Assessing the Non-Genetic and Genetic Factors Affecting Refraction in the Aging Adult Population

$Assessing the Non-Genetic and Genetic Factors Affecting Refraction in the Aging Adult Population$

ASSESSING THE NON-GENETIC AND GENETIC FACTORS AFFECTING REFRACTION IN THE AGING ADULT POPULATION

by Samantha Bomotti

A dissertation submitted to Johns Hopkins University in conformity with the requirements for the degree of Doctor of Philosophy

Baltimore, MD January, 2018

Abstract

Refractive errors are the most common form of visual impairment in the world, and are becoming an increasing public health burden as the world’s population ages.

Refractive errors arise from changes in the shape of the eye, such as axial length and corneal curvature, or from aging of the lens. Refraction is a quantitative trait underlying refractive errors. The goal of this project was to investigate the non-genetic and genetic factors affecting refraction, a complex multifactorial trait, in the aging adult population.

We used phenotypic data available from the population-based Beaver Dam Eye

Study (BDES) consisting of 4,972 individuals aged 43-86 years at baseline to identify and characterize the association of nuclear sclerosis, among other factors, with refraction and changes in refraction. We then imputed exome array data in a subset of BDES participants to enhance our coverage of protein-coding regions and identify variants associated with refraction or either of its biological determinants, axial length and corneal curvature. Finally, we conducted a heritability analysis to determine whether the heritability of refraction varied by nuclear sclerosis severity, and to quantify the genetic or environmental influences shared between refraction and nuclear sclerosis.

We determined nuclear cataract is the primary contributor to the myopic shift observed in older persons, as only those with nuclear cataract experienced a myopic shift while those without nuclear cataract did not. Sex, diabetes, and baseline refractive error, but not education, were also associated with refractive changes. While we were unable to detect novel genetic loci associated with refraction, axial length, or corneal curvature, we demonstrated that variants associated in prior studies are also associated with these same phenotypes in the BDES, validating these genetic variants as contributors to the etiology

ii of these phenotypes. Finally, the heritability of refraction was consistent across differing distributions of nuclear sclerosis severity, and approximately one third to one half of the genetic effects contributing to refraction and nuclear sclerosis were shared. The results of these studies can help better characterize the underlying etiology of refraction for improved preventative measures and targeted treatments designed to counteract the development of refractive errors in older persons.

Advisors: Dr. Priya Duggal, PhD

Dr. Alison P. Klein, PhD

Thesis Readers: Dr. Priya Duggal, PhD

Dr. Alison P. Klein, PhD

Dr. Joan Bailey-Wilson, PhD

Dr. Ingo Ruczinski, PhD

iii

Acknowledgements

I want to begin by thanking my high school chemistry professor, Mrs. Viscardo.

Her passion and enthusiasm for chemistry, and encouragement to pursue science in my own future, directed my own path toward science in college and graduate school. To this day I consider her the best teacher I ever had.

I would also like to thank my MPH advisor, Dr. Sharon Kardia, for her diligent mentorship during my tenure at the University of Michigan. Her patience, wisdom, and good humor were invaluable as I navigated my way through my very first genetic epidemiology project, from which I gained an even greater appreciation and enthusiasm for the field.

Of course, My PhD would not be possible without the superior mentorship I received during my time at Johns Hopkins. Firstly, I want to thank Dr. Robert

Wojciechowski for always being so kind and approachable to me, particularly when I was a new student and knew few others in the program. I also sincerely appreciate his persistent collaboration with other researchers on my behalf, and how he facilitated networking opportunities for me, when I was first searching for a project.

Dr. Terri Beaty has remained a consistently reliable, honest, and truly supportive mentor throughout my tenure at Johns Hopkins. I have deeply valued her wisdom and support and would not have navigated through the program nearly as smoothly without her.

My Thesis Committee members, Dr. Bryan Lau, Dr. Alison Klein, and Dr. Priya

Duggal, undoubtedly deserve my gratitude for dedicating so much of their time to frequent (and often long) meetings with me, to reading through drafts of my material, and

iv steering me in the right direction when I began to go of course. Their guidance has made me a more confident and better epidemiologist, and I cannot thank them enough for the endless amount of support and direction they provided to bring me to where I am now.

My primary advisor in particular, Dr. Duggal, deserves a round of applause for meeting with me weekly and responding to my needs in such a timely manner in spite of her incredibly busy schedule.

I sincerely thank all of my oral exam committee members (Dr. Chris Ladd-

Acosta, Dr. Priya Duggal, Dr. Elizabeth Platz, Dr. Rasika Mathias, Dr. Alison Klein, Dr.

Bryan Lau, and Dr. Elizabeth Colantuoni) for their time and for their positive encouragement before and during my oral exams. I also want to give a special thank you to Dr. Colantuoni, who met with me several additional times after I passed my oral exams to discuss my project and offer expert advice regarding my statistical methods. Her patience and kindness, and constant willingness to answer any questions I had no matter how busy she was with her own students, meant a great deal to me and certainly helped advance my own abilities as an epidemiologist. I appreciate her dedication.

I was honored during my tenure to be offered a TA and Lead TA position by Drs.

Amber D’Souza and Stephan Ehrhardt. Their confidence in my abilities as a teacher, and as an epidemiologist, helped convince me of my own abilities. My experiences as TA and

Lead TA under their skilled and passionate guidance will be considered among my best and most fulfilling experiences at Johns Hopkins.

I owe such a debt of gratitude to Fran Burman and Matt Miller, without whom I would have been completely lost. They simply know everything about everything related to the department. I never once asked them a question they could not answer. They were

v constantly my allies and always worked in my best interest, for which I cannot thank them enough. They also acted as in-house counselors during stressful times. I am convinced all students would be completely lost without them and my experience here was definitely enhanced by my interactions with them.

All of my professors, Thesis readers, and any faculty who had a positive influence on my education while I was at Johns Hopkins absolutely deserve my sincerest gratitude for their diligence, effort, and dedication to the education and well-being of all of their students, including me. There are simply too many individuals who have improved my experience here in some way to name, but I want them all to know that I appreciate all they did and do for all of us.

Most importantly, I want to thank my family and friends. I would have no emotional support system without them, and I certainly would not have survived graduate school without their ardent love and support. I especially want to thank my parents. I would not be where I am or who I am without them. They have given me everything I have and everything I could ever need or want. Their unwavering confidence in me, even when I had none in myself, has allowed me to get to where I am. I am incredibly lucky to have such supportive and loving parents who have dedicated so much of their time and energy to my progress and well-being. Finally, I have to thank my fiancé Tim. Tim is my closest confidant, best friend, and greatest source of happiness. Regardless of anything else going on, he has never failed to make me smile with his words of comfort and sense of humor. His unconditional love and support, even during the most challenging times, has become a persistently shining light in my life that has provided me with the strength necessary to face each day and achieve my goals.

TABLE OF CONTENTS

Abstract ...... ii Acknowledgements ...... iv

CHAPTER ONE

Introduction and Specific Aims

Refraction and Refractive Errors ...... 2 The Burden of Refractive Errors in an Aging Population ...... 7 Overview of the Project ...... 7 Aim 1: Non-Genetic Factors Affecting Refraction and Changes in Refraction in Older Adults ...... 10 Aim 2: Genetic Factors Affecting Refraction in Older Adults ...... 13 Aim 3: Shared Genetic and Environmental Effects of Refraction and Nuclear Sclerosis ...... 16 Thesis Organization ...... 17 References ...... 18

CHAPTER TWO

vii

Refraction and Change in Refraction Over a 20-Year Period in the Beaver Dam Eye

Study

Abstract ...... 27 Introduction ...... 28 Methods ...... 29 The Beaver Dam Eye Study ...... 29 Trait Definitions ...... 30 Exclusion Criteria ...... 30 Statistical Analysis ...... 31 Inverse Probability Weighting ...... 32 Results ...... 32 Study Participants ...... 32 Age and Birth Cohort ...... 33 Nuclear Sclerosis ...... 33 Risk Factors Associated with Refraction ...... 34 Discussion ...... 35 Acknowledgements ...... 39 References ...... 40

CHAPTER THREE

Imputation of Exome Array Variants to the Haplotype Reference Consortium Panel

to Detect Additional Loci Associated with Ocular Phenotypes in the Beaver Dam

Eye Study

viii

Abstract ...... 51 Introduction ...... 52 Methods ...... 53 The Beaver Dam Eye Study ...... 53 Trait Definitions ...... 54 Quality Control for Samples for Axial Length, Corneal Curvature, and Refraction ..... 55 Quality Control for Genotyping and Imputation ...... 55 Coverage ...... 57 Genotype Concordance ...... 57 Single Variant Analysis ...... 57 Meta-Analysis of BDES Groups ...... 58 Gene-Based Analysis ...... 59 Association of Known Genome-Wide Association Analysis (GWAS) Loci with Axial Length, Corneal Curvature, and Refraction ...... 59 Annotations ...... 60 Results ...... 60 Coverage ...... 60 Genotype Concordance ...... 61 Study Participants in First and Second BDES Groups...... 61 Single Variant and Gene-Based Analyses ...... 61 Meta-Analysis of BDES Groups ...... 62 Association with Known GWAS Genetic Loci ...... 62 Discussion ...... 64 Acknowledgements ...... 70 References ...... 71

CHAPTER FOUR

Bivariate Heritability Shows Evidence of Shared Genetic Effects of Refraction and

Nuclear Sclerosis: The Beaver Dam Eye Study

Abstract ...... 87 Introduction ...... 88 Methods ...... 89 The Beaver Dam Eye Study ...... 89 Trait Definitions ...... 90 Exclusion Criteria ...... 91 Statistical Analysis ...... 91 Results ...... 92 Study Participants ...... 92 Heritability of Refraction at Visits 1 and 3 ...... 93 Heritability of Refraction by Nuclear Sclerosis Severity ...... 94 Shared Effects of Refraction and Nuclear Sclerosis ...... 94 Discussion ...... 95 Acknowledgements ...... 98 References ...... 99

CHAPTER FIVE

Conclusions, Future Directions, and Public Health Significance

Summary and Discussion of Results ...... 110 Strengths of the Studies ...... 113 Limitations of the Studies ...... 114 Future Directions ...... 115

Public Health Significance ...... 118 References ...... 120

APPENDIX

Methods ...... 125 Quality Control and Imputation for the Second Beaver Dam Eye Study (BDES) Group ...... 125 Samples for Axial Length, Corneal Curvature, and Refraction for the Second BDES Group ...... 126 Meta-Analysis of BDES Groups ...... 126 Results ...... 127 Single Variant Analysis ...... 127 Gene-Based Analysis ...... 128 Meta-Analysis of BDES Groups ...... 130 References ...... 133

LIST OF TABLES

CHAPTER TWO

Table 1. Characteristics of Participants of the Beaver Dam Eye Study at Baseline, 1988- 1990 ...... 45 Table 2. Estimated Associations of Various Factors With Spherical Equivalent and Amount of Change in Spherical Equivalent Among Participants of the Beaver Dam Eye Study, Beaver Dam, Wisconsin, 1988-2010 ...... 46

Table 3. Mean Changes in Spherical Equivalent by Age and Nuclear Sclerosis Grade Across Visits Among Participants of the Beaver Dam Eye Study, Beaver Dam, Wisconsin, 1988-2010 ...... 47

CHAPTER THREE

Table 1. Mean Imputation R2 by Minor Allele Frequency Bin for Imputed Exome Array ...... 77 Table 2. Coverage of Genotyped and Imputed Variants by Chromosome ...... 78

Table 3. Association Results for Variants in Previously Published Loci for Corneal Curvature and Spherical Equivalent ...... 80 Table 4. Comparison of Variants Associated With Ocular Phenotypes Before and After Imputation of the Exome Array ...... 82

CHAPTER FOUR

Table 1. Ocular and Demographic Characteristics of Participants of the Beaver Dam Eye Study ...... 105 Table 2. Relative Pairs Contributing to the Heritability Analyses for Spherical Equivalent and Nuclear Sclerosis Among Participants of the Beaver Dam Eye Study ...... 106

xii

Table 3. Bivariate Heritability Analysis of Spherical Equivalent and Nuclear Sclerosis Among Participants of the Beaver Dam Eye Study ...... 107

APPENDIX

Supplementary Table S1. Characteristics of the Beaver Dam Eye Study Participants Contributing to the Axial Length Analyses ...... 135 Supplementary Table S2. Characteristics of the Beaver Dam Eye Study Participants Contributing to the Corneal Curvature Analyses ...... 136

Supplementary Table S3. Characteristics of the Beaver Dam Eye Study Participants Contributing to the Spherical Equivalent Analyses ...... 137 Supplementary Table S4. Top Single Variant Association Analysis Results for Axial Length ...... 138 Supplementary Table S5. Top Single Variant Association Analysis Results for Corneal Curvature...... 145 Supplementary Table S6. Top Single Variant Association Analysis Results for Spherical Equivalent ...... 147 Supplementary Table S7. Top Gene-Based Association Analysis Results for Corneal Curvature and Spherical Equivalent ...... 154 Supplementary Table S8. Top Results of Meta-Analysis of First and Second Beaver Dam Eye Study Groups for Axial Length, Corneal Curvature, and Spherical Equivalent ...... 156 Supplementary Table S9. Association Results for Variants in Previously Published Loci for Axial Length ...... 158

LIST OF FIGURES

xiii

CHAPTER ONE

Figure 1. Location of the Focal Point in a Human Myopic Eye and a Human Hyperopic Eye ...... 4 Figure 2. The Axial Length, or Distance from the Anterior to the Posterior Poles, of the Human Eye ...... 5

Figure 3. Radius of Curvature of the Cornea, from the Corneal Surface to the Lens ...... 6

CHAPTER TWO

Figure 1. Refraction by Age and Year of Birth in the Beaver Dam Eye Study ...... 48

CHAPTER THREE

Figure 1. Coverage of the Non-Imputed and Imputed Exome Array on Chromosome 22 ...... 83

CHAPTER FOUR

Figure 1. Heritability of Refraction by Nuclear Sclerosis Severity at Visit 3 in the Beaver Dam Eye Study ...... 108

xiv

APPENDIX

Supplementary Figure S1. Principal Component Analysis of Participants of the Beaver Dam Eye Study Who Were Genotyped on the Illumina Exome Array ...... 159 Supplementary Figure S2. Quantile-Quantile Plots of P-Values for the Single Variant Association Analyses of Axial Length, Corneal Curvature, and Refraction ...... 161 Supplementary Figure S3. Overlapping Histograms of the Distributions of Axial Length, Corneal Curvature, and Refraction in the First and Second Groups Genotyped in the Beaver Dam Eye Study ...... 167

CHAPTER ONE

Introduction and Specific Aims

Refraction and Refractive Errors

Images from our environment must be focused on the retina for the human eye to see clear images. The retina is a sensory tissue at the back of the eyeball containing photoreceptor cells sensitive to light and trigger nerve impulses upon activation. These impulses are sent to the brain via the optic nerve connected to the retina so a visual image can be formed and perceived. When images are focused on the retina without the help of corrective lenses or other means, the eye is considered emmetropic and there is no refractive error (1). Refractive errors, myopia and hyperopia, occur when the eye does not correctly refract light and images are not focused on the retina (1). A person is said to be myopic, or nearsighted, if images are focused in front of the retina instead of on the retina itself. A person is said to be hyperopic, or farsighted, if images are focused behind the retina (Figure 1) (1). Refraction is a quantitative trait underlying refractive errors, which are defined by values of refraction that exceed established thresholds. An eye is considered myopic if refraction is more negative than an established threshold and hyperopic if refraction is more positive than an established threshold.

These refractive errors arise because changes in the shape of the eye prevent clear focus of images. Changes to the shape of the eye can include the length of the eyeball or changes in corneal shape (1). The length of the eyeball from the anterior (front of the cornea) to the posterior (back of the retina) pole is known as the axial length of the eye

(Figure 2). Elongation of the eye, or longer axial length, can induce myopia if not accommodated by other compensatory mechanisms. Decreases in axial length can similarly induce hyperopia. Changes in corneal shape can also lead to refractive errors.

Corneal curvature describes the shape of the front surface of the cornea (Figure 3). The radius of curvature of the cornea is the distance from the center of the cornea to the lens,

2 and can increase or decrease as the cornea becomes steeper or flatter. A steeper corneal curvature can lead to myopia if not corrected by other ocular mechanisms. A flatter corneal curvature can likewise lead to hyperopia. Refractive errors can arise from a lack of coordination between the position of the retina relative to the cornea, which is governed by axial length, and the refractive power of the eye, which is governed by corneal curvature (2–9).

Refractive errors can also arise from aging of the lens. Opacification of the lens with age can lead to cataracts, which are the leading cause of blindness worldwide (10).

Nuclear cataracts, which form from yellowing and hardening of the central zone

(nucleus) of the lens, are the most common form of age-related cataract (11,12). Such cataracts cause light to scatter toward the retina, leading to glare and reduced visual acuity resulting from an inability of images to focus clearly on the retina. A primary effect of cataract is a myopic shift in refraction in older individuals (13–15). The etiology of axial length, corneal curvature, and nuclear sclerosis can all inform the etiology of refraction and refractive errors, due to their strong influence on refraction.

Figure 1. Location of the focal point (where images are focused) in a human myopic eye

(top) and a human hyperopic eye (bottom). Adapted from http://lasereyesurgeryjb.weebly.com/myopia-and-hyperopia.html (16).

Figure 2. The axial length, or distance from the anterior to the posterior poles, of the human eye (red line). Adapted from http://4uxj7d6q.adtddns.asia/label-eye-diagram.html

(17).

Figure 3. Radius of curvature of the cornea, from the corneal surface to the lens (red line). Adapted from http://4uxj7d6q.adtddns.asia/label-eye-diagram.html (17).

The Burden of Refractive Errors in an Aging Population

Approximately 65% of all people suffering from visual impairment due to uncorrected refractive errors are over 50 years of age, yet this age group comprises only

20% of the world’s population. Uncorrected refractive errors, or cases in which the eye cannot focus images clearly, are responsible for 43% of visual impairment worldwide, resulting from excessively blurred vision and even functional blindness in some cases

(18,19). The global economic burden associated with visual impairment resulting from refractive errors is estimated to be US$202 billion each year (20). Although refractive errors can be treated with spectacles, contacts, or surgery, studies have demonstrated that refractive errors are too often not properly corrected, leaving many in developed nations like the United States with mild visual impairment (21). Over one quarter of those 40 years and older in the United States and western Europe have myopia alone, while 10% of individuals in the same age group have detectable hyperopia (21). With an increasing elderly population in many developed countries, including the United States, more people will be at risk for age-related eye disorders like refractive errors in the coming decades

(19,22).

Concerted public health action is required to address this mounting burden of eye disorders among the growing aged population. Understanding the underlying etiology of refractive errors can allow us to develop tailored clinical measures that can be provided for preventative and treatment services in this population sector. The overarching goal of this dissertation is to assess the non-genetic and genetic factors affecting refraction among the aging adult population.

Overview of the Project

The following sections describe the background and motivation behind each proposed specific aim, each of which is presented in its corresponding section. All three aims used data available from the Beaver Dam Eye Study (BDES). The BDES was initiated in 1987 in the township of Beaver Dam, Wisconsin to determine the prevalence and incidence of age-related ocular disorders causing vision loss among elderly individuals, and to identify and characterize potential causes or risk factors for these conditions. A private census identified nearly 6,000 individuals over the age of 40, almost

5,000 of whom participated in the baseline examination. More than half of these individuals were related in extended pedigrees. The BDES includes ocular, demographic, and other clinical measures taken every five years between 1988 and 2010. The BDES also includes genetic data from a subset of individuals genotyped on the Illumina exome array.

The first aim of this dissertation focused on the effect of non-genetic factors on changes in refraction over time. This study involved a longitudinal analysis assessing the relationship between changes in refraction and relevant demographic and clinical factors suggested to influence refractive changes in past studies. The association between changes in refraction and age-related nuclear sclerosis, a disorder characterized by changes in the density and opacity of the lens that develops into nuclear cataract in its most severe form, was the primary focus. Axial length was not measured until fifteen years after the baseline visit in the BDES (visit 4; 2003-2005), and thus could not be analyzed longitudinally.

The second aim of this dissertation focused on the genetic factors influencing refraction and two of its biological determinants, axial length and corneal curvature.

Genotypes available from the Illumina exome array were imputed to enhance the coverage of the regions targeted by the exome array, and to improve our ability to detect novel genetic loci associated with refraction, axial length, and corneal curvature. To increase our power to detect novel loci, we conducted gene-based tests in addition to single variant tests. We also attempted to replicate previously identified loci associated with these phenotypes. To evaluate the practical utility of imputation from the exome array, we compared results obtained from the imputed exome array with results obtained from the original (i.e. non-imputed) exome array to determine whether imputation improved detection of loci in the exome.

The focus of the third aim of this dissertation was to determine whether the heritability of refraction varied depending on the distribution of nuclear sclerosis, and to quantify the shared genetic and environmental effects of refraction and nuclear sclerosis.

The demonstrated genetic component of both traits, in addition to the strong influence of nuclear sclerosis on refraction, suggested nuclear sclerosis severity may have a strong influence on the heritability of refraction. Heritability estimates for refraction were obtained at two time points, and within each category of nuclear sclerosis severity to identify differences in heritability estimates by nuclear sclerosis severity. The heritability of nuclear sclerosis was also estimated at both time points. A bivariate heritability analysis was performed to identify any shared genetic influences between refraction and nuclear sclerosis, which should reveal the presence of genetic factors contributing to both phenotypes that could inform future genetic association analyses.

Aim 1: Non-Genetic Factors Affecting Refraction and Changes in Refraction in

Older Adults

Understanding the factors contributing to refraction and changes in refraction among older persons is critical to developing improved preventative eye care services and effective treatments to counteract changes resulting in vision impairment. However, the reasons behind changes in refraction in particular over time in adulthood are not yet completely understood. Several population-based cross-sectional studies were conducted to examine the potential clinical, demographic, and environmental factors affecting refraction among older adults. These studies consistently suggested a decrease in myopia prevalence and increase in hyperopia prevalence among those over the age of 40, as well as a higher prevalence of myopia among females and those with higher family income, education level, or nuclear sclerosis grade at baseline (23–30). Evidence further suggested severe nuclear sclerosis (nuclear cataract), which becomes more prevalent with age (31–33), was associated with a higher prevalence of myopia and may cause a myopic shift in refraction among older individuals.

This supposed myopic shift follows what appears to be a hyperopic shift observed between ages 40 and 70, as evidenced by a higher prevalence of hyperopia (23,34). While nuclear cataract has been consistently associated with a higher prevalence of myopia in older persons (35–38), axial length has been associated with the higher prevalence of hyperopia prior to the increase in myopia. Axial length is the primary determinant of refraction (6,8,39,40). Wong et al (35) observed a higher prevalence of hyperopia in older

(over 70) compared to younger (under 50) persons, predominantly explained by shorter axial length. The Reykjavik Eye Study found that refraction showed the largest (negative)

10 correlation with axial length of any of the ocular refractive components studied among older individuals(41). A more recent study among Chinese Americans found axial length to be the strongest determinant of refractive error(42).

In spite of evidence supporting the association between these etiologic factors and refractive shifts over time, the etiology of the observed variation in refraction with age was not clear due to the lack of longitudinal data. Changes in the prevalence of refractive errors in cross-sectional studies do not necessarily imply age-related shifts within individuals over time, but rather could indicate cohort effects commonly observed due to differences in environmental exposures experienced by individuals born at different times. Longitudinal studies investigating changes in refraction among adults over time were necessary and therefore undertaken.

Similar to the cross-sectional studies, population-based longitudinal studies investigating change in refraction among older adults consistently reported individuals becoming more hyperopic over time before the age of 70, and then becoming more myopic after the age of 70 (43–48). The observation of a myopic shift among older individuals has repeatedly been attributed to increasing nuclear sclerosis severity with age; namely, nuclear cataract (43–48). In fact, nuclear cataract is considered to be the principal cause of the myopic shift in older individuals (13,34). However, the reasons for the hyperopic shift commonly seen among those under age 70 are still not understood. A previous study conducted in Australia among individuals over the age of 40 found no association of axial length with a ten-year hyperopic shift among those under age 65, in spite of prior evidence of an association between axial length and hyperopia in previous cross-sectional studies (43). Furthermore, associations between the change in refraction

11 over time and sex, diabetes, or education are less consistent across longitudinal studies and require further characterization (43–48).

Our lack of firm evidence supporting associations between these factors and refractive changes may be due in part to the limited availability of longitudinal data at the time of these studies. The largest change in refraction measured by any of these longitudinal studies spanned ten years (43,46). Also, these longitudinal studies restricted their analyses to associations between a single measure of change in refraction and baseline values of the covariates only (43–46). This limits our ability to examine long- term trajectories of refraction across time, or to understand how changes in certain time- varying covariates may contribute to changes in refraction over time. Longitudinal studies covering a longer period of time and allowing all relevant measures to vary across time points may improve the characterization of the long-term trajectory of refraction in adulthood.

We will evaluate the effects of nuclear sclerosis and other factors on refraction and changes in refraction using a large population-based cohort of European-Americans over age 40 from the BDES. The BDES provides a unique opportunity to analyze the relationship between changing refraction and measures of relevant clinical, environmental, and demographic factors across each of five visits over a period of 20 years. Accurate knowledge of the factors influencing the course of refraction can be useful in predicting future eye care needs among an aging population, and improve our understanding of long-term effects of refractive error treatments. To address our question, we will complete Aim 1:

Aim 1: To evaluate the association of non-genetic factors with changing refraction over a period of 20 years among adults over the age of 40 in the BDES.

Analyses on changing refraction with age have consistently ignored potential emigrative selection bias, which may be created by non-random loss to follow-up. The primary reason for loss to follow-up among individuals in the BDES cohort (and other related cohorts conducted among the elderly) is mortality (49). Past studies demonstrate a higher risk of mortality among older individuals with nuclear cataract compared to those without nuclear cataract who are typically younger (50–54). As a result, individuals lost to follow-up due to death in the BDES (or other cohorts) are more likely to be older and have had nuclear cataract compared with those who remained in the study, among other differences. This introduces a systematic bias in which individuals who are lost to follow- up are not necessarily representative of individuals who remain in the study. This bias must be accounted for to produce more accurate results of the associations of non-genetic factors with changing refraction to enhance our understanding of the etiology of the progression of refraction in older adults. Inverse probability of selection weights will be incorporated to account for this bias in Aim 1.

Aim 2: Genetic Factors Affecting Refraction in Older Adults

Investigating the genetic factors affecting refraction can provide further insight into the biological mechanisms underlying refraction among older adults, which may explain some of the etiology of refraction for future targeted treatment. Genetic heritability estimates range from 50 to 88% among European-American and Australian cohorts (including the BDES) (2,55–57), suggesting a substantial genetic component to

13 refraction. Genetic association studies have identified numerous genes associated with refraction that are involved in several common biological pathways, including those known to mediate extracellular matrix and connective tissue remodeling (58). However, in spite of the discovery of nearly 40 genes associated with refraction identified through numerous genome-wide association studies (GWAS) (3,4,59–73), only ~3.4% of the genetic variation in refraction has been accounted for to date, leaving a significant proportion of ‘missing heritability’ yet to be explained (58,60).

Refractive errors arise from a lack of precise coordination between the position of the retina relative to the cornea, which is determined by axial length, and the refractive power of the eye, which is governed by corneal curvature (2–9). Ocular axial length and corneal curvature both have a significant genetic component, with heritability estimates reaching as high as 94% and 95%, respectively (2,8,9,55–57,74–79). Evidence collectively indicates overlapping genetic mechanisms due to shared etiologies between axial length, corneal curvature, and refraction. The genetic correlation between axial length and corneal curvature has been estimated to be as high as 64% among older adults

(9). In fact, two recent GWAS discovered two novel genetic loci associated with corneal curvature that were also found to be associated with axial length (80,81). Moreover, approximately 50% of the variation in spherical equivalent, a common measurement for refraction, can be attributed to genetic factors influencing axial length. Several GWAS have indeed identified genetic variants associated with axial length that are also associated with refraction (2,6,8,55,61,82). Heritability estimates from the BDES for refraction, axial length, and corneal curvature were 58%, 67%, and 95%, respectively

(56). In the context of the BDES, a large proportion of these heritability estimates

14 represent shared genetic effects between traits. The close correlation among refraction and its two biological determinants, axial length and corneal curvature, suggests identifying genes associated with axial length and corneal curvature in addition to refraction could account for some proportion of the missing heritability and enhance our understanding of the genetic architecture of refraction.

Few association studies have identified genes associated with axial length or corneal curvature relative to the number of association studies investigating refraction or refractive error, and most that have focused on the discovery of common variants through

GWAS (3,4,9,61,67,80,83–85). Lower frequency variants may have larger effect sizes relative to common variants previously identified through GWAS (58,60). The exome array genetic data available on a subset of the BDES cohort is enriched with less common variants and provides a unique opportunity to discover previously unidentified rare

(minor allele frequency (MAF) < 1%) and low frequency (1% ≤ MAF ≤ 5%) variants associated with axial length, corneal curvature, or refraction.

Genotype imputation is an efficient statistical method for establishing denser coverage of the genome and is most commonly used on genome-wide arrays (86–88).

Imputation from the exome array may provide an efficient method for increasing our coverage of the exome, thus improving our ability to detect additional novel associations with these ocular phenotypes without the expense of exome sequencing. The imputed exome array can provide genotypes of low frequency or rare variants not often found on genome-wide arrays that could help explain a larger proportion of missing heritability and further elucidate the genetic mechanisms underlying refraction. To achieve this, we will complete Aim 2:

Aim 2: To improve coverage of the exome through imputation of the exome array, and to use the imputed data to identify novel genetic loci associated with a) refraction, and two key biological determinants of refraction: b) axial length and c) corneal curvature, among adults over the age of 40 in the BDES.

Aim 3: Shared Genetic and Environmental Effects of Refraction and Nuclear

Sclerosis

The heritability of refraction has been well-established across many populations

(2,55–57) and was estimated to be 58% (95% confidence interval: 0.33, 0.83) in the

BDES (56). The proportion of additive genetic effects contributing to nuclear sclerosis has been estimated to range from 35% to 48% (89–93). While numerous genetic loci have been associated with refraction and validated (3,4,59–73,94–96), fewer nuclear sclerosis loci have been identified and successfully replicated in genetic association studies (97), leaving the underlying genetic mechanisms of nuclear sclerosis relatively unknown. As previously mentioned, the primary factor influencing the myopic shift observed in older adults is nuclear cataract (34,43–48). The established effect of nuclear sclerosis severity on refractive shifts, in addition to the heritability of these complex, polygenic traits (2,55–57,89–91,93,98), strongly suggests refraction and nuclear sclerosis may share genetic and/or environmental effects. We sought to determine whether the genetic effects underlying refraction were influenced by differences in the distribution of nuclear sclerosis severity as a result of these shared effects, and to quantify shared genetic and/or environmental effects between refraction and nuclear sclerosis. To answer these questions, we completed Aim 3:

Aim 3: To determine whether differences in the distribution of nuclear sclerosis severity alters the heritability of refraction over time as a result of shared effects, to determine if the heritability of refraction differs across strata of nuclear sclerosis severity due to shared effects, and to quantify the shared genetic and/or environmental effects between refraction and nuclear sclerosis among individuals over 40 years of age in the BDES.

Thesis Organization

Chapters 2-4 are manuscripts describing the original research undertaken to address specific Aims 1-3, respectively. Concluding remarks, including a discussion of major findings, strengths and limitations, public health implications, and future directions of the studies described here, are presented in chapter 5. The appendix provides supplementary methods and results corresponding to the study described in chapter 3 for

Aim 2.

References

1. National Eye Institute. Facts About Refractive Errors. 2010;(https://nei.nih.gov/health/errors/errors). (Accessed August 25, 2017) 2. Dirani M, Shekar SN, Baird PN. Evidence of Shared Genes in Refraction and Axial Length: The Genes in Myopia (GEM) Twin Study. Invest. Ophthalmol. Vis. Sci. 2008;49(10):4336–4339. 3. Cheng C-Y, Schache M, Ikram MK, et al. Nine Loci for Ocular Axial Length Identified through Genome-Wide Association Studies, Including Shared Loci with Refractive Error. Am. J. Hum. Genet. 2013;93(2):264–277. 4. Fan Q, Barathi V a, Cheng C-Y, et al. Genetic Variants on Chromosome 1q41 Influence Ocular Axial Length and High Myopia. PLoS Genet. 2012;8(6):1–14.

5. Atchison DA, Jones CE, Schmid KL, et al. Eye Shape in Emmetropia and Myopia. Investig. Opthalmology Vis. Sci. 2004;45(10):3380–3386. 6. Young TL, Metlapally R, Shay AE. Complex Trait Genetics of Refractive Error. Arch. Ophthalmol. 2007;125(1):38–48. 7. Baird PN, Schäche M, Dirani M. The GEnes in Myopia (GEM) Study in Understanding the Aetiology of Refractive Errors. Prog. Retin. Eye Res. 2010;29(6):520–542. 8. Meng W, Butterworth J, Malecaze F, et al. Axial Length of Myopia: A Review of Current Research. Ophthalmologica. 2011;225(3):127–134. 9. Guggenheim J a., Zhou X, Evans DM, et al. Coordinated Genetic Scaling of the Human Eye: Shared Determination of Axial Eye Length and Corneal Curvature. Investig. Ophthalmol. Vis. Sci. 2013;54(3):1715–1721. 10. CDC. Common Eye Disorders. Vis. Heal. Initiat. 2015;(https://www.cdc.gov/visionhealth/basics/ced/index.html). (Accessed August 25, 2017) 11. Klein BE, Klein R, Linton K. Prevalence of Age-Related Lens Opacities in a Population. The Beaver Dam Eye Study. Ophthalmology. 1992;99(4):546–552. 12. Mitchell P, Cumming RG, Attebo K, et al. Prevalence of Cataract in Australia: The Blue Mountains Eye Study. Ophthalmology. 1997;104(4):581–588. 13. Brown NAP, Hill AR. Cataract : The Relation Between Myopia and Cataract Morphology. Br. J. Ophthalmol. 1987;71:405–414. 14. Brown N. The Morphology of Cataract and Visual Performance. Eye. 1993;7:63– 67. 15. Pesudovs K, Elliott DB. Refractive Error Changes in Cortical, Nuclear, and

Posterior Subcapsular Cataracts. Br J Ophthalmol. 2003;87:964–967. 16. Laser Eye Surgery in Developing Countries and Communities. (http://lasereyesurgeryjb.weebly.com/myopia-and-hyperopia.html). (Accessed August 30, 2017) 17. Label Eye Diagram. (http://4uxj7d6q.adtddns.asia/label-eye-diagram.html). (Accessed August 30, 2017) 18. WHO. What is a Refractive Error? 2017;(http://www.who.int/features/qa/45/en/). (Accessed August 4, 2017) 19. WHO. Visual Impairment and Blindness. 2017;(http://www.who.int/mediacentre/factsheets/fs282/en/). (Accessed August 4, 2017) 20. Smith TST, Frick KD, Holden BA, et al. Potential Lost Productivity Resulting From the Global Burden of Uncorrected Refractive Error. Bull. World Health Organ. 2009;87(6):431–437. 21. Kempen JH, Mitchell P, Lee KE, et al. The Prevalence of Refractive Errors Among Adults in the United States, Western Europe, and Australia. Arch. Ophthalmol. 2004;122:495–505. 22. He W, Goodkind D, Kowal P. An Aging World : 2015. Washington, DC: 2016. 23. Kurokawa T, Katai N. Refractive Assocaitions with Cataract: the Blue Mountains Eye Study. Invest. Ophthalmol. Vis. Sci. 1999;40(12):3021–3026. 24. Katz J, Tielsch JM, Sommer A. Prevalence and Risk Factors for Refractive Errors in an Adult Inner City Population. Invest. Ophthalmol. Vis. Sci. 1997;38(2):334– 340. 25. Mirshahi A, Ponto K a., Hoehn R, et al. Myopia and Level of Education. Results from the Gutenberg Health Study. Ophthalmology. 2014;121(10):2047–2052. 26. Aine E. Refractive Errors in a Finnish Rural Population. Acta Ophthalmol. 1984;62(6):944–954. 27. Sperduto RD, Seigel D, Roberts J, et al. Prevalence of myopia in the United States. Arch. Ophthalmol. 1983;101(3):405–407. 28. Lavery JR, Gibson JM, Shaw DE, et al. Refraction and Refractive Errors in an Elderly Population. Ophthalmic Physiol Opt. 1988;8(4):394–396. 29. Wang Q, Klein BEK, Klein R, et al. Refractive Status in the Beaver Dam Eye Study. Investig. Ophthalmol. Vis. Sci. 1994;35(13):4344–4347. 30. Attebo K, Ivers RQ, Mitchell P. Refractive Errors in an Older Population: The Blue Mountains Eye Study. Ophthalmology. 1999;106(6):1066–1072.

31. Klein BEK, Klein R, Lee KE, et al. Incidence of Age-Related Cataract over a 15- Year Interval The Beaver Dam Eye Study. Ophthalmology. 2008;115(3):477–482. 32. Leske MC, Wu S, Nemesure B, et al. Incidence and Progression of Lens Opacities in the Barbados Eye Studies. Ophthalmology. 2000;107(7):1267–1273. 33. Leske MC, Chylack MPHLT, Wu S. Incidence and Progression of Nuclear Opacities in the Longitudinal Study of Cataract. Ophthalmology. 1996;103(5):705–712. 34. Samarawickrama C, Wang JJ, Burlutsky G, et al. Nuclear Cataract and Myopic Shift in Refraction. Am. J. Ophthalmol. 2007;144(3):457–459. 35. Wong TY, Foster PJ, Ng TP, et al. Variations in Ocular Biometry in an Adult Chinese Population in Singapore: The Tanjong Pagar Survey. Invest Ophthalmol Vis Sci. 2001;42(1):73–80. 36. Wong TY, Foster PJ, Johnson GJ, et al. Refractive Errors, Axial Ocular Dimensions, and Age-Related Cataracts: The Tanjong Pagar Survey. Invest Ophthalmol Vis Sci. 2003;44(4):1479–1485. 37. Warrier S, Wu HM, Newland HS, et al. Ocular Biometry and Determinants of Refractive Error in Rural Myanmar: The Meiktila Eye Study. Br J Ophthalmol. 2008;92:1591–1594. 38. Shufelt C, Fraser-bell S, Ying-lai M, et al. Opalescence in an Adult Population : The Los Angeles Latino Eye Study. Invest Ophthalmol Vis Sci. 2005;46(12):4450– 4460. 39. Foster PJ, Jiang Y. Epidemiology of Myopia. Eye. 2014;28(2):202–208. 40. Flitcroft DI. Emmetropisation and the Aetiology of Refractive Errors. Eye. 2014;28(2):169–179. 41. Olsen T, Arnarsson A, Sasaki H, et al. On the Ocular Refractive Components: The Reykjavik Eye Study. Acta Ophthalmol. Scand. 2007;85(4):361–366. 42. Richter GM, Wang M, Jiang X, et al. Ocular Determinants of Refractive Error and Its Age- and Sex-Related Variations in the Chinese American Eye Study. JAMA Ophthalmol. 2017;135(7):724–732. 43. Fotedar R, Mitchell P, Burlutsky G, et al. Relationship of 10-Year Change in Refraction to Nuclear Cataract and Axial Length Findings From an Older Population. Ophthalmology. 2008;115(8):1273–1278. 44. Guzowski M, Wang JJ, Rochtchina E, et al. Five-Year Refractive Changes in an Older Population: The Blue Mountains Eye Study. Ophthalmology. 2003;110(7):1364–1370. 45. Lee KE, Klein BE, Klein R. Changes in Refractive Error Over a 5-Year Interval in

the Beaver Dam Eye Study. Invest. Ophthalmol. Vis. Sci. 1999;40(8):1645–1649. 46. Lee KE, Klein BEK, Klein R, et al. Changes in Refraction Over 10 Years in an Adult Population: The Beaver Dam Eye Study. Invest. Ophthalmol. Vis. Sci. 2002;43(8):2566–2571. 47. Gudmundsdottir E, Arnarsson A, Jonasson F. Five-Year Refractive Changes in an Adult Population: Reykjavik Eye Study. Ophthalmology. 2005;112(4):672–677. 48. Panchapakesan J, Rochtchina E, Mitchell P. Myopic Refractive Shift Caused by Incident Cataract: The Blue Mountains Eye Study. Ophthalmic Epidemiol. 2003;10(4):241–247. 49. Klein R, Lee KE, Gangnon RE, et al. Incidence of Visual Impairment Over a 20- Year Period: The Beaver Dam Eye Study. Ophthalmology. 2013;120(6):1210– 1219. 50. Wang JJ, Mitchell P, Simpson JM, et al. Visual Impairment, Age-Related Cataract, and Mortality. Arch. Ophthalmol. 2001;119:1186–1190. 51. Fong CSU, Mitchell P, Rochtchina E, et al. Correction of Visual Impairment by Cataract Surgery and Improved Survival in Older Persons: The Blue Mountains Eye Study Cohort. Ophthalmology. 2013;120(9):1720–1727. 52. Song E, Sun H, Xu Y, et al. Age-Related Cataract, Cataract Surgery and Subsequent Mortality: A Systematic Review and Meta-Analysis. PLoS One. 2014;9(11):e112054. 53. Cugati S, Cumming R, Smith W, et al. Visual Impairment, Age-Related Macular Degeneration, Cataract, and Long-term Mortality. Arch. Ophthalmol. 2007;125(7):917–924. 54. Knudtson M, Klein BEK, Klein R. Age-Related Eye Disease, Visual Impairment, and Survival. Arch. Ophthalmol. 2006;124:243–249. 55. Dirani M, Chamberlain M, Shekar SN, et al. Heritability of Refractive Error and Ocular Biometrics: The Genes in Myopia (GEM) Twin Study. Invest. Ophthalmol. Vis. Sci. 2006;47(11):4756–4761. 56. Klein AP, Suktitipat B, Duggal P, et al. Heritability Analysis of Spherical Equivalent, Axial Length, Corneal Curvature, and Anterior Chamber Depth in the Beaver Dam Eye Study. Arch. Ophthalmol. 2009;127(5):649–655. 57. Chen CYC, Scurrah KJ, Stankovich J, et al. Heritability and Shared Environment Estimates for Myopia and Associated Ocular Biometric Traits: The Genes in Myopia (GEM) Family Study. Hum. Genet. 2007;121:511–520. 58. Wojciechowski R. Nature and Nurture: The Complex Genetics of Myopia and Refractive Error. Clin. Genet. 2011;79(4):301–320.

59. Hawthorne F a., Young TL. Genetic Contributions to Myopic Refractive Error: Insights from Human Studies and Supporting Evidence from Animal Models. Exp. Eye Res. 2013;114:141–149. 60. Hysi PG, Wojciechowski R, Rahi JS, et al. Genome-Wide Association Studies of Refractive Error and Myopia, Lessons Learned, and Implications for the Future. Investig. Opthalmology Vis. Sci. 2014;55(5):3344–3351. 61. Chen JH, Chen H, Huang S, et al. Endophenotyping Reveals Differential Phenotype-Genotype Correlations Between Myopia-Associated Polymorphisms and Eye Biometric Parameters. Mol. Vis. 2012;18:765–778. 62. Hysi PG, Young TL, Mackey D, et al. A Genome-Wide Association Study for Myopia and Refractive Error Identifies a Susceptibility Locus at 15q25. Nat. Genet. 2010;42(10):902–905.

63. Khor CC, Miyake M, Chen LJ, et al. Genome-Wide Association Study Identifies ZFHX1B as a Susceptibility Locus for Severe Myopia. Hum. Mol. Genet. 2013;22(25):5288–5294. 64. Kiefer AK, Tung JY, Do CB, et al. Genome-Wide Analysis Points to Roles for Extracellular Matrix Remodeling, the Visual Cycle, and Neuronal Development in Myopia. PLoS Genet. 2013;9(2):1–8. 65. Li Z, Qu J, Xu X, et al. A Genome-Wide Association Study Reveals Association Between Common Variants in an Intergenic Region of 4q25 and High-Grade Myopia in the Chinese Han Population. Hum. Mol. Genet. 2011;20(14):2861– 2868. 66. Li Y-J, Goh L, Khor C-C, et al. Genome-Wide Association Studies Reveal Genetic Variants in CTNND2 for High Myopia in Singapore Chinese. Ophthalmology. 2011;118(2):368–375. 67. Miyake M, Yamashiro K, Tabara Y, et al. Identification of Myopia-Associated WNT7B Polymorphisms Provides Insights into the Mechanism Underlying the Development of Myopia. Nat. Commun. 2015;6(6689):1–9. 68. Nakanishi H, Yamada R, Gotoh N, et al. A Genome-Wide Association Analysis Identified a Novel Susceptible Locus for Pathological Myopia at 11q24.1. PLoS Genet. 2009;5(9):1–7. 69. Shi Y, Qu J, Zhang D, et al. Genetic Variants at 13q12.12 are Associated with High Myopia in the Han Chinese Population. Am. J. Hum. Genet. 2011;88(6):805– 813. 70. Shi Y, Gong B, Chen L, et al. A Genome-Wide Meta-Analysis Identifies Two Novel Loci Associated with High Myopia in the Han Chinese Population. Hum. Mol. Genet. 2013;22(11):2325–2333. 71. Simpson CL, Wojciechowski R, Oexle K, et al. Genome-Wide Meta-Analysis of

Myopia and Hyperopia Provides Evidence for Replication of 11 Loci. PLoS One. 2014;9(9):1–19. 72. Solouki AM, Verhoeven VJM, van Duijn CM, et al. A Genome-Wide Association Study Identifies a Susceptibility Locus for Refractive Errors and Myopia at 15q14. Nat. Genet. 2010;42(10):897–901. 73. Verhoeven VJM, Hysi PG, Wojciechowski R, et al. Genome-Wide Meta-Analyses of Multiancestry Cohorts Identify Multiple New Susceptibility Loci for Refractive Error and Myopia. Nat. Genet. 2013;45(3):314–318. 74. Hammond CJ, Snieder H, Gilbert CE, et al. Genes and Environment in Refractive Error: The Twin Eye Study. Investig. Ophthalmol. Vis. Sci. 2001;42(6):1232– 1236. 75. Biino G, Palmas MA, Corona C, et al. Ocular Refraction: Heritability and Genome-Wide Search for Eye Morphometry Traits in an Isolated Sardinian Population. Hum. Genet. 2005;116(3):152–159. 76. Lyhne N, Sjølie a K, Kyvik KO, et al. The Importance of Genes and Environment for Ocular Refraction and its Determiners: A Population Based Study Among 20- 45 Year Old Twins. Br. J. Ophthalmol. 2001;85(12):1470–1476. 77. Teikari J, O’Donnell JJ, Kaprio J, et al. Genetic and Environmental Effects on Oculometric Traits. Optom. Vis. Sci. 1989;66(9):594–599. 78. Dirani M, Islam A, Shekar SN, et al. Dominant Genetic Effects on Corneal Astigmatism: The Genes in Myopia (GEM) Twin Study. Invest Ophthalmol Vis Sci. 2008;49(4):1339–1344. 79. Valluri S, Minkovitz JB, Budak K, et al. Comparative Corneal Topography and Refractive Variables in Monozygotic and Dizygotic Twins. Am. J. Ophthalmol. 1999;127(2):158–163. 80. Chen P, Miyake M, Fan Q, et al. CMPK1 and RBP3 are Associated with Corneal Curvature in Asian Populations. Hum. Mol. Genet. 2014;23(22):6129–6136. 81. Guggenheim J a, McMahon G, Kemp JP, et al. A Genome-Wide Association Study for Corneal Curvature Identifies the Platelet-Derived Growth Factor Receptor alpha Gene as a Quantitative Trait Locus for Eye Size in White Europeans. Mol. Vis. 2013;19:243–253. 82. Paget S, Vitezica ZG, Malecaze F, et al. Heritability of Refractive Value and Ocular Biometrics. Exp. Eye Res. 2008;86(2):290–295. 83. Schache M, Richardson AJ, Mitchell P, et al. Genetic Association of Refractive Error and Axial Length with 15q14 but not 15q25 in the Blue Mountains Eye Study Cohort. Ophthalmology. 2013;120(2):292–297. 84. Han S, Chen P, Fan Q, et al. Association of Variants in FRAP1 and PDGFRA with

Corneal Curvature in Asian Populations from Singapore. Hum. Mol. Genet. 2011;20(18):3693–3698. 85. Mishra A, Yazar S, Hewitt AW, et al. Genetic Variants Near PDGFRA Are Associated with Corneal Curvature in Australians. Invest. Ophthalmol. Vis. Sci. 2012;53(11):7131–7136. 86. Delaneau O, Marchini J, Zagury J-F. A Linear Complexity Phasing Method for Thousands of Genomes. Nat. Methods. 2012;9(2):179–181. 87. Delaneau O, Zagury J, Marchini J. Improved Whole-Chromosome Phasing for Disease and Population Genetic Studies. Nat. Methods. 2013;10(1):5–6. 88. Das S, Forer L, Schönherr S, et al. Next-Generation Genotype Imputation Service and Methods. Nat. Genet. 2016;48(10):1284–1287. 89. Yonova-Doing E, Forkin ZA, Hysi PG, et al. Genetic and Dietary Factors Influencing the Progression of Nuclear Cataract. Ophthalmology. 2016;123(6):1–8. 90. Hammond CJ, F.R.C.Ophth., Snieder H, et al. Genetic and Environmental Factors in Age-Related Nuclear Cataracts in Monozygotic and Dizygotic Twins. N. Engl. J. Med. 2000;342(24):1786–1790. 91. Congdon N, Broman KW, Lai H, et al. Nuclear Cataract Shows Significant Familial Aggregation in an Older Population after Adjustment for Possible Shared Environmental Factors. Invest Ophthalmol Vis Sci. 2004;45(7):2182–2186. 92. Heiba IM, Elston RC, Klein BEK, et al. Genetic Etiology of Nuclear Cataract: Evidence for a Major Gene. Am. J. Med. Genet. 1993;47(8):1208–1214. 93. Klein AP, Duggal P, Lee KE, et al. Support for Polygenic Infuences on Ocular Refractive Error. Investig. Ophthalmol. Vis. Sci. 2005;46(2):442–446. 94. Klein AP, Duggal P, Lee KE, et al. Linkage Analysis of Quantitative Refraction and Refractive Errors in the Beaver Dam Eye Study. Invest. Ophthalmol. Vis. Sci. 2011;52(8):5220–5225. 95. Klein AP, Duggal P, Lee KE, et al. Confirmation of Linkage to Ocular Refraction on Chromosome 22q and Identification of a Novel Linkage Region on 1q. Arch. Ophthalmol. 2007;125(1):80–85. 96. Wojciechowski R. Nature and Nurture: The Complex Genetics of Myopia and Refractive Error. Clin. Genet. 2011;79(4):301–320. 97. Liao J, Su X, Chen P, et al. Meta-Analysis of Genome-Wide Association Studies in Multiethnic Asians Identifies Two Loci for Age-Related Nuclear Cataract. Hum. Mol. Genet. 2014;23(22):6119–6128. 98. Klein AP, Duggal P, Lee KE, et al. Polygenic Effects of Cigarette Smoking Account for a Portion of the Familial Aggregation of Nuclear Sclerosis. Am. J.

Epidemiol. 2005;161(8):707–713.

CHAPTER TWO

Refraction and Change in Refraction Over a 20-Year Period in the Beaver Dam Eye

Study

Abstract

Hyperopic shifts in refraction have been consistently reported in adults over 40, followed by myopic shifts after age 70. Although previous studies attribute the myopic shift to nuclear cataract, contributions of other potential risk factors in older adults are incompletely understood. This may be due to limited longitudinal data and the use of baseline measures to assess risk factors in previous studies. To better characterize the trajectory of refraction in older adults, we evaluated the etiologic factors underlying changes in refraction measured over a 20-year period (1988-2010) among adults over age

40 from Beaver Dam, Wisconsin. Only those individuals with nuclear cataract experienced a myopic shift in refraction, showing a 0.245 diopter decrease (95% confidence interval [-0.417 diopters, -0.074 diopters]) over a five-year period. Individuals with mild and moderate nuclear sclerosis experienced varying degrees of hyperopic shifts over a five-year period (0.231 D [0.206, 0.257] and 0.251 D [0.215, 0.287], respectively).

Sex, diabetes status, and baseline refraction, but not education, were associated with refraction, although the effects were small. Nuclear cataract is the primary contributor to the myopic shift among older individuals. Other risk factors also contribute to refraction.

Introduction

Refractive errors, namely myopia (nearsightedness) and hyperopia

(farsightedness), are the most common causes of visual impairment in the United States and the world (1–4). An estimated 153 million individuals worldwide are visually impaired from uncorrected refractive errors, and nearly two-thirds of these individuals are over the age of 50 (5). In addition, over 25% of those 40 years and older in western

European countries and in the United States have refractive errors resulting in impaired vision (4,6). The heavily affected aging population (7) therefore demands accurate data on the natural history of refraction among older adults to anticipate their future eye care needs and to tailor treatments to fit their needs.

Population-based studies investigating changes in refraction among older adults have consistently reported individuals becoming hyperopic with age before the age of 70, and then more myopic after the age of 70. The reasons underlying the observed hyperopic shift prior to the myopic shift have been attributed to age-related decreases in axial length, which is established as the most important biological determinant of refraction

(4,8–10). The observed myopic shift among older individuals has been attributed primarily to the effects of nuclear sclerosis severity, namely age-related nuclear cataract

(11–17). Sex, diabetes, baseline refraction, and education have also been associated with refractive changes, but the studies have been equivocal (11–16).

Previous studies have only examined changes in refraction over periods of five or ten years due to the limited longitudinal available at the time, and these studies used baseline values of relevant factors only (11–15). This study will characterize the long- term refractive changes in adulthood using time-updated measures collected from each

28 visit over 20 years. We sought to evaluate the associations of etiologic factors including nuclear sclerosis, sex, diabetes, and education with refraction and refractive shifts over a

20-year period among European-Americans over age 40 from the Beaver Dam Eye Study

(BDES).

Methods

The Beaver Dam Eye Study

Recruitment and study design procedures for the BDES have been described in detail previously (18–21). Briefly, a private census was conducted in the city and township of Beaver Dam, Wisconsin beginning in 1987. The census identified 5,924 residents between the ages of 43 and 84 from 3,715 households. A total of 602 pedigrees were reconstructed from 2,783 eligible participants who had confirmed familial relationships. Of the 5,924 eligible residents, 83.1% (n = 4,926) participated in the baseline examination. Ninety-nine percent of the population was of European ancestry by self-report.

The baseline ocular examination included a standardized evaluation of refraction using the Humphrey 530 refractor (Humphrey, Humphrey Inc., San Leandro, CA) (18).

The standard formula for refraction (sphere + 0.5 x cylinder) was used to calculate the mean spherical equivalent in diopters (D) for each eye. Grades of the severity of nuclear lens opacity (nuclear sclerosis) were assigned by trained photograders using a five-point scale based on slitlamp lens photographs (22,23). Age, sex, years of education, and diabetes status were collected from a personal history questionnaire conducted at the baseline examination. An individual was considered to have diabetes if there was a self-

29 report of diabetes in conjunction with treatment (insulin or diet), or elevated glucose or glycosylated hemoglobin levels.

Follow-up examinations were conducted every five years. A total of 3,721 (five years), 2,962 (ten years), 2,375 (15 years), and 1,913 (20 years) individuals participated in the follow-up examinations, respectively. A majority (68.9%) of losses to follow-up were due to death. Measurements made during the baseline examination were repeated using similar procedures for all follow-up examinations. Sex and years of education were carried forward from the baseline visit.

The Institutional Review Board at the University of Wisconsin approved the study. Written informed consent was obtained from all subjects prior to enrollment. The study was performed in accordance with the tenets of the Declaration of Helsinki.

Trait Definitions

Refraction was measured as a quantitative trait of spherical equivalent at each visit and was the primary outcome. Values of spherical equivalent < -0.5 D indicated myopia and values > +0.5 D indicated hyperopia. Shifts toward myopia were defined as changes in spherical equivalent over time resulting in a negative change, while shifts toward hyperopia were defined as changes in spherical equivalent resulting in a positive change (12,24). Nuclear sclerosis grade was categorized based on grade at each visit: mild (grade of 1 or 2), moderate (grade of 3), or severe (grade of 4 or 5). Severe nuclear sclerosis (grades of 4 or 5) indicate nuclear cataract.

Exclusion Criteria

Individuals of non-European ancestry (n = 31), individuals with differences in baseline spherical equivalent between the right and left eyes > ±4 D (n = 23), or

30 individuals without baseline examination (n = 45) measures were excluded from the analysis. In addition, eyes that had undergone refractive surgery, eyes with no lens, eyes with an intraocular lens, or eyes with best corrected visual acuity of 20/200 or worse at the time of the baseline examination were excluded from the study because of reduced reliability and increased variability of refraction measurements (n = 319). Eyes developing one of these conditions at subsequent visits were censored upon development of the specific condition. If a participant failed to attend a visit but returned for a subsequent visit, only the absentee visit was removed. Finally, individuals with missing data for spherical equivalent, nuclear sclerosis, education, or diabetes were excluded (n =

113), resulting in 4,441 individuals for analysis. The full BDES cohort includes all individuals in the study cohort in addition to individuals excluded from the analysis (n =

4,972).

Statistical Analysis

Longitudinal analysis was modeled using generalized estimating equations with robust variance estimation in Stata version 13 (25–27). The model was:

Refractionij = β0+ β1agei1 + β2(ageij-1-agei1) + β3(nuclear sclerosisij-1 = moderate) + β4(nuclear sclerosisij-1 = severe) + β5(ageij-1-agei1)*(nuclear sclerosisij-1 = moderate) + β6(ageij-1-agei1)*(nuclear sclerosisij-1 = severe) + β7educationi1 + β8sexi1 + β9diabetesij-1 + β10refractioni1 + εij where ‘i’ represents the individual and ‘j’ represents visit number. After consideration of the correlation structure of the data, an independent working correlation structure was employed (28). Refraction was used as a time-varying outcome and age, nuclear sclerosis severity, and diabetes were included as time-varying covariates. All time-varying covariates were lagged by one visit, and values for the outcome began at visit 2 to establish temporality. Age was partitioned into baseline age and change in age from

31 baseline to distinguish baseline cohort effects from longitudinal effects of aging. Sex, education, and baseline refraction were carried forward from baseline to subsequent visits. Interactions between risk factors were considered. Baseline age, education, and baseline refraction were centered at their mean value. Beta coefficients for each factor in the model were reported and used to estimate the change in refraction for each increase in level of the relevant factor. These beta coefficients were also used to estimate the five and

20-year changes in refraction within each category of nuclear sclerosis severity in Table

3. The left and right eyes behaved similarly, so only the results of the right eye are presented. Birth cohorts (Figure 1) were defined by year of birth and grouped into four- year intervals to evenly distribute samples across groups. Birth cohorts with less than 200 observations were discarded in each figure.

Inverse Probability Weighting

Inverse probability weighting was employed to standardize the estimation of change in refraction from the subset of individuals who were not lost to follow-up over the course of the study to the original BDES cohort recruited at baseline (29–31). We calculated three inverse probability weights to account for the three reasons for loss to follow-up: 1) death, 2) refusal to participate or selective participation in the interview component of the visit only, and 3) censoring following insertion of an intraocular lens or removal of lens, refractive surgery, or best corrected visual acuity of 20/200 or worse.

These weights were combined and incorporated into the longitudinal analysis.

Results

Study Participants

Table 1 shows the distributions of the study cohort (n = 4,441) and the full BDES cohort (n = 4,972). The full BDES cohort was slightly older at baseline and had a higher proportion of individuals with diabetes and nuclear cataract (Table 1), but otherwise there are negligible differences between the two groups at baseline. Because the BDES recruited adults over 40 years of age at baseline, nearly one-third (32.5%) of the study cohort was lost to follow-up due to death at some point during the course of the study. A total of 16.1% refused to continue to participate or participated in the interview component of the examination only, and 24.5% were censored due to insertion of an intraocular lens or removal of lens, refractive surgery, or best corrected visual acuity of

20/200 or worse.

Age and Birth Cohort

Figure 1A shows the changes in refraction by age and birth cohort among all participants. Each birth cohort followed similar patterns of changing refraction with age, as evidenced by the consistent slopes representing these changes in refraction. The trajectories of refraction with age, regardless of birth cohort, showed a hyperopic shift among those under the age 70 followed by a transition to a myopic shift among those older than 70. However, values of refraction at a given age did differ across birth cohorts, with individuals born more recently being more myopic compared to individuals born earlier. After about age 70, refraction no longer appeared to differ across cohorts.

Nuclear Sclerosis

The relationship of changing refraction with age varied by nuclear sclerosis severity in the BDES (Figures 1B-1D). Individuals with mild nuclear sclerosis consistently showed positive slopes, indicating hyperopic changes in refraction with age,

33 regardless of birth cohort (Figure 1B). Individuals with moderate nuclear sclerosis followed a similar pattern (Figure 1C). Only those with nuclear cataract (severe nuclear sclerosis) had negative slopes, indicating myopic shifts in refraction with age (Figure

1D). As participants aged, nuclear sclerosis severity increased.

This longitudinal model suggests the trajectory of refraction with age differs depending on nuclear sclerosis severity (Table 2). An annual 0.049 D (95% confidence interval (CI): -0.083 D, -0.015 D) decrease (myopic shift) in refraction was observed among individuals with nuclear cataract after adjustment for baseline age, baseline refraction, sex, education, and diabetes status. Among individuals with mild (0.046 D,

95% CI: 0.041 D, 0.051 D) or moderate (0.050 D, 95% CI: 0.043 D, 0.057 D) nuclear sclerosis, a positive increase (hyperopic shift) in refraction each year was observed after inclusion of the same covariates in the model (Table 2). Those with mild and moderate nuclear sclerosis experienced a 0.231 D (95% CI: 0.206 D, 0.257 D) and 0.251 D (95%

CI: 0.215 D, 0.287 D) increase in refraction, respectively, over a five-year period while those with nuclear cataract experienced a 0.245 D (95% CI: -0.417 D, -0.074 D) decrease in refraction over the same time period (Table 3). Similar patterns of increasing refraction for those without nuclear cataract and decreasing refraction for those with nuclear cataract were observed over the 20-year period (Table 3).

Risk Factors Associated With Refraction

Sex, diabetes status, and baseline refraction were significantly associated with refraction (Table 2). Males had a 0.149 D (95% CI: -0.211 D, -0.087 D) lower refraction compared to females, whereas diabetics had a 0.237 D (95% CI: 0.122 D, 0.352 D) higher refraction compared to non-diabetics, adjusted for other factors (Table 2).

Education level was not associated with refraction. None of these factors were significantly associated with change in refraction resulting from an interaction with change in age.

Discussion

This study confirms a hyperopic shift among individuals aged 40 to 70, followed by a clear myopic shift after age 70 (11–15). These ocular shifts were observed independent of birth cohort, indicating changes in refraction over time in older adults are primarily due to the longitudinal effects of aging. The year in which the individuals were born did not alter their trajectory of changing refraction in adulthood. However, refraction was affected by birth cohort; prior to the age of 70, individuals born more recently were more myopic. These cohort effects largely disappeared after age 70, but this is likely due to loss to follow-up from death.

Previous studies established that nuclear sclerosis severity, which worsens with age (32–39), largely influences the myopic shifts among older individuals (16,17). Only those with nuclear cataract showed a myopic shift independent of birth cohort, while those without nuclear cataract showed a hyperopic shift. Although the observed myopic shift was small in magnitude over one year or even five years, it demonstrates a change in the trajectory of refraction among those with nuclear cataract distinct from those without nuclear cataract. An average participant with a baseline refraction of 0.25 D in this study would have been nearly twice as hyperopic after five years if they did not have nuclear cataract, while they would no longer be hyperopic after five years if they did have nuclear cataract. Over a 20 year-period, our study indicated that those without nuclear cataract

35 experienced approximately a 1 D hyperopic shift in refraction, while those with nuclear cataract experienced a roughly 1 D myopic shift in refraction. Even though this change occurs over a long period of time, it indicates an etiologic effect of nuclear cataract on refractive shifts. As older individuals continue to age, they will be more likely to develop nuclear cataract and thus more likely to experience a myopic shift in refraction. This information allows clinicians to predict future changes in refraction in their patients, and should inform preventative care and treatment of older individuals at risk for nuclear cataract to minimize vision loss.

Several cross-sectional studies established a significant association between nuclear lens opacity and prevalence of myopia (40–43). Lee et al (11) documented the strong relationship between nuclear sclerosis severity and myopic changes in refraction in the BDES over a five-year interval, which was also validated in a ten-year interval (12).

The Blue Mountains Eye Study confirmed this finding (14,15) and concluded nuclear cataract was the principal cause of the myopic shift in refraction among older people

(17). In this current 20-year study, we confirm nuclear cataract is a principal contributor to the observed myopic shift in the aging population by demonstrating changes in refraction with age vary significantly by nuclear sclerosis severity, and only those who have developed nuclear cataract experience a directional change in their trajectory of refraction while those without nuclear cataract do not.

Changes in the lens as individuals age likely explain the observed myopic shifts.

Lens thickness and steepness of curvature increase with age, which would imply trends toward myopia (44–47). Decreases in the refractive index of the lens with increasing age usually compensate for these changes, acting as an emmetropizing mechanism which

36 leads to a hyperopic shift among many individuals known as the lens paradox (44–47).

However, previous research has suggested the focal length of the lens increases before age 65 and decreases thereafter (48). A decrease in focal length can lead to an increase in refractive index, potentially contributing to a myopic shift in older persons. An increase in the degree of light scatter in the lens has also been related to the increases in refractive index and resulting myopic shifts among those with nuclear cataract (46,49).

Although several studies have established a cross-sectional association between shorter axial length and hyperopia (40,50,51), longitudinal studies showing an association between shortening axial length and hyperopic shifts in refraction among older individuals are limited (15). We lack longitudinal measures of axial length from baseline to evaluate axial length as a primary determinant of the hyperopic shift prior to the myopic shift induced by nuclear cataract.

Of the factors associated with refraction, diabetes is the most modifiable factor associated with refraction in this study. Those with diabetes showed a higher refraction compared to non-diabetics, indicating having diabetes may be a risk factor for refractive errors. Education level, which is commonly used as a surrogate for near work, did not affect refraction in this study. Although education has been repeatedly associated with myopia in cross-sectional studies (52–59), the level of education received does not appear to influence refraction or changes in refraction in older adults. This is likely because older adults received a majority of their education at a younger age, so while education may have had a larger impact on myopia when they were young, it does not have a large influence on refraction or refractive shifts at older ages. Being male was associated with a lower refraction compared to women in this study, and baseline refraction influenced the

37 trajectory of refraction beyond baseline, as expected. Although sex and baseline refraction are not modifiable, understanding the effects of both modifiable and nonmodifiable risk factors can provide insight into the likely changes in the trajectory of refraction among individuals who fit these criteria.

Diabetes and sex have been previously identified as risk factors for nuclear cataract in older populations (60–62). Since nuclear cataract is an established risk factor for refraction, it makes sense that diabetes and sex would also influence refractive change in addition to nuclear cataract. To avoid over-adjustment for factors potentially in the causal pathway for nuclear cataract and refractive shifts, we conducted a sensitivity analysis where we removed sex and diabetes from the model. The associations between nuclear sclerosis, age, and change in refraction were unchanged, indicating diabetes and sex in the final model did not affect the reported findings.

There are several strengths to this study. The original cohort recruited for the baseline examination constituted 83% of all individuals in Beaver Dam in 1987 between the ages of 43 and 86, making the study sample representative of older adults in Beaver

Dam at the time. This makes our study results generalizable to other elderly European

American populations. We had access to 20 years of follow-up measurements, which allowed us to model the trajectories of refraction across multiple time points over longer periods of time, and use time-updated values of all relevant factors. To our knowledge, this is the longest amount of time covered for any population-based study investigating changes in refraction (12,15).

The etiology of refraction and refractive shifts among older individuals is important to characterize and understand to anticipate future public health needs. As this

38 and other studies have shown, myopia in particular is becoming more common among older individuals, largely due to nuclear cataract (63,64). There is ultimately a greater need for enhanced screening and preventative eye care services to address this growing burden in older adults. For example, those at higher risk for cataract can be more closely monitored for corrective measures before uncorrected refractive errors become too severe. Information on how refraction changes in older persons can also be useful for refractive surgeons trying to predict changes in refraction that are likely to occur in their patients, and tailor surgeries and other treatments accordingly.

Acknowledgements

This work was supported by the National Eye Institute at the National Institutes of Health

(grants R01EY021531, U10006594, and 1T32EI022303); and by the Research to Prevent

Blindness Unrestricted Grant to the University of Wisconsin Department of

Ophthalmology and Visual Sciences.

We thank all staff and investigators of the Beaver Dam Eye Study.

Conflict of interest: none declared.

References

1. Hysi PG, Wojciechowski R, Rahi JS, et al. Genome-Wide Association Studies of Refractive Error and Myopia, Lessons Learned, and Implications for the Future. Investig. Opthalmology Vis. Sci. 2014;55(5):3344–3351. 2. National Eye Institute. Fact Sheet: Refractive Errors. 2010;(https://nei.nih.gov/sites/default/files/health- pdfs/HVM09_Fact_Sheet_Final_tagged.pdf). (Accessed November 9, 2015) 3. Centers for Disease Control and Prevention. Refractive Errors. Vis. Heal. Initiat. 2013;(http://www.cdc.gov/visionhealth/basic_information/eye_disorders.htm#a2). (Accessed November 9, 2015) 4. Meng W, Butterworth J, Malecaze F, et al. Axial Length of Myopia: A Review of Current Research. Ophthalmologica. 2011;225(3):127–134. 5. Resnikoff S, Pascolini D, Mariotti S, et al. Global Magnitude of Visual Impairment Caused by Uncorrected Refractive Errors in 2004. Bull. World Health Organ. 2008;86(1):63–70. 6. Kempen JH, Mitchell P, Lee KE, et al. The Prevalence of Refractive Errors Among Adults in the United States, Western Europe, and Australia. Arch. Ophthalmol. 2004;122:495–505. 7. He W, Goodkind D, Kowal P. An Aging World : 2015. Washington, DC: 2016. 8. Young TL, Metlapally R, Shay AE. Complex Trait Genetics of Refractive Error. Arch. Ophthalmol. 2007;125(1):38–48. 9. Foster PJ, Jiang Y. Epidemiology of Myopia. Eye. 2014;28(2):202–208. 10. Flitcroft DI. Emmetropisation and the Aetiology of Refractive Errors. Eye. 2014;28(2):169–179. 11. Lee KE, Klein BE, Klein R. Changes in Refractive Error Over a 5-Year Interval in the Beaver Dam Eye Study. Invest. Ophthalmol. Vis. Sci. 1999;40(8):1645–1649. 12. Lee KE, Klein BEK, Klein R, et al. Changes in Refraction Over 10 Years in an Adult Population: The Beaver Dam Eye Study. Invest. Ophthalmol. Vis. Sci. 2002;43(8):2566–2571. 13. Gudmundsdottir E, Arnarsson A, Jonasson F. Five-Year Refractive Changes in an Adult Population: Reykjavik Eye Study. Ophthalmology. 2005;112(4):672–677. 14. Guzowski M, Wang JJ, Rochtchina E, et al. Five-Year Refractive Changes in an Older Population: The Blue Mountains Eye Study. Ophthalmology. 2003;110(7):1364–1370. 15. Fotedar R, Mitchell P, Burlutsky G, et al. Relationship of 10-Year Change in

Refraction to Nuclear Cataract and Axial Length Findings From an Older Population. Ophthalmology. 2008;115(8):1273–1278. 16. Panchapakesan J, Rochtchina E, Mitchell P. Myopic Refractive Shift Caused by Incident Cataract: The Blue Mountains Eye Study. Ophthalmic Epidemiol. 2003;10(4):241–247. 17. Samarawickrama C, Wang JJ, Burlutsky G, et al. Nuclear Cataract and Myopic Shift in Refraction. Am. J. Ophthalmol. 2007;144(3):457–459. 18. Linton KL, Klein BE, Klein R. The Validity of Self-Reported and Surrogate- Reported Cataract and Age-Related Macular Degeneration in the Beaver Dam Eye Study. Am. J. Epidemiol. 1991;134(12):1438–1446. 19. Klein R, Klein BE, Linton KL, et al. The Beaver Dam Eye Study: Visual Acuity. Ophthalmology. 1991;98(8):1310–1315. 20. Klein R, Klein BE, Lee KE. Changes in Visual Acuity in a Population. The Beaver Dam Eye Study. Ophthalmology. 1996;103(8):1169–1178. 21. Klein R, Klein BE, Lee KE, et al. Changes in Visual Acuity in a Population Over a 10-year Period : The Beaver Dam Eye Study. Ophthalmology. 2001;108(10):1757– 1766. 22. Klein BE, Klein R, Linton KL, et al. Assessment of Cataracts from Photographs in the Beaver Dam Eye Study. Ophthalmology. 1990;97(11):1428–1433. 23. Klein BE, Klein R, Linton K. Prevalence of Age-Related Lens Opacities in a Population. The Beaver Dam Eye Study. Ophthalmology. 1992;99(4):546–552. 24. Chen F, Duggal P, Klein BEK, et al. Variation in PTCHD2, CRISP3, NAP1L4, FSCB, and AP3B2 Associated with Spherical Equivalent. Mol. Vis. 2016;22:783– 796. 25. Stata Corporation. Stata Statistical Software: Release 13. 2013; 26. Gardiner JC, Luo Z, Roman LA. Fixed Effects , Random Effects and GEE: What are the Differences? Stat. Med. 2008;28:221–239. 27. Liang K-Y, Zeger SL. Longitudinal Data Analysis Using Generalized Linear Models. Biometrika Trust. 1986;73(1):13–22. 28. Kurland BF, Johnson LL, Egleston BL, et al. Longitudinal Data with Follow-up Truncated by Death : Match the Analysis Method to Research Aims. Stat. Sci. 2009;24(2):211–222. 29. Cole SR, Hernan MA. Constructing Inverse Probability Weights For Marginal Structural Models. Am. J. Epidemiol. 2008;168(6):656–664. 30. Robins JM, Rotnitzky A, Zhao LP. Analysis of Semiparametric Regression Models for Repeated Outcomes in the Presence of Missing Data. J. Am. Stat.

Assoc. 1995;90(429):106–121. 31. Troxel AB, Lipsitz SR, Brennan TA. Weighted Estimating Equations with Nonignorably Missing Response Data. Biometrics. 1997;53(3):857–869. 32. Wang JJ, Mitchell P, Simpson JM, et al. Visual Impairment, Age-Related Cataract, and Mortality. Arch. Ophthalmol. 2001;119:1186–1190. 33. Fong CSU, Mitchell P, Rochtchina E, et al. Correction of Visual Impairment by Cataract Surgery and Improved Survival in Older Persons: The Blue Mountains Eye Study Cohort. Ophthalmology. 2013;120(9):1720–1727. 34. Song E, Sun H, Xu Y, et al. Age-Related Cataract, Cataract Surgery and Subsequent Mortality: A Systematic Review and Meta-Analysis. PLoS One. 2014;9(11):e112054. 35. Cugati S, Cumming R, Smith W, et al. Visual Impairment, Age-Related Macular Degeneration, Cataract, and Long-term Mortality. Arch. Ophthalmol. 2007;125(7):917–924. 36. Knudtson M, Klein BEK, Klein R. Age-Related Eye Disease, Visual Impairment, and Survival. Arch. Ophthalmol. 2006;124:243–249. 37. Klein BEK, Klein R, Lee KE, et al. Incidence of Age-Related Cataract over a 15- Year Interval The Beaver Dam Eye Study. Ophthalmology. 2008;115(3):477–482. 38. Leske MC, Chylack MPHLT, Wu S. Incidence and Progression of Nuclear Opacities in the Longitudinal Study of Cataract. Ophthalmology. 1996;103(5):705–712. 39. Leske MC, Wu S, Nemesure B, et al. Incidence and Progression of Lens Opacities in the Barbados Eye Studies. Ophthalmology. 2000;107(7):1267–1273. 40. Wong TY, Foster PJ, Ng TP, et al. Variations in Ocular Biometry in an Adult Chinese Population in Singapore: The Tanjong Pagar Survey. Invest Ophthalmol Vis Sci. 2001;42(1):73–80. 41. Wong TY, Foster PJ, Johnson GJ, et al. Refractive Errors, Axial Ocular Dimensions, and Age-Related Cataracts: The Tanjong Pagar Survey. Invest Ophthalmol Vis Sci. 2003;44(4):1479–1485. 42. Warrier S, Wu HM, Newland HS, et al. Ocular Biometry and Determinants of Refractive Error in Rural Myanmar: The Meiktila Eye Study. Br J Ophthalmol. 2008;92:1591–1594. 43. Shufelt C, Fraser-bell S, Ying-lai M, et al. Opalescence in an Adult Population : The Los Angeles Latino Eye Study. Invest Ophthalmol Vis Sci. 2005;46(12):4450– 4460. 44. Hemenger RP, Garner LF, Ooi CS. Change With Age of the Refractive Index

Gradient of the Human Ocular Lens. Invest Ophthalmol Vis Sci. 1995;36(3):703– 707. 45. Ooi C, Grosvenor T. Mechanisms of Emmetropization in the Aging Eye. Optom. Vis. Sci. 1995;72(2):60–66. 46. Brown N, Koretz JF, Bron AJ. The Development and Maintenance of Emmetropia. Eye. 1999;13:83–92. 47. Brown NAP, Hill AR. Cataract : The Relation Between Myopia and Cataract Morphology. Br. J. Ophthalmol. 1987;71:405–414. 48. Glasser A, Campbell MCW. Biometric, Optical and Physical Changes in the Isolated Human Crystalline Lens with Age in Relation to Presbyopia. Vision Res. 1999;39:1991–2015. 49. Brown N. The Morphology of Cataract and Visual Performance. Eye. 1993;7:63– 67. 50. Olsen T, Arnarsson A, Sasaki H, et al. On the Ocular Refractive Components: The Reykjavik Eye Study. Acta Ophthalmol. Scand. 2007;85(4):361–366. 51. Richter GM, Wang M, Jiang X, et al. Ocular Determinants of Refractive Error and Its Age- and Sex-Related Variations in the Chinese American Eye Study. JAMA Ophthalmol. 2017;135(7):724–732. 52. Kurokawa T, Katai N. Refractive Assocaitions with Cataract: the Blue Mountains Eye Study. Invest. Ophthalmol. Vis. Sci. 1999;40(12):3021–3026. 53. Katz J, Tielsch JM, Sommer A. Prevalence and Risk Factors for Refractive Errors in an Adult Inner City Population. Invest. Ophthalmol. Vis. Sci. 1997;38(2):334– 340. 54. Mirshahi A, Ponto K a., Hoehn R, et al. Myopia and Level of Education. Results from the Gutenberg Health Study. Ophthalmology. 2014;121(10):2047–2052. 55. Aine E. Refractive Errors in a Finnish Rural Population. Acta Ophthalmol. 1984;62(6):944–954. 56. Sperduto RD, Seigel D, Roberts J, et al. Prevalence of myopia in the United States. Arch. Ophthalmol. 1983;101(3):405–407. 57. Lavery JR, Gibson JM, Shaw DE, et al. Refraction and Refractive Errors in an Elderly Population. Ophthalmic Physiol Opt. 1988;8(4):394–396. 58. Wang Q, Klein BEK, Klein R, et al. Refractive Status in the Beaver Dam Eye Study. Investig. Ophthalmol. Vis. Sci. 1994;35(13):4344–4347. 59. Attebo K, Ivers RQ, Mitchell P. Refractive Errors in an Older Population: The Blue Mountains Eye Study. Ophthalmology. 1999;106(6):1066–1072.

60. Richter GM, Choudhury F, Torres M, et al. Risk Factors for Incident Cortical, Nuclear, Posterior Subcapsular, and Mixed Lens Opacities: The Los Angeles Latino Eye Study. Ophthalmology. 2012;119(10):2040–2047. 61. McCarty CA, Mukesh BN, Fu CL, et al. The Epidemiology of Cataract in Australia. Am. J. Ophthalmol. 1999;128(4):446–465. 62. McCarty CA, Nanjan MB, Taylor HR. Attributable Risk Estimates for Cataract to Prioritize Medical and Public Health Action. Investig. Ophthalmol. Vis. Sci. 2000;41(12):3720–3725. 63. Saw SM, Gazzard G, Shin-Yen EC, et al. Myopia and Associated Pathological Complications. Ophthalmic Physiol. Opt. 2005;25:381–391. 64. Wojciechowski R. Nature and Nurture: The Complex Genetics of Myopia and Refractive Error. Clin. Genet. 2011;79(4):301–320.

Tables and Figures Table 1. Characteristics of Participants of the Beaver Dam Eye Study at Baseline, 1988-1990 Characteristic Included BDES Participants Total BDES Participants (N = 4,441)a (N = 4,972)b Age (years), mean (SD) 61.2 (10.8) 62.0 (11.2) Female Gender, N(%) 4,441 4,972 Female 2,480 (55.8) 2,784 (56.0) Male 1,961 (44.2) 2,188 (44.0) Education (years), mean (SD) 12.0 (2.8) 12.0 (2.9) Type II Diabetes, N(%) 4,441 4,901 No 4,077 (91.8) 4,460 (91.0) Yes 364 (8.2) 441 (9.0) Nuclear Sclerosis Grade, N(%) 4,441 4,540 1 or 2 2,513 (56.6) 2,577 (56.8) 3 1,362 (30.7) 1,369 (30.2) 4 or 5 566 (12.7) 594 (13.1) Spherical Equivalent (D), mean (SD) 0.25 (2.3) 0.25 (2.3) Abbreviations: D, diopters; SD, standard deviation. aAll individuals remaining in the study sample after exclusions. bThe total number of individuals varies by trait due to missingness.

Table 2. Estimated Associations of Various Factors With Spherical Equivalent and Amount of Change in Spherical Equivalent Among Participants of the Beaver Dam Eye Study, Beaver Dam, Wisconsin, 1988-2010 Factor Na Multivariable Modelb Effect Sizec 95% CI

Baseline Age (years) 7,748 -0.018 -0.022, -0.014 Change in Age from Baseline (years) Mild Nuclear Sclerosis 2,476 0.046 0.041, 0.051

Moderate Nuclear Sclerosis 3,845 0.050 0.043, 0.057

Severe Nuclear Sclerosis 1,427 -0.049 -0.083, -0.015

Mild Nuclear Sclerosis 2,476 Moderate vs. Mild Nuclear Sclerosis 3,845 -0.054 -0.104, -0.004 Severe vs. Mild Nuclear Sclerosis 1,427 -1.358 -1.152, -1.197 Sex Female 4,266 Male 3,482 -0.149 -0.211, -0.087

Education (years) 7,748 0.005 -0.006, 0.017 Diabetes Status No 6,848 Yes 900 0.237 0.122, 0.352

Baseline Spherical Equivalent (D) 7,748 1.014 1.000, 1.028 Abbreviations: CI, confidence interval; D, diopters. aPerson-visits remaining after exclusions, by category. bModel contains all factors listed in the table. Change in age, nuclear sclerosis, and diabetes status are lagged by one visit. cEstimates represent the mean change in refraction with each one-unit change in each factor.

Table 3. Mean Changes in Spherical Equivalent by Age and Nuclear Sclerosis Grade Across Visits Among Participants of the Beaver Dam Eye Study, Beaver Dam, Wisconsin, 1988-2010 Na Mean One- 95% CI Mean 5-Year 95% CI Mean 20-Year 95% CI Year Change Changec (D) Changec (D) (D)c

Multivariable Modelb Nuclear Sclerosis Grade 1 or 2 2,476 0.046 0.041, 0.051 0.231 0.206, 0.257 0.926 0.823, 1.028

3 3,845 0.050 0.043, 0.057 0.251 0.215, 0.287 1.006 0.862, 1.149

4 or 5 1,427 -0.049 -0.083, -0.015 -0.245 -0.417, -0.074 -0.981 -1.667, -0.295

Abbreviations: CI, confidence interval; D, diopters. aPerson-visits included in the model, by nuclear sclerosis category. bModel includes baseline age (centered), change in age from baseline, nuclear sclerosis, an interaction between change in age and nuclear sclerosis, sex, education level (centered), diabetes status, and baseline spherical equivalent (centered). Change in age, nuclear sclerosis, and diabetes status are lagged by one visit. cThe mean one-year estimate represents the mean change in refraction with each one year increase in age from baseline, which varies by nuclear sclerosis severity, adjusted for all other factors in the model. The adjusted mean five and 20-year changes in refraction were estimated by multiplying the estimates for change in age from baseline by five or 20 years, separately by each category of nuclear sclerosis severity.

A B

2.0 2.0

1.5 1.5

1.0 1.0

0.5 0.5

0 0

Spherical Equivalent (D) -0.5 Spherical Equivalent (D) -0.5

-1.0 -1.0 40 50 60 70 80 Age (Years) 40 50 60 70 80 Age (Years) 1908-1912 1923-1927 1938-1942 1913-1917 1928-1932 1943-1947 1918-1922 1928-1932 1938-1942 1918-1922 1933-1937 1923-1927 1933-1937 1943-1947

C D

2.0 2.0

1.5 1.5

1.0 1.0

0.5 0.5

0 0

Spherical Equivalent (D)

-0.5 Spherical Equivalent (D) -0.5 -1.0 50 60 70 80 90 -1.0 Age (Years) 65 70 75 80 85 Age (Years) 1908-1912 1923-1927 1938-1942 1913-1917 1928-1932 1943-1947 1908-1912 1918-1922 1928-1932 1918-1922 1933-1937 1913-1917 1923-1927

Figure 1. Lowess curves of spherical equivalent (D) by age and year of birth for participants of the Beaver Dam Eye Study (Beaver Dam, Wisconsin, 1988-2010) among: (A) all participants, (B) among those with mild (grades 1 or 2) nuclear sclerosis, (C) among those with moderate (grade 3) nuclear sclerosis, or (D) among those with severe (grades 4 or 5) nuclear sclerosis.

CHAPTER THREE

Imputation of Exome Array Variants to the Haplotype Reference Consortium Panel

to Detect Additional Loci Associated with Ocular Phenotypes in the Beaver Dam

Eye Study

Abstract

We performed a whole genome imputation from exome array data to determine if genome-wide imputation of rare and low frequency variants improved the coverage of the exome array and enabled detection of novel variants associated with three ocular phenotypes: refraction, axial length, and corneal curvature. We imputed data from 1,871

European-Americans from the Beaver Dam Eye Study who were genotyped on the

Illumina exome array to the Haplotype Reference Consortium reference panel.

Imputation increased the number of variants tenfold, from 95,376 to 1,024,985 high quality variants. Imputation provided denser coverage in regions targeted by the exome array, as expected. We improved the genome-wide coverage to 1.8 variants per 5 kilobases, but imputation was not uniformly distributed across the genome. No novel variants or genes associated with our phenotypes were identified. Associations were identified for previously known loci with corneal curvature (13q12.12 and 4q12) and spherical equivalent (15q14 and 2q13). However, all associated loci were identified on the original exome array. Overall, imputation of the exome array did improve our coverage of targeted protein-coding regions, but it did not allow us to detect novel variants with refraction, axial length, or corneal curvature.

Introduction

Genotype imputation provides an efficient statistical method for inferring genotypes at variant sites not available on existing arrays, most frequently genome-wide arrays containing common variants (minor allele frequency (MAF) > 0.05) (1–3). Unlike genome-wide arrays, the exome array, a cost-effective intermediate between genome- wide arrays and exome sequencing, consists primarily of rare (MAF < 0.01) and low frequency (0.01 ≤ MAF ≤ 0.05) variants in selectively targeted exons. Imputation from exome arrays has historically been less successful (4–7). However, several studies have reported that larger, population-specific reference panels enhance resulting imputation quality, especially for rare and low frequency variants (4,5,8–10).

The Haplotype Reference Consortium (HRC) reference panel consists of 64,976 haplotypes from 32,611 individuals of predominantly European ancestry (11). Imputation from an exome array to the large and dense HRC reference panel may improve our ability to identify novel loci in association analyses that would not be detectable with the exome array alone.

The Beaver Dam Eye Study (BDES) is a population-based ocular cohort study of primarily European-American adults from Beaver Dam, Wisconsin. A subset of the

BDES participants was genotyped on the Illumina exome array. We sought to determine whether imputation of the exome array genotypes to the HRC reference panel would expand our coverage of the regions targeted by the exome array so we could detect genetic loci not detectable with the non-imputed exome array in the association analysis of three ocular phenotypes: refraction, axial length, and corneal curvature. We evaluated

52 imputation quality and coverage, and compared association results obtained from the imputed data with results obtained from the non-imputed data.

Refractive errors, myopia (nearsightedness) and hyperopia (farsightedness), are the most common visual disorders in the world, resulting in blurred vision (12–15). Older individuals are disproportionately affected (16). Only ~3.4% of the genetic variation in refraction has been accounted for to date, leaving a significant proportion of missing heritability yet to be explained (12,17). Axial length and corneal curvature are primary biological determinants of refraction, and all three traits are highly heritable with shared etiologies (15,18–33). Identifying variants associated with axial length and corneal curvature in addition to refraction could account for a proportion of the missing heritability and enhance our understanding of the genetic architecture of refraction.

Methods

The Beaver Dam Eye Study

The BDES has been described previously (34–39). Briefly, the baseline examination of the BDES included a personal history questionnaire and a complete ocular examination. The eye exam included a standardized evaluation of refraction using the Humphrey 530 refractor (Humphrey, Humphrey Instruments Inc., San Leandro, CA)

(34). A standard formula was used to calculate the mean spherical equivalent, a measure of refraction, in diopters (D) for each eye (sphere + 0.5 x cylinder). Nuclear lens opacity, or nuclear sclerosis, was measured by grading slitlamp lens photographs using a standardized protocol. Grades were assigned by trained photograders using a five-point scale of severity based on degree of nuclear lens opacity, with grades of four or five

53 indicating nuclear cataract (40,41). Eyes with no lens, with an intraocular lens, or with best corrected visual acuity of 20/200 or worse were excluded from the further study (for refraction only) because of reduced reliability and increased variability of refraction.

During the fourth visit, ocular biometry measurements, including axial length and corneal curvature, were obtained using partial coherence laser interferometry

(IOLMaster, Carl Zeiss Meditec, Jena, Germany). Age, sex, height and years of education were obtained at all visits. Data were collected according to the tenets of the Declaration of Helsinki, and all protocols were prospectively approved by the institutional review board at the University of Wisconsin at Madison. All study participants provided informed consent.

Trait Definitions

Axial length, corneal curvature, and spherical equivalent (a continuous measure of refraction) were each analyzed independently. Analyses of axial length and corneal curvature were conducted using measurements of the right eye at visit 4, followed by analyses using measures of the left eye at visit 4 to ensure consistency of results (19).

Spherical equivalent was measured as the average spherical equivalent between the right and left eyes at baseline (42). Nuclear sclerosis was measured as the sum of the nuclear lens opacity score of both eyes at baseline. Age, sex, and education were associated with axial length in this and previous studies (19,22,43–46), while age, sex, and height influenced corneal curvature (19,22,31,32,45,46). Our decision to exclude height as a covariate for axial length was made to avoid over-adjustment due to collinearity between axial length and height in our study. Age, sex, education, and nuclear sclerosis were associated with spherical equivalent in our study at baseline (19,42,47). Baseline height

54 was used for corneal curvature instead of visit 4 height measures to avoid capturing age- related shrinkage. Otherwise, visit 4 values were used for axial length and corneal curvature, while baseline covariate values were used for spherical equivalent.

Residual values of all three phenotypes were obtained using linear regression of each phenotype on their respective covariates and used as quantitative outcomes in our association analyses. The distributions of all three phenotypes were approximately normally distributed both before and after adjustment for covariates following standard quality control procedures.

Quality Control for Samples for Axial Length, Corneal Curvature, and Refraction

One individual from each first- or second-degree relative pair was removed following imputation on 1,871 individuals (n = 117), as were all individuals with missing values of the relevant phenotype and covariates. Individuals with differences in axial length or corneal curvature > 3 standard deviations from the mean between the two eyes were excluded from each respective dataset. Similarly, individuals with differences in spherical equivalent > ±4 D between the left and right eyes were removed from the third dataset. Final sample sizes for axial length, corneal curvature, and spherical equivalent were 874, 883, and 1,552 individuals, respectively.

Quality Control for Genotyping and Imputation

A nested subset of 2,032 individuals from the BDES was genotyped on the

Illumina Infinium HumanExome-12 v1.1 BeadChip (Illumina, Illumina, Inc., San Diego,

CA) exome array. Individuals were originally selected based on the tail ends of the distribution for the quantitative traits of intraocular pressure and spherical equivalent. The distribution of both traits was normal following selection. Genotyping was conducted at

55 the Genetic Resource Core Facility, Johns Hopkins Institute of Genetic Medicine.

Genotype calling was performed using Illumina’s GenTrain clustering algorithm in

GenomeStudio.

Among the 1,908 individuals who were successfully genotyped (sample call rate

> 98%), some individuals were removed for quality control: unresolved sex inconsistencies (n = 15), Mendelian errors (n = 2), and unexpected duplicates (n = 8).

Principal component analysis using SMARTPCA in EIGENSTRAT version 4.2

(Supplementary Figure S1) (48) identified 12 individuals of non-European ancestry who were excluded from further analysis.

Non-autosomal variants (n = 5,465), variants with call rates < 98% (n = 6,114), variants out of Hardy-Weinberg Equilibrium (P < 1.0 × 10-6) according to the Hardy-

Weinberg Equilibrium exact test (49) (n = 68), monomorphic variants (n = 135,499), indels (n = 61), and duplicate variants (n = 318) were excluded. The remaining 1,871 individuals and 95,376 variants were imputed to the HRC reference panel (n = 32,611) using SHAPEIT version 2 and Minimac version 3 on the publicly available Michigan

Imputation Server (1–3,11).

To discriminate well and poorly imputed variants across MAF bins, and to account for the low marker density and high proportion of rare variants on the exome array which could not provide a strong backbone of common variants for imputation, we used an imputation R2 threshold of 0.6 for all downstream analyses (6,11). Imputation R2 is the imputation metric provided by Minimac and defined as the squared correlation between imputed genotypes and true, unobserved genotypes. The imputed dataset consisted of a total of 39,015,721 unique variants (including those genotyped on the

56 exome array), of which 1,024,985 (2.63%) remained after applying the imputation R2 filter of 0.6 to imputed variants.

Coverage

Coverage was defined as the number of variants in the imputed dataset per 1 and

5 kilobases (kb). Coverage was estimated by chromosome, across the genome, and across the exome using base-pair estimates from the UCSC Known Gene human annotation

(hg19; UCSC Genome Browser, University of California Santa Cruz, Santa Cruz,

California). Coverage was further characterized visually by chromosome using Integrated

Genomics Viewer version 2.3 (50,51).

Genotype Concordance

Genotype concordance was assessed in a subset of variants genotyped on the exome array and included in the HRC reference panel, defined as the agreement between the observed and most likely imputed genotypes. For rare (MAF < 0.01) and low frequency (0.01 ≤ MAF ≤ 0.05) variants, concordant homozygous major genotypes were excluded from the calculation to help correct for sensitivity of these concordance values to MAF for less common variants (7,9). Concordance values were obtained after converting imputed dosages to hard call genotypes by applying a threshold of 0.9 to the maximum value in each genotype probability triple. Genotype concordance was used to evaluate the quality of the imputed genotypes. Concordance values for genotypes exceeding the imputation R2 threshold of 0.6 are presented. Genotype concordance for common and rare or low frequency variants was calculated in PLINK and VCFtools, respectively (52,53).

Single Variant Analysis

We conducted a single variant analysis under an additive model using linear regression in Mach2qtl (54,55). Imputed dosages were used to account for imputation uncertainty. DosageConverter version 1 (DosageConverter, University of Michigan, Ann

Arbor, MI) was used to convert the Variant Call Files from the Michigan Imputation

Server to dose files.

Single variant analyses were performed on rare (0.0057 ≤ MAF < 0.01), low- frequency (0.01 ≤ MAF ≤ 0.05) and common (MAF > 0.05) variants for axial length and corneal curvature. Rare variants were defined as 0.0032 ≤ MAF < 0.01 for spherical equivalent. Variants with a MAF below the 0.0057 (axial length and corneal curvature) or

0.0032 (spherical equivalent) thresholds (minor allele count < 10) were not considered

(42,56,57). A Bonferroni-corrected significance threshold of P < 1.09 × 10-6 for axial length and corneal curvature, and P < 9.25 × 10-7 for spherical equivalent, was used based on the number of independent markers in each dataset (i.e. those with linkage disequilibrium r2 < 0.20). Variants with MAF > 0.01 and effect sizes ≥ 1.5 for axial length, ≥ 0.4 for corneal curvature, or ≥ 3.0 for spherical equivalent could be detected with sufficient power (80%) in these single variant analyses (58). Significantly associated variants and suggestively associated variants (P < 1.0 × 10-4) are reported.

Meta-Analysis of BDES Groups

A second, non-overlapping group of individuals from the BDES was selected for genotyping separately on the Illumina exome array and imputed to the HRC reference panel. A detailed overview of the selection, genotyping, and imputation of this second

BDES group can be found in the appendix (59,60), along with details regarding the meta- analysis of the two BDES groups. A meta-analysis of the single variant results from the

58 first and second BDES groups was performed to increase our power to detect significant associations. Meta-analysis was conducted instead of a joint analysis because the exome arrays used in each group were different.

Gene-Based Analysis

We conducted an optimized sequence kernel association test (SKAT-O) using the default weight √푤푗 = 훽(푀퐴퐹푗; 푎1 = 1, 푎2 = 25) to up-weight rare variants presumed to be causal (61,62). The a1 and a2 parameters represent the Beta distribution shape parameters α and β, respectively. The gene-based SKAT-O test was conducted in the

Efficient and Parallelizable Association Container Toolbox (EPACTS, University of

Michigan, Ann Arbor, MI) program, using genotype dosages to account for imputation uncertainty. Gene groups were defined using variant annotations in ANNOVAR under

GRCh37/hg19 (63). Genes with more than one rare or low frequency variant (MAF ≤

0.05) were included in the analysis. Genes with a cumulative minor allele frequency

(CMAF, the sum of MAF values of all variants within the gene) > 0.01 are reported as reliable findings. A total of 13,854 autosomal genes were tested for their association with axial length and corneal curvature, and 13,868 genes were tested for their association with spherical equivalent. This translated to a Bonferroni-corrected significance threshold of P < 3.61 × 10-6. Significantly associated genes and suggestively associated genes (P <

1.0 × 10-4) are reported.

Association of Known Genome-Wide Association Study (GWAS) Loci with Axial

Length, Corneal Curvature, and Refraction

We identified previously published GWAS loci for (1) axial length, (2) corneal curvature, (3) spherical equivalent, or (4) myopia. We assessed whether (1) the exact

59 variant, or (2) variants in strong linkage disequilibrium (r2 ≥ 0.7) with these previously published GWAS loci were significantly associated with one or more of the three phenotypes (axial length, corneal curvature, spherical equivalent) analyzed in this study at P < 0.01. Ensembl release 85 (Ensembl, Wellcome Trust Genome Campus, Hinxton,

UK) was used to identify variants in LD with previously published GWAS variants.

Annotations

SeattleSeq Annotation Server 138 version 9.01 was used to annotate reported genetic variants from the single variant analyses under GRCh37/hg19 (SeattleSeq

Annotation Server 138, University of Washington, Seattle, WA). Chromosomal locations and amino acid alterations of variants were obtained using dbSNP build 147 (dbSNP

Short Genetic Variations, National Center for Biotechnology Information, Bethesda,

MD).

Results

We retained1,024,985 of 39,015,721 variants including all of the original exome array variants from the filtered imputed dataset (2.63%; Table 1). This is more than ten times the number of variants prior to imputation (n = 95,376). While the exome array contained 63.4% (n = 60,482) rare variants and 27.1% (n = 25,856) common variants, the final imputed dataset contained 15.6% (n = 160,209) rare variants and 72.5% (n =

743,282) common variants (Table 1).

Coverage

The estimated mean coverage of the genome was 1.78 variants per 5 kb (Table 2).

The genomic coverage values for each chromosome varied by the number of genes

60 present on each chromosome. Coverage of the exome in particular was far denser at 65.8 variants per 5 kb (Table 2). The imputed variants were primarily in coding regions surrounding the original exome array variants, enhancing the resolution of these regions, but did not extend into non-coding regions not targeted by the exome array (Figure 1).

Areas completely omitted by the exome array, like the short arm of chromosome 22, could not be imputed (Figure 1).

Genotype Concordance

In general, genotype concordance for the 76,177 variants directly assayed and imputed increased with increasing MAF. Concordance values for rare (n = 42,386), low frequency (n = 8,166), and common (n = 25,625) variants were 0.97, 0.97, and 0.98, respectively, after applying an imputation R2 threshold of 0.6.

Study Participants in First and Second BDES Groups

With the exception of sex and height, there appeared to be significant (albeit small) differences in age, education, and axial length between the two genotyped groups for axial length at P < 0.05 (Supplementary Table S1). The genotyped groups for corneal curvature showed no significant differences, except in age (Supplementary Table S2).

The genotyped groups for spherical equivalent showed significant differences with respect to age, education, sex composition, nuclear sclerosis, and spherical equivalent

(Supplementary Table S3). Individuals in the first BDES group were more myopic compared to those in the second BDES group, as expected from the sampling scheme.

Single Variant and Gene-Based Analyses

No significant associations of single variants or genes (CMAF > 0.01) with axial length, corneal curvature, or spherical equivalent were identified. Suggestively associated

61 variants are reported in Supplementary Tables S4-S7. Results of the right and left eyes were similar for both axial length and corneal curvature, so only the results of the right eye are reported for suggestively associated variants and genes.

Meta-Analysis of BDES Groups

No significant associations were detected in the meta-analysis. Suggestively associated variants are reported in Supplementary Table S8. Results were similar for both eyes, so results for the right eye are reported for axial length and corneal curvature.

Association with Known GWAS Genetic Loci

We confirmed several previously published genotyped and imputed variants associated with corneal curvature and spherical equivalent in this study at P < 0.01 (Table

3). However, all imputed variants previously identified in GWAS and associated with our phenotypes of interest were in LD with variants from the same region genotyped on the exome array and associated with one of the phenotypes in this study (Table 4).

We identified three variants in LD significantly associated with corneal curvature at P < 0.01 (Table 3). These variants had previously been associated with high myopia in the mitochondrial intermediate peptidase (MIPEP) gene (MIM 602241) on chromosome

13q12.12 (64). The minor allele A of the first variant (rs9318086, g.24432467G>A), which was genotyped on the exome array, was associated with 0.04 mm decrease in corneal curvature (MAF = 0.44, 95% CI = [-0.06 mm, -0.01 mm], p = 4.8 × 10-3) despite leading to an increased odds of high myopia in the published study (64). The two remaining variants, rs9510902 (g.24430312G>A) and rs1886970 (g.24440498T>C), were imputed and in high LD with rs9318086 (r2 = 0.71 and 0.75, respectively). The minor allele A at rs9510902 led to an increase in corneal curvature in this study (MAF = 0.42, β

= 0.05 mm, 95% CI = [0.02 mm, 0.08 mm], p = 2.4 × 10-3). The minor allele C of rs1886970 led to a similar increase in corneal curvature (MAF = 0.42, β = 0.04 mm, 95%

CI = [0.01 mm, 0.07 mm], p = 2.9 × 10-3). Both variants increased the odds of high myopia in the published study (64), but were not associated with spherical equivalent in this study. Although different phenotypes were being analyzed in each study, the results of the latter two variants were consistent since an increase in corneal curvature is associated with myopia.

A variant in an intergenic region on chromosome 4q12 near the platelet derived growth factor receptor alpha (PDGFRA) gene (MIM 473190) genotyped on the exome array was significantly associated with corneal curvature in a previous GWAS and replicated in this study at P < 0.01 (Table 3) (65). In the BDES, the minor allele C at rs2114039 (g.55092626T>C) was associated with a 0.03 mm decrease in corneal curvature (MAF = 0.28, 95% CI = [-0.06 mm, -0.01 mm], p = 0.01). Two imputed variants in high LD with each other (r2 = 1.0) and in moderate LD with rs2114039 (r2 =

0.66) on chromosome 4q12 near the PDGFRA gene were previously associated with corneal curvature and successfully replicated in this study. The minor allele A at the first variant (rs17084051, g.55087581C>A; imputation R2 = 0.721) led to a 0.05 mm increase in corneal curvature in this study (MAF = 0.27, 95% CI = [0.01 mm, 0.08 mm], p =

0.01). This variant was also subsequently associated with corneal curvature in a combined meta-analysis of 9,383 Europeans and Asians (p = 4.50 × 10-14) (32). The second imputed variant that replicated a previous finding with corneal curvature showed a similar effect to rs17084051. The minor allele A of rs1800813 (g.55094467G>A; imputation R2 = 0.723) was associated with higher corneal curvature (MAF = 0.27, β =

0.05 mm, 95% CI = [0.01 mm, 0.08 mm], p = 0.01). The allele frequency values were quite different between studies because the effect allele in the published study was the G instead of A (Table 3) (31).

Two variants in high LD on chromosome 15q14 near the gap junction protein, delta-2 (GJD2) gene (MIM 607058) that were associated with spherical equivalent in previously published studies replicated for spherical equivalent at P < 0.01 in this study

(Table 3) (66,67). Of these two replicated variants, rs634990 (g.35006073T>C) had been genotyped on the exome array and rs685352 (g.35008335A>G) was imputed. The imputed variant was in high LD with rs634990 (r2 = 0.851). The minor allele C at rs634990 was associated with lower spherical equivalent in this study (MAF = 0.47, β = -

0.27 D, 95% CI = [-0.45, -0.09], p = 3.8 × 10-3). The minor allele G at rs685352

(imputation R2 = 0.912) was associated with a 0.29 D increase in spherical equivalent

(MAF = 0.47, 95% CI = [0.10, 0.49], p = 3.1× 10-3).

In addition, one variant previously associated with myopia in the protease, serine,

56 (PRSS56) gene (MIM 613858) on chromosome 2q13 was also associated with spherical equivalent in this study at P < 0.01 (Table 3) (68). This variant had been genotyped on the exome array. The minor allele G at rs1550094 (g.233385396A>G, p.Ala30Thr) was associated with lower spherical equivalent in this study (MAF = 0.31, β

= -0.32 D, 95% CI = [-0.52, -0.12], p = 2.4 × 10-3), which is consistent with the increased risk for myopia reported in the published study (68). We were unable to replicate any variants previously associated with axial length in this study (Supplementary Table S9).

Discussion

Imputation of the exome array to the HRC reference panel improved our coverage of regions targeted by the exome array with high quality imputed variants, but we were unable to detect novel loci associated with refraction, axial length, or corneal curvature.

The high concordance between exome array genotypes and imputed genotypes after applying the imputation R2 threshold of 0.6 suggests imputation of the exome array to the

HRC reference panel produced high quality imputed variants across the full allele frequency spectrum. This is consistent with a recent study comparing the quality of imputation of an exome array genotyped from a European sample to 1000 Genomes phase 3 and HRC reference panels and showed imputation quality was dramatically improved, especially for rare variants, when imputed to the HRC (69). More common variants were successfully imputed than rare variants, as expected.

Regions originally targeted by the exome array were denser with genotypes following imputation to the HRC reference panel. Regions outside of those targeted by the exome array were still sparsely covered with genotypes, as expected. Regions not targeted by the exome array could not be reliably imputed since imputation relies on establishing patterns of LD between variants. The exome array, which consisted primarily of rare and low frequency variants, did not provide as many variants that would be in high LD with other variants. The proportion of high quality imputed variants

(2.63%) was low as a result. A previous study evaluating the imputation quality of low density arrays when imputing to 1000 Genomes reported similarly low proportions of variants retained for low density arrays (6). The accurately imputed variants enhanced coverage in regions targeted by the exome array, but did not extend coverage to other regions of the genome.

We did not detect any previously published loci associated with corneal curvature in this study at P < 0.01 not detected using the original non-imputed exome array. The first three variants associated with corneal curvature in this study were previously associated with high myopia in a case-control GWAS conducted among Han Chinese participants (64). All three variants (one of which was genotyped on the exome array) lie within the MIPEP gene on chromosome 13q12.12. MIPEP is expressed in several tissues throughout the body, including retinal pigment epithelial cells in the human eye (64). The product of this gene plays a large role in the maturation of oxidative phosphorylation

(OXPHOS)-related proteins (70,71). It has been suggested by previous investigators that

MIPEP likely plays a role in eye development, and mutations in this and related genes can affect myopia onset (64). The association of these variants with both high myopia and corneal curvature supports the claim that these phenotypes share an overlapping genetic mechanism. These variants were originally associated with high myopia among

Han Chinese (64). There is much evidence to suggest the 13q12.12 region, and specifically the MIPEP gene, plays a large role in ocular mechanisms related to refraction and its biological determinants across populations. This could be clinically relevant for those at high risk for myopia or related ocular disorders.

Although the two imputed variants from 13q12.12 showed consistent effects with previously published results (both showed an increased odds of myopia in the published study and an increase in corneal curvature in our study), the genotyped variant

(rs9318086) showed a decrease in corneal curvature. The reason for this inconsistency is unclear, as the two other variants were in high LD with this genotyped variant but showed opposite effects. It is possible a decrease in corneal curvature associated with this

66 variant may be accompanied by other changes to the eye, such as axial elongation, which overcompensates for decreases in corneal curvature and results in myopia instead of hyperopia.

The second locus associated with corneal curvature in our study was located on chromosome 4q12 near the PDGFRA gene. The PDGFRA gene encodes the alpha- isoform of PDFGR, a cell surface tyrosine kinase receptor suggested to play a role in organ development, wound healing, and tumor progression (70–72). There is also evidence of its expression in the eye (73). While there is limited evidence to directly support the function of this gene in the cornea, the role of PDGFRA in growth and development could extend to the growth and maintenance of the cornea itself.

With the exception of rs2114039, the replicated corneal curvature variants showed opposing directions of effect in our study compared to published studies

(31,32,65). This could be attributed to the inclusion of Asians and children in the published GWAS, in addition to their lack of adjustment for height. Individuals of Asian descent, and children, typically have different distributions of height compared to adults of European descent, suggesting their distributions of corneal curvature may differ as well since height and corneal curvature are positively correlated. Results of a sensitivity analysis unadjusted for height were similar to the results adjusted for height in the current study, suggesting adjustment for height did not impact the effect sizes of these reported loci in the BDES. Differences in the distributions of corneal curvature between distinct ancestral groups, and between children and adults, likely contributed to the differences in the effect sizes observed.

We successfully replicated three variants previously associated with spherical equivalent in our study, but no additional loci were replicated as a result of imputation

(42,66,68). Two of the three replicated variants (rs634990 and rs1550094) were genotyped on the exome array and the third (rs685352) was imputed and in high LD (r2 =

0.851) with rs634990. The first and third variants (rs634990 and rs685352) are on chromosome 15q14, 39 kb upstream from the 3’ end of the GJD2 gene. The GJD2 protein is a gap junction protein that is neuron-specific and present in photoreceptors, amacrine cells, and bipolar cells in the retina (66,70,71). Mutations in the GJD2 gene have been repeatedly associated with refractive error and degenerative myopia, likely due to its critical role in intercellular transport between the cells of the eye (66,70). The consistency of the direction and magnitude of effect between previous GWAS and the current study for rs634990, but not the imputed variant in high LD (r2 = 0.851) with this variant, remains unclear. The imputed variant was initially detected in a cohort of individuals of European ancestry (n = 15,608), similar to the BDES (66). The same variant was also detected in a cohort of 26,615 Europeans in a more recent meta-analysis

(67). Although the minor allele, MAF, adjusted covariates, and directions of effect presented for this variant are consistent between these two published studies, the magnitudes of effect are less consistent (66,67).

The third replicated variant (rs1550094) on chromosome 2q13 was previously associated with a higher risk of myopia in a large cohort of 26,615 Europeans (68). This variant lies within the PRSS56 gene, which encodes a protein with trypsin-like serine protease activity (70). Like GJD2, PRSS56 similarly shows possible links to ocular morphology. The primary disease associated with this gene is Microphthalmia, a

Mendelian developmental disorder characterized by small malformed eyes and resulting hyperopia due to short axial length among other clinical features (71,72).

While we successfully replicated variants previously associated with spherical equivalent and corneal curvature, we were unable to replicate any previously associated variants with axial length in our study. There are several possible reasons for this, including differences in ancestry between our study and previously published studies, and overall sample size. One study combined 12,531 Europeans and 8,216 Asians in a meta- analysis (44) and another study conducted their meta-analysis among 4,944 Asians, 929 of whom were children (43). Both studies adjusted for height in their analyses, so we performed a sensitivity analysis adjusted for height in the BDES and this showed negligible differences. Similarly, we modeled our analysis with and without education, which had a negligible effect on the associations. Our sample size was much smaller compared to the sample sizes of both referenced studies (43,44). Of the variants previously associated with axial length in our dataset, the lowest MAF was 0.30. Using this MAF, the sample size, and mean axial length from our axial length sample, our power to detect significant associations (at P < 0.01) only exceeded 80% if effect sizes were greater than 0.20. None of the effect sizes reported in the current study reached

0.20; thus, we did not have sufficient power to detect associations with variants previously reported with axial length.

Imputation of the exome array did not improve our ability to detect additional loci associated with our phenotypes of interest. We were unable to detect novel variants significantly associated with axial length, corneal curvature, or spherical equivalent after imputation. Any imputed variants associated with our phenotypes that had been

69 previously identified in GWAS were located in the same regions in high LD with associated variants genotyped on the exome array. It is possible that our inability to detect additional associations was limited by our modest sample sizes. Imputation of the exome array may also be more successful with other phenotypes. Further studies are warranted to understand any potential advantages to imputing an exome array to a large reference panel like the HRC in terms of detection of novel genetic loci.

Acknowledgements

This work was supported by the National Eye Institute at the National Institutes of Health

(grants R01EY021531, U10006594, and 1T32EI022303); and by the Research to Prevent

Blindness Unrestricted Grant to the University of Wisconsin Department of

Ophthalmology and Visual Sciences.

We thank all staff and investigators of the Beaver Dam Eye Study.

Conflict of interest: none declared.

References

1. Delaneau O, Marchini J, Zagury J-F. A Linear Complexity Phasing Method for Thousands of Genomes. Nat. Methods. 2012;9(2):179–181. 2. Delaneau O, Zagury J, Marchini J. Improved Whole-Chromosome Phasing for Disease and Population Genetic Studies. Nat. Methods. 2013;10(1):5–6. 3. Das S, Forer L, Schönherr S, et al. Next-Generation Genotype Imputation Service and Methods. Nat. Genet. 2016;48(10):1284–1287. 4. Martin AR, Tse G, Bustamante CD, et al. Imputation-Based Assessment of Next Generation Rare Exome Variant Arrays. Pac Symp Biocomput. 2014;241–252. 5. Kim YJ, Lee J, Kim B, et al. A New Strategy for Enhancing Imputation Quality of Rare Variants from Next- Generation Sequencing Data Via Combining SNP and Exome Chip Data. BMC Genomics. 2015;16(1109):1–11. 6. Pistis G, Porcu E, Vrieze SI, et al. Rare Variant Genotype Imputation with Thousands of Study-Specific Whole-Genome Sequences: Implications for Cost- Effective Study Designs. Eur. J. Hum. Genet. 2015;23:975–983. 7. Nelson SC, Doheny KF, Pugh EW, et al. Imputation-Based Genomic Coverage Assessments of Current Human Genotyping Arrays. G3. 2013;3:1795–1807. 8. Li L, Li Y, Browning SR, et al. Performance of Genotype Imputation for Rare Variants Identified in Exons and Flanking Regions of Genes. PLoS One. 2011;6(9). 9. Tantoso E, Wong L, Li B, et al. Evaluating the Coverage and Potential of Imputing the Exome Microarray with Next-Generation Imputation Using the 1000 Genomes Project. PLoS One. 2014;9(9):1–9. 10. Zheng HF, Rong JJ, Liu M, et al. Performance of Genotype Imputation for Low Frequency and Rare Variants from the 1000 Genomes. PLoS One. 2015;10(1):1– 10. 11. McCarthy S, Das S, Kretzschmar W, et al. A Reference Panel of 64,976 Haplotypes for Genotype Imputation. Nat. Genet. 2016;1–5. 12. Hysi PG, Wojciechowski R, Rahi JS, et al. Genome-Wide Association Studies of Refractive Error and Myopia, Lessons Learned, and Implications for the Future. Investig. Opthalmology Vis. Sci. 2014;55(5):3344–3351. 13. National Eye Institute. Fact Sheet: Refractive Errors. 2010;(https://nei.nih.gov/sites/default/files/health- pdfs/HVM09_Fact_Sheet_Final_tagged.pdf). (Accessed November 9, 2015) 14. Centers for Disease Control and Prevention. Refractive Errors. Vis. Heal. Initiat. 2013;(http://www.cdc.gov/visionhealth/basic_information/eye_disorders.htm#a2).

(Accessed November 9, 2015) 15. Meng W, Butterworth J, Malecaze F, et al. Axial Length of Myopia: A Review of Current Research. Ophthalmologica. 2011;225(3):127–134. 16. Resnikoff S, Pascolini D, Mariotti S, et al. Global Magnitude of Visual Impairment Caused by Uncorrected Refractive Errors in 2004. Bull. World Health Organ. 2008;86(1):63–70. 17. Wojciechowski R. Nature and Nurture: The Complex Genetics of Myopia and Refractive Error. Clin. Genet. 2011;79(4):301–320. 18. Dirani M, Chamberlain M, Shekar SN, et al. Heritability of Refractive Error and Ocular Biometrics: The Genes in Myopia (GEM) Twin Study. Invest. Ophthalmol. Vis. Sci. 2006;47(11):4756–4761. 19. Klein AP, Suktitipat B, Duggal P, et al. Heritability Analysis of Spherical Equivalent, Axial Length, Corneal Curvature, and Anterior Chamber Depth in the Beaver Dam Eye Study. Arch. Ophthalmol. 2009;127(5):649–655. 20. Chen CYC, Scurrah KJ, Stankovich J, et al. Heritability and Shared Environment Estimates for Myopia and Associated Ocular Biometric Traits: The Genes in Myopia (GEM) Family Study. Hum. Genet. 2007;121:511–520. 21. Dirani M, Shekar SN, Baird PN. Evidence of Shared Genes in Refraction and Axial Length: The Genes in Myopia (GEM) Twin Study. Invest. Ophthalmol. Vis. Sci. 2008;49(10):4336–4339. 22. Chen JH, Chen H, Huang S, et al. Endophenotyping Reveals Differential Phenotype-Genotype Correlations Between Myopia-Associated Polymorphisms and Eye Biometric Parameters. Mol. Vis. 2012;18:765–778. 23. Hammond CJ, Snieder H, Gilbert CE, et al. Genes and Environment in Refractive Error: The Twin Eye Study. Investig. Ophthalmol. Vis. Sci. 2001;42(6):1232– 1236.

24. Biino G, Palmas MA, Corona C, et al. Ocular Refraction: Heritability and Genome-Wide Search for Eye Morphometry Traits in an Isolated Sardinian Population. Hum. Genet. 2005;116(3):152–159. 25. Lyhne N, Sjølie a K, Kyvik KO, et al. The Importance of Genes and Environment for Ocular Refraction and its Determiners: A Population Based Study Among 20- 45 Year Old Twins. Br. J. Ophthalmol. 2001;85(12):1470–1476. 26. Teikari J, O’Donnell JJ, Kaprio J, et al. Genetic and Environmental Effects on Oculometric Traits. Optom. Vis. Sci. 1989;66(9):594–599. 27. Dirani M, Islam A, Shekar SN, et al. Dominant Genetic Effects on Corneal Astigmatism: The Genes in Myopia (GEM) Twin Study. Invest Ophthalmol Vis Sci. 2008;49(4):1339–1344.

28. Valluri S, Minkovitz JB, Budak K, et al. Comparative Corneal Topography and Refractive Variables in Monozygotic and Dizygotic Twins. Am. J. Ophthalmol. 1999;127(2):158–163. 29. Guggenheim J a., Zhou X, Evans DM, et al. Coordinated Genetic Scaling of the Human Eye: Shared Determination of Axial Eye Length and Corneal Curvature. Investig. Ophthalmol. Vis. Sci. 2013;54(3):1715–1721. 30. Young TL, Metlapally R, Shay AE. Complex Trait Genetics of Refractive Error. Arch. Ophthalmol. 2007;125(1):38–48. 31. Chen P, Miyake M, Fan Q, et al. CMPK1 and RBP3 are Associated with Corneal Curvature in Asian Populations. Hum. Mol. Genet. 2014;23(22):6129–6136. 32. Guggenheim J a, McMahon G, Kemp JP, et al. A Genome-Wide Association Study for Corneal Curvature Identifies the Platelet-Derived Growth Factor Receptor alpha Gene as a Quantitative Trait Locus for Eye Size in White Europeans. Mol. Vis. 2013;19:243–253. 33. Paget S, Vitezica ZG, Malecaze F, et al. Heritability of Refractive Value and Ocular Biometrics. Exp. Eye Res. 2008;86(2):290–295. 34. Linton KLP, Klein BEK, Klein R. The Validity of Self-Reported and Surrogate- Reported Cataract and Age-Related Macular Degeneration in the Beaver Dam Eye Study. Am. J. Epidemiol. 1991;134(12):1438–1446. 35. Klein R, Klein BE, Lee KE, et al. Changes in Visual Acuity in a Population Over a 10-year Period : The Beaver Dam Eye Study. Ophthalmology. 2001;108(10):1757– 1766. 36. Klein R, Klein BEK, Lee KE, et al. Changes in Visual Acuity in a Population Over a 15-Year Period: The Beaver Dam Eye Study. Ophthalmology. 2006;142:539– 549. 37. Klein R, Klein BE, Linton KL, et al. The Beaver Dam Eye Study: Visual Acuity. Ophthalmology. 1991;98(8):1310–1315. 38. Klein R, Klein BE, Lee KE. Changes in Visual Acuity in a Population. The Beaver Dam Eye Study. Ophthalmology. 1996;103(8):1169–1178. 39. Klein R, Lee KE, Gangnon RE, et al. Incidence of Visual Impairment Over a 20- Year Period: The Beaver Dam Eye Study. Ophthalmology. 2013;120(6):1210– 1219. 40. Klein BE, Klein R, Linton KL, et al. Assessment of Cataracts from Photographs in the Beaver Dam Eye Study. Ophthalmology. 1990;97(11):1428–1433. 41. Klein BE, Klein R, Linton K. Prevalence of Age-Related Lens Opacities in a Population. The Beaver Dam Eye Study. Ophthalmology. 1992;99(4):546–552.

42. Chen F, Duggal P, Klein BEK, et al. Variation in PTCHD2, CRISP3, NAP1L4, FSCB, and AP3B2 Associated with Spherical Equivalent. Mol. Vis. 2016;22:783– 796. 43. Fan Q, Barathi V a, Cheng C-Y, et al. Genetic Variants on Chromosome 1q41 Influence Ocular Axial Length and High Myopia. PLoS Genet. 2012;8(6):1–14. 44. Cheng C-Y, Schache M, Ikram MK, et al. Nine Loci for Ocular Axial Length Identified through Genome-Wide Association Studies, Including Shared Loci with Refractive Error. Am. J. Hum. Genet. 2013;93(2):264–277. 45. Miyake M, Yamashiro K, Tabara Y, et al. Identification of Myopia-Associated WNT7B Polymorphisms Provides Insights into the Mechanism Underlying the Development of Myopia. Nat. Commun. 2015;6(6689):1–9. 46. Schache M, Richardson AJ, Mitchell P, et al. Genetic Association of Refractive Error and Axial Length with 15q14 but not 15q25 in the Blue Mountains Eye Study Cohort. Ophthalmology. 2013;120(2):292–297. 47. Klein AP, Duggal P, Lee KE, et al. Linkage Analysis of Quantitative Refraction and Refractive Errors in the Beaver Dam Eye Study. Invest. Ophthalmol. Vis. Sci. 2011;52(8):5220–5225. 48. Price A, Patterson NJ, Plenge RM, et al. Principal Components Analysis Corrects for Stratification in Genome-Wide Association Studies. Nat. Genet. 2006;38(8):904–909. 49. Wigginton JE, Cutler DJ, Abecasis GR. A Note on Exact Tests of Hardy-Weinberg Equilibrium. Am. J. Hum. Genet. 2005;76(5):887–893. 50. Robinsion JT, Thorvaldsdottir H, Winckler W, et al. Integrative Genomics Viewer. Nat. Biotechnol. 2011;29(1):24–26. 51. Thorvaldsdottir H, Robinson JT, Mesirov JP. Integrative Genomics Viewer (IGV): High-Performance Genomics Data Visualization and Exploration. Briefing. 2012;14(2):178–192. 52. Purcell S, Neale B, Todd-Brown K, et al. PLINK: A Tool Set for Whole-Genome Association and Population-Based Linkage Analyses. Am. J. Hum. Genet. 2007;81(3):559–575. 53. Danecek P, Auton A, Abecasis G, et al. The Variant Call Format and VCFtools. Bioinformatics. 2011;27(15):2156–2158. 54. Li Y, Willer CJ, Ding J, et al. MaCH : Using Sequence and Genotype Data to Estimate Haplotypes and Unobserved Genotypes. Genet. Epidemiol. 2010;34:816– 834. 55. Pei Y, Zhang L, Li J, et al. Analyses and Comparison of Imputation-Based Association Methods. PLoS One. 2010;5(5):1–10.

56. Chen F, Klein AP, Klein BEK, et al. Exome Array Analysis Identifies CAV1/CAV2 as a Susceptibility Locus for Intraocular Pressure. Invest. Ophthalmol. Vis. Sci. 2014;56(1):544–551. 57. Soderholm M, Almgren P, Jood K, et al. Exome Array Analysis of Ischaemic Stroke: Results from a Southern Swedish Study. Eur. J. Neurol. 2016;23:1722– 1728. 58. Gauderman W, Morrison J. QUANTO 1.1: A Computer Program for Power and Sample Size Calculations for Genetic-Epidemiology Studies. 2006; 59. Fritsche LG, Igl W, Cooke Bailey JN, et al. A Large Genome-Wide Association Study of Age-Related Macular Degeneration Highlights Contributions of Rare and Common Variants. Nat. Genet. 2016;48(2):134–143. 60. Cavalli-sforza LL. The Human Genome Diversity Project: Past , Present and Future. Nat. Genet. 2005;6:333–340. 61. Lee S, Wu MC, Lin X. Optimal Tests for Rare Variant Effects in Sequencing Association Studies SUP. Biostatistics. 2012;13(4):762–775. 62. Lee S, Miropolsky L, Wu M. SKAT: SNP-set (Sequence) Kernel Association Test: R Package Version 0.95. 2014; 63. Wang K, Li M, Hakonarson H. ANNOVAR: Functional Annotation of Genetic Variants from High-Throughput Sequencing Data. Nucleic Acids Res. 2010;38(16):1–7. 64. Shi Y, Qu J, Zhang D, et al. Genetic Variants at 13q12.12 are Associated with High Myopia in the Han Chinese Population. Am. J. Hum. Genet. 2011;88(6):805– 813. 65. Han S, Chen P, Fan Q, et al. Association of Variants in FRAP1 and PDGFRA with Corneal Curvature in Asian Populations from Singapore. Hum. Mol. Genet. 2011;20(18):3693–3698.

66. Solouki AM, Verhoeven VJM, van Duijn CM, et al. A Genome-Wide Association Study Identifies a Susceptibility Locus for Refractive Errors and Myopia at 15q14. Nat. Genet. 2010;42(10):897–901. 67. Verhoeven VJM, Hysi PG, Saw SM, et al. Large Scale International Replication and Meta-Analysis Study Confirms Association of the 15q14 Locus with Myopia. The CREAM Consortium. Hum. Genet. 2012;131(9):1467–1480. 68. Kiefer AK, Tung JY, Do CB, et al. Genome-Wide Analysis Points to Roles for Extracellular Matrix Remodeling, the Visual Cycle, and Neuronal Development in Myopia. PLoS Genet. 2013;9(2):1–8. 69. Iglesias AI, Lee SJ Van Der, Bonnemaijer PWM, et al. Haplotype Reference Consortium Panel : Practical implications of Imputations With Large Reference

Panels. Hum. Mutat. 2017; 70. GeneCards Human Genome Database. 2017;(http://www.genecards.org/). (Accessed July 5, 2017) 71. Online Mendelian Inheritance in Man. 2017;(http://www.omim.org/). (Accessed July 5, 2017) 72. Consortium U. UniProt. 2017;(www.uniprot.org). (Accessed July 5, 2017) 73. Qian J. Tissue-Specific Gene Expression and Regulation. 2008;(http://bioinfo.wilmer.jhu.edu/tiger/). (Accessed July 6, 2017)

Tables and Figures Table 1. Mean Imputation R2 by Minor Allele Frequency Bin for Imputed Exome Array All Markers Imputation R2 ≥ 0.6 MAF Bin Total Number of Mean Total Number of Markers Mean Markersa (SD) Imputation (% of Total)b (SD) Imputation R2 R2 All Markers 39,015,721 0.054 (0.15) 1,024,985 (2.63) 0.823 (0.13) MAF > 0.05 5,318,341 0.248 (0.27) 743,282 (13.98) 0.816 (0.13) 0.01 ≤ MAF ≤ 0.05 2,336,568 0.123 (0.20) 121,494 (5.20) 0.816 (0.13) MAF < 0.01 31,360,812 0.015 (0.08) 160,209 (0.51) 0.858 (0.14) Abbreviations: MAF, minor allele frequency; SD, standard deviation. aMarkers genotyped on the exome array were given imputation R2 = 1 and included with imputed variants. bPercentages are calculated using the total number of variants in the imputed dataset.

Table 2. Coverage of Genotyped and Imputed Variants by Chromosome Chromosome Number of Total Number of Total Number of Total Variants per Variants per Variants at Basepairs in Basepairs in Number of 5kb (Genome) 5kb (Exome) Imputation Genomeb Exomeb Genesc R2 ≥ 0.6a All 1,024,985 2,881,033,286 77,870,963 ~30,000 1.78 65.813 1 95,674 249,250,621 8,079,409 >3000 1.92 59.209 2 79,177 243,199,373 5,781,424 >2500 1.63 68.475 3 74,275 198,022,430 4,706,998 ~1900 1.88 78.898 4 57,450 191,154,276 3,364,332 ~1600 1.50 85.381 5 56,545 180,915,260 3,820,351 ~1700 1.56 74.005 6 103,650 171,115,067 4,241,245 ~1900 3.03 122.193 7 49,061 159,138,663 4,049,692 ~1800 1.54 60.574 8 42,153 146,364,022 2,909,471 >1400 1.44 72.441 9 36,111 141,213,431 3,430,407 >1400 1.28 52.634 10 46,998 135,534,747 3,398,919 >1400 1.73 69.137 11 68,360 135,006,516 4,439,924 ~2000 2.53 76.983 12 51,709 133,851,895 4,144,621 >1600 1.93 62.381 13 24,929 115,169,878 1,655,075 ~800 1.08 75.311 14 29,548 107,349,540 2,665,222 ~1200 1.38 55.433 15 31,910 102,531,392 2,897,969 ~1200 1.56 55.056 16 30,316 90,354,753 3,248,662 ~1300 1.68 46.659 17 37,031 81,195,210 4,348,983 >1600 2.28 42.574 18 20,642 78,077,248 1,377,184 >600 1.32 74.943 19 37,596 59,128,983 4,500,567 >1700 3.18 41.768 20 23,332 63,025,520 2,034,342 >900 1.85 57.345 21 10,960 48,129,895 888,164 >400 1.14 61.700

22 17,558 51,304,566 1,888,002 >800 1.71 46.499 Abbreviations: kb, kilobase. aVariants genotyped on the exome array were given imputation R2 = 1 for all coverage calculations. bEstimates obtained from the UCSC Known Gene human annotation (hg19). cEstimates obtained from https://www.ncbi.nlm.nih.gov/books/NBK22266/.

Table 3. Association Results for Variants in Previously Published Loci for Corneal Curvature and Spherical Equivalent GWAS BDES BDES Gene GWAS Sample GWAS GWAS BDES Locusa GWAS Trait GWAS Variantb Effect Size Effect P- [MIM] Size P-value MAF MAF (CI) Size (CI)c value Corneal Curvature MIPEP 0.44/0.50 -0.04 rs9318086 6,311/3,222 OR: 1.32 1.91 × 4.81 × 13q12.12 [602241] High Myopia (control/ (-0.06, - 0.44 (g.24432467G>A) (control/case) (1.19, 1.46) 10-16 10-3 (64) case) 0.01) MIPEP 0.46/0.51 0.05 rs9510902 6,311/3,222 OR: 1.25 6.32 × 2.41 × 13q12.12 [602241] High Myopia (control/ (0.02, 0.42 (g.24430312G>A) (control/case) (1.12, 1.38) 10-14 10-3 (64) case) 0.08) MIPEP 0.44/0.50 0.04 rs1886970 6,311/3,222 OR: 1.27 2.31 × 2.92 × 13q12.12 [602241] High Myopia (control/ (0.01, 0.42 (g.24440498T>C) (control/case) (1.15, 1.41) 10-14 10-3 (64) case) 0.07) Normalized β: -0.13 -0.03 4q12 rs2114039 1.33 × 4q12 Corneal 10,008 (-0.19, - 0.27-0.30 (-0.06, - 0.01 0.28 (65) (g.55092626T>C) 10-9 Curvature Score 0.06) 0.01) Normalized β: -0.15 0.05 4q12 4q12 rs17084051 2.23 × Corneal 10,008 (-0.22, - 0.20-0.26 (0.01, 0.01 0.27 (65) (g.55087581C>A) 10-8 Curvature Score 0.09) 0.08) β: -0.13 0.05 4q12 4q12 Corneal rs17084051 4.50 × 9,383 (-0.16, - 0.20-0.26 (0.01, 0.01 0.27 (32) Curvature (g.55087581C>A) 10-14 0.09) 0.08) rs1800813 0.05 4q12 4q12 Corneal β: 0.10 7.50 × (g.55094467G>A) 9,913 0.77 (0.01, 0.01 0.27 (31) Curvature (0.06, 0.14) 10-9 0.08) Spherical Equivalent β: -0.23 -0.27 15q14 Spherical rs634990 2.21 × 3.79 × 15q14 15,608 (-0.23, - 0.47 (-0.45, - 0.47 (66) Equivalent (g.35006073T>C) 10-14 10-3 0.23) 0.09) 15q14 Spherical rs685352 β: -0.21 4.19 × 0.29 3.12 × 15q14 15,608 0.44 0.47 (66) Equivalent (g.35008335A>G) (-0.21, - 10-12 (0.10, 10-3

0.21) 0.49) rs685352 β: -0.08 0.29 15q14 Spherical 2.09 × 3.12 × 15q14 (g.35008335A>G) 26,615 (-0.01, - 0.46 (0.10, 0.47 (67) Equivalent 10-10 10-3 0.06) 0.49) rs1550094 PRSS56 -0.32 (g.233385396A>G/ HR: 1.09 1.30 × 2.42 × 2q13 [613858] Myopia 45,771 0.30 (-0.52, - 0.31 p.Ala30Thr (1.07, 1.11) 10-15 10-3 (68) 0.12)

Abbreviations: CI, confidence interval; GWAS, genome-wide association study; HR, hazard ratio; MAF, minor allele frequency; MIM, Mendelian Inheritance in Man; OR, odds ratio. aReported in GRCh37 (hg19). bVariants listed in bold were genotyped on the exome array. The remaining variants were imputed. cSingle variant associations of corneal curvature adjusted for age and sex measured at visit 4, and baseline height; single variant associations for spherical equivalent adjusted for baseline age, sex, education, and nuclear sclerosis.

Table 4. Comparison of Variants Associated With Ocular Phenotypes Before and After Imputation of the Exome Array Ocular Phenotype in the Number of Variants Number of Variants Number of Loci Identified BDES Associated (Non-Imputed Associated (Imputed Exome From Imputation of Exome Exome Array)a Array)a Arrayb Axial Length 0 0 0 Corneal Curvature 2 7 0 Spherical Equivalent 2 4 0 aVariants associated with phenotype at P < 0.01. bNumber of loci associated from imputation of the exome array that were not associated prior to imputation.

Figure 1. Coverage plots showing the geographic distribution of variants across: (A) all of chromosome 22 and (B) a portion of chromosome 22 (22q12.1-22q12.2) for the imputed (top) and non-imputed exome array (bottom) datasets. Chromosomal location is shown at the top of each plot, with the corresponding locations of all chromosomal genes along the bottom. Variants are represented in the center by p-values obtained by conducting the single variant analyses described in this study for axial length of the right eye. The magnitudes of the p-values are inversely proportional to the size of the data points shown. All variants with imputation R2 ≥ 0.6 were included. Variants with MAF < 0.57% were not excluded (as they were in the single variant analysis) to accurately depict full coverage of the variants. Variants genotyped on the exome array were given imputation R2 = 1 and are included in the coverage plots of the imputed data. Imputation appears to have made the regions covered by the exome array denser, without extending far beyond those regions to cover those not included on the exome array.

CHAPTER FOUR

Bivariate Heritability Shows Evidence of Shared Genetic Effects of Refraction and

Nuclear Sclerosis: The Beaver Dam Eye Study

Abstract

Nuclear cataract is a primary contributor to refractive shifts observed in older persons. Both refraction and nuclear sclerosis are heritable, polygenic traits that likely share genetic and environmental effects. We sought to determine whether the heritability of refraction changed across time points with differing distributions of nuclear sclerosis, or across categories of nuclear sclerosis severity, due to these shared effects. We also aimed to quantify the shared genetic and environmental effects between refraction and nuclear sclerosis in related European-Americans over age 40 from the Beaver Dam Eye

Study. The heritability of refraction was consistent across time points despite different distributions of nuclear sclerosis, and across categories of nuclear sclerosis severity.

Refraction and nuclear sclerosis showed a significant genetic, but not environmental, correlation. Approximately one third to one half of the genetic effects were shared between these traits. The proportion of genetic effects contributing to refraction is unaffected by differences in nuclear sclerosis severity. However, there are common genetic influences contributing to both refraction and nuclear sclerosis. The genetic mechanisms of nuclear sclerosis remain largely unknown. Identifying novel loci associated with nuclear sclerosis could improve our understanding of the etiology of nuclear sclerosis and ultimately that of refraction.

Introduction

As the world’s population ages (1), refractive errors, or disorders in which the eye cannot properly focus images, are becoming an increasingly important public health problem among older adults. Refractive errors, including myopia and hyperopia, are the most common causes of visual impairment in the United States and the world (2–5).

Although a majority of those affected by uncorrected refractive errors are over 50 (6), only 20% of the world’s population falls in this age group (7,8). Individuals over the age of 40 in western countries like the United States are also particularly impacted, with nearly one third of these individuals suffering from myopia alone (5,9).

Refraction changes with age in older populations. Individuals become more hyperopic with age until age 70, after which they become more myopic. Several studies, including our own unpublished study, suggest the primary contributor to the myopic shift observed in older adults is nuclear cataract (10–17). Nuclear cataract is a severe form of nuclear sclerosis (nuclear lens opacification) that develops with age and is the most common type of age-related cataract. Studies have even claimed nuclear cataract is the cause of the myopic shift in refraction in this age group (16,18).

Refraction has a significant genetic component, with heritability estimates reaching as high as 88% in some adult populations (19–22). The estimated heritability of refraction was 58% among European-Americans over the age of 40 in the Beaver Dam

Eye Study (BDES) (20). Nuclear sclerosis has been suggested to have a genetic component in past studies as well, with heritability estimates ranging from 35 to 48%

(23–27). Since severe nuclear sclerosis (i.e. nuclear cataract) results in changes in refraction, and refraction and nuclear sclerosis have been established as complex,

88 heritable polygenic traits with both genetic and environmental factors contributing to their etiology (19–25,27,28), these traits no doubt share genetic and/or environmental influences. The purpose of this study was to determine whether differences in the distribution of nuclear sclerosis alters the heritability of refraction over time as a result of shared effects, to determine if the heritability of refraction differs across strata of nuclear sclerosis severity due to shared effects, and to quantify the shared genetic and/or environmental effects between refraction and nuclear sclerosis.

Methods

The Beaver Dam Eye Study

The population-based BDES cohort was initiated in 1987. Recruitment and study design procedures have been described previously (29–32). Briefly, a private census was conducted in the city and township of Beaver Dam, Wisconsin, which identified 5,924 residents between the ages of 43 and 84 from 3,715 households. A total of 602 pedigrees were reconstructed from 2,783 eligible participants who had confirmed familial relationships. Of the 5,924 eligible residents, 4,926 (83.14%) participated in the baseline examination between 1988 and 1990. Ninety-nine percent of the population was

European.

A complete ocular examination was conducted at the baseline visit, which included a standardized evaluation of refraction using the Humphrey 530 refractor

(Humphrey, Humphrey Inc., San Leandro, CA) (29). The standard formula for refraction

(sphere + 0.5 x cylinder) was used to calculate the mean spherical equivalent in diopters

(D) for each eye. Grades of the severity of nuclear lens opacity (nuclear sclerosis) were

89 assigned by trained photograders using a 5-point scale based on slitlamp lens photographs. Grades of 4 or 5 indicated nuclear cataract (33,34). Age, sex, total number of years of education, and smoking status were also collected at baseline. An individual was categorized as a never, former, or current smoker based on self-report.

Follow-up examinations were conducted every five years. The total number of individuals participating in the BDES at the five, ten, fifteen, and 20-year follow-up examinations were as follows: 3,721, 2,962, 2,375, and 1,913. A majority (68.9%) of losses to follow-up were due to death. Measurements made during the baseline examination were repeated using similar procedures for all follow-up examinations with the exception of sex and education, which were carried forward to subsequent visits.

The Institutional Review Board at the University of Wisconsin approved the

BDES. Written informed consent was obtained from all subjects and the study was performed in accordance with the tenets of the Declaration of Helsinki.

Trait Definitions

Refraction was estimated continuously using spherical equivalent measures.

Values of spherical equivalent < -0.5 D indicated myopia and values > +0.5 D indicated hyperopia. Shifts toward myopia were defined as changes in spherical equivalent over time resulting in a negative change, while shifts toward hyperopia were defined as changes in spherical equivalent resulting in a positive change (11,35). Nuclear sclerosis was treated as a continuous variable when estimating its heritability, and as a categorical variable when used to categorize individuals based on nuclear sclerosis severity. Nuclear sclerosis severity was categorized as follows: mild (grade of 1 or 2), moderate (grade of

3), or severe (grade of 4 or 5). Grades of 4 or 5 indicated nuclear cataract.

Exclusion Criteria

Of the 4,972 individuals included in our dataset, individuals not present for the baseline examination (n = 45), of non-European ancestry (n = 31), and with differences in baseline spherical equivalent between the right and left eyes > ±4 D (n = 23) were removed. Eyes that had undergone refractive surgery, eyes with no lens, eyes with an intraocular lens, or eyes with best corrected visual acuity of 20/200 or worse at baseline were excluded (n = 319). Eyes that developed one of these conditions at a subsequent visit were censored at that visit. Individuals missing spherical equivalent, nuclear sclerosis, or other relevant measures were excluded (n = 113). The cohort was then restricted to individuals with complete data at visits 1 and 3 (n = 2,110) who had at least one other family member with complete data in the dataset, leaving 719 individuals for further analysis.

Statistical Analysis

Narrow-sense heritability, or the proportion of variation attributable to additive genetic effects, was calculated for spherical equivalent and nuclear sclerosis at the baseline (visit 1) and second follow-up (visit 3) visits. We calculated heritability among a subset of individuals who participated in both visits to facilitate direct comparison among the same group of individuals at different time points (n = 719). The distribution of nuclear sclerosis at each visit was different, allowing us to determine whether heritability estimates for refraction vary by differing distributions of nuclear sclerosis over time. Data from later visits were not used due to the limited number of individuals with family members remaining after attrition. We also compared heritability estimates among the same group of individuals ten years apart based on age (n = 314) to ensure any patterns

91 we observed over time based on visit were not influenced by secular trends. We estimated the heritability of refraction among individuals when they were 55-60 years old and again ten years later when they were 65-70 years old. Sample sizes in older age groups were not sufficient for heritability estimation. Due to the extremely small number of individuals with nuclear cataract at visit 1 (n = 27), we estimated the heritability of refraction across nuclear sclerosis strata at visit 3 only. Residual values were obtained from linear regression with covariates found to be significantly associated with spherical equivalent (age and education) or nuclear sclerosis (age, sex, and smoking status).

Residuals were calculated as the summation of the products of all covariate β coefficients, plus the intercept. These residuals were used in SOLAR to estimate the narrow-sense heritability under a variance component framework (36).

Bivariate heritability analyses were also conducted at visits 1 and 3 using SOLAR to determine if there was evidence for shared genetic and/or environmental effects between spherical equivalent and nuclear sclerosis (36). The genetic correlation (ρg) and environmental correlation (ρe) between these two phenotypes quantified any shared genetic effects or shared environmental effects, respectively. Analyses were conducted using measures of the right eye and then were repeated using measures of the left eye to ensure consistency of results.

Results

Study Participants

Table 1 shows the distributions of relevant characteristics among the group of individuals who were members of families and contributed to the heritability analyses at

92 both visits (n = 719). The familial group at visit 3 was more hyperopic and older compared to what they had been at visit 1, as expected. In addition, the familial group had a higher percentage of individuals with nuclear cataract at visit 3 compared to visit 1.

Comparison of the age groups (55-60 years and 65-70 years) showed similar patterns

(Table 1).

Table 2 shows the number of each type of relative pair contributing to the heritability analyses at visits 1 and 3, in addition to the number and type of relative pairs categorized by nuclear sclerosis severity at visit 3. A total of 719 individuals from 209 pedigrees were included in the heritability analyses at visits 1 and 3. Siblings and first cousins comprised the largest number of relative pairs contributing to the heritability analyses at visits 1 and 3 (Table 2). A total of 314 individuals from 106 pedigrees were included in the heritability analysis at ages 55-60 and 65-70, and they showed a similar distribution of relative pairs to the 719 individuals at visits 1 and 3 (Table 2).

Heritability of Refraction at Visits 1 and 3

We estimated the heritability (h2) of refraction at visits 1 and 3, adjusted for age and education (Table 3). Results of the right and left eyes were consistent, so only the results of the right eye are reported. The heritability for refraction at visit 1 (h2 = 0.80,

95% confidence interval (CI): 0.60, 1.00) overlapped with the heritability at visit 3 (h2 =

0.71, 95% CI: 0.51, 0.91) (Table 3). Unadjusted heritability estimates at visits 1 and 3 were similar. To ensure secular trends were not altering these results, the heritability of refraction was also estimated by age group. The heritability for refraction among those aged 55-60 (h2 = 0.95, 95% CI: 0.64, 1.00) overlapped with the heritability (h2 = 0.83,

95% CI: 0.52, 1.00) among the same individuals aged 65-70 (i.e. ten years older). These results were consistent with the results by visit, showing consistent heritability over time.

Heritability of Refraction by Nuclear Sclerosis Severity

The heritability of refraction was estimated within each category of nuclear sclerosis severity (mild, moderate, or severe) at visit 3, adjusted for age and education

(Figure 1). Heritability point estimates for those with mild (h2 = 0.53, 95% CI: 0, 1), moderate (h2 = 0.80, 95% CI: 0.51, 1), or severe nuclear sclerosis (h2 = 0.60, 95% CI: 0,

1) at visit 3 appeared to vary, but their confidence intervals overlapped (Figure 1). The heritability estimates for refraction among those with mild or severe nuclear sclerosis at visit 3 were not significantly different from zero, reflecting the small number of pedigrees contributing to these estimates. A similar analysis conducted among all individuals with at least one family member with complete data at visit 1 (including those who did not participate at visit 3; n = 1,787) showed a similar pattern of overlapping confidence intervals across nuclear sclerosis strata following an increase in sample size: mild (h2 =

0.58, 95% CI: 0.38, 0.78), moderate (h2 = 0.43, 95% CI: 0.12, 0.74), or severe nuclear sclerosis (h2 = 0.86, 95% CI: 0.33, 1). Unadjusted heritability estimates showed similar trends.

Shared Effects of Refraction and Nuclear Sclerosis

The genetic correlations between refraction (adjusted for age and education) and nuclear sclerosis (adjusted for age, sex, and smoking status) were significant at visits 1 and 3 (ρg = 0.46, 95% CI: 0.03, 0.89 and ρg = 0.33, 95% CI: 0.06, 0.60, respectively), but the environmental correlations were not (ρe = -0.20, 95% CI: -0.53, 0.13 and ρe = -0.37,

95% CI: -0.76, 0.02, respectively) (Table 3). This remained true when comparing refraction and nuclear sclerosis unadjusted for any covariates.

Discussion

The heritability estimates for refraction were consistent across visits ten years apart (visits 1 and 3). This indicates the heritability of refraction does not change appreciably over time within a population. The distribution of nuclear sclerosis differed at these two visits, which is not surprising since all individuals included in the analysis had aged ten years since visit 1. As a result, the number of individuals with moderate or severe nuclear sclerosis increased. However, differences in the distribution of nuclear sclerosis did not alter the estimated heritability of refraction over time. Heritability estimates for refraction were also consistent across age groups ten years apart (55-60 year-olds vs. 65-70 year-olds), further supporting the results obtained by comparing heritability estimates at visits 1 and 3. This suggests the proportion of additive genetic effects contributing to refraction is unaffected by changes in nuclear sclerosis severity.

This remained true when comparing heritability estimates for refraction unadjusted for age or education.

Furthermore, the heritability of refraction did not appear to significantly differ across strata of nuclear sclerosis severity at visit 3. However, reduced sample sizes within each stratum at visit 3 resulted in these non-significant heritability estimates. We therefore estimated the heritability of refraction among those with mild, moderate, and severe nuclear sclerosis at visit 1 (including those who did not participate at visit 3) to increase our sample size and improve our ability to draw inferences. The heritability of

95 refraction appeared consistent across nuclear sclerosis strata at visit 1, supporting our conclusion that the genetic effects contributing to refraction are consistent across varying degrees of nuclear sclerosis severity. The consistency of the heritability of refraction over time and across nuclear sclerosis strata suggests differences in the distribution of nuclear sclerosis across populations do not significantly impact the heritability of refraction. This may be because those with varying degrees of nuclear sclerosis share the same underlying genetic effects for refraction, but those with milder nuclear sclerosis have not yet developed nuclear cataract while older individuals have.

The bivariate heritability analyses showed heritability estimates for refraction and nuclear sclerosis individually, and they are consistent with previous findings within and outside of the BDES (19–27). As expected, this analysis showed a significant genetic correlation between refraction and nuclear sclerosis at both visits. The environmental correlation was not significant at either visit. However, the environmental correlation showed a negative correlation between refraction and nuclear sclerosis, suggesting more negative refraction (i.e. myopia) is correlated with more severe nuclear sclerosis, which we expect. These results indicate approximately one third to one half of the genetic factors influencing refraction also influence nuclear sclerosis. It is not surprising that these traits have common genetic influences, considering the strong association between nuclear sclerosis and refraction (10,11,13,14).

In summary, the heritability of refraction was consistent over time in spite of differences in the distribution of nuclear sclerosis. The heritability of refraction was also consistent across categories of nuclear sclerosis severity. Finally, this study showed roughly one third to one half of the genetic influences underlying refraction are shared by

96 nuclear sclerosis. Taken together, the results of this study demonstrate that while refraction and nuclear sclerosis share some underlying genetic influences, the large proportion of genetic effects contributing to refraction remain unaffected by the distribution of nuclear sclerosis severity.

There are several strengths to this study. We had access to refraction and nuclear sclerosis measurements among thousands of individuals (including those with nuclear cataract) across five visits spanning a period of 20 years. Although sample sizes at visits

4 and 5 were too small to be included in the heritability analyses, particularly after categorization by nuclear sclerosis severity, sample sizes at visits 1 and 3 were sufficient to allow for a comparison of heritability estimates across a ten-year period within the same cohort. In addition, 83% of all individuals in Beaver Dam in 1987 over the age of

40 participated in the baseline examination, making the study sample representative of older adults in Beaver Dam at the time. The results of our study are therefore generalizable to other similar adult European-American populations.

Refraction and nuclear sclerosis share common genetic influences. Given the relative lack of genome-wide association studies conducted to investigate the genetic mechanisms of nuclear sclerosis (37,38), this phenotype is an ideal candidate for future genetic association studies. Only two genetic loci have been significantly associated with age-related nuclear cataract to date in genome-wide association studies, neither of which have been associated with refraction (38). Genetic studies identifying novel loci associated with nuclear sclerosis can help elucidate the biological pathways underlying the etiology of this trait, which is such a strong risk factor for refractive change.

Identification of novel loci may ultimately explain some of the missing heritability of

97 refraction that remains in spite of the discovery of nearly 40 associated genetic loci to date (2,39–57).

Acknowledgements

This work was supported by the National Eye Institute at the National Institutes of Health

(grants R01EY021531, U10006594, and 1T32EI022303); and by the Research to Prevent

Blindness Unrestricted Grant to the University of Wisconsin Department of

Ophthalmology and Visual Sciences.

We thank all staff and investigators of the Beaver Dam Eye Study.

Conflict of interest: none declared.

References

1. He W, Goodkind D, Kowal P. An Aging World : 2015. Washington, DC: 2016. 2. Hysi PG, Wojciechowski R, Rahi JS, et al. Genome-Wide Association Studies of Refractive Error and Myopia, Lessons Learned, and Implications for the Future. Investig. Opthalmology Vis. Sci. 2014;55(5):3344–3351. 3. National Eye Institute. Fact Sheet: Refractive Errors. 2010;(https://nei.nih.gov/sites/default/files/health- pdfs/HVM09_Fact_Sheet_Final_tagged.pdf). (Accessed November 9, 2015) 4. Centers for Disease Control and Prevention. Refractive Errors. Vis. Heal. Initiat. 2013;(http://www.cdc.gov/visionhealth/basic_information/eye_disorders.htm#a2). (Accessed November 9, 2015)

5. Meng W, Butterworth J, Malecaze F, et al. Axial Length of Myopia: A Review of Current Research. Ophthalmologica. 2011;225(3):127–134. 6. Resnikoff S, Pascolini D, Mariotti S, et al. Global Magnitude of Visual Impairment Caused by Uncorrected Refractive Errors in 2004. Bull. World Health Organ. 2008;86(1):63–70. 7. WHO. What is a Refractive Error? 2017;(http://www.who.int/features/qa/45/en/). (Accessed August 4, 2017) 8. WHO. Visual Impairment and Blindness. 2017;(http://www.who.int/mediacentre/factsheets/fs282/en/). (Accessed August 4, 2017) 9. Kempen JH, Mitchell P, Lee KE, et al. The Prevalence of Refractive Errors Among Adults in the United States, Western Europe, and Australia. Arch. Ophthalmol. 2004;122:495–505. 10. Lee KE, Klein BE, Klein R. Changes in Refractive Error Over a 5-Year Interval in the Beaver Dam Eye Study. Invest. Ophthalmol. Vis. Sci. 1999;40(8):1645–1649. 11. Lee KE, Klein BEK, Klein R, et al. Changes in Refraction Over 10 Years in an Adult Population: The Beaver Dam Eye Study. Invest. Ophthalmol. Vis. Sci. 2002;43(8):2566–2571. 12. Gudmundsdottir E, Arnarsson A, Jonasson F. Five-Year Refractive Changes in an Adult Population: Reykjavik Eye Study. Ophthalmology. 2005;112(4):672–677. 13. Guzowski M, Wang JJ, Rochtchina E, et al. Five-Year Refractive Changes in an Older Population: The Blue Mountains Eye Study. Ophthalmology. 2003;110(7):1364–1370. 14. Fotedar R, Mitchell P, Burlutsky G, et al. Relationship of 10-Year Change in Refraction to Nuclear Cataract and Axial Length Findings From an Older

Population. Ophthalmology. 2008;115(8):1273–1278. 15. Panchapakesan J, Rochtchina E, Mitchell P. Myopic Refractive Shift Caused by Incident Cataract: The Blue Mountains Eye Study. Ophthalmic Epidemiol. 2003;10(4):241–247. 16. Samarawickrama C, Wang JJ, Burlutsky G, et al. Nuclear Cataract and Myopic Shift in Refraction. Am. J. Ophthalmol. 2007;144(3):457–459. 17. Bomotti S, Duggal P, Klein BEK, et al. Change in Refraction Over a 20-Year Period in the Beaver Dam Eye Study. 2017. 18. Brown NAP, Hill AR. Cataract : The Relation Between Myopia and Cataract Morphology. Br. J. Ophthalmol. 1987;71:405–414. 19. Dirani M, Chamberlain M, Shekar SN, et al. Heritability of Refractive Error and Ocular Biometrics: The Genes in Myopia (GEM) Twin Study. Invest. Ophthalmol. Vis. Sci. 2006;47(11):4756–4761. 20. Klein AP, Suktitipat B, Duggal P, et al. Heritability Analysis of Spherical Equivalent, Axial Length, Corneal Curvature, and Anterior Chamber Depth in the Beaver Dam Eye Study. Arch. Ophthalmol. 2009;127(5):649–655. 21. Chen CYC, Scurrah KJ, Stankovich J, et al. Heritability and Shared Environment Estimates for Myopia and Associated Ocular Biometric Traits: The Genes in Myopia (GEM) Family Study. Hum. Genet. 2007;121:511–520. 22. Dirani M, Shekar SN, Baird PN. Evidence of Shared Genes in Refraction and Axial Length: The Genes in Myopia (GEM) Twin Study. Invest. Ophthalmol. Vis. Sci. 2008;49(10):4336–4339. 23. Yonova-Doing E, Forkin ZA, Hysi PG, et al. Genetic and Dietary Factors Influencing the Progression of Nuclear Cataract. Ophthalmology. 2016;123(6):1–8. 24. Hammond CJ, F.R.C.Ophth., Snieder H, et al. Genetic and Environmental Factors in Age-Related Nuclear Cataracts in Monozygotic and Dizygotic Twins. N. Engl. J. Med. 2000;342(24):1786–1790. 25. Congdon N, Broman KW, Lai H, et al. Nuclear Cataract Shows Significant Familial Aggregation in an Older Population after Adjustment for Possible Shared Environmental Factors. Invest Ophthalmol Vis Sci. 2004;45(7):2182–2186. 26. Heiba IM, Elston RC, Klein BEK, et al. Genetic Etiology of Nuclear Cataract: Evidence for a Major Gene. Am. J. Med. Genet. 1993;47(8):1208–1214. 27. Klein AP, Duggal P, Lee KE, et al. Support for Polygenic Infuences on Ocular Refractive Error. Investig. Ophthalmol. Vis. Sci. 2005;46(2):442–446. 28. Klein AP, Duggal P, Lee KE, et al. Polygenic Effects of Cigarette Smoking Account for a Portion of the Familial Aggregation of Nuclear Sclerosis. Am. J.

100

Epidemiol. 2005;161(8):707–713. 29. Linton KL, Klein BE, Klein R. The Validity of Self-Reported and Surrogate- Reported Cataract and Age-Related Macular Degeneration in the Beaver Dam Eye Study. Am. J. Epidemiol. 1991;134(12):1438–1446. 30. Klein R, Klein BE, Linton KL, et al. The Beaver Dam Eye Study: Visual Acuity. Ophthalmology. 1991;98(8):1310–1315. 31. Klein R, Klein BE, Lee KE. Changes in Visual Acuity in a Population. The Beaver Dam Eye Study. Ophthalmology. 1996;103(8):1169–1178. 32. Klein R, Klein BE, Lee KE, et al. Changes in Visual Acuity in a Population Over a 10-year Period : The Beaver Dam Eye Study. Ophthalmology. 2001;108(10):1757– 1766. 33. Klein BE, Klein R, Linton KL, et al. Assessment of Cataracts from Photographs in the Beaver Dam Eye Study. Ophthalmology. 1990;97(11):1428–1433. 34. Klein BE, Klein R, Linton K. Prevalence of Age-Related Lens Opacities in a Population. The Beaver Dam Eye Study. Ophthalmology. 1992;99(4):546–552. 35. Chen F, Duggal P, Klein BEK, et al. Variation in PTCHD2, CRISP3, NAP1L4, FSCB, and AP3B2 Associated with Spherical Equivalent. Mol. Vis. 2016;22:783– 796. 36. Almasy L, Blangero J. Multipoint Quantitative-Trait Linkage Analysis in General Pedigrees. Am. J. Hum. Genet. 1998;62(5):1198–211. 37. Ritchie MD, Verma SS, Hall MA, et al. Electronic Medical Records and Genomics (eMERGE) Network Exploration in Cataract: Several New Potential Susceptibility Loci. Mol. Vis. 2014;20:1281–1295. 38. Liao J, Su X, Chen P, et al. Meta-Analysis of Genome-Wide Association Studies in Multiethnic Asians Identifies Two Loci for Age-Related Nuclear Cataract. Hum. Mol. Genet. 2014;23(22):6119–6128. 39. Klein AP, Duggal P, Lee KE, et al. Linkage Analysis of Quantitative Refraction and Refractive Errors in the Beaver Dam Eye Study. Invest. Ophthalmol. Vis. Sci. 2011;52(8):5220–5225. 40. Klein AP, Duggal P, Lee KE, et al. Confirmation of Linkage to Ocular Refraction on Chromosome 22q and Identification of a Novel Linkage Region on 1q. Arch. Ophthalmol. 2007;125(1):80–85. 41. Hawthorne F a., Young TL. Genetic Contributions to Myopic Refractive Error: Insights from Human Studies and Supporting Evidence from Animal Models. Exp. Eye Res. 2013;114:141–149. 42. Chen JH, Chen H, Huang S, et al. Endophenotyping Reveals Differential

101

Phenotype-Genotype Correlations Between Myopia-Associated Polymorphisms and Eye Biometric Parameters. Mol. Vis. 2012;18:765–778. 43. Fan Q, Barathi V a, Cheng C-Y, et al. Genetic Variants on Chromosome 1q41 Influence Ocular Axial Length and High Myopia. PLoS Genet. 2012;8(6):1–14. 44. Hysi PG, Young TL, Mackey D, et al. A Genome-Wide Association Study for Myopia and Refractive Error Identifies a Susceptibility Locus at 15q25. Nat. Genet. 2010;42(10):902–905. 45. Khor CC, Miyake M, Chen LJ, et al. Genome-Wide Association Study Identifies ZFHX1B as a Susceptibility Locus for Severe Myopia. Hum. Mol. Genet. 2013;22(25):5288–5294. 46. Kiefer AK, Tung JY, Do CB, et al. Genome-Wide Analysis Points to Roles for Extracellular Matrix Remodeling, the Visual Cycle, and Neuronal Development in Myopia. PLoS Genet. 2013;9(2):1–8. 47. Li Z, Qu J, Xu X, et al. A Genome-Wide Association Study Reveals Association Between Common Variants in an Intergenic Region of 4q25 and High-Grade Myopia in the Chinese Han Population. Hum. Mol. Genet. 2011;20(14):2861– 2868. 48. Li Y-J, Goh L, Khor C-C, et al. Genome-Wide Association Studies Reveal Genetic Variants in CTNND2 for High Myopia in Singapore Chinese. Ophthalmology. 2011;118(2):368–375. 49. Miyake M, Yamashiro K, Tabara Y, et al. Identification of Myopia-Associated WNT7B Polymorphisms Provides Insights into the Mechanism Underlying the Development of Myopia. Nat. Commun. 2015;6(6689):1–9. 50. Nakanishi H, Yamada R, Gotoh N, et al. A Genome-Wide Association Analysis Identified a Novel Susceptible Locus for Pathological Myopia at 11q24.1. PLoS Genet. 2009;5(9):1–7. 51. Shi Y, Qu J, Zhang D, et al. Genetic Variants at 13q12.12 are Associated with High Myopia in the Han Chinese Population. Am. J. Hum. Genet. 2011;88(6):805– 813. 52. Shi Y, Gong B, Chen L, et al. A Genome-Wide Meta-Analysis Identifies Two Novel Loci Associated with High Myopia in the Han Chinese Population. Hum. Mol. Genet. 2013;22(11):2325–2333. 53. Simpson CL, Wojciechowski R, Oexle K, et al. Genome-Wide Meta-Analysis of Myopia and Hyperopia Provides Evidence for Replication of 11 Loci. PLoS One. 2014;9(9):1–19. 54. Solouki AM, Verhoeven VJM, van Duijn CM, et al. A Genome-Wide Association Study Identifies a Susceptibility Locus for Refractive Errors and Myopia at 15q14. Nat. Genet. 2010;42(10):897–901.

102

55. Verhoeven VJM, Hysi PG, Wojciechowski R, et al. Genome-Wide Meta-Analyses of Multiancestry Cohorts Identify Multiple New Susceptibility Loci for Refractive Error and Myopia. Nat. Genet. 2013;45(3):314–318. 56. Cheng C-Y, Schache M, Ikram MK, et al. Nine Loci for Ocular Axial Length Identified through Genome-Wide Association Studies, Including Shared Loci with Refractive Error. Am. J. Hum. Genet. 2013;93(2):264–277. 57. Wojciechowski R. Nature and Nurture: The Complex Genetics of Myopia and Refractive Error. Clin. Genet. 2011;79(4):301–320.

103

Intended to be Blank

104

Tables and Figures Table 1. Ocular and Demographic Characteristics of Participants of the Beaver Dam Eye Study Cohorta Nb Mean Mean Age in Female Sex Mean Mean Nuclear Nuclear Spherical Years (%) Education in Sclerosis Cataract (%) Equivalent (D) (SD) Years Grade (SD) (SD) (SD) Visit 1 719 0.24 (2.1) 57.7 (8.8) 52.0 11.7 (2.2) 2.2 (0.8) 3.8

Visit 3 719 0.54 (2.2) 67.8 (8.8) 52.0 11.7 (2.2) 2.8 (0.7) 15.9

Ages 55-60 314 0.06 (1.9) 57.3 (1.6) 48.4 12.2 (1.9) 2.3 (0.7) 1.3

Ages 65-70 314 0.61 (2.0) 66.9 (1.5) 48.4 12.2 (1.9) 2.7 (0.7) 8.0

Abbreviations: D, diopters; SD, standard deviation. aCharacteristics of all individuals with at least one other family member with phenotype data at both visits and both age groups are presented. The same measures are reported at visits 1 and 3 and at ages 55-60 and 65-70 for the same individuals. bThe number of individuals with at least one other family member who also has complete phenotype data.

105

Table 2. Relative Pairs Contributing to the Heritability Analyses for Spherical Equivalent and Nuclear Sclerosis Among Participants of the Beaver Dam Eye Study Cohort Number of Na Parent- Siblings Avuncular Half First Fourth Pedigrees Offspring Siblings Cousins Degree or Higher Visits 1 and 3 Total 209 719 39 241 95 12 304 170 Visit 3 Mild Nuclear 46 131 1 26 2 2 40 28 Sclerosis Moderate 116 326 8 94 21 1 96 20 Nuclear Sclerosis Nuclear 26 62 0 11 1 1 13 1 Cataract Age Groups 55-60 and 65-70 Total 106 314 0 87 7 2 107 53 aThe number of individuals with at least one other family member who also has complete phenotype data.

106

Table 3. Bivariate Heritability Analysis of Spherical Equivalent and Nuclear Sclerosis Among Participants of the Beaver Dam Eye Study Cohorta Na Heritability 95% CI Heritability 95% CI Genetic 95% CI Environmental 95% CI of of Nuclear Correlation Correlation c Spherical Sclerosis (ρg) (ρe) Equivalentb Visit 1 719 0.80 0.60, 1.00 0.27 0.03, 0.51 0.46 0.03, 0.89 -0.20 -0.53, 0.13 Visit 3 719 0.71 0.51, 0.91 0.57 0.33, 0.81 0.33 0.06, 0.60 -0.37 -0.76, 0.02 Abbreviations: CI, confidence interval. aThe number of individuals with at least one other family member who also has complete phenotype data. bAdjusted for age and education. cAdjusted for age, sex, and smoking status.

107

Visit 3 1.0

0.8

0.6

0.4

0.2

Heritability of Spherical Equivalent (95% CI) 0

All Mild Cataract Moderate Nuclear Sclerosis Severity

Figure 1. Narrow-sense heritability of spherical equivalent by nuclear sclerosis severity at visit 3, with error bars representing 95% confidence intervals. Heritability was measured among all family members at visit 3, and then among those with mild (grade 1 or 2), moderate (grade 3), or severe (grade 4 or 5) nuclear sclerosis, also known as nuclear cataract.

108

CHAPTER FIVE

Conclusions, Future Directions, and Public Health Significance

109

Summary and Discussion of Results

The complex etiology of refraction in older adults encompasses an array of non- genetic and genetic factors, contributing to the variation in refraction and influencing the refractive changes over time. Nuclear sclerosis severity was confirmed as an integral factor affecting refractive changes with age in the Beaver Dam Eye Study (BDES).

Specifically, nuclear cataract was the principal factor responsible for the myopic shifts in refraction observed following hyperopic shifts in older persons. Only those with nuclear cataract exhibited this change in trajectory of refraction from a hyperopic shift to a myopic shift. Nuclear cataract is characterized by morphologic changes to the lens, including lens thickening, changes in curvature, and opacification of the central zone

(nucleus), all of which result in myopic shifts (1–5). The strong association between nuclear sclerosis and changes in refraction has been well established in this and past studies (6–11).

Both refraction and nuclear sclerosis are polygenic traits with established genetic components and environmental influences (12–21). Since nuclear sclerosis (specifically nuclear cataract) is such a strong risk factor for refractive change, refraction and nuclear sclerosis clearly share some of their genetic and environmental influences. We estimated the heritability of refraction across time points and across strata of nuclear sclerosis severity to determine if the proportion of genetic effects contributing to refraction varied significantly depending on the distribution of nuclear sclerosis. The heritability of refraction remained consistent over time in spite of the different distributions of nuclear sclerosis at each time point. The heritability of refraction was also consistent across strata of nuclear sclerosis severity (mild, moderate, and severe nuclear sclerosis). This suggests the proportion of genetic effects underlying variation in refraction are unaffected by

110 differing distributions of the severity of nuclear sclerosis. We found about one third to one half of the genetic effects influencing refraction and nuclear sclerosis are shared between these traits. This provides further evidence that genetic loci associated with one of the traits may also contribute to the phenotypic variation observed in the other trait.

This should motivate future genetic association studies of nuclear sclerosis in particular, which has not been studied as extensively as refraction (22,23).

We were unable to detect novel genetic loci associated with refraction or two of its biological determinants, axial length and corneal curvature, in our imputed exome array analysis. We were thus unable to explain any more of the missing heritability remaining for refraction, a polygenic trait with many loci contributing to its etiology. We did replicate previously published loci associated with refraction (spherical equivalent and myopia) and corneal curvature, which validates these loci as contributors to the etiology of these traits (24–30). However, these loci were also replicated on the non- imputed exome array, indicating the imputation did not allow us to gain additional insight into the genetic mechanisms of these traits despite the denser exomic coverage provided.

Furthermore, we were unable to replicate any previously published loci associated with axial length in the non-imputed or imputed exome array. The modest sample size in this study limited our ability to detect these associations.

Because axial length was not measured in the BDES until visit 4 (2003-2005), were unable to assess the longitudinal effects of axial length on changing refraction.

Axial length, the largest determinant and endophenotype of refraction (31–34), has been found to be strongly associated with refraction in previous cross-sectional studies.

Specifically, shorter axial length predicted higher prevalence of hyperopia (35,36). We

111 lacked sufficient longitudinal data on axial length to confirm that axial length is the primary contributor to the hyperopic shift in refraction observed prior to the myopic shift in older adults.

Sex, diabetes status, and baseline refraction, but not education level, were associated with refraction in the BDES. These results were largely consistent with results of previous studies in the BDES (6,7). The lack of association between education and refraction indicates refraction and refractive shifts occur independently of education level, and are primarily driven by other factors in older individuals. It is possible the environmental exposures accumulated over decades among older individuals supersede the effects of any education these individuals received earlier in life. The significant association between baseline refraction and refraction at later visits showed the trajectory of refraction can vary depending on baseline refractive status. Sex and diabetes status also appeared to influence refraction in older adults, although the effects sizes were small. This suggests these factors should be taken into account when predicting refraction in this age group. The association of diabetes with refraction in particular can encourage individuals to adhere to a healthy lifestyle and avoid diabetes onset to minimize any negative effects that diabetes may have on refraction in older age.

Sex and diabetes are also risk factors for nuclear cataract, indicating the associations reported in this study represent the direct effects of sex and diabetes on refraction, not the indirect effects that would operate through nuclear cataract as a risk factor for refraction (and therefore a mediator between sex or diabetes and refraction). A sensitivity analysis excluding sex and diabetes from the longitudinal model revealed negligible effects on the association between nuclear sclerosis and refraction.

112

Strengths of the Studies

A vast majority (83%) of all eligible individuals living in Beaver Dam in 1987 participated in the baseline examination for the BDES. Those participating in the baseline examination were therefore representative of all adults over 40 in Beaver Dam that year.

The lack of an immigrative selection bias makes the results of our studies generalizable to other populations of similar age distribution and ancestry. A majority (68.9%) of the losses to follow-up that occurred at subsequent visits were due to death, indicating retention of participants was high among those who lived. We accounted for selection bias induced by loss to follow-up in our longitudinal analysis by applying inverse probability weights accounting for loss to follow-up due to death, drop-out, or clinical intervention (including cataract or refractive surgeries, removal of lens, etc.). Losses to follow-up were not accounted for in previous longitudinal studies of changing refraction

(6–9,37). In addition, we had access to data collected every five years over a 20-year period, allowing us to characterize changes in refraction over a longer period of time than any previous longitudinal study (7,9).

Our access to exome array genotypes from a subset of BDES participants presented a unique opportunity to identify rare (minor allele frequency (MAF) < 0.01) and low frequency (0.01 ≤ MAF ≤ 0.05) variants associated with refraction, axial length, and corneal curvature. To our knowledge, only one exome array analysis has been conducted using spherical equivalent as the phenotypic measure, and no exome array analyses have been conducted for axial length or corneal curvature (38). A majority of the genetic loci interrogated in previous studies involving these phenotypes primarily

113 comprised common variants in genome-wide association studies (GWAS). Furthermore, we had free access to the Haplotype Reference Consortium reference panel housed on the

Michigan Imputation Server. This large, predominantly European reference panel is ten times the size of the 1000 Genomes phase 3 reference panel, so we were able to accurately impute our exome array with thousands of high quality variants for downstream analysis.

We were fortunate that 2,783 of the 5,924 individuals eligible for the BDES had confirmed familial relationships and were members of one of 602 extended pedigrees, allowing us to use data collected in the BDES to conduct heritability analyses for refraction and nuclear sclerosis using a variance component framework. This was critical to our third study involving univariate and bivariate heritability analyses of refraction and nuclear sclerosis. The large number of individuals with familial relationships allowed us to maintain a sufficient sample size for the heritability analyses at the baseline second follow-up visits, facilitating a comparison of heritability measures across time points.

Limitations of the Studies

Since ocular biometric trait measures including axial length and corneal curvature were not collected until beginning at visit 4 of the BDES, we were unable to investigate the association of baseline axial length with changes in refraction in Aim 1. This would have allowed us to establish temporality and better characterize how axial length affected future changes in refraction, particularly among those without nuclear cataract. We were unable to examine the longitudinal trajectory of axial length and how it affected the trajectory of refraction. As more data are collected in the BDES, several repeated measures of axial length will become available for future analyses.

114

The lack of axial length or corneal curvature measures until visit 4 also restricted our sample sizes for these traits in our imputed exome array analyses. Loss to follow-up, primarily due to death, reduced the number of individuals with axial length or corneal curvature measures at visit 4. In addition, only a subset of the BDES participants was genotyped on the exome array, further limiting the number of individuals available for analysis. Our sample sizes for axial length and corneal curvature were modest as a result, which reduced our power to detect rare (MAF < 0.01) and low frequency (0.01 ≤ MAF ≤

0.05) variants in particular that were associated with either trait.

Losses to follow-up in the BDES also restricted our ability to measure heritability beyond visit 3 in our heritability analyses. We estimated heritability at visits 1 and 3, but not beyond visit 3 due to the limited number of individuals remaining at visits 4 and 5 with at least one other family member who was also still participating in the study. The number of individuals remaining with family members at visits 4 and 5 were especially insufficient following categorization by nuclear sclerosis severity, making the estimation of heritability impossible.

Future Directions

While genetic loci influencing refraction continue to be extensively investigated, genetic studies evaluating the genetic architecture of nuclear sclerosis are currently lacking. Although several studies have investigated whether genes previously associated with congenital cataract are also associated with age-related cataract, results of these studies have been inconsistent and lack replication (39–42). One GWAS meta-analysis identified two genes associated with age-related nuclear cataract in Asians, but similar findings were not available in populations of European descent (23). Another GWAS

115 examining all age-related cataract subtypes did not identify variants of genome-wide significance (22). The genetic correlation between nuclear sclerosis and refraction presents nuclear sclerosis as an ideal candidate for future genetic association studies. The genetic loci associated with nuclear sclerosis may elucidate the underlying biological mechanisms responsible for the development of nuclear cataract in older persons. It is also possible that identification of genetic loci associated with nuclear sclerosis may explain some of the missing heritability of refraction and ultimately enhance our understanding of the etiology of refraction, due to their shared genetic influences. In fact, bivariate association analyses involving nuclear sclerosis and refraction as the primary outcomes could increase statistical power to detect variants associated with both traits that may not be detectable in univariate analyses. This would provide evidence of pleiotropy.

Although we were unable to detect novel loci associated with either axial length or corneal curvature in our study, additional genetic studies for both traits are also warranted. Relatively few GWAS have interrogated common variants associated with axial length or corneal curvature compared to refraction (25,27,43–49). As such, additional GWAS in larger and more diverse samples will be necessary to identify novel loci associated with axial length and corneal curvature also influencing refraction. A bivariate analysis can also serve to increase the statistical power to detect associations and identify loci specifically associated with both biometric traits (pleiotropy). Genetic loci affecting both traits would likely also influence refraction. In addition, exome array analyses for all three phenotypes should be conducted in larger samples with sufficient power to detect associations with less common variants (MAF < 0.05). This may reveal

116 some of the suggestively significant variants from our study to be significantly associated variants in larger studies. Sequencing studies should also be undertaken to identify variants not included in genotyping arrays that may be associated with refraction, axial length, or corneal curvature.

Although several studies have investigated the association of axial length and refraction (35,36,50), few studies have reported on changes in axial length over time.

Several population-based studies, including the BDES, did not begin measuring biometric traits like axial length until years after the study was initiated. Longitudinal data on axial length in older adults is therefore limited. Future studies documenting the trajectory of axial length over time within elderly individuals, in addition to evaluating factors associated with changes in axial length, can shed new light on the etiology of axial length and allow us to develop a better understanding of how and why axial length changes over time as people age. The trajectory of axial length can also be compared to the trajectory of refraction to determine whether shortening of axial length with age predicts hyperopic shifts in refraction in the same cohort of individuals. This would provide stronger support for the hypothesis that axial length shortens with age, leading to hyperopic shifts in refraction prior to the myopic shift that occurs due to nuclear cataract. A comparison of changes in axial length with changes in refraction could also provide deeper insight into how axial length changes once individuals develop nuclear cataract, and how the trajectory of axial length differs between those with and without nuclear cataract. This could help elucidate how the shape of the eye changes upon the development of nuclear cataract, and how these changes ultimately influence refraction and refractive errors.

117

Public Health Significance

Refractive errors are the most prevalent eye disorders in the world (31,51–53). As the world’s population ages (54), refractive errors are becoming an increasing burden that requires accurate and detailed knowledge of the underlying etiology of refraction, including non-genetic and genetic factors that can influence the onset of refractive errors in older individuals, to anticipate future public health needs. Knowledge of how these factors affect refraction and induce refractive errors can prepare clinical service providers to develop preventative screening measures aimed at identifying early signs of refractive errors and taking necessary steps to minimize the clinical effects before they become more severe. For example, accurate data on the factors influencing refractive errors should motivate early and frequent assessment of refractive errors among elderly individuals who are at higher risk for developing refractive errors due to age-related factors like nuclear cataract. A deep understanding of factors influencing refraction would allow implementation of early intervention measures to minimize loss of vision.

This is especially important because studies have shown vision impairment to be significantly associated with other clinical and mental comorbidities, many of which are expensive and time-consuming to treat in their own right (55,56). Minimizing vision impairment can also minimize other comorbidities, which can have considerable positive emotional and financial implications for the patients.

These studies can also inform treatment options for those already suffering from advanced refractive errors. Information obtained regarding non-genetic factors affecting refraction can be applied by refractive surgeons, who need to be informed of the changes in refraction that are most likely to occur in their patients so they can better inform their

118 patients of potential changes to their vision in the future, and tailor treatments to fit their needs. A comprehensive understanding of the genetic architecture of refraction, which includes genetic loci underlying correlated traits such as axial length, corneal curvature, and nuclear sclerosis, could elucidate the biological mechanisms influencing the onset of refractive errors, some of which may be suitable targets for drug therapies. The development of drug therapies could replace the need for spectacles or contacts for those whose refractive errors can be corrected by targeting specific molecular pathways identified through etiologic research. This would greatly benefit those who struggle to pay for their spectacles or contacts, or who dislike wearing glasses on their face or putting contacts in their eyes on a daily basis.

119

References

1. Hemenger RP, Garner LF, Ooi CS. Change With Age of the Refractive Index Gradient of the Human Ocular Lens. Invest Ophthalmol Vis Sci. 1995;36(3):703– 707. 2. Ooi C, Grosvenor T. Mechanisms of Emmetropization in the Aging Eye. Optom. Vis. Sci. 1995;72(2):60–66. 3. Brown N, Koretz JF, Bron AJ. The Development and Maintenance of Emmetropia. Eye. 1999;13:83–92. 4. Brown NAP, Hill AR. Cataract : The Relation Between Myopia and Cataract Morphology. Br. J. Ophthalmol. 1987;71:405–414. 5. Brown N. The Morphology of Cataract and Visual Performance. Eye. 1993;7:63– 67. 6. Lee KE, Klein BE, Klein R. Changes in Refractive Error Over a 5-Year Interval in the Beaver Dam Eye Study. Invest. Ophthalmol. Vis. Sci. 1999;40(8):1645–1649. 7. Lee KE, Klein BEK, Klein R, et al. Changes in Refraction Over 10 Years in an Adult Population: The Beaver Dam Eye Study. Invest. Ophthalmol. Vis. Sci. 2002;43(8):2566–2571. 8. Guzowski M, Wang JJ, Rochtchina E, et al. Five-Year Refractive Changes in an Older Population: The Blue Mountains Eye Study. Ophthalmology. 2003;110(7):1364–1370. 9. Fotedar R, Mitchell P, Burlutsky G, et al. Relationship of 10-Year Change in Refraction to Nuclear Cataract and Axial Length Findings From an Older Population. Ophthalmology. 2008;115(8):1273–1278. 10. Panchapakesan J, Rochtchina E, Mitchell P. Myopic Refractive Shift Caused by Incident Cataract: The Blue Mountains Eye Study. Ophthalmic Epidemiol. 2003;10(4):241–247. 11. Samarawickrama C, Wang JJ, Burlutsky G, et al. Nuclear Cataract and Myopic Shift in Refraction. Am. J. Ophthalmol. 2007;144(3):457–459. 12. Dirani M, Chamberlain M, Shekar SN, et al. Heritability of Refractive Error and Ocular Biometrics: The Genes in Myopia (GEM) Twin Study. Invest. Ophthalmol. Vis. Sci. 2006;47(11):4756–4761. 13. Klein AP, Suktitipat B, Duggal P, et al. Heritability Analysis of Spherical Equivalent, Axial Length, Corneal Curvature, and Anterior Chamber Depth in the Beaver Dam Eye Study. Arch. Ophthalmol. 2009;127(5):649–655. 14. Chen CYC, Scurrah KJ, Stankovich J, et al. Heritability and Shared Environment Estimates for Myopia and Associated Ocular Biometric Traits: The Genes in

120

Myopia (GEM) Family Study. Hum. Genet. 2007;121:511–520. 15. Dirani M, Shekar SN, Baird PN. Evidence of Shared Genes in Refraction and Axial Length: The Genes in Myopia (GEM) Twin Study. Invest. Ophthalmol. Vis. Sci. 2008;49(10):4336–4339. 16. Yonova-Doing E, Forkin ZA, Hysi PG, et al. Genetic and Dietary Factors Influencing the Progression of Nuclear Cataract. Ophthalmology. 2016;123(6):1–8. 17. Hammond CJ, F.R.C.Ophth., Snieder H, et al. Genetic and Environmental Factors in Age-Related Nuclear Cataracts in Monozygotic and Dizygotic Twins. N. Engl. J. Med. 2000;342(24):1786–1790. 18. Congdon N, Broman KW, Lai H, et al. Nuclear Cataract Shows Significant Familial Aggregation in an Older Population after Adjustment for Possible Shared Environmental Factors. Invest Ophthalmol Vis Sci. 2004;45(7):2182–2186. 19. Heiba IM, Elston RC, Klein BEK, et al. Genetic Etiology of Nuclear Cataract: Evidence for a Major Gene. Am. J. Med. Genet. 1993;47(8):1208–1214. 20. Klein AP, Duggal P, Lee KE, et al. Support for Polygenic Infuences on Ocular Refractive Error. Investig. Ophthalmol. Vis. Sci. 2005;46(2):442–446. 21. Klein AP, Duggal P, Lee KE, et al. Polygenic Effects of Cigarette Smoking Account for a Portion of the Familial Aggregation of Nuclear Sclerosis. Am. J. Epidemiol. 2005;161(8):707–713. 22. Ritchie MD, Verma SS, Hall MA, et al. Electronic Medical Records and Genomics (eMERGE) Network Exploration in Cataract: Several New Potential Susceptibility Loci. Mol. Vis. 2014;20:1281–1295. 23. Liao J, Su X, Chen P, et al. Meta-Analysis of Genome-Wide Association Studies in Multiethnic Asians Identifies Two Loci for Age-Related Nuclear Cataract. Hum. Mol. Genet. 2014;23(22):6119–6128. 24. Shi Y, Qu J, Zhang D, et al. Genetic Variants at 13q12.12 are Associated with High Myopia in the Han Chinese Population. Am. J. Hum. Genet. 2011;88(6):805– 813. 25. Han S, Chen P, Fan Q, et al. Association of Variants in FRAP1 and PDGFRA with Corneal Curvature in Asian Populations from Singapore. Hum. Mol. Genet. 2011;20(18):3693–3698. 26. Guggenheim J a, McMahon G, Kemp JP, et al. A Genome-Wide Association Study for Corneal Curvature Identifies the Platelet-Derived Growth Factor Receptor alpha Gene as a Quantitative Trait Locus for Eye Size in White Europeans. Mol. Vis. 2013;19:243–253. 27. Chen P, Miyake M, Fan Q, et al. CMPK1 and RBP3 are Associated with Corneal Curvature in Asian Populations. Hum. Mol. Genet. 2014;23(22):6129–6136.

121

28. Solouki AM, Verhoeven VJM, van Duijn CM, et al. A Genome-Wide Association Study Identifies a Susceptibility Locus for Refractive Errors and Myopia at 15q14. Nat. Genet. 2010;42(10):897–901. 29. Verhoeven VJM, Hysi PG, Saw SM, et al. Large Scale International Replication and Meta-Analysis Study Confirms Association of the 15q14 Locus with Myopia. The CREAM Consortium. Hum. Genet. 2012;131(9):1467–1480. 30. Kiefer AK, Tung JY, Do CB, et al. Genome-Wide Analysis Points to Roles for Extracellular Matrix Remodeling, the Visual Cycle, and Neuronal Development in Myopia. PLoS Genet. 2013;9(2):1–8. 31. Meng W, Butterworth J, Malecaze F, et al. Axial Length of Myopia: A Review of Current Research. Ophthalmologica. 2011;225(3):127–134. 32. Young TL, Metlapally R, Shay AE. Complex Trait Genetics of Refractive Error. Arch. Ophthalmol. 2007;125(1):38–48. 33. Foster PJ, Jiang Y. Epidemiology of Myopia. Eye. 2014;28(2):202–208. 34. Flitcroft DI. Emmetropisation and the Aetiology of Refractive Errors. Eye. 2014;28(2):169–179. 35. Wong TY, Foster PJ, Ng TP, et al. Variations in Ocular Biometry in an Adult Chinese Population in Singapore: The Tanjong Pagar Survey. Invest Ophthalmol Vis Sci. 2001;42(1):73–80. 36. Olsen T, Arnarsson A, Sasaki H, et al. On the Ocular Refractive Components: The Reykjavik Eye Study. Acta Ophthalmol. Scand. 2007;85(4):361–366. 37. Gudmundsdottir E, Arnarsson A, Jonasson F. Five-Year Refractive Changes in an Adult Population: Reykjavik Eye Study. Ophthalmology. 2005;112(4):672–677. 38. Chen F, Duggal P, Klein BEK, et al. Variation in PTCHD2, CRISP3, NAP1L4, FSCB, and AP3B2 Associated with Spherical Equivalent. Mol. Vis. 2016;22:783– 796. 39. Zhou P, Luo Y, Liu X, et al. Down-Regulation and CpG Island Hypermethylation of CRYAA in Age-Related Nuclear Cataract. FASEB J. 2012;26(12):4897–4902. 40. Zhou Z, Wang B, Luo Y, et al. Major Intrinsic Protein (MIP) Polymorphism is Associated with Age-Related Cataract in Chinese. Mol. Vis. 2011;17:2292–2296.

41. Shiels A, Hejtmancik JF. Molecular Genetics of Cataract. In: Progress in Molecular Biology and Translational Science. Elsevier Inc.; 2015:203–218. 42. Shi Y, Shi X, Jin Y, et al. Mutation Screening of HSF4 in 150 Age-Related Cataract Patients. Mol. Vis. 2008;14:1850–1855. 43. Chen JH, Chen H, Huang S, et al. Endophenotyping Reveals Differential Phenotype-Genotype Correlations Between Myopia-Associated Polymorphisms

122

and Eye Biometric Parameters. Mol. Vis. 2012;18:765–778. 44. Fan Q, Barathi V a, Cheng C-Y, et al. Genetic Variants on Chromosome 1q41 Influence Ocular Axial Length and High Myopia. PLoS Genet. 2012;8(6):1–14. 45. Miyake M, Yamashiro K, Tabara Y, et al. Identification of Myopia-Associated WNT7B Polymorphisms Provides Insights into the Mechanism Underlying the Development of Myopia. Nat. Commun. 2015;6(6689):1–9. 46. Schache M, Richardson AJ, Mitchell P, et al. Genetic Association of Refractive Error and Axial Length with 15q14 but not 15q25 in the Blue Mountains Eye Study Cohort. Ophthalmology. 2013;120(2):292–297. 47. Cheng C-Y, Schache M, Ikram MK, et al. Nine Loci for Ocular Axial Length Identified through Genome-Wide Association Studies, Including Shared Loci with Refractive Error. Am. J. Hum. Genet. 2013;93(2):264–277. 48. Guggenheim J a., Zhou X, Evans DM, et al. Coordinated Genetic Scaling of the Human Eye: Shared Determination of Axial Eye Length and Corneal Curvature. Investig. Ophthalmol. Vis. Sci. 2013;54(3):1715–1721. 49. Mishra A, Yazar S, Hewitt AW, et al. Genetic Variants Near PDGFRA Are Associated with Corneal Curvature in Australians. Invest. Ophthalmol. Vis. Sci. 2012;53(11):7131–7136. 50. Richter GM, Wang M, Jiang X, et al. Ocular Determinants of Refractive Error and Its Age- and Sex-Related Variations in the Chinese American Eye Study. JAMA Ophthalmol. 2017;135(7):724–732. 51. Hysi PG, Wojciechowski R, Rahi JS, et al. Genome-Wide Association Studies of Refractive Error and Myopia, Lessons Learned, and Implications for the Future. Investig. Opthalmology Vis. Sci. 2014;55(5):3344–3351. 52. National Eye Institute. Fact Sheet: Refractive Errors. 2010;(https://nei.nih.gov/sites/default/files/health- pdfs/HVM09_Fact_Sheet_Final_tagged.pdf). (Accessed November 9, 2015) 53. Centers for Disease Control and Prevention. Refractive Errors. Vis. Heal. Initiat. 2013;(http://www.cdc.gov/visionhealth/basic_information/eye_disorders.htm#a2). (Accessed November 9, 2015) 54. He W, Goodkind D, Kowal P. An Aging World : 2015. Washington, DC: 2016. 55. Saw SM, Gazzard G, Shin-Yen EC, et al. Myopia and Associated Pathological Complications. Ophthalmic Physiol. Opt. 2005;25:381–391. 56. Wojciechowski R. Nature and Nurture: The Complex Genetics of Myopia and Refractive Error. Clin. Genet. 2011;79(4):301–320.

123

APPENDIX

124

Methods

Quality Control and Imputation for the Second Beaver Dam Eye Study (BDES)

Group

Recruitment of individuals for the second genotyped BDES group has been described previously (1). Briefly, a nested subset of 1,816 individuals from the BDES was genotyped as part of the International Age-related Macular Degeneration (AMD)

Genomic Consortium (1). These individuals were selected largely based on not having been genotyped in the first BDES group (with the exception of a small group of individuals who were genotyped in both cohorts for quality control purposes). Data were collected according to the Declaration of Helsinki and all protocols were approved by local ethics committees. All study participants provided informed consent (1). Genotypes from 1,816 BDES individuals were generated for 569,645 variants using a customized

Illumina HumanCoreExome array (“HumanCoreExome_Goncalo_15038949_A”

Illumina, Illumina, Inc., San Diego, CA). The customized exome array included 302,188 variants from HumanCore content (genome content), 250,494 variants from Exome1.0 content (exome content), and 29,858 custom variants (1). Individuals were genotyped centrally at the Center for Inherited Disease Research at Johns Hopkins University

School of Medicine and genotype calling was performed using the same clustering algorithm as the first BDES genotyped group.

Samples with call rates below 98% (n = 23), blind duplicates (n = 19), sex mismatches (n = 17), Mendelian inconsistencies (n = 0), unexpected duplicates (n =3), or data in both BDES groups (n = 87) were excluded. Individuals of non-European ancestry

(n = 26) were identified using samples from the Human Genome Diversity Project (1,2)

125 and also removed. Non-autosomal variants (n = 16,430), variants with call rate < 98% (n

= 81,250), monomorphic variants (n = 164,127), variants out of Hardy-Weinberg

Equilibrium (P < 1.0 × 10-6) according to the Hardy-Weinberg exact test (n = 6) (3), indels (n = 1,280), and duplicate variants (n = 5,867) were excluded prior to imputation.

The 1,641 individuals and 300,685 variants that passed quality control were imputed to the Haplotype Reference Consortium reference panel using SHAPEIT version 2 for phasing and Minimac version 3 for imputation on the Michigan Imputation Server (4–7).

Samples for Axial Length, Corneal Curvature, and Refraction for Second BDES

Group

The association analysis for the second BDES group was restricted to variants with an imputation R2 ≥ 0.6 to remain consistent with methods used in the first BDES group. Related individuals (r2 > 0.20), individuals with missing phenotype or covariate measures, and individuals with large differences in ocular measures between the two eyes

(> 3 SD from the mean for axial length and corneal curvature, > ±4 D for spherical equivalent) were removed prior to analysis. Final sample sizes for axial length, corneal curvature, and spherical equivalent were 568, 566, and 1,263 individuals, respectively.

All three phenotypes showed a normal distribution before and after adjustment for covariates.

Meta-Analysis of BDES Groups

Imputed files for the second BDES dataset were converted to dose files using

DosageConverter (DosageConverter, University of Michigan, Ann Arbor, MI), and a single variant analysis was conducted as described previously for the first BDES group in

Mach2qtl (8,9). Variants in the second BDES group with minor allele frequency (MAF)

126

< 0.88% for axial length and corneal curvature or MAF < 0.40% for spherical equivalent

(which translates to a minor allele count < 10 in each sample) were discarded (10–12).

Meta-analysis of the results from the two imputed BDES groups was conducted in

METAL, with weights based on sample size (13). The standard genome-wide significance threshold of P < 5.0 × 10-8 was applied (14,15).

Results

Single Variant Analysis

The genomic inflation factor (λ) for the axial length, corneal curvature, and spherical equivalent samples were 1.02 (right eye), 0.99 (right eye), and 1.02, respectively (Supplementary Figure S2). Results between the right and left eyes were similar, so the results of the right eye are presented. No variants were significantly associated with any of the phenotypes in our study. A rare missense variant (rs3544090, g.68709884G>A, MAF = 1.03%) in the RAD17 checkpoint clamp loader component gene (MIM 603139) on chromosome 5q13.2 was suggestively associated with axial length (Supplementary Table S4). This variant, which substitutes an arginine for a lysine

(p.Arg593Lys), led to a 1.24 mm increase (95% CI = [0.68, 1.78], p = 1.05 × 10-5) in axial length among carriers compared to non-carriers of the minor allele.

Multiple intronic variants within the glycoprotein M6A (GPM6A) gene (MIM

601275) on 4q22 were suggestively associated with corneal curvature. The most strongly associated intronic variant (rs80258802, g.176859667T>A, MAF = 3.96%) led to a 0.19 mm (95% CI = [-0.23, -0.15], p = 4.42 × 10-6) decrease in corneal curvature with each

127 additional copy of the minor allele (Supplementary Table S5). The other variants in the

GPM6A gene had similar effects.

Two intronic variants in the vacuolar protein sorting 18 homolog (VPS18) gene

(MIM 608551) on chromosome 15q21.1 were suggestively associated with spherical equivalent (Supplementary Table S6). The variant with the stronger association

(rs12591803, g.41191095G>A, MAF = 0.35%) led to a 4.47 D (95% CI = [3.53, 5.42], p

= 2.31 × 10-6) increase in spherical equivalent with each additional copy of the minor allele, making individuals with the minor allele more hyperopic compared to those without the minor allele. The effect of the other variant in the VPS18 gene was similar.

Gene-Based Analysis

No genes with cumulative MAF > 0.01 were significantly associated with any phenotype in this study. However, the nucleosome assembly protein 1-like 4 (NAP1L4) gene (MIM 601651) on chromosome 11p15.4 was significantly associated with spherical equivalent (CMAF = 0.10%, p = 1.88 × 10-6). Two rare missense variants contributed to this association, both of which were genotyped on the exome array (Supplementary Table

S7). This result is consistent with the results of a previously published gene-based analysis conducted on spherical equivalent using the non-imputed exome array from the

BDES (10). This gene encodes a member of the nucleosome assembly protein (NAP) family (16) and functions as a histone chaperone that facilitates nucleosome formation by mediating histone dimerization (16,17). NAP1L4 is located near a known tumor- suppressor gene region on chromosome 11p15.5 that has been associated with diseases including Beckwith-Wiedemann syndrome and Wilms Tumor Susceptibility-5 (16,17).

The role of NAP1L4 in refraction or any related ocular process is unclear. However, this

128 gene contained only two variants with a CMAF << 1% in our study and is therefore interpreted with caution.

The cytohesin 1-interacting protein (CYTIP) gene (MIM 604448) on chromosome

2q24.1, which contained three rare variants (two missense and one intronic) in the gene- based analysis, was significantly associated with corneal curvature (Supplementary Table

S7; CMAF = 0.23%, P = 1.43 × 10-6). The intronic variant was imputed and the two missense variants were genotyped on the exome array. The CYTIP gene encodes a protein that binds to cytohesin-1 (CYTH1) and modulates the activation of ADP-ribosylation factor (ARF) genes by CYTH1. The purpose of this gene is allegedly to sequester

CYTH1 in the cytoplasm, which is expressed in natural killer and peripheral T cells and regulates adhesiveness of integrins at the plasma membrane of lymphocytes (16,17). Not surprisingly, the CYTIP gene is found in lymph nodes, peripheral blood leukocytes, bone marrow, and other immune-related tissues. However, there is no evidence currently available to directly support ocular expression or function (18,19). This gene was only suggestively associated with corneal curvature in a similar analysis conducted using the non-imputed exome array data (CMAF = 0.15%, p = 5.81 × 10-6), in which only the two missense variants contributed to the analysis. This indicates that the presence of the third, imputed variant in this gene allowed us to detect a significant association we would have otherwise considered suggestive. However, these results should be interpreted with caution due to the low frequency of the rare variants in this gene.

No other genes were found to be significantly associated with any of the three phenotypes. Suggestively associated genes (P < 1.0 × 10-4) for corneal curvature and spherical equivalent are reported in Supplementary Table S7. For corneal curvature,

129 results from the left and right eyes were similar, so only the results of the right eye are presented except for the results for the CYTIP gene. The p-value for this gene narrowly exceeded the threshold for significance in the left eye but not the right, so the results of the left eye are reported for this gene only (Supplementary Table S7). No genes with more than one eligible variant were suggestively associated with axial length in this study.

Meta-Analysis of BDES Groups

No significant associations were detected in the meta-analysis of the two BDES groups. Suggestively associated variants for each phenotype are provided in

Supplementary Table S8 (P < 1.0 × 10-4). Results of the right and left eyes were similar, so the results of the right eye are presented. The lack of significant associations could be due to differences in how each group was sampled. Individuals were recruited for the first

BDES group if they had extreme values of either spherical equivalent or intraocular pressure. Recruitment for the second BDES group was not selective. Differences were particularly evident between the two genotyped groups when comparing spherical equivalent distributions (Supplementary Table S3). Although small, significant differences in axial length between the groups were also observed (Supplementary Table

S1). As indicated in Supplementary Tables S1 and S3, the distributions of axial length and spherical equivalent in particular were wider in the first BDES group. This is supported by Supplementary Figure S3, in which the distributions of each ocular phenotype are presented from each group as overlapping histograms. The distributions showed significant overlap, but the distribution of the first BDES group was wider than the second. These differences remained following adjustment for relevant covariates.

130

Although single variant analysis results between these two groups were consistent for some common variants, including those reported in Supplementary Table S8, the variants most strongly associated with these phenotypes in our single variant analysis of the first BDES group were not strongly associated in the second BDES group. This includes rare variants and low frequency variants. This indicates that the individuals occupying the tail ends of the distributions of these phenotypes in the first BDES group may be driving the associations of these select variants. The narrower distribution and thus reduced variability in the second BDES group lacks individuals with more extreme values of the phenotypes, which may be the reason why the strongly associated variants in the first BDES group are not as strongly associated in the second group.

To determine whether other factor(s) may have contributed to differences in distributions and results between the two BDES groups, we conducted sensitivity analyses using axial length. Axial length was used because one factor we had interest in evaluating was cataract surgery, and spherical equivalent measures were no longer recorded in the BDES for individuals once they had cataract surgery. We conducted sensitivity analyses by (1) excluding those with cataract surgery, and (2) by excluding those with early or late AMD from each group, then comparing the results of the most strongly associated variants in the first BDES group with results of the same variants in the second group. We also removed three individuals who were members of families that showed a high LOD score (LOD > 1.0) in a previous unpublished linkage analysis for spherical equivalent within the BDES from the first group and compared results.

P-values for the most strongly associated variants in the first BDES group with axial length remained small and not comparable to p-values for these variants in the

131 second BDES group among those without cataract, among those without early or late

AMD, or without any family members with high LOD scores (data not shown). Removal of select individuals in each case did not significantly alter the results beyond lowering the p-values slightly (likely due to a reduction in sample size). Differences in adjusted mean axial length between the groups remained significant following removal of those with cataract, AMD, or family members, indicating that removal of these individuals did not significantly alter the distribution of axial length. Similar results were observed when comparing distributions of spherical equivalent after removing those with early or late

AMD from both groups, or removing family members with high LOD scores from the first BDES group. These factors therefore do not appear to contribute significantly to the differences we observe between the two BDES groups.

132

References

1. Fritsche LG, Igl W, Cooke Bailey JN, et al. A Large Genome-Wide Association Study of Age-Related Macular Degeneration Highlights Contributions of Rare and Common Variants. Nat. Genet. 2016;48(2):134–143. 2. Cavalli-sforza LL. The Human Genome Diversity Project: Past , Present and Future. Nat. Genet. 2005;6:333–340. 3. Wigginton JE, Cutler DJ, Abecasis GR. A Note on Exact Tests of Hardy-Weinberg Equilibrium. Am. J. Hum. Genet. 2005;76(5):887–893. 4. Delaneau O, Marchini J, Zagury J-F. A Linear Complexity Phasing Method for Thousands of Genomes. Nat. Methods. 2012;9(2):179–181. 5. Delaneau O, Zagury J, Marchini J. Improved Whole-Chromosome Phasing for Disease and Population Genetic Studies. Nat. Methods. 2013;10(1):5–6. 6. Das S, Forer L, Schönherr S, et al. Next-Generation Genotype Imputation Service and Methods. Nat. Genet. 2016;48(10):1284–1287. 7. McCarthy S, Das S, Kretzschmar W, et al. A Reference Panel of 64,976 Haplotypes for Genotype Imputation. Nat. Genet. 2016;1–5. 8. Li Y, Willer CJ, Ding J, et al. MaCH : Using Sequence and Genotype Data to Estimate Haplotypes and Unobserved Genotypes. Genet. Epidemiol. 2010;34:816– 834. 9. Pei Y, Zhang L, Li J, et al. Analyses and Comparison of Imputation-Based Association Methods. PLoS One. 2010;5(5):1–10. 10. Chen F, Duggal P, Klein BEK, et al. Variation in PTCHD2, CRISP3, NAP1L4, FSCB, and AP3B2 Associated with Spherical Equivalent. Mol. Vis. 2016;22:783– 796. 11. Chen F, Klein AP, Klein BEK, et al. Exome Array Analysis Identifies CAV1/CAV2 as a Susceptibility Locus for Intraocular Pressure. Invest. Ophthalmol. Vis. Sci. 2014;56(1):544–551. 12. Soderholm M, Almgren P, Jood K, et al. Exome Array Analysis of Ischaemic Stroke: Results from a Southern Swedish Study. Eur. J. Neurol. 2016;23:1722– 1728. 13. Willer CJ, Li Y, Abecasis GR, et al. METAL: Fast and Efficient Meta-Analysis of Genomewide Association Scans. Bioinformatics. 2010;26(17):2190–2191. 14. Consortium TIH. The International HapMap Project. Nature. 2003;426:789–796. 15. Pe’er I, Yelensky R, Altshuler D, et al. Estimation of the Multiple Testing Burden for Genomewide Association Studies of Nearly All Common Variants. Genet.

133

Epidemiol. 2008;32:381–385. 16. Online Mendelian Inheritance in Man. 2017;(http://www.omim.org/). (Accessed July 5, 2017) 17. GeneCards Human Genome Database. 2017;(http://www.genecards.org/). (Accessed July 5, 2017) 18. Consortium U. UniProt. 2017;(www.uniprot.org). (Accessed July 5, 2017) 19. Qian J. Tissue-Specific Gene Expression and Regulation. 2008;(http://bioinfo.wilmer.jhu.edu/tiger/). (Accessed July 6, 2017) 20. Cheng C-Y, Schache M, Ikram MK, et al. Nine Loci for Ocular Axial Length Identified through Genome-Wide Association Studies, Including Shared Loci with Refractive Error. Am. J. Hum. Genet. 2013;93(2):264–277. 21. Fan Q, Barathi V a, Cheng C-Y, et al. Genetic Variants on Chromosome 1q41 Influence Ocular Axial Length and High Myopia. PLoS Genet. 2012;8(6):1–14.

134

Supplementary Tables and Figures Supplementary Table S1. Characteristics of the Beaver Dam Eye Study Participants Contributing to the Axial Length Analyses Characteristica First BDES Group Second BDES Group P-Valueb BDES Full (N = 874) (N = 568) (N = 1,946)c Age (years), mean ± SD (range) 70.1 ± 8.2 (58-94) 71.3 ± 8.1 (58-94) 0.006 70.3 ± 7.9 (58-96) Female Gender, N (%) 526 (60.2) 316 (55.6) 0.087 1,116 (57.3) Education (years), mean ± SD (range) 13.2 ± 2.8 (5-25) 12.7 ± 2.6 (6-24) 0.001 12.9 ± 2.6 (5-25) Height (cm), mean ± SD (range) 167.4 ± 9.1 168.0 ± 9.1 0.221 167.5 ± 9.1 (143.0-208.3) (147.3-196.9) (137.2-208.3) Axial Length of Right Eye (mm), 23.8 ± 1.1 23.5 ± 0.8 <0.001 23.7 ± 1.1 mean ± SD (range) (20.5-28.4) (21.3-26.9) (13.8-37.4) Axial Length of Left Eye (mm), 23.8 ± 1.2 23.5 ± 0.8 <0.001 23.7 ± 1.2 mean ± SD (range) (20.7-28.2) (21.4-27.4) (17.0-39.5) Abbreviations: cm, centimeters; mm, millimeters; SD, standard deviation. aAll measurements were taken from visit 4 except height, which was measured at baseline. bP-value from a 2 sample t-test (or z-test for proportions) comparing the equivalence of means (or proportions) in the two BDES groups. cAll BDES individuals with reliable measures for axial length in both eyes, age, sex, and education at visit 4.

135

Supplementary Table S2. Characteristics of the Beaver Dam Eye Study Participants Contributing to the Corneal Curvature Analyses b First BDES Group Second BDES Group P-Value BDES Full Characteristica (N = 883) (N = 566) (N = 1,941)c 70.1 ± 8.1 71.3 ± 8.1 0.006 70.3 ± 8.0 Age (years), mean ± SD (range) (58-94) (58-94) (58-96) Female Gender, N (%) 536 (60.7) 316 (55.8) 0.066 1112 (57.3) 167.4 ± 9.1 168.0 ± 9.1 0.221 167.6 ± 9.1 Height (cm), mean ± SD (range) (143.0-208.3) (147.3-196.9) (137.2-208.3) 7.7 ± 0.3 7.7 ± 0.3 0.999 7.7 ± 0.3 Corneal Curvature of Right Eye (mm), mean ± SD (range) (7.0-8.6) (7.0-8.7) (6.3-8.7) 7.7 ± 0.3 7.7 ± 0.3 0.999 7.7 ± 0.3 Corneal Curvature of Left Eye (mm), mean ± SD (range) (6.9-8.7) (7.0-8.7) (6.9-8.7) Abbreviations: cm, centimeters; mm, millimeters; SD, standard deviation. aAll measurements were taken from visit 4 except height, which was measured at baseline. bP-value from a 2 sample t-test (or z-test for proportions) comparing the equivalence of means (or proportions) in the two BDES groups. cAll BDES individuals with reliable measures for corneal curvature in both eyes, age, sex, and height at visit 4.

136

Supplementary Table S3. Characteristics of the Beaver Dam Eye Study Participants Contributing to the Spherical Equivalent Analyses Characteristica First BDES Group Second BDES P-Valueb BDES Full c (N = 1,552) Group (N = 4,299) (N = 1,263) Age (years), mean ± SD (range) 59.3 ± 10.5 61.5 ± 10.3 <0.001 60.9 ± 10.7 (43-85) (43-86) (43-86) Female Gender, N (%) 896 (57.7) 669 (53.0) 0.011 2,397 (55.8) Education (years), mean ± SD (range) 12.7 ± 3.0 (2-26) 12.1 ± 2.7 (3-24) <0.001 12.1 ± 2.8 (0-26) Nuclear Sclerosis of Right Eye, mean ± SD (range) 2.3 ± 0.9 2.4 ± 0.9 0.003 2.0 ± 0.9 (1.0-5.0) (1.0-5.0) (1.0-5.0) Nuclear Sclerosis of Left Eye, mean ± SD (range) 2.3 ± 0.9 2.4 ± 0.9 0.003 2.0 ± 0.9 (1.0-5.0) (1.0-5.0) (1.0-5.0) Nuclear Sclerosis, mean ± SD (range) 4.7 ± 1.6 4.9 ± 1.6 0.001 4.8 ± 1.6 (2.0-9.0) (2.0-10.0) (2.0-10.0) Spherical Equivalent of Right Eye (D), mean ± SD (range) -0.3 ± 3.0 0.55 ± 1.2 <0.001 0.21 ± 2.3 (-13.5-11.0) (-4.5-4.5) (-13.0-9.5) Spherical Equivalent of Left Eye (D), mean ± SD (range) -0.2 ± 3.0 0.61 ± 1.1 <0.001 0.28 ± 2.3 (-13.5-11.0) (-3.3-5.0) (-13.5-11.0) Spherical Equivalent (D), mean ± SD (range) -0.2 ± 2.9 0.58 ± 1.1 <0.001 0.24 ± 2.3 (-13.0-10.3) (-3.9-3.9) (-13.0-10.3) Abbreviations: cm, centimeters; mm, millimeters; SD, standard deviation. aAll measurements were taken from the baseline visit. bP-value from a 2 sample t-test (or z-test for proportions) comparing the equivalence of means (or proportions) in the two BDES groups. cAll BDES individuals with reliable measures for spherical equivalent in both eyes, age, sex, education and nuclear sclerosis at baseline.

137

Supplementary Table S4. Top Single Variant Association Analysis Results for Axial Length MAF in Gene Chr:Position dbSNP DNA (Amino BDES HWE P- βc SE P-valued [MIM] (Base-Pair)a rsID Acid) Alterationb [Minor Valuee Allele] RAD17 g. 68709884G>A 5:68709884 35440980 0.0103 [A] 1.24 0.280 1.05E-05 1.00 [603139] (p.Arg593Lys) g. 33723383T>C Intergenic 6:33723383 1536500 0.2042 [T] -0.28 0.065 1.79E-05 0.04 (N/A) g. 33721933C>T Intergenic 6:33721933 9461905 0.2042 [T] -0.28 0.065 1.97E-05 0.04 (N/A) g.33726990C>A Intergenic 6:33726990 9461907 0.2037 [A] -0.28 0.065 2.15E-05 0.04 (N/A) g.33727614C>A Intergenic 6:33727614 62398943 0.1974 [A] -0.28 0.066 2.40E-05 0.02 (N/A) g.33727885C>T Intergenic 6:33727885 1536501 0.1974 [T] -0.28 0.066 2.40E-05 0.02 (N/A) g.29538188T>G Intergenic 12:29538188 140931349 0.0669 [G] -0.52 0.124 2.56E-05 1.00 (N/A) g.29539479G>A Intergenic 12:29539479 117870308 0.0681 [A] -0.52 0.123 2.64E-05 1.00 (N/A) SERPINB11 g.61389943G>T 18:61389943 9961025 0.0601 [T] -0.48 0.115 3.14E-05 0.83 [615682] (Intronic) g.33719877C>T Intergenic 6:33719877 943463 0.2031 [T] -0.27 0.065 3.38E-05 0.04 (N/A) g.33721457C>T Intergenic 6:33721457 1998949 0.2065 [T] -0.27 0.065 3.46E-05 0.06 (N/A) g.61400127T>C Intergenic 18:61400127 4510126 0.0584 [C] -0.51 0.123 3.70E-05 0.66 (N/A) g.61397708C>T Intergenic 18:61397708 142447836 0.0584 [T] -0.50 0.122 3.79E-05 0.66 (N/A) g.33729718C>G Intergenic 6:33729718 11751943 0.1951 [G] -0.28 0.067 3.87E-05 0.00 (N/A)

138

g.61391738T>C Intergenic 18:61391738 17071630 0.0584 [C] -0.49 0.119 3.92E-05 0.66 (N/A) TACC3 g.1724789C>T 4:1724789 62285103 0.1791 [T] -0.37 0.091 4.31E-05 0.30 [605303] (Intronic) MROH9 g.170961360G>A 1:170961360 189860161 0.0155 [A] -0.85 0.209 4.53E-05 1.00 [N/A] (p.Val362Met) OVCH1-AS1 g.29547807T>C 12:29547807 76670607 0.0755 [C] -0.44 0.109 4.69E-05 0.75 [N/A] (Intronic) TMEM129 g.1721790G>A 4:1721790 17132038 0.1791 [A] -0.40 0.098 4.80E-05 0.30 [615975] (Intronic) OVCH1-AS1 g.29551576A>G 12:29551576 76588580 0.0755 [G] -0.43 0.107 5.03E-05 0.75 [N/A] (Intronic) IP6K3 g.33713779C>T 6:33713779 602399 0.1974 [T] 0.27 0.067 5.15E-05 0.01 [606993] (Intronic) OVCH1,OVCH1 g.29580751A>G -AS1 12:29580751 7301354 0.1350 [G] -0.31 0.077 5.63E-05 0.84 (Intronic) [N/A] IP6K3 g.33711420C>T 6:33711420 652049 0.1974 [T] 0.27 0.068 5.70E-05 0.01 [606993] (Intronic) IP6K3 g.33710229C>T 6:33710229 570749 0.1974 [T] 0.27 0.068 5.87E-05 0.01 [606993] (Intronic) IP6K3 g.33710993G>A 6:33710933 487835 0.1974 [A] 0.27 0.068 5.90E-05 0.01 [606993] (Intronic) g.106403951T>C none 3:106403951 7622407 0.4737 [C] 0.22 0.054 5.97E-05 0.07 (N/A) IP6K3 g.33709316G>A 6:33709316 791904 0.1979 [A] 0.27 0.068 6.22E-05 0.01 [606993] (Intronic) g.33709316G>A Intergenic 6:33732987 4996879 0.2037 [A] -0.27 0.066 6.30E-05 0.01 (N/A) g.71206914G>T Intergenic 4:71206194 72653362 0.1121 [T] 0.37 0.093 6.38E-05 0.15 (N/A) g.71206061C>T Intergenic 4:71206061 10022704 0.1121 [T] 0.37 0.093 6.39E-05 0.15 (N/A)

139

OVCH1-AS1 g.29570774C>A 12:29570774 1035608 0.1350 [A] -0.31 0.078 6.41E-05 0.84 [N/A] (Intronic) PTPDC1 g.96860820G>A 9:96860820 61755091 0.0109 [A] -1.00 0.250 6.42E-05 1.00 [N/A] (p.Val604Ile) OVCH1-AS1 g.29574103C>T 12:29574103 34137040 0.1350 [T] -0.31 0.078 6.43E-05 0.84 [N/A] (Intronic) g.71205477A>G Intergenic 4:71205477 10031338 0.1121 [G] 0.38 0.094 6.43E-05 0.15 (N/A) g.71205630A>G Intergenic 4:71205630 10031477 0.1121 [G] 0.37 0.093 6.44E-05 0.15 (N/A) g.71205692T>C Intergenic 4:71205692 10033835 0.1121 [C] 0.37 0.093 6.45E-05 0.15 (N/A) CABS1 g.71201943A>G 4:71201943 13039 0.1127 [G] 0.38 0.094 6.45E-05 0.15 [N/A] (STOP) OVCH1-AS1 g.29573986T>C 12:29573986 11831164 0.1350 [C] -0.31 0.078 6.48E-05 0.84 [N/A] (Intronic) OVCH1-AS1 g.29574756G>A 12:29574756 11829375 0.1350 [A] -0.31 0.078 6.51E-05 0.84 [N/A] (Intronic) g.106404147A>C Intergenic 3:106404147 11928487 0.4737 [C] 0.21 0.053 6.51E-05 0.08 (N/A) g.71206397T>C Intergenic 4:71206397 78523196 0.1121 [C] 0.37 0.093 6.52E-05 0.15 (N/A) OVCH1-AS1 g.29571618G>C 12:29571618 11614744 0.1350 [C] -0.31 0.078 6.52E-05 0.84 [N/A] (Intronic) OVCH1-AS1 g.29572985A>G 12:29572985 7297659 0.1350 [G] -0.31 0.078 6.53E-05 0.84 [N/A] (Intronic) g.71205482G>A Intergenic 4:71205482 2332002 0.1121 [A] 0.38 0.094 6.55E-05 0.15 (N/A) OVCH1-AS1 g.29566537T>A 12:29566537 77743180 0.0755 [A] -0.41 0.104 6.55E-05 0.75 [N/A] (Intronic) g.71205495T>C Intergenic 4:71205495 1871712 0.1121 [C] 0.38 0.094 6.56E-05 0.15 (N/A) g.71206382T>C Intergenic 4:71206382 72653364 0.1121 [C] 0.37 0.093 6.56E-05 0.15 (N/A)

140

OVCH1-AS1 g.29576517C>T 12:29576517 2042512 0.1350 [T] -0.31 0.078 6.57E-05 0.84 [N/A] (Intronic) g.71203806G>A Intergenic 4:71203806 9990693 0.1127 [A] 0.37 0.094 6.57E-05 0.15 (N/A) g.71204662A>G Intergenic 4:71204662 12649847 0.1127 [G] 0.37 0.093 6.57E-05 0.15 (N/A) g.71203303T>C Intergenic 4:71203303 6841488 0.1127 [C] 0.37 0.094 6.59E-05 0.15 (Intronic) OVCH1-AS1 g.29567981G>A 12:29567981 56808425 0.1350 [A] -0.32 0.079 6.59E-05 0.84 [N/A] (Intronic) g.71203397G>A Intergenic 4:71203397 6858667 0.1127 [A] 0.37 0.094 6.60E-05 0.15 (N/A) g.71206651T>A Intergenic 4:71206651 9992266 0.1121 [A] 0.37 0.093 6.60E-05 0.15 (N/A) g.71205183G>A Intergenic 4:71205183 9996508 0.1127 [A] 0.37 0.093 6.60E-05 0.15 (N/A) OVCH1-AS1 g.29578048C>T 12:29578048 11610605 0.1350 [T] -0.31 0.077 6.62E-05 0.84 [N/A] (Intronic) g.71205214C>T Intergenic 4:71205214 922661 0.1127 [T] 0.37 0.093 6.63E-05 0.15 (N/A) OVCH1-AS1 g.29577707G>A 12:29577707 73276833 0.1350 [A] -0.31 0.077 6.64E-05 0.84 [N/A] (Intronic) OVCH1-AS1 g.29573833G>C 12:29573833 11838057 0.1350 [C] -0.31 0.079 6.66E-05 0.84 [N/A] (Intronic) OVCH1,OVCH1 g.29581702G>C -AS1 12:29581702 7978026 0.1350 [C] -0.31 0.077 6.71E-05 0.84 (Intronic) [N/A] g.33738442C>T Intergenic 6:33738442 943472 0.1962 [T] -0.27 0.068 6.72E-05 0.00 (N/A) OVCH1,OVCH1 g.29581776A>G -AS1 12:29581776 7977826 0.1350 [G] -0.31 0.077 6.72E-05 0.84 (Intronic) [N/A] OVCH1,OVCH1 g.29585875G>A 12:29585875 11833098 0.1350 [A] -0.30 0.076 6.77E-05 0.84 -AS1 (Intronic)

141

[N/A] OVCH1-AS1 g.29577767G>A 12:29577767 10219521 0.1350 [A] -0.31 0.078 6.79E-05 0.84 [N/A] (Intronic) OVCH1-AS1 g.29577238A>G 12:29577238 7954043 0.1350 [G] -0.31 0.078 6.80E-05 0.84 [N/A] (Intronic) OVCH1,OVCH1 g.29586116G>T -AS1 12:29586116 73278706 0.1350 [T] -0.30 0.076 6.83E-05 0.84 (Intronic) [N/A] OVCH1,OVCH1 g.29585181A>G -AS1 12:29585181 1369899 0.1350 [G] -0.30 0.076 6.83E-05 0.84 (Intronic) [N/A] OVCH1,OVCH1 g.29585485T>C -AS1 12:29585485 73276902 0.1350 [C] -0.30 0.076 6.85E-05 0.84 (Intronic) [N/A] OVCH1-AS1 g.29578160G>A 12:29578160 10219541 0.1350 [A] -0.31 0.078 6.87E-05 0.84 [N/A] (Intronic) OVCH1-AS1 g.29568380C>G 12:29568380 76344495 0.0755 [G] -0.41 0.103 6.96E-05 0.75 [N/A] (Intronic) OVCH1-AS1 g.29568401C>A 12:29568401 73276822 0.0755 [A] -0.41 0.103 6.97E-05 0.75 [N/A] (Intronic) OVCH1,OVCH1 g.29584609G>A -AS1 12:29584609 7295922 0.1350 [A] -0.31 0.077 6.99E-05 0.84 (Intronic) [N/A] IP6K3 g.33700376T>G 6:33700376 688209 0.1939 [G] 0.27 0.068 7.12E-05 0.01 [606993] (Intronic) ASPN,CENPP g.95228663C>T [608135, 9:95228663 41278695 0.0195 [T] -0.74 0.187 7.14E-05 1.00 (p.Gly193Glu) 611505] LEMD2 g.33739960G>A 6:33739960 10947436 0.1962 [A] -0.27 0.068 7.29E-05 0.00 [616312] (3' UTR) OVCH1,OVCH1 g.29593220G>T -AS1 12:29593220 11050231 0.1350 [T] -0.30 0.076 7.34E-05 0.84 (Intronic) [N/A]

142

OVCH1-AS1 g.29572954C>T 12:29572954 73276826 0.0755 [T] -0.41 0.102 7.34E-05 0.75 [N/A] (Intronic) OVCH1,OVCH1 g.29591184A>C -AS1 12:29591184 7302425 0.1350 [C] -0.30 0.076 7.45E-05 0.84 (Intronic) [N/A] g.30915386G>A Intergenic 2:30915386 530076 0.4983 [A] 0.22 0.056 7.54E-05 0.85 (N/A) CENPP g.95200623A>G 9:95200623 138565395 0.0195 [G] -0.79 0.199 7.60E-05 1.00 [611505] (Intronic) OVCH1-AS1 g.29575453G>C 12:29575453 11829966 0.0755 [C] -0.40 0.102 7.60E-05 0.75 [N/A] (Intronic) g.33731469T>C Intergenic 6:33731469 943464 0.2122 [C] -0.26 0.066 7.74E-05 0.00 (N/A) OVCH1-AS1 g.29577620C>T 12:29577620 73276831 0.0755 [T] -0.40 0.101 7.91E-05 0.75 [N/A] (Intronic) OVCH1-AS1 g.29564564C>T 12:29564564 11835469 0.0755 [T] -0.41 0.104 7.93E-05 0.75 [N/A] (Intronic) g.30914717C>T Intergenic 2:30914717 544923 0.4983 [T] 0.22 0.055 7.96E-05 0.85 (N/A) OVCH1-AS1 g.29565110A>G 12:29565110 114766844 0.0755 [G] -0.42 0.106 8.19E-05 0.75 [N/A] (Intronic) g.30917949C>G Intergenic 2:30917949 490259 0.4983 [G] 0.21 0.053 8.24E-05 0.82 (N/A) g.30920548G>T Intergenic 2:30920548 572801 0.4983 [T] 0.21 0.053 8.44E-05 0.82 (N/A) g.30920293A>T Intergenic 2:30920293 489249 0.4983 [T] 0.21 0.053 8.47E-05 0.82 (N/A) OVCH1-AS1 g.29566637G>A 12:29566637 73276815 0.0755 [A] -0.41 0.104 8.48E-05 0.75 [N/A] (Intronic) g.30930165C>T Intergenic 2:30930165 472161 0.4983 [T] 0.20 0.052 8.50E-05 0.82 (N/A) LEMD2 g.33743174C>T 6:33743174 2395447 0.1962 [T] -0.27 0.067 8.54E-05 0.00 [616312] (Intronic)

143

OVCH1-AS1 g.29567278A>C 12:29567278 73276818 0.0755 [C] -0.41 0.103 8.67E-05 0.75 [N/A] (Intronic) OVCH1,OVCH1 g.29584845C>A -AS1 12:29584845 1436313 0.0755 [A] -0.39 0.100 8.85E-05 0.75 (Intronic) [N/A] g.30931045G>A Intergenic 2:30931045 575493 0.4983 [A] 0.20 0.052 8.99E-05 0.82 (N/A) OVCH1,OVCH1 g.29598616G>C -AS1 12:29598616 10161540 0.1350 [C] -0.30 0.076 9.04E-05 0.84 (Intronic) [N/A] g.71206626A>G Intergenic 4:71206626 10034512 0.1241 [G] 0.36 0.092 9.31E-05 0.22 (N/A) BACH2 g.90902266T>C 6:90902266 2501714 0.3730 [C] 0.26 0.067 9.35E-05 0.20 [605394] (Intronic) g.30922184G>C Intergenic 2:30922184 543552 0.4983 [C] 0.21 0.053 9.36E-05 0.82 (N/A) OVCH1,OVCH1 g.29589719T>G -AS1 12:29589719 137975967 0.0755 [G] -0.38 0.098 9.54E-05 0.75 (Intronic) [N/A] OVCH1,OVCH1 g.29584846G>A -AS1 12:29584846 1436312 0.0755 [A] -0.39 0.100 9.56E-05 0.75 (Intronic) [N/A] OVCH1,OVCH1 g.29596365T>C -AS1 12:29596365 78265994 0.0755 [C] -0.38 0.097 9.95E-05 0.75 (Intronic) [N/A] Abbreviations: Chr, chromosome; dbSNP, Single Nucleotide Polymorphism Database; HWE, Hardy-Weinberg Equilibrium; MAF, minor allele frequency; MIM, Mendelian Inheritance in Man; SE, standard error. aChr:Position reported in GRCh37. bAmino acid alteration reported in dbSNP. cAxial Length measurements were adjusted for age, sex, and education. dBonferroni-corrected genome-wide significance threshold = P < 1.09 × 10-6. Threshold for suggestive significance = P < 1 × 10-4. eHWE exact test used.

144

Supplementary Table S5. Top Single Variant Association Analysis Results for Corneal Curvature Gene Chr:Position DNA (Amino Acid) MAF in BDES HWE P- dbSNP rsID βc SE P-valued [MIM] (Base-Pair)a Alterationb [Minor Allele] Valuee GPM6A g.176859667T>A 4:176859667 80258802 0.0396 [A] -0.18 0.041 1.14E-05 0.08 [601275] (Intronic) GPM6A g.176866310G>A 4:176866310 28402107 0.0379 [A] -0.19 0.043 1.15E-05 0.16 [601275] (Intronic) GPM6A g.176862038G>A 4:176862038 10222704 0.0391 [A] -0.19 0.042 1.21E-05 0.07 [601275] (Intronic) GPM6A g.176858247T>G 4:176858247 12498242 0.0391 [G] -0.18 0.042 1.25E-05 0.07 [601275] (Intronic) ANKFN1 g.54400595A>G 17:54400595 80152904 0.0136 [G] -0.31 0.072 1.50E-05 1.00 [N/A] (Intronic) ANKFN1 g.54400604A>C 17:54400604 79757657 0.0136 [C] -0.31 0.072 1.50E-05 1.00 [N/A] (Intronic) ANKFN1 g.54426389C>A 17:54426389 181805720 0.0142 [A] -0.29 0.068 2.09E-05 1.00 [N/A] (Intronic) ANKFN1 g.54455923G>A 17:54455923 76404130 0.0142 [A] -0.29 0.069 2.28E-05 1.00 [N/A] (Intronic) ANKFN1 g.54441957T>G 17:54441957 75489175 0.0142 [G] -0.29 0.069 2.30E-05 1.00 [N/A] (Intronic) ANKFN1 g.54483155G>A 17:54483155 187423937 0.0142 [A] -0.30 0.070 2.57E-05 1.00 [N/A] (Intronic) ANKFN1 g.54534855G>A 17:54534855 77486483 0.0130 [A] -0.30 0.072 3.32E-05 1.00 [N/A] (Intronic) GPM6A g.176859026A>T 4:176859026 7687921 0.0413 [T] -0.12 0.029 4.82E-05 0.06 [601275] (Intronic) GPM6A g.176859021C>A 4:176859021 7663888 0.0413 [A] -0.12 0.029 4.89E-05 0.06 [601275] (Intronic) g.69184522G>A Intergenic 3:69184522 7634162 0.0142 [A] -0.27 0.067 4.96E-05 1.00 (N/A) MOV10L1 g.50584130G>A 22:50584130 55747387 0.0379 [A] -0.13 0.032 6.30E-05 1.00 [605794] (p.Ala840Thr)

145

DNAH11 g.21633505T>C 7:21633505 1880304 0.3834 [C] -0.05 0.013 6.36E-05 0.70 [603339] (Intronic) g.112702124T>A Intergenic 3:112702124 983386 0.1846 [A] 0.07 0.017 7.50E-05 0.70 (N/A) FBN3 g.8150662A>G 19:8150662 7260399 0.4909 [G] -0.06 0.015 8.53E-05 0.23 [608529] (Intronic) g.112706125C>A Intergenic 3:112706125 1488191 0.1801 [A] 0.07 0.017 9.05E-05 0.48 (N/A) g.69187537G>C Intergenic 3:69187537 76476421 0.0136 [C] -0.29 0.074 9.23E-05 1.00 (N/A) g.75758122C>T Intergenic 14:75758122 12895416 0.3313 [T] 0.05 0.012 9.51E-05 0.92 (N/A) g.69173340T>C Intergenic 3:69173340 72924888 0.0142 [C] -0.23 0.060 9.61E-05 1.00 (N/A) g.112705874A>G Intergenic 3:112705874 984706 0.1806 [G] 0.07 0.017 9.74E-05 0.43 (N/A) g.75755071G>C Intergenic 14:75755071 4903304 0.3313 [C] 0.05 0.013 9.80E-05 0.92 (N/A) g.75753532A>G Intergenic 14:75753532 11159132 0.3313 [G] 0.05 0.013 9.82E-05 0.92 (N/A) g.75753791A>G Intergenic 14:75753791 10134421 0.3313 [G] 0.05 0.013 9.85E-05 0.92 (N/A) Abbreviations: Chr, chromosome; dbSNP, Single Nucleotide Polymorphism Database; HWE, Hardy-Weinberg Equilibrium; MAF, minor allele frequency; MIM, Mendelian Inheritance in Man; SE, standard error. aChr:Position reported in GRCh37. bAmino acid alteration reported in dbSNP. cCorneal curvature measurements were adjusted for age, sex, and baseline height. dBonferroni-corrected genome-wide significance threshold = P < 1.09 × 10-6. Threshold for suggestive significance = P < 1 × 10-4. eHWE exact test used.

146

Supplementary Table S6. Top Single Variant Association Analysis Results for Spherical Equivalent Gene Chr:Position DNA (Amino Acid) MAF in BDES HWE P- dbSNP rsID βc SE P-valued [MIM] (Base-Pair)a Alterationb [Minor Allele] Valuee VPS18 g.41191095G>A 15:41191095 12591803 0.0035 [A] 4.47 0.947 2.31E-06 1.00 [608551] (Intronic) VPS18 g.41193968G>A 15:41193968 61250004 0.0035 [A] 4.59 0.973 2.39E-06 1.00 [608551] (Intronic) g.76022274A>C Intergenic 11:76022274 2851466 0.4233 [C] -0.51 0.116 1.01E-05 0.24 (N/A) g.76022722C>T Intergenic 11:76022722 736984 0.4233 [T] -0.51 0.115 1.06E-05 0.24 (N/A) g.76022282G>T Intergenic 11:76022282 2851467 0.4233 [T] -0.51 0.116 1.14E-05 0.24 (N/A) g.76022544C>T Intergenic 11:76022544 2618089 0.4394 [T] -0.51 0.116 1.14E-05 0.40 (N/A) g.44248257G>A TCTE1 6:44248257 9381305 0.1385 [A] 0.60 0.138 1.32E-05 0.25 (Intronic) AARS2 g.44278592A>G 6:44278592 10807297 0.1382 [G] 0.62 0.143 1.74E-05 0.25 [612035] (Intronic) TCTE1 g.44253650A>G 6:44253650 2297335 0.1385 [G] 0.58 0.135 1.76E-05 0.25 [186975] (Intronic) TCTE1 g.44256175T>G 6:44256175 9381308 0.1385 [G] 0.58 0.135 1.85E-05 0.25 [186975] (Intronic) TCTE1 g.44255777T>C 6:44255777 9394998 0.1385 [C] 0.58 0.135 1.86E-05 0.25 [186975] (Intronic) TCTE1 g.44255778C>A 6:44255778 9394999 0.1385 [A] 0.58 0.135 1.86E-05 0.25 [186975] (Intronic) TCTE1 g.44253391C>T 6:44253391 9381307 0.1385 [T] 0.58 0.135 1.87E-05 0.25 [186975] (Intronic) TCTE1 g.44253765A>G 6:44253765 2297336 0.1385 [G] 0.58 0.135 1.88E-05 0.25 [186975] (p.Phe261Ser) AARS2 g.44266984G>A 6:44266984 1123523 0.1385 [A] 0.59 0.139 1.91E-05 0.25 [612035] (3' UTR)

147

TCTE1 g.44251045C>A 6:44251045 4714774 0.1385 [A] 0.58 0.136 1.91E-05 0.25 [186975] (Intronic) TCTE1 g.44248390C>T 6:44248390 9381306 0.1385 [T] 0.59 0.137 1.93E-05 0.25 [186975] (Intronic) TCTE1 g.44261187A>C 6:44261187 12205227 0.1385 [C] 0.59 0.137 1.93E-05 0.25 [186975] (Intronic) TCTE1 g.44259390G>A 6:44259390 12215176 0.1385 [A] 0.58 0.137 1.94E-05 0.25 [186975] (Intronic) TCTE1 g.44248521A>T 6:44248521 9369469 0.1385 [T] 0.59 0.137 1.95E-05 0.25 [186975] (Intronic) TCTE1 g.44260169T>C 6:44260169 9395000 0.1385 [C] 0.58 0.137 2.02E-05 0.25 [186975] (Intronic) TCTE1 g.44252253C>T 6:44252253 12664825 0.0892 [T] 0.78 0.187 2.94E-05 0.20 [186975] (Intronic) g.76020672C>T Intergenic 11:76020672 1107843 0.4707 [T] -0.47 0.114 3.48E-05 0.58 (N/A) g.46777003C>T Intergenic 20:46777003 227880 0.4849 [T] -0.45 0.109 4.17E-05 0.27 (N/A) g.46776892G>T Intergenic 20:46776892 2758951 0.4849 [T] -0.45 0.109 4.18E-05 0.27 (N/A) g.46776912C>G Intergenic 20:46776912 151051 0.4849 [G] -0.45 0.109 4.19E-05 0.27 (N/A) g.76025070T>C Intergenic 11:76025070 2851155 0.4849 [C] -0.44 0.109 4.52E-05 0.46 (N/A) g.46776607A>G Intergenic 20:46776607 151050 0.4849 [G] -0.44 0.108 4.54E-05 0.27 (N/A) g.46777272G>A Intergenic 20:46777272 2758954 0.4849 [A] -0.44 0.108 4.59E-05 0.27 (N/A) g.76027936G>T Intergenic 11:76027936 745870 0.4797 [T] 0.45 0.111 4.60E-05 0.68 (N/A) g.76020398A>G Intergenic 11:76020398 11236698 0.4716 [G] -0.46 0.114 4.82E-05 0.58 (N/A) g.46778390T>C Intergenic 20:46778390 11908506 0.4849 [C] -0.39 0.097 5.18E-05 0.27 (N/A)

148

g.46779102T>C Intergenic 20:46779102 6094991 0.4849 [C] -0.39 0.097 5.19E-05 0.27 (N/A) g.46779442T>C Intergenic 20:46779442 6066649 0.4849 [C] -0.39 0.097 5.19E-05 0.27 (N/A) g.46779235T>C Intergenic 20:46779235 11086243 0.4849 [C] -0.39 0.097 5.19E-05 0.27 (N/A) g.46778503T>C Intergenic 20:46778503 11700195 0.4849 [C] -0.39 0.097 5.20E-05 0.27 (N/A) g.46779491C>T Intergenic 20:46779491 11698870 0.4849 [T] -0.39 0.097 5.20E-05 0.27 (N/A) g.46778325T>C Intergenic 20:46778325 6066645 0.4849 [C] -0.39 0.097 5.20E-05 0.27 (N/A) g.46778574G>A Intergenic 20:46778574 11700186 0.4849 [A] -0.39 0.097 5.23E-05 0.27 (N/A) g.46777647A>G Intergenic 20:46777647 6063216 0.4849 [G] -0.40 0.098 5.23E-05 0.27 (N/A) g.46778984T>C Intergenic 20:46778984 6063218 0.4849 [C] -0.39 0.097 5.24E-05 0.27 (N/A) g.46778604A>C Intergenic 20:46778604 6066646 0.4849 [C] -0.39 0.097 5.24E-05 0.27 (N/A) g.46777514T>A Intergenic 20:46777514 6063215 0.4849 [A] -0.40 0.098 5.25E-05 0.27 (N/A) g.46780313G>A Intergenic 20:46780313 6066651 0.4849 [A] -0.39 0.097 5.26E-05 0.27 (N/A) g.46778744A>G Intergenic 20:46778744 6094990 0.4849 [G] -0.39 0.097 5.27E-05 0.27 (N/A) g.76025889C>T Intergenic 11:76025889 2618091 0.4897 [T] -0.44 0.108 5.30E-05 0.07 (N/A) g.46764459T>C Intergenic 20:46764459 6066638 0.4849 [C] -0.43 0.106 5.78E-05 0.27 (N/A) ST14 g.130059701G>A 11:130059701 150984123 0.0035 [A] 3.41 0.848 5.87E-05 1.00 [606797] (p.Glu170Lys) g.46764627G>A Intergenic 20:46764627 1133029 0.4849 [A] -0.43 0.107 5.87E-05 0.27 (N/A)

149

g.113901194C>T Intergenic 10:113901194 34820178 0.2732 [T] 0.51 0.126 5.89E-05 0.73 (N/A) g.76027338G>A Intergenic 11:76027338 2618086 0.4868 [A] -0.44 0.109 5.90E-05 0.07 (N/A) g.113901637T>C Intergenic 10:113901637 58175211 0.2732 [C] 0.51 0.126 5.95E-05 0.73 (N/A) g.46763414G>C Intergenic 20:46763414 6090807 0.4849 [C] -0.43 0.107 5.97E-05 0.27 (N/A) g.113902454C>T Intergenic 10:113902454 72836628 0.2732 [T] 0.51 0.126 6.00E-05 0.73 (N/A) g.113902418A>C Intergenic 10:113902418 2792744 0.2732 [C] 0.51 0.126 6.08E-05 0.73 (N/A) g.113902827C>T Intergenic 10:113902827 2803606 0.2735 [T] 0.51 0.126 6.24E-05 0.73 (N/A) GPAM g.113933886G>A 10:113933886 2255141 0.2742 [A] -0.43 0.107 6.72E-05 0.77 [602395] (Intronic) GPAM g.113934384C>G 10:113934384 2803619 0.2742 [G] -0.43 0.107 6.76E-05 0.77 [602395] (Intronic) g.76029879T>C Intergenic 11:76029879 2618084 0.4774 [C] -0.46 0.116 6.77E-05 0.06 (N/A) GPAM g.113937941T>G 10:113937941 2487294 0.2742 [G] -0.43 0.107 6.88E-05 0.77 [602395] (Intronic) TMEM151B g.44246049G>A 6:44246049 325017 0.1678 [A] -0.52 0.131 6.92E-05 0.74 [N/A] (3' UTR) GPAM g.113936855T>C 10:113936855 2792759 0.2742 [C] -0.43 0.107 7.04E-05 0.77 [602395] (Intronic) GPAM g.113939584A>G 10:113939584 2803621 0.2742 [G] -0.43 0.107 7.13E-05 0.86 [602395] (Intronic) TMEM151B g.44246766T>C 6:44246766 12943 0.1678 [C] -0.52 0.130 7.26E-05 0.74 [N/A] (3' UTR) TCTE1 g.44253617T>C 6:44253617 324142 0.1678 [C] -0.51 0.128 7.37E-05 0.74 [186975] (Intronic) TCTE1 g.44254793C>T 6:44254793 324145 0.1678 [T] -0.51 0.127 7.37E-05 0.74 [186975] (Intronic)

150

TMEM151B g.44245930T>C 6:44245930 325018 0.1678 [C] -0.52 0.131 7.39E-05 0.74 [N/A] (3' UTR) TCTE1 g.44255459A>G 6:44255459 324146 0.1678 [G] -0.50 0.127 7.41E-05 0.74 [186975] (p.Pro35Leu) TCTE1 g.44248262T>A 6:44248262 694349 0.1678 [A] -0.51 0.129 7.48E-05 0.74 [186975] (Intronic) TCTE1 g.44250853T>A 6:44250853 324139 0.1678 [C] -0.51 0.128 7.49E-05 0.74 [186975] (Intronic) TCTE1 g.44250165G>A 6:44250165 516582 0.1678 [A] -0.51 0.128 7.50E-05 0.74 [186975] (p.Gly326Gly) TCTE1 g.44252628T>C 6:44252628 324140 0.1678 [C] -0.51 0.128 7.50E-05 0.74 [186975] (Intronic) TCTE1 g.44248364T>A 6:44248364 570469 0.1678 [A] -0.51 0.129 7.51E-05 0.74 [186975] (Intronic) TCTE1 g.44248646A>G 6:44248646 844315 0.1678 [G] -0.51 0.129 7.51E-05 0.74 [186975] (Intronic) TCTE1 g.44254687T>C 6:44254687 324144 0.1678 [C] -0.50 0.127 7.55E-05 0.74 [186975] (Intronic) TCTE1 g.44256020C>T 6:44256020 324147 0.1678 [T] -0.51 0.128 7.56E-05 0.74 [186975] (Intronic) GPAM g.113936244C>A 10:113936244 2255400 0.2742 [A] -0.42 0.107 7.59E-05 0.77 [602395] (Intronic) GPAM g.113922728G>A 10:113922728 2803611 0.2742 [A] -0.42 0.107 7.60E-05 0.77 [602395] (Intronic) TCTE1 g.44250838T>C 6:44250838 324138 0.1678 [C] -0.51 0.128 7.60E-05 0.74 [186975] (Intronic) GPAM g.113921354A>G 10:113921354 2250802 0.2735 [G] -0.42 0.107 7.70E-05 0.73 [602395] (Intronic) TCTE1 g.44258034T>A 6:44258034 614715 0.1678 [A] -0.51 0.128 7.76E-05 0.74 [186975] (Intronic) TCTE1 g.44258324A>G 6:44258324 557373 0.1678 [G] -0.51 0.128 7.90E-05 0.74 [186975] (Intronic) GPAM g.113940329C>T 10:113940329 2792751 0.2735 [T] -0.42 0.107 8.14E-05 0.91 [602395] (p.Ile43Val)

151

GPAM g.113947040T>C 10:113947040 4918722 0.2735 [C] -0.43 0.109 8.34E-05 0.91 [602395] (Intronic) GPAM g.113949664C>T 10:113949664 10787429 0.2735 [T] -0.43 0.109 8.43E-05 0.91 [602395] (Intronic) GPAM g.113921159A>T 10:113921159 2792736 0.2735 [T] -0.42 0.107 8.45E-05 0.73 [602395] (Intronic) DNAH14 g.225561187G>A 1:225561187 61851862 0.0722 [A] -0.80 0.203 8.53E-05 0.61 [603341] (Intronic) GPAM g.113921825A>G 10:113921825 2792735 0.2774 [G] -0.42 0.107 8.62E-05 0.82 [602395] (Intronic) GPAM g.113944940A>T 10:113944940 2263608 0.2738 [T] -0.43 0.108 8.73E-05 0.82 [602395] (Intronic) GPAM g.113944271G>A 10:113944271 2419604 0.2738 [A] -0.42 0.108 8.76E-05 0.82 [602395] (Intronic) KCNMB4 g.70827901A>G 12:70827901 138742128 0.0090 [G] -2.44 0.623 8.81E-05 1.00 [605223] (3' UTR) GPAM g.113917085A>T 10:113917085 2254537 0.2735 [T] -0.42 0.107 8.82E-05 0.73 [602395] (p.Pro681Pro) GPAM g.113931690T>C 10:113931690 2792703 0.2784 [C] -0.42 0.107 8.83E-05 0.91 [602395] (Intronic) GPAM g.113916835C>A 10:113916835 2254532 0.2735 [A] -0.42 0.107 8.85E-05 0.73 [602395] (Intronic) GPAM g.113916302C>T 10:113916302 2803608 0.2735 [T] -0.42 0.107 8.90E-05 0.73 [602395] (Intronic) GPAM g.113917833C>T 10:113917833 6585139 0.2735 [T] -0.42 0.107 8.93E-05 0.73 [602395] (Intronic) GPAM g.113950418C>T 10:113950418 7096937 0.2738 [T] -0.43 0.110 8.98E-05 0.82 [602395] (Intronic) GPAM g.113947678G>A 10:113947678 4917628 0.2738 [A] -0.43 0.109 9.04E-05 0.82 [602395] (Intronic) GPAM g.113910721G>A 10:113910721 1129555 0.2738 [A] -0.42 0.108 9.42E-05 0.69 [602395] (Intronic) g.113903510G>A Intergenic 10:113903510 2792743 0.3985 [A] 0.48 0.123 9.92E-05 0.70 (N/A)

152

Abbreviations: Chr, chromosome; dbSNP, Single Nucleotide Polymorphism Database; HWE, Hardy-Weinberg Equilibrium; MAF, minor allele frequency; MIM, Mendelian Inheritance in Man; SE, standard error. aChr:Position reported in GRCh37. bAmino acid alteration reported in dbSNP. cSpherical equivalent measurements were adjusted for age, sex, education, and nuclear sclerosis. dBonferroni-corrected genome-wide significance threshold = P < 9.25 × 10-7. Threshold for suggestive significance = P < 1 × 10-4. eHWE exact test used.

153

Supplementary Table S7. Top Gene-Based Association Analysis Results for Corneal Curvature and Spherical Equivalenta Gene Amino Acid Locusb P-Valuec Number of Variants CMAFd Start Locatione End Locatione [MIM] Alterationf Corneal Curvature g.158272652T>C (p.Asp206Gly), CYTIP g.158290914G>C 2q24 1.43 × 10-6 3 0.0023 158272652 158341335 [604448] (p.Gln83Glu), g.158289499T>G (Intronic) g.176859021C>A (Intronic), g.176862038G>A (Intronic), g.176858247T>G GPM6A (Intronic), 4q34 1.78 × 10-5 6 0.0781 176560091 176866310 [601275] g.176859026A>T (Intronic), g.176866310G>A (Intronic), g.176859667T>A (Intronic) Spherical Equivalent g.2992748G>A NAP1L4 (p.Thr111Ile), 11p15 1.88× 10-6 2 0.0019 2972505 2992748 [601651] g.2972505G>A (p.Ala369Val) g.220254611G>A DNPEP 2q35 1.00 × 10-5 2 0.0006 220254611 220256545 (Intronic), [611367] g.220256545G>T

154

(Intronic) g.140595001A>G PCDHB13 (p.Asn436Asp), 5q31 8.50 × 10-5 2 0.0013 140594470 140595802 [604967] g.140594470G>A (p.Gly259Ser) Abbreviations: CMAF, cumulative minor allele frequency; MIM, Mendelian Inheritance in Man. aGene-based associations for corneal curvature adjusted for age, sex, and baseline height. Gene-based associations for spherical equivalent adjusted for age, sex, education, and nuclear sclerosis. bGene locations reported in GRCh37. cBonferroni-corrected genome-wide significance threshold = P < 3.61 × 10-6. Threshold for suggestive significance = P < 1 × 10-4. dCumulative minor allele frequency summed over all variants in the gene region in the analyzed sample. eChromosomal positions for the variants included in each gene, reported in GRCh37. fAmino acid alteration reported in dbSNP.

155

Supplementary Table S8. Top Results of Meta-Analysis of First and Second Beaver Dam Eye Study Groups for Axial Length, Corneal Curvature, and Spherical Equivalent Varianta Chr Gene Meta-Analysis Het I2 Het P-Value MAF MAF [MIM] P-Valueb (Directions (First (Second of Effect) BDES) BDES) Axial Length rs140506609 4 WFS1 3.45 × 10-5 26.8 0.2426 (++) 0.0309 0.0352 (g.6284151G>A) [606201] Corneal Curvature rs9507184 13 Intergenic 4.71 × 10-5 0.0 0.9697 (++) 0.4230 0.3869 (g.24473642T>C) rs75752641 13 Intergenic 4.80 × 10-5 0.0 0.9694 (++) 0.4230 0.3887 (g.24473240A>G) rs6825981 4 FRAS1 4.82 × 10-5 0.0 0.4398 (++) 0.3773 0.3578 (g.79346759A>G) [607830] rs12251836 10 IL2RA 5.90 × 10-5 0.0 0.4128 (++) 0.4371 0.3657 (g.6091281T>A) [147730] rs12414777 10 CNNM2 7.90 × 10-5 0.0 0.9443 (--) 0.2033 0.2094 (g.104697781C>T) [607803] Spherical Equivalent rs61743204 21 SON 2.35 × 10-5 0.0 0.7964 (--) 0.1134 0.1029 (g.34926043G>A, [182465] p.Met1502Ile) rs8133982 21 ITSN1 7.18 × 10-5 0.0 0.9448 (++) 0.1134 0.1006 (g.35018470A>G) [602442] rs116321235 6 Intergenic 8.03 × 10-5 0.0 0.9287 (--) 0.0190 0.0190 (g.31318225G>A) rs116472963 6 Intergenic 8.31 × 10-5 0.0 0.9296 (--) 0.0187 0.0194 (g.31233460C>A) rs3761347 21 Intergenic 8.64 × 10-5 0.0 0.7734 (++) 0.1144 0.1160

156

(g.34864284T>C) Abbreviations: Chr, chromosome; Het, heterozygosity; MAF, minor allele frequency; MIM, Mendelian Inheritance in Man; SE, standard error. aGene locations reported in GRCh37. bSignificance threshold = P < 5.0 × 10-8. Threshold for suggestive significance = P < 1 × 10-4.

157

Supplementary Table S9. Association Results for Variants in Previously Published Loci for Axial Length Locusa Gene GWAS GWAS Variantb GWAS GWAS GWAS GWAS BDES BDES BDES [MIM] Trait Sample Effect P- MAF Effect P-Value MAF Size Size Value Size (SE)c (SE) 15q14 15q14 Axial rs11073058 20,747 0.07 4.3 × 0.45 -0.095 0.14 0.47 (20) Length (g.34989626G>T) (0.01) 10-11 (0.065)

3q12 CMSS1 Axial rs9811920 20,747 0.08 4.9 × 0.40 0.017 0.80 0.41 [N/A] Length (g.99844293G>A) (0.01) 10-11 (0.065) (20) 1q41 1q41 Axial rs4373767 4,944 -0.16 2.7 × 0.30 -0.058 0.38 0.46 (21) Length (g.219759682C>T) (0.02) 10-10 (0.067)

5q11.2 PARP8 Axial rs282544 4,944 -0.10 4.2 × 0.35 0.076 0.23 0.30 [N/A] Length (g.50026465T>C) (0.02) 10-6 (0.064) (21) 1q41 1q41 Axial rs4428898 4,944 -0.14 9.1 × 0.30 -0.043 0.41 0.48 (21) Length (g.219739966G>A) (0.02) 10-9 (0.052) 22q12.2 ZNRF3 Axial rs12321 20,747 -0.05 4.1 × 0.46 -0.005 0.92 0.44 [612062] Length (g.29453193G>C) (0.01) 10-8 (0.052) (20) Abbreviations: GWAS, genome-wide association study; HR, hazard ratio; MAF, minor allele frequency; MIM, Mendelian Inheritance in Man; OR, odds ratio; SE, standard error. aReported in GRCh37 (hg19). bSingle variant associations of axial length adjusted for age measured at visit 4, sex, and education. cAll effects are reported as beta coefficients.

158

A B

Supplementary Figure S1. Principal component (PC) plots of PC2 v. PC1 for the Beaver Dam Eye Study sample and: (A) the HapMap CEPH Utah (CEU), Gujarati Indian (GIH), Han Chinese (CHB), Japanese (JPT), Mexican (MXL), Toscani (TSI), and Yoruba (YRI) populations, and (B) the HapMap, CEU, and TSI populations. Individuals from the BDES are indicated by black dots. Individuals from the HapMap populations are indicated by red triangles (CEU), dark magenta triangles (GIH), chocolate brown triangles (CHB), blue triangles (JPT), gold triangles (MXL), green triangles (TSI), and dark turquoise triangles (YRI). Twelve

159 individuals from the BDES do not aggregate with the CEU and TSI HapMap populations (both of which are European populations), and were therefore removed prior to imputation.

160

161

162

163

164

165

Supplementary Figure S2. QQ-plots for the: (A) axial length of the right eye (λ = 1.02), (B) axial length of the left eye (λ = 1.02), (C) corneal curvature of the right eye (λ = 0.99), (D) corneal curvature of the left eye (λ = 1.01), and (E) spherical equivalent (λ =

1.02). The observed p-values (-log10 transformed) are plotted against the p-values (-log10 transformed) expected given no association. There does not appear to be notable population stratification for any phenotype. The red line represents the expected relationship between observed and expected p-values given no association. Deviations from the red line suggest association.

166

167

168

169

Supplementary Figure S3. Overlapping histograms of the distributions of: (A) axial length of the right eye, (B) corneal curvature of the right eye, and (C) mean spherical equivalent for the first (blue) and second (red) Beaver Dam Eye Study groups. The sample sizes for the first and second BDES groups are 874 and 568, 883 and 566, and 1,552 and 1,263 for axial length, corneal curvature, and spherical equivalent, respectively. The first BDES group shows a wider distribution of axial length and spherical equivalent compared to the second BDES group. The distributions of corneal curvature in each group are more similar.

170

Samantha Bomotti

Personal Data

Home Address: 3018 N. Calvert St. #2 Baltimore, MD 21218

Phone: (702) 824-4279

Email: [email protected]

Date of Birth: December 9, 1988 (Anchorage, AK)

Education

2018 Doctor of Philosophy Department of Epidemiology Bloomberg School of Public Health Johns Hopkins University Baltimore, MD Advisors: Priya Duggal, PhD Alison P. Klein, PhD Thesis: Assessing the Non-Genetic and Genetic Factors Affecting Refraction Among the Aging Adult Population

2013 Master of Public Health Department of Epidemiology Certificate in Public Health School of Public Health Genetics University of Michigan Ann Arbor, MI Thesis: Epigenetic Markers of Renal Function in African Americans

2011 Bachelor of Science University of Arizona Majors: Biochemistry and Tucson, AZ Molecular Biophysics, Honors Thesis: The Fibroblast Molecular and Cellular Growth Factor Pathway: A

171

Biology Necessary and Sufficient Means of of Inducing Endothelial Cell Development in Nascent Chicken Embryos Research Experience

2015-Present Doctoral Dissertation Refraction in the Beaver Dam Eye Study Priya Duggal, PhD, Alison Klein, PhD Johns Hopkins University Baltimore, MD

2015 F31 National Research Service Award Training Grant Writer Submitted to the National Institutes of Health Terri H. Beaty, PhD, Robert Wojciechowski, PhD Johns Hopkins University Baltimore, MD

2014-Present Research Associate Genetic Variants of Stargardt Disease Hendrik Scholl, MD, Robert Wojciechowski, PhD Johns Hopkins University Baltimore, MD

2013-2014 Research Associate Genetic Variants of Cleft Lip/Cleft Palate Joan E. Bailey-Wilson, PhD National Human Genome Research Institute Baltimore, MD

2012-2013 Master’s Thesis Epigenetic Markers of Chronic Kidney Disease Sharon Kardia, PhD University of Michigan Ann Arbor, MI

2011-2012 Research Associate Triggers for Myocardial Infarction Among Colombian Patients Ana Baylin, MD, Dr.PH University of Michigan

172

Ann Arbor, MI

2009-2011 Bachelor’s Thesis The Genetics of the Fibroblast Growth Factor Pathway Parker Antin, PhD University of Arizona Tucson, AZ

2009 Research Associate Potential Causes of Alzheimer’s Disease and the Effects of Chemotherapies on Cancer Patients Bryan Spangelo, PhD University of Nevada, Las Vegas

Teaching Experience

2016 Lead Teaching Assistant, Epidemiological Methods, Second Term Johns Hopkins Bloomberg School of Public Health

2015-2016 Teaching Assistant, Genetic Epidemiology, First Term (online and on-site) Johns Hopkins Bloomberg School of Public Health

2015 Lead Teaching Assistant, Genetic Epidemiology, Second Term (online and on-site) Johns Hopkins Bloomberg School of Public Health

2014 Teaching Assistant, Epidemiological Methods, Second Term Johns Hopkins Bloomberg School of Public Health

Professional Activities

2016-Present Member: The American Society of Human Genetics

2016 Member: Genomics of Common Diseases

2014-2015 Member: The Association for Research in Vision and Ophthalmology

2012-2015 Member: International Genetic Epidemiology Society

173

2014 Attendee: 55th Annual Short Course on Medical and Experimental Mammalian Genetics

2014 Attendee: 25th Annual Wilmer Eye Institute Research Meeting

2009-2011 Member: American Society of Biochemistry and Molecular Biology

2009-Present Member: Golden Key International Honors Society

Honors and Awards

2013-2017 Eye and Vision Genomics Training Program (T32) National Institutes of Health, National Eye Institute

2011-2013 University of Michigan Dean’s Award for Full Tuition Waiver University of Michigan

2011 Summa Cum Laude with Honors University of Arizona

2007-2011 Arizona Excellence Scholarship University of Arizona

2009 Arizona Michael Wells Research Endowment University of Arizona

2008-2011 Arizona Dean’s List with Distinction University of Arizona

Academic Service

2009-2011 Ambassador College of Chemistry and Biochemistry, University of Arizona Tucson, AZ

2010 Hospital Volunteer University Medical Center Radiation Oncology Department Tucson, AZ

174

Publications

Original Reports

Strauss R.W., Ho A., Munoz B., Cideciyan A.V., Sahel J.A., Sunness J.S., Birch D.G., Bernstein P.S., Michealides M., Traboulsi E.I., Zrenner E., Sadda S., Ervin A.M., West S., Scholl H.P., Progression of Stargardt Disease Study Group. The natural history of the progression of atrophy secondary to Stargardt disease (ProgStar) studies: design and baseline characteristics: ProgStar Report No. 1. Ophthalmology 2016; 123: 817-828

Bomotti S.M., Smith J.A., Zagel A.L., Taylor J.Y., Turner S.T., Kardia S.L.R. Epigenetic markers of renal function in african americans. Nursing Research and Practice 2013

Bomotti S.M., Duggal P., Chen F., Loomis S.J., Klein B.E.K., Lee, K.E., Truitt B., Klein R., Iyengar S.K., Klein A.P. Variation in RAD17 and ADGRG7 associated with axial length in the Beaver Dam Eye Study. In Review

Loomis S.J., Klein A.P., Lee K.E., Chen F., Bomotti S.M., Truitt B., Iyengar S.K., Klein R., Klein B.E.K., Duggal P. Variation in RNF149 is associated with nuclear lens opacity. In Review

Fujinami K., Strauss R.W., Chiang J. (Pei-Wen), Audo I., Bernstein P.S., Birch D.G., Bomotti, S.M., Ervine, A.M., Jacobson S.G., Mansfield B.C., Marino M.J., Sahel J., Mohand-Said S., Sunness J.S., Traboulsi E.I., West S., Wojciechowski R., Zrenner E., Michaelides M., Scholl H.P., ProgStar Study Group. Detailed genetic characteristics of an international large cohort of patients with Stargardt disease: ProgStar Study Report #7. In Review

Bomotti S.M., Duggal P., Klein B.E.K., Lee K.E., Klein R., Klein A.P. Change in refraction over a 20-year period in the Beaver Dam Eye Study. In Preparation

Bomotti S.M., Imputation of exome array variants to the haplotype reference consortium panel. In Preparation

Presentations

2017 Poster: American Society of Human Genetics Orlando, FL Imputation of exome array variants to the

175

Haplotype Reference Consortium (HRC)

2017 Poster: American Society of Human Genetics Orlando, FL Genetic characteristics of an international large cohort with Stargardt disease: the progression of atrophy secondary to Stargardt disease (ProgStar) study

2017 Poster: The Maryland Genetics, Epidemiology, and Medicine Training Program: Genetics Research Day Baltimore, MD Identifying rare and low-frequency variants associated with axial length in the Beaver Dam Eye Study (BDES)

2016 Poster: American Society of Human Genetics Vancouver, British Columbia Identifying rare and low-frequency variants associated with axial length in the Beaver Dam Eye Study (BDES)

2016 Poster: The Maryland Genetics, Epidemiology, and Medicine Training Program: Genetics Research Day Baltimore, MD The distribution of ABCA4 variants in Stargardt disease from the ProgStar studies

2015 Poster: International Genetic Epidemiology Society Baltimore, MD The distribution of ABCA4 variants in Stargardt disease from the ProgStar studies

2015 Poster: The Association for Research in Vision and Ophthalmology Denver, CO ABCA4 variants in Stargardt disease: preliminary results from the ProgStar study

176

2015 Poster: 26th Annual Wilmer Eye Institute Research Meeting Baltimore, MD ABCA4 variants in Stargardt disease: preliminary results from the ProgStar study

2015 Poster: The Maryland Genetics, Epidemiology, and Medicine Training Program: Genetics Research Day Baltimore, MD ABCA4 variants in Stargardt disease: preliminary results from the ProgStar study

2013 Poster: International Genetic Epidemiology Society Chicago, IL Analyses of WES data in multiplex Syrian oral clefts families

2012 Poster: International Genetic Epidemiology Society Stevenson, WA Epigenetic markers of renal function in african americans

2012 Poster: Department of Epidemiology, University of Michigan Ann Arbor, MI Epigenetic markers of renal function in african americans

2011 Poster: College of Chemistry and Biochemistry, University of Arizona Tucson, AZ A necessary and sufficient means of inducing endothelial cell development in nascent chicken embryos

177