0%) 15 (5.1%)


Nulliparous 167 (43.8%) 12 (42.4%) 155 (43.9%)

0.88 Parous 216 (56.2%) 16 (57.6%) 200 (56.1%)

Previous Preeclampsia Diagnosis

No 370 (97.3%) 24 (91.7%) 346 (97.7%)


Yes 13 (2.7%) 4 (8.3%) 9 (2.3%)

Gestational Diabetes

No 355 (93.4%) 25 (90.3%) 330 (93.6%)


Yes 28 (6.6%) 3 (9.7%) 25 (6.4%)

Use of ART

No 349 (90.8%) 23 (75.7%) 326 (91.7%)



Chronic Hypertension

No 368 (97.1%) 22 (87.6%) 346 (97.7%)


Yes 15 (2.9%) 6 (12.5%) 9 (2.3%)

Use of Multivitamins During Pregnancy§

No 106 (27.4%) 8 (30.6%) 98 (27.2%)

Yes 272 (72.6%) 20 (69.4%) 252 (72.8%) 0.73

Use of Calcium Supplements During Pregnancy§

No 324 (85.9%) 20 (72.9%) 304 (86.7%)


Yes 54 (14.1%) 8 (27.1%) 46 (13.3%)

Use of Iron Supplement During Pregnancy§

No 336 (88.7%) 24 (91.7%) 312 (88.5%)


Yes 42 (11.3%) 4 (8.3%) 38 (11.5%)

Infant Sex

Female 166 (44.7%) 17 (77.2%) 149 (42.8%)


Male 217 (55.3%) 11 (22.8%) 206 (57.2%)


sFlt-1Expression|| (ng/mL) 5.79 (3.85, 9.12) 6.30 (4.15, 9.88) 5.72 (3.83, 9.06) 0.29

Maternal PLGF

Expression# (pg/mL) 448 (284, 645) 275 (421, 132) 456 (292, 651) <0.01 Maternal sFlt-1/PLGF

Ratio|| 13.8 (8.27, 22.9) 23.2 (12.0, 39.9) 13.3 (8.11, 22.3) 0.03

*p-value for weighted t-test, Chi-square test, or Fisher Exact test, where appropriate. † = 10 missing; ‡ = 9 missing; § = 5 missing, || = 16 missing; # = 15 missing.

Abbreviations: IQR, interquartile range; BMI, body mass index; ART, assisted reproductive technology

Medians (IQRs) of SG-adjusted trace metals in 3rd study visit urine samples by


Table 2. Weighted distribution of specific gravity-adjusted trace metals from 3rd study visit urine samples (µg/L) by preeclampsia case status (N=383).

Metals LOD

N (%) < LOD

Preeclamptic Median (IQR) or N detected (weighted %)

Non-preeclamptic Median (IQR) or N detected (weighted %)

As 0.30 0 (0) 15.9 (6.05, 21.7) 17.9 (9.59, 32.6)

Ba 0.10 4 (1.03) 2.25 (0.93, 4.11) 1.93 (0.98, 3.34)

Cd 0.06 214 (55.9) 0.09 (0.06, 0.14) 0.08 (0.04, 0.14)

Cu 2.50 32 (8.21) 9.62 (8.77, 11.7) 8.96 (6.73, 12.1)

Hg 0.05 32 (8.21) 0.50 (0.24, 0.76) 0.51 (0.27, 0.97)

Mn 0.08 6 (1.54) 0.91 (0.53, 1.18) 0.73 (0.51, 1.13)

Mo 0.30 0 (0) 45.6 (36.2, 69.6) 51.3 (37.1, 68.8)

Ni 0.80 54 (13.8) 3.40 (1.86, 3.97) 2.84 (1.88, 3.97)

Pb 0.10 92 (23.6) 0.34 (0.16, 0.64) 0.35 (0.15, 0.62)

Se 5.00 3 (0.08) 36.3 (31.2, 46.0) 37.0 (29.6, 45.6)

Sn 0.10 24 (6.15) 0.47 (0.28, 0.98) 0.63 (0.35, 1.22)

Tl 0.02 61 (15.6) 0.13 (0.09, 0.16) 0.13 (0.08, 0.18)

Zn 2.00 0 (0) 294 (206, 398) 242 (146, 364)

Be 0.04 356 (91.3) 3 (9.72) 31 (9.61)

Cr 0.40 330 (84.6) 7 (21.5) 50 (13.2)

U 0.01 342 (87.7) 4 (15.3) 43 (11.3)

W 0.20 309 (79.2) 4 (15.3) 73 (19.5)

Shading denotes metals with >70% of samples below the limit of detection.

1.4.2 Urinary Trace Metals and Preeclampsia Risk


Table 3. Adjusted association between urinary metals and the HR (95% CI) of preeclampsia.

Adjusted HR (95% CI) p Single Contaminant Models

As 0.73 (0.46, 1.16) 0.18

Ba 1.05 (0.62, 1.79) 0.85

Cd 0.94 (0.54, 1.64) 0.83

Cu 0.71 (0.23, 2.24) 0.56

Hg 0.90 (0.63, 1.28) 0.55

Mn 1.26 (0.75, 2.12) 0.39

Mo 0.47 (0.21, 1.04) 0.06

Ni 0.89 (0.50, 1.59) 0.69

Pb 0.97 (0.67, 1.40) 0.86

Se 0.28 (0.08, 0.94) 0.04

Sn 0.82 (0.48, 1.38) 0.45

Tl 0.80 (0.47, 1.37) 0.42

Zn 0.94 (0.44, 2.02) 0.88

Be 1.46 (0.32, 6.74) 0.63

Cr 3.48 (1.02, 11.8) 0.05

U 0.99 (0.23, 4.21) 0.99

W 1.77 (0.49, 6.37) 0.38

Principal Components Analysis Models

PC1: Cu, Se, and Zn 0.89 (0.35, 2.29) 0.81

PC2: Cd, Mn, and Pb 1.63 (0.74, 3.61) 0.22

PC3: As, Hg, and Sn 0.75 (0.39, 1.46) 0.40

Shading denotes metals with >70% of samples below the limit of detection.



Table 4. Adjusted relationship between urinary trace metals and the percent change (95% CI) in circulating maternal angiogenic biomarkers.

sFlt-1 PlGF sFlt-1/PlGF Ratio

% Change (95% CI) p % Change (95% CI) p % Change (95% CI) p

Single Contaminant Models

As -0.88 (-7.07, 5.73) 0.79 0.05 (-6.47, 7.03) 1.00 -1.96 (-10.5, 7.39) 0.67 Ba -0.95 (-7.24, 5.78) 0.78 -2.77 (-9.23, 4.15) 0.42 2.06 (-6.99, 12.0) 0.67 Cd -0.45 (-6.72, 6.24) 0.89 -6.99 (-13.1, -0.47) 0.04 6.64 (-2.71, 16.9) 0.17 Cu 11.5 (0.18, 24.1) 0.05 -10.6 (-20.1, -0.001) 0.05 23.7 (6.44, 43.8) <0.01 Hg 4.99 (-0.95, 11.3) 0.10 -1.65 (-7.48, 4.56) 0.59 6.17 (-2.22, 15.3) 0.15 Mn 1.45 (-5.89, 9.36) 0.71 -1.44 (-8.89, 6.62) 0.72 1.88 (-8.37, 13.3) 0.73 Mo 0.69 (-9.49, 12.0) 0.90 4.99 (-6.09, 17.4) 0.39 -5.72 (-18.9, 9.60) 0.44 Ni 2.27 (-6.46, 11.8) 0.62 -0.51 (-9.39, 9.24) 0.92 4.22 (-8.12, 18.2) 0.52 Pb -2.34 (-7.03, 2.58) 0.35 -7.20 (-11.8, -2.33) <0.01 4.80 (-2.24, 12.4) 0.19 Se -4.07 (-19.5, 14.3) 0.64 -18.2 (-31.8, -1.80) 0.03 15.9 (-9.49, 48.3) 0.24 Sn 5.15 (-1.17, 11.9) 0.11 4.98 (-1.61, 12.0) 0.14 0.22 (-8.20, 9.42) 0.96 Tl -4.52 (-11.1, 2.56) 0.20 -4.35 (-11.3, 3.08) 0.24 -0.30 (-9.89, 10.3) 0.95 Zn -2.67 (-11.22, 6.7) 0.56 -5.17 (-13.9, 4.41) 0.28 3.45 (-9.16, 17.8) 0.61 Be -6.72 (-25.2, 16.3) 0.54 15.3 (-8.42, 45.2) 0.23 -17.5 (-39.5, 12.6) 0.23 Cr -9.09 (-25.4, 10.8) 0.34 -24.5 (-38.2, -7.77) <0.01 5.25 (-20.4, 39.1) 0.72 U -17.7 (-33.3, 1.48) 0.07 -9.16 (-27.1, 13.1) 0.39 -11.1 (-33.9, 19.5) 0.44 W 1.09 (-14.0, 18.9) 0.90 -1.62 (-16.9, 16.5) 0.85 4.77 (-16.6, 31.7) 0.69

Principal Components Analysis


1.4.4 Trace Metals Mixtures, Preeclampsia Risk, and Angiogenic Biomarkers

Using PCA, we identified three primary PCs, which explained 46.0% of the variance in urinary metals. PC1 was characterized by higher loading of essential metals (Cu, Se, and Zn); PC2 was characterized by higher loading of toxic metals (As, Mn, and Pb); and PC3 was

characterized by higher loading of seafood-associated metals (As, Hg, and Sn) [156-158]. These factor loading patterns have been previously observed in our examination of urinary trace metals and preterm birth in this study population [146]. Factor loadings and communalities for urinary metals are displayed in

Appendix 1.5.

With respect to preeclampsia, neither PC1 (Cu, Se, and Zn) (HR: 0.89, 95% CI: 0.35, 2.29), PC2 (Cd, Mn, and Pb) (HR: 1.63, 95% CI: 0.74, 3.61), nor PC3 (As, Hg, and Sn) (HR: 0.75, 95% CI: 0.39, 1.46) were significantly associated with preeclampsia (Table 3). Interactions between PCs loaded by



and PCs loaded by essential metals (PC1) were further investigated (Figure 4). While there was no statistically significant interaction between PC1 (Cu, Se, and Zn) and PC2 (Cd, Mn, and Pb) (pLRT = 0.12), the association between PC2 (Cd, Mn, and Pb) and preeclampsia (HR: 3.54, 95%

CI: 1.09, 11.5) was significant among individuals with low levels of PC1 (Cu, Se, and Zn). On the other hand, this association was null among individuals with higher levels of PC1 (Cu, Se, and Zn). Similar trends were observed among the individual trace metals loaded that loaded onto PC1 and PC2 (Appendix 1.6).

After adjusting for potential confounders, PC1 (Cu, Se, and Zn) (% Change: -13.5%, 95% CI: -22.6, -3.41) and PC2 (Cd, Mn, and Pb) (% Change: -11.9, 95% CI: -20.7, -2.12) were

associated with lower circulating PlGF levels (Table 4). The associations between PCs and angiogenic biomarkers were similar in both crude and adjusted models with BMI excluded (Appendix 1.4).

1.5 Discussion


with preeclampsia risk, but both PC1 (Cu, Se, and Zn) and PC2 (Cd, Mn, and Pb) were associated with lower levels of circulating PlGF. In addition, while we did not observe a

statistically significant interaction between PC1 (Cu, Se, and Zn) and PC2 (Cd, Mn, and Pb), the association between PC2 (Cd, Mn, and Pb) and preeclampsia was significant among individuals with low levels of PC1 (Cu, Se, and Zn).

In single-contaminant models, detection of Cr in the urine was associated with an increased risk of preeclampsia and a reduction in circulating PlGF levels, suggesting that Cr is associated with a reduction in placentation and placental function. While the effect estimates were relatively imprecise, other studies have also noted higher levels of total urinary or hair Cr in preeclampsia cases compared to normotensive controls [99, 136]. Urinary Cr levels in this cohort were low, with most samples being below the LOD, but the primary source of exposure to Cr in the general population is from food, particularly from meat and fish products [159]. Recent toxicologic studies have also demonstrated that hexavalent Cr exposure leads to increased expression of oxidative stress markers and apoptotic signaling in trophoblast cells and mouse placenta [160-162], providing further molecular-level evidence for a potential link between Cr exposure and preeclampsia. Despite the consistency of association between urinary Cr,

angiogenic biomarkers, and preeclampsia, we did not feel that it was appropriate to apply mediation analysis methodologies to this cohort given (1) the small number of samples with Cr above the LOD, (2) the relatively high degree of imprecision in the estimates, and (3) the lack of clear connection between Cr exposure and preeclampsia in current epidemiologic and

toxicologic literature. Nevertheless, we feel that these results warrant further research. Although we did not find associations between other toxic metals and the risk of


levels. These results may provide additional mechanistic evidence for previous studies that have reported associations between prenatal Cd and Pb exposure and preeclampsia [101, 109, 110]. Both Cd and Pb are ubiquitous environmental exposures. Both may be found in contaminated air, soil, or water. Cd levels tend to be higher in certain foods, such as rice and grains, and both Cd and Pb can be found in drinking water systems or housing, particularly in areas with old or failing infrastructure [163-165]. Similarly, the PC loaded by toxic metals (PC2: Cd, Pb, and Mn) was not associated with preeclampsia risk, but was associated with lower circulating PlGF levels.


uptake of essential metals in attempt to buffer the higher levels oxidative stress associated with the disease [168].

In addition to PC1 (Cu, Se, and Zn) and PC2 (Cd, Mn, and Pb), we identified a third PC that was characterized by loading from seafood-associated metals (PC3: As, Hg, and Sn). Neither PC3 (As, Hg, and Sn), nor the individual metals that comprise it, were associated with

preeclampsia incidence or levels of circulating angiogenic biomarkers. Notably, while inorganic As is considered a toxic metal, non-toxic organic forms of As, such as arsenobetaine, are widely found in seafood [169]. In populations with frequent seafood consumption, including those within the United States, much of the total urinary As may be organic arsenicals from such foods [170-172]. In addition, although seafood-associated metals (PC3: As, Hg, and Sn) grouped independently of toxic metals (PC2: Cd, Mn, and Pb), they should not be considered non-toxic. For instance, pregnant women are advised against consuming certain types of seafood on the basis that it may contain high levels of methylmercury [173]. However, disentangling the harmful effects of these trace metals from the beneficial effects of seafood consumption (i.e. fatty acids) is complex and beyond the scope of this analysis.





Related subjects :