Maternal Dietary Factors and Whole-Exome Sequencing of Primary Congenital Glaucoma and Anterior Segment Defects of the Eye

MATERNAL DIETARY FACTORS AND WHOLE-EXOME SEQUENCING OF PRIMARY CONGENITAL GLAUCOMA AND ANTERIOR SEGMENT DEFECTS OF THE EYE

Kristin Jean Voltzke

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 Epidemiology.

Chapel Hill 2019

Approved by:

Andrew F. Olshan

Suzan L. Carmichael

Tania A. Desrosiers

Kari E. North

Robert E. Meyer

ABSTRACT

Kristin Jean Voltzke: Maternal Dietary Factors and Whole-Exome Sequencing of Primary Congenital Glaucoma and Anterior Segment Defects of the Eye (Under the direction of Andrew F. Olshan)

There are few studies on the risk factors of primary congenital glaucoma (PCG) and anterior segment defects (ASDs) of the eye in a US-based population. The goal of this dissertation was to evaluate maternal diet, a modifiable risk factor, and to explore genes which have not previously been evaluated within PCG cases. Each of these exposures were evaluated within the National Birth Defects Prevention Study (NBDPS), a large population-based case- control study in the United States. Specific Aim 1 evaluated maternal diet through intake levels of individual micronutrients, a diet quality score, and the creation and evaluation of dietary patterns. Inverse, albeit imprecise, associations were seen for a number of individual micronutrients. Diet quality did not appear to be associated with having a child with PCG or

ASDs. Mothers who adhered to the Prudent and Mexican dietary patterns appeared to have lower odds of having a child with these defects. In Specific Aim 2, samples from a subset of PCG cases from the NBDPS underwent whole-exome sequencing. The goal of this aim was to identify rare mutations in genes which previously have not been investigated in relation to PCG. We found a similar proportion (13.5%) of families with variants in CYP1B1, the most commonly mutated gene in PCG. In addition, we found a number of variants present in individual families which occurred in genes that are known to underly other Mendelian conditions, some with phenotypic

iii overlap, or genes which are known to play a role in eye development signaling pathways. These genes, and specific variants, should be further explored through in vitro, in vivo, and other epidemiologic studies to better understand the genetic mechanisms in PCG cases. The findings of this dissertation add to the sparse literature on risk factors for PCG and ASDs.

I would like to dedicate this work to my family for their constant encouragement and to my amazing fiancé, Patrick Moore, for your continuous love and support.

ACKNOWLEDGMENTS

First and foremost, I would like to express my appreciation to the women and families who participated in the National Birth Defects Prevention Study and the study staff at each center without whom this work would not be possible. I would also like to thank the members of the North Carolina Center of the National Birth Defects Prevention Study, the University of

Washington Center for Mendelian Genomics, and Mary Jenkins and Lynn Almli at the CDC for their helpful feedback and support of this work. Finally, I would like to express my most sincere gratitude to my dissertation committee for their guidance and encouragement throughout the dissertation process.

TABLE OF CONTENTS

LIST OF TABLES ...... ix

LIST OF FIGURES ...... x

LIST OF ABBREVIATIONS ...... xi

CHAPTER 1 – INTRODUCTION AND SPECIFIC AIMS ...... 1

CHAPTER 2. - BACKGROUND ...... 5

2.1 Development of the eye and associated structures ...... 5

2.2 Congenital Glaucoma ...... 8

2.3 Risk Factors of primary and congenital glaucoma and anterior chamber defects ...... 15

2.4 Maternal diet ...... 21

2.5 Section 2 summary ...... 23

CHAPTER 3. – STUDY DESIGN AND METHODS ...... 24

3.1 Study overview ...... 24

3.2 National Birth Defects Prevention Study ...... 24

3.3 Methodology for Specific Aim 1 ...... 27

3.4 Methodology for Specific Aim 2 ...... 36

CHAPTER 4. – MATERNAL DIET AS A RISK FACTOR FOR PRIMARY CONGENITAL GLAUCOMA AND DEFECTS OF THE ANTERIOR SEGMENT OF THE EYE ...... 42

4.1 Introduction ...... 42

4.2 Materials and Methods ...... 44

vii

4.3 Results ...... 47

4.4. Discussion ...... 54

CHAPTER 5. – EXOME SEQUENCING TO IDENTIFY NOVEL GENES UNDERLYING PRIMARY CONGENITAL GLAUCOMA IN THE NATIONAL BIRTH DEFECTS PREVENTION STUDY...... 59

5.1 Introduction ...... 59

5.2 Materials and Methods ...... 60

5.3 Results ...... 62

5.4. Discussion ...... 68

CHAPTER 6. – DISCUSSION AND CONCLUSIONS ...... 76

6.1 Summary of Specific Aims ...... 76

6.2 Strengths and Limitations ...... 78

6.3 Public Health Implications and Future Studies ...... 80

6.4 Conclusions ...... 81

APPENDIX 1. SUPPLEMENTARY TABLES FOR CHAPTER 4 ...... 83

Supplemental Table 1. ORs and 95% CIs for individual nutrients for combined primary congenital glaucoma and anterior segment defects, stratified by periconceptional vitamin use ...... 83

Supplemental Table 2. ORs and 95% CIs for individual nutrients, stratified by case group status ...... 87

REFERENCES ...... 91

viii

LIST OF TABLES

Table 2.1. Diagnostic Criteria for types of congenital glaucoma ...... 8

Table 3.1. Response rates by PCG and ACDs by center and total...... 26

Table 3.2. Variables and planned initial coding schemes for confounders in Aim 1...... 34

Table 3.3. List of micronutrients evaluated for associations with PCG and ASDs...... 35

Table 3.4 Filters used in determination of candidate variants using different models of inheritance...... 39

Table 4.1. Demographic and clinical characteristics of primary congenital glaucoma and anterior segment defect cases and controls, National Birth Defects Prevention Study, 2000-2011...... 50

Table 4.2. ORs and 95% CIs for individual nutrients for combined primary congenital glaucoma and anterior segment defects, National Birth Defects Prevention Study, 2000-2011...... 52

Table 4.3 Association between dietary pattern class and having a child with primary congenital glaucoma or anterior segment defects, National Birth Defects Prevention Study, 2000-2011 ...... 54

Table 5.1 Comparison of demographic characteristics of WES samples to all PCG cases within NBDPS...... 65

Table 5.2 Pathogenic CYP1B1 variants in 5 primary congenital glaucoma families in the National Birth Defects Prevention Study...... 66

Table 5.3 Identification of novel genes in primary congenital glaucoma families in the National Birth Defects Prevention Study...... 67

LIST OF FIGURES

Figure 2.1. Developmental stages of fetal eye development...... 6

Figure 2.2. Diagram of normal aqueous fluid flow and disrupted aqueous fluid flow ...... 7

Figure 2.3. Complexity of the trabecular meshwork...... 7

Figure 2.4. Childhood Glaucoma Research Network Classification System of childhood glaucoma...... 10

Figure 3.1. Study centers participating in the National Birth Defects Prevention Study...... 25

Figure 3.2. Directed Acyclic Graph for maternal diet and congenital glaucoma...... 33

LIST OF ABBREVIATIONS

AR Arkansas

ASDs Anterior Segment Defects

BIC Bayesian information criteria

CA California

CATI Computer Assisted Telephone Interview

CDC-MACDP Centers for Disease Control-Metropolitan Atlanta Congenital Defects Program

CGRN Childhood Glaucoma Research Network

CI Confidence Interval

CYP1B1 Cytochrome P450 family 1 subfamily B member 1

DAG Directed Acyclic Graph

DQI-P Diet Quality Index for Pregnancy

EFA Exploratory Factor Analysis

FFQ Food Frequency Questionnaire

FOCX1 Forkhead Box C1

GPATCH3 G-patch Domain Containing 3

HWE Hardy-Weinberg Equilibrium

IA Iowa

IOP Intraocular Pressure

LCA Latent Class Analysis

LRT Likelihood Ratio Test

LTBP2 Latent Transforming Growth Factor -Binding Protein 2

MA Massachusetts

MAF Minor Allele Frequency

MYOC Myocilin

NBDPS National Birth Defects Prevention Study

NC North Carolina

NHGRI National Human Genome Research Institute

NISC NIH Intramural Sequencing Center

NJ New Jersey

NY New York

OR Odds Ratio

PAX6 Paired Box 6

PCA Principal Components Analysis

PCG Primary Congenital Glaucoma

PITX2 Paired Like Homeodomain 2

ROS Reactive Oxygen Species

SES Socioeconomic Status

SNP Single Nucleotide Polymorphism

SNV Single Nucleotide Variant

SSRI Selective Serotonin Reuptake Inhibitor

TEK TEK Receptor Tyrosine Kinase

TIGR Trabecular Meshwork Inducible Glucocorticoid Response Protein

TX Texas

USDA United States Department of Agriculture

xii

CHAPTER 1 – INTRODUCTION AND SPECIFIC AIMS

Congenital glaucoma encompasses both primary congenital glaucoma (PCG) and related defects of the anterior segment (ASDs) of the eye. By definition, PCG affects children younger than four with approximately 80% of cases being diagnosed before a child’s first birthday1. In the

United States the prevalence rate of PCG is approximately 1 in 10,000 live births per year2,3, but is higher in cultures where inbreeding (consanguinity) is common3. Children present with raised intraocular pressure (IOP), loss of corneal transparency, photophobia, and buphthalmos

(enlargement of the eyeball)2. PCG results in blindness in approximately 10% of cases and reduced vision in approximately half of all cases4. ASDs are a collection of developmental disorders which affect structures in anterior segment of the eye which include the iris, cornea, and trabecular meshwork5. Due to structural changes in the eye, children with some ASDs are at an increased risk of developing glaucoma. Prevalence rates of individual ASDs are even more rare than PCG with rates ranging from 1 in 64,000 (aniridia) to 1 in 200,000 (Axenfeld-Rieger Syndrome)6.

Much of the work on risk factors for congenital glaucoma has focused on genetics and molecular factors, however the etiology remains only partially understood. Linkage studies of large, multigenerational families were used to identify areas of interest within the genome. The two genes identified include Cytochrome P450 Family 1 Subfamily B Member 1 (CYP1B1)7 and

Latent transforming growth factor -binding protein 2 (LTBP2)8. Mutations in the CYP1B1 gene are thought to be the most common cause of congenital glaucoma worldwide7. There is biological plausibility for this because CYP1B1 plays an important role in elimination of reactive oxidative species (ROS), as well as the modulation and development of the trabecular meshwork, an area of 1

tissue responsible for draining the aqueous humor from the base of the eye. Defects in CYP1B1 result in difficulties in achieving intraocular pressure control, a common risk factor for glaucoma.

Mutations in this gene have been reported in 90-100% cases of Slovakian Roma Gypsies, 44% of cases from India, and 50% of cases from Brazil9,10. Mutations in LTBP2, the second locus of interest identified through linkage studies, were identified in cases from Pakistani families and cases of Gypsy descent8,11. Despite strong evidence pointing towards these two genes being causal, the generalizability of these findings remains uncertain. For example, mutations in CYP1B1 are only present in 14.6% of affected individuals from the United States10, a more ethnically diverse population than the original families studied from Pakistani and of Gypsy descent.

Additionally, PCG is generally thought to be transmitted as an autosomal-recessive trait with variable penetrance4-8, however, this model has been challenged with reports of transmission in successive generations, unequal sex distribution among affected individuals, and lower-than- expected number of affected siblings in familial cases3,6-7,9-11. Reduced penetrance has been reported in certain families, suggesting existence of a modifier, or modifiers, located elsewhere in the genome12,14.

Moreover, very little work has been done on modifiable risk factors of congenital glaucoma. Diet has been previously implicated in abnormal development of the eye. Several nutrients such as folic acid, vitamin D, and omega-3 fatty acids have been shown to produce fetal abnormalities, including malformations of the eye, neural tube, orofacial clefts, and heart12–17.

Other nutrients like zinc, vitamin E and vitamin A have been identified for their contribution to eye development18–20. Additionally, animal studies have shown that mice with CYP1B1 mutations show similar phenotypes to children with primary congenital glaucoma including irregular collagen distribution in the trabecular meshwork, increased intraocular pressure, and increased

oxidative stress21. The animal studies showed that the structural abnormalities in the trabecular meshwork could be prevented with the administration of the free radical scavenger N- acetylcysteine21. This suggests that other molecules which act as free radical scavengers or have anti-oxidant capacities may have the ability to modify PCG risk.

Despite the focus on specific genetic loci as a risk factor for PCG and ASDs, there is still a general lack of epidemiologic data on these defects in a non-selected (non-familial) group.

Therefore, the National Birth Defects Prevention Study (NBDPS) provides an excellent population-based study sample to explore potential genetic and non-genetic risk factors.

The purpose of this study is to address the gaps in the literature on the etiology of PGC and related ASDs through the following aims:

Aim 1: Evaluate maternal nutrition as a risk factor for PCG and ASDs. A comprehensive approach will be taken to determine whether maternal nutrition is a risk factor for PCG and ASDs.

To begin, intake of individual micronutrients will be compared between cases and controls to determine whether higher intake is associated with reduced risk. Second, the quality of maternal diets will be calculated with Diet Quality Index for Pregnancy (DQI-P) scores based on self- reported diet in the NBDPS telephone interview22. These scores will be split into quartiles based on distribution in controls to determine whether those with poor diets are at an increased risk for giving birth to children with these birth defects. Finally, latent class analysis (LCA)23 will be used to classify women into dietary patterns to determine whether specific patterns are associated with an increased risk of giving birth to children with these birth defects.

Aim 2: Comprehensively investigate genetic variation in whole-exome sequencing data of a large cohort of nationally-representative cases of PCG in the United States. Much is known about causal genetic variants of PCG in populations where consanguineous relationships are

common, however studies in the United States have failed to find these same causal variants suggesting that alternate variants or mechanisms likely underlie cases from the United States. We plan to utilize the strengths of whole-exome sequencing data of family trios to describe rare, pathogenic mutations in genes which previously have not been investigated for PCG focusing on de novo, compound heterozygote, and autosomal recessive inheritance patterns.

CHAPTER 2. - BACKGROUND

2.1 Development of the eye and associated structures

2.1.1 Early embryology

The human eye is derived from various embryonic tissues including the neural ectoderm, surface ectoderm, and mesenchymal cells predominantly arising from neural crest cells. The eyes first appear in the fourth week as a pair of lateral grooves which evaginate out from the neural groove24 (Figure 2.1, A). These outpocketings grow out towards the surface ectoderm until only a few mesenchymal cells separate them from the surface ectoderm24. The surface ectoderm near the optic vesicles thickens to form the lens placode (Figure 2.1, A). Almost as soon as it appears, it invaginates until it pinches off and becomes a hollow lens vesicle24 (Figure 2.1, B). Around day

32, the outermost layer of cells of the optic vesicles invaginate to form a goblet-shaped optic cup24

(Figure 2.1, B). In the 6th week of development mesenchymal cells begin to migrate to the space between the anterior epithelium of the lens vesicle and the surface ectoderm.2 These cells then condense into a dense layer which later gives rise to the corneal endothelium. In the seventh week the mesenchyme surrounding the lens splits into two layers to encloses a new cavity which becomes the anterior chamber of the eye2,24. The thick posterior wall of the anterior chamber breaks down to create the posterior chamber and eventually forms an opening called the pupil24 (Figure

2.1, E). By the end of the third month the cells in the choroid tissue differentiate into the iris24. As the cells of the iris elongate, flatten, and separate from each other, space opens allowing

extracellular fibers to form the trabecular meshwork near the edge of the cornea (Figure 2.2 and

2.3).

Figure 2.1. Developmental stages of fetal eye development.

2.1.2 Aqueous humor flow

The aqueous humor is a clear fluid which flows from the posterior chamber to the anterior chamber and out of the eye through the trabecular meshwork and into the canal of Schlemm (see

Figure 2.2)25,26. The eye is avascular, so adequate flow is necessary to supply nutrients and remove metabolic waste26. The amount of aqueous humor produced by the ciliary body in the eye must be equal to the amount of aqueous humor leaving the eye to maintain normal intraocular pressure

(IOP)26.

Studies in both primates and humans have shown that increased IOP is closely linked to the death of retinal ganglion cells27,28 which are the cells that take information from the visual field to the brain. An exact mechanism for this death is not yet elucidated, however possible candidates include strain on the extracellular components surrounding these cells28 or through signaling pathways29. The death of these cells is thought to be the primary risk factor for blindness in

glaucoma patients30. In fact, the only proven methods to treat glaucoma to date focus on reducing

IOP31.

Figure 2.2. Diagram of (2A) normal aqueous fluid flow and (2B) disrupted aqueous fluid flow found in glaucoma. Borrowed from Karmel M. Glaucoma Pipeline Drugs: Targeting the Trabecular Meshwork. EyeNet Magazine. 2013. https://www.aao.org/eyenet/article/glaucoma- pipeline-drugs-targeting-trabecular-meshw.

Figure 2.3. Complexity of the trabecular meshwork. Each color represents a different layer of the trabecular meshwork and the black arrow denotes the aqueous fluid flow. Adapted from Karmel M. Glaucoma Pipeline Drugs: Targeting the Trabecular Meshwork. EyeNet Magazine. 2013. https://www.aao.org/eyenet/article/glaucoma-pipeline-drugs-targeting-trabecular-meshw.

2.2 Congenital Glaucoma

Primary congenital glaucoma (PCG) cases are diagnosed early in childhood and develop without overt structural eye defects6. In the past, glaucoma that developed secondary to recognizable ocular or systemic findings were called “developmental glaucoma”, however new terminology is moving towards more unambiguous categories that can be seen in Table 2.1. The flow-chart in Figure 2.1 shows the decision tree physicians use to classify glaucoma in children.

In this study, we plan to focus on cases of PCG and cases specific to the anterior segment including

Peters anomaly, Axenfeld-Rieger anomaly, aniridia, and lens coloboma. These defects fit into the category of glaucoma associated with non-acquired ocular anomalies in Table 2.1. Because we are only including four of the conditions which fit into this category, we have decided to refer to them as anterior segment defects instead of the collective category name throughout the rest of this report.

Table 2.1. Diagnostic Criteria for types of congenital glaucoma included in this study according to the Childhood Glaucoma Research Network (CGRN).

Congenital Glaucoma Diagnostic Criteria Primary Congenital Glaucoma (PCG)  Isolated angle anomalies (+/- mild congenital iris anomalies)  Meets glaucoma definition (usually with ocular enlargement)  Subcategories based on age of onset (1) Neonatal or newborn onset (0-1 month) (2) Infantile onset (>1-24 months) (3) Late onset or late-recognized (>2 years)  Spontaneously arrested cases with normal IOP but typical signs of PCG may be classified as PCG.

Glaucoma Associated with Non-Acquired  Includes conditions of predominantly Ocular Anomalies ocular anomalies present at birth which may or may not be associated with systemic signs  Meets glaucoma definition  List common ocular anomalies: Axenfeld Rieger anomaly (Syndrome if systemic associations) Peters anomaly (Syndrome if systemic associations) Ectropion uveae Congenital iris hypoplasia Aniridia Persistent fetal vasculature/PFV (if glaucoma present before cataract surgery) Oculodermal melanocytosis (Nevus of Ota) Posterior polymorphous dystrophy Microphthalmos Microcornea Ectopia lentis Simple ectopia lentis (no systemic associations, possible Type 1 FBN mutation) Ectopia lentis et pupillae

Figure 2.4. Childhood Glaucoma Research Network Classification System of childhood glaucoma.

2.2.1 Primary congenital glaucoma descriptive epidemiology

2.2.1.1 Clinical characteristics

Primary congenital glaucoma can be classified into three subtypes: neonatal or newborn, infantile, or late-onset. Neonatal or newborn cases of PCG are diagnosed within the first month of life. Infantile cases are diagnosed from the first month to two years of age and late-onset are those cases diagnosed after two32. Primary congenital glaucoma typically presents within the first 12

months of age2. Approximately 25% of infants with PCG and related defects are diagnosed at birth, and more than 80% are diagnosed within the first year of life1. These defects are usually bilateral

(65-80%).

Children with PCG present with raised IOP, loss of corneal transparency, photophobia, and buphthalmos2. It results in blindness in approximately 10% of cases and reduced vision in approximately half of all cases2. Children are treated with IOP-lowering medications and angle surgery33. Management of the disease is more complicated than in adults due to differences in anatomy which are compounded by additional anatomic variations due to congenital anomalies, a more aggressive healing response, and a less predictable postoperative course33.

2.2.1.2 Prevalence and mortality in the United States

Primary congenital glaucoma occurs in approximately 1 in 10,000 live births in Western countries, with increased rates being seen in populations with higher rates of consanguinity34. By definition, PCG affects children younger than four with approximately 80% of cases being diagnosed before a child turns one1. For a general ophthalmologist practicing in a non-specialist center in North America, this corresponds to diagnosing a new case once every 5 years35. There appears to be no racial or geographic predilection36. In the United States, boys are more commonly affected than girls at a ratio of 3:22,37.

Congenital glaucoma can lead to irreversible vision loss and is not thought of as a fatal disease27. Instead, the goal of both medical and surgical management is often to prevent further deterioration36. Children with these defects are generally identified early in life and benefit from advances in new instrumentation to measure intraocular pressure and visual field defects38. In addition to surgery, children will often be treated with medical therapies which are usually well tolerated and rarely lead to serious complications39. 11

2.2.1.3 Prevalence and mortality outside of the United States

There is minimal data on the incidence of primary congenital glaucoma in the developing world. Reports from sub-Saharan regions of Africa indicate that congenital glaucoma may account for 0.7-4% of all new glaucoma cases35,40–42. Primary congenital glaucoma is far more frequent in certain ethnic and religious groups where consanguineous relationships are common. In Saudi

Arabia the prevalence is reported to be 1 in 2,500 live births and is even higher (1 in 1,200) in

Gypsies in Slovakia43.

The limited resources and lack of experienced medical personnel in the developing world also have impacts on the care and management of affected individuals. In the United States, affected children are often diagnosed within the first few months of life, but in remote places like rural Ethiopia, one research team found the average age of diagnosis to be over 3 years of age35.

Additionally, management of primary congenital glaucoma has significantly lower success rates in these countries when compared to those that are more developed 35. There are no long term follow-up studies of primary congenital glaucoma in the developing world35, but the life expectancy of a child with primary congenital glaucoma is likely lower due to the fact that blindness is closely linked to survival38.

2.2.2 Anterior segment dysgenesis

Anterior chamber dysgenesis refers to a collection of developmental disorders which affect structures in the anterior segment of the eye which include the iris, cornea, and trabecular meshwork5. These disorders can refer to isolated ocular anomalies or a part of systemic defects which are characterized by both autosomal dominant and autosomal recessive patterns of inheritance and often show variable expressivity and/or incomplete penetrance44.

Glaucoma that occurs as a result of the structural defects is common, although not always present, in individuals with anterior segment defects. This project includes some individuals who have been diagnosed with anterior segment defects either with, or without, glaucoma. This next section will define the types of anterior segment defects that are included in our study sample.

2.2.2.1 Peters anomaly

Peters anomaly is a heterogeneous set of disorders and is associated with a range of ocular abnormalities and varying amounts of corneal opacification5. There is very little published information about the incidence of Peters anomaly in the literature. A study published in 2014 by

Kurilec and Zaidman extrapolated data from New York to estimate an incidence of 1.1-1.5 per

100,000 births in the United States45. Both autosomal dominant and autosomal recessive inheritance patterns have been described46,47. Similar to Axenfeld-Rieger syndrome, which will be described in the next section, approximately 50% of patients with Peters anomaly will develop glaucoma5.

A number of mutations have been linked to Peters anomaly including in Paired box 6

(PAX6)46, Forkhead box C1 (FOXC1)48,49, Paired like homeodomain 2 (PITX2)49, and

CYP1B150,51.

2.2.2.2 Axenfeld-Rieger Syndrome

Axenfeld-Rieger Syndrome is a heterogeneous group of disorders characterized by ocular and systemic features6. It is an extremely rare disease with a prevalence of approximately

1:200,000 and is inherited in an autosomal dominant pattern with high penetrance and variable expressivity52. A number of structural ocular anomalies, all which can restrict fluid drainage from the eye, have been identified including strands or bands of the iris, high iris insertion, and

microscopic abnormalities of the trabecular meshwork6. Consequently, as many as 50% of

Axenfeld-Rieger Syndrome patients develop elevated intraocular pressure and glaucoma53,54.

The CYP1B1 gene has been linked to Axenfeld-Rieger syndrome51, but linkage analyses have also identified two other genes: PITX2 and FOXC155.

2.2.2.3 Aniridia

Aniridia translates to absence of the iris, although all patients affected by this disorder have at least some iris tissue6. The prevalence of this ocular disease is between 1/64,000 to 1/96,000.

Both autosomal dominant and sporadic occurrences have been identified. It is thought that the secondary glaucoma in these patients results from the abnormal irises collapsing against the cornea and obstructing outflow of the aqueous humor resulting in increased intraocular pressure6.

Phenotypes similar to aniridia have been found in animal models where mutations occur in the common ancestor of the PAX6 gene56,57, a gene that is highly conserved and central to ocular development58. Sporadic cases are thought to be caused by a large deletion of chromosome 11p13 which covers the PAX6 gene and occasionally also includes a deletion of the neighboring gene6.

2.2.2.4 Lens coloboma

Colobomas are missing pieces of tissue resulting from defects in closure of fetal tissue59.

They can occur in the lens, iris, ciliary body, choroid, and optic nerve60. The size and shape of colobomas in the lens can vary, but is characterized by notching of the equator of the lens61.

Occasionally lens coloboma can appear in combination with other ocular defects59,61, but may also be manifestations of systemic diseases62,63. The literature on lens coloboma is extremely sparse with no reports on incidence or prevalence estimates; instead, it is made up mostly of case reports60,61.

2.3 Risk Factors of primary and congenital glaucoma and anterior chamber defects

2.3.1. Genetic exposures

Much of the research on primary congenital glaucoma and anterior segment defects have focused on the genetics of the defects. Primary congenital glaucoma is familial in 10-40% of cases39. PCG is most often seen in families where consanguinity is present. In these families, cases are most often seen among siblings and in some families. High rates of concordance have been seen among monozygotic twins and discordant among dizygotic twins64–68. Primary congenital glaucoma is generally inherited as an autosomal recessive trait with variable expression and penetrance that ranges from 40-100%69, meaning that not everyone with the mutation presents with disease or presents in the same way. However, this model of inheritance has been challenged with reports of transmission in successive generations, unequal sex distribution among affected individuals, and lower-than-expected number of affected siblings in familial cases64,67,68,70–72.

Primary congenital glaucoma has been mapped to three loci within the genome: GLC3A73,

GLC3B74, and GLC3C75. In 1995, Sarfarazi et al mapped GLC3A to a 2.5 centimorgan (cM) location on the 21st position on the p-arm of chromosome 2 (labeled: 2p21)73. The causative gene in this region was later found to be Cytochrome P450 Family 1 Subfamily B Member 1, or

CYP1B1. The second primary congenital glaucoma site, GLCA3B, was mapped to a 3 cM location at chromosomal location 1p3674. This location within the genome has been implicated in malignant diseases and shown to frequently participate in recombination events74. The final primary congenital glaucoma site, GLC3C, was mapped to chromosomal location 14q24.3 within a 2.9 cM space75.

2.3.1.1 Cytochrome P450 Family 1 Subfamily B Member 1 (CYP1B1)

CYP1B1 is a gene that encodes for a member of the cytochrome P450 family of enzymes which are known for their metabolic abilities76. In particular, the CYP1B1 protein is known to function in the metabolism of potentially toxic foreign compounds but is also known to metabolize endogenous compounds like retinoids. In normal eye development, CYP1B1 protein is expressed in the ciliary body, the non-pigmented epithelium, the iris, and the trabecular meshwork77. In fact,

CYP1B1 null mice exhibit abnormalities in the trabecular meshwork and ocular drainage structures78.

In the Human Gene Mutation Database79, an online database of published germline mutations that underlie or are closely associated with human inherited disease, there are a total of

148 mutations in CY1B1 which have been linked to primary congenital glaucoma80. Mutations in

CYP1B1 are inherited in an autosomal recessive pattern and include missense and nonsense mutations, duplications, insertions, and deletions81. This gene is most consistently identified mutated gene in PCG cases, however frequencies of specific types or patterns of mutations within

CYP1B1 vary among ethnic groups81. For example, the current largest study of sequence variants in a cohort of nonconsanguineous PCG cases (n=47 cases) from the United States found that

CYP1B1 was the main causative gene in the US population, but mutations in this gene were less common than in Arabic or Romany populations10. The authors claimed that this finding suggested that other genes may be accountable for PCG in a US population and that more studies were needed10. The sequencing data from this study will be used as the replication set for our findings.

Mutations in CYP1B1 have also been discovered to be associated with other phenotypes such as aniridia82, Peters anomaly51, Axenfeld-Rieger Syndrome83, and other glaucomas84,85.

2.3.1.2 Latent transforming growth factor  binding protein 2 (LTBP2)

LTBP2 is an extracellular matrix protein that interacts with fibrillin containing microfibrils11,86,87. It is expressed in the trabecular meshwork and ciliary zonules in the eyes11,87.

The LTBP2 gene is located near the GLC3C locus on the q-arm of chromosome 14 (14q24.3)69.

In 2009, two papers were published identifying LTBP2 as the disease-causing gene in

Iranian11, Pakistani and Gypsy families8. The current largest study of sequence variants in a cohort of nonconsanguineous PCG cases from the United States included 47 families, but did not identify any disease-causing mutations in this gene10. Interestingly, case-control studies of over 300 Han

Chinese, 46 Turkish and 94 British individuals also showed no deleterious mutations in LTBP2 in

CYP1B1-negative patients88,89. Currently there are six reported mutations listed as linked to primary congenital glaucoma in the Human Gene Mutation Database90.

2.3.1.3 Myocilin (MYOC)

The MYOC gene has previously been associated with juvenile and primary open-angle glaucoma91,92. It codes for a protein which was originally named the trabecular meshwork inducible glucocorticoid response protein (TIGR)69, however the name was later changed to myocilin (MYOC) by Kubota et al in 199793. Mutations in MYOC have been linked to the reduced secretions of myocilin and the formation of insoluble aggregates in cultured trabecular meshwork cells94, however the exact causal mechanism behind how these mutations cause glaucoma is yet to be elucidated. Disease causing MYOC mutations have been identified in families in both the presence and absence of CYP1B1 mutations95. There is also evidence of interactions between

CYP1B1 and MYOC in primary congenital glaucoma96. A case of PCG was found to be a double heterozygote for CYP1B1 and MYOC loci, which suggests a digenic mode of inheritance95, while another study has suggested that CYP1B1 may be able to modify MYOC expression97. 17

Mutations in MYOC are more rare than other mutations in primary congenital glaucoma, with only three being reported in the Human Gene Mutation Database98. There were no disease- causing causing mutations within the cohort of 47 primary congenital glaucoma cases in the US10 and only two in a molecular study of primary congenital glaucoma in South Korea99.

2.3.1.4 Paired-like homeodomain transcription factor 2 (PITX2)

PITX2 encodes a transcription factor important in embryonic development and tissue morphogenesis6. Mouse models have shown that PITX2 is required for normal ocular development100,101 Additionally, mice that only have one functional copy of PITX2 show the same ocular malformations of Axenfeld-Rieger syndrome and associated glaucoma102. Mutations in

PITX2 have also been associated with Peters anomaly54.

PITX2 is also known to interact directly with FOXC1, another transcription factor, and act as a negative regulator. This means that mutations in PITX2 could lead to resulting phenotypes, at least in part, by affecting function or expression of FOXC1103.

2.3.1.5 Forkhead box C1 (FOXC1)

The forkhead box family of transcription factors are important regulators of embryogenesis, cell migration, differentiation, and fate determination104. FOXC1 is expressed in the trabecular meshwork during development and remains present in the adult iris105. It is also expressed in the heart which is a tissue that is affected in Axenfeld Rieger syndrome106. The mechanisms by which FOXC1 mutations cause Axenfeld-Rieger syndrome are yet to be elucidated, but may relate to haploinsufficiency which is when an individual only has a one functioning copy of a gene which can’t produce enough gene product6. FOXC1 mutations have also been linked to aniridia107 and Peters anomaly108. Underscoring the overlapping etiology of

anterior chamber defects, there are reports of multigenerational families in which multiple members share a single base pair change but members are diagnosed with Axenfeld Rieger syndrome, Peters anomaly, or both48,108.

The functional feature of FOXC1 is a conserved 110-amino acid forkhead domain, which allows the protein to bind to DNA and mediates protein-protein interactions6. Many of the mutations found in Axenfeld-Rieger syndrome cases are located in this region of the gene108,109, however some are found in other regions of the gene109.

2.3.1 Non-genetic exposures

Very little is known about non-genetic risk factors for primary congenital glaucoma and anterior chamber defects. Many of the studies come out of countries with high rates of consanguinity. One case-control study of 52 PCG cases and 38,151 population controls from

Hungary investigated the association between maternal demographic and socioeconomic status

(SES) and primary congenital glaucoma110. They found a higher association with disease among unmarried (OR=10.1, 95% CI= 5.5-18.6) and unemployed mothers (OR=13.6, 95% CI= 5.7-

32.5)110. Mothers of high employment status showed an inverse association with disease (OR=0.2,

95% CI= 0.1-0.5) whereas those with a low employment status were more likely to have children born with the defect (OR=12.2, 95% CI= 6.1-24.6)110. They also found increased associations among certain maternal occupations (agriculture, stay at home with children) and inverse associations among others (industry, trade, other). Hungarian gypsies make up the lower SES class in this country, possibly explaining the large estimates of association found in this study. The same group published a study two years earlier that showed that compared to population controls, PCG cases had a shorter gestational age (p=0.0005), were more likely to be a preterm birth (OR=4.0,

95% CI= 2.2-7.3), and a low birthweight (OR=6.7, 95% CI= 3.7-12.3)111. In addition, cases of 19

PCG were more likely to be a higher birth order number compared to population controls

(p=0.03)111. This group did not find any significant associations between acute infectious diseases, but did find that mothers of PCG cases were more likely (OR=10.8, 95% CI= 2.6-45.0) to have thyroid diseases111. The use of pregnancy supplements was also investigated in this study, however limited use by case mothers mean that any interpretations must be made cautiously. Inverse associations with PCG were found for mothers who used iron (OR=0.2, 95% CI=0.1-0.4), and folic acid (OR=0.3, 95% CI= 0.2-0.6), vitamin D (OR=0.4, 95% CI= 0.2-0.8) supplements111.

There were also suggestions of inverse associations with calcium, vitamin E, and multivitamin supplement use111. The authors concede that many of the findings from this study are consistent with the Gypsy demographics of several of the mothers. Although it is not reported how many of the 52 cases are of Gypsy descent, the authors believe that at least 28 of them are of Gypsy origin and only 6 are confirmed to be non-Gypsy111.

A recent analysis of NBDPS data by Forestieri et al112 is one of the first studies to examine a wide array of potential risk factors for PCG in a population-based sample. Consistent with previous reports, PCG cases were more likely to be male and have lower birth weight compared to controls. In this population, smoking and regular alcohol use were found to be inversely associated with PCG, differing from the Hungarian study110,111 which had found an increased risk.

Other findings that differed from previous reports were null findings for opioid use, antibacterial medication use during pregnancy, and higher birth order. Novel findings included a higher risk among those who used antihypertensive medications or NSAIDs during pregnancy, and Non-

Hispanic Black mothers.

2.4 Maternal diet

2.4.1 Overview

A woman’s diet before and during pregnancy are important because she will be providing nutrition for both herself and the developing fetus. It also represents an important modifiable behavior which has been shown to have direct and indirect effects on the child. For example, in mice, high fat and high sugar diets, characteristics typical of a Western-style diet, have been shown to change metabolic patterns in the mother which has consequences for maternal nutrient allocation for the fetus113. At the fetal level, maternal nutrition has been shown to have direct effects on gene expression114,115, gene methylation116,117, and eye development118.

2.4.2 Measures of diet

Traditionally diet research has focused around the associations between disease and specific nutrients, however focusing only on individual nutrients will ignore the complex combinations of nutrients that have interactive or synergistic effects within the body. There is also a high intercorrelation of some nutrients, and thus lack of ability to untangle independent effects when included in the same statistical model, which provides a methodologic limitation of this type of single-nutrient analysis119. For those reasons, much of the current literature focuses on more holistic approaches to measures of diet.

Dietary pattern indexes measure maternal intake compared to some type of scale of diet quality. Types of diet quality scores include: the Diet Quality Index for Pregnancy (DQI-P)120, the

Healthy Eating Index121, the Alternative Health Eating Index122, the Recommended Food Score123, and the Mediterranean Diet Score22. Broadly, diets are broken out into components and individuals

receive a score for each component. These component scores are then added together for a total score.

Another holistic approach to measuring diet is to create dietary patterns through data- driven approaches. These approaches do not rely on prior criteria which diets are measured against, instead they allow the data to determine variables or groups that best represent diets which emerge within a study population. These types of models include: factor analysis, principal components analysis, mixture modeling, and latent class analysis.

2.4.3 Maternal Diet and Other Birth Defects

There are no published studies on the association between maternal diet and primary congenital glaucoma or anterior chamber defects, however there are a number of studies which have investigated the association of maternal diet and other birth defects.

Within the NBDPS, a higher diet quality has been shown to be associated with a reduced risk of anencephaly124, cleft lip124, cleft palate124, and some conotruncal and septal heart defects125.

A trend toward a reduced risk was also seen with increasing DQI-P scores in both microtia126 and spina bifida124. A better diet quality was also found to be associated with a reduced risk of gastroschisis among Hispanic women.22

Data-driven dietary patterns have also been investigated within the NBDPS. In Sotres-

Alvarez et al. four dietary patterns were discovered and were named: Prudent, Western, Low-

Calorie Western, and Mexican23. Among mothers who did not use vitamin/mineral supplements, mothers in the Western, the Low-Calorie Western, and the Mexican dietary pattern classes were all significantly more likely to have a child with a neural tube defect compared to mothers in the

Prudent class23. They were also more likely to have septal and some conotruncal heart defects compared to mothers in the Prudent class23. 22

2.5 Section 2 summary

Very little is known about non-genetic risk factors for PCG and ACDs. The majority of populations that have been used to study these are not generalizable to the United States because they are in ethnically homogenous populations where consanguinity is high with potentially different exposure patterns9,43,73. Another comprehensive investigation on non-genetic risk factors has been done within NBDPS, a population-based case-control study in the United States. Maternal diet represents an interesting modifiable risk factor that has yet to be studied in relation to PCG and ASDs. This is despite the fact that maternal nutrition has been shown to have direct effects on gene expression114,115, gene methylation116,117, and eye development118. At the population level, maternal diet quality has been shown to be inversely associated with some birth defects such as anencephaly124, cleft lip124, cleft palate124, and some conotruncal and septal heart defects125.

Additionally, certain diet patterns were found to be associated with increases in neural tube defects, septal and some conotruncal heart defects23.

Overall, the genetics of PCG and ASDs have provided some etiologic clues. However, work in more ethnically heterogeneous populations, like those in the United States, have failed to find the same disease-causing variants. While mutations in CY1B1 are found in populations from the United States, they are found at a much lower frequency than populations of Slovakian Roma

Gypsies, Pakistani, Indian, and Brazilian descent9,10. These findings have suggested that alternative mechanisms or genes underly PCG cases in the United States. Our sequencing aim will focus on novel genes which have not previously been explored in relation to PCG.

CHAPTER 3. – STUDY DESIGN AND METHODS

3.1 Study overview

Broadly, this goal of this dissertation is to investigate potential risk factors for primary congenital glaucoma and anterior segment defects of the eye using data from the National Birth

Defects Prevention Study (NBDPS). For Specific Aim 1, maternal nutrition data collected during maternal interviews will be used to investigate the relationship between maternal diet and PCG and ASDs. In Specific Aim 2, data from whole-exome sequencing will be used to describe the genetic landscape of of a subsample of PCG cases in the NBDPS.

3.2 National Birth Defects Prevention Study

This study uses data from the National Birth Defects Prevention Study which is a population-based case-control study conducted in 10 states across the United States of America

(Figure 3.1)127. Infants included in NBDPS had an estimated delivery date between October 1,

1997 and December 31, 2011. Active population-based surveillance programs were employed in each of the participating states to identify cases. Cases were considered eligible if they had one or more of the over 30 major birth defects identified by this study, however any infants that were recognized or strongly suspected to be related to single-gene defects or chromosomal abnormalities were excluded127, which is done because the purpose of NBDPS is to investigate unknown causes of birth defects. Cases are made up of livebirths and stillbirths greater than 20 weeks gestation or at least 500 grams and, where available, elective terminations of prenatally- diagnosed defects at any gestational age127. Controls are live-births who were selected at random from vital records (IA, MA, NC, NJ, UT, CDC-MACDP) or hospital birth records (AR, CA, NY, 24

TX)127. Controls were not matched to cases; however, they were selected from the same geographical region and time periods as cases. Participation rates varied from center to center and over time, but overall 67% of cases and 65% of controls who were invited to participate chose to.127 Additional details on the study design have been published extensively.127 The response rates for PCG and ASDs can be found in Table 3.1.

Figure 3.1. Study centers participating in the National Birth Defects Prevention Study128.

Table 3.1. Response rates by PCG and ASDs by center and total.

Center Response Rate

Arkansas 68%

California 63%

Georgia 70%

Iowa 51%

Massachusetts 60%

New Jersey 100%

New York 78%

North Carolina 71%

Texas 73%

Utah 75%

Totals 67%

Upon enrollment, mothers were asked to complete a detailed computer assisted telephone interview (CATI). Sections covered in the interview include demographics, pregnancy history, maternal conditions and illnesses, family history, lifestyle and behavioral factors, medication use, multivitamin use before and during pregnancy, environmental exposures, occupational history, physical activity, and maternal nutrition. Biological specimens were also collected in the form of buccal cells for DNA extraction and genetic analysis127. In 2017 the CDC led an effort for whole- exome sequencing of a subset of these samples.

Cases were initially clinically diagnosed by a health care provider not affiliated with the study, and were subsequently ascertained and abstracted by birth defects registry staff at each study

site. More extensive birth defect classifications were done for cases corresponding to mothers who chose to participate in the NBDPS. Cases were classified as: isolated, multiple, or complex. If the child had a single major defect with or without any number of minor defects which weren’t recognized as a syndrome, they were considered an isolated case127. When more than one major defect was found that were not known to be related, subjects were considered to have multiple defects127. Complex cases were defined as “a pattern of major defects that are embryologically related and likely represent an early problem in morphogenesis, often akin to a developmental field defect”127. More detailed guidelines can be found in Rasmussen et al.129

The NBDPS study was approved by the institutional review boards of the University of

North Carolina at Chapel Hill, the Centers for Disease Control and Prevention, and the collaborating study centers.

3.3 Methodology for Specific Aim 1

3.3.1 Study population

For this aim, the study population includes all cases of primary congenital glaucoma and anterior segment defects enrolled in the NBDPS between January 1, 2000, to December 31, 2011.

Mothers with a reported total energy intake of < 500 kcals or >5000 kcals were excluded. In addition, plural births, those resulting from donor egg or sperm, those with a gestational age of <

15 weeks, or those cases with co-occurring anopthalmia or micropthalmia were excluded.

3.3.2 Exposure assessment

3.3.2.1 Exposure assessment overview

Maternal interviews were conducted via telephone using a standardized computer-based interview. Topics covered were mentioned above and include, briefly, demographic, occupational,

environmental, behavioral, and dietary data. Dietary data was measured through the use of a shortened Willet food frequency questionnaire (FFQ)23. Interviews were administered between 6 weeks past the estimated date of delivery of the infant and no later than 2 years after this date.

The primary exposure of interest in Specific Aim 1 is maternal diet. It will be evaluated in the following three ways using data collected as a part of the FFQ.

3.3.2.2 Individual micronutrients

The intakes of individual micronutrients were calculated based on answers to the FFQ.

Mothers were asked how frequently they consumed each of the 58 foods during the year prior to pregnancy. In total, there were 16 possible response categories for frequency ranging from never or less than once a month to 6 or more times per day. Intakes in grams per day were then calculated by multiplying the consumption frequency to a standard portion size which was listed on the questionnaire for each of the items. The estimates of each nutrient were determined based on the

US Department of Agriculture National Nutrient Database, version 20130. Results will be stratified by periconceptional vitamin supplementation.

3.3.2.3 Diet Quality Index for Pregnancy (DQI-P)

There are limited valid biochemical indices to measure whether an individual is getting enough of certain nutrients in their diet. One such measure is the use of hemoglobin concentration as a marker for iron status. When these indices are available, they lack the ability to assess more than one nutrient at a time. Composite score measures of diet quality are able to capture the complexity of diet by combining intakes of food and nutrient components reported in dietary assessment tools like food frequency questionnaires and comparing each participant’s diet to standard guidelines like the Dietary Guidelines for Americans and the Food Guide Pyramid. Two 28

of the most common diet quality composite scores are Diet Quality Index and the Healthy Eating

Index, however neither of these measures incorporate the changes in dietary recommendations when a woman is pregnant. In 2002 Bodnar and Siega-Riz adapted the Diet Quality Index to incorporate these changes and developed the Diet Quality Index for Pregnancy (DQI-P)120.

Adaptations include dropping the meal pattern component but adding a sweets component. In addition, it scores each component based on quartiles rather than absolute values. The DQI-P is based on 8 dietary components which include grains, vegetables, fruits, folate, iron, calcium, percentage of calories from fat120. Each component has a maximum of 10 possible points and is weighted equally.

Since it was developed, the DQI-P has been used in multiple racial and ethnic populations including African Americans120, Caucasians120,131, and Native Americans131. One study showed that among low-income pregnant women, Native Americans had significantly lower DQI-P scores compared to Caucasians (51.8 vs. 54.2), but overall all women in the population were not meeting dietary recommendations131. Others have shown that individuals who live more than 4 miles from a supermarket132 and those who are classified as obese had an increased odds (OR=2.16, 95% CI=

1.2-4.0; OR=1.87, 95% CI= 1.37-2.55, respectively) of being in the lowest compared to the highest tertile of DQI-P score133. As mentioned before, the DQI-P has been used in the NBDPS where it was seen that higher DQI-P scores were associated with lower risk of neural tube defects and orofacial clefts124, as well as gastroschisis among Hispanics22.

In this study, we will include the adapatation of the DQI-P that has been used in recent

NBDPS studies by Carmichael et al124 and Feldkamp et al22.

The component scores that make up the DQI-P are each calculated individually. For the folate, calcium, and iron components, NBDPS researchers have previously calculated nutrient

values using the United States Department of Agriculture (USDA) nutrient database (version 19) for all nutrients except choline, for which the USDA version 20 was used because it was considered more complete134,135. For the grains, vegetables, and fruits components, food items were mapped from the FFQ items to the specific food groups in the DQI-P as outlined in Carmichael et al124. To calculate servings per day for grains, vegetables, and fruits, the following formula was used:

 (servings/day x grams/serving for each food item in the group)

(mean grams/serving of all foods in the component)

To calculate servings per day for sweets, the following formula was used:

 (servings/day x grams/serving for each non-soda sweet food item)

(mean grams/serving of all foods in the component) + servings/day of soda

To account for variability in proportion sizes of different food items within each component, the equations above divided the sums of grams per day by the mean grams per serving sizes of all foods within that component.

For grains, vegetables, fruits, folate, iron, and calcium, participants in the lowest decile of received a score of 0, those in the second lowest decile received a 1, up to participants in the highest decile who received a score of 9. Sugar intake will be scored in the opposite manner where individuals in the lowest decile will be given scores of 9 and those in the highest decile will be given a score of 0. Percentage of calories from fat has four categories:  30%, >30% to  35%,

>35% to 40%, and >40% which correspond to scores of 9, 7, 4, and 0 points, respectively. Total

DQI-P scores will then be calculated by summing the component scores. Mothers will then be divided into quartiles and compared to reference, or the lowest quartile of DQI-P scores.

3.3.2.4 Latent Class Analysis (LCA)

Identifying dietary patterns offers additional advantages over individual nutrients and diet quality scores. These patterns allow for interactions between single nutrients while not measuring against prevailing hypotheses or guidance from current dietary recommendations.

Data-driven approaches that have been used to derive dietary patterns include principal components analysis (PCA)136, exploratory factor analysis (EFA)137, and cluster analysis138. These methods are considered data reduction methods and do not consider dietary patterns as unobserved random variables (latent variables)139. In PCA and EFA, food items are grouped together according to their correlations to each other and individuals have a score for each dietary pattern. Latent variable models offer many advantages when used in dietary pattern analyses including the ability to account for correlated errors, the ability to adjust for covariates, and multi-group analyses140.

Latent class analysis (LCA)141, a type of latent variable model, was used in this study.

Broadly, in LCA individuals are assumed to belong to one of a set number of mutually exclusive classes for which membership is unknown. The latent model is built from a multinomial logistic model and takes into account associations between input variables and determines latent classes, or groups, which can explain these associations. In our study, the latent classes are the dietary patterns and the input variables are intake of nutrients reported by the mothers in the FFQ. Each participant is then given a predicted probability of belonging to each group based on levels of their input variables. This method then allows for an estimation of the risk of an outcome for each group compared to a referent group139.

In this study, we used the same approach as Sotres-Alvarez et al23 where dietary patterns were derived from 64 food items relative to total consumption and adjusted for energy intake from control data only. Relative consumption was categorized into four levels: no consumption and tertiles of nonzero consumption. For foods there were consumed by less than 20% of the women, binary variables will be created: consumed or not consumed. Ubiquitously consumed foods did not have a nonconsumption category and instead have quartiles of nonzero consumption. Dietary classes were interpreted and named based on conditional food intake probabilities. We began by specifying four classes (dietary patterns) to look for similarities with previous literature in the

NBDPS, but determined that this was the best choice based on the Bayesian information criteria

(BIC).

Two sets of parameters were estimated: regression coefficients predicting class membership and conditional probabilities of the observed indicators given the class. These two sets of probabilities were then used via Bayes’ theorem to predict each mother’s probability of belonging to each class. Mothers were then classified for the class for which they have the highest probability of belonging to.

3.3.3 Covariate selection

We used findings from Foresteiri et al, along with an extensive literature review on maternal nutrition and congenital birth defects22,23,124,125 to develop a directed acyclic graph

(DAG; Figure 3.2). In this analysis, Table 3.1 outlines the variables and coding schemes for the adjustment set that will be used for this analysis. The variable categories were chosen based on power limitations of this study and functional form analyses.

Figure 3.2. Directed Acyclic Graph for maternal diet and congenital glaucoma.

Table 3.2. Variables and planned initial coding schemes for confounders in Aim 1.

Variable Coding Scheme Variable Range Maternal Age Continuous 14 - 49 Maternal Race/Ethnicity Categorical Non-Hispanic White Non-Hispanic Black Hispanic Other Periconceptional vitamin Categorical Yes supplementation* No Maternal BMI Categorical <18.5 18.5-<25 25-<30 ≥30 Total kcal Intake Continuous 503 - 4986 Study Center Categorical Arkansas California Iowa Massachusetts New Jersey New York Texas CDC/Atlanta North Carolina Utah * beginning three months prior to conception through three months after conception

3.3.4 Modeling strategy

3.3.4.1 Individual nutrients

Functional form analyses will be done to evaluate continuous, quadratic, spline, and categorical forms of each of the 16 micronutrients listed in Table 3.4. Simple logistic regression will be used to calculate odds ratios (ORs) and 95% confidence intervals (CIs). We will calculate

ORs and 95% CIs for each micronutrient in two models: 1) adjusting for total energy intake only, and 2) adjusting for total energy intake and additional confounders.

Table 3.3. List of micronutrients evaluated for associations with PCG and ASDs.

Betaine (mg) Magnesium (mg)

Calcium (mg) Retinol (µg)

Alpha Carotene (µg) Riboflavin (mg)

Beta Carotene (µg) Vitamin A, RAE (µg)

Choline, total Vitamin B12 (µg)

Copper (mg) Vitamin B6 (µg)

Folate, DFE (µg) Vitamin E (mg)

Iron (mg) Zinc (mg)

3.3.4.2 DQI-P

A DQI-P score will be calculated for each mother based on the sum of their individual component scores as described above. The distribution of the DQI-P score among controls will be used to determine quartiles. We plan to calculate ORs and 95% CIs using simple logistic regression for each of the quartiles using the lowest quartile as the reference.

3.3.4.3 LCA

Each mother will be assigned to a dietary pattern for which they have the highest probability of belonging to. This nominal predictor will be as the exposure variable in a simple logistic regression model.

3.3.5 Power Calculations

Minimal detectable odds ratios (ORs) were calculated for all primary congenital glaucoma and related anterior chamber defects together (~175 cases) and separately for primary congenital glaucoma cases individually (~105 cases). These estimates do not contain the full number of 35

interviewed cases, and instead take into account that some cases may need to be excluded due to incomplete data on demographic information. All calculations were performed in EpiSheet

We plan to divide our data into quartiles of each individual nutrient mentioned above according to the distribution in the control pattern. We also plan to use this approach for the DQI as was done in previous NBDPS studies22,124. In the most conservative of the planned analyses, we will have approximately 107 cases of primary congenital glaucoma without anterior chamber defects. With this analysis plan, we should have adequate power (=80%) to detect ORs of approximately 0.5 or 1.80 when exposure frequencies are 25% (quartiles) among mothers in the control group. For the dietary patterns analysis, the prevalence of the patterns has been estimated within the NBDPS control population in a previous study. There were four dietary patterns identified using latent class analysis including the Prudent, the Western, the Low-Calorie Western, and the Mexican dietary patterns which ranged from 15% (Mexican) to 32% (Prudent Based on the prevalences of dietary patterns and this conservative case number, we would have power

(=80%) to detect minimum ORs of 1.75-1.9623.

3.4 Methodology for Specific Aim 2

3.4.1 Study population

The study population for Specific Aim 2 will include complete family trios (mother/father/baby) from the NBDPS from whom buccal (cheek) samples were collected and children were diagnosed with PCG. For this dissertation, we will not look at whole-exome sequencing of ASD trios due to differences in inheritance patterns. Any PCG cases that were conceived using donor egg, sperm, or embryo were excluded and all family members had to have at least 200 ng of DNA available to be eligible for sequencing. In total, this includes 38 complete trios. 36

3.4.2 Exposure Assessment

3.4.2.1 Buccal cell collection & DNA isolation

One of the aims of the NBDPS was to “better understand the role that genetic and epigenetic factors play in the etiology of birth defects”127. After completion of the interview, families were mailed a buccal cell collection kit which included documents for informed consent,

6 cytobrushes (2 per family member), instructions for use, and a postage-paid return mailer. In

2002, two study centers with low participation rate began adding a $20 incentive for completion of the buccal cell collection kits. By 2008 this was implemented in all centers127. Collection kits were modified in mid-2003 to replace the cytobrush packaging after concerns about DNA degradation during transit127.

Prior to 2003, DNA was extracted at each center using different methods127, however after

2003 a centralized laboratory was established to improve consistency and standardize extraction and quality control methods. For each participant, one cytobrush remained at the local center142.

At the centralized laboratory DNA was extracted from one of the cytobrushes using either a phenol-chloroform method or Gentra Purgene (Qiagen). Amounts of DNA isolated was assessed by quantitative real-time PCR using the human RNaseP gene and STR markers were used to assess DNA quality. Quality control standards included 0.1 ng/uL DNA, successful PCR amplification after at least two attempts with at least one STR marker, and alleles consistent with the reported family relationship127.

3.4.2.2 Whole-exome sequencing and alignment

Specimens from case trios that had at least 200 ng of genomic DNA were sent to Dr. Larry

Brody’s lab at National Human Genome Research Institute (NHGRI). Specimen concentrations

were measured by Qubit and DNA quality was checked by running a small portion on an agarose gel. Samples were prepared for sequencing at the NIH Intramural Sequencing Center (NISC) using

100 ng of genomic DNA and the Accel-NGS 2S Plus DNA Library Kit. Additional modifications included modified PCR cycling conditions to improve recovery of high GC regions and swapping a polymerase at the post-capture library amplification step.

The Nimblegen SeqCap EZ Exome + UTR Library (Version 3.0) was used for whole- exome sequencing. Briefly, this kit shears DNA, targeted fragments are captured by probe hybridization, which are then amplified. Samples were then pooled and sequenced on an Illumina

HiSeq 2500 instrument. Image analysis and base calling was performed using Illumina Genome

Analyzer Pipeline software (version 1.18.64.0) with default parameters

Sequence reads were aligned to the human reference sequence (NCBI build 37/hg19) using the Illumina aligner “ELAND” (Efficient Large-scale Alignment of Nucleotide Databases). When at least one read in a pair maps to a unique location in the genome, that read and its pair are then subjected to a more accurate alignment with Novoalign v3.02.07. The aligned lane bam files are merged, sorted, and indexed. Sequenced bases with a probability of an incorrect base call of 1 in

100143, or a phred quality scores less of than 20, where ignored.

3.4.2.3 Identification of Candidate Variants by the University of Washington Center for

Mendelian Genomics

Our collaborators at the University of Washington Center for Mendelian Genomics (UW-

CMG) reviewed the whole-exome sequencing data and pedigree files. Using these two files, they confirmed the sex of all participants and the relationships reported within triads. All of the triads in this study population included affected children and non-affected parents. Using this information, they investigated de novo, compound heterozygote, autosomal recessive, X-linked 38

recessive, and X-linked de novo models of inheritance to generate a list of candidate genes and variants which will meet the filters outlined in Table 3.5.

Table 3.4 Filters used in determination of candidate variants using different models of inheritance.

Filter Level Requirement Genotype  Read depth  4  Genotype quality  20 Variant  Frequency  0.05 in UW-CMG subjects  MAF < 0.05 in NHLBI GO Exome Sequency Project (ESP6500) and/or 1,000 Genomes  Not annotated in SeattleSeq137 as: - Intergenic - Intron - Synonymous or Synonymous-coding - Untranslated region (utr) - 3’ Untranslated region (utr-3) - 5’ Untranslated region (utr-5) - Near gene- 3’ end - Near gene- 5’ end - No coding Person  All members in each family must have the correct genotype given the inheritance model

After these analyses are complete, the UW-CMG group provided us with a written report outlining the list of candidate variants within each family. Specific attention was paid to genes which are known to underly other Mendelian conditions, including some which have phenotypic overlap. In addition, we provided a comprehensive list of genes important in fetal eye development and/or those that are in a signaling pathway that interacts with genes known to be causal in PCG (CYP1B1,

LTBP2, MYOC, etc).

3.4.3 Initial analysis plans for Aim 2

Initial plans for Aim 2 included three sub-aims. In sub-aim 1 we had intended to focus on calculating risk using a log-linear model, in common SNPs present in genes which have previously been identified as causal for PCG and ACDs in the literature. Sub-aim 2 was to focus on less common SNPs and single nucleotide variants (SNVs) in genes outlined in sub-aim 1, but also in de novo mutations identified in at least three families in the UW-CMG analysis. Sub-aim 3 was to investigate gene-environment interactions.

Upon receiving the UW-CMG report, we discovered that the only gene with variants in more than one family was CYP1B1, though there were multiple genes with biologic plausibility found in individual families. Thus, we decided that the efficient and effective focus of Aim 2 would be a descriptive analysis of the findings, with a special emphasis on genes which hadn’t previously been described in PCG patients.

3.4.4 Updated analysis for Aim 2

The goal for Aim 2 is to evaluate the prevalence of variants in genes which have previously been linked to PCG such as CYP1B1, LTBP2, MYOC, and FOXC1. We will investigate whether any variants are novel, compare the prevalence of variants in each gene to previous literature, and evaluate if there appears to be segregation of variants by maternal race/ethnicity.

Further, we plan to utilize the strengths of whole-exome sequencing to identify rare, pathogenic variants in genes which previously have not been investigated for PCG. Efforts will be focused on genes which are known to underly other Mendelian conditions, especially those with phenotypic overlap, and genes which were identified in eye development signaling pathways or are known to interact with genes previously linked to PCG. The role of each gene and corresponding protein will be investigated for expression levels and function in the developing 40

mammalian eye. A literature review will be done on each variant to determine whether it has previously been described and if there are in vitro or in vivo functional studies.

CHAPTER 4. – MATERNAL DIET AS A RISK FACTOR FOR PRIMARY CONGENITAL GLAUCOMA AND DEFECTS OF THE ANTERIOR SEGMENT OF THE EYE 4.1 Introduction

Childhood glaucoma encompasses cases of primary congenital glaucoma (PCG) and glaucoma with non-acquired ocular anomalies144. These conditions are rare, with prevalence ranging from 1 in 10,000 (PCG) to 1 in 100,000 (Peters anomaly45) live births. Two diagnostic differences between these groups is if glaucoma co-occurs with ocular defects and the timing of glaucoma presentation. PCG typically presents in the absence of co-occurring ocular anomalies.

Approximately 25% of cases are diagnosed at birth and close to 80% are diagnosed within the first year of life1. Glaucoma with non-acquired ocular anomalies includes a collection of conditions where glaucoma presents secondary to anterior segment dysgenesis (ASDs), which are defects in the front portion of the eye present at birth. This group of heterogeneous disorders includes conditions such as Axenfeld-Rieger anomaly, Peters anomaly, and aniridia, among others144. Both

PCG and ASDs result in a restriction of the flow of the aqueous humor, the clear fluid which supplies nutrients and removes metabolic waste from the eye, leading to an increased intraocular pressure. This can cause blindness if left untreated or if poorly controlled 2. At the genetic level, there is some evidence of a shared etiology for PCG and ASDs. Mutations in some of the same genes, including CYP1B1 and FOXC1, have been shown to cause both PCG and ASDs107,108.

Much of the research on PCG and ASDs has focused on the genetics of these conditions and little is known about non-genetic risk factors. Many of the findings for non-genetic risk factors are mixed and dependent on the country of origin for the study sample. In a series of studies from

Hungary, increased risks of PCG were seen among infants of lower birth weight, higher birth order, mothers who used tobacco or alcohol during pregnancy, unmarried mothers, and unemployed mothers 110,111. In a recent analysis of US-based data from the National Birth Defects Prevention

Study our research group also found a higher risk of PCG among lower birth weight infants and reported a number of new findings such as an increased odds among Non-Hispanic Blacks, increased odds with certain medications, and inverse associations with maternal smoking and alcohol intake112.

There are many studies showing an association between maternal periconceptional diet and birth defects. Among these are studies of defects in tissues that arise from the neural crest cells, the same cells that differentiate into the trabecular meshwork. For example, epidemiologic studies have shown that diet quality124 and certain dietary patterns23,145 are associated with neural tube defects (NTDs). While there are no studies on the role of maternal diet and congenital glaucoma, there have been studies linking maternal deficiencies of folic acid15,17, zinc19 and vitamin C20 to other structural malformations of the eye. In addition, animal studies have shown structural abnormalities in the trabecular meshwork could be prevented by molecules that act as free radical scavengers or have anti-oxidant capacities21. Apart from one study which reported total folate intake, the current literature on maternal nutrition and congenital glaucoma is limited to prenatal folic acid and/or vitamin use and findings are equivocal 111,112. Vogt et al111 found inverse associations for all vitamin/mineral supplements investigated; some were statistically significant

(iron, folic acid, and vitamin D) and others were only suggestive with large confidence intervals

(calcium, vitamin E, and multivitamins). Conversely, Forestieri et al112 found the suggestion of an increased odds among those mothers who used folic acid supplements during early pregnancy.

Interestingly, in Forestieri112 et al total folate intake was suggestive of a U-shaped pattern where

inverse associations were seen for middle quartiles and no association was seen in the highest quartile. In this study, we expanded on these findings by comprehensively evaluating maternal diet as a risk factor for PCG and ASDs.

4.2 Materials and Methods

Study Population

Data for this analysis come from the National Birth Defects Prevention Study (NBDPS), a large population-based multicenter case-control study of over 30 major birth defects in the United

States. Detailed methods have been reported previously 127. Briefly, case infants were identified through birth defects surveillance systems in ten states (AR, CA, GA, IA, MA, NJ, NY, NC, TX, and UT) and controls were live-born infants without major birth defects who were selected at random from vital records (IA, MA, NC, NJ, UT, CDC-MACDP) or hospital birth records (AR,

CA, NY, TX). Controls were not matched to cases; however, they were selected from the same geographical region and time periods as cases. The NBDPS was approved by the Institutional

Review Board (IRB) at the Centers for Disease Control and Prevention, and each study center also received local IRB approval.

Case Classification

Clinical geneticists at each study center reviewed abstracted medical records to confirm that identified cases met eligibility criteria, which included diagnosis by an ophthalmologist within the first 12 months of life. Cases with defects of known etiology, including single-gene disorders and chromosomal abnormalities were excluded. Cases of PCG and ASDs (Peters anomaly,

Axenfeld-Rieger anomaly, aniridia, lens coloboma, and aphakia/spherophakia) were eligible for

NBDPS from birth years 2000-2011. We included both isolated and non-isolated PCG and ASD cases in our study, since previous investigations in this population demonstrate that the distribution

of demographic and clinical risk factors are quite similar between cases with isolated vs. non- isolated defects.

Risk Factor Ascertainment

Upon enrollment in NBDPS, mothers were asked to complete a standardized computer assisted telephone interview (CATI) between 6 weeks to 24 months after the infant’s estimated date of delivery. Interviews included a range of demographic, clinical, lifestyle, and behavioral factors. Participation rates varied from center to center and over time, but overall 67% of cases and 65% of controls who were invited chose to participate. On average case mothers were interviewed approximately 5 months later than control mothers (12.5 months vs. 7.5 months after the infant’s date of birth). Part of the interview included a 58-question version of the validated

Willet food frequency questionnaire146 (FFQ) modified for NBDPS, which was used to assess dietary intake for the year before pregnancy. Mothers were asked to report their average frequency of consumption based on the serving size for each food item. These values were used to estimate the average daily intake, with nutrient calculations being derived from the U.S. Department of

Agriculture, version 25 nutrient database134.

Mothers were also asked about vitamin and mineral supplement use from 3 months before they became pregnant until the end of their pregnancy. Frequency of use and start- and stop-dates were recorded. If a mother did not remember specific dates, she was asked for the approximate duration of use. Each product was reviewed to determine whether it contained folic acid and if so, it was included in the category of vitamin use. In this analysis, we restricted use to the periconceptional period, which includes the 2 months before conception to 2 months after conception- reflecting the timing of major embryologic eye development.

Assessment of Pre-Pregnancy Dietary Quality and Dietary Patterns

The Diet Quality for Pregnancy Index (DQI-P) scores were calculated similarly to

Carmichael et al124. This score combines six positively scored components (grains, folate, fruits, vegetables, iron, and calcium) and two negatively scored components (percentage of calories from sweets and fats)120,147. Since the original DQI-P was developed, minor changes have been made over time such as the inclusion of sweets, the exclusion of meal patterns, and the use of quartiles among controls to rank each component and then sum instead of using absolute values 124,148.

Latent class analysis (LCA) was used to derive dietary patterns based on participant’s consumption of food items from the FFQ after adjusting for caloric intake. LCA is a statistical method for identifying subgroups, or “classes”, that are not directly observed within a sample or population. This method has previously been described in an analysis of NBDPS data by Sotres-

Alvarez et al23. Briefly, dietary patterns, adjusting for energy intake, were derived using dietary intake data among controls. Regression coefficients that predicted class membership and conditional probabilities of the observed indicators given the class were used to predict posterior probabilities of class membership for both cases and controls. Each subject was then assigned to the class for which they had the highest probability of belonging. Despite including slightly different birth years for the control group (1997-2005 in Sotres-Alvarez et al vs. 2000-2011 in the current analysis), the four dietary classes we saw had similar characteristics to previous work within NBDPS. Women in the Prudent class were more likely to consume fruits, vegetables and whole wheat breads. Those in the Mexican dietary pattern were more likely to consume tortillas, refried beans, avocados, and similar to the Prudent dietary pattern, fruits and vegetables. Finally, high intake of regular sodas, French fries, and white bread consumption characterized those in the

Western and Low-Calorie Western dietary patterns. These patterns differed from each other in

terms of calorie intake (those in the Low-Calorie Western consumed 222 fewer calories) and patterns of probabilities of consumption (those in the Low-Calorie Western dietary pattern had higher extreme probabilities, such as higher levels of no consumption).

Statistical Analysis

We excluded non-singleton births (6 cases; 275 controls), pregnancies conceived with use of donor egg or sperm (1 case; 28 controls), mothers with energy intake <500 or > 5,000 kcal (4 cases; 154 controls), and mothers with more than one food item missing from the FFQ (9 cases;

448 controls). Finally, cases with co-occurring anophthalmia or microphthalmia were also excluded to minimize heterogeneity within the outcome group (15 cases; 0 controls).

We used logistic regression to estimate odds ratios (ORs) and 95% confidence intervals

(CIs) for individual micronutrients, DQI-P scores, and dietary patterns. Nutrient intakes and DQI-

P scores were divided into quartiles based on the distribution among controls. All analyses were adjusted for total energy intake (Kcal). Additional analyses estimated adjusted odds ratios (aORs) and were further adjusted for covariates identified by a Directed Acyclic Graph, including maternal age (continuous, years), maternal race/ethnicity (non-Hispanic White, non-Hispanic Black,

Hispanic, Other), periconceptional vitamin supplementation (yes, no), maternal BMI (<18.5, 18.5-

<25, 25-<30, ≥30), and study center. A sensitivity analysis was done stratifying by periconceptional vitamin supplementation, based on the a priori assumption that risks may be different in women reporting vitamin use149.

All analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC).

4.3 Results

There were 152 cases (91 PCG and 61 ASD; all livebirths) and 9,178 controls eligible for this analysis. Cases were more likely to be male and delivered preterm (Table 4.1). Among term

infants, cases were more likely to be low birth weight. The majority of case infants with PCG

(86%) and ASDs (84%) did not have another birth defect (i.e., the PCG or ASD was considered isolated). The majority of mothers self-identified as Non-Hispanic White (59%), but proportionally Non-Hispanic Black ethnicity was more common among PCG cases compared to control mothers. Compared to control mothers, case mothers were also less likely to report smoking during pregnancy, have a slightly higher BMI, and a higher age among ASD cases only.

Periconceptional vitamin use was high in the study sample with no apparent differences between cases and controls (p=0.87). Control mothers had slightly higher daily energy intake (1604.5 kcal) compared to mothers of cases with PCG (1550.5 kcal) and ASD (1490 kcal) though they were not statistically different (p=0.46 and p=0.20; respectively).

Associations with various micronutrients and diet quality are shown in Table 4.2. Most were consistent with a null finding and were not monotonic across quartiles. However, some of the aORs suggested inverse, although imprecise, associations with increasing micronutrient intake.

Compared to the lowest quartile of intake, all three remaining quartiles of calcium, choline, and

Vitamin E were associated with consistently lower odds of PCG or ASDs. For example, the aORs for PCG or ASDs were lower for those in the highest quartile of choline intake (aOR 0.79, 95%

CI 0.40-1.55) compared to those in the lowest quartile. Intake of copper and folate showed lower odds ratios among those in the middle quartiles, but increased aORs among those in the highest quartile. Both zinc and Vitamin A were suggestive of this U-shaped pattern, though ORs in the highest quartile of intake were close to, but below, the null value. There did not appear to be any association between DQI-P and PCG or ASDs (Q3 aOR 1.03, 95% CI 0.61-1.72; Q4 aOR 0.91,

95% CI 0.49-1.66 vs. the lowest quartile). Findings were similar with stratification by periconceptional vitamin use and case status (Supplemental Tables 1 & 2).

Mothers who belonged to the Prudent and Mexican dietary patterns had lower odds (aOR

0.82, 95% CI 0.52-1.29, aOR 0.80, 95% CI 0.36-1.78) of having a child with PCG or ASD compared to mothers in the Western dietary pattern, but again, confidence intervals included 1.0

(Table 4.3). Mothers in the Low Calorie Western dietary pattern had similar odds compared to those in the Western dietary pattern (aOR 1.04, 95% CI 0.67-1.64).

Table 4.1. Demographic and clinical characteristics of primary congenital glaucoma and anterior segment defect cases and controls, National Birth Defects Prevention Study, 2000-2011.

Controls Primary Congenital Anterior Segment (n= 9178) Glaucoma Defects (n=91) (n=61) Characteristic n (%) or mean (SD) n (%) or mean (SD) n (%) or mean (SD) Infant Sex Female 4445 (48.5) 41 (45.1) 28 (45.9) Male 4723 (51.5) 50 (55.0) 33 (54.1) Missing 10 0 0 Gestational Age < 37 weeks 751 (8.2) 10 (11.0) 7 (11.5) 37- 41 weeks 8312 (90.6) 79 (86.8) 54 (88.5) ≥ 42 weeks 115 (1.3) 2 (2.2) 0 (0) Birth Weighta <2500 grams 152 (1.8) 5 (6.7) 3 (5.7) ≥2500 grams 8158 (98.2) 70 (93.3) 50 (94.3) Missing 117 6 1 Maternal

Race/Ethnicity Non-Hispanic White 5282 (57.6) 46 (50.6) 37 (60.7) Non-Hispanic Black 975 (10.6) 23 (25.3) 8 (13.1) Hispanic 2279 (24.8) 16 (17.6) 12 (19.7) Other 642 (7.0) 6 (6.6) 4 (6.6) Gravidity Primigravid 2714 (29.6) 28 (30.8) 18 (29.5) Multigravid 6462 (70.4) 63 (69.2) 43 (70.5) Missing 2 0 0 Maternal age at delivery ≤ 24 2998 (32.7) 35 (38.5) 19 (31.2) 25-34 4922 (53.6) 49 (53.9) 28 (45.9) ≥ 35 1258 (13.7) 7 (7.7) 14 (23.0) Paternal age at delivery ≤ 24 1956 (21.9) 22 (25.9) 12 (19.7) 25-34 4792 (53.7) 41 (48.2) 28 (45.9) ≥ 35 2179 (24.4) 22 (25.9) 21 (34.4) Missing 251 6 0 Maternal smoking No 7518 (82.4) 84 (93.3) 52 (85.3) Yes 1608 (17.6) 6 (6.7) 9 (14.8) Missing 52 1 0

Maternal BMI

(kg/m2) <18.5 465 (5.3) 1 (1.2) 2 (3.5) 18.5 - <25 4621 (52.5) 48 (55.8) 31 (54.4) 25 - <30 2054 (23.4) 18 (20.9) 12 (21.1) ≥ 30 1656 (18.8) 19 (22.1) 12 (21.1) Missing 382 5 4 Perinconceptional

Vitamin Use No 2011 (22.1) 18 (20.0) 13 (21.3) Yes 7073 (77.9) 72 (80.0) 48 (78.7) Missing 94 1 0 Daily Energy 1604.5 (688.5) 1550.5 (703.0) 1490.1 (669.9) Intake (kcal) a Restricted to term birth (gestational age ≥37 weeks) only

Table 4.2. ORs and 95% CIs for individual nutrients for combined primary congenital glaucoma and anterior segment defects, National Birth Defects Prevention Study, 2000-2011.

Nutrient Cases Controls ORa 95% CI aORb 95% CI (%) (%) Calcium (mg) Q1 49 (32.2) 2295 (25.0) 1 1 Q2 39 (35.7) 2294 (25.0) 0.80 0.52, 1.23 0.79 0.50, 1.25 Q3 35 (23.0) 2295 (25.0) 0.72 0.45, 1.15 0.73 0.44, 1.20 Q4 29 (19.1) 2294 (25.0) 0.60 0.34, 1.06 0.63 0.34, 1.16 Betaine (mg) Q1 36 (23.7) 2295 (25.0) 1 1 Q2 40 (26.3) 2295 (25.0) 1.18 0.75, 1.87 1.14 0.71, 1.85 Q3 32 (21.1) 2293 (25.0) 1.01 0.62, 1.66 1.01 0.60, 1.70 Q4 44 (29.0) 2295 (25.0) 1.56 0.95, 2.57 1.52 0.88, 2.60 Choline, total Q1 49 (32.2) 2295 (25.0) 1 1 Q2 37 (24.3) 2294 (25.0) 0.78 0.50, 1.21 0.84 0.53, 1.33 Q3 32 (21.1) 2295 (25.0) 0.68 0.41, 1.13 0.70 0.41, 1.18 Q4 34 (22.4) 2294 (25.0) 0.77 0.40, 1.45 0.79 0.40, 1.55 Copper (mg) Q1 46 (30.3) 2253 (24.6) 1 1 Q2 42 (27.6) 2333 (25.4) 0.93 0.60, 1.44 0.93 0.59, 1.47 Q3 26 (17.1) 2297 (25.0) 0.61 0.36, 1.05 0.65 0.37, 1.15 Q4 38 (25.0) 2295 (25.0) 0.98 0.52, 1.83 1.08 0.54, 2.17 Iron (mg) Q1 43 (28.3) 2294 (25.0) 1 1 Q2 44 (29.0) 2295 (25.0) 1.10 0.71, 1.70 1.12 0.72, 1.75 Q3 26 (17.1) 2291 (25.0) 0.68 0.40, 1.14 0.62 0.36, 1.07 Q4 39 (25.7) 2298 (25.0) 1.13 0.65, 1.97 1.18 0.67, 2.09 Magnesium (mg) Q1 51 (33.6) 2295 (25.0) 1 1 Q2 37 (24.3) 2294 (25.0) 0.75 0.48, 1.16 0.74 0.47, 1.18 Q3 28 (18.4) 2295 (25.0) 0.58 0.34, 0.97 0.59 0.34, 1.04 Q4 36 (23.7) 2294 (25.0) 0.79 0.41, 1.51 0.90 0.44, 1.85 Riboflavin (mg) Q1 46 (30.3) 2292 (25.0) 1 1 Q2 39 (25.7) 2291 (25.0) 0.88 0.57, 1.37 0.91 0.58, 1.43 Q3 32 (21.1) 2287 (24.9) 0.75 0.46, 1.24 0.73 0.43, 1.22 Q4 35 (23.0) 2308 (25.2) 0.88 0.49, 1.60 0.87 0.46, 1.63 Vitamin B6 (mg) Q1 48 (31.6) 2310 (25.2) 1 1 Q2 35 (23.0) 2290 (25.0) 0.77 0.49, 1.22 0.72 0.45, 1.15 52

Q3 33 (21.7) 2281 (24.9) 0.77 0.47, 1.26 0.80 0.48, 1.33 Q4 36 (23.7) 2297 (25.0) 0.91 0.50, 1.68 0.91 0.49, 1.72 Vitamin B12 (µg) Q1 39 (25.7) 2303 (25.1) 1 1 Q2 45 (29.6) 2289 (24.9) 1.25 0.80, 1.94 1.25 0.79, 1.96 Q3 27 (17.8) 2297 (25.0) 0.80 0.47, 1.34 0.77 0.45, 1.32 Q4 41 (27.0) 2289 (24.9) 1.37 0.80, 2.34 1.30 0.75, 2.28 Vitamin E (mg) Q1 50 (32.9) 2290 (25.0) 1 1 Q2 39 (25.7) 2300 (25.1) 0.79 0.51, 1.21 0.82 0.52, 1.28 Q3 30 (19.7) 2297 (25.0) 0.61 0.37, 1.00 0.59 0.35, 1.00 Q4 33 (21.7) 2291 (25.0) 0.69 0.39, 1.23 0.76 0.41, 1.39 Zinc (mg) Q1 51 (33.6) 2297 (25.0) 1 1 Q2 31 (20.4) 2293 (25.0) 0.65 0.41, 1.03 0.70 0.43, 1.12 Q3 31 (20.4) 2290 (25.0) 0.69 0.42, 1.14 0.72 0.42, 1.21 Q4 39 (25.7) 2298 (25.0) 0.96 0.53, 1.76 0.94 0.50, 1.77 Vitamin A, RAE (µg) Q1 49 (32.2) 2295 (25.0) 1 1 Q2 31 (20.4) 2294 (25.0) 0.67 0.42, 1.06 0.61 0.37, 0.98 Q3 32 (21.1) 2295 (25.0) 0.72 0.45, 1.16 0.72 0.44, 1.18 Q4 40 (26.3) 2294 (25.0) 1.00 0.60, 1.67 0.94 0.55, 1.63 Carotene, alpha (µg) Q1 46 (30.3) 2294 (25.0) 1 1 Q2 45 (29.6) 2295 (25.0) 0.98 0.65, 1.48 1.09 0.70, 1.68 Q3 26 (17.1) 2294 (25.0) 0.58 0.35, 0.93 0.58 0.34, 0.99 Q4 35 (23.0) 2295 (25.0) 0.81 0.51, 1.28 0.87 0.53, 1.45 Carotene, beta (µg) Q1 42 (27.6) 2294 (25.0) 1 1 Q2 44 (29.0) 2295 (25.0) 1.06 0.69, 1.63 1.15 0.74, 1.80 Q3 31 (20.4) 2295 (25.0) 0.77 0.48, 1.23 0.83 0.50, 1.37 Q4 35 (23.0) 2294 (25.0) 0.91 0.57, 1.46 0.96 0.57, 1.63 Retinol (µg) Q1 41 (27.0) 2295 (25.0) 1 1 Q2 36 (23.7) 2294 (25.0) 0.93 0.59, 1.47 0.90 0.56, 1.43 Q3 37 (24.3) 2295 (25.0) 1.01 0.63, 1.61 0.96 0.59, 1.57 Q4 38 (25.0) 2294 (25.0) 1.16 0.69, 1.95 1.09 0.63, 1.87 Folate, DFE (µg) Q1 43 (28.3) 2294 (25.0) 1 1 Q2 35 (23.0) 2295 (25.0) 0.87 0.55, 1.38 0.86 0.53, 1.38 53

Q3 35 (23.0) 2294 (25.0) 0.92 0.57, 1.48 0.85 0.52, 1.41 Q4 39 (25.7) 2295 (25.0) 1.14 0.67, 1.93 1.18 0.69, 2.02 DQI-P Q1 34 (22.5) 2056 (22.6) 1 1 Q2 47 (31.1) 2206 (24.2) 1.32 0.84, 2.08 1.38 0.87, 2.19 Q3 37 (24.5) 2334 (25.6) 1.01 0.62, 1.66 1.03 0.61, 1.72 Q4 33 (21.9) 2522 (27.7) 0.88 0.50, 1.54 0.91 0.49, 1.66 a Model was adjusted for energy intake (continuous) b Model was adjusted for maternal age (continuous), race/ethnicity (categorical), periconceptional vitamin supplementation (categorical), BMI (categorical), study center (categorical), and energy intake (continuous)

Table 4.3 Association between dietary pattern class and having a child with primary congenital glaucoma or anterior segment defects, National Birth Defects Prevention Study, 2000-2011.

Dietary Pattern Cases Controls ORa 95% CI aORb 95% CI Western 40 (26.3) 2300 (25.1) 1 1 Prudent 46 (30.3) 3024 (33.0) 0.86 0.56, 1.32 0.82 0.52, 1.29 Mexican 16 (10.5) 1404 (15.3) 0.69 0.38, 1.24 0.80 0.36, 1.78 Low-Calorie 50 (32.9) 2450 (26.7) 1.14 0.75, 1.74 1.04 0.67, 1.64 Western a Model was adjusted for energy intake (continuous) b Model was adjusted for maternal age (continuous), race/ethnicity (categorical), periconceptional vitamin supplementation (categorical), BMI (categorical), study center (categorical), and energy intake (continuous)

4.4. Discussion

The importance of maternal diet in the etiology of birth defects has been well demonstrated22,23,124,125,148. Our study is the first to report on the association between maternal nutrition and PCG and ASDs. Our results indicate that higher intake of some nutrients and certain dietary patterns may be associated with a lower odds of PCG and ASDs. No association between these defects and the DQI-P were noted.

We found consistently lower odds in the top three quartiles of calcium, choline, and vitamin

E intake. Calcium has been shown to be important in eye development in zebrafish150 and has been 54

implicated to impact the function of trabecular meshwork and retinal ganglion cells151,152. There are currently no epidemiologic studies on the association between calcium or vitamin E and congenital glaucoma. Deficiencies in vitamin E, a potent lipophilic antioxidant153, during pregnancy have been linked to increased fetal demise 154 in animal studies and early miscarriage in human studies155. Finally, choline has been studied for its beneficial effects in retinal progenitor cell development156. Our study is the first to report on the association between choline intake and risk of anterior segment defects of the eye. In contrast to our findings, a recent study of diet and anophthalmia (absence of one or both eyes) and microphthalmia (abnormally small eye(s)) in

NBDPS observed slightly elevated odds with increasing total choline levels148. The benefits of choline may therefore be limited to certain structures of the eye. The human eye is derived from various embryonic tissues including the neural ectoderm, surface ectoderm, and mesenchymal cells predominantly arising from neural crest cells, the latter contributing substantially to the anterior structures. This developmental pattern could explain why higher levels of choline have been found to be associated with reduced odds of neural tube defects157.

There is limited research on the use of vitamin/mineral supplements amongst PCG case mothers. Vogt et al111 found a lower odds among mothers that reported use of iron and vitamin D supplements during pregnancy. There did not appear to be a consistent pattern among higher quartiles of dietary iron intake in the mothers within NBDPS. We did not investigate vitamin D intake within our study population. Similar to findings from studies of other birth defects145,157,

Vogt et al111 found an lower odds of having a child with PCG in their study population among mothers who used folic acid supplements during pregnancy. In our study there was a lower odds of PCG and ASDs among those in the second and third quartiles compared to the lowest dietary folate intake level, and those in the highest quartile had a slightly increased odds of PCG and

ASDs. These findings are similar to previous NBDPS work by our group showing that mothers who took periconceptional folic acid-containing supplements less than once per day or daily had increased odds of having a child with PCG (aOR 1.23, 95% CI 0.72-2.10; aOR 1.37, 95% CI 0.73-

2.57)112. These findings may be explained by the differing study periods and when mandatory folic acid fortification was implemented in each country. The study period for Vogt et al111 (1980- 2002) predated Hungary’s folic acid food fortification program which began in August 1998158, whereas the NBDPS study population used in congenital glaucoma analyses (2000-2011) occurred after fortification programs were implemented in the United States8.

We did not find an association between DQI-P and PCG and ASDs in this study. These findings differ from literature on more common birth defects like gastroschisis22, orofacial clefts23, and neural tube defects23 that found higher diet quality was associated with reduced risk of these birth defects. The sample size of cases in each of these studies ranged from nearly two- to ten- times the number of cases in this analysis and therefore it’s possible that our study was underpowered.

Studies that have examined dietary patterns in birth defects research have found that

“healthier diets” (those rich in fruits and vegetables23,145 or specific nutrients like folate and choline159) were associated with reduced risks of birth defects. Similarly, we found that those mothers who belonged to the Prudent dietary pattern, characterized by greater fruit and vegetable intake, had lower odds of having a child with PCG or ASDs. Interestingly, we found similar odds ratios among those in the Prudent and Mexican dietary patterns. This may be due to higher levels of free radical scavengers or anti-oxidants, which have been shown to modify PCG phenotypes in animal studies21, that are present in the foods characteristic of both patterns.

This study has many strengths. NBDPS is a multi-year, population-based case- control study which includes participants from 10 different states. These participants reflect racial/ethnic, geographic, and socioeconomic diversity that is generally representative of the US population160.

Furthermore, methods are rigorous and standardized across centers ensuring thorough and uniform data collection. Individuals who chose to participate in the study were not substantially different than those who chose not to be interviewed, which further provides support of the representativeness of the study sample112. Clinical review and classification of case infants provides a rigorously defined case sample for this study. Another strength of this study is the fact that maternal diet was measured in various ways. This approach yields a more comprehensive analysis of maternal nutrition. Though many epidemiologic studies evaluate the impact of single nutrients, this approach does not account for correlations between nutrient intake, nor the complex interactions in bioavailability between nutrients. In addition to the individual nutrients, we also used a composite score measuring diet quality which allowed for the assessment of food groups and incorporates dietary recommendations specific for pregnant women. Lastly, our data-driven approach (latent class analysis) identifies empirical dietary patterns inherent within our study sample, which allows for exploration of patterns that may not necessarily align with a “gold standard” based on recommended intake of specific nutrients or food items.

Despite the size of NBDPS, the rarity of PCG and ASDs still resulted in a relatively small number of cases and imprecise effect estimates. Nonetheless, this national study is the largest to examine maternal diet in relation to these birth defects. NBDPS cases are limited to infants diagnosed within the first year of life. Although most infants with PCG are diagnosed within the first year of life 1, our case sample may be incomplete and could miss milder cases that are not diagnosed until after one year of age or those who are diagnosed later for other reasons (e.g.,

reduced access to healthcare). Further, though the FFQ is validated, this study still includes self- reported data on each mother’s diet in the year before pregnancy. Some foods may be believed to be healthier than others, possibly resulting in the potential for differential recall bias between case and control mothers. If case mothers are more likely to report some of these foods (e.g. fruits, vegetables, whole grains), we would have seen more positive associations for increasing micronutrient intake. Finally, as little is known about the risk factors for PCG and ASDs, there may be residual confounding. One such possibility is heterogeneity in underlying genetic susceptibility. The most consistently mutated gene in congenital glaucoma is CYP1B1, with many of the mutations being described in different ethnic populations69. These patients may develop congenital glaucoma regardless of non-genetic exposures. Ideally, future studies would account for known genetic variants when investigating non-genetic exposures like maternal diet.

In summary, this study evaluated specific maternal dietary factors, diet patterns, and diet quality as risk factors for PCG and ASDs in a large, population-based case-control study. We found that higher intake of some nutrients and certain dietary patterns were associated with lower odds of PCG and ASDs. These findings, if replicated, add to the accumulating evidence of the importance of healthy maternal nutrition during pregnancy.

CHAPTER 5. – EXOME SEQUENCING TO IDENTIFY NOVEL GENES UNDERLYING PRIMARY CONGENITAL GLAUCOMA IN THE NATIONAL BIRTH DEFECTS PREVENTION STUDY 5.1 Introduction

Primary congenital glaucoma (PCG) is diagnosed early in childhood and often occurs without overt structural eye defects6. Children present with raised intraocular pressure (IOP), loss of corneal transparency, photophobia, and enlargement of the eye2. Raised IOP in this context is thought to be a result of reduced flow of aqueous humor out of the anterior chamber of the eye through the trabecular meshwork, a porous structure of tissues161.

In the US the prevalence of PCG is approximately 1 in 10,000 live births per year2,3, but has been reported to be as high as 1in 1,250 in communities with high consanguinity 34.

Historically PCG was considered inherited as an autosomal recessive trait; however,this model of inheritance has been recently challenged by reports of transmission in successive generations, unequal sex distribution among affected individuals, and lower-than-expected numbers of affected siblings in familial cases64,67,68,70–72.

A few genes are known to be responsible for some cases of PCG, though the most commonly mutated gene is is CYP1B1, a cytochrome P450 enzyme. Frequencies of specific

CYP1B1 mutations are known to vary with reports of 90-100% cases of Slovakian Roma

Gypsies, 50% of cases from Brazil, 44% of cases from India, and 20% of cases from Japan having CYP1B1 mutations 9,10,162. Despite these high prevalences in some populations, pathogenic variants have only been found in 16-25% of families from the US10,163. Mutations in other genes, such as LTBP2 and MYOC, have been found in consanguineous families affected by

PCG, but studies in families from the US have failed to find the same disease-causing mutations10. This finding suggests that variants in genes other than CYP1B1 may cause PCG in the general US population.

In this study, we will describe the prevalence of CYP1B1 variants identified used whole- exome sequencing data of complete trios (mother, father, and PCG-affected child) from the

National Birth Defects Prevention Study.,This is a population-based case-control study that has been shown to be generally representative of the US population160. In addition, we used these data to discover variants in genes which have not previously been linked to PCG cases in the current literature.

5.2 Materials and Methods

Study Population

Data for this analysis come from the National Birth Defects Prevention Study (NBDPS), a large multicenter population-based case-control study of over 30 major structural birth defects in the US. Detailed methods have been reported previously127. Briefly, case infants were identified through birth defects surveillance systems in ten states (AR, CA, GA, IA, MA, NJ, NY, NC, TX, and UT) and were diagnosed up to one year after delivery. Clinical geneticists at each study center reviewed abstracted medical records to confirm that identified cases met eligibility criteria. Cases were then systematically reviewed again by a study-wide clinician to perform further classification according to study protocols (cite Rasmussen paper). Cases with defects of known etiology, including single-gene disorders and chromosomal abnormalities were excluded.

This analysis only included WES data from PCG cases (ICD-9 codes 743.200-743.204) and their parents. A total of 38 complete trios (mother-father-child) were included in this analysis

Cases of PCG were eligible for NBDPS from birth years 2000-2011, though all cases included in this analysis were enrolled after 2004.

The NBDPS was approved by the Institutional Review Board (IRB) at the Centers for

Disease Control and Prevention, and each study center also received local approval.

Buccal Cell Collection and DNA Isolation

Families that participated in the NBDPS completed a telephone interview between 6 weeks and 18 months after delivery, and were subsequently mailed a buccal cell collection kit which included documents for informed consent, 6 cytobrushes (2 per eligible family member), instructions for use, and a postage-paid return mailer. DNA was extracted from one of the cytobrushes using either a phenol-chloroform method or Gentra Purgene (Qiagen). DNA quantity was assessed by quantitative real-time PCR using the human RNaseP gene, and STR markers were used to assess DNA quality. Quality control standards included 0.1 ng/uL DNA, successful PCR amplification after at least two attempts with at least one STR marker, and alleles consistent with the reported family relationship127.

Whole-exome sequencing and alignment

Samples were prepared for sequencing at the NIH Intramural Sequencing Center (NISC) using 100 ng of genomic DNA and the Accel-NGS 2S Plus DNA Library Kit. Modifications to the library kit protocol included modified PCR cycling conditions to improve recovery of high

GC regions and swapping a polymerase at the post-capture library amplification step.

The Nimblegen SeqCap EZ Exome + UTR Library (Version 3.0) was used for whole- exome sequencing. Samples were then pooled and sequenced on an Illumina HiSeq 2500 instrument. Image analysis and base calling were performed using Illumina Genome Analyzer

Pipeline software (version 1.18.64.0) with default parameters

Sequence reads were aligned to the human reference sequence (NCBI build 37/hg19) using the Illumina aligner “ELAND” (Efficient Large-scale Alignment of Nucleotide Databases).

When at least one read in a pair maps to a unique location in the genome, that read and its pair are then subjected to a more accurate alignment with Novoalign v3.02.07. The aligned lane bam files were merged, sorted, and indexed. Sequenced bases with a probability of an incorrect base call of 1 in 100143, or a phred quality scores less of than 20, were ignored.

Identification of Genes

The University of Washington Center for Mendelian Genomics reviewed the whole-exome sequencing data and pedigree files, and confirmed the sex of all participants and the relationships reported within triads. More detailed methods can be found in Jenkins et al164. Briefly, variant filtration was conducted under various Mendelian inheritance models using GEMINI 0.20.2165.

Variant filters included depth ≥ 6, genotype quality of ≥ 20, alternative allele ≤ 0.005 across any single population within reference databases (ExAC v0.3166), Exome Sequencing Project 6200

(v2), 1000 Genomes (phase 3), or the UK10K 2016-02-15 release167), and a predicted medium to high impact of the gene/protein. Finally, any variants that did not pass the strand bias filter in

GEMINI were excluded. To further narrow our focus on genes which could create PCG phenotypes, we decided to focus on those which are known to underlie other Mendelian conditions with phenotypic overlap, genes with roles in eye development, or those which are in the same pathways as genes in which mutations have been reported in PCG cases or other anterior segment defects of the eye (e.g. CYP1B1, LTBP2, FOXC1, MYOC, PAX6 or PITX269).

5.3 Results

A total of 38 complete trios with a PCG-affected child and unaffected parents underwent whole-exome sequencing (WES). Table 1 shows a comparison of demographic characteristics

between the WES samples and all PCG cases that are enrolled within NBDPS that were conceived without donor egg or sperm. There were far fewer mothers who identified as Hispanic among WES samples. Both maternal and paternal age at delivery were slightly higher among the

WES samples and there were slight differences in the proportion of samples from each of the

NBDPS centers.

There was one discrepancy on the sex of the affected child in one of the families. Due to our inability to resolve this through a review of collected clinical and interview data, we excluded that family from further analysis, leaving a total sample size of 37 trios. Two families had high levels of contamination (freemix=0.035 and 0.09; as described in Jun et al168) and are identified as such throughout the rest of this paper.

Five of 37 families (13.5%) carried known CYP1B1 variants (Table 1). Each of these families had compound heterozygous changes which included variants that have been reported in children with PCG previously. There were four frameshift changes, four missense mutations, and two stop gain, but only two variants were identified in more than one family. A 13-bp deletion

(p.Arg355fs) resulting in a pathogenic frameshift change was identified in two of five families

(both Non-Hispanic White). In addition, a missense mutation (p.Glu387Lys) was identified in one Non-Hispanic White and one Non-Hispanic Black family. In one family, which had no suspected pathogenic CYP1B1 variants, the affected child had a de novo frameshift variant

(c.1141dupG) in FOXC1 which is predicted to leads to an early termination of the amino acid sequence.

A number of individual families included cases that harbored variants in genes that are known to be important in eye development or are known to cause other Mendelian conditions with phenotypic overlap. De novo variants thought to result in changes to the protein structure

were found in CRYBB2, RXRA, GLI2, POU4F1, PIEZO2, and B3GALT6. One homozygous recessive variant was found in POMT2. A number of compound heterozygous changes were found in ALDH3A2, CREBBP, USH2A, UGT1A8, AND SLC4A11. Information about functional changes, whether each variant has been previously identified, and both CADD and PolyPhen scores of variants within these genes can be found in Table 3.

Table 5.1 Comparison of demographic characteristics of WES samples to all PCG cases within NBDPS. All PCG cases WES PCG samples Demographic Characteristics (n=136) (n=38) Maternal Race/Ethnicity Non-Hispanic White 63 (46.3) 23 (60.6) Non-Hispanic Black 31 (22.8) 9 (23.7) Hispanic 32 (23.5) 2 (5.3) Asian/Pacific Islander 7 (5.2) 2 (5.3) Other 3 (2.2) 2 (5.3) Maternal Age at Delivery <=24 53 (39.0) 9 (23.7) 25-34 71 (52.2) 25 (65.8) >=35 12 (8.8) 4 (10.5) Paternal Age At Delivery <=24 31 (24.0) 7 (18.9) 25-34 67 (51.9) 21 (56.8) >=35 31 (24.0) 9 (24.3) Study Center Arkansas 22 (16.2) 5 (13.2) California 14 (10.3) 2 (5.3) Iowa 12 (8.8) 4 (10.5) Massachusetts 19 (20.0) 6 (15.8) New Jersey 1 (0.7) New York 8 (5.9) 2 (5.3) Texas 11 (8.1) 1 (2.6) CDC/Atlanta 27 (19.9) 9 )23.7) North Carolina 5 (3.7) 1 (2.6) Utah 17 (12.5) 8 (21.1)

Table 5.2 Pathogenic CYP1B1 variants in 5 primary congenital glaucoma families in the National Birth Defects Prevention Study.

Functional Family Reported Race CYP1B1 Variant and Amino Acid Change Change SNP GLAUC_B1_10 Non-Hispanic White p.Arg355fs Frameshift rs72549380 (c.1064_1076delGAGTGCAGGCAGA) p.Thr404fs Frameshift rs587778873 (c.1200_1209dupTCATGCCACC) GLAUC_B1_12 Non-Hispanic Black p.Glu387Lys (c.1159G>A) Missense rs55989760 p.Pro437Leu (c.1310C>T) Missense rs56175199 GLAUC_B1_13 Non-Hispanic White p.Gln19* (c.55C>T) Stop Gain rs72549388 p.Glu387Lys (c.1159G>A) Missense rs55989760 GLAUC_B1_16 Non-Hispanic White p.Arg290fs (c.868dup) Frameshift rs587778875

66 p.Trp57* (c.171G>A) Stop Gain rs72549387 GLAUC_B1_33 Non-Hispanic White p.Asn423Tyr (c.1267A>T) Missense rs104893629 p.Arg355fs Frameshift rs72549380 (c.1064_1076delGAGTGCAGGCAGA)

Table 5.3 Identification of novel genes in primary congenital glaucoma families in the National Birth Defects Prevention Study.

Inheritance Amino Acid Change Existing CADD Pattern Family Gene Function Variation Score PolyPhen2 Compound GLAUC_B1_1 ALDH3A2 Missense/ p.Ala360Val/ rs759058578/ 15.8/ Possibly Heterozygote Missense p.Ser505Gly rs200501128 9.2 Damaging/ Benign GLAUC_B1_2 CREBBP Missense p.Ser893Leu/ rs142047649/ 12.6/ Benign/ Missense p.Thr856Ala rs373531233 16.7 Possibly Damaging GLAUC_B1_5 USH2A Frameshift/ p.Thr5135fs/ rs770984400/ - / Unknown/ Missense p.Asp778Tyr rs142898216 21.9 Probably Damaging GLAUC_B1_11 UGT1A8 Missense/ p.Asp151Asn rs14739512/ 10.2/ Possibly Missense p.Lue228Ile rs45568432 0.7 Damaging/ Benign GLAUC_B1_21 SLC4A11 Missense/ p.Met856Lys -/ 23.5/ Possibly 67 Stop Gain p.Arg730* rs772409032 37 Damaging/ Unknown De novo GLAUC_B1_7 CRYBB2 Missense p.Trp59Arg 21.8 Probably Damaging GLAUC_B1_9 RXRA Stop Gain c.1101_1102ins* - Unknown GLAUC_B1_27ǂ GLI2 Missense p.Ala1398Thr rs768730033 10.2 Benign GLAUC_B1_32 POU4F1 In-frame c.237_239delGCC - Unknown Deletion GLAUC_B1_34 PIEZO2 Missense p.Ile1441Val 14.3 Probably Damaging GLAUC_B1_37 B3GALT6 Missense p.Thr266Phe 9.8 Possibly Damaging Homozygous GLAUC_B1_30 POMT2 Missense p.Arg659Gln rs770606360 34 Probably Recessive Damaging ǂOne parent in this family had high freemix levels (0.035)

5.4. Discussion

To date, most of the published literature on genetic risk factors for PCG has included reports on mutations within CYP1B1, LTBP2, FOXC1, and MYOC. Consistent with previous studies10, we show in our sample that CYP1B1 is likely the main causative gene for primary congenital glaucoma in some families, but that mutations in this gene are less common among cases from the US (13.5%) than have been previously reported in other populations. We also identified one family which had a mutation in FOXC1, which has previously been implicated in primary congenital glaucoma patients169. It has previously been suggested that there may be other, novel genes which cause primary congenital glaucoma in more ethnically heterogeneous populations.

We have included variants in a number of genes which have not previously been investigated in association with primary congenital glaucoma, but which are known to be important in eye development or are known to cause Mendelian conditions with phenotypic overlap.

CRYBB2 variant

The CRYBB2 gene encodes beta–crystallins which are major proteins of the vertebrate lens and are important for maintaining the transparency and refractory index of the lens. The specific variant (p.Trp59Arg) that we found has not been previously described, though is expected to be pathogenic by PolyPhen and CADD. It occurs in exon 4 near the conversion of a

β-sheet to a α-helix. A similar variant (p.Trp59Cys) has been described in congenital cataracts, a condition where the lens becomes opaque. This area of the protein has been previously shown to be important in correct protein folding170.

CREBBP variant

During eye development, the CREBBP gene is upregulated along with αA-crystallins as the primary lens fibers differentiate171. The CREBBP protein has intrinsic histone

acetyltransferase activity allowing for transcriptional activation172, and it can also act as a sort of scaffold to stabilize proteins of the transcription complex173. Mutations in CREBBP have been linked to Rubinstein-Taybi syndrome (RSTS) which is characterized by distinctive dysmorphic features, short stature, and moderate to severe intellectual disability174,175. A recent report by

Menke et al175 described a number of patients who had missense mutations in a specific region of

CREBBP who did not exhibit the typical phenotypes of RSTS patients. Interestingly, in the last there 50 years, there have been two reports of glaucoma in children diagnosed with RSTS; one with juvenile glaucoma176 and one with congenital glaucoma177.

SLC4A11 variant

There are multiple molecular actions proposed for the SLC4A11 protein, including a number of different ion channels and potential aquaporin activities178. Studies have shown that

SLC4A11 is present in the corneal endothelium, however there are no reports of expression in the trabecular meshwork or aqueous outflow pathways178. Dysfunction in this gene has been strongly linked to congenital hereditary endothelial dystrophy (CHED), a rare disorder of the corneal endothelium in which opacification of the cornea may be present at birth178. There are only a few published case reports of glaucoma in patients diagnosed with CHED178,179. Alsaif et al180 recently published whole exome sequencing results of CYP1B1-negative cases of children from

Saudi Arabia where they reported one family with a SLC4A11 mutation which they determined had been misdiagnosed as primary congenital glaucoma.

POMT2 variant

The protein encoded by POMT2 is part of the O-mannosyltransferase (POMT) enzyme complex which plays a role in glycosylation of α-dystroglycan. It is particularly present in the skeletal muscles, testes, and fetal brain. The rare variant found in one family from our sample

(p.Arg659Gln) has been listed in ExAC Browser181 with an allele frequency of 1.65 x 10-5.

Variants in POMT2 have been reported in individuals affected by Walker-Warburg syndrome, or congenital muscular dystrophies that affect the muscles, brain, and anterior portion of the eyes182,183. Eye abnormalities include microphthalmia (small eyes), buphthalmos (enlarged eyes), cataracts, and problems with the optic nerve184. The POMT2 protein is made up of 750 amino acids. The homozygous recessive missense variant in this study results in an amino acid change from a positively charged amino acid (Arg) to a polar amino acid (Gln) at position 659.

B3GALT6 variant

B3GALT6 encodes for an enzyme involved in biosynthesis of glycosaminoglycans

(GAG) which are long chains of unbranched polysaccharides with a repeating disaccharide unit that are the most abundant heteropolysaccharide in the human eye185. GAGs are important for their ability to fill in space, including in the extracellular matrix between cells, and to organize water molecules. They also make up a portion of the cornea, where the transparency is due to uniform distribution of collagen fibrils which are regulated by proteoglycans186,187. Human studies have GAGs play an important role in age-related diseases of the cornea185. B3GALT6 mutations have been reported in spondyloepimetaphyseal dysplasia (SEMD), a group of skeletal disorders, some of which show mild craniofacial dysmorphism including prominent eyes188.

There have not been any reports of variants in B3GALT6 in congenital glaucoma cases. The missense variant identified in our study (p.Thr266Phe) results in an amino acid change in exon

X; the protein has 329 amino acids.

PIEZO2 variant

This gene encodes a large mechanically-activated cation channel with more than thirty transmembrane domains189. Mutations have been reported in patients with Warden-Walker

Syndrome and Distal Arthrogryposis Type 3, both of which may present with ocular abnormalities. A recent study of 14 patients with microphthalmia and/or anophthalmia identified one with a missense variant in PIEZO2190. Similarly, we found a missense mutation resulting in an amino acid change (p.Ile1441Val) occurring approximately halfway through this large protein, which is predicted to be damaging even though both amino acids are hydrophobic.

RXRA variant

Retinoic acid, a metabolite of vitamin A, serves as a signaling molecule during embryonic development of a number of tissues including the eye191. Retinoic acid receptor alpha

(RXRA) is a receptor located in the nucleus that can mediate the biological effects of retinoids191.

Exposure to retinoids, like isotretinoin, early in pregnancy have been implicated with a spectrum of neural crest-related phenotypes including craniofacial malformations and malformations of the

CNS192. The biological mechanism behind these defects may overlap with ocular defects. AFor example, a number of variants near RXRA193 have been associated with central corneal thickness, which have been shown to be lower in glaucoma patients than the general population. This suggests that variants near RXRA can cause ocular phenotypes. The variant found in our study includes an insertion that results in an early termination and a truncation of the RXRA protein.

Stop gain variants are of particular interest because they can have potentially profound impacts on protein structure194.

GLI2 variant

GLI2 encodes a C2H2-type zinc finger which has the ability to act as a mediator of Sonic hedgehog (shh) signaling which is important in embryonic development. Mutations in this gene have been found in patients with holoprosencephaly-like features195. Holoprosencephaly (HPE) is the most common forebrain defect in humans, which can manifest in a failure of separation of

the eye and forebrain195,196. The variant found in this study resulted in an amino acid change from a hydrophobic to an uncharged amino acid toward the end of this protein. PolyPhen-2 predicted this would result in a benign change and the CADD score (CADD= 10.2) indicated that this would be a likely benign change.

POU4F1 variant

The POU4F1 gene encodes a member of the POU-IV class of neural transcription factors that are expressed in a subset of retinal ganglion cells197–199. Currently, there have not been any reports of mutations in POU4F1 and congenital eye defects. This gene was selected because it is important in retinal ganglion cell survival, cell fate, and axonal targeting during development199.

CADD scores and PolyPhen-2 are not calculated for deletions, like the one found in this sample, though the deletion occurs early in the protein (base 237 of 5,199 bases).

ALDH3A2 variant

The ALDH2A2 gene is a member of the aldehyde dehydrogenase family which catalyze the conversion of fatty aldehydes with medium- to very-long-chain fatty acids. According to the

Human Protein Atlas200, ALDH3A2 is not expressed in the eye but is present in a number of other tissues. This gene is frequently mutated in another Mendelian condition called Sjogren-Larsson

Syndrome in which patients present with mental retardation, spastic di- or tetraplegia, and ichthyosis201. The variants discovered in this population were both missense variants that were inherited in a compound heterozygote manner, with one variant predicted to be possibly damaging and another thought to be benign.

USH2A variant

USH2A encodes for a basement membrane protein called usherin which is present in the inner ear and retina. Among other genes, USH2A has been implicated in Usher syndrome which

can include retinitis pigmentosa (RP), an eye condition that can result in gradual vision loss202,203. Although vision loss is reported in both PCG and patients with RP, the age at onset for visual loss in RP patients isn’t until approximately 10 years of age204. The compound heterozygote variants identified in our sample include both a frameshift and a missense variants that result in an amino acid change. The frameshift variant results in an amino acid change from a threonine to a proline which have different structures and chemical properties.

UGT1A8 variant

UGT1A8 encodes an enzyme that is part of the glucuronidation pathway that transforms small lipophilic molecules like steroids, hormones, and water-soluble, excretable metabolites205.

It is associated with other drug metabolism pathways that include cytochrome P450 enzymes (if which CYP1B1 belongs) and is mostly expressed in hepatic cells205,206. Most of the literature describing conditions associated with this gene are in phenotypes that include conditions of the gastrointestinal tract and the liver, such as pericholangitis205 and Gilbert syndrome207. The variants identified in this population are both missense variants. The variant that occurs in approximately the middle of the protein (p.Ala360Val) is predicted to be potentially damaging, whereas the other occurs at amino acid 505 of 530 and is thought to be benign.

Conclusions

There are many strengths to this study, most of which are related to the design of NBDPS which is a multi-year, population-based case-control study which includes participants from 10 different states. These participants reflect racial/ethnic, geographic, and socioeconomic diversity that has been shown to be generally representative of the US population160. Furthermore, methods are rigorous and standardized across centers ensuring thorough and uniform data collection. Individuals who chose to participate in the study were not substantially different than

those who chose not to be interviewed, which further provides support of the representativeness of the study sample112. Clinical review and classification of case infants provides a rigorously defined case sample. This study also benefited from the availability of WES over more traditional genotyping methods that are often used in epidemiological studies. This is an important strength of this paper since genotyping requires a priori hypotheses about important genes, important areas of the genome, or are chosen because of their availability on already validated arrays that are commercially available. Much like the strengths of genome-wide association studies (GWAS)208, WES allows for a comprehensive scan of the the exome in a less biased way to identify important and novel factorsFurther, in this analysis we had WES available from complete trios which allows for high confidence in the inheritance patterns of each of these variants and likely fewer false positive results.

There are some limitations of this study. NBDPS is one of the largest studies of birth defects in the United States, though cases are limited to infants diagnosed within the first year of life. Although close to 80% of infants with PCG are diagnosed within the first year of life1, this means that our case sample might miss close to 20% of cases. Those missing may include milder cases that are not diagnosed until after one year of age or those who are diagnosed later for other reasons (e.g., reduced access to healthcare). Despite a few reports209,210 in the literature of PCG cases being inherited in an autosomal dominant manner, we did not investigate this inheritance pattern in the current analysis. Finally, we did not attempt to replicate these findings in another study sample at this time. Though finding the variants identified within this study in other families would lead to an increased confidence in their importance and thus lead to a higher priority for validation and/or animal studies, these rare variants may be specific to individual

families and difficult to replicate without other large studies with WES findings of which there are few in the United States.

In conclusion, this study confirms previous findings that, although CYP1B1 is the main causative gene for primary congenital glaucoma in some families and that the prevalence of mutations is less common in the US compared to other populations. We identified a number of variants in other genes which have not previously been investigated in association with primary congenital glaucoma but are known to underly other Mendelian conditions with phenotypic overlap or are important in eye development, and have detected mutations that appear to have functional consequences. Additionally, many of the proteins encoded by these genes play important roles in eye deveelopment. Future studies are needed to replicate these findings in cohorts of patients without mutations in known disease-causing genes. Experimental studies to explore the mechanisms associated with the new variants we reported may also be warranted.

CHAPTER 6. – DISCUSSION AND CONCLUSIONS

6.1 Summary of Specific Aims

There are few epidemiologic studies that have evaluated risk factors for PCG and ASDs.

Of the studies that are available, many have occurred in populations with limited generalizability to the United States.

In Aim 1 we evaluated the association between maternal dietary factors and PCG and

ASDs. We estimated associations between intake of specific micronutrients and a measure of diet quality (DQI-P score). Additionally, we used latent class analysis to categorize women into dietary patterns and evaluated whether associations were present between specific dietary patterns and PCG and ASDs.

In Aim 2, we used whole-exome sequencing of a subset of PCG cases to describe the genetic variation in a US-based population. Complete family trios allowed for evaluation of a variety of inheritance patterns. We assessed the prevalence of pathogenic CYP1B1 variants and described variants in genes which hadn’t previously been linked to PCG.

6.1.1 Aim 1

Among individual micronutrients, few odds ratios were statistically significant and most were not monotonic across quartiles. However, we found some inverse, yet imprecise, associations with increasing intake. For example, compared to the lowest quartile of intake, all three quartiles of calcium, choline, and vitamin E were associated with consistently lower odds.

Copper and folate showed lower odds among those in the middle quartiles of intake, but an

increased odds among those in the highest. Both zinc and vitamin A were suggestive of this U- shaped pattern, though aORs in the highest quartile compared to the lowest quartile showed values close tothe null value. There did not appear to be any association between DQI-P score and having a child with PCG or ASDs.

Using latent class analysis we were able to identify the same four dietary patterns as previous NBDPS studies. Those mothers who belonged to the Prudent and Mexican dietary patterns had weakly lower odds ratios compared to mothers in the Western dietary pattern. Mothers in the Low-Calorie Western dietary pattern had similar odds compared to those in the Western dietary pattern.

6.1.2 Aim 2

Similar to previous work, CYP1B1 was found to be the main causative gene for primary congenital glaucoma in our sample, though the overall prevalence was low (13.5%) relative to other studies. All of the CYP1B1 mutations present in our population had been reported before.

One of the families harbored a de novo variant in the FOXC1 gene, a gene which has previously been linked to PCG and ASDs. In families that had no CYP1B1 mutations, there were variants which were predicted to result in moderate-to-severe functional changes in a number of genes which have not previously been investigated in association with primary congenital glaucoma but are known to underlie other Mendelian conditions with potential phenotypic overlap or are known to be involved in eye development. De novo variants were found in CRYBB2, RXRA,

GLI2, POU4F1, PIEZO2, and B3GALT6, one homozygous recessive variant was found in

POMT2, and compound heterozygous changes were found in ALDH3A2, CREBBP, USH2A,

UGT1A8, AND SLC4A11.

6.2 Strengths and Limitations

6.2.1 Strengths

A major strength of this study is the use of the NBDPS study population. NBDPS is a multi-year, population-based case control study which includes participants from 10 different states. This large sample size is required for rare outcomes like PCG and ASDs. These participants show diversity on a number of scales including various racial/ethnic groups, geographic locations, and socioeconomic status. Methods are rigorous and standardized across centers ensuring thorough and uniform data collection on a number of relevant covariates.

Clinical review and classification of case infants provide a rigorously defined case sample for this study. In addition, there are no other US-based studies of PCG and ASDs of this size and scope.

In the analysis for Aim 1, we were able to comprehensively evaluate maternal diet using a variety of measures. Not only did we evaluate individual micronutrient level information, we were able to evaluate a composite score of more than one nutrient at a time which also incorporated dietary recommendations for pregnant women. Similarly, we utilized a data-driven approach which allows for interactions between multiple nutrients, but doesn’t measure against a prevailing hypothesis of what defines quality foods.

One of the major strengths of Aim 2 is that whole-exome sequencing data was used to evaluate genetic variation in these individuals. Candidate gene and GWAS approaches, which are often used in large epidemiologic studies of birth defects, require researchers to have a prior hypothesis about specific genes, and regions of genes, to investigate. Whole-exome sequencing does not require this, and allows for an evaluation of all of the protein coding (exonic) regions of the genome. Only samples from complete trios, where samples from each individual passed quality 78

control measures, underwent whole-exome sequencing. This allowed for high confidence in the inheritance patterns because we have the genomic sequence of each unaffected parent. We chose to focus on variants that were predicted to have a medium-to-high functional impact on the protein based on multiple software sources (CADD-scores, PolyPhen, etc).

6.2.2 Limitations

Despite the size of NBDPS, the rarity of PCG and ASDs still result in a relatively small number of cases and imprecise effect estimates. In addition, NBDPS cases are limited to infants diagnosed within the first year of life. Although approximately 80% of PCG infants are diagnosed within the first year of life1, our case sample may be incomplete and could include infants with more severe forms or those who are diagnosed early for other reasons (e.g., better access to healthcare). Finally, as such little is known about the risk factors for PCG and ASDs, there may be unmeasured confounding. This may further be complicated by heterogeneity in underlying genetic susceptibility. The most consistently mutated gene in congenital glaucoma is CYP1B1, with many of the mutations being described in different ethnic populations69. These patients may develop congenital glaucoma regardless of non-genetic exposures. Ideally, future studies should evaluate potential modification by variant/gene status when investigating non-genetic exposures like maternal diet. This should be a high priority for researchers.

Specific to Aim 1, the FFQ is a validated measure of dietary intake, however this study still includes self-reported data on each mother’s diet in the year before pregnancy. Due to the period of time between the maternal interview and the time period of interest, there is the possibility of misclassification for specific foods and/or micronutrients. Further, some foods may be believed to be healthier than others, potentially resulting in differential recall bias between case and control mothers. However, if case mothers are more likely to report some of these foods (e.g. fruits, 79

vegetables, whole grains), we would have seen more associations above one for increasing micronutrient intake than what we observed. We also chose to combine both PCG and ASD case groups for Aim 1 because of the suggestion of a shared etiology and to improve our power to observe associations. Despite sub-analyses which split case groups apart and showed similar patterns in association measures, there is a possibility that we did not have the precision in these small individual case groups to detect different associations between the groups.

Despite having whole-exome sequencing data on complete trios, we were further limited when looking for rare variants in a subset (n=38) of PCG cases from NBDPS in Aim 2. While this is one of the larger studies of whole-exome sequencing data from complete families of PCG cases, we found a number of variants that were only present in individual families. In addition, in this study we only focused on variants that would have been inherited assuming a compound heterozygote, de novo, or autosomal recessive pattern. X-linked de novo and X-linked recessive patterns were also investigated but did not identify any genes that were known to underlie a

Mendelian condition with potential phenotypic overlap or were known to be important in eye development. Autosomal dominant inheritance patterns have been reported in PCG210 but were not investigated in this analysis. Using the same statistical methods to prioritize variants outlined in

Aim 2, all dominant alleles inherited from either parent would be identified for further investigation which would likely result in a large number of false positive variants.

6.3 Public Health Implications and Future Studies

6.3.1 Public Health Impact

There is very little work on the risk factors for PCG and ASDs in a racially, ethnically, and geographically diverse population like that in the United States. Our findings from Aim 1 add to the accumulating evidence of the importance of maternal diet during pregnancy. In addition, in 80

Aim 2 where we restricted to only PCG cases, we identified some novel genes which previously have not been investigated for their role in PCG. These discoveries may improve the accuracy of risk assessment information, identify potential therapeutic targets, and provide information about the biological mechanisms underlying this birth defect.

6.3.2 Future Studies

There are multiple directions for further research in both Aims of this project. This is the first report of the association between maternal diet and PCG and ASD, therefore other studies are needed to replicate these findings. Larger studies with more cases, or meta-analytic methods to combine estimates from multiple studies, will allow for more precise estimates of association.

In addition, future studies could attempt to explore new methods of measuring food intake or better quantify bioavailability and/or metabolism of individual micronutrients, especially those that would translate into large, case-control studies. Other studies which have whole-exome sequencing data of patients without disease-causing mutations should replicate the findings of in

Aim 2. In addition, experimental studies to explore the mechanisms associated with the new variants we reported, or evaluating the role of these specific genes, are also warranted. In silico functional studies, tissue culture studies, and proteomics, similar to the evaluation of TEK variants in Souma et al210, could be employed. The CRISPR-Cas 9 system211 could be used to recreate these variants in animal models which could then be evaluated for phenotypic similarities with PCG cases.

6.4 Conclusions

In conclusion, this study was able to determine that increasing amounts of specific micronutrients and a Prudent or Mexican dietary pattern may be associated with a decreased odds of PCG and ASDs. As this is the first report of the association between maternal nutrition and PCG 81

and ASDs, other investigations are needed to replicate these findings. However, these findings, along with previous research, support the continued recommendation of diets high in fruits and vegetables, like the Prudent and Mexican dietary patterns, among pregnant women. We found that

CYP1B1 was the most commonly mutated gene among PCG cases in our population, but the overall prevalence was low. We also identified a number of other genes which may be etiologically relevant in PCG cases from the United States. Other studies which have individuals without known disease-causing mutations should evaluate whether variations in these genes are present.

APPENDIX 1. SUPPLEMENTARY TABLES FOR CHAPTER 4

Supplemental Table 1. ORs and 95% CIs for individual nutrients for combined primary congenital glaucoma and anterior segment defects, stratified by periconceptional vitamin use

Non- Non- Vitamin Vitamin Cases Controls Vitamin Vitamin Cases Controls Nutrient Users Users (%) (%) Users Users (%) (%) ORa 95% CI ORa 95% CI Calcium

(mg) 8 554 41 1720 1 1 Q1 (25.8) (27.6) (34.2) (24.3) 11 449 0.65, 28 1824 0.40, 1.68 0.65 Q2 (35.5) (22.3) 4.32 (23.3) (25.8) 1.07 4 447 0.16, 31 1793 0.44, 0.57 0.74 Q3 (12.9) (23.7) 2.04 (25.8) (25.4) 1.24 8 531 0.29, 20 1736 0.26, 1.01 0.5 Q4 (25.8) (26.4) 3.55 (16.7) (24.5) 0.98 Betaine

(mg) 4 601 32 1673 1 1 Q1 (12.9) (29.9) (26.7) (23.7) 11 439 1.27, 29 1834 0.53, 4.06 0.89 Q2 (35.5) (21.8) 12.94 (24.2) (25.9) 1.48 4 454 0.37, 28 1813 0.54, 1.54 0.93 Q3 (12.9) (22.6) 6.33 (23.3) (25.6) 1.58 12 517 1.35, 31 1753 0.68, 4.69 1.19 Q4 (38.7) (25.7) 16.28 (25.8) (24.8) 2.10 Choline,

total 6 478 43 1797 1 1 Q1 (19.4) (23.8) (35.8) (25.4) 10 405 0.68, 27 1864 0.38, 1.95 0.63 Q2 (32.3) (20.1) 5.56 (22.5) (26.4) 1.05 6 488 0.28, 26 1782 0.38, 0.96 0.66 Q3 (19.4) (24.3) 3.27 (21.7) (25.2) 1.15 9 640 0.26, 24 1630 0.34, 1.08 0.72 Q4 (29.0) (31.8) 4.50 (20.0) (23.1) 1.49 Copper

(mg) 5 465 41 1768 1 1 Q1 (16.1) (23.1) (34.2) (25.0) 10 401 0.79, 32 1907 0.47, 2.39 0.77 Q2 (32.3) (19.9) 7.23 (26.7) (27.0) 1.26 6 469 0.35, 20 1805 0.29, 1.26 0.54 Q3 (19.4) (23.3) 4.50 (16.7) (25.5) 1.00 10 676 0.38, 27 1593 0.45, 1.52 0.93 Q4 (32.3) (33.6) 6.06 (22.5) (22.5) 1.92 83

Iron (mg) 8 456 35 1821 1 1 Q1 (25.8) (22.7) (29.2) (25.8) 8 478 0.37, 36 1797 0.69, 1.02 1.13 Q2 (25.8) (23.8) 2.81 (30.0) (25.4) 1.83 444 0.06, 24 1818 0.45, 2 (6.5) 0.29 0.78 Q3 (22.1) 1.41 (20.0) (25.7) 1.37 13 633 0.46, 25 1637 0.53, 1.43 1.02 Q4 (41.9) (31.5) 4.43 (20.8) (23.1) 1.96 Magnesium

(mg) 9 500 42 1776 1 1 Q1 (29.0) (24.9) (35.0) (25.1) 7 436 0.32, 30 1831 0.44, 0.89 0.72 Q2 (22.6) (21.7) 2.49 (25.0) (25.9) 1.18 5 452 0.19, 23 1822 0.32, 0.62 0.58 Q3 (16.1) (22.5) 2.03 (19.2) (25.8) 1.03 10 623 0.24, 25 1644 0.35, 0.9 0.75 Q4 (32.3) (31.0) 3.36 (20.8) (23.2) 1.62 Riboflavin

(mg) 7 530 39 1746 1 1 Q1 (22.6) (26.4) (32.5) (24.7) 8 417 0.56, 31 1848 0.47, 1.58 0.77 Q2 (25.8) (20.7) 4.48 (25.8) (16.1) 1.26 4 454 0.22, 28 1804 0.43, 0.8 0.73 Q3 (12.9) (22.6) 2.93 (23.3) (25.5) 1.26 12 610 0.62, 22 1675 0.33, 2.15 0.66 Q4 (38.7) (30.3) 7.43 (18.3) (23.7) 1.33 Vitamin B 6 (mg) 6 473 42 1816 1 1 Q1 (19.4) (23.5) (35.0) (25.7) 8 441 0.57, 27 1829 0.39, 1.68 0.65 Q2 (25.8) (21.9) 4.98 (22.5) (25.9) 1.07 474 0.09, 31 1784 0.45, 2 (6.5) 0.45 0.77 Q3 (23.6) 2.38 (25.8) (25.2) 1.32 15 623 1.02, 20 1644 0.26, 3.48 0.55 Q4 (48.4) (31.0) 11.87 (16.7) (23.2) 1.16 Vitamin

B12 (µg) 6 478 33 1806 1 1 Q1 (19.4) (23.8) (27.5) (25.5) 7 457 0.45, 38 1809 0.76, 1.36 1.23 Q2 (22.6) (22.7) 4.12 (31.7) (25.6) 1.99 469 0.15, 24 1802 0.47, 3 (9.7) 0.64 0.83 Q3 (23.3) 2.69 (20.0) (25.5) 1.45 15 607 0.92, 25 1656 0.55, 2.83 1.05 Q4 (48.4) (30.2) 8.67 (20.8) (23.4) 1.98 Vitamin E

(mg) 84

7 479 43 1793 1 1 Q1 (22.6) (23.8) (35.8) (25.4) 10 446 0.55, 29 1825 0.42, 1.5 0.68 Q2 (32.3) (22.2) 4.05 (24.2) (25.8) 1.11 6 494 0.24, 24 1784 0.34, 0.79 0.59 Q3 (19.4) (24.6) 2.56 (20.0) (25.2) 1.03 8 592 0.22, 24 1671 0.34, 0.84 0.66 Q4 (25.8) (29.4) 3.13 (20.0) (23.6) 1.27 Zinc (mg) 8 527 43 1750 1 1 Q1 (25.8) (26.2) (35.8) (24.7) 7 414 0.44, 24 1856 0.33, 1.27 0.55 Q2 (22.6) (20.6) 3.64 (20.0) (26.2) 0.93 484 0.13, 28 1786 0.40, 3 (9.7) 0.52 0.7 Q3 (24.1) 2.14 (23.3) (25.3) 1.22 13 586 0.68, 25 1681 0.35, 2.3 0.72 Q4 (41.9) (29.1) 7.80 (20.8) (23.8) 1.48 Vitamin A,

RAE (µg) 10 519 39 1758 1 1 Q1 (32.3) (25.8) (32.5) (24.9) 4 464 0.14, 27 1806 0.43, 0.47 0.71 Q2 (12.9) (23.1) 1.52 (22.5) (25.5) 1.18 5 430 0.21, 27 1839 0.44, 0.65 0.75 Q3 (16.1) (21.4) 2.01 (22.5) (26.0) 1.27 12 598 0.43, 27 1670 0.50, 1.21 0.92 Q4 (38.7) (29.7) 3.35 (22.5) (23.6) 1.69 Carotene, alpha (µg) 10 657 36 1613 1 1 Q1 (32.3) (32.7) (30.0) (22.8) 9 427 0.56, 36 1844 0.55, 1.38 0.88 Q2 (29.0) (21.2) 3.44 (30.0) (26.1) 1.41 5 427 0.26, 21 1844 0.31, 0.77 0.53 Q3 (16.1) (21.2) 2.27 (17.5) (26.1) 0.91 7 500 0.34, 27 1772 0.44, 0.92 0.75 Q4 (22.6) (24.9) 2.48 (22.5) (25.1) 1.26 Carotene, beta (µg) 10 611 32 1660 1 1 Q1 (32.3) (30.4) (26.7) (23.5) 10 467 0.54, 34 1808 0.62, 1.31 1 Q2 (32.3) (23.2) 3.18 (28.3) (25.6) 1.63 4 456 0.17, 27 1810 0.49, 0.54 0.82 Q3 (12.9) (22.7) 1.73 (22.5) (25.6) 2.39 7 477 0.32, 27 1795 0.51, 0.9 0.89 Q4 (22.6) (23.7) 2.50 (22.5) (25.4) 1.53 Retinol

(µg) 8 475 33 1799 1 1 Q1 (25.8) (23.6) (27.5) (25.4)

6 456 0.28, 30 1819 0.58, 0.82 0.96 Q2 (19.4) (22.7) 2.41 (25.0) (25.7) 1.60 4 444 0.17, 33 1823 0.67, 0.59 1.12 Q3 (12.9) (22.1) 2.03 (27.5) (25.8) 1.88 13 636 0.50, 24 1632 0.56, 1.43 1.04 Q4 (41.9) (31.6) 4.04 (20.0) (23.1) 1.93 Folate,

DFE (µg) 8 479 35 1801 1 1 Q1 (25.8) (23.8) (29.2) (25.5) 7 482 0.32, 28 1784 0.53, 0.91 0.88 Q2 (22.6) (24.0) 2.58 (23.3) (25.2) 1.48 5 472 0.21, 30 1800 0.59, 0.69 1 Q3 (16.1) (23.5) 2.24 (25.0) (25.5) 1.69 11 578 0.43, 27 1688 0.59, 1.33 1.08 Q4 (35.5) (28.7) 4.07 (22.5) (23.9) 1.98 DQI-P 6 454 28 1581 1 1 Q1 (19.4) (22.8) (23.5) (22.5) 9 429 0.55, 38 1764 0.77, 1.58 1.27 Q2 (29.0) (21.5) 4.54 (31.9) (25.1) 2.09 8 497 0.40, 29 1806 0.57, 1.21 0.99 Q3 (25.8) (24.9) 3.68 (24.4) (25.7) 1.73 8 615 0.28, 24 1881 0.44, 0.97 0.84 Q4 (25.8) (30.8) 3.34 (20.2) (26.8) 1.62 aModel was adjusted for energy intake (continuous)

Supplemental Table 2. ORs and 95% CIs for individual nutrients, stratified by case group status

Controls PCG Controls ASD PCG PCG ASD ASD (%) cases (%) cases Nutrient Cases cases Cases cases 95% 95% (%) ORa (%) ORa CI CI Calcium

(mg) 28 2295 21 2295 1 1 Q1 (30.8) (25.0) (34.4) (25.0) 20 2294 0.40, 19 2294 0.46, 0.73 0.88 Q2 (22.0) (25.0) 1.31 (31.2) (25.0) 1.67 22 2295 0.45, 13 2295 0.28, 0.82 0.59 Q3 (24.2) (25.0) 1.49 (21.3) (25.0) 1.25 21 2294 0.40, 2294 0.13, 0.81 8 (13.1) 0.35 Q4 (23.1) (25.0) 1.65 (25.0) 0.94 Betaine

(mg) 20 2295 16 2295 1 1 Q1 (22.0) (25.0) (26.2) (25.0) 26 2295 0.76, 14 2295 0.46, 1.37 0.94 Q2 (28.6) (25.0) 2.48 (23.0) (25.0) 1.95 16 2293 0.46, 16 2293 0.57, 0.9 1.17 Q3 (17.6) (25.0) 1.77 (26.2) (25.0) 2.40 29 2295 0.95, 15 2295 0.56, 1.8 1.25 Q4 (31.9) (25.0) 3.43 (24.6) (25.0) 2.77 Choline,

total 26 2295 23 2295 1 1 Q1 (37.7) (25.0) (37.7) (25.0) 21 2294 0.48, 16 2294 0.34, 0.87 0.66 Q2 (23.1) (25.0) 1.58 (26.2) (25.0) 1.29 20 2295 0.46, 12 2295 0.22, 0.88 0.47 Q3 (22.0) (25.0) 1.69 (19.7) (25.0) 1.04 24 2294 0.55, 10 2294 0.12, 1.22 0.35 Q4 (26.4) (25.0) 2.71 (16.4) (25.0) 1.03 Copper

(mg) 27 2253 19 2253 1 1 Q1 (29.7) (24.6) (31.2) (24.6) 23 2333 0.48, 19 2333 0.53, 0.86 1.03 Q2 (25.3) (25.4) 1.54 (31.2) (25.4) 2.00 17 2297 0.34, 2297 0.22, 0.67 9 (14.8) 0.52 Q3 (18.7) (25.0) 1.33 (25.0) 1.27 24 2295 0.46, 14 2295 0.33, 1.03 0.91 Q4 (26.4) (25.0) 2.28 (23.0) (25.0) 2.48

Iron (mg) 24 2294 19 2294 1 1 Q1 (26.4) (25.0) (31.2) (25.0)

24 2295 0.62, 20 2295 0.55, 1.1 1.07 Q2 (26.4) (25.0) 1.98 (32.8) (25.0) 2.06 14 2291 0.34, 12 2291 0.30, 0.69 0.65 Q3 (15.4) (25.0) 1.37 (19.7) (25.0) 1.41 29 2298 0.84, 10 2298 0.21, 1.66 0.55 Q4 (31.9) (25.0) 3.29 (16.4) (25.0) 1.45 Magnesium

(mg) 27 2295 24 2295 1 1 Q1 (29.7) (25.0) (39.3) (25.0) 23 2294 14 2294 0.29, 0.89 1.59 0.59 Q2 (25.3) (25.0) (23.0) (25.0) 1.17 18 2295 0.37, 10 2295 0.18, 0.72 0.42 Q3 (19.8) (25.0) 1.41 (16.4) (25.0) 0.96 23 2294 0.44, 13 2294 0.29, 1.01 0.55 Q4 (25.3) (25.0) 2.33 (21.3) (25.0) 1.55 Riboflavin

(mg) 26 2292 20 2292 1 1 Q1 (28.6) (25.0) (32.8) (25.0) 23 2291 0.53, 16 2291 0.41, 0.94 0.8 Q2 (25.3) (25.0) 1.67 (26.2) (25.0) 1.57 16 2287 0.36, 2287 0.38, 0.7 16 (6.2) 0.8 Q3 (17.6) (24.9) 1.37 (24.9) 1.66 26 2308 0.62, 2308 0.16, 1.28 9 (14.8) 0.44 Q4 (28.6) (25.2) 2.67 (25.2) 1.24 Vitamin B 6 (mg) 28 2310 20 2310 1 1 Q1 (30.8) (25.2) (32.8) (25.2) 21 2290 0.44, 14 2290 0.37, 0.79 0.75 Q2 (23.1) (25.0) 1.42 (23.0) (25.0) 1.52 18 2281 0.37, 15 2281 0.40, 0.71 0.85 Q3 (19.8) (24.9) 1.36 (24.6) (24.9) 1.81 24 2297 0.48, 12 2297 0.28, 1.03 0.75 Q4 (26.4) (25.0) 2.21 (19.7) (25.0) 2.03 Vitamin B 12 (µg) 22 2303 17 2303 1 1 Q1 (24.2) (25.1) (27.9) (25.1) 27 2289 0.75, 18 2289 0.57, 1.34 1.13 Q2 (24.9) (24.9) 2.39 (29.5) (24.9) 2.22 12 2297 0.31, 15 2297 0.47, 0.64 0.98 Q3 (13.2) (25.0) 1.34 (24.6) (25.0) 2.08 30 2289 0.94, 11 2289 0.32, 1.84 0.79 Q4 (33.0) (24.9) 3.60 (18.0) (24.9) 1.97 Vitamin E

(mg) 32 2290 18 2290 1 1 Q1 (35.2) (25.0) (29.5) (25.0) 17 2300 0.29, 22 2300 0.64, 0.54 1.22 Q2 (18.7) (25.1) 0.98 (36.1) (25.1) 2.33 88

19 2297 0.33, 11 2297 0.27, 0.61 0.62 Q3 (20.9) (25.0) 1.13 (18.0) (25.0) 1.29 23 2291 0.37, 10 2291 0.21, 0.76 0.57 Q4 (25.3) (25.0) 1.54 (16.4) (25.0) 1.52

Zinc (mg) 29 2297 22 2297 1 1 Q1 (31.9) (25.0) (36.1) (25.0) 16 2293 0.32, 15 2293 0.35, 0.6 0.69 Q2 (17.6) (25.0) 1.14 (24.6) (25.0) 1.37 19 2290 0.42, 12 2290 0.25, 0.78 0.56 Q3 (20.9) (25.0) 1.50 (19.7) (25.0) 1.24 27 2298 0.62, 12 2298 0.21, 1.3 0.57 Q4 (29.7) (25.0) 2.75 (19.7) (25.0) 1.57 Vitamin A,

RAE (µg) 29 2295 20 2295 1 1 Q1 (31.9) (25.0) (32.8) (25.0) 17 2294 0.34, 14 2294 0.36, 0.63 0.72 Q2 (18.7) (25.0) 1.15 (23.0) (25.0) 1.44 14 2295 0.28, 18 2295 0.47, 0.55 0.94 Q3 (15.4) (25.0) 1.08 (29.5) (25.0) 1.87 31 2294 0.74, 2294 0.20, 1.39 9 (14.8) 0.49 Q4 (34.1) (25.0) 2.60 (25.0) 1.26 Carotene, alpha (µg) 29 2294 17 2294 1 1 Q1 (31.9) (25.0) (27.9) (25.0) 28 2295 0.57, 17 2295 0.51, 0.97 1 Q2 (30.8) (25.0) 1.63 (27.9) (25.0) 1.97 11 2294 0.19, 15 2294 0.45, 0.38 0.91 Q3 (12.1) (25.0) 0.77 (24.6) (25.0) 1.82 23 2295 0.47, 12 2295 0.36, 0.82 0.78 Q4 (25.3) (25.0) 1.45 (19.7) (25.0) 1.66 Carotene, beta (µg) 28 2294 14 2294 1 1 Q1 (30.8) (25.0) (23.0) (25.0) 24 2295 0.50, 20 2295 0.74, 0.87 1.47 Q2 (26.4) (25.0) 1.50 (32.8) (25.0) 2.91 16 2295 0.31, 15 2295 0.54, 0.59 1.14 Q3 (17.6) (25.0) 1.09 (24.6) (25.0) 2.38 23 2294 0.48, 12 2294 0.44, 0.87 0.98 Q4 (25.3) (25.0) 1.56 (19.7) (25.0) 2.21 Retinol

(µg) 25 2295 16 2295 1 1 Q1 (27.5) (25.0) (26.2) (25.0) 17 2294 0.39, 19 2294 0.63, 0.73 1.23 Q2 (18.7) (25.0) 1.37 (31.2) (25.0) 2.43 20 2295 0.49, 17 2295 0.55, 0.92 1.14 Q3 (22.0) (25.0) 1.70 (27.9) (25.0) 2.35

29 2294 0.80, 2294 0.25, 1.51 9 (14.8) 0.65 Q4 (31.9) (25.0) 2.86 (25.0) 1.66 Folate,

DFE (µg) 26 2294 17 2294 1 1 Q1 (28.8) (25.0) (27.9) (25.0) 18 2295 0.41, 17 2295 0.53, 0.75 1.06 Q2 (19.8) (25.0) 1.39 (27.9) (25.0) 2.11 18 2294 0.42, 17 2294 0.53, 0.79 1.1 Q3 (19.8) (25.0) 1.50 (27.9) (25.0) 2.27 29 2295 0.76, 10 2295 0.28, 1.44 0.71 Q4 (31.9) (25.0) 2.74 (16.4) (25.0) 1.78

DQI-P

22 2056 12 2056 1 1 Q1 (24.4) (22.6) (19.7) (22.6) 23 2206 0.55, 24 2206 0.95, 0.99 1.93 Q2 (25.6) (24.2) 1.80 (39.3) (24.2) 3.91 23 2334 0.52, 14 2334 0.49, 0.96 1.11 Q3 (25.6) (25.6) 1.78 (23.0) (25.6) 2.50 22 2522 0.43, 11 2522 0.33, 0.88 0.86 Q4 (24.4) (27.7) 1.78 (18.0) (27.7) 2.23 aModel was adjusted for energy intake (continuous)

REFERENCES

1. Allingham R, Damji K, Freedman S, Mori S, Shafranov G. Shield’s Textbook of Glaucoma. In: Allingham R, Damji K, Freedman S, Mori S, Shafranov G, eds. Shield’s Textbook of Glaucoma. 5th ed. Philadelphia: Lippincott Williams and Wilkins; 2005:235- 252.

2. DeLuise VP, Anderson DR. Primary infantile glaucoma (Congenital glaucoma). Surv Ophthalmol. 1983;28(1):1-19.

3. Alanazi FF, Song JC, Mousa A, et al. Primary and secondary congenital glaucoma: Baseline features from a registry at King Khaled Eye Specialist Hospital, Riyadh, Saudi Arabia. Am J Ophthalmol. 2013;155(5):882-889.

4. Clark RA, Kozak A, Epley KD. Glaucoma, Congenital Or Infantile. EyeWiki. https://eyewiki.aao.org/Glaucoma,_Congenital_Or_Infantile. Published 2017. Accessed December 1, 2018.

5. Ito YA, Walter MA. Genomics and anterior segment dysgenesis: A review. Clin Exp Ophthalmol. 2014;42(1):13-24.

6. Lewis CJ, Hedberg-Buenz A, DeLuca AP, Stone EM, Alward WLM, Fingert JH. Primary congenital and developmental glaucomas. Hum Mol Genet. 2017;26(R1):R28-R36.

7. Stoilov I, Akarsu AN, Sarfarazi M. Identification of three different truncating mutations in cytochrome P4501B1 (CYP1B1) as the principal cause of primary congenital glaucoma (Buphthalmos) in families linked to the GLC3A locus on chromosome 2p21. Hum Mol Genet. 1997;6(4):641-647.

8. Ali M, McKibbin M, Booth A, et al. Null mutations in LTBP2 cause primary congenital glaucoma. Am J Hum Genet. 2009;84(5):664-671.

9. Plasilova M, Ferakova E, Kadasi L, et al. Linkage of autosomal recessive primary congenital glaucoma to the GLC3A locus in Roms (Gypsies) from Slovakia. Hum Hered. 1998;48(1):30-33.

10. Lim S-H, Tran-Viet K-N, Yanovitch TL, et al. CYP1B1, MYOC, and LTBP2 mutations in primary congenital glaucoma patients in the United States. Am J Ophthalmol. 2013;155(3):508-517.

11. Narooie-Nejad M, Paylakhi SH, Shojaee S, et al. Loss of function mutations in the gene encoding latent transforming growth factor beta binding protein 2, LTBP2, cause primary congenital glaucoma. Hum Mol Genet. 2009;18(20):3969-3977.

12. Yoo TK, Oh E, Hong S. Is vitamin D status associated with open-angle glaucoma? A cross-sectional study from South Korea. Public Health Nutr. 2014;17(4):833-843.

13. Nguyen CTO, Bui B V., Sinclair AJ, Vingrys AJ. Dietary omega 3 fatty acids decrease intraocular pressure with age by increasing aqueous outflow. Investig Ophthalmol Vis Sci. 2007;48(2):756-762.

14. Kim HT, Kim JM, Kim JH, et al. The relationship between vitamin D and glaucoma: A Kangbuk Samsung Health Study. Korean J Ophthalmol. 2016;30(6):426-433.

15. Nelson M, Evans H. Pteroylglutamic acid and reproduction in the rat. J Nutr. 1949;38(1):11-24.

16. Nelson M, Wright H, Asling H, Evans H. Multiple congenital abnormalities resulting from transitory deficiency of pteroylglutamic acid during gestation in the rat. J Nutr. 1955;56(3):349-69.

17. Baird C, Nelson M. Congenital cardiovascular anomalies induced by pteroylglutamic acid deficiency during gestation in the rat. Circ Res. 1954;2(6):544-554.

18. Fernandes KA, Harder JM, Williams PA, et al. Using genetic mouse models to gain insight into glaucoma: Past results and future possibilities. Exp Eye Res. 2015;141:42-56.

19. Russel R, Cox M, Solomons N. Zinc and the special senses. Ann Intern Med. 1983;99(2):227-239.

20. Nielsen J, Carlton W. Colobomatous microphthalmos in a New Zealand white rabbit, arising from a colongy with suspected vitamin E deficiency. Lan Anim Sci. 1995;45(3):320-322.

21. Zhao Y, Wang S, Sorenson CM, et al. Cyp1b1 mediates periostin regulation of trabecular meshwork development by suppression of oxidative stress. Mol Cell Biol. 2013;33(21):4225-4240.

22. Feldkamp ML, Krikov S, Botto LD, Shaw GM, Carmichael SL. Better diet quality before pregnancy is associated with reduced risk of gastroschisis in Hispanic women. J Nutr. 2014;144(11):1781-1786.

23. Sotres-Alvarez D, Siega-riz AM, Herring AH, et al. Maternal dietary patterns are associated with risk of neural tube and congenital heart defects. Am J Epidemiol. 2013;177(11):1279-1288.

24. Larsen WJ. Development of the Eyes. In: Sherman LS, Potter SS, Scott WJ, eds. Human Embryology. Third. Philadelphia: Churchill Livingstone; 2001:379-389.

25. Braunger BM, Fuchshofer R, Tamm ER. The aqueous humor outflow pathways in glaucoma: A unifying concept of disease mechanisms and causative treatment. Eur J Pharm Biopharm. 2015;95(Pt B):173-181.

26. Goel M, Picciani RG, Lee RK, Bhattacharya SK. Aqueous humor dynamics: A review. Open Ophthalmol J. 2010;4:52-59.

27. Weinreb R, Aung T, Medeiros F. The pathophysiology and treatment of glaucoma. J Am Med Assoc. 2014;311(18):1901-1911.

28. Tamm ER, Ethier CR, Dowling JE, et al. Biological aspects of axonal damage in glaucoma: A brief review. Exp Eye Res. 2017;157:5-12.

29. Syc-Mazurek SB, Fernandes KA, Libby RT. JUN is important for ocular hypertension- induced retinal ganglion cell degeneration. Cell Death Dis. 2017;8(7):e2945.

30. Li X, Zhao F, Xin M, et al. Regulation of intraocular pressure by microRNA cluster miR- 143/145. Sci Rep. 2017;7(1):915.

31. Boland MV, Ervin A-M, Friedman DS JH et al. Comparative effectiveness of treatments for open-angle glaucoma: A systematic review for the U.S. Preventive Services Task Force. Ann Intern Med. 2013;158(4):271-279.

32. Zagora SL, Funnell CL, Martin FJ, et al. Primary congenital glaucoma outcomes: Lessons from 23 years of follow-up. Am J Ophthalmol. 2015;159(4):788-796.

33. Tan Y-L, Chua J, Ho C-L. Updates on the surgical management of pediatric glaucoma. Asia Pac J Ophthalmol (Phila). 2016;5(1):85-92.

34. Mandal A, Netland, P. The Pediatric Glaucomas. Elsevier Inc; 2006.

35. Moore DB, Tomkins O, Ben-Zion I. A review of primary congenital glaucoma in the developing world. Surv Ophthalmol. 2013;58(3):278-285.

36. Ho CL, Walton DS. Primary congenital glaucoma: 2004 update. J Pediatr Ophthalmol Strabismus. 2004;41(5):271-88.

37. Shaffer R. Genetics and the congenital glaucomas. Am J Ophthalmol. 1965;60(6):981-994.

38. Gilbert C, Foster A. Childhood blindness in the context of VISION 2020 - The right to sight. Bull World Health Organ. 2001;79(3):227-232.

39. Abdolrahimzadeh S, Fameli V, Mollo R, Contestabile MT, Perdicchi A, Recupero SM. Rare diseases leading to childhood glaucoma: Epidemiology, pathophysiogenesis, and management. Biomed Res Int. 2015;2015:11 pages.

40. Omoti AE, Uhumwangho OM. Problems of management of primary congenital glaucoma in Benin City, Nigeria. Niger Postgrad Med J. 2007;14(4):310—313.

41. Onwasigwe EN, Ezegwui IR, Onwasigwe CN, Aghaji AE. Management of primary congenital glaucoma by trabeculectomy in Nigeria. Ann Trop Paediatr. 2008;28(1):49-52.

42. Ellong A, Mvogo CE, Bella-Hiag AL, Mourney EN, Ngosso A, Litumbe CN. Prevalence of glaucomas in a Black Cameroonian population. Sante. 2006;16(2):83-88.

43. Bejjani BA, Lewis RA, Tomey KF, et al. Mutations in CYP1B1, the gene for cytochrome P4501B1, are the predominant cause of primary congenital glaucoma in Saudi Arabia. Am J Hum Genet. 1998;62(2):325-333.

44. Reis LM, Semina E V. Genetics of anterior segment dysgenesis disorders. Curr Opin Ophthalmol. 2011;22(5):314-324.

45. Kurilec JM, Zaidman GW. Incidence of Peters Anomaly and congenital corneal opacities interfering with vision in the United States. Cornea. 2014;33:848-850.

46. Hanson IM, Fletcher JM, Jordan T, et al. Mutations at the PAX6 locus are found in heterogeneous anterior segment malformations including Peters’ anomaly. Nat Genet. 1994;6(2):168-173.

47. Waring G, Rodrigues M, Laibson P. Anterior chamber cleavage syndrome. A stepladder classification. Surv Ophthalmol. 1975;20(1):3-27.

48. Weisschuh N, Wolf C, Wissinger B, Gramer E. A novel mutation in the FOXC1 gene in a family with Axenfeld – Rieger syndrome and Peters’ anomaly. Clin Genet. 2008;74(5):476-480.

49. Weisschuh N, Dressler P, Schuettauf F, Wolf C, Wissinger B, Gramer E. Novel mutations of FOXC1 and PITX2 in patients with Axenfeld-Rieger malformations. Invest Ophthalmol Vis Sci. 2006;47(9):3846-3852.

50. Vincent A, Billingsley G, Priston M, et al. Phenotypic heterogeneity of CYP1B1: Mutations in a patient with Peters’ anomaly. J Med Genet. 2001;38(5):324-326.

51. Vincent A, Billingsley G, Priston M, et al. Further support of the role of CYP1B1 in patients with Peters anomaly. Mol Vis. 2006;12:506-510.

52. Alward WLM. Axenfeld-Rieger syndrome in the age of molecular genetics. Am J Ophthalmol. 2000;130(1):107-115.

53. Shields MB. Axenfeld-Rieger syndrome: A theory of mechanism and distinctions from the iridocorneal endothelial syndrome. Trans Am Ophthalmol Soc. 1983;81:736-784.

54. Doward W, Perveen R, Lloyd IC, Ridgway AE, Wilson L, Black GC. A mutation in the RIEG1 gene associated with Peters’ anomaly. J Med Genet. 1999;36(2):152-155.

55. Pasutto F, Mauri L, Popp B, et al. Whole exome sequencing reveals a novel de novo FOXC1 mutation in a patient with unrecognized Axenfeld-Rieger syndrome and glaucoma. Gene. 2015;568(1):76-80.

56. Hogan BL, Horsburgh G, Cohen J, Hetherington CM, Fisher G, Lyon MF. Small eyes (Sey): a homozygous lethal mutation on chromosome 2 which affects the differentiation of both lens and nasal placodes in the mouse. J Embryol Exp Morphol. 1986;97:95-110.

57. Quiring R, Walldorf U, Kloter U, Gehring WJ. Homology of the eyeless gene of drosophila to the small eye gene in mice and aniridia in humans. Science. 1994;265(5173):785-789.

58. Cvekl A, Callaerts P. PAX6: 25th anniversary and more to learn. Exp Eye Res. 2017;156:10-21.

59. Onwochei BC, Simon JW, Bateman JB, Couture KC, Mir E. Ocular colobomata. Surv Ophthalmol. 2000;45(3):175-194.

60. Hornby SJ, Adolph S, Gilbert CE, Dandona L, Foster A. Visual acuity in children with coloboma: Clinical features and a new phenotypic classification system. Ophthalmology. 2000;107(3):511-520.

61. Bavbek T, Ogut MS, Kazokoglu H. Congenital lens coloboma and associated pathologies. Doc Ophthalmol. 1993;83(4):313-322.

62. Schlote T, Volker M, Knorr M, Thiel HJ. Lens coloboma and lens dislocation in Stickler (Marshall) syndrome. Klin Monbl Augenheilkd. 1997;210(4):227-228.

63. Sahu S, Yadav R, Gupta S, Puri LR. Bilateral ectopia lentis with isolated lens coloboma in Marfan syndrome. GMS Ophthalmol Cases. 2016;6:6-8.

64. S D-E. Congential Deformities. St. Louis: CV Mosby; 1969.

65. Westerlund E. Clinical and Genetic Studies on the Primary Glaucoma Diseases. Copenhagen: NYT Norsdic Forlag, Arnold Busk; 1947.

66. Gencik A. Epidemiology and genetics of primary congenital glaucoma in Slovakia. Description of a form of primary congenital glaucoma in gypsies with autosomal- recessive inheritance and complete penetrance. Dev Ophthalmol. 1989;16:76-115.

67. Francois J. Heredity in Opthamology. St. Louis: CV Mosby; 1961.

68. Francois J. Congenital glaucoma and its inheritance. Opthalmologica. 1972;181(2):61-73.

69. Faiq M, Sharma R, Dada R, Mohanty K, Saluja D, Dada T. Genetic, biochemical and clinical insights into primary congenital glaucoma. J Curr Glaucoma Pract. 2013;7(2):66- 84.

70. Jay M, Rich N. Genetic implications of congenital glaucoma. Metab Pediatr Ophthalmol. 1978;2:257-258.

71. Demenias F, Bonaiti C, Briard M, Feingold J, Frezal J. Congenital glaucoma: Genetic models. Hum Genet. 1979;46(3):305-317.

72. Gencik A, Gencikova A, Gerinec A. Genetic heterogeneity of congenital glaucoma. Clin Genet. 1980;17(4):214-248.

73. Sarfarazi M, Akarsu AN, Hossain A, et al. Assignment of a locus (GLC3A) for primary congenital glaucoma (Buphthalmos) to 2p21 and evidence for genetic heterogeneity. Genomics. 1995;30(2):171-177.

74. Akarsu AN, Turacli ME, Aktan SG, et al. A second locus (GLC3B ) for primary congenital glaucoma (Buphthalmos) maps to the 1p36 region. Hum Mol Genet. 1996;5(8):1199-1203.

75. Stoilov IR, Costa VP, Vasconcellos JPC, et al. Molecular genetics of primary congenital glaucoma in Brazil. Invest Ophthalmol Vis Sci. 2002;43(6):1820-1827.

76. Sarfarazi M, Stoilov I. Molecular genetics of primary congenital glaucoma. Eye (Lond). 2000;14(Pt 3B):422-428.

77. Stoilov I, Akarsu AN, Alozie I, et al. Sequence analysis and homology modeling suggest that primary congenital glaucoma on 2p21 results from mutations disrupting either the hinge region or the conserved core structures of cytochrome P4501B1. Am J Hum Genet. 1998;62(3):573-584.

78. Libby RT, Smith RS, Savinova O V, et al. Modification of ocular defects in mouse developmental glaucoma models by tyrosinase. Science. 2003;299(5612):1578-1581.

79. Stenson PD, Mort M, Ball E V., et al. The Human Gene Mutation Database: Towards a comprehensive repository of inherited mutation data for medical research, genetic diagnosis and next-generation sequencing studies. Hum Genet. 2017;136(6):665-677.

80. Cardiff University. The Human Gene Mutation Database: CYP1B1. http://www.hgmd.cf.ac.uk/ac/gene.php?gene=CYP1B1. Published 2017. Accessed December 1, 2018.

81. Li N, Zhou Y, Du L, Wei M, Chen X. Overview of Cytochrome P450 1B1 gene mutations in patients with primary congenital glaucoma. Exp Eye Res. 2011;93(5):572-579.

82. Alzuhairy S, Abu-Amero KK, Al-Shahwan S, Edward DP. A novel CYP1B1 mutation with congenital glaucoma and total aniridia. Ophthalmic Genet. 2013;36(1):89-91.

83. Chavarria-Soley G, Michels-Rautenstrauss K, Caliebe A, Kautza M, Mardin C, Rautenstrauss B. Novel CYP1B1 and known PAX6 mutations in anterior segment dysgenesis (ASD). J Glaucoma. 2006;15(6):499-504.

84. Acharya M, Mookherjee S, Bhattacharjee A, et al. Primary role of CYP1B1 in Indian juvenile-onset POAG patients. Mol Vis. 2006;12:399-404.

85. Pasutto F, Chavarria-Soley G, Mardin CY, et al. Heterozygous loss-of-function variants in CYP1B1 predispose to primary open-angle glaucoma. Investig Ophthalmol Vis Sci. 2010;51(1):249-254.

86. Haji-Seyed-Javadi R, Jelodari-Mamaghani S, Paylakhi SH, et al. LTBP2 mutations cause Weill-Marchesani and Weill-Marchesani-like syndrome and affect disruptions in the extracellular matrix. Hum Mutat. 2012;33(8):1182-1187.

87. Jelodari-Mamaghani S, Haji-Seyed-Javadi R, Suri F, et al. Contribution of the latent transforming growth factor-β binding protein 2 gene to etiology of primary open angle glaucoma and pseudoexfoliation syndrome. Mol Vis. 2013;19:333-347.

88. Chen X, Chen Y, Wang L, et al. CYP1B1 genotype influences the phenotype in primary congenital glaucoma and surgical treatment. Br J Ophthalmol. 2014;98(2):246-251.

89. Sharafieh R, Child AH, Khaw PT, Fleck B, Sarfarazi M. LTBP2 gene analysis in the GLC3C-linked family and 94 CYP1B1-negative cases with primary congenital glaucoma. Ophthalmic Genet. 2013;34(1-2):14-20.

90. University C. The Human Gene Mutation Database: LTBP2. http://www.hgmd.cf.ac.uk/ac/gene.php?gene=LTBP2. Published 2017. Accessed October 17, 2017.

91. Stone EM, Fingert JH, Alward WLM, et al. Identification of a gene that causes primary open angle glaucoma. Science. 1997;275(5300):668-670.

92. Sheffield VC, Stone EM, Alward WL, et al. Genetic linkage of familial open angle glaucoma to chromosome 1q21-q31. Nat Genet. 1993;4(1):47-50.

93. Kubota R, Noda S, Wang Y, et al. A novel myosin-like protein (myocilin) expressed in the connecting cilium of the photoreceptor: molecular cloning, tissue expression, and chromosomal mapping. Genomics. 1997;41(3):360-369.

94. Jacobson N, Andrews M, Shepard a R, et al. Non-secretion of mutant proteins of the glaucoma gene myocilin in cultured trabecular meshwork cells and in aqueous humor. Hum Mol Genet. 2001;10(2):117-125.

95. Kaur K, Reddy ABM, Mukhopadhyay A, et al. Myocilin gene implicated in primary congenital glaucoma. Clin Genet. 2005;67(4):335-340.

96. Mookherjee S, Acharya M, Banerjee D, Bhattacharjee A, Ray K. Molecular basis for involvement of CYP1B1 in MYOC upregulation and its potential implication in glaucoma pathogenesis. PLoS One. 2012;7(9):e45077.

97. Do T, Shei W, Chau PTM, et al. CYP1B1 and MYOC Mutations in Vietnamese primary congenital glaucoma patients. J Glaucoma. 2015;25(5):491-498.

98. University C. Human Gene Mutation Database: MYOC. http://www.hgmd.cf.ac.uk/ac/gene.php?gene=MYOC. Published 2017. Accessed December 1, 2018.

99. Suh W, Kee C. A clinical and molecular genetics study of primary congenital glaucoma in South Korea. Br J Ophthalmol. 2012;96(11):1372-1377.

100. Gage PJ, Suh H, Camper SA. Dosage requirement of Pitx2 for development of multiple organs. Development. 1999;126(20):4643-4651.

101. Evans AL, Gage PJ. Expression of the homeobox gene Pitx2 in neural crest is required for optic stalk and ocular anterior segment development. Hum Mol Genet. 2005;14(22):3347- 3359.

102. Chen L, Gage PJ. Heterozygous Pitx2 null mice accurately recapitulate the ocular features of axenfeld-rieger syndrome and congenital glaucoma. Investig Ophthalmol Vis Sci. 2016;57(11):5023-5030.

103. Phillips JC, del Bono E a, Haines JL, et al. A second locus for Rieger syndrome maps to chromosome 13q14. Am J Hum Genet. 1996;59(3):613-619.

104. Lehmann OJ, Sowden JC, Carlsson P, Jordan T, Bhattacharya SS. Fox’s in development and disease. Trends Genet. 2003;19(6):339-344.

105. Mears AJ, Jordan T, Mirzayans F, et al. Mutations of the forkhead/winged-helix gene, FKHL7, in patients with Axenfeld-Rieger anomaly. Am J Hum Genet. 1998;63(5):1316- 1328.

106. Swiderski RE, Reiter RS, Nishimura DY, et al. Expression of the Mf1 gene in developing mouse hearts: Implication in the development of human congenital heart defects. Dev Dyn. 1999;216(1):16-27.

107. Khan AO, Aldahmesh MA, Al-Amri A. Heterozygous FOXC1 mutation (M161K) associated with congenital glaucoma and aniridia in an infant and a milder phenotype in her mother. Ophthalmic Genet. 2008;29(2):67-71.

108. Honkanen RA, Nishimura DY, Swiderski RE, et al. A family with Axenfeld-Rieger syndrome and Peters Anomaly caused by a point mutation (Phe112Ser) in the FOXC1 gene. Am J Ophthalmol. 2003;135(3):368-375.

109. Nishimura DY, Swiderski RE, Alward WL, et al. The forkhead transcription factor gene FKHL7 is responsible for glaucoma phenotypes which map to 6p25. Nat Genet. 1998;19(2):140-147.

110. Puhó EH, Vogt G, Csáky-Szunyogh M, Metneki J, Czeizel AE. Maternal demographic and socioeconomic characteristics of live-born infants with isolated ocular congenital abnormalities. Ophthalmic Epidemiol. 2008;15(4):257-263.

111. Vogt G, Horvath-Puho E, Czeizel AE. A population-based case-control study of isolated primary congenital glaucoma. Am J Med Genet Part A. 2006;140(11):1148-1155.

112. Forestieri NE, Desrosiers TA, Freedman SF, et al. Risk factors for primary congenital glaucoma in the National Birth Defects Prevention Study. Accepted: Am J Med Genet. 2019.

113. Musial B, Vaughan OR, Fernandez-twinn DS, et al. A Western-style obesogenic diet alters maternal metabolic physiology with consequences for fetal nutrient acquisition in mice. J Physiol. 2017;595(14):4875-4892.

114. Tajima M, Ikarashi N, Okaniwa T, et al. Consumption of a high-fat diet during pregnancy changes the expression of Cytochrome P450 in the livers of infant male mice. Biol Pharm Bull. 2013;36(4):649-657.

115. Yu M, Jiang M, Yang C, et al. Maternal high-fat diet affects Msi/Notch/Hes signaling in neural stem cells of offspring mice. J Nutr Biochem. 2014;25(2):227-231.

116. O'Neill RJ, Vrana PB, Rosenfeld CS. Maternal methyl supplemented diets and effects on offspring health. Front Genet. 2014;5:289.

117. Lan X, Cretney EC, Kropp J, et al. Maternal diet during pregnancy induces gene expression and DNA methylation changes in fetal tissues in sheep. Front Genet. 2013;4:49.

118. Sandkam BA, Deere KA, Ashley M, Grether GF, Rodd FH, Fuller RC. Exploring visual plasticity : Dietary carotenoids can change color vision in guppies (Poecilia reticulata). J Comp Physiol A Neuroethol Sens Neural Behav Physiol. 2016;202(7):527-534.

119. Hu FB. Dietary pattern analysis: a new direction in nutritional epidemiology. Curr Opin Lipidol. 2002;13(1):3-9.

120. Bodnar LM, Siega-Riz AM. A Diet Quality Index for Pregnancy detects variation in diet and differences by sociodemographic factors. Public Health Nutr. 2002;5(6):801-809.

121. Guenther PM, Casavale KO, Kirkpatrick SI, et al. Update of the Healthy Eating Index: HEI-2010. J Acad Nutr Diet. 2014;113(4):569-580.

122. Chiuve SE, Fung TT, Rimm EB, et al. Alternative dietary indices both strongly predict risk of chronic disease. J Nutr. 2012;142(6):1009-1018.

123. Kant AK, Schatzkin A, Graubard BI, Schairer C. A prospective study of diet quality and mortality in women. J Am Med Assoc. 2000;283(16):2109-2115.

124. Carmichael SL, Yang W, Feldkamp ML, et al. Higher diet quality reduces risks of neural tube defects and orofacial clefts. Arch Pediatr Adolesc Med. 2012;166(2):121-126.

125. Botto LD, Krikov S, Carmichael SL, et al. Lower rate of selected congenital heart defects with better maternal diet quality : A population-based study. Arch Dis Child Fetal Neonatal Ed. 2016;101(1):F43-9.

126. Ma C, Shaw GM, Scheuerle AE, Canfield MA. Association of microtia with maternal nutrition. Birth Defects Res A Clin Mol Teratol.2012;94(12):1026-1032.

127. Reefhuis J, Gilboa SM, Anderka M, et al. The National Birth Defects Prevention Study: A review of the methods. Birth Defects Res Part A Clin Mol Teratol. 2015;103(8):656-669.

128. National Birth Defects Prevention Study. http://www.nbdps.org/. Accessed November 30, 2017.

129. Rasmussen SA, Olney RS, Holmes LB, et al. Guidelines for case classification for the National Birth Defects Prevention Study. Birth Defects Res Part A Clin Mol Teratol. 2003;67(3):193-201.

130. Service AR. USDA National Nutrient Database for Standard Reference, Release 20.; 2007.

131. Watts V, Rockett H, Baer H, Leppert J, Colditz G. Assessing diet quality in a population of low-income pregnant women : A comparison between Native Americans and Whites. Matern Child Health J. 2007;11(2):127-136.

132. Laraia BA, Siega-Riz AM, Kaufman JS, Jones SJ. Proximity of supermarkets is positively associated with diet quality index for pregnancy. Prev Med. 2004;39(5):869-875.

100

133. Laraia BA, Bodnar LM, Siega-riz AM, Carolina N. Pregravid body mass index is negatively associated with diet quality during pregnancy. Public Health Nutr. 2007;10(9):920-926.

134. Research Service USD of AA. USDA National Nutrient Database for Standard Reference, Release 20. Nutrient Data Laboratory Home Page.

135. Service USD of AAR. USDA Database for the Choline Content of Common Foods, Release 2. http://www.nal.usda.gov/fnic/foodcomp/Data/ Choline/Choln02.pdf.2008%0A24. Published 2008.

136. Newby P, Tucker K. Empirically derived eating patterns using factor or cluster analysis: A review. Nutr Rev. 2004;62(5):177-203.

137. Hoffman K, Schulze M, Boeing H, Altenburg H. Dietary patterns: Report of an international workshop. Public Health Nutr. 2002;5(1):89-90.

138. Everitt B, Landau S, Leese M. Cluster Analysis. (Leese M, ed.). London, New York: Oxford University Press; 2001.

139. Sotres-Alvarez D. Latent class analysis is useful to classify pregnant women into dietary patterns. J Nutr. 2010;140(12):2253-2259.

140. Fahey M, Thane C, Bramwell G, Coward W. Conditional Gaussian mixture modeling for dietary pattern analysis. J R Stat Soc. 2007;170(1):149-166.

141. Hagenaars JA, McCutcheon AL. Applied Latent Class Analysis. (JA H, AL M, eds.). Cambridge: Cambridge University Press; 2002.

142. Rasmussen SA, Lammer EJ, Shaw GM, et al. Integration of DNA sample collection into a multi-site birth defects case-control study. Teratology. 2002;66(4):177-184.

143. Illumina. Quality Scores for Next-Generation Sequencing; 2011.

144. Thau A, Lloyd M, Freedman S, Beck A, Grajewski A, Levin A V. New classification system for pediatric glaucoma : Implications for clinical care and a research registry. Curr Opin Ophthalmol. 2018;29(5):385-394.

145. Vujkovic M. The maternal Mediterranean dietary pattern is associated with a reduced risk of spina bifida in the offspring. BJOG. 2009;116(3):408-415.

146. Willett WC, Reynolds RD, Cottrell-Hoehner, S. Sampson L, Browne ML. Validation of a semi-quantitative food frequency questionnaire: Comparison with a 1-year diet. J Am Diet Assoc. 1987;87(1):43-47.

101

147. Haines PS, Siega-Riz AM, Popkin BM. A diet quality index revised: A measurement instrument for populations. J Am Diet Assoc. 1999;99(6):697-704.

148. Weber KA, Yang W, Carmichael SL, Shaw GM. Nutrient intake in women before conception and risks of anophthalmia and microphthalmia in their offspring. Birth Defects Res. 2018;110(10):863-870.

149. Carmichael SL, Shaw GM, Yang W, et al. Correlates of intake of folic acid-containing supplements among pregnant women. Am J Obstet Gynecol. 2006;194(1):203-210.

150. Webb SE, Miller AL. Calcium signalling during zebrafish embryonic development. BioEssays. 2000;22(2):113-123.

151. He Y, Ge J, Tombran-Tink J. Mitochondrial defects and dysfunction in calcium regulation in glaucomatous trabecular meshwork cells. Invest Ophthalmol Vis Sci. 2008;49(11):4912- 4922.

152. Lin Y, DL E, Liton P. Intralysosomal iron induces lysosomal membrane permeabilization and cathepsin D-mediated cell death in trabecular meshwork cells exposed to oxidative stress. Invest Ophthalmol Vis Sci. 2010;51(12):6483-6495.

153. Burton G, Joyce A, Ingold K. First proof that vitamin E is major lipid-soluble, chain- breaking antioxidant in human blood plasma. Lancet. 1982;2(8293):327.

154. McDougall M, Choi J, Kim H, et al. Lethal dysregulation of energy metabolism during embryonic vitamin E deficiency. Free Radic Biol Med. 2017;104:324-332.

155. Shamim A, Schulze K, Merrill R, et al. First-trimester plasma tocopherols are associated with risk of miscarriage in rural Bangladesh. Am J Clin Nutr. 2015;101(2):294-301.

156. Wang Y, Surzenko N, Friday W, Zeisel S. Maternal dietary intake of choline in mice regulates development of the cerebral cortex in the offspring. FASEB J. 2016;30(4):1566- 1578.

157. Shaw GM, Carmichael SL, Yang W, Selvin S, Schaffer DM. Periconceptional dietary intake of choline and betaine and neural tube defects in offspring. Am J Epidemiol. 2004;160(2):102-109.

158. Czeizel AE, Merhala Z. Bread fortification with folic acid, vitamin B12, and vitamin B6 in Hungary. Lancet. 1998;352(9135):1225.

159. Obermann-borst SA, Vujkovic M, Vries JH De, Wildhagen MF, Looman CW, Jonge R De. A maternal dietary pattern characterised by fish and seafood in association with the risk of congenital heart defects in the offspring. BJOG. 2011;118(10):1205-1215.

102

160. Cogswell ME, Bitsko RH, Anderka M, et al. Control selection and participation in an ongoing, population-based, case- control study of birth defects. Am J Epidemiol. 2009;170(8):975-985.

161. Abu-Hassan DW, Acott TS, Kelley MJ. The trabecular meshwork: A basic review of form and function. J Ocul Biol. 2014;2(1).

162. Mashima Y, Suzuki Y, Sergeev Y, et al. Novel Cytochrome P4501B1 (CYP1B1) Gene mutations in Japanese patients with primary congenital glaucoma. Invest Ophthalmol Vis Sci. 2001;42(10):2211-2216.

163. Reis LM, Tyler RC, Weh E, et al. Analysis of CYP1B1 in pediatric and adult glaucoma and other ocular phenotypes. Mol Vis. 2016;22:1229-1238.

164. Jenkins MM, Almli LM, Pangilinan F, et al. Exome sequencing of family trios from the National Birth Defects Prevention Study: Tapping into a rich resource of genetic and environmental data. 2019; unpublished.

165. Paila U, Chapman BA, Kirchner R, Quinlan AR. GEMINI : Integrative exploration of genetic variation and genome annotations. PLOS Comput Biol. 2013;9(7):e1003153.

166. Lek M, Karczewski KJ, Eric V, et al. Analysis of protein-coding genetic variation in 60,706 humans. Nature. 2016;536(7616):285-291.

167. The UK10K Consortium. The UK10K project identifies rare variants in health and disease. Nature. 2015;526(7571):82-90.

168. Jun G, Flickinger M, Hetrick KN, et al. Detecting and estimating contamination of human DNA samples in sequencing and array-based genotype data. Am J Clin Genet. 2012;91(5):839-848.

169. Chakrabarti S, Kaur K, Rao KN, et al. The transcription factor gene FOXC1 exhibits a limited role in primary congenital glaucoma. Invest Ophthalmol Vis Sci. 2009;50(1):75- 83.

170. Zhao WJ, Xu J, Chen XJ, Liu HH, Yao K, Yan Y Bin. Effects of cataract-causing mutations W59C and W151C on βB2-crystallin structure, stability and folding. Int J Biol Macromol. 2017;103:764-770.

171. Yang Y, Wolf L V., Cvekl A. Distinct embryonic expression and localization of CBP and p300 histone acetyltransferases at the mouse A-crystallin locus of the lens. J Mol Biol. 2007;369(4):917-926.

172. Chen Q, Dowhan DH, Liang D, Moore DD, Overbeek PA. CREB-binding protein / p300 co-activation of crystallin gene expression. J Biol Chem. 2002;277(27):24081-24089.

103

173. Goodman RH, Smolik S. CBP/p300 in cell growth, transformation, and development. Genes Dev. 2000;14(13):1553-1577.

174. Hennekam R. Rubinstein-Taybi syndrome. Eur J Hum Genet. 2006;14(9):981-985.

175. Menke LA, Belzen MJ Van, Alders M, et al. CREBBP Mutations in individuals without Rubinstein – Taybi Syndrome phenotype. Am J Med Genet Part A. 2016;170(10):2681- 2693.

176. Levy NS. Juvenile glaucoma in the Rubinstein-Taybi syndrome. J Pediatr Ophthalmol. 1976;13(3):141-143.

177. McKusick VA. Personal Communication. Baltimore: MD; 1968.

178. Patel SP, Parker MD. SLC4A11 and the pathophysiology of Congenital Hereditary Endothelial Dystrophy. Biomed Res Int. 2015;2015:475392.

179. Mullaney P, Risco J, Teichmann K, Millar L. Congenital hereditary endothelial dystrophy associated with glaucoma. Ophthalmology. 1995;102(2):186-192.

180. Alsaif HS, Khan AO, Patel N, Alkuraya H, Hashem M, Abdulwahab F. Congenital glaucoma and CYP1B1: An old story revisited. Hum Genet. 2018:[Epub ahead of print].

181. POMT2. ExAC Browser. http://exac.broadinstitute.org/variant/14-77745128-C-T. Accessed February 9, 2019.

182. Van Reeuwijk J, Janssen M, van den Elzen C, et al. POMT2 mutations cause alpha- dystroglycan hypoglycosylation and Walker-Warburg syndrome. J Med Genet. 2005;42(12):907-912.

183. Bouchet C, Gonzales M, Vuillaumier-Barrot S, et al. Molecular heterogeneity in fetal forms of Type II Lissencephaly. Hum Mutat. 2007;28(10):1020-1027.

184. Vajsar J, Schachter H. Walker-Warburg syndrome. Orphanet J Rare Dis. 2006;1:29.

185. Pacella E, Pacella F, Paolis G De, et al. Glycosaminoglycans in the human cornea: Age- related changes. Ophthalmol Eye Dis. 2015;7:1-5.

186. Maurice D. Epithelial potential of the cornea. Exp Eye Res. 1967;6(2):138-140.

187. Millodot M. Dictionary of Optometry and Visual Science. 7th Ed. Oxford: Butterworth- Heinemann; 2008.

188. Nakajima M, Mizumoto S, Miyake N, et al. Mutations in B3GALT6 , which encodes a glycosaminoglycan linker region enzyme cause a spectrum of skeletal and connective tissue disorders. Am J Hum Genet. 2013;92(6):927-934. 104

189. Science WI of. Gene Cards: PIEZO2 Gene. Human Gene Database. https://www.genecards.org/cgi-bin/carddisp.pl?gene=PIEZO2. Published 2019. Accessed March 8, 2019.

190. Matias-Perez D, Garcia-Montano LA, Cruz-Aguilar M, et al. Identification of novel pathogenic variants and novel gene- phenotype correlations in Mexican subjects with microphthalmia and / or anophthalmia by next-generation sequencing. J Hum Genet. 2018;63(11):1169-1180.

191. Cvekl A, Wang WL. Retinoic acid signaling in mammalian eye development. Exp Eye Res. 2009;89(3):280-291.

192. Mondal D, Shenoy SR, Antisena M. Retinoic acid embryopathy. Int J Appl Basic Med Res. 2017;7(4):264-265.

193. Sng CCA, Ang M, Barton K. Central corneal thickness in glaucoma. Curr Opin Ophthalmol. 2017;28(2):120-126.

194. Pagel KA, Pejaver V, Lin GN, et al. When loss-of-function is loss of function : assessing mutational signatures and impact of loss-of-function genetic variants. Bioinformatics. 2017;33(14):i389-i398.

195. Roessler E, Du Y, Mullor JL, et al. Loss-of-function mutations in the human GLI2 gene are associated with pituitary anomalies and holoprosencephaly-like features. Proc Natl Acad Sci USA. 2003;100(23):13424-13429.

196. Roessler E, Muenke M. Midline and laterality defects: left and right meet in the middle. BioEssays. 2001;23(10):888-900.

197. Freund C, Horsford DJ, Mcinnes RR. Transcription factor genes and the developing eye : A genetic perspective. Hum Mol Genet. 1996;5:1471-1488.

198. Xie B, Zhang X, Hashimoto T, et al. Differentiation of retinal ganglion cells and photoreceptor precursors from mouse induced pluripotent stem cells carrying an ATOH7/MATH5 lineage reporter. PLoS One. 2014;9(11):e112175.

199. Muzyka VV, Brooks M, Badea TC. Postnatal Developmental Dynamics of Cell Type Specification Genes in Brn3a / Pou4f1 Retinal Ganglion Cells. Neural Development. 2018;13(1):15.

200. ALDH3A2. The Human Protein Atlas. https://www.proteinatlas.org/ENSG00000072210- ALDH3A2/tissue. Accessed March 8, 2019.

105

201. Naganuma T, Takagi S, Kanetake T, et al. Disruption of the Sjögren-Larsson Syndrome gene ALDH3A2 in mice increases keratinocyte growth and retards skin barrier. J Bio. 2016;291(22):11676-11688.

202. Huang L, Mao Y, Yang J, Li Y, Li Y, Yang Z. Mutation screening of the USH2A gene in retinitis pigmentosa and USHER patients in a Han Chinese population. Eye. 2018;32(10):1608-1614.

203. Jian Seyedahmadi B, Rivolta C, Keene JA, Berson EL, Dryja TP. Comprehensive screening of the USH2A gene in Usher syndrome type II and non-syndromic recessive retinitis pigmentosa. Exp Eye Res. 2004;79(2):167-173.

204. Disorders NO for R. Usher syndrome. Rare Disease Database. https://rarediseases.org/rare-diseases/usher-syndrome/. Published 2018. Accessed March 8, 2019.

205. UGT1A8. Human Gene Database. https://www.genecards.org/cgi- bin/carddisp.pl?gene=UGT1A8. Published 2019. Accessed March 8, 2019.

206. Iao ZX, Unome KN, Ahara TY, et al. Comparative studies of human UDP- glucuronosyltransferase 1A8 and 1A9 proximal promoters using single base substitutions. Drug Metab Pharmacokinet. 2014;29(1):90-93.

207. Ehmer U, Kalthoff S, Fakundiny B, et al. Gilbert Syndrome redefined: A complex genetic haplotype influences the regulation of glucuronidation. Hepatology. 2012;55(6):1912- 1921.

208. Hirschhorn JN, Daly MJ. Genome-wide association studies for common diseases and complex traits. Nat Rev Genet. 2005;6(2):95-108.

209. Cascella R, Strafella C, Germani C, et al. The genetics and the genomics of primary congenital glaucoma. Biomed Res Int. 2015;2015: 7 pages.

210. Souma T, Tompson SW, Thomson BR, et al. Angiopoietin receptor TEK mutations underlie primary congenital glaucoma with variable expressivity. J Clin Invest. 2016;126(7):2575-2587.

211. Adli M. The CRISPR tool kit for genome editing and beyond. Nat Commun. 2018;9(1):1911.

106