Genetic Contribution to Heterogeneity in Brain Morphology

with Applications to Schizophrenia

by

Tristram Alexander Pierrepont Lett

A thesis submitted in conformity with the requirements

for the degree of Doctor of Philosophy

Institute of Medical Science

University of Toronto

© Copyright by Tristram Lett 2014

ii

Genetic Contribution to Heterogeneity in Brain Connectivity and

Plasticity in Schizophrenia

Tristram Alexander Pierrepont Lett

Doctor of Philosophy

Institute of Medical Science

University of Toronto

2014

Abstract

Schizophrenia is a highly heterogeneous disorder. Differences among patients with schizophrenia

have been reported in clinical features, cognitive functioning, and brain structure. This thesis

investigates the impact of the well supported genetic risk variants on brain structure and also

considers the role of imaging-genetic changes on heterogeneous phenotypes relevant to

schizophrenia. In the first study, the genome-wide supported NRXN1 gene, associated with

schizophrenia and autism spectrum disorders, was examined with magnetic resonance imaging

(MRI) volumetric measures and measures of sensorimotor function. Study two investigated the

effect of the genome-wide identified schizophrenia risk variant in the MIR137 gene for

association with age-at-onset and brain structures implicated in disease severity, as well as white

matter fractional anisotropy (FA) and cortical thickness. Risk allele carriers had earlier age-at-

onset and aberrant brain structure suggesting that MIR137 may predict a more severe

schizophrenia subphenotype. Study three examined the role of the GAD1 gene on brain structure

and executive function. Among patients and controls the GAD1 variant predicted changes in

white matter FA and multiple working memory tasks. Using voxel-wise mediation analysis, we

were able to infer the functional relevance of GAD1 on structural connectivity by showing these

iii

FA changes statistically cause impaired working memory performance. In the final study, we

investigated relationships among polygenic additive risk, brain-wide measures, and cognition in

schizophrenia patients and healthy controls. Schizophrenia patients with low polygenic risk were

similar in brain structure and cognitive performance to healthy controls. In contrast, high

polygenic risk in patients led to large reductions in cortical thickness and white matter skeleton

FA, in addition to impaired semantic memory and motor functioning. Taken together, these

studies suggest that neuroimaging and genetics can be used to meaningful disentangle

heterogeneity of schizophrenia phenotypes, and may move towards neurobiological based

treatment options.

iv

§

I would like to dedicate this thesis in loving memory of Margaret Depew

§

v

Acknowledgments

It is my pleasure to thank the many people for who made this thesis possible.

I would like to thank my PhD supervisors Drs. James Kennedy and Aristotle Voineskos. Dr.

Kennedy has been instrumental for my scientific development. He has taught me to navigate

through the complexity of scientific field, and I never wanted for any intellectual need under his

tutelage. Dr. Voineskos has been a true mentor as well as supervisor, and he was the inspiration

for my pursuit as a scientist. He provided the means and stellar direction necessary to maximize

my potential. As a result of their outstanding supervision, I believe I am well-prepared for a

career as an independent investigator. I have been extremely fortunate work and learn from them

these past few years.

I also want to thank my PhD committee members Drs. Jeff Daskalakis and Gary Remington. I

have benefitted greatly from their expertise and scientific input at every stage of my doctoral

training. Dr. Remington has provided invaluable clinical perspective to my research. I am

extremely grateful to Dr. Daskalakis who has acted in every way as supervisor and outstanding

mentor to me. He has providing novel understanding of translational and brain stimulation

research. It has been an honour to work in his lab.

I am grateful to all past and present members of the Psychiatric Neurogenetics Laboratory (Dr.

Kennedy), the Kimel Family Translational Imaging-genetics Laboratory (Dr. Voineskos), and

Temerty Centre for Therapeutic Brain Intervention (Dr. Daskalakis). Drs. Mallar Chakravarty

and Arun Tiwari deserve special mention.

vi

I am also thankful for the support provided by the Institute of Medical Science, and the Centre

for Addiction and Mental Health.

Last, I thank my wife Eva Brandl, my parents, my grandmother, and my siblings for their love

and support. Also, I appreciate the patience of my daughter, Hanna, who has spent many hours

bouncing on my knee while I wrote this thesis.

vii

Table of Contents

Acknowledgments ........................................................................................................................... v

Table of Contents .......................................................................................................................... vii

List of Figures .............................................................................................................................. xiv

List of Appendices ....................................................................................................................... xvi

List of Abbreviations .................................................................................................................. xvii

Chapter 1 ......................................................................................................................................... 1

1 Introduction ................................................................................................................................ 1

1.1 Overview ............................................................................................................................. 1

1.2 Schizophrenia ...................................................................................................................... 2

1.2.1 Impact, Epidemiology and Prognosis ..................................................................... 2

1.2.2 Treatment of Schizophrenia .................................................................................... 3

1.3 Sources of Schizophrenia Heterogeneity ............................................................................ 5

1.3.1 Onset of Psychotic Symptoms ................................................................................ 6

1.3.2 Cognitive Symptoms ............................................................................................... 8

1.3.3 Brain Structure ...................................................................................................... 10

1.4 Cortical Thickness and Diffusion Tensor Imaging Measures of White Matter

Structure ............................................................................................................................ 12

1.4.1 Analytic Approaches to Cortical Thickness and Diffusion Tensor Imaging ........ 12

1.4.2 Genetic Contribution to Cortical Thickness and White Matter Fractional

Anisotropy (FA) .................................................................................................... 15

1.5 Functional Integration in Schizophrenia ........................................................................... 16

1.6 Schizophrenia Genetics: An Update ................................................................................. 17

1.7 Important Genetic Modifiers of Schizophrenia Phenotypes ............................................. 23

1.7.1 Neurexin-1 (NRXN1) ............................................................................................ 23

1.7.2 Glutamate Decarboxylase 1 (GAD1) .................................................................... 23

viii

1.7.3 Brain-derived Neurotrophic Factor (BDNF) ........................................................ 24

1.7.4 MicroRNA 137 (MIR137) ..................................................................................... 25

1.7.5 L-type Voltage-dependent Calcium Channel CAv1.2 (CACNA1C) ..................... 26

1.7.6 Zinc-Finger 804A (ZNF804A) .............................................................................. 27

1.8 Multivariate Approaches to Neuroimaging ...................................................................... 29

1.8.1 Complex Genetic Analysis on Candidate Imaging Phenotypes ........................... 29

1.8.2 Single Variant Analysis of Whole Brain Imaging Phenotypes ............................. 32

1.8.3 Complex Genetic Analysis of Whole Brain Imaging Phenotypes ........................ 34

1.9 Application of polygenic risk models to imaging genetic studies in psychiatry .............. 38

1.10 Outline of Experiments ..................................................................................................... 40

Chapter 2 ....................................................................................................................................... 41

2 Overview of Experiments, and Hypothesis .............................................................................. 41

2.1 Neurexin-1 and Frontal Lobe White Matter: An Overlapping Intermediate Phenotype

for Schizophrenia and Autism Spectrum Disorders .......................................................... 41

2.1.1 Background ........................................................................................................... 41

2.1.2 Hypothesis ............................................................................................................. 41

2.2 The Genome-Wide Supported MicroRNA-137 Variant Predicts Phenotypic

Heterogeneity within Schizophrenia ................................................................................. 42

2.2.1 Background ........................................................................................................... 42

2.2.2 Hypothesis ............................................................................................................. 42

2.3 Glutamate Decarboxylase 1 (GAD1) Variant Predicts a Neuroanatomical and

Working Memory Susceptibly Mechanism Relevant to Schizophrenia. .......................... 43

2.4 Background ....................................................................................................................... 43

2.5 Hypothesis ......................................................................................................................... 43

2.6 Additive Genetics Risk Predicts Widespread Changes in Brain Structure that Cause

Poorer Cognitive Function ................................................................................................ 44

2.6.1 Background ........................................................................................................... 44

2.6.2 Hypothesis ............................................................................................................. 44

ix

Chapter 3 ....................................................................................................................................... 45

3 Neurexin-1 and Frontal Lobe White Matter: An Overlapping Intermediate Phenotype for

Schizophrenia and Autism Spectrum Disorders ...................................................................... 45

3.1 Abstract ............................................................................................................................. 46

3.2 Introduction ....................................................................................................................... 47

3.3 Results ............................................................................................................................... 49

3.3.1 Genotypes ............................................................................................................. 49

3.3.2 Cognitive ............................................................................................................... 52

3.3.3 In silico Analysis ................................................................................................... 52

3.4 Discussion ......................................................................................................................... 52

3.5 Materials and Methods ...................................................................................................... 56

3.5.1 Participants ............................................................................................................ 56

3.5.2 Neuroimaging ....................................................................................................... 57

3.5.3 Image Processing .................................................................................................. 57

3.5.4 Genetics ................................................................................................................. 58

3.5.5 Cognitive Assessment ........................................................................................... 59

3.5.6 Statistical Analysis ................................................................................................ 59

3.5.7 In Silico Analysis .................................................................................................. 60

3.6 Acknowledgements ........................................................................................................... 60

Chapter 4 ....................................................................................................................................... 74

4 The Genome-Wide Supported MicroRNA-137 Variant Predicts Phenotypic Heterogeneity

Within Schizophrenia ............................................................................................................... 74

4.1 Abstract ............................................................................................................................. 75

4.2 Introduction ....................................................................................................................... 76

4.3 Subjects and Methods ....................................................................................................... 78

4.3.1 Participants for Genetic Investigation of Age-at-onset Phenotypes ..................... 78

4.3.2 Participants for Genetic Investigation of Neuroimaging Phenotypes ................... 79

x

4.3.3 Image Acquisition ................................................................................................. 80

4.3.4 Cortical Volumes Processing ................................................................................ 80

4.3.5 Cortical Thickness Mapping ................................................................................. 80

4.3.6 Tract-Based Spatial Statistics (TBSS) .................................................................. 81

4.3.7 Whole-Brain TBSS Analysis ................................................................................ 82

4.3.8 Genetics ................................................................................................................. 82

4.3.9 Statistical Analysis ................................................................................................ 82

4.3.10 Mediation Analysis ............................................................................................... 84

4.4 Results ............................................................................................................................... 85

4.4.1 Genetics ................................................................................................................. 85

4.4.2 Age-at-onset .......................................................................................................... 85

4.4.3 Neuroimaging ....................................................................................................... 86

4.4.4 Mediation Analysis ............................................................................................... 88

4.5 Discussion ......................................................................................................................... 88

4.6 Acknowledgements ........................................................................................................... 92

4.7 Conflict of interest ............................................................................................................ 93

Chapter 5 ..................................................................................................................................... 106

5 GAD1 variant predicts a neuroanatomical and working memory susceptibly mechanism

relevant to schizophrenia ........................................................................................................ 106

5.1 Abstract ........................................................................................................................... 107

5.2 Introduction ..................................................................................................................... 108

5.3 Methods ........................................................................................................................... 109

5.3.1 Participants .......................................................................................................... 109

5.3.2 Genetics ............................................................................................................... 110

5.3.3 Image Acquisition ............................................................................................... 111

5.3.4 Cortical Thickness Mapping ............................................................................... 111

xi

5.3.5 Tract-Based Spatial Statistics (TBSS) ................................................................ 112

5.3.6 Assessment of Working Memory ....................................................................... 112

5.3.7 Statistics .............................................................................................................. 113

5.3.8 Voxel-wide mediation analysis ........................................................................... 114

5.4 Results ............................................................................................................................. 115

5.4.1 Participants .......................................................................................................... 115

5.4.2 Association between GAD1 and brain structure ................................................. 116

5.4.3 Association between GAD1 and working memory ............................................ 116

5.4.4 Voxel-wise mediation analysis ........................................................................... 117

5.5 Discussion ....................................................................................................................... 117

Chapter 6 ..................................................................................................................................... 126

6 Additive Genetics Risk Predicts Widespread Changes in Brain Structure Leading to

Poorer Cognitive Function ..................................................................................................... 126

6.1 Abstract ........................................................................................................................... 127

6.2 Introduction ..................................................................................................................... 128

6.3 Methods ........................................................................................................................... 132

6.3.1 Participants .......................................................................................................... 132

6.3.2 Image Acquisition ............................................................................................... 133

6.3.3 Cortical Thickness Mapping ............................................................................... 133

6.3.4 Tract-Based Spatial Statistics (TBSS) ................................................................ 134

6.3.5 Neuroimaging Dimension Reduction ................................................................. 134

6.3.6 Genetics ............................................................................................................... 135

6.3.7 Additive Model ................................................................................................... 135

6.3.8 Neuropsychological Assessment ........................................................................ 136

6.3.9 Statistical Analysis .............................................................................................. 136

6.3.10 Voxel-wise mediation analysis ........................................................................... 137

xii

6.4 Results ............................................................................................................................. 138

6.4.1 Demographics ..................................................................................................... 138

6.4.2 Genetics ............................................................................................................... 138

6.4.3 The effect of additive risk on whole brain measure of cortical thickness and

white matter FA .................................................................................................. 139

6.4.4 The effect of additive risk on general brain structure ......................................... 139

6.4.5 The effect of extreme additive risk loading on general brain structure and

cognitive performance ........................................................................................ 140

6.4.6 Voxel-wise mediation analysis ........................................................................... 140

6.5 Discussion ....................................................................................................................... 141

Chapter 7 ..................................................................................................................................... 159

7 General Discussion & Future Direction ................................................................................. 159

7.1 Summary of Results ........................................................................................................ 159

7.2 Can Imaging-genetics Dissect Clinically Meaningful Heterogeneity within

Schizophrenia? ................................................................................................................ 160

7.3 What Benefits does Translational Research Address? .................................................... 163

7.4 Can Imaging-genetics explain enough of the Heterogeneity to Guide Treatment

Decisions? ....................................................................................................................... 165

7.5 Limitations ...................................................................................................................... 166

7.6 Future Directions ............................................................................................................ 169

7.6.1 Functional Relevance of Genetic Variation ........................................................ 169

7.6.2 In Silico Prediction of SNP Function: Insight from ENCODE ........................... 170

7.6.3 Combining in vivo Biomarkers ........................................................................... 172

7.6.4 Towards Neurobiological Treatment .................................................................. 173

7.6.5 Conclusion .......................................................................................................... 174

References ................................................................................................................................... 175

Appendices .................................................................................................................................. 209

xiii

List of Tables

Table 3-1. Demographic Characteristics ...................................................................................... 61

Table 3-S1. Locations and Minor Allele Frequency in Toronto and Hapmap (CEU) Samples ... 67

Table 3-S2. T-test between rs1045881 T-Carriers Vs C/C and Demographics ........................... 68

Table 3-S3. Chi-squared Tests of Region by Genotype or Allele Interactions of rs1045881 and

rs858932 ....................................................................................................................................... 69

Table 3-S4. Haplotype Association between Frontal Lobe White Matter and rs1045881 (T/C) and

rs858932 (G/C) ............................................................................................................................ 70

Table 3-S5. Reported deletions within NRXN1 in Developmental Disorders, Schizophrenia and

Autism Spectrum Disorders .......................................................................................................... 71

Table 4-S1. Demographics for age at onset samples .................................................................... 98

Table 4-S2. Demographics and clinical characteristics for the Toronto imaging-genetics sample

....................................................................................................................................................... 99

Table 5-1. Demographics ........................................................................................................... 121

Table 5-2. The association between working memory related tasks and GAD1 genotype,

diagnosis, and their interaction ................................................................................................... 122

Table 6-1. Count and frequency of risk alleles by diagnosis ...................................................... 156

Table 6-2. Demographics and clinical characteristics ............................................................... 157

Table 6-3. The effect of high additive risk allele loading on general fluid intelligence (gF) and its

components ................................................................................................................................. 158

xiv

List of Figures

Figure 3-1. Reported Deletions in the Neurexin-1α gene ............................................................. 62

Figure 3-2. The effect of rs1045881 on combined hemispheric volume of brain regions with total

brain volume (TBV) and age as covariates ................................................................................... 63

Figure 3-3. The effect of rs858932 on right and left thalamic volume with TBV and age as

covariates ...................................................................................................................................... 64

Figure 3-S1. The effect of rs858932 on combined hemispheric volume of brain regions with

TBV and age as covariates ............................................................................................................ 66

Figure 4-1. MIR137 rs1625579 risk variant homozygotes have earlier age-at-onset of

schizophrenia ................................................................................................................................ 94

Figure 4-2. MIR137 risk variant predicts poorer structural brain phenotypes in schizophrenia .. 95

Figure 4-3. Effect of MIR137 rs1625579 genotype on voxel-based white matter integrity in

patients with schizophrenia .......................................................................................................... 96

Figure 4-4. The effect of MIR137 risk variant on mean whole-brain fractional anisotropy (FA)

across the lifespan for four ‘diagnosis-genotype’ groups ............................................................ 97

Figure 4-S1. MIR137 rs1625579 risk variant homozygotes have earlier age-at-onset of

schizophrenia ............................................................................................................................. 101

Figure 4-S2. The main effect of MIR137 rs1625579 genotype on voxel-based white matter

integrity in healthy controls and patients with schizophrenia ................................................... 102

Figure 4-S3. Effect of MIR137 rs1625579 genotype by diagnosis interaction on voxel-based

white matter integrity in healthy controls and patients with schizophrenia .............................. 103

Figure 4-S4. Mediation Model .................................................................................................... 104

xv

Figure 4-S5. Mediation model examining the associations between MIR137 rs1625579 genotype,

age at onset and mean whole brain fractional anisotropy (FA) .................................................. 105

Figure 5-1. GAD1 rs3749034 risk A-allele predicts lower TBSS skeleton white matter FA in

healthy controls (N=115) and patients with schizophrenia (N=80) ............................................ 123

Figure 5-2. Higher TBSS skeleton white matter FA correlates with better digit span performance

..................................................................................................................................................... 124

Figure 5-3. Skeletal white matter FA regions that mediate the effect of GAD1 rs3749034 on digit

span performance ........................................................................................................................ 125

Figure 6-1. Greater additive risk predicts poorer white matter fractional anisotropy ............... 146

Figure 6-2. Significant additive score-by-diagnosis interaction for vertex-wide cortical thickness

..................................................................................................................................................... 147

Figure 6-3. The first principal component (PC1[FA]) of skeleton FA across additive model

scores in schizophrenia patients and healthy controls ................................................................ 148

Figure 6-4. PC1 Cortical Thickness across additive model scores in schizophrenia patients and

healthy controls ........................................................................................................................... 149

Figure 6-5. High additive risk allele load predicts lower PC1 fractional anisotropy in

schizophrenia patients ................................................................................................................. 150

Figure 6-6. High additive risk allele load predicts lower PC1 fractional anisotropy in

schizophrenia patients ................................................................................................................. 151

Figure 6-7. High additive risk allele load predicts lower poorer verbal fluency in schizophrenia

patients ........................................................................................................................................ 152

Figure 6-8. High additive risk allele load predicts lower PC1 of motor coordination in

schizophrenia patients ................................................................................................................. 153

Figure 6-9. Voxel-wise mediation analysis of verbal fluency in schizophrenia patients ........... 154

xvi

List of Appendices

Appendix 1: Lett TA, Tiwari AK, Meltzer HY, Lieberman JA, Potkin SG, Voineskos AN,

Kennedy JL, Müller DJ. The putative functional rs1045881 marker of neurexin-1 in

schizophrenia and clozapine response. Schizophr Res. 2011 Nov;132(2-3):121-4.

Appendix 2: Lett TA, Voineskos AN, Kennedy JL, Levine B, Daskalakis ZJ. Treating working

memory deficits in schizophrenia: a review of the neurobiology. Biol Psychiatry. 2014 Mar

1;75(5):361-70.

xvii

List of Abbreviations

AAO age-at-onset

ADHD attention deficit hyperactivity disorder

ADNI Alzheimer’s Disease Neuroimaging Initiative

AKT1 v-akt murine thymoma viral oncogene homolog 1

AMPA alpha-Amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid

ANCOVA analysis of covariance

ASD autism spectrum disorder

BA Brodmann area

BDNF brain-derived neurotrophic factor

BOLD blood-oxygen-level dependent

B-SNIP Bipolar Schizophrenia Network on Intermediate Phenotypes

CACNA1C calcium channel, voltage-dependent, L-type, α 1C subunit

CACNA1D calcium channel, voltage-dependent, L type, alpha 1D subunit

CACNA1E calcium channel, voltage-dependent, R type, alpha 1E subunit

CACNA1S calcium channel, voltage-dependent, L type, alpha 1S subunit

CACNA2D2 calcium channel, voltage-dependent, alpha 2/delta subunit 2

CACNA2D4 calcium channel, voltage-dependent, alpha 2/delta subunit 4

CACNB2 calcium channel, voltage-dependent, beta 2 subunit

xviii

CATIE Clinical Antipsychotic Trials for Intervention Effectiveness

CBT cognitive behavioral therapy

CCA canonical component analysis

CIRS-G Clinical Information Rating Scale for Geriatrics

CNTRICS Cognitive Neuroscience Treatment Research to Improve Cognition

Schizophrenia

CNV copy number variant

COMT catecholamine-O-methyltransferanse

CONSIST Cognitive and Negative Symptoms in Schizophrenia Trial

COWAT controlled oral word association test

CNTNAP2 contactin associated protein-like 2

CSF cerebral spinal fluid

CRT cognitive remediation therapy

DISC1 disrupted in schizophrenia 1

DLPFC dorsolateral prefrontal cortex

DPYD dihydropyrimidine dehydrogenase

DTI diffusion tensor imaging

EDTA ethylenediametetraaecidic acid

EEG electroencephalography

ENCODE Encyclopedia of DNA Elements

ENIGMA Enhancing Neuro Imaging Genetics through Meta-Analysis

xix

ERBB4 v-erb-b2 avian erythroblastic leukemia viral oncogene homolog 4

EZH2 enhancer of zeste 2 polycomb repressive complex 2 subunit

FA fractional anisotropy

FDR false discovery rate

FSL FMRIB Software Library

FWE family-wise error

GABA gamma-Aminobutyric acid

GAD1 glutamate decarboxylase 1

GAD67 glutamic acid decarboxylase-67

GRF Gaussian random field

GRM3 glutamate receptor, metabotropic 3

GSEA gene set enrichment analysis

GWAS genome-wide association study

HARDI high-angular-resolution diffusion imaging

HOXC8 homeobox C8

ICA independent component analysis

ITIH3 inter-alpha-trypsin inhibitor heavy chain 3

IQ intelligence quotient

LD linkage disequilibrium

LICI long interval intracortical inhibition

xx

LSD1 histone lysine-specific demethylase 1

LNS letter-number span

LPBA40 LONI probabilistic brain atlas

LRRTM2 leucine-rich repeat transmembrane protein

MAF Minor Allele Frequency

MATRICS Measurement and Treatment Research to Improve Cognition in

Schizophrenia

MIB1 mindbomb E3 ubiquitin protein ligase 1

MITF microphthalmia-associated transcription factor

MDR multifactor dimensionality reduction

MIR137 microRNA-137

MLA machine learning algorithms

MHC major histocompatibility complex

MMP16 matrix metallopeptidase 16 (membrane-inserted)

MMSE mini mental status exam

MRI magnetic resonance imaging

MTHFR methylenetetrahydrofolate reductase (NAD(P)H)

MULM mass-univariate linear modeling

NDEL1 nudE neurodevelopment protein 1-like 1

NLGN neuroligin

NMDA N-methyl-D-aspartate receptor

xxi

NRG1 neuregulin 1

NRXN1 neurexin-1

OR odds ratio

PCA principal component analysis

PFC prefrontal cortex

PLS partial least squares

PGC psychiatric genomics consortium

PV parvalbumin

RBANS repeatable battery for the assessment of neuropsychological status

RDoC Research Domain Criteria

RGS4 regulator of G-protein signaling 4

ROI region of interest

RRR reduced rank regression

rTMS repetitive transcranial magnetic stimulation

SDCCAG8 serologically defined colon cancer antigen 8

SGB stochastic gradient boosting

SPM statistical parameter mapping

SNP single nucleotide polymorphism

SNV single nucleotide variants

SVM support vector machines

xxii

TBV total brain volume

TFCE threshold-free clustering enhancement

TMT trail making test

TMS transcranial magnetic stimulation

TCF4 transcription factor 4

TURNS Treatment Units for Research on Neurocognition and Schizophrenia

TBSS tract-based spatial statistics

WBP1L WW domain binding protein 1-like

WTAR Wechsler test for adult reading

VIPR2 vasoactive intestinal peptide receptor 2

ZNF804A zinc finger protein 804A

1

Chapter 1

1 Introduction

1.1 Overview

Schizophrenia is a highly heritability disorder, although genetic studies of the disorder poses a

number of significant challenges (Burmeister, McInnis et al. 2008). A condition of any genetics

analysis is a valid and accurate phenotype; however, currently there are no neurobiological-based

tests for schizophrenia. Rather, the clinical diagnosis is usually made by structured interview

based on diagnostic criteria. Furthermore, the symptom heterogeneity among patients with

schizophrenia suggests that there are different subgroups within the disorder. The boundaries of

diagnostic criteria are not distinct, especially when symptoms in patients are not clear.

Environmental conditions can have significant impact in expression of schizophrenia including

prenatal or perinatal events, including but not limited to hypoxemia, infections, birth in winter,

and high-expressed negative emotions in families (Lewis and Levitt 2002). Difficulty arises in

overlapping symptoms with other disorders; for example, psychotic symptoms can be part of a

variety of diagnoses other than schizophrenia, such as mania or depression with psychotic

symptoms.

Schizophrenia is likely a spectrum of disorders that overlaps with autism spectrum disorder

(ASD), bipolar disorder, and schizoaffective disorder. Interestingly, genetic association studies

have mirrored these patterns. First, genetic associations have identified with diagnostic criteria

such as clinical symptoms. For example, the genome wide association study (GWAS) identified

variant rs1344709 of the zinc finger protein 804A (ZNF804A) gene is associated with psychosis

2

(Steinberg, Mors et al. 2010) and psychotic symptoms within bipolar patients (Lett, Zai et al.

2011). Second, there is a high overlap in genetic association findings between different

psychiatric disorders. For instance, there are more genetic variants common to bipolar disorder

and schizophrenia than those differentiating the two diseases (Carroll and Owen 2009, Cardno

and Owen 2014).

An approach to overcome the obstacles of phenotypic heterogeneity in psychiatric disorders is

the concept of intermediate and endophenotypes (e.g. neurophysiological, neuroanatomical, and

cognitive phenotypes). These are closer linked to biological mechanisms and genetic effects that

underlie schizophrenia. They are refined phenotypes with lower environmental and other

confounding influences compared to this complex diagnosis (Gottesman and Gould 2003,

Meyer-Lindenberg and Weinberger 2006). Understanding how schizophrenia risk genes impact

on specific phenotypes of the disorder such as changes in brain structure and how one risk gene

affects multiple subphenotypes might help to develop more biologically-based treatment options

in the future.

1.2 Schizophrenia

1.2.1 Impact, Epidemiology and Prognosis

Schizophrenia is a disabling brain disorder characterized by a diverse array of clinical features

and substantial comorbidity. It is one of the most devastating diseases for several reasons: (a) the

condition is one of the major contributors to global burden of disease (Murray and Lopez 1997).

This burden of the disease primarily due to the early mean onset of the disease in early adulthood

(discussed in detail in the following sections), and despite optimal treatment, approximately two-

3

thirds of affect individuals have chronic and unstable symptoms (American Psychiatric

Association 1994). Furthermore, 20-30 percent of patients fail to respond to treatment (Kane,

Honigfeld et al. 1988, Kane, Meltzer et al. 2007). (b), it is a common disorder. Systematic review

of the prevalence of schizophrenia estimate the lifetime morbid risk at 7.2 per 1000 persons

(Saha, Chant et al. 2005). (c), premature mortality. Twelve percent of the mortality of patients is

due to completed suicide (Brown 1997). Even after accounting for suicide, patients generally

have a much shorter (15-20 years) life expectancy than the general population (Tiihonen,

Lonnqvist et al. 2009). (d), the tremendous societal cost of patient care. Mental illness is the

second leading cause of disability and premature death in Canada (Waddell, McEwan et al.

2005), and the WHO ranks schizophrenia as one of the top ten causes of disability in developed

countries (Barbato and Organization 1998). The economic burden of schizophrenia in the USA

was estimated at $75 billion in 2012 dollars (Kennedy, Altar et al. 2014). (e), Current

pharmaceutical treatment of schizophrenia, antipsychotic medication, is effective against positive

symptoms (e.g., psychosis including hallucinations and delusions) but has little or no effect on

cognitive impairments (e.g., reduced executive functioning and concentration) (Mishara and

Goldberg 2004, Goldberg, Goldman et al. 2007, Keefe, Bilder et al. 2007, Keefe, Sweeney et al.

2007). This is particularly distressing since cognition is a key determinant of patient long-term

outcome and mortality in schizophrenia (Green 1996). (f), the heterogeneity of the schizophrenia

symptoms suggests a complex disorder (discussed below); therefore, no single treatment will be

efficacious for all patients.

1.2.2 Treatment of Schizophrenia

The introduction of chlorpromazine in the 1950s provided the first effective pharmacological

treatment of schizophrenia, and established antipsychotics as the primary intervention in many

4

psychiatric disorders. Numerous typical antipsychotics or first-generation antipsychotics were

subsequently developed, and collectively demonstrated successful treatment of psychotic

symptoms, in particular delusions and hallucinations. The benefits of first generation

antipsychotics were offset by the common occurrence of elevated prolactin and extrapyramidal

symptoms, such as parkinsonism and tardive dyskinesia. In the early 1970s, clozapine was

introduced as the first atypical antipsychotic or second-generation antipsychotic, which marked

an advantage over typical medications via increased efficacy and fewer neurological side effects

(Kane 1988, Foussias and Remington 2010). Among the second-generation antipsychotics that

have been developed, arguably only clozapine shows an increased efficacy of treatment

compared with first generation antipsychotics (Kane and Correll 2010). Unfortunately, most

second generation antipsychotic medication (in particular clozapine and olanzapine) have been

linked to substantial drug-induced weight gain, which confers risks of metabolic syndrome,

leading to diabetes mellitus type II and cardiovascular disease (Lett, Wallace et al. 2012).

Furthermore, cardiovascular disease is the leading cause of death in schizophrenia (Newcomer

2007).

In general, most second generation antipsychotic medication use a ‘pharmacological shotgun’

approach with strong affinity for the serotonin 5-HT2 receptor and concomitant affinities for

dopaminergic, muscarinic, histaminergic and adrenergic receptors (Meltzer, Matsubara et al.

1989). In contrast, first generation antipsychotics are more specific, and have greater affinity to

the dopamine D2 receptor. There are exceptions to this generalization including phenothiazines

(e.g., chlorpromazine) that have a more diverse binding profile and second generation

antipsychotics that have high D2 and D3 antagonism (e.g., risperidone and ziprasidone)

(Nasrallah 2008). Although second generation antipsychotics have preferential action on 5-HT2

receptors, the rapid dissociation of most second generation antipsychotics from the D2 receptor

5

has been suggested to account for the atypicality (Kapur and Remington 2001, Kapur and

Seeman 2001). Perhaps the major distinction is that second generation antipsychotics constitute a

major improvement in the avoidance of extrapyramidal symptoms through reduced D2 receptor

occupancy (Kapur, Zipursky et al. 2000).

Since antipsychotic medications mainly target positive symptoms, many adjunctive therapies

have been developed including: novel pharmacologic compounds (particularly for cognitive

dysfunction and negative symptoms), cognitive behavioral therapy, cognitive remediation

therapy, repetitive transcranial magnetic stimulation (rTMS), and electroconvulsive therapy (in

catatonic patients). For in-depth review please see (Lett, Voineskos et al. 2014) and the

Discussion section.

1.3 Sources of Schizophrenia Heterogeneity

It is now clear that the disease trajectory of schizophrenia carries tremendous heterogeneity.

There is wide variability from patient to patient in: age of onset of psychotic symptoms, rate of

onset, inter-episode residual impairment, long-term outcome, treatment response, treatment

emergent side effects, functional and structural brain abnormalities, and the severity or absence

of core symptoms of the disorder including positive symptoms, negative symptoms, and

cognitive impairment (Carpenter Jr and Kirkpatrick 1988, DeLisi, Hoff et al. 1991, DeLisi 1992,

Shenton, Dickey et al. 2001, Lett, Wallace et al. 2012, Lett, Brandl et al. 2014, Lett, Voineskos

et al. 2014). Therefore, it has been suggested that schizophrenia is a clinical syndrome rather

than a single disease entity (Carpenter Jr and Kirkpatrick 1988). The following section will

discuss some aspects of the heterogeneity within schizophrenia that may provide key insights

6

into the pathogenesis of schizophrenia. Here we focus on: (i) onset of illness, (ii) cognitive

dysfunction, and (iii) brain structure.

1.3.1 Onset of Psychotic Symptoms

There are many measures of age-at-onset (AAO) of schizophrenia (e.g., 1st psychotic episode, 1st

psychiatric interview, 1st hospitalization, 1st antipsychotic treatment, onset of negative symptoms,

and others), although, they tend to be highly correlated. There is a prodromal stage in which

cognitive symptoms are present (e.g. general fluid intelligence (gF), executive functioning), and

the prodromal cognitive symptoms may contribute to heterogeneity in patterns of cognitive

changes across illness phases and among individuals (Mesholam-Gately, Giuliano et al. 2009).

Furthermore, there are large cognitive deficits across most measures in first-episode

schizophrenia (Heinrichs and Zakzanis 1998, Mesholam-Gately, Giuliano et al. 2009, Rajji,

Ismail et al. 2009). The AAO of psychotic symptoms normally occurs between the ages of 15-24

years old (Messias, Chen et al. 2007). The mean AAO in men is approximately three years

earlier than in women. Furthermore, women tend to have a bimodal distribution of AAO with

second peak around 55-64 years of age. Schizophrenia is rare in children less than 12 years of

age (approximately 1%), and the prevalence in adolescents (12-18 years) is approximately 12-

33% (Kumra, Oberstar et al. 2008). However, it is likely that many patients have psychotic

symptoms before the age of 18 and seek treatment later (Schimmelmann, Schmidt et al. 2013).

The prevalence of late onset (greater than 40 years of age) schizophrenia is more difficult to

determine due to confounding factors associated with aging, although estimates range from 15%

to 32% of patients (Harris and Jeste 1988).

7

The AAO may have tremendous effects on social and vocational functioning. For instance, it

could be the difference between a patient finishing a degree, having a job, or being married.

Perhaps more importantly, AAO can provides can provide key insights into long-term clinical

outcome of schizophrenia. There is a relatively strong relationship between earlier AAO and

more severe cognitive symptoms of schizophrenia. Meta-analysis reveals that individuals with

youth-onset schizophrenia demonstrate larger deficits than those with first-episode schizophrenia

on arithmetic, executive function, IQ, psychomotor speed of processing, and verbal memory

(Rajji, Ismail et al. 2009). In contrast, late-onset patients schizophrenia patients tend to have

relatively conserved cognitive function (Rajji, Ismail et al. 2009). This is particularly important

as cognitive performance is an established predictor of functional outcome (Green 1996). Men

have greater negative symptom burden, worse clinical outcome, and poorer social functioning

than women (Abel, Drake et al. 2010), and these differences could be attributed to earlier AAO.

More specifically, early onset patients have greater symptoms at the onset of psychosis

(Ballageer, Malla et al. 2005). Early onset patients display further prognostic criteria including

more neurodevelopmental deficits and lower premorbid adjustment (Hans, Auerbach et al. 2000,

Joa, Johannessen et al. 2009), poorer social outcome (Szymanski, Lieberman et al. 1995,

Lauronen, Miettunen et al. 2007), and worse treatment response as well as poorer longitudinal

clinical outcome (Reichert, Kreiker et al. 2008, Vyas, Patel et al. 2011).

There is increasing evidence for a neurobiological basis to AAO of psychosis. Genetic factors

significantly contribute to the AAO of psychotic symptoms with a moderately high degree of

heritability (H2=0.33±0.9) (Hare, Glahn et al. 2010). There have been two genome-wide analysis

of AAO, although the results have been mixed (Wang, Liu et al. 2011, Bergen, O'Dushlaine et al.

2014). AAO also had an effect on clinically relevant brain structure. Earlier AAO has been

directly linked to enlarge left lateral ventricles, and together this suggested to be early predictors

8

of poor outcome (Crow 1980, DeLisi, Hoff et al. 1991, DeLisi 1992, Ho, Andreasen et al. 2003).

Furthermore, earlier AAO has been associated with reduced hippocampal volume (Giedd,

Jeffries et al. 1999), reduced gray matter volume (Marsh, Harris et al. 1997, Zipursky, Lambe et

al. 1998), and lower fractional anisotropy (FA) in the frontal cortex (Kumra, Ashtari et al. 2004,

Kyriakopoulos, Perez-Iglesias et al. 2009). All structure changes have arguably been associated

with a poorer outcome in schizophrenia (Ho, Andreasen et al. 2003, Szeszko, Robinson et al.

2008).

1.3.2 Cognitive Symptoms

Since the earliest conceptualization of schizophrenia, cognitive impairment has been viewed as a

core feature of schizophrenia (Bleuler 1950, Kraepelin 1971). Despite notable heterogeneity

among individuals with schizophrenia (Seidman 1990, Kremen, Seidman et al. 2004), it is

regarded as a core feature of the disorder (Elvevag and Goldberg 2000, Weickert, Goldberg et al.

2000). Neurocognitive dysfunction is prevalent in at least 70% of patients before disease onset

and after chronic treatment (Palmer, Heaton et al. 1997, Heinrichs and Zakzanis 1998,

Mesholam-Gately, Giuliano et al. 2009). Cognitive dysfunction is present in healthy relatives of

schizophrenia patients, and it has been suggested as a biomarker of schizophrenia (Barrantes-

Vidal, Aguilera et al. 2007). Compared to healthy controls, schizophrenia outpatients have a

generalized cognitive impairment. However, patient performance on the vast majority of

neurocognitive tests tends to show no deterioration over time; for example, performance on full-

scale IQ, attention, verbal and non-verbal memory, and visual skills are remarkably stable (Gold,

Arndt et al. 1999, Kurtz 2005). Cognitive dysfunction in schizophrenia shows high prevalence

and is relatively stable over time. For instance, measurements of neurocognitive impairment have

9

been more closely linked to community outcome, social problem solving, and social skill

acquisition rather than symptoms (Green 1996, Green, Kern et al. 2000, Gold 2004).

Currently approved pharmaceutical treatments for schizophrenia are typically effective for

positive symptoms, but have little or no effect on cognitive impairment (Keefe, Bilder et al.

2007). Although antipsychotic medications show small effects on cognitive performance with

treatment, some studies show therapeutic advantages of atypical antipsychotics compared to

typical antipsychotics(Woodward, Purdon et al. 2005); however, the large, multisite Clinical

Antipsychotic Trials for Intervention Effectiveness (CATIE) trial failed to find any advantage of

atypical antipsychotics in treating cognition (Keefe, Bilder et al. 2007). Clozapine, the only

second generation antipsychotic medication with any efficacy for treatment resistant

schizophrenia (Kane, Honigfeld et al. 1988), is no longer considered superior to other

antipsychotic mediation for cognitive deficits (Harvey, Sacchetti et al. 2008). These results were

driven by multiple pharmacologic initiatives such as the Measurement and Treatment Research

to Improve Cognition in Schizophrenia (MATRICS) (Green, Nuechterlein et al. 2004),

Treatment Units for Research on Neurocognition and Schizophrenia (TURNS) (Buchanan,

Freedman et al. 2007), and Cognitive Neuroscience Treatment Research to Improve Cognition

Schizophrenia (CNTRICS) (Barch and Smith 2008). Other pharmacological agents have been

studied as adjunctive treatment options for cognitive symptoms. For instance, results from the

Cognitive and Negative Symptoms in Schizophrenia Trial (CONSIST) suggest that either

glycine (binds to allosteric site of the N-methyl-D-aspartate receptor (NMDAR)) or D-

cycloserine (partial NMDAR agonist) were not effective in treating cognitive impairments

(Buchanan, Javitt et al. 2007). Catecholamine-O-methyltransferanse (COMT) has been directly

associated with prefrontal cortex dopamine turnover and working memory performance (Meyer-

Lindenberg, Kohn et al. 2005). COMT inhibitors, such as tolcapone, are a promising target

10

although they have unfortunately also been associated with hepatotoxicity (Goff, Hill et al.

2011). The selective agonist of the GABA (gamma-Aminobutyric acid) A receptor, MK-0777,

was shown to be effective in treating working memory deficits in a study with limited sample

size (Lewis, Cho et al. 2008); however, a subsequent study failed to replicate the enhancement of

working memory by MK-0777 in schizophrenia (Buchanan, Keefe et al. 2011). Beyond

pharmacological agents, cognitive remediation therapy consistently showed improvement in

social cognition (effect size ~ 0.65) and working memory (effect size ~ 0.35) (McGurk,

Twamley et al. 2007, Wykes, Huddy et al. 2011, Lett, Voineskos et al. 2014). Furthermore, there

is promising evidence that rTMS or anodal direct current stimulation may be effective at

improving cognitive functioning in schizophrenia (Utz, Dimova et al. 2010, Mulquiney, Hoy et

al. 2011, Barr, Farzan et al. 2013, Lett, Voineskos et al. 2014).

In summary, neurocognitive dysfunction predicts functional outcome in schizophrenia, and while

there are some promising treatment strategies, no staple treatment has been established yet.

Therefore, it is imperative for further investigation into the neurobiological mechanism of

dysfunction, and potentially which patients may best respond to a given treatment.

1.3.3 Brain Structure

Several meta-analyses have demonstrated replicable abnormalities in cross sectional structural

MRI studies of patients with first-episode and chronic schizophrenia. The most robust findings

comparing cases and controls are reductions in whole brain and gray matter volume (particularly

in limbic, paralimbic, and frontal cortical region as well as the thalamus), and enlargement of the

lateral ventricles (Wright, Rabe-Hesketh et al. 2000, Shenton, Dickey et al. 2001, Honea, Crow

et al. 2005, Steen, Mull et al. 2006, Vita, De Peri et al. 2006). Meta-analysis of diffusion tensor

11

imaging (DTI) MRI studies of white matter fractional anisotropy (FA) complement the gray

matter findings with lower FA in deep white matter frontal lobe (traversing white matter tracts

including the genu of the corpus callosum, cingulum bundle, left anterior thalamic radiation, left

inferior fronto-occipital fasciculus), and deep white matter in the left temporal lobe (splenium of

the corpus callosum, fornix, left inferior longitudinal fasciculus) (Ellison-Wright and Bullmore

2009). Furthermore, progressive brain abnormalities have been observed in longitudinal studies

in schizophrenia. Meta-analyses show a yearly decrease in whole brain gray matter, frontal white

matter, parietal white matter, and temporal white matter as well as increase in lateral ventricles

even compared to controls (Olabi, Ellison-Wright et al. 2011). It is important to note that both

white and gray matter volume decreases are inversely correlated to antipsychotic treatment, but

not duration of illness or severity in longitudinal studies of chronic and first episode and chronic

patients (Ho, Andreasen et al. 2011). Together, this suggests that antipsychotic treatment may be

causing these neuroanatomical changes, and is an important source of heterogeneity. The only

regions that moderated the effects of age (greater in patients versus controls) and illness duration

were the hippocampi (Fusar-Poli, Smieskova et al. 2013, Torres, Portela-Oliveira et al. 2013).

Furthermore, AAO predicts significant heterogeneity in frontal gray matter (please see the AAO

section) (Olabi, Ellison-Wright et al. 2011). Brain structure may also be an important

determinant of clinical symptoms within schizophrenia. For example, schizophrenia patients with

a deficit subtype have greater mean diffusivity in the right inferior longitudinal fasciculus, right

arcuate fasciculus, and left unicinate fasciculus compared to matched first episode, non-deficit

schizophrenia patients (Voineskos, Foussias et al. 2013).

12

1.4 Cortical Thickness and Diffusion Tensor Imaging Measures of White Matter Structure

The following sections will highlight the two major MRI methods employed in this thesis to

examine the genetic influence on in vivo brain structure: cortical thickness and DTI. The field of

structural neuroimaging is dynamic and the methodology discussed builds on earlier imaging

modalities applied in schizophrenia research including computed tomography (e.g., (Johnstone,

Crow et al. 1976)), volumetric analyses (e.g., (Smith, Calderon et al. 1984)), and voxel-based

morphometry (e.g., (Kubicki, Shenton et al. 2002)). For review of these modalities, please see

(Shenton, Dickey et al. 2001, Fusar-Poli, Radua et al. 2012). Next, we will discuss genetics

contribution of cortical thickness and white matter DTI, and its particular relevance in

schizophrenia heterogeneity.

1.4.1 Analytic Approaches to Cortical Thickness and Diffusion Tensor Imaging

Neuroimaging studies have shown volumetric reductions in a number of cortical regions (notably

in frontal and temporal cortices) as well as several subcortical regions (particularly in

hippocampal volumes). Advances in morphological imaging allow the estimation of structural

features in a sub-voxel range. This enables the reconstruction of cortical surfaces that allows

estimation of cortical thickness, surface area, and sulcal depth (Dale, Fischl et al. 1999, Fischl,

Sereno et al. 1999, Fischl and Dale 2000, Han, Jovicich et al. 2006). These measures are of

particular interest to schizophrenia research since: (a), abnormal architecture of the cortex has

been identified in post mortem studies of schizophrenia (Weinberger and Lipska 1995,

Rajkowska, Selemon et al. 1998, Selemon, Mrzljak et al. 2003); (b), they can provide enhance

understanding of the pathogenesis of the disorder (Murray and Lewis 1987, Weinberger 1987,

Lewis and Lieberman 2000, Lewis and Levitt 2002, Lewis, Hashimoto et al. 2005); and (c),

13

cortical volume are composed of surface area and cortical thickness that may under distinct

genetic (Panizzon, Fennema-Notestine et al. 2009), environmental (Raznahan, Cutter et al.

2010), and cellular (Chenn and Walsh 2002) influences. In particular, cortical thickness

represents the density and arrangement of cells (neurons, glia, and nerved fibers) (Chenn and

Walsh 2002). Aberrant neurogenesis, neuronal migration, differentiation, synaptogenesis, and

mechanisms involved in synaptic pruning have all been implicated in schizophrenia and reflect

changes in cortical thickness (Jakob and Beckmann 1986, Arnold 1999). Therefore, cortical

thickness is a means to potentially model neurodevelopmental contribution to schizophrenia with

high regional specificity. The most common approaches employ automated or semi-automated

procedures. In brief, a three dimensional polygonal mesh is applied over the cortical surface.

Cortical thickness is determined to be the distance between the white matter surface and the gray

matter-cerebral spinal fluid intersection. At each vertex (from the mesh) a scalar value (ranging

from approximately 1.5 to 4.5 mm in healthy controls) is outputted in millimeters representing

the cortical thickness at that point. Limitations of this method include the determining the inner

and out surface areas accurately using the resolutions of T1-weighted scan (normally around 1

mm3), and the fine detail of sulcal regions can be obscured by partial volume effects (regions of

poor definition due to the osmotic movement of water).

In addition to gray matter abnormalities observed in schizophrenia, microscopic and molecular

studies demonstrate oligodendrocyte abnormalities in schizophrenia (Hakak, Walker et al. 2001,

Uranova, Orlovskaya et al. 2001). Oligiodendrocytes are neuroglia that play a major role in

establishing the conductivity, and in protection of neuronal axons travelling within white matter

tracts by forming myelin sheets (Cotter, Pariante et al. 2001). Furthermore, they support cellular

metabolism as well as function in both neuronal migration and synaptic signalling regulation

(Fields and Burnstock 2006). All of these functions have also been reported to be disrupted in

14

schizophrenia (Lewis, Hashimoto et al. 2005). Therefore, it is conceivable that glial

abnormalities have a substantial contribution to the reduced neuronal size, reduced synaptic field

density, and functional dysconnectivity observed in schizophrenia (Benes, Davidson et al. 1986).

DTI MRI is thought to be, arguably, an indicator of white matter tract integrity, and currently is

the mainstay for assessing in vivo measures of white matter structure. The basic principal of DTI

is that diffusion weighted magnetic resonance sequence (gradient pulses) excite protons from

water molecules to vibrate in phase. The molecular diffusion of water in white matter tracts is

organized providing an estimation of diffusion. The degree of organized movement along white

matter tracts can be represented as a vector. The most common method of characterizing

diffusion within a voxel is via diffusion tensor (Basser, Mattiello et al. 1994). Three orthogonal

axes correspond to diffusivities along each axis over time (eigenvalues). When diffusion

eigenvalues are equal, the diffusion tensor is isotropic (equal diffusion in all directions);

whereas, unequal eigenvalues are anisotropic. Abnormalities detected using DTI-based measures

of anisotropy are thought to reflect coherence of white matter fibers; however, changes in density

and crossing of interconnecting fibers may affect the degree of anisotropy. Myelin is considered

to be the major (but not only) barrier to diffusion of white matter tracts (Beaulieu 2002). One of

the widely used metrics of diffusion anisotropy is fractional anisotropy (FA) (Basser, Mattiello et

al. 1994). FA measures the fraction of the tensor that is due to anisotropic diffusion. The FA

index is normalized (ranging from 0 to 1) and FA maps are created that can distinguish voxels

contain white matter fibers from gray matter and CSF.

There are two major categories of analyses that can be performed on DTI data: deterministic and

probabilistic. Deterministic tractography reconstructs three dimensional fiber trajectories of

anisotropic structures using voxel-based estimates of the continuous fiber orientation field

(Conturo, Lori et al. 1999). Deterministic methods normally involve either region of interest

15

(ROI) analyses or applying clustering algorithms to reconstruct white matter tracts. The average

FA value (or other DTI metrics) of the white matter tracts identified can then be assessed as

quantitative variables for statistical analysis. Probabilistic methods are generated by seed points

with random perturbations to the trajectory direction (Descoteaux, Deriche et al. 2009).

Probabilistic fiber tracking methods create a distribution of possible pathways that are weighted

by their likelihood. A newer voxel-based probabilistic approach is tract-based spatial statistics

(TBSS) (Smith, Jenkinson et al. 2006). TBSS creates a white matter FA skeleton for each subject

using peak FA values on a mean group template. TBSS provides batter alignment of tracts across

subjects and is robust again registration errors (Smith, Jenkinson et al. 2006). Voxel-wide

analyses can then be performed using a variety of statistical methodologies.

1.4.2 Genetic Contribution to Cortical Thickness and White Matter Fractional Anisotropy (FA)

Since there is a considerable genetic contribution to schizophrenia (See Chapter 1.6), it is of

strong interest if cortical thickness and FA are heritable (the proportion of variance of a given

traits that can be explained by genetics). The degree of heritability of any morphological

structure is dynamic, depending point of development and stability of environmental influences;

nevertheless, twin studies have provided estimations of heritability in adulthood.

Total cortical thickness and cortical surface have both been shown to highly heritable

(approximately 80%), and may be genetically distinct (Panizzon, Fennema-Notestine et al. 2009,

Rimol, Panizzon et al. 2010). Within schizophrenia patients, cortical thickness has also shown

high heritability (Goldman, Pezawas et al. 2009). However, first degree relatives of these

patients only demonstrated marginal differences from healthy controls (Goldman, Pezawas et al.

2009). These results suggest that cortical thickness may not be a strong intermediate phenotype

16

of schizophrenia, but within the disease cortical thickness still carries significant contribution

from genetic factors. DTI twin studies have consistently shown substantial heritability (40-80%)

among white matter tracts in different stages of development (Pfefferbaum, Sullivan et al. 2001,

Chiang, Barysheva et al. 2009, Brouwer, Mandl et al. 2010, Kochunov, Glahn et al. 2010,

Chiang, McMahon et al. 2011, Geng, Prom-Wormley et al. 2012). Meta-analytic studies support

heritability of white matter FA. Furthermore, graph theoretical approaches applied to FA

demonstrate brain topography is also a moderate heritable explaining (57-68%) (Bohlken, Mandl

et al. 2014). Results from the ENIGMA (Enhancing Neuro Imaging Genetics through Meta-

Analysis) consortium pooled five samples (4 twin and 1 family sample; N=2248) demonstrated

that all white matter tracts except the cortico-spinal tract are heritable (ranging from 40-

70%)(Kochunov, Jahanshad et al. 2014). Additive genetic variance explained over 50% of the

inter-subject variance in FA values (Kochunov, Jahanshad et al. 2014). Results from the B-SNIP

(Bipolar Schizophrenia Network on Intermediate Phenotypes) consortium supports the high

degree of white matter FA heritability (Skudlarski, Schretlen et al. 2013). Furthermore, first

degree relatives of patients had significant reductions in FA compared to healthy control,

suggesting that FA may be strong intermediate phenotype (Skudlarski, Schretlen et al. 2013).

Taken together, these studies suggest that both cortical thickness and white matter FA have

substantial contributions from genetic factors, although FA may be a more suitable intermediate

phenotype.

1.5 Functional Integration in Schizophrenia

There is overwhelming evidence that schizophrenia is, at least in part, a disorder of abnormal

functional integration of the brain (Pettersson-Yeo, Allen et al. 2011). This dysconnectivity of

brain processes may be due to aberrant wiring during development, due to aberrant plasticity, or

17

both. Features of this dysfunction are abnormal functional connectivity (e.g. frontotemporal

connectivity (Meyer-Lindenberg, Olsen et al. 2005), gamma synchrony (Spencer, Nestor et al.

2003)), abnormal structural connectivity (e.g. white matter integrity (Kubicki, Park et al. 2005,

Voineskos, Lobaugh et al. 2010), corpus callosum morphology (Woodruff, McManus et al.

1995), reduced brain asymmetry (Sommer, Ramsey et al. 2001)), synaptic plasticity (e.g.

pharmacological induced schizophrenia symptomology (Kapur 2003), and reduced dendritic

field size and density (Glantz and Lewis 2000). Neural connectivity is dynamically regulated by

signaling pathways relating neuronal activity to the expression of key activity regulated genes

(Flavell and Greenberg 2008, Leslie and Nedivi 2011). Genes associated with schizophrenia are

common to all of the abovementioned points of dysfunction suggesting that genetic

underpinnings potentiate dysconnectivity. Moreover, neural dysconnectivity may be a causative

factor in the more intractable deficits of schizophrenia, such as working memory functioning

(Tan, Choo et al. 2005). Thus, understanding how these schizophrenia liability genes influence

functional integration may describe an important underlying susceptibility mechanism of

schizophrenia.

1.6 Schizophrenia Genetics: An Update

The current knowledge on the complex genetic architecture of schizophrenia is rapidly

expanding. There are numerous excellent reviews on the genetic basis of schizophrenia (for

example, (Burmeister, McInnis et al. 2008, Sullivan, Daly et al. 2012, Horvath and Mirnics

2014)). This section will primarily focus on some of the most recent and exciting genome-wide

genetic findings.

Evidence from over 40 years of epidemiological and genetic studies has demonstrated that

schizophrenia is a complex genetic disorder, with both genetic and environmental determinants.

18

Twin and family studies of schizophrenia have demonstrated an estimated heritability ranging

from 0.60 to 0.80. The presence of a first degree relative with schizophrenia increases the

lifetime risk to 6.5%. Further, if one parent is affected the risk is 13%, and if both are affected

the risk increases to 50%. Meta-analyses of candidate gene studies and genome-wide analyses

have discovered: (a) common variants with low effect size (minor allele frequency (MAF)>0.1;

OR ~ 1.1-1.2), (b) rare variants (MAF <0.1) with greater effect sizes, and (c) rare copy number

variants (CNVs, insertions or deletion greater than 100 BP; OR ~ 4-20).

Recently, there has been a paradigm shift in genetic study of schizophrenia. The field is moving

towards large consortia examining risk of over tens of thousands of patients. Very large samples

are necessary to reliably detect the rare occurrence or small effect of these variants, and consortia

has been crucial for achieving improved statistical power. Perhaps the most ambitious of the

many consortia in psychiatry is the Psychiatric Genomics Consortium (PGC) (Psychiatric 2009).

The PGC has, currently, collected approximately 125,000 cases and controls with GWAS date

for mega-analyses of schizophrenia and other neuropsychiatric disorders (Giusti-Rodriguez and

Sullivan 2013). Most recently, it was found that 8300, mostly common SNPs, additively

accounted for 32% of the variance in schizophrenia liability (Ripke, O'Dushlaine et al. 2013),

suggesting massive sample size may be effective at explaining a high proportion of the ‘missing

heritability’ of schizophrenia (Manolio, Collins et al. 2009).

The most recent schizophrenia GWAS from the PGC identified 22 regions with genome-wide

significance including 13 novel regions and replication of previous GWAS findings (Ripke,

O'Dushlaine et al. 2013). The replication regions included the MHC (major histocompatibility

complex), CACNA1C (calcium channel, voltage-dependent, L-type, α 1C subunit), ITIH3 (inter-

alpha-trypsin inhibitor heavy chain 3), MIR137 (microRNA-137), MMP16 (matrix

19

metallopeptidase 16 (membrane-inserted)), SDCCAG8 (serologically defined colon cancer

antigen 8), and WBP1L (WW domain binding protein 1-like) (Ferreira, O'Donovan et al. 2008,

International Schizophrenia Consortium, Purcell et al. 2009, Shi, Levinson et al. 2009,

Stefansson, Ophoff et al. 2009, Psychiatric 2011, Ripke, Sanders et al. 2011, Hamshere, Walters

et al. 2013, Ripke, O'Dushlaine et al. 2013). Interestingly, MIR137 and CACNA1C provide key

insight into molecular pathways that may be disrupted in schizophrenia (discussed in detail in the

preceding sections). The MIR137 region has emerged as one of the best clues into the genetic

etiology of schizophrenia. First, it is strongly associated with schizophrenia in independent

genome-wide analyses similar effect sizes (combined analyses: OR = 0.89±0.02, p = 1.72 x 10-

12) (Ripke, O'Dushlaine et al. 2013). Second, microRNA-137 targets other schizophrenia risk

variants. Fourteen out of the 22 identified risk regions as well as CNVs associated with

schizophrenia (e.g. NRXN1) are putative targets of microRNA-137 (http://www.targetscan.org).

Moreover, this dynamic regulation has been confirmed in vitro for a number of these variants

(Kwon, Wang et al. 2011, Kim, Parker et al. 2012). Third, microRNA-137 has been implicated in

neurodevelopment, adult neural stem cell maturation, and dendritic arborisation (Szulwach, Li et

al. 2010, Sun, Ye et al. 2011, Willemsen, Valles et al. 2011, Sun, Gong et al. 2012).

There is now compelling evidence for disruption of the calcium signaling pathway increasing

liability multiple neuropsychiatric disorders. The CACNA1C and CACNB2 regions are associated

with schizophrenia in GWAS (Ripke, O'Dushlaine et al. 2013), and these regions also conferred

cross-disorder association in autism spectrum disorder, attention deficit-hyperactivity disorder,

bipolar disorder, major depressive disorder, and schizophrenia (Cross-Disorder Group of the

Psychiatric Genomics, Smoller et al. 2013). CACNA1C had been associated with schizophrenia

and bipolar disorder in multiple independent GWAS cohorts (Ferreira, O'Donovan et al. 2008,

Nyegaard, Demontis et al. 2010, Lett, Zai et al. 2011, Psychiatric 2011, Schizophrenia

20

Psychiatric Genome-Wide Association Study 2011, Hamshere, Walters et al. 2013, Ruderfer,

Fanous et al. 2013). In pathways based analysis, 20 of the 67 gene regions (CACNA1C,

CACNA1D, CACNA1E, CACNA1S, CACNA2D2, CACNA2D4, CACNB2) in the calcium activity

genes set were associated with the cross-disorder association (α < 1 x 10-3) (Cross-Disorder

Group of the Psychiatric Genomics, Smoller et al. 2013). Furthermore, calcium channel signaling

is critically important in learning, memory, and synaptic plasticity suggesting a common source

of neurocognitive dysfunction potentially independent of any particular neuropsychiatric

disorder(Moosmang, Haider et al. 2005, Baumgartel and Mansuy 2012, Bading 2013).

There are multiple lines of evidence implicating a role of the immune system in the neurobiology

of schizophrenia. Epidemiological studies point to maternal infections leading to an

inflammatory response that may elevate the risk for schizophrenia. For instance, influenza

infections lead to seven-fold increase in schizophrenia risk during the first trimester (Brown,

Begg et al. 2004). Prenatal exposure to rubella is associated with a five-fold elevated risk of

psychosis in adulthood (Brown, Cohen et al. 2000). Toxoplasma gondii infections are associated

with a two-fold increase of schizophrenia risk (Mortensen, Norgaard-Pedersen et al. 2007,

Torrey, Bartko et al. 2012). Last, maternal genital infections are associated with a five-fold

increase in disease liability (Babulas, Factor-Litvak et al. 2006). GWAS provides convergent

evidence for common SNPs being involved in immune function contributing to schizophrenia

etiology.

The region on the 6p22.1 that includes the major histocompatibility complex (MHC) has the

most consistently replicated association with schizophrenia (International Schizophrenia

Consortium, Purcell et al. 2009, Shi, Levinson et al. 2009, Stefansson, Ophoff et al. 2009, Irish

Schizophrenia Genomics and the Wellcome Trust Case Control 2012, Jia, Wang et al. 2012).

21

Indeed, the most recent PGC consortia findings pointing to the 6p22.1 region have been

overwhelmingly the strongest association (p<10-18 to P<10-31) (Schizophrenia Working Group of

the Psychiatric Genomics 2014). Despite the fact that 6p22.1 encompasses over 200 genes

spanning 8 Mb, most of the markers are in strong linkage disequilibrium. This creates substantial

difficulty in assessing the function significance of any specific marker as well as its relationship

to any particular gene. It is important to note that genes in this region may function beyond

immune regulation. 6p22.1 also includes histone protein genes that may be relevant in regulation

of transcription, or DNA repair via epigenetic regulation (Costa, Dong et al. 2007) and

antimicrobial defense (Kawasaki and Iwamuro 2008). Furthermore, the MHC region, particularly

the MHC class I family (MHCI), has an integral function in brain development and

neuroplasticity (Huh, Boulanger et al. 2000, Boulanger 2009, Deverman and Patterson 2009).

For example, MHCI proteins including TNF-α, IL-6, and IL-1β are essential for adult neural

stem cell regulation in the subventricular zone of the lateral ventricles and subgranular zone of

the hippocampus (Carpentier and Palmer 2009). This regulation is similar to the function of

microRNA-137 suggesting a potential common risk pathway (See Chapter 4.2). Taken together,

the GWAS findings in the 6p22.1 region provide one of most interesting clues into the genetic

and environmental contribution to the etiology of schizophrenia. Furthermore, since proteins

expressed in the 6p22.1 region have a critical and diverse role in neurodevelopment and

plasticity, they may predict heterogeneous features related to schizophrenia including differences

in brain structure or cognitive performance.

There is also building evidence that rare genomic variation may be playing a role in

neuropsychiatric disorders. CNVs are over-represented in schizophrenia and other

neuropsychiatric disorders. The 22q11.2 CNVs are robustly associated with schizophrenia

(Karayiorgou, Simon et al. 2010, Levinson, Duan et al. 2011). Other CNV have been

22

consistently been associated with schizophrenia including the NRXN1 and VIPR2 (vasoactive

intestinal peptide receptor 2) gene regions (Rujescu, Ingason et al. 2009, Vacic, McCarthy et al.

2011) as well as other multiplex gene regions (1q21, 3q29, 7q11, 15q11, 15q13, 16q13, 16p11,

17q12) (Walsh, McClellan et al. 2008, Malhotra and Sebat 2012, Sullivan, Daly et al. 2012). In

general, CNVs tend to be non-specific, and are risk factors for multiple disorders including

schizophrenia, mental retardation, and autism spectrum disorder. In contrast to CNVs, there is

less evidence for rare single nucleotide variants (SNV) associated with schizophrenia. Whole

exome sequencing studies have provided evidence for increased de novo SNVs in schizophrenia

including modest associations in the DPYD (dihydropyrimidine dehydrogenase) gene, a region

with SNPs that are in linkage disequilibrium (LD) with the MIR137 GWAS variants. However,

there have been some mixed findings, and further research in larger samples will provide better

evidence for the role of SNV in schizophrenia.

It is likely that the genetic architecture of schizophrenia includes contribution from common

variants, highly penetrant CNVs, and potentially exome SNVs. Future research will improve our

understanding. For example, preliminary results from the latest PGC schizophrenia mega-

analysis (25000 schizophrenia patients and 28000 controls) increased the number of significant

genome-wide associations to 62 (Anderson‐ Schmidt, Beltcheva et al. 2013), and many of the

new genome-wide associations have been identified by candidate gene analyses. As the genetic

findings become more concrete, it will be imperative to assess the biological relevance of these

genomic findings.

23

1.7 Important Genetic Modifiers of Schizophrenia Phenotypes

1.7.1 Neurexin-1 (NRXN1)

The NRXN1 gene is one of the largest known human genes (1.1 Mb) with 24 exons, located on

chromosome 2p16.3 (Südhof 2008). The NRXN1 gene encodes the neurexin-1α and neurexin-1β

proteins that function as pre-synaptic neural adhesion molecules. Neurexin-1α is reported to

interact with postsynaptic neuroligins mediating GABAergic and glutamatergic synapse function

(Südhof 2008). It also binds to leucine-rich repeat transmembrane protein (de Wit, Sylwestrak et

al. 2009), instructing presynaptic and mediating postsynaptic differentiation of glutamatergic

synapses. Substantial evidence implicates deletions in the NRXN1 gene in ASD (Feng, Schroer et

al. 2006, Szatmari, Paterson et al. 2007, Kim, Kishikawa et al. 2008, Marshall, Noor et al. 2008,

Morrow, Yoo et al. 2008, Yan, Noltner et al. 2008, Glessner and Hakonarson 2009), and in

schizophrenia (Vrijenhoek, Buizer-Voskamp et al. 2008, Glessner, Wang et al. 2009, Kirov,

Rujescu et al. 2009, Need, Ge et al. 2009, Ikeda, Aleksic et al. 2010, Shah, Tioleco et al. 2010).

Furthermore, common variants in NRXN1 have been linked to antipsychotic response in

schizophrenia patients (Souza, Meltzer et al. 2010, Lett, Tiwari et al. 2011, Jenkins, Apud et al.

2014).

1.7.2 Glutamate Decarboxylase 1 (GAD1)

The major determinant of GABA in the neocortex is glutamic acid decarboxylase-67 (GAD67;

encoded by the GAD1 gene). Convergent evidence suggests a compelling role for the GAD1

gene in cognition and dorsolateral prefrontal cortex (DLPFC) dysfunction in schizophrenia.

GAD1 codes for the glutamic acid decarboxylase (GAD67) enzyme that metabolizes glutamate to

GABA. One of the most consistent findings in schizophrenia is down-regulation of GAD1

mRNA and protein in the prefrontal cortex (Torrey, Barci et al. 2005). Furthermore, in

24

schizophrenia patients, DNA methylation profile of GAD1 in the prefrontal cortex (PFC) shows

an eight-fold increase in the promoter region leading to repressed expression (Huang and

Akbarian 2007). Genetic variation in the 5’ promoter and untranslated region of GAD1 was

associated with child-onset schizophrenia, and with an increased rate of cortical gray matter loss

over a two to-eight year period (Addington, Gornick et al. 2004). Last, optogenetics has revealed

that inhibition of fast-spiking parvalbumin (PV) interneurons results in suppression of gamma

activity (Sohal, Zhang et al. 2009), and there is a growing body of evidence suggesting abnormal

gamma-band oscillations are an endophenotype of schizophrenia related to cognition (Spencer,

Nestor et al. 2004, Lewis, Hashimoto et al. 2005, Spencer, Salisbury et al. 2008, Haenschel,

Bittner et al. 2009, Spencer 2009, Farzan, Barr et al. 2010, Farzan, Barr et al. 2010, Hall, Taylor

et al. 2011).

1.7.3 Brain-derived Neurotrophic Factor (BDNF)

Brain-derived neurotrophic factor (BDNF) is one of the key regulators of neuroplasticity,

synaptic structure, memory function and consolidation. Post-mortem studies have identified

reduced BDNF expression in the hippocampus and prefrontal cortex of schizophrenia patients

(Green, Matheson et al. 2011), and reduced BDNF levels in schizophrenia patients have been

associated with cognitive performance and clinical outcome (Chen da, Wang et al. 2009,

Vinogradov, Fisher et al. 2009). In healthy controls, it has been reported that the BDNF rs6265

SNP interacts with age to predict differences in cortical thickness, white matter FA, and episodic

memory relevant to Alzheimer’s disease (Voineskos, Lerch et al. 2011). Furthermore, there was

a significant schizophrenia diagnosis by rs6265 genotype interaction observed in resting and

working-memory related hippocampal regional cerebral blood flow, as well as on hippocampal

prefrontal coupling (Eisenberg, Ianni et al. 2013). The rs6265 SNP has also been shown to be

25

predict lower hippocampal volume, particularly within patients with schizophrenia in two

independent samples (Szeszko, Lipsky et al. 2005, Smith, Thornton et al. 2012).

1.7.4 MicroRNA 137 (MIR137)

MicroRNA-137 serves as a regulator of adult neural stem cell maturation and migration (Smrt,

Szulwach et al. 2010, Szulwach, Li et al. 2010, Sun, Ye et al. 2011) in the subventricular zones

in proximity to the lateral ventricles and the subgranular zone of the hippocampus. A single

nucleotide polymorphism (SNP), rs1625579, near the MIR137 gene (1p21.3) achieved genome-

wide significance for association with schizophrenia in approximately 50,000 subjects (p=1.6 x

10-11) (Ripke, Sanders et al. 2011), and the rs1198588 SNP (high LD with rs1625579; R2=0.79)

was further associated in the latest PCG results (Ripke, O'Dushlaine et al. 2013). Microdeletions

of the MIR137 region have also been associated with ASD and intellectual disability (Pinto,

Delaby et al. 2014). The risk allele of rs1625579 has consistently been associated with aberrant

dorsolateral prefrontal cortex (DLPFC) connectivity (Whalley, Papmeyer et al. 2012, Liu, Zhang

et al. 2014, van Erp, Guella et al. 2014). Furthermore, the variant along with severe negative

symptoms predicts an impaired neurocognitive subtype of schizophrenia(Green, Cairns et al.

2012). MIR137 has also been functionally shown to specifically regulate genes with replicated

genome-wide evidence for a role in schizophrenia, most notably CACNA1C (calcium channel,

voltage-dependent, L type, alpha 1C subunit), TCF4 (transcription factor 4) (Kwon, Wang et al.

2011) and ZNF804A (Zinc-Finger 804A) (Kim, Parker et al. 2012). The known role of

microRNAs as potent disease modifiers (Karres, Hilgers et al. 2007, Kim, Inoue et al. 2007, Lee,

Samaco et al. 2008, Williams, Valdez et al. 2009) raises the question of whether genetic

variation in the MIR137 gene might play a critical role in phenotypic expression of

schizophrenia.

26

1.7.5 L-type Voltage-dependent Calcium Channel CAv1.2 (CACNA1C)

The CACNA1C gene encodes the alpha subunit of the L-type voltage-dependent calcium channel

CAv1.2. The rs1006737 variant of CACNA1C was associated with bipolar disorder in a meta-

analysis of several large independent GWAS (p = 7.0x10-8, OR=1.18) (Ferreira, O'Donovan et al.

2008). In addition, the CACNA1C gene has recently been reported to be associated with

schizophrenia and unipolar affective disorder (Green, Grozeva et al. 2010, Nyegaard, Demontis

et al. 2010). Furthermore, recent results from the Cross-Disorder Group of the Psychiatric

Genomics Consortium show that two markers in genes involved in calcium regulation,

CACNA1C and CACNB2, reach genome-wide significance across five disorders (ASD, attention

deficit hyperactivity disorder (ADHD), bipolar disorder, major depressive disorder, and

schizophrenia) (Cross-Disorder Group of the Psychiatric Genomics, Smoller et al. 2013). At a

lower significance threshold (P<10-3), 20 of the 67 calcium active genes were associated with

these disorders, suggesting that calcium channels may have a pleiotropic effect on

psychopathology. The risk allele of the CACNA1C rs1006737 marker has been associated with

increased CACNA1C expression in the DLPFC, increased hippocampal activity during emotional

processing, and increased PFC activity during the n-back working memory task (Bigos, Mattay

et al. 2010). During reward and fear processing, healthy controls and first degree relative had

increased amygdala activation, while bipolar patients were reported to have reduced ventrolateral

PFC activation. Healthy risk variant carriers also showed reduced activation of the hippocampus

and the subgenual prefrontal cortex during an episodic memory task (Erk, Meyer-Lindenberg et

al. 2010). Most recently, it was found in healthy controls during the n-back working memory

task that the risk allele was associated with decreased activation in the DLPFC and increased

functional coupling between the DLPFC and the medial temporal lobe (Paulus, Bedenbender et

al. 2013). Furthermore, in schizophrenia patents and healthy controls the risk allele was

27

associated with poor working memory performance, whereas the association was in the opposite

direction in bipolar patients (Zhang, Shen et al. 2012). In summary, these results indicate that

CACNA1C may be a modulator of prefrontal function and cortical connectivity although the

direction of effects and their localization showed significant variance across subjects and tasks.

Considering the inconsistent direction of effect of the rs1006737 variant, more research is

necessary to understand how CACNA1C may impact working memory in different disease

populations.

1.7.6 Zinc-Finger 804A (ZNF804A)

The ZNF804A gene encodes the zinc-finger protein 804A that is expressed broadly in the brain,

especially in the developing hippocampus and cortex, as well as the adult cerebellum (Donohoe,

Morris et al. 2010). The rs1344706 variant has been implicated in schizophrenia in several

GWAS, and results became more significant when patients with bipolar disorder were included,

even though the odds ratio was still only slightly above one (p = 9.96 x 10-9; OR = 1.12)

(O'Donovan, Craddock et al. 2008). Furthermore, the variant may be particularly associated with

bipolar patients with psychotic symptoms (Lett, Zai et al. 2011). The conserved region around

the rs1344706 variant is a potential binding site for transcription factors, Myt1L zinc-finger

protein and POU3F1/Oct-6, which are involved in oligodendrocyte differentiation and

proliferation (Riley, Thiselton et al. 2010). Alternatively, the mouse homolog of ZNF804A,

zfp804a, is a target for HOXC8 (homeobox C8), suggesting that the ZNF804Agene may be

involved in early neurodevelopment; however the exact function of ZNF804A is unknown

(Chung, Lee et al. 2010). In an fMRI imaging-genetics study employing the n-back working

memory paradigm, healthy individuals (N=115) carrying the ZNF804A risk genotypes exhibited

no changes in regional activity although there was a pronounced gene dosage-dependent

28

alteration in functional connectivity (Esslinger, Walter et al. 2009). That is, risk allele carriers

had reduced connectivity between the right and left DLPFC and increased right DLPFC and left

hippocampal connectivity. Evidence for prefrontal-hippocampal connectivity as an intermediate

phenotype for schizophrenia comes from an independent replication study where it was shown

that the ZNF804A risk variant modulates connectivity in healthy controls (N=153), healthy

siblings (N=178) and patients with schizophrenia (N=78) (Rasetti, Sambataro et al. 2011).

Moreover, the impact of the ZNF804A risk variant on prefrontal-hippocampal connectivity was

specific to the working memory task, whereas right-left DLPFC connectivity was modulated by

the risk variants during working memory as well as during a face-matching paradigm and during

resting state (Esslinger, Kirsch et al. 2011). Indeed, during a theory of mind task (a measure of

social cognition) the risk variant was also associated with altered connectivity between medial

PFC and the angular gyrus (BA 39) (Walter, Schnell et al. 2011). Further, healthy controls

homozygous for the risk variant were reported to have reduced cortical thickness in the left

posterior cingulate cortex, left superior temporal gyrus and right anterior cingulate cortex, all

regions involved in attentional control and working memory (Voineskos, Lerch et al. 2011).

Together, these results support the notion that ZNF804A is conferring risk on basic brain

processes involved in proper cognitive function. ZNF804A may also impact heterogeneity within

schizophrenia. In a two-stage cognitive study examining cognitive function in schizophrenia

patients (N=297; N=165) and controls (N=165; N=1475), there was a significant gene-by-

diagnosis interaction in both episodic and working memory (Walters, Corvin et al. 2010). In both

samples, schizophrenia patients with the ZNF804A risk genotype performed worse in multiple

working memory and episodic memory tasks, although no effect was observed in healthy

controls. Moreover, this finding was stronger in patients with lower IQ. Notably, these findings

have been independently replicated in a Japanese sample (Hashimoto, Ohi et al. 2010). This

29

suggests that first, intermediate phenotypes may be sensitive to subtle effects of GWAS variants;

and second, ZNF804A may act in concert with other schizophrenia risk variants to impact

working memory function.

1.8 Multivariate Approaches to Neuroimaging

Bridging the gap between complex genetics information and whole brain neuroimaging measures

requires sophisticated multivariate analysis. Currently, within the field of psychiatric imaging-

genetics, there are many different strategies each with different assumptions, advantages, and

limitations. Three main experimental designs include: (1) complex genetics analyses (many

variants) on candidate imaging phenotypes, (2) single variant analysis of whole brain imaging

phenotypes, and (3) complex genetic analysis of whole brain imaging phenotypes. Furthermore,

there are a few strategies incorporating these imaging-genetic strategies to clinical and cognitive

phenotypes. Multivariate approaches either apply methods of data reduction (e.g. principal

component analysis (PCA)) to reduce multiple comparisons penalties for a priori hypothesis

testing, or data driven approaches to agnostically identify variables of interest (e.g., graph theory

analysis, machine learning).

1.8.1 Complex Genetic Analysis on Candidate Imaging Phenotypes

Identification of complex genetic variation influencing human brain structure may reveal

biological mechanisms underlying schizophrenia symptomology and cognitive dysfunction. One

strategy is to select a set of markers, in terms of genetic analysis, and tests these loci against

predefined neuroimaging phenotypes relevant to schizophrenia (e.g. mean hippocampal

volumes) based on a specific hypothesis. Selecting a specific brain-imaging phenotype as a

30

quantitative dependent variable and performing genome-wide analysis has been a successful

strategy that is particular useful for meta-analyses. For example, two meta-analyses of cohorts

with genetics and brain imaging have demonstrated significant associations have been observed

for hippocampal volume, intracranial volume, and total brain volume (Bis, DeCarli et al. 2012,

Stein, Medland et al. 2012). Furthermore, a gene expression and co-expression atlas has been

created that demonstrate gene transcription is highly variable across structures, but relatively

conserved across individuals (Hawrylycz, Lein et al. 2012). Therefore, providing a necessary

connection between expression and neuroanatomy, and thus, better rationale for selecting

candidate imaging phenotypes. The ENIGMA Consortium now has amalgamated advance

fMRI, structural MRI, and DTI-MRI scans on approximately 25000 subjects across healthy

controls and multiple neuropsychiatric disorders (Thompson, Stein et al. 2014). Currently,

genome-wide analyses have yielded very promising results, and in the near future more complex

multivariate analyses will be performed over multiple subgroups and imaging modalities.

Beyond genome-wide analyses, there have been many efforts to model complex genetics systems

or pathways of genes. A common approach is gene set enrichment analysis (GSEA) in which a

set of SNPs is selected based on common biological pathway (or ontology), then phenotype-

genotype associations are determine based on whether associations are enriched compared to the

null distribution (Subramanian, Tamayo et al. 2005).Thereby, pathways of interest can be

identified that may reveal be biological relevant, and potentially lead to novel treatment. Many

specific issues with this method have been addressed including biases due to gene size and LD

(Li, Gui et al. 2011). However, it should be noted that GSEA normally based on differences in

gene expression profile between psychiatric patients and healthy individuals. Gene expression

has its own drawbacks (Gunawardana and Niranjan 2013), and tissue selection may not be valid

31

for the candidate imaging phenotype. More recently, polygenic risk scores have been employed

in brain imaging based on a priori disease association.

Schizophrenia has long been theorized to be a polygenic disorder (Gottesman and Shields 1967),

and a polygenic model may influence individual differences across domains of brain

development and function; thus, polygenic risk may reflect within disease heterogeneity of

psychiatric disorders, and potentially shared genetic liability across disorders (e.g. bipolar

disorder and schizophrenia). There are many different types of strategies for deriving polygenic

scores including, but not limited to: quantitative models, linear regression, allele count, shrinkage

estimations, log-risk models, and model simulations (for a detailed review, see (Dudbridge

2013)). In imaging-genetics analysis, Whalley and colleagues employed the polygenic developed

by Purcell et al. (International Schizophrenia Consortium, Purcell et al. 2009) and found that

polygenic load was associated with increased limbic activity characteristic of bipolar disorder,

but was independent of diagnosis (Whalley, Papmeyer et al. 2012). Another polygenic risk score

selected variation associated with schizophrenia based on the meta-analyses from the

Schizophrenia Research Forum (www.schizophreniaforum.org) database, reported that

cumulative genetics risk explained 3.6% of the total variance in DLPFC activation (Walton,

Turner et al. 2013). A drawback of these polygenic models and GSEA is that they currently do

not take into account gene-gene interactions.

An alternative is multifactor dimensionality reduction (MDR) which was developed to identify

combinations of gene-gene and gene-environment interactions predicting a given phenotype.

MDR is a non-parametric machine learning strategy, similar to random forest and decision tree

models, that was specifically designed to detect interactions in the absence of marginal effects

(Moore, Asselbergs et al. 2010). To date, there are no published studies applying this data-driven

32

strategy to brain imaging, although it has been applied to quantitative phenotypes((Winham

2013)). Some drawbacks to MDR are that it can be difficult to interpret complex interactions, the

exploratory nature of MDR requires independent replication, and gene-gene interaction should

be validated in vitro. Furthermore, it has been argued the majority of genetics variance in

complex traits is due to additive effects (Hill, Goddard et al. 2008). Nevertheless, MDR remains

a potentially powerful strategy for examining epistasis that may have broad applications in

imaging-genetics.

1.8.2 Single Variant Analysis of Whole Brain Imaging Phenotypes

Perhaps the greatest advances in multivariate analyses in imaging-genetics is from whole brain

analyses of imaging-phenotypes. Two (minor carriers versus non-carries) or three genotypic

groups are created based on a single variant, then analyzed again using complex neuroimaging

phenotypes (e.g., voxel-wise FA). One of the major challenges of neuroimaging data is the high

degree correlation between brain regions even across imaging modalities. Therefore, in analysis

of candidate imaging phenotypes, it is difficult to know if a genetic association is a true effect, or

rather due to variance in the region of interest (ROI) explained by another region. Similar to

complex genetics analysis, there are two basic strategies: data reduction through incorporating

the high degree of covariance or clustering strategies to reduce the multiple comparison burden.

The latter is effectively utilized by many neuroimaging packages including FSL and SPM

(Friston, Holmes et al. 1994).

Component based analyses has been successfully used in fMRI studies, and are becoming more

common in structural neuroimaging. Principal component analysis (PCA) produces a linear set

of orthogonal principal components (latent variables) that explain the maximal variances in a set

33

of variables (e.g. a set of ROIs). Independent component analyses (ICA) extracts statistically

independent components and non-Gaussian, thus revealing hidden factors that can be a

particularly useful in noise reduction. In fMRI, ICA models provide an approximation of “true”

sources, thus within individuals it can correct for sources of systematic bias, such as cardiac

pulses in blood-oxygen-level dependent (BOLD) signals. Furthermore, several multi-subject ICA

methods have been developed that have been successfully applied to resting state fMRI data to

extract brain networks independent of a specific task (Calhoun and Adali 2012). ICA has also

been applied to DTI data and independent components have been identified that may represent

distinct white matter tracts (Wahl, Li et al. 2010), although more research is necessary to

understand the biological basis of the components. PCA on white matter tracts reveals that the

first factor explains approximately 45% of the variance that may be related to processing speed

and general fluid intelligence (Penke, Maniega et al. 2012). Thus, component based methods

may reveal biological relevant processes in terms of cognitive function. A major drawback of

these approaches is that the latent variable derived in these methods are derived from the sample.

Therefore, the results are not readily comparable among studies.

Recent advances in network-based statistical analysis and graph theory have had led to feasible

means to assess the human connectome via neuroimaging that could reveal how genetic factors

affect structural (and functional) connectivity (Hulshoff Pol and Bullmore 2013). Graph theory

permits the calculation of summary measures connectivity that underlie normal organization

processes that may break down in neuropsychiatric disease (Fornito, Zalesky et al. 2013).

Therefore, it is an effective, data-driven means to identify neural circuitry that may be dependent

on genetic factors. Indeed, small-work networks identified via resting state EEG have revealed

that local and global interconnectedness are moderately heritable (ranging from 32-89%) (Smit,

Stam et al. 2008), and similar heritability has been observed in global interconnectivity in

34

children via resting state fMRI (van den Heuvel, van Soelen et al. 2013). Furthermore,

genetically mediated relationships following small world architecture have been observed in

cortical thickness (Schmitt, Lenroot et al. 2008). Results from graph theoretical approaches can

be used to identify crucial network topography liable in neuropsychiatric disorder and

heterogeneity of this network can be examined in following genetic analysis. For example, DTI

analysis comparing complex networks of schizophrenia and healthy subjects revealed preserved

small world organization, yet longer path length in frontal and temporal regions. These results

suggest that schizophrenia patients have poor global integration in these regions resulting in a

limited capacity to integrate information across brain regions (van den Heuvel, Mandl et al.

2010). The regions with group differences in node global path length could then be used as seed

points for further DTI analysis examining the effect of a given genetic risk variant. An advantage

to this approach would be the identified DTI tract could be also applied to cognitive and clinical

phenotypes. An alternative would be to examine differences in global and regional connectivity

based on genotype. For instance, the autism risk gene CNTNAP2 (contactin associated protein-

like 2) was examined in 328 healthy individuals using high-angular-resolution diffusion imaging

(HARDI) revealed altered path length, small worldness, and global efficiency in risk carriers.

This approach provides detailed information on brain interconnectivity (Dennis, Jahanshad et al.

2011); however, it is difficult to assess what these measure represent in the phenotypic

expression of complex neuropsychiatric disorders.

1.8.3 Complex Genetic Analysis of Whole Brain Imaging Phenotypes

One of the ultimate goals of imaging-genetics is to analyze the whole variation of the genome

(greater than three billion base pairs) and voxel-wise imaging approaches (upwards of 140000

voxels). The advantages would include novel, data-drive findings as well as greater confidence

35

that the results are not confounded by regions that are not analyzed (either genetic or

anatomical). The complexity of the data, rather than the quality, poses significant challenges. It is

conceivable whole brain analysis of whole genome sequencing data would need to correct for

approximately 5 x 1014 multiple comparisons. Therefore, it is necessary to intelligently reduce

the complexity both genetic and imaging data. The approaches mentioned in the previous

sections can be combined to increase power. For instance, polygenic risk scores could be

regressed against clustering or component analyses of brain imaging. It should be noted that to

date no imaging-genetics studies have performed this type analysis. Alternatively, multivariate

methods have been developed to analyze this complex and confounded data.

There are at least three data driven approaches that could be utilized. The first is mass-univariate

linear modeling (MULM) in which all linear model combinations are fit. This approach has been

successful applied in imaging-genetic analysis (Stein, Hua et al. 2010). Beyond the need for

stringent correction for experiment-wise error, a drawback of MULM is that it independently

tests phenotypes and genotypes. Therefore, it is unable to capture cumulative effects from

multiple markers, and it does capitalize on power gains from confounded quantitative

phenotypes. The second approach is similar to component analyses in which latent variables are

extracted simultaneously from both imaging and genetics data to produce new genotype-

phenotype variables that are optimized using cost functions (e.g., partial least squares (PLS),

canonical component analysis (CCA), and reduced rank regression (RRR)). PLS maximizes the

covariance between latent variables, whereas CCA maximizes the correlation between them.

When the number of variables is significantly larger than the sample size (such as in imaging-

genetics), these methods become effectively equal (Le Floch, Guillemot et al. 2012). In RRR,

response and predictor variables are reduced and ranked according to latent variables and linear

regressions are performed. Given the data reduction strategies of these approaches, they have

36

more power with limited samples size, and they are especially effective for analyzing

confounded data; although, they may over-fit the data. PLS has successfully used to identify

genotype-phenotype differences between schizophrenia patients and controls through examining

genetic variation in the dopamine receptor D1 gene and DLPFC activation during the Serial Item

Recognition fMRI paradigm (Tura, Turner et al. 2008). Furthermore, PLS analysis has used to

parcellate genotype-phenotype associations in myelin associated genes and white matter FA that

predicted cognitive performance impaired in schizophrenia (Voineskos, Felsky et al. 2013). Last,

sparse RRR has been employed for genome-wide detection of markers associated with voxel-

wise longitudinal changes in the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database

(Vounou, Janousova et al. 2012).

The last approach is to employ data-mining strategies such as machine learning. Machine

learning is an iterative artificial system through which efficiency and effectiveness is improved

over time. There are a wide variety of techniques (e.g., support vector machines (SVM),

Gaussian random field (GRF), graphical models, autoregression, and others) (Nilsson 1996).

Major advantages of theses technique is that it is truly agnostic (does not require hypothesis

testing including what phenotype to examine), and the quality of the results increases with more

information. There have been a few studies applying machine learning algorithms (MLA) to

imaging-genetics studies. In fMRI analysis, MLA can be used to train classifiers to decode

stimuli, behaviors and other variables ((Pereira, Mitchell et al. 2009). For instance, components

obtained from ICA of resting state fMRI can classify patients with schizophrenia and healthy

controls (Shinkareva, Ombao et al. 2006). Furthermore, SVM had been used to distinguish

bipolar and schizophrenia patients from healthy controls using gene expression data as well as

demographic and clinical data (Struyf, Dobrin et al. 2008). Indeed, schizophrenia patients and

healthy controls have been successfully classified using structural MRI ((Davatzikos, Ruparel et

37

al. 2005), fMRI (Costafreda, Fu et al. 2011), and DTI (Ingalhalikar, Kanterakis et al. 2010).

More recently, SVM MLA has been used on voxel-wise ICA of fMRI resting state data and

genotypes from 384 SNPs, and imaging-genetic data provided better classification between

schizophrenia and healthy controls than either method alone (Yang, Liu et al. 2010). However, in

all of the mentioned studies the comparison group were healthy controls and given the overlap of

schizophrenia (clinical, neuroanatomical, genetic) it is likely that the high degree of sensitivity

and specificity would not remain comparing neuropsychiatic groups (e.g. schizophrenia versus

bipolar patients). MLAs can also be used to increase the power of voxel-wise genome-wide

association studies. A novel method of combining GRF for imaging data to SNP data is via least

square kernel machine. This allows for the joint effect of SNPs on imaging traits and assessment

of epistasis among SNPs (Ge, Feng et al. 2012). The approach was applied on the ADNI

database, and the top associations overlapped with Alzheimer’s candidate genes including

GRIN2B, although it did not survive multiple comparison testing (Ge, Feng et al. 2012). The

same approach was applied to data from the IMAGEN Consortium. The top variant in the

neuroplastin gene was associated with frontal and temporal lop thinning leading to verbal and

non-verbal intellectual disabilities in adolescents (Desrivieres, Lourdusamy et al. 2014).

Moreover, epistatic interactions were examined in over 14 candidate genes (810 SNPs) using

stochastic gradient boosting (SGB) (Andreasen, Wilcox et al. 2012). SGB identified SNPs that

had the strongest relationship with multiple measures of brain changes. Novel epistatic

interactions were discovered as well as five of the 17 SNPs identified by GSB had previously

implicated in cognitive processes relevant to schizophrenia, suggesting that this machine learning

algorithm may uncover meaningful epistasic relationships between genes (Andreasen, Wilcox et

al. 2012). Taken together, both approaches suggesting that MLA is a viable means to discovery

38

novel associations and a potentially powerful means to circumvent the large penalties studying

how complex genetic variation impacts voxel-wise imaging data.

There are caveats that should be considered using MLA. First, similar to component based

approaches, it is uncertain if the results from be over-fitted models. Second, there is no

consensus on which MLA method should be employed in imaging-genetic analysis. Therefore, it

is difficult to interpret the results among studies. Third, MLA approaches are by definition

exploratory, and confidence in the results requires replication or converging evidence. Indeed,

the majority of diagnostic studies using MLA tend to have a high degree of sensitivity and

specificity (Orru, Pettersson-Yeo et al. 2012). Given the small sample size of these studies it is

uncertain if the effect is rather due to population structure.

1.9 Application of polygenic risk models to imaging genetic studies in psychiatry

Lately, there has been an increasing trend towards applying polygenic risk models in imaging-

genetics. Earlier fMRI studies have divided their sample according to risk variant groups to

assess ‘epistasis’. Tan et al. reported that GRM3 risk allele homozygotes had greater DLPFC

activation in COMT rs4680 Val homozygotes but not Met homozygotes (Tan, Chen et al. 2007).

Similar three-way ‘interactions’ were observed in the NRG1, ERBB4, and AKT1 gene variants

(Nicodemus, Callicott et al. 2010). Furthermore, polygene influences on DLPFC activation have

been observed with DISC1 and NDEL1 (Nicodemus, Callicott et al. 2010), MTHFR and COMT

(Roffman, Weiss et al. 2008), and RGS4 and COMT (Buckholtz, Sust et al. 2007). Polygenic risk

scores have only recently been applied to functional and structural neuroimaging studies. The

most popular method involves selecting a group of SNPs at different p-value thresholds based on

large case-control disease association studies. Next, a set of risk scores (at different significance

39

thresholds) is created in which each risk variant is weighted by their odds ratio and accrued as a

polygenic score. Then, the polygenic scores are tested against brain phenotypes. Polygenic risk

for schizophrenia was also association with total brain volume and white matter across more than

14,000 SNPs among both healthy controls and schizophrenia patients (Terwisscha van

Scheltinga, Bakker et al. 2013). Walton et al. found a polygenic risk score, based on 600 SNPs

nominally associated with schizophrenia, predicted left DLPFC inefficiency in healthy controls

(Walton, Turner et al. 2013). Whalley and colleagues subsequently reported polygenic risk

scores derived from the PGC Bipolar Working Group (PGC-BD) predicting increased limbic

activation among both healthy controls and individuals with familial risk of bipolar disorder;

however, there was no significant association with limbic activation within each group (Whalley,

Papmeyer et al. 2012). In a subsequent study, polygenic score for major depressive disorder, but

not bipolar disorder, was associated with decreased white matter FA across healthy controls and

individuals with familial risk for mood disorders (Whalley, Sprooten et al. 2013). Furthermore,

in a relatively large healthy control sample (N=438), polygenic risk was associated with reduced

cortical thickness in the left medial prefrontal cortex (Holmes, Lee et al. 2012).

In general, these studies have provided important first steps to understanding polygenic risk on

brain function and structure, although there are some important caveats. First, the p-value

thresholds were arbitrarily selected; thus, there are potentially spurious associations included in

the polygenic scores. Second, to date, none of the studies included independent replication of

their imaging findings. Given the exploratory nature of these analyses, replication would greatly

bolster confidence in their results. Third, the scores were weighted according to disease

associations. It could be argued that disease risk may not be valid when assessing disease

heterogeneity. For instance, the effect on any particular brain region of a risk variant could be

different within schizophrenia than in healthy controls. Last, to assess the clinical utility of

40

polygenic risk score, it may be necessary to assess if the polygenic association with brain

phenotypes are mediating clinical and behavioral heterogeneity. Considering the wide variety of

methods to create polygenic scores (Dudbridge 2013), it may necessary to tailor the polygenic

methodology based to the specific hypothesis of the imaging-genetic study.

1.10 Outline of Experiments

The succeeding chapter will provide a brief background and hypothesis for the four manuscripts

in the thesis (Chapter 3-6). Chapters three and four have been published in the journals PLoS

One and Molecular Psychiatry, respectively. Chapter five and six are about to be submitted to

peer reviewed journals. Each of the studies are standalone articles; therefore, sections contained

in each studies may overlap with each other as well as with material presented in the

Introduction. Additionally, the appendix contains two manuscripts published in the journals

Schizophrenia Research and Biological Psychiatry. These articles complement the work

presented, but they are beyond the scope the specific hypotheses outlined in this thesis.

41

Chapter 2

2 Overview of Experiments, and Hypothesis

This thesis is composed of four independent projects that sought to examine heterogeneity

relevant to or within schizophrenia through combining neuroimaging and genetics, with

particular focus on how these impact variability in clinical and cognitive functioning.

2.1 Neurexin-1 and Frontal Lobe White Matter: An Overlapping Intermediate Phenotype for Schizophrenia and Autism Spectrum Disorders

2.1.1 Background

Structural variation in the neurexin-1 (NRXN1) gene increases risk for both autism spectrum

disorders (ASD) and schizophrenia. However, the manner in which NRXN1 gene variation may

be related to brain morphology to confer risk for ASD or schizophrenia is unknown. This study

examines the NRXN1 gene on brain structure and cognitive function, thereby attempting to

identify a neural and cognitive susceptibility mechanism by which the NRXN1 gene confers risk

for both schizophrenia and ASD.

2.1.2 Hypothesis

The intermediate phenotype approach permits us to examine how shared genetic underpinnings

of these two disorders may confer risk in the brain. Therefore, we used this approach to

investigate 11 single nucleotide polymorphisms (SNPs) of the NRXN1 gene lying within regions

overlapped by numerous deletions implicated in ASD and schizophrenia, and their effects on

brain morphometry in healthy individuals. Given that such deletions confer susceptibility to both

42

schizophrenia and ASD, we hypothesized that NRXN1 polymorphisms would confer an

intermediate phenotype related to schizophrenia and ASD, via effects on neural structures and

cognitive function altered in both disorders.

2.2 The Genome-Wide Supported MicroRNA-137 Variant Predicts Phenotypic Heterogeneity within Schizophrenia

2.2.1 Background

There is notable heterogeneity in the phenotypic presentation of schizophrenia including, but not

limited to, the onset of illness, severity of positive and negative symptoms, neurological soft

signs and cognition, course of illness, response to treatment, and functional and structural brain

abnormalities. This phenotypic heterogeneity has been a central challenge for schizophrenia

research and other neuropsychiatric disorders. MicroRNA regulate genetic expression and

translation over networks of gene, and thus, they are potent disease modifiers. A single

nucleotide polymorphism, rs1625579, near the MIR137 gene (microRNA 137; 1p21.3) is a top

genome-wide significance for association with schizophrenia, and MIR137 has also been

functionally shown to specifically regulate genes with replicated genome-wide significant

evidence for a role in schizophrenia, most notably CACNA1C and TCF4.

2.2.2 Hypothesis

The identification of the genetic sources of phenotype heterogeneity, such as the effects of a

genetic risk variant on phenotypes such as age-at-onset, or brain structure, may lead to early

identification of disease trajectory. Such identification, before disease progression, could then

serve as a platform to test earlier interventions, particularly within the subgroup at-risk for poorer

outcome. Given the recently established role of MIR137 as a central player in coordinating the

43

timing and expression of schizophrenia risk genes we hypothesized that MIR137 may be an

important determinant of age-at-onset of psychosis and brain structure in schizophrenia.

2.3 Glutamate Decarboxylase 1 (GAD1) Variant Predicts a Neuroanatomical and Working Memory Susceptibly Mechanism Relevant to Schizophrenia.

2.4 Background

Working memory dysfunction is a central feature of schizophrenia and many other psychiatric

disorders. In schizophrenia patients, working memory deficits are associated with dysfunction of

dorsolateral prefrontal cortex (DLPFC) as well as DLPFC connectivity with other regions and

disruption of neurotransmitter input such as GABA inhibitory neurotransmission. Convergent

evidence suggests a compelling role for the glutamate decarboxylase 1 (GAD1) gene in working

memory and dorsolateral prefrontal cortex (DLPFC) dysfunction in schizophrenia.

2.5 Hypothesis

We developed a novel method of voxel-wise mediation analysis that was used to examine the

relationship of GAD1 genetic variation, brain structure, and working memory performance. We

first hypothesized that the GAD1 rs3749034 risk variant would predict brain structure changes

relevant to schizophrenia and working memory function. Next, we examined the relationship

between the risk variant and working memory performance across multiple tasks relevant to

working memory (letter-number span, digit-span, Stroop). Last, we hypothesized that the effect

of the risk variant on brain structure would mediate its effect on working memory performance.

44

2.6 Additive Genetics Risk Predicts Widespread Changes in Brain Structure that Cause Poorer Cognitive Function

2.6.1 Background

There is growing theoretical and empirical evidence that additive genetic variation accounts for a

considerable proportion of the variance in complex traits. Therefore, examination of additive

genetic risk across several common variants might provide a better explanation for the high

degree of heterogeneity in neurocognitive dysfunction in schizophrenia that depends on brain

network connectivity. There is evidence to suggest that neuroanatomical changes and

neurocognitive dysfunction within schizophrenia are likely dependent on genetic load. Further,

these anatomical changes may mediate neurocognitive dysfunction.

2.6.2 Hypothesis

We hypothesized that increasing additive genetic risk loading may produce a more ‘severe’ brain

phenotype that may predict cognitive function. Furthermore, as we have previously shown, the

effect of schizophrenia risk variants on brain structure may be greater within schizophrenia

patients compared to healthy controls. Therefore, we examined the accrued effect of five

common genetic variants, implicated in schizophrenia, brain structure and cognitive function, for

association with brain-wide measures of white matter fraction anisotropy (FA) and cortical

thickness in healthy controls and patients. To compare genetic subsets with differences in brain

structure, we then isolated subjects with either low or high risk allele loading for association with

our neurocognitive battery. Last, we employ a novel voxel-wise mediation analysis to understand

how high risk allele loading explains poorer cognitive functioning via worse brain structure.

45

Chapter 3

3 Neurexin-1 and Frontal Lobe White Matter: An Overlapping Intermediate Phenotype for Schizophrenia and Autism Spectrum Disorders

Contents of this chapter have been published as:

Voineskos AN, Lett TA et al. Neurexin-1 and frontal lobe white matter: an overlapping

intermediate phenotype for schizophrenia and autism spectrum disorders. PLoS One.

2011;6(6):e20982.

A link to the published paper can be found at:

http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0020982#pone-

0020982-g003

This work is open access under the Creative Commons Attribution (CC BY) license

46

3.1 Abstract

Background: Structural variation in the neurexin-1 (NRXN1) gene increases risk for both autism

spectrum disorders (ASD) and schizophrenia. However, the manner in which NRXN1 gene

variation may be related to brain morphology to confer risk for ASD or schizophrenia is

unknown.

Method/ Principal Findings: 53 healthy individuals between 18-59 years of age were

genotyped at 11 single nucleotide polymorphisms of the NRXN1 gene. All subjects received

structural MRI scans, which were processed to determine cortical gray and white matter lobar

volumes, and volumes of striatal and thalamic structures. Each subject’s sensorimotor function

was also assessed. The general linear model was used to calculate the influence of genetic

variation on neural and cognitive phenotypes. Finally, in silico analysis was conducted to assess

potential functional relevance of any polymorphisms associated with brain measures. A

polymorphism located in the 3’ untranslated region of NRXN1 significantly influenced white

matter volumes in whole brain and frontal lobes after correcting for total brain volume, age and

multiple comparisons. Follow-up in silico analysis revealed that this SNP is a putative

microRNA binding site that may be of functional significance in regulating NRXN1 expression.

This variant also influenced sensorimotor performance, a neurocognitive function impaired in

both ASD and schizophrenia.

Conclusions: Our findings demonstrate that the NRXN1 gene, a vulnerability gene for SCZ and

ASD, influences brain structure and cognitive function susceptible in both disorders. In

conjunction with our in silico results, our findings provide evidence for a neural and cognitive

susceptibility mechanism by which the NRXN1 gene confers risk for both schizophrenia and

ASD.

47

3.2 Introduction

Autism Spectrum Disorders (ASDs) and schizophrenia are highly heritable disorders with

genetic factors comprising the majority of the known risk (Carroll and Owen 2009). Currently,

the gene with the best evidence for shared susceptibility for schizophrenia and ASD is the

Neurexin-1 (NRXN1) gene, one of the largest known human genes (1.1 Mb) with 24 exons,

located on chromosome 2p16.3 (Südhof 2008). The NRXN1 gene encodes the neurexin-1α and

neurexin-1β proteins that function as pre-synaptic neural adhesion molecules. Neurexin-1α is

reported to interact with postsynaptic neuroligins (NLGNs) mediating GABAergic and

glutamatergic synapse function (Südhof 2008). It also has been reported to bind to leucine-rich

repeat transmembrane protein (LRRTM2) (de Wit, Sylwestrak et al. 2009), instructing

presynaptic and mediating postsynaptic differentiation of glutamatergic synapses. Substantial

evidence implicates deletions in the NRXN1 gene in ASD (Feng, Schroer et al. 2006, Szatmari,

Paterson et al. 2007, Kim, Kishikawa et al. 2008, Marshall, Noor et al. 2008, Morrow, Yoo et al.

2008, Yan, Noltner et al. 2008, Glessner and Hakonarson 2009) and schizophrenia (Vrijenhoek,

Buizer-Voskamp et al. 2008, Glessner, Wang et al. 2009, Kirov, Rujescu et al. 2009, Need, Ge et

al. 2009, Rujescu, Ingason et al. 2009, Ikeda, Aleksic et al. 2010, Shah, Tioleco et al. 2010).

NRXN1 has also been associated with mental retardation (Zweier, de Jong et al. 2009, Ching,

Shen et al. 2010), nicotine dependence (Bierut, Madden et al. 2007, Nussbaum, Xu et al. 2008,

Novak, Boukhadra et al. 2009), alcoholism (Yang, Chang et al. 2005) and vertebral anomalies

(Zahir, Baross et al. 2008). Therefore, it is apparent that disruptions of the NRXN1 gene,

especially deletions, confer risk to a range of neurodevelopmental phenotypes, including ASDs,

schizophrenia, and mental retardation.

48

The results of neuroimaging studies suggest that schizophrenia and ASD patients also share

neural vulnerability, most notably in the frontal lobe and in frontal lobe circuitry (Minshew and

Keller 2010, Pettersson-Yeo, Allen et al. 2011). Therefore, genes that confer susceptibility to

both schizophrenia and ASD might contribute to altered brain structure and/or function common

to both disorders. Although few studies have included both ASD and schizophrenia patients,

overlapping findings between these illnesses occur most prominently in the frontal lobe and in

fronto-striatal circuitry (Minshew and Keller 2010, Pettersson-Yeo, Allen et al. 2011). Grey and

white matter in ASD has been associated with increased cortical grey to white matter ratio and

decreased volumes beyond childhood (Courchesne, Karns et al. 2001, Acosta and Pearl 2004).

Although both increases and decreases in grey and white matter volumes in ASD have been

reported, white matter abnormalities in the frontal lobe remain some of the most consistent

neuroimaging findings in ASD (McAlonan, Daly et al. 2002, Herbert, Ziegler et al. 2003,

Barnea-Goraly, Kwon et al. 2004, Herbert, Ziegler et al. 2004, McAlonan, Cheung et al. 2005,

Sundaram, Kumar et al. 2008, McAlonan, Cheung et al. 2009, Mengotti, D'Agostini et al. 2010).

Thus, developmental abnormalities in white matter growth seems important in the

etioneuropathology of ASD (Williams and Minshew 2007). Structural MRI findings in

schizophrenia populations are typically characterized by decreases in temporal and frontal lobe

volumes, and some reductions in total brain volume and parietal volumes (McCarley, Wible et

al. 1999, Shenton, Dickey et al. 2001). Although findings have not always been consistent, a

recent meta-analysis of 17 studies confirmed a frontal lobe white matter deficit in patients with

schizophrenia (Di, Chan et al. 2009). Furthermore, cytoarchitectural alterations of the prefrontal

cortex have been found in schizophrenia, and decreased thalamic volume and altered prefrontal-

thalamic circuitry are common findings in this disorder (Goldman-Rakic and Selemon 1997,

Jones 1997, Danos, Baumann et al. 2003, Brickman, Buchsbaum et al. 2004, James, James et al.

49

2004, McIntosh, Job et al. 2004, Rose, Chalk et al. 2006). Altogether, these findings suggest

abnormalities of frontal, thalamic, and striatal structure that may be shared in the neuropathology

of schizophrenia and ASD. Neurocognitively, sensorimotor deficits are shared by both disorders.

Such deficits are typically apparent in ASD patients (Sigman and Ungerer 1981). Cognitive

assessment (Rajji and Mulsant 2008) and birth cohort studies (Welham, Isohanni et al. 2009) also

identify impaired sensorimotor function in schizophrenia.

The intermediate phenotype approach permits us to examine how shared genetic underpinnings

of these two disorders may confer risk in the brain (Gottesman and Gould 2003, Meyer-

Lindenberg and Weinberger 2006, Tan, Callicott et al. 2008). Therefore, we used this approach

to investigate 11 single nucleotide polymorphisms (SNPs) of the NRXN1 gene lying within

regions overlapped by numerous deletions implicated in ASD and schizophrenia, and their

effects on brain morphometry in healthy individuals. Given that such deletions confer

susceptibility to both schizophrenia and ASD, we hypothesized that NRXN1 polymorphisms

would confer an intermediate phenotype related to schizophrenia and ASD, via effects on neural

structures and cognitive function altered in both disorders.

3.3 Results

3.3.1 Genotypes

Concordance for the 10% of re-genotyping of all 11 SNPs (Figure 3-1) was 100%. No SNP

deviated significantly from Hardy-Weinberg equilibrium. Four SNPs (rs10208208, rs12623467,

rs10490162, 10490227) were not included in further analysis since their minor allele frequency

(MAF) was below 15% (Table 3-S1). Furthermore, none of the SNPs was in linkage

50

disequilibrium (LD) (not shown), and their MAF was similar to the Hapmap CEU population

(Thorisson, Smith et al. 2005). For rs1045881 since only one TT homozygote was in the sample,

we combined T-allele carriers (T/T and T/C) and collectively analyzed in one cell. Post hoc

independent t-tests of rs1045881 genotype (T-Carriers vs. C/C) revealed no significant

differences in any demographics measured (Table 3-S2).

For lobar gray matter volumes, no genotype by brain region interactions or main effects of

genotype were found following repeated measure ANCOVAs conducted for each of the seven

SNPs with MAF > 15%, with age and total brain volume as covariates. Therefore, no follow-up

analysis was performed. When examining white matter volumes, we found that for each lobe, a

minimum of 85% of the variance in one hemisphere was explained by the white matter volume

of the other hemisphere (P<0.001, R2 (Pearson)>0.85); therefore, we combined lobar white matter

volumes across hemispheres. For lobar white matter volumes, a genotype by white matter lobe

volume interaction was found following repeated measures ANCOVA, at the rs1045881

(F2.25=5.498, p = 0.004) and rs858932 (F4.56= 3.802 , p=0.004) polymorphisms (Bonferroni

corrected alpha of 0.0071). We did not observe significant white matter region volume by

genotype interactions in any other NRXN1 variants examined. The results for the rs1045881 and

rs858932 SNPs were followed up using separate ANCOVAs for white matter volume at each

lobe. The rs1045881 variant was significantly associated with frontal lobe white matter volume

(Bonferroni corrected alpha = 0.0125 for four brain regions): F1,49=8.231, p=0.006; (Figure 3-2),

where ‘CC’ homozygotes demonstrated reduced frontal white matter volumes compared to ‘T’

allele carriers. Consistent with the direction of effect in frontal lobe, the rs1045881 was

nominally associated (as it did not survive Bonferroni correction) with change in parietal lobe

51

white matter volume (F1,49=4.089, p = 0.049). No association of this SNP with temporal or

occipital lobe white matter volume was observed.

The follow-up ANCOVA examining rs858932 genotype also predicted frontal lobe white matter

volume (F2,51=5.472, p=0.007), where ‘GG’ individuals had lower frontal lobe white matter

volume and nominal association in the parietal lobe also occurred in the same direction, but did

not survive Bonferroni correction (F48,2=3.719, p = 0.032; Figure 3-S1). Frontal lobe white

matter volumes were also associated at the allelic level: both the ‘C’ allele of rs1045881

(χ2=7.184, p=0.0074) and the ‘G’ allele of rs858932 (χ2=4.121, p=0.0432) predicted lower

frontal white matter volume (Table 3-S3). Similar results were shown in the haplotype analysis

(p(Global)<0.001; Table 3-S4).

Repeated measures analysis for striatal and thalamic structures revealed a significant volume by

region interaction for the rs858932 SNP only (F14,336 = 3.4, p < 0.001; Greenhouse-Geiser

correction: F4,99 = 3.4, p = 0.01). Follow-up ANCOVAs at left and right caudate, putamen,

globus pallidus, and thalamus revealed that this interaction was driven by the influence of the

rs858932 SNP on thalamic volumes only: for left thalamus (F2,48 = 8.9, p = 0.001), and for right

thalamus (F2,48 = 7.3, p = 0.002), significant at the Bonferroni corrected alpha for eight

comparisons (alpha= 0.0063, Figure 3-3). Here, ‘GG’ individuals had significantly lower

thalamic volumes compared to ‘T’ allele carriers. No significant effects were observed at

caudate, putamen, or globus pallidus.

52

3.3.2 Cognitive

Repeated measures ANCOVA showed a main effect of the rs1045881 SNP on sensorimotor

function (F1,49 = 4.8, p = 0.03). The ‘C/C’ homozygotes had reduced finger tapping scores

compared to ‘T’ allele carriers, consistent with the directional effect on white matter volumes.

No association was observed for the rs858932 SNP (F1,48 = 0.4, p = 0.67). No task by genotype

interaction was observed for either polymorphism.

Frontal lobe white matter volume was highly correlated with finger tapping (FT) score even after

accounting for age effects (Dominant Hand: R2 = 0.404, p = 0.003; Non-Dominant Hand: R2 =

0.469, p = 0.001).

3.3.3 In silico Analysis

The rs1045881 SNP is located in the 3’UTR of Neurexin-1. In silico prediction by miRBase

analysis revealed the presence of the C-allele creates a binding site for the microRNA hsa-miR-

1274a and hsa-miR-339-5p. Furthermore, alteration in exon splicing enhancer and other motifs

were observed. The rs858932 SNP was not sufficiently near any splice site (i.e. intron/exon

border) for in silico prediction.

3.4 Discussion

We found that genetic variation in the 3’untranslated region of the NRXN1 gene predicted an

intermediate risk phenotype in healthy individuals relevant to schizophrenia and ASD. Our

primary finding at the rs1045881 SNP in the 3’UTR of Neurexin1 demonstrated that the ‘C’

allele predicts reduced frontal white matter volume and sensorimotor function. Furthermore, our

53

in silico analysis suggested presence of the same ‘C’ allele predicted microRNA binding, thus

providing a potential mechanism for this allele’s effects on brain structure and cognitive

function. The gene variants that influenced brain morphology in our study are located in the

regions of NRXN1 susceptible to deletion in schizophrenia and ASD. The effects of these genetic

variants localized to brain structure and cognitive function that demonstrate overlapping

susceptibility in both schizophrenia and ASD, namely frontal lobe white matter abnormalities, as

shown in recent meta-analyses (Di, Chan et al. 2009, Radua, Via et al. 2010) and sensorimotor

function (Curcio 1978, Braff and Geyer 1990, Geyer, Swerdlow et al. 1990, Peng, Mansbach et

al. 1990). To our knowledge, this work provides the first evidence in vivo of how variation in the

NRXN1 gene may confer a potential neural risk mechanism for schizophrenia and ASD.

Schizophrenia and ASD patients share sensorimotor deficits and soft neurological signs

(Dumontheil, Burgess et al. 2008). Such shared deficits are almost certainly neurodevelopmental

in nature, as in ASD they present at a very early age, and when present in schizophrenia, they are

often present before illness onset. White matter, likely through myelination, plays a key role in

ensuring appropriate sensorimotor development, and motor tasks and motor speed are tightly

correlated with white matter indices on MRI (Barnea-Goraly, Menon et al. 2005, Takarae,

Minshew et al. 2007). Our finding correlating white matter volumes with sensorimotor

performance is consistent with previous investigations (Herbert, Ziegler et al. 2004, Douaud,

Smith et al. 2007, Shukla, Keehn et al. 2010). Moreover, the same NRXN1 allele that predicted

microRNA binding (and thus presumably increased enzymatic breakdown of NRXN1 mRNA and

reduced NRXN1 translation) also correlates with reduced white matter volumes and altered

sensorimotor function.

54

Our second finding was that the intronic rs858932 SNP, also located in a deletion site

(Vrijenhoek, Buizer-Voskamp et al. 2008, Rujescu, Ingason et al. 2009, Ching, Shen et al. 2010),

similarly influenced frontal lobe white matter volume, but also prominently influenced left and

right thalamic volumes. We consider this finding more preliminary due to the lower minor allele

frequency at this variant in our sample. Nevertheless, association of this variant with thalamic

volumes is consistent with overlapping neural vulnerability for ASD and schizophrenia as well

(Shenton, Dickey et al. 2001, Brambilla, Hardan et al. 2003), and suggests that the NRXN1 gene

may influence thalamocortical circuitry that is vulnerable in both disorders.

Little is known about how specific types of deletions within the NRXN1 gene may relate to a

given neuropsychiatric phenotype. Our in silico analysis demonstrated the 3’UTR SNP as a

putative microRNA binding site for hsa-miR-339 and hsa-miR-1274, thus suggesting a

functional role for this region of the gene that may relate to mRNA expression of NRXN1. This is

interesting since expression of miR-339 microRNA has been reported to be dysregulated in the

cortex of psychotic patients (Moreau, Bruse et al. 2011). Reduced NRXN1 mRNA may influence

white matter alterations by concomitant reductions in binding to the NRXN1 binding partner,

LRRTM2, which mediates postsynaptic differentiation of glutamatergic synapses (de Wit,

Sylwestrak et al. 2009, Ko, Fuccillo et al. 2009, Siddiqui, Pancaroglu et al. 2010). Glutamatergic

dysfunction is well established in schizophrenia (Coyle 1996); further, NXRN1 expression is

induced by AMPA receptors, and mediates recruitment of NMDA receptors, a hallmark of

synapse maturation(Thyagarajan and Ting 2010). Glutamatergic dysfunction can also lead to

white matter abnormalities. Oligodendrocytes possess glutamatergic receptors (both AMPA and

NMDA), and are highly sensitive to any form of stress or toxicity (McTigue and Tripathi 2008).

Therefore, NRXN1 may influence frontal white matter in schizophrenia and ASD through

55

disrupted interaction with its glutamatergically-related binding partners, or possibly via direct

glutamatergic involvement as the NRXN1 knock out mouse demonstrates decreased excitatory

synaptic strength and decreased prepulse inhibition (Etherton, Blaiss et al. 2009).

Recent imaging-genetics studies (Scott-Van Zeeland, Abrahams et al. 2010, Tan, Doke et al.

2010) have implicated a neurexin superfamily member, the contactin-associated protein-like 2

(CNTNAP2) gene in brain structure and function providing evidence for neural susceptibility

patterns relevant to ASD. These studies demonstrated volumetric reductions for CNTNAP2 risk

allele carriers particularly in frontal lobe (Scott-Van Zeeland, Abrahams et al. 2010, Tan, Doke

et al. 2010) and also showed altered frontal connectivity. One of these two studies (Scott-Van

Zeeland, Abrahams et al. 2010) demonstrated strong effects with sample sizes smaller than ours.

Our findings, in conjunction with the recent imaging-genetics findings of CNTNAP2 demonstrate

the value of examining common variants within known ASD risk genes to understand neural

susceptibility mechanisms conferred by these risk genes. The ‘added-value’ of this approach lies

in the neural localization of gene effects, providing information regarding how the genes may

confer brain risk patterns for these disorders.

There are several limitations in our study that should be considered. First, we imposed a

dominant model by combining genotypic groups C/T and T/T at rs1045881; however concern

regarding this model can be mitigated by our findings that allelic association analysis supported

such a model. Second, one could argue that our finding may constitute a ‘winner’s curse’, and

therefore we would encourage replication efforts. A third limitation of our study is that while

there was a clear effect of this putative risk variant on frontal lobe white matter volume, in a

direction consistent with cognitive function findings and in silico prediction, various MRI studies

56

have reported either reductions or increases in frontal lobe white matter for both populations.

Finally, given that we measured gray and white matter volumes for cortical lobar structures, we

were somewhat limited in obtaining more localized regional specificity for effects of NRXN1

variation. More detailed parcellation, white matter voxel-based morphometry, or other white

matter imaging techniques such as diffusion tensor imaging, magnetization transfer imaging, or

T2 techniques should help clarify further the manner in which NRXN1 influences frontal white

matter.

In summary, we found that variants within the NRXN1 gene influence brain morphometry with a

susceptibility pattern relevant to both schizophrenia and ASD. This finding is consistent with the

fact that NRXN1 is a vulnerability gene for both disorders. In addition to reporting that the

rs1045881 gene variant is associated with frontal white matter volume and sensorimotor

performance, we provide a putative mechanistic explanation for its effects in the brain. Taken

together, our findings provide evidence that genetic variation in NRXN1, a risk gene for

schizophrenia and ASD, may confer neural and cognitive susceptibility common to both

disorders.

3.5 Materials and Methods

3.5.1 Participants

Fifty-three healthy volunteers (15 women, 38 men) (Table 3-1) met the following eligibility

criteria: age between 18 and 59; right handedness; absence of any history of a mental disorder,

current substance abuse or a history of substance dependence, positive urine toxicology, history

of head trauma with loss of consciousness, seizure, or another neurological disorder; no first

57

degree relative with a history of psychotic mental disorder. All participants were assessed with

the Edinburgh handedness inventory (Oldfield 1971) for handedness, Wechsler Test for Adult

Reading (WTAR) for IQ, and Hollingshead index for socio-economic status (Hollingshead

1975). They were interviewed by a psychiatrist, and completed the Structured Clinical Interview

for DSM-IV Disorders (First MB 1995) . They also completed a urine toxicology screen. The

study was approved by the Research Ethics Board of the Centre for Addiction and Mental Health

(Toronto, Canada) and all participants provided informed, written consent.

3.5.2 Neuroimaging

High resolution magnetic resonance images were acquired as part of a multi-modal imaging

protocol using an eight-channel head coil on a 1.5 Tesla GE Echospeed system (General Electric

Medical Systems, Milwaukee, WI), which permits maximum gradient amplitudes of 40 mT/m.

Axial inversion recovery prepared spoiled gradient recall images were acquired: echo time (TE)

= 5.3, repetition time (TR) = 12.3, time to inversion (TI) = 300, flip angle = 20, number of

excitations (NEX) = 1 (124 contiguous images, 1.5 mm thickness).

3.5.3 Image Processing

Each subject’s T1 image was submitted to the CIVET pipeline (version 1.1.7)

(http://wiki.bic.mni.mcgill.ca/index.php/CIVET) developed at the Montreal Neurologic Institute

(Ad-Dab'bagh, Einarson et al. 2006). The processing steps included registration to the symmetric

ICBM 152 template (Mazziotta, Toga et al. 2001) with a 12-parameter linear

transformation(Collins 1994), correction for inhomogeneity artifact (Sled, Zijdenbos et al. 1998),

skull stripping(Smith, Zhang et al. 2002), tissue classification into white and grey matter,

58

cerebrospinal fluid and background (Zijdenbos, Forghani et al. 2002, Tohka, Zijdenbos et al.

2004) and neuroanatomical segmentation using ANIMAL (Collins, Holmes et al. 1995). Total

volumes for each cortical lobe and subcortical structures were estimated for each individuals by

non-linearly warping each T1 image towards a segmented atlas (Chakravarty, Sadikot et al.

2008). Volume (mL) was extracted from each of these regions using the RMINC package

(version 0.4) for reading and analyzing MINC2 output files. Total gray matter, white matter, and

CSF volumes were calculated, along with lobar cortical gray and white matter volumes (i.e., left

and right frontal, temporal, parietal, occipital), along with volumes of subcortical structures

related to the fronto-striato-thalamic loop implicated in both schizophrenia and ASD including

left and right caudate, putamen, globus pallidus, and thalamus.

3.5.4 Genetics

Genomic data was extracted from ethylenediametetraaecidic acid (EDTA) anticoagulated venous

blood according to standard procedures. Eleven SNPs were genotyped on an Applied Biosystems

ABI 7500 Real-Time PCR system, using Taqman 5’ nuclease assay. Genotyping accuracy was

assessed by running 10% of the sample in duplicate. Eleven SNPs were selected across the

NRXN1 gene (NC_000002.11). Each marker is located in reported regions within which multiple

rare deletions associated with ASD and schizophrenia (Figure 3-1, Table 3-S5).

The program Haploview 4.2 (Barrett, Fry et al. 2005) was used to determine pair-wise LD

between all SNPs with blocks determined by the Gabriel et al. method (Gabriel, Schaffner et al.

2002). Haploview 4.2 was also used to determine whether SNPs were in Hardy Weinberg

equilibrium.

59

3.5.5 Cognitive Assessment

Fifty-two of the study participants completed cognitive testing that included the finger-tapping

test (Reitan and Wolfson 1985, Reitan and Wolfson 1993, Lezak 1995). Although cognitive

deficits in ASD are not as well-characterized as those in schizophrenia, sensorimotor function is

disrupted in both disorders (Flashman, Flaum et al. 1996, Honey, Pomarol-Clotet et al. 2005,

Goldman, Wang et al. 2009, Mostofsky, Powell et al. 2009). Therefore, we used the finger-

tapping test to assess sensorimotor function.

3.5.6 Statistical Analysis

Statistical analysis was performed using SPSS for Windows 15.0. To test for effects of NRXN1

genotype on brain morphometry, three separate repeated measures ANCOVA (for cortical lobar

gray matter, cortical lobar white matter, and subcortical structures) were performed with

genotype as the between group factor, brain region volume as the within group factor, and age

and total brain volume (TBV) as covariates. To ensure adequate power, only markers with a

minor allele frequency (MAF) greater than 15% were tested. We used a Bonferroni correction

based on multiple comparisons of 7 SNPs to determine significance (alpha = 0.0071). Where the

repeated measures ANCOVA revealed a significant volume by genotype interaction, follow-up

ANCOVAs were performed and Bonferroni correction applied. When significant effect of a

genotype on brain volume was found, UNPHASED 3.1 was then used to examine allelic

association with brain phenotypes. Haplotype quantitative analysis of frontal lobe white matter

volume and the rs1045881 and rs858932 NRXN1 variants were calculated using haplotype score

(Methods S1). Finally, for those genotypes that significantly predicted brain measures, repeated

measures ANCOVA for sensorimotor function was performed (dominant and nondominant

60

finger-tapping scores as within group measures) with age as covariate. For any gene variant that

predicted both brain measures and cognitive performance, the relationship between that brain

measure and cognitive performance was examined using a linear regression model, accounting

for age effects.

3.5.7 In Silico Analysis

In order to enhance the understanding of the biological meaningfulness of the genetic

associations, we used in silico methods to predict potential function of the SNPs investigated in

this study. Depending on their location, SNPs were assessed for alteration in transcription factor

binding using MatInspector (Genomatix; promoter and intron 1). Presence of splicing enhancers,

repressors or intronic regulatory elements (intronic and exonic, synonymous and

nonsynonymous SNPs) were determined using F-SNP (http://compbio.cs/queensu.ca/F-SNP) and

Human Splicing Finder (http://www.umd.be/HSF). 3’UTR SNPs were also assessed for

alteration in microRNA binding sites (http://www.targetscan.org/).

3.6 Acknowledgements

The authors would like to thank Dielle Miranda for her help with this study. This work was

supported by the Canadian Institutes of Health Research Clinician Scientist Award (ANV);

APA/APIRE Astra-Zeneca Young Minds in Psychiatry Award (ANV), NARSAD (ANV, AKT,

TKR) and the Centre for Addiction and Mental Health (AKT). Dr. A. Voineskos and Mr. T. Lett

had full access to all of the data in the study and take responsibility for the integrity of the data

and the accuracy of the data analysis. The funders had no role in study design, data collection

and analysis, decision to publish, or preparation of the manuscript.

61

Table 3-1. Demographic Characteristics.

Mean± St. Dev. Range

Age 39.0 ± 13.1 19-59

Education (years) 15.6 ± 2.0 12-20

IQ (WTAR) 118.2 ±7.7 92-127

Socioeconomic Statusa 50.0 ± 9.8 27-66

WTAR, Wechsler Test of Adult Reading. aComposed of four factors are education, occupation,

sex, and marital status.

62

Figure 3-1. Reported Deletions in the Neurexin-1α gene. Figure contains the location of gene,

markers, and reported deletion in: developmental disorders (green; Ching et al.(Ching, Shen et

al. 2010)), schizophrenia (red; Rujescu et al. (Rujescu, Ingason et al. 2009), Vrijenhoek et al.

(Vrijenhoek, Buizer-Voskamp et al. 2008), Magri et al. (Magri, Sacchetti et al. 2010), Ikeda et al.

(Ikeda, Aleksic et al. 2010), Need et al.(Need, Ge et al. 2009)), and autism spectrum disorders

(blue; Pinto et al. (Pinto, Pagnamenta et al. 2010), Glessner et al. (Glessner, Wang et al. 2009).

The Autism Chromosome Rearrangement Database (Marshall, Noor et al. 2008)). Figure adapted

from the UCSC genome browser (GRCh37/hg19 assembly) (Kent, Sugnet et al. 2002).

63

Figure 3-2. The effect of rs1045881 on combined hemispheric volume of brain regions with

total brain volume (TBV) and age as covariates. Brain regions: (A) Frontal Lobe, (B)

Temporal Lobe, (C) Occipital Lobe, and (D) Parietal Lobe. Frontal lobe white matter volume

was significantly greater in T allele carriers (T/T +T/C) (ANCOVA F1,52 = 8.197, p = 0.006),

while other regions are non-significant after correcting for multiple comparisons. Covariates

appearing in the model are evaluated at the following values: TBV = 1364768.17, Age = 39.04,

(*) denotes significance of P<0.0125. Error bars represent +/- standard error of the marginal

means and percentages reflect the percent change in each brain region.

64

Figure 3-3. The effect of rs858932 on right and left thalamic volume with TBV and age as

covariates. There are approximately 10% and 9% percent differences between the G/G to G/C

and G/G and C/C genotypes for both thalamic hemispheres, respectively. Covariates appearing in

the model are evaluated at the following values: TBV = 1364768.17, Age = 39.04, (*) denotes

significance of P<0.0063. Error bars represent +/- standard error of the marginal means.

65

Methods 3-S1. Haplotype Analysis.

Haplotype quantitative analysis of frontal lobe white matter volume and the rs1045881 and

rs858932 NRXN1 variants were calculated using haplotype score algorithm in haplostats in the R

programming language (http://mayoresearch.mayo.edu/mayo/research/schaid_lab/software.cfm).

Schaid et al.(Schaid, Rowland et al. 2002) developed a score statistic that can test the

associations between haplotypes and a wide variety of traits, including binary, ordinal,

quantitative, and Poisson. This method also allows for adjustment for non-genetic covariates. In

our analysis, we used haplo.score to compute the global score statistic (that tests the significance

of association of all haplotypes) and haplotype specific statistic (that compares each haplotype

with selected common haplotypes). Our dependent variable was frontal lobe white matter

volume. Our covariates were the TBV and age of the subjects. All haplotypes with a frequency

less than 5% were dropped from the score test.

66

Figure 3-S1. The effect of rs858932 on combined hemispheric volume of brain regions with

TBV and age as covariates. Brain regions: (A) Frontal Lobe, (B) Temporal Lobe, (C) Occipital

Lobe, and (D) Parietal Lobe. Frontal and parietal lobe white matter volumes were significantly

greater in G allele carriers (T/T +T/C) (ANCOVA F2,52 = 7.074, p = 0.002; ANCOVA F2,52 =

5.724, p = 0.006). Other region are non-significant after correcting for multiple comparisons.

Covariates appearing in the model are evaluated at the following values: TBV = 1364768.17,

Age = 39.04, (*) denotes significance of P<0.0125. Error bars represent +/- standard error of the

marginal means and percentages reflect the percent change in each brain region.

67

Table 3-S1. Locations and Minor Allele Frequency in Toronto and Hapmap (CEU)

Samples.

Marker Positiona Alleles Strand Location MAF MAF (CEU)b

rs1995584 51263149 A/G + 5’ Pro A: 0.446 A: 0.446

rs10490162 51247657 A/G - Intron G: 0.054 G: 0.102

rs12623467 51225089 C/T + Intron T: 0.027 T: 0.050

rs2193225 51079482 A/G - Intron A: 0.446 A: 0.496

rs858932 50930063 C/G - Intron G: 0.420 G: 0.442

rs11125321 50852016 A/G + Intron G: 0.375 G: 0.403

rs2351765 50793780 A/C + Intron A: 0.312 A: 0.274

rs6721498 50713011 A/G + Intron G: 0.464 G: 0.492

rs10490227 50659515 A/G - Intron A: 0.143 A: 0.093

rs10208208 50593914 G/T + Intron T: 0.000 T: 0.175

rs1045881 50148972 A/G - 3’UTR T: 0.205 T: 0.129

MAF = Minor Allele Frequency; aAccording to dbSNP build 131; bHapmap CEU Sample

68

Table 3-S2. T-test between rs1045881 T-Carriers Vs C/C and Demographics.

Genotype N Mean Std. Dev. t df p-value

Age TT + TC 20 38.55 12.959 -0.209 51 0.836

CC 33 39.33 13.411

Education TT + TC 18 15.89 1.568 0.893 48 0.377

CC 32 15.38 2.136

WTAR TT + TC 19 118.53 6.230 0.370 50 0.713

CC 33 117.70 8.527

MMSE TT + TC 20 29.60 0.821 1.202 50 0.235

CC 32 29.28 0.991

CIRSG TT + TC 18 1.28 1.841 -0.232 48 0.817

CC 32 1.41 1.898

SE Status a TT + TC 17 51.12 8.667 0.736 43 0.466

CC 28 48.89 10.457

WTAR, Wechsler Test of Adult Reading; MMSE, Mini Mental State Examination; CIRS-G,

Cumulative Illness Rating Scale - Geriatrics; SE, Socioeconomic status. aComposed of four

factors: education, occupation, sex, and marital status.

69

Table 3-S3. Chi-squared Tests of Region by Genotype or Allele Interactions of rs1045881

and rs858932. Analysis was performed by Unphased 3.1 with total brain volume and age as

confounding factors.

rs1045881 rs858932

Allelic

(C vs T)a

Genotypic

(T-Carriers vs C/C)b

Allelic

(G vs C)c

Genotypic

(G/G vs G/C vs C/C)d

Region χ2 p-value χ2 p-value χ2 p-value χ2 p-value

Frontal Lobe 7.1840 0.0074 8.4151 0.0037 4.1213 0.0423 10.0033 0.0067

Temporal Lobe 1.8624 0.1723 2.4474 0.1177 2.4568 0.1170 3.9295 0.1402

Occipital Lobe 0.4720 0.4921 0.1402 0.5639 0.8973 0.3435 1.8958 0.3876

Parietal Lobe 2.8906 0.0891 4.2641 0.0389 4.0856 0.0432 8.5304 0.0140

Bold values are significant after Bonferroni correction alpha for multiple comparisons

(α=0.0125). a(T:C=21:85); b(T-carriers:C/C = 20:33); c(G:C = 45:61); d(G/G:G/C:C/C=6:33:14).

70

Table 3-S4. Haplotype Association between Frontal Lobe White Matter and rs1045881

(T/C) and rs858932 (G/C).

a Age and Total Brain Volume are covariates.

Haplotype Test of Overall Association with Frontal Lobe White Mattera

Global Score Statistic df p-value

21.92665 3 7 x 10-5

Estimates of Haplotype Main Effectsa

Haplotype Haplotype Score Freq p-value

T/G 2.12585 0.09992 0.03352

T/C 2.49622 0.09819 0.01255

C/G -3.9051 0.32461 9 x 10-5

C/C 1.10893 0.47728 0.26746

71

Table 3-S5. Reported deletions within NRXN1 in Developmental Disorders, Schizophrenia

and Autism Spectrum Disorders.

Start Stop Diagnosis

Ching et al. (2010) (Ching, Shen et al. 2010)

46938685 52015885 Moderate mental retardation

50128256 54050713 Global developmental delays, suspected autism

50897002 51212385 Gross motor delay, hypotonia

50936914 51167934 PDD-NOS, hypotonia

50920082 51059469 VACTERL

51059410 51316396 PDD-NOS, motor coordination delays

51090504 51212385 Autism, moderate mental retardation

50522892 50827767 Mild mental retardation

50689280 50853329 Language delay, prenatal substance exposure

50714297 50853329 PDD-NOS

50735499 50811018 Hypotonia, muscle weakness, large birth weight

50735499 50801233 Poor weight gain, mild craniofacial dysmorphism

Rujescu et al. (2009) (Rujescu, Ingason et al. 2009)

50856110 50900862 Schizophrenia

50890216 51116653 Schizophrenia, social contact problems in childhood

51147600 51225851 Schizophrenia

50071499 50208992 Schizophrenia, chronic, positive symptoms, low IQ (82),

low educational level

50822312 50948557 Schizophrenia, chronic, negative symptoms, episode

of aggression

51101161 51344213 Schizophrenia, alcoholism

72

50735657 50800548 Schizophrenia

50786446 50900862 Schizophrenia

51002576 51250922 Schizophrenia, episodic with partial remission and negative

symptoms

51024962 51251873 Schizophrenia

51211406 51299436 Schizophrenia, myoclonic seizures in the right shoulder

50711199 50756435 Schizophrenia

50836690 50936258 Schizophrenia

50850456 51225851 Schizophrenia, mental retardation (mild), Tardive dyskinesia

Vrijenhoek et al. (2008) (Vrijenhoek, Buizer-Voskamp et al. 2008)

51063670 51300517 Schizophrenia

Magri et al. (2010) (Magri, Sacchetti et al. 2010)

50952424 51280162 Schizophrenia

Ikeda et al. (2010) (Ikeda, Aleksic et al. 2010)

50743926 50911879 Schizophrenia

Need et al. (2009) (Need, Ge et al. 2009)

49999148 51113178 Schizophrenia

Pinto et al. (2010) (Pinto, Pagnamenta et al. 2010)

50912249 50955087 Autism Spectrum Disorder

50493827 50677835 Autism Spectrum Disorder

50539877 50730546 Autism Spectrum Disorder

50990306 51222043 Autism Spectrum Disorder

51002576 51157742 Autism Spectrum Disorder

50735657 50804497 Autism Spectrum Disorder

50822312 50886363 Autism Spectrum Disorder

73

50822312 50900862 Autism Spectrum Disorder

Glessner et al. (2009) (Glessner, Wang et al. 2009)

51120644 51147600 Autism Spectrum Disorder

The Autism Chromosome Rearrangement Database (Marshall, Noor et al. 2008)

50371853 50727153 Autism Spectrum Disorder

50722055 50801053 Autism Spectrum Disorder

50273117 50443987 Autism Spectrum Disorder

51086655 59969199 Autism Spectrum Disorder

51759493 52031003 Autism Spectrum Disorder

74

Chapter 4

4 The Genome-Wide Supported MicroRNA-137 Variant Predicts Phenotypic Heterogeneity Within Schizophrenia

Contents of this chapter have been published as:

Lett TA et al. The genome-wide supported microRNA-137 variant predicts phenotypic

heterogeneity within schizophrenia. Mol Psychiatry. 2013 Apr;18(4):443-50

A link to the published paper can be found at:

http://www.nature.com.myaccess.library.utoronto.ca/mp/journal/v18/n4/full/mp201317a.html

Reprint by permission from Nature Publishing Group

75

4.1 Abstract

We examined the influence of the genome-wide significant schizophrenia risk variant rs1625579

near the microRNA-137 (MIR137) gene on well-established sources of phenotypic variability in

schizophrenia: age at onset of psychosis, and brain structure. We found that the MIR137 risk

genotype strongly predicts an earlier age-at-onset of psychosis across four independently

collected samples of patients with schizophrenia (n=510; F1,506=17.7, p = 3.1x10-5). In an

imaging-genetics subsample that included additional matched controls (n=213), patients with

schizophrenia who had the MIR137 risk genotype had reduced white matter integrity

(F3,209=13.6, p=3.88x10-8) throughout the brain as well as smaller hippocampi, and larger lateral

ventricles; the brain structure of patients who were carriers of the protective allele was no

different from healthy control subjects on these neuroimaging measures. Our findings suggest

that MIR137 substantially influences variation in phenotypes that are thought to play an

important role in clinical outcome and treatment response. Finally, the possible consequences of

genetic risk factors may be distinct in patients with schizophrenia compared to healthy controls.

Keywords: schizophrenia; age-at-onset; imaging; genetics; MIR137; heterogeneity

76

4.2 Introduction

There is notable heterogeneity in the phenotypic presentation of schizophrenia including, but not

limited to, the onset of illness, severity of positive and negative symptoms, neurological soft

signs and cognition, course of illness, response to treatment, and functional and structural brain

abnormalities(Carpenter Jr and Kirkpatrick 1988, DeLisi 1992, Shenton, Dickey et al. 2001).

This phenotypic heterogeneity has been a central challenge for schizophrenia research and other

neuropsychiatric disorders.

MicroRNAs (miRNAs) may be critically important genetic mechanisms contributing to

phenotypic heterogeneity.. Individual, small non-coding miRNAs regulate hundreds of genes in

tandem, and may fine-tune the activity of entire biological pathways (Chen and Rajewsky 2007).

Therefore, miRNAs may play an especially important role in contributing to phenotypic

heterogeneity via regulation of, or interaction with, risk gene pathways in neuropsychiatric

disorders (Kwon, Wang et al. 2011, Kim, Parker et al. 2012, Miller, Zeier et al. 2012). MiRNAs

function as crucial regulators of gene expression, and have been identified as potent disease

modifiers(Karres, Hilgers et al. 2007, Kim, Inoue et al. 2007, Lee, Samaco et al. 2008, Williams,

Valdez et al. 2009). MiRNAs function as crucial regulators of gene expression, and have been

identified as potent disease modifiers. For instance, knock out of miR-8 results in elevated

neuronal apoptosis and behavioral defects(Karres, Hilgers et al. 2007); a mouse model of

amyotrophic lateral sclerosis, miR-206 slows the progression of motor neuron

degeneration(Williams, Valdez et al. 2009); inhibition of microRNAs increase the toxicity of

CAG repeats in a spinocerebellar ataxia type 1 animal model(Lee, Samaco et al. 2008), and the

microRNA-133b regulates maturation and function of midbrain dopaminergic neurons relevant

to Parksinon’s disease(Kim, Inoue et al. 2007).

77

MicroRNA-137 (miR-137) serves as a regulator of adult neural stem cell maturation and

migration(Smrt, Szulwach et al. 2010, Szulwach, Li et al. 2010, Sun, Ye et al. 2011) in the

subventricular zones in proximity to the lateral ventricles and the subgranular zone of the

hippocampus. MiR-137 is also a regulator of gliogenesis(Silber, Lim et al. 2008). A single

nucleotide polymorphism, rs1625579, near the MIR137 gene (microRNA 137; 1p21.3) recently

achieved genome-wide significance for association with schizophrenia in a study of

approximately 50,000 subjects (p=1.6 x 10-11)(Ripke, Sanders et al. 2011). This polymorphism is

in the intronic region of the MIR137HG gene, MIR137 host gene (non-protein coding), that

includes MIR137. MIR137 has also been functionally shown to specifically regulate genes with

replicated genome-wide significant evidence for a role in schizophrenia, most notably CACNA1C

(calcium channel, voltage-dependent, L type, alpha 1C subunit) and TCF4 (transcription factor

4) (Kwon, Wang et al. 2011). The known role of microRNAs as potent disease modifiers raises

the question of whether genetic variation in the MIR137 gene might play a critical role in

phenotypic expression of schizophrenia, a psychiatric disorder known to show extensive

phenotypic heterogeneity.

In schizophrenia, age-at-onset of psychosis(DeLisi 1992) and brain structure(Shenton, Dickey et

al. 2001) are two well-established phenotypic measures, which are heritable,

heterogeneous(Brans, van Haren et al. 2008, Hare, Glahn et al. 2010), and related to disease

severity and outcome (Suvisaari, Haukka et al. 1998, Lieberman, Chakos et al. 2001, Lieberman,

Tollefson et al. 2005, Mitelman, Canfield et al. 2010). Brain structure including lateral ventricle

volume increases and hippocampal volume reductions in schizophrenia are well-established

sources of phenotypic heterogeneity and lateral ventricle volume, in particular, may predict

disease outcome (Lieberman, Chakos et al. 2001, Ho, Andreasen et al. 2003). A meta-analysis of

studies comparing schizophrenia patients and healthy controls showed reduced hippocampal

78

volumes and increased ventricular volumes in patients relative to controls(Steen, Mull et al.

2006). Similarly, cortical thickness reductions(Ehrlich, Brauns et al. 2011) and white matter

integrity changes(Voineskos, Foussias et al. 2013) can be also be heterogeneous among patients

with schizophrenia. Across these studies, the range of structural changes in schizophrenia

patients overlapped with controls.

The identification of the genetic sources of phenotypic heterogeneity, such as the effects of a

genetic risk variant on phenotypes such as age-at-onset, or brain structure, may lead to early

identification of disease trajectory. Such identification, before disease progression, could then

serve as a platform to test earlier interventions, particularly within the subgroup at-risk for poorer

outcome. Given the recently established role of MIR137 as a central player in coordinating the

timing and expression of schizophrenia risk genes (Kwon, Wang et al. 2011), we hypothesized

that MIR137 may be an important determinant of age-at-onset of psychosis and brain structure in

schizophrenia.

4.3 Subjects and Methods

4.3.1 Participants for Genetic Investigation of Age-at-onset Phenotypes

For the age-at-onset analysis, four independently collected samples were investigated. In total,

510 patients (346 Male, 154 Female) diagnosed with schizophrenia or schizoaffective disorder

according to DSM-III or DSM-IV criteria were included. Patients with a history of substance

abuse or dependence and those with a head injury with loss of consciousness >30 minutes or

neurological disorders were excluded from the study. Subjects were recruited from three clinical

sites in North America: the Centre for Addiction and Mental Health in Toronto, Canada (JLK,

Toronto schizophrenia sample: n=278; ANV, Toronto imaging-genetics sample: n=96). Case

Western Reserve University in Cleveland, Ohio (HYM, n=85); and Hillside Hospital in Glen

79

Oaks, New York (JAL, n=51). No overlapping subjects were present between the two Toronto

samples. Age-at-onset was ascertained in the same manner in all samples, recorded as the year of

the first psychotic episode. Ethnicity was assessed by self report.

4.3.2 Participants for Genetic Investigation of Neuroimaging Phenotypes

Participants from the Toronto imaging-genetics sample were recruited at the Centre for

Addiction and Mental Health (CAMH) in Toronto, Canada, via referrals, study registries, and

advertisements. All clinical assessments occurred at CAMH while DT-MRI scans were

performed at a nearby general hospital. Ninety-two of the 96 patients with a diagnosis of

schizophrenia or schizoaffective disorder and 121 healthy control subjects in this sample

completed all imaging, and genetics, protocols. All participants were administered the Structured

Clinical Interview for DSM-IV Disorders (First MB 1995) to determine diagnosis, and were

interviewed by a psychiatrist to ensure diagnostic accuracy. IQ was measured using the Wechsler

Test for Adult Reading (WTAR)(Wechsler 2001) and all participants were screened with the

Mini Mental Status Exam (MMSE) for dementia(Folstein, Folstein et al. 1975) and a urine

toxicology screen. Comorbid physical illness burden was measured by administration of the

Clinical Information Rating Scale for Geriatrics (CIRS-G)(Miller, Paradis et al. 1992).

Medication histories were initially recorded via self-report, and then verified either by the

patient’s treating psychiatrist or chart review. All subjects received urine toxicology screens and

anyone with current substance abuse or any history of substance dependence was excluded.

Individuals with previous head trauma with loss of consciousness, or neurological disorders were

also excluded. A history of a primary psychotic disorder in first-degree relatives was an

additional exclusion criterion for controls.

80

4.3.3 Image Acquisition

High-resolution axial inversion recovery-prepared spoiled gradient recall MR images were

acquired using a 1.5-T GE Echospeed system (General Electric Medical Systems, Milwaukee,

WI; echo time (TE): 5.3, repetition time (TR): 12.3, time to inversion: 300, flip angle 20,

number of excitations=1; 124 contiguous images, 1.5 mm thickness). For DTI acquisition, a

single-shot spin-echo planar sequence was used with diffusion gradients applied in 23 non-

collinear directions, b=1000 s/mm2, and two b=0 images. Fifty-seven slices were acquired for

whole-brain coverage oblique to the axial plane (2.6 mm isotropic voxels; field of view was

330 mm, 128 × 128 mm2 acquisition matrix; TE=85.5 ms, TR=15 000 ms; the sequence was

repeated three times to improve signal-to-noise ratio).

4.3.4 Cortical Volumes Processing

Automated measures of total cerebral and lateral ventricles volumes were derived via the CIVET

pipeline (version 1.1.10 developed at the Montreal Neurologic Institute)(Lerch, Pruessner et al.

2005, Lerch, Pruessner et al. 2008). Hippocampal volumes were processed using the FMRIB

Integrated Registration and Segmentation Tool (FIRST v1.2) automated subcortical

segmentation pipeline(Patenaude, Smith et al. 2011). Brain tissue volume, normalized for subject

head size, was estimated with SIENAX(Smith, Zhang et al. 2002), also part of the FSL toolkit.

4.3.5 Cortical Thickness Mapping

All MRIs were submitted to the CIVET pipeline. T1 images were registered to the ICBM152

nonlinear template with a 9-parameter linear transformation, intensity inhomogeneity corrected

(Sled, Zijdenbos et al. 1998) and tissue classified (for grey matter, white matter, and cerebral

spinal fluid)(Zijdenbos, Forghani et al. 2002, Tohka, Zijdenbos et al. 2004). Deformable models

81

were used to create white and gray matter surfaces for each hemisphere separately, resulting in 4

surfaces of 40,962 vertices each(MacDonald, Kabani et al. 2000, Kim, Singh et al. 2005). From

these surfaces, the t-link metric was derived for determining the distance between the white and

gray surfaces(Lerch and Evans 2005). The thickness data were blurred using a 20-mm surface-

based diffusion kernel in preparation for statistical analyses. Unnormalized, native-space

thickness values were used in all analyses owing to the poor correlation between cortical

thickness and brain volume (Ad-Dab'bagh, Singh et al. 2005).

4.3.6 Tract-Based Spatial Statistics (TBSS)

All diffusion tensor imaging (DTI) analysis was done using tools implemented in the FSL toolkit

v.4.1.10 (Smith, Jenkinson et al. 2004). All three repetitions were merged for each subject’s 4D

DTI volume. The resulting images were corrected for motion and eddy current distortion, and

then averaged. After brain extraction and skull stripping using BET (Smith, Zhang et al. 2002),

fractional anisotropy (FA) images were created by fitting a tensor model at each voxel using

DTIFit. FA quantifies directionality of water diffusion on a scale from zero (random diffusion) to

one (diffusion in one direction). Voxel-wise analysis of the FA data was carried out using Tract-

Based Spatial Statistics (TBSS, v1.2) (Smith, Jenkinson et al. 2006). TBSS projects all subjects'

fractional anisotropy (FA) data onto a mean FA tract skeleton, before applying voxelwise cross-

subject statistics. FA images then underwent nonlinear registration to the FMRIB58_FA target

image. Next, the mean FA image was iteratively generated from scans of healthy controls and

patients with schizophrenia separately. Each group was then aligned to MNI 152 standard space

using an affine transformation. An average white matter skeleton was then generated from the

mean of all subjects’ transformed FA images at a threshold of 0.2. For group comparisons, each

82

subject’s FA data was projected onto the white matter skeleton and voxel-wise statistics were

calculated using randomise (v2.1) with 10,000 permutations.

4.3.7 Whole-Brain TBSS Analysis

To compare genotype-by-diagnosis groups in all patients with schizophrenia and matched

controls, we extracted global whole-brain skeleton FA values. Univariate analysis of covariance

(ANCOVAs) were applied in the Statistical Program for the Social Sciences v. 15.0 software

(SPSS; Chicago, Illinois), comparing diagnosis-genotype group, with age as a covariate.

4.3.8 Genetics

Genotyping of the rs1625579 polymorphism was performed using a standard ABI (Applied

Biosystems Inc.) 5’ nuclease Taqman assay-on-demand protocol in a total volume of 10 µL.

Postamplification products were analyzed on the ABI 7500 Sequence Detection System (ABI,

Foster City, California, USA) and genotype calls were performed manually. Results were

verified independently by laboratory personnel blind to demographic and phenotypic

information.(Lahiri and Nurnberger 1991). Genotyping accuracy was assessed by repeating 10%

of the sample.

4.3.9 Statistical Analysis

Given the low frequency of GG genotype, these cases were collapsed with GT genotypes and

referred to as ‘G allele carriers’ for all statistical tests. Analyses were performed examining the

relationship between MIR137 genotype group (T allele homozygotes or G allele carriers) for age-

at-onset and brain morphology.

The influence of MIR137 as a predictor of age-at-onset was studied by ANCOVA with sex and

sample site as covariates using SPSS (Version 15). In complementary analyses, we further

83

performed a survival analysis (Cox proportional hazards model) to estimate the effect of

genotype on age-at-onset while controlling for sex. Ethnicity (Caucasian versus non-Caucasian),

sex (male versus female), and MIR137 genotype were compared among the four samples using

Chi-squared tests. To further exclude the potential confound of population stratification, we then

examined only subjects identified as Caucasian (3/4 grandparents from Caucasian descent) in a

separate analysis. Adherence to Hardy-Weinberg equilibrium was determined using Haploview

4.2(Barrett, Fry et al. 2005).

To assess the effect of MIR137 on brain morphology, we examined subjects from the Toronto

imaging-genetics sample: schizophrenia subjects that underwent structural MRI protocols (n=92)

and a healthy control sample (n=121). Repeated measure ANCOVAs were performed with

MIR137 genotype and diagnosis as the between-group factors, region as the within-group factors,

with age and total brain volume as a covariates. Using this model, we evaluated total brain

volume, ventricular volumes and hippocampal volumes. Using this model, we evaluated

between-subject interaction (genotype by diagnosis) and within-subject interaction (region by

genotype by diagnosis). For cortical thickness analyses, diagnosis and genotype were used as

between group factors, with age and sex as covariates. Finally, three separate TBSS analyses

were preformed. First, we examined a diagnosis by MIR137 genotype interaction, and main

effects of diagnosis and genotype on the FA skeleton. In follow-up analyses, separately in the

schizophrenia and control samples, we used TBSS to examine voxel-wise associations between

white matter integrity and MIR137 genotype with age as a covariate. An FDR correction of

q=0.05 was applied for cortical thickness and a family-wise error rate of 5% for white matter FA

comparisons. For lateral ventricle and hippocampal volumes, analyses were deemed significant

after Bonferroni correction for multiple comparisons. Multiple comparisons corrections were

performed within each analytic imaging modality. Post hoc partial correlations were performed

84

between hippocampal volumes, lateral ventricular volumes, and whole-brain white matter FA

after controlling for age and total brain volume.

4.3.10 Mediation Analysis

In post hoc analysis, a mediation model was used to evaluate if age-at-onset mediates the effect

of MIR137 rs1625579 genotype on white matter integrity. Others have shown that earlier age-at-

onset predicts reduced white matter integrity (Kyriakopoulos, Perez-Iglesias et al. 2009). We

used the multiple regression approach describe by Baron and Kenny (Baron and Kenny 1986).

There are four steps to establishing mediation (Figure 4-S4). This analysis is accomplished with

three regression equations: the dependent variable (white matter integrity) is regressed on the

independent variable (e.g. MIR137 genotype); the mediator (e.g. age-at-onset) is regressed on the

independent variable; and the dependent variable is regressed on both the mediator and

independent variables. Perfect mediation is defined as the case where the independent variable is

found to have no effect in the third equation (i.e., regression coefficient = 0); partial mediation is

the case where there is a significant reduction in the effect of the independent variable on the

dependent variable in the third equation. The Sobel test was used to assess the indirect effect of

the independent variable on the dependent variable via the mediator (Baron and Kenny 1986).

This test gives a Z score reflecting effect size and an associated p value.

85

4.4 Results

4.4.1 Genetics

For the schizophrenia subjects, the frequencies for rs1622579 genotype were 2.7% GG (n=14),

29.8% GT (n=152), and 67.5% TT (n=344). For the control subjects, the frequencies for

rs1622579 genotype were 4.1% GG (n=5), 33.1% GT (n=40), and 62.8% TT (n=76). There was

no significant deviation from Hardy-Weinberg equilibrium in either controls or patients with

schizophrenia (p>0.05). The minor allele frequency of rs1625579 in our schizophrenia

population (0.18) was virtually the same as the frequency observed in the Ripke et al. GWAS

(0.18) (Ripke, Sanders et al. 2011). In the Caucasian subsample our minor allele frequency was

0.164, which is nearly identical to the Hapmap CEU population (0.165; Hapmap Build #27)

(Gibbs, Belmont et al. 2003).

4.4.2 Age-at-onset

Frequencies and distribution of demographic data for the four age-at-onset samples are shown in

Table 4-S1. We found a significantly earlier mean age-at-onset of psychosis for T risk allele

homozygotes (20.8±5.8 years) compared to protective G allele carriers (23.4±8.5 years) after

covarying for sex and sample site: F1,506=17.7, p=3.1x10-5 (Figure 4-1a). The effect size of

rs1625579 genotype (Cohen’s D=0.38[95% CI: 0.20-0.57]) was greater than the effect size of

sex (Cohen’s D=0.20[0.02-0.39]), which to date has been considered among the most powerful

predictors of age-at-onset in schizophrenia(DeLisi 1992). A similar pattern was observed when

examining age-at-onset over time (using time dependent covariates). Our Cox regression model

showed a significant effects of progression to age-at-onset within schizophrenia patients across

the lifespan when grouped by rs1625579 genotype (β=0.37, p=1.43x10-4; Figure 4-1b) and sex

(β=0.23, p=0.015). A follow-up analysis was performed to exclude potential confounds of ethnic

86

admixture. Caucasian patients (n=379) yielded significant association between age-at-onset and

rs1625579 genotype (F1,375=12.7, p= 4.1x10-4; Cox-regression: p=0.003). It has been suggested

that schizophrenia with early (12 years or less; i.e. childhood-onset schizophrenia) or late

(greater than 45 years; i.e. paraphrenia) age-at-onset may constitute different forms of the

disorder based on severity and course of illness(Jeste, Harris et al. 1995, Hollis 2000). Thus, we

reanalyzed the sample excluding patients in these categories: early age-at-onset (n=15 [11 TT; 4

TG/GG]) and late age-at-onset (n=3 [0 TT; 3 TG/GG]). When considering our sample without

these individuals (n=492), the effect of rs1625579 genotype remained significant (F1,488=12.0,

p=5.7x10-4; Cox-regression: p=8.6x10-4; Figure 4-S1). Finally, in the subsample with brain

imaging data (n=92), we again found that T allele homozygotes had an earlier age-at-onset (22.5

± 6.9 years) compared to G allele carriers (27.8 ± 11.9 years; F1,89 = 7.8, p=0.006).

4.4.3 Neuroimaging

In the Toronto imaging-genetics sample, schizophrenia patients were not different from healthy

controls on age, sex, handedness, total brain volume, mini mental state examination (MMSE),

and CIRS-G, but had fewer years of education (p<0.05), and lower IQ (p<0.05; Table 4-S2).

There were no significant genotype differences on any demographic measure, PANSS score,

duration of illness, abnormal involuntary movement scale, chlorpromazine equivalent, MMSE,

and CIRS-G (p>0.05; Table 4-S2) with the exception of age-at-onset. Repeated measures

ANCOVA of the left and right lateral ventricles with age and total brain volume as covariates

revealed a significant MIR137 genotype by diagnosis interaction (F1,205=4.6, p=0.03). Further,

there was a significant within-group interaction between genotype and diagnosis (F1,205=4.3, p

=0.04). T allele homozygotes in the schizophrenia sample had larger left lateral ventricle

volumes than G allele carriers (F1,90=4.5, p=0.04). Repeated measures ANCOVA of left and right

87

hippocampal volumes with age and total brain volume as covariates revealed a main effect of

genotype (F1,205=4.2, p=0.05) and a significant genotype by diagnosis interaction (F1,205=4.2,

p=0.04). On follow-up within group analyses, repeated measure ANCOVA showed that T allele

homozyogtes in the schizophrenia sample had lower hippocampal volume (F1,87=5.6, p=0.02),

but there was no such effect in the control sample (F1,116=0.004, p=1.0). For both lateral ventricle

and hippocampal volumes, schizophrenia patients who carried the protective G allele were not

significantly different from healthy controls (Puncorrected>0.05; Figure 4-2a-d).

We found no effect of genotype or genotype by diagnosis interaction on cortical thickness at any

vertex using false discovery rate correction of 5%(Genovese, Lazar et al. 2002). Voxel-wise

statistical analysis of white matter integrity showed a prominent main effect of genotype on FA

throughout the brain using a family-wise error rate of 5% (Figure 4-S2). Most striking was the

widespread interaction between MIR137 genotype and diagnosis that was observed throughout

the brain (Figure 4-S3); we therefore proceeded to examine the effect of genotype separately in

the schizophrenia and control groups. In the schizophrenia sample, T allele homozygotes had

substantially lower FA than G allele carriers in an evident whole brain, rather than tract-based

pattern (Figure 4-3). No genotypic effect was found in the control group. Given the evident

effect of genotype across much of the white matter skeleton, we applied a whole brain TBSS

analysis to measure the aggregate of mean FA white matter skeleton for each subject, which

allowed us to treat white matter integrity as a single measurement. There was a global effect of

group (F3,209=13.60, p=3.88x10-8), and post hoc pairwise comparisons revealed that patients with

schizophrenia homozygous for the T risk allele have significantly lower whole-brain white

matter FA than all other genotype by diagnosis groups (Figure 4-3). Protective G allele carriers

in the schizophrenia group were no different in whole-brain white matter FA or in decline of FA

across the lifespan compared to the control group (p=0.72-0.84; Figure 4-3). Post hoc analysis

88

revealed that in patients homozygous for the risk allele, whole-brain white matter FA was

correlated with left and right ventricular volume (r=-0.54, p<0.001; r=-0.47, p<0.001).

4.4.4 Mediation Analysis

The effect of MIR137 genotype on whole-brain white matter FA was not significantly influenced

by controlling for age-at-onset. The main effect of genotype on FA (t=3.9, p=0.002) was similar

to the direct effect (t=3.4, p=0.009). Furthermore, the indirect effect of genotype on white matter

FA via age-at-onset was not significant (Z=0.90, p=0.37; Figure 4-S5) indicating that the effect

of MIR137 on mean whole-brain white matter FA was not significantly mediated by age-at-

onset. Age and age-at-onset alone were both significant predictors of whole-brain mean FA

(tage=-7.0,p=4.7x10-10; tAAO=2.1,p=0.04) and the model predicted 35% of the variance

(F2,92=24.6;p=3.1x10-9;R2=0.35). By adding MIR137 genotype to the model, we were then able

to explain 44% of the variance in FA (F3,92=23.1;p=3.8x10-11;R2=0.44).

4.5 Discussion

Our findings support an important role for the MIR137 rs1625579 variant in determining

phenotypic heterogeneity in schizophrenia via its effects on age-at-onset and brain structure.

Age-at-onset is a known predictor of disease severity in schizophrenia (DeLisi 1992). Similarly,

differences in brain structure in patients with schizophrenia have been related to disease severity

and disparate outcomes (Braff, Freedman et al. 2007) since the early classification based on

ventricular volume (Crow 1980), and correlation with functional outcomes (Ho, Andreasen et al.

2003). More recently, white matter FA has been found to be reduced in patients with poor

outcomes, but to a lesser degree in schizophrenia patients with good outcome (Mitelman,

Newmark et al. 2006). Therefore, our findings of T risk allele homozygotes with an earlier age-

at-onset, larger left lateral ventricle volume, smaller left hippocampal volume, and lower white

89

matter integrity, provide convergent evidence for MIR137 genotype as an important mechanism

of phenotypic variation in schizophrenia.

MiRNAs have emerged as integral regulators of expression of neuronal gene pathways involved

in brain function, plasticity and development (Olde Loohuis, Kos et al. 2012). There is increasing

evidence that miRNAs are implicated in both neurodevelopmental and neurodegenerative

disorders (Beveridge, Gardiner et al. 2009, Miller and Wahlestedt 2010, Geekiyanage and Chan

2011, Geekiyanage, Jicha et al. 2011, Moreau, Bruse et al. 2011, Santarelli, Beveridge et al.

2011). The role of MIR137 as a regulator of adult neural stem cell migration and maturation in

the subventricular zones and hippocampus (Smrt, Szulwach et al. 2010, Szulwach, Li et al. 2010,

Sun, Ye et al. 2011) supports the influence of this miRNA in neurodevelopmental processes

relevant to schizophrenia. Mir-137 interacts with several neurodevelopmental genes including

MIB1, MITF, TLX, LSD1, and EZH2 (Smrt, Szulwach et al. 2010, Szulwach, Li et al. 2010, Sun,

Ye et al. 2011, Willemsen, Valles et al. 2011). MIR137 is also involved in microRNA-

transcription factor regulatory networks within the Notch signalling pathway in neuroglia (Sun,

Gong et al. 2012). Therefore, one pathway through which MIR137 influences phenotypic

heterogeneity in schizophrenia may occur via neurodevelopmental gene networks.

The MIR137 gene has also been functionally shown to specifically regulate genome-wide

significant schizophrenia-associated liability genes such as CACNA1C and TCF4 (Kwon, Wang

et al. 2011). CACNA1C, a gene coding for a voltage-gated calcium channel, can influence

hippocampal function during an episodic memory task (Bigos, Mattay et al. 2010). CACNA1C

may also be relevant to white matter variation since calcium channel abnormalities have been

associated with white matter lesions in the brain (Matute 2010). Disruption of calcium channel

homeostasis can also play an important role in increasing oxidative stress, which is of particular

90

relevance to oligodendrocytes, which are the most susceptible cells to oxidative stress in the

central nervous system (McTigue and Tripathi 2008). TCF4 is of particular relevance to

oligodendrocytes at the neurodevelopmental stages since it promotes the initial stages of

oligodendrocyte differentiation (Emery 2010). In general, miRNAs form a positive feedback

loop during oligodendrocyte differentiation, such that key miRNAs induced early in

differentiation act to inhibit the expression of genes that promote oligodendrocyte precursor cell

maintenance, thus further inhibiting proliferation and promoting differentiation. These regulatory

interactions with other schizophrenia risk genes may provide a mechanistic explanation for our

findings of association of the MIR137 variant with hippocampal volume and white matter

integrity.

Our findings provide the first compelling evidence that MIR137 plays a sizeable role in

determining heterogeneity among schizophrenia patients in age-at-onset and brain structure.

Recent investigations have demonstrated association of this variant with symptom and cognitive

measures in schizophrenia. One recent publication found the MIR137 rs1625579 variant

associated with cognitive dysfunction in patients with high negative symptom burden(Green,

Cairns et al. 2012); however, the authors found protective G allele identified in the GWAS study

was associated with cognitive impairment (Ripke, Sanders et al. 2011). The risk variant has also

been recently associated with positive symptoms in a mixed sample of bipolar disorder I,

schizoaffective disorder, and schizophrenia patients (Cummings, Donohoe et al. 2012).

However, in both studies the associations were relatively weak, which is often the case in

explorations of genetic association with symptom or cognitive measures (Meyer-Lindenberg and

Weinberger 2006).

91

Age at onset has not been associated with gray matter deficits in several cross-sectional studies

(Lim, Harris et al. 1996, Marsh, Harris et al. 1997, Zipursky, Lambe et al. 1998). Larger

ventricular size has been associated with earlier age-at-onset (DeLisi, Hoff et al. 1991), and

differential effects have been observed in white matter FA between adolescent and adult onset

groups (Kyriakopoulos, Perez-Iglesias et al. 2009). In general, earlier onset patients (age-at-onset

before 18 years) are shown to have reduced hippocampal volume, increased lateral ventricle

volume, and decreased white matter integrity compared to matched controls (Giedd, Jeffries et

al. 1999, Kumra, Ashtari et al. 2004). However, it remains unclear whether these brain structure

abnormalities are related to the higher incidence of premorbid abnormalities (Hollis 1995,

Vourdas, Pipe et al. 2003), worse cognitive performance (Rajji, Ismail et al. 2009), and poorer

functional outcome (Hollis 2000) observed in early onset schizophrenia. Our mediation analysis

suggests that the effect of MIR137 genotype on earlier age-at-onset may not be driving lower

white matter FA. Rather, MIR137 may be an underlying neurobiological mechanism linking age-

at-onset and brain morphology within schizophrenia.

Phenotype choice notwithstanding, study of a single marker represents only a limited description

of the genetic variation within or near one gene. Furthermore, little is known regarding the

mechanism by which MIR137 may direct phenotypic heterogeneity. Therefore, future studies are

necessary to understand the precise interplay between miRNAs and transcription factors within

schizophrenia. Although MIR137 variation had a consistent effect across our schizophrenia

samples on both age-at-onset and neuroimaging phenotypes it is not entirely clear why a stronger

association was not found in the healthy control imaging sample, in line with the intermediate

phenotype approach. Although a main effect of genotype on brain structure was found in the

entire sample, the effect was more prominently observed in schizophrenia patients. One

speculative explanation might be that our finding is due to the interaction of MIR137 with other

92

schizophrenia risk genes, including one or more of ZNF804A, CACNA1C, and TCF4 (Kwon,

Wang et al. 2011, Kim, Parker et al. 2012), which might in turn influence age-at-onset and brain

structure. Therefore, the consequence of MIR137 variation may differ in schizophrenia patients

compared to healthy controls, and the mechanisms underlying such differences deserve further

exploration.

Phenotypic heterogeneity has impeded the study of neuropsychiatric disorders in general and

schizophrenia in particular. It poses a major challenge for consistent findings when comparing

schizophrenia patients to matched controls on any number of clinical or neurobiological

variables. Importantly, our data suggest MIR137 genotype may predict a schizophrenia

subphenotype with earlier age at onset and compromised brain structure. A better understanding

of phenotypic heterogeneity may lead to early identification of patients with more severe disease

trajectory for whom more aggressive treatment strategies may improve long-term outcome. In

addition, our results may help to provide a model for the role of miRNAs in phenotypic

heterogeneity of psychiatric disorders. Although it seems unlikely that a single genetic variant

could account for the phenotypic diversity of a disorder as complex as schizophrenia, MIR137

appears to be an important factor influencing the phenotypic expression in schizophrenia, and

potentially other related disorders.

4.6 Acknowledgements

This work was supported by the Canadian Institutes of Health Research Clinician Scientist

Award (ANV); NARSAD (ANV, TKR), Ontario Mental Health Foundation (ANV) and the

Centre for Addiction and Mental Health (CAMH) and the CAMH Foundation thanks to the

Kimel Family, Koerner New Scientist Award, and Paul E. Garfinkel New Investigator Catalyst

Award. We also like to thank Mr. Gabriel Oh for manuscript comments. Mr. Lett, Dr. Kennedy,

93

and Dr. Voineskos had full access to all of the data in the study and take responsibility for the

integrity of the data and the accuracy of the data analysis. No sponsor or funder played any role

in the design and conduct of the study; collection, management, analysis, and interpretation of

the data; and preparation, review, or approval of the manuscript.

4.7 Conflict of interest

We report the following conflicts of interest: JAL has received research funding or is a member

of the advisory board of Allon, Alkermes Bioline, GlaxoSmithKline Intracellular Therapies,

Lilly, Merck, Novartis, Pfizer, Pierre Fabre, Psychogenics, F. Hoffmann-La Roche LTD,

Sepracor (Sunovion) and Targacept. HYM reports having received research funding or is a

member of the advisory board of Novartis, Janssen, ACADIA, TEVA, Lilly, Jazz

Pharmaceuticals, Sunovion, Dainippon Sumitomo, Envivo. JLK has been a consultant to

GlaxoSmithKline, Sanofi-Aventis, and Dainippon-Sumitomo.

94

Figure 4-1. MIR137 rs1625579 risk variant homozygotes have earlier age-at-onset of

schizophrenia. (a) The mean age-at-onset of psychosis in T risk allele homozygotes (20.8 +/-

5.7 years) was significantly earlier than in G allele carriers (23.7 +/- 9.1 years; F1,540=21.4,

p=3.1x10-5). Circles represent each subject, and dotted lines mean AAO for each genotype

group. (b) Predicting survival curves of MIR137 rs1625579 genotype and AAO are based on a

Cox regression with sex as covariate. “Proportion surviving” refers to the proportion of

participants free of psychosis or not yet having onset of psychotic symptoms. There are

significant associations with both MIR137 genotype (β=0.37, p=1.4x10-4) and sex (β=0.23,

p=0.015).

95

Figure 4-2. MIR137 risk variant predicts poorer structural brain phenotypes in

schizophrenia. Marginal mean of (a) left lateral ventricle volume, (b) right lateral ventricle

volume, (c) left hippocampus volume, and (d) right hippocampal volume in control and

schizophrenia imaging samples. Covariates in the model are cerebral volume (1042204.12 mm3)

and age (45.9 years). Error bars represent s.e.m. * represents a significant difference between Scz

TT and other genotype-diagnostic groups (Scz GG/GT, Cnt TT, Cnt GG/GT) at an uncorrected

alpha of 0.05. SCZ = schizophrenia; CNT = control.

96

Figure 4-3. Effect of MIR137 rs1625579 genotype on voxel-based white matter integrity in

patients with schizophrenia. White matter regions in which schizophrenia subjects carrying the

T risk allele homozygotes have reduced fractional anisotropy than G allele carriers. Areas

coloured from red to yellow correspond to p-values ranging from 0.05 to less than 0.01 following

correction for multiple comparisons using family-wise error of 5%. Significant regions are

mapped onto the standard Montreal Neurological Institute atlas MN152 1-mm brain template.

Numbers refer to Z coordinates.

97

Figure 4-4. The effect of MIR137 risk variant on mean whole-brain fractional anisotropy

(FA) across the lifespan for four ‘diagnosis-genotype’ groups. There was a global effect of

group (F3,209=13.60, p=3.88x10-8). Post hoc pairwise comparisons revealed that patients with

schizophrenia that were T risk allele homozygotes had significantly lower mean FA than each of

the other genotype by diagnosis groups (CNT TT: p=6.05x10-8; CNT GT/GG: p=5.14x10-7; SCZ

GT/GG: p=1.05x10-5). No other pairwise comparisons were significant (p>0.05). Age was a

covariate in the model (45.9 years). SCZ = schizophrenia; CNT = control.

98

Table 4-S1. Demographics for age at onset samples

Sample Site Subjects, n Ethnicity, n Sex, n MIR137

genotype, n

AAO,

mean±SD

Toronto schizophrenia 278 203C,

75NC

193M, 85F 198 TT, 80

GT or GG

21.18±5.96

Toronto imaging-

genetics

96 74C, 22NC 63M, 33F 58 TT, 38

GT or GG

24.66±9.43

Case Western Reserve 85 61C,24NC 57M, 28F 48 TT, 37

GT or GG

20.72±5.86

Hillside Hospital in Glen

Oaks

51 41C, 10NC 33M, 18F 40 TT, 11

GT or GG

19.96±5.72

Total 510 379C,

131NC

346M, 164F 344 TT,

166 GT or

GG

21.64±6.86

AAO, Age-at-Onset of schizophrenia; C, Caucasian; F, Female; L, Left-handed; M, Male; NC,

Non-Caucasian; R, Right-handed.

99

Table 4-S2. Demographics and clinical characteristics for the Toronto imaging-genetics sample

Control Schizophrenia

TT,

mean±SD

GT or GG ,

mean±SD

Total,

mean±S

D

TT,

mean±SD

GT or GG ,

mean±SD

Total,

mean±SD

Age 44.95±19.7

5

45.18±17.8

7

45.03±1

9.00

44.18±16.71 48.78±17.6

2

45.72±17.

25

TBV (cm3) 1077.3±124

.1

1078.2±122

.0

1076.3±

124.1

1055.7±149.

7

1071.9±124

.4

1060.0±1

39.4

Education 15.53±1.58 15.20±2.32 15.40±1.

88

13.13±3.07 13.00±2.65 13.07±2.8

7

WTAR (IQ) 117.58±7.8

8

118.02±7.7

1

117.74±

7.79

107.60±17.2

3

110.47±13.

50

108.74±1

5.81

MMSE 29.31±0.89 29.44±0.89 29.36±0.

89

28.52±2.01 28.87±1.14 28.66±1.7

1

CIRS-G 2.05±2.08 1.78±2.13 1.95±2.1

0

2.42±2.09 2.61±2.27 2.48±2.15

AAO* 22.46±6.90 27.84±11.8

9

24.68±9.6

1

DOI 20.58±16.45 21.11±16.2

7

20.63±16.

29

AIMS 1.32±2.82 0.78±2.90 1.09±2.82

Chlorpromazine equivalent

(mg)

379.77±306.

99

361.75±209

.94

372.33±2

69.82

100

AAO, Age at Onset of schizophrenia; AIMS, Abnormal Involuntary Movement Scale; C,

Caucasian; CIRS-G, Cumulative Illness Rating Scale – Geriatrics; DOI – Duration of Illness, F,

Female; L, Left-handed; M, Male; MMSE, Mini Mental State Examination; NC, Non-Caucasian;

PANSS, Positive and Negative Syndrome Scale; FGA, first generation antipsychotic; SGA,

second generation antipsychotic; R, Right-handed; TBV, Total Brain Volume; WTAR, Wechsler

Test of Adult Reading. * denotes significant difference (p<0.05) in AAO by MIR137 genotype.

All other variables are non-significant (p>0.05).

Antipsychotic Treatment 45 SGA,4

FGA,1 Both

31 SGA,2

FGA,1 Both

76 SGA,6

FGA,2

Both

PANSS 52.96±14.78 55.84±16.0

5

54.13±15.

20

Positive 14.21±5.52 14.41±6.26 14.32±5.7

7

Negative 13.59±5.39 15.57±6.41 14.37±5.8

3

General 25.16±6.49 25.86±7.07 25.44±6.6

6

Sex, n 43M,32F 24M, 22F 67M,54

F

34M, 20F 26M, 12F 60M,32F

Handedness,

n

70R,4L 42R,3L 112R,7L 54R, 4L 34R, 4L 88R,8L

Ethnicity, n 71C, 4NC 44C, 2NC 115C,6N

C

41C, 13NC 31C, 7NC 72C,20N

C

101

Figure 4-S1. (a) Distribution of age-at-onset of psychosis based on MIR137 rs1625579.

Genotype is significantly associated with age-at-onset of psychotic symptoms (F1,488 = 12.0,

p=5.7x10-4). Bars represent estimated marginal mean for each genotypic group with sex and

sample site as a covariate. (b) Predicting survival curves of MIR137 rs1625579 genotype and

age-at-onset of psychosis, based on Cox regression with sex as covariate. “Proportion surviving”

refers to the proportion of participants free of psychosis or not yet having onset of psychotic

symptoms. There are significant associations with both MIR137 genotype (B=0.317, p=8.6x10-4)

and sex (B=0.247, p=0.009). Participants with an age-at-onset of 12 years or younger (n=15) or

over 45 years (n=3) were excluded for a total n=492 in this analysis.

102

Figure 4-S2. The main effect of MIR137 rs1625579 genotype on voxel-based white matter

integrity in healthy controls and patients with schizophrenia. Areas coloured from red to yellow

correspond to corrected p-values ranging from 0.05 to less than 0.01 at a family-wise error rate

of 5%. Significant regions are mapped onto the standard Montreal Neurological Institute atlas

MN152 1-mm brain template. Numbers refer to Z coordinates

103

Figure 4-S3. Effect of MIR137 rs1625579 genotype by diagnosis interaction on voxel-based

white matter integrity in healthy controls and patients with schizophrenia. Areas coloured from

red to yellow correspond to corrected p-values ranging from 0.05 to less than 0.01 at a family-

wise error rate of 5%. Significant regions are mapped onto the standard Montreal Neurological

Institute atlas MN152 1-mm brain template. Numbers refer to Z coordinates.

104

Figure 4-S4. Mediation Model. The models describe a causal relationship in which the mediator

causes the outcome and not vice versa. Each path (a,b,c, and c’) represents the four steps to

establishing mediation. First, demonstrate that the independent variable is correlated with the

dependent variable (path c; total effect). Second, show that the independent variable is correlated

with the mediator (path a). Third, show that the mediator affects the dependent variable (path b).

Fourth, demonstrate that the effect of the independent variable on the dependent variable (path

c’; direct effect) is significantly reduced or eliminated when the mediator is controlled for. The

indirect effect is the remaining variance explained the independent variable explains after

removing the variance explained by the mediator variable (path ab).

105

Figure 4-S5. Mediation model examining the associations between MIR137 rs1625579

genotype, age at onset and mean whole brain fractional anisotropy (FA). *Mean whole brain FA

has been corrected for age (45.9 years).

106

Chapter 5

5 GAD1 variant predicts a neuroanatomical and working memory susceptibly mechanism relevant to schizophrenia

107

5.1 Abstract

Cortical GABAergic dysfunction has been implicated as a key component in the

pathophysiology of schizophrenia and in working memory impairment. We examined the

influence of the functional rs3749034 variant in the glutamic acid decarboxylase 1 (GAD1) gene

on brain structure and working memory performance in schizophrenia patients and healthy

controls (N=197). We found that the rs3749034 A-allele carrier risk group predicted voxel-wise

lower white matter fractional anisotropy (FA) in frontal cortex region (Pcorrected<0.05), and

working memory performance (Digit-Span: p=0.005, LNS: p=0.026) as well as selective

attention (Stroop Ratio: p=0.009). White matter FA in the frontal cortex was associated with

digit-span performance. Last, our voxel-wise mediation analysis revealed that the effect of the

GAD1 risk variant on poorer digit-span performance was statistically caused by lower white

matter FA. Our findings converge on variation in the GAD1 predicting a susceptibility

mechanism through which genetic variation leads to reduced white matter FA and aberrant

working memory. These results also provide a plausible mechanism through which aberrant

GABA signaling in schizophrenia may potentiate working memory dysfunction.

108

5.2 Introduction

Working memory dysfunction is a central feature of schizophrenia and other psychiatric

disorders. In schizophrenia patients, working memory deficits are associated with dysfunction of

dorsolateral prefrontal cortex (DLPFC) as well as DLPFC connectivity with other regions and

disruption of neurotransmitter input such as GABA inhibitory neurotransmission (Meyer-

Lindenberg, Poline et al. 2001, Callicott, Mattay et al. 2003, Lewis and Moghaddam 2006).

There is a genetic basis to working memory dysfunction. Unaffected co-twins of schizophrenia

patients perform significantly worse than controls on spatial and verbal working memory tasks

(Cannon, Huttunen et al. 2000, Tuulio-Henriksson, Haukka et al. 2002, Pirkola, Tuulio-

Henriksson et al. 2005). For example, the letter-number-sequencing task (a measure of working

memory) has been identified as an endophenotype of schizophrenia with a moderately high

heritability (h2=0.39) (Greenwood, Braff et al. 2007). Therefore, understanding the effect of

inhibitory neurotransmission in shaping structural connectivity in the DLPFC and other brain

regions may provide insights into the neuroanatomical changes underlying working memory

impairment in schizophrenia.

Convergent evidence suggests a compelling role for the glutamate decarboxylase 1 (GAD1) gene

in working memory and dorsolateral prefrontal cortex (DLPFC) dysfunction in schizophrenia.

GAD1 codes for the glutamic acid decarboxylase (GAD67) enzyme that metabolizes glutamate to

GABA and is responsible for the production of the majority of GABA in the brain (Lewis,

Hashimoto et al. 2005). Downregulation of GAD67 in the parvalbumin-positive (PV)

interneurons of the DLFPC is a well-replicated postmortem finding in schizophrenia (Torrey,

Barci et al. 2005). Optogenetics has revealed that inhibition of fast-spiking parvalbumin

interneurons results in suppression of gamma activity (Cardin, Carlen et al. 2009, Sohal, Zhang

109

et al. 2009). Moreover, there is a growing body of evidence suggesting that abnormal gamma-

band oscillations are an endophenotype of schizophrenia related to working memory (Spencer,

Nestor et al. 2004, Lewis, Hashimoto et al. 2005, Spencer, Salisbury et al. 2008, Haenschel,

Bittner et al. 2009, Spencer 2009, Farzan, Barr et al. 2010, Farzan, Barr et al. 2010, Hall, Taylor

et al. 2011). The GAD1 rs3749034 SNP is associated with downregulation of GAD67 in the

DLPFC (Straub, Lipska et al. 2007), and reduced cortical thickness in the left parahippocampal

gyrus (Brauns, Gollub et al. 2013). Further, a 5’-promoter SNP in linkage disequilibrium with

rs3749034 is associated with variation in working memory performance and DLPFC function in

vivo (Straub, Lipska et al. 2007). In the mouse PFC, GAD67 deficits in the parvalbumin

interneurons has been causally link to inhibition transmission and network disinhibition

(Lazarus, Krishnan et al. 2013).

In the present study, we set out to provide convergent evidence that GABA signaling is integral

to brain structure related to working memory, and working memory performance. We first

applied an imaging-genetics approach to examine association between the GAD1 rs3749034 risk

variant and brain structure. Next, we examined the relationship between the risk variant and

working memory performance in the LNS and forward digit-span tasks, as well as the Stroop

selective attention task. Last, we connected the effect of GAD1 on working memory through

changes in brain structure via voxel-wise mediation analyses.

5.3 Methods

5.3.1 Participants

Participants from the Toronto imaging-genetics sample were recruited at the Centre for

Addiction and Mental Health (CAMH) in Toronto, Canada, via referrals, study registries, and

advertisements. All clinical assessments occurred at CAMH while diffusion tensor magnetic

110

resonance imaging (MRI) scans were performed at a nearby general hospital. Eighty patients

with a diagnosis of schizophrenia or schizoaffective disorder and 115 healthy control subjects in

this sample completed all imaging and genetics protocols; three healthy controls did not

complete DT-MRI protocols. Seventy one patients and 118 healthy controls completed cognitive

testing, and genetics protocols. All participants were identified as Caucasian based on self-

reported ethnicity of three out of four grandparents.

All participants were administered the Structured Clinical Interview for DSM-IV Disorders

(First MB 1995) to confirm diagnosis, and were interviewed by a psychiatrist to ensure

diagnostic accuracy. IQ was measured using the Wechsler Test for Adult Reading (WTAR)

(Wechsler 2001) and all participants were screened with the Mini Mental Status Exam (MMSE)

for dementia (Folstein, Folstein et al. 1975) and a urine toxicology screen. The Hand Dominance

Questionnaire was used to examine handedness. Comorbid physical illness burden was measured

by administration of the Clinical Information Rating Scale for Geriatrics (CIRS-G) (Miller,

Paradis et al. 1992). All subjects received urine toxicology screens and anyone with current

substance abuse or any history of substance dependence was excluded. Individuals with previous

head trauma with loss of consciousness, or neurological disorders were also excluded. A history

of a primary psychotic disorder in first-degree relatives was an additional exclusion criterion for

controls.

5.3.2 Genetics

Genotyping of the rs3749034 variant (GAD1) was performed using a standard ABI (Applied

Biosystems Inc.) 5’ nuclease Taqman assay-on-demand protocol in a total volume of 10 µL.

Postamplification products were analyzed on the ABI 7500 Sequence Detection System (ABI,

Foster City, California, USA) and genotype calls were performed manually. Results were

111

verified independently by laboratory personnel blind to demographic and phenotypic information

(Lahiri and Nurnberger 1991). Genotyping accuracy was assessed by repeating 10% of the

sample with 100% accordance in all genotype calls.

5.3.3 Image Acquisition

High-resolution axial inversion recovery-prepared spoiled gradient recall MR images were

acquired using a 1.5-T GE Echospeed system (General Electric Medical Systems, Milwaukee,

WI; echo time (TE): 5.3, repetition time (TR): 12.3, time to inversion: 300, flip angle 20,

number of excitations=1; 124 contiguous images, 1.5 mm thickness). For DTI acquisition, a

single-shot spin-echo planar sequence was used with diffusion gradients applied in 23 non-

collinear directions, b=1000 s/mm2, and two b=0 images. Fifty-seven slices were acquired for

whole-brain coverage oblique to the axial plane (2.6 mm isotropic voxels; field of view was

330 mm, 128 × 128 mm2 acquisition matrix; TE=85.5 ms, TR=15 000 ms; the sequence was

repeated three times to improve signal-to-noise ratio).

5.3.4 Cortical Thickness Mapping

All MRIs were submitted to the CIVET pipeline. T1 images were registered to the ICBM152

nonlinear template with a 9-parameter linear transformation, intensity inhomogeneity corrected

(Sled, Zijdenbos et al. 1998) and tissue classified (for grey matter, white matter, and cerebral

spinal fluid) (Zijdenbos, Forghani et al. 2002, Tohka, Zijdenbos et al. 2004). Deformable models

were used to create white and gray matter surfaces for each hemisphere separately, resulting in 4

surfaces of 40,962 vertices each (MacDonald, Kabani et al. 2000, Kim, Singh et al. 2005). From

these surfaces, the t-link metric was derived for determining the distance between the white and

gray surfaces (Lerch and Evans 2005). The thickness data were blurred using a 20-mm surface-

based diffusion kernel in preparation for statistical analyses. Unnormalized, native-space

112

thickness values were used in all analyses owing to the poor correlation between cortical

thickness and brain volume (Ad-Dab'bagh, Singh et al. 2005).

5.3.5 Tract-Based Spatial Statistics (TBSS)

All diffusion tensor imaging (DTI) analyses were done using tools implemented in the FSL

toolkit v.4.1.10 (Smith, Jenkinson et al. 2004). The three repetitions for each subject’s 4D DTI

volume were merged. The images were corrected for motion and eddy current distortion, and

averaged. After skull stripping using BET (Smith, Zhang et al. 2002), fractional anisotropy (FA)

images were created by fitting a tensor model at each voxel using DTIFit. FA quantifies

directionality of water diffusion on a scale from zero (random diffusion) to one (diffusion in one

direction). Voxel-wise analysis of the FA data was carried out using Tract-Based Spatial

Statistics (TBSS, v1.2) (Smith, Jenkinson et al. 2006). TBSS projects all subjects' fractional

anisotropy (FA) data onto a mean FA tract skeleton, before applying voxel-wise cross-subject

statistics. FA images then underwent nonlinear registration to the FMRIB58_FA target image.

Next, the mean FA image was iteratively generated from scans and was then aligned to MNI 152

standard space using an affine transformation. An average white matter skeleton was then

generated from the mean of all subjects’ transformed FA images at a threshold of 0.2. For group

comparisons, each subject’s FA data was projected onto the white matter skeleton and voxel-

wise statistics were calculated using randomise (v2.1) with 10,000 permutations.

5.3.6 Assessment of Working Memory

All subjects underwent a battery of cognitive tests administered over approximately 1.5 hours.

This battery includes a wide range of cognitive domains with varying degrees of impairment in

schizophrenia (Rajji, Ismail et al. 2009), and has been previously described (Voineskos, Rajji et

al. 2012, Voineskos, Felsky et al. 2013). We chose two working memory span tasks (verbal

113

working memory: the Letter-Number Span task (LNS); non-verbal working memory: the Digit

Span task (Digit-span)) (Randolph, Tierney et al. 1998, Hale, Hoeppner et al. 2002). The LNS

requires an understanding of order of the stimuli related to previous learning, whereas Digit-span

requires on the repetition of the forward order. We further assessed selective attention using the

Stroop Neuropsychological Screening Test (Stroop 1935, Golden and Freshwater 1978). We

assessed the Stroop interference effect by using the reaction time of the colour-word task (time

per item), and a ratio score (i.e. the Stroop difference score divided by the latency to colour-word

control items). This ratio score provides a more conservative estimate of Stroop interference

because it controls for differences in overall response latencies, both between and within groups

(Plude and Hoyer 1981, Perlstein, Carter et al. 1998).

5.3.7 Statistics

Given the low frequency of the AA genotype, these cases were grouped with GA genotypes and

referred to as ‘A allele carriers’ for all statistical tests. Analyses were performed examining the

relationship between GAD1 genotype group (G allele homozygotes or A allele carriers) for

cognitive performance and brain morphology. Adherence to Hardy-Weinberg equilibrium was

determined using Haploview 4.2(Barrett, Fry et al. 2005) for all healthy controls.

Differences in demographic characteristics between healthy controls and schizophrenia patients

were assessed by used independent t-tests for continuous variables and χ2 tests for count

variables (Table 5-1) using SPSS (Version 15). For cortical thickness analyses, diagnosis and

genotype were used as between group factors, with age as a covariate. For our TBSS analysis,

we examined a diagnosis by GAD1 genotype interaction, and main effects of diagnosis and

genotype on the FA skeleton with age as a covariate. An FDR correction of q=0.05 was applied

114

for cortical thickness and a family-wise error rate of 5% for white matter FA comparisons.

Multiple comparison corrections were performed within each analytic imaging modality.

To determine potential confounders of working memory, we first evaluated if age, sex, IQ, and

three measures of education (participant, father, mother) were predictors of task performance

(LNS, Digit-Span) in healthy controls (N=117) or schizophrenia patients (N=71) via the general

linear model (GLM). Diagnostic groups were evaluated separately, and covariates of no interests

were only included in our model if they were below a significance threshold of p = 0.1. The

influence of GAD1 as a predictor of working memory performance, diagnosis, and their

interaction was then analyzed via the GLM with our covariates of no interest. Four GLMs were

performed with LNS score, Digit-Span, Stroop (Time/Item), and Stroop ratio. Associations were

deemed significant at P=0.05 after Bonferroni correction for four comparisons.

5.3.8 Voxel-wide mediation analysis

Voxel-wise mediation analysis was performed in MATLAB (R2013b). We used the multiple

regression approach described by Baron and Kenny (Baron and Kenny 1986), and applied this

approach across the entire TBSS FA skeleton. A 4D image containing the TBSS skeleton of the

subjects was loaded into MATLAB, and transformed into an array of all non-zero voxels across

each subject. To remove the confounding effects of age, we then regressed out the effect of mean

centered age for all voxels. Our mediation analysis was accomplished with three regression

equations applied across all voxels. First, we regressed the independent variable (risk group)

against white matter FA. A z-score was produced for each non-zero voxel and was used to

produce a 3D image of z-scores (‘Path A’). We then applied TFCE in the FSL ‘fslmaths’

function with E=2, H=1, and the neighbourhood-connectivity parameter = 26 as recommended in

TBSS analysis (Smith and Nichols 2009). 10,000 permutations (i.e. randomization analysis)

115

were then performed and the maximum z-statistics for each permutation was used to assess

significance accounting for FWE. Second, we regressed the mediator variable (white matter FA

at each voxel) against cognitive performance at each voxel (‘Path B’). A 3D image of z-scores

were produced, and we tested significance using the same TFCE and permutation test technique.

Third, we regressed the independent variable (risk group) on cognitive performance (‘Path C’).

A significant association in all three sets of regressions then allowed us to proceed with the Sobel

equation to assess the indirect effect of the independent variable on the dependent variable via

the mediator (white matter FA at each voxel). We used the unstandardized regression

coefficients (beta) and the standard errors (SE) from ‘Path A’ and ‘Path B’ in order to produce a

z-value at each white matter FA voxel (Sobel equation: z-value = beta(Path A)*beta(Path B)/ √(beta(path

B)2 *SE(Path A)

2 + beta(Path A)2 *SE(Path B)

2)). A 3D image of z-values were produced, we applied

TFCE, and significant mediation was assessed using the max z-value of 10,000 permutations. It

should be noted that resampling strategies to assess significance of the Sobel equation are

considered to be a better alternative than parameter tests that impose distribution assumptions

(Preacher and Hayes 2008).

5.4 Results

5.4.1 Participants

Age, height, sex, and handedness were not significantly different between schizophrenia patients

and healthy controls. Patients had lower IQ, education, level of education of each parent (p<0.05;

Table 5-1). Patients also weighed more than controls (p<0.05; Table 5-1). There was no

significant deviation from Hardy-Weinberg equilibrium for the rs3749034 genotype in healthy

controls (p>0.05). The rs3749034 minor allele frequency of the healthy controls (0.22) was

similar to Hapmap CEU population (0.18; Hapmap Build #27) (Gibbs, Belmont et al. 2003).

116

5.4.2 Association between GAD1 and brain structure

Because GAD1 has previously been associated with gray matter volume and cortical thickness,

we performed vertex-wise analysis of cortical thickness. We found no effect of genotype or

genotype by diagnosis interaction on cortical thickness after covarying for age and at any vertex

using false discovery rate correction of 5% (Genovese, Lazar et al. 2002). Since the effect of

GAD1 on white matter fraction anisotropy is unknown, we tested GAD1 and the interaction with

diagnosis on TBSS skeleton FA after covarying for age and correcting for FWE. We found a

significant main effect of GAD1 on voxel-wise FA predominately in the prefrontal cortex

including the genu of the corpus callosum (Figure 5-1). There was no significant GAD1

genotype-by-diagnosis interaction on FA after correcting for FWE.

5.4.3 Association between GAD1 and working memory

We included age and IQ as covariates of no interest in our analyses since they were significantly

associated with our working memory outcome measures (Digit-Span and LNS) in schizophrenia

patients and healthy controls (p<0.05). Sex and education (participant, father, and mother) were

not associated with Digit-span or LNS performance in either patients or controls (p>0.1);

therefore, they were not included in as covariates.

We tested the effect of the GAD1 rs3749034 SNP on working memory and attention across three

cognitive tasks. There was a significant effect of GAD1 on digit span performance (F1,188 = 7.97,

p = 0.005) and a nominally significant effect on the LNS task performance (F1,188 = 5.03, p =

0.026), after covarying for subject age and IQ. There also was a significant association with

Stroop ratio (F1,188 = 7.03, p=0.009), but not time per item. No significant GAD1 genotype-by-

diagnosis interaction was observed for any cognitive task (Table 5-2).

117

5.4.4 Voxel-wise mediation analysis

Since there was a significant association between main effect of GAD1 risk genotype and

skeleton white matter FA, as well as working memory, we employed voxel-wise mediation

analysis across all subjects. For mediation to occur, we first needed to establish if skeleton whiter

matter FA predicted working memory performance (corrected for IQ). Higher skeleton FA

particularly in corpus callosum and right superior longitudinal fasciculus was associated with

better Digit-span performance after covarying for the effect of age and correcting for family-wise

error (pcorrected<0.05; Figure 5-2). Skeleton FA did not predict LNS performance or Stroop ratio

after correcting for family-wise error; therefore, we did not perform any mediation analyses for

the tasks. White matter skeleton FA mediated the association between GAD1 risk genotype and

Digit-span performance, particularly in the prefrontal and right inferior parietal regions,

suggesting that lower FA in this regions is statistically causing poorer working memory

(p[sobel]corrected<0.05).

5.5 Discussion

We have provided convergent evidence that the GAD1 risk variant leads to lower white matter

FA that may be related to poor working memory performance. Carriers of the GAD1 risk allele

had poorer working memory, and lower white matter FA in the prefrontal cortices. Moreover,

these effects tended to be independent from schizophrenia diagnosis. Further, we showed that

white matter FA positively correlates with digit-span performance. We can statistically infer a

causal relationship in which white matter FA mediates the effect on GAD1 on digit-span

performance. These results build upon previous work associated GABA inhibitory

neurotransmission in structural dysconnectivity and working memory dysfunction in

schizophrenia.

118

To the best of our knowledge, our study is the first to examine the effect of GAD1 genotype on

white matter FA. Our finding that GAD1 risk genotype predicted lower FA, predominately in the

genu of the corpus callosum, is consistent with the literature on top-down modulation of

posterior brain regions during inhibition and attention tasks (Hopfinger, Buonocore et al. 2000,

Erickson, Prakash et al. 2009). This is bolstered by our convergent cognitive associations that

risk genotype had poorer working memory across multiple tasks. Furthermore, we found the

effect of GAD1 on FA and working memory performance was consistent across diagnostic

groups. This suggests that genotypic variation in GAD1 acts as a modifier of brain structure and

related cognitive function. GAD1 may explain some of the heterogeneity of brain structure and

cognitive function reported within schizophrenia, despite the potential confounders within

disease. We were unable to replicate previous association between the GAD1 risk variant and

cortical structure (Brauns, Gollub et al. 2013). This may due to stringent FDR correction in our

vertex-wise analysis, or different methodologies.

The GAD1 rs3749034 risk variant has repeatedly been associated with reduced expression in the

prefrontal cortex (Straub, Lipska et al. 2007, Hyde, Lipska et al. 2011). The rs3749034 variant is

associated with decoupling of progressive switching from expression of GAD25 to GAD67 that is

required for proper maturation of the GABA function in the prefrontal cortex (Hyde, Lipska et

al. 2011). Therefore, GAD1 genotype may predict progressive neurodevelopmental changes

similar to what is observed in schizophrenia. Furthermore, decreased GAD1 expression in PV

interneurons in the prefrontal cortex is consistently associated with schizophrenia (Lewis, Curley

et al. 2012), although the density of PV interneurons in the DLPFC patients is no different than

healthy controls ((Woo, Miller et al. 1997, Hashimoto, Volk et al. 2003, Tooney and Chahl

2004)). Therefore, PV interneurons may not be altered in schizophrenia, rather that GAD1 is

reduced in a subset of these neurons. It is also important to note that GAD1 expression is

119

activity-regulated (Benson, Huntsman et al. 1994), and lower expression may be due reduced

cortical activity of secondary factors related to schizophrenia. Nevertheless, lower cortical GAD1

expression has not been associated with potential confounders (e.g. antipsychotic medication,

age of onset, functional outcome, and duration of illness) (Hashimoto, Arion et al. 2008, Curley,

Arion et al. 2011). Therefore, reduced GAD1 expression may be a core component of

schizophrenia, and our association with rs3749034 could be a marker of this disease process.

Our finding may have clinical implications in the treatment of working memory dysfunction.

Arguably the best supported treatment for working memory dysfunction (and general cognitive

function) in schizophrenia is cognitive remediation therapy (Lett, Voineskos et al. 2014).

Although the neurobiology mechanisms of CRT needs further research, CRT has been shown to

increase brain activation in the frontocortical regions associated with working memory function

(Wexler, Anderson et al. 2000, Wykes, Huddy et al. 2011, Penadés, Pujol et al. 2013). Also,

CRT has been associated with increased FA in the genu of the corpus callosum (Penadés, Pujol

et al. 2013). Nevertheless, not all schizophrenia patients respond to CRT with improved working

memory function. Bilateral rTMS treatment to the DLPFC has also been associated with

improved working memory performance, and GABAergic inhibitory function in schizophrenia

patients (Barr, Farzan et al. 2011, Barr, Farzan et al. 2013). Long-term rTMS treatment was also

associated with increased prefrontal cortex white matter FA (Allendorfer, Storrs et al. 2012,

Peng, Zheng et al. 2012). Considering our associations among GAD1, prefrontal cortex white

matter FA, and working memory, it could be speculated that GAD1 rs3749034 genotype may be

an important genetic predictor of adjunctive treatments including CRT and rTMS. Furthermore,

GAD1 genotype may be a potential confounder in future treatment trials, and therefore,

segregating subjects by genotype may improve trial outcome.

120

This study has a number of limitations. First, the evidence for association between GAD1 SNPs

and schizophrenia is relatively weak with no reported genome-wide findings, although there is

strong evidence for disruption of GAD1 expression in schizophrenia. Our a priori hypothesis was

that GAD1 is an important modifier of the disease process, and we selected the rs3749034 based

on it known effect on GAD1 expression. That is, we sought to indirectly model for how altered

GAD1 expression may lead to neuroanatomical differences and working memory dysfunction via

imaging-genetics. Second, since the minor allele frequency of the rs3749034 variants is

relatively low, for statistical analysis we inferred a dominant model by grouping minor allele risk

homozygotes and heterozygotes. Third, we limited our analysis only to Caucasian subjects;

therefore, we are unable to assess the effect of GAD1 in other ethnic groups.

Working memory dysfunction is one of the most intractable symptoms of schizophrenia, with

severe consequences on functional outcome (Lett, Voineskos et al. 2014). We have provided

evidence that GAD1 may predict lower working memory function via changes in white matter

FA, and likely explains some of the heterogeneity in working memory dysfunction. In addition,

our results suggests that the relationship between inhibitory signaling and working memory

dysfunction may be independent of schizophrenia.

121

Table 5-1. Demographics

Controls (N= 115) Schizophrenia (N = 80)

Mean SD Mean SD t193 P

Age (years) 46.79 19.32 45.49 16.00 0.49 0.63

IQ (WTAR score) 117.92 8.11 110.64 15.16 4.24 <0.05

Education (Years) 15.32 1.95 13.25 2.64 6.27 <0.05

Level of Education, Father 4.88 1.99 3.91 2.37 3.05 <0.05

Level of Education, Mother 4.67 1.75 3.79 2.12 3.93 <0.05

Weight (kg) 75.91 14.19 81.17 19.15 -2.13 <0.05

Height (m) 1.71 0.10 1.70 0.10 0.58 0.56

Count Frequency Count Frequency χ2 p

Handedness (R) 105 92.1% 71 89.9% 0.23 0.59

Sex (M) 61 53.0% 48 60.0% 0.93 0.38

122

Table 5-2. The association between working memory related tasks and GAD1 genotype,

diagnosis, and their interaction.

GAD1 Diagnosis GAD1*Diagnosis

Working Memory Task F1,188 P F1,188 P F1,188 P

Letter-Number Span 5.03 0.026 18.04 <0.0001 2.89 0.091

Digit Span 7.97 0.005 5.06 0.026 0.02 0.89

Stroop (Time/Item) 1.17 0.28 24.41 <0.0001 0.17 0.68

Stroop Ratio 7.03 0.009 12.50 0.001 2.50 0.12

**Covariates included in the model include age, and IQ (WTAR).

123

Figure 5-1. GAD1 rs3749034 risk A-allele predicts lower TBSS skeleton white matter FA in

healthy controls (N=115) and patients with schizophrenia (N=80). There was a significant

main effect of GAD1 genotype on prefrontal FA, and no significant effect of GAD1 genotype-

by-diagnosis interaction. Areas in yellow correspond to p<0.05 after correction for family-wise

error. Green represents the mean FA skeleton overlaid on the FMRIB58_FA standard space

image.

124

Figure 5-2. Higher TBSS skeleton white matter FA correlates with better digit span

performance. Areas in yellow correspond to p<0.05 after correction for family-wise error.

Green represents the mean FA skeleton overlaid on the FMRIB58_FA standard space image.

125

Figure 5-3. Skeletal white matter FA regions that mediate the effect of GAD1 rs3749034 on

digit span performance. Significant p-values indicated broad areas of the white matter skeleton

FA that mediated the effect of GAD1 risk A-allele risk genotype on poor digit span performance.

We created a z-statistic from voxel-wise Sobel tests for mediation based on the beta coefficients

and standard error from (i) GAD1 regressed on TBSS skeleton FA (Figure 5-1), and (ii) TBSS

skeleton FA regressed on digit-span performance (Figure 5-2). The z-values then underwent

threshold free clustering enhancement, and p-values are derived using permutation testing

(N=10000). Areas in yellow correspond to p<0.05 after correction for family-wise error. Green

represents the mean FA skeleton overlaid on the FMRIB58_FA standard space image.

126

Chapter 6

6 Additive Genetics Risk Predicts Widespread Changes in Brain Structure Leading to Poorer Cognitive Function

127

6.1 Abstract

There is growing theoretical and empirical evidence that additive genetic variation accounts for

the majority of the variance in complex traits. In healthy controls and schizophrenia patients

(N=198), we examined the association between an additive genetic model and brain structure via

brain-wide analysis of cortical thickness (vertex-wise analysis), and white matter FA (tract-based

spatial statistics), as well as cognitive performance. Our additive model included risk alleles

from MIR137 (rs1622579), CACNA1C (rs1006737), ZNF804A (rs1344706), GAD1 (rs3749034),

and BDNF (rs6265). Voxel-wise white matter FA mediation analysis was performed on

cognitive domains significant associated with additive genetic risk. We found that additive

schizophrenia risk score predicted white matter integrity throughout the brain (pcorrected<0.001),

and there was a significant model-by-diagnosis interaction predominately in the corpus callosum.

There was also a significant vertex-wise interaction between our additive risk score and

diagnosis in cortical thickness. High genetic risk loading predicted poor cognitive performance

and the effect was greater among schizophrenia patients for verbal fluency (F1,64=9.8, p=0.003;

interaction, F1,64=4.7, p=0.031) and motor functioning (F1,64=5.4, p=0.020; interaction,

F1,64=10.1, p=0.002)). Voxel-wise FA mediation analyses showed that genetic risk loading on

verbal fluency was statistically caused by white matter changes predominately in the corpus

callosum (Pcorrected[Sobel] < 0.001). Our findings suggest that our additive genetic risk model

predicts changes in brain structure and cognitive function. Furthermore, in a novel manner, our

results directly link our genetics association with cognitive performance through changes in

white matter FA.

128

6.2 Introduction

Schizophrenia is a chronic and severe brain disorder that affects approximately 1% of the

population, and the disease carries a high degree of heritability (>80%). It is a complex genetic

disorder and genetic predisposition is likely to be determined both through genetic pathways and

environmental risk factors. To date, common risk gene variants only have shown limited effect

sizes explaining disease association (Ripke, Sanders et al. 2011, Ripke, O'Dushlaine et al. 2013),

and the same has been largely true for neuroimaging phenotypes and cognitive phenotypes

relevant to schizophrenia (Greenwood, Braff et al. 2007, Brans, van Haren et al. 2008, Lencz,

Knowles et al. 2014). The genetic variation in complex traits, such as brain morphology,

consists of many components due to additive, dominant, and interaction effects of genes. There

is growing theoretical and empirical evidence that additive genetic variation accounts for the

considerable portion of genetic variance (Hill, Goddard et al. 2008). Therefore, examination of

additive genetic risk across several common variants might provide a better explanation for the

high degree of heterogeneity in neurocognitive dysfunction in schizophrenia that depends on

brain network connectivity (Chan and Gottesman 2008, Rao, Di et al. 2008, Chan, Wang et al.

2009, Lett, Voineskos et al. 2014). Furthermore, assessing additive genetics may better means to

assessing multi-gene risk without necessarily needing large sample size necessary for examining

statistical interactions, especially when considering more than three variants (Smolka, Buhler et

al. 2007, Puls, Mohr et al. 2009, Button, Ioannidis et al. 2013). Last, associations between risk

variants and neuroimaging phenotypes are rarely consistent among studies that may be explained

by method or sample heterogeneity. It could also be speculated that the effect of any one genetic

risk variant on brain structure may, in part, depend on the additive effect other risk variants.

129

In the present study, we investigate additive effects accrued across five risk variants implicated

in neuroanatomical and neurocognitive heterogeneity within schizophrenia. The rs1625579

variant near the microRNA-137 (MIR137) gene is among the top genome-wide associated SNPs

for schizophrenia, and MIR137 has been shown to regulate numerous other genome-wide

associated genes including: calcium channel, voltage-dependent, L-type, alpha 1C subunit

(CACNA1C) and zinc-finger 804A (ZNF804A)(Kwon, Wang et al. 2011, Ripke, Sanders et al.

2011, Kim, Parker et al. 2012, Ripke, O'Dushlaine et al. 2013). Recently, we have demonstrated

that the MIR137 risk variant (rs1625579) predicts clinical and neuroanatomical heterogeneity

within schizophrenia including poorer white matter structure, increased lateral ventricle volumes,

and lower hippocampal volume, as well as an earlier age-at-onset of psychosis. MIR137 has also

reported to be associated with frontal activation and cognitive function within schizophrenia

(Green, Cairns et al. 2012, Whalley, Papmeyer et al. 2012, van Erp, Guella et al. 2014). The

CACNA1C rs1006737 risk variant has been associated with poorer neurocognition in

schizophrenia patients but not healthy controls; however, the variant has been associated with

episodic memory circuit activation in healthy controls (Erk, Meyer-Lindenberg et al. 2010, Hori,

Yamamoto et al. 2012). It was recently reported that rs1006737 also confers greater frontolimbic

dysfunction in relatives of psychiatric patients compared to controls (Erk, Meyer-Lindenberg et

al. 2013). The ZNF804A rs1344706 risk allele has been associated with poorer episodic memory

and working memory in patients with schizophrenia but not healthy controls (Hashimoto, Ohi et

al. 2010, Walters, Corvin et al. 2010). Although in healthy controls during working memory, the

ZNF804A rs1344706 SNP risk allele carriers had reduced connectivity between the right and left

DLPFC, and increased right DLPFC and left hippocampal connectivity (Esslinger, Walter et al.

2009). ZNF804A is also associated with DLPFC greater inefficiency in siblings and patients with

schizophrenia than healthy controls; however, aberrant DLPFC-hippocampal connectivity was

130

only observed siblings and patients (Rasetti, Sambataro et al. 2011). In healthy controls, our

group has reported that individuals homozygous for the risk variant have reduced cortical

thickness in the left posterior cingulate cortex, left superior temporal gyrus and right anterior

cingulate cortex, all regions associated with cognitive dysfunction within schizophrenia

(Voineskos, Lerch et al. 2011).

We further included two risk genes to our additive model that have been strongly associated with

schizophrenia in postmortem studies as well as brain structure and cognitive function.

Convergent evidence suggests a compelling role for the glutamate decarboxylase 1 (GAD1) gene

in cognition and dorsolateral prefrontal cortex (DLPFC) dysfunction in schizophrenia. The major

determinant of GABA in the neocortex is glutamic acid decarboxylase-67 (GAD67; encoded by

the GAD1 gene). One of the most consistent findings in schizophrenia is down-regulation of

GAD1 mRNA and protein in the prefrontal cortex (Torrey, Barci et al. 2005), and in the PFC,

eight-fold increase of GAD1 methylation has been reported (Huang and Akbarian 2007).

Conserved 2q31 chromosomal configuration is associated with GAD1 transcription, and this

spatial genome architecture is disrupted in the prefrontal cortex of schizophrenia patients

compared to controls (Bharadwaj, Jiang et al. 2013). GAD1 variants has been shown to be

associated with childhood-onset psychosis within schizophrenia, and increased rate of cortical

gray matter loss (Addington, Gornick et al. 2004). GAD1 influences multiple cognitive domains

including declarative memory, attention and working memory in families with schizophrenia,

and prefrontal activation during working memory (Straub, Lipska et al. 2007). Our final gene of

interest, the brain-derived neurotrophic factor (BDNF), is one of the key regulators of

neuroplasticity, synaptic structure, memory function and consolidation. Post-mortem studies

have identified reduced BDNF expression in the hippocampus and prefrontal cortex of

schizophrenia patients (Green, Matheson et al. 2011), and reduced BDNF levels in schizophrenia

131

patients have been associated with cognitive performance and clinical outcome (Chen da, Wang

et al. 2009, Vinogradov, Fisher et al. 2009). In healthy controls, we recently have reported that

the BDNF rs6265 SNP interacts with age to predict differences in cortical thickness, white matter

FA, and episodic memory relevant to Alzheimer’s disease (Voineskos, Lerch et al. 2011).

Furthermore, there was a significant schizophrenia diagnosis by rs6265 genotype interaction

observed in resting and working-memory related hippocampal regional cerebral blood flow, as

well as on hippocampal prefrontal coupling (Eisenberg, Ianni et al. 2013).

There is evidence to suggest that neuroanatomical changes and neurocognitive dysfunction

within schizophrenia are likely dependent on genetic load. Further, these anatomical changes

may mediate neurocognitive dysfunction. We hypothesize that increasing additive genetic risk

loading may produce a more ‘severe’ brain phenotype that may predict cognitive function.

Furthermore, as we have previously shown, the effect of schizophrenia risk variants on brain

structure may be greater within schizophrenia patients (Lett, Chakavarty et al. 2013). Therefore,

we examine the accrued effect of five common genetic variants, implicated in schizophrenia,

brain structure and cognitive function, for association with brain-wide measures of white matter

fraction anisotropy (FA) and cortical thickness in healthy controls and patients. To compare

genetic subsets with differences in brain structure, we then isolated subjects with either low or

high risk allele loading for association with our neurocognitive battery. Last, we employ a novel

voxel-wise mediation analysis to understand how high risk allele loading explains poorer

cognitive functioning via worse brain structure.

132

6.3 Methods

6.3.1 Participants

Participants from the Toronto imaging-genetics sample were recruited at the Centre for

Addiction and Mental Health (CAMH) in Toronto, Canada, via referrals, study registries, and

advertisements. All clinical assessments occurred at CAMH while DT-MRI scans were

performed at a nearby general hospital. Eighty-nine patients with a diagnosis of schizophrenia or

schizoaffective disorder and 109 healthy control subjects in this sample completed all imaging,

cognitive and genetics protocols. All participants were administered the Structured Clinical

Interview for DSM-IV Disorders (First MB 1995) to determine diagnosis, and were interviewed

by a psychiatrist to ensure diagnostic accuracy. IQ was measured using the Wechsler Test for

Adult Reading (WTAR) (Wechsler 2001) and all participants were screened with the Mini

Mental Status Exam (MMSE) for dementia (Folstein, Folstein et al. 1975) and a urine toxicology

screen. Comorbid physical illness burden was measured by administration of the Clinical

Information Rating Scale for Geriatrics (CIRS-G) (Miller, Paradis et al. 1992). Medication

histories were initially recorded via self-report, and then verified either by the patient’s treating

psychiatrist or chart review. All subjects received urine toxicology screens and anyone with

current substance abuse or any history of substance dependence was excluded. Individuals with

previous head trauma with loss of consciousness, or neurological disorders were also excluded.

A history of a primary psychotic disorder in first-degree relatives was an additional exclusion

criterion for controls. In previous imaging-genetics studies, we have examined MIR137

rs1625579 in healthy controls and schizophrenia, as well as ZNF804A rs1344706 and BDNF

rs6265 in healthy controls (Voineskos, Lerch et al. 2011, Voineskos, Lerch et al. 2011, Lett,

Chakavarty et al. 2013).

133

6.3.2 Image Acquisition

High-resolution axial inversion recovery-prepared spoiled gradient recall MR images were

acquired using a 1.5-T GE Echospeed system (General Electric Medical Systems, Milwaukee,

WI; echo time (TE): 5.3, repetition time (TR): 12.3, time to inversion: 300, flip angle 20,

number of excitations=1; 124 contiguous images, 1.5 mm thickness). For DTI acquisition, a

single-shot spin-echo planar sequence was used with diffusion gradients applied in 23 non-

collinear directions, b=1000 s/mm2, and two b=0 images. Fifty-seven slices were acquired for

whole-brain coverage oblique to the axial plane (2.6 mm isotropic voxels; field of view was

330 mm, 128 × 128 mm2 acquisition matrix; TE=85.5 ms, TR=15 000 ms; the sequence was

repeated three times to improve signal-to-noise ratio).

6.3.3 Cortical Thickness Mapping

All MRIs were submitted to the CIVET pipeline. T1 images were registered to the ICBM152

nonlinear template with a 9-parameter linear transformation, intensity inhomogeneity corrected

(Sled, Zijdenbos et al. 1998) and tissue classified (for grey matter, white matter, and cerebral

spinal fluid) (Zijdenbos, Forghani et al. 2002, Tohka, Zijdenbos et al. 2004). Deformable models

were used to create white and gray matter surfaces for each hemisphere separately, resulting in 4

surfaces of 40,962 vertices each (MacDonald, Kabani et al. 2000, Kim, Singh et al. 2005). From

these surfaces, the t-link metric was derived for determining the distance between the white and

gray surfaces (Lerch and Evans 2005). The thickness data were blurred using a 20-mm surface-

based diffusion kernel in preparation for statistical analyses. Unnormalized, native-space

thickness values were used in all analyses owing to the poor correlation between cortical

thickness and brain volume (Ad-Dab'bagh, Singh et al. 2005).

134

6.3.4 Tract-Based Spatial Statistics (TBSS)

All diffusion tensor imaging (DTI) analysis was done using tools implemented in the FSL toolkit

v.4.1.10 (Smith, Jenkinson et al. 2004). The three repetitions for each subject’s 4D DTI volume

were merged. The images were corrected for motion and eddy current distortion, and averaged.

After skull stripping using BET (Smith, Zhang et al. 2002), fractional anisotropy (FA) images

were created by fitting a tensor model at each voxel using DTIFit. FA quantifies directionality of

water diffusion on a scale from zero (random diffusion) to one (diffusion in one direction).

Voxel-wise analysis of the FA data was carried out using Tract-Based Spatial Statistics (TBSS,

v1.2) (Smith, Jenkinson et al. 2006). TBSS projects all subjects' fractional anisotropy (FA) data

onto a mean FA tract skeleton, before applying voxel-wise cross-subject statistics. FA images

then underwent nonlinear registration to the FMRIB58_FA target image. Next, the mean FA

image was iteratively generated from scans of healthy controls and patients with schizophrenia

separately. Each group was then aligned to MNI 152 standard space using an affine

transformation. An average white matter skeleton was then generated from the mean of all

subjects’ transformed FA images at a threshold of 0.2. For group comparisons, each subject’s FA

data was projected onto the white matter skeleton and voxel-wise statistics were calculated using

randomise (v2.1) with 10,000 permutations.

6.3.5 Neuroimaging Dimension Reduction

To assess the effect of our additive model on general brain structure, we employed a region of

interest (ROI) approach and performed separate factor analyses on average skeletal FA and

average cortical thickness among regions. We extracted the mean cortical thickness from 56

anatomical regions in the LONI Probabilistic Brain Atlas (LPBA40) (Shattuck, Mirza et al.

2008).From the TBSS FA skeleton, we extracted values of mean FA from 50 white-matter

135

regions from the ICBM-DTI-81 atlas (Wakana, Caprihan et al. 2007). We applied separate factor

analyses for the mean skeleton FA and mean cortical thickness values using SPSS (version 15)

across all subjects, and extracted principal components that explained greater than 10% of the

variance.

6.3.6 Genetics

Genotyping of the rs1622579 (MIR137HG), rs1006737 (CACNA1C), rs1344706 (ZNF804A),

rs3749034 (GAD1), and rs6265 (BDNF) SNPs were performed using a standard ABI (Applied

Biosystems Inc.) 5’ nuclease Taqman assay-on-demand protocol in a total volume of 10 µL.

Postamplification products were analyzed on the ABI 7500 Sequence Detection System (ABI,

Foster City, California, USA) and genotype calls were performed manually. Results were

verified independently by laboratory personnel blind to demographic and phenotypic

information.(Lahiri and Nurnberger 1991). Genotyping accuracy was assessed by repeating 10%

of the sample with 100% accordance in genotype calls.

6.3.7 Additive Model

Risk scores for each gene variant were based on previous association with schizophrenia,

neuroimaging phenotypes, and cognitive function. The score range [0, 0.5, 1] corresponded to

the number of risk alleles. For each participant, we then calculated an additive risk score based

on the addition of the risk scores from each variant. The scores were not weighted for each

variant since it is unlikely these variants would affect cortical thickness and white matter FA in

proportion to their associations with schizophrenia.

136

6.3.8 Neuropsychological Assessment

All subjects underwent a battery of cognitive tests administered over approximately 1.5 hours.

This battery includes a wide range of cognitive domains with varying degrees of impairment in

schizophrenia (Rajji, Ismail et al. 2009), and have been previously described (Voineskos, Rajji et

al. 2012). From the Repeatable Battery for the Assessment of Neuropsychological Status

(RBANS) battery, we selected measures of attention (Digit Span), processing efficiency (Digit-

Symbol Coding), immediate memory (Story memory), delayed memory (Story recall),

visuospatial construction (Line Orientation); we further selected standalone measures of verbal

fluency (Controlled oral word association test, COWAT; total words for F+A+S), working

memory (Letter-number span, LNS), processing speed (Trail Making Test A, TMT-A), and

executive function (Trail Making Test B, TMT-B) (Reitan and Wolfson 1985, Ruff, Light et al.

1989, Wechsler 2001, Hale, Hoeppner et al. 2002, Dickinson 2008). To assess motor

functioning, participants underwent test for fine motor speed (Finger tapping; dominant and non-

dominant hand) and visual-motor coordination (Grooved pegboard; dominant and non-dominant

hand) (Halstead 1947, Matthews and Klove 1964).

6.3.9 Statistical Analysis

Analysis of variance (ANOVA), χ2 Goodness-of-fit, and t-tests were performed using SPSS

(Version 15). Adherence to Hardy-Weinberg equilibrium for each marker was assessed using χ2

Goodness-of-fit. To test if our additive risk score was higher in patients with schizophrenia

compared to controls we used a one-way t-test. ANOVAs were performed to test for significant

differences between patients and controls for: age, education, IQ, MMSE, CIRS-G. χ2

Goodness-of-fit tests were used to assess significant differences in: sex, handedness, and

ethnicity (Caucasian versus non-Caucasian). For each brain measure, our independent variables

137

were diagnosis (control or schizophrenia), additive risk score, their interactions and age as a

covariate of no interest. Vertex-wise cortical thickness analyses were performed separately on

left and right cortices, and we applied a false discovery rate (FDR) of q=0.05. TBSS analyses

were corrected for multiple comparisons using family-wise error (FWE).

6.3.10 Voxel-wise mediation analysis

Voxel-wise mediation analysis was performed in MATLAB (R2013b). We used the same

multiple regression approach described by Baron and Kenny (Baron and Kenny 1986), although

we applied this approach across the entire TBSS FA skeleton. A 4D image containing the TBSS

skeleton of the subjects was loaded into MATLAB, and transformed into an array of all non-zero

voxel across each subject. To remove confound of age, we then regressed out the effect of age

for all voxels. Our mediation analysis was accomplished with three regression equations applied

across all voxels. First, we regressed the independent variable (risk group) against white matter

FA. A z-score was produced for each non-zero voxel and was used to produce a 3D image of z-

scores (‘Path A’). We then applied TFCE in the FSL ‘fslmaths’ function with E=2, H=1, and the

neighbourhood-connectivity parameter = 26 as recommend in TBSS analysis (Smith and Nichols

2009). 10,000 permutations (i.e. randomization analysis) were then performed and the max z-

statistics for each permutation was used to assess significance accounting for FWE. Second, we

regressed the mediator variable (white matter FA at each voxel) against cognitive performance at

each voxel (‘Path B’). A 3D image of z-scores were produced, and we tested significance using

the same TFCE and permutation test technique. Third, we regressed the independent variable

(risk group) on cognitive performance (‘Path C’). A significant association in all three sets of

regressions then allowed us to proceed with the Sobel equation to assess the indirect effect of the

independent variable on the dependent variable via the mediator (white matter FA at each voxel).

138

We used the unstandardized regression coefficients (beta) and the standard errors (SE) from

‘Path A’ and ‘Path B’ in order to produce a z-value at each white matter FA voxel (Sobel

equation: z-value = beta(Path A)*beta(Path B)/ √(beta(path B)2 *SE(Path A)

2 + beta(Path A)2 *SE(Path B)

2)). A

3D image of z-values were produced, we applied TFCE, and significant mediation was assessed

using the max z-value of 10,000 permutations. It should be noted that resampling strategies to

assess significance of the Sobel equation are considered to be a better alternative than parameter

tests that impose distribution assumptions (Preacher and Hayes 2008).

6.4 Results

6.4.1 Demographics

Frequencies and distribution of demographic data for our sample are shown in Table 6-2.

Schizophrenia patients were not different than controls with respect to age, IQ, sex, and

handedness, but had lower education, Mini Mental Status Exam (MMSE) score, Cumulative

Illness Rating Scale – Geriatrics (CIRS-G), and proportion of Caucasian subjects (p<0.05).

6.4.2 Genetics

None of the SNPs significantly deviated from Hardy-Weinberg equilibrium in the healthy

controls (p>0.05). There was no significant difference in the frequency for any of the SNPs

between healthy controls and schizophrenia patients, although it should be noted that all markers

were in the direction consistent with previous associations (Table 6-1). Our additive risk score

did not differ significantly between healthy controls (mean= 2.46±0.76) and schizophrenia

patients (mean=2.59±0.75), t196= 1.13, p(1-sided) = 0.131).

139

6.4.3 The effect of additive risk on whole brain measure of cortical thickness and white matter FA

Vertex-wise analysis of cortical thickness revealed a significant score-by-diagnosis interaction

(q<0.05) predominately in the inferior parietal cortex (MNI co-ordinate [47 -73 12]): F1,197=16.5,

p=5.0x10-7) and right insular cortex (MNI co-ordinate [40 -7 6]): F1,197=9.8, p=0.002; Figure 6-

2). There was no main effect of additive risk (q>0.05). Voxel-wise TBSS of white matter FA

showed a prominent main effect of our additive risk score on FA throughout the brain (p<0.05 to

p<0.001) after correcting for family-wise error. Furthermore, there was a significant score-by-

diagnosis interaction predominately in the corpus callosum (P<0.05; Figure 6-1).

6.4.4 The effect of additive risk on general brain structure

To assess the effect of our additive model on general brain structure, we performed separate

factor analyses on average skeletal FA and average cortical thickness among regions. For cortical

thickness, the first principal component, PC1[CT], explained 62.6% of the variance. From the

TBSS FA skeleton the first principal component, PC1[FA], explained 46.5% of the variance. No

other components explained greater than 10% of the variance. PC1[FA] and PC1[CT]

significantly correlated with each other after removing the effect of age (r=0.30, p=1.6x10-5).

There was no significant main effect of additive risk score on PC1[CT] was observed, but a

significant diagnosis-by score interaction (t=2.42, p=0.016; Figure 6-4). Within schizophrenia

patients, risk score predicted 4.9% (R2=0.049, t=2.12, p=0.037) of the variance in PC1[CT].

There was a significant main effect of additive risk score on PC1[FA] (t=2.49 p=0.014; Figure 6-

3), and there was no significant diagnosis-by-score interaction. Within schizophrenia patients,

additive risk score predicted 6.7% (R2=0.067, t=2.50, p=0.014) of the variance in PC1[FA].

140

6.4.5 The effect of extreme additive risk loading on general brain structure and cognitive performance

To examine the impact of multiple simultaneous risk allele hits on brain structure and cognitive

function, we re-analyzed our data including only subjects with high additive risk allele loading

(score>3; n=35) or low risk loading (score=<1.5; n=32). Significant diagnosis-by-score

interaction for PC1[CT] (F1,67=4.79, p=0.032; Figure 6-5), and a significant main effect of high

risk groups interactions were observed for PC1[FA] (F1,67=6.41, p=0.014; Figure 6-6).

The first principal component of our cognitive tasks, general fluid intelligence (gF), predicted

51.3% of the variance in the entire sample (Eigenvalue =4.62). There was no significant main

effect of high additive risk or diagnosis-by-group interactions for gF. In follow-up analyses in

our cognitive tasks composing gF, we did find a significant main effect association with verbal

fluency after Bonferroni correction for eight multiple comparisons (F1,67=9.84, p=0.003), and

nominally significant diagnosis-by-group interaction (F1,67=5.21, p=0.026; Figure 6-7). There

were no other significant associations (Table 6-3). The first principal component of our

psychomotor tasks (PC1[MC]) explained 70.1% of the variance in our entire sample

(Eigenvalue=2.81). There was a significant diagnosis-by-group interaction

(F1,63=10.07,p=0.002), and main effect of high additive risk on PC1[MC] (F1,63=5.77, p=0.020;

Figure 6-8). Because PC1[MC] explains a high degree of the variance, we did not examine these

tasks individually (dominant/non-dominant hand for grooved pegboard and finger-tapping).

6.4.6 Voxel-wise mediation analysis

Since there were significant associations between high risk allele loading and lower FA as well

poorer neurocognitive performance, we employed a novel approach of voxel-wise mediation

analysis on the TBSS skeleton. This allows for the testing of statistical inferences on whether the

141

additive genetics effect on voxel-wise FA is causing the poorer verbal fluency and PC1[MC].

We analyzed only schizophrenia patients since the majority of effects on both cognition and

brain structure occurs in this group (Figure 6-5 to 6-8). High genetic risk loading strongly

predicted lower white matter skeleton FA after family-wise error correction for multiple

comparisons (Figure 6-9A). Poorer white matter skeleton FA predicted worse performance on

the controlled oral word association test (COWAT) for verbal fluency (Figure 6-9B). White

matter skeleton FA mediated the genetic association with verbal fluency over a widespread area,

suggesting that these regions (or FA regions in high correlation) are causally explaining the poor

performance in these schizophrenia patients (Figure 6-9C). For example, high risk allele loading

explained 33% of the variance in verbal fluency (R2=0.33, F1,24=11.12, p=0.003). When we co-

varied for the top associated voxel with high risk allele loading (MNI co-ordinates [108 148 98];

R2=0.50, p=7.5x10-5) the variance explained was 4% (R2=0.04, p=0.36); thus, 29% of the

variance was mediated after correction for family-wise error (PFWEcorrected[Sobel] < 0.001). There

were no significant associations between white matter skeleton FA voxels and PC1[MC] after

correcting for family-wise error; therefore, we did not perform any mediation analysis for motor

coordination.

6.5 Discussion

To our knowledge, this is the first study to apply additive genetic modelling on neuroimaging

phenotypes, and to causally show the effect of the additive risk on cognition using voxel-wide

mediation. Our findings support an important role for additive genetic risk in determining brain

structure and cognitive function within schizophrenia. Increasing number of schizophrenia risk

alleles lead to widespread reductions in FA, and in schizophrenia patients there were reductions

in cortical thickness. Our principal component analyses demonstrated that one latent variable

142

explains a large percentage of the variance in both skeletal FA and cortical thickness. There were

significant main effect and diagnosis-by-score interaction on our principal component for

skeletal FA and cortical thickness, respectively. This suggests that additive risk broadly affects

brain structure, which we found to be especially true within schizophrenia patients. Lastly, our

results show that schizophrenia patients with high risk allele loading may constituent a more

severely impaired subgroup with reductions in white matter FA, cortical thickness motor

coordination and verbal fluency scores. Moreover, we showed a direct statistical relationship

suggesting that the reduced white matter FA in these individuals causes the impairment of verbal

fluency.

Our results show a differential effect of additive genetic risk between healthy controls and

patients with schizophrenia. There was a prominent main effect and diagnosis interaction of

additive risk loading in white FA. Conversely, we only detected an additive risk loading by

diagnosis interaction effect in cortical thickness. In fact, healthy subjects with high additive risk

loading tended to have greater thickness in our principal component analysis. Cortical thickness

reductions have a high genetic contribution, and there are cortical thickness reductions observed

in schizophrenia; however, reductions in thickness there are only trend-level reductions observed

in sibling of patients (Goldman, Pezawas et al. 2009, Panizzon, Fennema-Notestine et al. 2009,

Chen, Fiecas et al. 2013, Xiao, Lui et al. 2013). Thus, cortical thickness may not be a poor

schizophrenia intermediate phenotype. Moreover, cortical thickness has a high degree of

plasticity that may be dependent on schizophrenia treatment (Lett, Voineskos et al. 2014). Poor

outcome patients have more pronounced cortical thinning, and higher intake of antipsychotic

medication correlates with less cortical thinning over time (van Haren, Schnack et al. 2011). It

could be speculated that healthy individuals with high additive risk, may have higher compensate

via greater cortical thickness. Nevertheless, we observed interaction effects in both cortical

143

thickness and white matter FA, suggesting that the effect of additive risk loading is greater

within schizophrenia patients. One potential explanation is that each risk gene included in our

additive model has been shown to be an important modifier of disease. Our previous results

showed that MIR137 had significant effects on white matter integrity, hippocampal volumes, and

lateral ventricle volumes, but only in schizophrenia (Lett, Chakavarty et al. 2013). Moreover in

each gene in our model, diagnosis-by-genotype interactions have been reported in cognitive

function (Hashimoto, Ohi et al. 2010, Walters, Corvin et al. 2010, Green, Cairns et al. 2012,

Hori, Yamamoto et al. 2012), imaging intermediate phenotypes (Addington, Gornick et al. 2004,

Smith, Thornton et al. 2012, Eisenberg, Ianni et al. 2013, Erk, Meyer-Lindenberg et al. 2013),

and genome regulation (Mill, Tang et al. 2008, Bharadwaj, Jiang et al. 2013). Taken together,

our results suggest that additive risk loading may be distinct within schizophrenia in comparison

to controls.

Although each one of the variants included in our model have been reported to have distinct

effects on brain structure and function, we found that an accrual of risk variants led to greater

effect on white matter FA and cortical thickness. The effect of additive risk was predominately

within schizophrenia patients, and therefore, may have important implications in explaining the

high degree of variability in brain structure finding within the disease (Ho, Andreasen et al.

2003, Kubicki, Westin et al. 2005, Kubicki, McCarley et al. 2007). It could be speculated that the

association between one variant and brain structure may be dependent on the additive risk from

other schizophrenia risk variants. Because our model takes the effects of multiple SNPs into

account, it may show effects which are closer to the actual genetic risk on brain structure

compared to analyses of single variants. It should be noted that although our SNPs could be

explain in terms of gene network interactions, none of the SNPs interact directly with each other.

That is, there are only indirect interactions through their gene products. We did find that patients

144

with risk allele loading had robust reductions in cortical thickness and white matter FA.

Importantly, we were able to causally connect this change in white matter FA to poorer verbal

fluency. This has a number of important consequences. First, due to the high variability of brain

structure within and between patients groups, assumptions are made to what constitute poor brain

structure (e.g., lower FA). Our results definitely show that, in our sample, lower FA leads to

worse verbal fluency. Second, while the molecular genetic significance of our findings may

require more study, we do explain some of the neuroanatomical basis for language dysfunction

within schizophrenia.

We were unable to find statistically significant mediation of white matter structure and our motor

coordination first principal component (PC1[MC]). The voxel-wise skeleton FA analysis did not

significantly predict PC1[MC]; therefore, no voxels explain enough of the variance in PC1[MC]

for any meaningful mediation. It is possible that the effect was masked by our stringent

correction for multiple comparisons. Alternatively, other brain imaging modalities, such as

cortical surface area or subcortical volumes, may better correlated with PC1[MC], and thus,

potentially mediate the effect of high additive risk on motor coordination. Unfortunately, we

were unable to perform similar analysis in our vertex-wise analysis of cortical thickness.

This study has a number of important limitations. Many different analyses were conducted;

nevertheless, our study was strictly guided by our a priori hypotheses testing, and rigorous

correction for multiple comparison within each imaging modality. Moreover, our cognitive

analyses was performed in relatively smaller, genetically homogenous subgroups, and thus may

be potentially inconclusive. The fact that verbal fluency findings association did mediate through

brain structure support the validity of our findings. Furthermore, our subgroup analyses does

specifically address our research question. Namely, the impact of simultaneous risk alleles on

145

brain structure and cognitive function. Last, we did not perform gene x gene interaction analysis.

Given that five SNPs were included in our model we would need a much larger sample to have

enough power to detect epistasis. Moreover, empirical and theoretic evidence suggest that

complex traits, such as morphology, are mainly due to additive genetic variance (Hill, Goddard

et al. 2008). Therefore, our approach may be a useful alternative to examining gene interaction

without massive sample sizes. These caveats notwithstanding, the data does suggests that

additive genetic risk has meaning impact on brain structure, and it may lead to reduced cognitive

function. Furthermore, the consequences of high additive risk for schizophrenia may differ in

schizophrenia patients compared to controls, and further research is necessary to explain the

mechanism underlying these differences.

To the best of our knowledge, this is the first study to examine the additive effect of these variant

on cortical structure and cognitive function. Via additive genetic modeling, we identified a

subgroup of patients with schizophrenia characterized by widespread reductions in white matter

FA, cortical thickness, motor coordination, and verbal fluency. Early identification of patients

with neurobiological markers of more severe disease trajectory may lead to better outcome either

through novel treatment tailored to these markers, or identification of patient whom may benefit

form more aggressive treatment strategies. Moreover, our additive modelling approach may

account for core features underlying the diverse features of schizophrenia and other psychiatric

disorders that may be difficult to detect with a single gene variant, and move us toward

biological or molecular subtyping of this heterogeneous disorder.

146

Figure 6-1. Greater additive risk predicts poorer white matter fractional anisotropy.

Greater additive risk score predicted reduced fractional anisotropy across multiple brain regions,

and the effect was larger within schizophrenia patients. Areas coloured from red to yellow

correspond to p-values ranging from 0.05 and lower following correction for multiple

comparisons using family-wise error. Significant regions are mapped onto the standard Montreal

Neurological Institute atlas MN152 1-mm brain template.

147

Figure 6-2. Significant additive score-by-diagnosis interaction for vertex-wide cortical

thickness. Schizophrenia subjects with greater additive risk have reduced cortical thickness.

Areas coloured from blue to yellow correspond to p-values ranging from 0.05 and lower

following FDR correction for multiple comparisons at q=0.05.

148

Figure 6-3. The first principal component (PC1[FA]) of skeleton FA across additive model

scores in schizophrenia patients and healthy controls.

149

Figure 6-4. PC1 Cortical Thickness across additive model scores in schizophrenia patients

and healthy controls.

150

Figure 6-5. High additive risk allele load predicts lower PC1 fractional anisotropy in

schizophrenia patients. HC = Healthy Control; SCZ = Schizophrenia

151

Figure 6-6. High additive risk allele load predicts lower PC1 fractional anisotropy in

schizophrenia patients. HC = Healthy Control; SCZ = Schizophrenia

152

Figure 6-7. High additive risk allele load predicts lower poorer verbal fluency in

schizophrenia patients. HC = Healthy Control; SCZ = Schizophrenia

153

Figure 6-8. High additive risk allele load predicts lower PC1 of motor coordination in

schizophrenia patients. HC = Healthy Control; SCZ = Schizophrenia

154

Figure 6-9. Voxel-wise mediation analysis of verbal fluency in schizophrenia patients. Areas

corresponding from red to yellow correspond to p-values ranging from 0.05 and lower following

family-wise error correction for multiple comparisons. 10000 permutations were performed. (A)

High additive genetic risk loading predicted lower FA across multiple brain regions including the

corpus callosum. (B) White matter FA predicted verbal fluency across multiple white matter

tracts. (C) Significant p-values indicated broad areas of the white matter skeleton FA partially

mediate the effect of high genetics risk loading on cognitive function. We create a z-statistic

from voxel-wise Sobel tests for mediation based on the beta coefficients and standard error from

analyses preformed in A and B. The z-values then undergo threshold free clustering

enhancement, and p-values are derived using permutation testing.

155

156

Table 6-1. Count and frequency of risk alleles by diagnosis

157

Table 6-2. Demographics and Clinical Characteristics

Control (N=109) Schizophrenia (N=89)

Mean SD Mean SD T(196) P

Age (years) 45.62 19.01 45.47 17.42 0.06 0.95

Education 15.35 1.87 13.22 2.82 6.4 <0.05

IQ (WTAR) 111.44 7.85 108.39 16.27 1.7 0.09

MMSE 29.33 0.91 28.67 1.75 3.4 <0.05

CIRSG 1.89 2.00 2.49 2.17 2.0 <0.05

Sex (M) 62 56.90% 58 34.80% χ2 = 1.4 0.24

Handedness (R) 101 92.70% 81 91.00% χ2 = 2.5 0.28

Ethnicity (C) 103 94.50% 70 78.70% χ2 = 11.1 <0.05

Age at Onset (years) 25.15 9.52

Duration (years) 20.35 16.30

AIMS 0.99 2.38

PANSS 53.90 15.21

Positive 14.01 5.70

Negative 14.43 5.89

General 25.45 6.79

AAO, Age at Onset of schizophrenia; AIMS, Abnormal Involuntary Movement Scale; C,

Caucasian; CIRS-G, Cumulative Illness Rating Scale – Geriatrics; DOI – Duration of Illness, F,

Female; M, Male; MMSE, Mini Mental State Examination; PANSS, Positive and Negative

Syndrome Scale; R, Right-handed; WTAR, Wechsler Test of Adult Reading.

158

Table 6-3. The effect of high additive risk allele loading on general fluid intelligence (gF)

and its components.

Neurocognitive Task Group Diagnosis Group*Diagnosis

F(1,64) P F(1,64) P F(1,64) P

gF (PC1) 1.219 0.274 24.535 <0.001 1.301 0.259

COWAT 9.84 0.003 4.86 0.031 5.21 0.026

LNS 2.25 0.139 16.3 <0.001 1.224 0.273

TMT-A 0.079 0.779 10.716 0.002 0.755 0.388

TMT-B 0.031 0.862 10.273 0.002 0.067 0.796

Digit Span 2.435 0.124 5.892 0.018 0.474 0.494

Digit Symbol Coding 0.299 0.587 17.096 <0.001 1.108 0.297

Story Memory 0.033 0.857 9.837 0.003 0.234 0.631

Story Recall 0.303 0.584 18.31 <0.001 0.003 0.954

Line Orientation 0.741 0.393 5.24 0.026 0.875 0.353

159

Chapter 7

7 General Discussion & Future Direction

7.1 Summary of Results

The experiments presented in the previous chapters constitute several key findings. We

demonstrated that common genetic variants in the NRXN1 gene impact brain volume and

sensorimotor functioning integral to schizophrenia and other psychiatric disorders. Next, we

showed that the MIR137 risk variant robustly predicts across four independent sample the age-at-

onset of psychosis, a predictor of clinical outcome and cognitive function within schizophrenia.

The MIR137 risk variant also predicted neuroanatomical correlates of schizophrenia severity

(lateral ventricle and hippocampal volumes). Moreover, MIR137 was associated with worse

white matter FA across the lifespan. Importantly, patients carrying the protective genotype were

no different than healthy controls in cortical structure. Therefore, we demonstrated the MIR137

gene that influences neural development and regulates other strongly associated schizophrenia

risk variants plays a sizeable role in explaining heterogeneity among schizophrenia patients via

age-at-onset and brain structure. In the next project, we demonstrated that the GAD1 risk variant

was associated with verbal working memory, non-verbal working memory, and selective

attention among schizophrenia patients and healthy controls. GAD1 was also associated with

white matter FA in the prefrontal cortex. Furthermore, our results suggest that the GAD1

association with FA carries functional relevance to non-verbal working memory performance

using our voxel-wise mediation analysis. Thus, providing a potentially casual mechanism

through which a schizophrenia candidate gene influences some of the variance of working

memory functioning within the disorder. We then successfully demonstrated that additive genetic

160

risk among five genes associated with schizophrenia heterogeneity strongly predicts cortical

thickness and white matter FA, particularly within schizophrenia patients. Using PCA on whole

brain measures of cortical thickness and white matter FA, additive risk was associated with core

changes in brain structure that provide a potential mechanism through which multiple common

risk variants enact a general effect on brain structure. Comparing genetically distinct

schizophrenia subgroups based on polygenic risk loading, we found that high risk loading

predicted reduced motor functioning, and poorer verbal fluency performance. Therefore, our

additive risk model may serve as a paradigm in which genetics can help reveal heterogeneity

within schizophrenia that may well predict different disease trajectory.

In summary, the results of the presented studies demonstrate that variation in schizophrenia risk

genes has appreciable effects on brain structure that may well affect cognitive function and

clinical heterogeneity within the disorder.

7.2 Can Imaging-genetics Dissect Clinically Meaningful Heterogeneity within Schizophrenia?

When imaging-genetics strategies were first developed over ten years ago it held tremendous

promise to better understanding the etiology of schizophrenia, and thus, provide better

description of the novel neurophysiological treatment. That is, characterizing the neural systems

affected by risk gene variants to understand quantitative measure of brain structure and function

related to psychiatric disease could provide novel treatment avenues via: novel molecular targets

of medication or better neuroanatomical outcome factors more specifically identifying patients

than may respond to current treatments. Importantly, imaging-genetics can target specific core

symptoms of schizophrenia, such as working memory deficits, in which there is arguably little

amelioration of dysfunction with antipsychotics treatment (Lett, Voineskos et al. 2014).

161

Unfortunately to date, neurobiological research has not translated into novel treatment options in

psychiatric research have become standard of care.

The intermediate phenotype approach has been the dominant imaging-genetics method in

psychiatry. In the most common application, genetic variants associated with a disorder are

examined in healthy controls. Since common variants are present in healthy controls,

intermediate phenotypes should provide subclinical changes in brain structure and function

outside of confounds of the psychiatric disorder (e.g. socioeconomic status, medication, and/or

comorbidities). We have successfully used this approach with NRXN1, GAD1, and our additive

polygenic model. However, it could be argued that even in the presence of confounding factors

of schizophrenia (antipsychotic treatment, co-morbidities, stressful life events, and others),

meaningful imaging-genetic data can be derived in patient populations. Indeed, MIR137 had a

relatively small effect on brain structure in health controls, but in patients with schizophrenia the

risk genotype predicted early age-at-onset, larger lateral ventricles, lower hippocampal volume,

and lower white matter FA throughout the brain. This genotype-by-diagnosis interaction may be

due to the effect of microRNA-137 as a regulator of neurodevelopmental genes, epigenetics

machinery, and other genes strongly associated with schizophrenia liability genes (See Chapter

4.2). Importantly, the findings suggest the effect of common gene variants on brain structure and

cognitive function may be distinct within the disorder. It is possible the effect of a single risk

variant may be dependent on the presence of other risk variants. Our additive risk model

demonstrated some effects across healthy controls and schizophrenia patients although the

majority of effect was within schizophrenia. The effect sizes of white matter FA and cognitive

associations were greater in patients. Moreover, the association between additive risk loading

and cortical thickness was in opposite direction between controls and patients. The polygenic

risk by diagnosis interaction may be partially explained by the genetic contribution to cortical

162

thickness. There is a high degree of cortical thickness heritability among healthy co-twins; thus,

likely determined through genetic variation (Panizzon, Fennema-Notestine et al. 2009, Rimol,

Panizzon et al. 2010). Among co-twins discordant for schizophrenia, cortical thickness

reductions observed in affected twins are not prevalent in the healthy sibling (Goldman, Pezawas

et al. 2009, Panizzon, Fennema-Notestine et al. 2009, Rimol, Panizzon et al. 2010) suggesting

that cortical thickness may be a poor intermediate phenotype. Alternatively, since cortical

thickness carries a high degree of plasticity it is also possible that the increased cortical thickness

may be a protective mechanism in healthy controls with high risk allele loading. Taken together,

our findings suggest that the intermediate phenotype approach (i.e., in healthy controls), although

useful, may be missing clinically relevant molecular genetic-by-diagnosis interactions.

Our results also suggest that the relationship between schizophrenia risk variants and brain

structure may be dependent on the presence of other schizophrenia risk variants. The MIR137

gene product functionally regulates other schizophrenia risk genes including NRXN1, CACNA1C,

and ZNF804A (Kwon, Wang et al. 2011, Kim, Parker et al. 2012); however, the functional

relevance of the rs1625579 variant is unclear. Furthermore, the association between disrupted

mir-137 expression and schizophrenia is controversial (Guella, Sequeira et al. 2013, Wright,

Turner et al. 2013). It should be noted though that mir-132, an upstream regulator of mir-137, is

associated with schizophrenia (Miller, Zeier et al. 2012). Therefore, the relationship between

rs1625779 genotype and the regulation of schizophrenia associated genes by mir-137 needs

further research. Alternatively, it could be hypothesized that the cumulative (or additive) effect

of the rs1625579 MIR137 variant and other common risk variants in neurodevelopmental gene

pathway may be driving the strong association within schizophrenia. Our additive modelling

results suggest that indeed while any one of the risk variants may independently have different

associations with brain structure, together they may have broad effects on the brain since

163

polygenic risk predicted the first principal components of cortical thickness, and white matter

FA. Moreover, the results in voxel-wise FA mediation analysis suggest additive risk loading may

be clinical relevant in understanding how genetic effect on brain structure lead to verbal fluency

dysfunction in schizophrenia patients. In summary, the relative importance and the effect on the

brain of gene variants may be different in schizophrenia patients compared to healthy controls.

7.3 What Benefits does Translational Research Address?

Clinical diagnosis based on DSM and ICD currently relies on presenting symptoms that may not

reflect the neurobiology. It has been suggested that the boundaries between different diagnoses

fail to align with the clinical neuroscience and genetics. For example, the genetic effects shared

between schizophrenia and bipolar disorder are greater than the genetic effects differentiating the

disorders (Lichtenstein, Yip et al. 2009). Moreover, current diagnostic categories have not been

predictive of treatment response. One key strategy is to use translational research to describe one

core component of disease pathology or neural circuit that can be used to drive further research

or novel treatment targets. The best example is the NIMH Research Domain Criteria (RDoC)

initiative which addresses some these concerns by developing novel ways of classifying mental

disorders based on clinical neuroscience and genetics (Insel, Cuthbert et al. 2010, Cuthbert and

Insel 2013). Translational research among imaging, genetic, and neurocognitive data allows

modelling complex relation that may provide novel findings.

An important benefit would be driving new treatment strategies, and target individuals that may

better respond to treatment. Our NRXN1 findings suggest an intermediate phenotype with

disrupted sensorimotor function and frontal white matter volume that overlaps dysfunction

common to ASD and schizophrenia (See Chapter 3.4). In schizophrenia subjects, the rs1045881

NRXN1 marker was also associated with response to the antipsychotic medication clozapine in

164

schizophrenia, and interestingly, this result was driven by improvement of negative symptoms

(Lett, Tiwari et al. 2011). Previous structural MRI brain imaging findings have reported that a

generalized reduction in frontal lobe white matter correlates with greater severity of negative

symptoms (Sanfilipo, Lafargue et al. 2000, Wolkin, Choi et al. 2003, Voineskos, Foussias et al.

2013). Thus, our NRXN1 imaging intermediate phenotype may be applicable to the treatment of

negative symptoms in schizophrenia. Given these findings are correlative it difficult to assess,

with confidence, if there is a direct relationship between the genetics effect on white matter and

negative symptoms within schizophrenia.

It may be necessary to not only describe heterogeneous phenotypes within schizophrenia, but

also to relate the genetic effect among different phenotypes. For example, we have shown that

the GAD1 rs3749034 risk variant, known to be a predictor of GABA levels in vivo, leads to

lower prefrontal white matter integrity and, at least statistically, causing working memory

dysfunction (See Chapter 5.5). Our voxel-wise mediation analysis importantly takes into account

variation in the entire white matter FA skeleton. Most imaging genetic modalities are highly

correlated; therefore, association in region of interest approaches may well due to the true

association in other regions of the brain. Our GAD1 voxel-wise mediation analysis identifies all

regions associated with working memory dysfunction. Therefore, this approach potentially

allows attachment of functional significance of cognitive performance tasks to white matter FA

imaging-genetics data. Since GAD1 is associated with lower FA that leads to impaired working

memory performance, this may be integral to future studies in which prefrontal white matter FA

could be used an outcome measure or early detector of treatment response (such as a GABAegic

pharmacological agent (Lett, Voineskos et al. 2014)).

165

7.4 Can Imaging-genetics explain enough of the Heterogeneity to Guide Treatment Decisions?

The integration of research on neuroimaging and genetics holds tremendous promise for

improving outcome in schizophrenia by facilitating: (a) development of novel therapies based on

improved understanding of pathogenic mechanisms underlying core symptoms of schizophrenia,

and (b) more sensitive measure of treatment response in genetically defined patients groups.

However to date, common risk gene variants only have shown limited effect sizes explaining

disease association (Ripke, Sanders et al. 2011, Ripke, O'Dushlaine et al. 2013), and the same

has been largely true for neuroimaging phenotypes and cognitive phenotypes relevant to

schizophrenia (Greenwood, Braff et al. 2007, Brans, van Haren et al. 2008, Lencz, Knowles et al.

2014). It could be suggested that for clinically meaning for associations that a greater proportion

of the variance needs to explained. In the NRXN1 study (See Chapter 3.4) we were only able to

explain some proportion of the variation with rs1045581 risk allele homozygotes predicting

approximately 6% reduction in frontal lobe white matter volume in healthy controls. In contrast,

our MIR137 results (See Chapter 4.4.4) showed that within schizophrenia patients the risk

genotype predicted 9% of the variance in average white matter FA across the lifespan, and the

effect on age-at-onset of psychotic symptoms was approximately double that of sex (a well-

established predictor of age-at-onset). Our additive model results indicate that a clinically

relevant subgroup could be described based on polygenic risk. These results suggest enough of

the variance can be explained, but may be necessary for prospective studies to test the clinical

application of an imaging-genetics based algorithm for prediction of long term outcome in

schizophrenia.

166

7.5 Limitations

Beyond the specific limitations discussion in each chapter, there are a number of important

limitations with the approaches and findings that should be considered. Most important is that

our positive findings require replication. Imaging-genetics studies has previously tended to have

sample sizes that are generally small, leading to low power and inflated effect size of genetic

association in discovery samples (Button, Ioannidis et al. 2013). For example, Munafo et al.

demonstrate that the effects of the 5-HTTLPR (serotonin-transporter-linked polymorphic region)

on amygdala activation in the discovery samples are usually much higher than in any replication

(Munafo, Freimer et al. 2009). It is possible that this “winner’s curse” occurred with our NRXN1

findings (Zollner and Pritchard 2007, Ioannidis 2008, Kraft 2008). Furthermore, considering the

modest sample size, our genotypic associations with brain volumes would be under-powered in

NRXN1 SNPs with a minor allele frequency less than 15%. Consequently, we excluded these

SNPs from our analyses to reduce penalties for multiple comparison. However, in all of

subsequent imaging analyses of this thesis our sample sizes were comparatively large (n~200).

Moreover, in each study there has been convergent evidence from clinical or cognitive

phenotypes to support the hypothesis. For example, our MIR137 age-at-onset findings were in

the same direction over four independent samples suggesting a true effect. The voxel-wise

mediation approach employed in final two studies (See Chapter 5.4 and Chapter 6.4) provides

even greater support since it is unlikely genetics association with a whole brain voxel-wise

measure would mediate the cognitive effects since (i) the stringent multiple comparison

correction performed, and (ii) the direction of the effect in the mediation model. The latter is

particularly important because it means the mediation not correlation (or moderation) between

our cognitive variables and white matter FA was not driving the effects. Therefore, we can state

167

that within our own sample the genetic association with skeleton white matter FA structure has

functional relevance in cognitive performance.

Another important limitation was that we did not link the functional relevance of the genetic

variants via altered gene expression and cell biology. While some of the variant we examined

have well established effects on gene expression or function (e.g. BDNF rs6265 variant), it is

difficult to assess function on the genome-wide identified variants. Therefore, it is possible that

these variants are not be affecting expression or function of the gene products, rather it may

variant in linkage disequilibrium with genome-wide identified variant. This has a number of

important consequences. First, we have made an assumption that proximity of the variant to the

gene suggests a functional role although it may be associated with another gene. For example,

the MIR137 rs1625579 has SNPs in linkage disequilibrium (R2>0.8) over a region of

approximately 13,000 base pairs included in MIR137HG which codes for two microRNAs (mir-

137 and mir-2682). Second, we may be selecting the wrong variant thereby introducing noise

into our analysis leading to type II error. Third, it is difficult to ascertain how the action of the

variant leadings to dysfunction, such whether more or less expression leads to increased

schizophrenia liability. The Encyclopedia of DNA Elements (ENCODE) project should clarify

the functional relevance of many of the genome wide findings, and provide more target avenues

for in vitro analyses (Consortium, Bernstein et al. 2012). It is important to note that imaging-

genetic findings may provide some functional significance. For instance, we independently show

that rs16225579 MIR137 risk homozygotes lead to poorer brain structure and earlier age-at-

onset. Moreover, our voxel-wise mediation analyses indicate that even though may not have the

true risk locus, we do explain enough of the variance for genetic association to be useful in

explaining brain function.

168

The last three studies in this thesis examined imaging-genetic associations within patients with

schizophrenia. Antipsychotic mediation may be an important confounder of our imaging findings

since it may have both direct and indirect consequences on brain structure. Grey matter loss,

higher neuronal density, and reduced glial cell number similar to what is histologically observed

in schizophrenia was reported in non-human primates exposed to olanzapine or haloperidol over

a two year period (Dorph-Petersen, Pierri et al. 2005, Konopaske, Dorph-Petersen et al. 2007).

Furthermore, longitudinal first episode schizophrenia study showed progressive decline of white

and gray matter volume correlating with antipsychotic medication dose (Ho, Andreasen et al.

2011). Longitudinal reductions in hippocampal volume and BDNF levels has also been reported

after eight months of antipsychotic treatment in first episode patients (Rizos, Papathanasiou et al.

2014). The effect of medication on white matter FA is more controversial with reports of

increase (Garver, Holcomb et al. 2008) and decrease FA with treatment (Wang, Cheung et al.

2013). Furthermore, antipsychotic medication causes significant weight gain in patients with

schizophrenia (Lett, Wallace et al. 2012), which may have significant impact on white matter FA

(Kuswanto, Sum et al. 2014). However, the lack of longitudinal, within-subject studies suggests

that further research is necessary. We found that chlorpromazine equivalents were not

significantly difference between groups. It would be expected the medication effects would

confound our results, and therefore not necessarily invalidate our significant findings. Moreover,

it is unlikely that the confounding effect of medication would be greater than the large effect size

of some of results (e.g. MIR137). Nevertheless, we cannot be certain if our genetics by disease

interactions may in part due to pharmacogenetic interactions, but our findings would arguably

remain just as clinically relevant.

169

7.6 Future Directions

Imaging-genetics of schizophrenia is a dynamic field. Despite the aforementioned limitations it

has provided promising insights into how genetics may influence structure and function relevant

to the etiology and treatment of schizophrenia. Genome-wide identified variants may be of

especially high importance, since these schizophrenia risk variants might provide deeper insight

into the genetic mechanisms of phenotypic heterogeneity in the context of psychiatric disorders.

Future imaging-genetics studies will benefit from better description at the gene function (e.g.

regulation, expression, and interactions) and the brain function (e.g. connectivity, plasticity, and

relationship to behavior). The following section will discuss some of the recent advances in

which translational research may lead ultimately lead to novel treatment options.

7.6.1 Functional Relevance of Genetic Variation

Major leaps in increasing quality and affordability of newer sequencing technologies permit

unprecedented detail in the genetic variability. At the genomic level, whole-genome sequencing

is becoming affordable (currently $1000 USD per subject) with a relatively high degree of

coverage (average of 30 times), and such analyses are leading to discoveries of rare disruptive

mutations in synaptic pathways underlying schizophrenia (Fromer, Pocklington et al. 2014,

Purcell, Moran et al. 2014). The amount of information collected from each genome

(approximately three billion base pairs) and the complexity of voxel-wise brain imaging

approaches (e.g., 140,000 voxel for each skeleton FA) poses significant challenges in

multivariate analysis. One potential method could be to apply an additive risk model across a

gene network (e.g., neuronal plasticity, calcium regulation) in which sequencing data has

identified rare, functional variants and common risk associated variants. Next, this genetic

170

information could be applied as a score to brain imaging phenotypes and symptom profiles to

provide a potential pathophysiological link between gene networks and phenotypes of interest.

One of the major concerns of genomic information is understanding the functional significance

of this static variation on the dynamic transciptome. RNA-sequencing and epigenetic methods

that have been very successfully applied to other domains such as cancer research are inherently

difficult in psychiatric research because of the challenges involved in sampling brain tissue.

These techniques are being applied to postmortem brain samples of healthy individuals across

the life span including: Brain Cloud (Colantuoni, Lipska et al. 2011, Numata, Ye et al. 2012),

Human Brain Transcriptome (Kang, Kawasawa et al. 2011), and the BrainSpan project

(http://www.brainspan.org/). These freely accessible resources provide invaluable information

regarding the timing of gene expression in the brain and the impact of genetic variants of interest

although there are caveats. First, epigenetic profile and RNA expression pattern vary

considerably, and the neurophysiology of the brain tissue can change dramatically depending on

the post-mortem interval and the pH of the tissue (Pidsley and Mill 2011). Second, in the

majority of the adult post-mortem samples clinical and neurocognitive assessments or in MRI

imaging may not have been collected. Third, most of the samples are not within clinical

populations. It is reasonable to suggest that the dynamic regulation of gene expression is

different between healthy controls and schizophrenia patients; therefore, the impact of any

variant may be dependent on diagnosis.

7.6.2 In Silico Prediction of SNP Function: Insight from ENCODE

It could be argued that our imaging-genetic analyses provide some degree of functional

relevance to significantly associated SNPs. However, establishing how these variants impact

gene expression would provide broader understanding of their impact in gene systems. We were

171

able to demonstrate that NRXN1 rs1045881 likely affects binding of mir-339-5p, and thus may

affect expression of NRXN1. Furthermore, GAD1 rs3749034 has been associated with reduced

expression in vivo and in vitro. The BDNF rs6265 is a missense variant changing Valine to

Methionine at amino acid 66. In contrast, the functional relevance of the genome-wide identified

variants is poorly understood. It is unlikely that the top associated GWAS variants directly affect

gene expression, rather it may be another variant that is LD. Consequently, these SNPs could be

introducing noise into imaging-genetics analyses since the true association may be another SNP.

Recently, the Encyclopedia of DNA Elements (ENCODE) project is providing additional

insights into the true associated SNP via in silico prediction of protein binding. In silico analysis

using regulomeDB examining all SNPs with an R2 value greater than 60% (1000 genome phase

1, EUR population) from the GAD1, CACNA1C, MIR137, and ZNF804A markers reveals some

putative role in the expression of the co-localized genes. The GAD1 rs376255 variant likely

affects the binding of the repressor protein CTCF, and is in LD (R2=0.99) and 1080 bp away

from the GAD1 rs3749034. The putative relevance of the GWAS identified markers is less clear.

The MIR137 rs1625579 shares considerable variance with other two markers. 74% of the

variance (R2=0.74) in rs1625579 is explained by rs9324383 which is 17 kbp away. The

rs9324383 likely affects the binding of the transcription factor GATA2. The effect of GATA2 is

particularly interesting since the putative regulator of MIR137 and schizophrenia risk factor miR-

132 also regulates GATA2 expression (Miller, Zeier et al. 2012). 64% of the variance is also

shared by the rs4294451 MIR137 SNP that is 107 kbp away and likely affects the binding of

HNF4A, JUND, and P300. The rs1006737 marker also shared considerable variance with two

SNPs: rs139758774 and rs7308351. 62% of the variance at rs10066737 is explained by

rs139758774 that is 22 kbp away and affects the binding of the immune response transfactor

TRIM28. Further, 61% of variance is explained by rs7308351 that is 42 kbp away, and likely

172

affects the binding of ZNF263. Last, there were no markers sharing more than 60% of the

variance with ZNF804A rs1344706 with any putative functional relevance. Potentially we would

have greater power in our imaging-genetic analysis by examining these putatively functional

markers. Nevertheless, in vitro analyses are necessary to confirm the functional influence of

these markers.

7.6.3 Combining in vivo Biomarkers

Structural imaging can provide “a window” through which neural connectivity can be assessed;

however, other methods may be better at understanding how circuits function. Cortical inhibitory

tone can be measured through combining transcranial magnetic stimulation (TMS) and

electroencephalogram (EEG) (Daskalakis, Christensen et al. 2002). Impaired network

synchronized activity has been reported in schizophrenia patients (Spencer, Nestor et al. 2003,

Spencer, Nestor et al. 2004, Spencer, Salisbury et al. 2008). In particular, long interval

intracortical inhibition (LICI; a measure closely associated with GABAB receptor

neurotransmission) at the DLPFC has been strongly correlated with performance on the N-back

(r=0.63, p=0.04) and LNS (r=0.68, p=0.005) working memory tasks in healthy controls

(Daskalakis, Farzan et al. 2008, Hoppenbrouwers, De Jesus et al. 2012). This is in line with

recent findings that age related working memory decline in rats correlates with GABAB receptor

expression, and the decline is reversible with administration of the CGP55845 GABAB receptor

antagonist (Banuelos, Beas et al. 2014). Furthermore, we recently found that LICI strongly

predict general fluid intelligence (gF), and in particular, working memory in healthy controls

(unpublished data; LNS: F1,22 =12.92, p=0.002; Digit-Span: F1,22 = 16.57, p=7.0x10-4). We also

found that frontothalamic FA was strongly associated with working memory only after covarying

for LICI and our overall model predicted approximately 65% of variance in working memory

173

(unpublished data; R2=0.65, F3,18=11.0, p=2.4x10-4). These results along with our GAD1 findings

(See Chapter 5) suggest multiple lines of evidence pointing to aberrant GABAergic signaling

potentiating working memory dysfunction in schizophrenia patients. Our findings also

demonstrate that combining in vivo biomarkers is an efficient strategy to dissecting complex

neural circuitry that may be at risk in psychiatric disorders. We are currently expanding our

TMS-EEG sample for genetic analysis. One important question is whether cortical inhibition and

white matter structure mediates the association between genetic variants and heterogeneous

phenotypes relevant to schizophrenia.

7.6.4 Towards Neurobiological Treatment

One of the major challenges in neuropsychiatric research is to identify biomarkers predicting or

directing better treatment response for evidence based decision making on treatment options. The

integration of research on neuroimaging and genetics holds tremendous promise for improving

outcome in schizophrenia by facilitating: (a) development of novel therapies based on improved

understanding of pathogenic mechanisms underlying core symptoms of schizophrenia, and (b)

more sensitive measure of treatment response in genetically defined patients groups. For

example, pharmacogenetic and non-pharmaceutical approaches (e.g., repetitive transcranial

magnetic stimulation (rTMS), cognitive remediation therapy (CRT), psychoeducation, cognitive

behavioral therapy (CBT)) could be employed in congruence with imaging methods.

Our GAD1 findings may be of particular relevance to neurobiological-guided treatment as they

point to aberrant GABAergic signaling potentiating PFC white matter structure changes and

working memory dysfunction relevant to schizophrenia (See Chapter 5). An inverse relationship

between FA in the genu of the corpus callosum and TMS-induced interhemispheric signal

propagation suggest functional asymmetry of the DLPFC depends white matter structure

174

(Voineskos, Farzan et al. 2010). TMS-induced LICI leads to suppression of gamma-band

oscillations in the DLPFC (Farzan, Barr et al. 2009). Also, rTMS treatment to the DLPFC

normalizes differences in gamma-band oscillatory activity between schizophrenia patients and

controls (Barr, Farzan et al. 2011). A recent 4-week, randomized, double-blind pilot study

suggests that rTMS to the DLPFC is effective at improving working memory performance in

schizophrenia patients (Barr, Farzan et al. 2013), and these findings are currently being followed-

up in large clinical trial. Considering the effects of GAD1 genotype on DLPFC inhibitory

function and working memory, it may be an important predictor of response to rTMS treatment

in schizophrenia patients. Furthermore, rTMS treatment has been shown to increase white matter

FA (Allendorfer, Storrs et al. 2012, Peng, Zheng et al. 2012). Therefore, changes in prefrontal

FA may well be a clinically relevant marker of plasticity, and potentially an important mediator

of response to rTMS treatment.

7.6.5 Conclusion

The combination of neuroimaging and genetics is a powerful strategy to parse out schizophrenia

heterogeneity. It provides clues into the neurobiological mechanism underlying psychiatric

disorders through characterization of at risk structures present in patients and controls. The

results of this thesis suggest that imaging-genetics can describe clinically meaningful

schizophrenia heterogeneity. Several previously identified variants in schizophrenia risk genes

explained significant differences in core features of the disorder and related brain abnormalities.

The multimodal approach of combing genetics, brain imaging, and clinical characterization was

more informative than any of the approaches alone in these studies. In the near future, it is likely

that this integrated approach will lead to better prognostic markers, and novel therapeutic

strategy for this complex and devastating disorder.

175

References

Abel, K. M., et al. (2010). "Sex differences in schizophrenia." Int Rev Psychiatry 22(5): 417-428.

Acosta, M. T. and P. L. Pearl (2004). "Imaging data in autism: from structure to malfunction."

Seminars in Pediatric Neurology 11(3): 205-213.

Ad-Dab'bagh, Y., et al. (2006). The CIVET Image-Processing Environment: A Fully Automated

Comprehensive Pipeline for Anatomical Neuroimaging Research. 12th Annual Meeting of the

Organization of Human Brain Mapping, Florence, Italy.

Ad-Dab'bagh, Y., et al. (2005). Native space cortical thickness measurement and the absence of

correlation to cerebral volume. Organization of Human Brain Mapping, Toronto, Neuroimage.

Addington, A., et al. (2004). "GAD1 (2q31. 1), which encodes glutamic acid decarboxylase

(GAD67), is associated with childhood-onset schizophrenia and cortical gray matter volume

loss." Molecular psychiatry 10(6): 581-588.

Allendorfer, J. B., et al. (2012). "Changes in white matter integrity follow excitatory rTMS

treatment of post-stroke aphasia." Restor Neurol Neurosci 30(2): 103-113.

American Psychiatric Association, A. (1994). "Diagnostic and statistical manual of mental

disorders."

Anderson‐Schmidt, H., et al. (2013). "Selected rapporteur summaries from the XX world

congress of psychiatric genetics, Hamburg, Germany, october 14–18, 2012." American Journal

of Medical Genetics Part B: Neuropsychiatric Genetics 162(2): 96-121.

Andreasen, N. C., et al. (2012). "Statistical epistasis and progressive brain change in

schizophrenia: an approach for examining the relationships between multiple genes." Mol

Psychiatry 17(11): 1093-1102.

Arnold, S. E. (1999). "Neurodevelopmental abnormalities in schizophrenia: insights from

neuropathology." Dev Psychopathol 11(3): 439-456.

Babulas, V., et al. (2006). "Prenatal exposure to maternal genital and reproductive infections and

adult schizophrenia." Am J Psychiatry 163(5): 927-929.

Bading, H. (2013). "Nuclear calcium signalling in the regulation of brain function." Nat Rev

Neurosci 14(9): 593-608.

Ballageer, T., et al. (2005). "Is adolescent-onset first-episode psychosis different from adult

onset?" J Am Acad Child Adolesc Psychiatry 44(8): 782-789.

Banuelos, C., et al. (2014). "Prefrontal cortical GABAergic dysfunction contributes to age-

related working memory impairment." J Neurosci 34(10): 3457-3466.

176

Barbato, A. and W. H. Organization (1998). Schizophrenia and public health, Division of Mental

Health and Prevention of Substance Abuse, World Health Organization.

Barch, D. M. and E. Smith (2008). "The cognitive neuroscience of working memory: relevance

to CNTRICS and schizophrenia." Biol Psychiatry 64(1): 11-17.

Barnea-Goraly, N., et al. (2004). "White matter structure in autism: preliminary evidence from

diffusion tensor imaging." Biological Psychiatry 55(3): 323-326.

Barnea-Goraly, N., et al. (2005). "White Matter Development During Childhood and

Adolescence: A Cross-sectional Diffusion Tensor Imaging Study." Cerebral Cortex 15(12): 1848

-1854.

Baron, R. M. and D. A. Kenny (1986). "The moderator-mediator variable distinction in social

psychological research: conceptual, strategic, and statistical considerations." J Pers Soc Psychol

51(6): 1173-1182.

Barr, M. S., et al. (2011). "The effect of repetitive transcranial magnetic stimulation on gamma

oscillatory activity in schizophrenia." PLoS One 6(7): e22627.

Barr, M. S., et al. (2013). "Can repetitive magnetic stimulation improve cognition in

schizophrenia? Pilot data from a randomized controlled trial." Biol Psychiatry 73(6): 510-517.

Barrantes-Vidal, N., et al. (2007). "Working memory in siblings of schizophrenia patients."

Schizophr Res 95(1-3): 70-75.

Barrett, J. C., et al. (2005). "Haploview: analysis and visualization of LD and haplotype maps."

Bioinformatics (Oxford, England) 21(2): 263-265.

Basser, P. J., et al. (1994). "MR diffusion tensor spectroscopy and imaging." Biophys J 66(1):

259-267.

Baumgartel, K. and I. M. Mansuy (2012). "Neural functions of calcineurin in synaptic plasticity

and memory." Learn Mem 19(9): 375-384.

Beaulieu, C. (2002). "The basis of anisotropic water diffusion in the nervous system - a technical

review." NMR Biomed 15(7-8): 435-455.

Benes, F. M., et al. (1986). "Quantitative cytoarchitectural studies of the cerebral cortex of

schizophrenics." Arch Gen Psychiatry 43(1): 31-35.

Benson, D. L., et al. (1994). "Activity-dependent changes in GAD and preprotachykinin mRNAs

in visual cortex of adult monkeys." Cereb Cortex 4(1): 40-51.

Bergen, S. E., et al. (2014). "Genetic modifiers and subtypes in schizophrenia: investigations of

age at onset, severity, sex and family history." Schizophr Res 154(1-3): 48-53.

Beveridge, N., et al. (2009). "Schizophrenia is associated with an increase in cortical microRNA

biogenesis." Molecular psychiatry 15(12): 1176-1189.

177

Bharadwaj, R., et al. (2013). "Conserved chromosome 2q31 conformations are associated with

transcriptional regulation of GAD1 GABA synthesis enzyme and altered in prefrontal cortex of

subjects with schizophrenia." J Neurosci 33(29): 11839-11851.

Bierut, L. J., et al. (2007). "Novel genes identified in a high-density genome wide association

study for nicotine dependence." Human Molecular Genetics 16(1): 24-35.

Bigos, K. L., et al. (2010). "Genetic variation in CACNA1C affects brain circuitries related to

mental illness." Archives of general psychiatry 67(9): 939.

Bis, J. C., et al. (2012). "Common variants at 12q14 and 12q24 are associated with hippocampal

volume." Nat Genet 44(5): 545-551.

Bleuler, E. (1950). "Dementia praecox or the group of schizophrenias."

Bohlken, M. M., et al. (2014). "Heritability of structural brain network topology: A DTI study of

156 twins." Hum Brain Mapp.

Boulanger, L. M. (2009). "Immune proteins in brain development and synaptic plasticity."

Neuron 64(1): 93-109.

Braff, D. L., et al. (2007). "Deconstructing schizophrenia: an overview of the use of

endophenotypes in order to understand a complex disorder." Schizophr Bull 33(1): 21-32.

Braff, D. L. and M. A. Geyer (1990). "Sensorimotor gating and schizophrenia. Human and

animal model studies." Arch Gen Psychiatry 47(2): 181-188.

Brambilla, P., et al. (2003). "Brain anatomy and development in autism: review of structural

MRI studies." Brain Res Bull 61(6): 557-569.

Brans, R. G., et al. (2008). "Heritability of changes in brain volume over time in twin pairs

discordant for schizophrenia." Arch Gen Psychiatry 65(11): 1259-1268.

Brauns, S., et al. (2013). "Genetic variation in GAD1 is associated with cortical thickness in the

parahippocampal gyrus." J Psychiatr Res 47(7): 872-879.

Brickman, A. M., et al. (2004). "Thalamus size and outcome in schizophrenia." Schizophrenia

Research 71(2-3): 473-484.

Brouwer, R. M., et al. (2010). "Heritability of DTI and MTR in nine-year-old children."

Neuroimage 53(3): 1085-1092.

Brown, A. S., et al. (2004). "Serologic evidence of prenatal influenza in the etiology of

schizophrenia." Arch Gen Psychiatry 61(8): 774-780.

Brown, A. S., et al. (2000). "Nonaffective psychosis after prenatal exposure to rubella." Am J

Psychiatry 157(3): 438-443.

Brown, S. (1997). "Excess mortality of schizophrenia. A meta-analysis." Br J Psychiatry 171:

502-508.

178

Buchanan, R. W., et al. (2007). "Recent advances in the development of novel pharmacological

agents for the treatment of cognitive impairments in schizophrenia." Schizophr Bull 33(5): 1120-

1130.

Buchanan, R. W., et al. (2007). "The Cognitive and Negative Symptoms in Schizophrenia Trial

(CONSIST): the efficacy of glutamatergic agents for negative symptoms and cognitive

impairments." Am J Psychiatry 164(10): 1593-1602.

Buchanan, R. W., et al. (2011). "A randomized clinical trial of MK-0777 for the treatment of

cognitive impairments in people with schizophrenia." Biol Psychiatry 69(5): 442-449.

Buckholtz, J. W., et al. (2007). "fMRI evidence for functional epistasis between COMT and

RGS4." Mol Psychiatry 12(10): 893-895, 885.

Burmeister, M., et al. (2008). "Psychiatric genetics: progress amid controversy." Nature Reviews

Genetics 9(7): 527-540.

Button, K. S., et al. (2013). "Power failure: why small sample size undermines the reliability of

neuroscience." Nature Reviews Neuroscience.

Calhoun, V. D. and T. Adali (2012). "Multisubject independent component analysis of fMRI: a

decade of intrinsic networks, default mode, and neurodiagnostic discovery." IEEE Rev Biomed

Eng 5: 60-73.

Callicott, J. H., et al. (2003). "Complexity of prefrontal cortical dysfunction in schizophrenia:

more than up or down." Am J Psychiatry 160(12): 2209-2215.

Cannon, T. D., et al. (2000). "The inheritance of neuropsychological dysfunction in twins

discordant for schizophrenia." Am J Hum Genet 67(2): 369-382.

Cardin, J. A., et al. (2009). "Driving fast-spiking cells induces gamma rhythm and controls

sensory responses." Nature 459(7247): 663-667.

Cardno, A. G. and M. J. Owen (2014). "Genetic relationships between schizophrenia, bipolar

disorder, and schizoaffective disorder." Schizophr Bull 40(3): 504-515.

Carpenter Jr, W. T. and B. Kirkpatrick (1988). "The heterogeneity of the long-term course of

schizophrenia." Schizophrenia Bulletin 14(4): 645.

Carpentier, P. A. and T. D. Palmer (2009). "Immune influence on adult neural stem cell

regulation and function." Neuron 64(1): 79-92.

Carroll, L. S. and M. J. Owen (2009). "Genetic overlap between autism, schizophrenia and

bipolar disorder." Genome Med 1(10): 102.

Chakravarty, M. M., et al. (2008). "Towards a validation of atlas warping techniques." Medical

Image Analysis 12(6): 713-726.

179

Chan, R. C. and Gottesman, II (2008). "Neurological soft signs as candidate endophenotypes for

schizophrenia: a shooting star or a Northern star?" Neurosci Biobehav Rev 32(5): 957-971.

Chan, R. C., et al. (2009). "Neurological soft signs and their relationships to neurocognitive

functions: a re-visit with the structural equation modeling design." PLoS One 4(12): e8469.

Chen, C. H., et al. (2013). "Genetic topography of brain morphology." Proc Natl Acad Sci U S A

110(42): 17089-17094.

Chen da, C., et al. (2009). "Decreased levels of serum brain-derived neurotrophic factor in drug-

naive first-episode schizophrenia: relationship to clinical phenotypes." Psychopharmacology

(Berl) 207(3): 375-380.

Chen, K. and N. Rajewsky (2007). "The evolution of gene regulation by transcription factors and

microRNAs." Nature Reviews Genetics 8(2): 93-103.

Chenn, A. and C. A. Walsh (2002). "Regulation of cerebral cortical size by control of cell cycle

exit in neural precursors." Science 297(5580): 365-369.

Chiang, M. C., et al. (2009). "Genetics of brain fiber architecture and intellectual performance." J

Neurosci 29(7): 2212-2224.

Chiang, M. C., et al. (2011). "Genetics of white matter development: a DTI study of 705 twins

and their siblings aged 12 to 29." Neuroimage 54(3): 2308-2317.

Ching, M. S. L., et al. (2010). "Deletions of NRXN1 (neurexin-1) predispose to a wide spectrum

of developmental disorders." American Journal of Medical Genetics. Part B, Neuropsychiatric

Genetics: The Official Publication of the International Society of Psychiatric Genetics 153B(4):

937-947.

Chung, H. J., et al. (2010). "Mouse Homologue of the Schizophrenia Susceptibility Gene

ZNF804A as a Target of Hoxc8." Journal of Biomedicine & Biotechnology 2010.

Colantuoni, C., et al. (2011). "Temporal dynamics and genetic control of transcription in the

human prefrontal cortex." Nature 478(7370): 519-523.

Collins, D. (1994). "The crime of failing to record instructions." N Z Med J 107(971): 40-41.

Collins, D., et al. (1995). "Automatic 3D model-based neuroanatomical segmentation." Hum

Brain Mapp 3: 190-208.

Consortium, E. P., et al. (2012). "An integrated encyclopedia of DNA elements in the human

genome." Nature 489(7414): 57-74.

Conturo, T. E., et al. (1999). "Tracking neuronal fiber pathways in the living human brain." Proc

Natl Acad Sci U S A 96(18): 10422-10427.

180

Costa, E., et al. (2007). "Reviewing the role of DNA (cytosine-5) methyltransferase

overexpression in the cortical GABAergic dysfunction associated with psychosis vulnerability."

Epigenetics 2(1): 29-36.

Costafreda, S. G., et al. (2011). "Pattern of neural responses to verbal fluency shows diagnostic

specificity for schizophrenia and bipolar disorder." BMC Psychiatry 11: 18.

Cotter, D. R., et al. (2001). "Glial cell abnormalities in major psychiatric disorders: the evidence

and implications." Brain Res Bull 55(5): 585-595.

Courchesne, E., et al. (2001). "Unusual brain growth patterns in early life in patients with autistic

disorder: an MRI study." Neurology 57(2): 245-254.

Coyle, J. T. (1996). "The glutamatergic dysfunction hypothesis for schizophrenia." Harvard

Review of Psychiatry 3(5): 241-253.

Cross-Disorder Group of the Psychiatric Genomics, C., et al. (2013). "Identification of risk loci

with shared effects on five major psychiatric disorders: a genome-wide analysis." Lancet

381(9875): 1371-1379.

Crow, T. J. (1980). "Molecular pathology of schizophrenia: more than one disease process?" Br

Med J 280(6207): 66-68.

Cummings, E., et al. (2012). "Mood congruent psychotic symptoms and specific cognitive

deficits in carriers of the novel schizophrenia risk variant at MIR-137." Neuroscience Letters.

Curcio, F. (1978). "Sensorimotor functioning and communication in mute autistic children." J

Autism Child Schizophr 8(3): 281-292.

Curley, A. A., et al. (2011). "Cortical deficits of glutamic acid decarboxylase 67 expression in

schizophrenia: clinical, protein, and cell type-specific features." Am J Psychiatry 168(9): 921-

929.

Cuthbert, B. N. and T. R. Insel (2013). "Toward the future of psychiatric diagnosis: the seven

pillars of RDoC." BMC Med 11: 126.

Dale, A. M., et al. (1999). "Cortical surface-based analysis. I. Segmentation and surface

reconstruction." Neuroimage 9(2): 179-194.

Danos, P., et al. (2003). "Volumes of association thalamic nuclei in schizophrenia: a postmortem

study." Schizophrenia Research 60(2-3): 141-155.

Daskalakis, Z. J., et al. (2002). "Evidence for impaired cortical inhibition in schizophrenia using

transcranial magnetic stimulation." Arch Gen Psychiatry 59(4): 347-354.

Daskalakis, Z. J., et al. (2008). "Evaluating the relationship between long interval cortical

inhibition, working memory and gamma band activity in the dorsolateral prefrontal cortex."

Clinical EEG and Neuroscience 39(3): 150-155.

181

Davatzikos, C., et al. (2005). "Classifying spatial patterns of brain activity with machine learning

methods: application to lie detection." Neuroimage 28(3): 663-668.

de Wit, J., et al. (2009). "LRRTM2 interacts with Neurexin1 and regulates excitatory synapse

formation." Neuron 64(6): 799-806.

DeLisi, L. E. (1992). "The significance of age of onset for schizophrenia." Schizophr Bull 18(2):

209-215.

DeLisi, L. E., et al. (1991). "Brain morphology in first-episode schizophrenic-like psychotic

patients: a quantitative magnetic resonance imaging study." Biological Psychiatry 29(2): 159-

175.

Dennis, E. L., et al. (2011). "Altered structural brain connectivity in healthy carriers of the

autism risk gene, CNTNAP2." Brain Connect 1(6): 447-459.

Descoteaux, M., et al. (2009). "Deterministic and probabilistic tractography based on complex

fibre orientation distributions." IEEE Trans Med Imaging 28(2): 269-286.

Desrivieres, S., et al. (2014). "Single nucleotide polymorphism in the neuroplastin locus

associates with cortical thickness and intellectual ability in adolescents." Mol Psychiatry.

Deverman, B. E. and P. H. Patterson (2009). "Cytokines and CNS development." Neuron 64(1):

61-78.

Di, X., et al. (2009). "White matter reduction in patients with schizophrenia as revealed by

voxel-based morphometry: An activation likelihood estimation meta-analysis." Progress in

Neuro-Psychopharmacology and Biological Psychiatry 33(8): 1390-1394.

Dickinson, D. (2008). "Digit symbol coding and general cognitive ability in schizophrenia: worth

another look?" Br J Psychiatry 193(5): 354-356.

Donohoe, G., et al. (2010). "The psychosis susceptibility gene ZNF804A: associations,

functions, and phenotypes." Schizophr Bull 36(5): 904-909.

Dorph-Petersen, K. A., et al. (2005). "The influence of chronic exposure to antipsychotic

medications on brain size before and after tissue fixation: a comparison of haloperidol and

olanzapine in macaque monkeys." Neuropsychopharmacology 30(9): 1649-1661.

Douaud, G., et al. (2007). "Anatomically related grey and white matter abnormalities in

adolescent-onset schizophrenia." Brain 130(9): 2375 -2386.

Dudbridge, F. (2013). "Power and predictive accuracy of polygenic risk scores." PLoS Genet

9(3): e1003348.

Dumontheil, I., et al. (2008). "Development of rostral prefrontal cortex and cognitive and

behavioural disorders." Developmental Medicine and Child Neurology 50(3): 168-181.

182

Ehrlich, S., et al. (2011). "Associations of cortical thickness and cognition in patients with

schizophrenia and healthy controls." Schizophrenia Bulletin.

Eisenberg, D. P., et al. (2013). "Brain-derived neurotrophic factor (BDNF) Val(66)Met

polymorphism differentially predicts hippocampal function in medication-free patients with

schizophrenia." Mol Psychiatry 18(6): 713-720.

Ellison-Wright, I. and E. Bullmore (2009). "Meta-analysis of diffusion tensor imaging studies in

schizophrenia." Schizophr Res 108(1-3): 3-10.

Elvevag, B. and T. E. Goldberg (2000). "Cognitive impairment in schizophrenia is the core of

the disorder." Crit Rev Neurobiol 14(1): 1-21.

Emery, B. (2010). "Regulation of oligodendrocyte differentiation and myelination." Science

Signalling 330(6005): 779.

Erickson, K. I., et al. (2009). "Top-down attentional control in spatially coincident stimuli

enhances activity in both task-relevant and task-irrelevant regions of cortex." Behav Brain Res

197(1): 186-197.

Erk, S., et al. (2013). "Hippocampal and Frontolimbic Function as Intermediate Phenotype for

Psychosis: Evidence from Healthy Relatives and a Common Risk Variant in CACNA1C." Biol

Psychiatry.

Erk, S., et al. (2010). "Brain function in carriers of a genome-wide supported bipolar disorder

variant." Arch Gen Psychiatry 67(8): 803-811.

Esslinger, C., et al. (2011). "Cognitive state and connectivity effects of the genome-wide

significant psychosis variant in ZNF804A." Neuroimage 54(3): 2514-2523.

Esslinger, C., et al. (2009). "Neural mechanisms of a genome-wide supported psychosis variant."

Science 324(5927): 605.

Etherton, M. R., et al. (2009). "Mouse neurexin-1alpha deletion causes correlated

electrophysiological and behavioral changes consistent with cognitive impairments." Proc Natl

Acad Sci U S A 106(42): 17998-18003.

Farzan, F., et al. (2010). "Evidence for gamma inhibition deficits in the dorsolateral prefrontal

cortex of patients with schizophrenia." Brain 133(Pt 5): 1505-1514.

Farzan, F., et al. (2010). "Reliability of long-interval cortical inhibition in healthy human

subjects: a TMS-EEG study." J Neurophysiol 104(3): 1339-1346.

Farzan, F., et al. (2009). "Suppression of gamma-oscillations in the dorsolateral prefrontal cortex

following long interval cortical inhibition: a TMS-EEG study." Neuropsychopharmacology

34(6): 1543-1551.

Feng, J., et al. (2006). "High frequency of neurexin 1beta signal peptide structural variants in

patients with autism." Neuroscience Letters 409(1): 10-13.

183

Ferreira, M. A., et al. (2008). "Collaborative genome-wide association analysis supports a role

for ANK3 and CACNA1C in bipolar disorder." Nat Genet 40(9): 1056-1058.

Fields, R. D. and G. Burnstock (2006). "Purinergic signalling in neuron-glia interactions." Nat

Rev Neurosci 7(6): 423-436.

First MB, S. R., Gibbon M, Williams JBW (1995). Strucutured Clinical Interview for DSM-IV

Axis I Disorders, Patient Edition (SCID-P), version 2. New York, Biometrics Research.

Fischl, B. and A. M. Dale (2000). "Measuring the thickness of the human cerebral cortex from

magnetic resonance images." Proc Natl Acad Sci U S A 97(20): 11050-11055.

Fischl, B., et al. (1999). "Cortical surface-based analysis. II: Inflation, flattening, and a surface-

based coordinate system." Neuroimage 9(2): 195-207.

Flashman, L. A., et al. (1996). "Soft signs and neuropsychological performance in

schizophrenia." The American Journal of Psychiatry 153(4): 526-532.

Flavell, S. W. and M. E. Greenberg (2008). "Signaling mechanisms linking neuronal activity to

gene expression and plasticity of the nervous system." Annual review of neuroscience 31: 563.

Folstein, M. F., et al. (1975). ""Mini-mental state". A practical method for grading the cognitive

state of patients for the clinician." J Psychiatr Res 12(3): 189-198.

Fornito, A., et al. (2013). "Graph analysis of the human connectome: promise, progress, and

pitfalls." Neuroimage 80: 426-444.

Foussias, G. and G. Remington (2010). "Antipsychotics and schizophrenia: from efficacy and

effectiveness to clinical decision-making." Can J Psychiatry 55(3): 117-125.

Friston, K. J., et al. (1994). "Statistical parametric maps in functional imaging: a general linear

approach." Hum Brain Mapp 2(4): 189-210.

Fromer, M., et al. (2014). "De novo mutations in schizophrenia implicate synaptic networks."

Nature 506(7487): 179-184.

Fusar-Poli, P., et al. (2012). "Neuroanatomical maps of psychosis onset: voxel-wise meta-

analysis of antipsychotic-naive VBM studies." Schizophr Bull 38(6): 1297-1307.

Fusar-Poli, P., et al. (2013). "Progressive brain changes in schizophrenia related to antipsychotic

treatment? A meta-analysis of longitudinal MRI studies." Neurosci Biobehav Rev 37(8): 1680-

1691.

Gabriel, S. B., et al. (2002). "The structure of haplotype blocks in the human genome." Science

(New York, N.Y.) 296(5576): 2225-2229.

Garver, D. L., et al. (2008). "Compromised myelin integrity during psychosis with repair during

remission in drug-responding schizophrenia." Int J Neuropsychopharmacol 11(1): 49-61.

184

Ge, T., et al. (2012). "Increasing power for voxel-wise genome-wide association studies: the

random field theory, least square kernel machines and fast permutation procedures." Neuroimage

63(2): 858-873.

Geekiyanage, H. and C. Chan (2011). "MicroRNA-137/181c regulates serine

palmitoyltransferase and in turn amyloid beta, novel targets in sporadic Alzheimer's disease." J

Neurosci 31(41): 14820-14830.

Geekiyanage, H., et al. (2011). "Blood serum miRNA: Non-invasive biomarkers for Alzheimer's

disease." Exp Neurol.

Geng, X., et al. (2012). "White matter heritability using diffusion tensor imaging in neonatal

brains." Twin Res Hum Genet 15(3): 336-350.

Genovese, C. R., et al. (2002). "Thresholding of statistical maps in functional neuroimaging

using the false discovery rate." Neuroimage 15(4): 870-878.

Geyer, M. A., et al. (1990). "Startle response models of sensorimotor gating and habituation

deficits in schizophrenia." Brain Res Bull 25(3): 485-498.

Gibbs, R. A., et al. (2003). "The international HapMap project." Nature 426(6968): 789-796.

Giedd, J. N., et al. (1999). "Childhood-onset schizophrenia: progressive brain changes during

adolescence." Biological Psychiatry.

Giusti-Rodriguez, P. and P. F. Sullivan (2013). "The genomics of schizophrenia: update and

implications." J Clin Invest 123(11): 4557-4563.

Glantz, L. A. and D. A. Lewis (2000). "Decreased dendritic spine density on prefrontal cortical

pyramidal neurons in schizophrenia." Arch Gen Psychiatry 57(1): 65-73.

Glessner, J. T. and H. Hakonarson (2009). "Common variants in polygenic schizophrenia."

Genome Biology 10(9): 236.

Glessner, J. T., et al. (2009). "Autism genome-wide copy number variation reveals ubiquitin and

neuronal genes." Nature 459(7246): 569-573.

Goff, D. C., et al. (2011). "The treatment of cognitive impairment in schizophrenia." Pharmacol

Biochem Behav 99(2): 245-253.

Gold, J. M. (2004). "Cognitive deficits as treatment targets in schizophrenia." Schizophr Res

72(1): 21-28.

Gold, S., et al. (1999). "Longitudinal study of cognitive function in first-episode and recent-onset

schizophrenia." Am J Psychiatry 156(9): 1342-1348.

Goldberg, T. E., et al. (2007). "Cognitive improvement after treatment with second-generation

antipsychotic medications in first-episode schizophrenia: is it a practice effect?" Arch Gen

Psychiatry 64(10): 1115-1122.

185

Golden, C. J. and S. M. Freshwater (1978). "Stroop color and word test."

Goldman-Rakic, P. S. and L. D. Selemon (1997). "Functional and Anatomical Aspects of

Prefrontal Pathology in Schizophrenia." Schizophrenia Bulletin 23(3): 437-458.

Goldman, A. L., et al. (2009). "Widespread reductions of cortical thickness in schizophrenia and

spectrum disorders and evidence of heritability." Arch Gen Psychiatry 66(5): 467-477.

Goldman, S., et al. (2009). "Motor stereotypies in children with autism and other developmental

disorders." Developmental Medicine and Child Neurology 51(1): 30-38.

Gottesman, II and J. Shields (1967). "A polygenic theory of schizophrenia." Proc Natl Acad Sci

U S A 58(1): 199-205.

Gottesman, I. I. and T. D. Gould (2003). "The endophenotype concept in psychiatry: etymology

and strategic intentions." American Journal of Psychiatry 160(4): 636-645.

Gottesman, I. I. and T. D. Gould (2003). "The endophenotype concept in psychiatry: etymology

and strategic intentions." The American Journal of Psychiatry 160(4): 636-645.

Green, E. K., et al. (2010). "The bipolar disorder risk allele at CACNA1C also confers risk of

recurrent major depression and of schizophrenia." Mol Psychiatry 15(10): 1016-1022.

Green, M., et al. (2012). "Genome-wide supported variant MIR137 and severe negative

symptoms predict membership of an impaired cognitive subtype of schizophrenia." Molecular

psychiatry.

Green, M. F. (1996). "What are the functional consequences of neurocognitive deficits in

schizophrenia?" Am J Psychiatry 153(3): 321-330.

Green, M. F., et al. (2000). "Neurocognitive deficits and functional outcome in schizophrenia:

are we measuring the "right stuff"?" Schizophr Bull 26(1): 119-136.

Green, M. F., et al. (2004). "Approaching a consensus cognitive battery for clinical trials in

schizophrenia: the NIMH-MATRICS conference to select cognitive domains and test criteria."

Biol Psychiatry 56(5): 301-307.

Green, M. J., et al. (2011). "Brain-derived neurotrophic factor levels in schizophrenia: a

systematic review with meta-analysis." Mol Psychiatry 16(9): 960-972.

Greenwood, T. A., et al. (2007). "Initial heritability analyses of endophenotypic measures for

schizophrenia: the consortium on the genetics of schizophrenia." Arch Gen Psychiatry 64(11):

1242-1250.

Guella, I., et al. (2013). "Analysis of miR-137 expression and rs1625579 in dorsolateral

prefrontal cortex." J Psychiatr Res 47(9): 1215-1221.

186

Gunawardana, Y. and M. Niranjan (2013). "Bridging the gap between transcriptome and

proteome measurements identifies post-translationally regulated genes." Bioinformatics 29(23):

3060-3066.

Haenschel, C., et al. (2009). "Cortical oscillatory activity is critical for working memory as

revealed by deficits in early-onset schizophrenia." J Neurosci 29(30): 9481-9489.

Hakak, Y., et al. (2001). "Genome-wide expression analysis reveals dysregulation of

myelination-related genes in chronic schizophrenia." Proc Natl Acad Sci U S A 98(8): 4746-

4751.

Hale, J. B., et al. (2002). "Analyzing digit span components for assessment of attention

processes." Journal of Psychoeducational Assessment 20(2): 128-143.

Hall, M. H., et al. (2011). "The early auditory gamma-band response is heritable and a putative

endophenotype of schizophrenia." Schizophr Bull 37(4): 778-787.

Halstead, W. C. (1947). "Brain and intelligence; a quantitative study of the frontal lobes."

Hamshere, M. L., et al. (2013). "Genome-wide significant associations in schizophrenia to

ITIH3/4, CACNA1C and SDCCAG8, and extensive replication of associations reported by the

Schizophrenia PGC." Mol Psychiatry 18(6): 708-712.

Han, X., et al. (2006). "Reliability of MRI-derived measurements of human cerebral cortical

thickness: the effects of field strength, scanner upgrade and manufacturer." Neuroimage 32(1):

180-194.

Hans, S. L., et al. (2000). "Social adjustment of adolescents at risk for schizophrenia: the

Jerusalem Infant Development Study." J Am Acad Child Adolesc Psychiatry 39(11): 1406-1414.

Hare, E., et al. (2010). "Heritability of age of onset of psychosis in schizophrenia." Am J Med

Genet B Neuropsychiatr Genet 153B(1): 298-302.

Harris, M. J. and D. V. Jeste (1988). "Late-onset schizophrenia: an overview." Schizophr Bull

14(1): 39-55.

Harvey, P. D., et al. (2008). "A randomized double-blind comparison of ziprasidone vs.

clozapine for cognition in patients with schizophrenia selected for resistance or intolerance to

previous treatment." Schizophr Res 105(1-3): 138-143.

Hashimoto, R., et al. (2010). "The impact of a genome-wide supported psychosis variant in the

ZNF804A gene on memory function in schizophrenia." Am J Med Genet B Neuropsychiatr

Genet 153B(8): 1459-1464.

Hashimoto, T., et al. (2008). "Alterations in GABA-related transcriptome in the dorsolateral

prefrontal cortex of subjects with schizophrenia." Mol Psychiatry 13(2): 147-161.

Hashimoto, T., et al. (2003). "Gene expression deficits in a subclass of GABA neurons in the

prefrontal cortex of subjects with schizophrenia." J Neurosci 23(15): 6315-6326.

187

Hawrylycz, M. J., et al. (2012). "An anatomically comprehensive atlas of the adult human brain

transcriptome." Nature 489(7416): 391-399.

Heinrichs, R. W. and K. K. Zakzanis (1998). "Neurocognitive deficit in schizophrenia: a

quantitative review of the evidence." Neuropsychology 12(3): 426-445.

Herbert, M. R., et al. (2003). "Dissociations of cerebral cortex, subcortical and cerebral white

matter volumes in autistic boys." Brain: A Journal of Neurology 126(Pt 5): 1182-1192.

Herbert, M. R., et al. (2004). "Localization of white matter volume increase in autism and

developmental language disorder." Annals of Neurology 55(4): 530-540.

Hill, W. G., et al. (2008). "Data and theory point to mainly additive genetic variance for complex

traits." PLoS Genetics 4(2): e1000008.

Ho, B. C., et al. (2003). "Progressive structural brain abnormalities and their relationship to

clinical outcome: a longitudinal magnetic resonance imaging study early in schizophrenia." Arch

Gen Psychiatry 60(6): 585-594.

Ho, B. C., et al. (2011). "Long-term antipsychotic treatment and brain volumes: a longitudinal

study of first-episode schizophrenia." Arch Gen Psychiatry 68(2): 128-137.

Hollingshead, A. B. (1975). Four Factor Index of Social Status, Yale University. New Haven, Ct.

Hollis, C. (1995). "Child and adolescent (juvenile onset) schizophrenia. A case control study of

premorbid developmental impairments." Br J Psychiatry 166(4): 489-495.

Hollis, C. (2000). "Adult outcomes of child- and adolescent-onset schizophrenia: Diagnostic

stability and predictive validity." American Journal of Psychiatry 157(10): 1652-1659.

Holmes, A. J., et al. (2012). "Individual differences in amygdala-medial prefrontal anatomy link

negative affect, impaired social functioning, and polygenic depression risk." J Neurosci 32(50):

18087-18100.

Honea, R., et al. (2005). "Regional deficits in brain volume in schizophrenia: a meta-analysis of

voxel-based morphometry studies." Am J Psychiatry 162(12): 2233-2245.

Honey, G. D., et al. (2005). "Functional dysconnectivity in schizophrenia associated with

attentional modulation of motor function." Brain 128(11): 2597 -2611.

Hopfinger, J. B., et al. (2000). "The neural mechanisms of top-down attentional control." Nat

Neurosci 3(3): 284-291.

Hoppenbrouwers, S. S., et al. (2012). "Inhibitory deficits in the dorsolateral prefrontal cortex in

psychopathic offenders." Cortex.

Hori, H., et al. (2012). "Effects of the CACNA1C risk allele on neurocognition in patients with

schizophrenia and healthy individuals." Sci Rep 2: 634.

188

Horvath, S. and K. Mirnics (2014). "Schizophrenia as a Disorder of Molecular Pathways." Biol

Psychiatry.

Huang, H. S. and S. Akbarian (2007). "GAD1 mRNA expression and DNA methylation in

prefrontal cortex of subjects with schizophrenia." PLoS One 2(8): e809.

Huh, G. S., et al. (2000). "Functional requirement for class I MHC in CNS development and

plasticity." Science 290(5499): 2155-2159.

Hulshoff Pol, H. and E. Bullmore (2013). "Neural networks in psychiatry." Eur

Neuropsychopharmacol 23(1): 1-6.

Hyde, T. M., et al. (2011). "Expression of GABA signaling molecules KCC2, NKCC1, and

GAD1 in cortical development and schizophrenia." J Neurosci 31(30): 11088-11095.

Ikeda, M., et al. (2010). "Copy number variation in schizophrenia in the Japanese population."

Biological Psychiatry 67(3): 283-286

Ingalhalikar, M., et al. (2010). "DTI based diagnostic prediction of a disease via pattern

classification." Med Image Comput Comput Assist Interv 13(Pt 1): 558-565.

Insel, T., et al. (2010). "Research domain criteria (RDoC): toward a new classification

framework for research on mental disorders." Am J Psychiatry 167(7): 748-751.

International Schizophrenia Consortium, et al. (2009). "Common polygenic variation contributes

to risk of schizophrenia and bipolar disorder." Nature 460(7256): 748-752.

Ioannidis, J. P. (2008). "Why most discovered true associations are inflated." Epidemiology

19(5): 640-648.

Irish Schizophrenia Genomics, C. and C. the Wellcome Trust Case Control (2012). "Genome-

wide association study implicates HLA-C*01:02 as a risk factor at the major histocompatibility

complex locus in schizophrenia." Biol Psychiatry 72(8): 620-628.

Jakob, H. and H. Beckmann (1986). "Prenatal developmental disturbances in the limbic

allocortex in schizophrenics." J Neural Transm 65(3-4): 303-326.

James, A. C., et al. (2004). "Cerebellar, Prefrontal Cortex, and Thalamic Volumes Over Two

Time Points in Adolescent-Onset Schizophrenia." Am J Psychiatry 161(6): 1023-1029.

Jenkins, A., et al. (2014). "Identification of Candidate Single-Nucleotide Polymorphisms in

NRXN1 Related to Antipsychotic Treatment Response in Patients with Schizophrenia."

Neuropsychopharmacology.

Jeste, D. V., et al. (1995). "Clinical and neuropsychological characteristics of patients with late-

onset schizophrenia." Am J Psychiatry 152(5): 722-730.

Jia, P., et al. (2012). "A bias-reducing pathway enrichment analysis of genome-wide association

data confirmed association of the MHC region with schizophrenia." J Med Genet 49(2): 96-103.

189

Joa, I., et al. (2009). "Baseline profiles of adolescent vs. adult-onset first-episode psychosis in an

early detection program." Acta Psychiatr Scand 119(6): 494-500.

Johnstone, E. C., et al. (1976). "Cerebral ventricular size and cognitive impairment in chronic

schizophrenia." Lancet 2(7992): 924-926.

Jones, E. G. (1997). "Cortical Development and Thalamic Pathology in Schizophrenia."

Schizophrenia Bulletin 23(3): 483-501.

Kane, J., et al. (1988). "Clozapine for the treatment-resistant schizophrenic. A double-blind

comparison with chlorpromazine." Arch Gen Psychiatry 45(9): 789-796.

Kane, J. M. (1988). "The role of neuroleptics in manic-depressive illness." J Clin Psychiatry 49

Suppl: 12-14.

Kane, J. M. and C. U. Correll (2010). "Pharmacologic treatment of schizophrenia." Dialogues

Clin Neurosci 12(3): 345-357.

Kane, J. M., et al. (2007). "Aripiprazole for treatment-resistant schizophrenia: results of a

multicenter, randomized, double-blind, comparison study versus perphenazine." J Clin

Psychiatry 68(2): 213-223.

Kang, H. J., et al. (2011). "Spatio-temporal transcriptome of the human brain." Nature

478(7370): 483-489.

Kapur, S. (2003). "Psychosis as a state of aberrant salience: a framework linking biology,

phenomenology, and pharmacology in schizophrenia." Am J Psychiatry 160(1): 13-23.

Kapur, S. and G. Remington (2001). "Atypical antipsychotics: new directions and new

challenges in the treatment of schizophrenia." Annu Rev Med 52: 503-517.

Kapur, S. and P. Seeman (2001). "Does fast dissociation from the dopamine d(2) receptor

explain the action of atypical antipsychotics?: A new hypothesis." Am J Psychiatry 158(3): 360-

369.

Kapur, S., et al. (2000). "Relationship between dopamine D(2) occupancy, clinical response, and

side effects: a double-blind PET study of first-episode schizophrenia." Am J Psychiatry 157(4):

514-520.

Karayiorgou, M., et al. (2010). "22q11.2 microdeletions: linking DNA structural variation to

brain dysfunction and schizophrenia." Nat Rev Neurosci 11(6): 402-416.

Karres, J. S., et al. (2007). "The conserved microRNA miR-8 tunes atrophin levels to prevent

neurodegeneration in Drosophila." Cell 131(1): 136-145.

Kawasaki, H. and S. Iwamuro (2008). "Potential roles of histones in host defense as

antimicrobial agents." Infect Disord Drug Targets 8(3): 195-205.

190

Keefe, R. S., et al. (2007). "Neurocognitive effects of antipsychotic medications in patients with

chronic schizophrenia in the CATIE Trial." Arch Gen Psychiatry 64(6): 633-647.

Keefe, R. S., et al. (2007). "Effects of olanzapine, quetiapine, and risperidone on neurocognitive

function in early psychosis: a randomized, double-blind 52-week comparison." Am J Psychiatry

164(7): 1061-1071.

Kennedy, J. L., et al. (2014). "The social and economic burden of treatment-resistant

schizophrenia: a systematic literature review." Int Clin Psychopharmacol 29(2): 63-76.

Kent, W. J., et al. (2002). "The human genome browser at UCSC." Genome Res 12(6): 996-

1006.

Kim, A. H., et al. (2012). "Experimental validation of candidate schizophrenia gene ZNF804A as

target for hsa-miR-137." Schizophr Res 141(1): 60-64.

Kim, H.-G., et al. (2008). "Disruption of neurexin 1 associated with autism spectrum disorder."

American Journal of Human Genetics 82(1): 199-207.

Kim, J., et al. (2007). "A MicroRNA feedback circuit in midbrain dopamine neurons." Science

317(5842): 1220-1224.

Kim, J. S., et al. (2005). "Automated 3-D extraction and evaluation of the inner and outer cortical

surfaces using a Laplacian map and partial volume effect classification." Neuroimage 27(1): 210-

221.

Kirov, G., et al. (2009). "Neurexin 1 (NRXN1) deletions in schizophrenia." Schizophrenia

Bulletin 35(5): 851-854.

Ko, J., et al. (2009). "LRRTM2 functions as a neurexin ligand in promoting excitatory synapse

formation." Neuron 64(6): 791-798.

Kochunov, P., et al. (2010). "Genetics of microstructure of cerebral white matter using diffusion

tensor imaging." Neuroimage 53(3): 1109-1116.

Kochunov, P., et al. (2014). "Multi-site study of additive genetic effects on fractional anisotropy

of cerebral white matter: Comparing meta and megaanalytical approaches for data pooling."

Neuroimage 95: 136-150.

Konopaske, G. T., et al. (2007). "Effect of chronic exposure to antipsychotic medication on cell

numbers in the parietal cortex of macaque monkeys." Neuropsychopharmacology 32(6): 1216-

1223.

Kraepelin, E. (1971). Dementia praecox and paraphrenia, Krieger Publishing Company.

Kraft, P. (2008). "Curses--winner's and otherwise--in genetic epidemiology." Epidemiology

19(5): 649-651; discussion 657-648.

191

Kremen, W. S., et al. (2004). "Heterogeneity of schizophrenia: a study of individual

neuropsychological profiles." Schizophr Res 71(2-3): 307-321.

Kubicki, M., et al. (2007). "A review of diffusion tensor imaging studies in schizophrenia." J

Psychiatr Res 41(1-2): 15-30.

Kubicki, M., et al. (2005). "DTI and MTR abnormalities in schizophrenia: analysis of white

matter integrity." Neuroimage 26(4): 1109-1118.

Kubicki, M., et al. (2002). "Voxel-based morphometric analysis of gray matter in first episode

schizophrenia." Neuroimage 17(4): 1711-1719.

Kubicki, M., et al. (2005). "The application of DTI to investigate white matter abnormalities in

schizophrenia." Ann N Y Acad Sci 1064: 134-148.

Kumra, S., et al. (2004). "Reduced frontal white matter integrity in early-onset schizophrenia: a

preliminary study." Biol Psychiatry 55(12): 1138-1145.

Kumra, S., et al. (2008). "Efficacy and tolerability of second-generation antipsychotics in

children and adolescents with schizophrenia." Schizophr Bull 34(1): 60-71.

Kurtz, M. M. (2005). "Neurocognitive impairment across the lifespan in schizophrenia: an

update." Schizophr Res 74(1): 15-26.

Kuswanto, C. N., et al. (2014). "Increased body mass index makes an impact on brain white-

matter integrity in adults with remitted first-episode mania." Psychol Med 44(3): 533-541.

Kwon, E., et al. (2011). "Validation of schizophrenia-associated genes CSMD1, C10orf26,

CACNA1C and TCF4 as miR-137 targets." Mol Psychiatry.

Kyriakopoulos, M., et al. (2009). "Effect of age at onset of schizophrenia on white matter

abnormalities." The British Journal of Psychiatry 195(4): 346-353.

Lahiri, D. K. and J. I. Nurnberger, Jr. (1991). "A rapid non-enzymatic method for the preparation

of HMW DNA from blood for RFLP studies." Nucleic Acids Res 19(19): 5444.

Lauronen, E., et al. (2007). "Outcome and its predictors in schizophrenia within the Northern

Finland 1966 Birth Cohort." Eur Psychiatry 22(2): 129-136.

Lazarus, M. S., et al. (2013). "GAD67 Deficiency in Parvalbumin Interneurons Produces Deficits

in Inhibitory Transmission and Network Disinhibition in Mouse Prefrontal Cortex." Cereb

Cortex.

Le Floch, E., et al. (2012). "Significant correlation between a set of genetic polymorphisms and a

functional brain network revealed by feature selection and sparse Partial Least Squares."

Neuroimage 63(1): 11-24.

Lee, Y., et al. (2008). "miR-19, miR-101 and miR-130 co-regulate ATXN1 levels to potentially

modulate SCA1 pathogenesis." Nat Neurosci 11(10): 1137-1139.

192

Lencz, T., et al. (2014). "Molecular genetic evidence for overlap between general cognitive

ability and risk for schizophrenia: a report from the Cognitive Genomics consorTium

(COGENT)." Mol Psychiatry 19(2): 168-174.

Lerch, J. P. and A. C. Evans (2005). "Cortical thickness analysis examined through power

analysis and a population simulation." Neuroimage 24(1): 163-173.

Lerch, J. P., et al. (2008). "Automated cortical thickness measurements from MRI can accurately

separate Alzheimer's patients from normal elderly controls." Neurobiology of aging 29(1): 23-30.

Lerch, J. P., et al. (2005). "Focal decline of cortical thickness in Alzheimer's disease identified by

computational neuroanatomy." Cerebral Cortex 15(7): 995-1001.

Leslie, J. H. and E. Nedivi (2011). "Activity-regulated genes as mediators of neural circuit

plasticity." Progress in neurobiology 94(3): 223-237.

Lett, T. A., et al. (2014). Imaging genetics of working memory. Neuroimaging Genetics. A. R.

Hariri, K. L. Bigos and D. R. Weinberger, Oxford University Press.

Lett, T. A., et al. (2013). "The genome-wide supported microRNA-137 variant predicts

phenotypic heterogeneity within schizophrenia." Mol Psychiatry 18(4): 443-450.

Lett, T. A., et al. (2011). "The putative functional rs1045881 marker of neurexin-1 in

schizophrenia and clozapine response." Schizophr Res 132(2-3): 121-124.

Lett, T. A., et al. (2014). "Treating Working Memory Deficits in Schizophrenia: A Review of the

Neurobiology." Biol Psychiatry 75(5): 361-370.

Lett, T. A., et al. (2012). "Pharmacogenetics of antipsychotic-induced weight gain: review and

clinical implications." Mol Psychiatry 17(3): 242-266.

Lett, T. A., et al. (2011). "ANK3, CACNA1C and ZNF804A gene variants in bipolar disorders

and psychosis subphenotype." World J Biol Psychiatry 12(5): 392-397.

Levinson, D. F., et al. (2011). "Copy number variants in schizophrenia: confirmation of five

previous findings and new evidence for 3q29 microdeletions and VIPR2 duplications." Am J

Psychiatry 168(3): 302-316.

Lewis, D. A., et al. (2008). "Subunit-selective modulation of GABA type A receptor

neurotransmission and cognition in schizophrenia." Am J Psychiatry 165(12): 1585-1593.

Lewis, D. A., et al. (2012). "Cortical parvalbumin interneurons and cognitive dysfunction in

schizophrenia." Trends Neurosci 35(1): 57-67.

Lewis, D. A., et al. (2005). "Cortical inhibitory neurons and schizophrenia." Nat Rev Neurosci

6(4): 312-324.

Lewis, D. A. and P. Levitt (2002). "Schizophrenia as a disorder of neurodevelopment." Annual

review of neuroscience 25(1): 409-432.

193

Lewis, D. A. and J. A. Lieberman (2000). "Catching up on schizophrenia: natural history and

neurobiology." Neuron 28(2): 325-334.

Lewis, D. A. and B. Moghaddam (2006). "Cognitive dysfunction in schizophrenia: convergence

of gamma-aminobutyric acid and glutamate alterations." Arch Neurol 63(10): 1372-1376.

Lezak, M. (1995). Neuropsychological Assessment. Oxford, Oxford University Press, New

York.

Li, M. X., et al. (2011). "GATES: a rapid and powerful gene-based association test using

extended Simes procedure." Am J Hum Genet 88(3): 283-293.

Lichtenstein, P., et al. (2009). "Common genetic determinants of schizophrenia and bipolar

disorder in Swedish families: a population-based study." Lancet 373(9659): 234-239.

Lieberman, J., et al. (2001). "Longitudinal study of brain morphology in first episode

schizophrenia." Biol Psychiatry 49(6): 487-499.

Lieberman, J. A., et al. (2005). "Antipsychotic drug effects on brain morphology in first-episode

psychosis." Arch Gen Psychiatry 62(4): 361-370.

Lim, K. O., et al. (1996). "Gray matter deficits in young onset schizophrenia are independent of

age of onset." Biol Psychiatry 40(1): 4-13.

Liu, B., et al. (2014). "The Impact of MIR137 on Dorsolateral Prefrontal-Hippocampal

Functional Connectivity in Healthy Subjects." Neuropsychopharmacology.

MacDonald, D., et al. (2000). "Automated 3-D extraction of inner and outer surfaces of cerebral

cortex from MRI." Neuroimage 12(3): 340-356.

Magri, C., et al. (2010). "New copy number variations in schizophrenia." PloS One 5(10):

e13422.

Malhotra, D. and J. Sebat (2012). "CNVs: harbingers of a rare variant revolution in psychiatric

genetics." Cell 148(6): 1223-1241.

Manolio, T. A., et al. (2009). "Finding the missing heritability of complex diseases." Nature

461(7265): 747-753.

Marsh, L., et al. (1997). "Structural magnetic resonance imaging abnormalities in men with

severe chronic schizophrenia and an early age at clinical onset." Arch Gen Psychiatry 54(12):

1104-1112.

Marshall, C. R., et al. (2008). "Structural variation of chromosomes in autism spectrum

disorder." American Journal of Human Genetics 82(2): 477-488.

Matthews, C. and H. Klove (1964). "Instruction manual for the adult neuropsychology test

battery." Madison, WI: University of Wisconsin Medical School.

194

Matute, C. (2010). "Calcium dyshomeostasis in white matter pathology." Cell calcium 47(2):

150-157.

Mazziotta, J., et al. (2001). "A four-dimensional probabilistic atlas of the human brain." J Am

Med Inform Assoc 8(5): 401-430.

McAlonan, G. M., et al. (2009). "Differential effects on white-matter systems in high-

functioning autism and Asperger's syndrome." Psychological Medicine 39(11): 1885-1893.

McAlonan, G. M., et al. (2005). "Mapping the brain in autism. A voxel-based MRI study of

volumetric differences and intercorrelations in autism." Brain: A Journal of Neurology 128(Pt 2):

268-276.

McAlonan, G. M., et al. (2002). "Brain anatomy and sensorimotor gating in Asperger's

syndrome." Brain: A Journal of Neurology 125(Pt 7): 1594-1606.

McCarley, R. W., et al. (1999). "MRI anatomy of schizophrenia." Biological Psychiatry 45(9):

1099-1119.

McGurk, S. R., et al. (2007). "A meta-analysis of cognitive remediation in schizophrenia." Am J

Psychiatry 164(12): 1791-1802.

McIntosh, A. M., et al. (2004). "Voxel-based morphometry of patients with schizophrenia or

bipolar disorder and their unaffected relatives." Biological Psychiatry 56(8): 544-552.

McTigue, D. M. and R. B. Tripathi (2008). "The life, death, and replacement of oligodendrocytes

in the adult CNS." J Neurochem 107(1): 1-19.

Meltzer, H. Y., et al. (1989). "Classification of typical and atypical antipsychotic drugs on the

basis of dopamine D-1, D-2 and serotonin2 pKi values." J Pharmacol Exp Ther 251(1): 238-246.

Mengotti, P., et al. (2010). "Altered white matter integrity and development in children with

autism: A combined voxel-based morphometry and diffusion imaging study." Brain Research

Bulletin.

Mesholam-Gately, R. I., et al. (2009). "Neurocognition in first-episode schizophrenia: a meta-

analytic review." Neuropsychology 23(3): 315-336.

Messias, E. L., et al. (2007). "Epidemiology of schizophrenia: review of findings and myths."

Psychiatr Clin North Am 30(3): 323-338.

Meyer-Lindenberg, A., et al. (2005). "Midbrain dopamine and prefrontal function in humans:

interaction and modulation by COMT genotype." Nature neuroscience 8(5): 594-596.

Meyer-Lindenberg, A., et al. (2001). "Evidence for abnormal cortical functional connectivity

during working memory in schizophrenia." American Journal of Psychiatry 158(11): 1809-1817.

Meyer-Lindenberg, A. and D. R. Weinberger (2006). "Intermediate phenotypes and genetic

mechanisms of psychiatric disorders." Nature Reviews Neuroscience 7(10): 818-827.

195

Meyer-Lindenberg, A. S., et al. (2005). "Regionally specific disturbance of dorsolateral

prefrontal-hippocampal functional connectivity in schizophrenia." Arch Gen Psychiatry 62(4):

379-386.

Mill, J., et al. (2008). "Epigenomic profiling reveals DNA-methylation changes associated with

major psychosis." Am J Hum Genet 82(3): 696-711.

Miller, B. H. and C. Wahlestedt (2010). "MicroRNA dysregulation in psychiatric disease." Brain

research 1338: 89-99.

Miller, B. H., et al. (2012). "MicroRNA-132 dysregulation in schizophrenia has implications for

both neurodevelopment and adult brain function." Proc Natl Acad Sci U S A 109(8): 3125-3130.

Miller, M. D., et al. (1992). "Rating chronic medical illness burden in geropsychiatric practice

and research: application of the Cumulative Illness Rating Scale." Psychiatry Res 41(3): 237-

248.

Minshew, N. J. and T. A. Keller (2010). "The nature of brain dysfunction in autism: functional

brain imaging studies." Current Opinion in Neurology 23(2): 124-130.

Mishara, A. L. and T. E. Goldberg (2004). "A meta-analysis and critical review of the effects of

conventional neuroleptic treatment on cognition in schizophrenia: opening a closed book." Biol

Psychiatry 55(10): 1013-1022.

Mitelman, S. A., et al. (2010). "Progressive ventricular expansion in chronic poor-outcome

schizophrenia." Cogn Behav Neurol 23(2): 85-88.

Mitelman, S. A., et al. (2006). "White matter fractional anisotropy and outcome in

schizophrenia." Schizophr Res 87(1-3): 138-159.

Moore, J. H., et al. (2010). "Bioinformatics challenges for genome-wide association studies."

Bioinformatics 26(4): 445-455.

Moosmang, S., et al. (2005). "Role of hippocampal Cav1.2 Ca2+ channels in NMDA receptor-

independent synaptic plasticity and spatial memory." J Neurosci 25(43): 9883-9892.

Moreau, M. P., et al. (2011). "Altered microRNA expression profiles in postmortem brain

samples from individuals with schizophrenia and bipolar disorder." Biol Psychiatry 69(2): 188-

193.

Morrow, E. M., et al. (2008). "Identifying autism loci and genes by tracing recent shared

ancestry." Science (New York, N.Y.) 321(5886): 218-223.

Mortensen, P. B., et al. (2007). "Toxoplasma gondii as a risk factor for early-onset

schizophrenia: analysis of filter paper blood samples obtained at birth." Biol Psychiatry 61(5):

688-693.

Mostofsky, S. H., et al. (2009). "Decreased connectivity and cerebellar activity in autism during

motor task performance." Brain: A Journal of Neurology 132(Pt 9): 2413-2425.

196

Mulquiney, P. G., et al. (2011). "Improving working memory: exploring the effect of transcranial

random noise stimulation and transcranial direct current stimulation on the dorsolateral

prefrontal cortex." Clin Neurophysiol 122(12): 2384-2389.

Munafo, M. R., et al. (2009). "5-HTTLPR genotype and anxiety-related personality traits: a

meta-analysis and new data." Am J Med Genet B Neuropsychiatr Genet 150B(2): 271-281.

Murray, C. J. and A. D. Lopez (1997). "Global mortality, disability, and the contribution of risk

factors: Global Burden of Disease Study." Lancet 349(9063): 1436-1442.

Murray, R. M. and S. W. Lewis (1987). "Is schizophrenia a neurodevelopmental disorder?" Br

Med J (Clin Res Ed) 295(6600): 681-682.

Nasrallah, H. A. (2008). "Atypical antipsychotic-induced metabolic side effects: insights from

receptor-binding profiles." Mol Psychiatry 13(1): 27-35.

Need, A. C., et al. (2009). "A genome-wide investigation of SNPs and CNVs in schizophrenia."

PLoS Genetics 5(2): e1000373.

Newcomer, J. W. (2007). "Metabolic considerations in the use of antipsychotic medications: a

review of recent evidence." J Clin Psychiatry 68 Suppl 1: 20-27.

Nicodemus, K. K., et al. (2010). "Evidence of statistical epistasis between DISC1, CIT and

NDEL1 impacting risk for schizophrenia: biological validation with functional neuroimaging."

Hum Genet 127(4): 441-452.

Nilsson, N. J. (1996). "Introduction to machine learning. An early draft of a proposed textbook."

Novak, G., et al. (2009). "Association of a polymorphism in the NRXN3 gene with the degree of

smoking in schizophrenia: a preliminary study." The World Journal of Biological Psychiatry:

The Official Journal of the World Federation of Societies of Biological Psychiatry 10(4 Pt 3):

929-935.

Numata, S., et al. (2012). "DNA methylation signatures in development and aging of the human

prefrontal cortex." Am J Hum Genet 90(2): 260-272.

Nussbaum, J., et al. (2008). "Significant association of the neurexin-1 gene (NRXN1) with

nicotine dependence in European- and African-American smokers." Human Molecular Genetics

17(11): 1569-1577.

Nyegaard, M., et al. (2010). "CACNA1C (rs1006737) is associated with schizophrenia."

Molecular psychiatry 15(2): 119-121.

O'Donovan, M. C., et al. (2008). "Identification of loci associated with schizophrenia by

genome-wide association and follow-up." Nature Genetics 40(9): 1053-1055.

Olabi, B., et al. (2011). "Are there progressive brain changes in schizophrenia? A meta-analysis

of structural magnetic resonance imaging studies." Biol Psychiatry 70(1): 88-96.

197

Olde Loohuis, N. F., et al. (2012). "MicroRNA networks direct neuronal development and

plasticity." Cell Mol Life Sci 69(1): 89-102.

Oldfield, R. C. (1971). "The assessment and analysis of handedness: the Edinburgh inventory."

Neuropsychologia 9(1): 97-113.

Orru, G., et al. (2012). "Using Support Vector Machine to identify imaging biomarkers of

neurological and psychiatric disease: a critical review." Neurosci Biobehav Rev 36(4): 1140-

1152.

Palmer, B. W., et al. (1997). "Is it possible to be schizophrenic yet neuropsychologically

normal?" Neuropsychology 11(3): 437-446.

Panizzon, M. S., et al. (2009). "Distinct genetic influences on cortical surface area and cortical

thickness." Cerebral Cortex 19(11): 2728-2735.

Patenaude, B., et al. (2011). "A Bayesian model of shape and appearance for subcortical brain

segmentation." Neuroimage 56(3): 907-922.

Paulus, F. M., et al. (2013). "Association of rs1006737 in CACNA1C with alterations in

prefrontal activation and fronto-hippocampal connectivity." Hum Brain Mapp.

Penadés, R., et al. (2013). "Brain Effects of Cognitive Remediation Therapy in Schizophrenia: A

Structural and Functional Neuroimaging Study." Biological Psychiatry.

Peng, H., et al. (2012). "High-frequency rTMS treatment increases white matter FA in the left

middle frontal gyrus in young patients with treatment-resistant depression." J Affect Disord

136(3): 249-257.

Peng, R. Y., et al. (1990). "A D2 dopamine receptor agonist disrupts sensorimotor gating in rats.

Implications for dopaminergic abnormalities in schizophrenia." Neuropsychopharmacology 3(3):

211-218.

Penke, L., et al. (2012). "Brain white matter tract integrity as a neural foundation for general

intelligence." Mol Psychiatry 17(10): 1026-1030.

Pereira, F., et al. (2009). "Machine learning classifiers and fMRI: a tutorial overview."

Neuroimage 45(1 Suppl): S199-209.

Perlstein, W. M., et al. (1998). "The Stroop task and attention deficits in schizophrenia: a critical

evaluation of card and single-trial Stroop methodologies." Neuropsychology 12(3): 414-425.

Pettersson-Yeo, W., et al. (2011). "Dysconnectivity in schizophrenia: where are we now?"

Neurosci Biobehav Rev 35(5): 1110-1124.

Pfefferbaum, A., et al. (2001). "Genetic regulation of regional microstructure of the corpus

callosum in late life." Neuroreport 12(8): 1677-1681.

198

Pidsley, R. and J. Mill (2011). "Epigenetic studies of psychosis: current findings, methodological

approaches, and implications for postmortem research." Biol Psychiatry 69(2): 146-156.

Pinto, D., et al. (2014). "Convergence of genes and cellular pathways dysregulated in autism

spectrum disorders." Am J Hum Genet 94(5): 677-694.

Pinto, D., et al. (2010). "Functional impact of global rare copy number variation in autism

spectrum disorders." Nature 466(7304): 368-372.

Pirkola, T., et al. (2005). "Spatial working memory function in twins with schizophrenia and

bipolar disorder." Biol Psychiatry 58(12): 930-936.

Plude, D. J. and W. J. Hoyer (1981). "Adult age differences in visual search as a function of

stimulus mapping and processing load." Journal of gerontology 36(5): 598-604.

Preacher, K. J. and A. F. Hayes (2008). "Asymptotic and resampling strategies for assessing and

comparing indirect effects in multiple mediator models." Behav Res Methods 40(3): 879-891.

Psychiatric, G. C. B. D. W. G. (2011). "Large-scale genome-wide association analysis of bipolar

disorder identifies a new susceptibility locus near ODZ4." Nat Genet 43(10): 977-983.

Psychiatric, G. C. S. C. (2009). "A framework for interpreting genome-wide association studies

of psychiatric disorders." Mol Psychiatry 14(1): 10-17.

Puls, I., et al. (2009). "A model comparison of COMT effects on central processing of affective

stimuli." Neuroimage 46(3): 683-691.

Purcell, S. M., et al. (2014). "A polygenic burden of rare disruptive mutations in schizophrenia."

Nature 506(7487): 185-190.

Radua, J., et al. (2010). "Voxel-based meta-analysis of regional white-matter volume differences

in autism spectrum disorder versus healthy controls." Psychological Medicine: 1-12.

Rajji, T. K., et al. (2009). "Age at onset and cognition in schizophrenia: meta-analysis." Br J

Psychiatry 195(4): 286-293.

Rajji, T. K. and B. H. Mulsant (2008). "Nature and course of cognitive function in late-life

schizophrenia: a systematic review." Schizophr Res 102(1-3): 122-140.

Rajkowska, G., et al. (1998). "Neuronal and glial somal size in the prefrontal cortex: a

postmortem morphometric study of schizophrenia and Huntington disease." Arch Gen Psychiatry

55(3): 215-224.

Randolph, C., et al. (1998). "The Repeatable Battery for the Assessment of Neuropsychological

Status (RBANS): preliminary clinical validity." J Clin Exp Neuropsychol 20(3): 310-319.

Rao, H., et al. (2008). "A regulation role of the prefrontal cortex in the fist-edge-palm task:

evidence from functional connectivity analysis." Neuroimage 41(4): 1345-1351.

199

Rasetti, R., et al. (2011). "Altered cortical network dynamics: a potential intermediate phenotype

for schizophrenia and association with ZNF804A." Arch Gen Psychiatry 68(12): 1207-1217.

Raznahan, A., et al. (2010). "Cortical anatomy in human X monosomy." Neuroimage 49(4):

2915-2923.

Reichert, A., et al. (2008). "The psychopathological and psychosocial outcome of early-onset

schizophrenia: preliminary data of a 13-year follow-up." Child Adolesc Psychiatry Ment Health

2(1): 6.

Reitan, R. M. and D. Wolfson (1985). The Halstead-Reitan Neuropsychological Test Battery:

Theory and clinical interpretation. Tucson, AZ, Neuropsychology Press.

Reitan, R. M. and D. Wolfson (1993). The Halstead-Reitan Neuropsychological Test Battery:

Theory and clinical interpretation. Tucson, AZ, Neuropsychology Press.

Riley, B., et al. (2010). "Replication of association between schizophrenia and ZNF804A in the

Irish Case-Control Study of Schizophrenia sample." Molecular psychiatry 15(1): 29-37.

Rimol, L. M., et al. (2010). "Cortical thickness is influenced by regionally specific genetic

factors." Biol Psychiatry 67(5): 493-499.

Ripke, S., et al. (2013). "Genome-wide association analysis identifies 13 new risk loci for

schizophrenia." Nature Genetics 45(10): 1150-1159.

Ripke, S., et al. (2011). "Genome-wide association study identifies five new schizophrenia loci."

Nat Genet 43(10): 969-976.

Rizos, E., et al. (2014). "A longitudinal study of alterations of hippocampal volumes and serum

BDNF levels in association to atypical antipsychotics in a sample of first-episode patients with

schizophrenia." PLoS One 9(2): e87997.

Roffman, J. L., et al. (2008). "Interactive effects of COMT Val108/158Met and MTHFR C677T

on executive function in schizophrenia." Am J Med Genet B Neuropsychiatr Genet 147B(6):

990-995.

Rose, S. E., et al. (2006). "Evidence of altered prefrontal-thalamic circuitry in schizophrenia: an

optimized diffusion MRI study." Neuroimage 32(1): 16-22.

Ruderfer, D. M., et al. (2013). "Polygenic dissection of diagnosis and clinical dimensions of

bipolar disorder and schizophrenia." Mol Psychiatry.

Ruff, R. M., et al. (1989). "Selective Reminding Tests: a normative study of verbal learning in

adults." J Clin Exp Neuropsychol 11(4): 539-550.

Rujescu, D., et al. (2009). "Disruption of the neurexin 1 gene is associated with schizophrenia."

Human Molecular Genetics 18(5): 988-996.

200

Saha, S., et al. (2005). "A systematic review of the prevalence of schizophrenia." PLoS Med

2(5): e141.

Sanfilipo, M., et al. (2000). "Volumetric measure of the frontal and temporal lobe regions in

schizophrenia: relationship to negative symptoms." Arch Gen Psychiatry 57(5): 471-480.

Santarelli, D. M., et al. (2011). "Upregulation of dicer and microRNA expression in the

dorsolateral prefrontal cortex Brodmann area 46 in schizophrenia." Biological Psychiatry 69(2):

180-187.

Schaid, D. J., et al. (2002). "Score tests for association between traits and haplotypes when

linkage phase is ambiguous." American Journal of Human Genetics 70(2): 425-434.

Schimmelmann, B. G., et al. (2013). "Treatment of adolescents with early-onset schizophrenia

spectrum disorders: in search of a rational, evidence-informed approach." Current opinion in

psychiatry 26(2): 219-230.

Schizophrenia Psychiatric Genome-Wide Association Study, C. (2011). "Genome-wide

association study identifies five new schizophrenia loci." Nat Genet 43(10): 969-976.

Schizophrenia Working Group of the Psychiatric Genomics, C. (2014). "Biological insights from

108 schizophrenia-associated genetic loci." Nature 511(7510): 421-427.

Schmitt, J. E., et al. (2008). "Identification of genetically mediated cortical networks: a

multivariate study of pediatric twins and siblings." Cereb Cortex 18(8): 1737-1747.

Scott-Van Zeeland, A. A., et al. (2010). "Altered functional connectivity in frontal lobe circuits

is associated with variation in the autism risk gene CNTNAP2." Science Translational Medicine

2(56): 56ra80

Seidman, L. J. (1990). "The neuropsychology of schizophrenia: a neurodevelopmental and case

study approach." J Neuropsychiatry Clin Neurosci 2(3): 301-312.

Selemon, L. D., et al. (2003). "Regional specificity in the neuropathologic substrates of

schizophrenia: a morphometric analysis of Broca's area 44 and area 9." Arch Gen Psychiatry

60(1): 69-77.

Shah, A. K., et al. (2010). "Rare NRXN1 promoter variants in patients with schizophrenia."

Neuroscience Letters 475(2): 80-84.

Shattuck, D. W., et al. (2008). "Construction of a 3D probabilistic atlas of human cortical

structures." Neuroimage 39(3): 1064-1080.

Shenton, M. E., et al. (2001). "A review of MRI findings in schizophrenia." Schizophr Res 49(1-

2): 1-52.

Shi, J., et al. (2009). "Common variants on chromosome 6p22.1 are associated with

schizophrenia." Nature 460(7256): 753-757.

201

Shinkareva, S. V., et al. (2006). "Classification of functional brain images with a spatio-temporal

dissimilarity map." Neuroimage 33(1): 63-71.

Shukla, D. K., et al. (2010). "White matter compromise of callosal and subcortical fiber tracts in

children with autism spectrum disorder: a diffusion tensor imaging study." Journal of the

American Academy of Child and Adolescent Psychiatry 49(12): 1269-1278, 1278.e1261-1262.

Siddiqui, T. J., et al. (2010). "LRRTMs and neuroligins bind neurexins with a differential code to

cooperate in glutamate synapse development." The Journal of Neuroscience: The Official

Journal of the Society for Neuroscience 30(22): 7495-7506.

Sigman, M. and J. Ungerer (1981). "Sensorimotor skills and language comprehension in autistic

children." J Abnorm Child Psychol 9(2): 149-165.

Silber, J., et al. (2008). "miR-124 and miR-137 inhibit proliferation of glioblastoma multiforme

cells and induce differentiation of brain tumor stem cells." BMC Med 6: 14.

Skudlarski, P., et al. (2013). "Diffusion tensor imaging white matter endophenotypes in patients

with schizophrenia or psychotic bipolar disorder and their relatives." Am J Psychiatry 170(8):

886-898.

Sled, J. G., et al. (1998). "A nonparametric method for automatic correction of intensity

nonuniformity in MRI data." IEEE Trans Med Imaging 17(1): 87-97.

Smit, D. J., et al. (2008). "Heritability of "small-world" networks in the brain: a graph theoretical

analysis of resting-state EEG functional connectivity." Hum Brain Mapp 29(12): 1368-1378.

Smith, G. N., et al. (2012). "Hippocampal volume and the brain-derived neurotrophic factor

Val66Met polymorphism in first episode psychosis." Schizophr Res 134(2-3): 253-259.

Smith, R. C., et al. (1984). "Nuclear magnetic resonance in schizophrenia: a preliminary study."

Psychiatry Res 12(2): 137-147.

Smith, S. M., et al. (2006). "Tract-based spatial statistics: voxelwise analysis of multi-subject

diffusion data." Neuroimage 31(4): 1487-1505.

Smith, S. M., et al. (2004). "Advances in functional and structural MR image analysis and

implementation as FSL." Neuroimage 23: S208-S219.

Smith, S. M. and T. E. Nichols (2009). "Threshold-free cluster enhancement: addressing

problems of smoothing, threshold dependence and localisation in cluster inference." Neuroimage

44(1): 83-98.

Smith, S. M., et al. (2002). "Accurate, robust, and automated longitudinal and cross-sectional

brain change analysis." Neuroimage 17(1): 479-489.

Smolka, M. N., et al. (2007). "Gene-gene effects on central processing of aversive stimuli." Mol

Psychiatry 12(3): 307-317.

202

Smrt, R. D., et al. (2010). "MicroRNA miR-137 regulates neuronal maturation by targeting

ubiquitin ligase mind bomb-1." Stem Cells 28(6): 1060-1070.

Sohal, V. S., et al. (2009). "Parvalbumin neurons and gamma rhythms enhance cortical circuit

performance." Nature 459(7247): 698-702.

Sommer, I., et al. (2001). "Handedness, language lateralisation and anatomical asymmetry in

schizophrenia: meta-analysis." Br J Psychiatry 178: 344-351.

Souza, R. P., et al. (2010). "Influence of neurexin 1 (NRXN1) polymorphisms in clozapine

response." Hum Psychopharmacol 25(7-8): 582-585.

Spencer, K. M. (2009). "The functional consequences of cortical circuit abnormalities on gamma

oscillations in schizophrenia: insights from computational modeling." Front Hum Neurosci 3: 33.

Spencer, K. M., et al. (2003). "Abnormal neural synchrony in schizophrenia." J Neurosci 23(19):

7407-7411.

Spencer, K. M., et al. (2004). "Neural synchrony indexes disordered perception and cognition in

schizophrenia." Proc Natl Acad Sci U S A 101(49): 17288-17293.

Spencer, K. M., et al. (2008). "Gamma-band auditory steady-state responses are impaired in first

episode psychosis." Biol Psychiatry 64(5): 369-375.

Steen, R. G., et al. (2006). "Brain volume in first-episode schizophrenia: systematic review and

meta-analysis of magnetic resonance imaging studies." Br J Psychiatry 188: 510-518.

Stefansson, H., et al. (2009). "Common variants conferring risk of schizophrenia." Nature

460(7256): 744-747.

Stein, J. L., et al. (2010). "Voxelwise genome-wide association study (vGWAS)." Neuroimage

53(3): 1160-1174.

Stein, J. L., et al. (2012). "Identification of common variants associated with human hippocampal

and intracranial volumes." Nat Genet 44(5): 552-561.

Steinberg, S., et al. (2010). "Expanding the range of ZNF804A variants conferring risk of

psychosis." Molecular psychiatry 16(1): 59-66.

Straub, R. E., et al. (2007). "Allelic variation in GAD1 (GAD67) is associated with

schizophrenia and influences cortical function and gene expression." Mol Psychiatry 12(9): 854-

869.

Stroop, J. R. (1935). "Studies of interference in serial verbal reactions." Journal of experimental

psychology 18(6): 643.

Struyf, J., et al. (2008). "Combining gene expression, demographic and clinical data in modeling

disease: a case study of bipolar disorder and schizophrenia." BMC Genomics 9: 531.

203

Subramanian, A., et al. (2005). "Gene set enrichment analysis: a knowledge-based approach for

interpreting genome-wide expression profiles." Proc Natl Acad Sci U S A 102(43): 15545-

15550.

Südhof, T. C. (2008). "Neuroligins and neurexins link synaptic function to cognitive disease."

Nature 455(7215): 903-911.

Sullivan, P. F., et al. (2012). "Genetic architectures of psychiatric disorders: the emerging picture

and its implications." Nat Rev Genet 13(8): 537-551.

Sun, G., et al. (2011). "miR-137 forms a regulatory loop with nuclear receptor TLX and LSD1 in

neural stem cells." Nat Commun 2: 529.

Sun, J., et al. (2012). "Uncovering microRNA and transcription factor mediated regulatory

networks in glioblastoma." PLoS Computational Biology 8(7): e1002488.

Sundaram, S. K., et al. (2008). "Diffusion tensor imaging of frontal lobe in autism spectrum

disorder." Cerebral Cortex (New York, N.Y.: 1991) 18(11): 2659-2665.

Suvisaari, J. M., et al. (1998). "Age at onset and outcome in schizophrenia are related to the

degree of familial loading." Br J Psychiatry 173: 494-500.

Szatmari, P., et al. (2007). "Mapping autism risk loci using genetic linkage and chromosomal

rearrangements." Nature Genetics 39(3): 319-328.

Szeszko, P. R., et al. (2005). "Brain-derived neurotrophic factor val66met polymorphism and

volume of the hippocampal formation." Mol Psychiatry 10(7): 631-636.

Szeszko, P. R., et al. (2008). "Clinical and neuropsychological correlates of white matter

abnormalities in recent onset schizophrenia." Neuropsychopharmacology 33(5): 976-984.

Szulwach, K. E., et al. (2010). "Cross talk between microRNA and epigenetic regulation in adult

neurogenesis." J Cell Biol 189(1): 127-141.

Szymanski, S., et al. (1995). "Gender differences in onset of illness, treatment response, course,

and biologic indexes in first-episode schizophrenic patients." Am J Psychiatry 152(5): 698-703.

Takarae, Y., et al. (2007). "Atypical involvement of frontostriatal systems during sensorimotor

control in autism." Psychiatry Research: Neuroimaging 156(2): 117-127.

Tan, G. C. Y., et al. (2010). "Normal variation in fronto-occipital circuitry and cerebellar

structure with an autism-associated polymorphism of CNTNAP2." Neuroimage.

Tan, H. Y., et al. (2008). "Intermediate phenotypes in schizophrenia genetics redux: is it a no

brainer?" Molecular psychiatry 13(3): 233-238.

Tan, H. Y., et al. (2007). "Epistasis between catechol-O-methyltransferase and type II

metabotropic glutamate receptor 3 genes on working memory brain function." Proc Natl Acad

Sci U S A 104(30): 12536-12541.

204

Tan, H. Y., et al. (2005). "fMRI study of maintenance and manipulation processes within

working memory in first-episode schizophrenia." Am J Psychiatry 162(10): 1849-1858.

Terwisscha van Scheltinga, A. F., et al. (2013). "Genetic schizophrenia risk variants jointly

modulate total brain and white matter volume." Biol Psychiatry 73(6): 525-531.

Thompson, P. M., et al. (2014). "The ENIGMA Consortium: large-scale collaborative analyses

of neuroimaging and genetic data." Brain Imaging Behav.

Thorisson, G. A., et al. (2005). "The International HapMap Project Web site." Genome Research

15(11): 1592-1593.

Thyagarajan, A. and A. Y. Ting (2010). "Imaging activity-dependent regulation of neurexin-

neuroligin interactions using trans-synaptic enzymatic biotinylation." Cell 143(3): 456-469.

Tiihonen, J., et al. (2009). "11-year follow-up of mortality in patients with schizophrenia: a

population-based cohort study (FIN11 study)." Lancet 374(9690): 620-627.

Tohka, J., et al. (2004). "Fast and robust parameter estimation for statistical partial volume

models in brain MRI." Neuroimage 23(1): 84-97.

Tooney, P. A. and L. A. Chahl (2004). "Neurons expressing calcium-binding proteins in the

prefrontal cortex in schizophrenia." Prog Neuropsychopharmacol Biol Psychiatry 28(2): 273-

278.

Torres, U. S., et al. (2013). "Structural brain changes associated with antipsychotic treatment in

schizophrenia as revealed by voxel-based morphometric MRI: an activation likelihood

estimation meta-analysis." BMC Psychiatry 13: 342.

Torrey, E. F., et al. (2005). "Neurochemical markers for schizophrenia, bipolar disorder, and

major depression in postmortem brains." Biol Psychiatry 57(3): 252-260.

Torrey, E. F., et al. (2012). "Toxoplasma gondii and other risk factors for schizophrenia: an

update." Schizophr Bull 38(3): 642-647.

Tura, E., et al. (2008). "Multivariate analyses suggest genetic impacts on neurocircuitry in

schizophrenia." Neuroreport 19(6): 603-607.

Tuulio-Henriksson, A., et al. (2002). "Heritability and number of quantitative trait loci of

neurocognitive functions in families with schizophrenia." Am J Med Genet 114(5): 483-490.

Uranova, N., et al. (2001). "Electron microscopy of oligodendroglia in severe mental illness."

Brain Res Bull 55(5): 597-610.

Utz, K. S., et al. (2010). "Electrified minds: transcranial direct current stimulation (tDCS) and

galvanic vestibular stimulation (GVS) as methods of non-invasive brain stimulation in

neuropsychology--a review of current data and future implications." Neuropsychologia 48(10):

2789-2810.

205

Vacic, V., et al. (2011). "Duplications of the neuropeptide receptor gene VIPR2 confer

significant risk for schizophrenia." Nature 471(7339): 499-503.

van den Heuvel, M. P., et al. (2010). "Aberrant frontal and temporal complex network structure

in schizophrenia: a graph theoretical analysis." J Neurosci 30(47): 15915-15926.

van den Heuvel, M. P., et al. (2013). "Genetic control of functional brain network efficiency in

children." Eur Neuropsychopharmacol 23(1): 19-23.

van Erp, T. G., et al. (2014). "Schizophrenia miR-137 Locus Risk Genotype Is Associated with

Dorsolateral Prefrontal Cortex Hyperactivation." Biol Psychiatry 75(5): 398-405.

van Haren, N. E., et al. (2011). "Changes in cortical thickness during the course of illness in

schizophrenia." Arch Gen Psychiatry 68(9): 871-880.

Vinogradov, S., et al. (2009). "Is serum brain-derived neurotrophic factor a biomarker for

cognitive enhancement in schizophrenia?" Biol Psychiatry 66(6): 549-553.

Vita, A., et al. (2006). "Brain morphology in first-episode schizophrenia: a meta-analysis of

quantitative magnetic resonance imaging studies." Schizophr Res 82(1): 75-88.

Voineskos, A. N., et al. (2010). "The role of the corpus callosum in transcranial magnetic

stimulation induced interhemispheric signal propagation." Biol Psychiatry 68(9): 825-831.

Voineskos, A. N., et al. (2013). "Oligodendrocyte genes, white matter tract integrity, and

cognition in schizophrenia." Cereb Cortex 23(9): 2044-2057.

Voineskos, A. N., et al. (2013). "Neuroimaging evidence for the deficit subtype of

schizophrenia." JAMA Psychiatry 70(5): 472-480.

Voineskos, A. N., et al. (2011). "The brain-derived neurotrophic factor Val66Met polymorphism

and prediction of neural risk for Alzheimer disease." Arch Gen Psychiatry 68(2): 198-206.

Voineskos, A. N., et al. (2011). "The ZNF804A gene: characterization of a novel neural risk

mechanism for the major psychoses." Neuropsychopharmacology 36(9): 1871-1878.

Voineskos, A. N., et al. (2010). "Diffusion tensor tractography findings in schizophrenia across

the adult lifespan." Brain 133(5): 1494-1504.

Voineskos, A. N., et al. (2012). "Age-related decline in white matter tract integrity and cognitive

performance: a DTI tractography and structural equation modeling study." Neurobiol Aging

33(1): 21-34.

Vounou, M., et al. (2012). "Sparse reduced-rank regression detects genetic associations with

voxel-wise longitudinal phenotypes in Alzheimer's disease." Neuroimage 60(1): 700-716.

Vourdas, A., et al. (2003). "Increased developmental deviance and premorbid dysfunction in

early onset schizophrenia." Schizophr Res 62(1-2): 13-22.

206

Vrijenhoek, T., et al. (2008). "Recurrent CNVs disrupt three candidate genes in schizophrenia

patients." American Journal of Human Genetics 83(4): 504-510.

Vyas, N. S., et al. (2011). "Neurobiology and phenotypic expression in early onset

schizophrenia." Early Interv Psychiatry 5(1): 3-14.

Waddell, C., et al. (2005). "A public health strategy to improve the mental health of Canadian

children." Can J Psychiatry 50(4): 226-233.

Wahl, M., et al. (2010). "Microstructural correlations of white matter tracts in the human brain."

Neuroimage 51(2): 531-541.

Wakana, S., et al. (2007). "Reproducibility of quantitative tractography methods applied to

cerebral white matter." Neuroimage 36(3): 630-644.

Walsh, T., et al. (2008). "Rare structural variants disrupt multiple genes in neurodevelopmental

pathways in schizophrenia." Science 320(5875): 539-543.

Walter, H., et al. (2011). "Effects of a genome-wide supported psychosis risk variant on neural

activation during a theory-of-mind task." Mol Psychiatry 16(4): 462-470.

Walters, J. T., et al. (2010). "Psychosis susceptibility gene ZNF804A and cognitive performance

in schizophrenia." Arch Gen Psychiatry 67(7): 692-700.

Walton, E., et al. (2013). "Cumulative genetic risk and prefrontal activity in patients with

schizophrenia." Schizophr Bull 39(3): 703-711.

Wang, K. S., et al. (2011). "Genome-wide association analysis of age at onset in schizophrenia in

a European-American sample." Am J Med Genet B Neuropsychiatr Genet 156B(6): 671-680.

Wang, Q., et al. (2013). "White-matter microstructure in previously drug-naive patients with

schizophrenia after 6 weeks of treatment." Psychol Med 43(11): 2301-2309.

Wechsler, D. (2001). Wechsler Test of Adult Reading: WTAR, Psychological Corporation.

Weickert, T. W., et al. (2000). "Cognitive impairments in patients with schizophrenia displaying

preserved and compromised intellect." Arch Gen Psychiatry 57(9): 907-913.

Weinberger, D. R. (1987). "Implications of normal brain development for the pathogenesis of

schizophrenia." Arch Gen Psychiatry 44(7): 660-669.

Weinberger, D. R. and B. K. Lipska (1995). "Cortical maldevelopment, anti-psychotic drugs, and

schizophrenia: a search for common ground." Schizophr Res 16(2): 87-110.

Welham, J., et al. (2009). "The antecedents of schizophrenia: a review of birth cohort studies."

Schizophr Bull 35(3): 603-623.

Wexler, B. E., et al. (2000). "Preliminary evidence of improved verbal working memory

performance and normalization of task-related frontal lobe activation in schizophrenia following

cognitive exercises." Am J Psychiatry 157(10): 1694-1697.

207

Whalley, H. C., et al. (2012). "Impact of a microRNA MIR137 Susceptibility Variant on Brain

Function in People at High Genetic Risk of Schizophrenia or Bipolar Disorder."

Neuropsychopharmacology.

Whalley, H. C., et al. (2012). "The influence of polygenic risk for bipolar disorder on neural

activation assessed using fMRI." Transl Psychiatry 2: e130.

Whalley, H. C., et al. (2013). "Polygenic risk and white matter integrity in individuals at high

risk of mood disorder." Biol Psychiatry 74(4): 280-286.

Willemsen, M. H., et al. (2011). "Chromosome 1p21.3 microdeletions comprising DPYD and

MIR137 are associated with intellectual disability." J Med Genet 48(12): 810-818.

Williams, A. H., et al. (2009). "MicroRNA-206 delays ALS progression and promotes

regeneration of neuromuscular synapses in mice." Science 326(5959): 1549-1554.

Williams, D. L. and N. J. Minshew (2007). "Understanding autism and related disorders: what

has imaging taught us?" Neuroimaging Clinics of North America 17(4): 495-509, ix.

Winham, S. (2013). "Applications of multifactor dimensionality reduction to genome-wide data

using the R package 'MDR'." Methods Mol Biol 1019: 479-498.

Wolkin, A., et al. (2003). "Inferior frontal white matter anisotropy and negative symptoms of

schizophrenia: a diffusion tensor imaging study." Am J Psychiatry 160(3): 572-574.

Woo, T. U., et al. (1997). "Schizophrenia and the parvalbumin-containing class of cortical local

circuit neurons." Am J Psychiatry 154(7): 1013-1015.

Woodruff, P. W., et al. (1995). "Meta-analysis of corpus callosum size in schizophrenia." J

Neurol Neurosurg Psychiatry 58(4): 457-461.

Woodward, N. D., et al. (2005). "A meta-analysis of neuropsychological change to clozapine,

olanzapine, quetiapine, and risperidone in schizophrenia." Int J Neuropsychopharmacol 8(3):

457-472.

Wright, C., et al. (2013). "Potential Impact of miR-137 and Its Targets in Schizophrenia." Front

Genet 4: 58.

Wright, I. C., et al. (2000). "Meta-analysis of regional brain volumes in schizophrenia." Am J

Psychiatry 157(1): 16-25.

Wykes, T., et al. (2011). "A meta-analysis of cognitive remediation for schizophrenia:

methodology and effect sizes." Am J Psychiatry 168(5): 472-485.

Xiao, Y., et al. (2013). "Altered Cortical Thickness Related to Clinical Severity But Not the

Untreated Disease Duration in Schizophrenia." Schizophr Bull.

Yan, J., et al. (2008). "Neurexin 1alpha structural variants associated with autism." Neuroscience

Letters 438(3): 368-370.

208

Yang, H.-C., et al. (2005). "A genome-wide scanning and fine mapping study of COGA data."

BMC Genetics 6 Suppl 1: S30.

Yang, H., et al. (2010). "A Hybrid Machine Learning Method for Fusing fMRI and Genetic

Data: Combining both Improves Classification of Schizophrenia." Front Hum Neurosci 4: 192.

Zahir, F. R., et al. (2008). "A patient with vertebral, cognitive and behavioural abnormalities and

a de novo deletion of NRXN1alpha." Journal of Medical Genetics 45(4): 239-243.

Zhang, Q., et al. (2012). "The effects of CACNA1C gene polymorphism on spatial working

memory in both healthy controls and patients with schizophrenia or bipolar disorder."

Neuropsychopharmacology 37(3): 677-684.

Zijdenbos, A. P., et al. (2002). "Automatic "pipeline" analysis of 3-D MRI data for clinical trials:

application to multiple sclerosis." IEEE Trans Med Imaging 21(10): 1280-1291.

Zipursky, R. B., et al. (1998). "Cerebral gray matter volume deficits in first episode psychosis."

Arch Gen Psychiatry 55(6): 540-546.

Zollner, S. and J. K. Pritchard (2007). "Overcoming the winner's curse: estimating penetrance

parameters from case-control data." Am J Hum Genet 80(4): 605-615.

Zweier, C., et al. (2009). "CNTNAP2 and NRXN1 are mutated in autosomal-recessive Pitt-

Hopkins-like mental retardation and determine the level of a common synaptic protein in

Drosophila." American Journal of Human Genetics 85(5): 655-666.

209

Appendices

Appendix 1: Lett TA, Tiwari AK, Meltzer HY, Lieberman JA, Potkin SG, Voineskos AN,

Kennedy JL, Müller DJ. The putative functional rs1045881 marker of neurexin-1 in

schizophrenia and clozapine response. Schizophr Res. 2011 Nov;132(2-3):121-4.

Appendix 2: Lett TA, Voineskos AN, Kennedy JL, Levine B, Daskalakis ZJ. Treating working

memory deficits in schizophrenia: a review of the neurobiology. Biol Psychiatry. 2014 Mar

1;75(5):361-70.

The putative functional rs1045881 marker of neurexin-1 in schizophrenia andclozapine response

Tristram A.P. Lett a, Arun K. Tiwari a, Herbert Y. Meltzer b, Jeffrey A. Lieberman c, Steven G. Potkin d,Aristotle N. Voineskos a, James L. Kennedy a, Daniel J. Müller a,⁎a Neurogenetics Section, Centre for Addiction and Mental Health, Toronto, Canadab Psychiatry and Pharmacology, Vanderbilt University, Nashville, TN, USAc Psychiatry, Columbia University, New York City, NY, USAd Department of Psychiatry & Human Behavior, University of California, Irvine, CA, USA

a b s t r a c ta r t i c l e i n f o

Article history:Received 15 May 2011Received in revised form 9 August 2011Accepted 12 August 2011Available online 3 September 2011

Keywords:Neurexin-1 geneNXRN1SchizophreniaGeneticsPharmacogeneticsAntipsychotic medicationClozapineProspective treatmentAntipsychotic responseBrief Psychiatric Rating Scale

Neurexin-1 (NRXN1) modulates recruitment of NMDA receptors. Furthermore, clozapine reduceshyperactivity of NMDA receptors. Thus, regulation of the NRXN1 gene may mediate the efficacy of clozapineat reducing cortical hyperactivity. We examined the putative functional SNP, rs1045881, for association withschizophrenia, and the potential role of this SNP in clozapine response. The rs1045881 variant was notsignificantly associated with schizophrenia (N=302 case–control pairs), but with clozapine response(N=163; p=0.030). Baseline and BPRS scores after six months revealed a trend for rs1045881 genotype bytreatment interaction (p=0.079). In the post hoc analysis, a significant association between BPRS negativesymptoms score and genotype was observed (p=0.033). These results suggest that the rs1045881 NRXN1polymorphism may influence clozapine response.

© 2011 Elsevier B.V. All rights reserved.

1. Introduction

Deletions in the neurexin-1 (NRXN1; 2p16.3; gene size=1.1 mb)gene have been strongly associated with the etiology of schizophre-nia, and autism spectrumdisorder (Voineskos et al., 2011). TheNRXN1gene encodes the neurexin-1α protein that functions as a pre-synaptic neural adhesion molecule reported to interact with post-synaptic neuroligins mediating GABAergic and glutamatergic synapsefunction (Südhof, 2008). Neurexin-1α knockout mice exhibit anelectrophysiological phenotype consistent with a network disruptionthat presents as a presynaptic loss of synaptic strength in excitatorysynapses of the hippocampus (Kehrer et al., 2008; Etherton et al.,2009). Recent evidence suggests NRXN1 also binds to leucine-richrepeat transmembrane protein (LRRTM2), that modulates postsyn-aptic differentiation of glutamatergic synapses (de Wit et al., 2009).

Therefore, NRXN1 may, at least partially, direct excitatory synapseformation. These findings are interesting in light of reports thatclozapine prevents phencyclidine-induced functional hyperactivity ofN-methyl D-aspartate receptors (NMDAR) in pyramidal cells in ratmedial prefrontal cortex (Arvanov and Wang, 1999; Ninan et al.,2003). Furthermore, clozapine is reported to differentially regulatedendritic spine formation and synaptogenesis in the rat hippocampalneurons (Critchlow et al., 2006). Recently, we have reported thatrs1045881 is located in a putative miRNA binding site that influencesfrontal lobe structural white matter volume and sensorimotorfunction (Voineskos et al., 2011). Altogether, variation in regulationof the NRXN1 gene may influence response to clozapine treatedschizophrenia patients.

In this study, we analyzed associations of rs1045881 in schizo-phrenia (SCZ) matched case-controls. We have a strong a priorihypothesis to examine the association between this high-interestmarker and SCZ because of our in silico, neuroimaging andneurobehavioral findings. Second, given the effect of clozapine onNMDAR function, and the role of NRXN1 in NMDAR recruitment, weexamine the role of the rs1045881 in prospectively assessedEuropean-American schizophrenia patients for clozapine treatmentresponse.

Schizophrenia Research 132 (2011) 121–124

⁎ Corresponding author at: Neurogenetics Section, Centre for Addiction and MentalHealth, R31 250 College Street, Toronto, Ontario, Canada M5T 1R8. Tel.: +1 (416) 5358501; fax: +1 (416) 979 4666.

E-mail address: daniel_mueller@camh.net (D.J. Müller).

0920-9964/$ – see front matter © 2011 Elsevier B.V. All rights reserved.doi:10.1016/j.schres.2011.08.007

Contents lists available at SciVerse ScienceDirect

Schizophrenia Research

j ourna l homepage: www.e lsev ie r.com/ locate /schres

2. Experimental/materials and methods

2.1. Participants

The case–control study was composed of 302 matched pairs thatwere recruited at the Centre for Addiction andMental Health. Patientsand controls were matched for sex, ethnicity, and age at recruitment(Table S1). Research interviews were conducted using the StructuredClinical Interview for Diagnostic and Statistical Manual of MentalDisorder IV (DSM-IV) Disorders. Patients with a history of majorneurological disorders, major substance abuse, and head injury withsignificant loss of consciousness were excluded from the study. Eachindividual of the control group was screened for history of majorpsychiatric disorders using the SCID-I, and only those without majorpsychiatric disorders were entered as healthy controls. Our pairedcase–control sample had over 80% power to detect an odds ratio aslow as 1.65 (α=0.05, minor allele frequency=0.141, dominantmodel; Quanto v1.2.4 (Gauderman, 2002)).

Our clozapine response sample consisted of 169 European–Americanschizophrenia patients obtained from three research clinics: CaseWestern Reserve University in Cleveland, Ohio (HY Meltzer, N=68);Hillside Hospital in Glen Oaks, New York (JA Lieberman, N=73); andUniversity of California at Irvine (SG Potkin, N=28). These subjects hadDSM-III-R or DSM-IV diagnoses of SCZ andmet the criteria for treatmentrefractoriness or intolerance to traditional antipsychotic therapy. After a2- to 4-week washout period, patients were treated with clozapine andevaluated prospectively for 6 months and clozapine blood levels weremonitored. Treatment response was evaluated using the 18-item BriefPsychiatric Rating Scale (BPRS) at the time of enrolment into the study(baseline) and after 6 months of clozapine treatment. Differences inresponse rates across clinical sites were not observed (χ2=0.901,df=2, P=0.637, Table S2). Therefore, data from the three clinical siteswere analyzed together. Our sample had 80% power to detect an oddsratio of 2.0 at a non-responder frequency of 40% (unmatched case–control design; α=0.05, minor allele frequency=0.141, dominantmodel; Quanto v1.2.4 (Gauderman, 2002)). In our categorical responsemeasure sample, treatment response was analyzed as a dichotomousvariable. Responders were defined as a reduction N20% on the overallBPRS score after 6 monthsof treatmentbasedon the criteria proposedby(Kane et al., 1988). Quantitative treatment response data was availableonly for a subset (total BPRS [N=91]; positive/negative symptomssubscale [N=87]). All experimental procedures were approved by localethics committee and all patients signed informed consent prior to theirparticipation, in accordance with the Declaration of Helsinki.

2.2. Genetics

Genomic data was extracted from venous blood (Lahiri andNurnberger, 1991). The rs1045881 variant was genotyped, usingTaqman 5′ nuclease assay (Applied Biosystems; Foster City, CA, USA).Genotyping accuracy was assessed by repeating 10% of the sample, andresults showed 100% concordance.

2.3. Statistical analysis

Analysis of SCZ cases versus matched controls was done using log-likelihood χ2 ratio test both in terms of allele frequencies andgenotype frequencies in UNPHASED 3.1 (Dudbridge, 2008). Haplo-view 4.2 was used to determine Hardy Weinberg equilibrium (HWE)(Barrett et al., 2005).

To test the effect of NRXN1 genotype on quantitative treatmentresponse, a repeated measure analysis of variance (RM ANOVA) testswere performed. NRXN1 genotype was the between-group factor, andBPRS scores at baseline and 6 months were the within-group factor.These analyses were performed using Statistical Package for the SocialSciences 15.0.0 (Chicago, IL, USA).

3. Results

3.1. Association with Schizophrenia

The rs1045881 polymorphism was in HWE in both cases andcontrols (pN0.05). No significant allelic or genotypic associationsbetween rs1045881 and SCZ in our matched case–control sampleswere detected (p=0.37; p=0.27, respectively; Table 1). Further-more, we found no significant associations in European–Americansalone suggesting that our results are not masked by populationstratification (Table S3).

3.2. Influence of NRXN1 on clozapine response

The rs1045881 variant did not deviate from HWE in responders ornon-responders groups (pN0.05). Furthermore, therewasno significantdifference in gender, age at onset, and treatment duration based onrs1045881 genotype (Table S4). Our categorical analysis found thers1045881C allele of NRXN1 to be associated with clozapine response(p=0.012, OR=2.199 [1.185–4.080]; Table 2). Additionally, the C/Cgenotype showed association with treatment response under adominant model (p=0.030, OR=2.153 [1.077–4.304]; Table 2). Thiswas further supported by the trend observed in our quantitative mea-sure of treatment response (RM ANOVA: F1,87 Within-subject=3.151p=0.079; Fig. 1; Table S4).

In post hoc analysis, we examined positive and negative symptomsubscales of the BPRS. RM ANOVA of negative symptoms showed asignificant genotype association (F1,85 Between-subject=4.686,p=0.033) and a trend for genotype by treatment response (F1,85Within-subject=3.293, p=0.075; Fig. 2; Table S5). In contrast, therewas no genotype or genotype by treatment response effect forpositive symptoms (Fig. 2; Table S5).

4. Discussion

Our results show an association between the rs1045881 andclozapine response. The rs1045881T allele was over-represented inthe non-responder group, suggesting that the rs1045881 variant ofNRXN1 may influence clozapine response. This is consistent withrecent findings by our group that found two other markers of NRXN1to be nominally associated with clozapine response (Souza et al.,2010) and our observed trend of association between rs1045881 andquantitative total BPRS treatment response. There was no observableeffect of rs1045881T on change in negative symptoms; althoughoverall, T-allele carriers had lower negative symptom scores. Thetrend seen in the C allele homozygotes suggests their responsivenessto clozapine treatment.

The core clinical symptoms of schizophrenia are negativesymptoms (Andreasen, 1982) which do not respond well to existingtreatment (Murphy et al., 2006). Previous structural and diffusiontension imaging MRI brain imaging finding have reported that ageneralized reduction in frontal lobe white matter correlate withgreater severity of negative symptoms (Sanfilipo et al., 2000; Wolkinet al., 2003). This suggests that the increased frontal lobewhitematterwe previously reported in T-allele carriers could be related to the

Table 1The association between schizophrenia diagnosis and genotypic and allele frequenciesof rs1045881.

Case Control

Schizophrenia versuscontrol

T 88 99 χ2 Allelic p-valueC 514 501 0.811 0.368T/T+T/C 82 94 χ2 Genotypic p-valueC/C 219 206 1.214 0.270

T-allele carriers versus C/C homozygotes were compared in genotypic test because cellswith less than five T/T homozygotes were observed. χ2 = Likelihood ratio chi squared.

122 T.A.P. Lett et al. / Schizophrenia Research 132 (2011) 121–124

lower negative symptoms scores in our clozapine response sample(Voineskos et al., 2011).

Our findings are also concordant with the notion that NMDAantagonism may account for negative symptoms in SCZ (Olney et al.,1999). In silico analysis predicts that the rs1045881T allele eliminatesmiRNA binding sites, thus reducing mRNA decay and T allele carriersmay therefore have increased levels of NRXN1mRNA and protein. Thisincrease in NRXN1 could explain the protective effect observed in Tallele carriers against negative symptoms through facilitation ofNMDAR recruitment. This is important because dysfunction ofNMDARs has been implicated in memory function, cognitivedisturbances and negative symptoms (Olney et al., 1999). Thus, itcould be argued that clozapine action on negative symptoms may beattenuated in T-allele carriers because NRXN1 may already beexpressed at high levels. Alternatively, the clozapine response in C/Chomozygotes could be explained by higher initial negative symptomsproviding greater opportunity to respond to medication.

The association between NRXN1 and SCZ has only previously beenreported with respect to deletions within the gene. It is possible thatmultiple insults to the NRXN1 gene culminating across a threshold maybe involved inSCZpathogenesis. For instance, a recent studybyShahet al.(2010), in which the promoter region of NRXN1 was re-sequenced,reported that rare point mutations in the promoter region in addition tochromosomal alterations may contribute to the etiology of SCZ.Therefore, we may not have captured enough of the variance in NRXN1with a single SNP to observe a significant associationwith schizophrenia.

Our study has some limitations. First, we imposed a dominantmodel by combining genotypic groups T/T and T/C of rs1045881. Ourmodel would need to be confirmed by further analysis; however,results from our previous imaging analysis support such a model(Voineskos et al., 2011), and the allelic association we observed inclozapine response supports a dominant model. Second, quantitativeresponse data was only available for only a subset of our clozapine

response sample; therefore, our analysis could be underpowered.Furthermore, our sample size for categorical response was relativelylow compared tomost studies, even thoughwe had over 80% power todetect association with SCZ and responders to clozapine treatment.

In conclusion, we show that a common variant within the NRXN1gene with a predicted functional effect may be associated withclozapine response in European-American SCZ patients. This tenta-tively suggests with replication and further work, that NRXN1screening may provide more efficient treatment strategies throughpersonalized medicine.

Supplementarymaterials related to this article can be found onlineat 10.1016/j.schres.2011.08.007.

Role of funding sourceCIHR operating grant to DJM (Genetics of antipsychotics induced metabolic

syndrome, MOP 89853); NARSAD Young Investigator Award to DJM, CIHR MichaelSmith New Investigator Salary Prize for Research in Schizophrenia to DJM, OMHF New

Fig. 1. Interaction between clozapine treatment duration and rs1045881 genotype. Thepercent difference BPRS scores at baseline and after 6 months of clozapine treatment islisted for T allele carriers and C/C genotypes. (N[T/T+T/C]=28, N[C/C]=63; BPRS =Brief Psychiatric Ratings Scale). Repeated measures ANOVA of baseline and 6 monthBPRS scores reveal a trend in rs1045881 genotype by treatment response (F1,87 Within-subject=3.151 p=0.079).

Fig. 2.Mean positive (A) and negative (B) symptom subscale from the BPRS for baselineto 6 months by genotype. (N [T/T+T/C]=27, N [C/C]=60; BPRS = Brief PsychiatricRatings Scale). (A) Repeated measures ANOVA were not significant for within- orbetween-subject effects. Negative symptoms showed a significant genotype association(F1,85 (Between-subject)=4.686, p=0.033) and a trend for genotype by treatmentresponse (F1,85 (Within-subject)=3.293, p=0.075). * denotes pb0.05.

Table 2NRXN1 rs1045881 marker allelic and genotypic associations with clozapine response.

Non-responders Responders

Allele T C T C χ2 Allelic p-value OR (95% CI)27 107 21 183 6.301 0.012 2.199 (1.185–4.080)

Genotype T/T+T/C C/C T/T+T/C C/C χ2 Genotypic p-value OR (95% CI)24 43 21 81 4.736 0.030 2.153 (1.077–4.304)

OR = odds ratio.

123T.A.P. Lett et al. / Schizophrenia Research 132 (2011) 121–124

Investigator Fellowship to DJM.; the funding sources have no further role in studydesign; in the collection, analysis and interpretation of data; in the writing of thereport; and in the decision to submit the paper for publication.

ContributorsTAPL wrote the first draft of the manuscript performed the molecular genetic

analysis, statistical analysis, and managed the literature searches and analysis. AKTundertook statistical analysis and writing the manuscript. Author ANV contributed tothe literature search and writing of the manuscript. HYM, JAL, and SGP collected andclinically characterized the sample. Authors JLK and DJM designed the study and wrotethe protocol. All authors contributed to and have approved the final manuscript.

Conflict of interestHYM has received grants or is a consultant to Abbott Labs, ACADIA, Bristol Myers

Squibb, Eli Lilly, Janssen, Pfizer, Astra Zeneca, Glaxo Smith Kline, Memory, Cephalon,Minster, Aryx and BiolineRx. HYM is a shareholder of ACADIA. JAL reports that he serveson the Advisory Board of Bioline, GlaxoSmithKline, Intracellular Therapies, Eli Lilly,Pierre Fabre, Psychogenics and Wyeth. He does not receive financial compensation orsalary support for his participation as an advisor. He receives grant support from Allon,Forest Labs, Merck and Pfizer; he holds a patent from Repligen. JKL has one timehonorarium from Eli Lilly Corporation, and has been a consultant to GSK, Sanofi-Aventisand Dainippon-Sumitomo.

AcknowledgmentsNone.

References

Andreasen, N.C., 1982. Negative symptoms in schizophrenia. Definition and reliability.Arch. Gen. Psychiatry 39 (7), 784–788.

Arvanov, V.L., Wang, R.Y., 1999. Clozapine, but not haloperidol, prevents the functionalhyperactivity of N-methyl-D-aspartate receptors in rat cortical neurons induced bysubchronic administration of phencyclidine. J. Pharmacol. Exp. Ther. 289 (2),1000–1006.

Barrett, J.C., Fry, B., Maller, J., Daly, M.J., 2005. Haploview: analysis and visualization ofLD and haplotype maps. Bioinformatics (Oxford, England) 21 (2), 263–265.

Critchlow, H.M., Maycox, P.R., Skepper, J.N., Krylova, O., 2006. Clozapine and haloperidoldifferentially regulate dendritic spine formation and synaptogenesis in rathippocampal neurons. Mol. Cell. Neurosci. 32 (4), 356–365.

de Wit, J., Sylwestrak, E., O'Sullivan, M.L., Otto, S., Tiglio, K., Savas, J.N., Yates, J.R.,Comoletti, D., Taylor, P., Ghosh, A., 2009. LRRTM2 interacts with Neurexin1 andregulates excitatory synapse formation. Neuron 64 (6), 799–806.

Dudbridge, F., 2008. Likelihood-based association analysis for nuclear families andunrelated subjects with missing genotype data. Hum. Hered. 66 (2), 87–98.

Etherton, M.R., Blaiss, C.A., Powell, C.M., Südhof, T.C., 2009. Mouse neurexin-1alphadeletion causes correlated electrophysiological and behavioral changes consistentwith cognitive impairments. Proc. Natl. Acad. Sci. U. S. A. 106 (42), 17998–18003.

Gauderman, W.J., 2002. Sample size requirements for association studies of gene–geneinteraction. Am. J. Epidemiol. 155 (5), 478–484.

Kane, J., Honigfeld, G., Singer, J., Meltzer, H., 1988. Clozapine for the treatment-resistantschizophrenic. A double-blind comparison with chlorpromazine. Arch. Gen.Psychiatry 45 (9), 789–796.

Kehrer, C., Maziashvili, N., Dugladze, T., Gloveli, T., 2008. Altered Excitatory–InhibitoryBalance in the NMDA-hypofunction model of schizophrenia. Front. Mol. Neurosci.1, 6.

Lahiri, D.K., Nurnberger Jr., J.I., 1991. A rapid non-enzymatic method for the preparationof HMW DNA from blood for RFLP studies. Nucleic Acids Res. 19 (19), 5444.

Murphy, B.P., Chung, Y.C., Park, T.W., McGorry, P.D., 2006. Pharmacological treatment ofprimary negative symptoms in schizophrenia: a systematic review. Schizophr. Res.88 (1–3), 5–25.

Ninan, I., Jardemark, K.E., Wang, R.Y., 2003. Olanzapine and clozapine but nothaloperidol reverse subchronic phencyclidine-induced functional hyperactivity ofN-methyl-D-aspartate receptors in pyramidal cells of the rat medial prefrontalcortex. Neuropharmacology 44 (4), 462–472.

Olney, J.W., Newcomer, J.W., Farber, N.B., 1999. NMDA receptor hypofunction model ofschizophrenia. J. Psychiatr. Res. 33 (6), 523–533.

Sanfilipo, M., Lafargue, T., Rusinek, H., Arena, L., Loneragan, C., Lautin, A., Feiner, D.,Rotrosen, J., Wolkin, A., 2000. Volumetric measure of the frontal and temporal loberegions in schizophrenia: relationship to negative symptoms. Arch. Gen. Psychiatry57 (5), 471–480.

Shah, A.K., Tioleco, N.M., Nolan, K., Locker, J., Groh, K., Villa, C., Stopkova, P., Pedrosa, E.,Lachman, H.M., 2010. Rare NRXN1 promoter variants in patients with schizophre-nia. Neurosci. Lett. 475 (2), 80–84.

Souza, R.P., Meltzer, H.Y., Lieberman, J.A., Le Foll, B., Kennedy, J.L., 2010. Influence ofneurexin 1 (NRXN1) polymorphisms in clozapine response. Hum. Psychopharma-col. Clin. Exp. 25 (7–8), 582–585.

Südhof, T.C., 2008. Neuroligins and neurexins link synaptic function to cognitivedisease. Nature 455 (7215), 903–911.

Voineskos, A.N., Lett, T.A., Lerch, J.P., Tiwari, A.K., Ameis, S.H., Rajji, T.K., Muller, D.J.,Mulsant, B.H., Kennedy, J.L., 2011. Neurexin-1 and frontal lobe white matter: anoverlapping intermediate phenotype for schizophrenia and autism spectrumdisorders. PloS One 6 (6), e20982.

Wolkin, A., Choi, S.J., Szilagyi, S., Sanfilipo, M., Rotrosen, J.P., Lim, K.O., 2003. Inferiorfrontal white matter anisotropy and negative symptoms of schizophrenia: adiffusion tensor imaging study. Am. J. Psychiatry 160 (3), 572–574.

124 T.A.P. Lett et al. / Schizophrenia Research 132 (2011) 121–124

REVIEW

Treating Working Memory Deficits in Schizophrenia:A Review of the NeurobiologyTristram A. Lett, Aristotle N. Voineskos, James L. Kennedy, Brian Levine, andZafiris J. Daskalakis

Cognitive deficits are a core feature of schizophrenia. Among these deficits, working memory impairment is considered a centralcognitive impairment in schizophrenia. The prefrontal cortex, a region critical for working memory performance, has been demonstratedas a critical liability region in schizophrenia. As yet, there are no standardized treatment options for working memory deficits inschizophrenia. In this review, we summarize the neuronal basis for working memory impairment in schizophrenia, including dysfunctionin prefrontal signaling pathways (e.g., γ-aminobutyric acid transmission) and neural network synchrony (e.g., gamma/theta oscillations).We discuss therapeutic strategies for working memory dysfunction such as pharmacological agents, cognitive remediation therapy, andrepetitive transcranial magnetic stimulation. Despite the drawbacks of current approaches, the advances in neurobiological andtranslational treatment strategies suggest that clinical application of these methods will occur in the near future.

Key Words: Cognition, EEG, neurophysiology, schizophrenia, TMS,working memory

Schizophrenia is a common and chronic psychiatric disordercharacterized by delusions, hallucinations with concomitantcognitive, organizational, and motivational impairments.

Currently approved pharmaceutical treatments for schizophreniaare typically effective for positive symptoms but have little or noeffect on cognitive impairment (1). This is of particular concern,because cognitive performance is a key determinant of long-termoutcome and mortality in schizophrenia (2). Cognitive dysfunctionin schizophrenia shows high prevalence, is relatively stable overtime, and is independent of psychotic symptoms (3). Moreover,cognitive dysfunction is present in healthy relatives of schizo-phrenia patients, and it has been suggested as a biomarker ofschizophrenia (4). As a consequence, disturbances in criticalcognitive process, such as working memory, are regarded as acore feature of schizophrenia.

Of the demonstrated neurocognitive deficits in schizophrenia,research has focused on working memory, which has beendefined as the ability to transiently hold and manipulate infor-mation to guide goal-directed behavior (5). The contents ofworking memory are constitutively updated, monitored, andmanipulated in response to immediate processing demands (5).Working memory prolongs the impact of experience beyondimmediately accessible information to enable the incorporation ofinformation from long-term memory, lexical labels, and otherevents into goal-oriented behavior (6). The dorsolateral prefrontalcortex (DLPFC) is crucial to working memory function in healthyadults (7). In schizophrenia patients, working memory deficitsare associated with dysfunction of DLPFC as well as DLPFC

connectivity with other regions and disruption of neurotransmit-ter input (e.g., γ-aminobutyric acid [GABA], glutamate, anddopamine) (8–10). Working memory in schizophrenia might alsohave a genetic basis. Schizophrenia patients and their unaffectedco-twins perform significantly worse than control subjects onspatial working memory tasks (11). The letter-number-sequencingtask (a measure of working memory) has been identified asan endophenotype of schizophrenia with a heritability of .39(.25–.52) (12). Thus, improved identification of circuit disruption(from DLPFC to other regions) can help provide insights into thepathophysiology of working memory impairment in schizophre-nia and the development of novel therapeutic interventions.

In this article, we review the neuropsychological and neuro-anatomical basis of working memory and its relationship toschizophrenia. Next, the therapeutic approaches for treatmentof working memory deficits in schizophrenia are discussed,including pharmacological interventions and cognitive remedia-tion therapy (CRT). Finally, we establish the neurophysiologicalbasis for working memory deficits and present repetitive trans-cranial magnetic stimulation (rTMS) as a potential novel ther-apeutic strategy.

Working Memory

A key benefit of studying working memory is that psychologyand cognitive neuroscience have built a comprehensive frame-work for understanding the cognitive architecture of workingmemory and its neural correlates. For instance, studies in non-human primates suggest that lesions to the prefrontal cortex(PFC) cause marked reduction in working memory function andthat subdivisions of the PFC might represent multiple workingmemory domains, each having its own specialized processing orcontent-specific storage (13,14). According to the seminal theo-retical model by Baddeley (5), working memory functions can befractionalized into specialized systems that serve as buffers for thestorage and manipulation of information. The model is comple-mented by empirical evidence that most primate electrophysiol-ogy and neuroimaging studies, regardless of experimentalprocedure, report delay-period activity in the PFC [for examples,see (15,16)].

The PFC is an integral component of executive functioning(e.g., complex attention, planning, and mental flexibility) (17).The PFC contributes to working memory by exerting top-down control through filtering and strategic reorganization of

From the Centre for Addiction and Mental Health (TAL, ANV, JLK, ZJD);Institute of Medical Science (TAL, ANV, JLK, ZJD); Department ofPsychiatry (ANV, JLK, ZJD); Department of Psychology (BL), Universityof Toronto; and the Rotman Research Institute (BL), Baycrest CentreToronto, Toronto, Ontario, Canada.

Address correspondence to Zafiris J. Daskalakis, M.D., Ph.D., TemertyChair in Therapeutic Brain Intervention, Professor of Psychiatry,University of Toronto, Centre for Addiction and Mental Health(CAMH), 1001 Queen Street West, Toronto, Ontario, M6J 1H4 Canada;E-mail: Jeff.Daskalakis@camh.ca.

Received Jan 3, 2013; revised and accepted Jul 22, 2013.

0006-3223/$36.00 BIOL PSYCHIATRY 2014;75:361–370http://dx.doi.org/10.1016/j.biopsych.2013.07.026 & 2014 Society of Biological Psychiatry

information (18). Therefore, working memory performance woulddepend on efficient communication to the PFC and its capacity toinhibit extraneous information. Top-down attention relies onparietal and prefrontal regions that largely overlap with activationduring working memory tasks in both these regions (19). More-over, high global brain connectivity to the DLPFC predicts betterworking memory performance as well as general fluid intelligence(20). The PFC thus acts as a flexible hub by which frontalconnectivity is adjusted according to task demands (21).

More recently, research has emphasized recruitment ofextrafrontal regions involved in perceptual or long-termrepresentations in orchestration with DLPFC (22). Electroence-phalagram (EEG) studies show theta coupling between pre-frontal and parietal cortices is increased with more complexmanipulation (23), memory load (24), and predict individualworking memory capacity (25). Theta phase synchronybetween the prefrontal and temporal cortices occurs duringthe maintenance phase of working memory (26) in addition toencoding and retrieval (27). Further evidence for medialtemporal lobe involvement comes from intracranial EEGrecording in human epilepsy patients that shows cross-frequency coupling of oscillatory activity in the hippocampusbetween beta/gamma range and the theta band, and theprecision of coupling predicts working memory performance (28).This sustained phase synchronization between higher-order sen-sory, frontal, and temporal cortices and the hippocampus providesa mechanism for working memory maintenance by which activityin different brain regions is sustained in the absence of directsensory output (26). An important consequence of these findingsis that working memory depends on network-level activation andcoordination.

Working Memory and Schizophrenia

Schizophrenia patients are cognitively compromised on theorder of magnitude 1.0–1.8 SDs below the normal mean (29).Patients with an earlier onset have more severe cognitive deficitsthat persist throughout the course of the disorder (30). Cognitiveimpairments are present in the prodromal period and mightcontribute to heterogeneity in patterns of cognitive changesacross illness phases and among individuals (31). Meta-analyses inschizophrenia demonstrate large deficits in all 3 domains ofworking memory (phonological, visuospatial, and central execu-tive) with no clear differences across domains or tasks (32,33).There was also no consistent association between duration ofillness, antipsychotic medication, or symptom profile and workingmemory in schizophrenia (33).

The DLPFC has been identified as a key liability region forworking memory dysfunction in schizophrenia (34). In an earlystudy, healthy individuals demonstrated increased blood flow tothe DLPFC during the Wisconsin Card Sorting Task that was notobserved in medication-free schizophrenia patients (35). How-ever, recent neuroimaging studies generated conflicting findingswith regard to DLPFC activation during working memory tasks.Both “task-related hypofrontality” and “task-related hyperfrontal-ity” have been reported in patients with schizophrenia relative tohealthy subjects (34). These discrepancies are potentially drivenby study differences in task performance or difficulty, although itis possible that the findings are confounded by coupling andactivation in other cortical regions. For example, stronger activa-tion of deep brain structures [e.g., thalamus (36)] and the anteriorcingulate cortex (37) in schizophrenia patients might be a product

of compensatory mechanism for working memory deficits. There-fore, working memory dysfunction could be a result of reducedfunction of specific regions but also an impairment to engagefunctional networks synchronized to a given cognitive task.

The disruption of working memory networks in schizophrenia isstill poorly understood. As reviewed in the preceding text, dynamicnetwork connectivity is necessary for proper working memoryfunctioning. Given the functional and anatomical “dysconnectivity”observed in schizophrenia (38), especially to the DLPFC (39),working memory deficits in schizophrenia could be due todysfunction of establishing or changing brain networks. Thus,establishing a link between functional integration and workingmemory deficits is crucial to developing novel, neurobiological-based interventions to enhance working memory performance.

Current Treatments of Working Memory Deficits

Therapeutic strategies for working memory deficits in schizo-phrenia are of great interest, considering their predictive value forfunctional outcome. Nonpharmacological and pharmacologicaltreatment strategies have been investigated but demonstratemixed results.

Antipsychotic TreatmentPharmacological studies have examined differences in effects

of antipsychotic medications on cognitive functioning. Althoughshowing small effects toward improved cognitive performancewith treatment, some studies show therapeutic advantages ofatypical antipsychotics compared with typical antipsychotics (40);however, the large, multisite CATIE trial (Clinical AntipsychoticTrials of Intervention Effectiveness) failed to find any advantage ofatypical antipsychotics in treating cognition (1). Clozapine, theatypical antipsychotic agent for treatment resistant-schizophrenia(41), is no longer considered superior to other atypical antipsy-chotic agents for cognitive deficits (42). These results were drivenby multiple pharmacological initiatives, such as the MATRICS(Measurement and Treatment Research to Improve Cognition inSchizophrenia) (43), TURNS (Treatment Units for Research onNeurocognition and Schizophrenia) (44), and CNTRICS (CognitiveNeuroscience Treatment Research to Improve Cognition Schizo-phrenia) (45). These initiatives highlight continuing interest andcommitted resources currently dedicated for novel therapies forcognitive deficits in schizophrenia and, in particular, workingmemory deficits. It should be noted that the long-term con-sequences of antipsychotic treatment might be detrimental tocognition. Progressive declines in working memory performanceare observed in nonhuman primates undergoing chronic treat-ment of haloperidol over a 6-month period (46). Additionally, graymatter loss, higher neuronal density, and reduced glial cellnumber similar to that histologically observed in schizophreniawas reported in nonhuman primates exposed to olanzapine orhaloperidol over a 2-year period (47,48). A longitudinal first-episode schizophrenia study showed progressive decline of whiteand gray matter volume correlating with antipsychotic medica-tion dose (49). Thus, the evidence does not support a benefitfrom antipsychotic medication with regard to cognitive deficitsbut rather indicates a potential negative effect on workingmemory in schizophrenia during long-term treatment.

Pharmacological TargetsThe pharmacology of working memory dysfunction might

provide critical understanding for the development of new

362 BIOL PSYCHIATRY 2014;75:361–370 T.A. Lett et al.

www.sobp.org/journal

treatments (Figure 1) (50). Blockade of the glutamate-mediatedexcitatory neurotransmission by N-methyl-D-aspartate receptor(NMDAR) antagonists mimics positive and negative symptomsas well as cognitive deficits in schizophrenia. These findingssuggest that enhancing NMDAR neurotransmission mightreverse cognitive deficits (51,52). Furthermore, NMDAR ablationon GABA interneurons impairs hippocampal theta rhythmleading to impaired working memory (53). The NMDAR activa-tion also subserves persistent DLPFC neuronal firing duringworking memory (54), suggesting that glutamate function andconnectivity is integral to working memory performance.Results from the CONSIST (Cognitive and Negative Symptomsin Schizophrenia Trial), however, suggest that either glycine(binds to allosteric site of the NMDAR) or D-cycloserine (partialNMDA agonist) were not effective in treating cognitive impair-ments (55). Vis-à-vis dopamine, early preclinical work shows thatdopamine neurotransmission might be augmented to treat work-ing memory deficits in schizophrenia. Increased availability of PFCdopamine D1 receptors has been reported in schizophrenia andmight reflect a compensatory upregulation, due to reduced PFCdopamine release; furthermore, the increased expression hasbeen directly associated with poor working memory performance(56,57). In nonhuman primates, intermittent long-term D1 recep-tor agonist treatment yielded persistent improvements inhaloperidol-induced working memory deficits (46). The selectiveD1 receptor agonist, dihydrexidine, was reported to be well-tolerated in schizophrenia subjects (58). Single-dose administra-tion of dihydrexidine was reported to have no effect on neuro-cognition (59). Nevertheless, intermittent D1 receptor agonisttreatment remains a promising strategy. Catecholamine-O-meth-yltransferase has been directly associated with PFC dopamineturnover and working memory performance (60). Catecholamine-O-methyltransferase inhibitors, such as tolcapone, are a promisingtarget, although they have unfortunately also been associatedwith hepatotoxicity (61). Finally, GABAergic inhibitory neurotrans-mission in the DLPFC is altered in schizophrenia (62) and isintegral to organizing gamma oscillations associated with work-ing memory load (63). The major determinant of GABA in theneocortex, glutamic acid decarboxylase, is consistently down-regulated in postmortem studies of patients with schizophrenia(64). The selective agonist of the GABAA receptor, MK-0777, wasshown to be effective in treating working memory deficits andcould potentially modulate frontal gamma activity in a study withlimited sample size (65). A subsequent study failed to replicatethe enhancement of working memory by MK-0777 in schizophre-nia (66); however, modulating GABA neurotransmission remains apromising target.

Other pharmacological strategies including galantamine, acombined acetylcholinesterase inhibitor and allosteric potentiatorof the nicotinic receptor, show modest effect across severalcognitive domains (67) with no improvement in working memoryas confirmed by a recent Cochrane review (68). Furthermore,other pharmacological treatment attempts have failed, including:the novel neuropeptide davunetide; the nicotinic agonist vareni-cline; and pregnenolone (69–71). Although some improvement inworking memory performance was shown with pergolide, mino-cycline, amphetamine, and recombinant human erythropoietin,none of these findings have been replicated in a controlled study(72–75).

In summary, pharmacological investigations for working mem-ory deficits in schizophrenia could benefit from the use of novelagents, because existing studies have demonstrated limitedtreatment effects (61).

Cognitive TrainingPerhaps the best-supported strategy targeting working mem-

ory deficits (and cognitive dysfunction in general) in schizophre-nia is CRT. Cognitive remediation therapy employs drill or practiceexercises, teaching strategies to improve cognitive functioning, aswell as compensatory strategies and group discussions (76).A number of studies have investigated effects of CRT on differentcognitive domains (e.g., attention/vigilance, processing speed,verbal working memory, or social cognition) (77). Computerizedand noncomputerized training methods for these differentdomains of cognitive function have been described.

Although the neural mechanisms of action remain poorly under-stood, CRT might influence cortical connectivity and brain structurerelevant to the specific training involved. Wykes et al. (78) found that

Figure 1. Hypothesized prefrontal cortical circuit highlighting synapsesthat are implicated in working memory dysfunction and targets ofpharmacological cognitive enhancers. Direct and indirect disruptions ofdopamine, glutamate, and γ-aminobutyric acid (GABA) neurotransmittersignaling are reported in schizophrenia, and these synapses are integral toworking memory function. In the prefrontal cortex, chandelier cells(parvalbumin-containing, fast-spiking GABA interneurons) mediate GABAneurotransmission at the axon initial segment of pyramidal cells (excita-tory neurons) to GABAA receptors, including the α2 subunit. Pyramidalcells release glutamate to N-methyl-D-aspartate receptors (NMDARs) onchandelier cell forming a feedback mechanism responsible for gammaoscillation activity in the prefrontal cortex. Pyramidal cells synapse onbasket cells (parvalbumin-containing, fast-spiking GABA interneurons)with reciprocal GABAergic synapses of basket cells on the soma ofpyramidal interneurons. Pyramidal cells also release glutamate on striataldopamine neurons (and other regions, such as the hippocampus andventral tegmental area) leading to activation of dopamine D1 receptors(D1R) on prefrontal chandelier and pyramidal cells, thereby augmentingthe activation timing of these neurons. Examples of pharmacologicaltargets to improve working performance include: 1) restoring glutamatesignaling with glycine (binding to the allosteric site of the NMDAR) orD-cycloserine (partial NMDA agonist); 2) selectively increasing dopaminesignaling to the D1R with dihydrexidine (D1R agonist); and3) increasing GABAergic tone through agonism of the GABAA α2 subunitby MK-0777. For further review, please see Lewis and Gonzalez-Burgos(50) and Lisman et al. (149). GABA(A)α1, GABAA receptor including the α1subunit; GABA(A)α2, GABAA receptor including the α2 subunit.

T.A. Lett et al. BIOL PSYCHIATRY 2014;75:361–370 363

www.sobp.org/journal

schizophrenia patients undergoing CRT for executive functioning (n¼ 6) over 3 months showed increased brain activation in fronto-cortical regions associated with working memory compared withcontrol therapy patients. Increased activation of the inferior frontalcortex after 10 weeks of verbal memory training in eight patientswith schizophrenia was associated with verbal working memoryimprovement (79). A 2-year randomized control trial of cognitiveenhancement therapy (combined neurocognitive and social cogni-tive remediation) in 121 patients reported strong, lasting effect oncognition and global functioning (Cohen’s d � 1.00), although therewas no association with working memory (80). Subsequently, it wasreported that the cognitive enhancement therapy group hadpreservation of left hippocampal, parahippocampal, and fusiformgyrus gray matter volume and increased amygdala volume (81).Furthermore, less gray matter loss in the parahippocampus andfusiform gyrus as well as greater amygdala gray matter volume wasrelated to improved cognition (81). In poor-reading children, 100hours of remedial reading training normalized left frontal fractionalanisotropy to that of normal reading children (82). More recently,strategy-learning-based CRT in 30 schizophrenia subjects normalizedactivation toward the pattern of healthy control subjects during anN-back working memory functional magnetic resonance imagingparadigm (83). Moreover, after CRT these subjects had increasedfractional anisotropy in the genu of the corpus callosum that wascorrelated with total cognition and executive function, although CRTwas not associated with working memory improvement (83). Takentogether, these results suggest that CRT initiates learning-inducedplasticity in cognitively compromised populations.

The most recent meta-analyses indicate that CRT can provide amoderate improvement in global cognition (effect size is approx-imately .4–.5) (77,84). Despite the difference between CRTapproaches in terms of methods used and targeted cognitivedomains, studies have shown consistent effect sizes (85); more-over, no single method (e.g., remediation approach, CRT duration)was superior in terms of cognitive outcome (77,84). Althoughsome CRT studies have shown no effect on working memory(77,86), computer-based programs that focused on the remedia-tion of verbal working memory in schizophrenia through auditorytraining exercises have shown promise (87–89). For instance,Fisher et al. (88) demonstrated significant improvements on theletter–number span working memory task in patients withschizophrenia after 50 hours of auditory training exercises,compared with a control group. This training also improvedauditory psychophysical performance that was related toimproved verbal working memory and global cognition. More-over, the active group had significantly elevated peripheral levelsof the brain-derived neurotrophic factor (BDNF), indicating CRTmight induce neuroplasticity. Six-month follow-up revealed aclear improvement of global cognition with lasting effects onauditory function, visual processing, and cognitive control (89).Although a recent study failed to replicate these results (90), CRTmight have enduring therapeutic value. Particularly when CRT isprovided with adjunctive psychiatric rehabilitation, such as socialgroup exercises (80), it is shown to be more effective (77).

The observed effects on neuroplasticity together with moder-ate effect sizes of the training suggest that CRT might best servein combination with neurophysiological-based interventionsinducing neuroplasticity, such as rTMS or with pharmacologicalapproaches. For example, CRT might act in concert with othermethods of inducing neuroplasticity to reinforce working memorypathways. However, to the best of our knowledge, there are nopublished studies examining CRT in combination with othercognitive neuroenhancement techniques.

TMS and Cognition

Transcranial magnetic stimulation is an investigational tool toexamine physiological brain processes in relation to cognitionand psychiatric illness (91). For example, rTMS intervention to theDLPFC transiently impairs encoding and retrieval mechanismswith visuospatial (92) and verbal stimuli (93); however, the samestimulation can facilitate cognition in picture naming, objectnaming, speed during reasoning puzzle, and cognitive reactiontasks (94–97). In this regard, rTMS modulates cortical excitabilitythrough local inhibitory circuits that could facilitate or inhibitbrain networks relevant to the cognitive task (98). Combining TMSwith EEG (TMS-EEG) allows measurement of both temporal andspatial activations at the targeted brain region (99). A linkbetween regional neurophysiology and cognitive function canbe examined, by assessing how TMS-induced modulation ofcortical activation of the DLPFC relates to working memoryfunction. For instance, long interval cortical inhibition (a measureclosely associated with GABAB receptor neurotransmission) hasbeen strongly correlated with performance on the N-back(r ¼ .63, p ¼ .04) and letter–number-sequencing (r ¼ .68, p ¼ .005)working memory tasks in healthy control subjects (100–102).Furthermore, this suggests that long interval cortical inhibitionmight be important in modulating high-frequency oscillations inthe DLPFC that influences working memory (102). Therefore, TMS-EEG measures of cortical inhibition and DLPFC synchrony mightprovide key insights into working memory function. The con-nection between targeted area and brain functioning is of greatimportance, because psychiatric disorders, such as schizophrenia,have abnormal neural oscillations and synchrony. This has beendemonstrated in rhythm-generating networks of GABA interneur-ons and in cortico–cortical connections (103).

Brain Networks Synchrony

Theories of schizophrenia emphasize deficits in the coordina-tion of distributed neuronal oscillatory activity that lead toworking memory dysfunction (104). Patients with schizophreniahave abnormal gamma oscillations and gamma–theta couplingthat might underlie independent cognitive and functional impair-ments (104–106). Gamma frequency oscillations are particularlyinteresting, because of their integral relationship with higherbrain processes (107). They provide a temporal structure forinformation processing in the brain, mediating storage and recallof information (108). One GABA interneuron typically connectsextensively with several pyramidal neurons forming neuronalnetworks that fire contemporaneously, a process that can beenrecorded over the surface of the cortex as gamma oscillatoryactivity (Figure 1) (109). The kinetics of inhibitory interneurons inthe cortex are such that their firing rate is much higher than thatof pyramidal cells, permitting higher rates of pyramidal cell firingcompared with baseline (110). Finally, inhibitory interneuronsform synaptic connections with pyramidal neurons at the cellbody, a synaptic relationship that allows greater control ofpyramidal neuron firing compared with synaptic terminations atmore distal regions of the neuron (111). As a result of this patternof connectivity, inhibitory interneurons exert fine control over thefiring of pyramidal neuron networks, which translates into high-frequency gamma oscillatory activity on EEG (112). Gammaoscillations are also coupled to theta rhythms (110), and thiscoupling has been found to be essential for working memory (27).This suggests that specific interplay between large ensembles ofneurons has clinical significance.

364 BIOL PSYCHIATRY 2014;75:361–370 T.A. Lett et al.

www.sobp.org/journal

In patients with schizophrenia, aberrant gamma oscillatoryactivity has been reported during a cognitive control task,compared with healthy subjects (107). Furthermore, inability tosupport stimulus-driven gamma oscillations in schizophreniapatients has been associated with working memory dysfunction(104). Excessive frontal activation of gamma oscillations werereported in schizophrenia and correlated with working memorytask difficulty (113). Finally, the increase in gamma oscillationswas associated with a later maintenance phase of workingmemory and induced gamma and theta activity during retrieval(114). Taken together, these results suggest that schizophreniapatients are not properly able to coordinate cortical activity that isappropriate cognitive demand.

Modulating Network Plasticity

There are several mechanisms for TMS to induce and measurecortical plasticity. The TMS activation of a population of neurons inthe same synaptic pathway (homosynaptic) or in different path-ways (heterosynaptic) is modulating synaptic efficacy either by:increased synaptic strength (long-term potentiation [LTP]); ordecreased synaptic strength (long-term depression) (115). Throughthese mechanisms, low-frequency rTMS will cause a decrease inbrain excitability (116); in contrast, high-frequency rTMS causesincreased brain excitability (117). Similarly, other magnetic brainstimulation protocols can produce changes in excitation orinhibition (Table S1 in Supplement 1). Furthermore, studies ofthe motor cortex have shown that TMS protocols potentiate lastingeffects on this excitability for 30 min to several hours (118). Arecent study suggests that one paradigm, known as paired-associated stimulation (PAS), might enhance motor learning at 1week post-PAS (119). The PAS-25, peripheral nerve stimulation 25msec before rTMS, induced LTP, leading to enduring enhancementof evoked motor potential. These lasting effects are particularlyexciting, because rTMS can modulate brain network oscillatoryactivity, thus providing evidence that PAS could trigger structuraland functional changes necessary for long-term improvement ofmotor performance. Similar findings have been reported in animalstudies. A recent study by Benali et al. (120) found that theintermittent theta-burst stimulation (a type of rTMS) (121) to the ratneocortex differently modulates gamma oscillations and proteinexpression. Intermittent theta-burst stimulation (excitatory)enhanced neural firing and EEG gamma power by reducingparvalbumin expression in fast spiking GABA interneurons; incontrast, continuous theta-burst stimulation (inhibitory) ratheraffected pyramidal neurons calbindin D-28k expression.

Taken together, the lasting cortical plasticity induced by TMS ispromising, especially because it can alleviate difficult-to-treatfacets of complex psychiatric disorders, such as cognitive deficitsin schizophrenia.

Remodeling of Connectivity in Schizophrenia by rTMS

There is overwhelming evidence that schizophrenia is, at least inpart, a disorder of dysconnectivity of the brain (122). This abnormalfunctional integration of processes might be due to aberrant wiringduring development or aberrant synaptic plasticity or both. Thisincludes abnormal functional connectivity [e.g., frontotemporalconnectivity, gamma synchrony (123,124)], abnormal structuralconnectivity [e.g., white matter integrity, reduced brain asymmetry(125–127)], and synaptic plasticity [e.g., pharmacological-inducedschizophrenia symptomology, reduced dendritic field size and

density (128,129)]. Genetic factors common to all of these pointsof dysfunction [e.g., disrupted-in-schizophrenia 1 (DISC1), glutamatedecarboxylase 1 (GAD1), neuregulin 1 (NRG1), microRNA 137(MIR137), and zinc finger protein 804A (ZNF804A) genes (130,131)]all point to the possibility that schizophrenia patients are predis-posed to dysconnectivity. Moreover, neural dysconnectivity mightbe a causative factor in the more intractable deficits of schizo-phrenia, such as working memory functioning (132). Importantly,rTMS as an external intervention might be used to activate neuraldevelopmental pathways sidestepping the normal modes ofsynaptic plasticity. In this regard, rTMS might galvanize plasticityin brain networks that are compromised in schizophrenia. Forinstance, intermittent theta-burst stimulation could remediateabnormalities of gamma oscillations of cognitive processing inschizophrenia patients (105). Indeed, high-frequency rTMS to theDLFPC results in reduced frontal gamma oscillation in schizophreniapatients during the N-back working memory task (Figure 2A) (133).

Activity-Dependent Regulation of Molecular Factorsby rTMS

In most cases, molecular factors that regulate plasticity relateneuronal activation to expression of activity-dependent genes (134).Knowledge of the molecular factors involved in rTMS induction ofneural plasticity is necessary to understand how rTMS might be usedto shape lasting effects on neural circuitry. Activity-dependent geneexpression is integral in the refinement of neuronal network indevelopment as well as in the adaptive, long-lasting modificationsnecessary for mature brain function, such as learning and memory. Ithas been well-established that the cyclic adenosine monophos-phate-response element binding-protein (CREB) plays a central role,at least in part, in mediating activity-dependent neuroplasticity(Figure 2B) (135). Ji et al. (136) reported that rTMS stimulation tothe rat brain activated CREB, leading to increased expression inparaventricular nucleus of the thalamus, cingulate cortex, and frontalcortex. Furthermore, CREB functionally regulates the BDNF gene(137), and theta-burst induction of LTP causes upregulation of BDNF(138). Most recently, it was shown that rTMS treatment to humanneuroblastoma cell lines (SH-SY5Y) resulted in activation of CREB(139). Interestingly, CREB regulates cellular fate by inducing expres-sion of the small, noncoding microRNA, miR-132, that controls themessenger RNA stability or translation of many genes involved inepigenetic regulation and neuronal morphogenesis including:dihydropyrimidinase-like 3 (DPYSL3), Rho GTPase activating protein32 (ARHGAP32), GATA binding protein 2 (GATA2), DNA(cytosine-5-)-methyltransferase-3-alpha (DNMT3A), and methyl CpG bindingprotein 2 (MECP2) (140–142).

Taken together, CREB and downstream factors might play acritical role in rTMS-induced plasticity. It should be noted that manyof the molecular factors potentially modulated by rTMS(e.g., MECP2, BDNF, and miR-132) are also schizophrenia risk factorsrelated to neuroplasticity. This relationship suggests that rTMS couldrescue normal function of neuroplasticity networks in schizophrenia;however, further research is imperative to establish causal relation-ships between rTMS gene networks involved in plasticity.

Treatment of Working Memory Deficits inSchizophrenia with rTMS

To date, there are only two published clinical trials examiningthe efficacy of rTMS for treatment of working memory dysfunctionin schizophrenia. In a 4-week sham-controlled rTMS trial, patients

T.A. Lett et al. BIOL PSYCHIATRY 2014;75:361–370 365

www.sobp.org/journal

(n = 13 active; n = 12 sham) that received 20-Hz stimulation to theDLPFC (Brodmann area 46/6) had improved performance on the 3-back condition of the N-back working memory task (143). More-over, working memory performance in the active treatmentschizophrenia group normalized to that of healthy control subjects(143). In contrast, a 3-week sham-controlled 10-Hz rTMS trial to theleft posterior medial frontal gyrus in schizophrenia patients (n ¼13 active; n ¼ 12 sham) and control subjects (n ¼ 11 active; n ¼11 sham) reported no significant effect of treatment or treatment� diagnostic interaction in the 2-back condition (144). Thedisparity of results between these studies could be due to anumber of factors. First, rTMS-induced differences in gammaoscillations are reported to be more pronounced in the 3-backworking memory task (133); thus, treatment might be specific to

high working memory load. Second, rTMS treatment of workingmemory might be more efficacious when targeted to Brodmannarea 46/9. Last, 4 weeks of 20-Hz rTMS stimulation (in contrast to 3weeks of 10 Hz) might be a more effective mode of treatment.Anodal direct current stimulation (tDCS) treatment has beenpreviously associated with improvement in global cognitivefunction, attention, and enhancement of working memory [forreview, see (145)]. A single study has examined anodal tDCS fortreatment of working memory dysfunction in schizophreniapatients (n ¼ 12) and reported improvements in reaction timebut not accuracy (146). Transcranial alternating current (tACS) caninduce or disrupt theta phase-coupling and therefore might play arole in working memory function. In healthy subjects, tACSartificially induced frontoparietal phase coupling leading to

Figure 2. Neuroplasticity induction by repetitive transcranial magnetic stimulation (rTMS). (A) Topographical illustration of change in gamma power(30–50 Hz) during the 3-back working memory task (left, sham rTMS treatment to DLPFC; right, active treatment to DLPFC). Modified, with permission,from Barr et al. (150). (B) Potential molecular mechanism through which rTMS might induce plasticity. Synaptic activation of L-Type voltage-gated calciumchannel (VGCC) by rTMS leads to increased intracellular calcium initiating a signaling cascade causing activation of the transcription factor cAMP-responseelement binding protein (CREB) by phosphorylation, a hallmark of long-term potentiation (LTP)/long-term depression (LTD). Examples of downstreamchanges in gene expression and the effect these genes have on mediating neuronal plasticity are listed. Red, green, and yellow stars correspond to factorsshown to be associated with schizophrenia, modulated by rTMS, and involved in mediating LTP or LTD, respectively. BA, Brodmann area; BDNF, brain-derived neurotrophic factor; CaM, Calmodulin; CaMKII, CaM kinases II; DNMT3A, DNA (cytosine-5-)-methyltransferase 3 alpha; GABAAR, γ-aminobutyric acid(GABA) A receptor; GATA2, GATA binding protein 2; MECP2, methyl CpG binding protein 2 (Rett syndrome); miR-132, microRNA 132; NMDAR, N-methyl-D-aspartic acid receptor; p250GAP, ARHGAP32 Rho GTPase activating protein 32; TrkB, TrkB receptor.

366 BIOL PSYCHIATRY 2014;75:361–370 T.A. Lett et al.

www.sobp.org/journal

improved working memory, whereas desynchonization impairedworking memory (147). These early results suggest tDCS and tACStarget DLPFC functioning and connectivity, and thus, these treat-ments warrant further investigation. In major depressive disorder,mixed results are observed, although depressive patients are notas cognitively impaired as schizophrenia patients, and workingmemory has not been the primary outcome measure in any study(148). Therefore, results in major depressive disorder patientsmight not generalize to schizophrenia.

Conclusions

Several adjunctive pharmacological agents have shown prom-ising results, yet no agent has demonstrated efficacy in largeclinical trials. Cognitive remediation therapy fairly consistentlyshows improved cognition in schizophrenia, although the effectsseem to depend on domain. For example, CRT has large effect onsocial cognition (effect size is approximately .65), whereas meta-analyses reveal more moderate effects on working memory(effect size is approximately .35). Thus, CRT might be mosteffective in conjunction with other working memory treatments,such as rTMS, to produce large and durable effects whereby theDLPFC and related circuitry would be activated by rTMS andengaged by CRT concomitantly. Furthermore, cognitive enhance-ment drugs could enhance the efficacy of rTMS treatment ofworking memory. It could be speculated that GABAA receptoragonists, such as MK-0777, that affect gamma oscillatory tonecould act in concert with rTMS activation of the DLPFC tospecifically target local GABA signaling and coupling to otherbrain regions.

There are convergent lines of evidence suggesting that rTMSto the DLPFC might be efficacious treatment for working memorydeficits at multiple levels, including: synaptic (e.g., GABA signal-ing); cellular (e.g., GABA interneurons); neurophysiological (e.g.,inhibition); neural network (e.g., gamma oscillations); and func-tional neuroanatomy (e.g., DLPFC). Therefore, rTMS treatment forworking memory deficits in schizophrenia should garner moreresearch, both as an investigative to tool to understand howdysfunction might occur and as a powerful mechanism to induceneuroplasticity.

This work was supported by the Canadian Institutes of HealthResearch Clinician Scientist Award (ZJD, ANV); National Alliance forResearch on Schizophrenia and Depression (ANV), Ontario MentalHealth Foundation (ZJD, ANV) and the Centre for Addiction andMental Health, the Brain and Behaviour Research Foundation, andthe Centre for Addiction and Mental Health Foundation and theCampbell Institute, thanks to the Temerty Family, Grant Family,Kimel Family, Koerner New Scientist Award, and Paul E. GarfinkelNew Investigator Catalyst Award.

ZJD received external funding through Neuronetics and Brains-way and Aspect Medical and a travel allowance through Pfizer andMerck. ZJD has also received speaker funding through Sepracor andAstraZeneca and served on the advisory board for Hoffmann-LaRoche Limited. JLK has received honoraria from Eli Lilly, Roche, andNovartis. TAL, ANV, and BL report no biomedical financial interestsor potential conflicts of interest.

Supplementary material cited in this article is available online athttp://dx.doi.org/10.1016/j.biopsych.2013.07.026.

1. Keefe RS, Bilder RM, Davis SM, Harvey PD, Palmer BW, Gold JM, et al.(2007): Neurocognitive effects of antipsychotic medications in

patients with chronic schizophrenia in the CATIE Trial. Arch GenPsychiatry 64:633–647.

2. Green MF (1996): What are the functional consequences of neuro-cognitive deficits in schizophrenia? Am J Psychiatry 153:321–330.

3. Gold JM (2004): Cognitive deficits as treatment targets in schizo-phrenia. Schizophr Res 72:21–28.

4. Barrantes-Vidal N, Aguilera M, Campanera S, Fatjo-Vilas M, Guitart M,Miret S, et al. (2007): Working memory in siblings of schizophreniapatients. Schizophr Res 95:70–75.

5. Baddeley A (1992): Working memory. Science 255:556–559.6. Mesulam MM (1998): From sensation to cognition. Brain 121:

1013–1052.7. D’Esposito M, Postle BR, Rypma B (2000): Prefrontal cortical con-

tributions to working memory: Evidence from event-related fMRIstudies. Exp Brain Res 133:3–11.

8. Callicott JH, Mattay VS, Verchinski BA, Marenco S, Egan MF, WeinbergerDR (2003): Complexity of prefrontal cortical dysfunction in schizophre-nia: More than up or down. Am J Psychiatry 160:2209–2215.

9. Meyer-Lindenberg A, Poline JB, Kohn PD, Holt JL, Egan MF,Weinberger DR, et al. (2001): Evidence for abnormal cortical func-tional connectivity during working memory in schizophrenia. Am JPsychiatry 158:1809–1817.

10. Lewis DA, Moghaddam B (2006): Cognitive dysfunction in schizo-phrenia: Convergence of gamma-aminobutyric acid and glutamatealterations. Arch Neurol 63:1372–1376.

11. Pirkola T, Tuulio-Henriksson A, Glahn D, Kieseppa T, Haukka J, KaprioJ, et al. (2005): Spatial working memory function in twins withschizophrenia and bipolar disorder. Biol Psychiatry 58:930–936.

12. Greenwood TA, Braff DL, Light GA, Cadenhead KS, Calkins ME, DobieDJ, et al. (2007): Initial heritability analyses of endophenotypicmeasures for schizophrenia: The consortium on the genetics ofschizophrenia. Arch Gen Psychiatry 64:1242–1250.

13. Petrides M (1996): Specialized systems for the processing ofmnemonic information within the primate frontal cortex. PhilosTrans R Soc Lond B Biol Sci 351:1455–1461; discussion 1461–1452.

14. Goldman-Rakic PS (1996): Regional and cellular fractionation ofworking memory. Proc Natl Acad Sci U S A 93:13473–13480.

15. Funahashi S, Chafee MV, Goldman-Rakic PS (1993): Prefrontal neuro-nal activity in rhesus monkeys performing a delayed anti-saccadetask. Nature 365:753–756.

16. Wilson FA, Scalaidhe SP, Goldman-Rakic PS (1993): Dissociation ofobject and spatial processing domains in primate prefrontal cortex.Science 260:1955–1958.

17. Fuster JM (2008): The Prefrontal Cortex. Boston: Academic Press.18. Postle BR (2006): Working memory as an emergent property of the

mind and brain. Neuroscience 139:23–38.19. Corbetta M, Shulman GL (2002): Control of goal-directed and

stimulus-driven attention in the brain. Nat Rev Neurosci 3:201–215.20. Cole MW, Yarkoni T, Repovs G, Anticevic A, Braver TS (2012): Global

connectivity of prefrontal cortex predicts cognitive control andintelligence. J Neurosci 32:8988–8999.

21. Miller EK, Cohen JD (2001): An integrative theory of prefrontal cortexfunction. Annu Rev Neurosci 24:167–202.

22. D’Esposito M (2007): From cognitive to neural models of workingmemory. Philos Trans R Soc Lond B Biol Sci 362:761–772.

23. Sauseng P, Klimesch W, Schabus M, Doppelmayr M (2005): Fronto-parietal EEG coherence in theta and upper alpha reflect centralexecutive functions of working memory. Int J Psychophysiol 57:97–103.

24. Payne L, Kounios J (2009): Coherent oscillatory networks supportingshort-term memory retention. Brain Res 1247:126–132.

25. Kopp F, Schroger E, Lipka S (2006): Synchronized brain activityduring rehearsal and short-term memory disruption by irrelevantspeech is affected by recall mode. Int J Psychophysiol 61:188–203.

26. Sarnthein J, Petsche H, Rappelsberger P, Shaw GL, von Stein A(1998): Synchronization between prefrontal and posterior associa-tion cortex during human working memory. Proc Natl Acad Sci U S A95:7092–7096.

27. Sauseng P, Klimesch W, Doppelmayr M, Hanslmayr S, Schabus M,Gruber WR (2004): Theta coupling in the human electroencephalo-gram during a working memory task. Neurosci Lett 354:123–126.

28. Axmacher N, Henseler MM, Jensen O, Weinreich I, Elger CE, Fell J(2010): Cross-frequency coupling supports multi-item working memoryin the human hippocampus. Proc Natl Acad Sci U S A 107:3228–3233.

T.A. Lett et al. BIOL PSYCHIATRY 2014;75:361–370 367

www.sobp.org/journal

29. Heaton RK, Gladsjo JA, Palmer BW, Kuck J, Marcotte TD, Jeste DV(2001): Stability and course of neuropsychological deficits in schiz-ophrenia. Arch Gen Psychiatry 58:24–32.

30. Rajji TK, Ismail Z, Mulsant BH (2009): Age at onset and cognition inschizophrenia: meta-analysis. Br J Psychiatry 195:286–293.

31. Mesholam-Gately RI, Giuliano AJ, Goff KP, Faraone SV, Seidman LJ(2009): Neurocognition in first-episode schizophrenia: A meta-ana-lytic review. Neuropsychology 23:315.

32. Heinrichs RW, Zakzanis KK (1998): Neurocognitive deficit in schizophre-nia: A quantitative review of the evidence. Neuropsychology 12:426–445.

33. Forbes NF, Carrick LA, McIntosh AM, Lawrie SM (2009): Workingmemory in schizophrenia: A meta-analysis. Psychol Med 39:889–905.

34. Manoach DS (2003): Prefrontal cortex dysfunction during workingmemory performance in schizophrenia: Reconciling discrepant find-ings. Schizophr Res 60:285–298.

35. Weinberger DR, Berman KF, Zec RF (1986): Physiologic dysfunctionof dorsolateral prefrontal cortex in schizophrenia. I. Regional cerebralblood flow evidence. Arch Gen Psychiatry 43:114–124.

36. Minzenberg MJ, Laird AR, Thelen S, Carter CS, Glahn DC (2009): Meta-analysis of 41 functional neuroimaging studies of executive functionin schizophrenia. Arch Gen Psychiatry 66:811–822.

37. Bor J, Brunelin J, Sappey-Marinier D, Ibarrola D, d’Amato T, Suaud-Chagny MF, et al. (2011): Thalamus abnormalities during workingmemory in schizophrenia. An fMRI study. Schizophr Res 125:49–53.

38. Stephan KE, Baldeweg T, Friston KJ (2006): Synaptic plasticity anddysconnection in schizophrenia. Biol Psychiatry 59:929–939.

39. Zhou Y, Liang M, Jiang T, Tian L, Liu Y, Liu Z, et al. (2007): Functionaldysconnectivity of the dorsolateral prefrontal cortex in first-episodeschizophrenia using resting-state fMRI. Neurosci Lett 417:297–302.

40. Woodward ND, Purdon SE, Meltzer HY, Zald DH (2005): A meta-analysis of neuropsychological change to clozapine, olanzapine,quetiapine, and risperidone in schizophrenia. Int J Neuropsychophar-macol 8:457–472.

41. Kane J, Honigfeld G, Singer J, Meltzer H (1988): Clozapine for thetreatment-resistant schizophrenic. A double-blind comparison withchlorpromazine. Arch Gen Psychiatry 45:789–796.

42. Harvey PD, Sacchetti E, Galluzzo A, Romeo F, Gorini B, Bilder RM,et al. (2008): A randomized double-blind comparison of ziprasidonevs. clozapine for cognition in patients with schizophrenia selectedfor resistance or intolerance to previous treatment. Schizophr Res105:138–143.

43. Green MF, Nuechterlein KH, Gold JM, Barch DM, Cohen J, Essock S,et al. (2004): Approaching a consensus cognitive battery for clinicaltrials in schizophrenia: The NIMH-MATRICS conference to selectcognitive domains and test criteria. Biol Psychiatry 56:301–307.

44. Buchanan RW, Freedman R, Javitt DC, Abi-Dargham A, Lieberman JA(2007): Recent advances in the development of novel pharmaco-logical agents for the treatment of cognitive impairments inschizophrenia. Schizophr Bull 33:1120–1130.

45. Barch DM, Smith E (2008): The cognitive neuroscience of workingmemory: Relevance to CNTRICS and schizophrenia. Biol Psychiatry 64:11–17.

46. Castner SA, Williams GV, Goldman-Rakic PS (2000): Reversal ofantipsychotic-induced working memory deficits by short-term dop-amine D1 receptor stimulation. Science 287:2020–2022.

47. Dorph-Petersen KA, Pierri JN, Perel JM, Sun Z, Sampson AR, Lewis DA(2005): The influence of chronic exposure to antipsychotic medi-cations on brain size before and after tissue fixation: A comparisonof haloperidol and olanzapine in macaque monkeys. Neuropsycho-pharmacology 30:1649–1661.

48. Konopaske GT, Dorph-Petersen KA, Pierri JN, Wu Q, Sampson AR,Lewis DA (2007): Effect of chronic exposure to antipsychoticmedication on cell numbers in the parietal cortex of macaquemonkeys. Neuropsychopharmacology 32:1216–1223.

49. Ho BC, Andreasen NC, Ziebell S, Pierson R, Magnotta V (2011): Long-term antipsychotic treatment and brain volumes: A longitudinalstudy of first-episode schizophrenia. Arch Gen Psychiatry 68:128–137.

50. Lewis DA, Gonzalez-Burgos G (2006): Pathophysiologically basedtreatment interventions in schizophrenia. Nat Med 12:1016–1022.

51. Moghaddam B (2004): Targeting metabotropic glutamate receptorsfor treatment of the cognitive symptoms of schizophrenia. Psycho-pharmacology (Berl) 174:39–44.

52. Malhotra AK, Pinals DA, Adler CM, Elman I, Clifton A, Pickar D, et al.(1997): Ketamine-induced exacerbation of psychotic symptoms andcognitive impairment in neuroleptic-free schizophrenics. Neuropsy-chopharmacology 17:141–150.

53. Korotkova T, Fuchs EC, Ponomarenko A, von Engelhardt J, Monyer H(2010): NMDA receptor ablation on parvalbumin-positive interneur-ons impairs hippocampal synchrony, spatial representations, andworking memory. Neuron 68:557–569.

54. Wang M, Yang Y, Wang C-J, Gamo NJ, Jin LE, Mazer JA, et al. (2013):NMDA receptors subserve persistent neuronal firing during workingmemory in dorsolateral prefrontal cortex. Neuron 77:736–749.

55. Buchanan RW, Javitt DC, Marder SR, Schooler NR, Gold JM, McMahonRP, et al. (2007): The Cognitive and Negative Symptoms inSchizophrenia Trial (CONSIST): The efficacy of glutamatergic agentsfor negative symptoms and cognitive impairments. Am J Psychiatry164:1593–1602.

56. Abi-Dargham A, Mawlawi O, Lombardo I, Gil R, Martinez D, Huang Y,et al. (2002): Prefrontal dopamine D1 receptors and working memoryin schizophrenia. J Neurosci 22:3708–3719.

57. Abi-Dargham A, Moore H (2003): Prefrontal DA transmission at D1receptors and the pathology of schizophrenia. Neuroscientist 9:404–416.

58. Mu Q, Johnson K, Morgan PS, Grenesko EL, Molnar CE, Anderson B,et al. (2007): A single 20 mg dose of the full D1 dopamine agonistdihydrexidine (DAR-0100) increases prefrontal perfusion in schizo-phrenia. Schizophr Res 94:332–341.

59. George MS, Molnar CE, Grenesko EL, Anderson B, Mu Q, Johnson K,et al. (2007): A single 20 mg dose of dihydrexidine (DAR-0100), a fulldopamine D1 agonist, is safe and tolerated in patients withschizophrenia. Schizophr Res 93:42–50.

60. Meyer-Lindenberg A, Kohn PD, Kolachana B, Kippenhan S,McInerney-Leo A, Nussbaum R, et al. (2005): Midbrain dopamineand prefrontal function in humans: Interaction and modulation byCOMT genotype. Nat Neurosci 8:594–596.

61. Goff DC, Hill M, Barch D (2011): The treatment of cognitive impair-ment in schizophrenia. Pharmacol Biochem Behav 99:245–253.

62. Lewis DA, Volk DW, Hashimoto T (2004): Selective alterations inprefrontal cortical GABA neurotransmission in schizophrenia: A noveltarget for the treatment of working memory dysfunction. Psycho-pharmacology (Berl) 174:143–150.

63. Howard MW, Rizzuto DS, Caplan JB, Madsen JR, Lisman J,Aschenbrenner-Scheibe R, et al. (2003): Gamma oscillations corre-late with working memory load in humans. Cereb Cortex 13:1369–1374.

64. Torrey EF, Barci BM, Webster MJ, Bartko JJ, Meador-Woodruff JH,Knable MB (2005): Neurochemical markers for schizophrenia, bipolardisorder, and major depression in postmortem brains. Biol Psychiatry57:252–260.

65. Lewis DA, Cho RY, Carter CS, Eklund K, Forster S, Kelly MA, et al.(2008): Subunit-selective modulation of GABA type A receptorneurotransmission and cognition in schizophrenia. Am J Psychiatry165:1585–1593.

66. Buchanan RW, Keefe RS, Lieberman JA, Barch DM, Csernansky JG,Goff DC, et al. (2011): A randomized clinical trial of MK-0777 for thetreatment of cognitive impairments in people with schizophrenia.Biol Psychiatry 69:442–449.

67. Schubert MH, Young KA, Hicks PB (2006): Galantamine improvescognition in schizophrenic patients stabilized on risperidone. BiolPsychiatry 60:530–533.

68. Singh J, Kour K, Jayaram MB (2012): Acetylcholinesterase inhibitorsfor schizophrenia. Cochrane Database Syst Rev 1:CD007967.

69. Javitt DC, Buchanan RW, Keefe RS, Kern R, McMahon RP, Green MF,et al. (2012): Effect of the neuroprotective peptide davunetide(AL-108) on cognition and functional capacity in schizophrenia.Schizophr Res 136:25–31.

70. Hong LE, Thaker GK, McMahon RP, Summerfelt A, Rachbeisel J, FullerRL, et al. (2011): Effects of moderate-dose treatment with vareniclineon neurobiological and cognitive biomarkers in smokers and non-smokers with schizophrenia or schizoaffective disorder. Arch GenPsychiatry 68:1195–1206.

71. Ritsner MS, Gibel A, Shleifer T, Boguslavsky I, Zayed A, Maayan R,et al. (2010): Pregnenolone and dehydroepiandrosterone as anadjunctive treatment in schizophrenia and schizoaffective disorder:

368 BIOL PSYCHIATRY 2014;75:361–370 T.A. Lett et al.

www.sobp.org/journal

An 8-week, double-blind, randomized, controlled, 2-center, parallel-group trial. J Clin Psychiatry 71:1351–1362.

72. McClure MM, Harvey PD, Goodman M, Triebwasser J, New A,Koenigsberg HW, et al. (2010): Pergolide treatment of cognitivedeficits associated with schizotypal personality disorder: Continuedevidence of the importance of the dopamine system in theschizophrenia spectrum. Neuropsychopharmacology 35:1356–1362.

73. Levkovitz Y, Mendlovich S, Riwkes S, Braw Y, Levkovitch-Verbin H,Gal G, et al. (2010): A double-blind, randomized study of minocyclinefor the treatment of negative and cognitive symptoms in early-phase schizophrenia. J Clin Psychiatry 71:138–149.

74. Barch DM, Carter CS (2005): Amphetamine improves cognitivefunction in medicated individuals with schizophrenia and in healthyvolunteers. Schizophr Res 77:43–58.

75. Ehrenreich H, Hinze-Selch D, Stawicki S, Aust C, Knolle-Veentjer S,Wilms S, et al. (2007): Improvement of cognitive functions in chronicschizophrenic patients by recombinant human erythropoietin. MolPsychiatry 12:206–220.

76. Cicerone KD, Dahlberg C, Malec JF, Langenbahn DM, Felicetti T,Kneipp S, et al. (2005): Evidence-based cognitive rehabilitation:Updated review of the literature from 1998 through 2002. Arch PhysMed Rehabil 86:1681–1692.

77. Wykes T, Huddy V, Cellard C, McGurk SR, Czobor P (2011): A meta-analysis of cognitive remediation for schizophrenia: Methodologyand effect sizes. Am J Psychiatry 168:472–485.

78. Wykes T, Brammer M, Mellers J, Bray P, Reeder C, Williams C, et al.(2002): Effects on the brain of a psychological treatment: Cognitiveremediation therapy: Functional magnetic resonance imaging inschizophrenia. Br J Psychiatry 181:144–152.

79. Wexler BE, Anderson M, Fulbright RK, Gore JC (2000): Preliminaryevidence of improved verbal working memory performance andnormalization of task-related frontal lobe activation in schizophreniafollowing cognitive exercises. Am J Psychiatry 157:1694–1697.

80. Hogarty GE, Flesher S, Ulrich R, Carter M, Greenwald D, Pogue-GeileM, et al. (2004): Cognitive enhancement therapy for schizophrenia:Effects of a 2-year randomized trial on cognition and behavior. ArchGen Psychiatry 61:866–876.

81. Eack SM, Hogarty GE, Cho RY, Prasad KM, Greenwald DP, HogartySS, et al. (2010): Neuroprotective effects of cognitive enhancementtherapy against gray matter loss in early schizophrenia: Results froma 2-year randomized controlled trial. Arch Gen Psychiatry 67:674–682.

82. Keller TA, Just MA (2009): Altering cortical connectivity: Remediation-induced changes in the white matter of poor readers. Neuron 64:624–631.

83. Penadés R, Pujol N, Catalán R, Massana G, Rametti G, García-Rizo C,et al. (2013): Brain effects of cognitive remediation therapy inschizophrenia: A structural and functional neuroimaging study. BiolPsychiatry 73:1015–1023.

84. McGurk SR, Twamley EW, Sitzer DI, McHugo GJ, Mueser KT (2007): Ameta-analysis of cognitive remediation in schizophrenia. Am JPsychiatry 164:1791–1802.

85. Vinogradov S, Fisher M, Nagarajan S (2013): Cognitive training inschizophrenia: Golden age or wild west? Biol Psychiatry 73:935–937.

86. Dickinson D, Tenhula W, Morris S, Brown C, Peer J, Spencer K, et al.(2010): A randomized, controlled trial of computer-assisted cognitiveremediation for schizophrenia. Am J Psychiatry 167:170–180.

87. Adcock RA, Dale C, Fisher M, Aldebot S, Genevsky A, Simpson GV,et al. (2009): When top-down meets bottom-up: Auditory trainingenhances verbal memory in schizophrenia. Schizophr Bull 35:1132–1141.

88. Fisher M, Holland C, Merzenich MM, Vinogradov S (2009): Usingneuroplasticity-based auditory training to improve verbal memory inschizophrenia. Am J Psychiatry 166:805–811.

89. Fisher M, Holland C, Subramaniam K, Vinogradov S (2010):Neuroplasticity-based cognitive training in schizophrenia: An interimreport on the effects 6 months later. Schizophr Bull 36:869–879.

90. Murthy NV, Mahncke H, Wexler BE, Maruff P, Inamdar A, ZucchettoM, et al. (2012): Computerized cognitive remediation training forschizophrenia: An open label, multi-site, multinational methodologystudy. Schizophr Res 139:87–91.

91. Hallett M (2000): Transcranial magnetic stimulation and the humanbrain. Nature 406:147–150.

92. Rossi S, Cappa SF, Babiloni C, Pasqualetti P, Miniussi C, Carducci F,et al. (2001): Prefrontal [correction of Prefontal] cortex in long-termmemory: An “interference” approach using magnetic stimulation.Nat Neurosci 4:948–952.

93. Rami L, Gironell A, Kulisevsky J, Garcia-Sanchez C, Berthier M,Estevez-Gonzalez A (2003): Effects of repetitive transcranial magneticstimulation on memory subtypes: A controlled study. Neuropsycho-logia 41:1877–1883.

94. Mottaghy FM, Sparing R, Topper R (2006): Enhancing picture namingwith transcranial magnetic stimulation. Behav Neurol 17:177–186.

95. Cappa SF, Sandrini M, Rossini PM, Sosta K, Miniussi C (2002): The roleof the left frontal lobe in action naming: rTMS evidence. Neurology59:720–723.

96. Boroojerdi B, Phipps M, Kopylev L, Wharton CM, Cohen LG, GrafmanJ (2001): Enhancing analogic reasoning with rTMS over the leftprefrontal cortex. Neurology 56:526–528.

97. Hansenne M, Laloyaux O, Mardaga S, Ansseau M (2004): Impact oflow frequency transcranial magnetic stimulation on event-relatedbrain potentials. Biol Psychol 67:331–341.

98. Rossini PM, Rossi S (2007): Transcranial magnetic stimulation.Neurology 68:484–488.

99. Rogasch NC, Fitzgerald PB (2013): Assessing cortical network proper-ties using TMS-EEG. Hum Brain Mapp 34:1652–1669.

100. Daskalakis ZJ, Farzan F, Barr MS, Rusjan PM, Favalli G, LevinsonAJ, et al. (2008): Evaluating the relationship between longinterval cortical inhibition, working memory and gamma bandactivity in the dorsolateral prefrontal cortex. Clin EEG Neurosci 39:150–155.

101. Hoppenbrouwers SS, De Jesus DR, Stirpe T, Fitzgerald PB,Voineskos AN, Schutter DJLG, et al. (2013): Inhibitory deficits inthe dorsolateral prefrontal cortex in psychopathic offenders. Cortex49:1377–1385.

102. Daskalakis ZJ, Farzan F, Barr MS, Maller JJ, Chen R, Fitzgerald PB(2008): Long-interval cortical inhibition from the dorsolateral pre-frontal cortex: A TMS-EEG study. Neuropsychopharmacology 33:2860–2869.

103. Uhlhaas PJ, Singer W (2010): Abnormal neural oscillations andsynchrony in schizophrenia. Nat Rev Neurosci 11:100–113.

104. Light GA, Hsu JL, Hsieh MH, Meyer-Gomes K, Sprock J, Swerdlow NR,et al. (2006): Gamma band oscillations reveal neural network corticalcoherence dysfunction in schizophrenia patients. Biol Psychiatry 60:1231–1240.

105. Farzan F, Barr MS, Levinson AJ, Chen R, Wong W, Fitzgerald PB, et al.(2010): Evidence for gamma inhibition deficits in the dorsolateralprefrontal cortex of patients with schizophrenia. Brain 133:1505–1514.

106. Kirihara K, Rissling AJ, Swerdlow NR, Braff DL, Light GA (2012):Hierarchical organization of gamma and theta oscillatory dynamics inschizophrenia. Biol Psychiatry 71:873–880.

107. Cho RY, Konecky RO, Carter CS (2006): Impairments in frontal corticalgamma synchrony and cognitive control in schizophrenia. Proc NatlAcad Sci U S A 103:19878–19883.

108. Herrmann CS, Munk MH, Engel AK (2004): Cognitive functions ofgamma-band activity: Memory match and utilization. Trends Cogn Sci8:347–355.

109. Bartos M, Vida I, Jonas P (2007): Synaptic mechanisms of synchron-ized gamma oscillations in inhibitory interneuron networks. Nat RevNeurosci 8:45–56.

110. Canolty RT, Edwards E, Dalal SS, Soltani M, Nagarajan SS, Kirsch HE,et al. (2006): High gamma power is phase-locked to theta oscillationsin human neocortex. Science 313:1626–1628.

111. Papp E, Leinekugel X, Henze DA, Lee J, Buzsaki G (2001): The apicalshaft of CA1 pyramidal cells is under GABAergic interneuronalcontrol. Neuroscience 102:715–721.

112. Sohal VS, Zhang F, Yizhar O, Deisseroth K (2009): Parvalbuminneurons and gamma rhythms enhance cortical circuit performance.Nature 459:698–702.

113. Barr MS, Farzan F, Tran LC, Chen R, Fitzgerald PB, Daskalakis ZJ (2010):Evidence for excessive frontal evoked gamma oscillatory activity inschizophrenia during working memory. Schizophr Res 121:146–152.

114. Haenschel C, Bittner RA, Waltz J, Haertling F, Wibral M, Singer W, et al.(2009): Cortical oscillatory activity is critical for working memory asrevealed by deficits in early-onset schizophrenia. J Neurosci 29:9481–9489.

T.A. Lett et al. BIOL PSYCHIATRY 2014;75:361–370 369

www.sobp.org/journal

115. Thickbroom GW (2007): Transcranial magnetic stimulation andsynaptic plasticity: Experimental framework and human models.Exp Brain Res 180:583–593.

116. Chen R, Classen J, Gerloff C, Celnik P, Wassermann EM, Hallett M, et al.(1997): Depression of motor cortex excitability by low-frequencytranscranial magnetic stimulation. Neurology 48:1398–1403.

117. Pascual-Leone A, Valls-Sole J, Wassermann EM, Hallett M (1994):Responses to rapid-rate transcranial magnetic stimulation of thehuman motor cortex. Brain 117:847–858.

118. Ziemann U, Paulus W, Nitsche MA, Pascual-Leone A, Byblow WD,Berardelli A, et al. (2008): Consensus: Motor cortex plasticity proto-cols. Brain Stimul 1:164–182.

119. Rajji TK, Liu SK, Frantseva MV, Mulsant BH, Thoma J, Chen R, et al.(2011): Exploring the effect of inducing long-term potentiation in thehuman motor cortex on motor learning. Brain Stimul 4:137–144.

120. Benali A, Trippe J, Weiler E, Mix A, Petrasch-Parwez E, Girzalsky W,et al. (2011): Theta-burst transcranial magnetic stimulation alterscortical inhibition. J Neurosci 31:1193–1203.

121. Huang YZ, Edwards MJ, Rounis E, Bhatia KP, Rothwell JC (2005): Thetaburst stimulation of the human motor cortex. Neuron 45:201–206.

122. Pettersson-Yeo W, Allen P, Benetti S, McGuire P, Mechelli A (2011):Dysconnectivity in schizophrenia: Where are we now? NeurosciBiobehav Rev 35:1110–1124.

123. Meyer-Lindenberg AS, Olsen RK, Kohn PD, Brown T, Egan MF,Weinberger DR, et al. (2005): Regionally specific disturbance ofdorsolateral prefrontal-hippocampal functional connectivity inschizophrenia. Arch Gen Psychiatry 62:379–386.

124. Spencer KM, Nestor PG, Niznikiewicz MA, Salisbury DF, Shenton ME,McCarley RW (2003): Abnormal neural synchrony in schizophrenia.J Neurosci 23:7407–7411.

125. Voineskos AN, Lobaugh NJ, Bouix S, Rajji TK, Miranda D, Kennedy JL,et al. (2010): Diffusion tensor tractography findings in schizophreniaacross the adult lifespan. Brain 133:1494–1504.

126. Kubicki M, Park H, Westin CF, Nestor PG, Mulkern RV, Maier SE, et al.(2005): DTI and MTR abnormalities in schizophrenia: Analysis of whitematter integrity. Neuroimage 26:1109–1118.

127. Sommer I, Ramsey N, Kahn R, Aleman A, Bouma A (2001): Handed-ness, language lateralisation and anatomical asymmetry in schizo-phrenia: Meta-analysis. Br J Psychiatry 178:344–351.

128. Kapur S (2003): Psychosis as a state of aberrant salience: A frameworklinking biology, phenomenology, and pharmacology in schizophre-nia. Am J Psychiatry 160:13–23.

129. Glantz LA, Lewis DA (2000): Decreased dendritic spine density onprefrontal cortical pyramidal neurons in schizophrenia. Arch GenPsychiatry 57:65–73.

130. Ayalew M, Le-Niculescu H, Levey DF, Jain N, Changala B, Patel SD,et al. (2012): Convergent functional genomics of schizophrenia: Fromcomprehensive understanding to genetic risk prediction. Mol Psy-chiatry 17:887–905.

131. Lett TA, Chakavarty MM, Felsky D, Brandl EJ, Tiwari AK, Goncalves VF,et al. (2013): The genome-wide supported microRNA-137 variantpredicts phenotypic heterogeneity within schizophrenia. Mol Psy-chiatry 18:443-450.

132. Tan HY, Choo WC, Fones CS, Chee MW (2005): fMRI study ofmaintenance and manipulation processes within working memoryin first-episode schizophrenia. Am J Psychiatry 162:1849–1858.

133. Barr MS, Farzan F, Arenovich T, Chen R, Fitzgerald PB, Daskalakis ZJ(2011): The effect of repetitive transcranial magnetic stimulation ongamma oscillatory activity in schizophrenia. PLoS One 6:e22627.

134. Flavell SW, Greenberg ME (2008): Signaling mechanisms linkingneuronal activity to gene expression and plasticity of the nervoussystem. Annu Rev Neurosci 31:563.

135. Deisseroth K, Bito H, Tsien RW (1996): Signaling from synapse tonucleus: Postsynaptic CREB phosphorylation during multiple forms ofhippocampal synaptic plasticity. Neuron 16:89–101.

136. Ji RR, Schlaepfer TE, Aizenman CD, Epstein CM, Qiu D, Huang JC, et al.(1998): Repetitive transcranial magnetic stimulation activates specificregions in rat brain. Proc Natl Acad Sci U S A 95:15635–15640.

137. Tao X, Finkbeiner S, Arnold DB, Shaywitz AJ, Greenberg ME (1998):Ca2� influx regulates BDNF transcription by a CREB family tran-scription factor-dependent mechanism. Neuron 20:709–726.

138. Muller D, Djebbara-Hannas Z, Jourdain P, Vutskits L, Durbec P,Rougon G, et al. (2000): Brain-derived neurotrophic factor restoreslong-term potentiation in polysialic acid-neural cell adhesionmolecule-deficient hippocampus. Proc Natl Acad Sci U S A 97:4315–4320.

139. Hellmann J, Juttner R, Roth C, Bajbouj M, Kirste I, Heuser I, et al.(2012): Repetitive magnetic stimulation of human-derived neuron-like cells activates cAMP-CREB pathway. Eur Arch Psychiatry ClinNeurosci 262:87–91.

140. Vo N, Klein ME, Varlamova O, Keller DM, Yamamoto T, Goodman RH,et al. (2005): A cAMP-response element binding protein-inducedmicroRNA regulates neuronal morphogenesis. Proc Natl Acad Sci U S A102:16426–16431.

141. Wayman GA, Davare M, Ando H, Fortin D, Varlamova O, Cheng HY,et al. (2008): An activity-regulated microRNA controls dendriticplasticity by down-regulating p250GAP. Proc Natl Acad Sci U S A 105:9093–9098.

142. Miller BH, Zeier Z, Xi L, Lanz TA, Deng S, Strathmann J, et al. (2012):MicroRNA-132 dysregulation in schizophrenia has implications forboth neurodevelopment and adult brain function. Proc Natl Acad SciU S A 109:3125–3130.

143. Barr MS, Farzan F, Rajji TK, Voineskos AN, Blumberger DM, ArenovichT, et al. (2013): Can repetitive magnetic stimulation improvecognition in schizophrenia? Pilot data from a randomized controlledtrial. Biol Psychiatry 73:510–517.

144. Guse B, Falkai P, Gruber O, Whalley H, Gibson L, Hasan A, et al. (2012):The effect of long-term high frequency repetitive transcranialmagnetic stimulation on working memory in schizophrenia andhealthy controls-A randomized placebo-controlled, double-blindfMRI study. Behav Brain Res 237C:300–307.

145. Utz KS, Dimova V, Oppenlander K, Kerkhoff G (2010): Electrifiedminds: Transcranial direct current stimulation (tDCS) and galvanicvestibular stimulation (GVS) as methods of non-invasive brainstimulation in neuropsychology—a review of current data and futureimplications. Neuropsychologia 48:2789–2810.

146. Mulquiney PG, Hoy KE, Daskalakis ZJ, Fitzgerald PB (2011): Improvingworking memory: Exploring the effect of transcranial random noisestimulation and transcranial direct current stimulation on the dorso-lateral prefrontal cortex. Clin Neurophysiol 122:2384–2389.

147. Polania R, Nitsche MA, Korman C, Batsikadze G, Paulus W (2012): Theimportance of timing in segregated theta phase-coupling forcognitive performance. Curr Biol 22:1314–1318.

148. Guse B, Falkai P, Wobrock T (2010): Cognitive effects of high-frequency repetitive transcranial magnetic stimulation: A systematicreview. J Neural Transm 117:105–122.

149. Lisman JE, Coyle JT, Green RW, Javitt DC, Benes FM, Heckers S, et al.(2008): Circuit-based framework for understanding neurotransmitterand risk gene interactions in schizophrenia. Trends Neurosci 31:234–242.

150. Barr MS, Farzan F, Rusjan PM, Chen R, Fitzgerald PB, Daskalakis ZJ(2009): Potentiation of gamma oscillatory activity through repetitivetranscranial magnetic stimulation of the dorsolateral prefrontalcortex. Neuropsychopharmacology 34:2359-2367.

370 BIOL PSYCHIATRY 2014;75:361–370 T.A. Lett et al.

www.sobp.org/journal