Genetics of Obsessive-Compulsive Disorder: from Phenotypes to Pharmacogenetics

GENETICS OF OBSESSIVE-COMPULSIVE DISORDER: FROM PHENOTYPES TO PHARMACOGENETICS

Gwyneth Ching Mung Zai

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Institute of Medical Science University of Toronto

Thesis Title: Genetics of Obsessive-Compulsive Disorder: From Phenotypes to Pharmacogenetics

Name: Gwyneth Ching Mung Zai

Degree: Doctor of Philosophy

Department: Institute of Medical Science

Institution: University of Toronto

Year of Convocation: 2016

Overall Abstract

Background: Obsessive-compulsive disorder (OCD) is a debilitating neuropsychiatric disorder that is characterized by a diverse clinical presentation. Evidence suggests genetic involvement in the etiology of OCD; however, genetic association studies have yielded mixed results partly due to clinical heterogeneity.

Aims: We therefore investigated the genetics of OCD subphenotypes: age at onset (AAO), Yale-

Brown Obsessive-Compulsive Scale (Y-BOCS) severity score and symptom dimensions, family history of obsessive-compulsive and related disorders (OCRDs), psychiatric comorbidities, and drug response. We first examined these subphenotypes to ascertain clinically homogeneous dimensions of OCD. We then analyzed these subphenotypes and antidepressant response for genetic association to identify marker(s) for each subphenotype.

Methods: The sample consists of 560 OCD individuals. For the subphenotypic analyses, admixture analysis (STATA) was performed to analyze AAO and factor analysis (SPSS) was applied to reduce the Y-BOCS symptom checklist. Family history and psychiatric comorbidity

ii were obtained using the modified Family History Index and SCID-IV interviews respectively.

The candidate gene study investigated markers, mostly in the remote regulatory regions, of 17 candidates for association with subphenotypes in 497 OCD participants. Genome-wide association study was conducted in a subset of this sample to examine AAO and Y-BOCS severity. Retrospective antidepressant response data was available in 222 OCD individuals.

Statistical analyses were performed using SPSS, PLINK, and R programs, comparing genotype frequencies between different subphenotype groups.

Results: We identified early (≤8), intermediate (9-17), and late onset (≥18 years) OCD groups, and a 5- or 6-factor model by weighing Y-BOCS symptoms. Our sample revealed the following clinical differences: females had a higher percentage of comorbid lifetime OCRDs and the early onset group was associated with greater symmetry/order and contamination/cleaning symptoms.

However, our genetic studies revealed no significant findings across all OCD subphenotypes and antidepressant response after correction for multiple comparisons.

Conclusions: Interesting results, although negative, suggest the role of serotonergic and glutamatergic system genes in addition to a newly identified candidate in influencing OCD subphenotypes and antidepressant response. Although replication studies are warranted, these results contribute to the future development of genetic tests to assess for OCD risk and antidepressant response, leading to better outcome for OCD patients.

Word Count: 350

iii

Acknowledgments

I would like to give my special and sincere thanks to my family, especially my parents, Mr. Henry and Mrs. Elizabeth Zai, my brothers, Dr. Clement Zai and Mr. Felix Zai, and my fiancé, Mr. Sunny Chan, my friends and colleagues from medical school and psychiatry residency, my clinical and research supervisors and mentors for their tremendous help and support throughout my Ph.D. study. I dedicate this Ph.D. thesis to them.

Regarding research, I would like to especially thank my supervisors and mentors, Dr. James L. Kennedy and Dr. Margaret (Peggy) A. Richter, for their guidance, teaching, and insightful comments on my projects as well as advice on my future career path. I would like to express my gratitude to my international mentors (Dr. Katharine Phillips, Dr. Matthew Rudorfer, and Dr. Mark Vawter) and research supervisors (Dr. David Pauls, Professor Barbara Sahakian, and Professor Trevor W. Robbins) for their time, guidance, and support. I would also like to thank the members of my Ph.D. advisory committee, Dr. Joanne Knight and Dr. Paul Arnold, final PAC external advisor, Dr. Paul Sandor, and other members of my Ph.D. examination committee, Dr. Mary Seeman, and Dr. Daniel Geller. For the peer-reviewed process of the original article chapters of this thesis, I would like to thank Dr. Blair Simpson, Dr. Don Black, Dr. Lea Davis, and Dr. Dan Rujescu for their time and constructive feedback. Within the Institute of Medical Science, I would like to acknowledge the director, graduate coordinators, and administrative staff for their hard work and support. For the research team, many thanks to all members of the Psychiatric Neurogenetics Laboratory (Dr. Kennedy’s lab) at the Centre for Addiction and Mental Health (lab manager – Ms. Natalie Freeman; IMPACT study manager – Ms. Nicole Braganza; IMPACT study research analyst – Mr. Sheraz Cheema; lab technicians – Mr. Sajid Shaikh, Mr. David Sibony, and Ms. Maria Tampakeras; visiting postdoctoral fellows – Dr. Vanessa Gonçalves and Dr. Arun Tiwari; colleagues – Dr. Vincenzo de Luca and Dr. Daniel Mueller), the CAMH scientific computing team (manager – Mr. David Rotenberg), and the Frederick W. Thompson Anxiety Disorders Centre at the Sunnybrook Health Sciences Centre (research assistant – Ms. Marissa Williams; database manager – Ms. Karen Wigg; therapist – Ms. Eliza Burroughs; administrative staff – Ms. Carmen Costa and Ms. Amanda Calzolaio), for making my doctoral experience stimulating and exciting. I would further like to acknowledge iv my international collaborators (Brazil: Dr. Roseli Shavitt, Dr. Euripedes Miguel, Dr. Carolina Cappi, and Dr. Maria Conceição do Rosario; USA: Dr. Carol Mathews, Dr. Gregory Hanna, Dr. Thomas Fernandez, Dr. Lea Davis, Dr. James Knowles, and Dr. Dongmei Yu; Spain: Dr. Pino Alonso and Dr. Xavier Estivill; Netherlands: Dr. Danielle Cath and Dr. Eske Derks; Italy: Dr. Cristina Cavallini; Switzerland: Dr. Edna Grünblatt and Dr. Susanne Walitza; South Africa: Dr. Dan Stein and Dr. Christine Lochner; Mexico: Dr. Humberto Nicolini; Canada: Dr. Evelyn Stewart; UK: Professor Naomi Fineberg, Dr. Sam Chamberlain, Dr. Becky Inkster, Dr. Annette Brühl, Dr. Annemieke Apergis-Schoute, Dr. Sharon Morein-Zamir, and Ms. Matilde Vaghi) and their administrative staff for their contributions in my career development and Ph.D. thesis. Finally, I would like to give my sincere thanks and gratitude to all research participants for their time and effort.

Regarding funding, I would like to acknowledge the Ontario Ministry of Health and Long-Term Care (PGY-5 residency salary from 2012-2014), the Canadian Institutes of Health Research (CIHR) Postdoctoral Fellowship (salary from 2014-2016), the W. Garfield Weston Doctoral Fellowship (fellowship travel expense and salary from 2014-2015), the Younger Foundation (conference allowance in 2014) and the Frederick W. Thompson Anxiety Disorders Centre (travel allowance in 2014-2015) for supporting my salary, research allowance, and/or conference expenses. Research funding was supported by a grant from the Ontario Mental Health Foundation (Drs. Richter and Kennedy) and a private donation from the Frederick W. Thompson family (Dr. Richter).

Contributions

Chapter 1: Dr. Gwyneth Zai conducted a thorough literature search on the genetics of obsessive- compulsive disorder (OCD) phenotypes and pharmacogenetics, and she completed the entire chapter. A large part of this chapter was taken from a recent review on the pharmacogenetics of OCD, for which Dr. Gwyneth Zai is the first author, and she performed a thorough literature search and review, wrote, submitted, and revised the manuscript for publication. This review article has been published in the journal of Pharmacogenomics 2014; 15(8):1147-1157.

Chapter 2: Dr. Gwyneth Zai designed the aims, objectives, and hypothese (with guidance from faculty) for this PhD thesis.

Chapter 3: Dr. Gwyneth Zai designed the experiment (with guidance from faculty), performed data collection, corresponded with co-authors, performed all statistical analyses, wrote and revised the manuscript, which is ready for publication.

Chapter 4: Dr. Gwyneth Zai designed the experiment (with guidance from faculty), performed part of the data collection and all genotyping, corresponded with co-authors, performed all the statistical analyses, wrote and revised the manuscript, which is ready for publication after the completion of replication study.

Chapter 5: Dr. Gwyneth Zai designed the experiment (with guidance from faculty), performed part of the data collection, corresponded with co-authors, performed all the statistical analyses, wrote and revised the manuscript, which is ready for publication after the completion of replication study.

Chapter 6: Dr. Gwyneth Zai designed the experiment (with guidance from faculty), performed data collection and most of the genotyping, corresponded with co-authors, performed all the statistical analyses, wrote and revised the manuscript, which is ready for publication after replication study.

Chapter 7: Dr. Gwyneth Zai wrote the entire final chapter, summarizing the results, and providing limitations and future directions to extend the findings from the PhD thesis.

Table of Contents

Acknowledgments ...... iv

Contributions ...... vi

Table of Contents ...... vii

List of Abbreviations ...... xii

List of Tables ...... xix

List of Figures ...... xxii

List of Appendices ...... xxv

Chapter 1 GENERAL INTRODUCTION: LITERATURE REVIEW ...... 1

1 Obsessive-Compulsive Disorder (OCD) ...... 2

1.1 Historical Development and Treatment of Clinical OCD ...... 5

1.2 Treatment of OCD ...... 5

1.3 Gender Differences in OCD ...... 8

1.4 Genetics of OCD ...... 9

1.4.1 Genetics of Gender Differences in OCD ...... 15

1.4.2 Genetics of OCD Subphenotypes ...... 17

1.4.3 Pharmacogenetics of OCD ...... 47

1.5 Genes of Interest ...... 70

1.5.1 Significance of Remote Regulatory Regions ...... 85

1.5.2 Genetic Methodology ...... 87

Chapter 2 AIMS, OBJECTIVES, HYPOTHESES, AND IMPLICATIONS ...... 89

2 Thesis Aims, Hypotheses, and Implications ...... 89

2.1 Aims and Objectives ...... 89

2.1.1 Aims ...... 90

2.1.2 Objectives ...... 90

vii

2.2 Hypotheses ...... 91

2.2.1 Clinical Phenomenology of OCD ...... 91

2.2.2 Genetics of OCD Phenotypes ...... 91

2.2.3 Genetics of SRI Response in OCD ...... 92

2.3 Originality ...... 92

2.4 Implications ...... 92

Chapter 3 CLINICAL PHENOMENOLOGY OF OCD ...... 95

3 An Update on the Clinical Phenomenology of Obsessive-Compulsive Disorder ...... 95

3.1 Abstract ...... 96

3.2 Introduction ...... 97

3.3 Methods ...... 101

3.3.1 Diagnostic Criteria and Clinical Sample ...... 101

3.3.2 Familiality Data to Determine the Presence of Absence of Family History of OCRDs ...... 103

3.3.3 Antidepressant Response Data Description ...... 104

3.3.4 Data Analysis ...... 105

3.4 Results ...... 109

3.4.1 Does Gender Contribute to Clinical Diversity in OCD? ...... 109

3.4.2 Does AAO Present with Multiple Normal Distributions? ...... 111

3.4.3 Y-BOCS Symptom Dimension – Factor Model and Clusters ...... 111

3.4.4 Rate of Psychiatric Comorbidities ...... 112

3.4.5 Rate of Significant Family History of OCRDs ...... 112

3.4.6 Can Clinical Characteristics Predict Antidepressant Response? ...... 113

3.5 Discussion ...... 128

3.5.1 Gender Does Not Contribute to Difference in Clinical Presentations ...... 128

3.5.2 AAO has a Tri-modal Distributions ...... 129

viii

3.5.3 Factor and Cluster Analyses Confirmed a 5- and 6-Factor Model of Y-BOCS Symptom Dimensions ...... 130

3.5.4 High Rates of Psychiatric Comorbidities ...... 132

3.5.5 High Rate of Family History of OCRDs ...... 132

3.5.6 Antidepressant Response Rate is Higher than Published Rates and Differences were Observed between SRI Responders and Non-Responders ...... 133

3.5.7 Limitations of the Clinical Study ...... 135

3.5.8 Conclusion ...... 136

Chapter 4 CANDIDATE GENE STUDY OF OCD PHENOTYPES ...... 137

4 Candidate Gene Study of Obsessive-Compulsive Disorder Phenotypes ...... 137

4.1 Abstract ...... 138

4.2 Introduction ...... 138

4.3 Methods ...... 149

4.3.1 Diagnostic Criteria and Sample for Candidate Gene Analysis ...... 149

4.3.2 Choosing of SNPs of Candidate Genes and Genotyping ...... 153

4.3.3 Quality Control and Statistical Analyses ...... 154

4.4 Results ...... 155

4.5 Discussion ...... 179

Chapter 5 GWAS OF OCD PHENOTYPES ...... 183

5 Genome-Wide Association Study of Obsessive-Compulsive Disorder Phenotypes ...... 183

5.1 Abstract ...... 184

5.2 Introduction ...... 184

5.3 Methods ...... 187

5.3.1 Diagnostic Criteria and Sample ...... 187

5.3.2 Genotyping and Statistical Analyses ...... 187

5.4 Results ...... 189

5.5 Discussion ...... 209 ix

Chapter 6 PHARMACOGENETICS OF OCD ...... 212

6 Multi-Gene Pharmacogenetic Study of Antidepresant Response in Obsessive-Compulsive Disorder ...... 212

6.1 Abstract ...... 213

6.2 Introduction ...... 213

6.3 Methods ...... 217

6.3.1 Diagnostic Criteria and Sample ...... 217

6.3.2 Response Data ...... 218

6.3.3 Choosing of SNPs and Genotyping ...... 219

6.3.4 Statistical Analyses ...... 221

6.4 Results ...... 221

6.5 Discussion ...... 232

Chapter 7 GENERAL DISCUSSION ...... 235

7 General Discussion ...... 235

7.1 Summary of Findings and Implications ...... 236

7.1.1 Clinical OCD (Chapter 3) ...... 237

7.1.2 Genetics of OCD Phenotypes (Chapter 4 and Chapter 5) ...... 238

7.1.3 Pharmacogenetics of OCD (Chapter 6) ...... 240

7.2 Limitations and Considerations ...... 243

7.2.1 Sample Size and Genomic Coverage ...... 243

7.2.2 Corrections for Multiple Comparisons ...... 245

7.2.3 Diagnostic Uncertainty ...... 247

7.2.4 Use of Factor Analysis to Identify OCD Subtypes or Symptom Dimensions .... 248

7.3 Future Directions ...... 248

7.3.1 Trait and Subphenotype Analysis ...... 249

7.3.2 Future Genetic Studies ...... 250

7.4 Concluding Remarks ...... 256

References ...... 257

Appendices ...... 334

Appendix I ...... 335

Appendix II ...... 336

Appendix III ...... 349

Appendix IV ...... 350

List of Abbreviations

χ2 Chi-squared

5-HT Serotonin

AAO Age at onset

ABLIM1 Actin-binding LIM protein family, member 1 gene

ACE Angiotensin-converting enzyme gene

Akt Protein kinase B

AKT1 V-Akt murine thymoma viral oncogene homolog 1 gene

ANKFN1 Ankyrin-repeat and fibronectin type III domain containing 1 gene

ARRDC4 Arrestin domain containing 4

BDD Body dysmorphic disorder

BDKRB2 Bradykinin receptor B2 gene

BDNF Brain-derived neurotrophic factor gene

BTBD3 BTB (POZ) domain containing 3 gene

BTBD9 BTB (POZ) domain containing 9 gene

C9orf68 Chromosome 9 open reading frame 68 gene

CBT Cognitive behavioural therapy

CDH9 Cadherin 9, type 2 (T1-cadherin) gene

CGI-I Clinical Global Impression – Improvement

ChIP Chromatin immuno-precipitation

CI Confidence interval

CIHR Canadian Institutes of Health Research

CNV Copy-number variation

COMT Catechol-O-methyltransferase gene

CRHR1 Corticotropin releasing hormone receptor 1

xii

CSTC Cortico-striato-thalamo-cortical

CYP Cytochrome P450 genes

DAT1 Dopamine transporter gene, also known as SLC6A3

DDR1 Discoidin domain receptor tyrosine kinase 1 gene

DHS DNase hypersensitive site

DISP1 Dispatched homolog 1 (drosophila) gene

DLGAP1 Discs, large (drosophila) homolog-associated protein 1 gene

DLGAP2 Discs, large (drosophila) homolog-associated protein 2 gene

DLGAP3 Discs, large (drosophila) homolog-associated protein 3 gene

DNA Deoxyribonucleic acid

DNase Deoxyribonuclease

DRD1 Dopamine D1 receptor gene

DRD2 Dopamine D2 receptor gene

DRD3 Dopamine D3 receptor gene

DRD4 Dopamine D4 receptor gene

DRD5 Dopamine D5 receptor gene

DSM-5 Diagnostic and Statistical Manual of Mental Disorders, 5th Edition

DSM-IV Diagnostic and Statistical Manual of Mental Disorders, 4th Edition

EAAC1 Excitatory amino acid carrier 1

EFNA5 Ephrin-A5 gene

ERBB4 Erb-b2 receptor tyrosine kinase 4

EurMAP European minor allele frequencies

ENCODE ENCyclopedia Of DNA Elements

ERBB4 V-erb-b2 avian erythroblastic leukemia viral oncogene homolog 4 gene

ERE6 Estrogen response element 6 gene

ERK Extracellular signal-regulated kinase xiii

ESRα Estrogen receptor alpha gene

EurMAF European minor allele frequencies

FAIM2 Fas apoptotic inhibitory molecule 2 gene

FBAT Family-based association test

FHI Family History Interview

FKBP5 FK506 binding protein 5 gene

FMM Finite Mixture Models

FS Functional significance

FUT2 Fucosyltransferase 2 gene

GABHS Group A β-hemolytic streptococcal infection

GAD1 Glutamate decarboxylase 1 gene

GAD2 Glutamate decarboxylase 2 gene

GENDEP Genome-based Therapeutic Drugs for Depression study

GI Gastrointestinal

GNB3 Guanine nucleotide binding protein (G protein), beta polypeptide 3 gene

GPC6 Glypican 6 gene

GPCR G-protein-coupled receptor

GRIA1 Glutamate receptor, ionotropic, AMPA 1 gene

GRIA3 Glutamate receptor, ionotropic, AMPA 3 gene

GRIK2 Glutamate receptor, ionotropic, kainate 2 gene

GRIK4 Glutamate receptor, ionotropic, kainate 4 gene

GRIN2B Glutamate receptor, ionotropic, N-methyl D-aspartate 2B gene

GSK3β Glycogen synthase kinase 3 beta gene

GWAS Genome-wide association study

HD Hoarding disorder

HLA Human leukocyte antigen xiv

HOXB8 Homeobox B8 gene

HTR1A Serotonin 1A receptor gene

HTR1B Serotonin 1D beta receptor gene

HTR2A Serotonin 2A receptor gene

HTR2C Serotonin 2C receptor gene

HTR6 Serotonin 6 receptor gene

HTTLPR Serotonin transporter-linked polymorphic region

ICD-10 International Classification of Diseases, 10th Edition

IER3 Immediate early response 3 gene

IL1B Interleukin 1, beta gene

INPP-1 Inositol-polyphosphatase-1 gene

Ins/del Insertion/deletion polymorphism

IOCDFGC International OCD Foundation Genetic Collaborative Group

KCNN3 Potassium intermediate/small conductance calcium-activated channel, subfamily N, member 3 gene

LD Linkage disequilibrium

LMX1A LIM homeobox transcription factor 1, alpha gene

MAOA Monoamine oxidase A gene

MAPK Mitogen-activated protein kinase

MARS Munich Antidepressant Response Signature project

MDD Major depressive disorder

MDS Multi-dimensional scaling

MOG Myelin oligodendrocyte glycoprotein gene

MPDZ Multiple PDZ domain protein gene

NaSSA Noradrenergic and specific serotonergic antidepressant

NF-κB Nuclear factor kappa-light-chain-enhancer of activated B cells

NGFR Nerve growth factor receptor gene

NIEHS National Institute of Environmental Health Sciences

NIMH National Institute of Mental Health

NMDA N-methyl-D-aspartate

NPSR1 Neuropeptide S receptor 1 gene

NRG1 Neuregulin 1 gene

NRG3 Neuregulin 3

NSAIDs Non-steroidal anti-inflammatory drugs

NTM Neurotrimin gene

NTRK2 Neurotrophic tyrosine kinase, receptor, type 2 gene

NTRK3 Neurotrophic tyrosine kinase, receptor, type 3 gene

OCRDs Obsessive-compulsive and related disorders

OCD Obsessive-compulsive disorder

OLIG1 Oligodendrocyte lineage transcription factor 1 gene

OLIG2 Oligodendrocyte lineage transcription factor 2 gene

OR Odds ratio

PANDAS Pediatric Autoimmune Neuropsychiatric Disorders Associated with (group A β-hemolytic) Streptococcal infection

PANS Pediatric Acute-onset Neuropsychiatric Syndrome

PCA Principal component analysis

PCDH10 Protocadherin 10 gene

PCDH1D Protocadherin-1D gene

PGC Psychiatric Genomics Consortium

PI3K Phosphoinositide 3 kinase

PKC Protein kinase C

PLC-γ1 Phospholipase C gamma-1 gene

xvi

PLCB1 Phospholipase C, beta 1 (phosphoinositide-specific)

PTPN21 Protein-tyrosine phosphatase, nonreceptor-type, 21 gene

PTPRD Protein-tyrosine phosphatase, receptor-type, delta

RNA Ribonucleic acid

SAPAP3 SAP90/PSD95-associated protein 3 gene

SCID Structured Clinical Interview for the DSM-IV

SD Standard deviation

SIADH Syndrome of inappropriate antidiuretic hormone

SLC1A1 Glutamate transporter, neuronal, gene

SLC18A1 Vesicular monoamine transporter

SLC6A2 Norepinephrine transporter gene

SLC6A4 Serotonin transporter gene, also known as 5HTT

SLITRK5 SLIT and NRTK-like family, member 5 gene

SNRI Serotoin-norepinephrine reuptake inhibitor

SNP Single nucleotide polymorphism

SPD Skin picking disorder

SRI Serotonin reuptake inhibitor

SSRI Selective serotonin reuptake inhibitor

STAR*D Sequenced Treatment Alternatives to Relieve Depression study

TCA Tricyclic antidepressant

TDT Transmission disequilibrium test

TF Transcription factor

TFBS Transcription factor binding site

TH Tyrosine hydroxylase

TIAM1 T-cell lymphoma invasion and metastasis 1 gene

TNFA Tumor necrosis factor alpha gene xvii

TPH1 Tryptophan hydroxylase 1 gene

TPH2 Tryptophan hydroxylase 2 gene

TSS Transcription start site

TTM Trichotillomania

USCS University of California, Santa Cruz genome browser

VNTR Variable number tandem repeat

WHO World Health Organization

Y-BOCS Yale-Brown Obsessive-Compulsive Scale

xviii

List of Tables

Table 1.1. Common Symptoms of OCD...... 4

Table 1.2. Pharmacological Treatment of OCD...... 7

Table 1.3. Treatment Response Categories in OCD...... 7

Table 1.4. Summary Table of Genetic Studies of Gender in OCD...... 19

Table 1.5. Summary Table of Genetic Studies of AAO in OCD...... 25

Table 1.6. Summary Table of 5-Factor Y-BOCS Symptom Dimensions in OCD...... 32

Table 1.7. Summary Table of 4-Factor Y-BOCS Symptom Dimensions in OCD...... 35

Table 1.8. Summary Table of 3-, 6-, and Mixed-Factor Y-BOCS Symptom Dimensions in OCD...... 37

Table 1.9. Summary Table of Genetic Studies of Y-BOCS Symptom Dimensions in OCD...... 39

Table 1.10. Summary Table of Genetic Studies of Y-BOCS Symptom Severity in OCD...... 41

Table 1.11. Psychiatric Comorbidity of OCD (Percentage of total number of individuals reported with OCD or in the general population)...... 44

Table 1.12. Genetic Studies of Psychiatric Comorbidity in OCD...... 45

Table 1.13. Summary of Pharmacogenetic Studies in OCD...... 49

Table 1.14. Summary of Candidate Gene Studies of OCD...... 71

Table 3.1. Subject Demographics...... 109

Table 3.2. Comparison of Clinical Characteristics between Males and Females...... 110

Table 3.3. Comparison of Clinical Characteristics between Age At Onset (AAO) Groups...... 116

Table 3.4. Y-BOCS Symptom Dimension Factor Analysis (weighing all symptoms as equal). 118

xix

Table 3.5. Y-BOCS Symptom Dimension Factor Analysis (double weighing target symptoms)...... 119

Table 3.6. Comparison of Clinical Characteristics between Familial and Non-Familial (OCD) OCD Individuals...... 125

Table 3.7. Comparison of Clinical Characteristics between Familial and Non-Familial (OCRDs) OCD Individuals...... 126

Table 3.8. Comparison of Clinical Characteristics between SRI Responders and Non- Responders...... 127

Table 4.1. Chosen Genes...... 144

Table 4.2. Chosen SNPs...... 146

Table 4.3. Subject Demographics...... 155

Table 4.4. Subject Demographics (100% Caucasian)...... 156

Table 4.5. Investigated SNPs...... 156

Table 4.6. Genetic Results of Gender Differences in OCD...... 162

Table 4.7. Genetic Results of Age At Onset (AAO) Groups in OCD...... 164

Table 4.8. Genetic Results of Linear Regression of Age At Onset (AAO) in OCD...... 165

Table 4.9. Genetic Results of Linear Regression of Y-BOCS Total Severity Scores in OCD. .. 167

Table 4.10. Genetics Results of Linear Regression of 5-Factor Y-BOCS Symptom Dimensions in OCD...... 169

Table 4.11. Genetic Results of Linear Regression of 6-Factor Y-BOCS Symptom Dimensions in OCD...... 171

Table 4.12. Genetic Results of Psychiatric Comorbidities in OCD...... 173

Table 4.13. Genetic Results of Family History in OCD...... 176 xx

Table 4.14. Genetic Results of Family History (combined OCRDs) in OCD...... 178

Table 5.1. Genome-Wide Association Study (GWAS) Quality Control...... 189

Table 5.2. Demographic Data of the Entire OCD Sample...... 189

Table 5.3. Top Hits of Age At Onset (AAO) in OCD...... 194

Table 5.4. Top Hits of Yale-Brown Obsessive-Compulsive Scale (Y-BOCS) Severity Score in OCD...... 195

Table 5.5. Population Description for the MDS Plots...... 198

Table 6.1. Brief Summary of Pharmacogenetic Studies of OCD...... 215

Table 6.2. Investigated Genes with Their SNPs for the Single Taqman Assay Protocol...... 216

Table 6.3. Genes and SNPs for the QuantStudio Array Protocol...... 217

Table 6.4. Demographic Data of the Entire OCD Sample...... 222

Table 6.5. Subject Demographics for the Single Taqman Assay Protocol...... 222

Table 6.6. Final Demographic Data for the QuantStudio OCD Pharmacogenetic Sample...... 223

Table 6.7. Pharmacogenetic Results for the Single Taqman Assay Protocol...... 225

Table 6.8. Result of Genetic Variations across Remote Regulatory Regions in 14 OCD Candidate Genes and SRI Response...... 229

xxi

List of Figures

Figure 1.1. Genes Associated with OCD...... 10

Figure 1.2. Annotation of Disease-Associated Variants from ENCyclopedia Of DNA Elements (ENCODE) Data (Kavanagh et al., 2013)...... 88

Figure 2.1. Thesis Hypothesis...... 94

Figure 3.1. Normal Distributions of Age At Onset (AAO)...... 115

Figure 3.2. Y-BOCS Symptom Dimension Cluster Analysis (weighing all symptoms as equal)...... 120

Figure 3.3. Y-BOCS Symptom Dimension Cluster Analysis (double weighing target symptoms)...... 121

Figure 3.4. Current Psychiatric Comorbidities of OCD Sample. Percentage is indicated in brackets...... 122

Figure 3.5. Lifetime Comorbid OCRDs of OCD Sample. Percentage is indicated in brackets. 123

Figure 3.6. Family History of OCD Sample ...... 124

Figure 4.1. Age At Onset (AAO) Distribution with Normal Curve...... 159

Figure 4.2. Log10 Transformed Age At Onset (AAO) Distribution with Normal Curve...... 160

Figure 4.3. Yale-Brown Obsessive-Compulsive Scale (Y-BOCS) Severity Score Distribution with Normal Curve...... 161

Figure 4.4. Linear Regression of Age At Onset (AAO) in OCD...... 166

Figure 4.5. Linear Regression of Y-BOCS Total Severity Scores in OCD...... 168

Figure 4.6. Linear Regression of 5-Factor Y-BOCS Symptom Dimensions in OCD...... 170

Figure 4.7. Linear Regression of 6-Factor Y-BOCS Symptom Dimensions in OCD...... 172

xxii

Figure 5.1. Frequency Distribution of Age At Onset (AAO)...... 190

Figure 5.2. Frequency Distribution of Log10 Transformation of Age At Onset (AAO)...... 191

Figure 5.3. Frequency Distribution of Yale-Brown Obsessive-Compulsive Scale (Y-BOCS) Severity Score...... 192

Figure 5.4. Quantile-Quantile Plot for Age At Onset (AAO) – Left (Initial QQ Plot) and Right (QQ Plot After Permutation). The genomic inflation value (lambda) is 1.003295...... 196

Figure 5.5. Quantile-Quantile Plot for Yale-Brown Obsessive-Compulsive Scale (Y-BOCS) Severity Score – Left (Initial QQ Plot) and Right (QQ Plot After Permutation). The genomic inflation value (lambda) is 1.048644...... 197

Figure 5.6. Multi-Dimensional Scale (MDS) Plot – Initial...... 199

Figure 5.7. Multi-Dimensional Scale (MDS) Plot – Removal 1...... 200

Figure 5.8. Multi-Dimensional Scale (MDS) Plot – Removal 2...... 201

Figure 5.9. Multi-Dimensional Scale (MDS) Plot – Final...... 202

Figure 5.10. Manhattan Plot for Age At Onset (AAO) – Left (Initial Manhattan Plot) and Right (After Permutation)...... 203

Figure 5.11. Manhattan Plot for Yale-Brown Obsessive-Compulsive Scale (Y-BOCS) Severity Score – Left (Initial Manhattan Plot) and Right (After Permutation)...... 204

Figure 5.12. LocusZoom Plot for Age At Onset (AAO) Result on Chromosome 6...... 205

Figure 5.13. LocusZoom Plot for Age At Onset (AAO) Result on Chromosome 10...... 206

Figure 5.14. LocusZoom Plot for Yale-Brown Obsessive-Compulsive Scale (Y-BOCS) Severity Score Result on Chromosome 14...... 207

Figure 5.15. LocusZoom Plot for Yale-Brown Obsessive-Compulsive Scale (Y-BOCS) Severity Score Result on Chromosome 9...... 208

xxiii

Figure 6.1. Result of Genetic Variations across Remote Regulatory Regions in 14 OCD Candidate Genes and Antidepressant Response with only Cacausians. Red box consists of SNP within the gene instead of remote regulatory regions...... 231

Figure 7.1. Summary of Genetic Results in OCD Subphenotypes...... 242

xxiv

List of Appendices

Appendix I. Yale-Brown Obsessive Compulsive Scale (Y-BOCS) Symptom Checklist and Severity Score…………………………………………………………………………. 334

Appendix II. Family History Interview (FHI)…………………………………………….. 335

Appendix III. Pharmacogenetics Questionnaire…………………………………………... 348

Appendix IV. Supplementary Tables……………………………………………………....349

xxv 1

Chapter 1 GENERAL INTRODUCTION: LITERATURE REVIEW

Psychiatric disorders are complex, often chronic, and are multi-factorial, with a range of genetic and environmental contributions. Efforts in delineating the genetic etiology of psychiatric disorders have been hampered by the complex genetic transmission patterns, multiple biological mechanisms, gene-gene and gene-environment interactions, and phenotypic heterogeneity challenges that are commonly faced by psychiatric geneticists. Thus, new research paradigms incorporating the use of high-throughput genetic technologies, computational biostatistical methodologies, response and tolerability to different treatment modalities, and brain imaging techniques, have been adopted in an attempt to elucidate the etiologic mechanisms in psychiatric disorders.

Genetics comprises the study of inheritance patterns of variations across genes in living organisms. Genes are regions of deoxyribonucleic acid (DNA) within the human genome that code for ribonucleic acid (RNA) and protein products. Variations across genes are what make each one of us unique in terms of, for example, eye and hair colours, size, shape, and behaviours. The following major milestones in molecular genetics mark the expansion in our understanding of many medical and psychiatric disorders. The concept of heredity came from the ancient Greek physician, Hippocrates, who theorized that “seeds” from different parts of the body were transmitted to newly conceived embryos, also known as pangenesis, which was later supported by Charles Darwin in 1868 (Holterhoff, 2014). The key ingredient of DNA was isolated in 1869 by a Swiss physician, Dr. Johannes Friedrich Miescher (Bernhard, 1970). The first genetic laws, Mendelian laws of inheritance, were reported by Dr. Gregor Mendel in 1866 after he spent a large amount of time crossing pea plants (Castle, 1903). An early geneticist, Dr. Walter Sutton, described and termed the chromosome in 1902 (Crow and Crow, 2002). The link between chromosomes and heredity was confirmed in 1941, revealing that genes code for proteins, by two geneticists, Dr. Edward Tatum and Dr. George Beadle (Raju, 1999; Satzinger, 2008). The discovery of the double-helix structure of DNA in 1953 by Dr. James Watson and Dr. Francis Crick marks the beginning of molecular genetics (Watson and Crick, 1953). With advancing

molecular technologies, the construction of the complete human genetic library, the Human Genome Project, became a reality in 2001 (Lander et al., 2001). The project identified approximately 23,000 protein-coding genes, which only account for 1.5% of the entire genome, while the remaining so-called “junk DNA” did not appear to have any known functional significance. It was not until 2012 that the ENCyclopedia Of DNA Elements (ENCODE) project announced the importance of these “junk DNA” as having regulatory influence on gene expression (ENCODE Project Consortium, 2012). Thus, the field of molecular genetics has grown exponentially, and large-scale collaborative projects such as the Human Genome Project and ENCODE have provided powerful tools for the investigations of complex traits.

Psychiatric genetics, the study of the role of genetics in psychiatric disorders, has mainly focused on humans, with a limited number of animal models given the complexity of behaviour presented in psychiatric conditions. The founder of psychiatric genetics, Dr. Francis Galton, began his search for hereditary factors of “mental characteristics” in the late nineteenth century (Schulze et al., 2004). With advances in technology and computational capacity, psychiatric genetics has now entered a new era of exploration of the genetic architecture of psychiatric disorders.

Based on the current understanding of psychiatric genetics, the present thesis has investigated genetic contributions that may play a role in the heritability of obsessive-compulsive disorder (OCD) with the main focus being on the identification of different molecular genetic markers that given its heterogeneous nature, may influence different phenotypic expressions of OCD.

1 Obsessive-Compulsive Disorder (OCD)

Obsessive-compulsive disorder (OCD) is a common psychiatric disorder that is characterized by chronic and debilitating symptoms of obsessions and/or compulsions (APA, 2013). Obsessions are described as intrusive, repetitive, and persistent thoughts, urges, or images that cause significant distress and great difficulty in suppression or mental dismissal. Compulsions are excessive and repetitive ritualistic behaviours or mental acts that cannot be

resisted and are perhaps performed to reduce the severe anxiety caused by the obsessions. These obsessions and/or compulsions are excessive, irrational, and ego-dystonic in nature and patients with OCD devote one hour or more per day to these symptoms, which significantly impair their daily functioning. Common examples of obsessions and their related compulsions are presented in Table 1.1 and include: fear of contamination and cleaning rituals, need for symmetry and rearrangement, aggressive thoughts about one’s safety and checking behaviours. This condition is common, affecting 1 to 3% of the general population (APA, 2000; Ruscio et al., 2010), with subclinical rates of obsessions and/or compulsions ranging from 13% to 49% (Fullana et al., 2009).

According to the World Health Organization (WHO 2001), OCD is one of the top ten leading causes of disability worldwide and it is also the fourth most frequently diagnosed psychiatric disorder. Furthermore, the National Comorbidity Study Replication epidemiological study reported OCD as the anxiety disorder with the highest percentage of “serious” classifications, which was defined using several indicators such as a 12-month suicide attempt rate, substantial work disability, or role impairment (Kessler et al., 2005). Thus, increasing public awareness of this disorder, and optimizing assessment and treatment regimens are necessary to reduce the burden of OCD globally.

Given that OCD is an internalizing disease (individuals who suffer from this type of disorder often keep or internalize their problems/issues/symptoms) with limited public awareness in society, it has not been the focus of research and/or treatment until more recently. Most patients suffering from OCD deem themselves as the only ones who exhibit OCD symptoms, thereby creating an invisible barrier to seeking psychiatric help because of shamefulness and stigma. Understanding the etiology of OCD requires increased public awareness of obsessive- compulsive symptoms, increased screening, and more active patient participation; and these are just some of the reasons why research in OCD has been lagging behind other severe and persistent mental illnesses. Greater public awareness of OCD will encourage research participation, which in turn will improve the statistical power of our research sample in identifying genetic factors of OCD. Improving the understanding of the underlying biological mechanism of OCD will help guide research in targeting treatment for this debilitating mental

illness. Discoveries from research will bring about innovations in healthcare, which lead to improved outcome for OCD patients.

OCD was grouped as one of the anxiety disorders in the DSM-IV (Diagnostic and Statistical Manual of Mental Disorders, 4th Edition) (APA, 2000), but became a new category, the “obsessive-compulsive and related disorders” (OCRDs), in the recent DSM-5 (APA, 2013). The new section for OCRDs in the DSM-5 includes OCD, body dysmorphic disorder (BDD), hoarding disorder (HD), trichotillomania (TTM) or hair-pulling disorder, and excoriation (skin- picking) disorder (SPD) (APA, 2013). These disorders share clinical phenomenology, patterns of comorbidity, genetic loading, and etiology with OCD (Monzani et al., 2014; Pallanti et al., 2014). It is worth noting that hoarding was previously considered as a subtype or symptom of OCD in the DSM-IV (APA, 2000). Given its relatively distinct clinical profile, it is now widely accepted as a separate disorder (APA, 2013).

Given its complex clinical presentations with diverse symptom characteristics, OCD is known for its heterogeneous nature (Pauls et al., 2014). Therefore, it seems important to identify different homogeneous subgroups or phenotypes of OCD for genetic analysis.

Table 1.1. Common Symptoms of OCD.

Obsession Examples Associated Compulsions

Fear of Concerns with dirt, germs, contaminants, Cleaning, washing contamination household items, animals Concerns with getting self or other ill Symmetry and Recurrent thoughts of needing to do things in a Ordering, arranging, organizing exactness certain way Checking, reassurance seeking, Aggressive Fear of harming self or others monitoring Recurrent violent images Religious Concerns with sacrilege and blasphemy Praying, asking for forgiveness Excessive concerns with morality, scrupulosity Somatic Excessive concerns with illness of disease Physical checking Excessive concerns with body part or aspect of appearance Excessive concerns with discarding waste or Collecting Hoarding junk Recurrent forbidden or perverse sexual Avoiding situations Sexual thoughts, images, or impulses 4

Pathologic doubt Recurrent worries of making mistakes or doing Checking, performing rituals and completeness something wrong Superstition- Fear of certain "bad" or "unlucky" numbers or Counting related colours

Common symptoms of OCD with specific examples of obesssions and their associated compulsions are reported in this table.

1.1 Historical Development and Treatment of Clinical OCD

OCD was once known as obsessive compulsive neurosis. Symptoms of OCD were first described as early as the 1400s addressing symptoms from superstitious notions to scrupulosity (Climacus, 1982). The first “self-help” writings came from the late 1600s by Richard Baxter and colleagues (Clifford, 1716). The first attempted treatment for OCD symptoms was described by an astrological healer named Richard Napier with zodiac signs and positioning of the symbols of the Sun, Moon, planets, etc. in the 1600s (MacDonald, 1981). Medical treatment for these symptoms known as “bad thoughts” was described in the 1700s by Ms. Hannah Allen with “bloodletting” (Allen, 1683). Other treatments in history include exorcism (Jenike et al., 1986; Berrios, 1989) as individuals suffering from “bad thoughts” were thought to be possessed by the Devil (Aardema and O’Connor, 2007). Sigmund Freud later developed the diagnostic category of obsessional neurosis in 1894 that captured symptoms of OCD (May-Tolzmann, 1998). He attributed obsessive-compulsive behaviours to unconscious conflicts that manifest as symptoms (Jenike et al., 1986; Berrios, 1989) and used psychoanalytical psychotherapy to treat OCD. A paradigm shift occurred in the 1970s when behavioural psychology (and cognitive psychology to a lesser extent) began to overshadow Freudian theory and became the dominant psychotherapeutic model for the treatment of OCD (Rachman, 2009).

1.2 Treatment of OCD

Returning to the present decade, treatment of OCD comprises two different streams, psychotherapy and pharmacotherapy. For mild to moderate severity of OCD, cognitive behavioural therapy (CBT) with exposure and response prevention strategy is the first-line

treatment (CPA, 2006). Its response rate is reported to be as high as 86% (Foa et al., 2005). However, psychotherapy requires schedule flexibility, time commitment, psychological mindedness, high motivation, ability to express feelings/affect, ability to tolerate change/anxiety, and absence of severe psychopathology (Sadock et al., 2007). Therefore, some OCD patients are not suitable for psychotherapy and some prefer pharmacotherapy. The first- and second-line pharmacological treatments of OCD are antidepressants, selective serotonin reuptake inhibitors (SSRIs), serotonin-norepinephrine reuptake inhibitors (SNRIs), and tricyclic antidepressants (TCAs) (Table 1.2) (CPA, 2006; Fineberg et al., 2015; Katzman et al., 2014). The efficacy of antidepressants is lower than that of CBT, ranging between 40% and 60% with reduction of symptoms (Foa et al., 2005; McDougle et al., 1993), which is usually defined as OCD symptom reduction of at least 25% to 35% using the Yale-Brown Obsessive-Compulsive Scale (Y-BOCS) severity score (Tolin et al., 2005) (Table 1.3). Patients who take these antidepressants can experience side effects including gastrointestinal (GI) disturbances (nausea, vomiting, diarrhea, constipation), headaches, insomnia, sedation, fatigue, sexual dysfunction, dry mouth, sweating, anxiety, prolonged QTc intervals, GI bleeding when combined with non-steroidal anti- inflammatory drugs (NSAIDs), serotonin syndrome when combined with other serotonergic agents, in addition to rare adverse events such as osteoporosis and fractures in the elderly, hyponatremia due to syndrome of inappropriate antidiuretic hormone (SIADH), and agranulocytosis. In practice, most patients need to try several antidepressants prior to achieving a satisfactory response. The long therapeutic lag before benefits are seen in OCD pharmacotherapy translates into medication trials requiring 10 to 12 weeks duration each. This factor in combination with the typical pattern of trying a number of different medications before a good response is achieved means that many patients endure periods of many months of ongoing disability even after starting medications. Currently, there are no practice guidelines that differentiate patients according to their response or tolerance to psychotropic medications and/or psychotherapy. Thus, the ability to identify in advance the right medication or therapy that will result in the best response would make a substantial difference to clinical practice. One of the goals of future genetic testing is to utilize genetic materials to determine whether a patient will respond to and/or tolerate a proposed treatment (i.e., medication, psychotherapy).

Table 1.2. Pharmacological Treatment of OCD.

Lowest Highest FDA Off-Labelled Class Generic Name Brand Name Consensus Guideline Evidence Recommended Approved Highest Dose Dose Dose SSRI Fluoxetine Prozac First-line First-line Level 1 40mg/day 80mg/day 80mg/day SSRI Fluvoxamine Luvox First-line First-line Level 1 150mg/day 300mg/day 300mg/day SSRI Sertraline Zoloft First-line First-line Level 1 50mg/day 200mg/day 250mg/day SSRI Paroxetine Paxil First-line First-line Level 1 40mg/day 60mg/day 60mg/day SNRI Venlafaxine Effexor First-line Second-line Level 2 150mg/day 225mg/day 375mg/day SSRI Citalopram Celexa First-line Second-line Level 2 20mg/day 60mg/day 60mg/day SSRI Escitalopram Cipralex First-line Third-line Level 4 10mg/day 20mg/day 50mg/day TCA Clomipramine Anafranil Second-line Second-line Level 1 150mg/day 250mg/day 250mg/day NaSSA Mirtazapine Remeron Second-line Second-line Level 2 15mg/day 45mg/day 45mg/day SSRI: selective serotonin reuptake inhibitor SNRI: serotonin-norepinephrine reuptake inhibitor TCA: tricyclic antidepressant NaSSA: noradrenergic and specific serotonergic antidepressant

Table 1.3. Treatment Response Categories in OCD.

Treatment Response Categories in OCD (Pallanti et al., 2002)

I Recovery with Y-BOCS scores ≤8 II Remission with Y-BOCS score ≤16 III Full response with ≥35% reduction in Y-BOCS score IV Partial response with 25-35% reduction in Y-BOCS score V Non-response (resistant) with <25% reduction in Y-BOCS score VI Relapase with symptoms return (25% increase in Y-BOCS score) after 3 months of "adequate" treatment or remission VII Refractory with no change (improvement or worsening) with any conventional therapy Treatment response categories using the Y-BOCS score was defined by Pallanti et al. in 2002.

1.3 Gender Differences in OCD

OCD affects both genders equally (APA, 2000), although a higher rate in male children and adolescents has been reported (Fineberg et al., 2013). Evidence suggests that gender may play an important role in the neurobiological aspects, psychosocial factors, and behavioural patterns of several psychiatric disorders. Recent clinical studies of OCD postulated that gender may moderate the genetic expression of this complex disorder, resulting in gender-specific phenotypes (Cherian et al., 2014; de Mathis et al., 2011; Torresan et al., 2013).

Gender-related differences exist in the clinical characteristics of this illness but results have been inconsistent across studies. The following gender differences have been reported in OCD studies (Cherian et al., 2014; de Mathis et al., 2011): clinical phenomenology (Cherian et al., 2014; de Mathis et al., 2011; Jaisoorya et al., 2009; Tukel et al., 2013), age at onset (AAO) (Cherian et al., 2014; de Mathis et al., 2011; Weissman et al., 1994), and co-morbidities (Cherian et al., 2014; Tukel et al., 2013). Men appear to have a higher frequency of aggressive (Tukel et al., 2013), sexual, religious, symmetry/exactness obsessions, in addition to checking, ordering/arranging, and hoarding compulsions, while obsessions related to contamination and washing/cleaning compulsions, as well as somatic obsessions and compulsions, appear to be more common among women (de Mathis et al., 2008; Jaisoorya et al., 2009; Rosso et al., 2012; Tukel et al., 2013). OCD males, especially those with early onset OCD, were reported to have a higher rate of comorbid tics and attention deficit hyperactivity disorder (Leckman et al., 1994). However, one study did not find gender-related phenomenological differences in OCD individuals (Matsunaga et al., 2000).

Additional clinical studies have also found that OCD females tend to have an abrupt onset of illness with episodic presentation following stressful life events, whereas males with OCD generally have an insidious onset with a more chronic course (de Mathis et al., 2011; Nestadt et al., 1998; Rosso et al., 2012) and higher symptom severity (Nestadt et al., 1998; Rosso et al., 2012).

1.4 Genetics of OCD

The biological and genetic underpinning of OCD was first reported in 1936, describing the heredity of “obsessional neurosis” (Lewis, 1936). Up to 2015, there have been over 500 peer-reviewed publications on the human genetics of OCD. Evidence supporting the genetic basis of OCD originates from family aggregation and twin studies (Pauls, 2010; Pauls et al., 2014), which have provided heritability estimates ranging from 27% to 47% in adults and 45% and 65% in children (van Grootheest et al., 2008). Although knowledge is very limited, there appear to be genetic differences between OCD and the other OCRDs. Twin and family studies have estimated a heritability ranging from 51.1% to 71.4% for hoarding (Monzani et al., 2013; Mathews et al., 2007) and 31.6% to 78% for trichotillomania (Monzani et al., 2014; Novak et al., 2009). Two studies have estimated that 40% and 44% of the vulnerability for skin picking and body dysmorphic concerns respectively was due to genetic factors (Monzani et al., 2012). However, notwithstanding the high genetic loading shared by these disorders, there may also be important genetic distinctions, which may influence medication response differentially across these conditions. For example, Monzani et al. (Monzani et al., 2014) identified a genetic vulnerability factor for trichotillomania and skin picking disorder that is distinct from the others; and hoarding has long been described as showing a pattern of relatively poor response to pharmacotherapy (Saxena et al., 2011).

Evidence from five genetic linkage analyses and over 100 genetic candidate gene association studies supports the relevance of several candidate regions and genes to OCD (Figure 1.1), with the most consistent reports being in the glutamatergic and serotonergic systems (Mathews et al., 2012; Pauls, 2010). The following sections will highlight the major findings from candidate gene studies of OCD.

region 9p24.2 /(5HT1.) glutamate transporter (SLC1A1) region 12q13.1 glutamate receptor, region 11q14.1 ionotropic, N-methyl brain-derived neurotrophic D-aspartate 2. factor (.DNC) (GRIN2.)

region 3q13.3 dopamine receptor D3 (DRD3) region 13q14.2 region 21q22.11 serotonin receptor 2A (5HT2A) region 17q11.2 oligodendrocyte lineage serotonin transporter transcription factor 2 (hLIG2) (SLC6A4)

Adapted from: Pato et al. (2002) J Clin Psychiatry 63(Suppl. 6):30-3

Figure 1.1. Genes Associated with OCD. This figure illustrates genes located in their associated chromosomal regions with at least one significant association with OCD. Copright permission has been given by Lundbeck Institute to use this figure. 10

Glutamatergic System:

Evidence implicates abnormality in the glutamatergic system as part of the etiology of OCD with the most robust genetic results coming from genes that are involved in this system. Support of its involvement comes from the most consistently implicated neuropathophysiological model of OCD, which postulates a disturbance in the cortico-striato-thalamo-cortical (CSTC) circuit (Pauls et al., 2014). Convergent evidence from animal and human research proposes an abnormal threshold of activation in this circuit, leading to hyperactivation of the orbitofrontal- subcortical pathway (Pauls et al., 2014). Research studies have postulated an alteration of glutamatergic neurotransmission in the CSTC circuit that involves the orbitofrontal cortex, which mediates exaggerated concerns related to danger, harm, symmetry, or hygiene as seen in OCD patients with obsessions, and this in turn is thought to generate compulsions in order to neutralize these obsessions (Pauls et al., 2014). Temporary relief of anxiety reinforces the behaviour in a positive feedback loop, which leads to eventual debilitating OCD symptoms. Further evidence implicating the glutamatergic system in the etiology of OCD comes from magnetic resonance spectroscopy studies, knock-out mouse models, genetic association studies, and clinical trials of glutamatergic agents (Wu et al., 2012).

The first published genome-wide association study (GWAS) in OCD by the International OCD Foundation Genetics Collaborative group did not detect genome-wide significant association between any tested markers and OCD diagnosis (Stewart et al., 2013). However, interesting trends were observed in several glutamatergic system genes including the ‘discs large (drosophila) homolog-associated protein 1’ (DLGAP1) and glutamate receptor, ionotropic, kainate 2 (GRIK2) genes. Although a second GWAS by the OCD Collaborative Genetics Association Study also did not find any genome-wide significant results (Mattheisen et al., 2015), the authors detected a trend in the top-hit marker on chromosome 9 near the protein- tyrosine phosphatase, receptor-type, delta (PTPRD) gene (P=4.13×10-7), which promotes glutamatergic synaptic differentiation. Despite a non-significant trend in a glutamatergic system gene, the neuronal glutamate transporter gene (SLC1A1) from a meta-analysis (Taylor, 2013), another recent but more in-depth meta-analysis by Stewart et al. (Stewart et al., 2013) of association between nine previously examined single nucleotide polymorphisms (SNPs) across

the 3’ region of the SLC1A1 gene [first reported by Arnold et al. (Arnold et al., 2006)] and OCD illness revealed a consistent moderately significant finding in one of the SNPs, rs301443 (P=0.046), with another SNP showing modest association when controlled for gender (rs12682897; P=0.012). Furthermore, another glutamate system gene, glutamate receptor, ionotropic N-methyl D-aspartate 2B (GRIN2B), has also been reported to have a significant association with the diagnosis of OCD (Arnold et al., 2004).

Serotonergic System:

The serotonergic system has been extensively studied in OCD given that SSRIs are the first-line pharmacological treatment of OCD based on clinical trials (CPA, 2006; Koran et al., 2007; Soomro et al., 2008). SSRIs increase the level of neurotransmitter serotonin in the synaptic cleft that binds to postsynaptic receptors by inhibiting its reuptake into presynaptic cells. This is a widely accepted mechanism of action of SSRIs that reduces depression and anxiety in a complex interaction with other neurotransmitter systems.

Genes related to the serotonergic neurotransmitter system have been examined extensively in relation to the genetic risk of developing OCD given its functional implication, the role of serotonin in the proposed neuro-circuitry, and the significance of selective serotonin reuptake inhibitors as the first-line pharmacological treatment of OCD. A recent meta-analysis of OCD genetic association studies (Taylor, 2013) showed nominally significant results for a serotonin transporter promoter (SLC6A4) polymorphism (HTTLPR; OR=1.251, 95% confidence interval [CI] 1.048-1.492, P=0.003) and the serotonin 2A receptor (HTR2A) rs6311 marker (OR=1.219, 95% CI 1.037-1.433, P=0.003).

Other Systems:

Interest in the dopaminergic system has come from the treatment of OCD using antipsychotics as augmenting agents to SSRI antidepressants in addition to dopamine interaction with the serotonergic system (Pauls et al., 2014). Further evidence from animal, neuroimaging,

and neurochemical studies implicates dopaminergic dysfunction in the pathoetiology of OCD (Koo et al., 2010). Although Taylor (2013) in his recent meta-analysis of OCD genetic association studies, showed a nominally significant finding for the catechol-O-methyltransferase (COMT) rs4680 marker (OR=1.200, 95% CI 1.001-1.438, P=0.010), and non-significant trends for two additional dopaminergic system genes, the dopamine transporter (DAT1) and dopamine D3 receptor (DRD3), no further genetic replication studies support the dopamine system in OCD. Genes involved in the gabaergic system have been investigated with inconsistent and non- replicated results (Taylor, 2013). Similar lack of support was reported from the brain-derived neurotrophic factor (BDNF) gene including the most studied Val66Met (rs6265) variant (Zai et al., 2015), as well as genes related to neuroplasticity (Taylor, 2013).

Candidate Genes Identified through Animal Studies: Although human behaviours are highly complex, important confounds are difficult to manipulate for ethical limitations when using invasive techniques in human subjects. Animal models provide the ideal tool to directly modify neural circuitry, in the expectation that it may lead to the manifestation of behaviours similar to those of human beings with OCD. Abnormalities within CSTC circuitry, which involves the orbitofrontal cortex, anterior cingulate cortex, and striatum, have been consistently replicated in OCD (Monteiro and Feng, 2015; Saxena and Rauch, 2000; Ting and Feng, 2011). Based on these solid findings, development of novel genetic animal models of OCD that can manipulate this circuit have been used to further our understanding of the etiology of OCD. Four genetic animal models have been reported: SAP90/PSD95-associated protein 3 (SAPAP3) null mice, SLIT and NRTK-like family, member 5 gene (SLITRK5) null mice, homeobox B8 (HOXB8) null mice, and SLC1A1/EAAC1 null mice. SAPAP3 encodes a protein, which is highly expressed in the striatum. It is involved in postsynaptic scaffolding at excitatory (glutamatergic) synapses. Evidence suggests that disruption of the glutamatergic neurotransmitter system in the CSTC circuitry plays a vital role in the etiology of OCD (Wu et al., 2012). Abnormal behaviours consistent with an increase in anxiety and compulsive self-grooming were observed in mice that had a deletion of this gene, SAPAP knock-out or null mice (Welch et al., 2007). More interestingly, the same study showed that these behaviours were alleviated with repeated administration of fluoxetine. Following this

discovery, three genetic association studies have examined the role of this gene in OCD. Bienvenu et al. (2009) tested the association of SAPAP3 in 383 families with grooming disorders and reported nominally significant association with at least one grooming disorders but not with OCD itself. Another study sequenced SAPAP3 in patients with TTM and OCD (77 TTM, 44 OCD+TTM, 44 OCD) and detected significantly higher percentage of novel nonsynonymous heterozygous variants than healthy controls (4.2% versus 1.1%) (Züchner et al., 2009). The last study investigated seven variants within SAPAP3 in 172 OCD patients, 45 TTM patients, and 153 healthy controls (Boardman et al., 2011) and observed an initial positive association between rs11583978 and TTM, which did not survive correction for multiple testing. Additionally, they detected significant association between earlier AAO and the A-T-A-T (rs11583978-rs7541937- rs6662980-rs4652867) haplotype when compared with the C-G-G-G haplotype, implicating this gene in the development of early onset OCD. SLITRK5 encodes for a protein that regulates neurite outgrowth and is important in neuronal survival. Shmelkov et al. (2010) knocked out SLITRK5 in mice, which then exhibited obsessive-compulsive-like behaviours including excessive self-grooming and increased anxiety in the open maze test. This study showed that these behaviours were once again reversed using chronic fluoxetine treatment, demonstrating similarities with the SAPAP3 null mice. Although SLITRK5 is widely expressed in the CNS, preferential increase of neuronal activity in the orbitofrontal cortex of the SLITRK5 null mice was observed, which is consistent with human OCD functional imaging findings (Grünblatt et al., 2014). This mouse model also showed a reduction in striatal volume in addition to a decrease in dendritic complexity (Shmelkov et al., 2010). Striatal dysfunction has been implicated in OCD (Grünblatt et al., 2014).

HOXB8 encodes a nuclear protein that functions as a sequence-specific transcription factor with an important role in establishing body patterning during embryonic development. HOXB8 knockout mice developed compulsive self-grooming behaviour and fur loss in addition to excessive grooming of wildtype (normal) caged-mice (Greer and Capecchi, 2002).

The last animal model has shown the greatest genetic contribution in OCD. The neuronal excitatory amino acid carrier 1 (EAAC1), which is encoded by SLC1A1, plays a vital function in transporting glutamate and regulating postsynaptic action of glutamate, which is essential in maintaining extracellular glutamate concentrations below oxidative stress and toxic levels

(Aoyama et al., 2006; Scimemi et al., 2009). Thirty percentage of EAAC1 null mice exhibited an increase in aggression in addition to excessive self-grooming and fur loss (Aoyama et al., 2006), which was reduced with N-acetyl-cysteine (NAC) treatment, a prodrug of glutathione that modulates the glutamatergic system (Aoyama et al., 2006; Berman et al., 2010). This finding further suggests the role of glutamate in the etiology of OCD and provides additional treatment avenues for this debilitating disorder.

Although these animal models have demonstrated compulsive grooming behaviours, further genetic studies in human are necessary to translate these intriguing findings that may help to understand the underlying mechanism and improve the treatment of OCD.

Summary:

In summary, these genetic findings in OCD are exciting but premature in resolving the heterogeneity of OCD. This is because of small sample sizes with limited statistical power, and difficulty in replication. Furthermore, many genes across different neurotransmitter systems confer only a small contribution to the risk of developing OCD, suggesting a polygenic model of liability (Browne et al., 2014).

The difficulty in generating replicable results in the genetics of OCD risk is partly due to the relatively high clinical heterogeneity of OCD; thus, researchers have undertaken strategies, aiming to reduce the phenotypic variability of OCD.

1.4.1 Genetics of Gender Differences in OCD

Evidence suggests important gender differences in human behaviour (i.e., emotion recognition, mentalizing, social behaviours) (Christov-Moore et al., 2014) and in psychiatric disorders (Arnold, 2003; Brady and Randall, 1999; Figueira and Ouakinin, 2010; Kornstein, 1997; Rucklidge, 2008). Although the rate of OCD is equal in males and females as stated earlier, studies have reported OCD males having an earlier onset (Cherian et al., 2014; de Mathis et al., 2011; Weissman et al., 1994), greater severity (Nestadt et al., 1998; Rosso et al., 2012), 15

and a more chronic course of OCD (de Mathis et al., 2011; Nestadt et al., 1998; Rosso et al., 2012) than females. Thus, gender may moderate the genetic expression of this complex illness, resulting in specific phenotypes with OCD males perhaps having a higher genetic loading than females. The most consistent findings for the genetics of a gender effect in OCD have been reported in SLC1A1 and COMT (Table 1.4; for further details, please refer to Appendix IV Table 1).

Although only four studies have examined SLC1A1, all showed significant results. Arnold et al. (2006) first reported a preferential transmission of rs301434 (P=0.0007-0.002), rs301435 (P=0.001-0.009), rs3087879 (P=0.002-0.006), and the rs301434-rs301435 haplotype (P=0.0006-0.001) to males and the total sample of 157 OCD families. Within the same year, Dickel et al. (2006) detected another SNP, rs3780412, was associated with early-onset males (P=0.002) in a sample of 71 OCD trios, which was replicated in another 57 OCD trios, over- transmission of rs3780412 C allele to males but not to females (permuted P=0.045) (Stewart et al., 2007). Shugart et al. (2009) later reported that another SNP, rs301443, was associated with OCD males (P=0.00027) in 378 OCD families. However, a recent meta-analysis of 3’ SNPs within this gene in males only reported no significant results (rs12682807; uncorrected P=0.012 but non-significant after correction) (Stewart et al., 2013).

Five studies investigated BDNF for gender effect in OCD and only one presented negative results. Hemmings et al. (2006) investigated multiple candidate genes in OCD phenotypes and specifically for gender, the authors reported significant findings for the rs6265 marker in OCD males (P=0.036 – total sample of 132 OCD and 218 controls). This polymorphism was further examined and results were as follows: the Met allele conferring OCD risk to males (P=0.047) (Hemmings et al., 2008), Met/Met genotype associating with later onset OCD in females (P=0.031) (Katerberg et al., 2009), and Val/Val genotype predicting more severe OCD symptoms in OCD females (P=0.031) (Hemmings et al., 2008); however, one study failed to report positive findings for this marker (Márquez et al., 2013). Nonetheless, the latter study examined additional SNPs and found that the rs7124442 C allele was associated with OCD females (P=0.018). Alonso et al. (2008) tested eight SNPs within BDNF including rs6265 in addition to 46 SNPs within the neurotrophic tyrosine kinase, receptor, type 2 (NTRK2) gene and

only reported association between the NTRK2 rs2378672 and OCD females (P<0.0001) but not with BDNF in a case-control study comprising of 215 OCD patients and 342 controls.

The most investigated gene is COMT with seven positive and three negative studies in addition to a meta-analysis. Karayiorgou et al. (1997) was the first group to report an increased frequency of the rs4680 Met allele and Met/Met genotype in OCD males (P=0.0002), which was replicated in three case-control studies (Denys et al., 2006; Pooley et al., 2007; Poyurovsky et al., 2005) in addition to a meta-analysis (P<0.0001) (Pooley et al., 2007); however, two later studies detected significant association of the same Met allele with OCD females (P=0.048 with 56 trios) (Alsobrook et al., 2002) (P=0.039) (Katerberg et al., 2010). Two studies detected positive results in the total OCD sample but did not observed differences in gender (Kinnear et al., 2001; Schindler et al., 2000) and Hemmings et al. (2006) only reported an association between rs362204 and AAO in OCD males (P=0.040). Liu et al. (2011) only published data for the rs2075507 SNP and demonstrated that the G/G genotype was more frequent in OCD males (P<0.001) and OCD females (P=0.026) than in healthy controls; however, there was no difference between the gender groups.

1.4.2 Genetics of OCD Subphenotypes

Clinical, epidemiological, treatment, interventional, and genetic studies have consistently observed the heterogeneous nature of OCD (Pauls et al., 2014). Previous research has attempted to define a more phenotypically homogeneous subtype or subgroup of OCD (de Mathis et al., 2006) in order to identify specific genetic and environmental factors that influence the risk of developing OCD.

Evidence suggests that OCD is a clinically and genetically heterogeneous disorder and researchers have focused on the identification of valid subgroups of OCD (Leckman et al., 2010). The examined subgroups include early onset, gender effects (male versus female), symptom dimensions (different OCD symptoms based on Y-BOCS symptom checklist), neuropsychological performance (i.e., cognitive flexibility), neuroimaging data (i.e., caudate, orbitofrontal cortex, anterior cingulate cortex), and treatment response (mainly SRIs). To date,

only two such subtypes have been accepted into the DSM-5 and these specifiers are the degree of insight and the presence of a lifetime diagnosis of tic disorder (APA, 2013). The following sub- sections describe the different OCD phenotypes that were examined in genetic association studies within this thesis: age at onset (AAO), symptom dimensions and severity, psychiatric comorbidities, familiality, and SRI treatment response.

Table 1.4. Summary Table of Genetic Studies of Gender in OCD.

Gene Results Reference GRIK2 - Mas et al., 2014 GRIN2B -- Hemmings et al., 2006; Mas et al., 2014 GRIA1 - Mas et al., 2014 GRIA3 - Mas et al., 2014 SLC1A1 ++++/- Arnold et al., 2006; Dickel et al., 2006; Mas et al., 2014; Shugart et al., 2009; Stewart et al., 2007 DLGAP3 - Mas et al., 2014 GAD1 - Mas et al., 2014 GAD2 + Mas et al., 2014 Baca-Garćia et al., 2005; Dickel et al., 2007; Liu et al., 2011; Lochner et al., 2004; Mas et al., 2014; Voyiaziaki SLC6A4 ++/--- et al., 2011 HTR1B +++/- Dickel et al., 2007; Kim et al., 2009; Lochner et al., 2004; Mas et al., 2014 HTR2A ++/--- Dickel et al., 2007; Enoch et al., 2001; Hemmings et al., 2006; Mas et al., 2014; Walitza et al., 2002 HTR2C +/-- Cavallini et al., 1998; Hemmings et al., 2006; Mas et al., 2014 HTR6 - Hemmings et al., 2006 SLC18A1 - Mas et al., 2014 TPH2 - Mas et al., 2014 DAT/SLC6A3 +/---- Billett et al., 1998; Hemmings et al., 2006; Lochner et al., 2004; Mas et al., 2014; Miguita et al., 2007 DRD1 - Hemmings et al., 2006 DRD2 +/--- Billett et al., 1998; Denys et al., 2006; Hemmings et al., 2006; Mas et al., 2014 DRD3 --- Billett et al., 1998; Hemmings et al., 2006; Mas et al., 2014 DRD4 +++/- Billett et al., 1998; Camarena et al., 2007; Hemmings et al., 2006; Taj et al., 2013 Alsobrook et al., 2002; Denys et al., 2006; Hemmings et al., 2006; Karayiorgou et al., 1997; Katerberg et al., +++++++/---- (+ COMT 2010; Kinnear et al., 2001; Liu et al., 2011; Lochner et al., 2004; Mas et al., 2014; Pooley et al., 2007; meta-analysis) Poyurovsk et al., 2005; Schindler et al., 2000 MAOA ++/- Camarena et al., 2001; Lochner et al., 2004; Mas et al., 2014 Alonso et al., 2008; Dickel et al., 2007; Hemmings et al., 2006; Hemmings et al., 2008; Katerberg et al., 2009; BDNF ++++/--- Márquez et al., 2013; Mas et al., 2014 NTRK2 +/- Alonso et al., 2008; Mas et al., 2014 NTRK3 - Mas et al., 2014 AKT1 - Mas et al., 2014 GSK3β - Mas et al., 2014 NGFR - Mas et al., 2014

ERBB4 - Mas et al., 2014 NRG1 - Mas et al., 2014 OLIG1 - Mas et al., 2014 OLIG2 - Mas et al., 2014 LMX1A - Mas et al., 2014 BDKRB2 - Mas et al., 2014 CDH9 - Mas et al., 2014 KCNN3 - Mas et al., 2014 EFNA5 - Mas et al., 2014 TH - Lochner et al., 2004 ACE + Hemmings et al., 2006 ESRα - Hemmings et al., 2006 INPP-1 - Hemmings et al., 2006 PLC-γ1 - Hemmings et al., 2006 HOXB8 - Hemmings et al., 2006 + indicates significant study - indicates negative study

1.4.2.1 The Significance of Age at Onset (AAO) in OCD

The consensus of age at onset (AAO) in OCD is usually between the ages of 10 and 37 years although two thirds of OCD cases had emerged by the age of 22 years (Fineberg et al., 2013).

Most research studies to date have arbitrarily and subjectively defined early onset as less than 10 to 18 years depending on the research groups (Chabane et al., 2005; Janowitz et al., 2009; Maina et al., 2008) and late onset as on or after age 17 to 30 years (Grant et al., 2007; Maina et al., 2008); however, many did not examine AAO closely to calculate the early versus late onset cutoff using analytical approaches. Furthermore, some studies have defined AAO as age at symptom onset (Butwicka and Gmitrowicz, 2010; de Mathis et al., 2009), while some set AAO as age when first diagnosed (Maina et al., 2008; Taylor, 2011; Tukel et al., 2005). The arbitrary and different definitions of AAO used in previous studies will give rise to further clinical heterogeneity and pose significant challenges in comparing different datasets and in combining them in order to increase sample size.

Several studies have since performed the admixture analysis to identify various normal distribution curves within the dataset. This provides an objective method to calculate the exact point where the normal distribution curves meet. Only four such studies have used admixture analysis to determine the best-fitting model for AAO in the OCD population. The first study was conducted by Delorme et al. (2005) in 161 OCD subjects and the admixture analysis detected two distributions with mean ages of 11.1 years for early-onset and 23.5 years as late-onset. Delorme and colleagues (2005) noticed significantly greater Tourette’s syndrome and family history of OCD in early-onset group and higher prevalence of generalized anxiety disorder and MDD in the late-onset group. This study supported the notion that early- and late-onset groups may exhibit different phenotypic characteristics. The second study used mixture analysis in a subset of the current OCD sample (N=196), which was conducted by our group (De Luca et al., 2011). De Luca et al. (2011) detected two normal distributions with a cut-off at 15 years and found that the early-onset OCD patients were assessed at an earlier age and have more frequent panic attacks than the late-onset subjects. The third study performed admixture analysis in 377 OCD individuals and also reported a bimodal distribution for AAO with the best-fitting cut-off

AAO for early-onset as less than or equal to 19 years (Anholt et al., 2014). Anholt et al. (2014) found that the early-onset group has higher OCD symptom severity across all OCD symptom dimensions in addition to greater ADHD symptoms and higher rates of comorbid bipolar disorder. The authors suggested that the early-onset group may be associated with greater functional impairment and poor clinical prognosis. A very recent paper examined AAO using two different definitions, onset when symptom first appeared and onset when symptom first met diagnostic criteria of OCD (Albert et al., 2015). The authors conducted admixture analysis in 483 OCD subjects with two different sets of AAO. They identified a trimodal distribution for age at symptom onset with mean ages of 6.9 years for early-onset, 14.99 years for the intermediate onset, and 27.7 years for the late-onset group; and reported early- and intermediate- onset as having an insidious onset (rather than acute onset) and a chronic course of illness (rather than episodic) when comparing to the late-onset group. For the age at diagnosis, they confirmed a bimodal distribution with means 18.0 years and 29.5 years. Early-onset group for both AAO definitions was found to have a higher prevalence of younger male subjects and greater severity and rate of symptoms related to religion, symmetry or exactness, contamination and cleaning/washing, in addition to repeating and counting.

A review and meta-analysis (Taylor, 2011) of clinical characteristics of early- versus late- onset OCD indicated that early-onset OCD: a) occurs more often in males (de Mathis et al., 2011; Jaisoorya et al., 2009; Narayanaswamy et al., 2012; Nestadt et al., 1998; Rasmussen and Eisen, 1992); b) associates with higher level of symptom severity and greater prevalence of types of obsessive-compulsive symptoms; c) has more comorbidities including tics and other OCRDs; and associates with a higher prevalence of OCD in first-degree relatives. A recent multi-center study indicated that OCD patients with earlier AAO had significantly longer duration of illness and higher disability in work, social, and family life (P=0.017-0.035) (Dell’Osso et al., 2013). Nakamae (2011) reported individuals with early-onset OCD appear to be more SRI treatment resistant when compared to later onset.

The rationale for separating out the early onset group is that the heritability of childhood OCD is higher than the adult OCD population, ranging from 27% to 47% in adults and 45% and 65% in children (van Grootheest et al., 2008). The concept of early-onset disease as having a more familial pattern of genetic influence over later onset of the same illness is well-documented

[i.e., bipolar disorder (Rende et al., 2007; Strober et al., 1988), major depressive disorder (Kendler et al., 2007; Levinson et al., 2003; Sullivan et al., 2000; Weissman et al., 1984), mood disorders (Orvaschel, 1990)]. Furthermore, specifically in OCD, earlier onset was postulated to have a higher familial and genetic loading than late onset OCD (Chabane et al., 2005; do Rosario-Campos et al., 2005; Hemmings et al., 2004). Thus, by examining the genetics of different AAO groups as identified using the admixture methodology, we may have a better chance of identifying genetic markers, which may predispose to early rather than late onset OCD.

Mixed and inconsistent genetic results have been reported across different neurotransmitter system candidate genes (Table 1.5; for further details, please refer to Appendix IV Table 1). Mathews et al. (2012) published the only genome-wide linkage study in 33 childhood-onset OCD families and reported a LOD score linkage of greater than two on chromosome 1p36, 2p14, 5q13, 6p25, and 10p13. Although 47 genes have been explored in relation to AAO in OCD, most were single study or with very inconsistent and conflicting findings (Table 1.5). Genes described below are considered to have a more robust result. Two significant meta-analyses examined SLC6A4 5HTTLPR (Bloch et al., 2008) and reported an over-transmission of the La allele to early-onset OCD (P=0.00021) (Walitza et al., 2014) when almost all the individual samples were too small to detect any positive results. In addition to SLC6A4, HTR2A was investigated in nine published articles with mixed results (Table 1.5 and for additional details of each study, please refer to Appendix IV Table 1). Only one study examined linkage and reported a LOD score of two on chromosome 9q where SLC1A1 is located (Hanna et al., 2002) and three out of five studies reported significant findings of different SNPs that were associated with AAO [rs3780412 and rs301430 with early-onset (P=0.04) (Dickel et al., 2006); rs10491734 A allele (P=0.023) and A/A genotype (P=0.009) with early onset (Wu et al., 2013); rs301430 C/C genotype with later onset (P=0.017) (Dallaspezia et al., 2014)]. Regarding BDNF, the first significant study with several SNPs was reported by Hall et al. (2003) and only two of eight subsequent studies supported the role of the rs6265 marker in AAO but with gender bias (Hemmings et al., 2006; Katerberg et al., 2009). Thus, with such scattered genetic findings, it is important to replicate some of these findings in a much larger sample of OCD, which can be stratified into AAO groups.

Table 1.5. Summary Table of Genetic Studies of AAO in OCD.

Gene Results Reference GRIK2 - Mas et al., 2014 GRIN2B +/- Hemmings et al., 2006; Mas et al., 2014 GRIA1 - Mas et al., 2014 GRIA3 - Mas et al., 2014 SLC1A1 +++/-- Dallaspezia et al., 2014; Dickel et al., 2006; Mas et al., 2014; Veenstra-VanderWeele et al., 2001; Wu et al., 2013 DLGAP3 - Mas et al., 2014 GAD1 - Mas et al., 2014 GAD2 + Mas et al., 2014 SLC6A4 (++ meta-analyses)/-- Bloch et al., 2008 (meta-analysis); Denys et al., 2006; Dickel et al., 2007; Liu et al., 2011; Lochner et al., ------2004; Lochner et al., 2008; Mas et al., 2014; Walitza et al., 2004; Walitza et al., 2014 HTR1B +/------Denys et al., 2006; Dickel et al., 2007; Hemmings et al., 2004; Hemmings et al., 2006; Lochner et al., 2004; Lochner et al., 2008; Mas et al., 2014; Wendland et al., 2007 HTR2A ++++/----- Denys et al., 2006; Dickel et al., 2007; Hemmings et al., 2004; Hemmings et al., 2006; Lochner et al., 2004; Lochner et al., 2008; Mas et al., 2014; Walitza et al., 2002; Walitza et al., 2012 HTR2C -- Hemmings et al., 2006; Mas et al., 2014 HTR6 - Hemmings et al., 2006 SLC18A1 - Mas et al., 2014 TPH1 - Walitza et al., 2004 TPH2 -- Mossner et al., 2006; Mas et al., 2014 DAT/SLC6A3 +/------Hemmings et al., 2004; Hemmings et al., 2006; Lochner et al., 2004; Lochner et al., 2008; Mas et al., 2014; Miguita et al., 2007; Walitza et al., 2008 DRD1 - Hemmings et al., 2006 DRD2 - Mas et al., 2014 DRD3 +/- Hemmings et al., 2006; Mas et al., 2014 DRD4 ++/---- Hemmings et al., 2004; Hemmings et al., 2006; Lochner et al., 2004; Lochner et al., 2008; Mas et al., 2014; Walitza et al., 2008 COMT ++/----- Hemmings et al., 2004; Hemmings et al., 2006; Liu et al., 2011; Lochner et al., 2004; Lochner et al., 2008; Mas et al., 2014; Walitza et al., 2008 MAOA ----- Camarena et al., 2001; Hemmings et al., 2006; Lochner et al., 2004; Lochner et al., 2008; Mas et al., 2014 BDNF +++/------Dickel et al., 2007; Hall et al., 2003; Hemmings et al., 2006; Hemmings et al., 2008; Katerberg et al., 2009; Márquez et al., 2013; Mas et al., 2014; Mossner et al., 2005; Wendland et al., 2007 NTRK2 - Mas et al., 2014

NTRK3 - Mas et al., 2014 AKT1 - Mas et al., 2014 GSK3β - Mas et al., 2014 NGFR - Mas et al., 2014 ERBB4 - Mas et al., 2014 NRG1 - Mas et al., 2014 OLIG1 - Mas et al., 2014 OLIG2 - Mas et al., 2014 LMX1A - Mas et al., 2014 BDKRB2 - Mas et al., 2014 CDH9 - Mas et al., 2014 KCNN3 - Mas et al., 2014 EFNA5 - Mas et al., 2014 SAPAP3 - Mas et al., 2014 TNFα + Lüleyap et al., 2012 NPSR1 + Lennertz et al., 2013 TH -- Lochner et al., 2004; Lochner et al., 2008 ACE - Hemmings et al., 2006 ESRα - Hemmings et al., 2006 INPP-1 - Hemmings et al., 2006 PLC-γ1 + Hemmings et al., 2006 HOXB8 + Hemmings et al., 2006 + indicates significant study - indicates negative study

1.4.2.2 Clinical Heterogeneity of OCD: Symptom Dimensions and Severity

One of the main challenges in OCD is the large number of symptom presentations and the varying degree of symptom severity in patients who suffer from OCD. The following section will provide rationale for exploring these subphenotypes in genetics.

1.4.2.2.1 OCD Symptom Dimensions

The majority of OCD individuals endorse both obsessions and compulsions whereas approximately 25% and 5% have obsessions and compulsions only respectively (APA, 2010).

The National Comorbidity Survey Replication epidemiological study reported distinct symptoms of OCD including checking (79.3%), hoarding (62.3%), ordering (57.0%), moral concerns (43.0%), sexual/religious concerns (30.2%), contamination (25.7%), harming (24.2%), concerns related to illness (14.3%), and others (19.0%), with a high 81% of individuals endorsing multiple symptoms (Ruscio et al., 2010).

Given the vast number of possible OCD symptoms for the diagnosis of OCD, each individual with OCD may present very differently. In order to capture all of the OCD symptoms and to avoid over-simplification, factor analysis has been conducted to reduce the number of OCD symptoms. A factor analysis is used to group variables, in this case, different OCD symptoms, according to shared variance. OCD symptoms that are explained by the same variance are grouped together into each factor, symptom dimension. A different approach is the Ward’s method in the hierarchical cluster analysis to join variables (i.e., OCD symptoms) into clusters such that the variance within a cluster is minimized. Each variable begins as its own cluster and clusters are then merged in such a way as to reduce the variability within a cluster if the merging process results in the minimum increase in the error sum of squares. By reducing multiple OCD symptoms into several symptom dimensions, it aids clinician to interpret clinical outcome easier and helps genetic research to increase statistical power by limiting symptom stratification.

Several validated scales have been developed to capture OCD symptom dimensions. The Yale-Brown Obsessive-Compulsive Scale (Y-BOCS) severity score and symptom checklist is considered to be the gold standard to assess for OCD symptom severity and it is widely used to generate symptom dimensions (Goodman et al., 1989). Previous studies have consistently identified a four- or five-factor of symptom dimensions using the factor analysis based on principal component analysis (PCA) of the Y-BOCS symptom checklist at a category- or item- level including symmetry/ordering, hoarding, contamination/cleaning, aggressive/checking+/- sexual/religious concerns (Table 1.1).

The first study to use statistical approach to reduce the number of Y-BOCS symptom category was conducted in 1994 (Baer et al., 1994) and the 3-factor model explained 48% of the variance for symmetry/hoarding, contamination/cleaning, and pure obsessions. Since 1994, over 30 studies (Table 1.6, Table 1.7, Table 1.8; for full details, please refer to Appendix IV Table 2) have utilized the data reduction technique to calculate the minimal OCD symptom categories for comparing clinical features and treatment outcome. Eighteen studies have identified a 5-factor model for the Y-BOCS symptom checklist and the most common factors include contamination/cleaning, hoarding, aggressive/checking, symmetry/ordering, and sexual/religious (Table 1.6). Foa et al. (2002) used the Obsessive-Compulsive Inventory (Foa et al., 2002), which is a self-report measure of obsessive-compulsive traits with 18 items and six subscales, and the same five factors were reported. The second most common is the 4-factor model with aggressive/checking/sexual, hoarding, symmetry/ordering/counting, and contamination/cleaning as the most consistently identified factors (Table 1.7). The highest percentage of variance that explained the factor model was reported by Cullen et al. (2007) at 76.7% with the following factors: pure obsessions, contamination, symmetry/order, and hoarding. A meta-analysis, which included the above study, consisting of 5,124 OCD individuals, confirmed a four-factor model that explained 79% of the variance (Bloch et al., 2008). Four studies detected three factors and two reported six factors; moreover, the number of identified factors changes with the use of individual item of OCD symptoms versus using category of OCD symptoms (Table 1.8).

Other validated scales such as the Dimensional Y-BOCS (Rosario-Campos et al., 2006), the self-report Dimensional Obsessive-Compulsive Scale (Abramowitz et al., 2010), the self- report Obsessive-Compulsive Inventory (Foa et al., 2002), the self-report Clark-Beck Obsessive-

Compulsive Inventory (Clark et al., 2005), the self-report Maudsley Obsessional Compulsive Inventory (Hodgson et al., 1977), the self-report Padua Inventory (Sanavio et al., 1988), the self- report Padua Inventory-Revised (Van Oppen et al., 1995), the self-report Vancouver Obsessive Compulsive Inventory (Thordarson et al., 2004), and the self-report Obsessional Intrusive Thoughts Inventory (García-Soriano et al., 2011) have been used but either due to the length or the over-generalization of these inventories, Y-BOCS remains the most extensively utilized in the clinical OCD population.

Symptom dimensions are of clinical utility given that they respond differently to various treatments. Bloch et al. (Bloch et al., 2014) recently conducted a meta-analysis reporting poor treatment outcome in OCD patients with hoarding symptoms. Treatment implication for different symptom subtypes have previously been reported (Nakamae, 2011). OCD individuals with sexual, religious, and harm-related (aggressive) obsessions and checking compulsions appeared to have improved SRI response (>60% with very much improved) (Landeros-Weisenberger et al., 2010), whereas individuals with hoarding symptoms tend to be SRI resistant (Nakamae, 2011).

Studies have also reported clinical differences among different types of symptom dimension. Several studies indicated that somatic and contamination obsessions and washing/cleaning compulsions occur more frequently in female OCD patients (de Mathis et al., 2008; Jaisoorya et al., 2009; Rosso et al., 2012; Tukel et al., 2013), whereas males appear to have a higher frequency of aggressive (Tukel et al., 2013), sexual, religious, symmetry/exactness obsessions, in addition to checking, ordering/arranging, and hoarding compulsions (de Mathis et al., 2008; Jaisoorya et al., 2009; Rosso et al., 2012; Tukel et al., 2013).

A twin study of non-clinical obsessive-compulsive dimensional traits found that the hoarding dimension had the lowest common genetic factor loading and was 54.5% influenced by specific genetic effects (Iervolino et al., 2011); this confirms the separation of hoarding away from the OCD diagnostic criteria in DSM-5 (APA, 2013). Another study identified two statistically separate OCD subgroups according to OCD symptoms (Schooler et al., 2008). Schooler et al. (2008) reported that one group of OCD individuals had significantly higher rates of OCD in the families, earlier onset and more severe OCD symptoms, in addition to more

psychiatric comorbidities. Additional family studies reported different heritability for specific OCD symptom dimensions including significant correlations between relatives in the dimensions of hoarding (Hasler et al., 2005; Katerberg et al., 2010; Pinto et al., 2007), unacceptable/taboo thoughts (Hasler et al., 2005; Katerberg et al., 2010; Pinto et al., 2007), contamination and symmetry (Katerberg et al., 2010), in addition to checking compulsions (Hasler et al., 2005).

Several studies have examined the genetics of OCD symptom dimensions (Table 1.9; for further details, please refer to Appendix IV Table 1) and OCD symptom severity (Table 1.10; ; for further details, please refer to Appendix IV Table 1) but these studies were mostly under- powered and reported inconsistent results with no replication. Thirteen genetic studies have examined Y-BOCS symptom checklist dimensions in our search for published articles. The most consistent results were reported in the COMT rs4680 SNP, which has been found to be associated with hoarding (Hemmings et al., 2006), obsessional/checking dimension (Lochner et al., 2008), and somatic and sensory phenomena (Katerberg et al., 2010); nonetheless, Liu et al. (2011) failed to replicate these findings. Although two studies detected significant associations between SLC6A4 5HTTLPR and similar factor, repeating/counting and symmetry/repeating/counting/ordering, their results implicated opposite genotypes, L/L genotype (P=0.0013) from the study conducted by Cavallini et al. (2002) and S/S genotype by Hasler et al. (2006). Hemmings et al. (2006) demonstrated that the DRD4 rs1800955 and ESRα rs9340799 were associated with hoarding but two studies have failed to replicate the DRD4 results (Lochner et al., 2008; Taj et al., 2013) and Alonso et al. (2012) reported a different marker, rs34535804, predicting lower risk of having contamination/cleaning factor symptoms (P=0.0001).

We aimed to investigate the effect of OCD candidate genes in Y-BOCS symptom checklist generated symptom dimensions because: 1) Limited number of published genetic studies (13 in total) have explored Y- BOCS symptom dimensions 2) Previous studies investigated commonly examined markers within the exons of the candidate genes or SNPs with no known functional role 3) Studies all have relatively small sample size 4) Results were inconsistent and have not been replicable.

5) Reducing clinical heterogeneity may identify risk genetic variations with strong effect size. 6) Biological mechanisms for symptom dimensions may be different (i.e., as in the case of hoarding symptom, which is now considered a distinct disorder) 7) Diagnostic criteria and/or definition of OCD may change over time.

Table 1.6. Summary Table of 5-Factor Y-BOCS Symptom Dimensions in OCD.

Variance Study Scoring * Analysis Technique Number 5 Factors Explained Contamination/cleaning/repeating, Tek and symmetry/ordering/somatic, Range=0-1 PCA, current symptoms 45 OCD 65.5% Ulug, 2001 aggressive/counting/hoarding, sexual/religious, checking/hoarding 180 OCD, Contamination/cleaning, hoarding, Cavallini et PCA, (category-level), lifetime Range=0-1 112 sexual/religious/aggressive/checking, 59.87% al., 2002 symptoms controls symmetry/ordering, repeating/counting Symmetry/ordering, hoarding, Mataix-Cols Range=0-2 PCA, category-level 117 OCD contamination/cleaning, aggressive/checking, N/A et al., 2002 sexual/religious Mataix-Cols PCA, category-level, current Aggressive/checking, contamination/washing, Range=0-1 153 OCD 63.7% et al., 2002 symptoms symmetry/ordering, hoarding, sexual/somatic Range=0-4 Symmetry/ordering, hoarding, Foa et al., (Obsessive- Confirmatory PCA, current 215 OCD contamination/cleaning, aggressive/checking, N/A 2002 Compulsive symptoms sexual/religious obsessions Inventory) Aggressive/sexual/religious obsessions, Denys et PCA, item-level, current Range=0-2 150 OCD contamination/cleaning, somatic/checking, 42.5% al., 2004 symptoms symmetry/ordering, high risk aggressive/checking Contamination/cleaning, Denys et PCA, item-level, current aggressive/sexual/religious/somatic, somatic Range=0-2 335 OCD 41.7% al., 2004 symptoms obsessions and checking, symmetry and prefectionism, high risk aggressive and checking Symmetry/ordering, hoarding, Pinto et al., Range=0-1 and Exploratory PCA, category- 293 OCD doubt/somatic/checking, contamination/cleaning, 65.5% 2007 0-2 level taboo thoughts (aggressive/sexual/religious) Contamination/cleaning, harm/checking, Stein et al., Exploratory PCA, item-level, Range=0-2 434 OCD aggressive/sexual/religious, hoarding/symmetry, N/A 2007 current symptoms somatic/hypochondriacal Pinto et al., PCA, item-level, lifetime Taboo thoughts, symmetry/ordering, hoarding, Range=0-1 485 OCD N/A 2008 symptoms contamination/cleaning, doubt/checking

PCA, categorical-level, lifetime Symmetry/ordering, taboo thoughts, hoarding, 64.1% symptoms doubt/checking, contamination/cleaning

Contamination/cleaning, harm/checking, Stein et al., Range=0-2 Exploratory PCA, item-level 466 OCD hoarding/symmetry, religious/sexual, N/A 2008 somatic/hypochondriacal

Y-BOCS and Thoughts and Behaviour Inventory (TBI; Schooler et Fear of hurting self/others, repetition/symmetry, Range=0-1 Slattery et al., 2004), 398 OCD N/A al., 2008 contamination, checking/counting, hoarding confirmatory PCA, item-level, lifetime symptoms

Contamination/cleaning, hoarding, Jang et al., symmetry/ordering, Range=0-2 PCA, current symptoms 144 OCD 65.41% 2010 aggressive/sexual/religious/checking, repeating/counting 200 OCD, Hoarding, contamination/cleaning, Liu et al., PCA, category-level, current Range=0-1 403 symmetry/ordering, aggressive/checking, 71.8% 2011 symptoms controls somatic/repeating Doubts/checking/repeating, Cherian et contamination/washing/cleaning, hoarding/collecting, N/A PCA, category-level 545 OCD 62% al., 2012 symmetry/ordering, forbidden/sexual/religious/aggressive Brakoulias Hoarding, contamination/cleaning, doubt/checking, Range=0-2 PCA, category-level 154 OCD 64.9% et al., 2013 symmetry/ordering, unacceptable/taboo thoughts 173 OCD, Taj et al., Hoarding, doubts/checking, symmetry/ordering, N/A PCA, category-level 201 65% 2013 contamination, forbidden thoughts controls Hoarding, contamination/cleaning, Brakoulias N/A PCA, category-level 154 OCD symmetry/ordering, unacceptable/taboo thoughts, 64.9% et al., 2014 doubt/checking Doubts about harm/checking, unacceptable/taboo Williams et Exploratory PCA, category- Range=0-2 238 OCD thoughts, contamination/cleaning, hoarding, 68.5% al., 2014 level, current symptoms symmetry/ordering

Multidimensional item 269 OCD, Martoni et response theory model, Hoarding, washing, symmetry, rituals, forbidden Range=0-1 120 N/A al., 2015 category-level, current thoughts controls symptoms

Scoring: investigators assigned a score of 1 if a symptom category was present, 0 if it was absent, and 2 was assigned in some studies for the most prominent symptom category PCA: principal component analysis N/A: not applicable Highlighted in grey indicates the use of other OCD symptom scales than Y-BOCS symptom checklist * Scoring of 0 indicates the absence of symptom; scoring of 1 indicates the presence of symptom; scoring of 2 indicates the presence of target or principal symptom

Table 1.7. Summary Table of 4-Factor Y-BOCS Symptom Dimensions in OCD.

Analysis Variance Study Scoring * Number 4 Factors Technique Explained PCA, Number Leckman et category-level, 2 OCD samples Obsessions/checking, symmetry/ordering, cleanliness/washing, of 63.5% al., 1997 lifetime (208 and 98) hoarding symptoms symptoms PCA, current Aggressive/sexual/religious/somatic/checking, Summerfeldt Range=0- symptoms, 203 OCD symmetry/repeating/ordering/counting, contamination and N/A et al., 1997 1 including cleaning/washing, hoarding miscellaneous PCA as per Number 100 OCD, 466 Alsobrook II Leckman et al. Aggressive/checking/sexual, hoarding, of first-degree ~70% et al., 1999 (1997), symmetry/ordering/counting/ritual, contamination symptoms relatives category-level Confirmatory PCA, Aggressive/sexual/religious/somatic/checking, Summerfeldt Range=0- category-level, 203 OCD symmetry/repeating/ordering/counting, contamination and N/A et al., 1999 1 current cleaning/washing, hoarding symptoms PCA, item- Symmetry/exactness/ordering/arranging/repeating/counting/touching, Feinstein et Range=0- level, current 160 OCD contamination/cleaning/aggressive/checking, hoarding, 54.1% al., 2003 1 baseline sexual/religious symptoms PCA, Aggressive/sexual/religious/checking, Hasler et al., Range=0- category-level, 169 OCD symmetry/repeating/counting/ordering/arranging, 64.7% 2005 1 lifetime contamination/cleaning, hoarding symptoms Kim et al., Range=0- PCA, 124 OCD, 171 Hoarding/repeating, contamination/cleaning, aggressive/sexual, 62.8% 2005 2 category-level controls religious/somatic PCA, Hasler et al., Range=0- category-level, Aggressive/religious/checking, symmetry/repeating/counting/ordering, 153 OCD 65.0% 2006 1 lifetime contamination/cleaning, hoarding symptoms PCA, 221 Cullen et al., Range=0- category-level, definite/probable Pure obsessions, contamination, symmetry/order, hoarding 76.7% 2007 1 lifetime OCD 35

symptoms

Wendland et Range=0- PCA, 347 OCD, 749 Obsessions/checking, symmetry/ordering, contamination/cleaning, N/A al., 2007 1 category-level controls hoarding PCA and cluster Aggressive/sexual/religious/somatic/checking, Hasler et al., Range=0- analysis, 418 OCD symmetry/repeating/counting/ordering/arranging, 64.0% 2007 1 category-level, contamination/cleaning, hoarding lifetime symptoms PCA, Matsunaga Range=0- category-level, Contamination/washing, hoarding, symmetry/repeating/ordering, 343 OCD 57.7% et al., 2008 2 current aggressive/checking symptoms Albert et al., Range=0- PCA, lifetime Symmetry/repeating/ordering/counting/checking, 329 OCD 58% 2010 1 symptoms aggressive/religious/sexual/somatic, contamination/cleaning, hoarding

58 OCD trios, Lennertz et Range=0- PCA, Symmetry/ordering/repeating/counting, contamination/washing, 236 OCD, 310 53.9% al., 2014 1 category-level hoarding, aggressive/sexual/religious controls Scoring: investigators assigned a score of 1 if a symptom category was present, 0 if it was absent, and 2 was assigned in some studies for the most prominent symptom category PCA: principal component analysis N/A: not applicable * Scoring of 0 indicates the absence of symptom; scoring of 1 indicates the presence of symptom; scoring of 2 indicates the presence of target or principal symptom

Table 1.8. Summary Table of 3-, 6-, and Mixed-Factor Y-BOCS Symptom Dimensions in OCD.

Analysis Variance Reference Scoring Number 3, 6, or Mixed Factors Technique Explained PCA, category- Baer et al., Range=0- Symmetry and hoarding, contamination and cleaning, pure level, current 107 OCD 48% 1994 2 obsessions symptoms Hantouche PCA, category- and N/A level, current 615 OCD Predominantly compulsive, predominately obsessive, mixed 32.5% Lancrenon, symptoms 1996 53 OCD, 96 non- Confirmatory Contamination/cleaning/repeating/checking/somatic, Wu et al., Range=0- OCD psychiatric factor analysis, aggressive/sexual/religious, N/A 2007 1 patients, 419 category-level hoarding/ordering/counting/symmetry students Cluster analysis, Lochner et Range=0- Contamination/washing, hoarding/symmetry/ordering, item-level, current 261 OCD N/A al., 2008 1 obsessional/checking symptoms Taboo, contamination/cleaning, rituals and superstition, fear Katerberg et Range=0- PCA, item-level, 373 OCD, 462 of harm, intolerance of uncertainty, somatic and sensory 66% al., 2010 1 lifetime symptoms controls phenomena PCA, category- Wiliams et Range=0- 2 OCD samples (74 Contamination/washing, hoarding, sexuality/reassurance, level, current 59.1% al., 2012 2 and 54) aggression/mental, symmetry/perfectionism, doubt/checking symptoms 5 factors (item-level): taboo, contamination/cleaning, doubts, 59% 1224 OCD, 52 OCD- superstitions/rituals, symmetry/hoarding Confirmatory PCA, Katerberg et Range=0- affected item-level, lifetime 4 factors (categorical-level): al., 2010 1 multigenerational symmetry/ordering/arranging/counting/repeating, symptoms 65.5% families aggressive/sexual/religious/checking, contamination/cleaning, hoarding Exploratory factor 5 factors (category-level): Zhang et al., Range=0- analysis (PCA), symmetry/arranging/repeating/counting, 512 OCD 67.1% 2013 1 category- and contamination/cleaning, hoarding, aggressive/checking, item-level, lifetime somatic/religious/sexual;

symptoms 6 factors (item-level): contamination/cleaning, symmetry/arranging/repeating/counting/superstition, hoarding, 66.1% doubt/checking, somatic, religious/sexual/mental

Scoring: investigators assigned a score of 1 if a symptom category was present, 0 if it was absent, and 2 was assigned in some studies for the most prominent symptom category PCA: principal component analysis N/A: not applicable * Scoring of 0 indicates the absence of symptom; scoring of 1 indicates the presence of symptom; scoring of 2 indicates the presence of target or principal symptom

Table 1.9. Summary Table of Genetic Studies of Y-BOCS Symptom Dimensions in OCD.

Gene Dimension Results Reference GRIN2B Contamination/cleaning +/- Alonso et al., 2012; Hemmings et al., 2006 SLC6A4 Repeating/counting, ++/--- Cavallini et al., 2002; Hasler et al., 2006; Kim et al., 2005; symmetry/repeating/counting/ordering Lochner et al., 2008; Wendland et al., 2007 HTR1B --- Hemmings et al., 2006; Kim et al., 2009; Lochner et al., 2008 HTR2A -- Hemmings et al., 2006; Lochner et al., 2008 HTR2C - Hemmings et al., 2006 HTR3A - Lennertz et al., 2014 HTR3B - Lennertz et al., 2014 HTR3C - Lennertz et al., 2014 HTR3D - Lennertz et al., 2014 HTR3E Contamination/washing + Lennertz et al., 2014 HTR6 - Hemmings et al., 2006 DAT/SLC6A3 -- Hemmings et al., 2006; Lochner et al., 2008 DRD1 - Hemmings et al., 2006 DRD2 - Hemmings et al., 2006 DRD3 - Hemmings et al., 2006 DRD4 Hoarding +/-- Hemmings et al., 2006; Lochner et al., 2008; Taj et al., 2013 COMT Hoarding, obsessional/checking, somatic and sensory +++/- Hemmings et al., 2006; Katerberg et al., 2010; Liu et al., 2011; phenomena Lochner et al., 2008 MAOA - Lochner et al., 2008 BDNF -- Hemmings et al., 2006; Wendland et al., 2007 TH - Lochner et al., 2008 ACE - Hemmings et al., 2006 ESRα Hoarding, contamination/cleaning ++ Alonso et al., 2011; Hemmings et al., 2006 ESRβ - Alonso et al., 2011 INPP-1 - Hemmings et al., 2006 PLC-γ1 - Hemmings et al., 2006 HOXB8 - Hemmings et al., 2006 + indicates significant study - indicates negative study

1.4.2.2.2 OCD Symptom Severity

As mentioned in the previous section, Y-BOCS is a clinician-rated, gold standard measure to assess for OCD symptom severity (Goodman et al., 1989). This scale consists of 10 items with each item scoring between 0 (no symptoms) to 4 (extreme), up to a total score ranging from 0 to 40. The scale is divided into two sections, one for obsessions and the other for compulsions, and includes questions related to the time spent on obsessions/compulsions, the interference that the obsessions/compulsions have caused, the amount of distress the patient experienced, the level of resistance, and control over the obsessions/compulsions. The scoring system is as follows: subclinical for 0-7, mild for 8-15, moderate for 16-23, severe for 24-31, and 32-40 for extreme. The severity score allows clinicians to assess treatment outcome.

Higher severity of symptoms has been associated with a greater genetic loading in psychiatric disorders (OCD – Schooler et al., 2008; ADHD – Stergiakouli). Thus, five in total, have looked into the genetics of OCD severity (Table 1.10). Two positive studies demonstrated that the BDNF rs6265 Val/Val genotype was associated with higher severity in OCD females (P=0.013-0.045; Hemmings et al., 2006) (P=0.031; Hemmings et al., 2008); but the samples likely overlapped given that they were conducted by the same group. The study by Márquez et al. (2013) that examined the same SNP was however negative. Three genes, GRIN2B, HTR2A, and DRD1 were found to be associated with OCD severity in one study (Hemmings et al., 2006) but they have not been replicated yet.

Given the scarcity of genetic studies exploring Y-BOCS severity in OCD sample in addition to the potential higher genetic loading in more severe cases of OCD, we aimed to investigate the effect of OCD candidate genes based on OCD severity.

Table 1.10. Summary Table of Genetic Studies of Y-BOCS Symptom Severity in OCD.

Gene Results Reference GRIN2B + Hemmings et al., 2006 HTR1B - Hemmings et al., 2006 HTR2A +/- Hemmings et al., 2006; Walitza et al., 2012 HTR2C - Hemmings et al., 2006 HTR6 - Hemmings et al., 2006 DAT/SLC6A3 - Hemmings et al., 2006 DRD1 + Hemmings et al., 2006 DRD2 - Hemmings et al., 2006 DRD3 - Hemmings et al., 2006 DRD4 - Hemmings et al., 2006 COMT +/- Hemmings et al., 2006; Katerberg et al., 2010 BDNF ++/- Hemmings et al., 2006; Hemmings et al., 2008; Márquez et al., 2013 ACE - Hemmings et al., 2006 ESRα - Hemmings et al., 2006 INPP-1 - Hemmings et al., 2006 PLC-γ1 - Hemmings et al., 2006 HOXB8 - Hemmings et al., 2006 + indicates significant study - indicates negative study

1.4.2.3 Psychiatric Comorbidities in OCD

Individuals suffering from OCD often present with additional comorbid psychiatric conditions. According to a recent study (Cross-Disorder Group of the Psychiatric Genomics Consortium, 2013), a set of genetic markers contribute to the risk of developing several psychiatric disorders (major depressive disorder, bipolar disorder, schizophrenia, autism spectrum disorder, and attention-deficit hyperactivity disorder), implicating a common genetic pathway with genetic markers having pleiotropic effect that may lead to different psychiatric presentations. Since genetic studies require large dataset to identify multiple common small gene effect, cross disorder studies are valuable to identify common genetic variants that predispose an individual to a group of psychiatric disorders.

OCD is highly comorbid with other psychiatric disorders that may have a negative impact on OCD outcome and increase therapeutic challenge (de Mathis et al., 2013; Ruscio et al., 2010). Studies have shown that OCD individuals with comorbid psychiatric disorders predict a poorer short- and long-term treatment outcome in addition to worse quality of life (Jakubowski et al., 2013). Large-scale OCD epidemiological studies reported a lifetime comorbidity rate ranging from 76% to 91% and a current comorbidity rate between 42% and 55% with anxiety and mood disorders as the most frequent conditions in individuals who suffer from OCD (Ruscio et al., 2010; Hofmeijer-Savink et al., 2013). Table 1.11 reported frequencies of psychiatric comorbidity across several large OCD studies (Murphy et al., 2013).

Higher rates of comorbid depression (de Mathis et al., 2011) and eating disorders (Torresan et al., 2009) were observed in OCD females (de Mathis et al., 2011) whereas a higher frequency of comorbid social phobia (Jaisoorya et al., 2009) and tic disorder (de Mathis et al., 2011) was noted in OCD males. In a blinded family study comprising of 80 cases and 73 control probands in addition to 343 OCD cases and 300 control first-degree relatives, body dysmorphic disorder, hypochondriasis, eating disorders, and pathologic “grooming” conditions including nail biting, skin picking, and trichotillomania occurred more frequently in OCD cases than controls (Bienvenu et al., 2000). In the Johns Hopkins OCD family study, the authors reported higher frequency of all anxiety and affective disorders except for bipolar disorder in OCD cases than controls (Nestadt et al., 2001), which was supported by other family studies (Carter et al., 2004).

A Brazilian family study showed higher age-corrected recurrence risks of OCD (22.7% versus 0.9%, odds ratio [OR]=32.5, 95% confidence interval [CI]=4.5-230.8, P=0.0005) and chronic tics (11.6% versus 1.7%, OR=7.9, 95% CI=1.9-33.1, P=0.005) in OCD relatives than control relatives; furthermore, the best predictor of a diagnosis of OCD was found to be a comorbid diagnosis of tics in the relatives (OR=7.35, 95% CI=3.79-14.25, P<0.0001) (do Rosario-Campos et al., 2005).

One of the largest OCD studies from the OCD Collaborative Genetics Study (OCGS) (Nestadt et al., 2009), consisting of 706 OCD individuals, attempted to sub-classify OCD based on comorbidity and the authors reported three subgroups: (1) an OCD simplex class in which only major depressive disorder (MDD) is the most frequent comorbid disorder; (2) an OCD comorbid tic-related class in which tics are highly represented and affective disorders are rarer; and (3) an OCD comorbid affective and anxiety-related class in which panic disorder, affective disorders, and obsessive-compulsive personality disorder are more frequent.

Regarding treatment implication, treatment studies reported that OCD individuals with comorbid tics (Bloch et al., 2006; Masi et al., 2013; McDougle et al., 1994; Nakamae, 2011) and family history of tics (McDougle, 1997) have improved response to antipsychotic (including typical [haloperidol, pimozide], atypical [risperidone, olanzapine, quetiapine], and partial agonist [aripiprazole]) augmentation.

To date, seven studies have examined the genetic effect of a total of 19 genes in OCD all with comorbid major depressive disorder and tics except for one study (Table 1.12; for further details, please refer to Appendix IV Table 1). Wendland et al. (2007) used another approach and examined three markers within SLC6A4 and the BDNF Val66Met (rs6265) polymorphism in OCD with the total number of comorbid mood and anxiety disorders in addition to suicidality and experience of trauma. The authors detected significant association between the total number of comorbid anxiety disorders and 5HTTLPR (P=0.037) and SLC6A4 rs25531 (P=0.003) respectively. However, most were negative except for four genes, serotonin receptor 6 (HTR6), DRD4, COMT, and BDNF with no replications.

OCD patients with comorbid psychiatric conditions often require different treatment approaches and therefore, psychiatric comorbidities are of clinical importance in the 43

management of illness progression and relapse prevention. By determining the genetic basis of OCD with additional comorbid conditions, we may identify common pathways across different psychiatric disorders.

Table 1.11. Psychiatric Comorbidity of OCD (Percentage of total number of individuals reported with OCD or in the general population).

Lifetime Current General Population Comorbid Disorder Rates (%) Rates (%) Prevalence (%)

Obsessive-Compulsive and Related - - - Disorders Trichotillomania 7-36 - 1-2 Body Dysmorphic Disorder 6-12 - 1.7-1.8 Skin Picking 18 - 1.4 Hoarding 2-6 Tourette Syndrome 4-8 - 0.3-0.8 Any Mood Disorder 64-74 16-37 8-32 MDD 38-70 15-18 6-17 Bipolar Disorder 1-23 0.7-1 1-2 Dysthymia 6-24 0-6 3-13 Any Anxiety Disorder 46-70 37-38 15-38 Panic Disorder 6-23 7-9 4 Agoraphobia 1-18 1 5 PTSD 5-19 3 9 GAD 8-46 8-9 5-9 SAD 23-44 19 13 Specific Phobia 11-43 8-15 8-14 Any Substance Use Disorder 10-71 5-6 17-35 Any Psychotic Disorder 3-5 2-3 1.5 Any Eating Disorder 6-19 2-5 0.4-2 Any Impulse Control Disorder 15-39 11 0.3-11 Any Somatoform Disorder 6-8 6 0.1-10 Autism Spectrum Disorder 3 - 1 ADHD 10-13 3 2.5-5 Information adapted from Hasler et al. (2007), Hofmeijer-Savink et al. (2013), LaSalle et al. (2004), Miguel et al. (2008), Nestadt et al. (2001 and 2008), Pallanti and Grassi (2014), Pinto et al. (2006), Ruscio et al. (2010), and Torresan et al. (2013) for the OCD population and Kessler et al. (1994 and 2005) and APA (2013) for the general population.

Table 1.12. Genetic Studies of Psychiatric Comorbidity in OCD. Gene Disorder Results Reference GRIN2B MDD, tics - Hemmings et al., 2006* SLC6A4 Tics, mood and anxiety disorders --- Dickel et al., 2007^; Liu et al., 2011^; Wendland et al., 2007+ HTR1B Tics, MDD -- Dickel et al., 2007^; Hemmings et al., 2006* HTR2A Tics, MDD -- Dickel et al., 2007^; Hemmings et al., 2006* HTR2C Tics -- Cavallini et al., 1998^; Hemmings et al., 2006* HTR6 MDD, tics + Hemmings et al., 2006* DAT/SLC6A3 MDD, tics - Hemmings et al., 2006* DRD1 MDD, tics - Hemmings et al., 2006* DRD2 MDD, tics - Hemmings et al., 2006* DRD3 MDD, tics - Hemmings et al., 2006* DRD4 Tics, MDD +/- Camarena et al., 2007^; Hemmings et al., 2006* COMT MDD, tics +/- Hemmings et al., 2006*; Liu et al., 2011^ MAOA MDD - Camarena et al., 2001# BDNF Tics, MDD, mood and anxiety disorders +/-- Dickel et al., 2007^; Hemmings et al., 2006*; Wendland et al., 2007+ ACE MDD, tics - Hemmings et al., 2006* ESRα MDD, tics - Hemmings et al., 2006* INPP-1 MDD, tics - Hemmings et al., 2006* PLC-g1 MDD, tics - Hemmings et al., 2006* HOXB8 MDD, tics - Hemmings et al., 2006* + indicates significant result - indicates negative result Bold style indicates significant findings with the bolded disorder(s) * Hemmings et al. (2006) investigated genetic variations between OCD with and without comorbid MDD and tics ^ Camarena et al. (2007), Cavallini et al. (1998), Dickel et al. (2007), and Liu et al. (2011) examined genetic markers between OCD with and without comorbid tics # Camarena et al. (2001) tested genetic variants between OCD with and without MDD + Wendland et al. (2007) studied genetic polymorphisms between OCD according to the total number of comorbid mood and anxiety disorders in addition to suicidality and the experience of trauma.

1.4.2.4 Importance of Familiality in OCD

Individuals with a significant family history of mental disorders have a higher risk of developing psychiatric disorders. The rationale for OCD being a genetic brain disorder initially came from family studies because OCD runs in the family. It is widely accepted that the risk of OCD is higher when there is a greater number of relatives with OCD in the family. The unadjusted aggregate risk of OCD was found to be ranging from 8.2% to 13.8% as compared to 1-3% in the general population (Cavallaro et al., 2002; Hettema et al., 2010) and several family studies consisting of 343 OCD cases and 300 control relatives reported significantly higher lifetime prevalence of OCD when compared to control relatives (11.7% versus 2.7%; Nestadt et al., 2000). The rate of OCD among relatives of adults with OCD was approximately two times that of controls but the rate dramatically increased to 10-fold among relatives of children and/or adolescents with OCD (Pauls, 2010). Family studies reported prevalence rates of OCD ranging from 7% to 15% in first-degree relatives of child and adolescent probands with OCD, which supported an increased familial loading in OCD probands with earlier AAO (Nestadt et al., 2010).

Several clinical OCD studies have reported significant differences between familial and sporadic cases of OCD. Hanna et al. (2005) indicated that familial OCD individuals presented with ordering compulsions and aberrant grooming behaviours with skin picking more frequently than the sporadic type. Furthermore, a more recent study reported that familial OCD was associated with earlier onset and greater duration of untreated illness with more compulsive symptoms such as ordering and cognitive types, higher comorbidity including depression and anxiety disorders, in addition to non-response to treatment (Viswanath et al., 2011).

To date, only three studies have examined the genetic effect of familiality or family history in OCD (Denys et al., 2006; Katerberg et al., 2009; Katerberg et al., 2010). Denys et al. (2006) reported a significant association between positive family history of OCD and the HTR2A rs6311 G allele (P=0.039-0.043) in 156 OCD patients. On the contrary, Katerberg et al. (2009; 2010) examined the BDNF rs6265 and COMT rs4680 markers respectively and did not detect any differences in allele and genotype frequencies between presence and absence of family history of obsessive-compulsive symptoms. However, no study to date has explored family

history of other psychiatric disorders in the genetics of OCD. Therefore, it is noteworthy to target OCD individuals with significant family history of OCRDs for genetic study given the higher probability of genetic loading in the familial group.

1.4.3 Pharmacogenetics of OCD

A large part of this section has been taken directly from a review paper (Zai et al., 2014; Pharmacogenomics 2014; 15(8):1147-1157) with granted copyright permission.

The role of pharmacogenetics has become increasingly important in the treatment of mental illness. Evidence suggests that inter-individual genetic variation may largely determine drug response and adverse effects. Pharmacogenetic research in major depressive disorder (MDD) and schizophrenia has provided solid background to utilize genetic data in the treatment of these disorders. Genetic factors influencing antidepressant response have been established and well-studied in MDD but not in anxiety disorders (Tiwari et al., 2009) or the newly categorized obsessive-compulsive and related disorders including obsessive-compulsive disorder (OCD).

Antidepressants are the first-line treatment of MDD, anxiety disorders, and obsessive- compulsive and related disorders. More than 40% of patients do not typically respond to the initial treatment (McDougle et al., 1993) and the duration of the initial trial may take up to six weeks for MDD and up to 12 weeks for anxiety, and obsessive-compulsive and related disorders prior to switching to another medication. Both patients and clinicians are making treatment decisions for psychiatric disorders in a trial-and-error fashion, without any particular guidance. Pharmacogenetics may significantly contribute to a better prediction of drug response and side effect tolerance by personalizing the choice of the best antidepressant for a particular patient, taking into account one’s genetic information that influences drug metabolism, drug transport and drug mechanism. This should ultimately decrease the duration of patient suffering, reduce the number of medication trials, lower health care costs, and improve patients’ quality of life.

Compared to the genetics of OCD, pharmacogenetics concept is a more recent approach in psychiatry. There are only approximately 20 published articles or abstracts reporting on the 47

effect of genetic variations in antidepressant response in OCD patients (Qin et al., 2015; Zai et al., 2014).

A summary and review of previous studies on the pharmacogenetics of OCD can be found in (Table 1.13). A total of 18 published articles were found in literature search using these specific search terms: genetics, pharmacogenetics, obsessive-compulsive gene response, treatment response, drug response, antidepressant response, antidepressant, OCD, obsessive- compulsive disorder, side effects, and adverse effects.

Pharmacogenetics can be separated into two categories, pharmacokinetic factors including the cytochrome P450 liver enzyme system, and pharmacodynamic factors including the brain candidate gene systems. Table 1.13 presents all pharmacogenetic studies in OCD to- date.

Table 1.13. Summary of Pharmacogenetic Studies in OCD.

Gene Polymorphism Study design Medication Sample Results Main finding Reference CYP2D6 *4 (rs3892097) 12 weeks Venlafaxine up Initially 91 patients and - No association of Van *6 (rs5030655) randomized to 300 mg or final analysis with 74 polymorphisms Nieuwerburgh *10 (rs1065852) double blind paroxetine up OCD* patients, m/f: with treatment et al., 2009 *41 (rs28371725) prospective to 60 mg 39.2%/60.8%, mean age response; plasma Response 36.3±11.6 levels of drugs definition: >25% significantly higher reduction of Y- in patients with at BOCS scores least one allele with reduced activity, plasma levels of metabolites significantly lower in cases with two alleles with reduced activity CYP2D6 AmpliChip CYP450 >10 weeks Fluoxetine, 39 patients; m/f: - No overall Müller et al., test, which detects up unless patient sertraline, 41%/59%; mean age significant 2012 to 33 alleles including reported robust fluvoxamine, 46.0±10.2; ethnicity: association 7 duplications in clinical response paroxetine, 87.5% European between CYP2D6 and 3 Response citalopram ancestry metabolizer status alleles in CYP2C19 definition: based and treatment CYP2C19 - on clinical global response but poor impression – metabolizer was improvement associated with (CGI-I) scale lack of response to antidepressants

CYP2D6 *3 (rs35742686) >10 weeks Fluoxetine, 184 patients; m/f: 71/113; - No significant Brandl et al., *4 (rs3892097) unless patient sertraline, mean age 38±11.4; association but 2014 *10 (rs1065852) reported robust fluvoxamine, ethnicity CYP2D6 non- *17 (rs28371706) clinical response paroxetine, European/others: 166/18; extensive *41 (rs28371725) Response citalopram, mean age at onset metabolism (48% definition: based venlafaxine 14±8.6; mean Y-BOCS versus 22% with on clinical global 29.8±5.9 ≥4 trials) when impression – compared with improvement extensive (CGI-I) scale metabolizers Side effect (P=0.007) and definition: without CYP2D6 non- CYP2C19 *2 (rs4244285) - significant side extensive *3 (rs4986893) effects as “well- metabolism was *17 (rs12248560) tolerated” or “mild associated with side effects” and greater venlafaxine significant side side effects effects as (P=0.022) “marked” or “severe side effects”

SLC6A4 44-bp ins/del 12 weeks Clomipramine 34 family trios with 34 - Transmission McDougle et (5-HTTLPR) prospective 150-250 mg, patients diagnosed with disequilibrium test al., 1998 (rs25531) Response fluvoxamine OCD*; mean age indicated L-allele definition: ≥35% 150-300 mg, 33.0±8.5 years (gender transmitted more improvement on fluoxetine 40- information not provided) frequently to non- the Y-BOCS and 80 mg, responders than S- final Y-BOCS sertraline 100- allele (P=0.052) scores of <16 200 mg or and a final CGI paroxetine 40- score of ‘much 60 mg improved’ or ‘very much improved’; 10 weeks Fluoxetine >60 72 OCD patients*/ 72 - No association with Billett et al., retrospective mg matched controls; mean treatment response 1997 Response or age 36.3±8.4 years definition: 25% clomipramine 50

reduction in >150 mg or symptom severity equivalent dose of fluvoxamine, paroxetine or sertraline 12 weeks Fluvoxamine 181 OCD patients*; m/f: - No association with Di Bella et al., prospective 200-300 mg 93/88; 92 patients treatment 2002 Response treated with response; definition: Y- fluvoxamine/191 significant time per BOCS reduction controls; mean age genotype >35% 33.4±11.5 interaction for Y- BOCS compulsions score (P=0.028) and significant time per genotype interaction in non- tic disorder patients for total Y- BOCS score (P=0.041) and Y- BOCS compulsions score (P=0.007) 12 weeks Venlafaxine up 88 OCD patients*; + S/L genotype Denys et al., prospective to 300 mg or m/f:34/54 (paroxetine associated with 2007 randomized paroxetine up N=40); mean age of better response in double-blind to 60 mg responders/non- venlafaxine-treated Response responders patients (P=0.008) definition: >25% 34.1±11.3/31.7±12.0 reduction of Y- BOCS scores 8 weeks Different SRIs 113 OCD patients and - No association Zhang et al., prospective families (information 2004 response regarding age and assessment with gender not provided) Y-BOCS scale, (further details not available) 14 weeks Clomipramine 41 OCD patients*; m/f: - No association Miguita et al., 51

STin2 prospective 23/18; age distribution: - No association 2011 Response age 20-30: N=22, age definition: >40% 31-40: N=13, age>41: reduction in Y- N=6 BOCS score HTR1B 861G/C 12 weeks Venlafaxine up 88 OCD patients*; - No association Denys et al., (rs6296) prospective to 300 mg or m/f:34/54 (paroxetine 2007 randomized paroxetine up N=40); mean age of double-blind to 60 mg responders/non- Response responders definition: >25% 34.1±11.3/31.7±12.0 reduction of Y- BOCS score 14 weeks Clomipramine, 41 OCD patients*; - No association Miguita et al., prospective mean dose m/f:23/18; age 2011 Response 235.5 mg distribution: age 20-30: definition: >40% N=22, age 31-40: N=13, reduction in Y- age>41: N=6 BOCS score Response Clomipramine, 239 OCD patients (60 - No association Corregiari et definition: fluoxetine, with response data); al., 2012 responders with fluvoxamine, m/f:36/24 Y-BOCS score>9 citalopram, and Sheehan sertraline, Disability paroxetine Scale>10 for ≥2 months and non- responders with <25% reduction in Y-BOCS score for ≥2 SSRIs and clomipramine for ≥8 weeks HTR2A -1438G/A 12 weeks Venlafaxine up 88 OCD patients*; m/f: - G/G genotype Denys et al., (rs6311) prospective to 300 mg or 34/54 (paroxetine associated with 2007 randomized paroxetine up N=40); mean age of better response in double-blind to 60 mg responders/non- paroxetine-treated Response responders patients (P=0.013) definition: >25% 34.1±11.3/31.7±12.0 reduction of Y- 52

BOCS score 8 weeks Different SRIs 113 OCD patients and + Homozygosity Zhang et al., prospective families (information associated with 2004 response regarding age and treatment response assessment with gender not provided) (P=0.030) Y-BOCS scale (further details not available) -1438G/A (rs6311) 12 weeks Fluvoxamine 58 OCD patients*; m/f: - No association Tot et al., and prospective 100-300 mg, 20/38; mean age 27±5 2003 102T/C Response fluoxetine 20- and 83 healthy controls (rs6313) definition: CGI 80 mg, rating “very sertraline 100- much” or “much 200 mg improved 102T/C (rs6313) Response Clomipramine, 239 OCD patients (60 + Higher CC Corregiari et definition: fluoxetine, with response data); genotype in non- al., 2012 responders with fluvoxamine, m/f:36/24 responders Y-BOCS score>9 citalopram, (P<0.01) and Sheehan sertraline, Disability paroxetine Scale>10 for ≥2 516C/T (rs6305) months and non- - No association responders with <25% reduction in Y-BOCS score for ≥2 SSRIs and clomipramine for ≥8 weeks 102T/C (rs6313) 14 weeks Clomipramine, 41 OCD patients*; m/f: - No association Miguita et al., 516C/T (rs6305) prospective mean dose 23/18; age distribution: - 2011 Response 235.5 mg age 20-30: N=22, age definition: >40% 31-40: N=13, age>41: reduction in Y- N=6 BOCS score BDNF rs11030096 12 weeks Fluoxetine 60- 131 OCD patients*; m/f: - No association Real et al., prospective 80 mg, 61/62; mean age after permutation 2009 Response fluvoxamine 33.3±10.5 correction rs925946 definition: >35% 200-300 mg, or - No association 53

rs10501087 reduction in Y- clomipramine - No association BOCS score 225-300 mg after permutation correction Val66Met - No association (rs6265) after permutation correction rs12273363 - No association rs908867 - No association of SNP, haplotype containing G-allele associated with poorer response (OR=2.27; P=0.027) rs1491850 + C-allele associated with better response (OR=2.40, P=0.005); haplotype containing T-allele associated with poorer response (OR=2.27; P=0.027) rs1491851 - No association SLC1A1 rs301434, rs301435, + Association of A- Real et al., rs3087879 allele of rs301434 2010 with pharmacological resistance in patients without SLE, association of A-G-C haplotype with pharmacological resistance in patients without SLE

COMT Details not shown 8 weeks Different SRIs 113 OCD patients and - No association Zhang et al., prospective families (information 2004 response regarding age and assessment with gender not provided) Y-BOCS scale (further details not available) Val158Met (rs4680) 14 weeks Clomipramine, 41 OCD patients*; m/f: - No association Miguita et al., prospective; mean dose 23/18; age distribution: 2011 response 235.5mg age 20-30: N=22, age definition: >40% 31-40: N=13, age>41: reduction in Y- N=6 BOCS score 10 weeks Citalopram up 64 patients; m/f: 31/33; + Association of Vulink et al., prospective to 60mg/day mean age of Met/Met with 2012 Response plus quetiapine responders/non- responders to definition: 450mg/day responders: citalopram as responder as ≥ versus 33.0±11.7/38.2±12.7; compared to non- 25% reduction in citalopram mean age at onset of responders Y-BOCS score versus placebo responders/non- (P=0.007) responders: 15.9±8.9/15.0±4.8; mean Y-BOCS of responders/non- responders: 26.3±4.3/27.8±4.7; mean HAM-A of responders/non- responders: 12.5±7.5/14.6±6.3; mean HAM-D of responders/non- responders: 8.7±4.8/9.8±3.1 36-month Different SSRIs 171 OCD (91 with drug - No association Umehara et prospective (Details not response data) patients; al., 2015 treatment study; shown) m/f: 81/90; mean age of: Response 33.8±11.0; median definition: “A” duration of illness: 6; group responders mean Y-BOCS: 25.8±6.0 55

as >35% reduction in Y- BOCS score with only SSRI treatment, “B” group responders as >35% reduction in Y- BOCS score with antipsychotics and SSRI treatment, and “C” as non- responders GRIN2B rs7972211 Details not SSRIs (details 1598 OCD cases from a + P=2.71×10-5 Qin et al., provided but not provided) family-based GWAS with (OR=0.65 [95% 2015 according to the ~800 cases as confidence interval primary author, responders according to 0.49-0.87]) response details the primary author were provided by patient with “yes” or “no” response to a SSRI SLC6A3 40 bp VNTR UTR 14 weeks Clomipramine, 41 OCD patients*; m/f: - No association Miguita et al., region prospective; mean dose 23/18; age distribution: 2011 30 bp VNTR intron 8 response 235.5mg age 20-30: N=22, age - VNTR intron 14 definition: >40% 31-40: N=13, age>41: - SLC6A2 1287G/A reduction in Y- N=6 - (rs5569) BOCS score DRD2 Details not shown 8 weeks Different SRIs 113 OCD patients and - No association Zhang et al., prospective families (information 2004 response regarding age and assessment with gender not provided) Y-BOCS scale (further details not available) Taq1A (rs1800497) 10 weeks Citalopram up 64 patients; m/f: 31/33; - No association Vulink et al., prospective to 60mg/day mean age of 2012 Response plus quetiapine responders/non-

definition: 450mg/day responders: responder as ≥ versus 33.0±11.7/38.2±12.7; 25% reduction in citalopram mean age at onset of Y-BOCS score versus placebo responders/non- responders: 15.9±8.9/15.0±4.8; mean Y-BOCS of responders/non- responders: 26.3±4.3/27.8±4.7; mean HAM-A of responders/non- responders: 12.5±7.5/14.6±6.3; mean HAM-D of responders/non- responders: 8.7±4.8/9.8±3.1 DRD4 Details not shown 8 weeks Different SRIs 113 OCD patients and - No association Zhang et al., prospective families (information 2004 response regarding age and assessment with gender not provided) Y-BOCS scale (further details not available) VNTR Naturalistic Different SRIs 146 patients; m/f: 98/48; - No association Viswanath et follow-up; mean age of: 28.6±9.1; al., 2013 Response median duration of definition: illness: 6; mean Y- responders as BOCS: 25.8±6.0 >35% reduction in Y-BOCS score and CGI-I of 1 or 2, partial responders as 25-35% reduction in Y-BOCS score, and non- responders as <25% reduction 57

in Y-BOCS score MAOA Details not shown 8 weeks Different SRIs 113 OCD patients and - No association Zhang et al., prospective families (information 2004 response regarding age and assessment with gender not provided) Y-BOCS scale (further details not available) GWAS Illumina Treatment SRIs including 804 OCD patients + • DISP1 Qin et al., HumanOmniExpress response using a clomipramine, rs17162912 2015 – 12v1 with 730,525 five-point scale of citalopram, (P=1.76×10-8; SNPs “no response”, escitalopram, OR=0.39, 95% “could not fluoxetine, CI=0.26-0.58) tolerate”, fluvoxamine, • PCDH10 “minimal”, paroxetine, rs7675822 “moderate sertraline, (P=2.86×10-6; improvement”, venlafaxine, OR=0.65, 95% and “total and duloxetine CI=0.51-0.83) remission” with and rs1911877 “moderate (P=8.41×10-6; improvement” or OR=0.66, 95% “total remission” CI=0.52-0.84) as responders • GRIN2B and “no rs7972211 response” or -5 (P=2.71×10 ; “minimal OR=0.65, 95% response” as CI=0.49-0.87) non-responders • GPC6 rs17253738 (P=2.13×10-5; OR=0.59, 95% CI=0.43-0.82), rs95116369 (P=4.38×10-5; OR=0.61, 95% CI=0.44-0.84), and rs3891616 (P=8.39×10-5; OR=0.63, 95% 58

CI=0.46-0.87) • PLCB1 rs722665 (P=8.47×10-5; OR=1.61, 95% CI=1.25-2.08) and rs2423366 (P=0.0001) • PKC rs11158347 (P=5.18×10-5; OR=1.83, 95% CI=1.39-2.41) + indicates significant study - indicates negative study

1.4.3.1 Pharmacokinetic Factors

Numerous previous studies have examined the effects of cytochrome P450 liver enzymes in drug metabolism (Caley, 2011). CYP2D6 metabolizes most antidepressants including all tricyclic antidepressants, the serotonin norepinephrine reuptake inhibitor (SNRI), venlafaxine, and most selective serotonin reuptake inhibitors (SSRIs), predominantly fluoxetine and paroxetine. CYP2C19 metabolizes several tricyclic antidepressants and SSRIs including escitalopram and citalopram. Furthermore, genetic variations in CYP2D6 and CYP2C19 alter the metabolism of their substrates (i.e., medications), thereby affecting drug efficiency and tolerability. Although other cytochrome P450 enzymes such as CYP1A2, CYP2B6, CYP2C9, and CYP3A4 may also be important to the metabolism of antidepressants including tricyclic antidepressants, bupropion, trazodone, and some SSRIs (of variable degree), very few pharmacogenetic studies of these enzymes and antidepressant response have been published in the literature (Altar et al., 2013) and none has been investigated in OCD. Therefore, we have limited this review to CYP2D6 and CYP2C19.

The genotypes of CYP2D6 allow for a prediction of its enzymatic activity levels in the absence of confounding factors such as co-medication. It is grouped into four different metabolism classifications including “poor” with no CYP2D6 activity, “intermediate” with metabolism between poor and extensive, “extensive” with normal activity, and “ultra-rapid” with an increased metabolism (Jann et al., 2000). Having a poor metabolism status indicates that the individual will metabolize the substrate (i.e., SSRIs, venlafaxine) at a very low rate. This can in turn increase side effects given the higher cumulative drug concentration that remains in the individual’s body. On the other hand, an individual with an ultra-rapid metabolism status will break down the substrate (i.e., SSRIs, tricyclic antidepressants) at a fast rate, resulting in low cumulative serum concentration of the medication and likely a sub-therapeutic level. For metabolism status, on average, 10% of the general population are poor metabolizers, 35% are intermediate metabolizers, and 7% are ultra-rapid metabolizers. However, ethnicity plays an important factor in the CYP2D6 metabolism variability (Neafsey et al., 2009). The prevalence of CYP2D6 poor metabolizers ranges from approximately 2-10% in Caucasians, 2-7% in Hispanics, and 1-2% in Asians, Africans, and African Americans (Bernard et al., 2006).

Regarding CYP2C19, the same classifications have been used based on the level of enzyme activity. Similarly, inter-ethnic differences have been observed in CYP2C19 metabolism, with Asians having a much higher prevalence (ranging from 12-23%) of poor metabolizers when compared to Caucasian (1-6%) and Black African (1-7.5%) populations (Desta et al., 2002).

Recently, two studies from our group reported influence of genetic variations across these two liver enzymes in SSRI response and OCD (Brandl et al., 2014; Mueller et al., 2012). The first study by Müller and colleagues (Mueller et al., 2012) included 39 patients with OCD who were assessed for SSRI antidepressant response (including fluoxetine, sertraline, fluvoxamine, paroxetine, and citalopram) based on the Clinical Global Impression – Improvement (CGI-I) scale. Although no overall significant association was detected between CYP2D6 or CYP2C19 metabolizer status and treatment response, impairment in CYP2D6 function (poor metabolizer status) was associated with lack of response to SSRIs. The second study (Brandl et al., 2014) investigated the association of CYP2D6 and CYP2C19 status with treatment response and side effects in 184 OCD subjects. Although this study did not show significant association of metabolism status with response to specific medications, the authors noted that patients with CYP2D6 non-extensive metabolism had a significantly increased number of failed SSRI trials when compared to extensive metabolizers (48% of non-extensive metabolizers had required ≥4 trials versus 22% of extensive metabolizers with ≥4 trials; P=0.007). Furthermore, the use of venlafaxine in individuals with CYP2D6 non-extensive metabolism was significantly associated with the occurrence of side effects (P=0.022).

To summarize, only two cytochrome P450 liver enzyme genes, CYP2D6 and CYP2C19 have been studied in antidepressant response and OCD. The genetic status of CYP2D6 appears to modify antidepressant response in OCD patients with non-responders being more common among non-extensive metabolizers. However, the sample size of these studies is relatively small and further validation of these findings is required.

1.4.3.2 Pharmacodynamic Factors

Pharmacodynamic factors affect drug actions at a molecular level within our brain. Psychotropic medications have various mechanism of actions for which numerous neurotransmitter systems are involved. The most well-known and most studied systems consist of the serotonergic, glutamatergic, and dopaminergic systems, which interact with one another and are associated with drug response and adverse effects of medications (Gvozdic et al., 2012). In this section, we present a general overview of each of the above mentioned systems in addition to other systems of interest. With regard to brain candidate genes across these systems and pharmacogenetics of OCD, there has been little development in this area recently.

1.4.3.2.1 Serotonergic system genes

The serotonergic system has been postulated to be involved in the pathogenesis of OCD given that SSRIs are the first-line treatment of OCD based on extensive clinical trials (Koran et al., 2007). SSRIs increase the level of serotonin in the synaptic cleft by inhibiting its reuptake into the presynaptic cells, thereby allowing more serotonin to bind to the postsynaptic receptors. This is thought to be the mechanism by which these medications reduce depression and anxiety, and engage in complex interaction with other neurotransmitter systems. Many genes that are involved in serotonin metabolism, transport, storage, and signalling cascade, have been studied extensively over the past decade in relation to OCD (Pauls, 2010). Candidate genes in the serotonergic system implicated in OCD include: serotonin transporter (SLC6A4) and its promoter (HTTLPR), serotonin 2A receptor (HTR2A), serotonin 2C receptor (HTR2C), serotonin 1B (1D beta) receptor (HTR1B), and tryptophan hydroxylase (TPH).

Most pharmacogenetic studies of antidepressant response came from MDD and findings for HTR2A appear to be most robust (Fabbri et al., 2013; Singh et al., 2013). The study by Corregiari et al. (Corregiari et al., 2012) consisted of 239 OCD patients and of those, 60 had antidepressant response data. Corregiari and colleagues (Corregiari et al., 2012) investigated one SNP across the HTR1B gene, 861G/C (rs6296), and two SNPs within the HTR2A gene, 102T/C (rs6313) and 516C/T (rs6305). Six different antidepressants were examined including clomipramine, fluoxetine, fluvoxamine, citalopram, sertraline, and paroxetine. The authors 62

defined responders as having a Yale-Brown Obsessive-Compulsive Scale (Y-BOCS) severity score of less than 9 and a Sheehan Disability Scale score of less than 10 after treatment with an SRI and being stable for at least two months. Patients with less than 25% reduction in the Y- BOCS severity score after at least two different SSRIs and clomipramine for at least 8 weeks were classified as non-responders. This study reported significantly higher CC genotype and C allele of the HTR2A 516C/T marker in non-responders (P<0.01 and P=0.01 respectively). The authors also investigated the relationship between these polymorphisms, study groups (responders, non-responders, and healthy controls), and endocrine and serotonin measures (serotonin, prolactin, growth hormone, and cortisol basal levels); they observed significantly higher cortisol and prolactin levels in response to citalopram treatment in individuals with the CC genotype of the HTR1B 861G/C marker (P<0.01 and P<0.001 respectively). This was the first study examining genetic association of endocrine response to SRI treatment and the authors concluded that serotonergic system genes may play an important role in influencing SRI response and regulating endocrine and serotonin level.

Overall, there were a limited number of OCD pharmacogenetic studies with relatively small sample size, providing consistent but not yet conclusive results implicating HTR2A and HTR1B in SRI antidepressant response in OCD.

1.4.3.2.2 Glutamatergic system genes

There is increasing evidence of the influence of the glutamatergic system on the etiology of many neuropsychiatric disorders including OCD (Kariuki-Nyuthe et al., 2014). Support for involvement of this system in the pathogenesis of OCD comes from research implicating the complex neurotransmitter network, the CSTC circuit, in the underlying neuro-anatomical basis of OCD; altered glutamatergic function is seen as one of the major drivers of observed over- activity in this brain circuit in OCD (Milad et al., 2012). Glutamate has also been implicated in this disorder based on evidence from magnetic resonance spectroscopy studies, knock-out mouse models, clinical studies of glutamatergic agents, and candidate gene association studies (Wu et al., 2012). We briefly summarize the evidence supporting glutamate involvement in the

neurobiology and pathoetiology of OCD below, but also refer interested readers to a recent extensive review article (Wu et al., 2012).

Glutamatergic medications such as phenobarbital, lamotrigine, topiramate, riluzole, N- acetylcysteine (NAC), memantine, ketamine, and D-cycloserine, have been used as an augmenting agent in the management of OCD (Grados et al., 2013). In particular, adjunctive topiramate and riluzole are considered third-line pharmacotherapy for OCD according to the Canadian Psychiatric Association anxiety disorders clinical practice guideline in 2006 (CPA, 2006), and use of glutamate antagonists such as riluzole and topiramate was also recommended by the American Psychiatric Association (Koran et al., 2007). Of interest, new glutamatergic modulating drugs have been developed by pharmaceutical companies, namely Eli Lilly (mGluR2/3 agonist) and Roche (glycine transporter inhibitor). These medications have had mixed results in clinical trials in schizophrenia, and are currently being investigated in OCD. A pilot clinical trial of the glycine transporter inhibitor (Sarcosine) has provided benefits in drug- naïve OCD patients but not in patients who were treatment-resistant or had prior SSRI trial(s) (Wu et al., 2011).

A very recent GWAS of drug response was conducted by Qin et al. (Qin et al., 2015) with 804 OCD cases. GWAS was performed using the Illumina HumanOmniExpress – 12v1 with 730,525 SNPs (Illumina Inc. ®, San Diego, CA, USA). The following is a list of the SRI antidepressants that were included in this GWAS: clomipramine, citalopram, escitalopram, fluoxetine, fluvoxamine, paroxetine, sertraline, venlafaxine, and duloxetine. Drug response status was determined using a five-point scale of “no response”, “could not tolerate”, “minimal response”, “moderate improvement”, and “total remission”. Responders were defined as those who have scored “moderate improvement” or “total remission” and non-responders refer to those who reported “no response” or “minimal response”. The authors concluded that one of the top SNPs was the rs7972211 marker near GRIN2B, which showed modest significant association with SSRI response (OR=0.65 [95% confidence interval 0.49-0.87], P=2.71×10-5). Moreover, three SNPs (rs17253738, rs9516369, and rs3891616) across the glypicans 6 (GPC6) gene detected strong signals of association with SSRI response. GPC6 functions to promote glutamate receptor clustering and receptivity in addition to inducing post-synaptic activity (Allen et al., 2012).

Only one candidate gene study examined a glutamatergic system gene, SLC1A1, in SRI response in OCD and Real et al. (2010) reported a significant association between the rs301434 A allele and non-response to fluoxetine, fluvoxamine, or clomipramine response in a 12-week prospective study with 131 OCD patients.

With clinical support from utilizing glutamatergic agents in the treatment of OCD and research evidence of glutamatergic system genes from genetic association studies, GWAS, and pharmacogenetics studies, further investigation of glutamatergic pathway genes is crucial in determining the etiology and management of OCD. As previously stated, only one study with positive finding has implicated a role of SLC1A1 in OCD treatment response (Real et al., 2010) but a recent GWAS implicated an enrichment of glutamate-related genes in SRI response (Qin et al., 2015). However, no published pharmacogenetic studies have investigated glutamate medication response in OCD to date.

1.4.3.2.3 Dopaminergic system genes

Disruption of the complex interactions across the CSTC circuit has been implicated in the pathogenesis of OCD, and dopaminergic neurotransmission is an important player in this complicated network (Pauls, 2010). Further evidence of dopamine influence is provided by neuroimaging, neurochemical, and candidate gene studies (Koo et al., 2010). Importantly, antipsychotic medications have established utility as augmentors to SRI antidepressants in the treatment of OCD (CPA, 2006).

Dopaminergic system genes have thus received some interest in genetic association and pharmacogenetic studies of OCD. Only three pharmacogenetic studies, examining the dopaminergic genes in OCD, dopamine D2 receptor (DRD2) and dopamine D4 receptor (DRD4), have been conducted and no associations were detected (Viswanath et al., 2013; Vulink et al., 2012); however, a COMT functional SNP was found to be associated with citalopram response (Vulink et al., 2012). However, a previous published study did not detect any significant association between SRI response and DRD2, DRD4, COMT, and MAOA in OCD (Zhang et al.,

2004). Two additional studies failed to demonstrate positive results between COMT and SRI response in OCD (Miguita et al., 2011; Umehara et al., 2015).

Vulink et al. (Vulink et al., 2012) (Table 1) examined the DRD2 Taq1A (rs1800497) and COMT Val158Met (rs4680) polymorphisms, with interesting results. This study recruited 64 OCD patients who were randomly assigned to citalopram plus quetiapine versus citalopram plus placebo. Although no significant differences were observed in genotype distributions or allele frequencies of DRD2 or COMT between responders and non-responders to adjunctive quetiapine treatment, 48% of the responders to citalopram had COMT Met/Met genotype as compared to 0% of the non-responders (χ2=10.06, P=0.007). The authors concluded that COMT Val158Met Met/Met genotype may predict better response to citalopram treatment.

A recent naturalistic study by Viswanath et al. (Viswanath et al., 2013) examined 146 OCD patients and the influence of the DRD4 VNTR polymorphism in SRI response. However, the authors did not detect any significant association.

Once again, inconsistent reports of the involvement of dopaminergic system genes suggest that further investigations are needed with significantly larger sample size in antidepressant response in OCD.

1.4.3.2.4 Other candidate genes that have not been examined in pharmacogenetics

Numerous additional candidate systems or genes have emerged in the recent years in the genetics of OCD, including the neurotrophic system genes, the neuronal-related genes, and additional top hits from the GWAS (Stewart et al., 2013) that were not previously known to be involved in neuropsychiatric disorders.

Neurotrophic factor genes such as the brain-derived neurotrophic factor (BDNF) and neurotrophin-3 receptor (NTRK3) have frequently been hypothesized in the etiology of many neuropsychiatric disorders in addition to OCD (Pauls, 2010). BDNF has been studied extensively and although a recent meta-analysis detected no significant association of its functional Val66Met polymorphism and OCD (Taylor, 2013), it is still considered an interesting 66

gene given its function in neurodevelopment, neurogenesis, and neuroplasticity (Autry and Monteggia, 2012). Very limited pharmacogenetic studies to-date, examining BDNF and other related genes, have revealed inconsistent and limited findings (Zai et al., 2014).

Neuronal-related genes such as the myelin-oligodendrocyte glycoprotein (MOG) (Zai et al., 2004) and the oligodendrocyte lineage transcription factor 2 (OLIG2) (Stewart et al., 2007) genes were examined in genetic association studies of OCD with promising findings. However, the MOG tetra-nucleotide repeat finding has not been studied for replication purpose and we (Zai et al., 2012) could not replicate the significant results in OLIG2 from Stewart et al. (Stewart et al., 2007).

Many new loci were identified as top hits in the recent GWAS of OCD (Stewart et al., 2013); however, further research is needed to clarify their functional status and their potential role in the pathoetiology of OCD and antidepressant response in OCD.

The GWAS (Qin et al., 2015) (previously described in the glutamatergic system genes section) detected significant top-hits with antidepressant response including: dispatched homolog 1 (drosophila) (DISP1) rs17162912 (P=1.76×10-8), ankyrin-repeat and fibronectin type III domain containing 1 (ANKFN1) rs9303380 (P=1.03×10-6), arrestin domain containing 4 (ARRDC4) rs12437601 (P=1.66×10-6), T-cell lymphoma invasion and metastasis 1 (TIAM1) rs16988159 (P=2.48×10-6), protocadherin 1D (PCDH1D) rs7676822 (P=2.86×10-6), uncharacterized LOC730101 (LOCT30101) rs723815 (P=3.50×10-6), protocadherin 10 (PCDH10) rs1911877 (P=8.41×10-6). Several genes are involved in neuronal cell-cell adhesion including the DISP1 gene (Roessler et al., 2009), which is located on a previously known micro- deletion region for mental retardation on chromosome 1q41-q42 (Shaffer et al., 2007), PCDH10, which was previously implicated in autistic spectrum disorder (Redies et al., 2012), and TIAM1 (Mack et al., 2012), and neurotrimin (NTM) (Liu et al., 2004).

1.4.3.2.5 Pharmacogenetics summary

No definitive results support a single genetic variation or gene that determines antidepressant response in OCD; however, several SNPs across candidate genes within the cytochrome P450, glutamatergic, and serotonergic systems appear interesting (Zai et al., 2014). These markers include: cytochrome P450 2D6 (CYP2D6), SLC6A4, HTR1B, HTR2A, SLC1A1, and GRIN2B. A recent GWAS of antidepressant response in OCD revealed one genome-wide significant hit for the marker near the dispatched homolog 1 (Drosophila) (DISP1) gene, which was previously implicated in neurological development (Qin et al., 2015). Future directions should focus on expanding the sample size and fine mapping these regions in order to refine our understanding of the functional details of the genetic variation and their relation to the pathophysiology of OCD.

Interest in the pharmacogenetics of OCD has increased over the past decade with a number of international research groups examining genetic factors that determine antidepressant response in OCD. Given the high prevalence and morbidity of OCD, however, there remains a relative paucity of research in pharmacogenetics of OCD and in OCD in general. Although there is no definitive data supportive a single genetic variation or gene that determines antidepressant response in OCD, many SNPs across candidate genes within the cytochrome P450, glutamatergic, and serotonergic systems appear interesting and potentially involved (Table 1.13). Future directions should focus on expanding the sample size and fine mapping these regions in order to refine our understanding of the functional details of the genetic variation and their relation to the pathophysiology of OCD. Neuropsychiatric disorders are likely the result of subtle changes in highly complex neurotransmitter systems. As such, far more extensive functional and pharmacological studies interrogating these systems more deeply are warranted to delineate the effects of genetic variations in antidepressant response.

There are several general limitations in these published pharmacogenetic studies including small sample size, retrospective study design, and a lack of details regarding subjects’ treatment history. Additional larger and well-characterized OCD samples in a prospective study design will increase statistical power to detect true genetic variant(s) that may predict antidepressant response in OCD and its related disorders. Encouraging international groups to join together in collaboration provides a framework to increase sample size, to assess for more 68

detailed treatment characteristics (i.e., drug-naïve, drug-free, or treatment resistant), to refine studied phenotypes, to delineate more homogeneous subgroups, and for sharing expertise. Recent advances in technologies such as GWAS, DNA/RNA/whole exome sequencing, and fine mapping allow researchers to conduct more powerful investigations of genetic variations across the entire genome. The recent ENCyclopedia Of DNA Elements (ENCODE) project has explored large regions between genes, and elucidated the importance of a majority of these stretches of non-expressed DNA sequence – previously known as “junk” DNA. These regions contain functional elements and regulators of a large number of known genes, thereby providing a new frame-work for researchers to examine DNA sequence outside of recognized coding regions (Kavanagh et al., 2013). It is easy to imagine that multiple genes will be involved in any given patient for their medication response and side effects, thus development of prediction algorithms that include markers within genes and markers at more remote locations that regulate gene activity will no doubt be necessary. Computational and bioinformatics advancements to address the complex facets of each patient will be required, since it is unlikely that a single clinician will be able to assemble and interpret all the pharmacokinetic, pharmacodynamic, and environmental factors as well as their interactions.

In addition to the strategy of narrowing to more homogeneous subgroups within OCD, there is merit in examining pharmacogenetic associations across disorders. For example, the genes involved in antidepressant response in depression or anxiety may well have similar effects in OCD. Furthermore, studies combining all psychiatric disorders, especially across obsessive- compulsive and related disorders that undergo antidepressant treatment may provide a platform to examine rare variant(s) for the prediction of treatment response and adverse effects by increasing the sample size and power to detect genes of small effect. Identification of genetic variations that predict response and tolerability of medications will likely lead to personalized medicine in the relatively near future, with attendant improvement in clinical outcome and reduction in morbidity and mortality.

1.5 Genes of Interest

This thesis utilized two different genetic approaches to examine OCD phenotypes. GWAS examines common genetic variants across the entire genome. The GWAS utilized genetic data generated from the genome-wide Illumina Human610-Quadv1_B SNP BeadChip array (Illumina Inc. ®, San Diego, CA, USA). The array uses tag SNPs to the genomic coverage of 924,000 randomly selected SNPs, which offer 80% of genomic coverge at a maximum r2 of 0.9 (Illumina Inc. ®). The required sample size in a case-control study for 80% power at an odds ratio risk of 1.3 with a minor allele frequency of 5% is 2,606 for this array (Illumina Inc. ®). It also offers an additional ~60,000 CNV-specific markers.

The second genetic approach that was conducted in this thesis is using candidate genes with specific rationale, which support the examination of the chosen genes. The rationales for choosing the following 17 genes of interest are described below and briefly summarized in Table 1.14.

As mentioned previously in the genetics and pharmacogenetics sections of this chapter, several important neurotransmitter systems have been consistently implicated in the genetic etiology of OCD. These systems include: glutamatergic, serotonergic, dopaminergic, and others such as neurotrophic factors and additional newly identified genes of unknown connection to OCD.

Table 1.14. Summary of Candidate Gene Studies of OCD.

Candidate Meta- SRI Gene Chromosome GWAS Gender AAO Y-BOCS Familiality Comorbidity Reference Gene Studies Analysis Response MOG 6p22.1 + + Zai et al., 2004 HTR1B 6q14.1 * - + + --- Corregiari et al., 2012; Denys et al., 2007; Mas et al., 2014; Miguita et al., 2011; Taylor, 2013 DLGAP2 8p23.3 GRIK2 6q16.3 ++ # Delorme et al., 2004; Sampaio et al., 2011; Stewart et al., 2013 SLC1A1 9p24.2 * +/- + + Stewart et al., 2013; Real et al., 2010; Taylor, 2013 BDNF 11p14 * -- ++ + +++ + + Hall et al., 2003; Katerberg et al., 2009; Márquez et al., 2013; Real et al., 2009; Rocha et al., 2010; Taylor, 2013; Tukel et al., 2014; Zai et al., 2015 GRIN2B 12p13.1 * - + + # Alonso et al., 2012; Qin et al., 2015; 71

Taylor, 2013 FAIM2 12q13.12 # Stewart et al., 2013 HTR2A 13q14.2 * + + +++-- Corregiari et al., 2012; Denys et al., 2007; Dickel et al., 2007; Miguita et al., 2011; Taylor, 2013; Tot et al., 2003; Zhang et al., 2004 SLITRK5 13q31.2 SLC6A4 17q11.2 *+ + Cengiz et al., 2015; Taylor, 2013 DLGAP1 18p11.31 - # - - + - Li et al., 2015; Stewart et al., 2013 FUT2 19q13.33 # Stewart et al., 2013 BTBD3 20p12.2 # Stewart et al., 2013 OLIG2 21q22.11 ++/- + Stewart et al., 2007; Zai et al., 2012; Zhang et al., 2015 COMT 22q11.21 * + + + Alsobrook et al., 2002; Taylor, 2013; Vulink et al., 2012 MAOA Xp11.3 * - Taylor, 2013 + indicates significant study - indicates negative study * indicates previous meta-analysis # indicates trend association with OCD 72

Glutamatergic System Genes:

Five glutamate-related candidate genes were selected for the genetics of OCD phenotypes and pharmacogenetics given their previous significant association with OCD diagnosis.

Glutamate receptor ionotropic kainite 2 (GRIK2) [MIM 113505] is the tenth top hit gene reported in the recent GWAS of OCD (Stewart et al., 2013). Functional studies have implicated its role in RNA editing of glutamate receptor 2 and 6 (Paschen et al., 1994) in addition to induction of long-term potentiation in the hippocampus (Contractor et al., 2001). It is mapped on chromosome 6q16.3 (Paschen et al., 1994) and two studies have implicated its involvement in the genetic etiology of OCD. Delorme et al. (2004) conducted a case-control study in 156 OCD patients and 141 controls in addition to 124 trios and detected a significant under-transmission of the I867 allele of the rs2238076 polymorphism in OCD (P<0.03). Sampaio et al. (2011) investigated the association of OCD and GRIK2 in 47 OCD small nuclear families and reported significant finding at the rs1556995 polymorphism (permutated P=0.03) and the two-marker rs1556995-rs1417182 haplotype (permutated P=0.01) with OCD diagnosis. Furthermore, significant associations of this gene have been reported in clozapine-induced obsessive- compulsive symptoms (Cai et al., 2013), autism (Jamain et al., 2002; Kim et al., 2007; Shuang et al., 2004), schizophrenia (Bah et al., 2004; Shibata et al., 2002), Huntington disease (Rubensztein et al., 1997) and mental retardation (Cordoba et al., 2015; Motazacker et al., 2007). Given the convergence of previous significant candidate gene studies and trend for association in the first GWAS of OCD in addition to the lack of study examining OCD phenotypes, further genetic analysis of this gene in OCD phenotypes was explored.

Neuronal glutamate transporter (SLC1A1) [MIM 133550] has been most consistently implicated in OCD and a meta-analysis investigating nine single nucleotide polymorphisms (SNPs) across the 3’ region of this gene supported its involvement in 306 OCD cases and 634 healthy controls (Stewart et al., 2013) although an earlier meta-analysis was negative (Taylor, 2013). A weak association was reported for the rs301443 variant (P=0.046; non-significant corrected P) with OCD and a slightly stronger association for the rs12682807 marker with OCD males only (P=0.012; non-significant corrected P) (Stewart et al., 2013). This gene has been mapped to chromosome 9p24.2 (Smith et al., 1994), which has been implicated as a 73

susceptibility region in OCD from previous linkage studies (Walitza et al., 2010). It encodes a high-affinity neuronal glutamate transporter excitatory amino acid carrier 1 (EAAC1) and is highly expressed in brained regions within the CSTC circuit (Kanai et al., 2004; Rothstein et al., 1994), which has been consistently implicated in the pathoetiology of OCD (Paul et al., 2014). In addition to its association with OCD diagnosis, Wu et al. (2013) detected a significant association of the rs10491734 polymorphism and two four-marker (rs10491734-rs2228622- rs301430-rs301443 A-A-C-C and A-G-C-C) haplotypes with early-onset OCD. The rs3056 polymorphism of this gene was reported to be associated with increased total (P=0.01), left (P=0.02), and right (P=0.02) thalamic volume in OCD patients (Arnold et al., 2009). Furthermore, one pharmacogenetic study reported a significant association between the rs301434 A allele and non-response to SRI response in 131 OCD patients (Real et al., 2010). Thus, additional genetic investigation of this gene is warranted to identify its role in the susceptibility risk of developing OCD and drug response in OCD.

Glutamate receptor, ionotropic N-methyl D-aspartate (NMDA) 2B (GRIN2B) [MIM 138252] is located on chromosome 12p13.1 (Endele et al., 2010) and encodes its receptor, which can be found at excitatory synapses throughout the central nervous system. It has important function in modulating synaptic plasticity, learning, and memory (Chen et al., 1999), maintaining long-term potentiation (Thomas et al., 1996), regulating neuronal morphogenesis (Setou et al., 2000) and cell death (Hardingham et al., 2002). Significant association of this gene was first reported by Arnold et al. (2004) for the rs890 marker (P=0.014) and the 5072G-5988T (rs890- rs1805502) haplotype (P=0.002). The same research team later reported a significant association of this gene with left orbitofrontal cortex (P=0.04) and right anterior cingulate cortex (P=0.02) volume in OCD patients (Arnold et al., 2009). In addition to OCD diagnosis and brain volume, Alonso et al. (2012) examined this gene and OCD subphenotypes and reported the following significant associations between: rs1805476 and OCD males (P=0.002), 4-SNP (rs1805476- rs1805501-rs1805502-rs1805477) haplotype and presence of contamination obsessions and cleaning compulsions (permutated P=0.023). Moreover, a recent GWAS detected association between the rs7972211 marker and SRI response (P=2.71×10-5). Although this gene has been implicated in OCD, only few studies have investigated its involvement in the pathogenesis of OCD.

SLIT and NTRK-like family member 5 (SLITRK5) [MIM 609680] is one of the top hits from the recent OCD GWAS (Stewart et al., 2013) and it is mapped on chromosome 13q31.2 (Aruga et al., 2003; Nagase et al., 1998). It was first implicated in OCD because the SLITRK5 knockout (deficient) mice were found to have impaired corticostriatal circuitry and develop obsessive- compulsive-like behaviours including excessive self-grooming and increased anxiety in the open maze test (Shmelkov et al., 2010). A member of the SLITRK family, SLITRK1 has been linked to Tourette’s syndrome (Abelson et al., 2005). This SLITRK gene family encodes integral transmembrane proteins with two N-terminal leucine-rich repeat domains, which are similar to those of SLIT proteins. Most of its members including SLITRK5 also have C-terminal regions that share homolog with neurotrophin receptors and are expressed predominantly in neuronal tissues. Aruga and Mikoshiba (2003) reported that SLITRK5 may have neurite-modulating activity but its exact function remains unknown. Given its functional implication and lack of genetic association studies in OCD, this gene has been studied in this thesis.

Discs large (drosophila) homolog-associated protein 1 (DLGAP1) [MIM 605445] is the fifth top hit gene and the strongest glutamatergic gene in the first published GWAS of OCD (Stewart et al., 2013). It maps on chromosome 18p11.31 and its encoded protein interacts with other members of synaptic ion channel clustering proteins, which can mediate the clustering of NMDA receptors and Shaker-type potassium channels (Kim et al., 1997; Satoh et al., 1997). DLGAP1 plays an important in the post-synaptic scaffolding of neuronal cells and synaptic transmission (Satoh et al., 1997). Only one association study has examined this gene more closely in OCD. Li et al. (2015) recently investigated this gene for OCD susceptibility and subphenotype risk in 320 Chinese Han OCD patients and 431 healthy controls. The author did not detect significant differences in allele or genotype frequencies of the rs11081062 polymorphism or when classifying by sex, AAO, and comorbidity; nonetheless, this marker was associated with the presence of contamination obsessions (P=0.021) and cleaning compulsions (P<0.001). In addition to OCD, studies have found positive association of this gene with schizophrenia (Li et al., 2013). Due to the limited number of genetic studies investigating DLGAP1 with potential functional implication, we attempted to explore this gene further in OCD phenotypes.

Discs large (drosophila) homolog-associated protein 2 (DLGAP2) [MIM 605438] is another glutamate-related gene, which is from the same family as DLGAP1 with similar functions (Ranta

et al., 2000). This gene is located on chromosome 8p23.3, which has been implicated in autism spectrum disorders (Chen et al., 2013; Pinto et al., 2010). A previous imaging genetic study reported the strongest trend between two SNPs in DLGAP2 (rs6558484 and rs7014992) and orbitofrontal cortex white matter volume (Wu et al., 2013). No study to date has examined this gene in OCD risk.

Serotonergic System Genes:

Three serotonin-related genes were chosen for the genetic study of OCD phenotypes. For the above list of genes, SLC6A4 appears to be most consistently implicated in antidepressant response in MDD, followed by HTR2A and HTR1A (Fabbri et al., 2013).

Serotonin 1B receptor (HTR1B) [MIM 182131] maps to chromosome 6q14.1 (Jin et al., 1992) and encodes a terminal auto-receptor that is involved in the regulation of serotonin release with wide physiological functions. The acute administration of non-selective (i.e., mCPP) or selective (i.e., sumatriptan) ligands of this receptor induces a transient worsening of OCD symptoms (Slassi, 2002). Evidence for its role in OCD came from significant genetic association studies with the first reported by Mundo et al. (2000). The authors observed a preferential transmission of the G allele of the rs6296 (G861C) polymorphism to OCD probands in 67 OCD families, which was confirmed in a subsequent family-based association study by the same research team (P=0.023) in 157 OCD families. Although a recent meta-analysis of the same marker in 11 datasets did not yield significant result (P=0.232) (Taylor, 2013), Mas et al. (2014) conducted a transmission disequilibrium study in 84 OCD trios and found that another marker, rs2000292, within this gene was associated with early-onset OCD males (P=0.0006) but not in females. Only three pharmacogenetics studies have been published, examining the rs6296 marker in SRI response in OCD and although they were all negative (Corregiari et al., 2012; Denys et al., 2007; Miguita et al., 2011), this gene has been implicated to predict antidepressant response based on an animal behavioural model using the forced swim test (Redrobe et al., 1996). Therefore, it is vital to examine other genetic variations within this gene in OCD phenotypes and SRI response.

Serotonin 2A receptor (HTR2A) [MIM 182135] encodes another abundant serotonin receptor with important role in many physiological processes including sleep, appetite, thermoregulation, pain perception, hormone secretion, and sexual behaviour. Abnormalities in the serotonergic neurotransmitter system have consistently been implicated in the etiology of OCD in addition to other psychiatric disorders. It is mapped to chromosome 13q14.2 (Sparkes et al., 1991) and has a specific role in modulating conflict anxiety, which is consistent with the cortical or “top-down” model of cognitive influence (Weisstaub et al., 2006). The initial genetic support of a genetic association was reported by Enoch et al. (1998) who found an association between the A allele of the rs6311 (G-1438A) promoter polymorphism and OCD. The significant results were later replicated in 55 OCD children/adolescents and 223 healthy controls by another group (Walitza et al., 2002). A family-based transmission disequilibrium study examined the same SNP in 54 OCD trios and detected nominal significant association in OCD subjects with comorbid tic disorder (P=0.05) (Dickel et al., 2007). Furthermore, this association was also confirmed with the meta-analysis in 19 datasets (mean OR 1.219, 99th percentile CI 1.037-1.433, P=0.002) (Taylor, 2013). Thus, markers across this gene should be further examined in OCD.

The serotonergic system genes have been widely studied in the pharmacogenetics of major depressive disorder (MDD) and the most robust findings supported a role of HTR2A in antidepressant response of MDD (Fabbri et al., 2013; Singh et al., 2013). Regarding OCD pharmacogenetics, five studies have examined HTR2A. Zhang et al. (2004) first reported that the homozygous genotypes of rs6311 (G-1438A) was associated with SRI response (P=0.030), which was supported by another study (Denys et al., 2007) observing a significant association between the G/G genotype and better paroxetine response (P=0.013); however, an earlier study detected no association (Tot et al., 2003). Conflicting results for other SNPs have been published. Miguita et al. (2011) found no significant association between clomipramine response and rs6313 or rs6305 in 41 OCD patients but Corregiari et al. (2012) detected significant association between the C/C genotype of rs6313 and non-responders (P<0.01) in 239 OCD patients. Therefore, there is a need to clarify the role of this gene in OCD SRI response.

Serotonin transporter (SLC6A4) [MIM 182138] encodes a high-affinity, sodium (Na+) and chloride (Cl-) dependent transporter that localizes in the presynaptic neuronal membranes, and is highly expressed in the central nervous system. SLC6A4 actively clears serotonin from the

synaptic cleft, which prevents serotonin from attaching to other serotonin receptor for down- stream actions and recycles it to the neurotransmitter pool (Ramamoorthy et al., 1993). It is the main target of SSRI antidepressants. This gene is localized to chromosome 17q11.2 (Ramamoorthy et al., 1993) and Lesch et al. (1996) reported functional significance between the variant of the 5HTTLPR polymorphism – the long allele associated with twice the basal activity than the short allele. The long allele was reported to be twice as common in OCD patients (N=169) than healthy controls (N=253), which was replicated to have a 2-fold over-transmission to OCD patients in a family association study with 175 OCD trios (Hu et al., 2006). Many studies have examined its effect on OCD risk and a meta-analysis of this long-short marker (5HTTLPR) was positive when combining 8 OCD datasets (mean OR 1.251, 99th percentile CI 1.048-1.492, P=0.001) (Taylor, 2013). Furthermore, another very recent small case-control study consisting of 80 OCD patients and 100 healthy controls reported that the G allele of another SNP, rs16965628, was associated with risk of OCD (P=0.02) (Cengiz et al., 2015).

Dopaminergic System Genes:

Catechol-O-methyltransferase (COMT) [MIM 116790] is localized to chromosome 22q11.21 (Brahe et al., 1996; Grossman et al., 1992). Its encoded protein, COMT, is an enzyme that catalyzes the transfer of methyl group from S-adenosylmethionine to catecholamines including the common neurotransmitters of dopamine, epinephrine, and norepinephrine. Karayiorgou et al. (1997; 1999) reported significant associations between the homozygous low activity genotype of the rs4680 (Val158Met) polymorphism and OCD males in two different studies, a case-control (73 OCD subjects and 148 healthy controls) and a family (110 OCD families) study; and another study detected nominal association of this low activity allele in OCD females (P=0.049) but not in males (Alsobrook et al., 2002). Significant finding was reported for this polymorphism in the recent meta-analysis using 25 OCD datasets (mean OR 1.200, 99th percentile CI 1.001-1.438, P=0.010). Although many studies have investigated COMT in relation to OCD risk, very few have examined other SNPs within this gene. Regarding pharmacogenetics of antidepressant response, Vulink et al. (2012) detected significantly more responders to citalopram with the rs4680 Met/Met genotype (P=0.007).

Monoamine oxidase A (MAOA) [MIM 309850] is the only examined gene that is localized to chromosome X, more specifically Xp11.3 (Levy et al., 1989). It encodes MAOA, which are expressed in the outer mitochondrial membrane. It functions to regulate amines by oxidizing neurotransmitters such as serotonin, norepinephrine, and dopamine, in addition to dietary amines, which is vital in maintaining mental states. Low levels of MAO activity and mutations in this gene have been associated with violent, criminal, and/or impulsive behaviour (Chen et al., 2004). MAOA inhibitors including phenelzine and tranylcypromine are third-line pharmacological treatment of OCD (CPA, 2006); however, given its side effect profile and dietary restriction, this drug class has not been widely prescribed. Very few studies have examined MAOA in OCD and Karayiorgou et al. (1999) reported a sexual dimorphic association between the allele with high MAOA enzymatic activity and OCD in 110 OCD families. However, a meta-analysis of six OCD datasets did not yield significant finding (P=0.483) (Taylor, 2013).

Neurotrophic and Neuroplasticity Gene:

Brain-derived neurotrophic factor (BDNF) [MIM 113505] is located at chromosome 11p14 (Hanson et al., 1992). BDNF is a member of the neurotrophin superfamily of survival- promoting molecules such as neurotrophin-3/4/5, and nerve growth factor (NGF), which play an important role in cell survival, differentiation, and cell death (Jones and Reichardt, 1990; Numakawa et al., 2010). Members of this family play a vital role in promoting neuronal growth and maintenance during normal development and differentiation of the vertebrate central nervous system (Hofer and Barde, 1988). BDNF specifically has trophic effects on neuronal plasticity and learning (Cowansage et al., 2010; Tyler et al., 2002). BDNF is therefore an attractive candidate for molecular analysis in OCD, based on potential functional relevance, and reported interaction between this gene and other neurotransmitters implicated in this disorder. BDNF is involved in both early brain development and adult synaptic formation and decay. There is a reciprocal relationship between BDNF and the serotonergic system: BDNF has important modulatory effects on serotonergic transmission (Celada et al., 1996; Martinowich and Lu, 2008; Mössner et al., 2000), and serotonergic inputs influence BDNF expression (Garcia et al., 2003;

Molteni et al., 2010; Nibuya et al., 1995). The interaction of serotonin and BDNF is particularly relevant given the large extant literature implicating involvement of serotonergic neurotransmission in OCD and the well described efficacy of serotonin reuptake inhibitors (SRIs) in the treatment of this disorder (Koran et al., 2007; Soomro et al., 2008; Zohar et al., 2000). Moreover, BDNF is also functionally interconnected with many neurotransmitters involved in OCD, including serotonin, glutamate, and dopamine. Several studies have indicated a possible effect of BDNF on the differentiation of serotonergic and dopaminergic neurons (Eaton et al., 1995; Studer et al., 1995; Zhou et al., 1994). BDNF enhances phosphorylation of the postsynaptic N-methyl D-aspartate (NMDA) receptor subunit 1 (Suen et al., 1997), an interaction that may be particularly relevant given the apparent role of glutamate in OCD based on genetic studies (Arnold et al., 2006; Rotge et al., 2010) and promising early reports of anti-obsessional efficacy of glutamate modulating agents (Wu et al., 2012). Further, two published reports describe lower peripheral BDNF serum (Maina et al., 2010) and plasma levels (Wang et al., 2011) in individuals with OCD, which was recently confirmed by another group (Fontenelle et al., 2012) and a meta-regression analysis by Suliman et al. (2013).

BDNF polymorphisms have been associated with a variety of psychiatric phenotypes such as mood disorders (Neves-Pereira et al., 2002; Sen et al., 2003; Sklar et al., 2002) and late age at onset in schizophrenia (Krebs et al., 2000; Muglia et al., 2003). Hall and colleagues first reported association for a number of BDNF polymorphisms in OCD, including the Met allele of the Val-66-Met polymorphism, a finding that was most significant in their early onset cases (Hall et al., 2003). Several other studies have also found significant association between OCD and BDNF including a five-marker protective haplotype containing the Val allele of the Val-66-Met polymorphism (P = 0.006) (Alonso et al., 2008) and Val as the protective allele in a case-control study (P < 0.01) (Rocha et al., 2010). Interestingly, another case-control study reported no significant differences in allele or genotype frequency between patients with OCD and healthy controls, however the Met/Met genotype was associated with later illness onset (P = 0.008) and a trend observed towards association of the Val/Val genotype and lower Y-BOCS scores (P = 0.051) in the female subgroup (Katerberg et al., 2009). Furthermore, Hemmings et al. (2013) recently detected that the Met allele increased the risk of OCD in individuals with a history of childhood emotional abuse. By contrast, the Val allele has also been associated with OCD diagnosis in the most recent study by Márquez et al. (2013) (P = 0.0001) and more severe OCD 80

in female OCD cases (Hemmings et al., 2008). The Val/Val genotype was significantly associated with OCD (Márquez et al., 2013; P = 0.0001) and hoarding symptoms (Timpano et al., 2011). However, there have also been six negative studies published in this population, including Dickel et al. (2007), Klaffke et al. (2006), Mossner et al. (2005), Tükel et al. (2012), Wendland and colleagues (2007), as well as our team in a previous study of both the Val-66-Met or BDNF-(GT)n polymorphisms with OCD (Zai et al., 2005). In addition to the Val-66-Met polymorphism, a recent study reported that in OCD patients the C allele of rs2883187 was associated with more severe symptoms and higher familial loading of OCD (Tükel et al., 2014). Despite these mixed findings and a recent negative meta-analysis of the Val-66-Met polymorphism (Zai et al., 2015), it appears possible that variable BDNF expression or function, due to DNA sequence variation, may contribute to the neuropathogenesis of OCD phenotypes. Furthermore, only one pharmacogenetic study has been published and Real et al. (2009) reported that the rs1491851 C allele was associated with better SRI response (P=0.005) whereas other SNPs within this gene was negative. Thus, further investigation is necessary to fully explore its role in OCD.

Myelin-Related Genes:

Brain imaging studies have identified potential neurobiological factors involved in the development of OCD (Kwon et al., 2009; MacMaster, 2010; Saxena and Rauch, 2000). Most studies have implicated abnormal link between the CSTC circuit, particularly the pathway that connects the orbito-frontal cerebral cortex to the basal ganglia (Graybiel and Rauch, 2000; Saxena and Rauch, 2000). Several early neuroimaging studies reported decreased white matter volume in OCD when compared with healthy controls (Breiter et al., 1994; Jenike et al., 1996; MacMaster et al., 1999; Rosenberg et al., 1997), which has led to questions regarding possible white matter dysfunction in OCD. This was recently confirmed by a recent meta-analysis of neuroimaging in OCD using multimodal voxel-based methodology (Radua et al., 2014), which reported widespread white matter abnormalities with the most robust finding of an increase in white matter volume and reduction in fractional anisotropy (FA) in the anterior midline tracts between the anterior parts of the cingulum bundle and body of corpus callosum. Diffusion tensor

imaging (DTI) studies comparing white matter of patients with OCD to healthy volunteers have reported lower FA bilaterally in the anterior cingulate gyrus (Szeszko et al., 2005), the rostrum of the corpus callosum, as well as the cingulum bundle (Saito et al., 2008). FA is reflective of the integrity of myelination and axonal function, with lower values reflecting disturbance in the unidirectional diffusion expected in normal axonal bundles. Cannistraro et al. (2007) reported greater left than right FA in the cingulum bundle and anterior limb of the internal capsule regions. Menzies et al. (2008) reported significantly reduced FA in a large region of right inferior parietal white matter and increased FA in the right medial frontal region in patients with OCD in addition to abnormal FA findings in their relatives. A recent article reviewed all DTI studies in OCD and confirmed that the majority of studies showed decreased FA in OCD patients when compared to healthy controls, with the most consistent white matter connectivity abnormalities found in the cingulate bundle, corpus callosum, and anterior limb of internal capsule (Koch et al., 2014). Thus, based on this literature, researchers have hypothesized that impairment in the development or maintenance of myelination might play a role in the etiology of OCD in addition to providing rationale for white matter as an endophenotype of OCD. This hypothesis has received some support from preliminary genetic studies implicating two myelination genes, the myelin oligodendrocyte glycoprotein (MOG; Zai et al., 2004) and the oligodendrocyte lineage transcription factor 2 (OLIG2; Stewart et al., 2007), in OCD.

Myelin oligodendrocyte glycoprotein (MOG) [MIM 159465] is mapped to chromosome 6p22.1 (Pham-Dinh et al., 1993), just distal to the human leukocyte antigen region, which contains genes that contribute to immune response. Its encoded protein is a minor component of the myelin sheath, which is expressed preferentially on the extracellular surface of oligodendrocytes (Pham- Dinh et al., 1993). It has been postulated that MOG functions as a cellular adhesion molecule, a regulator of oligodendrocyte microtubule stability, and/or a mediator of interactions between myelin and the immune system, particularly as an activator of the complement cascade (Johns and Bernard, 1999). It contains an immunoglobulin-like variable domain characteristic of members of the immunoglobulin super-family (Pham-Dinh et al., 1993) and variants of this gene may therefore be a target for immune-related demyelination and contribute to the development or progression of autoimmune disorders including multiple sclerosis (Lee and Linker, 2012; Reindl et al., 2013). An immune-related proposed mechanism of the development of OCD came from the Pediatric Autoimmune Neuropsychiatric Disorders Associated with (group A β-hemolytic) 82

Streptococcal (GABHS) infection (PANDAS), which has been characterized with an abrupt onset of OCD symptoms following GABHS infection (da Rocha et al., 2008; Snider and Swedo, 2003). Another term, Pediatric Acute-onset Neuropsychiatric Syndrome (PANS), has been proposed in 2012 to expand the clinical entity of the existing PANDAS by removing the cause (GABHS) (Swedo et al., 2012). Only one study has investigated genetic variants of this gene in OCD. Zai et al. (2004) a significant association between OCD and allele 2 at the tetranucleotide repeat (TAAA)n located in the 3’-untranslated region (P=0.022) in 160 small nuclear families with an OCD proband. The same allele 2 was found to be associated with highly Y-BOCS severity score (P=0.020). Additionally, the MOG haplotype C-C-13-2 [rs2251711-rs2071653- (CA)n-(TAAA)n] was significantly over-transmitted from parents to their OCD-affected offspring (P=0.011). Atmaca et al. (2010) provided further support for this gene in the etiology of OCD, reporting significantly larger total white matter volumes in OCD patients carrying the (Val142Leu) Val/Val genotype (P<0.01).

Oligodendrocyte lineage transcription factor 2 (OLIG2) [MIM 606386] encodes for a protein that is an essential regulator in the development of human white matter (myelin)-producing cells (Takebayashi et al., 2002) and it is highly expressed in the amygdala, caudate nucleus, and thalamus, brain regions that have been associated with OCD (Rauch and Savage, 1997; Saxena et al., 1998). It may act as a transcription factor required for oligodendrocyte progenitor cell differentiation into different lineages during re-myelination throughout life (Ligon et al., 2006). It is mapped on chromosome 21q22.11 (Georgieva et al., 2006; Wang et al., 2000). A family- based association study of this gene first reported an over-transmission of the rs9653711 C allele (P=0.0035), over-transmission of the G-A-T-T-C (rs762178-rs1059004-rs9653711-rs6517137- rs13046814) haplotype (P=0.020), and an under-transmission of the A-C-T-T-G haplotype (P=0.0035) in 66 families with 160 OCD probands (P=0.0035) (Stewart et al., 2007). Two studies have been published since with one negative and one positive result. Zai et al. (2012) failed to replicate these findings in 160 OCD small nuclear families using the family-based association approach. However, a recent study supported the initial findings and demonstrated that in addition to the rs9653711 polymorphism, rs762178 and rs1059004 were also associated with OCD diagnosis and early-onset OCD in a case-control study comprising of 400 OCD subjects and 459 healthy controls (Zhang et al., 2015). Thus, this gene may be an interesting candidate to examine OCD subphenotypes in OCD. 83

Other Genes:

One large study (Chapters 4 and 6) mainly focused on the remote regulatory regions (Zai et al., 2014), testing markers with potential functional impact on several interesting top-hit OCD candidate genes from the first GWAS (Stewart et al., 2013): fas apoptotic inhibitory molecule 2 (FAIM2), fucosyl-transferase 2 (FUT2), and BTB (POZ) domain containing 3 (BTBD3).

Fas apoptotic inhibitory molecule 2 (FAIM2) [MIM 604306] is the eighth top hit gene in the OCD GWAS (Stewart et al., 2013). It is mapped to chromosome 12q13.12 (Somia et al., 1999) and has been implicated in obesity given its role in apoptosis (Williams et al., 2012). Its encoded protein is a 35.1 kDa membrane protein, which is highly expressed in the hippocampus and it regulates Fas ligand-mediated apoptosis in neurons (Fernandez et al., 2007). Specifically, FAIM2 is a neuron-specific inhibitor of Fas/CD95-mediated apoptosis with potential neuroprotective effect (Reich et al., 2011). Evidence has shown that this gene is regulated by nutritional state and methylation levels of the FAIM2 promoter region have been found to be associated with obesity (Wu et al., 2015). However, the molecular mechanism by which this gene affects obesity has yet to be clarified. Moreover, its role in psychiatric disorders or drug response has not yet been determined.

Fucosyl-transferase 2 (FUT2) [MIM 182100] is one of the top hit genes from the GWAS of OCD (Stewart et al., 2013) and is localized onto chromosome 19q13.33 (Rouquier et al., 1995). It has been found to be positively predict plasma vitamin B12 levels in two GWASs (Hazra et al., 2008; Tanaka et al., 2009) and associated with Crohn’s disease (McGovern et al., 2010). Interestingly, vitamin B12 serum level was reported to be abnormally low in OCD patients when compared to healthy controls (Hermesh et al., 1988). A case report of vitamin B12 deficiency was associated with OCD symptoms (Sharma and Biswas, 2012). A recent study also implicated a low vitamin B12 level in OCD patients (Türksoy et al., 2014). Thus, this gene may be associated with OCD in the context of low vitamin B12 serum level. Exploration of this gene in psychiatric disorders or drug response has not been done.

BTB (POZ) domain containing 3 (BTBD3) [MIM 615566], localized on chromosome 20p12.2, is the strongest associated gene in the trio sample of the first GWAS of OCD with a P value of 3.84E-08 (Stewart et al., 2013). BTBD3 has multiple cellular functions including the regulation of dendritic orientation (Matsui et al., 2013), cytoskeleton dynamics, transcriptional regulation, ion channel assembly and gating, protein ubiquitination and degradation (Perez-Torrado et al., 2006). Its close family of transcription factors, BTBD9, has been found to be associated with Tourette syndrome, which is frequently comorbid with OCD (Riviere et al., 2009); however, its role in other psychiatric disorders or drug response remains currently unknown.

These candidate genes described above have been selected based on their previously significant genetic association with OCD and several were recent top hits from the first published GWAS of OCD (Stewart et al., 2013). The rationale for SNP selection was mainly based on the ENCyclopedia Of DNA Elements (ENCODE) project, which reported that 80.4% of the human genome displays some functionality in gene regulation, gene-gene, and gene-environment interaction (Kavanagh et al., 2013). This provides strong support for the examination of genetic variations across these remote regulatory regions in terms of OCD phenotypes and drug response in OCD.

1.5.1 Significance of Remote Regulatory Regions

The human genome provides the fundamental script for human biology. However, our current understanding of the genome is still underway given that less than 3% of the genome is devoted to protein coding. The rest of the genome was once thought to be made up of “junk” DNA with unknown function; nonetheless, recent advances in human genetics identified over 80% of the genome as containing sequences of non-coding ribonucleic acids (RNAs), alternatively spliced transcripts, and regulatory regions with important regulatory functions (ENCODE Project Consortium, 2012; Kavanagh et al., 2013; Nair and Howard, 2013) (Figure 1.2). The identification of these newly discovered regulatory elements has provided researchers new insights into the mechanisms of gene regulation.

Previous OCD genetic studies have mainly examined known genetic variations within or surrounding genes of interest (Pauls et al., 2014). Almost none has investigated remote regulatory regions of these candidate genes. As mentioned above, the notion that our genome is made up of mostly “junk” DNA was disproven after the ENCyclopedia Of DNA Elements (ENCODE) project reported 80.4% of human genome displaying some functionality into gene regulation, gene-gene, and gene-environment interactions (Kavanagh et al., 2013; Nair and Howard, 2013). This provides a strong evidence of support to examine genetic variations across these remote regulatory regions in psychiatric disorders.

Defining the regulatory regions that control gene activity has become a key biomedical issue. Recent technological advances have provided an increasing repertoire of methods to identify regulatory elements. Regulation of gene expression is controlled by the sequence- specific binding of transcription factors (TFs) that may act as activators, repressors, or both, activating or preventing transcription. As illustrated recently by the ENCODE project (Figure 1.2), chromatin immuno-precipitation (ChIP) is a powerful assay that is used to detect TFs that attach to TF-binding sites (regulatory elements), chromatin-associated markers, and transcription-associated protein complexes. This invaluable tool helps to catalog sequences with characteristics of regulatory elements (ENCODE Project Consortium, 2012). Examples of functional regulatory elements include RNA transcribed regions, TF-binding sites for cis- regulatory modules, repressors, activators, enhancers, and silencers, chromatic structure, and DNA methylation sites. The ongoing discovery of these functional regulatory elements will allow researchers to address specific underlying mechanism to further our understanding of psychiatric disorders. This will in turn accelerate gene discovery efforts, leading to novel therapeutics.

One of the major studies in this thesis examines genetic variations that have functional implications on their respective genes. There are many different publicly available search engines including functional significance (FS) score, European minor allele frequencies (EurMAF), BrainCloud (Colantuoni et al., 2011), UCSC Genome Browser (Rosenbloom et al., 2013), HaploReg (Ward et al., 2012), RegulomeDB (Boyle et al., 2012), and National Institute of Environmental Health Sciences (NIEHS) (Xu et al., 2009), which facilitate the identification of SNP with higher allele frequency for meaningful genetic analysis and known functions.

Thus, one of the goals of this PhD thesis is to utilize these publicly available databases to identify SNPs in the remote regulatory regions of OCD candidate genes and to examine the impact of these SNPs in OCD subphenotypes and drug response.

1.5.2 Genetic Methodology

Traditional genetic studies in psychiatric disorders have focused on the hypothesis-driven candidate gene approach. Candidates were chosen based on the putative biological mechanisms and drug targets of existing psychotropic medications. With the advances in genomic technology, genome-wide association studies (GWASs) have the ability to screen for genetic variations across the entire genome with varying degree of coverage. By using the latter approach and in contrast to the candidate gene studies, a hypothesis-free screening for genetic involvement can be made possible. Given the large amount of generated data, the significance level threshold has to be corrected to 5E-08 for multiple comparisons. Therefore, the studied sample size will have to be substantially enlarged in the genetics of psychiatric disorders given the relatively smaller effect size of genetic findings to date.

The goal of GWAS is to detect signals across the entire genome but it is unable to fine map the region of interest. Additional algorithms have been developed to test for specific hypotheses in the context of GWAS. Candidate gene studies have the advantage of tailoring the SNP selection to cover all common variations across each gene, taking into the consideration of functional significance of each SNP.

The increasing use of GWAS approach has shifted stand-alone institutional contributions to large-scale international collaboration because the sample sizes required for GWAS discovery and replication are beyond the reach of a single research group. This shift has led to the establishment of the Psychiatric Genomics Consortium (PGC) since the early 2007. PGC currently consists of over 800 investigators from 38 countries and over 900,000 samples have been collected for analysis. To date, there are only two GWASs of OCD diagnosis, which did not detect any genome-wide significant findings. Hence, in this thesis, we have explored OCD phenotypes in order to better understand the complexity of this illness.

Figure 1.2. Annotation of Disease-Associated Variants from ENCyclopedia Of DNA Elements (ENCODE) Data (Kavanagh et al., 2013). LD: linkage disequilibrium; TSS: transcription start site; TFBS: transcription factor binding site

Chapter 2 AIMS, OBJECTIVES, HYPOTHESES, AND IMPLICATIONS

2 Thesis Aims, Hypotheses, and Implications

The main purpose of this thesis is to aid in the discovery of genetic influence in the etiology of OCD by examining its subphenotypes. There are many approaches in genetic studies and this thesis will only focus on common genetic variations that affect risk of developing specific OCD subphenotypes in addition to predicting response to serotonin reuptake inhibitor (SRI) antidepressant treatment in OCD.

2.1 Aims and Objectives

Prior to investigating the genetic basis of OCD subphenotypes, a comprehensive examination of the clinical phenomenology of OCD was necessary as a precursor to genetic studies. Several OCD subphenotypes were explored including age at onset (AAO), Yale-Brown Obsessive-Compulsive Scale (Y-BOCS) symptom severity, symptom dimensions, comorbid psychiatric conditions, family history, and SRI response.

In determining the genetic architecture of OCD subphenotypes and drug response in OCD, serotonin system genes are obvious candidates given the extensive literature supporting serotonergic mechanisms as etiologically important in this illness. Moreover as the first-line agents for this condition, all SRIs exert their effects by modifying serotonergic neurotransmission, and this provides further support for the importance of this pathway. Additionally, evidence suggests that the neuropathology of OCD lies in the complex neurotransmitter network of the cortico-striato-thalamo-cortical (CSTC) circuit, where dopamine and glutamate dysfunctions also have been implicated in this disorder (Pauls, 2010; Pauls et al., 2014). Genotyping data derived from a genome wide association study (GWAS) of OCD (Stewart et al., 2013), of which our group was a contributor, provides an interesting opportunity to explore the genetic basis of OCD subphenotypes and drug response in OCD. 89

2.1.1 Aims

The following is a list of objectives for this PhD thesis:

1) To identify and refine clinically useful subphenotypes in OCD for genetic studies;

2) To develop further understanding of the complex molecular and genetic basis of OCD;

3) To conduct statistical analyses of complex traits in OCD using different genetic approaches including candidate gene study and non-hypothesis driven GWAS.

2.1.2 Objectives

There are two main sections of this thesis. The initial objective is to examine the complex clinical characteristics of OCD and to analyze them in an approach that reduces the dimensions of the data, therefore making it more amenable to genetic analysis.

The second objective of this thesis is to utilize genetic methodology to examine these OCD subphenotypes in order to identify genetic variants that contribute to:

1) The susceptibility risk of developing a particular subgroup of OCD (distinguishing OCD subphenotypes), and;

2) SRI response in the treatment of OCD.

This thesis will attempt to dissect the various shared and distinct genetic variations that influence the risk of developing certain subtypes of OCD, earlier onset, severity, familial form, additional comorbid psychiatric conditions, and predict response to SRIs.

2.2 Hypotheses 2.2.1 Clinical Phenomenology of OCD

The null hypotheses of the clinical phenomenology of OCD are that:

1. There are no differences within gender and AAO groups;

2. The Y-BOCS symptom checklist will generate the similar symptom dimensional factors as previous published studies;

3. The rates of significant family history and comorbid psychiatric condition(s) are the same as published reports.

The alternative hypotheses of the clinical phenomenology of OCD are that:

1. There are significant differences within gender and AAO groups;

2. The Y-BOCS symptom checklist will generate different symptom dimensional factors from previous published studies;

3. The rates of significant family history and comorbid psychiatric condition(s) are different from published reports.

2.2.2 Genetics of OCD Phenotypes

The null hypothesis of the genetics of OCD phenotypes is that genetic variations within the serotonergic system (i.e., serotonin transporter [5HTT/SLC6A4], serotonin 2A receptor [HTR2A]) and the glutamatergic system (i.e., glutamate receptor, ionotropic, N-methyl D- aspartate 2B [GRIN2B], glutamate transporter [SLC1A1]) candidate genes are not associated with any phenotypes (i.e., AAO, Y-BOCS severity, OCD symptom dimensions, psychiatric comorbidity, family history) in OCD.

The alternative hypothesis is that genetic variations within the serotonergic system (i.e., 5HTT/SLC6A4, HTR2A) and the glutamatergic system (i.e., GRIN2B, SLC1A1) candidate genes are associated with specific subphenotypes (i.e., AAO, Y-BOCS severity, OCD symptom dimensions, psychiatric comorbidity, family history) in OCD (Figure 2.1).

2.2.3 Genetics of SRI Response in OCD

The null hypothesis of the genetics of SRI response in OCD is that genetic variations within the serotonergic system (i.e., 5HTT/SLC6A4, HTR2A) and the glutamatergic system (i.e., GRIN2B, SLC1A1) candidate genes are not associated with SRI response in OCD.

2.3 Originality

This is the first attempt to utilize a GWAS approach to examine genetic susceptibility of OCD subphenotypes. Furthermore, given the recent interest in regulatory elements from the ENCyclopedia Of DNA Elements (ENCODE) project (ENCODE Project Consortium, 2012), it is important to examine regulatory regions in the vicinity of top OCD candidate genes, regions that can potentially alter gene function and protein activity.

2.4 Implications

This thesis is designed to investigate the genetic susceptibility and pharmacogenetics of OCD. Thus, we may better understand genetic factors for risk of developing various forms of OCD. The identification of genetic factors that contribute to SRI response is vitally important in identifying those who are responders. The clinicians will be assisted in choosing the best tolerated medications in order to prevent the significant morbidity and mortality in patients with 92

OCD. Furthermore, by developing an algorithm to determine the dimensions of OCD symptoms, it can help researchers and clinicians to identify a more homogenous OCD group with potentially different response to a given SRI.

Figure 2.1. Thesis Hypothesis. This figure illustrates the hypothesis of this Ph.D. thesis – different genetic variations may predispose to different phenotypes of OCD and there may be shared genetic factors influencing the risk of OCD phenotypes.

Chapter 3 CLINICAL PHENOMENOLOGY OF OCD

3 An Update on the Clinical Phenomenology of Obsessive-Compulsive Disorder

Gwyneth Zai 1,2,4, David Pauls 3, Vincenzo de Luca 1,2, Marissa Williams 4, Karen Wigg 4, James L. Kennedy 1,2, Margaret A. Richter 1,2,4

1 Neurogenetics Section, Centre for Addiction and Mental Health, Clarke Division, Toronto, ON, Canada

2 Department of Psychiatry, University of Toronto, Toronto, ON, Canada

3 Psychiatric and Neurodevelopmental Genetics Unit, Center for Human Genetic Research, Massachusetts General Hospital, Boston, MA, USA

4 Frederick W. Thompson Anxiety Disorders Centre, Department of Psychiatry, Sunnybrook Health Sciences Centre, Toronto, ON, Canada

* Corresponding author: Dr. James L. Kennedy, Director of the Neuroscience Department and

Head of Neurogenetics Section at the Centre for Addiction and Mental Health (CAMH),

Professor of the Department of Psychiatry and the Institute of Medical Science at University of

Toronto, Toronto, ON M5T 1R8, Canada (Phone: 1-416-979-4987; Fax: 1-416-979-4666; E- mail: [email protected])

3.1 Abstract

Background: Obsessive-compulsive disorder (OCD) is a complex and debilitating disorder with high rate of comorbid psychiatric illness. Evidence suggests that OCD is a heterogeneous condition. Existing genetic studies in psychiatric disorders have been limited by poor reproducibility and the use of current diagnostic systems, which lacks biological justification. The main goal of this chapter is to identify homogeneous subphenotypes of OCD that are clinically relevant and suitable for future genetic approaches with greater effect size in search for the underlying mechanism in the etiology of OCD.

Methods: We explored demographics, clinical presentations, and treatment outcome in a large Canadian OCD sample consisting of 560 OCD individuals and 273 family members. Examined variables included: gender, age at onset (AAO), OCD symptoms and severity as measured by the Yale-Brown Obsessive-Compulsive Scale (Y-BOCS), DSM-IV psychiatric comorbidity, family history of obsessive-compulsive and related disorders (OCRDs), and serotonin reuptake inhibitor (SRI) response. The distribution of AAO was explored using admixture analysis in the STATA program in order to identify statistically calculated subgroups of AAO, and symptom dimensional factors were derived using the factor and cluster analyses in the SPSS software on the Y-BOCS symptom checklist. Frequency of comorbid psychiatric disorders including lifetime OCRDs and current psychiatric disorders was calculated using the diagnoses that were determined from the SCID-IV interview. Family history of OCRDs including OCD, hoarding disorder (HD), trichotillomania (TTM), body dysmorphic disorder (BDD), and skin picking disorder (SPD) was examined using a modified version of the Family History Interview (FHI). Serotonin reuptake inhibitor (SRI) response was determined retrospectively using the Clinical Global Impression – Improvement (CGI-I) scale. Clinical measures were compared between gender, statistically derived AAO subgroups (early-, intermediate-, and late-onset), familial status, and SRI response status (responders and non-responders).

Results: Female OCD subjects in our sample had a higher percentage of comorbid lifetime OCRDs when compared with OCD males but no differences in other clinical characteristics were detected. AAO analysis showed a tri-modal distribution, separating early (≤7 years), 96

intermediate (8-17 years), and late (≥18 years) onset groups. The early onset group showed significantly more symmetry/order and contamination/cleaning symptoms. Factor analysis for Y-BOCS confirmed a 5-factor model (factor 1: symmetry/order; factor 2: aggression/sexual/religious; factor 3: hoarding; factor 4: contamination/cleaning; factor 5: somatic) when treating all symptoms equally, and detected a 6-factor model (factor 1: symmetry/order; factor 2: hoarding; factor 3: contamination/cleaning; factor 4: aggression/checking; factor 5: somatic; factor 6: sexual/religious) when double-weighting the principal/target symptoms (where aggression/sexual/religious dimension in the 5-factor model was divided), which were both confirmed with the cluster analysis. Fifty-nine percent and 24% of the OCD participants had one or more current comorbid psychiatric diagnoses and lifetime OCRDs respectively. Over 21% of the OCD individuals in our sample had a significant family history of OCRDs and only one individual had an overlapping comorbid and family history of the same condition (TTM). Familial OCD cases had less comorbidity. The SSRI/SRI antidepressant response rate in our OCD sample was 70.9%/74.7%. SRI responders had more symmetry/order symptoms, higher rates of comorbid mood disorders and family history of OCD and hoarding when compared to non-responders.

Conclusions: This is one of the largest North American OCD sample with an extensive investigation of clinical phenomenology. We derived clinically relevant and easily interpretable subphenotypes in OCD, which also detected significantly different clinical characteristics between the identified groups. The identification of more homogeneous subphenotypes will aid in genetic studies of OCD and should enable more definitive detection of genes contributing to different aspects of this complex disorder.

3.2 Introduction

Obsessive-compulsive disorder (OCD) is a chronic and debilitating psychiatric disorder that affects approximately 2% to 3% of the general population (APA, 2000; Ruscio et al., 2010). This disorder is characterized by obsessions that are recurrent intrusive thoughts, images, or urges, and compulsions, which are repetitive rituals or behaviours performed in order to counteract the anxiety-provoking obsessions (APA, 2013). These symptoms are typically 97

egodystonic and are frequently thought of as irrational and excessive by patients. The DSM-5 classifies OCD as one of the obsessive-compulsive and related disorders (OCRDs), which are a new category of disorders that have been separated from the DSM-IV anxiety disorders (APA, 2013). In addition to OCD, this new diagnostic category of OCRDs includes body dysmorphic disorder (BDD), hoarding disorder (HD), trichotillomania (TTM) or hair-pulling disorder, and excoriation (skin-picking) disorder (SPD) (APA, 2013). Hoarding used to be considered a symptom of OCD (APA, 2000) but given its distinct clinical presentation, treatment outcome, and genetic influence, it is now considered a separate disorder within OCRDs (APA, 2013).

The clinical presentation of DSM-IV OCD is highly heterogeneous given that there are many different symptom characteristics such as fear of contamination and cleaning rituals, obsessions and compulsions with symmetry and ordering, checking, hoarding, aggressive, religious, sexual, and somatic nature, as well as hoarding (APA, 2000) prior to the introduction of DSM-5. Furthermore, phenotypic heterogeneity has been supported by many epidemiological and clinical studies of OCD in terms of symptom subtype or dimension, age at onset (AAO), and treatment outcome.

To better understand the complexity of OCD, it may be important to delineate the diagnosis into different phenotypically distinct subtypes. Furthermore, unique and common pathophysiological mechanisms may be easier to identify by studying more homogeneous diagnostic subgroups of this disorder (Robins et al., 1970).

Researchers have postulated different schemes to ascertain different subgroups of OCD. OCD can be subdivided according to gender (Torresan et al., 2013), AAO (de Mathis et al., 2008; Janowitz et al., 2009; Taylor et al., 2011), comorbidities such as the presence or absence of schizotypal personality traits (Brakoulias et al., 2014; Jin Lee et al., 2006), obsessive-compulsive personality traits (Coles et al., 2008; Wetterneck et al., 2011), tics (Gomes de Alvarenga et al., 2012; Miguel et al., 1997), hoarding (Samuels et al., 2007; Torres et al., 2012), and other disorders (Murphy et al., 2010), in addition to treatment outcome (Nakamae et al., 2011; Sumitani et al., 2006), and symptom dimensions (Bloch et al., 2008; Fontenelle et al., 2005; Hasler et al., 2007). However, no one study has comprehensively examined a large OCD sample with regard to individuals’ demographics, clinical characteristics, and other OCD subphenotypes.

We aimed to explore the relationship of gender, age at onset (AAO), Yale-Brown Obsessive- Compulsive Scale (Y-BOCS) symptom checklist dimensions, psychiatric comorbid conditions, family history of obsessive-compulsive and related disorders (OCRDs), and serotonin reuptake inhibitor (SRI) response in a large well-characterized OCD sample. The rationale supporting our objective to examine gender and these OCD subphenotypes is explained below.

Although the gender ratio of OCD is 1:1 (APA, 2000), male children and adolescents have a slightly higher rate of OCD than females (Fineberg et al., 2013). Differences in clinical presentation between OCD males and females have previously been documented; nonetheless, there is one study that did not detect any gender differences (Matsunaga et al., 2000). Men suffering from OCD tend to have a chronic course of illness (de Mathis et al., 2011; Nestadt et al., 1998; Rosso et al., 2012) with greater symptom severity (Nestadt et al, 1998; Rosso et al., 2012) and they endorse more symptoms related to aggression (Tukel et al., 2013), sexual themes, religious themes, symmetry or exactness, and hoarding behaviours (de Mathis et al., 2008; Jaissorya et la., 2009; Rosso et al., 2012; Tukel et al., 2013). On the other hand, females with OCD have been reported to have an acute onset of symptoms and an episodic course of illness (de Mathis et al., 2011; Nestadt et al., 1998; Rosso et al., 2012); they also appear to develop symptoms associated with contamination and somatization more often than OCD males (de Mathis et al., 2008; Jaisoorya et al., 2009; Rosso et al., 2012; Tukel et al., 2013).

As mentioned previously, males appear to have an earlier onset of OCD and AAO for OCD has been reported to be in the range of 10 and 37 years although most cases have emerged by the age of 22 years (Fineberg et al., 2013). Previous studies have reported differences in the clinical characteristics of OCD depending on early- or late-onset; however, most have arbitrarily set the cut-off age for early- and late-onset without any validation (Chabane et al., 2005; Grant et al., 2007; Janowitz et al., 2009; Maina et al., 2008). Only four studies to date have examined AAO with a more in-depth approach using admixture analysis to identify subgroup of OCD (Albert et al., 2015; Anholt et al., 2014; Delorme et al., 2005; De Luca et al., 2011) and all confirmed a bimodal distribution in which differences in the clinical characteristics were observed between early- and late-onset groups. These independent studies implicated the potential utility of AAO to ascertain clinically informative subgroups of OCD. However,

100

variations in the differing clinical features for the AAO groups between the studies were substantial, suggesting the need for further research.

Variable clinical symptoms of OCD have been described and no one patient presents with the exact same severity and clusters of symptoms. The Y-BOCS symptom checklist has 74 items related to a distinct OCD symptom. Thus, this measure enables ascertainment of the breadth of symptoms for clinical outcome comparison, treatment guidance, and genetic studies. Multiple studies have utilized factor and cluster analyses to reduce the number of OCD symptoms identified with the Y-BOCS into meaningful and informative symptom dimensions. Although the majority of studies have identified a 5-factor model with factors including contamination/cleaning, hoarding, aggressive/checking, symmetry/ordering, and sexual/religious (Table 1.6), differences in clinical characteristics amongst the groups were reported (de Mathis et al., 2008; Jaisoorya et al., 2009; Rosso et al., 2012; Tukel et al., 2013). The identification of dimensional structure of OCD symptoms using this quantitative approach has the potential to advance our understanding of OCD. This is an attempt to derive OCD symptom dimensions in our large and well-characterized sample of OCD because these dimensional phenotypes may be useful in our future efforts to understand the natural history, genetics, neurobiology, treatment response, and outcomes of OCD (Leckman et al., 2007; Mataix-Cols et al., 2005).

Large-scale epidemiological studies have reported a high lifetime and current rate of comorbid psychiatric conditions in OCD patients, ranging from 76% to 91% and 42% and 55% respectively (Ruscio et al., 2010; Hofmeijer-Savink et al., 2013). Anxiety and mood disorders are amongst the most frequent comorbid conditions in individuals who suffer from OCD (Ruscio et al., 2010; Hofmeijer-Savink et al., 2013). Although these psychiatric disorders are distinct, they do share similar features and may form a basis for investigating the cross-disorder etiological overlap in the near future. However, most clinical studies only examine the difference in OCD presentation between the presence and absence of tics (do Rosario-Campos et al., 2005; Nestadt et al., 2009). The putative tic-related subtype may account for 10% to 40% of the childhood-onset OCD cases (Leckman et al., 1994). OCD individuals with a personal history of tics have been reported to show a male predominance with earlier onset and greater symmetry and hoarding symptoms (Baer, 1994; Leckman et al., 1997).

100

101

A higher rate of OCD in relatives of OCD patients, two times in OCD adults and 10-fold in OCD children and adolescents, when compared to the general population has been consistently reported (Mataix-Cols et al., 2013; Pauls, 2010; Pauls et al., 2014). When comparing clinical characteristics between familial and sporadic cases of OCD, familial OCD was associated with earlier onset of diagnosis and longer duration of untreated illness in addition to reporting different types of symptoms such as sexual obsessions and cognitive compulsions (Viswanath et al., 2011), ordering and arranging compulsions (Hanna et al., 2005) and higher comorbid OCRDs especially SPD and anxiety disorders (Hanna et al., 2005) in addition to major depression (Viswanath et al., 2011). Furthermore, familial OCD was found to be associated with treatment resistance (Pallanti and Grassi, 2014). Although multiple studies have supported the familial relationship between OCD and other OCRDs including BDD, TTM, SPD, and pathologic nail biting, no studies to date have examined the difference in clinical characteristics between the OCD individuals with and without a family history of OCRDs. These reported relationships of OCRDs may support the potential overlapping genetic risk between OCD and other OCRDs.

Hence, there is a critical need for deconstructing the current OCD diagnosis into different clinically relevant subgroups to determine the underlying mechanisms of OCD, to improve response to current treatment, and to provide basis for new therapies. Thus, the goal of this study is to examine the clinical phenomenology in a large OCD sample regarding gender, AAO, Y- BOCS symptom dimensions, psychiatric comorbidities, familiality, and SRI antidepressant treatment response.

3.3 Methods 3.3.1 Diagnostic Criteria and Clinical Sample

102

was defined as when the OCD subject first met diagnostic criteria of OCD (retrospective recall of when OCD symptoms first caused significant impairment or distress). The Yale-Brown Obsessive-Compulsive Scale (Y-BOCS) (Goodman et al., 1989) was completed for each participant to determine the severity of OCD symptoms. Lifetime severity of OCD symptoms was estimated by re-administering the Y-BOCS scale whiles individuals focused on the time period when the most severe OCD symptoms were experienced.

All assessments and interviews were performed by trained research assistants who were supervised by an experienced psychiatrist. All participants met DSM-IV criteria for the diagnosis of OCD as determined by the SCID and available medical records. Psychiatric comorbid conditions were also determined by the SCID interview and OCRDs were defined using the DSM-5 diagnostic criteria. The presence of tics was determined by DSM-IV diagnostic criteria for tic disorder and Tourette’s syndrome during the SCID-IV interview. We did not screen participant for any Axis II psychiatric disorders in this study protocol.

All subjects provided their written informed consent to participate in this study, and ethics approval was obtained from the local Research Ethics Board. Inclusion criteria included adults aged between 18 and 65 years who met DSM-IV criteria for OCD diagnosis with more than one year symptom duration or more than one year since onset if waxing/waning course. Exclusion criteria included any metabolic or chronic neurological disease (other than tic disorder), active substance dependence or abuse (OCD in the context of active substance use, i.e., include OCD participant if OCD predated substance use), schizophrenia, or schizoaffective disorder. Psychiatric comorbidities were obtained from the SCID assessment.

3.3.1.1 Description of Y-BOCS Symptom Checklist and Severity Score

The Y-BOCS (Goodman et al., 1989) is the gold-standard, validated and clinician- administered instrumental measure that consists of a symptom checklist and severity scale (Appendix I). The symptom checklist has 74 specific symptom items that are clustered into 15 different groups of obsessions or compulsions including two miscellaneous categories, each for obsessions and compulsions. Each checklist item is rated according to whether the symptom

102

103

occurs currently in the present (past week), in the past, or never occurs. The five most interfering, severe, or distressing obsessions and compulsions to a total of 10 are identified and ranked as the “target” or “principal” symptoms. The Y-BOCS 10-item index of severity measures the degree of control and resistance that the patient has over his/her obsessions and compulsions, time spent on these obsessions and compulsions, in addition to the interference and distress caused by them. This severity scale is divided into two subscales for obsessions and compulsions (five questions each). Each question is scored between 0 and 4 with 0 as none and 4 as most severe. The Y-BOCS severity score can range from 0 (non-clinical) to 40 (most severe). A score less than 7 is considered to be non-clinical and a score great than or equal to 16 indicates clinically significant OCD symptoms.

The Y-BOCS has two versions, a clinician-rated and self-report questionnaire. The clinician-rated version has been reported to have high inter-rater reliability and internal consistency (α=0.89) (Goodman et al., 1989; Steketee et al., 1996); therefore, we have chosen to use the clinician-reported version of Y-BOCS.

3.3.2 Familiality Data to Determine the Presence of Absence of Family History of OCRDs

Family history data for all available first- and second-degree relatives were obtained using the OCD Family History Interview (OCD-FHI), which is a semi-structured interview developed by the investigators (Drs. Richter and Kennedy; see Appendix II for details). It was modeled on the Family History Research Diagnostic Criteria (Andreasen et al., 1986), yielding diagnoses of OCD and relevant spectrum conditions (OCRDs). Reliability of information was also collected due to recall bias of the participants and their family members. The criteria of defining a significant (positive) family history of OCRDs included either of the following:

1) A clinical diagnosis of OCRDs, or

2) Reliability measure with an informant providing accurate information or close relationship between the informant and subject who provided reasonable details on their mental health that met criteria for an OCRD, or

103

104

3) If the informant reported having no or very little information on or regular indirect contact with the subject (not meeting reliability cut-off of good or excellent), then at least two members from the same family (including proband and first-degree relatives) reported on the same symptoms that met criteria for an OCRD separately.

The actual step-by-step algorithm for the modified OCD-FHI is listed in Appendix II.

3.3.3 Antidepressant Response Data Description

Two hundred and twenty-two of the 560 OCD patients had adequate antidepressant response data. Of the 222 sample, 217 Caucasian patients had drug response data for analysis. Six antidepressants including five selective serotonin reuptake inhibitors (SSRIs: fluoxetine, fluvoxamine, sertraline, paroxetine, citalopram) and clomipramine (serotonin reuptake inhibitor [SRI]), were examined for response or non-response.

The drug response data was collected by a questionnaire (Appendix III), which was developed by Drs. Kennedy and Richter, in addition to chart review retrospectively. Information including adequacy of trials (based on widely accepted criteria regarding optimal dose and duration), self-rated estimates of reliability of recollection, and response was obtained. Response to antidepressant medications was scored according to the Clinical Global Impression – Improvement Scale (CGI-I). Individuals deemed to be drug “responders” rated themselves as “much improved” or “very much improved” on at least one antidepressant; to be considered as “non-responders” had to give ratings of “minimal improvement”, “no change”, or “worse” in response to at least two different antidepressants.

The specific criteria for a complete and analyzable antidepressant treatment trial included the following:

1) Adequate duration of trial of at least 10 weeks

2) Adequate dosage of SRI (maximum dose achieved of: fluoxetine ≥20mg, fluvoxamine ≥150mg, paroxetine ≥20mg, sertraline ≥100mg, citalopram ≥20mg, or clomipramine ≥150mg) 104

105

3) Participant rated their recall of the drug trial as good or excellent

4) Individuals with intolerable side effects were excluded from the antidepressant response analysis given tolerability and response are different constructs and may have different genetic pathway

5) At least two failed trials of SRIs for coding as non-responder

3.3.4 Data Analysis

3.3.4.1 Gender Differences in Clinical Presentation of OCD

For each gender group, we investigated the relationship between gender and clinical characteristics including AAO, Y-BOCS severity, Y-BOCS symptom dimension factor, psychiatric comorbidities, family history of OCRDs, and SRI response using SPSS (version 20.0, Armonk, NY, USA). Levene’s test of homogeneity of variances was performed on continuous variables (age, AAO, Y-BOCS severity and symptom dimensions).

3.3.4.2 Identifying AAO subgroups using Admixture Analysis

We performed admixture analysis, a method for identifying the model that best fits the observed distribution of a continuous variable, AAO, as described in Delorme et al. (Delorme et al., 2005) using the DENORMIX (Kolenikov, 2001) module from the STATA program (version 11.0, Texas, USA). Admixture analysis has been used to reduce clinical heterogeneity. The DENORMIX analysis performs a decomposition of the AAO distribution into a mixture of normal components and estimates the number of components of the mixture. The maximum likelihood estimation of the finite normal mixture was applied to determine a theoretical model that best fits the observed distribution of AAO. The χ2 goodness-of-fit test was performed and the P value indicates the degree to which each model approximated the empirical distribution function of AAO in this sample. The model with the highest P value was selected to be the best- fitting model. The FMM procedure is then used to confirm the model, which fits finite mixture regression model to univariate outcomes using both maximum likelihood and Bayesian 105

106

techniques. Point of intersection between the normal distributions was calculated using the means and standard deviations of the identified components using the uniroot function in R program (version 3.2.1).

For each identified AAO group, we investigated the relationship between AAO and clinical characteristics including gender, age, Y-BOCS severity, Y-BOCS symptom dimension factor, psychiatric comorbidities, family history of OCRDs, and SRI response using SPSS (version 20.0, Armonk, NY, USA). Levene’s test of homogeneity of variances was performed on continuous variables (age, Y-BOCS severity and symptom dimensions).

3.3.4.3 Identifying Symptom Dimensions using Factor and Cluster Analyses on Y-BOCS Symptom Checklist

To evaluate whether this large OCD sample would generate a Y-BOCS severity score factor model similar to those previously reported, an exploratory factor analysis for data reduction was performed on seven obsession categories and eight compulsion categories with a total of 15 a priori categories according to the Y-BOCS Symptom Checklist (Appendix I) using SPSS (version 20.0, Armonk, NY, USA). Lifetime Y-BOCS symptoms (past or present/current) were used for this analysis as per Leckman et al. (Leckman et al., 1997). The miscellaneous obsessive and compulsive symptoms were excluded from the analysis since previous studies have shown that these symptoms are not clearly associated with previously described OCD symptom dimensions (Baer et al., 1994) and an exploratory factor analysis including these latter miscellaneous symptoms did not yield any meaningful results.

We utilized two different coding systems. First, symptoms present at the time of assessment, in the past, or both (i.e., lifetime symptoms) were coded as “1”, and symptoms that were never present were coded as “0”. Second, principal or target symptoms were coded as “2” to give a double weighting whereas symptoms present but not described as target or principal by subjects were coded as “1” and absence of symptoms were coded as “0”. Symptoms were summed for each category. Initial factors were extracted using the principal component analysis (PCA) method, and rotations based on whether the symptoms are related to one another were then performed by the PROMAX method using SPSS (version 20.0, Armonk, NY, USA) 106

107

because the oblique rotation provided by this method is most appropriate if the factors are expected to be correlated with each other (De Geus et al., 2004). Orthogonal rotation such as VARIMAX imposes restriction so that the factors cannot be correlated. Nonetheless, reanalysis using the VARIMAX method yielded similar results. Factor scores of the generated factors (symptom dimensions) for each subject were automatically calculated using SPSS (version 20.0, Armonk, NY, USA).

A cluster analysis using the Ward’s method in SPSS (version 20.0, Armonk, NY, USA) was also performed to compare with the factor analysis. Ward’s method is an agglomerative algorithm, applied in hierarchical cluster analysis for choosing the pair of clusters with the least incremental increase sum of squares to merge at each step. Symptom scores were generated based on the above described criteria (previous two paragraphs) for each of the 15 symptom categories in the Y-BOCS Symptom Checklist. Ward’s method for partitioning the data was used to minimize within-groups variance and to maximize between-groups variance (Abramowitz et al., 2003). Squared Euclidian distance was calculated as the similarity measure. Scores for each symptom category were standardized to correct for range differences within each category.

3.3.4.4 Determining the Presence or Absence of Comorbid Psychiatric Conditions

The frequency of current comorbid psychiatric conditions and lifetime/current OCRDs was determined by the SCID-IV DSM-IV diagnosis. Individuals with primary psychotic disorders and active substance use disorders that pre-dated the onset of OCD symptoms were excluded. Axis II diagnoses were not assessed.

3.3.4.5 Determining the Presence or Absence of Family History of OCRDs

To determine familiality, family history of OCD, OCD with hoarding, HD, TTM, and BDD was obtained for the proband and all identified first-degree relatives through interview of

107

108

the proband and any participating first-degree relatives were collected using the Family History Interview (FHI) questionnaire (Appendix II). Family history was obtained for OCD, OCD with hoarding, HD, TTM, and BDD. Reliability measures were also collected and data obtained were considered reliable if the informant was the subject who could recall his/her own symptoms or if the informant and subject: a) were reasonably close, b) had regular and frequent direct contact, and c) could provide details of subject’s symptoms. If the informant failed the reliability cut-off (no contact, not close to the subject, or indirect contact only with the subject), then two independent sources (including proband or any participating family members) would have to provide the same information or the participants indicated that the information provided had been confirmed. Familiality for each patient is coded either as a “0”, which indicates the absence of family history, or “1”, which refers to the presence of family history.

Familial form of OCD was first defined as OCD individual with a positive family history of OCD. The second familial analysis was defined as OCD individuals with a significant family history of OCRDs. We then examined the relationship between familial status and clinical characteristics including gender, Y-BOCS severity, Y-BOCS symptom dimension factor, family history of OCRDs, and current psychiatric comorbidities using SPSS (version 20.0, Armonk, NY, USA).

3.3.4.6 Determining Responder or Non-responder Status of Antidepressant Response

The response rate of the OCD sample was explored as a categorical dichotomous variable using frequency.

We investigated the relationship between SRI response status and clinical characteristics including gender, Y-BOCS severity, Y-BOCS symptom dimension factor, family history of OCRDs, and current psychiatric comorbidities using SPSS (version 20.0, Armonk, NY, USA).

108

109

3.4 Results

Subject demographics for the OCD sample have been reported in Table 3.1. In brief, the final number of OCD sample that had enough data for analysis is 519 with a mean age of 44 years, a mean AAO of 13.6 years, and a mean lifetime Y-BOCS severity score of 26.0. Almost 58% of the subjects were female and approximately 90% of the sample identified themselves as Caucasian.

Table 3.1. Subject Demographics.

Descriptors Mean (S.D.) N (final data set) 519 (probands) Age 44 (12.8) Gender 57.6% female Ethnicity 89.6% Caucasian 3.3% Asian 2.5% East Indian 0.7% African American 3.9% Others Age at Onset 13.6 (8.0) Yale-Brown Obsessive Compulsive Scale (Y-BOCS) severity score 26.0 (6.9) OCD subject demographics are shown in this table.

3.4.1 Does Gender Contribute to Clinical Diversity in OCD?

Subjects were divided according to their gender (Table 3.2). We observed significant greater number of comorbid lifetime history of SPD and TTM (P<0.001) in female OCD subjects than males (Table 3.2). Otherwise, there were no additional significant differences in clinical characteristics between males and females (Table 3.2) in terms of AAO, Y-BOCS severity and symptom dimensions as well as comorbid current psychiatric conditions, family history of OCRDs, and SRI response.

109

110

Table 3.2. Comparison of Clinical Characteristics between Males and Females.

Male Female Gender Group Mean±SD (N) Mean±SD (N) t P N 211 296 Age 42.7±13.1 (211) 44.9±12.6 (296) -1.967 0.063 AAO 13.2±7.3 (80) 13.6±8.5 (95) -0.266 0.791 Y-BOCS Severity 26.7±7.6 (89) 25.6±6.1 (112) 1.111 0.268 5-Factor 1 (Symmetry/Order) 0.22±1.0 (92) 0.00±1.1 (115) 1.553 0.122 5-Factor 2 0.13±1.0 (92) -0.03±1.0 (115) 1.128 0.261 (Aggression/Sexual/Religious) 5-Factor 3 (Hoarding) -0.06±1.0 (92) 0.12±1.0 (115) -1.276 0.203 5-Factor 4 (Contamination/Cleaning) 0.07±1.0 (92) -0.02±1.0 (115) 0.640 0.523 5-Factor 5 (Somatic) -0.06±1.1 (92) 0.04±1.0 (115) -0.728 0.467 6-Factor 1 (Symmetry/Order) 0.24±1.0 (92) 0.01±1.0 (115) 1.597 0.112 6-Factor 2 (Hoarding) -0.04±1.0 (92) 0.09±1.0 (115) -0.954 0.341 6-Factor 3 (Contamination/Cleaning) 0.05±1.0 (92) -0.01±1.0 (115) 0.366 0.715 6-Factor 4 (Aggression/Checking) 0.11±1.0 (92) -0.01±1.0 (115) 0.852 0.395 6-Factor 5 (Somatic) -0.04±1.1 (92) 0.01±1.0 (115) -0.329 0.743 6-Factor 6 (Sexual/Religious) 0.04±1.1 (92) 0.01±1.0 (115) 0.220 0.826 Disorders Percentage (N) Percentage (N) z/χ2 P Comorbid Lifetime OCRDs 15.5 (30) 42.6 (116) -0.308 0.757 - SPD 5.7 (11) 19.5 (53) -5.431 <0.001 - BDD 5.7 (11) 9.9 (27) -2.107 0.035 - TTM 3.6 (7) 12.5 (34) -3.967 <0.001 - HD 0.5 (1) 0.7 (2) -0.373 0.711 * Comorbid Current Psychiatric (194) (343) 0.699 0.403 Disorders (pure vs comorbid) - Anxiety disorders 55.4 (107) 60.7 (165) 0.450 0.646 - Mood disorders 38.5 (55) 33.1 (90) -0.142 0.889 - Tics/Tourette 8.3 (16) 4.8 (13) 1.911 0.056 - Others 8.3 (16) 15.8 (43) -1.898 0.057 - Pure OCD 9.3 (18) 11.8 (32) -0.836 0.401 Family History of OCRDs 19.7 (15/76) 21.6 (21/97) -0.308 0.757 - OCD 18.4 (13) 17.5 (13) 0.010 0.915 - HD 1.3 (1) 1.0 (1) 0.020 0.889 - TTM 0 (0) 1.0 (1) 0.670 0.414 - BDD 1.3 (1) 2.1 (2) 0.100 0.757 - SPD 0 (0) 0 (0) 0.000 1.000 SRI Response N N χ2 P Responders 33 57 0.423 0.515 Non-Responders 13 17 HD: hoarding disorder; TTM: trichotillomania; BDD: body dysmorphic disorder; SPD: skin picking disorder; SD: standard deviation Highlighted in yellow indicates significant P value * χ2 test was performed comparing the number of pure OCD and OCD with comorbid conditions Levene’s test of homogeneity of variances was not significant for age, Y-BOCS severity and symptom dimensions

110

111

3.4.2 Does AAO Present with Multiple Normal Distributions?

AAO in our sample ranged from 2 to 62 years with a mean of 13.6 and a standard deviation of 8.0. The overall sample showed a non-normal distribution of AAO in a histogram with kernel density and normal curve (Figure 3.1). Using the DENORMIX module in the STATA program (version 11.0, Texas, USA), the admixture analysis was performed and yielded a combination of one, two, three, or four normal theoretical distributions. The χ2 goodness-of-fit test for one-component was 81.824 (P=0.000), two-component was 8.488 (P=0.037), three- component was 6.116 (P=0.047), and four-component was 6.682 (P=0.035). The best-fit model with the highest P value (P=0.047) was the three-component model with means of 5.379, 11.583, and 22.291, standard deviations of 1.594, 3.426, and 8.726, and proportions of 0.196, 0.413, and 0.390, for the three normal distributions (Figure 3.2). The point of intersection between the first and second normal distributions is 8.265 and between the second and third normal distributions is 17.499. Therefore, subjects were divided into three separate groups according to AAO: early (≤8 years), intermediate (between 9 to 17 years), and late (≥18 years). Early onset group had significantly younger age, higher symmetry/order and contamination/cleaning symptom dimensional scores than intermediate or late onset group (Table 3.3). Otherwise, there were no significant differences in other clinical characteristics including Y-BOCS severity, psychiatric comorbidities, family history of OCRDs, and SRI response between the early, intermediate, and late AAO groups (Table 3.3).

3.4.3 Y-BOCS Symptom Dimension – Factor Model and Clusters

Table 3.4 and Table 3.5 provide the PROMAX rotated factor structure that created five distinct factors after treating all symptoms equally and six different factors after double- weighting target symptoms. Approximately 65% of the variance was explained by the 5-factor model whereas 70% of the variance for the 6-factor model. The cluster analyses confirmed both factor models that were generated from the PCA (Figure 3.2 and Figure 3.3). For the 5-factor model: factor 1 includes symmetry obsessions, ordering and arranging, counting, repeating, and checking compulsions; factor 2 comprises of sexual and religious obsessions, and aggressive obsessions and compulsions; factor 3 consists of hoarding and saving obsessions, hoarding and

111

112

collecting compulsions; factor 4 contains contamination obsessions and cleaning compulsions; and factor 5 includes somatic obsessions and compulsions. For the 6-factor model: factor 1 consists of symmetry obsession, ordering and arranging, repeating, and counting compulsions; factor 2 contains hoarding and saving obsessions, hoarding and collecting compulsions; factor 3 comprises of contamination obsessions and cleaning compulsions; factor 4 includes aggressive obsessions and compulsions, and checking compulsions; factor 5 includes somatic obsessions and compulsions; factor 6 consists of religious and sexual obsessions.

3.4.4 Rate of Psychiatric Comorbidities

Of 560 OCD participants, 520 had full SCID-IV data. Current comorbid psychiatric conditions were observed in 58.8% of the OCD participants and 23.7% had lifetime comorbid OCRDs (Figure 3.5 for current psychiatric diagnoses and Figure 3.6 for lifetime OCRDs). The top three current psychiatric comorbid conditions were social anxiety disorder, major depressive disorder, and hoarding disorder.

3.4.5 Rate of Significant Family History of OCRDs

We had 373 OCD subjects with one or more participating family members who provided FHI data. Approximately 86% of the OCD individuals had one informant, 10% had two, 4% had three, and 1% had four informants. Over 21% of the interview data was discarded due to unreliable recall. Within our OCD sample, 21.7% has significant family history of obsessive- compulsive and related disorders (OCRDs) including OCD, HD, TTM, and BDD (Figure 3.7). Positive family history was noted for the following disorders: OCD (15.5%; N=77), OCD with hoarding (1.4%; N=7), HD (2.0%; N=10), TTM (1.2%; N=6), and BDD (1.6%; N=8). Of the OCD subjects who reported positive family history of OCRDs, only one also had a diagnosis of TTM. Significant family history of OCRDs (except for OCD) did not appear to correlate with OCD subjects’ comorbidity status.

When comparing the clinical characteristics between familial and non-familial OCD individuals, first based on a positive family history of OCD only (Table 3.6), we detected trend 112

113

differences with more aggressive/sexual/religious symptoms, higher total comorbid psychiatric disorders and lower pure OCD cases in the familial OCD group. When combining the family history of other OCRDs (Table 3.7), we detected significantly lower number of pure OCD cases (participants who only had a diagnosis of OCD without any comorbid psychiatric conditions) in the familial group when compared to the non-familial group (P=0.035). There were also trend associations of higher comorbid anxiety disorders and tics in the familial group.

3.4.6 Can Clinical Characteristics Predict Antidepressant Response?

For the drug response data, of the 310 OCD subjects who completed the pharmacogenetic questionnaire, only 66% had an adequate trial of SRI (N=204) (also excluded those who could not tolerate the medication and therefore did not receive an adequate trial), which significantly reduce our sample. Of these 66% of OCD participants, over 92% had good or excellent recall (N=188).

After grouping all five SSRIs or all six SRIs (SSRIs and clomipramine) together, the response rate in our OCD sample is 70.9% and 74.7% respectively. When comparing clinical characteristics between responders and non-responders to SRI(s) (Table 3.8), we detected significantly more symmetry/order symptoms (P=0.008), higher rates of comorbid current psychiatric disorders (P=0.007) especially mood disorders (P=0.004), and higher rates of positive family history of OCD (P=0.042) and hoarding (P=0.002) in the SRI responder group. Responders had significantly higher rate of pure OCD (without any comorbid psychiatric disorders) (P=0.028).

113

114

Figure 3.1. Frequency Distribution of Age At Onset (AAO). This graph illustrates the frequency distribution of AAO in the OCD sample.

114

115

Age At Onset (AAO) Three Normal Distributions

0.5

0.4

0.3

0.2 Probability Density Probability

0.1

0 5.379058 11.58332 22.29146 62 Age at Onset (years)

Figure 3.1. Normal Distributions of Age At Onset (AAO). This graph illustrates three normal distributions within the AAO distribution in the OCD sample.

115

116

Table 3.3. Comparison of Clinical Characteristics between Age At Onset (AAO) Groups.

Early (≤8 years) Intermediate (between 9-17 years) Late (≥18 years) F P AAO Group Mean±SD (N) Mean±SD (N) Mean±SD (N) N 58 70 47 Gender 41.4% Male 50% Male 44.7% Male 0.291 0.864 Age 33.2±10.4 35.5±12.5 39.5±9.4 4.403 0.014 (4.371) (0.014) AAO 5.8±1.7 12.4±2.2 24.4±5.4 N/A N/A Y-BOCS Severity 26.4±7.1 (58) 26.5±6.9 (69) 25.3±6.6 (47) 0.500 0.608 5-Factor 1 (Symmetry/Order) 0.40±0.9 (58) 0.16±1.0 (70) -0.43±0.9 (47) 10.206 <0.001 (8.700) (<0.001) 5-Factor 2 0.25±1.0 (58) 0.16±1.0 (70) -0.14±0.8 (47) 2.305 0.103 (Aggression/Sexual/Religious) 5-Factor 3 (Hoarding) 0.09±1.0 (58) 0.14±1.0 (70) -0.11±1.0 (47) 0.928 0.397 5-Factor 4 (Contamination/Cleaning) 0.37±1.1 (58) -0.03±1.0 (70) -0.28±1.0 (47) 5.472 0.005 (5.724) (0.004) 5-Factor 5 (Somatic) 0.07±1.1 (58) 0.00±1.1 (70) -0.05±1.0 (47) 0.188 0.828 6-Factor 1 (Symmetry/Order) 0.37±1.0 (58) 0.16±1.1 (70) -0.29±0.9 (47) 6.004 0.003 (4.354) (0.014) 6-Factor 2 (Hoarding) 0.04±0.9 (58) 0.08±1.0 (70) -0.02±1.0 (47) 0.134 0.875 6-Factor 3 (Contamination/Cleaning) 0.38±1.1 (58) -0.09±1.0 (70) -0.21±1.0 (47) 5.219 0.006 (5.171) (0.007) 6-Factor 4 (Aggression/Checking) 0.31±1.0 (58) 0.01±0.9 (70) 0.05±0.9 (47) 1.615 0.202 6-Factor 5 (Somatic) -0.03±0.9 (58) 0.07±1.1 (70) -0.15±0.9 (47) 0.672 0.512 6-Factor 6 (Sexual/Religious) 0.04±1.1 (92) 0.19±1.0 (70) -0.20±0.9 (47) 2.110 0.124 2 Disorders Percentage (N) Percentage (N) Percentage (N) z/χ P Comorbid Lifetime OCRDs 28.6 (14/49) 22.6 (14/62) 15.6 (5/32) 2.422 0.298 - SPD 12.2 (6) 12.9 (8) 9.4 (3) - BDD 16.3 (8) 4.8 (3) 3.1 (1) - TTM 0 (0) 4.8 (3) 3.1 (1) - HD 0 (0) 0 (0) 0 (0) Comorbid Current Psychiatric Disorders 100 (61) 81.4 (57/70) 59.6 (28/47) 3.663 0.160 - Anxiety disorders 60.3 (35) 52.9 (37) 40.4 (19) 0.679 0.712

116

117

- Mood disorders 31.0 (18) 21.4 (15) 14.9 (7) 2.268 0.322 - Tics/Tourette 6.9 (4) 5.7 (4) 0 (0) - Others 6.9 (4) 1.4 (1) 4.3 (2) Family History of OCRDs 25.5 (13/51) 24.2 (15/62) 12.5 (4/32) 2.422 0.298 - OCD 21.6 (9) 17.7 (11) 9.4 (3) 2.046 0.360 - HD 2.0 (2) 4.8 (3) 3.1 (1) - TTM 1.9 (1) 0 (0) 0 (0) - BDD 1.9 (1) 1.6 (1) 0 (0) - SPD 0 (0) 0 (0) 0 (0) 2 SRI Response N N Percentage (N) χ P Responders 13 10 9 0.605 0.739 Non-Responders 5 6 3 HD: hoarding disorder; TTM: trichotillomania; BDD: body dysmorphic disorder; SPD: skin picking disorder Highlighted in yellow indicates significant P value and results remained significant after including gender and age as covariates in a linear regression model (in bracket) Levene’s test of homogeneity of variances was not significant for age, Y-BOCS severity and symptom dimensions

117

118

Table 3.4. Y-BOCS Symptom Dimension Factor Analysis (weighing all symptoms as equal).

118

119

Table 3.5. Y-BOCS Symptom Dimension Factor Analysis (double weighing target symptoms).

119

120

Y-BOCS Symptom Dimensions Cluster Analysis

Hoarding obsession Hoarding compulsion

Ordering/Arranging

Counting compulsion

Somatic compulsion

All symptoms treated equally as 1

Figure 3.2. Y-BOCS Symptom Dimension Cluster Analysis (weighing all symptoms as equal). This figure illustrates the grouped symptom dimensions using cluster analysis when weighing all symptoms equally.

120

121

Y-BOCS Symptom Dimensions Cluster Analysis

Hoarding obsession Hoarding compulsion

Ordering/Arranging

Counting compulsion

Somatic compulsion

Target Symptoms = 2 Identified Symptoms = 1 No Symptoms = 0 Figure 3.3. Y-BOCS Symptom Dimension Cluster Analysis (double weighing target symptoms). This figure illustrates the grouped symptom dimensions using cluster analysis when weighing target and other symptoms differently.

121

122

Figure 3.4. Current Psychiatric Comorbidities of OCD Sample. Percentage is indicated in brackets. This graph illustrates the count and percentage of current psychiatric comorbid conditions in the OCD sample.

122

123

Figure 3.5. Lifetime Comorbid OCRDs of OCD Sample. Percentage is indicated in brackets. This graph illustrates the count and percentage of lifetime obsessive-compulsive and related disorders in the OCD sample.

123

124

21.7% + Family History

1.4% 15.5% 2.0% 1.6% 1.2% OCD with OCD HD BDD TTM Hoarding

Figure 3.6. Family History of OCD Sample This figure illustrates the percentage of family history of obsessive-compulsive and related disorders in the OCD sample.

124

125

Table 3.6. Comparison of Clinical Characteristics between Familial and Non-Familial (OCD) OCD Individuals.

Familial Non-Familial Familial Group Mean±SD (N) Mean±SD (N) t P N 85 288 37.8% Male 39.9% Male -0.348 0.726 Gender (31/82) (115/288) Age 35.5±12.2 (80) 36.0±12.9 (274) 0.324 0.746 AAO 12.0±5.5 (20) 13.4±7.9 (112) 0.771 0.425 Y-BOCS Severity 27.3±4.6 (23) 26.9±6.7 (127) -0.221 0.826 5-Factor 1 (Symmetry/Order) -0.13±1.0 (23) 0.16±1.1 (130) 1.230 0.221 5-Factor 2 0.52±1.2 (23) 0.08±1.0 (130) -1.925 0.056 (Aggression/Sexual/Religious) 5-Factor 3 (Hoarding) -0.03±1.0 (23) 0.11±1.0 (130) 0.615 0.539 5-Factor 4 (Contamination/Cleaning) -0.02±1.1 (23) 0.13±1.1 (130) 0.584 0.560 5-Factor 5 (Somatic) 0.00±1.1 (23) 0.10±1.1 (130) 0.398 0.691 6-Factor 1 (Symmetry/Order) -0.01±1.1 (23) 0.14±1.0 (130) 0.683 0.496 6-Factor 2 (Hoarding) -0.13±0.9 (23) 0.09±1.0 (130) 1.017 0.311 6-Factor 3 (Contamination/Cleaning) -0.12±1.1 (23) 0.11±1.1 (130) 0.925 0.356 6-Factor 4 (Aggression/Checking) 0.23±1.1 (23) 0.14±0.9 (130) -0.432 0.666 6-Factor 5 (Somatic) 0.10±1.1 (23) 0.09±1.1 (130) -0.047 0.963 6-Factor 6 (Sexual/Religious) 0.55±1.2 (23) 0.01±1.0 (130) -2.271 0.025 Disorders Percentage (N) Percentage (N) z/χ2 P * Comorbid Current Psychiatric 62.4 (53/85) 52.1 (150/288) 2.791 0.095 Disorders (pure vs comorbid) - OCRDs 18.8 (16) 19.1 (55) -0.057 0.952 - Anxiety disorders 40.0 (34) 30.6 (88) 1.631 0.103 - Mood disorders 17.6 (15) 16.0 (46) 0.367 0.711 - Tics/Tourette 5.9 (5) 4.9 (14) 0.376 0.704 - Others 12.9 (11) 6.6 (19) 1.890 0.059 - Pure OCD 37.6 (32) 47.9 (138) -1.671 0.095 SRI Response N N χ2 P Responders 21 49 0.060 0.806 Non-Responders 6 16 HD: hoarding disorder; TTM: trichotillomania; BDD: body dysmorphic disorder; SPD: skin picking disorder Highlighted in yellow indicates significant P value Highlighted in grey indicates trend association * χ2 test was performed comparing the number of pure OCD and OCD with comorbid conditions

125

126

Table 3.7. Comparison of Clinical Characteristics between Familial and Non-Familial (OCRDs) OCD Individuals.

Familial Non-Familial Familial Group Mean±SD (N) Mean±SD (N) t P N 108 265 36.2% Male 40.8% Male -0.810 0.418 Gender (38/105) (108/265) Age 36.0±12.2 (102) 35.8±12.9 (252) -0.126 0.900 AAO 11.9±5.7 (26) 13.5±8.0 (106) 0.953 0.342 Y-BOCS Severity 28.3±5.8 (30) 26.7±6.6 (120) -1.249 0.214 5-Factor 1 (Symmetry/Order) 0.14±1.1 (30) 0.11±1.0 (123) -0.140 0.889 5-Factor 2 0.39±1.2 (30) 0.09±0.9 (123) -1.488 0.139 (Aggression/Sexual/Religious) 5-Factor 3 (Hoarding) 0.11±1.0 (30) 0.08±1.0 (123) -0.128 0.898 5-Factor 4 (Contamination/Cleaning) -0.03±1.1 (30) 0.14±1.1 (123) 0.737 0.462 5-Factor 5 (Somatic) 0.11±1.1 (30) 0.08±1.1 (123) -0.171 0.865 6-Factor 1 (Symmetry/Order) 0.18±1.1 (30) 0.11±1.0 (123) -0.326 0.745 6-Factor 2 (Hoarding) -0.01±0.9 (30) 0.08±1.0 (123) 0.423 0.673 6-Factor 3 (Contamination/Cleaning) -0.12±1.1 (30) 0.12±1.1 (123) 1.132 0.259 6-Factor 4 (Aggression/Checking) 0.15±1.1 (30) 0.15±1.0 (123) 0.041 0.968 6-Factor 5 (Somatic) 0.24±1.1 (30) 0.05±1.0 (123) -0.906 0.366 6-Factor 6 (Sexual/Religious) 0.39±1.1 (30) 0.02±1.0 (123) -1.751 0.082 2 Disorders Percentage (N) Percentage (N) z/χ P Comorbid Current Psychiatric 63.0 (68/108) 50.1 (135/265) 1.679 0.195 Disorders (pure vs comorbid) - OCRDs 22.2 (24) 17.7 (47) 1.001 0.317 - Anxiety disorders 39.8 (43) 29.8 (79) 1.868 0.061 - Mood disorders 15.7 (17) 16.6 (44) -0.204 0.841 - Tics/Tourette 8.3 (9) 3.8 (10) 1.817 0.069 - Others 11.1 (12) 6.8 (18) 1.391 0.165 - Pure OCD 37.0 (40) 49.1 (130) -2.114 0.035 2 SRI Response N N χ P Responders 26 44 0.206 0.650 Non-Responders 7 15 HD: hoarding disorder; TTM: trichotillomania; BDD: body dysmorphic disorder; SPD: skin picking disorder Highlighted in yellow indicates significant P value Highlighted in grey indicates trend association * χ2 test was performed comparing the number of pure OCD and OCD with comorbid conditions

126

127

Table 3.8. Comparison of Clinical Characteristics between SRI Responders and Non- Responders.

Responders Non-Responders SRI Response Mean±SD (N) Mean±SD (N) t P N 98 30 Gender 36.7% Male (33/90) 43.3% Male (13/30) -0.650 0.516 Age 38.1±11.6 (84) 38.4±11.5 (30) 0.090 0.928 AAO 13.3±8.2 (35) 13.6±7.9 (14) -0.180 0.857 Y-BOCS Severity 27.9±7.3 (91) 29.9±4.7 (30) -1.768 0.080 5-Factor 1 (Symmetry/Order) -0.16±1.1 (42) 0.43±1.0 (16) -2.760 0.008 5-Factor 2 (Aggression/Sexual/Religious) 0.22±0.9 (42) 0.19±1.1 (16) 0.136 0.892 5-Factor 3 (Hoarding) -0.07±1.0 (42) 0.25±1.1 (16) -1.423 0.165 5-Factor 4 (Contamination/Cleaning) 0.07±1.0 (42) 0.19±1.2 (16) -0.497 0.622 5-Factor 5 (Somatic) 0.16±1.1 (42) 0.40±1.4 (16) -0.861 0.398 6-Factor 1 (Symmetry/Order) -0.16±0.9 (42) 0.41±1.1 (16) -2.586 0.013 6-Factor 2 (Hoarding) -0.02±1.0 (42) 0.07±0.9 (16) -0.467 0.642 6-Factor 3 (Contamination/Cleaning) 0.05±1.0 (42) 0.10±1.1 (16) -0.222 0.826 6-Factor 4 (Aggression/Checking) 0.17±1.0 (42) 0.07±1.0 (16) 0.479 0.634 6-Factor 5 (Somatic) 0.11±1.0 (42) 0.60±1.6 (16) -1.585 0.123 6-Factor 6 (Sexual/Religious) 0.22±1.1 (42) 0.25±1.4 (16) -0.108 0.915 SRI Response Percentage (N) Percentage (N) z/χ2 P Pure OCD 45.9 (45/98) 23.3 (7/30) 2.204 0.028 * Comorbid Current Psychiatric Disorders (80) (35) 7.276 0.007 - OCRDs 16.3 (16) 16.7 (5) -0.044 0.968 - Anxiety disorders 35.7 (35) 46.7 (14) -1.080 0.280 - Mood disorders 15.3 (15) 40.0 (12) -2.901 0.004 - Tics/Tourette 6.1 (6) 3.3 (1) 0.588 0.555 - Others 8.2 (8) 10.0 (3) -0.314 0.757 Family History of OCRDs 40.0 (28/70) 45.5 (10/22) -0.453 0.653 - OCD 10.0 (7) 27.3 (6) -2.029 0.042 - HD 0 (0) 13.6 (3) -3.141 0.002 - TTM 0 (0) 4.5 (1) -1.079 0.073 - BDD 0 (0) 0 (0) 0.000 1.000 - SPD 0 (0) 0 (0) 0.000 1.000 HD: hoarding disorder; TTM: trichotillomania; BDD: body dysmorphic disorder; SPD: skin picking disorder Highlighted in yellow indicates significant P value Highlighted in grey indicates trend association * χ2 test was performed comparing the number of pure OCD and OCD with comorbid conditions

127

128

3.5 Discussion

To our knowledge, this is the single largest North American and most comprehensive clinical OCD study to date that explored demographical data and phenomenological characteristics of OCD including gender, AAO, Y-BOCS symptom dimensions and severity scores, psychiatric comorbidities, family history of OCRDs, and SRI antidepressant response. Our results were comparable to other large clinical OCD studies (Hofmeijer-Sevink et al., 2013; Torresan et al., 2013). The female OCD subjects in this study have a higher prevalence of psychiatric comorbidities especially OCRDs and the earlier AAO group appears to have higher symmetry/order and contamination/cleaning symptoms. Identified symptom dimensions from the 5- or 6-factor model were consistent with previous published studies (please refer to Chapter 1 Table 1.5a for details).

3.5.1 Gender Does Not Contribute to Difference in Clinical Presentations

Our current OCD sample did not detected any significant differences in AAO, symptom severity, symptom dimensions, and family history or OCRDs based on proband gender; however, we observed significantly higher rate of lifetime comorbid TTM and SPD, in females. Although we detected a significant difference of the rate of TTM between male and female individuals with OCD, the published gender ratio of TTM was approximately 10:1 (O’Sullivan et al., 1997; Stein et al., 2010), which was almost three times the rate of our current sample (3.5:1). The difference in the rate of SPD between males and females (3.4:1) was comparable to previous epidemiological studies with ¾ of cases were reported in female (Grant and Odlaug, 2009; Keuthen et al., 2001). However, OCRDs have recently been established as a diagnostic group and besides OCD, TTM and BDD, SPD and HD did not exist as a distinct disorder prior to the DSM-5. Changes made in DSM-5 may potentially affect future research study and there is a need to further explore the rates of OCRDs in the general population with the new diagnostic criteria.

Our findings for the remaining clinical characteristics appeared to contradict previous OCD studies except for one clinical study by Matsunaga et al. (2000), which detected no gender-

128

129

related phenomenological differences in OCD patients. Most studies reported notable gender differences. For example, OCD males had earlier AAO (de Mathis et al., 2011; Jaisoorya et al., 2009; Narayanaswamy et al., 2012; Nestadt et al., 1998; Rasmussen and Eisen, 1992), higher symptom severity (Nestadt et al., 1998; Rosso et al., 2012), more frequent aggressive (Tukel et al., 2013), sexual, religious, symmetry/exactness obsessions, checking, ordering/arranging, and hoarding compulsions (de Mathis et al., 2008; Jaisoorya et al., 2009; Rosso et al., 2012; Tukel et al., 2013). On the other hand, OCD females were observed to endorse higher frequency of contamination obsessions, washing/cleaning compulsions, and somatic obsessions and compulsions (de Mathis et al., 2008; Jaisoorya et al., 2009; Rosso et al., 2012; Tukel et al., 2013), and significantly more comorbid depression (Cherian et al., 2014). One reason for the general lack of differences between males and females in our current OCD sample is likely the sampling bias (which will be expanded in the limitation section below).

3.5.2 AAO has a Tri-modal Distributions

This current study demonstrated that the early-onset OCD group had younger age, higher frequency of symmetry/order and contamination/cleaning symptoms but did not differ in symptom severity, rate of comorbid psychiatric conditions, or rate of family history between the AAO groups. From a clinical perspective, individuals with early onset OCD often seek psychiatric help earlier given the longer duration of suffering when compared to later onset OCD patients and this clinical observation was supported by our finding. To date, only four clinical studies have examined AAO using admixture analysis in the OCD population. Delorme et al. (2005) identified two AAO groups, early-onset as defined by AAO less than or equal to 20 and late-onset as AAO greater than or equal to 21 years. The authors detected a higher rate of comorbid Tourette’s syndrome and greater percentage of positive family history of OCD in the early-onset group, and conversely a higher prevalence of generalized anxiety disorder and major depressive disorder in the late-onset group. The second study used mixture analysis in a subset of the current OCD sample (N=196), which was conducted by our group previously (De Luca et al., 2011). De Luca et al. (2011) detected two normal distributions with a cut-off at 15 years and found that the early-onset OCD patients were assessed at an earlier age and have more frequent panic attacks than the late-onset subjects. Similarly to the current larger sample, we also 129

130

observed an earlier age of assessment for the early-onset group when compared to the intermediate-onset group and for the intermediate-onset group when compared to the late-onset group. Although we did not detect any significant difference in psychiatric comorbidity, our results implicated a higher rate of anxiety disorders in the early-onset than the late-onset group. The third study was conducted by Anholt et al. (2014), who reported a bimodal distribution for AAO with the best-fitting cut-off AAO for early-onset as less than or equal to 19 years. The early-onset group was reported to have greater OCD symptom severity across all OCD symptom dimensions in addition to greater number of ADHD symptoms and higher rate of comorbid bipolar disorder. The last and most recent study utilized two sets of AAO as defined by age at which symptom first appeared and age at which symptom first met diagnostic criteria of OCD (Albert et al., 2015). Albert et al. (2015) also detected a bimodal distribution for AAO at which diagnosis was first met. However, they identified a trimodal distribution for AAO at symptom onset with means similar to our current study and a cut-off of 9 and 24 years (8 and 18 years in this study). The late-onset cut-off appeared to be comparable with these studies, ranging from 18 to 21 years except for Albert et al. (2015) who detected a cut-off at age 26 years. Nonetheless, we defined AAO as onset when OCD was first diagnosed and all three of the above mentioned studies identified a bimodal distribution for AAO using the same definition as our current study. The inconsistency between the present study and the others could be explained by sampling and recruitment bias. Our OCD sample had a higher mean of OCD Y-BOCS severity score (26 in this current study versus 23.6-25.4 for the other studies) and a younger AAO (13.6 years in this current study versus 17.1-22.1 years). These discrepancies suggested that different studies might have recruited different clinical OCD populations depending on the clinical settings and the city or country where the studies were conducted. Similar to the lack of gender differences, the general lack of differences between the AAO groups in the present study is also likely due to sampling and recruitment bias in a tertiary and highly specialized OCD clinical setting.

3.5.3 Factor and Cluster Analyses Confirmed a 5- and 6-Factor Model of Y-BOCS Symptom Dimensions

In this study, we demonstrated that OCD symptom dimensions could be reliably generated from the Y-BOCS symptom checklist using either the factor or cluster analysis. We 130

131

used two different coding methods, either weighing all OCD symptoms equally or double- weighing the target or principal symptoms, and identified five or six factors with the addition of somatic obsessions and compulsions to the consistently reported symptom factors or dimensions – symmetry/ordering, hoarding, contamination/cleaning, aggressive/checking+/-sexual/religious concerns from previous studies (Table 1.6, Table 1.7, Table 1.8). We detected an extra symptom dimension in both our 5- and 6-factor models, somatic subgroup, which is in concert with two previous studies (Stein et al., 2007; Stein et al., 2008). Somatic symptoms have been frequently grouped with aggressive/sexual/religious/checking obsessions and compulsions (Denys et al., 2004; Hasler et al., 2007; Lochner et al., 2008; Pinto et al., 2007). Liu et al. (2011), who weighed all symptoms equally, reported somatic and repeating as one symptom dimension, with hoarding, contamination/cleaning, symmetry/ordering, and aggressive/checking as the remaining four factors. Zhang et al. (2013) detected somatic symptoms as a separate dimension out of six different identified factors only when examining Y-BOCS symptom checklist at an item level rather than the commonly used categorical level. Most studies did not report somatic symptoms as the one of the prominent symptom dimensions within the 4 or 5 factors (Table 1.6 and Table 1.7).

Although the current study derived a 5-factor model that was consistent with most published studies, we did not replicate the identified symptom dimensions and the generated factors appeared to be consistent with only the minority of the studies. Furthermore, from a clinical perspective, somatic symptoms are very closely related to hypochondriasis. Recurrent health-related concerns with frequent visits to the family doctor’s office might not be perceived as a symptom of OCD but rather a somatic symptom within the DSM-IV somatization disorder category; thus, some of the previous studies might have under-detected this symptom domain. Another possible reason for a more common presentation of somatic symptom is the location of the recruitment sites. Toronto, the recruiting site for this study, is considered to be one of the most multi-cultural cities in the world with a large percentage, 49.1%, of Toronto’s population as Asians according to the 2011 National Household Survey (Statistics Canada, 2011). Many Asian individuals who suffer from anxiety or depression very often present with somatic complaints (Hinton et al., 2009; Lin and Cheung, 1999; Mumford, 1992). Due to stigma and misconception of mental illness, physical discomforts are more widely acceptable than emotional or behavioural issues in the Asian culture (Lin and Cheung, 1999). Our OCD sample consists of 3.3% of Asian 131

132

individuals, which might potentially affect the results of the factor analysis. Almost all of the previous studies included Caucasian OCD participants although many did not specify the ethnicity of their samples. Only one study has investigated the African American population and detected a 6-factor model that accounted for 59.1% of the variance: contamination and washing, hoarding, sexual obsessions and reassurance, aggression and mental compulsions, symmetry and perfectionism, and doubt and checking (Williams et al., 2012). Although the factors were similar to the previous studies, the authors reported significantly more contamination symptoms and excessive concerns with animals in African Americans than Europeans. These findings further implicate cultural differences in OCD symptom presentation.

3.5.4 High Rates of Psychiatric Comorbidities

Over 50% of our sample’s OCD participants have concurrent psychiatric disorders and over 20% have a lifetime history of OCRD(s). These numbers are consistent with previous published studies. Comorbidity rates in epidemiological studies vary widely, ranging from 19% to 90% depending on the methodological and semantic (i.e., lifetime versus current rates) differences (Overbeek et al., 2002; Pallanti et al., 2011). However, one key concept is that OCD individuals often present with comorbid psychiatric conditions, which can often influence treatment plan and outcome. Patients who are non-responders to pharmacological and psychological treatments are more likely to meet criteria for comorbid Axis I or Axis II disorders (Pallanti et al., 2011).

3.5.5 High Rate of Family History of OCRDs

Given that OCRDs formed a new and separate section in the recently published DSM-5 (APA, 2013), many previous studies have not investigated the presence or absence family history of OCRDs in OCD samples. Approximately 22% of our sample’s OCD probands reported having one or more first-degree relative who suffers from at least one OCRD. This is not surprising because of the similarities these disorders share. Many studies have reported a prevalence rate ranging from 7% to 37% of first-degree relatives of OCD individuals who also

132

133

have a diagnosis of OCD (Nestadt et al., 2010) and the present study detected 15.5% of OCD individuals having a positive family history of OCD. However, none has ever investigated the prevalence of OCRDs in first-degree relatives of OCD patients. This is likely because most research studies still utilize the DSM-IV diagnostic criteria, which do not account for SPD and HD and with slight differences in the diagnostic criteria of BDD and TTM. Hence, this is the first study, which reported 6.2% of OCD individuals with a positive family history of OCRDs other than OCD. Interestingly, after further exploration of the probands with positive family history of OCRDs, we only detected one individual with comorbid TTM who had a family history of TTM. The rest of the probands did not have congruent personal and family history of the same disorders (SPD, BDD, and HD). This is intriguing given that this study is the first to show a general liability of OCRDs in OCD patients without a comorbid diagnosis of other OCRDs.

Although the number of the comparison groups was small, we observed clinical differences between familial and non-familial form of OCD. The results for incorporating OCD only and other OCRDs together did not differ dramatically. Higher comorbidity was detected in the familial group in general when compared with the non-familial group, suggesting that family history may allow researchers to identify more homogeneous subgroups of OCD.

3.5.6 Antidepressant Response Rate is Higher than Published Rates and Differences were Observed between SRI Responders and Non-Responders

Although our OCD subjects were recruited from a tertiary and highly specialized OCD clinic, the SRI response rate is relatively higher, at over 70%, than other published rates of 40- 60% (Foa et al., 2005; Greist et al., 1995; McDougle et al., 1993). This could be explained by several factors: the definition of drug response and study design. Firstly, these three cited studies above (Foa et al., 2005; Greist et al., 1995; McDougle et al., 1993) have defined response as OCD symptom reduction of at least 25% to 35% according to the Y-BOCS severity score (Tolin et al., 2005), whereas our current study has employed the CGI-I scale to report response rate. Documented Y-BOCS severity score may provide a more objective SRI response that is specific

133

134

for OCD whereas the CGI-I scale offers a readily understood and practical measurement tool that can easily be administrated by any physicians but it lacks specificity to OCD and objectivity of a change in SRI response over time. Secondly, Y-BOCS severity scores are often documented prospectively by a physician who monitors the progress of an SRI treatment trial; however, in our current study, most of the response data have been generated retrospectively or by chart review; thus, recall bias could explain the small difference in SRI response. We attempted to address the limitation of recall bias by introducing a reliability measure. Thirdly, one additional possible reason for a higher response rate in this sample could be the increasing awareness of OCD in general, leading to referral of less severe cases to the tertiary clinic, which may be more representative of the general OCD population.

This study revealed clear clinical prognostic predictors of SRI response. Poor SRI response was associated with symmetry/order symptoms, higher number of comorbid current psychiatric disorders especially mood disorders, and family history of OCD and HD. Previous treatment studies have examined clinical factors that predict antidepressant response in OCD. Denys et al. (2003) reported the absence of previous treatment, moderate symptom severity (Y- BOCS severity score < 23), and low depressive symptoms (Hamilton Depression Rating Scale score between 6 and 15) as good prognostic factors of antidepressant response. Although we did not examine previous treatment history, we observed a trend for lower severity and a significantly lower number of comorbid current mood disorders in responders when compared with non-responders. Another study examined additional clinical predictors of SSRI response and detected higher level of baseline symptom severity and disability in non-responders (Tükel et al., 2006). Marazziti and Consoli (2010) provided a comprehensive review on treatment strategies for OCD and reported several robust positive clinical predictors of SSRI response, which included later onset of illness, less symptom severity, absence of previous treatment, and moderate severity of depression; however, the authors concluded that the strongest symptom predictor of poor treatment response was hoarding. Although inconsistent findings have been reported for comorbid tics as influencing SRI response (Husted et al., 2007; Skarphedinsson et al., 2015), the presence of tics (Bloch et al., 2006; Masi et al., 2013; McDougle et al., 1994; Nakamae, 2011) and family history of tics (McDougle, 1997) both predict a better response to antipsychotic augmentation strategy. Thus, in addition to examining SRI response, determining clinical predictors of antipsychotic response and response to other treatment modalities such as 134

135

psychotherapy and transmagnetic stimulation will allow us to move beyond standard treatment protocol to individualized medicine. Further investigation on other clinical and demographic variables including duration of illness and number of previous treatment is warranted to provide a comprehensive treatment guide that is tailored to each individual patient.

3.5.7 Limitations of the Clinical Study

One limitation of our study is regarding data collection. There are missing data within the current sample, which is often unavoidable in large research dataset. Many reasons could contribute to missing data and they are generally categorized into two main areas, participant factors and researcher factors. Participants may have lost interest given the lengthy assessment and complicated interviewing questions. The duration of the interview necessitated more than one visit for some subjects, and some probands did not return to complete the research assessment. Missing observations might occur due to human error or the lack of man-power to transfer written documents into electronic databases. Additionally, missing data could mislead our current findings and further reduce the power of our sample. Another limitation related to data collection is that the clinical data from this study including AAO, lifetime Y-BOCS severity, presence or absence of Y-BOCS OCD symptoms, family history, and comorbid psychiatric conditions, in addition to SRI antidepressant treatment response data were ascertained in a retrospective manner and chart review and there is a heightened chance of recall bias and unreliable assessment. We attempted to minimize the effect of recall bias by incorporating a reliability measure. Although using the reliability measure drastically reduces the number of positive family history and complete drug response data in our sample, this is a more stringent way to reduce recall bias but with the risk of increased missing data. For the family history interview, 21.5% of the interview data was discarded due to unreliable recall. For the drug response data, 33% had inadequate trial of SRI and an additional 8% of these individuals reported poor recall, which significantly reduce our sample. Prospective study will ultimately eliminate the risk of prematurely discontinuing SRI at a suboptimal dose (adequacy of trial) and recall bias by utilizing objective measures to quantify treatment response and to follow participants in addition to their relatives longitudinally. Therefore, prospective clinical and pharmacogenetic studies are warranted in order to clarify causality in defining clinical outcome 135

136

and treatment response. The third limitation of this study is the tertiary-centre sample, which may not fully represent the general OCD population but is rather a relatively more severe OCD sample with higher comorbidity; however, the rate of treatment responders is higher than the published rate, which may suggest either the contrary or methodological differences between the published studies. Nonetheless, our results suggested that this OCD sample is comparative to other large published OCD clinical studies with similar rates of major comorbidity.

3.5.8 Conclusion

The results of this study demonstrated that there may be phenotypically different and homogeneous subgroups of OCD, which may be scientifically relevant to be examined in genetic studies of OCD. Gender differences, early versus late AAO, symptom severity, symptom dimensional subtypes, positive versus negative family history, different psychiatric comorbidities, and antidepressant responders versus non-responders are clinically informative factors that may be associated with specific vulnerability pattern, illness chronicity, and illness severity. By combining associated genetic variations from ongoing OCD genetic studies with clinical and demographic risk factors, we may be able to develop a risk model for OCD. Once this model has been validated, it will be a useful tool for clinical application. A better understanding of clinically homogeneous OCD subgroups or subphenotypes has the potential to substantially improve treatment outcomes and to promote the development of new therapies. Further research should focus on the identification of homogeneous subgroups for genetic research and for treatment outcome study.

136

137

Chapter 4 CANDIDATE GENE STUDY OF OCD PHENOTYPES

4 Candidate Gene Study of Obsessive-Compulsive Disorder Phenotypes

Gwyneth Zai 1,2,4, Clement Zai 1,2, Vanessa Gonçalves 1,2, David Pauls 3, Karen Wigg 4, Margaret A. Richter 4, James L. Kennedy 1,2