This electronic thesis or dissertation has been downloaded from the King’s Research Portal at https://kclpure.kcl.ac.uk/portal/
Trajectories of DNA Methylation Associated with Clozapine Exposure and Clinical Outcomes
Gillespie, Amy Louise
Awarding institution: King's College London
The copyright of this thesis rests with the author and no quotation from it or information derived from it may be published without proper acknowledgement.
END USER LICENCE AGREEMENT
Unless another licence is stated on the immediately following page this work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International licence. https://creativecommons.org/licenses/by-nc-nd/4.0/ You are free to copy, distribute and transmit the work Under the following conditions: Attribution: You must attribute the work in the manner specified by the author (but not in any way that suggests that they endorse you or your use of the work). Non Commercial: You may not use this work for commercial purposes. No Derivative Works - You may not alter, transform, or build upon this work.
Any of these conditions can be waived if you receive permission from the author. Your fair dealings and other rights are in no way affected by the above.
Take down policy
If you believe that this document breaches copyright please contact [email protected] providing details, and we will remove access to the work immediately and investigate your claim.
Download date: 30. Sep. 2021 List of Tables
TRAJECTORIES OF DNA METHYLATION ASSOCIATED WITH CLOZAPINE EXPOSURE AND CLINICAL OUTCOMES
Amy Gillespie
INSTITUTE OF PSYCHIATRY PSYCHOLOGY AND NEUROSCIENCE, KING'S COLLEGE
- 1 -
List of Tables
Contents
List of Tables ...... 6
List of Figures ...... 7
List of Abbreviations ...... 8
Abstract ...... 10
Acknowledgements ...... 11
1 Introduction ...... 13
1.1 Schizophrenia ...... 13
1.2 Treatment-resistant schizophrenia ...... 13
1.2.1 Defining treatment-resistant schizophrenia ...... 14
1.2.2 Significance of treatment-resistant schizophrenia ...... 17
1.3 Clozapine ...... 17
1.4 What is DNA methylation and why study its role in clozapine? ...... 21
1.4.1 The genome and gene expression ...... 21
1.4.2 Genetic variation and schizophrenia ...... 24
1.4.3 The epigenome ...... 28
1.4.4 Epigenetic variation and schizophrenia ...... 31
2 Overall Method ...... 38
2.1 Longitudinal study ...... 38
2.1.1 Study design ...... 38
2.1.2 Participants ...... 41
2.1.3 Clinical assessments ...... 44
2.1.4 Blood samples ...... 47
2.1.5 Quality control (QC) assessments ...... 51
2.1.6 Power ...... 52
2.2 Cross-sectional dataset ...... 55
2.3 Summary ...... 55
- 2 -
List of Tables
3 Exposure to Clozapine and Associated DNA Methylation Changes ...... 56
3.1 Background: Exposure to clozapine ...... 56
3.1.1 Existing epigenetic research...... 57
3.2 Statistical method: Exposure to clozapine ...... 62
3.3 Results: Time on clozapine (weeks) ...... 64
3.3.1 Participant characteristics (time on clozapine) ...... 64
3.3.2 Global DNA methylation (time on clozapine) ...... 73
3.3.3 Differentially methylated positions (time on clozapine) ...... 74
3.3.4 Pathway analysis (time on clozapine) ...... 81
3.3.5 Candidate CpG site methylation (time on clozapine) ...... 84
3.3.6 Comparison to cross-sectional data (time on clozapine) ...... 85
3.4 Results: Clozapine levels ...... 87
3.4.1 Participant characteristics (clozapine levels) ...... 87
3.4.2 Global DNA methylation (clozapine levels) ...... 91
3.4.3 Differentially methylated positions (clozapine levels) ...... 91
3.4.4 Pathway analysis (clozapine levels) ...... 99
3.4.5 Candidate CpG site methylation (clozapine levels)...... 102
3.4.6 Comparison to cross-sectional data (clozapine levels) ...... 103
3.5 Discussion: Exposure to clozapine ...... 106
3.5.1 Time on clozapine ...... 106
3.5.2 Clozapine level ...... 111
3.5.3 Summary ...... 114
4 Therapeutic Response to Clozapine and Associated DNA Methylation Changes ...... 116
4.1 Background: Therapeutic response to clozapine ...... 116
4.1.1 Defining response to clozapine ...... 116
4.1.2 Physiological changes associated with response to clozapine ...... 118
4.1.3 Genetic variation associated with response to clozapine ...... 119
4.1.4 Epigenetic variation associated with response to clozapine ...... 123 - 3 -
List of Tables
4.2 Statistical method: Therapeutic response to clozapine ...... 124
4.3 Results: Percentage change in response to clozapine (between-individuals) ...... 128
4.3.1 Responders vs non-responders (categorical) ...... 128
4.3.2 Change in total symptom severity (continuous) ...... 152
4.3.3 Change in psychosocial functioning (continuous)...... 158
4.3.4 Change in quality of life (continuous) ...... 163
4.4 Results: Symptom severity (within individuals) ...... 168
4.4.1 Participant characteristics ...... 168
4.4.2 Total symptom severity ...... 168
4.4.3 Positive symptom severity ...... 180
4.4.4 Negative symptom severity ...... 190
4.5 Discussion: Therapeutic response to clozapine ...... 204
4.5.1 Responders vs non-responders ...... 204
4.5.2 Change in total symptoms (continuous) ...... 209
4.5.3 Change in psychosocial functioning ...... 210
4.5.4 Change in quality of life ...... 212
4.5.5 Total symptom severity ...... 213
4.5.6 Positive symptom severity ...... 217
4.5.7 Negative symptom severity ...... 218
4.5.8 Summary ...... 220
5 Adverse Reactions to Clozapine and Associated DNA Methylation Changes ...... 223
5.1 Background: Adverse reactions to clozapine ...... 223
5.1.1 Mechanisms of adverse reactions to clozapine ...... 225
5.2 Statistical method: Adverse reactions to clozapine ...... 227
5.3 Results: Increased appetite/weight ...... 230
5.3.1 Participant characteristics (increased appetite/weight) ...... 230
5.3.2 Global DNA methylation (increased appetite/weight) ...... 244
5.3.3 Differentially methylated positions (increased appetite/weight) ...... 245 - 4 -
List of Tables
5.3.4 Pathway analysis (increased appetite/weight) ...... 251
5.3.5 Candidate CpG site methylation (increased appetite/weight) ...... 254
5.4 Results: Drowsiness ...... 256
5.4.1 Participant characteristics (drowsiness) ...... 256
5.4.2 Global DNA methylation (drowsiness) ...... 270
5.4.3 Differentially methylated positions (drowsiness) ...... 271
5.4.4 Pathway analysis (drowsiness) ...... 276
5.4.5 Candidate CpG site methylation (drowsiness) ...... 279
5.5 Results: Cardiovascular symptoms ...... 280
5.5.1 Participant characteristics (cardiovascular symptoms) ...... 280
5.5.2 Global DNA methylation (cardiovascular symptoms) ...... 295
5.5.3 Differentially methylated positions (cardiovascular symptoms) ...... 296
5.5.4 Pathway analysis (cardiovascular symptoms) ...... 299
5.5.5 Candidate CpG site methylation (cardiovascular symptoms) ..... Error! Bookmark not defined.
5.6 Discussion: Adverse reactions to clozapine ...... 302
5.6.1 Increased appetite/weight ...... 302
5.6.2 Drowsiness ...... 304
5.6.3 Cardiovascular symptoms ...... 305
5.6.4 Summary ...... 305
6 General Discussion ...... 308
6.1 Summary of findings ...... 308
6.2 Strengths and limitations ...... 309
6.3 Comparison to broader literature ...... 316
6.4 Future research ...... 319
7 Conclusions ...... 321
REFERENCES ...... 322
APPENDIX I ...... 364 - 5 -
List of Tables
List of Tables
Table 1: Characteristics of participants included in Time on Clozapine Analysis ...... 65 Table 2: Top ten DMPs associated with time on clozapine (* indicates epigenome wide significance) ...... 76 Table 3: Top thirty over-represented pathways - time on clozapine ...... 82 Table 4: Top ten DMPs associated with clozapine level (* indicates epigenome wide significance) ... 93 Table 5: Top thirty over-represented pathways - clozapine level ...... 100 Table 6: Characteristics of responders to clozapine vs non-responders ...... 129 Table 7: Top ten DMPs associated with categorical response to clozapine classification (* indicates epigenome wide significance)...... 141 Table 8 Top thirty over-represented pathways - response group ...... 147 Table 9: Top ten DMPs associated with change in total symptom severity (continuous) (* indicates epigenome wide significance)...... 156 Table 10: Top ten DMPs associated with change in psychosocial functioning (* indicates epigenome wide significance) ...... 160 Table 11: Top ten DMPs associated with change in quality of life (* indicates epigenome wide significance) ...... 165 Table 12: Top ten DMPs associated with within-individual variation in total symptom severity (* indicates epigenome wide significance) ...... 170 Table 13: Top thirty over-represented pathways - total symptom severity ...... 178 Table 14: Top ten DMPs associated with within-individual variation in positive symptom severity (* indicates epigenome wide significance) ...... 182 Table 15: Top ten DMPs associated with within-individual variation in negative symptom severity (* indicates epigenome wide significance) ...... 192 Table 16: Characteristics of participants with and without increased appetite/weight at 12 weeks 231 Table 17: Top ten DMPs associated with increased appetite/weight (* indicates epigenome wide significance) ...... 247 Table 18: Top thirty over-represented pathways - increased appetite/weight ...... 252 Table 19: Characteristics of participants with and without drowsiness at 12 weeks ...... 257 Table 20: Top ten DMPs associated with drowsiness (* indicates epigenome wide significance) .... 273 Table 21: Top thirty over-represented pathways - drowsiness ...... 277 Table 22: Characteristics of participants with and without cardiovascular symptoms at 12 weeks .. 281 Table 23: Top ten DMPs associated with cardiovascular symptoms (* indicates epigenome wide significance) ...... 298 Table 24 Top thirty over-represented pathways - cardiovascular symptoms ...... 300
- 6 -
List of Figures
List of Figures
Figure 1: Image of DNA and chromosome structure, taken from Annunziato (2008) ...... 22 Figure 2: Image of gene expression taken from Clancy and Brown (2008) ...... 23 Figure 3: Flowchart of assessments and samples collected throughout the study ...... 40 Figure 4: Workflow for Illumina 450k BeadChip assays, from Infinium HD Methylation Assay Protocol GuideManual...... 49 Figure 5: Flowchart of QC assessments ...... 54 Figure 6: Pie charts illustrating key demographics across participants ...... 67 Figure 7: Pie charts comparing key demographics for responders vs non-responders...... 131 Figure 8: Pie charts comparing key demographics for those with and without increased appetite/weight ...... 233 Figure 9: Pie-charts of characteristics in those with and without drowsiness ...... 259 Figure 10: Pie charts comparing key demographics for those with and without cardiovascular symptoms ...... 283
- 7 -
List of Abbreviations
List of Abbreviations
AMPAR alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor
ATP adenosine triphosphate
BCD bisulfite converted DNA
BPRS Brief Psychiatric Rating Scale
CGI Clinical Global Impression
CNS central nervous system
CNV copy number variant
CpG cytosine-phosphate-guanine)
DLPFC dorsolateral prefrontal cortex
DMPs differentially methylated positions
DNA deoxyribonucleic acid
DNMTs DNA methyltransferases
EEG electroencephalogram
EWAS epigenome wide association study
GABA gamma-Aminobutyric acid
GAF Global Assessment of Functioning
GASS-C Glasgow Antipsychotic Adverse reaction Scale clozapine
GTP Guanosine-5'-triphosphate
GPCR g-coupled protein receptor
GWAS genome-wide association studies
HDACs histone deacetylases
5-HIAA 5-Hydroxyindoleacetic acid
HVA homovanillic acid
MAO monoamine oxidase
MAPK mitogen-activated protein kinase
- 8 -
List of Abbreviations
MHC major histocompatibility complex
MHPG 3-methoxy-4-hydroxyphenylglycol mQTLs methylation quantitative trait loci mRNA messengerRNA
MZ monozygotic
NAA n-acetyl aspartate
NMDAR N-methyl-D-aspartate (NMDA) receptor
PANSS Positive and Negative Syndrome Scale
PCP phencyclidine
PGC Psychiatric Genetics Consortium
PPI prepulse inhibition
PSP Personal and Social Performance scale
QC quality control
RNA Ribonucleic acid
SAM S-Adenosyl methionine
SLaM South London and Maudsley
SNP single nucleotide polymorphism
TSS transcription start site
UTR untranslated region
- 9 -
Abstract
Abstract
Clozapine is an atypical antipsychotic known to be uniquely effective in patients with treatment- resistant schizophrenia, but associated with numerous adverse reactions. Despite substantial research efforts, we lack a full understanding of the mechanistic processes underlying clozapine’s efficacy and identification of the most relevant mechanisms. Animal studies suggest that clozapine induces changes in DNA methylation (a dynamic epigenetic modification which can regulate gene expression). However, research in humans is extremely limited.
This thesis provides an original contribution to the literature as the first study to report positions of significant differences in longitudinal DNA methylation trajectories associated with clozapine exposure, therapeutic response and adverse reactions. 22 participants with schizophrenia were recruited and within-participant sampling was conducted before clozapine initiation and at three time-points over the subsequent six months. The Illumina 450k array was used to assay epigenome- wide DNA methylation in 85 samples of whole blood. I conducted multi-level regression models and pathway analysis.
Across all models, I identified 33 epigenome-wide significant (p<1*10-7) differently methylated positions (DMPS), with over-representation of DMPS indicated to have correlated methylation in brain tissue. DMPs located on genes involved in signal transduction, glutamatergic and GABAergic systems, calcium signalling, glucose homeostasis, immune function and transcription regulation were associated with both clozapine exposure and therapeutic efficacy. Additionally, DMPs located on genes involved in cell adhesion, glycine and neuronal development were associated specifically with therapeutic efficacy. DMPs located on genes involved in immune function and signal transduction were associated with drowsiness, increased appetite and weight, and cardiovascular symptoms. DMPs located on genes involved in energy metabolism were associated with increases in appetite and weight.
These results align with literature suggesting that clozapine normalises schizophrenia-associated disruption in glutamatergic and GABAergic systems, calcium signalling and immune function; however, further studies of gene expression and downstream effects are required to establish the causal role of DNA methylation.
- 10 -
Acknowledgements
Acknowledgements
Firstly, I would like to express my sincere gratitude to my supervisors, Dr James MacCabe and Professor Jonathan Mill: for giving me to the opportunity to do a dream PhD, for their continual support of my academic work, and for their scientific guidance, optimism and faith in me.
I cannot possibly name everyone at the Institute of Psychiatry, Psychology and Neuroscience who has supported me over the past few years; I was lucky enough to meet many of you, from all over the faculty, and my experience was much richer for it.
Special thanks go to:
The rest of the recruitment team who made this research possible through their tireless efforts: Grant McQueen, Alessia Avila, John McLally and Kyra Verena-Sendt. Alice Egerton and Gemma Modinos, women I look up to and who I’ve been lucky enough to have believe in me and push me to believe in myself All the security staff that enabled me to indulge in my night owl tendencies and work all hours of the night, and regularly lifted my spirits in the last month of doing so!
I’d like to also thank colleagues from at the University of Exeter, particularly Eilis Hannon for her patience and generosity in guiding someone new to R! Similarly, thanks go to the community of Stack Overflow; I’m not sure how anyone survived programming without this wealth of knowledge available 24/7!
I am also indebted to all the participants, who gave up their time and energy to be involved in this research, often during some of the most challenging periods of their life. Thank you.
More personally, I would not be here if I did not have the most incredible parents, who have consistently supported and championed me for as long as I can remember. It means the world to me to have their unquestioning belief in me and know that they will be always there for me, in any type of crisis – whether practical or a crisis of confidence. Thank you both!
I’m grateful to all my friends who have kept me going and been my cheerleaders throughout the rollercoaster of a PhD – this is especially true of Holly, Hannah, Sophie, and Aaron, who have never failed to lift my spirits!
Last, but certainly not least, to my fiancé Laurie: You were part of my inspiration for returning to academic research and I could not have asked for a more supportive and patient partner every step of the way since then; everything I have needed, you have provided without complaint, from morale
- 11 -
Acknowledgements boosts to editing to finding us a beautiful new home during the last month of my write-up. You helped get me to the finish line and I will be forever grateful.
- 12 -
Introduction
1 Introduction
In this chapter, I will provide a broad background to my thesis. For the first half of the chapter I will introduce the general issues surrounding treatment-resistant schizophrenia and clozapine and for the second half of the chapter I will introduce the concepts of genetic and epigenetic variation, their association with schizophrenia and antipsychotic use, and the limitations of the current literature. Subsequent chapters will provide more detailed reviews of the literature regarding associations with clozapine exposure, clozapine response and adverse reactions to clozapine.
1.1 Schizophrenia
Schizophrenia is a debilitating psychiatric illness, consisting of positive symptoms (e.g. persistent delusions), negative symptoms (e.g. blunted affect) and cognitive deficits. It has a complex, multi- faceted aetiology with both genetic and environmental risk factors known to be involved. It is the most common psychotic disorder (Perala et al., 2007); the lifetime prevalence rate of schizophrenia is 0.4% (Saha, Chant, Welham, & McGrath, 2005), and it is ranked as one of the top ten causes of disability in developed countries by the World Health Organisation (Murray & Lopez, 1996). Onset usually occurs during the early twenties (Sham, MacLean, & Kendler, 1994), and while it was initially conceptualised by Emil Kraepelin as “dementia praecox” (Kraepelin, 1899) - a degenerative disorder with no favourable outcome – it is now known to be treatable, with pharmacological treatment in the form of antipsychotics available and full recovery possible. Unfortunately, some patients are less responsive to treatment than others, with some patients showing no meaningful response to current antipsychotic treatment, other than clozapine. The majority of these patients are non-responsive from first onset, suggesting that treatment-resistance is an enduring trait (Demjaha et al., 2017).
1.2 Treatment-resistant schizophrenia
The existence of a group of “therapy resistant” patients, showing limited or no response to treatment has been evident since the 1950s. This is when the first antipsychotic – chlorpromazine – was discovered; a major breakthrough in psychiatric care, providing the first evidence-based treatment for reducing psychotic symptoms in patients with schizophrenia (Ban, 2007). However, since some of the first early antipsychotic trials (Cole, Goldberg, & Davis, 1966; Klein & Davis JM, 1969) most estimates have agreed that approximately a third of patients are treatment-resistant (R. R. Conley & Kelly, 2001; Elkis, 2007; Lehman et al., 2004). Some studies consider this an underestimate, reporting higher figures of 40 or 60% (Essock, Hargreaves, Covell, & Goethe, 1996; Juarez-Reyes et al., 1995). Similarly, Hegarty, Baldessarini, Tohen, Waternaux and Oepen (1994) found that a minimum of patients - only 35-45% - achieve good outcomes with conventional
- 13 -
Introduction antipsychotics; and this is without considering cognitive impairment, quality of life assessment, or the negative impact of adverse reactions, which would all reduce that number further. These differing estimates of the proportion of schizophrenia patients who are treatment-resistant are in part down to differing definitions of treatment-resistance.
1.2.1 Defining treatment-resistant schizophrenia
“Treatment resistance in schizophrenia is difficult to define and criteria still remain controversial” (Painuly, Gupta, & Avasthi, 2004)
Initially, treatment-resistance was equated with chronic hospitalisation (hospitalisation for more than two years) (Small, Kellams, Milstein, & Moore, 1975) on the assumption that remaining hospitalised reflected a lack of improvement in symptoms preventing discharge; however, hospitalisation can be required for numerous reasons including poor compliance or issues with housing – neither of which indicate an inferior response to antipsychotic treatment – and the threshold for admission differs between countries and has changed rapidly over the past 40 years. Since then, the key feature of treatment-resistance has been the persistence of positive symptoms despite adequate drug therapy. To rule out inadequate treatment as the reason for persistent symptoms, time frames and dosage for previous antipsychotic trials began to be specified. One of the first well-operationalised sets of criteria including these features was provided by Kane, Honigfeld, Singer and Meltzer (1988), reporting on a seminal trial discussed in the following section:
1. Persistent positive psychotic symptoms: item score > 4 (moderate) on at least two of four positive symptom items of the Brief Psychiatric Rating Scale (BPRS; rated on a 1-7 scale) - hallucinatory behaviour, suspiciousness, unusual thought content, and conceptual disorganization. 2. Current presence of at least moderately severe illness as rated by the total BPRS score (>4 on the 18-item scale) and a score of >4 (moderate) on the Clinical Global Impression (CGI) scale. 3. Presence of illness: no stable period of good social and/ or occupational functioning within the last five years (inability to maintain work and relationship). 4. Drug-refractory condition: at least three periods of treatment in the preceding five years with conventional antipsychotics (from at least two chemical classes) at doses equivalent to > 1,000 mg per day of chlorpromazine of six weeks each, without significant symptom relief; and either failure to improve by at least 20% in total BPRS score or intolerance to a six week prospective trial of haloperidol at a dose of 10-60 mg per day.
- 14 -
Introduction
Most current definitions are refined, condensed or modified versions of these criteria; however, this still allows for significant heterogeneity in definitions used by research groups. To address this, the Treatment Response and Resistance in Psychosis (TRRIP) working group has established consensus criteria to standardize the definition of treatment resistance, aiding study designs and facilitating comparison of results across different studies (Howes et al., 2016).
They recommend minimum requirements including:
at least moderately severe current symptoms resistance, assessed by interview using a standardised rating scale, lasting a minimum of 12 weeks at least moderate functional impairment with validated scale minimal improvement in past response, and assessment of past response using information from patient/carer/staff reports, pill counts and dispensing charts at least two past adequate treatment episodes with different antipsychotics at least six weeks at a therapeutic dose of at least 600mg of chlorpromazine equivalent per day with minimum and maximum durations and doses for each treatment episode recorded current adherence assessed using at least two sources and meeting at least 80% prescribed doses taken, with antipsychotic plasma levels monitored on at least one occasion.
Optimally, they advise prospective evaluation of treatment, a prospective trial of an antipsychotic demonstrating less than 20% symptom reduction over six weeks, and a long acting injectable antipsychotic trialled for at least four months, and antipsychotic serum levels measured at least twice two weeks apart.
Two key modifications since Kane et al., (1988) are: the shift away from a focus on positive psychotic symptoms and onto negative and cognitive symptoms (Suzuki et al., 2012), as research has since demonstrated that these can be the most persistent and disabling symptoms for treatment- resistant patients, contributing significantly to outcome (Conley & Kelly, 2001); and the gradual decrease in the duration of illness necessary for a diagnosis of treatment-resistance. From the original Kane criteria specifying five years with persistent symptoms, Brenner et al., (1990) considered two years sufficient, Peuskens (1999) accepted just one year as adequate, and now TRRIP specify only a minimum of 12 weeks to allow for two six-week antipsychotic trials. Concurrently, the number of failed drug trials necessary has also reduced from three to two. This push for earlier diagnosis of treatment-resistant schizophrenia is due to the combination of a) evidence indicating - 15 -
Introduction that patients who don’t respond to two adequate trials have less than 7% chance of responding to another (Kinon et al., 1993) and b) evidence indicating that delays in diagnosis of resistance and appropriate treatment lead to worse outcomes for treatment-resistant patients (Üçok et al., 2015).
Also note that the exact choice of standardised rating scale and specific cut-offs for determining e.g. “minimal improvement” are not specified. This is the area in which definitions of treatment- resistance vary most significantly. For example, many use the >20% reduction in symptoms used by Kane but some studies require reductions in symptom scores of at least 40% (Breier et al., 1994) or 50% (Farooq, Agid, Foussias, & Remington, 2013). Additionally, some long-term studies use broader ranges of outcome criteria than solely symptom reduction and this leads to some of the findings that poor response is the norm (Hegarty et al., 1994; Helgason, 1990; Meltzer, 1997). If the norm in psychiatric care and research was to aim for remission, which in non-psychiatric disorders is characterised as the disappearance of symptoms, and definite treatment-resistance as patients who do not achieve remission this would set a very different standard. Within psychiatry, remission is often defined not as the complete absence of symptoms, but by minimal symptoms (Doyle & Pollack, 2003); and within schizophrenia, even this definition of remission is rarely used to discuss treatment efficacy, instead assessing reductions in symptoms of 20-50% as discussed above. Meltzer (1990) proposed defining treatment-resistance as the failure of patients to return to their best premorbid level of functioning; an intuitive definition based on the aim of remission. However, the problem with this approach is that this unfortunately applies to the significant majority of patients with schizophrenia and thus does not provide any differentiation for guiding clinical decision making or prioritising those most in need.
Meltzer (1997) has proposed that “it may be more useful to evaluate treatment responsiveness as a continuum and include all relevant outcome criteria” (p. 3). Others have also challenged the dichotomous concept, with Brenner et al., (1990) proposing a continuum of response, with seven levels from full clinical remission to severely refractory. However, while this may capture clinical reality in a more nuanced way, dichotomous definitions allow for more powerful study design in research, clearly highlights a group with significant need (Peuskens, 1999), and are used by most research groups and therefore allow comparisons throughout the literature.
Ultimately, while there is no universal agreement on a complete set of criteria, there is broad consensus that consistent minimum requirements for a diagnosis of treatment-resistance include at least two periods of treatment with antipsychotics at doses equivalent to or greater than 600mg chlorpromazine, each for at least 6 weeks, without at least 20% reduction in symptoms; this is also
- 16 -
Introduction the basis for clinical guidelines diagnosing treatment-resistance and prescribing appropriate treatment (IPAP, 2004; Lehman et al., 2004; NICE, 2009).
Within treatment-resistant patients, there are further subdivisions. Treatment-resistant patients can be divided into those who are responsive to clozapine, and to those who are non-responsive to clozapine and are considered “ultra-treatment resistant” (Howes et al., 2016). Also, while a longitudinal cohort study of first-episode patients demonstrated that over 80% of patients with treatment-resistance appear to be resistant from first presentation, never showing response to antipsychotics (early onset treatment-resistant schizophrenia), a minority of patients appear to develop resistance on average five years after first presentation (delayed onset treatment-resistant schizophrenia) (Demjaha et al., 2017). These differing courses of illness and treatment response may reflect meaningful differences between patients.
1.2.2 Significance of treatment-resistant schizophrenia
Regardless of discrepancies in prevalence figures, it is undeniable that treatment-resistant schizophrenia is a severe problem; patients are amongst the most severely distressed and disabled of all mental health patients with high levels of suicidal ideation, a lower quality of life and increased comorbid substance abuse problems and subsequently require higher levels of health and social care (Kennedy, Altar, Taylor, Degtiar, & Hornberger, 2014). Fortunately, an effective pharmacological treatment does exist for these patients – clozapine – but, unfortunately, numerous issues remain.
1.3 Clozapine
After the initial discovery of chlorpromazine and the subsequent discovery of patients who did not respond, there was continued development of antipsychotic drugs, notably the development of clozapine in the 1960s (Hippius, 1989). Clozapine is the first atypical antipsychotic – an antipsychotic with lower antagonism of D2 dopamine receptors and relatively greater antagonism of D3 and 5- HT2A serotonin receptors - and was found to be a highly effective antipsychotic. Initial reports were positive, as patients experienced few of the extrapyramidal adverse reactions that typical antipsychotics (such as chlorpromazine and haloperidol) induced; but unfortunately, a number of patients in Finland died from agranulocytosis – severe acute lowering of white blood cell count - and the drug was withdrawn from the market for a few decades. Throughout this time, the group of patients failing to respond to the antipsychotics on offer remained untreated, with very few options.
Then, in the late 80s, a seminal multi-centre trial was carried out (Kane et al., 1988). Two- hundred-and-sixty-eight patients with treatment-resistant schizophrenia were recruited for a trial comparing clozapine to chlorpromazine, using the stringent criteria for treatment-resistance
- 17 -
Introduction outlined above. Patients with confirmed treatment-resistance were then randomly assigned to a six- week double-blind treatment trial with either clozapine or chlorpromazine. Their results showed that clozapine led to superior results across psychotic symptoms: within six weeks 30% of those on clozapine responded to treatment compared to 4% of those on chlorpromazine. In their conclusion, they stated:
We believe that the apparently increased comparative risk of agranulocytosis requires that the use of clozapine be limited to selected treatment-resistant patients for whom the potential benefits are judged to outweigh the risks…
Careful, regular monitoring of blood cell counts is necessary in patients receiving clozapine…
For individuals suffering from treatment-resistant schizophrenia, the availability of clozapine, a potentially helpful treatment, is, in our view, a useful therapeutic advance. If even a small proportion of these patients can go on to adjust to life in the community, with the associated reduced need for long-term institutionalization, this has significance for public health and health financing.
Clozapine was therefore brought back onto the market, but this time with strict regulations following recommendations by Kane et al. In most countries, its prescription is confined to patients with demonstrated treatment-resistance and there is regular monitoring of blood cell counts. Clozapine’s superior efficacy has been demonstrated again and again (Azorin et al., 2001; Green et al., 2000; Lieberman et al., 1994; McEvoy et al., 2006; Meltzer et al., 1989; Solanki, Singh, & Swami, 2007), with hospital admission studies (Kirwan, O’Connor, Sharma, & McDonald, 2017), studies of mortality and self-harm (Wimberley et al., 2017), and meta-analyses confirming this (Stefan Leucht et al., 2013; Siskind, Mccartney, Goldschlager, & Kisely, 2016); it is the most effective antipsychotic available, and its ability to reduce clinical symptoms is especially pronounced in those resistant to conventional treatment. A relatively recent meta-analysis contradicted this body of evidence, concluding there was no real difference in efficacy between clozapine and other antipsychotics (Samara et al., 2016), but Taylor (2017) convincingly refutes this conclusion. Firstly, this meta- analysis was diluted by trials which seem to have recruited non-refractory patients. Secondly, this meta-analysis included several studies using relatively low mean doses of clozapine, particularly in trials comparing clozapine to a second generation antipsychotic (with a mean dose of 392 mg per day across these trials). Both of these methodological factors seem to be driven by drug companies seeking to prove comparable effects to clozapine, and inclusion of these studies could have biased this meta-analysis to not find strong evidence for superiority of clozapine to other antipsychotics. - 18 -
Introduction
Meta-analyses such as Siskind et al., (2016) which were limited to studies with strict confirmation of true treatment-resistance (as the original Kane study did) show the strongest superiority of clozapine, and Siskind et al also found that limiting analysis to studies with equivalent doses of clozapine and comparator drug resulted in clear superiority of clozapine. Taylor states “the effect of clozapine in refractory psychosis is one of the few things in psychiatry that is clearly visible to the naked eye”.
However, patients can wait years after treatment-resistance is established to be prescribed clozapine. This is in part due to its notorious history and the perception of clozapine as dangerous. There is also reluctance due to the range of potential adverse reactions to clozapine; weight gain, hyper-salivation and sedation are all common, and – while not fatal in the same way as agranulocytosis – they all have significant impact on quality of life and can be highly distressing. Another problem is that the monitoring of cell counts has a very rigid timetable, requiring frequent blood collection; in the UK, for the first eighteen weeks patients must have blood drawn at least every week, then every two weeks until one year has passed, and then at least every four weeks indefinitely (“Joint Formulary Committee,” 2017). This can be disruptive to patients’ lives, and, combined with the potential adverse reactions, many patients do not wish to be prescribed clozapine. As a result of these combined factors, numerous papers have noted that the drug is heavily underutilised with clinicians appearing reluctant to prescribe (Mistry & Osborn, 2011; Üçok et al., 2015). This is particularly concerning when delays in initiation appear to be associated with polypharmacy (Grover, Hazari, Chakrabarti, & Avasthi, 2015; Thompson et al., 2016; Üçok et al., 2015) – which carries various risks – and with worse response to clozapine after initiation (Üçok et al., 2015). Moreover, once clozapine is prescribed, there is no guarantee that patients will respond. While clozapine is currently the best option available for treatment-resistant schizophrenia, a significant number of patients still do not experience reduction in symptoms (Cipriani, Boso, & Barbui, 2009). These “ultra-treatment resistant” patients currently have very limited options available to them. Therefore, while clozapine has provided an effective treatment for many patients with treatment-resistant schizophrenia, there are numerous problems regarding its use in clinical practice.
Our understanding of clozapine’s mechanism of action and the physiological systems involved is also limited, as is our understanding of what differs in patients who have different responses to clozapine. Kane et al., (1988) states:
“Given these findings, there is an obvious need for further research to explore the mechanisms by which clozapine accomplishes its clinical effects and
- 19 -
Introduction
to identify possible predictors that might help to select, if possible, the subgroup of patients most likely to benefit.”
Despite ongoing research over the last three decades since this paper, the specific mechanisms which underlie clozapine’s unique therapeutic efficacy and reliable predictors of appropriate patients have yet to be established. This creates barriers for both optimising clozapine treatment and developing new treatments. Improving our understanding of clozapine could help us identify novel targets for pharmacological interventions which more directly and selectively address the primary dysfunctions in treatment-resistant schizophrenia. Understanding the mechanism of clozapine and clozapine response could also give us insight into what differentiates between treatment-responsive patients (who don’t require clozapine), treatment-resistant patients (who only respond to clozapine) and ultra-treatment resistant patients (who do not respond to standard antipsychotics or clozapine); this distinction could be used to identify patient groups much earlier on in their psychiatric treatment, enabling prompt prescription of appropriate interventions.
My thesis extends the literature addressing clozapine’s mechanism and differing outcomes in treatment-resistant schizophrenia in a novel way: through analysing data on longitudinal trajectories of DNA methylation in patients following clozapine initiation. It uses this data to explore three broad questions:
What changes in DNA methylation are associated with exposure to clozapine? What changes in DNA methylation are associated with therapeutic response to clozapine? What changes in DNA methylation are associated with adverse reactions to clozapine?
Chapter two will outline the methodology of the study, including design, data collection and laboratory procedure for measuring DNA methylation. Chapters three, four and five will then address the above three questions in turn, each presenting a focused background, an outline of the specific statistical analysis used to address the question, the results of this analysis, and then a discussion. Chapter six will then summarise and discuss these results as a whole.
Before moving on to chapter two, the rest of the introduction will introduce the concepts of genetic and epigenetic variation and explain why I have chosen to study the role of DNA methylation in clozapine’s mechanism of action.
- 20 -
Introduction
1.4 What is DNA methylation and why study its role in clozapine?
The next section will introduce DNA methylation and its potential utility in furthering our understanding of clozapine – starting with a brief overview of genetic biology and its importance in schizophrenia and antipsychotic response, before moving onto epigenetics and DNA methylation.
1.4.1 The genome and gene expression
The genome is housed within deoxyribonucleic acid (DNA). DNA is formed of nucleotides which are made up of a nitrogenous base, a five carbon sugar, and a phosphate group. The nucleotides have one of four bases: cytosine, thymine, guanine and adenine. Each base is connected to another base via hydrogen bonds, and the phosphate groups bond with one of the carbons on the sugar, and form the two “sugar phosphate backbones” to DNA. As the hydrogen bonds connecting the nucleotides are not covalent, they can be broken and re-joined relatively easy – this is essential for gene expression, replication of DNA, and DNA sequencing. DNA wraps around a cluster of eight histone proteins, called a histone core, to form nucleosomes (Figure 1). Nucleosomes create the structure of chromatin, which chromosomes are made of. Inside the nucleus of each cell, there are 46 chromosomes. These chromosomes contain our entire DNA (Annunziatio, 2009).
Genes are distinct sequences of nucleotides which provide code for protein formation or another product such as RNA, with segments called exons (coding regions that are translated into mRNA) and introns (non-coding regions, which may be regulatory). For genes that are polymorphic – the nucleotide bases can vary - each version of the gene is called an allele and allelic variation can have functional consequences. When one nucleotide in a gene is known to vary, this is called a single nucleotide polymorphism (SNP) (Bush & Moore, 2012).
To study a gene, determining the nucleotides within a relevant section of DNA is necessary. The specific method of genotyping depends on the aim, methodological priorities and the resources available – for example, whether you want to assess a small number of specific genes based on an a priori hypothesis or whether you want to assess a huge number of genes across the genome. Genome sequencing involves determining the exact sequence of nucleotide bases within a gene, and can be done for individual genes or the whole genome, but the latter is incredibly costly. A more cost-efficient method for getting broad coverage across the genome, and a very popular method of gene-association research over the last decade, is genome-wide association studies (GWAS) (Bush & Moore, 2012). GWAS uses SNP arrays which have probes to identify a selection of common SNPs – loci where a SNP variant is present in at least 1% of the population – which function as tag SNPs. Tag SNPs can predict other SNPs in high linkage disequilibrium i.e. where there is non-random association between alleles of a SNP, and so possessing a specific allele of one SNP predicts the - 21 -
Introduction possession of a specific allele for another SNP. This allows for efficient use of resources; however, it means that any statistical associations with a SNP found through gene arrays may actually reflect a causal association with another SNP in the same linkage disequilibrium block.
Figure 1: Image of DNA and chromosome structure, taken from Annunziato (2008)
Gene expression is the process in which information from a gene is used to synthesise either a protein or another gene-product (Clancy & Brown, 2008). Ribonucleic acid (RNA) – generally a single strand of nucleotides, with uracil replacing thymine – is involved in the regulation and process of gene expression and protein synthesis. Gene expression occurs through transcription and translation. Transcription occurs in the nucleus and is triggered by an enzyme, RNA polymerase. This enzyme attaches to a region of the DNA – called the promoter region – which is upstream of the transcription start site (TSS) of the gene being expressed. Proteins called transcription factors bind - 22 -
Introduction to promoter regions and facilitate RNA polymerase binding. Once attached, RNA polymerase separates the two strands of DNA, and then reads the sequence, adding matching RNA nucleotides to the complementary nucleotides of one DNA strand. This produces a complementary antiparallel RNA strand called a primary transcript. This is modified into mRNA, which carries information from the DNA to the ribosome and serves as a template for protein synthesis through translation. Translation occurs in cytoplasm, when ribosome – ribosomal RNA and proteins – assemble around the mRNA. A small ribosomal subunit binds to the mRNA and reads the sequence, then ribosome induces the binding of complementary transferRNA to mRNA – these select, collect and carry amino acids to ribosome. The ribosome translocates – moves – along the mRNA creating an amino acid chain, until the polypeptide is released and folds to produce a specific protein (Figure 2).
Figure 2: Image of gene expression taken from Clancy and Brown (2008)
- 23 -
Introduction
1.4.2 Genetic variation and schizophrenia
Through twin and family studies, schizophrenia has been well-established as a highly heritable disorder with heritability estimates of approximately 80% (Cardno et al., 1999); this can be interpreted to mean that up to 80% of the phenotypic variation (variation in diagnosis of schizophrenia) in the population is attributable to genetic variation. Therefore, significant research resources have been invested in determining the location of genetic variation associated with schizophrenia, in the hope that this will provide insight into the biological pathways underlying the aetiology and course of schizophrenia.
Early studies focused on multi-generational families with numerous occurrences of schizophrenia, using a linkage study design – this was based on the assumption that there would be a small number of relevant genes of high penetrance (the higher the penetrance of a gene, the higher the likelihood that an individual with that gene will develop the disorder) (Riley, 2004). When this didn’t yield results, the field shifted towards the assumption that the genetic architecture of schizophrenia involves a large number of common variants of low penetrance that, in combination, increase risk (Becker, 2004). Common variants are those which are present in more than 1% of the population. Currently, to identify these common variants, GWAS using gene arrays (as described earlier) are dominant. In GWAS, large numbers of patients with schizophrenia and controls are recruited, blood samples are collected and DNA extracted to measure hundreds of thousands of SNPs across the genome (“genome wide”) via the probes on the arrays. By comparing the SNP variants of patients with the variants of controls, SNPs with allelic variation that are associated with having schizophrenia can be identified.
Well-powered, carefully conducted GWAS have recently proven to be more fruitful, with a ground-breaking study by the Psychiatric Genetics Consortium (PGC) in over 35,000 patients with schizophrenia identifying over 100 genetic regions associated with schizophrenia, 87 of these novel, which were collectively able to account for 33% of genetic liability (Ripke et al., 2014). 75% of these regions include protein-coding genes (35% of the regions include multiple associated genes), including genes associated with dopaminergic systems, glutamatergic neurotransmission, synaptic plasticity and voltage-gated calcium channel subunits – these are all biological systems which have previously been implicated in schizophrenia and which studies of rare genetic variants have also implicated. One of the strongest risk associations was in the major histocompatibility complex (MHC), which is located on chromosome 6 and contains genes involved in innate and acquired immunity. In particular, a recent study linked different genetic combinations of copy numbers at a locus near gene C4 and HERV status with increased RNA-expression data and increased risk of
- 24 -
Introduction schizophrenia (Sekar et al., 2016); this has been further linked to increased synaptic pruning during late adolescence, long postulated as a neurodevelopmental mechanism underlying schizophrenia. Even excluding the MHC region, Ripke et al., (2014) found enrichment for associations at enhancers active in tissues with immune functions.
As well as investigating associations with schizophrenia diagnosis, genetic studies have also studied other facets of schizophrenia including treatment-resistance and response to anti- psychotics.
1.4.2.1 Genetics of treatment-resistant schizophrenia
Two GWAS of treatment-resistant schizophrenia have been conducted. Both implicated genetic variants associated with the immune system: one compared treatment-resistant patients to healthy volunteers and found an association with polymorphisms of SLAMF1, a gene associated with lymphocyte activation (Liou et al., 2012), while another studied response to a single antipsychotic trial and found an association with polymorphisms of RTKN2, a gene believed to be involved in lymphopoiesis (Drago et al., 2014).
Outside of GWAS, there have been candidate-gene association studies. These follow the same basic design as GWAS, comparing patients to controls to identify genetic variation associated with schizophrenia, but in contrast to GWAS only compare genetic variation at pre-identified loci (candidate genes) instead of across the whole genome. Candidate gene studies are plagued by weak biological justification for the candidate gene, small underpowered samples, lack of consideration for underlying population structure, inadequate adjustment for explicit multiple-comparisons, inadequately conservative significance thresholds considering the high likelihood of type I errors, and subsequent lack of replication for reports of significant associations.
The strongest finding from candidate gene studies is an association between (proxies of) treatment-resistance with the minor allele of the Val66Met polymorphism of the BDNF gene. One study reporting robust associations when comparing clozapine users to non-clozapine users (Zhang, Lencz, et al., 2013) avoids some of the issues mentioned above, selecting candidates based on the CATIE trial (described below) and controlling for explicit multiple-comparisons. There is also some form of replication, with another study reporting an association with lack of response to a single antipsychotic trial (Zai et al., 2012). However, these are small samples by current standards, continue to have a high likelihood of type I error and are investigating potentially distinct phenotypes.
- 25 -
Introduction
During my thesis I conducted a systematic review (Gillespie, Samanaite, Mill, Egerton, & MacCabe, 2017) (Appendix I) of all studies comparing patients with treatment-resistant schizophrenia (defined as either a lack of response to two antipsychotic trials or clozapine prescription) and treatment-responsive schizophrenia (defined as known response to non-clozapine antipsychotics). There were no GWAS meeting this criteria and while studies implicated an association between treatment-resistance and numerous variants (the short allele of BDNF dinucleotide repeat polymorphism, C/C genotype of the serotonin 2A receptor (5-HT2A) gene T102C polymorphism, C/A genotype for the tryptophan hydroxylase enzyme (TPH1) gene, ccG10 alleles of the 5’UTR (ccG repeat) polymorphism of the reelin gene, and less homozygosity with the Bal I polymorphism of the dopamine receptor 3 (DRD3) gene), these were all findings from studies with small samples and “exploratory” analyses which did not report surviving correction for multiple testing, and none of these findings were replicated.
Another finding of my systematic review was that studies reported first and second degree relatives of treatment-resistant patients had a significantly higher morbidity risk of schizophrenia spectrum disorders compared to relatives of treatment-responsive patients, and a significantly higher familial-loading score was found for treatment-resistant patients. Similar findings are evident in the broader literature, as an earlier study investigating response to a single antipsychotic trial similarly found that first-degree relatives of non-responders to haloperidol had higher lifetime risk for schizophrenia spectrum disorders than relatives of responders (Silverman et al., 1987). More recent studies have found associations between a history of clozapine treatment and a higher polygenic risk score (Frank et al., 2015), and between treatment-resistance and higher genome-wide burden of rare duplications (Martin & Mowry, 2016). These cumulatively suggest that treatment- resistant schizophrenia may be a more heritable form of schizophrenia.
1.4.2.2 Genetics of anti-psychotic response
Numerous studies have investigated genetic predictors of anti-psychotic response; in chapter 4 I will specifically review the literature on genetic predictors of clozapine response but one of the largest studies spanning multiple antipsychotics is the CATIE (clinical antipsychotic trails of intervention effectiveness) study. A GWAS was conducted on data from 738 patients with schizophrenia (not resistant or first episode) prescribed either olanzapine, quetiapine, risperidone, ziprasidone or perphenazine, to identify associations between SNPs and PANSS scores (McClay et al., 2011). One overall conclusion was that there were no associations common to all antipsychotics, which coincides with the clinical observation that a good clinical response to one antipsychotic doesn’t predict the same response to another. Significant associations were found for specific
- 26 -
Introduction antipsychotics and subscales, for example negative symptom response to olanzapine was associated with ANKS1B, with the top SNPs explaining as much as 10-15% of the variance in antipsychotic response.
Candidate gene testing using this data-set was conducted by McClay et al., as well as by Xu et al., (2016), and Need et al., (2009), selecting different candidate genes and different measures of responses and reporting significant associations. However, we should remain cautious about the high rate of false positives in candidate-gene studies and require replication studies, particularly if findings are to be used for selecting and dosing medication. A review by Zhang and Malhotra, (2011) notes that none of the candidate genes indicated by previous literature (DRD2 -141C Ins/Del, DRD3 Ser9Gly, HTR2A -1438G/A , 5HTTLPR, COMT Val108Met) reached genome wide significance in the CATIE GWAS, highlighting the need for scepticism, and as Reynolds (2012) states there is a “striking paucity of consistently reproducible effects”.
1.4.2.3 Limitations of genetics research
The findings from the PGC have restored confidence that genetic research will prove a fruitful avenue in understanding schizophrenia. However, despite decades of genetics research for schizophrenia increasing in quality and power with each year, the fact remains that there have been no new successful treatments developed in decades, and key questions of aetiology remain. There are, of course, non-genetic and non-biological factors of significance to schizophrenia and schizophrenia outcomes e.g. (Morgan & Fisher, 2006) but there remains a fascinating discrepancy between the solid evidence for a genetic basis to schizophrenia and the lack of deep understanding provided by the research findings so far.
The genetic architecture of schizophrenia is evidently complex as, despite high heritability estimates, over half of patients with schizophrenia have no first or second degree relative with schizophrenia (Gottesman & Shields, 1982). It could be that our underlying assumptions – for example, the focus on common variants of low effect driving GWAS – are faulty. It could be that gene-gene interactions, gene-environment interactions or the regulatory influence of genes are more significant than previously realised. Relevant to the latter suggestion, a common finding from GWAS is that a high proportion of the variants associated with schizophrenia are involved in transcription regulation e.g. (Ripke et al., 2013). This suggests that we need a broader understanding of our genome, that doesn’t focus on DNA sequence in isolation. One piece of the puzzle is our epigenome.
- 27 -
Introduction
1.4.3 The epigenome
Epigenetic variation refers to “mechanisms that control chromatin remodelling and the accessibility of genes to transcriptional machinery” (Ibi & González-Maeso, 2015) or “changes in the physical structure of the chromatin, without a change in the DNA sequence itself” (Teroganova, Girshkin, Suter, & Green, 2016). The epigenome provides information regulating gene expression, providing the mechanism for genetic information to lead to phenotype. The existence of epigenetic mechanisms is demonstrable in the ability of somatic cells containing identical DNA sequences to vary widely in phenotype and function, through cellular differentiation guided by tissue-specific epigenetic variation (Li, 2002); the tens of thousands of genes in the human genome must be expressed at specific times in specific tissues to lead to specific protein formation. Epigenetic variation can effectively activate or silence genes, through their influence on processes involved in gene expression.
There are two primary mechanisms of epigenetic variation. The first is histone modification (Bannister & Kouzarides, 2011). The histones which DNA wraps around have tails which insert themselves into minor groves of the DNA and are accessible for chemical modification. These histone modifications can change the structure of chromatin. For example, acetylation loosens chromatin structure by creating negative charges which repel the negatively charged DNA phosphate group and vice versa for deacetylation. When loosely packed, chromatin becomes accessible for transcription factors and RNA polymerases (“euchromatin”), allowing gene expression; in contrast, when tightly packed it becomes less accessible (“heterochromatin”), limiting gene expression. Histone deacetylases (HDACs), histone acetylases and histone methyltransferases all regulate histone modifications. Inhibition of histone deacetylase activity seems to be the primary trigger for histone hyperacetylation, which happens in conjunction with DNA demethylation to increase transcriptional activity.
DNA methylation is the second form of epigenetic mediation (Jin, Li, & Robertson, 2011). This involves a methyl group (one carbon bound to three hydrogens) covalently binding to the 5’ cytosine within CpG (cytosine-phosphate-guanine) dinucleotide pairs in DNA sequences; this binding is catalysed by DNA methyltransferases (DNMTs), a group of enzymes. DNMTs establish and maintain DNA methylation patterns, MeCP2 is a methyl-CpG binding protein that recognises methylation patterns and mediates the process, and ten-eleven translocation (TET) proteins convert 5- methylycytosine into 5-hydroxymethylcytosine and cause subsequent demethylation (Schübeler, 2015). CpG sites can cluster together in regulatory areas of the gene, such as promoter regions, where transcription is initiated; these regions are called “CpG islands”. While most CpG sites are
- 28 -
Introduction methylated throughout the genome, CpG islands tend to have a high number of unmethylated CpG sites. Methylation of CpG sites appears to disrupt the binding of transcription factors and the subsequent process of gene expression, usually leading to reduced gene expression (Klose & Bird, 2006). Two mechanisms have been proposed: the “critical site” model which places emphasis on the reduced transcription of mRNA due to methylation at transcription-factor binding sites of specific cytosines, while the “methylation density” model places emphasis on the transcriptional potential of a gene due to a proportion of methylated cytosines across a region controlling chromatin structure.
However, while the majority of associations between methylation and gene expression are negative (i.e. higher levels of methylation are associated with lower levels of gene expression), there are a reasonable proportion of CpG sites which have a positive association between methylation and gene expression, particularly sites with trans (distant) effects (van Eijk et al., 2012). This has been supported by others study such as one which reports that almost half of genes with high levels of methylation at promoter regions show high expression rates, especially when these promoter regions are outside of CpG islands (Sarda, Das, Vinson, & Hannenhalli, 2017). van Eijk et al., (2012) also reports two other things of note: that only a fifth of the associations between methylation and expression identified in their study involved probes from the same gene, and that causal modelling indicated that in a minority of these associations expression levels regulated methylation rather than vice versa.
As well as directing gene expression, DNA methylation also controls and mediates genomic imprinting (information on whether the allele came from the mother or father, allowing only one- parent’s allele to be expressed due to hypermethylation of the other alleles CpG islands), chromosomal stability and genetic recombination (Robertson, 2005).
A key feature of the epigenome is the ability for dynamic rapid change and reversal, making it distinctly different from the genome and the nucleotide bases which make up somatic DNA. DNA methylation patterns undergo major changes during gametogenesis and embryogenesis (when all epigenetic markings are removed and then rewritten, with the likelihood of de novo epimutations one or two orders of magnitude greater than that for somatic DNA mutations), development, and throughout aging; they change in response to the cell environment; they change in response to environmental exposure, providing a physical mechanism for gene-environment interaction. A study in rats sparked considerable interest in the importance of epigenetics as a dynamic interaction between environment exposure and cellular biology when it reported that maternal nursing behaviour altered the epigenome of their offspring at a glucocorticoid receptor and subsequent HPA
- 29 -
Introduction response to stress, and that this persisted into adulthood unless reversed by cross-fostering (Weaver et al., 2004).
The dynamic nature of the epigenome can also be demonstrated by the reduction in concordance of DNA methylation and histone acetylation in monozygotic twins over their lifetime from age three to age fifty (Fraga et al., 2005). The dynamic nature of the epigenome provides plasticity in development and adaption to environment, with an obvious evolutionary function to increase chances of survival in changing environments.
It is important to note that the genome informs the epigenome, and – while perhaps less influential later in the life course (Gaunt et al., 2016)- there are genes which affect CpG methylation called methylation quantitative trait loci (mQTL) which account for approximately 20% of variance in DNA methylation (Gaunt et al., 2016; McRae et al., 2014). Therefore, the genome and epigenome should not be considered in isolation, but viewed as complementary systems which interact bi- directionally.
1.4.3.1 DNA Methylation Measurement
DNA methylation is one of the most studied epigenetic modifications (Portela & Esteller, 2010), in part because of the development of technologies that can efficiently perform epigenome-wide association studies (EWAS) in a similar way to genome-wide studies. The most common method for epigenome-wide analysis of DNA methylation is to analyse sodium-bisulfite treated DNA. Sodium- bisulfite treatment converts unmethylated cytosines to uracil, while methylated cytosine remains unchanged. The sites where the conversion to uracil have happened indicate that the sites are unmethylated. The sodium-bisulfite method allows an epigenome-wide approach in a way that previous methods such as southern blot hybridisation did not, because they relied on methylation- sensitive restriction enzymes which could only provide information on specific sites.
The most commonly used method is to use Illumina micro-arrays such as the Illumina 450k. After DNA has been bisulfite-treated, the DNA is amplified (creating thousands to millions of copies of the piece of DNA) and then applied to the beads on the micro-array, with each bead specific to a CpG site. Reversible dye-terminators are then applied to the micro-array, with each nucleotide base fluorescently labelled with a different colour. A laser is then used to excite the dyes and a photograph is taken from which the degree of methylation at each locus can be identified.
As with GWAS, the results for cases vs controls at each probe can be compared to identify positions with different levels of methylation associated with case status – these are called differentially methylated positions (DMPs).
- 30 -
Introduction
1.4.4 Epigenetic variation and schizophrenia
Numerous observations have indicated that epigenetic processes are relevant to understanding schizophrenia. The existence of discordant monozygotic twins, the fluctuating course of schizophrenia with remission and relapse, and parent of origin effects all indicate that inherited genes cannot provide all of the answers, but epigenetic differences between individuals and over time may be able to explain these phenomena. Similarly, epigenetics may be able to explain the cases of schizophrenia without first or second degree relatives, with epimutations occurring or being corrected during the reprogramming that occurs in gametogenesis and driving differences between parent and child. Information about methylation or other epigenetic modifications may also be able to refine the signal detected in GWAS; a relevant gene may only be associated with schizophrenia if methylated or unmethylated, so without knowledge of methylation levels there will be unmeasured noise reducing the association between a gene and schizophrenia. Understanding epigenetic processes involved in schizophrenia may also identify biological pathways underlying schizophrenic symptoms.
Specific evidence for the importance of methylation comes from a range of sources, as one review highlights (Maric & Svrakic, 2012). For example, increasing methylation levels via a high methionine diet worsens psychotic symptoms (Brune & Himwich, 1962), and schizophrenia patients have high plasma levels of homocysteine (demethylated methionine) during acute episodes (Petronijević et al., 2008) and increased levels of S-Adenosyl methionine (SAM) – a methyl group donor – in prefrontal cortex (Guidotti et al., 2007). Interestingly, research into discordant monozygotic twins has found that while they do not differ in underlying DNA sequence they show striking differences in methylation at DRD2, to the extent that an affected twin is more epigenetically similar to a non-related affected individual than their own unaffected twin (Mill et al., 2006). Similar research conducting an EWAS analysis of discordant monozygotic twins indicated no systemic changes in genome-wide epigenetic programming, but individual CpG sites showed considerable variability (Dempster et al., 2011)– the top eight were located by genes PUS3, SYNGR2, KDELR1, PDK3, PPARGC1A, ACADL, FLJ90650 and TUBB6. Pathway analysis revealed significant enrichment for the “nervous system development and function” functional network, the “psychological disorders” biological functional pathway, and the dopamine and glutamate receptor signally canonical pathways; all aetiologically relevant gene networks for schizophrenia. Additionally, post-mortem research (Mill et al., 2008) has found aberrant global hypermethylation and ~100 loci with disease-associated aberrant DNA methylation in the frontal cortex of schizophrenia patients;
- 31 -
Introduction these findings similarly implicate aetiologically relevant pathways including glutamatergic, GABAergic, stress response and neurodevelopmental pathways. More generally, there is evidence that genetic risk variants for schizophrenia may cause altered DNA methylation and subsequent disrupted gene regulation: fetal mQTLs have been shown to be enriched in schizophrenia risk loci identified through the PGC GWAS; analysis combining GWAS and EWAS data from a sample of schizophrenia patients found differential methylated CpG sites overlapped with schizophrenia risk loci from GWAS and these risk loci co-localise with mQTLs; and mQTLs in which the relevant CpG site is associated with differential gene expression are enriched for schizophrenic risk loci (Hannon et al., 2016; Hannon, Spiers, et al., 2015; van Eijk et al., 2015)
In 2016, Teroganova et al., conducted a systematic review of all studies measuring DNA methylation in schizophrenia using peripheral blood and saliva; there were 22 studies included, which ranged considerably in sample size and methodology, and which had inconsistent reporting of relevant confounds. Nine of these were candidate gene studies, with the most studied genes RELN, BDNF, COMT and HTR1A. 12/13 of the EWAS reported differential methylation associated with schizophrenia and the majority of differentially methylated regions showed hypermethylation (or did not specify direction of effect). The results implicated the glutamate, GABA and dopamine system, calcium signalling, immune function and inflammation, potassium voltage-gated channels, and monoamine systems, among others.
This research supports epigenetic models of schizophrenia, which have been discussed for well over a decade. Petronis, Paterson, and Kennedy (1999) demonstrated that an epigenetic model could integrate data supporting multiple hypotheses within schizophrenia (including neurodevelopmental, dopamine, and viral) and suggested that epigenetic alterations could mediate the genetic, developmental and environmental influences on risk for schizophrenia. A mouse model of adolescent exposure to isolation stress (Niwa et al., 2013) demonstrates how the interaction between a genetic risk factor (DISC-1) and environmental risk factor (isolation stress) can lead to hypermethylation of a gene which occurs alongside disruption of the mesocortical projection of dopaminergic neurons and behavioural changes indicative of a schizophrenia phenotype.
Epigenetic theories of psychosis have focused on epigenetic modifications associated with GABAergic and glutamatergic gene promoters. One suggestion is that promoter hypermethylation of loci within a gene for a GABAergic neuron leads to downregulation of telencephalic GABAergic function, disinhibiting pyramidal neurons and driving a hyper dopaminergic state through excitatory input to dopamine cells in the ventral tegmental area, or a hyper serotonergic state through excitatory input to serotonin cells in the raphe nucleus (Guidotti, Auta, Chen, Davis, Dong, Gavin,
- 32 -
Introduction
Grayson, 2012). Teroganova et al., (2016) did identify a study showing hypermethylation of a locus within GABRB2 (GABA A receptor beta 2) in peripheral blood, supporting this hypothesis and supporting the ability of peripheral blood studies to identify differential methylation at positions indicated by theory.
1.4.4.1 Epigenetic modifications as a result of anti-psychotics
Interest in epigenetic theories of schizophrenia as well as pharmacological research demonstrating the epigenetic effects of numerous medications has led to research into epigenetic mechanisms of psychiatric medication including anti-psychotics, many of which are known to have epigenetic consequences (Petronis et al., 1999). Boks et al., (2012) summarised the existing literature on epigenetic alterations induced by psychopharmacology and concluded that a number of antidepressants and antipsychotics had “compelling evidence”; in particular, clozapine was singled out as the antipsychotic with the strongest evidence for epigenetic alterations based on a number of animal studies (discussed further below).
There have been few studies of treatment-induced DNA methylation changes in humans. Mill et al., (2008) found a strong correlation between DNA methylation in the MEK1 gene promoter region and lifetime antipsychotic use in schizophrenia patients, supporting the proposal that antipsychotics have epigenetic consequences and suggesting that one mechanism of antipsychotic action could be epigenetic modifications, perhaps (for example) normalising aberrant DNA methylation in schizophrenia. Bönsch et al., (2012) found that the methylation status of patients on antipsychotics was similar to healthy controls. Another study (Rukova et al., 2014) studied six patients with schizophrenia who were in remission, comparing their DNA methylation profile before treatment to healthy controls (n=220) and then their DNA methylation profile after treatment to healthy controls, with the aim of identifying both sites with abnormal methylation that became normalised and sites with normal methylation that became abnormal. Focusing on the three male patients, they found significant methylation differences, implicating genes involved in transcription regulation, sphingolipid metabolism, protein degradation, and a potassium channel protein. They also studied fourteen patients who had not experienced remission, and found their methylation profiles did not differ significantly from healthy controls. However, this if an incredibly small sample, it is unclear if they control for any relevant confounds other than matching on age and sex, and the medication used as treatment is unspecified beyond being atypical antipsychotics.
- 33 -
Introduction
1.4.4.2 Limitations of existing DNA methylation research
1.4.4.2.1 Study design
Use of cross-sectional designs is a significant issue with nearly all of the human psychiatric epigenetic literature. In genetic studies, many concerns about confounding and causation can be dismissed: environmental exposures typically cannot cause changes in genetic variation, and similarly later outcomes cannot cause changes in genetic variation. Therefore, cross-sectional designs are appropriate as the independent variable (genetic variation) remains static and has a clear causal direction. In contrast, epigenetic variation can be caused by environmental exposures and could be either the cause or the result of outcomes of interest; therefore cross-sectional designs raise numerous concerns about causality and confounding in psychiatric epigenetics (Relton & Smith, 2012).
In addition, many genetic studies recruit participants on a very large scale to increase power with minimal need to collect detailed information on participants or ensure participant groups are matched for e.g. drug use, due to the static unidirectional nature of genetic variation. However, in epigenetic studies these variables could be highly relevant if they are potential causes of epigenetic variation and therefore potential confounding variables. In psychiatric epigenetics, there are numerous environmental exposures associated with disorders such as schizophrenia (psychiatric medication use, rate of smoking, lifestyle, social environment, childhood experiences etc.) that pose potential issues of confounding. Similarly, within schizophrenia, different patient groups may have different symptomatology, medication histories, and lifestyles.
Another concern with cross-sectional epigenetic studies is that of genetic confounding. The fact that epigenetic variation is not independent from genetic variation means that a significant proportion of inter-individual epigenetic variation can be explained by genomic variation. Therefore, a certain proportion of variation in methylation can be considered heritable, with any group differences more directly attributable to the inherited genetic variants.
Unfortunately, these issues make many previous studies difficult to interpret as conclusions of epigenetic variation causing schizophrenia/clinical outcomes may actually reflect epigenetic variation caused by schizophrenia/clinical outcome, epigenetic variation as a result of a confounding variable, or epigenetic variation purely as a result of genetic variation. Therefore, reviews and comment pieces (e.g. Heijmans & Mill, 2012; Relton & Smith, 2012) have called for increased use of longitudinal cohort studies, with a sample collected before the event of interest (e.g. onset of illness, initiation of an antipsychotic) and serial samples collected subsequently.
- 34 -
Introduction
1.4.4.2.2 Selection of tissue
Another significant issue in psychiatric epigenetic research is selection of tissue. A key consideration in all epigenetic research is the tissue being studied, as different tissue types show significant epigenetic variation; Lister & Ecker (2009) report on some of the first genome-wide, single base-pair resolution maps of methylated cytosine which demonstrate that tissue types in samples from mammals have widespread differences. This becomes a problem when studying psychiatric disorders such as schizophrenia because the primary tissue of interest is the brain where the core dysfunction is assumed to be located, but accessing samples of brain tissue from live patients is not possible. Post-mortem research provides access to brain tissue however epigenetic variation changes considerably throughout the lifespan and is likely to change after death. Current research therefore relies predominantly on peripheral tissue, such as whole blood, as this is far easier to access; however, there remains heated debate about the extent to which epigenetic variation in blood cells (usually lymphocytes) approximates epigenetic variation in brain tissue (Relton & Davey Smith, 2012; Relton & Smith, 2010). Even if a true association with schizophrenia / clinical outcomes is detected, the epigenetic variation in peripheral tissue may be due to effects considerably downstream of epigenetic variation in the brain. Therefore, it is unclear how much insight into the aetiology and course of schizophrenia can be gained from the loci and pathways implicated by peripheral tissue studies.
Nonetheless, recent studies comparing methylation in brain tissue and blood tissue provide some optimism. Firstly, results from EWAS using post-mortem brain tissue and results from the schizophrenia EWAS using peripheral tissue have been compared (Dempster et al., 2011; Mill et al., 2008) and five of the 100 top ranked genes in the latter (GGN, SLC117A, SMUG1, SOX1 and TCF7L2) were also found to be significantly associated with psychosis in the post mortem study; this supports the utility of peripheral blood to identify epigenetic variation simultaneously occurring in the tissue of interest. Similarly, a study of psychosis found that 94% of CpG SNPs sites methylated in brain were also methylated in blood - significantly more than expected by chance (Van Den Oord et al., 2016). Secondly, a few studies have compared methylation in brain tissue and blood tissue from the same individuals to directly ascertain correlation across these tissues. Analysis for 122 participants and using brain tissue from four regions – prefrontal cortex, entorhinal cortex, superior temporal gyrus and cerebellum found that, while the majority of CpG sites did not show a significant correlation between methylation in blood and methylation in brain, a subset of probes did show a strong correlation across tissue (Hannon, Lunnon, Schalkwyk, & Mill, 2015). Moreover, the authors created a searchable-database of their data to identify whether CpG sites implicated in blood-based EWAS appear to show correlated methylation in tissue from each of the four brain regions. A similar study - 35 -
Introduction on twelve patients also found that a subset of CpG sites – 7.9% - showed a statistically significant, large correlation across blood and brain tissue (Walton et al., 2015). These studies propose a way forward: we can’t assume that methylation at all CpG sites will give us biological insight into psychiatric disorders when we use peripheral blood, but we can use findings from studies such as these to identify the CpG sites which will give us most insight into what is occurring in the brain.
Distinct from providing biological insight, peripheral tissue has the benefit of being a good choice for potential biomarkers. If epigenetic variation in blood is consistently associated with outcomes in schizophrenia and can provide robust predictive validity, the underlying biological mechanism is not necessary for it to have significant clinical utility. For example, in cancer research, a study demonstrated that leukocyte DNA obtained from peripheral blood (significantly less invasive than existing methods involving analysis of pancreatic juice and primary tumour tissue) had utility in diagnosing pancreatic adenocarcinoma, with five CpG sites acting as good predictive discriminators (Pedersen et al., 2011).
These two key limitations – study design and use of tissue - contribute to scepticism about the utility of psychiatric epigenetic research, with arguments that the findings from most published studies provide minimal information of value due to the difficulties in deciding whether they the results truly tell us about the questions we are interested in. However, neither of these limitations is insurmountable: longitudinal study designs with rich data on potential confounders and databases identifying sites of epigenetic variation that correlate between brain and blood tissue can be used. More broadly, numerous fields of science are becoming increasingly aware of the need to prioritise independent replication of significant findings, and psychiatric epigenetics is no exception; indeed, epigenetic research has caught the public imagination and become somewhat of a buzzword, warranting extra caution about bold conclusions on preliminary data. To negate this, we must be honest about the limitations of our methods, highlight the appropriate caveats in our conclusions, and demand both validation of methodology and replication of associations.
1.4.4.3 My thesis
As stated above, my thesis extends the literature addressing clozapine’s mechanism of action, both in its therapeutic and adverse effects, through analysing data on DNA methylation changes over the first six months of clozapine treatment for patients with treatment-resistant schizophrenia. I chose to study DNA methylation based on the evidence presented above (and evidence explored further in later chapters) suggesting that not only are epigenetic mechanisms involved in the aetiology of schizophrenia but epigenetic mechanisms may explain the mechanisms underlying
- 36 -
Introduction antipsychotics effects, in particular clozapine - above and beyond the role of genetic variation in schizophrenia and treatment outcomes. I chose a longitudinal approach to address many of the concerns raised about previous psychiatric epigenetic studies; using a within-participant design means that individual differences are controlled for, significantly increasing the statistical power and ease of interpretation. I also use previously collected data on blood-brain correlations in DNA methylation to aid interpretation of the results. The next chapter will now outline my methodology in greater detail.
- 37 -
Overall Method
2 Overall Method
2.1 Longitudinal study
Note: My data was collected as part of a larger, ongoing project investigating a range of physiological variables related to treatment-resistance and clozapine funded by both the MRC and EU initiative, CRESTAR. This included genome-wide assays, gene-expression levels, protein levels, and structural and functional neuroimaging data; however, I shall only describe the methodology relevant to the collection of epigenetic and clinical data, as this forms the basis of my thesis and my independent work. I recruited and assessed a number of the participants recruited to this project, but data is also included from participants recruited by the broader recruitment team.
2.1.1 Study design
The study used a longitudinal design, recruiting participants before they commenced clozapine and then assessing them at three follow up visits over six months of taking clozapine. At each time point, myself or another member of the recruitment team, collected clinical information as well as a blood sample for DNA methylation analysis. This allowed me to analyse epigenome-wide DNA methylation before and after commencing clozapine and over time, as well as the relationship between DNA methylation and clinical outcomes (changes in symptoms and adverse reactions). As stated above, a primarily within-participant design was chosen to eliminate many of the issues with cross-sectional epigenetic studies; participants effectively act as their own control, with genetic variation (including mQTLs), demographic characteristics, early environment, age of onset, underlying illness etc. remaining stable across all time points and other variables in the environment such as general lifestyle likely to have minimal variation over six months (relative to the amount of variation seen between people). This increases the likelihood that any identified changes in methylation are due to clozapine, instead of alternative explanations.
A baseline assessment (Visit A) was carried out to capture information on their DNA methylation profile and clinical state before clozapine commenced, typically in the week before clozapine initiation (median= 4.5 days, four patients had baseline assessment over a month before initiation). The second follow up (Visit B), completed six weeks after the date of clozapine initiation (not six weeks after the baseline assessment), was intended to capture any changes in methylation patterns seen after clozapine reached a therapeutic plasma level; this timing was chosen based on typical timelines for clozapine titration. Plasma levels at this time-point were either assessed by the research or clinical team. If clozapine levels were sub-therapeutic at six weeks, a sample was collected at eight weeks after clozapine initiation (Visit C). It was not expected that participants
- 38 -
Overall Method would show a full response to clozapine at this early stage, having just met therapeutic thresholds so a full clinical assessment was not conducted. However, data on any changes in smoking habits was collected as this is known to induce significant changes in DNA methylation (at all other time points, this was part of the clinical assessment). At twelve weeks - after at least six weeks of clozapine at a therapeutic level - evidence demonstrates that patients who are responsive to clozapine will have shown a response; therefore, participants were visited at 12 weeks (Visit D) and a full clinical assessment was conducted. Participants were then categorised as a responder or non-responder to clozapine based on their reduction in PANSS score (>20% reduction = responder), alongside obtaining a blood sample to assess methylation now that response or non-response to clozapine was evident. Finally, six months after clozapine initiation (Visit E), another assessment was conducted to capture clinical state and DNA methylation profile after a stable period of time on clozapine. Information on current medication was collected at each time point.
Figure 3 provides an outline of the assessments conducted at each time point (further details on the specific assessments and reasons for these are provided later in the chapter).
- 39 -
Overall Method
•Clinical Assessments •Demographic information •Clinical information •Medical history •Alcohol, Drug, and Tobacco Screen Visit A •Glasgow Antipsychotic Adverse Reaction Scale (GAS) - clozapine edition •Positive and Negative Syndrome Scale (PANSS) baseline •Personal and Social Performance Scale (PSP) •Global Assessment of Functioning (GAF) •Operational Criteria Checklist (OPCRIT) •Blood samples •10ml EDTA tube for DNA analysis •4ml EDTA tube for cell counts (if not collected by clinical team)
•Clinical Assessments •Smoking questionnaire Visit B •Medication changes •Blood samples six weeks •10ml EDTA tube for DNA analysis •4ml EDTA tube for cell counts (if not collected by clinical team) •6ml EDTA tube for plasma clozapine levels (if not collected by clinical team)
•Clinical Assessments •Smoking questionnaire Visit C •Medication changes •Blood samples eight weeks •10ml EDTA tube for DNA analysis •4ml EDTA tube for cell counts (if not collected by clinical team) •6ml EDTA tube for plasma clozapine levels (if not collected by clinical team)
•Clinical Assessments •Medication changes •Alcohol, Drug, and Tobacco Screen •Glasgow Antipsychotic Adverse reaction Scale (GAS) - clozapine edition Visit D •Positive and Negative Syndrome Scale (PANSS) •Personal and Social Performance Scale (PSP) twelve weeks •Global Assessment of Functioning (GAF) •Blood samples •10ml EDTA tube for DNA analysis •4ml EDTA tube for cell counts (if not collected by clinical team) •6ml EDTA tube for plasma clozapine levels (if not collected by clinical team)
•Clinical Assessments •Medication changes •Alcohol, Drug, and Tobacco Screen •Glasgow Antipsychotic Adverse reaction Scale (GAS) - clozapine edition Visit E •Positive and Negative Syndrome Scale (PANSS) •Personal and Social Performance Scale (PSP) six months •Global Assessment of Functioning (GAF) •Blood samples •10ml EDTA tube for DNA analysis •4ml EDTA tube for cell counts (if not collected by clinical team) •6ml EDTA tube for plasma clozapine levels (if not collected by clinical team)
Figure 3: Flowchart of assessments and samples collected throughout the study
- 40 -
Overall Method
2.1.2 Participants
2.1.2.1 Eligibility
Eligible participants were adult patients (aged 18 or over) diagnosed with either schizophrenia or schizoaffective disorder (confirmed by OPCRIT), due to be prescribed clozapine by their clinical team. Members of the research team did not influence the prescription of clozapine in any way throughout the six months, so this was a naturalistic study. All patients were recruited from South London and Maudsley (SLaM) NHS Foundation Trust for which we had ethical approval.
Ideally, all participants would have been clozapine naïve. This would have eliminated the possibility of the baseline assessment of DNA methylation including any lasting changes in methylation induced by prior clozapine use, which could potentially reduce the power to identify methylation changes induced by clozapine. However, setting prior clozapine exposure as an exclusion criterion may have significantly limited recruitment and subsequent statistical power, an unacceptable sacrifice when already faced with a likely small sample size. Participants were therefore only excluded if they had taken clozapine at any point in the three months before the current clozapine initiation. Due to the limited prior research on the epigenetic effects of clozapine, it is unclear whether clozapine leads to long-term changes in methylation (and whether these continue after clozapine is discontinued), but this exclusion criterion was both practically viable for recruitment numbers while simultaneously limiting the potential residual effects of DNA methylation changes induced by recent clozapine use. Additionally, within-individual analysis will not be confounded by differences in prior clozapine exposure.
Participants were recruited from a variety of clinical teams within SLaM including community care teams, assertive outreach teams, locked inpatient wards, rehabilitation wards, and supported accommodation. A significant number of participants were recruited either from the National Psychosis Unit, an inpatient ward which takes referrals for complex cases of treatment-resistant schizophrenia, or the TREAT team at Maudsley Outpatients, which supports patients commencing clozapine while remaining in the community. As our recruitment wasn’t limited to a specific type of clinical setting this helped ensure our data was not restricted to particularly severe or particularly mild cases of schizophrenia, and the heterogeneity of our sample did not raise a significant issue due to the primarily within-participant design.
2.1.2.2 Recruitment
Potentially eligible participants were identified through contact between members of the research team (including myself) and a range of sources including pharmacists, consultant - 41 -
Overall Method psychiatrists, junior doctors, and research nurses from psychiatric teams across SLaM. We presented the study to clinical teams with the aims of increasing awareness about the study, encouraging staff to contact us if they had eligible patients, and feeding back educational information and any early results to the clinical teams.
Potentially eligible participants were first approached by a clinical staff member, and asked if they would be happy to talk to one of the research team about our study. I or another member of the clinical team would then arrange to meet with the patient and the study was described verbally to potential participants alongside presentation of the written information sheet and providing a copy of this for the participant to read over. They were encouraged to ask questions about the research, with the researchers clarifying matters such as the distinction between their clinical care and the research. If happy to take part, they were led through the consent sheet and consent was taken for each item. Once consent was gained, participants were informed that they could withdraw consent at any point with no repercussions.
Participants were offered financial reimbursement for their participation - £5 for each blood collection, £5 for each set of clinical assessments, and £30 for each scanning session.
2.1.2.2.1 Capacity
In our patient group, there was increased likelihood of lacking capacity to give informed consent as patients necessarily had a severe psychiatric illness with active symptoms. The Mental Capacity Act provides criteria for approving research which involves patients who do not have capacity to consent, as a protective measure to avoid engaging particularly vulnerable patients in research if unnecessary. This study met the Mental Capacity Act approval criteria for involving patients who do not have capacity to consent as: 1) the research is directed at their impairing condition (treatment resistant schizophrenia), 2) research of equal effectiveness cannot be carried out in only patients with capacity (as this would reduce generalisability of our results - creating a bias towards patients with less severe symptoms - which is vital if results are to be used to improve understanding and clinical care for patients who do lack or fluctuate in capacity); 3) the research has the potential to provide knowledge on the treatment of others with the same condition; 4) the research involves negligible risk to the participant, is not associated with reduced freedom and is not unduly invasive (no part of the procedure would be outside the scope of standard clinical care and where possible blood samples were collected during routine clinical blood collection).
The Mental Capacity Act is clear that capacity should be assumed, and only assessed if there is reason to doubt capacity, for a specific circumstance. A person is said to lack capacity if they meet two criteria: 1) an impairment or disturbance of the functioning of the mind or brain, and 2) they are - 42 -
Overall Method unable to achieve at least one of four decision-making elements. In this study, as a diagnosis of schizophrenia is necessary for clozapine prescription and necessary to be eligible for recruitment, capacity to consent was automatically assessed due to them necessarily meeting the first criteria. To demonstrate capacity, participants needed to show an ability to understand the information relevant to the decision, retain the information for long enough to make the decision, weigh the information relevant to the decision, and communicate the decision by any means. For example, asking relevant questions about the research based on information provided to help them make an informed decision (e.g. “how often will I need to provide blood samples? I already have blood tests so wouldn’t want many more”), would indicate the ability to do the above. This was then documented by the research team.
In cases where patients did lack capacity, positive assent from the patient was required at all times (e.g. they still were required to agree to take part), alongside consent from a consultee (someone non-professionally involved in caring for the patient, who would look out for their interests in the decision - this would normally be their next•of•kin but could be an independent patient advocate or representative.). The consultee would be given a copy of the information sheet for consultees, have the study and the role of the consultee explained to them, and be invited to ask any questions they may have; they would then make a decision about whether they believed the patient would consent if they had capacity. If there was any indication that the patient would not wish to participate (from the patient, their clinical team or carers, or any advance statement of their wishes) they would not be enrolled in the study.
Once the patient was enrolled in the study, if at any point the participant objected to the study, or indicated that they would like to be withdrawn, they were withdrawn from the study immediately. The research team maintained close contact with both the clinical team and the consultee, to ensure awareness of the patient’s wishes or any change in their feelings about participation. It was clearly explained to the consultee that if for any reason they suspected that the patient’s wishes about participation may have changed, that they should contact the study team so that this can be discussed and the patient can be withdrawn if necessary. If capacity improved to the extent where a patient was able to consent, at that stage consent would be sought and documented as explained above.
Alternatively, for those patients who initially consented when deemed to have capacity, researchers remained aware of the possibility that they may lose capacity and would need to be re- assessed and a consultee found if necessary. However, for all but three participants, mental state either improved throughout the study as they responded to clozapine, or remained relatively stable.
- 43 -
Overall Method
2.1.3 Clinical assessments
To complete these clinical assessments, information was collected primarily through interview with participants and clinical notes were used to provide supporting material where necessary e.g. when patients lacked the insight or concentration for clinical assessments, or were otherwise unable to provide a detailed medical history themselves.
2.1.3.1.1 Clinical Information
The clinical information form collected data on primary diagnosis, predominant symptoms (positive/negative/disorganised), date of illness onset, presence of co-morbid psychiatric diagnoses, prior clozapine use, number of hospitalisation, date of obtaining treatment resistant diagnosis, and date clozapine was started. This information was collected to confirm participant eligibility, allow description of the characteristics of the sample, and collect data to be used as covariates in analysis if necessary.
2.1.3.1.2 Medication history/changes
The medication history/medication changes form documented information on medication type, dosage, reason for use (primary or secondary symptoms), start and stop date, and whether medication was ongoing, for all CNS and non-CNS medications. This information was collected to allow for medication history or changes in medication over the course of the study to be controlled for in analysis, as other medications may induce changes in DNA methylation.
2.1.3.1.3 Alcohol, Drug, and Tobacco Assessment
The Alcohol, Drug and Tobacco Assessment includes 21+ questions collecting detailed information on history and current usage of tobacco, caffeine, alcohol, cannabis and other drugs (inhalant, cocaine, amphetamines, heroin, hallucinogens, ketamine and any other drugs including legal highs). Questions were included on frequency, amount/quantity and - in the case of cannabis - the type (hash, skunk etc.). Participants were also specifically asked about their intake of caffeine, tobacco and cannabis on the day of assessment, with tobacco intake especially relevant for methylation analysis.
2.1.3.1.4 Glasgow Antipsychotic Adverse reaction Scale (GASS-C)
The GASS-C includes 18 items outlining different adverse reactions (e.g. “I have felt breathless”) and requires participants to answer how often they have experienced the adverse reaction over the past week, with options of “Never”, “Once” “A few times”, and “Everyday” which score 0-3 respectively. Participants then report if any adverse reactions they have experienced were severe or
- 44 -
Overall Method distressing. There is also an additional option for writing down adverse reactions not included above. A total score of 0-16 is classified as “absent/mild adverse reactions”, 17-32 as “moderate adverse reactions” and 33-48 as “severe adverse reactions”. This scale was developed specifically for patients commencing clozapine, as previous scales for the adverse reactions of generic antipsychotics (ANNSERS, SRA-34) neglect common clozapine adverse reactions such as dry mouth, heartburn, and tachycardia, as well as being lengthy clinician-rated scales which require training. Patient-rated scales have been shown to be highly consistent with clinician rated scales (Takeuchi, Fervaha, & Remington, 2016). The original GASS is a patient-rated scale but did not specifically enquire about hypersalivation, heartburn or constipation (one of the most life-threatening common adverse reactions if left untreated), while mentioning adverse reactions irrelevant to clozapine such as hyperprolactinaemia. Therefore, the GASS-C was chosen as the primary measure of adverse reactions in participants.
2.1.3.1.5 Positive and Negative Syndrome Scale (PANSS)
The PANSS is a commonly used scale for assessing severity of schizophrenic symptoms, developed by Kay, Fiszbein, & Opler (1987) and extensively validated since. It has a total of 30 items, with each scored from 1-7 with higher scores indicating more severe symptoms (1=absent, 2=minimal or within normal range, 3=mild, 4=moderate, 5=moderate-severe, 6=severe, 7=severe). The first 7 items (P1-P7) cover positive symptoms (delusions, conceptual disorganisation, hallucinatory behaviour, excitement, grandiosity, suspiciousness, hostility) and form the Positive Syndrome scale, where scores can range from 7-49. The next 7 items (N1-N7) cover negative symptoms (blunted affect, emotional withdrawal, poor rapport, passive/apathetic social withdrawal, difficulty in abstract thinking, lack of spontaneity, stereotyped thinking) and form the Negative Syndrome scale, where scores can range from 7-49. The final 16 items (G1-G16) cover general psychopathology (somatic concern, anxiety, guilt feelings, tension, mannerisms and posturing, depression, motor retardation, uncooperativeness, unusual thought content, disorientation, poor attention, lack of judgement and insight, disturbance of volition, poor impulse control, preoccupation, active social withdrawal), and form the General Psychopathology scale, where scores can range from 16-112. The overall PANSS score across all 30 items ranges from 30-210.
Scores are primarily based upon a semi-structured interview with the participant. Responses to questions and observation of behaviour (physical, interpersonal, cognitive-verbal) throughout the interview enable scoring, combined with some additional information gained from reports of clinical teams/family. The PANSS rating manual provides detailed descriptions for scoring the severity of each item. The PANSS should be completed by trained researchers or clinicians and therefore all
- 45 -
Overall Method members of the research team were trained by more experienced researchers initially, shadowed assessments by experienced researchers, and conducted their first interviews under observation and scored participants together, before conducting them independently.
Participants were deemed responders or non-responders to clozapine based on their PANSS scores. The change in PANSS Total score between baseline (Visit A) and twelve weeks (Visit C) was the primary outcome measure for symptom change in participants, with the criteria for being a responder a minimum of a 20% reduction in PANSS total.
2.1.3.1.6 Personal and Social Performance Scale (PSP)
The PSP is a quick measure of four areas of personal and social functioning of patients with psychiatric disorders: socially useful activities, personal and social relationships, self-care and, disturbing and aggressive behaviour. It assesses the degree of difficulty a participant demonstrates with these four domains over a one month period. For each domain, a participant is judged as having absent, mild, manifest, marked, severe or very severe difficulties. The overall score ranges from 1 to 100, divided into 10 equal intervals to rate the degree of difficulty in social functioning. The score is given based upon how severe the difficulties are (e.g. “70-61 Manifest, but not marked difficulties in one or more of areas a-c, or mild difficulties in d”), with functioning in a wide range of areas (household, physical, social network) determining judgement of the exact number. A score of 71 to 100 indicates a mild degree of difficulty; from 31 to 70, varying degrees of difficulty; and from 1 to 30, functioning so poorly as to require intensive supervision. This assessment was completed by a member of the research team based on the semi-structured interview and reports from clinical staff and family members in clinical notes. This was a secondary measure of response to clozapine, as some clinicians and patients argue that functioning is a higher priority than symptom reduction alone, or it may capture changes that symptom scales do not; for example, hallucinations may score similarly but be better managed or less distressing, or a minor reduction in a prominent debilitating symptom may lead to a very significant improvement in functioning.
2.1.3.1.7 Global Assessment of Function (GAF)
The GAF is a tool from the Diagnostic and Statistical Manual of Mental Disorders IV (p34). It is a numeric scale (1-100) used to rate the psychological, social, and occupational functioning of adults, on a hypothetical continuum of mental-health illness e.g. 70-61 indicates “Some mild symptoms (e.g. depressed mood and mild insomnia) OR some difficulty in social, occupational, or school functioning (e.g., occasional truancy, or theft within the household), but generally functioning pretty well, has some meaningful interpersonal relationships.” This was another secondary measure of response, providing a global summary measure of functioning. It was also completed by a member of the - 46 -
Overall Method research team based on the semi-structured interview and reports from clinical staff and family members in clinical notes.
2.1.3.1.8 Operational Criteria Checklist (OPCRIT)
The OPCRIT generates diagnoses according to 12 operational diagnostic systems (e.g. DSM-III, DSM-III-R, Research Diagnostic Criteria, ICD-10). It includes 90 items, on Details and History, Appearance and Behaviour, Speech and Form of Thought, Affect and Associated Features, Abnormal Beliefs and Ideas, Abnormal Perceptions, Substance Abuse or Dependence, and General Appraisal. Items are given a numbered score based on a code (e.g. 0=no, 1=at least 4 days, 2=at least 1 week, 3=at least 2 weeks). This assessment was used to provide independent confirmation of the diagnosis of schizophrenia or schizoaffective disorder. It was completed by a member of the research team based on the semi-structured interview and reports from clinical staff and family members in clinical notes (especially for historical information).
2.1.4 Blood samples
2.1.4.1 Collection
Blood samples were collected by venepuncture, by trained phlebotomists. As all patients on clozapine must have regular blood samples for clinical purposes, sample collection was often co- ordinated alongside regular monitoring samples taken by the clinical team to ensure no additional burden or distress.
The following samples were collected at the same time:
10ml EDTA tube for DNA analysis 4ml EDTA tube for cell counts (if not collected by clinical team) 6ml EDTA tube for plasma clozapine levels (if not collected by clinical team)
Samples for cell counts and plasma clozapine levels were frequently collected by clinical teams, as part of standard clinical monitoring. As I did not need access to the sample to run specific analysis (unlike the samples for DNA analysis), only the results of the cell count and plasma clozapine level analysis, in these scenarios we did not collect these samples ourselves to reduce unnecessary burden on the participants and unnecessary use of resources; instead, we accessed their clinical records to gain the relevant results. However, unfortunately this information was often incompletely uploaded or not uploaded to clinical records. Additional samples were collected for other portions of the larger project.
- 47 -
Overall Method
2.1.4.2 Sample processing
Note: I did not carry out any of the sample processing, but shadowed every step taken for one of the batches of samples assayed.
All bloods reached our laboratory at the BRC Bioresource within two hours of collection. On reaching the lab, EDTA tubes for DNA and cell counts were immediately stored on ice and then either processed for DNA extraction or stored at -20oC before processing. All samples were labeled with linked-anonymised barcodes prior to storage. This allowed for a sample tracking system to transfer DNA samples for epigenetic analysis. 1.5ug of DNA diluted to 25ng/ul was sent from the BRC Bioresource labs to the Complex Disease Epigenetics Group labs within Exeter Medical School. Arrays were organised so that all samples from the same individual were placed on the same array to prevent batch effects being misconstrued for longitudinal changes. The official guidance for Illumina 450k microarrays was followed (Figure 4).
Day 1: Bisulfite treatment was used to convert unmethylated cytosine into uracil, while methylated cytosines remained as unconverted cytosines, due to different sensitivities of cytosine and 5-methylcytosine to deamination by bisulfite. The samples were then denatured (because the conversion only works on single-stranded DNA), a bisulfite conversion reagent added, and the samples were then incubated overnight in a thermal cycle.
Day 2: The reagent was then cleaned-up, samples centrifuged, and a desulfonation buffer was applied and samples were incubated at room temperature for 15 minutes. Samples were then cleaned and washed twice to remove the buffer, and the bisulfite converted DNA (BCD) was eluted by applying an elution buffer and centrifuging. The BCD samples were then transferred to a BCD plate.
- 48 -
Overall Method
Figure 4: Workflow for Illumina 450k BeadChip assays, from Infinium HD Methylation Assay Protocol GuideManual
Whole genome amplification was used to uniformly increase the amount of DNA while minimising locus bias and mutation rates. This entails random hexamer primers annealing to the template DNA, and when DNA synthesis starts Phi29 DNA polymerase displaces newly produced
- 49 -
Overall Method
DNA strands, generating newly synthesis single stranded DNA for more primers to anneal. To achieve this, a new plate for multi-sample amplification was created. Multi-sample amplification mix (MA1) was added first, followed by the 4 µl of the BCD samples, and then 0.1N sodium hydroxide (NaOH) to denature the BCD samples. Plates were vortexed, centrifuged and incubated, and then RPM reagent was added to neutralise the sample. Lastly, MSM was added and the plates were again vortexed and centrifuged.
The denatured samples were then incubated and isothermally amplified overnight in the Illumina Hybridization Oven, leading to a hyper-branched network.
Day 3: Enzymes are used to fragment these branches in a controlled way. Endpoint fragmentation was used to prevent over-fragmentation. The plate was centrifuged, fragmentation solution (FMS) added, the plate was vortexed, centrifuged and put on a heat block.
The DNA was then precipitated (purified/concentrated). PM1 was added to the plate, it was vortexed, incubated, centrifuged and then isopropanol added. The plate was then inverted to mix, incubated, centrifuged, and liquid was drained from the plate – after the plate had dried, blue pellets were left for each sample.
The DNA was then re-suspended in hybridisation buffer – RA1 was added, and the plates were incubated, vortex and centrifuged.
Finally, the single stranded DNA was hybridised to the Illumina 450k BeadChip assays. The assays were silicon based and each sample section contains oligonucleotide probe sequences that are attached to beads – they have microscopic DNA for the samples to hybridise to. The 450k array covers CpG sites on 99% of RefSeq genes, with an average of 17 CpG sites per gene region distributed across the promoter, 5'UTR (untranslated region which immediately precedes the translation initiation codon), first exon, gene body, and 3'UTR (untranslated region which immediately follows the translation termination codon). It covers 96% of CpG islands, with additional coverage in island shores and the regions flanking them.
To avoid batch effects creating the appearance of differences over the follow up period, all samples from a participant were included on the same assay/within the same batch. The samples were placed on the heat block to denature them, the plate was centrifuged and then the samples were added to the BeadChips. These were loaded into the hybridisation chamber inserts, which have PB2 as humidify buffer, and the chambers were incubated in the Illumina hybridization oven overnight. During hybridization, the amplified and fragmented DNA samples annealed to locus- specific 50mers – two bead types correspond to each CpG locus – one bead types corresponds to methylated cytosine and another to unmethylated. - 50 -
Overall Method
Day 4: After hybridisation, the BeadChips were washed in two PB1 reagent washes. Unhybridized and non-specifically hybridized DNA was washed away, leaving only strongly paired strands, and the BeadChip was prepared for staining and extension. Then each BeadChip was prepared for processing, with black-frames, clear spacers and clean glass back plates clamped to the BeadChip.
A robot was used to extend and stand the chips in capillary flow-through chambers, in a heated chamber rack. RA1 was used to further wash away unhybridized and non-specifically hybridized DNA sample. XC1 and XC2 were used. TEM reagent extended the primers hybridised to the DNA, incorporating labelled nucleotides into the extended primers. 95% formamide/1 mM EDTA was added to remove the hybridized DNA, and then XC3 reagent used to neutralise. The labelled extended primers could then undergo multilayer staining on the Chamber Rack with STM, ATM and XC3. Afterwards, the BeadChips were washed in PB1, coated with XC4 and dried using the vacuum desiccator.
The Illumina iScan System (two-channel high-resolution laser imagers) then imaged the BeadChip, using a laser to excite fluorophore (a fluorescent chemical compound) into emitting light, scanning at two wavelengths simultaneously. This produced a high resolution image which the scanner records. Signal intensity is dependent on the amount of target sample binding to the probes present at the spot, so is measured to generate regression coefficient values, a measure of the degree of methylation at each locus.
2.1.5 Quality control (QC) assessments
Ninety six samples from twenty-six participants with follow-up data were assayed at the laboratory. Six samples from three participants were excluded because they dropped out of the study before twelve-week follow-up and therefore had no response data. I ran comprehensive quality control (QC) assessment on the methylation data from the remaining samples leading to the final inclusion of eighty-five samples from twenty two participants (Figure 5).
Firstly, I checked median methylation intensities for each sample were not noticeably below (or above) the average, suggesting an issue had occurred with the sample at some stage. One sample with median methylation intensity below 2500 was excluded. I also checked that methylated and unmethylated intensities were roughly correlated when plotted against one another, as a sign of good samples, and there was good correlation. I then ran tests to assess if the bisulfite conversion for each sample had been successful. This is done by comparing sequences to bisulfite conversion control probes provided specifically by Illumina for QC purposes, and allows the calculation of percentage of bisulfite conversion for the sample. One sample with conversion of only 80% was - 51 -
Overall Method excluded. All the other samples had above 90% conversion, with all but one indicating conversion of 95% or above.
I then carried out a number of tests to identify any potential sample switches or mislabelling of samples – if any were identified, I looked back into records of sample collection and handling to correct mistakes. This included comparing reported gender with predicted gender based on probes indicating X or Y chromosome, as well as comparing methylation at SNPs where methylation is expected to stay stable on repeated samples from the same individual to ensure that there is expected consistency between samples from the same individual. One participant with three samples was excluded because there was a query which could not be resolved over whether the analysed samples belonged to the participant. The correlations between the remaining samples from the same participants are plotted. I also uploaded data to Horvath’s DNAm age calculator (Horvath, 2013) to run two additional quality assessment checks. Firstly, that samples were correctly predicted to come from human blood – this was true for all samples. Secondly, that the predicted age had a strong correlation with reported age – there was a correlation of 0.90, as expected. Horvath’s age calculator was also used to estimate cell composition for use as a covariate in analysis.
Probe filtering was then used to remove any probes that had failed to hybridise (indicated by over 1 % of samples with a detection p-value greater than 0.05), as well as probes represented by less than 3 bead counts on the array in 5% of samples. 1982 sites were removed because 1 % of samples had a detection p-value > 0.05 at these sites and 548 sites were removed due to a beadcount <3 at these sites in 5 % of samples. Regression coefficients were then normalised across all samples. For further analysis, probes that overlap with SNPs and those that cross hybridise to multiple genomic locations were removed. Probes on sex chromosomes were removed due to difficulties in interpreting differential methylation on the X chromosome caused by X inactivation in females; the sample was too small to do analysis divided by sex.
All plots can be found in Appendix II.
2.1.6 Power
This is one of the first studies to investigate longitudinal epigenetics changes induced by any antipsychotic, so effect sizes are hard to predict. However, the study benefited from increased statistical power due to a longitudinal within-participant design, with data collected to control for multiple potential known confounds. This is somewhat comparable to studies of discordant MZ twins, which use paired statistical analysis and can study the epigenome independent of underlying genetic structure in a similar way to a within-participant design. One such epigenome-wide study
- 52 -
Overall Method with 22 MZ twin pairs (44 individual samples) had 80% power to detect a Δβ (DNA methylation difference) of 0.06, using a strict Bonferroni correction (Dempster et al., 2011).
- 53 -
Overall Method
96 samples from 26 participants assayed at laboratory
3 participants excluded (2 samples each) due to no response data
Check median methylation: 1 sample excluded
Check correlation between methylated and unmethylated intensities: 0 samples excluded
Check bisulfite conversion: 1 sample excluded
Check matching gender and SNPs for repeated measures from same participant: 1 participant (3 samples) excluded due to unresolved mismatches
Check samples correctly predicted from human blood: 0 samples excluded
Check correlation between predicted and actual age: 0 samples excluded
85 samples from 22 participants left
1982 sites for each sample removed due to low hybridisation 548 sites removed for each sample due to low bead counts
Probes overlapping SNPs, cross-hybridising probes and probes on sex chromsomes removed
Figure 5: Flowchart of QC assessments
- 54 -
Overall Method
2.2 Cross-sectional dataset
To investigate the similarities between the current dataset and a previous dataset based on analysis of clozapine exposed patients, and see if DMPs were replicated, I was given access to data from a previously collected and analysed cross-sectional sample. The IMPACT (Improving Physical health and Substance Use in Severe Mental Illness trial) study and the CRESTAR clozapine clinics study both recruited large cohorts of patients with schizophrenia, primarily from SLaM NHS Trust. Combined, these projects provided cross-sectional samples of 155 chronic patients on clozapine and 135 chronic patients not on clozapine. The same method of bisulfite conversion and Illumina 450k array described above was used to determine epigenome-wide profiles of DNA methylation for all of these patients. An EWAS was conducted to determine differentially methylated CpG sites associated with clozapine exposure, controlling for age, sex, chip, smoking status, and predicted cell composition. I was provided with the results of the regression model.
2.3 Summary
To summarise, this longitudinal study recruited patients with treatment-resistant schizophrenia or schizoaffective disorder due to be initiated on clozapine and collected samples of whole blood alongside clinical assessments before initiation and at three subsequent time points. Clinical assessments covered symptom severity, psychosocial functioning and quality of life. We also collected information on medication use and smoking at each time point. DNA was extracted from whole blood samples, treated with sodium-bisulfite, and then processed using Illumina 450k arrays to profile DNA methylation at hundreds of thousands of CpG sites across the epigenome. A total of eighty-five samples from twenty-two participants passed QC and were included in subsequent analysis.
The following three chapters report on analysis of this data to investigate DNA methylation trajectories over the study period associated with 1) exposure to clozapine, 2) therapeutic efficacy and 3) adverse reactions.
- 55 -
Exposure to Clozapine and Associated DNA Methylation Changes
3 Exposure to Clozapine and Associated DNA Methylation Changes
3.1 Background: Exposure to clozapine
Before reporting on my analysis investigating changes in DNA methylation associated with exposure to clozapine, I will explore in greater depth previous research findings on the physiological consequences of clozapine exposure. First I will discuss broad literature from numerous fields, and then I will focus on epigenetic research, largely from rodent models.
As the first atypical antipsychotic, clozapine is known for being distinct from traditional antipsychotics in its pharmacology. Firstly, clozapine is a relatively weak antagonist of D2 dopamine receptors (Farde et al., 1992), whereas for the vast majority of antipsychotics greater effectiveness is strongly correlated with higher D2 occupancy (Yilmaz et al., 2012). In vitro data indicates approximately equivalent affinities for D2 and D1 receptors (Farde et al., 1992) which makes clozapine a greater D1 antagonist than most antipsychotics. It has an even higher D4 antagonism as well as an affinity for serotoninergic receptors, particularly 5-HT2A and 5-HT2C (Meltzer & Gudelsky, 1992). Some have suggested that the ratio between serotonergic and dopaminergic receptor affinities is critical to clozapine (Kapur, Zipursky, & Remington, 1999). Additionally, clozapine has an affinity for noradrenergic receptors, leading to increased noradrenaline plasma levels (Breier et al., 1994). In addition, one of the metabolites of clozapine, N-Desmethylclozapine (known as norclozapine), is pharmacologically active; it is a weak partial agonist of D2 and D3 receptors as well as being an agonist of the M1 and delta opioid receptors.
Clozapine appears to have a regional selectivity that also distinguishes it from other antipsychotics. Studies suggest a higher affinity for limbic D2 receptors than striatal D2 receptors, in contrast to most antipsychotics (Kessler et al., 2006). Clinical and neurophysiological studies have found clozapine to have more selective and restricted effects on the striatum than typical antipsychotics; for example, in the nucleus accumbens/ventral striatum, rodent studies (Deutch & Cameron, 1992) have found that clozapine increases gene expression in the shell regions (associated with limbic and prefrontal circuitry) whereas typical antipsychotics also increase gene expression in the core (associated with motor function – which would explain why typical antipsychotics are more likely to induce motor symptoms such as extra-pyramidal adverse reactions (Painuly et al., 2004).
Alongside this research into the pharmacological profile of clozapine, to understand the immediate consequences of clozapine on receptor binding, research has also investigated the effects of clozapine use on structural volume and regional perfusion. Garcia et al. (2015) identified 23 studies of structural and functional correlates of the use of clozapine in schizophrenia: the most
- 56 -
Exposure to Clozapine and Associated DNA Methylation Changes consistent findings were that clozapine use with associated with volume reductions in the caudate nucleus (part of the basal ganglia), reduced frontal lobe perfusion and decreased activation in the basal ganglia. Some recent longitudinal studies have studied other structural characteristics such as white matter integrity using diffusion tension imaging and found an increase after clozapine use (Ozcelik-Eroglu et al., 2014). Clozapine also appears to increase neuroplasticity, with rodent models demonstrating increased dendritic spine formation and synaptogenesis in the hippocampus (Critchlow, Maycox, Skepper, & Krylova, 2006).
However, interpreting many of these findings is challenging when considering other antipsychotic use, as – in many countries - all participants will necessarily have undergone two antipsychotic trials as a minimum which are likely to have caused structural and functional changes. Additionally, while some studies are longitudinal within-participant design, many are between- groups and comparing to patients on alternative antipsychotics. Therefore, it is often unclear whether changes are completely, partly, or not at all due to alternative antipsychotics, whether they are normalising or exacerbating changes caused by other antipsychotics, or whether changes are masked by other antipsychotics. The between-group design also raises the question of whether differences are due to clozapine use or due to an underlying difference between treatment-resistant patients who require clozapine and patients who respond to conventional treatment.
Genetic research in this area has largely focused on genetic variation which predicts outcomes in response to clozapine (discussed in Genetic variation associated with response to clozapine), as this is the main avenue for genetic variation providing insight into clozapine’s mechanism. Gene expression data on the other-hand can provide information about the direct effect of clozapine exposure. One study of whole-blood gene expression in 55 patients receiving clozapine and 97 patients not receiving clozapine did not find any significant associations, only a nominal association for a smaller subset of patients on clozapine monotherapy with a module of genes involved in functions such as blood coagulation and platelet degranulation (Harrison et al., 2016). However, a 2017 analysis of expression data from four microarray studies of post-mortem tissue identified three genes and four pathways associated with clozapine in all four studies (Lee et al., 2017): upregulation of glutamate-cysteine ligase modifier (GCLM) and zinc finger protein 652 (ZNF652) and downregulation of glycophorin-C (GYPC); clathrin-mediated endocytosis, SAPK/JNK signalling, 3- phosphoinositide signalling, and paxillin signalling pathways, which have been implicated in neuronal migration and synaptic function, among other phenomena. Epigenetic data is then another source of information about physiological changes induced by clozapine.
3.1.1 Existing epigenetic research
- 57 -
Exposure to Clozapine and Associated DNA Methylation Changes
Three studies in rodents have examined the impact of clozapine on epigenetic markers, and have driven speculation about epigenetic modifications driving clozapine’s unique efficacy in treatment-resistance as each compared clozapine to other antipsychotics which did not induce the same epigenetic changes (Aoyama et al., 2013; Dong, Nelson, Grayson, Costa, & Guidotti, 2008; Huang et al., 2007)
Huang et al., (2007) administered clozapine (5 mg/kg), haloperidol (0.5 mg/kg) or saline (placebo) to mice acutely or chronically (21 days) and then examined histone trimethylation of H3K4 at GAD1 (a GABAergic gene locus) in brain tissue. They found that clozapine exposure – but not haloperidol or placebo – significantly increased histone trimethylation of H3K4 at GAD1 (a GABAergic gene locus), with a threefold increase in the cerebral cortex after chronic treatment compared to placebo. They conducted an additional post-mortem comparison of human brain tissue from schizophrenia subjects treated with typical antipsychotics and schizophrenia subjects treated with clozapine demonstrating a similarly significant association between clozapine exposure and a two- fold increase in trimethylation at H3K4 at GAD1. They also found clozapine mediated Mll1 (H3K4- specific histone methyltransferase gene) expression and occupancy at GAD1 chromatin, with a twofold increase in occupancy after a single dose and with specificity to that locus, providing a mechanism for the increased histone methylation.
Within the same paper, a different set of post-mortem samples demonstrated that H3K4 methylation at GABAergic gene promoters in the human prefrontal cortex typically increases throughout development and into adulthood, but that in schizophrenia there was a pattern of decreased methylation; therefore, these findings combined suggest clozapine may normalise aberrant hypomethylation at this locus. However, Huang et al. did not find a corresponding significant increase in GAD1 mRNA levels induced by clozapine exposure, so the functional change in the cortical GABAergic system is unclear.
Dong et al., (2008) built upon this study design by first inducing hypermethylation of fronto- cortical and striatal reelin and GAD67 promotor regions with methione administration to rodents for 7 days, providing a model of the aberrant hypermethylation of DNA proposed in schizophrenia patients. Clozapine (3.8-15µmol/kg), sulpiride (12.5-50 µmol /kg), haloperidol (1.3-4 µmol/kg) or olanzapine (4-15 µmol/kg), each either alone or in combination with valproate, was then administered each twice a day for three days. They found that clozapine and sulpiride - but not haloperidol or olanzapine - induced dose-related reelin and GAD67 promoter demethylation in the frontal cortex and striatum, and the demethylation was particularly pronounced (in an interactive, not additive way) when valproate was also administered. Similar findings were reported for H3-K9
- 58 -
Exposure to Clozapine and Associated DNA Methylation Changes acetylation at reelin promoters in the frontal cortex; clozapine and sulpiride lead to increased acetylation, but there was a two-fold increase when valproate was co-administered. Examination of nuclear extracts found a similar increase in reelin DNA demethylation associated with clozapine and valproate, suggesting that demethylation is caused by activation of pre-existing nuclear demethylase enzymes. These epigenetic changes were not found in liver GABAergic neurons, indicating selectivity to central nervous system.
The authors suggest that clozapine is an antipsychotic which normalises the downregulation of GABAergic gene expression found in schizophrenia, through inducing DNA demethylation and subsequently increasing gene expression to normal levels; and valproate intensifies this effect. As these effects were not apparent for olanzapine, this does not seem to be a general mechanism of atypical antipsychotics. Alongside Huang et al reporting that typical antipsychotics such as haloperidol did not show equivalent epigenetic alternations to clozapine, the implication may be that clozapine has especially pronounced epigenetic consequences compared to other antipsychotics including other atypicals.
More recently, Aoyama et al., (2013) induced decreased histone H3 acetylation at lysine 9 residues in rodent prefrontal cortex, a shift in intracellular HDAC distribution to the nucleus, decreased phosphorylated -CaM kinase II in the nuclear extracts and decreased mRNA expression of GAD1, GABRA1, and PVALB, with 14 days of phencyclidine (PCP) administration - a non-competitive N-methyl-D-aspartate (NMDA) receptor antagonist. This produced behavioural deficits such as impaired recognition memory and social deficit, providing a useful model for schizophrenia. They then administered haloperidol (1.0mg/kg) or clozapine (10mg/kg), once a day for 7 days.
Clozapine but not haloperidol was able to significantly attenuate the behavioural deficits, normalise the acetylation levels of histone H3 at lysine 9 residues, and reverse the shift in intracellular HDAC distribution. All these effects were blocked if a dopamine D1 receptor antagonist (SCH-23390 hydro- chloride) was injected 30 minutes before clozapine administration. Clozapine also attenuated the decreases in prefrontal mRNA expression of PVALB, but not GAD1 or GABRA1. The combined findings of normalising nuclear extracts, acetylation, mRNA expression and behavioural deficits suggest a coherent role for clozapine inducing this chain of events.
Based on studies such as these, Guidotti et al (2012) suggested that clozapine (particularly when combined with valproate or other HDAC inhibitors) increases histone acetylation and recruits DNA demethylation enzymes, normalising GABAergic promoter hyper-methylation and the downregulated gene expression. For human studies, it is significantly easier to obtain epigenome
- 59 -
Exposure to Clozapine and Associated DNA Methylation Changes data on DNA methylation from peripheral tissue than obtaining data on histone modifications, so existing human studies of clozapine-induced epigenetic changes have measured DNA methylation.
During my thesis another paper of a similar sample size was published with a similar methodology (Kinoshita et al., 2017)– they collected blood samples from patients with treatment- resistant schizophrenia before clozapine initiation and after a year of clozapine treatment and measured changes in epigenome-wide DNA methylation and associations with symptom ratings. They report finding ~30,000 CpG sites with a significant change in DNA methylation, 44.8% of which indicated an increase in methylation, 40.7% of which were located in promoter regions and 32.5% in gene bodies. Furthermore they report that significant sites were notably in the CpG islands of promoter regions for GABA, glutamate and other schizophrenia-associated genes, and there was enrichment for cell substrate adhesion and cell matrix adhesion gene ontology.
Despite some of the similarities between this study and the design of my study, there are significant differences in methodology and analysis. Firstly, they do not report on any samples in between their baseline time-point and a year later, whereas I report on samples collected at three clinically significant time-points within the first six months allowing a closer look at trajectories of DNA methylation change. Secondly, in our study a subsection of samples had data on clozapine level allowing an additional analysis which directly reflects clozapine exposure. Thirdly, relevant to chapters four and five, they only report on a correlation with overall symptom percentage change, whereas I report on analysis of symptom change as both categorical and continuous variables and current symptom severity, as well as analysis of different sub-sets of symptoms (positive, negative, social), and they have not reported any analysis of adverse reactions to clozapine, whereas I report on associations with increased appetite/weight gain, drowsiness, and cardiovascular symptoms.
Moreover, their analysis does not control for numerous confounding variables of known importance – e.g. smoking status, cell counts, and microarray chip id – whereas I use a multi-level model including these variables as covariates. Additionally, while they report controlling for multiple comparisons, the details on this are ambiguous and their reported findings of ~30,000 significant findings (8.5% of all sites analysed) is highly unusual, whereas I use a clearly stated epigenome-wide significance threshold which is standard within the field (Lehne et al., 2015). Cumulatively, these methodological failings call their conclusions into question.
Therefore, this thesis remains an original contribution to the literature, providing a methodologically rigorous approach to a longitudinal dataset of clozapine users with comprehensive analysis of clinical outcomes.
- 60 -
Exposure to Clozapine and Associated DNA Methylation Changes
I will now present my analysis of our data on changes in DNA methylation associated with exposure to clozapine.
- 61 -
Exposure to Clozapine and Associated DNA Methylation Changes
3.2 Statistical method: Exposure to clozapine
To explore the effects of exposure to clozapine on DNA methylation, I conducted analysis of:
1) Within-individual length of time exposed to clozapine as a predictor of DNA methylation
Time exposed to clozapine: time on clozapine (in weeks) for each sample was recorded as the number of days between clozapine initiation and sample collection divided by seven. The time on clozapine for baseline samples, taken before exposure to clozapine, was recorded as 0 weeks.
2) Within-individual variation in clozapine level as a predictor of DNA methylation
Clozapine level: the plasma clozapine level for each sample was in mg/l.
To identify DNA methylation changes associated with clozapine exposure, I conducted mixed model regression analyses using lme4 and lmer, R packages for liner mixed-effects models (Bates, Maechler, Bolker, & Walker, 2014; Kuznetsova, Brockhoff, & Christensen, 2015). Time on clozapine or clozapine level was used a predictor of methylation at each CpG site. Both models controlled for baseline age, sex, predicted cell counts (CD8T, CD4T, NK, B, granulocytes and monocytes), smoking at time of sample collection (yes/no), chip ID, current medications1 (number of antipsychotics, benzodiazepines or other mood stabilisers, antidepressants, hyoscine, metformin, laxatives), number of total prior antipsychotics, as well as controlling for the repeated measures from individual participants as a random effect.
These analyses both produced a model with regression coefficients for each CpG site - representing the percentage change in methylation at the CpG site associated with an additional week on clozapine / increase in clozapine level of 1 – as well as a p value for the significance of this association.
Global DNA Methylation: I report global DNA methylation change associated with exposure to clozapine, calculated using the mean of the regression coefficients for all CpG sites which passed QC. To explore global DNA methylation at each time point, I report means of the uncorrected2 methylation data for all CpG sites which passed QC at each time point.
1 Not all medications were used as individual covariates as this would have involved an additional 20+ covariates and risked over-fitting the model.
2 Uncorrected methylation data was used because the model did not produce regression coefficients for each time point and this data provides additional information for interpretation; however note that this is not adjusted for covariates and it does not indicate within individual comparisons (as the regression model does) but comparisons between all samples. - 62 -
Exposure to Clozapine and Associated DNA Methylation Changes
Differentially methylated positions (DMPs): I report the means of the regression coefficients for the top ten and top one hundred differentially methylated positions (DMP). The direction of effect in the top one hundred DMPs is presented throughout the thesis, even when individual DMPs are not significant, as this may indicate broad patterns of hyper or hypo-methylation; this is of interest because previous research has proposed a wide-reaching dysfunction in regulation of DNA methylation and/or gene expression in schizophrenia.
I then summarise how many differentially methylated positions were below a standard threshold for epigenome wide significance (1*10-7) 3(Lehne et al., 2015) and how many were trend- level significance (5*10-5). I describe the ten most statistically significant DMPs, including the regression coefficient, p value, and uncorrected methylation data demonstrating changes over time on clozapine/demonstrating the correlation with clozapine level. Data from the UCSC Genome Browser was used for gene annotation. I report the location of the CpG site and a description of the function and expression of the associated gene; unless stated otherwise, all information is from the GeneCards database (www.genecards.org; Stelzer et al., 2016).
I also report an estimate of correlation between methylation at this CpG site in whole blood and methylation at this CpG site in brain tissue; this information was provided by the blood-brain DNA methylation comparison tool at http://epigenetics.iop.kcl.ac.uk/bloodbrain/ which is based on data from Hannon et al., (2015).
Pathway analysis: I used GOSeq to conduct pathway analysis on the results of each regression model and identify pathways with an over-representation of DMPs. GOSeq used the UCSC Browser gene annotations and ontology categories. DMPs with a p value of 5x10-4 were used and pathways were filtered to those including at least ten genes.
Candidate CpG site methylation: Based on previous literature reporting the differential methylation induced by exposure to clozapine (Aoyama et al., 2013; Dong et al., 2008; Huang et al., 2007; Kinoshita et al., 2017), I identified six genes which have been associated with clozapine exposure: GAD1, PVALB, RELN, GRIN2A, GRIN2D and GRM7. In total there were 182 CpG sites for these genes included in the 450k assay which passed QC. To adjust for multiple-comparisons, I used an adjusted p value of 0.05/182 = 0.275x10-3.
Comparison to cross-sectional data: I then compared the results of the models of clozapine exposure from my longitudinal dataset to results from the previous dataset of cross-sectional
3 As this is one of the first datasets of its kind, with this thesis intending to present exploratory analysis for discussion, significance thresholds were not further corrected for the multiple testing involved in conducted numerous models with the same sample. - 63 -
Exposure to Clozapine and Associated DNA Methylation Changes clozapine exposure, firstly to explore whether the cross-sectional model was validated by the longitudinal models, and then to explore whether the longitudinal model was validated by the cross- sectional model. To investigate replication of the findings from the cross-sectional model, I took all the top DMPs with a p value < 5*10-5 from the cross-sectional model and ran a correlation analysis on the regression coefficients from the cross-sectional model for these top DMPs and the regression coefficients from one of the longitudinal models for these DMPs; this was to establish if there was the same direction of effect for DMPs in both cohorts (i.e. if methylation was higher in the clozapine exposed group from the cross-sectional dataset, was methylation also higher in samples with greater levels of clozapine exposure in the longitudinal dataset). To investigate replication of the findings from both longitudinal models of clozapine exposure, I took all the top DMPs with a p value < 5*10-5 from a longitudinal model and ran a correlation analysis on the regression coefficients from the longitudinal model for these top DMPs and the regression coefficients from the cross-sectional model for these DMPs; again, this was to establish if there was the same direction of effect for DMPs in both cohorts (i.e. if methylation was higher in samples with greater levels of clozapine exposure in the longitudinal dataset, was methylation also higher in the clozapine exposed group from the cross- sectional dataset).
All analysis was conducted using R 3.1.2.
The results for each analysis model are reported separately. First I will report the results for analysis of time on clozapine, and then I will report the results for analysis of clozapine levels.
3.3 Results: Time on clozapine (weeks)
3.3.1 Participant characteristics (time on clozapine)
All eighty-five samples from twenty-two participants had information on date of collection so all samples were included in the analysis of time on clozapine. Information about the characteristics of this sample can be found in Table 1, with key details represented in subsequent pie charts (Figure 6): the majority of participants were male, had a diagnosis of paranoid schizophrenia, had no psychiatric comorbidities, had predominant negative symptoms, had not been previously exposed to clozapine, did not smoke, and experienced increased appetite/weight and drowsiness. There were ten responders to clozapine and twelve non-responders, and there was a mean decrease of 17.6% in PANSS Total (paired t-test comparing baseline total and twelve week total, t=8.420, df=78, p<0. 1x10-4) and a mean increase of 30.33 in psychosocial functioning and 39.84% in quality of life.
- 64 -
Exposure to Clozapine and Associated DNA Methylation Changes
Table 1: Characteristics of participants included in Time on Clozapine Analysis
Age (years) Mean 37.80 Minimum 19.94 Maximum 61.38 Gender Male 17 Female 5 Ethnicity White 10 Black 10 Asian 2 Diagnosis Schizophrenia 11 Schizoaffective disorder 3 Months between diagnosis of treatment- Mean 27.05 resistance and clozapine start Minimum 0 Maximum 111 Inpatient Psychiatric Admissions Mean 3.5 Minimum 0 Maximum 10 Psychiatric Comorbidities Affective 2 Anxiety 2 Substance Abuse 3 None 15 Prior Clozapine Exposure Yes 5 No 17 Number of prior typical antipsychotics Mean 0.77 Minimum 0 Maximum 2 Number of prior atypical antipsychotics Mean 2.91 Minimum 1 Maximum 7 Number of prior mood stabilisers Mean 0.23 Minimum 0 Maximum 1 Smoker at baseline Yes 3 No 19 Smoker at any time-point during study Yes 7 No 15
- 65 -
Exposure to Clozapine and Associated DNA Methylation Changes
Baseline PANSS Total Mean 78.32 Minimum 54 Maximum 119 Baseline PANSS Positive Mean 19.57 Minimum 9 Maximum 36 Baseline PANSS Negative Mean 21.864 Minimum 9 Maximum 39 Predominant Baseline Symptoms Positive 7 Negative 15 12w PANSS Total Mean 63.45 Minimum 35 Maximum 98 6m PANSS Total Mean 61.20 Minimum 37 Maximum 97 Responder at 12 weeks Yes 10 No 12 PANSS % Change at 12w Mean -17.60 Minimum -53.57 Maximum 26.47 PSP % Change at 12w Mean 30.33 Minimum -50 Maximum 162.50 GAF % Change at 12w Mean 39.84 Minimum -25 Maximum 185.71 Increased appetite/weight at 12w Yes 13 No 9 Cardiovascular symptoms at 12w Yes 4 No 18 Drowsiness at 12w Yes 15 No 7
- 66 -
Exposure to Clozapine and Associated DNA Methylation Changes
Figure 6: Pie charts illustrating key demographics across participants
- 67 -
Exposure to Clozapine and Associated DNA Methylation Changes
- 68 -
Exposure to Clozapine and Associated DNA Methylation Changes
- 69 -
Exposure to Clozapine and Associated DNA Methylation Changes
- 70 -
Exposure to Clozapine and Associated DNA Methylation Changes
- 71 -
Exposure to Clozapine and Associated DNA Methylation Changes
3.3.1.1 Details of Medication
The five most frequent antipsychotic medications that participants had been previously exposed to before baseline were risperidone (n=15, 68.18%), aripiprazole (n=13, 59.09%), olanzapine (n=11, 50%), amisulpride (n=6, 27.27%), quetiapine (n=6, 27.27%) and promethazine (n=6, 27.27%). The mean number of previously prescribed antipsychotics was 3.91, ranging from 2-10. At baseline, antipsychotic medication included olanzapine (n=6), aripiprazole (n=3), risperidone (n=3), promethazine (n=3), clonazepam (n=3), lithium (n=2), valproate (n=2), mirtazapine (n=2), lurasidone (n=1), quetiapine (n=1) and flupentixol (n=1).
During clozapine treatment, there were four patients on adjunctive antipsychotics; one on olanzapine and valproate throughout, one on clonazepam throughout and amisulpride initially, one on amisulpride throughout, and one on amisulpride initially. For affective symptoms, three patients took SSRIs (citalopram or sertraline) during their treatment, two took mirtazapine, one took amitriptyline and one took lithium. Five patients took metformin (a medication to treat diabetes and metabolic symptoms) at some point during their clozapine treatment, nine took hyoscine (a medication to treat hyper-salivation and nausea) and four took some form of laxative (e.g. senna, lactulose). Two participants took antibiotics at some point during their treatment.
- 72 -
Exposure to Clozapine and Associated DNA Methylation Changes
As described in the previous section on statistical method, this variability in medication was incorporated in analysis using covariates.
3.3.2 Global DNA methylation (time on clozapine)
Longer time on clozapine was associated with increased global DNA methylation. An additional week on clozapine was associated with an average increase in global DNA methylation of 0. 969x10-3 %. 51.18% of all CpG sites showed an increase in methylation associated with time exposed to clozapine, and 48.82 showed a decrease. Moreover, when exploring individual time points: at baseline, mean global DNA methylation was 48.285% and this decreased to 48.254% at six weeks, but then increased to 48.289% at twelve weeks, and further increased to 48.293% at six months.
- 73 -
Exposure to Clozapine and Associated DNA Methylation Changes
3.3.3 Differentially methylated positions (time on clozapine)
One DMP reached the threshold for epigenome wide significance (p<1*10-7) and eighty five DMPs were trend-level significant (p<5*10-5). Below, top ten DMPs (Table 2) are described in more detail, only when there they are either epigenome wide significant or there is prior evidence for correlation between methylation in blood and brain.
- 74 -
Exposure to Clozapine and Associated DNA Methylation Changes
- 75 -
Exposure to Clozapine and Associated DNA Methylation Changes
Table 2: Top ten DMPs associated with time on clozapine (* indicates epigenome wide significance)
CpG site Beta P value Mean methylation (%) Gene Location MAPINFO Significant blood-brain methylation correlation reported?
Baseline Six Twelve Six weeks weeks months cg23892310 -0.149x10-2 5.86E-08* 83.456 83.776 81.973 81.319 TBCD Body 80816702 Y – entorhinal cortex cg12590655 0.081 x10-2 1.49E-07 84.280 84.030 84.504 85.136 164340525 N 33.242 32.807 33.484 33.853 Y – superior temporal gyrus and cg00873351 0.117 x10-2 6.23E-07 CD300LB TSS1500 72527813 cerebellum cg12026992 0.141 x10-2 1.37E-06 79.915 79.882 80.186 82.427 90004011 N cg15766571 -0.090 x10-2 2.02E-06 9.839 9.942 8.994 7.909 C19orf43 Body 12845030 N cg05842815 0.079 x10-2 2.29E-06 87.111 87.581 87.607 88.727 440159 Y – entorhinal cortex cg25037165 -0.097 x10-2 2.85E-06 85.289 84.691 84.660 83.496 TEAD1 Body 12824283 Y – superior temporal gyrus 8.859 8.934 9.127 10.101 Y – entorhinal cortex and prefrontal cg05195750 0.077 x10-2 3.81E-06 ELP4;IMMP1L Body;TSS1500 31531557 cortex cg14060296 0.026 x10-2 3.82E-06 3.660 3.456 3.752 4.001 HSPA2 TSS200 65007096 N cg01421943 -0.137 x10-2 3.85E-06 81.771 83.111 81.605 80.710 DKK2 3'UTR 107843676 N
- 76 -
Exposure to Clozapine and Associated DNA Methylation Changes
The most statistically significant DMP associated with time on clozapine was cg23892310. Each additional week exposed to clozapine was associated with a decrease in methylation at this site of 0.149% (p=0.6x10-7). Uncorrected means show an average baseline methylation of 83.456%, with an increase to 83.776% at six weeks, then a decrease to 81.973% at twelve weeks and a further decrease to 81.319% at six months.
cg23892310 is located in the gene body of TBCD, a protein coding gene for beta-tubulin cofactor D. TBCD is located on chromosome 17 and expressed across numerous tissue including brain, with ubiquitous expression of mRNA and overexpression of protein in bone. Beta-tubulin cofactor D is one of four proteins involved in ensuring correctly folded beta-tubulin, capturing and stabilising beta-tubulin intermediates. It acts as a GTPase activating protein and is regulated by ARL2; the absence of ARL2 leads to microtubule disruption. It is also implicated in protein metabolism and associated with corneal diseases. This CpG site is reported to have a significant weak correlation between methylation in blood and methylation in prefrontal cortex (r = 0.271, p=0.0195).
- 77 -
Exposure to Clozapine and Associated DNA Methylation Changes
The third most statistically significant DMP associated with time on clozapine was cg00873351. Each additional week of exposure to clozapine was associated with an increase of 0.117% (p=0.6x10- 6). Uncorrected means show an average baseline methylation of 33.242%, with a decrease to 32.807% at six weeks, then an increase to 33.484% at twelve weeks and a further increase to 33.853% at six months.
cg00873351 is located within the TSS1500 (1500 bases upstream of the transcription start site) promoter region of CD300LB, a gene which codes for Leukocyte Mono-Ig-Like Receptor 5, a non- classical activating receptor of the immunoglobin superfamily expressed on myeloid cells. CD300LB is located on chromosome 17 and expressed in numerous tissues including brain, with overexpression of mRNA in whole blood and overexpression of protein in stomach. This CpG site is reported to have a significant weak correlation between methylation in blood and methylation in superior temporal gyrus (r = 0.236, p =0.042) and cerebellum (r=0.247, p=0.038).
- 78 -
Exposure to Clozapine and Associated DNA Methylation Changes
The sixth most statistically significant DMP associated with time on clozapine was cg05842815. Each additional week exposed to clozapine was associated with an increase of 0.079% (p=0.2x10-5). Uncorrected means show an average baseline methylation of 87.111%, with an increase to 87.581% at six weeks, then a further increase to 87.607% at twelve weeks and a further increase to 88.727% at six months.
cg05842815 is located on chromosome 5, outside of any characterised genes. This CpG site is reported to have a significant weak correlation between methylation in blood and methylation in entorhinal cortex (r=0.239, p=0.045).
The seventh most statistically significant DMP associated with time on clozapine was cg25037165. Each additional week exposed to clozapine was associated with a decrease of 0.097% (p= 0.2x10-5). Uncorrected means show an average baseline methylation of 85.289%, with a decrease to 84.691% at six weeks, then a decrease to 84.660% at twelve weeks and a further decrease to 83.496% at six months.
- 79 -
Exposure to Clozapine and Associated DNA Methylation Changes
cg25037165 is located within the gene body of TEAD1, which codes for transcriptional enhancer factor TEF-1. This gene is located on chromosome 11 and is expressed in numerous tissues including brain, with overexpression of mRNA in skeletal muscle and overexpression of protein in breast and ovary. Transcriptional enhancer factor TEF-1 is ubiquitous, directing the transactivation of numerous genes. This CpG site is reported to have a significant weak correlation between methylation in blood and methylation in superior temporal gyrus (r=0.277, p=0.016).
The eighth most statistically significant DMP associated with time on clozapine was cg05195750. Each additional week exposed to clozapine was associated with an increase of 0.0765% (p= 0.4x10-5). Uncorrected means show an average baseline methylation of 8.859%, with an increase to 8.934% at six weeks, then an increase to 9.127% at twelve weeks and a further increase to 10.101% at six months.
- 80 -
Exposure to Clozapine and Associated DNA Methylation Changes
cg05195750 is located with the gene body of ELP4, which codes for elongation protein 4 homolog. This gene is located on chromosome 11 and expressed across numerous tissues including brain, with wide expression of mRNA and overexpression of protein in breast and peripheral blood mononuclear cells. Elongation protein 4 homolog is a histone acetyltransferase complex that is involved in transcriptional elongation, chromatin organisation, and has been associated with epilepsy. This CpG site is reported to have a significant moderate correlation between methylation in blood and methylation in entorhinal cortex (r = 0.38, p=0.001) and a significant weak correlation between blood and prefrontal cortex (r = 0.234, p=0.045).
3.3.4 Pathway analysis (time on clozapine)
Pathway analysis indicated that the pathways in Table 3 were the top thirty pathways with an over-representation of DMPs associated with time on clozapine.
- 81 -
Exposure to Clozapine and Associated DNA Methylation Changes
Table 3: Top thirty over-represented pathways - time on clozapine
Category Name p value over- Number of differentially Number in Ontology representation methylated category GO:0055080 cation homeostasis 0.0006 23 470 BP
GO:0032446 protein modification by small protein conjugation 0.0008 30 640 BP
GO:0033005 positive regulation of mast cell activation 0.0011 3 12 BP
GO:0009312 oligosaccharide biosynthetic process 0.0011 3 13 BP
GO:0006700 C21-steroid hormone biosynthetic process 0.0013 4 18 BP
GO:0070647 protein modification by small protein conjugation or removal 0.0017 31 710 BP
GO:0004842 ubiquitin-protein ligase activity 0.0017 15 258 MF
GO:0033866 nucleoside bisphosphate biosynthetic process 0.0018 4 17 BP
GO:0034030 ribonucleoside bisphosphate biosynthetic process 0.0018 4 17 BP
GO:0034033 purine nucleoside bisphosphate biosynthetic process 0.0018 4 17 BP
GO:0008528 G-protein coupled peptide receptor activity 0.0018 8 106 MF
GO:0001653 peptide receptor activity 0.002 8 108 MF
GO:0006163 purine nucleotide metabolic process 0.002 54 1290 BP
GO:0072521 purine-containing compound metabolic process 0.002 55 1328 BP
GO:0008207 C21-steroid hormone metabolic process 0.003 4 27 BP
GO:0042623 ATPase activity 0.003 15 243 MF
GO:0004871 signal transducer activity 0.003 46 1434 MF
GO:0060089 molecular transducer activity 0.003 46 1434 MF
GO:0055065 metal ion homeostasis 0.003 20 434 BP
- 82 -
Exposure to Clozapine and Associated DNA Methylation Changes
Category Name p value over- Number of differentially Number in Ontology representation methylated category GO:0009150 purine ribonucleotide metabolic process 0.003 53 1269 BP
GO:0004930 G-protein coupled receptor activity 0.003 19 711 MF
GO:0032281 alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid selective glutamate receptor complex 0.003 4 24 CC
GO:0016887 ATPase activity 0.003 19 352 MF
GO:0009259 ribonucleotide metabolic process 0.003 53 1282 BP
GO:0016567 protein ubiquitination 0.003 26 596 BP
GO:0050801 ion homeostasis 0.003 23 535 BP
GO:0019693 ribose phosphate metabolic process 0.003 53 1285 BP
GO:0042181 ketone biosynthetic process 0.004 4 31 BP
GO:0046128 purine ribonucleoside metabolic process 0.004 48 1138 BP
GO:0042278 purine nucleoside metabolic process 0.005 48 1140 BP
- 83 -
Exposure to Clozapine and Associated DNA Methylation Changes
3.3.5 Candidate CpG site methylation (time on clozapine)
Of thirty one CpG sites located within GAD1 with methylation data available, one had a significant association with time on clozapine before but not after correction for multiple testing (0.275x10-3 Of fifteen CpG sites located within PVALB with methylation data available, none had a significant association with time on clozapine. Across all PVALB CpG sites, there was a mean increase in methylation of 0.0159% associated with each additional week on clozapine. Of twenty two CpG sites located within RELN with methylation data available, two had a significant association with time on clozapine before but not after correction for multiple testing (0.275x10-3 Of fifty three CpG sites located within GRIN2A with methylation data available, two had a significant association with time on clozapine before but not after correction for multiple testing (0.275x10-3 Of thirty nine CpG sites located within GRIN2D with methylation data available, three had a significant association with time on clozapine before but not after correction for multiple testing (0.275x10-3 Of thirty two CpG sites located within GRM7 with methylation data available, one had a significant association with time on clozapine before but not after correction for multiple testing (0.275x10-3 - 84 - Exposure to Clozapine and Associated DNA Methylation Changes 3.3.6 Comparison to cross-sectional data (time on clozapine) First I took the top DMPs (p<0.05*10-5) from the cross-sectional dataset and identified these within the longitudinal dataset. There was a significant very weak positive correlation between the regression coefficients from the longitudinal dataset and the cross-sectional data set of 0.053, p < 0.4x10-3, df =4528. I then took the top DMPs (p<0.05*10-5) from the longitudinal dataset and identified these within the cross-sectional dataset. There was a non-significant negative correlation between the regression coefficients from the longitudinal dataset and the cross-sectional data set of -0.024. - 85 - Exposure to Clozapine and Associated DNA Methylation Changes - 86 - Exposure to Clozapine and Associated DNA Methylation Changes 3.4 Results: Clozapine levels 3.4.1 Participant characteristics (clozapine levels) All twenty-two baseline samples with clozapine levels of 0 were included. Four participants had no further samples with corresponding clozapine level data. Of the eighteen other participants, there were forty two samples taken after clozapine use with corresponding clozapine level data. Combined with the baseline samples, this led to a total of sixty four samples included in the below analysis. Clozapine levels significantly increased with time (correlation with days on clozapine =0.688, df=64, p<0.1x10-5) – though when baseline samples are excluded, this disappears (0.126, df=42, p=0.41); this aligns with typical titration schedules in which patients should have reached a therapeutic dose by six weeks and alterations afterwards are individualised adjustments. Clozapine levels at twelve weeks are more variable in smokers (mean in smokers =0.434, sd=0.227, mean in non-smokers =0.403, sd=0.093). Those with lower clozapine levels at twelve weeks had greater reductions in total symptom severity at twelve weeks, though this correlation was not significant (0.251, df =13, p=0.3678). When baseline samples are included, higher clozapine levels are associated with lower total symptom severity (-0.346, df=48, p=0.0138); however, when baseline samples are excluded, there is no longer any association (0.0831, df=26, p=0.674) suggesting lower symptom severity is associated with being on clozapine at any dose and there is not a dose-severity association. - 87 - Exposure to Clozapine and Associated DNA Methylation Changes ) mg/l ( ) mg/l ( - 88 - Exposure to Clozapine and Associated DNA Methylation Changes ) mg/l ( ) mg/l ( - 89 - Exposure to Clozapine and Associated DNA Methylation Changes ) mg/l ( ) mg/l ( - 90 - Exposure to Clozapine and Associated DNA Methylation Changes 3.4.2 Global DNA methylation (clozapine levels) Higher clozapine levels were associated with higher levels of global DNA methylation. An increase in clozapine level was associated with an average increase in global DNA methylation of 0.0501%. 51.02% of all CpG sites showed an increase in methylation associated with clozapine level, and 48.98% showed a decrease. 3.4.3 Differentially methylated positions (clozapine levels) Three DMPs reached the threshold for epigenome wide significance (p<1*10-7), and two- hundred and thirty other DMPs were trend-level significant (p<5*10-5). Below, top ten DMPs (Table 4) are described in more detail, only when there they are either epigenome wide significant or there is prior evidence for correlation between methylation in blood and brain. - 91 - Exposure to Clozapine and Associated DNA Methylation Changes - 92 - Exposure to Clozapine and Associated DNA Methylation Changes Table 4: Top ten DMPs associated with clozapine level (* indicates epigenome wide significance) CpG site Beta P Gene Location MAPINFO Significant blood-brain methylation correlation reported? value 1.32E- N cg24651966 0.060 10* BBS12 Body 123664110 3.74E- Y – prefrontal cortex, entorhinal cortex, superior temporal gyrus cg11293029 0.080 08* SLC12A6 TSS1500 34630530 6.16E- N cg13193591 -0.054 08* 120377997 3.89E- N cg11657122 0.023 07 TNIP1 Body 150444590 4.29E- N cg16658473 -0.096 07 SHISA9 Body 12998870 4.44E- N cg08628431 0.044 07 TRIO Body 14506972 4.56E- Y – prefrontal cortex, entorhinal cortex, superior temporal gyrus cg15158376 -0.011 07 ENOSF1 TSS200 712731 4.79E- Y – prefrontal cortex and cerebellum cg06329022 -0.037 07 SPAG5 TSS1500 26926511 4.81E- Y – prefrontal cortex and superior temporal gyrus cg01731811 -0.061 07 PTPRN2 Body 157890171 5.19E- N cg08821738 -0.051 07 C1QTNF1 Body 77043179 - 93 - Exposure to Clozapine and Associated DNA Methylation Changes The most statistically significant DMP associated with clozapine level was cg24651966. An increase in clozapine level was associated with an increase in methylation at this site of 6.072% (p=0.1x10-9). Using uncorrected means there was a correlation of 0.261 between clozapine levels and methylation across samples. cg24651966 is located in the gene body of BBS12, which codes Bardet-Biedl Syndrome 12 protein. This gene is located on chromosome 4 and expressed across numerous tissues including brain, and defects can cause Bardet-Biedl Syndrome 12. Bardet-Biedl Syndrome 12 protein is a molecular chaperone involved in membrane trafficking and protein folding, as well as adipocyte differentiation. This CpG site is reported to show no correlation between methylation in blood and methylation in brain tissue. The second most significant differentially methylated CpG site associated with clozapine level was cg11293029. An increase in clozapine level was associated with an increase in methylation at - 94 - Exposure to Clozapine and Associated DNA Methylation Changes this site of 8.002% (p=0.4x10-7). Using uncorrected means there was a correlation of 0.352 between clozapine levels and methylation across samples. cg11293029 is located in the TSS1500 promoter region of SLC12A6, which codes solute carrier family 12 member 6. This gene is located on chromosome 15 and expressed across numerous tissues including brain, with overexpression of mRNA in whole blood and overexpression of protein in peripheral blood mononuclear cells, plasma and lymph nodes. Solute carrier family 12 member 6 is a member of the potassium-chloride cotransporter family; it is an integral membrane protein that lowers intracellular chloride to induce electrochemical changes. This CpG site is reported to have a significant moderate correlation between methylation at blood and methylation at prefrontal cortex (r=0.377, p=0.001), entorhinal cortex (r=0.313, p=0.008) and superior temporal gyrus (r=0.344, p=0.003). The third most significantly differentiated methylated CpG site associated with clozapine level was cg13193591. An increase in clozapine level was associated with a decrease in methylation at this - 95 - Exposure to Clozapine and Associated DNA Methylation Changes site of 5.409% (p=0.6x10-7). Using uncorrected means there was a correlation of -0.283 between clozapine levels and methylation across samples. cg13193591 is located on chromosome 11 outside of any characterised genes. This CpG site is reported to show no correlation between methylation in blood and methylation in brain tissue. The seventh most statistically significant DMP associated with clozapine level was cg15158376. An increase in clozapine level was associated with a decrease in methylation at this site of 1.106% (p=0.4x10-6). Using uncorrected means there was a correlation of -0.256 between clozapine levels and methylation across samples. - 96 - Exposure to Clozapine and Associated DNA Methylation Changes cg15158376 is located in the TSS200 of ENOSF1, which encodes antisense RNA to thymidylate synthase and is believed to regulate expression of the thymidylate synthase locus. The gene is located on chromosome 18 and expressed across numerous tissues including brain, with overexpression of protein in gallbladder and liver. This CpG site is reported to show a significant moderate correlation between methylation in blood and methylation in prefrontal cortex (r=0.439, p=0.1x10-3), entorhinal cortex (r=0.473, p=0.3x10-4), and superior temporal gyrus (r=0.327, p =0.004). The eighth most statistically significant DMP associated with clozapine level was cg06329022. An increase in clozapine level was associated with a decrease in methylation at this site of 3.654% (p=0.5x10-6). Using uncorrected means there was a correlation of -0.324 between clozapine levels and methylation across samples. - 97 - Exposure to Clozapine and Associated DNA Methylation Changes cg06329022 is located in the TSS1500 promoter region of SPAG5, which encodes sperm associated antigen 5. This gene is located on chromosome 17 and expressed across numerous tissues including brain, with overexpression of mRNA in testis and kidney and overexpression of protein in cervix, plasma, testis and bone. Sperm associated antigen 5 is implicated in regulator of mitotic spindles required for normal chromosome segregation. This CpG site is reported to show a significant weak correlation between methylation in blood and methylation in pre-frontal cortex (r=0.26, p=0.026) and cerebellum (r=0.27, p=0.023), with a trend at entorhinal cortex (r=0.217, p=0.069). The ninth most statistically significant DMP associated with clozapine level was cg01731811. An increase in clozapine level was associated with a decrease in methylation at this site of 6.091% (p=0.5x10-6). Using uncorrected means there was a correlation of -0.378 between clozapine levels and methylation across samples. - 98 - Exposure to Clozapine and Associated DNA Methylation Changes cg01731811 is located in the gene body of PTPRN2, which encodes protein tyrosine phosphatase, receptor type N2. This gene is located on chromosome 7 and expressed across numerous tissues including brain, with overexpression of mRNA in brain (cortex, anterior cingulate cortex, frontal cortex, nucleus accumbens and putamen) and overexpression of protein in cerebrospinal fluid and frontal cortex. Protein tyrosine phosphatase, receptor type N2 is required for normal accumulation of neurotransmitters in the brain, as well as secretion of pituitary hormones and secretory vesicles in hippocampus and insulin secretion. This CpG site is reported to show a significant moderate correlation between methylation in blood and methylation in prefrontal cortex (r=0.491, p=0.9x10-5) and superior temporal gyrus (r=0.471, p=0.2x10-4), with a weak correlation between methylation in blood and methylation in entorhinal cortex (r=0.254, p=0.032). 3.4.4 Pathway analysis (clozapine levels) Pathway analysis indicated that the pathways in Table 5Table 3 were the top thirty pathways with an over-representation of DMPs associated with clozapine level. - 99 - Exposure to Clozapine and Associated DNA Methylation Changes Table 5: Top thirty over-represented pathways - clozapine level Category Name p value over- Number of differentially Number in Ontology representation methylated category GO:0010035 response to inorganic substance 0.0003 40 361 BP GO:0051219 phosphoprotein binding 0.0009 10 47 MF GO:0010038 response to metal ion 0.0011 28 242 BP GO:0031929 TOR signalling 0.0021 11 46 BP GO:0032007 negative regulation of TOR signalling 0.0024 6 20 BP GO:0008092 cytoskeletal protein binding 0.0030 71 674 MF GO:0030100 regulation of endocytosis 0.0046 19 151 BP GO:0016712 oxidoreductase activity 0.0055 4 24 MF GO:2000505 regulation of energy homeostasis 0.0057 4 11 BP GO:0042743 hydrogen peroxide metabolic process 0.0064 7 38 BP GO:0045807 positive regulation of endocytosis 0.0066 12 81 BP GO:0007140 male meiosis 0.0067 6 33 BP GO:0033574 response to testosterone 0.0080 5 25 BP GO:0035270 endocrine system development 0.0082 19 131 BP GO:0030865 cortical cytoskeleton organization 0.0088 5 20 BP GO:0048168 regulation of neuronal synaptic plasticity 0.009 9 41 BP GO:0045132 meiotic chromosome segregation 0.009 5 22 BP GO:0010951 negative regulation of endopeptidase activity 0.001 17 194 BP GO:0007167 enzyme linked receptor protein signalling pathway 0.001 90 928 BP - 100 - Exposure to Clozapine and Associated DNA Methylation Changes Category Name p value over- Number of differentially Number in Ontology representation methylated category GO:0035412 regulation of catenin import into nucleus 0.011 6 22 BP GO:0043566 structure-specific DNA binding 0.011 23 211 MF GO:0009651 response to salt stress 0.011 5 24 BP GO:0006595 polyamine metabolic process 0.011 4 17 BP GO:0035413 positive regulation of catenin import into nucleus 0.011 4 11 BP GO:0006997 nucleus organization 0.011 11 87 BP GO:0035066 positive regulation of histone acetylation 0.012 4 17 BP GO:0003779 actin binding 0.012 38 342 MF GO:0043542 endothelial cell migration 0.012 17 120 BP GO:0050321 tau-protein kinase activity 0.012 4 12 MF GO:0009308 amine metabolic process 0.012 19 196 BP - 101 - Exposure to Clozapine and Associated DNA Methylation Changes 3.4.5 Candidate CpG site methylation (clozapine levels) Of thirty one CpG sites located within GAD1 with methylation data available, four had a significant association with clozapine level before but not after correction for multiple testing (0.275x10-3 Of fifteen CpG sites located within PVALB with methylation data available, one had a significant association with clozapine level after correction for multiple testing; cg19136704 showed an increase in methylation (6.358%, p= 0.5x10-4). Across all PVALB CpG sites, there was a mean decrease in methylation of 0.1995% associated with an increase in clozapine level. - 102 - Exposure to Clozapine and Associated DNA Methylation Changes Of twenty two CpG sites located within RELN with methylation data available, two had a significant association with clozapine level before but not after correction for multiple testing (0.275x10-3 Of fifty three CpG sites located within GRIN2A with methylation data available, six had a significant association with clozapine level before but not after correction for multiple testing (0.275x10-3 Of thirty nine CpG sites located within GRIN2D with methylation data available, five had a significant association with clozapine level before but not after correction for multiple testing (0.275x10-3 Of thirty two CpG sites located within GRM7 with methylation data available, two had a significant association with clozapine level before but not after correction for multiple testing (0.275x10-3 3.4.6 Comparison to cross-sectional data (clozapine levels) First I took the top DMPs (p<0.05*10-5) from the cross-sectional dataset and identified these within the longitudinal dataset. There was a highly significant negative correlation between the regression coefficients from the longitudinal dataset and the cross-sectional data set of -0.369, p <0.2x10-15, df =4528. - 103 - Exposure to Clozapine and Associated DNA Methylation Changes I then took the top DMPs (p<0.05*10-5) from the longitudinal dataset and identified these within the cross-sectional dataset. There was a highly significant negative correlation between the regression coefficients from the longitudinal dataset and the cross-sectional data set of -0.331, p = 0.3x10-6, df = 229. - 104 - Exposure to Clozapine and Associated DNA Methylation Changes - 105 - Exposure to Clozapine and Associated DNA Methylation Changes 3.5 Discussion: Exposure to clozapine To explore the effects of exposure to clozapine on DNA methylation, I have conducted mixed model regression analysis on 1) time exposed to clozapine within-individual as a predictor of DNA methylation, and 2) clozapine level within-individual as a predictor of DNA methylation. Systems involving signal transduction, GPCRs, AMPARs, calcium signalling at GABAergic interneurons, glucose homeostasis, immune function, beta-tubulin and transcription regulation are implicated. 3.5.1 Time on clozapine Analysis of the association between epigenome-wide DNA methylation and time exposed to clozapine identified one DMP that was significant at the epigenome-wide threshold (cg23892310) and eighty seven other DMPs showed trends towards significance. The majority of the top ten DMPs show increases in methylation over time, or an initial decrease and then increases; global DNA methylation indicates this pattern also. Three of the top ten DMPs are located outside of genes, but of the seven located within genes, five have been shown to demonstrate correlations between methylation in blood and methylation in at least one brain region tissue, predominantly entorhinal cortex. A high proportion of the biological systems indicated by the top ten DMPs and/or top thirty pathways associated with time on clozapine have been previously demonstrated to be dysfunctional in schizophrenia, and many have also been shown to be influenced by antipsychotics including clozapine. The top ranked DMP was located in the gene body of TBCD, which encodes a protein involved in the correct folding of beta-tubulin. Beta tubulin is a major cytoskeleton protein involved in neuronal differentiation, and a major component of axons, dendrites and dendritic spines – knock down studies highlight its importance for neurite density and protection against glutamate and glycine (Guo, Walss-Bass, & Ludueña, 2010). Post-mortem studies of schizophrenia have found decreased expression of beta-tubulin in ACC and increased expression in DLPFC, and it has been suggested that these abnormalities disrupt normal cell division and differentiation processes (Moehle, Luduena, Haroutunian, Meador-Woodruff, & McCullumsmith, 2012). Some have further argued that abnormal expression of cytoskeletal proteins including beta-tubulin could underlie findings of reduced white matter in schizophrenia and abnormalities in corpus callosum communication (Sivagnanasundaram, Crossett, Dedova, Cordwell, & Matsumoto, 2007). The third ranked DMP was in CD300LB, a gene encoding an activating receptor of the immunoglobin super family, which is highly expressed in mast cells and monocytes/macrophages, - 106 - Exposure to Clozapine and Associated DNA Methylation Changes both types of leukocytes (white blood cells), and involved in cytokine secretion (Enomoto et al., 2010). Similarly, the third most over-represented pathway was regulation of mast cell activation, both indicating an effect of clozapine on DNA methylation in immune related genes. There has been a long history of research implicating immune dysfunction in schizophrenia, with recent formulations suggesting that prenatal or early experiences induce an inflammatory state which leads to later disruptions in dopaminergic, glutamatergic and GABAergic networks (Anders & Kinney, 2015; Feigensona, Kusnecovb, & Silversteina, 2014; Müller, Weidinger, Leitner, & Schwarz, 2015); these reviews cite evidence of associations between schizophrenia and early immunological insults, increased inflammatory markers such as cytokines and microglia, and autoimmune diseases. The literature on clozapine’s interaction with the immune system is complex. On one hand, studies (Hu et al., 2012) demonstrate a protective, anti-inflammatory effect of clozapine; they found that clozapine led to reduced microglial activation and reduced production of microglial derived reactive oxygen or superoxide, and protection of cortical and mesencephalic neuron-glia cultures from inflammation related neurodegeneration. Similarly, Song, Lin, Kenis, Bosmans, & Maes (2000) found an immunosuppressive effect of clozapine, with increased production of an interleukin-T receptor antagonist and interleukin 10 (which inhibits synthesis of cytokines). Of course, clozapine is also known to induce agranulocytosis, a dangerous decrease in leukocytes which indicates an immunosuppressive effect. On the other hand, there are studies (Hefner, Shams, Unterecker, Falter, & Hiemke, 2016) which find that – within individuals - increased doses of clozapine were associated with elevated C reactive protein, which is a non-specific marker of inflammation. Similarly, Fernandez-Egea et al., (2016) found that while clozapine use was associated with decreased numbers of HLA-DR+ regulatory T-cells and CD4+ memory T cells compared to controls, it was also associated with increased numbers of NK cells, naïve B cells, CXCR5+ memory T cells and classical monocytes. The tenth DMP was located within DKK2, a gene encoding a protein which regulates Wnt/beta- catenin signalling. Wnt signalling involves signal transduction into cells via cell surface receptors; there are three Wnt signal transduction pathways which regulate gene transcription (canonical pathway), cell shape within the cytoskeleton (non-canonical planer cell polarity pathway) and intracellular calcium (non-canonical Wnt calcium pathway). While this DMP was not significant and is not indicated to show correlated methylation in blood and brain, clozapine has previously been demonstrated to influence Wnt signalling (Sutton, Honardoust, Mouyal, Rajakumar, & Rushlow, 2007); upregulation of numerous proteins was found in response to clozapine exposure, including dishevelled-3 (replicated in vivo and in vitro) which transduces Wnt signalling for canonical and planer cell polarity pathways, and is associated with the D2 receptor. DKK2 has also been associated - 107 - Exposure to Clozapine and Associated DNA Methylation Changes with Alzheimer’s, and it has been suggested (Inestrosa, Montecinos-Oliva, & Fuenzalida, 2012) that Wnt signalling dysfunction underlies both these disorders, leading to neurotoxicity of amyloid in Alzheimer’s disease and disruption of GABAergic neurons in schizophrenia. Four of the top ten DMPs are located within genes involved in regulation of gene expression – transcription factors, histone acetyltransferase, mRNA splicing and processes involving the nucleobases adenine and guanine – as well as one DMP and several over-represented pathways indicating an involvement in regulation of protein modification. The most overrepresented pathway associated with time on clozapine was cation homeostasis, while another overrepresented pathway was ion homeostasis in general. Cations are positively charged ions, and ion transport is critical to action potentials and nerve impulses - mediating conduction across the synapse - as well as T cell activation, pancreatic beta-cell insulin release, cardiac function and other biological functions. Other over-represented pathways were molecular transducer activity and signal transducer activity; signalling cascades through cells to transmit signals is critical for cell communication, transcription and translation of genes, post translation protein changes, and cell proliferation. Overrepresentation of all these pathways may indicate that a consequence of clozapine exposure is modification of synaptic signalling cascades via changes in DNA methylation. Animal studies have previously demonstrated the effects of clozapine on signal transduction cascades; for example, an animal study (Browning et al., 2005) found that clozapine increased phosphorylation in three mitogen-activated protein kinase (MAPK) signal transduction pathways which mediate glutamate receptor functioning, which appeared to drive changes in behavioural tests of conditioned avoidance response. Pathway analysis specifically indicated over-representation of DMPs within G-protein coupled receptor activity pathways (GPCRs). GPCRs play a pivotal role in post-receptor information transduction; ligands (signalling molecules) bind to the GCPRs, the associated G-protein is then activated and the protein’s alpha subunit dissociates from other subunits, exposing the site to enable interaction with other molecules. There has been considerable interest in GPCR dysfunction in schizophrenia, as there are GPCRs in dopaminergic, serotoninergic, noradrenergic, cholinergic, and glutamatergic pathways, all of which have all been implicated in schizophrenia dysfunction (Catapano & Manji, 2007). In two 2001 studies by the same team, it was demonstrated that untreated patients with schizophrenia have elevated dopamine induced GPCR function and that this is correlated with severity of symptoms but patients treated with clozapine (or haloperidol) do not display this dysfunction (Avissar, Barki-Harrington, et al., 2001; Avissar, Roitman, & Schreiber, 2001). - 108 - Exposure to Clozapine and Associated DNA Methylation Changes This suggests that clozapine acts to reduce and normalise GPCR function, and my analysis suggests that this could be associated with changes in DNA methylation of relevant genes. Pathway analysis also specifically indicated over-representation of DMPs within the alpha- amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) selective glutamate receptor complex pathway. AMPA receptors (AMPARs) are the most commonly found receptor in the nervous system; they are ionotropic transmembrane receptors for glutamate that mediate fast synaptic transmission in the CNS and they are integral to long-term potentiation. Over the last couple of decades there has been considerable interest in modulating AMPARs as a potential treatment for schizophrenia (Danysz, 2002), with numerous patent applications (Ward, Pennicott, & Beswick, 2015), driven by post-mortem studies such as Tucholski et al., (2013) which demonstrate increased GluR4 protein levels and disrupted regulation and localisation of GluR2, both of which are AMPAR subunits. Furthermore, activating AMPARs (and the subsequent depolarisation of the postsynaptic membrane) is a prerequisite for activation of NMDA receptors at glutamate synapses, and inhibiting AMPAR desensitisation enhances dopamine response; thus abnormalities of AMPARs may result in altered NMDA-mediated and dopaminergic transmission, both of which have been found in schizophrenia (discussed more later). Studies in rats have indicated that clozapine influences AMPAR function: Schmitt et al., (2003) found increased AMPAR binding in the striatum and frontal cortex after 21 days and in the anterior cingulate after six months of clozapine exposure, Meador-Woodruff, King, Damask, & Bovenkerk, (1996) found altered mRNA levels for AMPAR subunit genes in hippocampus and entorhinal cortex, and Healy and Meador-Woodruff, (1997) found altered mRNA levels for AMPAR subunit genes – increased GluR7 , decreased GluR3 and decreased GluR4 - in cortex and striatum. Pathway analysis implicated the ubiquitin protein system; this system targets proteins for degradation, vesicle trafficking, trafficking from the endoplasmic reticulum to plasma membrane, and recycling of receptors at cell membrane. Previous genetic studies have associated dysregulation of this system with psychosis (Bousman et al., 2010) and a post-mortem study of schizophrenia found reductions in protein ubiquitination and numerous associated proteins and ligases (Rubio, Wood, Haroutunian, & Meador-Woodruff, 2013); in the latter study, they also tested rats treated with haloperidol and found none of these alterations, indicating that antipsychotic treatment may normalise this disruption of the ubiquitin protein system. Pathway analysis also indicated ATPase activity. ATP, adenosine triphosphate, is a small molecule used and continuously recycled as a coenzyme for intracellular transfer, transporting chemical energy within cells for metabolism. ATP is used in signal transduction, restoration of cell - 109 - Exposure to Clozapine and Associated DNA Methylation Changes membrane ion gradients and transcription. A couple of bio-energetic studies have indicated that ATP abnormalities may be present in schizophrenia. Du et al., (2014) report abnormal storage and energy usage in the prefrontal cortex and a relative increase of glycolysis to ATP synthesis and suggested this would have implications for glutamatergic neurotransmission which is the predominant user of ATP. Similarly Jayakumar et al., (2010) found that antipsychotic naïve patients had high energy metabolism in basal ganglia (indicated by reduced PCr/ATP ratio, as PCr is converted to ATP in ATP consumption), which was reduced and normalised by antipsychotic treatment, with increased PCr/ATP ratios correlating with greater reductions in symptom severity. Finally, the eighth DMP was located within ELP4 which has been associated with epilepsy. This is intriguing as there is evidence of a bidirectional risk between epilepsy and schizophrenia (Fruchter et al., 2014; Wotton & Goldacre, 2012) and the literature has discussed potential common mechanism such as amygdala dysfunction leading to social cognition deficits in schizophrenia and mesial temporal lobe epilepsy (Okruszek et al., 2017), or shared glutamate dysfunction in both, with neuronal hyperexcitation leading to epileptic seizures and prefrontal receptor reduction involved in schizophrenia (Dedeurwaerdere, Boets, Janssens, Lavreysen, & Steckler, 2015). Similarly, one study (Vonberg & Bigdeli, 2016) found a positive association between SNPs associated with schizophrenia and those associated with epilepsy in GWAS data, suggesting common genetic liability. Furthermore, clozapine is known to induce seizures, with one study finding that 1.3% of patients on clozapine in the US have had seizures (Pacia & Devinsky, 1994); however, some case reports have found that using clozapine in patients with both epilepsy and schizophrenia has reduced seizures instead of increasing them as expected (Langosch & Trimble, 2002). One caveat to be aware of, for some of the results throughout this thesis, is that – due to the small sample size – some of the significant findings may be driven by outliers i.e. cpg sites where one individual has a particularly striking change in methylation over the study. cg05195750 is one example where, while there seems to be a general trend of increasing methylation, there is one participant in particular who may be driving the finding of significance. The candidate gene testing provided only weak support for an association between time exposed to clozapine and differential methylation at CpG sites on GAD1, GAD67, RELN, GRIN2A, GRIN2D, and GRM7; no associations survived correction for multiple testing. However, it may be that relevant changes in DNA methylation occur at CpG sites on genes involved in the broader networks for these genes. - 110 - Exposure to Clozapine and Associated DNA Methylation Changes Results from this model provide some replication of the results from previously collected cross- sectional data, with positively correlated direction of effect for their top DMPs. However, results from the cross-sectional data did not provide replication of the results from this model. 3.5.2 Clozapine level Analysis of the association between epigenome-wide DNA methylation and clozapine level identified three DMPs that were significant at the epigenome-wide threshold (cg24651966, cg11293029, cg13193591), and two hundred and thirty other DMPs which showed trends towards significance. The majority of the top ten DMPs show decreases in methylation associated with higher clozapine levels, however higher global DNA methylation was associated with higher clozapine levels. One of the top ten DMPs was located outside of genes, but of the nine located within genes all are expressed in brain tissue to some degree, and six sites are reported to display methylation in blood tissue which correlates to methylation in at least one brain region tissue. Again, a high proportion of the biological systems indicated by the top ten DMPs and/or top thirty pathways associated with clozapine level have been previously demonstrated to be dysfunctional in schizophrenia, and many have also been shown to be influenced by antipsychotics including clozapine. The top DMP was in BBS12, a gene associated with adipocyte differentiation and, the ninth DMP was in PTPRN2, a gene associated with insulin secretion. Adipose tissue plays a crucial role in regulating glucose homeostasis, with altered adiposity associated with insulin resistance and hyperglycaemia (Rosen & Spiegelman, 2006). This finding is perhaps unsurprising as clozapine is known for its metabolic effects and association with type 2 diabetes, discussed further in Adverse Reactions to Clozapine and Associated DNA Methylation Changes. Clozapine dose has previously been associated with fasting levels of glucose (Wysokiński & Kłoszewska, 2014), and animal studies suggest that clozapine stimulates insulin and glucagon secretions and blocks the typical feedback loop of increased glucose levels reducing glucagon secretions (Smith et al., 2014). As with the immune system, the challenge (particularly for future drug development) is to understand whether the effect of clozapine on glucose homeostasis is critical to therapeutic efficacy or a secondary effect with no benefit. BBS12 is also associated with Bardet-Biedl syndrome 12. This is a ciliopathic genetic disorder which causes widespread effects and is primarily characterised by obesity, retinitis pigmentosa, polydactyly, hypogonadism and renal failure. Interestingly, a missense mutation on a separate gene - - 111 - Exposure to Clozapine and Associated DNA Methylation Changes PCM1 – has been found to be associated with schizophrenia as well as Bardet-Biedl syndrome, and in mouse brain has been downregulated by clozapine but not haloperidol (Datta et al., 2010). The second DMP was located within SLC12A6, a gene that is part of the potassium-chloride cotransporter family; these proteins are neuron specific symporters which establish the chloride ion gradient by lowering intracellular chloride concentrations, critical for signal transduction which – as discussed above – is known to be disrupted in schizophrenia. This family of proteins specifically mediates synaptic inhibition and protects against excitotoxicity. PTPRN2, mentioned above, is also involved in regulating normal accumulation of neurotransmitters in the brain. Furthermore, the tenth DMP is located within C1QTNF1, a gene encoding a GPCR interacting protein, again implicating GPCR function. While not statistically significant or indicated to show correlated methylation between brain and blood, the fifth DMP, located within SHISA9 is associated with AMPAR desensitisation, and the sixth DMP, located within TRIO which encodes a protein which regulates AMPAR synaptic function, also again implicate AMPAR function. Pathway analysis also indicated over-representation of DMPs within enzyme-linked receptor protein signalling pathways; enzyme- linked receptor proteins include a number of neurotrophic factors, such as BDNF which has been associated with schizophrenia numerous times. Cumulatively, these results indicate an association between clozapine level and DNA methylation of genetic loci involved in numerous aspects of signal transduction. Pathway analysis indicated overrepresentation of DMPs in regulation of neuronal synaptic plasticity as well as cytoskeletal organisation and TOR signalling, which regulates cell growth in response to environmental cues. SHISHA9, mentioned above, regulates neuronal synaptic plasticity in the dentate gyrus (the section of the hippocampus where progenitors arise and new neuronal cells accumulate) and TRIO is also involved in development of hippocampal neurons. Focusing on the hippocampus, one of the few areas with continued neurogenesis into adulthood, a significant body of evidence from both in vivo and post mortem studies has indicated abnormalities in schizophrenia, from the prodromal stage on (Harrison, 2004); in particular, schizophrenia is associated with alterations in the organisation and functioning of neural circuits within the hippocampus, and connection between the hippocampus and other structures such as the prefrontal cortex. PTPRN2 is also specifically involved in hippocampal secretions. The hippocampus is part of the limbic system, interacting with numerous structures and receiving input from serotonin, noradrenaline, dopamine, and GABA; it plays a critical role in the consolidation of memory, which may be why hippocampal dysfunction has a stronger association with neuropsychological impairment than with psychotic symptoms (Harrison, 2004). There have been mixed findings for the effect of clozapine on hippocampal neurogenesis: some studies have found no effect on hippocampal volume or newly - 112 - Exposure to Clozapine and Associated DNA Methylation Changes dividing cells (Arango, Buchanan, Breier, MacMahon, & Carpenter, 2003; A Schmitt et al., 2003), some have found a short term increase in cell proliferation within the dental gyrus but no survival and integration of these newly generated cells (Halim, Weickert, McClintock, Weinberger, & Lipska, 2004), while others have found an increase in post synaptic protein density and dendritic spine density (Critchlow et al., 2006). Pathway analysis also indicated regulation of endocytosis, the process of cells interacting with environments via control of the lipid and protein composition of the plasma membrane, which has been previously linked to schizophrenia; in particular, Schubert, Föcking, Prehn, & Cotter, (2012) proposed that dysfunction of clathrin-mediated endocytosis (along with other clathrin-dependent processes) could be a core pathology which explains schizophrenia-related alterations in synaptic function, white matter and neurodevelopment. Pathway analysis indicated tau-protein kinase activity; tau protein kinase is an enzyme which catalyses the reaction between ATP (discussed above) and tau protein and it is involved in numerous pathways including – among others - Wnt signalling, axon guidance and insulin signalling, all of which have been associated with dysfunction in schizophrenia. The seventh DMP is located within ENOSF1, a gene encoding antisense RNA and associated with gene expression regulation, the eight DMP is within SPAG5, a gene associated with regulation of mitotic spindles required for normal chromosome segregation (mitotic spindles are a cytoskeletal structure involved in cell division, which are formed of microtubules polymerised from beta-tubulin, discussed above), and pathway analysis indicated overrepresentation of chromosome segregation and histone acetylation; all indicate regulation of gene expression and chromosomal structure. TRIO, mentioned above, has also been associated with intellectual disability. Numerous studies have reported associations between schizophrenia and intellectual disability: higher rates of schizophrenia within the intellectual disability community (Deb, Thomas, & Bright, 2001; MORGAN & Al, 2008), higher rates in relatives of those with intellectual disability (Greenwood, Husted, Bomba, Hodgkinson, & Bassett, 2004), and de novo mutations associated with schizophrenia have overlapped with genes implicated in intellectual disability and vice versa, indicating a shared genetic aetiology (Mccarthy et al., 2014; Rees et al., 2016). Owen, Donovan, Thapar, & Craddock, (2011) have proposed a neurodevelopmental hypothesis of schizophrenia linking the two, in which copy number variants impact upon brain development and various factors influence the severity and form of the impairment guiding individuals towards schizophrenia and/or intellectual disability. The candidate gene testing provided only weak support for an association between clozapine level and differential methylation at CpG sites on GAD1, RELN GRIN2A, GRIN2D, and GRM7; no - 113 - Exposure to Clozapine and Associated DNA Methylation Changes associations survived correction for multiple testing. However, there was support for an association between clozapine level and a CpG site in the TSS200 promoter region of PVALB, cg19136704. This DMP was found to increase in methylation with increased clozapine level. This is one of the few positive results for candidate testing within this thesis. PVALB encodes parvalbumin, a calcium binding protein present in GABAergic interneurons which is essential for generating high-frequency cortical gamma oscillations. Increasing evidence suggests these interneurons are dysfunctional in schizophrenia; Stedehouder & Kushner, (2016) present a review outlining some of the findings from in vivo, post mortem, rodent model and genetic studies of decreased protein, transcriptional changes, EEG abnormalities, abnormal maturation and CNVs implicating parvalbumin interneuron dysfunction as the major GABAergic dysfunction in schizophrenia. Similarly Gonzalez-Burgos, Cho, & Lewis, (2015) present a review of evidence from clinical, basic neuroscience and post-mortem studies implicating parvalbumin interneuron dysfunction in schizophrenia, specifically highlighting their role in cognitive and perceptual functions. While clozapine has previously been shown to regulate gene expression in rodent models of schizophrenia, the direction of effect in my analysis appears to be a different direction of effect to previous reports: Aoyama et al., (2013) had reported that clozapine increased prefrontal mRNA expression of PVALB which would typically indicate reductions in methylation of promoter regions instead of an increase. Results from this model provide some replication of results from previously collected cross- sectional data, with highly significant negatively correlated direction of effect for their top DMPs. Similarly, results from the cross-sectional data provided some replication of the results from my longitudinal clozapine levels analysis, with highly significant negatively correlated direction of effect for my top DMPs. The negative direction of effect is unexpected: it could indicate low levels of clozapine in the cross-sectional sample (blood may have been taken very early in clozapine titration, or the sample may have a bias towards patients with low levels of clozapine), or it could reflect differing methylation patterns over time (blood may have been taken before six weeks or after six months of clozapine exposure, and this may show a different relationship with clozapine level). 3.5.3 Summary To summarise, of the two analyses, one epigenome-wide significant hit came from the time on clozapine model - cg23892310, located in the gene body of TBCD (involved in beta tubulin)- while three epigenome-wide significant hits came from the clozapine levels model - cg24651966 which is located within the gene body of BBS12 (associated with adipocyte differentiation), cg11293029 which is located within the TSS1500 region of SLC12A6 (a neuron specific potassium chloride cotransporter which establishes chloride ion gradients for signal transduction), and cg13193591 - 114 - Exposure to Clozapine and Associated DNA Methylation Changes which is located outside of any genes. The only candidate gene hit was in the clozapine levels model – an increase in methylation at cg19136704 in PVALB. Also, the top regression coefficients from the levels model were significantly associated with previous cross-sectional data on clozapine exposure – whereas the time model was not associated – though this was a negative association, and it is unclear what this means. Overall this indicates a stronger relationship between changes in DNA methylation and clozapine level than time on clozapine. While no individual DMPs overlapped, some of the same biological systems were implicated in both analyses – signal transduction, especially ion function, GPCRs, AMPAR, immune function and regulation of gene expression and chromosomal organisation – all of which have been implicated in the pathology of schizophrenia, and the majority of which have been previously indicated to be affected by clozapine use. In the time model, the majority of top ranked DMPs showed increases in methylation or initial decreases then increases and this was true globally; in contrast, in the levels model, the majority of top ranked DMPs showed decreases in methylation, whereas globally there were still increases. The lack of direct overlap in individual DMPs and differing direction of methylation changes indicates that these analyses were measuring different variables – the analysis of time on clozapine is telling us about associations with the length of time exposed to clozapine whereas the analysis of clozapine level is telling us about associations with the amount of clozapine individuals are exposed to - and indeed, after baseline, clozapine level was not correlated with time on clozapine, so we should not necessarily expect a direct overlap. Of course, these are still not independent analyses and cannot act as replications of one another. Instead, these two analyses are intended to complement one another as different measures of clozapine exposure, with shared results indicating trajectories of change which are not associated solely with time passing but also directly associated with clozapine levels. Clozapine exposure appears to lead to changes in DNA methylation of genes involved in signal transduction, receptor function, immune function and gene regulation, which may lead to altered protein production and functional changes in these systems. However, one of the challenges in studying clozapine is that clozapine is known to have incredibly wide ranging effects and interactions and not all of these will be relevant to their therapeutic effect; some may be relevant to their undesired adverse outcomes, or some may be completely irrelevant. Therefore, analysis which focuses on changes which are associated with clinical outcomes is necessary. In the next chapter, I present my analysis on changes in DNA methylation associated with numerous measures of therapeutic response. - 115 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4 Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4.1 Background: Therapeutic response to clozapine Before reporting on my analysis investigating changes in DNA methylation associated with therapeutic response to clozapine, I will explore in greater depth previous research on clozapine response. First I will discuss the issue of defining response to clozapine. I will then move on to a review of literature on physiological changes and genetic predictors associated with clozapine response, based on a systematic review I co-conducted. 4.1.1 Defining response to clozapine One of the first challenges encountered when studying therapeutic response to antipsychotics is the heterogeneity in approaches to measuring response. Two systematic literature reviews I conducted demonstrate this, and highlight that there are two core areas of discussion. Firstly, which symptoms or outcome should be measured as a measure of response? In a review of all studies reporting on prospective predictors of clozapine response (Appendix III) the most commonly used primary outcome measure was the Brief Psychiatric Rating Scale (BPRS), however nine other measures were also used (in order of prevalence), with different focuses on positive or negative symptoms: Global Assessment of Functioning Scale (GAF/GAS), Clinical Global Impression (CGI), Positive and Negative Syndrome Scale (PANSS), Scale for the Assessment of Negative Symptoms (SANS), Scale for Assessment of Positive Symptoms (SAPS), Bunney-Hamburg Global Psychosis Rating, Schedule for Affective Disorders and Schizophrenia-Lifetime (SADS-L1), a positive symptom subscale (BPOS), a negative symptom subscale (BNEG), and two non-specific clinical interviews. More broadly, some argue we should shift to a focus on functionality which is often prioritised by the patient and carers or loved ones; however, this has less established forms of measurement. It is unclear how this relates to symptom severity; whether there is a threshold above which functionality is impossible, or below which functionality is inevitable. One option is studying pragmatic outcomes. In a review that I conducted of studies comparing treatment-resistant and treatment-responsive patients (Gillespie et al., 2017) some definitions of responsiveness required discharge from hospital. Another option is time to discontinuation or rehospitalisation, though one could argue this is a negative approach predicated on the expectation of relapse rather than hopeful for improvement. Similarly, cognition is increasingly being recognised as critical to recovery and functional outcomes, but as cognitive deficits are rarely addressed with antipsychotic treatment and - 116 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes do not abate with episodes in the way that psychotic symptoms do, there lacks an evidence base for defining remission via cognitive improvement. Secondly, should we measure change from baseline, or the passing of a threshold (which may be defined as remission)? Both of the above reviews found a range of approaches; a percentage reduction was the most common criteria for being deemed responsive, though some specified absolute threshold alternatively or additional (e.g. BPRS total scores ≤35). There are also decisions about whether to study response as a continuous variable or a categorical variable and, if a categorical variable, where to draw the line between who is classified a responder and who is classified a non-responder. For just the BPRS, the percentage change threshold for demonstrating response varied from 15-50%, though 20% was the most common. Opinions also differ on when response should be classified: the length of the clozapine trial before classifying response similarly varied from as low as three weeks to sixteen months. A review (Suzuki, Uchida, Watanabe, & Kashima, 2011) specifically focused on response to specific antipsychotic trials in the context of treatment-resistant schizophrenia proposed that treatment response should be defined “based on a CGI-Change score of ≤ 2, a ≥ 20% decrease on the total PANSS or Brief Psychiatric Rating Scale (BPRS) scores, and an increase of ≥ 20 points on the FACT-Sz or GAF.” This definition is largely based on previous studies, especially landmark ones such as the original one by Kane in 1988, building upon what is common to allow cross-study comparison. They justify their choice to stick with percentage reduction as most appropriate for treatment- resistant patients; in contrast to first-episode patients, absolute remission is very rare and the percentage reduction approach allows a reasonable number of patients to be classified as responders which is necessary for group comparisons. Moreover, studying change from baseline prevents patients who are simply less symptomatic being classified as responders because they are close to the threshold for response and can pass this without showing a notable response to medication. They also stress the importance of considering what can be realistically expected to change within the chosen time frame and suggest that this is why there has been a focus on positive symptoms, as these are known to be more amenable to measurable, rapid change in response to antipsychotics. They advocate for 6-week trials, based upon evidence that improvements can be seen in the early stages, though note that it can take longer for a clear response to be observed and that this may be particularly true in refractory patients. - 117 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4.1.2 Physiological changes associated with response to clozapine Note: This section is based on a systematic review I co-conducted and which has been submitted for publication. See Appendix III for details on methodology. A number of studies have measured physiological variables and symptoms before the initiation of clozapine as well as after a period of clozapine treatment, allowing measurement of changes associated with response to clozapine, however the findings are mixed. A couple of studies have looked at perfusion, with Ertugrul et al., (2009) reporting that across patients there was an increase in right frontal/caudate perfusion ratio, but in responders to clozapine there was an increase in ratio bilaterally with greater increases in bilateral frontal perfusion. Across patients there was also increases NAA/Cr ratio in the DLPFC. However, Rodriguez et al., (1997) found that responders to clozapine showed decreases in perfusion not only in the thalamus and basal ganglia but also dorsolateral prefrontal cortex and anterior pre-frontal cortex. Regarding neurotransmitter metabolites, Szymanski et al., (1993) reports no significant change in cerebrospinal fluid measures of HVA or 5-HIAA over the length of clozapine treatment- though an increase in the former and decrease in the latter can be observed, and the ratio seems to be significant, with a greater change in the ratio over treatment associated with response to clozapine. However, Pickar et al., (1992) reports decreases in HVA, particularly for responders (alongside decreases cortisol, and increases in prolactin and noradrenaline) and Green et al., (1993) also reports that HVA decrease in responders (while plasma dopamine and noradrenaline increase in all patients during clozapine treatment). In addition, Fleischhaker, Schulz and Remschmidt (1998) report increased serum serotonin and noradrenaline in all patients, and increased MHPG and noradrenaline in responders, and Evins et al., (1997) report increases in glutamate and aspartate after clozapine treatment, with the latter associated with improvements in negative symptoms. Ertugrul et al., (2007) report increases in monoamine oxidase B MAO-B and plasma serotonin, but decreases in platelet serotonin, highlighting the relevance of the tissue used to measure. However, many of these are small samples, with all bar one n<21. In summary, there are a lack of studies measuring longitudinal changes associated with clozapine response and those which have been conducted indicate potential longitudinal changes in perfusion, HVA, serotonin, MHPG, cortisol norepinephrine, glutamate, and aspartate associated with clozapine response but there is little consistency or replication. - 118 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4.1.3 Genetic variation associated with response to clozapine Note: This section is based on a systematic review I co-conducted and which was published after thesis submission and before corrections (Samanaite, Gillespie et al, 2018). See Appendix III for details on methodology. While the identification of genetic variants associated with clinical outcomes in response to clozapine does not directly tell us about what changes are occurring during clozapine treatment that are associated with clinical outcomes, they can direct us towards genomic locations and (more generally) biological pathways in which we might expect changes to be occurring. In particular, genes associated with response to clozapine are prime candidates for where we may expect to see differential DNA methylation, as clozapine interacts with the expression of these genes. Below outlines the adapted results of a systematic review I conducted in collaboration with a colleague, identifying all gene-association studies of clozapine response. The sixty-one studies identified included analysis of two-hundred-and-fifty-nine different gene variants. Two-hundred-and-forty-two of these were SNPs, sixteen were repeats or small insertion/deletions and one was a large chromosomal deletion (22q11.2). All of the identified studies took a candidate-gene approach, therefore suffering from the limitations described in the introduction to this thesis; no GWAS were identified. Out of the two-hundred-and-fifty-nine different genetic variants studied, thirty-seven variants (14%) have had significant findings, and eight of these have been replicated. Six of these eight replications were by an independent research group, but four of those six have also had failed replications. I limit comment to findings with at least one replication, as those without replication are at increased likelihood of being spurious findings. The details for all genetic studies, including those with non-significant or non-replicated findings, are provided in the appendices to the publication. Twelve SNPs within the DRD3 gene, encoding the D3 dopamine receptor, have been investigated in eight studies. Six studies only investigated the Ser9Gly polymorphism of rs6280 (Arranz et al., 2000; Barlas et al., 2009; Hwang et al., 2010; Malhotra et al., 1998; Potkin et al., 2003; Scharfetter et al., 1999; Shaikh et al., 1996; Xu et al., 2016) but Hwang et al., (2010) also reported on eight other SNPs in DRD3 and Xu et al. reported on an additional three others. While two studies independently replicated a finding of the Gly allele of the Ser9Gly polymorphism being associated with good response to clozapine (Scharfetter et al., 1999; Shaikh et al., 1996), all six subsequent studies found non-significant results. - 119 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Dopamine receptors are prevalent throughout the central nervous system. They are a class of G- protein-coupled receptors meaning their activity is mediated by G-proteins which inhibit adenylyl cyclase. There are five types (D1-D5), each encoded by different genes. The DRD3 gene encodes the D3 subtype, which is localised to limbic areas of the brain with maximum expression in the ventral striatum and nucleus accumbens, and associated with cognitive, emotional and endocrine functions. DRD3 mediates inhibitory neurotransmission and has strong affinity for both first and second generation antipsychotic drugs. The DRD3 gene is located on chromosome 3 and contains five exons. This gene can be alternative spliced resulting in multiple transcript variants encoding different isoforms, some of which would decay via nonsense-mediated decay. The Ser9Gly polymorphism is a single nucleotide polymorphism (rs6280) in exon 1 which can result in a serine-glycine substitution in the N terminus (codon 9) of the gene. Five SNPS within 5-HTR2A have been investigated in ten studies. The His allele of His452Tyr has been associated with good response to clozapine in four studies by two research groups (Arranz, Collier, Munro, & Sham, 1996; Arranz et al., 1998, 2000; Masellis et al., 1998). Two studies did not detect this association (Malhotra et al., 1996; Nöthen et al., 1995). The T allele of T102C has also been associated with good response to clozapine in three studies by a single research group (Arranz et al., 1995, 2000; Sodhi et al., 1995); however, five studies by other groups have failed to replicate these findings (Lin et al., 1999; Malhotra et al., 1996; Masellis et al., 1998; Nöthen et al., 1995; Potkin et al., 2003). The G-1438A SNP has also been shown to significantly predict clozapine response in two studies by the same group (Arranz et al., 2000; Arranz, Munro, & Owen, 1998). Again, there is no independent replication, with null findings reported in the second sample analysed by the same research group and from an independent research group (Masellis et al., 1998). The 5HTR2A receptor is a subtype of the 5HT2 receptor, the main excitatory receptor subtype among the serotonin receptors, and is another G-protein-coupled receptor. When stimulated, the G- protein subunits dissociate to imitate downstream effects. It enhances glutamate release, as well as having a complex range of interactions with the 5-HT1A, GABA, adenosine A1, AMPA, GluR2/3, mGlu5, and OX2 receptors. The receptor is encoded by the 5-HTR2A gene, located on chromosome 13, which includes three exons and two intros, with four transcription initiation sites and two alternative promoters, with a silencer element downstream from the second promoter element. It is expressed widely throughout the central nervous system, particularly the prefrontal, parietal and somatosensory cortex. The gene has over 250 polymorphisms, many of scientific interest, including His452Tyr (rs6314). His452Tyr is a SNP on exon 3 of the gene which can change between a C and a T, resulting in a missense substitution – switching between histidine and tyrosine. - 120 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Four SNPS within 5-HTR3A have been studied; the only SNP which has been studied twice is rs1062613, but one study found that good response to clozapine was associated with the T allele (Rajkumar et al., 2012) while the other found that good clozapine response was associated with the C allele (Souza, de Luca, Meltzer, Lieberman, & Kennedy, 2010) Five SNPs within the 5HTT (or SLC6A4) gene have been investigated in four studies by four independent groups , with the only independently replicated finding for an association of the HTTLPR polymorphism at rs25531 with clozapine response; Kohlrausch et al., (2010) found an association between good response and the long allele, but M. J. Arranz et al., (2000) do not report the direction of effect. 5-HTT is the sodium-dependent serotonin transporter, a monoamine transporter protein that transports serotonin to the presynaptic neurons from the synaptic cleft - thus terminating the action of serotonin - and recycles it. To function properly it requires membrane potential created by sodium-potassium adenosine triphosphates and therefore is dependent upon the concentration of potassium ion in the cytoplasm and the concentrations of sodium and chloride ions in the extracellular. It is encoded by the SLC6A4 gene. The 5-HTTLPR polymorphism at rs25531 is a degenerate repeat polymorphism within the promoter region of this gene. It is often repeated as the short or long variant, but it can be subdivided further, with one study identifying 14 allelic variants. The only other replicated finding is an association between the C allele of the C825T polymorphism in the G-protein-beta 3 (GNB3) gene and good clozapine response in two studies conducted by independent research groups (Kohlrausch et al., 2008; Müller et al., 2005). G-proteins intracellularly couple with most monoaminergic receptors and, upon receptor occupancy, activate to pass on information, integrating signals between these receptors and effector proteins which regulate protein activity; they are essential to converting external receptor signals into intracellular response. G-proteins are composed of an alpha, beta and gamma subunit, which are encoded by related genes, with the GNB3 gene encoding a beta subunit which regulates alpha subunits. C825T is a single nucleotide polymorphism in exon 10 of the GNB3 gene, and the T allele has been associated with a splice variant causing deletion of nucleotides in exon 9 and subsequent loss of amino acids within highly conserved gene units. This shorter protein is associated with increased intracellular signalling. This polymorphism has also been studied in relationship to antipsychotic induced weight gain, with findings that the T allele is associated with higher cholesterol in Japanese participants (Ishikawa et al., 2000). One meta-analysis of three published and two unpublished studies, cumulatively including 402 patients, found that the T allele had a trend for association with increased weight gain - 121 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes under a fixed-effects model, but when analysed under a random-effects model to account for significant heterogeneity, no significant association was found, even at a trend level (Souza et al., 2008). The five studies used samples from different ethnic populations, but stratifying by ethnicity did not reveal any significant associations either. They also conducted a much larger analysis of association with BMI in a non-psychiatric population, but similarly only found a trend under a fixed model, though this was in the same direction. Since then, one study of olanzapine use found an association with weight gain but another did not (Bishop, Ellingrod, Moline, & Miller, 2006; Park et al., 2009). Two studies have found significant gene-gene interactions (Hwang et al., 2011; Souza, Romano- Silva, et al., 2010) and seven have found significant haplotypes (Hwang et al., 2010; Hwang et al., 2005, 2006, 2007; Mitjans et al., 2015; Souza, et al., 2010), but none of these have been reported on more than once and therefore remain unreplicated. 4.1.3.1 Summary of genetic variation associated with response to clozapine Despite numerous studies associating genetic polymorphisms with clozapine response there has been little replication in independent cohorts and from independent research groups. Those conducted have mainly focused on polymorphisms within genes encoding dopamine or 5-HT receptors, with some significant findings. Many studies also investigated SNPs in relatively small patient samples for genetic studies. Of the two-hundred-and-fifty-nine polymorphisms investigated in relation to clozapine response a priori, replication by two or more independent research groups is available for the DRD3 Ser9Gly, 5HTR2A His452Tyr, 5HTT rs25531, and GNB3 C825T SNPs. However, findings of no association with clozapine response were also reported for DRD3, 5HTR2A His452Tyr and no findings were replicated by more than two independent groups. This starkly highlights the previously mentioned issues regarding candidate gene studies. In contrast to analysis investigating only single variants, a combination of six polymorphisms (T102C and His452Tyr of HTR2A gene, G-330T/ C-244T repeat and Cys23Ser of HTR2C gene, HTTLPR of SLC6A4 gene, G- 1018A of HRH1) predicted clozapine response with the retrospective positive predictive value of 76.7% and a sensitivity of 95% (Arranz et al., 2000) on which basis a pharmacogenetic test was developed, although it is no longer available. Since the majority of these studies were done, technology has advanced to GWAS. While this approach is beginning to be applied to identify SNPs contributing to response to non-clozapine antipsychotics, GWAS of clozapine response have yet been published. - 122 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4.1.4 Epigenetic variation associated with response to clozapine The only published study which has previously analysed epigenetic variation associated with response to clozapine was described above (Kinoshita et al., 2017). The methodological and analytical concerns raised above should again be borne in mind when interpreting these results, but they report finding an association between increased DNA methylation of the CREBBP (CREB binding protein) gene at cg05151055 and greater clinical improvement in response to clozapine, measured by reduction in PANSS as a continuous variable. The authors noted that CREBBP is a protein involved in histone acetyltransferase activity as well as stabilising protein interactions within the transcription complex, and pathway analysis of GWAS data has been previously associated genetic variation of CREBBP with schizophrenia symptoms and cognition. In my analysis, as well as investigating associations with change in PANSS as a continuous variable, I also investigate associations with categorical classification of responders and non- responders to clozapine, change in psychosocial functioning and quality of life, and associations with within-individual variations in total, positive and negative symptoms. I will now outline these analyses and their results. - 123 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4.2 Statistical method: Therapeutic response to clozapine To explore associations with therapeutic response to clozapine, I took two broad approaches. Firstly, I analysed change in clinical presentation at twelve weeks in four ways, all of which compared participants on measures of improvement in response to clozapine at follow up to investigate if this was associated with different trajectories in DNA methylation: 1) Between-individual classification of responder or non-responder to clozapine (categorical) as a predictor of DNA methylation trajectories over time Classification of responder or non-responder: responders had a >20% decrease in their baseline PANSS Total score at twelve week follow-up. 2) Between-individual variation in total symptom severity change between baseline and follow up (continuous) as a predictor of DNA methylation trajectories over time Total symptom severity change: percentage change in PANSS Total score between baseline and twelve week follow up was the measure of total symptom severity change. Negative percentages indicate improvement. 3) Between-individual variation in psychosocial functioning change between baseline and follow up (continuous) as a predictor of DNA methylation trajectories over time Psychosocial functioning change: percentage change in PSP score between baseline and twelve week follow up was the measure of psychosocial functioning change. Positive percentages indicate improvement. 4) Between-individual variation in quality of life change between baseline and follow up (continuous) as a predictor of DNA methylation trajectories over time Quality of life change: percentage change in GAF score between baseline and twelve week follow up was the measure of quality of life change. Positive percentages indicate improvement. Secondly, I analysed variation in symptom severity over the study period in three ways, all of which assessed whether fluctuations in symptom severity within individuals on clozapine are associated with differing DNA methylation: 1) Within-individual variation in total symptom severity as a predictor of DNA methylation Total symptom severity: PANSS Total score was the measure of total symptom severity at all time points with a clinical assessment. - 124 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 2) Within-individual variation in positive symptom severity as a predictor of DNA methylation Positive symptom severity: PANSS Positive score was the measure of positive symptom severity at all time points with a clinical assessment. 3) Within-individual variation in negative symptom severity as a predictor of DNA methylation Negative symptom severity: PANSS Negative score was the measure of negative symptom severity at all time points with a clinical assessment. To identify DNA methylation changes associated with therapeutic response to clozapine, I conducted mixed model regression analyses using lmer. For the prediction of different methylation trajectories between individuals with different changes in clinical presentation between baseline and twelve weeks, the interaction between response group or change in total symptom severity/psychosocial functioning/quality of life and weeks on clozapine was used a predictor of methylation at each CpG sites. These models controlled for baseline age, sex, predicted cell counts (CD8T, CD4T, NK, B, granulocytes and monocytes), smoking at time of sample collection (yes/no), chip ID, previous medications (number of prior atypical antipsychotics, number of prior typical antipsychotics, number of prior benzodiazepines or other mood stabilisers) and current medications (number of current antipsychotics, number of current benzodiazepines or other mood stabilisers, number of current antidepressants), as well as controlling for the repeated measures from individual participants as a random effect. These analyses all produced a model with regression coefficients for each CpG site - representing the percentage change in methylation at the CpG site associated with an additional week on clozapine for non-responders / those with an additional improvement at twelve weeks – as well as a p value for the significance of this association. For the prediction of differential methylation within individuals due to variation in symptom severity across time points, symptom severity (total/positive/negative) was used as a predictor of methylation at each CpG site. There models controlled for baseline age, sex, predicted cell counts (CD8T, CD4T, NK, B, granulocytes and monocytes), smoking at time of sample collection, chip ID, current medications (number of antipsychotics, benzodiazepines or other mood stabilisers, antidepressants, hyoscine, metformin, laxatives), number of total prior antipsychotics, as well as controlling for the repeated measures from individual participants. - 125 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes These analyses all produced a model with regression coefficients for each CpG site - representing the percentage change in methylation associated with an additional increase in symptom severity – as well as a p value for the significance of this association. For each model, I report global DNA methylation and differentially methylated positions (DMPs). Pathway analysis is also conducted for the between-participant categorical response model and the within-participant total symptom severity model. All of this is conducted as described in the previous chapter (Statistical method: Exposure to clozapine). Additionally, the categorical response model regression coefficients are used for candidate site testing and comparison to the cross-sectional dataset. Candidate CpG site methylation: Based on previous literature investigating independent replication of genes associated with response to clozapine (Chapter 4.1.3) and one study investigating genes with CpG sites associated with response to clozapine (Kinoshita et al., 2017), I identified five genes which have been associated with clozapine exposure: DRD3, 5HTR2A, 5HTT/SLC6A4, GNB3 and CREBBP. In total, there were 118 CpG sites for these genes included in the 450k assay which passed QC. To account for multiple-comparisons, the adjusted p value is 0.05/118 = 0.000424. Comparison to cross-sectional data: I then compared the results of the model of categorical clozapine response from my longitudinal dataset to results from the previous dataset of cross- sectional clozapine exposure, firstly to explore whether the cross-sectional model was validated by the longitudinal model, and then to explore whether the longitudinal model was validated by the cross-sectional model. To investigate replication of findings from the cross-sectional model, I took all the top DMPs with a p value < 5*10-5 from the cross-sectional model and ran a correlation analysis on the regression coefficients from the cross-sectional model for these top DMPs and the regression coefficients from one of the longitudinal models for these DMPs; this was to establish if there was the same direction of effect for DMPs in both cohorts (i.e. if methylation was higher in the clozapine exposed group from the cross-sectional dataset, was there an additional increase or decrease in methylation for non-responders in the longitudinal dataset). To investigate replication of findings from the longitudinal model of categorical clozapine response, I took all the top DMPs with a p value < 5*10-5 from the longitudinal model and ran a correlation analysis on the regression coefficients from the longitudinal model for these top DMPs and the regression coefficients from the cross-sectional model for these DMPs; again, this was to establish if there was the same direction of effect for DMPs in both cohorts (i.e. if there was an additional increase in methylation for non- - 126 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes responders in the longitudinal dataset, was methylation higher or lower in the clozapine exposed group from the cross-sectional dataset). The results for each analysis model are reported separately. First I will report the results for each between-participant models based on percentage change at twelve weeks - categorical classification of responder vs non-responder, change in total symptoms, change in psychosocial functioning, and then change in quality of life. Then I will report the results for each within-participant model of variation in symptom severity - total symptoms, positive symptoms and negative symptoms. - 127 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4.3 Results: Percentage change in response to clozapine (between-individuals) 4.3.1 Responders vs non-responders (categorical) 4.3.1.1 Participant characteristics Twelve weeks after clozapine initiation, ten participants were classified as responders to clozapine and twelve participants classified as non-responders. The characteristics of each group are compared in Table 6, with key details represented in subsequent pie charts (Figure 7): compared to responders, non-responders were more likely to be male, be white, have a diagnosis of paranoid schizophrenia, have not been previously exposed to clozapine, experience predominantly negative symptoms at baseline, experience drowsiness and cardiovascular symptoms, and not experience increased appetite/weight. Non-responders also had higher clozapine levels at twelve weeks (0.461 vs 0.342, p=0.09), and (when baseline samples are excluded) show fairly stable high levels of clozapine whereas responder start with lower levels which increase over time. However, none of these differences were statistically significant. Both groups show mean decreases in symptom severity over time, though the decreases at twelve weeks are smaller in non-responders and looking specifically at negative symptoms, non- responders show no decrease while responders show large decreases (-34.295% vs -3.686% total symptoms p<0.0001, -36.201% vs -13.142% positive symptoms p=0.042, -26.476% vs 2.668% negative symptoms p= 0.002). - 128 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Table 6: Characteristics of responders to clozapine vs non-responders Responders Non- Responders Age (years) Mean 38.77 36.99 p=.726 Minimum 19.94 21.08 Maximum 50.46 61.38 Gender Male 7 10 p=.816 Female 3 2 Ethnicity White 4 6 p=.267 Black 6 4 Asian 0 2 Diagnosis Schizophrenia 8 11 p=.865 Schizoaffective disorder 2 1 Months between diagnosis of Mean 24.78 29.10 p=.805 treatment-resistance and Minimum 0 0 clozapine start Maximum 100 111 Inpatient Psychiatric Mean 4.20 2.92 p=.371 Admissions Minimum 0 0 Maximum 10 7 Psychiatric Comorbidities Affective 1 1 p=.974 Anxiety 1 1 Substance Abuse 1 2 None 7 8 Prior Clozapine Exposure Yes 3 2 p=.816 No 7 10 Number of prior typical Mean 1.1 0.5 p=.08 antipsychotics Minimum 0 0 Maximum 2 1 Number of prior atypical Mean 3.1 2.75 p=.599 antipsychotics Minimum 1 1 Maximum 7 4 Number of prior mood Mean 0.2 0.25 p=.792 stabilisers Minimum 0 0 Maximum 1 1 Smoker at baseline Yes 1 2 p=1 No 9 10 Smoker at any time-point Yes 3 4 p=1 during study No 7 8 - 129 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Baseline PANSS Total Mean 78.3 78.3 p=.997 Minimum 54 55 Maximum 105 119 Baseline PANSS Positive Mean 20.10 19.13 p=.709 Minimum 9 12 Maximum 26 36 Baseline PANSS Negative Mean 20.70 22.83 p=.531 Minimum 9 17 Maximum 39 30 Predominant Baseline Positive 5 2 p=.226 Symptoms Negative 5 10 12w PANSS Total Mean 50.33 74.18 p=.001* Minimum 35 57 Maximum 74 98 6m PANSS Total Mean 51 69.55 p<.007* Minimum 37 44 Maximum 72 97 PANSS % Change at 12w Mean -34.30 -3.69 p<0.001* Minimum -53.57 -17.65 Maximum -21.33 26.47 PSP % Change at 12w Mean 39.33 21.33 p=.49 Minimum -50 -47.73 Maximum 160 162.50 GAF % Change at 12w Mean 70.50 11.97 p<.01* Minimum -8.33 -25 Maximum 185.71 29.03 Increased appetite/weight at Yes 8 5 p=.167 12w No 2 7 Cardiovascular symptoms at Yes 1 3 p=.724 12w No 9 9 Drowsiness at 12w Yes 8 7 p=.531 No 2 5 Clozapine Levels at 12w Mean 0.34 0.46 p=.095 Minimum 0.21 0.28 Maximum 0.50 0.75 - 130 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Figure 7: Pie charts comparing key demographics for responders vs non-responders - 131 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes - 132 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes - 133 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes - 134 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes ) mg/l ( - 135 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes ) mg/l ( ) mg/l ( - 136 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes - 137 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4.3.1.2 Global DNA methylation (Responders vs non-responders) Being a non-responder was associated with higher levels of global DNA methylation over time. For non-responders, an additional week on clozapine was associated with an average additional increase in global DNA methylation of 0.211x10-3 %. 50.31% of all CpG sites showed an additional increase in methylation associated with time exposed to clozapine for non-responders and 49.68% of all CpG sites showed a decrease. Moreover, when exploring individual time points, split by response: at baseline, mean global DNA methylation was 48.286% for responders and 48.284% for non-responders; at six weeks, mean global DNA methylation was 48.277% for responders and 48.236% for non-responders; at twelve weeks mean global DNA methylation was 48.292% for responders and 48.286% for non-responders; and at six months mean global DNA methylation was 48.294% for responders and 48.293% for non- responders. Both groups start and end with equivalent levels of global DNA methylation and show the same pattern of an initial decrease and then increases, however the graph below shows the initial decrease is more pronounced for non-responders. - 138 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4.3.1.3 Differentially methylated positions (responders vs non-responders) Two DMPs reached the threshold for epigenome wide significance (p<1*10-7), while one hundred and sixty eight other DMPs were trend-level significant (p<5*10-5). - 139 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes - 140 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Table 7: Top ten DMPs associated with categorical response to clozapine classification (* indicates epigenome wide significance) CpG site Beta P value Mean methylation Gene Location MAPINFO Significant blood-brain methylation correlation Baseline Six weeks Twelve weeks Six months reported? R: 73.24 R: 77.08 R: 74.43 R: 74.85 N cg15353603 -0.192 x10-2 5.97E-09* NR: 73.82 NR: 76.41 NR: 73.49 NR: 71.23 DPP4 Body 162915902 R: 17.17 R: 15.67 R: 17.25 R: 16.93 N cg18043267 0.202 x10-2 8.59E-08* NR: 15.58 NR: 17.27 NR: 19.49 C10orf125 Body 135170752 R: 66.69 R: 66.91 R: 66.79 R: 66.12 N cg10015974 0.221 x10-2 3.04E-07 NR: 65.15 NR: 66.07 NR: 68.60 NR: 68.86 199827580 R:83.32 R: 83.91 R: 83.20 R: 84.17 N cg27025387 -0.153 x10-2 5.24E-07 NR: 85.00 NR: 83.91 NR: 83.24 NR: 81.50 ANKRD11 Body 89352529 R: 26.61 R: 26.14 R: 26.67 R: 25.38 N cg08759026 0.138 x10-2 8.53E-07 NR: 24.83 NR: 25.08 NR: 27.43 NR: 26.94 MYEOV TSS200 69061454 R: 10.26 R: 10.29 R: 11.03 R: 10.81 Y – prefrontal cortex and entorhinal cortex cg01881899 0.127 x10-2 9.27E-07 NR: 9.30 NR: 10.10 NR: 11.08 NR: 12.21 ABCG1 Body 43652704 R: 7.14 R: 6.90 R: 7.15 R: 8.65 N cg06275059 0.086 x10-2 1.15E-06 NR: 6.81 NR: 7.08 NR: 7.06 NR: 8.62 PCDP1 TSS1500 120301779 R: 47.40 R: 49.81 R: 47.73 R: 48.15 N cg19466160 -0.365 x10-2 1.30E-06 NR: 45.23 NR: 48.80 NR: 41.33 NR: 40.49 DLG4 Body 7117160 R: 77.39 R: 77.78 R: 77.94 R: 78.55 N cg21157923 -0.145 x10-2 1.74E-06 NR: 79.03 NR: 77.47 NR: 77.33 NR: 75.85 LCN10 TSS1500 139637747 R: 31.78 R: 31.61 R: 31.79 R:31.91 Y – prefrontal cortex, entorhinal cortex, superior NR: 30.74 NR:31.64 NR: 31.99 NR:34.04 cg04408162 0.156x10-2 2.08E-06 KCNK9 Body 140643009 temporal gyrus - 141 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes The most statistically significant DMP associated with non-response to clozapine was cg15353603. For non-responders, each additional week exposed to clozapine was associated with an additional decrease in methylation at this site of 0.192% compared to responders (p=0.6x10-8). Uncorrected means show an average baseline methylation of 73.24% for responders and 73.82% for non-responders, 77.08% and 76.41% at six weeks, 74.43% and 73.49% at twelve weeks, and 74.85% and 71.23% at six months. In other words, responders and non-responders start off with similar levels of methylation but responders show an increase which tempers off but remains higher at the end of six months than at baseline, whereas non-responders show a smaller increase initially and then a decrease which means they end up with lower levels of methylation than at baseline. cg15353603 is located in the gene body of DPP4, which encodes dipeptidyl peptidase 4/ cell surface antigen CD26. This gene is located on chromosome 2 and is expressed across numerous tissues including brain, with overexpression of mRNA in small intense, kidney and prostate and overexpression of protein in serum, pancreas, placenta, urine and saliva. Dipeptidyl Peptidase 4 is a type II transmembrane enzymatic glycoprotein implicated in pancreatic disease, incretin synthesis secretion and inactivation, glucose homeostasis, protein digestion and absorption, cardiovascular - 142 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes disease, tumour biology, obesity and receptor binding and signalling. It is also associated with immune regulation as it is essential for T-cell-receptor mediated T-cell activation and acts as a positive regulator of T-cell co-activation, induces T-cell proliferation, and regulates lymphocyte- epithelial cell adhesion. DPP4 inhibitors are used for the treatment of type 2 diabetes. This CpG site is reported to show no correlation between methylation in blood and methylation in brain tissue. The second most statistically significant DMP associated with non-response to clozapine was cg18043267. For non-responders, each additional week exposed to clozapine was associated with an additional increase in methylation at this site of 0.202% compared to responders (p=0.9x10-7). Uncorrected means show an average baseline methylation of 17.174% for responders and 15.580% for non-responders, 15.670% and 15.701% at six weeks, 17.247% and 17.271% at twelve weeks, and 16.925% and 19.488% at six months. Responders show a decrease initially and then an increase taking them above baseline and then a final decrease ending below initial levels of methylation, while non-responders start at a lower level of methylation and this increases over time ending up higher than responders at any time point. - 143 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes cg18043267 is located in the gene body of C10orf125 or FUOM, which encodes fucose mutarotase. This gene is located on chromosome 10 and is expressed across numerous tissues including brain, with overexpression of mRNA in liver and kidney and overexpression of protein in liver and gall bladder. Fucose mutarotase is implicated in transport to the golgi and subsequent modification, protein metabolism, and fucose binding which is believed to play a critical role in cell- cell adhesion and recognition processes. This CpG site is reported to show no correlation between methylation in blood and methylation in brain tissue. The sixth most statistically significant DMP associated with non-response to clozapine was cg01881899. For non-responders, each additional week exposed to clozapine was associated with an additional increase in methylation at this site of 1.267% compared to responders (p=0.9x10-6). Uncorrected means show an average baseline methylation of 10.257% for responders and 9.298% for non-responders, 10.291% and 10.099% at six weeks, 11.029% and 11.077% at twelve weeks, and 10.812% and 12.207% at six months. Responders start off with a higher baseline methylation which increases and then reduces but remains above baseline, while non-responders show increase at each time ending with higher levels than responders. - 144 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes cg01881899 is located with the gene body of ABCG1, which encodes ATP-binding cassette sub- family G member 1. This gene is located on chromosome 21 and is expressed across numerous tissues including brain, with overexpression of mRNA in macrophages and overexpression of protein in adipocyte and breast. ATP-binding cassette sub-family G member 1 is a protein which transports various molecules – including macrophages and phospholipids - across extra and intra cellular membranes. This CpG site is reported to show significant weak correlation between methylation in blood and methylation in brain in the prefrontal cortex (r=0.239, p=0.040) and entorhinal cortex (r=0.26, p=0.028). The tenth most statistically significant DMP associated with non-response to clozapine was cg04408162. For non-responders, each additional week exposed to clozapine was associated with an additional increase in methylation at this site of 0.156% compared to responders (p=0.2x10-5). Uncorrected means show an average baseline methylation of 31.778% for responders and 30.735% for non-responders, 31.605% and 31.639% at six weeks, 31.787% and 31.985% at twelve weeks, and 31.914% and 34.040% at six months. Responders show a higher level of methylation at baseline which decreases and then increases, while non-responders show an increase which ends with a higher level of methylation than responders at all time points. - 145 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes cg04408162 is located in the gene body of KCNK9, which encodes potassium two pore domain channel subfamily K member 9 which functions as a pH-dependent potassium channel. This gene is located on chromosome 8 and is expressed in numerous tissue including brain, with overexpression of mRNA in brain (cerebellum) and overexpression of protein in liver. Variants of this gene have been associated with intellectual disability. This CpG site is reported to show significant moderate correlation between methylation in blood and methylation in brain tissue in prefrontal (r=0.415, p=0.0002), entorhinal (r=0.335, p=0.004) and superior temporal gyrus (r=0.295, p=0.0101). 4.3.1.4 Pathway analysis (responders vs non-responders) Pathway analysis indicated that the pathways in Table 3 were the top thirty pathways with an over-representation of DMPs associated with response classification. - 146 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Table 8 Top thirty over-represented pathways - response group Category Name p value over- Number of differentially Number in Ontology representation methylated category GO:0060080 regulation of inhibitory postsynaptic membrane potential 0.00016 5 11 BP GO:0010885 regulation of cholesterol storage 0.00026 5 13 BP GO:0010878 cholesterol storage 0.00036 5 14 BP GO:0032088 negative regulation of NF-kappaB transcription factor activity 0.00073 9 56 BP GO:1901618 organic hydroxy compound transmembrane transporter activity 0.00148 5 23 MF GO:0043433 negative regulation of sequence-specific DNA binding transcription factor activity 0.00237 14 111 BP GO:1901385 regulation of voltage-gated calcium channel activity 0.00327 4 12 BP GO:0015370 solute: sodium symporter activity 0.00336 7 46 MF GO:0009220 pyrimidine ribonucleotide biosynthetic process 0.00342 4 19 BP GO:0015812 gamma-aminobutyric acid transport 0.00347 4 13 BP GO:0010884 positive regulation of lipid storage 0.00463 4 17 BP GO:0009218 pyrimidine ribonucleotide metabolic process 0.00482 4 21 BP GO:0015665 alcohol transmembrane transporter activity 0.00496 3 11 MF GO:0003714 transcription corepressor activity 0.00533 20 184 MF GO:0072528 pyrimidine-containing compound biosynthetic process 0.00583 5 32 BP GO:0010883 regulation of lipid storage 0.00667 6 37 BP GO:0031664 regulation of lipopolysaccharide-mediated signalling pathway 0.00684 4 16 BP GO:0043254 regulation of protein complex assembly 0.007 20 224 BP GO:0046131 pyrimidine ribonucleoside metabolic process 0.007 4 29 BP - 147 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Category Name p value over- Number of differentially Number in Ontology representation methylated category GO:0015294 solute: cation symporter activity 0.007 9 77 MF GO:0015293 symporter activity 0.007 12 124 MF GO:0003084 positive regulation of systemic arterial blood pressure 0.008 3 13 BP GO:0000109 nucleotide-excision repair complex 0.008 3 11 CC GO:0030832 regulation of actin filament length 0.0095 12 114 BP GO:0019915 lipid storage 0.0099 7 56 BP GO:0006221 pyrimidine nucleotide biosynthetic process 0.0101 4 24 BP GO:0032984 macromolecular complex disassembly 0.0102 16 196 BP GO:0031663 lipopolysaccharide-mediated signalling pathway 0.0102 6 43 BP GO:0055094 response to lipoprotein particle 0.0106 3 10 BP GO:0051059 NF-kappaB binding 0.0107 4 24 MF - 148 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4.3.1.5 Candidate CpG site methylation (responders vs non-responders) Of eleven CpG sites located within DRD3 with methylation data available, two had a significant association with clozapine response before but not after correction for multiple testing (0.424x10- 3 Of twenty five CpG sites located within 5HTR2A with methylation data available, one had a significant association with clozapine response before but not after correction for multiple testing (0.424x10-3 Of fifteen CpG sites located within 5HTT with methylation data available, two had a significant association with clozapine response before but not after correction for multiple testing (0.424x10- 3 Of twelve CpG sites located within GNB3 with methylation data available, no CpG sites had a significant association with clozapine response. Across all GNB3 CpG sites, there was a mean additional increase in methylation of 0.0122% associated with non-response. Of fifty five CpG sites located within CREBBP with methylation data available, three had a significant association with clozapine response before but not after correction for multiple testing (0.424x10-3 - 149 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4.3.1.6 Comparison to cross-sectional data (responders vs non-responders) First I took the top DMPs (p<0.05*10-5) from the cross-sectional dataset and identified these within the longitudinal dataset. There was a highly significant moderate positive correlation between the regression coefficients from the longitudinal dataset and the cross-sectional data set of 0. 610, p<2.2e-16, df =4561. I then took the top DMPs (p<0.05*10-5) from the longitudinal dataset and identified these within the cross-sectional dataset. There was a highly significant strong positive correlation between the regression coefficients from the longitudinal dataset and the cross-sectional data set of 0.737, p = 2.2e-16, df = 165. - 150 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes - 151 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4.3.2 Change in total symptom severity (continuous) 4.3.2.1 Participant characteristics All participants had a score for percentage change in total symptom severity at twelve weeks so all twenty two participants and eighty five samples were included. Greater reductions in total symptom severity were correlated with greater improvements in quality of life (-0.620, df=19, p=0.003) and with greater improvements in psychosocial functioning (though this was not significant, -0.33 df=18 p=0.16), and improvements in psychosocial functioning and quality of life were correlated with one another (0.6, df=18, p=0.0052). - 152 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes - 153 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4.3.2.2 Global DNA methylation (change in total symptom severity %) Participants with greater reductions in total symptom severity showed greater reductions in global DNA methylation over time. Each additional decrease in symptom severity at twelve weeks was associated with an additional decrease in global DNA methylation over time of -0.00103%. 49.50% of all CpG sites showed an increase in methylation associated with greater reductions, and 50.50% showed a decrease. 4.3.2.3 Differentially methylated positions (change in total symptom severity %) Two DMPs reached the threshold for epigenome wide significance (p<1*10-7), while one hundred and eighty eight other DMPs were trend-level significant (p<5*10-5) . - 154 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes - 155 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Table 9: Top ten DMPs associated with change in total symptom severity (continuous) (* indicates epigenome wide significance) CpG site Beta P value Gene Location MAPINFO Significant blood-brain methylation correlation reported? cg10015974 0.245x10-2 3.15E-09* 199827580 N -0.190 N cg15353603 x10-2 2.96E-08* DPP4 Body 162915902 0.172 x10- Y – prefrontal cortex, entorhinal cortex, superior cg07829465 2 1.70E-07 TRIM31 Body 30079536 temporal gyrus 0.296 x10- N cg01558909 2 1.80E-07 HBM TSS200 215845 0.133 x10- N cg06623151 2 2.57E-07 ASCL2 TSS1500 2293201 0.120 x10- Y – prefrontal cortex cg10006515 2 2.66E-07 PDCD2L Body 34895882 0.179 x10- N cg22112035 2 4.10E-07 ANO1 Body 70002414 cg06275059 0.86 x10-3 7.96E-07 PCDP1 TSS1500 120301779 N cg03528137 0.40 x10-3 9.38E-07 111589969 N 0.116 x10- N cg10020520 2 9.64E-07 SETD1A Body 30976186 - 156 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes The most statistically significant DMP associated with reduction in total symptom severity at twelve weeks was cg10015974. Having an additional reduction in total symptom severity was associated with an additional decrease in methylation over time of -0.245% (p=0.3x10-8). cg10015974 is located on chromosome 1, outside of any characterised genes. This CpG site is reported to show no correlation between methylation in blood and methylation in brain tissue. The second most statistically significant DMP associated with reduction in total symptom severity at twelve weeks was cg15353603. Having an additional reduction in total symptom severity was associated with an additional increase in methylation over time of 0.19% (p=0.3x10-7). cg15353603 is described above – it is located within the gene body of DPP4. The third most statistically significant DMP associated with reduction in total symptom severity at twelve weeks was cg07829465. Having an additional reduction in total symptom severity was associated with an additional decrease in methylation over time of -0.172% (p=0.2x10-6). cg07829465 is located within the gene body of TRIM31, which encodes tripartite motif- containing protein 31. This gene is located on chromosome 6 and is expressed across numerous tissue including brain, with overexpression of mRNA in colon, bladder, small intense and stomach and overexpression of protein in adipocyte. Tripartite motif-containing protein 31 functions as an E3 ubiquitin-protein ligase, involved in regulation of cell growth, interferon gamma signalling, ligase activity and the innate immune system. This CpG site is reported to show a significant moderate correlation between methylation in blood and methylation in prefrontal cortex (r=0.336, p=0.0035), entorhinal cortex (r=0.43, p=0.0002), and superior temporal gyrus (r=0.346, p=0.0023). The sixth most statistically significant DMP associated with reduction in total symptom severity at twelve weeks was cg10006515. Having an additional reduction in total symptom severity was associated with an additional decrease in methylation over time of -0.12% (p=0.3x10-6). cg10006515 is located within the gene body of PDCD2L, which encodes programmed cell death 2 like. This gene is located on chromosome 19 and is expressed across numerous tissues including brain, with overexpression of protein in ovary, bone marrow and breast. Programmed cell death 2 like is involved in transcriptional regulation and cell cycle progression. This CpG site is reported to show a significant weak correlation between methylation in blood and methylation in prefrontal cortex (r=0.28, p=0.0155), with a trend for negative correlation in cerebellum (r=-0.211, p =0.078). - 157 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4.3.3 Change in psychosocial functioning (continuous) 4.3.3.1 Participant characteristics Two participants did not have a score for percentage change in psychosocial functioning at twelve weeks so the twenty remaining participants and their seventy eight samples were included in the below analysis. 4.3.3.2 Global DNA methylation (change in psychosocial functioning %) Participants with greater improvements in psychosocial functioning showed increased global DNA methylation over time. Each additional increase in psychosocial functioning at twelve weeks was associated with an average additional increase in global DNA methylation over time of 0.00103%. 48.88% of all CpG sites showed a decrease in methylation associated with greater improvements, and 51.12% showed an increase. 4.3.3.3 Differentially methylated positions (change in psychosocial functioning %) One DMP was epigenome wide significant (p<1*10-7), and one hundred and twenty two other DMPs were trend-level significant (p<*10-5). - 158 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes - 159 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Table 10: Top ten DMPs associated with change in psychosocial functioning (* indicates epigenome wide significance) CpG site Beta P value Gene Location MAPINFO Significant blood-brain methylation correlation reported? cg05195750 0.103x10-2 3.55E-08* ELP4;IMMP1L Body;TSS1500 31531557 Y – entorhinal cortex and prefrontal cortex cg20676542 0.31 x10-3 5.06E-07 TRIM25 TSS200 54991456 N 0.192 x10- Y – superior temporal gyrus cg15113009 2 5.50E-07 PKNOX2 Body 125297719 -0.115 Y – prefrontal cortex, entorhinal cortex, superior cg22867608 x10-2 7.20E-07 EGR2 Body 64573356 temporal gyrus and cerebellum 0.143 x10- N cg09595020 2 7.70E-07 IQCE 3'UTR 2654120 0.105 x10- Y – entorhinal cortex cg07950397 2 1.08E-06 OLFM2 Body 10022659 cg07572579 -0.76 x10-3 1.69E-06 68116058 N -0.175 N cg21215820 x10-2 1.77E-06 CLEC10A 3'UTR 6977901 0.321 x10- Y – entorhinal cortex cg14150252 2 1.98E-06 UGP2 Body;5'UTR 64069583 0.148 x10- N cg20951539 2 2.03E-06 PXT1 Body 36367819 - 160 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes The most statistically significant DMP associated with improvement in psychosocial functioning at twelve weeks was cg05195750. Having an additional improvement in psychosocial functioning was associated with an additional increase in methylation over time of 0.103% (p=0.4x10-7). cg05195750 is described above – it is located within the gene body of ELP4. The third most statistically significant DMP associated with improvement in psychosocial functioning at twelve weeks was cg15113009. Having an additional improvement in psychosocial functioning was associated with an additional increase in methylation over time of 0.192% (p=0.6x10-6). cg15113009 is located within the gene body of PKNOX2, which encodes homeobox protein PREP-2. This gene is located on chromosome 11 and is expressed in numerous tissues including brain, with overexpression of mRNA in cortex and overexpression of protein in liver. Homeobox protein PREP-2 is a sequence-specific transcription factor which plays a key role in cell proliferation, differentiation and death, and has also been associated with substance dependence. This CpG site is reported to show a significant negative correlation between methylation in blood and methylation in superior temporal gyrus (r=-0.243, p=0.035). The fourth most statistically significant DMP associated with improvement in psychosocial functioning at twelve weeks was cg22867608. Having an additional improvement in psychosocial functioning was associated with an additional decrease in methylation over time of 0.115% (p=0.7x10-6). cg22867608 is located within the gene body of EGR2, which encodes early growth response 2. This gene is located on chromosome 10 and is expressed in numerous tissues including brain, with overexpression of mRNA in tibial nerve. Early growth response 2 is a transcription factor which binds to two specific DNA sites located in the promoter region of HOXA4 and is associated with charcot- marie tooth disease which is a congenital disorder of hypo-myelination. This CpG site is reported to shown significant moderation correlations between methylation in blood and methylation in prefrontal cortex (r=0.495, p=7.47e-06), entorhinal cortex (r=0.371, p=0.0014), superior temporal gyrus (r=0.329, p=0.0039), and cerebellum (r=0.368, p=0.0016). The sixth most statistically significant DMP associated with improvement in psychosocial functioning at twelve weeks was cg07950397. Having an additional improvement in psychosocial functioning was associated with an additional increase in methylation over time of 0.105% (p=0.1x10-5). - 161 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes cg07950397 is located within the gene body of OLFM2, which encodes olfactomedin 2. This gene is located on chromosome 19 and is expressed across numerous tissue including brain, with overexpression of mRNA in brain (putamen, amygdala and nucleus accumbens) and overexpression of protein in vitreous humour and serum. Olfactomedin 2 is a protein involved in transforming growth factor beta and AMPAR complex organisation. This CpG site is reported to show a significant weak correlation between methylation in blood and methylation in entorhinal cortex (r=0.308, p=0.0091). The ninth most statistically significant DMP associated with improvement in psychosocial functioning at twelve weeks was cg14150252. Having an additional improvement in psychosocial functioning was associated with an additional increase in methylation over time of 0.321% (p=0.2x10-5). cg14150252 is located within the gene body of UGP2, a gene which encodes UDP-glucose pyrophosphorylase 2. This gene is located on chromosome 2 and is expressed in numerous tissues including brain, with overexpression of mRNA in skeletal muscle and overexpression of protein in liver. UDP-glucose pyrophosphorylase 2 is an enzyme involved in carbohydrate conversions of glucose. This CpG site is reported to demonstrate a trend towards negative correlation between methylation in blood and methylation in entorhinal cortex (r=-0.218, p=0.0677). - 162 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4.3.4 Change in quality of life (continuous) 4.3.4.1 Participant characteristics One participant did not have a score for percentage change in psychosocial functioning at twelve weeks so the twenty one remaining participants and their eighty two samples were included in the below analysis. 4.3.4.2 Global DNA methylation (change in quality of life %) Participants with greater improvements in quality of life showed increased global DNA methylation over time. Each additional increase in quality of life at twelve weeks was associated with an average additional increase in global DNA methylation of 0.000898%. 48.99% of all CpG sites showed a decrease in methylation associated with greater improvements, and 51.01% showed an increase. 4.3.4.3 Differentially methylated positions (change in quality of life %) One DMP reached the threshold for epigenome wide significance (p<1*10-7) and one hundred and forty eight DMPs were trend-level significant (p<5*10-5). - 163 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes - 164 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Table 11: Top ten DMPs associated with change in quality of life (* indicates epigenome wide significance) CpG site Beta P value Gene Location MAPINFO Significant blood-brain methylation correlation reported? cg14150252 0.408x10-2 5.94E-08* UGP2 Body;5'UTR 64069583 Y – entorhinal cortex 0.185 x10- N cg11641222 2 1.83E-07 SNTB1 Body 121566699 0.301 x10- Y – prefrontal cortex, entorhinal cortex, superior cg00190355 2 1.94E-07 SLC9A3 Body 497639 temporal gyrus 0.108 x10- Y – entorhinal cortex and superior temporal gyrus cg05195750 2 2.38E-07 ELP4;IMMP1L Body;TSS1500 31531557 0.166 x10- Y – prefrontal cortex, entorhinal cortex, superior cg07829465 2 3.39E-07 TRIM31 Body 30079536 temporal gyrus 0.191 x10- Y – prefrontal cortex, entorhinal cortex, superior cg11970204 2 4.52E-07 BTRC Body 103245907 temporal gyrus 0.213 x10- Y – prefrontal cortex, entorhinal cortex, superior cg13084525 2 5.30E-07 EHF TSS1500 34642463 temporal gyrus and cerebellum -0.157 N cg00182202 x10-2 5.36E-07 SRL Body 4257902 -0.266 N cg17859578 x10-2 5.64E-07 112816053 0.235 x10- N cg21768539 2 9.73E-07 COL23A1 Body 177929808 - 165 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes The most statistically significant DMP associated with improvement in quality of life at twelve weeks was cg14150252. Having an additional improvement in quality of life was associated with an additional increase in methylation over time of 0.408% (p=0.6x10-7). cg14150252 is described above - it is located within the gene body of UGP2. The third most statistically significant DMP associated with improvement in quality of life at twelve weeks was cg00190355. Having an additional improvement in quality of life was associated with an additional increase in methylation over time of 0.301% (p=0.2x10-6). cg00190355 is located within the gene body of SLC9A3, which encodes solute carrier family 9 member A3. This gene is located on chromosome 5 and is expressed across numerous tissues including brain, with overexpression of mRNA in colon, kidney and stomach and overexpression of protein in fetal gut, kidney and urine. Solute carrier family 9 member A3 is involved in sodium ion transportation, signal transduction and pH regulation. This CpG site is reported to show significant weak correlation between methylation in blood and methylation in prefrontal cortex (r=0.264, p=0.023), entorhinal cortex (r=0.366, p=0.0017), and superior temporal gyrus (r=0.262, p=0.023). The fourth most statistically significant DMP associated with improvement in quality of life at twelve weeks was cg05195750. Having an additional improvement in quality of life was associated with an additional increase in methylation over time of 0.108% (p=0.2x10-6). cg05195750 is described above – it is located within the gene body of ELP4. The fifth most statistically significant DMP associated with improvement in quality of life at twelve weeks was cg07829465. Having an additional improvement in quality of life was associated with an additional increase in methylation over time of 0.166% (p=0.3x10-6). cg07829465 is described above - it is located within the gene body of TRIM31. The sixth most statistically significant DMP associated with improvement in quality of life at twelve weeks was cg11970204. Having an additional improvement in quality of life was associated with an additional increase in methylation over time of 0.191% (p=0.5x10-6). cg11970204 is located within the gene body of BTRC, which encodes beta-transducin repeat containing E3 ubiquitin protein ligase. This gene is located on chromosome 10 and is expressed across numerous tissues including brain, with overexpression of protein in peripheral blood mononuclear cells and pancreatic juice. Beta-transducin repeat containing E3 ubiquitin protein ligase is involved in phosphorylation-dependent ubiquitination, IL-1 signalling, and ligase activity by targeting cells for degradation. This CpG site is reported to show significant weak correlations - 166 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes between methylation in blood and methylation in prefrontal cortex (r=0.306, p=0.008), entorhinal cortex (0.253, p=0.033), and superior temporal gyrus (r=0.309, p=0.007). The seventh most statistically significant DMP associated with improvement in quality of life at twelve weeks was cg13084525. Having an additional improvement in quality of life was associated with an additional increase in methylation over time of 0.213% (p=0.5x10-6). cg13084525 is located within the TSS1500 promoter region of EHF, which encodes ETS homologous factor. This gene is located on chromosome 11 and is expressed across numerous tissues including brain, with overexpression of mRNA in oesophagus, salivary gland and vagina and overexpression of protein in salivary gland and skin. ETS homologous factor is a transcriptional repressor involved in epithelial-specific expression. This CpG site is reported to show significant moderation correlations between methylation in blood and methylation in prefrontal cortex (r=0.432, p=0.0001) and entorhinal cortex (0.457, p=6.18e-05), and weak correlations between methylation in blood and methylation in superior temporal gyrus (r=0.254, p=0.028) and cerebellum (r=0.238, p=0.046). - 167 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4.4 Results: Symptom severity (within individuals) 4.4.1 Participant characteristics Twenty five of the samples did not have corresponding symptom severity score; twenty three of these because these were six/eight week samples when no clinical assessments were done. The remaining sixty samples from all twenty two participants were included in the following analysis. 4.4.2 Total symptom severity 4.4.2.1 Global DNA methylation (total symptom severity) Higher within-individual total symptom severity at any time point was associated with higher global DNA methylation. An increase in symptom severity was associated with an average increase in global DNA methylation of 0.0005%; however, 51.79% of all CpG sites showed a decrease in methylation associated with total symptom severity, and 48.20% showed an increase. 4.4.2.2 Differentially methylated positions (total symptom severity) Five DMPs reached the threshold for epigenome wide significance (p<1*10-7), while four hundred and fifty three other DMPs were trend-level significant (p<5*10-5). - 168 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes - 169 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Table 12: Top ten DMPs associated with within-individual variation in total symptom severity (* indicates epigenome wide significance) CpG site Beta P value Gene Location MAPINFO Significant blood-brain methylation correlation reported? -0.152x10- N cg02890365 2 6.68E-09* TUFT1;MIR554 Body;TSS200 151518189 0.229 x10- Y – entorhinal cortex, superior temporal gyrus cg23158483 2 1.92E-08* GLRA1 TSS1500 151305528 -0.153 Y – prefrontal cortex cg26129353 x10-2 6.30E-08* ATP8B1 Body 55316927 -0.145 N cg17651451 x10-2 8.22E-08* ADAMTS2 Body 178697400 0.109 x10- Y – entorhinal cortex cg23290344 2 8.38E-08* NEFM TSS1500;oneExon 24771466 cg03765206 0.85 x10-3 1.07E-07 NAV2 Body 19429376 N cg22359217 -0.21 x10-3 1.93E-07 FBXO4 Body 41925772 N cg08127293 -0.26 x10-3 2.21E-07 14379328 N -0.128 Y – entorhinal cortex cg00496462 x10-2 2.74E-07 93577610 cg08160619 -0.69 x10-3 3.16E-07 CORO1A 5'UTR 30195318 Y – prefrontal cortex - 170 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes The most statistically significant DMP associated with total symptom severity was cg02890365. Each within-individual increase in total symptom severity was associated with a decrease in methylation of 0.152% (p=0.7x10-8). Using uncorrected means there was a negative correlation of - 0.354 between total symptom severity and methylation at any time point across all samples. cg02890365 is located within the body of TUFT1, which encodes for tuftelin. This gene is located on chromosome 1 and is expressed in numerous tissues including brain, with overexpression of mRNA in skin and overexpression of protein in skin and cervix. Tuftelin is an acidic phosphorylated glycoprotein found in tooth enamel, which is also implicated in neurotrophin nerve growth factor mediated neuronal differentiation. This CpG site is reported to show no correlation between methylation in blood and methylation in brain tissue. The second most statistically significant DMP associated with total symptom severity was cg23158483. Each within-individual increase in total symptom severity was associated with an - 171 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes increase in methylation of 0.229% (p=0.2x10-7). Using uncorrected means there was a correlation of 0.515 between total symptom severity and methylation at any time point across all samples. cg23158483 is located within the TSS1500 promoter region of GLRA1, which encodes the glycine receptor subunit alpha-1. This gene is located on chromosome 5 and is expressed in numerous tissue including brain, with overexpression of mRNA in brain (hypothalamus, substantia nigra and spinal cord) and overexpression of protein in retina; this gene has also been associated with startle disease. The glycine receptor subunit alpha-1 is an inhibitory glycine receptor, one of the most widely distributed inhibitory receptors in the central nervous system key for mediating neurotransmission through postsynaptic inhibition. This CpG site is reported to show a significant weak correlation between methylation in blood and methylation in entorhinal cortex(r=0.331, p=0.005) and superior temporal gyrus (r=0.246, p=0.033), with a trend for correlation (r=0.211, p=0.072) at the prefrontal cortex. - 172 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes The third most statistically significant DMP associated with total symptom severity was cg26129353. Each within-individual increase in total symptom severity was associated with a decrease in methylation of 0.153% (p=0.6x10-7). Using uncorrected means there was a correlation of -0.180 between total symptom severity and methylation at any time point across all samples. cg26129353 is located within the gene body of ATP8B1, which encodes ATPase phospholipid transporting 8B1. This gene is located on chromosome 18 and is expressed in numerous tissues including brain, with overexpression of mRNA in colon and overexpression of protein in nasal epithelium and amniocyte. ATPase phospholipid transporting 8B1 is part of a family of proteins which transport phosphatidylserine and phosphatidylethanolamine from one side of a bilayer to another, and it is implicated in nucleotide binding and ion channel transport. This CpG site is reported to show a significant weak correlation between methylation in blood and methylation in prefrontal cortex (r=-0.274, p=0.018). - 173 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes The fourth most statistically significant DMP associated with total symptom severity was cg17651451. Each within-individual increase in total symptom severity was associated with a decrease in methylation of 0.145% (p=0.8x10-7). Using uncorrected means there was a correlation of -0.0003 between total symptom severity and methylation at any time point across all samples. cg17651451 is located within the gene body of ADAMTS2, which encodes ADAM metallopeptidase with thrombospondin type 1. This gene is located on chromosome 5 and is expressed in numerous tissues including brain, but with low levels of mRNA expression in brain, high levels in skin, bone, tendon and aorta and overexpression of protein in amniocyte and bone marrow stem cell. ADAM metallopeptidase with thrombospondin type 1 has a role in collagen biosynthesis and peptidase activity. This CpG site is reported to show no correlation between methylation in blood and methylation in brain tissue. The fifth most statistically significant DMP associated with total symptom severity was cg23290344. Each within-individual increase in total symptom severity was associated with an - 174 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes increase in methylation of 0.1904% (p=0.8x10-7). Using uncorrected means there was a correlation of 0.286 between total symptom severity and methylation at any time point across all samples. cg23290344 is located within the TSS1500 promoter region of NEFM, which encodes neurofilament medium. This gene is located on chromosome 8 and is expressed in numerous tissues including brain, with overexpression of mRNA in brain (frontal cortex, cortex, anterior cingulate cortex, hippocampus, substantia nigra) and overexpression of protein in spinal cord, frontal cortex and brain. Neurofilament medium functionally maintains neuronal calibre and is implicated in intracellular transport to axons and dendrites. This CpG site is reported to show significant weak correlation between methylation in blood and methylation in entorhinal cortex (r=0.316, p=0.007). The ninth most statistically significant DMP associated with total symptom severity was cg00496462. Each within-individual increase in total symptom severity was associated with a - 175 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes decrease in methylation of 0.128% (p=0.3x10-6). Using uncorrected means there was a correlation of -0.211 between total symptom severity and methylation at any time point across all samples. cg00496462 is located on chromosome 7, outside of any characterised genes. This CpG site is reported to show no significant correlation between methylation in blood and methylation in brain tissue, except for a trend in entorhinal cortex (r=0.222, p=0.063). The tenth most statistically significant DMP associated with total symptom severity was cg08160619. Each within-individual increase in total symptom severity was associated with a decrease in methylation of 0.069% (p=0.3x10-6). Using uncorrected means there was a correlation of -0.248 between total symptom severity and methylation at any time point across all samples. - 176 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes cg08160619 is located within the 5’UTR region of CORO1A, encodes coronin 1A. This gene is located on chromosome 16 and is expressed in numerous tissues including brain, with overexpression of mRNA in whole blood and spleen and overexpression of protein in lymph node, NK cells and T lymphocytes; variants of this gene have been associated with immunodeficiency. Coronin 1A, a member of the WD repeat protein family, is implicated in numerous cellular processes including signal transduction, gene regulation, and poly(A) RNA binding. This CpG site is reported to show significant weak correlations between methylation in blood and methylation in prefrontal cortex (r=0.253, p=0.0296) with a trend in superior temporal gyrus (r=0.208, p=0.073). 4.4.2.3 Pathway analysis (total symptom severity) Pathway analysis indicated that the pathways in Table 13 were the top thirty pathways with an over-representation of DMPs associated with total symptom severity. - 177 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Table 13: Top thirty over-represented pathways - total symptom severity Category Name p value over- Number of differentially Number in Ontology representation methylated category GO:0007156 homophilic cell adhesion 5.30E-07 50 130 BP GO:0098609 NA 9.08E-06 53 178 BP GO:0000902 cell morphogenesis 0.00063 163 1022 BP GO:0021761 limbic system development 0.00070 20 73 BP GO:0010876 lipid localization 0.00093 42 264 BP GO:0005509 calcium ion binding 0.00106 101 629 MF GO:0030030 cell projection organization 0.00107 158 1005 BP GO:0032989 cellular component morphogenesis 0.00146 167 1090 BP GO:0032990 cell part morphogenesis 0.00182 123 728 BP GO:0045956 positive regulation of calcium ion-dependent exocytosis 0.00194 7 14 BP GO:0048858 cell projection morphogenesis 0.00205 121 711 BP GO:0000904 cell morphogenesis involved in differentiation 0.00251 127 742 BP GO:0043330 response to exogenous dsRNA 0.00282 7 28 BP GO:0017158 regulation of calcium ion-dependent exocytosis 0.00324 10 30 BP GO:0022412 cellular process involved in reproduction in multicellular organism 0.00342 35 225 BP GO:0006869 lipid transport 0.00342 36 233 BP GO:0021766 hippocampus development 0.00346 14 51 BP GO:0031175 neuron projection development 0.00348 124 731 BP GO:0022610 biological adhesion 0.00358 150 992 BP GO:0004629 phospholipase C activity 0.00360 8 30 MF GO:0045921 positive regulation of exocytosis 0.00389 11 41 BP GO:0007155 cell adhesion 0.00418 149 989 BP - 178 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Category Name p value over- Number of differentially Number in Ontology representation methylated category GO:0048013 ephrin receptor signalling pathway 0.00503 10 32 BP GO:0015020 glucuronosyltransferase activity 0.00510 8 27 MF GO:0002922 positive regulation of humoral immune response 0.00542 4 10 BP GO:0007409 axonogenesis 0.00597 91 500 BP GO:0061564 axon development 0.00604 93 517 BP GO:0007411 axon guidance 0.00633 67 351 BP GO:0097485 neuron projection guidance 0.00633 67 351 BP GO:0034389 lipid particle organization 0.00635 5 14 BP - 179 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4.4.3 Positive symptom severity 4.4.3.1 Global DNA methylation (positive symptom severity) Higher within-individual positive symptom severity at any time point was associated with higher global DNA methylation. An increase in positive symptom severity was associated with an average increase in global DNA methylation of 0.00023%; however, 50.96% of all CpG sites showed a decrease in methylation associated with positive symptom severity, and 49.04% showed an increase. 4.4.3.2 Differentially methylated positions (positive symptom severity) Three DMPs reached the threshold for epigenome wide significance (p<1*10-7), and two hundred and thirty four DMPs were trend-level significant (p<5*10-5). - 180 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes - 181 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Table 14: Top ten DMPs associated with within-individual variation in positive symptom severity (* indicates epigenome wide significance) CpG site Beta P value Gene Location MAPINFO Significant blood-brain methylation correlation reported? cg03661817 0.298x10-2 2.42E-08* KIFC3 Body;1stExon;5'UTR 57831745 Y – prefrontal cortex and superior temporal gyrus cg02759966 0.305 x10-2 3.24E-08* 29139086 N cg08283882 0.134 x10-2 9.63E-08* EBF2 Body 25901017 N cg26620021 -0.298 x10-2 1.29E-07 MIR641;AKT2 TSS1500;5'UTR 40788926 Y - cerebellum cg17790273 0.482 x10-2 1.71E-07 SPAG9 3'UTR 49040709 N Y – prefrontal cortex, entorhinal cortex, superior cg16926712 0.270 x10-2 2.02E-07 INTS9 Body 28657193 temporal gyrus and cerebellum cg26724545 0.395 x10-2 2.39E-07 C5orf58 1stExon;5'UTR 169659950 N cg10298215 0.269 x10-2 2.42E-07 HLA-DQB2 Body 32730046 Y – superior temporal gyrus cg22673724 0.184 x10-2 3.66E-07 ATL3 TSS1500 63439590 Y – entorhinal cortex cg10561983 0.312 x10-2 3.71E-07 23454432 N - 182 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes The most statistically significant DMP associated with positive symptom severity was cg03661817. Each within-individual increase in positive symptom severity was associated with an increase in methylation of 0.298% (p=0.2x10-7). Using uncorrected means there was a correlation of 0.0772 between positive symptom severity and methylation at any time point across all samples. cg03661817 is located within the gene body of KIFC3, which encodes kinesin family member C3. This gene is located on chromosome 16 and is expressed across numerous tissues including brain, with overexpression of protein in monocytes, peripheral blood mononuclear cells, testis and adipocyte. Kinesin family member C3 is a microtubule motor which plays a role in the formation, maintenance and remodelling of bipolar mitotic spindle. This CpG site is reported to show significant moderate correlations between methylation in blood and methylation in prefrontal cortex (r=0.294, p=0.011) and superior temporal gyrus (r=0.282, r=0.014). The second most statistically significant DMP associated with positive symptom severity was cg02759966. Each within-individual increase in positive symptom severity was associated with an - 183 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes increase in methylation of 0.304% (p=0.3x10-7). Using uncorrected means there was a correlation of 0.0227 between positive symptom severity and methylation at any time point across all samples. cg02759966 is located on chromosome 16, outside of any characterised genes. This CpG site is reported to show no correlation between methylation in blood and methylation in brain tissue. The third most statistically significant DMP associated with positive symptom severity was cg08283882. Each within-individual increase in positive symptom severity was associated with an increase in methylation of 0.134% (p=0.096x10-7). Using uncorrected means there was a correlation of 0.5210 between positive symptom severity and methylation at any time point across all samples. - 184 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes cg08283882 is located within the body of EBF2, a gene which encodes early b-cell factor 2. This gene is located on chromosome 8 and is expressed across numerous tissues including brain, with overexpression of mRNA in tibial nerve and adipose tissue and overexpression of protein in pancreas, salivary gland and oral epithelium. Early b-cell factor 2 is a non-basic helix-loop-helix transcription factor implicated in developmental processes such as osteoblast differentiation. This CpG site is reported to show no correlation between methylation in blood and methylation in brain tissue. The fourth most statistically significant DMP associated with positive symptom severity was cg26620021. Each within-individual increase in positive symptom severity was associated with a decrease in methylation of 0.298% (p=0.1x10-6). Using uncorrected means there was a correlation of -0.1283 between positive symptom severity and methylation at any time point across all samples. - 185 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes cg26620021 is located within the TSS1500 promoter region of MIR641, which encodes microRNA 641, involved in post-transcriptional regulation of gene expression. This gene is located on chromosome 19 and is expressed across numerous tissues including brain. This CpG site is reported to show a significant negative correlation between methylation in blood and methylation in cerebellum (r=-0.316, p=0.007). The sixth most statistically significant DMP associated with positive symptom severity was cg16926712. Each within-individual increase in positive symptom severity was associated with an increase in methylation of 0.27% (p=0.2x10-6). Using uncorrected means there was a correlation of 0.4073 between positive symptom severity and methylation at any time point across all samples. - 186 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes cg16926712 is located within the gene body of INTS9, which encodes integrator complex subunit 9. This gene is located on chromosome 8 and is expressed across numerous tissues including brain, with overexpression of protein in heart, placenta, breast and monocytes. Integrator complex subunit 9 binds to the C-terminal domain of RNA polymerase II and is therefore involved in gene expression and translational regulation. This CpG site is reported to show significant strong correlations between methylation in blood and methylation in prefrontal cortex (r=0.698, p=4.69e-12), entorhinal cortex r=0.562, p=3.34e-07), superior temporal gyrus (r=0.688 p=9.39e-12) and cerebellum (r=0.759, p=1.64e-14). The eighth most statistically significant DMP associated with positive symptom severity was cg10298215. Each within-individual increase in positive symptom severity was associated with an increase in methylation of 0.269% (p=0.2x10-6). Using uncorrected means there was a correlation of 0.2094 between positive symptom severity and methylation at any time point across all samples. - 187 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes cg10298215 is located within the body of HLA-DQB2, which encodes major histocompatibility complex, class II, DQ beta 2. This gene is located on chromosome 6 and is expressed in numerous tissues including brain, with overexpression of mRNA in skin and overexpression of protein in NK cells; variants of this gene have been associated with two sleep disorders: Kleine-Levin syndrome and recurrent hypersomnia. Major histocompatibility complex, class II, DQ beta 2 is involved in the immune system; it presents peptides derived from extracellular proteins, and is implicated in interferon gamma signalling. This CpG site is reported to show significant weak correlations between methylation in blood and methylation in superior temporal gyrus (r=0.26 p=0.024) with a trend for a significant correlation with entorhinal cortex r=0.231, p=0.053). The ninth most statistically significant DMP associated with positive symptom severity was cg22673724. Each within-individual increase in positive symptom severity was associated with an increase in methylation of 0.184% (p=0.4x10-6). Using uncorrected means there was a correlation of 0.285 between positive symptom severity and methylation at any time point across all samples. - 188 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes cg22673724 is located within the TSS1500 promoter region of ATL3, which encodes atlastin GTPase 3. This gene is located on chromosome 11 and is expressed across numerous tissues including brain, with overexpression of mRNA central nervous system and dorsal root ganglia neurons; variants have been associated with neuropathy. Atlastin GTPase 3 is involved in formation of endoplasmic reticulum tubules network. This CpG site is reported to show a significant weak correlation between methylation in blood and methylation in entorhinal cortex r=0.269, p=0.023). - 189 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4.4.4 Negative symptom severity 4.4.4.1 Global DNA methylation (negative symptom severity) Higher within-individual negative symptom severity at any time point was associated with higher global DNA methylation. An increase in negative symptom severity was associated with an average increase in global DNA methylation of 0.00207%. 51.44% of all CpG sites showed a decrease in methylation associated with negative symptom severity, and 48.56% showed an increase. 4.4.4.2 Differentially methylated positions (negative symptom severity) Fourteen DMPs reached the threshold for epigenome wide significance (p<1*10-7), and seven hundred and thirty six DMPs were trend-level significant (p<5*10-5). - 190 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes - 191 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Table 15: Top ten DMPs associated with within-individual variation in negative symptom severity (* indicates epigenome wide significance) CpG site Beta P value Gene Location MAPINFO Significant blood-brain methylation correlation reported? cg00293996 0.266x10-2 2.98E-09* CLDN14 5'UTR;TSS1500 37915391 N -0.174 Y – entorhinal cortex, superior temporal gyrus and cg21288861 x10-2 8.80E-09* ITGB3 TSS200 45331061 cerebellum -0.545 Y – entorhinal cortex cg02518691 x10-2 1.11E-08* ADAMTS8 Body 130297149 0.199 x10- N cg27116770 2 1.17E-08* TRIM54 1stExon 27505755 cg01371207 -0.78 x10-3 1.87E-08* PFDN6;WDR46 TSS200 33257335 N 0.393 x10- Y – prefrontal cortex and superior temporal gyrus cg03321142 2 2.04E-08* IGSF9 Body 159899640 0.232 x10- N cg08682625 2 3.28E-08* LOC727677 Body 128470793 -0.199 Y – prefrontal cortex, entorhinal cortex, and cg14660007 x10-2 3.31E-08* SLAIN1 TSS1500 78271894 cerebellum -0.374 N cg19060383 x10-2 3.43E-08* NPHS1 Body 36334925 -0.309 N cg05797770 x10-2 5.49E-08* EPHA2 Body 16470667 - 192 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes - 193 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes The most statistically significant DMP associated with negative symptom severity was cg00293996. Each within-individual increase in negative symptom severity was associated with an increase in methylation of 0.265% (p=0.3x10-8). Using uncorrected means there was a correlation of 0.313 between negative symptom severity and methylation at any time point across all samples. cg00293996 is located within the 5’UTR region of CLDN14, which encodes claudin 14. This gene is located on chromosome 21 and is expressed in numerous tissues including brain, with overexpression of mRNA in liver and overexpression of protein in urine. Claudin 14 is an integral membrane protein and component of tight junction strands, and defects can cause deafness. This CpG site is reported to show no significant correlation between methylation in blood and methylation in brain tissue, but a trend towards significant correlation between blood and cerebellum (r=0.224, p=0.0598). The second most statistically significant DMP associated with negative symptom severity was cg21288861. Each within-individual increase in negative symptom severity was associated with a - 194 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes decrease in methylation of 0.174% (p=0.9x10-8). Using uncorrected means there was a correlation of -0.556 between negative symptom severity and methylation at any time point across all samples. cg21288861 is located within the TSS200 promoter region of ITGB3 which encodes integrin beta chain beta 3. This gene is located on chromosome 17 and is expressed in numerous tissues including brain, with overexpression of mRNA in artery and colon and overexpression of protein in platelets, peripheral blood mononuclear cells and neutrophils. Integrin beta chain beta 3 is an integral cell- surface protein involved in cell adhesion and cell-surface mediated signalling. This CpG site is reported to show significant moderate correlations between methylation in blood and methylation in entorhinal cortex (r=0.392, p=0.0007), superior temporal gyrus (r=0.407, p=0.0003) and cerebellum (r=0.341, p=0.0036). The third most statistically significant DMP associated with negative symptom severity was cg02518691. Each within-individual increase in negative symptom severity was associated with a - 195 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes decrease in methylation of 0.545% (p=0.1x10-7). Using uncorrected means there was a correlation of -0.280 between negative symptom severity and methylation at any time point across all samples. cg02518691 is located within the gene body of ADAMTS8, a gene which encodes ADAM metallopeptidase with thrombospondin type 1 motif 8. This gene is located on chromosome 11 and is expressed across numerous tissues including brain, with overexpression of mRNA in lung and artery and overexpression of protein in heart. ADAM metallopeptidase with thrombospondin type 1 motif 8 is an enzyme involved in heparin binding and integrin binding. This CpG site is reported to show a significant moderate correlation between methylation in blood and methylation in entorhinal cortex (r=0.323, p=0.006), with a trend towards significance at the prefrontal cortex (r=0.22, p=0.060). The fourth most statistically significant DMP associated with negative symptom severity was cg27116770. Each within-individual increase in negative symptom severity was associated with an - 196 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes increase in methylation of 0.199% (p=0.1x10-7). Using uncorrected means there was a correlation of 0.365 between negative symptom severity and methylation at any time point across all samples. cg27116770 is located within the 1st exon of TRIM54, which encodes tripartite motif containing 54. This gene is located on chromosome 2 and is expressed across numerous tissues including brain, with overexpression of mRNA in skeletal muscle and heart, and overexpression of protein in heart. Tripartite motif containing 54 has been implicated in signal transducer activity and microtubule binding. This CpG site is reported to show no correlation between methylation in blood and methylation in brain tissue. The fifth most statistically significant DMP associated with negative symptom severity was cg01371207. Each within-individual increase in negative symptom severity was associated with a decrease in methylation of 0.078% (p=0.2x10-7). Using uncorrected means there was a correlation of -0.500 between negative symptom severity and methylation at any time point across all samples. - 197 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes cg01371207 is located within the TSS200 promoter region of PFDN6, which encodes prefoldin subunit 6. This gene is located on chromosome 6 and is expressed in numerous tissues including brain, with overexpression of protein in fetal brain and peripheral blood mononuclear cells. Prefoldin subunit 6 is implicated in unfolded protein binding, chaperone binding, and tubulin folding. This CpG site is reported to show no correlation between methylation in blood and methylation in brain tissue. The sixth most statistically significant DMP associated with negative symptom severity was cg03321142. Each within-individual increase in negative symptom severity was associated with an increase in methylation of 0.393% (p=0.2x10-7). Using uncorrected means there was a correlation of 0.521 between negative symptom severity and methylation at any time point across all samples. - 198 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes cg03321142 is located within the body of IGSF9, which encodes immunoglobulin superfamily member 9. This gene is located on chromosome 1 and is expressed in numerous tissues including brain, with overexpression of mRNA in liver, skin, and colon and overexpression of protein in breast. Immunoglobulin superfamily member 9 is implicated in cell-cell adhesion, dendrite outgrowth and synapse maturation. This CpG site is reported to show significant weak correlations between methylation in blood and methylation in blood and prefrontal cortex (r=0.3, p=0.009) and superior temporal gyrus (r=0.24, p=0.038). The seventh most statistically significant DMP associated with negative symptom severity was cg08682625. Each within-individual increase in negative symptom severity was associated with an increase in methylation of 0.232% (p=0.3x10-7). Using uncorrected means there was a correlation of 0.125 between negative symptom severity and methylation at any time point across all samples. - 199 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes cg08682625 is located within the gene body of LOC727677, a gene which encodes miscRNA and is located on chromosome 8 but is otherwise uncharacterised; variants have been associated with colorectal and prostate cancer. This CpG site is reported to show no correlation between methylation in blood and methylation in brain tissue. The eighth most statistically significant DMP associated with negative symptom severity was cg14660007. Each within-individual increase in negative symptom severity was associated with a decrease in methylation of 0.199% (p=0.3x10-7). Using uncorrected means there was a correlation of -0.389 between negative symptom severity and methylation at any time point across all samples. - 200 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes cg14660007 is located within the TSS1500 promoter region of SLAIN1, which encodes SLAIN motif family member 1. This gene is located on chromosome 13 and is expressed across numerous tissue including brain, with overexpression of mRNA in brain (spinal cord, substantia nigra and hippocampus) and overexpression of protein in monocytes and lung. This CpG site is reported to show significant weak correlations between methylation in blood and methylation in prefrontal cortex (r=0.231, p=0.048), entorhinal cortex (r=0.303, p=0.010) and cerebellum (r=0.393, p=0.0007). The ninth most statistically significant DMP associated with negative symptom severity was cg19060383. Each within-individual increase in negative symptom severity was associated with a decrease in methylation of 0.374% (p=0.3x10-7). Using uncorrected means there was a correlation of -0.396 between negative symptom severity and methylation at any time point across all samples. - 201 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes cg19060383 is located within the body of NPHS1, which encodes renal glomerulus-specific cell adhesion receptor. This gene is located on chromosome 19 and is expressed across numerous tissues including brain, with overexpression of mRNA in kidney and pancreas and overexpression of protein in plasma, urine and peripheral blood mononuclear cells. Renal glomerulus-specific cell adhesion receptor is an immunoglobin cell adhesion molecule involved in the filtration barrier in the kidney. This CpG site is reported to show no correlation between methylation in blood and methylation in brain tissue. The tenth most statistically significant DMP associated with negative symptom severity was cg05797770. Each within-individual increase in negative symptom severity was associated with a decrease in methylation of 0.309% (p=0.5x10-7). Using uncorrected means there was a correlation of -0.202 between negative symptom severity and methylation at any time point across all samples. - 202 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes cg05797770 is located within the gene body of EPHA2, which encodes ephrin type-A receptor 2. This gene is located on chromosome 1 and is expressed in numerous tissues including brain, with overexpression of mRNA in oesophagus and overexpression of protein in amniocyte and adrenal. Ephrin type-A receptor 2 is a receptor involved in mediating development of the nervous system and implicated in G-protein couple receptors, protein tyrosine kinase activity and transferase activity This CpG site is reported to show no correlation between methylation in blood and methylation in brain tissue. - 203 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4.5 Discussion: Therapeutic response to clozapine To explore the effects of clozapine on DNA methylation associated with therapeutic response, I took three key analysis approaches. Firstly, multi-level model analyses exploring between-individual variation in the change between baseline and twelve weeks as predictors of DNA methylation trajectories over time, with a model of categorical classification of responder or non-responder and then secondly three models of continuous percentage change in total symptom severity, psychosocial functional and quality of life. Then, multi-level model analysis of within-individual variation in symptom severity (total, negative and positive) was used as a predictor of variation in DNA methylation. Systems involving signal transduction, NMDARs, glycine receptors, calcium signalling, neuronal development, cell adhesion, glucose homeostasis, immune function, microtubules and transcription regulation are implicated. 4.5.1 Responders vs non-responders Analysis of the association between epigenome-wide DNA methylation and categorical response classification identified two DMPs that were significant at the epigenome-wide threshold (cg15353603 and cg18043267) and one hundred and sixty eight other DMPs which showed trends towards significance. Global DNA methylation levels suggest both groups start and end with equivalent levels of global DNA methylation and show the same pattern of an initial decrease and then increase, but non-responders show a more pronounced initial decrease. One of the top ten differentially methylated sites is located outside of genes, but of the nine located within genes six are expressed in brain tissue to some degree, and one is reported to display methylation in blood tissue which correlates to methylation in at least one brain region tissue. Again, a high proportion of the biological systems indicated by the top ten DMPs and/or top thirty pathways associated with categorical response classification have been previously demonstrated to be dysfunctional in schizophrenia, and many have also been shown to be influenced by antipsychotics including clozapine. Several of the overrepresented pathways and the first, seventh, eighth, ninth and tenth DMPs are located within genes (respectively: DPP4, PCDP1, DLG4, LCN10 and KCNK9) which are associated with different aspects of signal transduction – receptor binding and signalling, ion channels and potassium channels, ligand transportation, regulation of voltage-gated calcium channel activity, calmodulin binding, cation symporter activity and cell-cell adhesion – some of which have already been previously discussed in the context of schizophrenia and clozapine. - 204 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Calmodulin is a calcium modulated protein which acts in the calcium signal transduction pathway by modifying interactions with target proteins; this can involve binding to target proteins, many of which cannot bind to calcium themselves so use calmodulin as a calcium sensor and signal transducer. Clozapine, among other antipsychotics, has been shown to increase calmodulin proteins and mRNA levels in an amphetamine model of schizophrenia (Rushlow, Seah, Sutton, Bjelica, & Rajakumar, 2009). Calmodulin mediates inflammation, metabolism, and immune response, as well as increasing the sensitivity of AMPARs through phosphorylation, all of which are discussed elsewhere in this thesis in the context of clozapine and schizophrenia. More broadly, dysfunction of calcium signalling has been long been implicated in schizophrenia. One review (Lidow, 2003) proposes calcium signalling as a unifying molecular pathology in schizophrenia: calcium is involved in regulation of dopamine synaptic uptake, is regulated by dopamine receptors, can inhibit NMDAR via regulation of NMDA voltage gated calcium channels, is required for myelination-associated proteins, and is required for proper formation of synaptic contacts, dendritic trees, cell death, and neuronal migration and all of these processes have been found to be disrupted in schizophrenia. Another (Berridge, 2014) has further linked calcium signalling to the NMDAR dysfunction theory of schizophrenia, as there appears to be decreased entry of calcium through NMDARs and repetitive calcium signalling of NMDARs is involved in maintaining GABAergic inhibitory interneurons. Genetic studies support these theories, with post-mortem gene expression data, GWAS and polygenic risk scores all implicating calcium related processes such as voltage gated calcium channels (Hertzberg, Katsel, Roussos, Haroutunian, & Domany, 2015; Purcell et al., 2014; Ripke et al., 2014) An epigenetic study of mQTLs associated with both differential methylation and gene expression specifically identified an association between schizophrenia risk and a SNP which regulates calcium homeostasis (van Eijk et al., 2015). Clozapine has been shown to normalise disrupted calcium systems in DISC1 knockout models (Park et al., 2015) and inhibit the current in calcium channels, facilitating influx and altering cells excitability and firing properties through action on calcium channels (Choi & Rhim, 2010). Potassium channels are the most common type of ion channel, conducting potassium ions rapidly and selectively down the electrochemical gradient in excitable neuronal cells. They shape action potentials and regulate secretion of hormones. Mice knock out models without genes encoding a voltage-dependent potassium channel subunit show reduced proteins and behavioural and pharmacological abnormalities similar to the schizophrenia phenotype, some of which were normalised by clozapine (Peltola et al., 2016). Similarly, Duncan, Chetcuti and Schofield, (2008) found clozapine (among other antipsychotics) downregulated genes involved in potassium channel subunit expression. - 205 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Cell adhesion molecules are critical to neuronal development – including synapse formation, neural connectivity and axonal/dendrite growth – and they have been specifically implicated in previous schizophrenia research (Jones & Murray, 1991). Early studies reported decreased expression of embryonic neural cell adhesion molecules in post mortem hippocampal samples of schizophrenia patients (Barbeau et al., 1995) and more recent genetic studies indicate associations between schizophrenia and genes regulating cell adhesion molecules: pathway analysis of the International Schizophrenia Consortium data found that only genes associated with cell adhesion molecules survived correction for multiple testing (Corvin et al., 2010), and other studies have commented on novel risk loci associated with cell adhesion (Collier, Eastwood, Malki, & Mokrab, 2016; Ripke et al., 2014). Furthermore, Kinoshita et al., (2017) reported that DMPs associated with clozapine use in humans were enriched for cell-adhesion genes and animal studies have found that clozapine elevates the level of spectrins (which associate with neural cell adhesion molecules) (Kedracka-Krok et al., 2015), suggesting this is a process clozapine targets, and my analysis indicates this may be linked to clozapine response. The eight ranked DMP is located within DLG4, a gene involved in NMDAR function and the ratio of excitatory to inhibitory synapses in hippocampal neurons. While this result was not statistically significant or at a site where correlation between methylation in brain and blood is indicated, it is of interest because of the considerable literature on NMDAR involvement in schizophrenia. NMDARs are ionotropic glutamate receptors, which are activated when glutamate and glycine bind to them and allow positively charged ions to flow through the cell membrane, and which NMDA selectively binds to. They are important for postsynaptic mediation of activity dependent synaptic plasticity and memory. Olney, Newcomer and Farber (1999) proposed the NMDAR hypofunction model of schizophrenia, partly based on early studies showing that NMDA channel blockers produced positive, negative and cognitive symptoms of schizophrenia in healthy controls (Krystal Karper, et al., 1994). Ketamine or PCP has since been used to model NMDAR hypofunction, which appears to increase glutamate levels in the short term but reduce levels in the long term, with upregulation of NMDA transporters. This leads to a schizophrenia-like phenotype: hyperactive subcortical dopamine and hypoactive prefrontal dopamine, reduced GABA synthesis enzyme levels, and neurotoxic effects (Jentsch & Roth, 1999). Evidence since then has further supported the hypothesis of NMDAR hypofunction in schizophrenia, with post-mortem samples and animal studies indicating hypofunction (Balu & Coyle, 2015) and genetic studies indicating disruption in NMDAR and glutamatergic transmission pathways (Allen et al., 2008; Ripke et al., 2014). There has been some suggestion that glutamatergic dysfunction is particularly true or exclusively true in treatment- resistant patients, especially clozapine-responsive patients (Gillespie et al., 2017; Goldstein, - 206 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Anderson, Pillai, Kydd, & Russell, 2015), while dopaminergic abnormalities are predominant in patients responsive to non-clozapine antipsychotics. This may explain why trials of glutamatergic drugs have found no overall effects, but substantial differences between sites (Buchanan et al., 2007). Interestingly, clozapine has been shown to be much more effective in alleviating symptoms in PCP models of schizophrenia than haloperidol (Jentsch & Roth, 1999), perhaps due to partial agonism of glycine-sites on NMDAR (Schwieler, Engberg, & Erhardt, 2004), enhanced activation of NMDAR through both increased release of D-serine and glutamate by glia (Tanahashi, Yamamura, Nakagawa, Motomura, & Okada, 2012) and antagonistic effects on glutamate causing NMDAR upregulation (Veerman, Schulte, Begemann, Engelsbel, & de Haan, 2014). DLG4 has previously found to be hypermethylated in schizophrenia, in a sample of patients either on clozapine or risperidone (Liao et al., 2014); in my analysis, responders started with a higher baseline level of methylation and ended up showing increases in methylation (increase, decrease, increase) while non-responders showed significant decreases (increase first). Pathway analysis indicated overrepresentation of DMPs involved in regulation of GABA transport. GABA is the most common inhibitory neurotransmitter in the CNS and a reduced level of mRNA encoding GAD67 (which regulates GABA synthesis, dependent upon the amount of excitatory activity received by GABA neurons) in the DLPFC is “one of the most replicated observations in post- mortem studies of schizophrenia”, alongside evidence for increased uptake of GABA due to decreased GABA transporters (Lewis & Sweet, 2009). Dysfunction of parvalbumin, a calcium binding protein involved in GABAergic interneuron signalling, was discussed in the previous chapter as I found methylation of a DMP in PVALB to be associated with clozapine levels. GABAergic dysfunction may be due to NMDAR hypofunction on GABAergic interneurons; this leads to a defective ability to respond to glutamate, with increased glutamate release exciting inhibitory interneurons to release GABA, which in turn increases production of NMDAR antagonists leading to a reduction in NMDAR function and further increased glutamate release i.e. a defective feedback loop. Genetic studies of schizophrenia have implicated associations with genes in GABA pathways, suggesting a genetic basis for this dysfunction (Collier et al., 2016). Furthermore, Ohno-Shosaku et al., (2011) found that, in cultured hippocampal neurons, clozapine depressed inhibitory transmission and post synaptic potentials, by antagonising postsynaptic GABA receptors and inhibiting presynaptic calcium and sodium channels. The top DMP was within DPP4, a gene involved in regulation of immune function via T-cell receptors. The relationship between schizophrenia and clozapine and immune function was - 207 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes discussed above, but T cells specifically interact with MHC molecules (also called HLA complex) which are receptors that aid identification of antigens. Debnath, (2015) specifically singles out altered T cell function in schizophrenia, as T cells can infiltrate the brain and dopamine can modulate the proliferation, trafficking and function of T cells, MHC pathways have been implicated in schizophrenia by some of the largest genetic studies to date (Ripke et al., 2014), and clozapine appears to regulate T cell levels (Chen et al., 2011; Fernandez-Egea et al., 2016; Steiner et al., 2010). DPP4 is also associated with glucose homeostasis, which as discussed above, has been implicated in schizophrenia and clozapine; similarly, the sixth DMP was within ABCG1, a gene associated with ATP transportation, and the tenth DMP was within a gene associated with intellectual disability, KCNK9. The candidate gene testing provided only weak support for an association between classification as non-responder to clozapine and differential methylation at CpG sites on DRD3, 5HTT, 5HTR2A and CREBBP; no associations survived correction for multiple testing. Results from this model provide some replication of results from previously collected cross- sectional data, with significant positively correlated direction of effect for their top DMPs. Similarly, results from the cross-sectional data provided replication of results from my longitudinal clozapine levels analysis, with significant positively correlated direction of effect for my top DMPs. As a positive correlation indicates a common direction of effect associated with non-response in the current sample, this could indicate high levels of non-response in the cross-sectional sample or again it could reflect differing methylation patterns over time (blood may have been taken before six weeks or after six months of clozapine exposure, and this may show a different relationship with response to clozapine). As in the previous chapter, some of the significant DMPs appear to be driven by outliers, and this should be considered when drawing conclusions. For example, methylation at both cg18043267 (the second most significant DMP) and cg04408162 (the tenth most significant DMP) rose sharply in one non-responder towards the end of the study, whereas for the rest of the sample the differences in methylation are much subtler. It may be that these CpG site were of significance for the clinical response in these individuals or, perhaps more likely, they may be spurious findings and thus should be treated with caution. They are presented here in this exploratory analysis but future work looking to replicate and find common patterns across groups must consider how to tackle outliers. - 208 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4.5.2 Change in total symptoms (continuous) Analysis of the association between epigenome-wide DNA methylation and change in total symptom severity identified two DMPs that were significant at the epigenome-wide threshold (cg10015974 and cg15353603), and one hundred and eighty eight other DMPs which showed trends towards significance. The majority of the top ten DMPs show additional increases in methylation associated with increased percentage reduction and global DNA methylation indicates this pattern also. Two of the top ten DMPs are located outside of genes, but of the eight located within genes seven are expressed in brain tissue to some degree, and three sites are reported to display methylation in blood tissue which correlates to methylation in at least one brain region tissue. Again, a high proportion of the biological systems indicated by the top ten DMPs and/or top thirty pathways associated with change in total symptom severity have been previously demonstrated to be dysfunctional in schizophrenia, and many have also been shown to be influenced by antipsychotics including clozapine. One gene in particular – SETD1A – has been previously associated with schizophrenia. Two of the top DMPs for continuous change in symptom severity were also within the top ten DMPs for categorical response classification, unsurprisingly as these analysed the same variables but in categorical and continuous forms: cg15353603 in DPP4 and cg06275059 in PCDP1, genes implicated in glucose homeostasis, receptor binding, immune activation via T cells, and calmodulin binding. Similarly, the seventh DMP was in ANO1, a gene associated with calcium chloride channel activity – while this does not directly overlap with the categorical response results, both implicate an effect on calcium signalling which – as discussed above – is thought to be disrupted in schizophrenia and potentially normalised by clozapine. ANO1 is also associated with glucose regulation, which was discussed in Chapter 3 as this system was implicated in results for clozapine levels. The tenth ranked DMP was within SETD1A. While this CpG site was not statistically significant and is not indicated to have correlated methylation between blood and brain, SETD1A is a gene which has specifically been associated with schizophrenia and so is of particular interest. SETD1A encodes a subunit of a histone methyltransferase protein complex, which mediates methylation of histone 3 at Lysine 4 (H3K4), a marker of active gene transcription. An association between schizophrenia and loss-of-function variants of SETD1A was first identified by an exome sequencing of 231 parent and proband trios (Takata et al., 2014), and this was confirmed by a study of over 4,000 schizophrenia cases (Singh et al., 2016) - 209 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Histone methylation processes are strongly associated with psychiatric disorders, including schizophrenia – as identified by pathway analysis of mega-analyse of genomic data - (Consortium, 2015), and H3K4 tri-methylation appears involved in neurodevelopment, and regulation of neuronal signalling - specifically GABAergic, glutamatergic and dopaminergic signalling (Dincer et al., 2015). Clozapine inducing an upregulation in H3K4 tri-methylation of GAD1 was one of the key findings by Huang et al., (2007), and this effect may come about through disruption of regulatory genes like SETD1A. The third DMP was in TRIM31, a gene associated with immune function via interferon gamma signalling. The general association between the immune system and both schizophrenia and clozapine is described above; interferon signalling specifically involves a type of cytokine made and released by host cells in response to pathogens, causing nearby cells to heighten defences. This includes upregulating antigen presentation by increasing MHC expression, leading to fever and other flu-like symptoms. Interferon gamma is a type II class interferon produced predominantly by T cells and is critical for innate and adaptive immunity, with immune-stimulatory and immunomodulatory effects and the ability to alter transcription of up to 30 genes. It is one of the cytokines which Song et al, (2000) found to increase in response to clozapine. TRIM31 is also involved in ubiquitin proteins and ligase activity, both systems discussed in Chapter 3, associated with time on clozapine. 4.5.3 Change in psychosocial functioning Analysis of the association between epigenome-wide DNA methylation and change in psychosocial functioning identified one DMP that was significant at the epigenome-wide threshold (cg05195750) and one hundred and twenty two other DMPs showed trends towards significance. The majority of the top ten DMPs show additional increases in methylation associated with greater improvements in psychosocial functioning and global DNA methylation shows the same pattern. One of the top ten DMPs is located outside of any genes, but of the nine located within genes all are expressed in brain tissue to some degree, and four sites are reported to display methylation in blood tissue which correlates to methylation in at least one brain region tissue. Again, a high proportion of the biological systems indicated by the top ten DMPs and/or top thirty pathways associated with change in psychosocial functioning have been previously demonstrated to be dysfunctional in schizophrenia, and many have also been shown to be influenced by antipsychotics including clozapine. - 210 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes The first DMP in ELP4, a histone acetyltransferase associated with chromatin organisation and risk of epilepsy, was also in the top ten DMPs associated with time on clozapine and is discussed in Chapter 3. Similarly, the fifth DMP was in IQCE, a gene associated with GPCRs and the sixth DMP was in OLFM2, a gene associated with AMPAR organisation: both of these are systems which were implicated in exposure to clozapine analysis and their dysfunction in schizophrenia was discussed in Chapter 3 also. The second DMP was in TRIM25, a gene associated with interferon signalling and the seventh DMP was in CLEC10A, a gene associated with inflammation. While the DMP in CLEC10A was not statistically significant or in a location indicated to show correlated methylation between blood and brain, it is not only associated with inflammation but carbohydrate binding (the ninth DMP in UGP2 is associated with carbohydrate conversion) and with cell adhesion, a process implicated by studies of genetic risk for schizophrenia. The third DMP is in PKNOX, a gene associated with cell proliferation. Changes in methylation of genes involved in systems of immune function, abnormal cell development, glucose homeostasis and cell adhesion were all implicated by analysis of change in total symptom severity and are discussed above also. PKNOX is also associated with substance dependence. A prospective population based study using Danish registers found that substance abuse was associated with an six fold increased risk of schizophrenia, with risk especially high for cannabis and alcohol and significant effects even after 15 years (Nielsen, Toftdahl, Nordentoft, & Hjorthøj, 2017). Similarly, a review (Thoma & Daum, 2013) reports that around half of patients with schizophrenia have comorbid substance dependence, particularly men with early onset. This review outlines numerous explanations: diathesis-stress model, in which drug use triggers genetic risk; cumulative risk factor model, in which poor cognitive abilities and low functioning lead to drug use; self-medication hypothesis, in which patients abuse drugs in an attempt to alleviate symptoms; affect regulation model, in which personality traits such as impulsivity and maladaptive coping mechanisms lead to drug use; altered reward processing hypothesis, in which patients overvalue the positive consequences of drugs; or the developmental dysfunction hypothesis, in which developmental deficits lead to a predisposition for both. Interestingly, studies of polygenic risk have found that the polygenic risk score for schizophrenia is associated with substance abuse disorders suggesting common genetic risk (Hartz et al., 2017). Some studies have reported that clozapine may be particularly efficacious for substance abuse, with a systematic review reporting that clozapine was superior in all existing comparison studies (Machielsen & de Haan, 2009); they suggest this may be due to its profile of low D2 receptor occupancy, high dissociation rate and a high D1/D2 receptor ratio. Mesholam-Gately, Gibson, Seidman, & Green (2014) suggest that clozapine may ameliorate dysfunction of the brain reward - 211 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes circuit, demonstrating that patients on clozapine show broader and stronger responses to rewarding stimuli. Interestingly, some studies report that patients with a history of substance abuse are better responders to clozapine treatment, even if they are equivalent in other characteristics and clinical severity at baseline (Buckley, Thompson, Way, & Meltzer, 1994; Kelly, Gale, & Conley, 2003) The fourth ranked DMP was in EGR2, a gene associated with disorders of hypo-myelination. Hypo-myelination has been frequently reported in schizophrenia (Carter, 2007), a recent review highlights myelination abnormalities as a key etiological mechanism of schizophrenia present from onset (Stedehouder & Kushner, 2016) and animal studies support that hypo-myelination is associated with schizophrenia like behaviours (e.g. impaired PPI) (Ishimoto et al., 2017) . A review (Corfas, Roy, & Buxbaum, 2004) suggests that hypo-myelination leads to reduced or irregular neural transmission (perhaps an adaptation of the brain, altering synaptic release or expression of receptors and other molecules), which then leads to dysfunctional perception and emotional processing, and/or that hypo-myelination may “unmask” other dysfunction in brain structure or connectivity. A review of genetic and neuroimaging studies reports that genes regulating myelination and oligodendrocyte function appear dysregulated in schizophrenia, and that density and distribution of oligodendrocytes in PFC appears altered along with reduced myelin content (Woo & Crowell, 2005), and recent pathway analysis of genetic data Duncan et al., (2014) implicates the glia-oligodendrocyte pathway in schizophrenia . Maas, Vallès and Martens, (2017) suggested that hypo-myelination may be caused by oxidative stress (triggered by genetic and environmental factors) impairing signal transduction processes for oligodendrocytes, necessary for the adolescent PFC myelination process. In comparison to haloperidol, clozapine has been found to enhance oligodendrocyte glucose metabolism, improving the energy supply and myelin lipid synthesis, leading to greater white matter integrity (Steiner et al., 2014) TRIM25, PKNOX, and EGR2 are also all genes for transcription factors, again highlighting the potential role of clozapine in regulating gene expression indirectly through methylation of genes which regulate transcription. 4.5.4 Change in quality of life Analysis of the association between epigenome-wide DNA methylation and change in quality of life identified one DMP (cg14150252) that was significant at the epigenome-wide threshold and one hundred and forty eight DMPs showing trends towards significance. The majority of the top ten DMPs showed additional increases in methylation associated with increased changes in quality of life and global DNA methylation shows this association also. One of the top ten DMPs is located outside of genes, but of the nine located within genes all are expressed in brain tissue to some degree, and - 212 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes six are reported to display methylation in blood tissue which correlates to methylation in at least one brain region tissue. Again, a high proportion of the biological systems indicated by the top ten DMPs and/or top thirty pathways associated with change in quality of life have been previously demonstrated to be dysfunctional in schizophrenia, and many have also been shown to be influenced by antipsychotics including clozapine. Three of the DMPs have been within the top ten DMPs of other analysis (cg14150252 in UGP2 associated with psychosocial functioning, cg05195750 in ELP4 associated with psychosocial functioning and time, and cg07829465 in TRIM31 associated with response classification). Furthermore, many of the individual DMPs and over-represented pathways implicate systems also implicated in previous analysis. The third DMP was in SLC9A3, a gene associated with sodium ion transport and signal transduction. The sixth DMP was in BTRC, a gene associated with ubiquitination and ligase activity. The seventh DMP was in EHF, a transcription repressor. The eight DMP was in SRL, a gene associated with calcium transport. Changes in methylation of genes involved in signal transduction, ubiquitination, ligase activity, regulation of transcription and calcium transport were all systems implicated by analysis of time exposed to clozapine and change in total symptom severity and are discussed above also. However, there were some novel systems implicated as well. BTRC is also associated with immune function and specifically interleukin (IL) signalling; interleukins are a group of thirty six cytokines first expressed by leukocytes, which the immune system relies upon for various forms of regulation, inhibiting or increasing production of other cytokines or lymphocytes. A meta-analysis of IL10 polymorphisms (IL10 inhibits synthesis of a number of cytokines) found an association with schizophrenia (Gao, Li, Chang, & Wang, 2014), and clozapine has been shown to increase IL10 (Ajami et al., 2014) ; clozapine has also been shown to inhibit neutrophil migration via IL8 (Capannolo et al., 2015), and both increase IL6 (B cell stimulatory factor) in healthy controls (Song et al., 2000) but decrease IL6 in macrophage cell cultures (Kracmarova & Pohanka, 2014). 4.5.5 Total symptom severity Analysis of the association between epigenome-wide DNA methylation and total symptom severity identified five DMPs that were significant at the epigenome-wide threshold (cg02890365, cg23158483, cg26129353, cg17651451 and cg23290344) and four hundred and fifty three other DMPs showing trends towards significance. The majority of the top ten DMPs showed decreases in methylation associated with total severity, however global DNA methylation showed an increase. Two of the top ten differentially methylated sites are located outside of genes, but of the eight - 213 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes located within genes seven are expressed in brain tissue to some degree, and four sites are reported to display methylation in blood tissue which correlates to methylation in at least one brain region tissue. The first and fifth ranked DMP are within TUFT1 and NEFM, genes implicated in neuronal differentiation and neuronal calibre (diameter, degree of myelination) and pathway analysis also indicated overrepresentation of DMPs in neuron project development and guidance, and axonogenesis and axon development and guidance specifically, and highlighted development in the limbic system and hippocampus. Furthermore, analysis indicated above that DMPs on genes involved in neuronal plasticity were associated with clozapine level. Abnormalities in neuronal development processes have long been implicated in schizophrenia, with an early review (Jones & Murray, 1991) highlighting not only structural abnormalities but potential defects in cell proliferation, migration, connectivity, axonal growth, myelination and neuronal density. Post-mortem studies (Lewis & Sweet, 2009) indicate increased neuronal density due to reductions in axon terminals and dendritic spines, lowered gene expression for proteins present in axon terminals, dendritic length and spine density, and suggest there may be exaggerated synaptic pruning during adolescence or deficits which adolescent synaptic pruning reveals. DISC1 knock out studies have resulted in neurons have aberrant development and dispositioning, enhanced excitability and accelerated formation, suggesting this gene (long associated with schizophrenia) regulates progress of neuronal integration (Duan et al., 2007); DISC1 appears to exert its affect via numerous protein interaction partners involved in neuronal migration, neural progenitor proliferation, neuron signalling and synaptic function (Bradshaw & Porteous, 2012). Interestingly, clozapine has been shown to reverse DISC-1 induced disruption of systems which regulate of axonal growth (Park et al., 2015). A number of studies have observed neurons developed from human induced pluripotent stem cells of patients with schizophrenia, and found a wide range of abnormalities: diminished neuronal connectivity and decreased neurite numbers, as well as decreased glutamate receptor expression and altered expression of Wnt signalling pathways (Brennand et al., 2011) (both discussed above); cytoskeletal remodelling, aberrant migration, increased oxidative stress, and dysfunctional neuronal maturation and cell adhesion gene expression (Brennand et al., 2015); differential expressed genes, copy number variants and differential methylation, with differential expression pathways (validated by post mortem study) indicating genes associated with neurodevelopment (Maschietto et al., 2015); and dopaminergic cells showing an impaired ability to differentiate and glutamatergic cells unable to mature, and disruption of mitochondrial connectivity, responsiveness and structure (which - 214 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes modulates morphogenesis and plasticity of spines and synapses) related to both (Robicsek et al., 2013). It is worth noting that a) many of these studies involve very small sample sizes and therefore may not reflect schizophrenia as a whole and b) cultured cells are more likely to have genomic instability and mutations, but there appears to be consistent implication of abnormal neuronal development in stem cells taken from schizophrenia patients. Some (Selemon & Zecevic, 2015) have suggested that prenatal maternal exposure or genetic liability limits early gestational cell proliferation, and then stress from adverse social situations in adolescence increases dopamine stimulation of meso-cortical pathways and leads to exaggerated synaptic pruning, and the combination of these two hits leads to schizophrenia; this is partly based on animal models showing that limiting early gestational cell proliferation can lead to schizophrenia- like behaviour, and dopamine hyper stimulation decreases pyramidal cell spine density and induces cognitive changes. These abnormalities may underlie dysfunction in signal transduction; or they may be caused by signalling dysfunction. The third DMP is within ATP8B1, a gene involved in ion channel transport, NEFM1 is involved in intracellular transport to axons and dendrites, and the tenth DMP is within CORO1A, involved in signal transduction generally. Additionally, pathway analysis also indicated overrepresentation of DMPs in genes involved in calcium ion binding and cell adhesion; cell adhesion specifically was the most over-represented pathway in this analysis and one of the most statistically significant across all analysis in this thesis. Different aspects of signal transduction have been implicated in all of the analyses, and have been discussed previously. The second ranked DMP is within the promoter region of GLRA1, which encodes a subunit of the glycine receptor. Glycine receptors are ionotropic receptors which produce fast responses when chloride current enters the neuron and induces hyperpolarisation; they are one of the most widely distributed inhibitory receptors, but there are also extra-synaptic excitatory glycine receptors which control migration of interneurons via activation of voltage-gated calcium channels, and some co- localise with GABA-A receptors on some hippocampal neurons. They are activated by a range of simple amino acids including glycine (biosynthesised from serine) but also b-alanine and taurine. The glycine system interacts with the glutamate system, as glycine is a required co-agonist for NMDARs and levels of occupancy and reuptake modulate calcium dependent neuronal excitation. A review of recent genetic studies highlighted noteworthy loci associated with schizophrenia which included SLC32A1, involved in glycine transport and localised to inhibitory terminals of glycinergic neurons, and NLGN4X which is localised to glycinergic post-synapses and regulates the number of glycine receptors (Collier et al., 2016). Yamamori et al., (2014) found that d-serine (from which L-serine is - 215 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes synthesised) and D/L serine ratios are lower in schizophrenia patients, and the ratio increases and normalises as symptoms improve and with clozapine treatment. A meta-analysis of clinical trials in schizophrenia found that glycine and d-serine particularly improve negative symptoms (Tuominen, Tiihonen, & Wahlbeck, 2005), and d-serine has been found to improve all PANSS measures in treatment-resistant schizophrenia as an adjunct to risperidone or olanzapine, with a third of patients becoming responders to antipsychotics (>20% reduction in BPRS) (Heresco-Levy et al., 2005) . D-cycloserine has also been shown to improve negative symptoms and memory consolidation (Goff et al., 2008), and D alanine has been shown to significantly improve CGI and PANSS scores (Tsai, Yang, Chang, & Chong, 2006), both as adjuncts. Tuominen et al. (2005) note that trials tend to show no improvement when glycine agonists are adjuncts to clozapine e.g. (Evins, Fitzgerald, Wine, Rosselli, & Goff, 2000), and suggest this is because clozapine is already working via this mechanism of action, a suggestion also made by others (Harvey & Yee, 2013; Tsai et al., 1999) who found no effect of serine on symptoms in clozapine treated patients and say that this is consistent with clozapine already activating the NMDA glycine receptor. The idea that clozapine may be an NMDAR glycine receptor agent is supported by animal studies which find that clozapine has similar effects to d-serine and glycine transporter inhibitors in restoring disrupted prepulse and latent inhibition (Lipina, Labrie, Weiner, & Roder, 2005). GLRA1 has also been associated with startle disease, an exaggerated startle response. Deficits in pre-pulse inhibition (in which a weaker pre-stimulus inhibits reaction to subsequent stronger stimuli) have been widely replicated in schizophrenia, with a recent review suggesting that underlying schizophrenia pathology is a cognitive, motor and sensory gating impairment (an inability to filter out redundant or unnecessary stimuli) (Swerdlow, Braff, & Geyer, 2016). This review also noted that intact PPI appears to be a good predictor of response, suggesting that deficits may be particularly relevant for treatment-resistant patients. Interestingly, in animal models, clozapine has been shown to reverse PPI deficits induced by dopamine agonists (Swerdlow & Geyer, 1993) or PCP (Bakshi, Swerdlow, & Geyer, 1994), with the later study demonstrating that this effect could not be attributed solely to antagonism of D1, D2 or 5-HT2 receptors. The tenth ranked DMP is located within CORO1A, a gene involved in immune function. The fourth ranked DMP is within ADAMTS2, a gene associated with collagen biosynthesis. The third DMP is located within ATP8B1, a gene involved in nucleotides, and CORO1A is involved in gene regulation. Changes in methylation of genes involved in immune function, , collagen biosynthesis and gene regulation were all systems implicated by analysis of time exposed to clozapine and change in total symptom severity or psychosocial function and are discussed above. - 216 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes 4.5.6 Positive symptom severity Analysis of the association between epigenome-wide DNA methylation and PANSS positive severity identified three DMPs that were significant at the epigenome-wide threshold (cg03661817, cg02759966 and cg08283882) and two hundred and thirty four other DMPs showing trends towards significance. The majority of the top ten DMPs and global DNA methylation showed increases in methylation associated with higher positive symptom severity. Two of the top ten DMPs are located outside of genes, but of the eight located within genes seven are expressed in brain tissue to some degree, and four sites are reported to display methylation in blood tissue which correlates to methylation in at least one brain region tissue. As previously noted for other analyses, some of the significant DMPs may be driven by outliers – this is most evident in the second DMP for this analysis, cg02759966, in which one individuals methylation levels are starkly different from the rest of the sample. The first ranked DMP is within KIFC3, a gene associated with microtubule and bipolar mitotic spindles, necessary for chromosomal segregation; my analysis indicated DMPs involved with beta- tubulin and microtubule function were also associated with exposure to clozapine, as described above. The eight ranked DMP is within HLA-DQB2, a gene associated with sleep disorders. Many psychiatric disorders include sleep disturbances, but early reviews of evidence suggested that disturbances were especially pronounced in schizophrenia (Benca, Obermeyer, Thisted, & Gillin, 1992) and more recent reviews confirm that sleep disturbances are common in schizophrenia patients, even from prodromal stages (Cohrs, 2008): this seems to come in the form of reduced sleep efficiency and total sleep time, and increased sleep latency (time to fall asleep), as well as potential differences in REM and slow wave sleep. The latter review noted that clozapine specifically appears to normalise these disturbances, and while this could be explained as a secondary effect of symptom reduction, the same direction of effect is seen in healthy controls suggesting a direct effect. The ninth ranked DMP is within ATL3, a gene associated with endoplasmic reticulum tubules – these are tubules within the network of intracellular membranes which synthesise and export proteins and membrane lipids. One of this network’s many functions is to regulate intracellular calcium homeostasis which, as discussed above, is disrupted in schizophrenia and affected by clozapine. It also interacts with the ubiquitin system, again, discussed above in context of disruption in schizophrenia. ATL3 has also been associated with neuropathy - nerve disorders of sensory, motor or autonomic systems. There is limited research linking this to schizophrenia, but (Neale, Olszewski, - 217 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes Gehl, Wroblewska, & Bzdega, 2005) discusses a common disruption of N-acetyl-aspartyl-glutamate in models of both neuropathic pain and schizophrenia, with NAAG peptidase inhibitors modulating both models. HLA-DQB2 is involved in immune function, in particular interferon gamma signalling, as part of the MHC. The third, fourth, and sixth ranked DMPs are within EBF2, MIR641 and INTS9, which are implicated in transcription regulation, micro RNA and RNA polymerase function. Both immune function and regulation of gene expression are systems which have been implicated by analysis of DMPS associated with exposure to clozapine and change in symptom severity, and have been discussed above. 4.5.7 Negative symptom severity Analysis of the association between epigenome-wide DNA methylation and negative symptom severity identified fourteen DMPs that were significant at the epigenome-wide threshold (cg00293996, cg21288861, cg02518691, cg271166770, cg01371207, cg03321142, cg08682625, cg14660007, cg19060383, cg05797770, cg17502527, cg00919806, cg05340048 and cg10594105) as well as seven hundred and thirty six other DMPs which showed trends towards significance, making this the analysis with the most statistically significant results within this thesis. The majority of the top ten DMPs showed decreases in methylation associated with negative symptom severity but global DNA methylation showed an increase. None of the top ten DMPs are located outside of genes, and of the ten located within genes nine are expressed in brain tissue to some degree, and four sites are reported to display methylation in blood tissue which correlates to methylation in at least one brain region tissue. These sites implicate pathways including synapse maturation, dendrite outgrowth, G-protein couple receptors, kinase activity, cell adhesion, cell-surface mediation signalling, signal transducer activity, RNA, microtubule binding, heparin binding, integrin binding, chaperone binding, tubule folding, kidney filtration. They have also been associated with deafness and colorectal and prostate cancer. The first ranked DMP is within CLDN14, a gene which has been associated with deafness. Early studies (Cooper, 1976) drew a link between schizophrenia and deafness, noting that samples of older patients with paranoid psychosis have a particularly high proportion who are hard of hearing, and suggesting that this may lead to social isolation or dysfunction with subsequent psychological impact, or that predisposed individuals with hearing impairments may misperceive auditory stimuli and develop subsequent inappropriate associations. Minimal research has been done in recent - 218 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes years, but one review (S Leucht et al., 2007) found two studies investigating an association between schizophrenia and ear disease or hearing loss, both of which reported positive findings. The sixth and tenth ranked DMP are within IGSF9 and EPHA2, two genes associated with synapse maturation, dendrite outgrowth and the development of the nervous system, some of which was discussed above in the context of schizophrenia and clozapine. Research has specifically investigated dendritic growth in schizophrenia; dendrites integrate synaptic inputs (up to 100,000) and determine the production of action potentials, and dendritic spines increase the ability to isolate signal specificity. They change in response to neural activity and long term potentiation and affect the communication and processing of cells. Combining PGC genetic data and proteomic data of schizophrenia samples has indicated disruptions in dendritic spine development (Pers et al., 2016), supporting previous models linking DISC1 to dendritic development (Duan et al., 2007; Kim et al., 2012) and one review of post-mortem studies which indicated that reduced density of dendritic spines was one of the most consistent findings in schizophrenia e.g. (Konopaske, Lange, Coyle, & FM, 2014), with some suggestion of reduced dendritic arborisation as well (Moyer, Shelton, & Sweet, 2015). This appears to be significant in schizophrenia as accelerated reduction in grey matter during adolescence is one of the predictors for transition to psychosis. Alterations in dendritic growth may be due to the disruption in schizophrenia of microtubules, calcium channel signalling and/or myelination discussed above. Critchlow et al., (2006) report that in cultures of rat hippocampal neurons clozapine increased primary dendritic spine density by 59%, whereas haloperidol led to reduced density. Similarly, Barros et al., (2009) report that clozapine was able to reverse dendritic spine deficits and associated behavioural deficits in gene knock out studies of ErbB2/B4-mediated NRG1 signalling. In contrast, Jaehne et al., (2015) found that clozapine was unable to reverse dendritic spine deficits in 14-3-3 knock out mice, but suggests that the cause of dendritic spine loss may be crucial to clozapine’s abilities to reverse this. EPHA2 is also associated with GPCRs and kinase activity, IGSF9 is associated with cell adhesion, and the second and fourth DMPs are within ITGB3 and TRIM54 which are genes associated with cell adhesion and signal transduction; all of these processes have been implicated and discussed above. ITGB3 and ADAMTS8, in which the third ranked DMP is located, are both genes associated with integrin binding. Integrins are produced by a wide variety of cells and play a role in cell adhesion and signal transduction primarily for extracellular matrix components. Integrins form cell adhesion complexes with other proteins to regulate kinase signalling and circulating leukocytes have cell membrane integrins which maintain their inactivity until inflammatory response is required. - 219 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes The ninth ranked DMP is within NPHS1, a gene associated with kidney filtration via immunoglobin cell adhesion molecules. Kidneys filter blood to make urine and regulate water, waste, nitrogen, ion concentrations, hormones, acid-base balance and blood pressure. The seventh ranked DMP is within LOC727677, a gene associated with colorectal and prostate cancer. One review has reported increased rates in schizophrenia, particularly associated with antipsychotic use (Seeman, 2010); however, the majority of individual studies report decreased rates of most cancers including colorectal or prostate cancer (Dalton et al., 2006; Howard et al., 2010; Lichtermann, Ekelund, Pukkala, Tanskanen, & Lönnqvist, 2001; Mortensen, 1994). TRIM54, PFDN6 – in which the fifth DMP is located - and LOC727677 are all genes associated with different aspects of gene and protein expression: miscRNA, microtubule binding and folding, and chaperone binding (proteins assisting primarily in protein folding and assembly). 4.5.8 Summary To summarise, two DMPs were epigenome-wide significant for an association with responder status - cg15353603 which is located within DPP4 (regulates glucose homeostasis and immune function) and cg18043267 which is located within C10orf125 (involved in cell-cell adhesion). The former was also epigenome-wide significant for an association with percentage change in symptom severity, unsurprisingly, as was cg10015974. Only one DMP was epigenome-wide significant for an association with percentage change in psychosocial functioning - cg05195750 in ELP4 (a histone acetyltransferase complex)– and one DMP was significantly associated with percentage change in quality of life –cg14150252 which is located within UGP2 (involved in carbohydrate conversion). None of the candidate genes for response were found to have significant DMPs in the categorical response model; however, the top regression coefficients associated with response status were significantly associated with previous cross-sectional data on clozapine exposure, providing a form of replication and some supporting evidence that the associations found are relevant to the therapeutic efficacy of clozapine and not spurious findings. Also, as might be expected some of the models produced the same DMPs. The categorical and continuous version of percentage change in PANSS both identified the same CpG sites in DPP4 and PCDP1, unsurprisingly as these were looking at the same variable coded differently. Similarly, but more interestingly, the models of change in quality of life and psychosocial function – overlapping and correlated constructs but based on distinct measures – both identified the same CpG sites in ELP4 and UGP2 within their top ten significant DMPs. Each of the analysis exploring symptom severity (total, positive or negative) found epigenome- wide significant DMPs. Five were associated with total symptoms - cg02890365 in TUFT1 (involved in - 220 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes neuronal differentiation), cg23158483 in GLRA1 (encoding a glycine receptor subunit), cg26129353 in ATP8B1 (involved in ion channel transport), cg17651451 in ADAMTS2 (involved in collagen biosynthesis) and cg23290344 in NEFM (involved in neuronal calibre and transport between axons and dendrites). Three were associated with positive symptoms - cg03661817 in KIFC3 (involved in bipolar mitotic spindles), cg02759966, and cg08283882 in EBF2 (a transcription factor) – and fourteen were associated with negative symptoms, making negative symptoms the strongest predictor of differential methylation of any variable studied in this thesis – the top ten were cg00293996 in CLDN14 (associated with deafness), cg21288861 in ITGB3 (involved in cell adhesion), cg02518691 in ADAMTS8 (involved in integrin binding), cg271166770 in TRIM54 (involved in signal transduction and microtubules), cg01371207 in PFDN6 (involved in tubulin folding), cg03321142 in IGSF9 (involved in cell adhesion and dendritic and synaptic development), cg08682625 in LOC727677, cg14660007 in SLAIN1, cg19060383 in NPHS1 (involved in cell adhesion and kidney filtration) and cg05797770 in EPHA2 (involved in GPCR function). The majority of DMPs were expressed in brain, and just under half have evidence for correlation between methylation in blood and brain. It is worth bearing in mind that the categorical response model and percentage change models are significantly less powered analyses because these compare between individuals, whereas the measures of symptom severity are exploring variation within individuals and therefore control for many more confounding variables related to individual differences. Again, while these models are exploring related variables, we can look at systems indicated by differing methodological approaches to find those most likely to reflect changes associated with a symptomatic response to clozapine. Both categorical and continuous measures of between-individual change in overall symptoms in response to clozapine indicate differential methylation in genes associated with receptor signalling and function, ion channels (in particular calcium channels), neuronal precursors, immune function, glucose homeostasis, and regulation of gene expression; within-individual variation of symptoms at any time point also indicate differential methylation genes associated with all of these except glucose homeostasis. Similarly, between-participant variation in percentage change in overall symptoms, psychosocial functioning and quality of life all indicated differential methylation in genes associated with receptor signalling function, calcium channels, immune function, glucose homeostasis and regulation of gene expression. Positive symptoms appear to be associated with differential methylation of genes indicated in signal transduction, ion channels, immune function and transcription regulation, while negative symptoms appear to be associated with signal transduction, receptor function, neuronal development and transcription regulation. Positive and - 221 - Therapeutic Response to Clozapine and Associated DNA Methylation Changes negative symptoms are quite different phenomenon, so it is to be expected that there would be deviation between the DMPs and systems implicated. While this analysis allows us to identify differential methylation associated with therapeutic response and therefore likely to indicate regions associated with mechanisms of clinical interest, the causal relationship is still uncertain; these changes in DNA methylation may be the cause of symptomatic change, or may simply reflect the processes of symptomatic change. Looking back at the previous chapter, models of exposure to clozapine and models of therapeutic response to clozapine both overlap in indicating differential methylation at loci involved in signal transduction, ion channels (especially calcium), GPCRs, glutamatergic function, microtubules or beta tubulin, and regulation of gene expression, as well as immune function and glucose homeostasis; as this analysis associates the latter two with response to clozapine, this suggest the effect on these systems may be critical to the therapeutic efficacy of clozapine and not just associated with adverse reactions. The next chapter will present my analysis on adverse reactions during clozapine treatment. - 222 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes 5 Adverse Reactions to Clozapine and Associated DNA Methylation Changes 5.1 Background: Adverse reactions to clozapine In this section I will first outline research on the prevalence of adverse reactions and implications for patients, and then review existing research on potential mechanisms underlying adverse reactions, which is predominantly genetic. Aside from the well-known but relatively rare cases of agranulocytosis, there are many more common adverse reactions of clozapine which also put patients at high risk. Patients with severe mental illness, including schizophrenia, have increased mortality rate and reduced life expectancy, with cardiovascular disease one of the leading causes of death, partly driven by increased rates of metabolic syndrome and associated obesity, glucose intolerance and hypertension. Weight gain and metabolic syndrome hold the potential for developing diabetes or cardiac problems, both of which increase the risk of a range of negative outcomes. In 2004, the American Diabetes Association highlighted the clinical significance of long-term antipsychotic use increasing risk for obesity related problems, which seems to be a particularly high risk for African-American patients (Blin & Micallef, 2001). Various reviews (Nasrallah & Newcomer, 2004) highlight the increased risk posed by clozapine. Along with olanzapine, clozapine has been reported as the second-generation-antipsychotic known to be most highly associated with weight gain (Pramyothin & Khaodhiar, 2010). Lamberti, Bellnier and Schwarzkopf (1992) found that half of patients on clozapine have metabolic adverse reactions, a significantly higher proportion than those in a matched comparison group within the healthy population. One naturalistic study of 307 patients found no increased risk for clozapine patients of any metabolic measures (body mass index, type 2 diabetes, hypertension, dyslipidaemia, or obesity) (Kelly et al., 2014). However, Allison et al., (1999) did a comprehensive review and meta-analysis of 81 studies investigating antipsychotic associated weight gain, and while most were associated with weight gain, clozapine was associated with the greatest weigh gain over a ten week period of treatment, with an average gain of 4.45 kg which is a clinically meaningful gain with associated risks. There are also adverse reactions such as constipation which appear mundane but can have severe consequences, including death, and are not systematically monitored despite their prevalence. There are numerous adverse reactions with less risk to life but still significant impact on quality of life. Many of these are at best inconvenient but at worst contribute to humiliation and stigma (Hodge & Jespersen, 2008) and impact on self-esteem. For example, hyper-salivation is one of the most common adverse reactions and this is very visible and easily recognisable as an adverse - 223 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes reaction of antipsychotics; this inhibits privacy and is generally disruptive to confidence in social situations. Adverse reactions such as these can cause a lack of medication adherence, potentially leading to relapse; approximately 20% of patients prescribed clozapine discontinue due to adverse reactions, with a larger group of patients showing covert partial adherence (Young, Bowers, & Mazure, 1998). Some of the most commonly reported adverse reactions, both by manufacturers and patients, are daytime drowsiness or sedation, dizziness or postural hypotension, tachycardia, myoclonus, hyper-salivation, anticholinergic effects including dry mouth and blurred vision, nausea, vomiting, gastric reflux or heartburn, constipation, nocturnal enuresis, polyuria, polydipsia, weight gain and sexual dysfunction. These have been reported to remain throughout the duration of treatment, with Yusufi et al., (2007) finding that 77% of patients on chronic treatment (at least three months but median of 30 months) had at least one moderate or severe adverse reaction, 19% reporting one or more severe adverse reactions, and every patient reporting at least one adverse reaction. The majority of adverse reactions (649/766 reported) were not reported when patients were asked an open question, only in response to systematic assessment, creating concern about untreated adverse reactions; this was especially true for adverse reactions which patients might feel embarrassed to report such as sexual dysfunction. However, there is a problem of under-prescription; Meltzer (2012) highlights that fewer than 5% of American patients with schizophrenia are treated with clozapine, which is less than a fifth of those who meet the criteria indicated for prescription. This may be driven by concerns about adverse reactions, but two large epidemiological studies have found clozapine to have the lowest mortality rate of any antipsychotic drug, a finding which is driven largely due to its significant reduction in suicide risk. In a study across all patients of Finland admitted to psychiatric care, found clozapine was associated with the lowest mortality risk, and this risk was significantly lower than for all other antipsychotics (Tiihonen et al., 2009)– potential explanations given were more intensive monitoring, increased effectiveness, increased treatment compliance and subsequent improved lifestyle. There were no differences in terms of death due to ischemic heart disease, despite known increases in cardiovascular adverse reactions. These results are surprising considering that there is a selection bias working against this finding – clozapine is reserved for the most unwell, most suicidal, most risky patients. Also, in an Australian observational cohort study of 355 patients, older age of commencement on clozapine was associated with increased risk of adverse reactions suggesting we should not allow concern about adverse reactions to delay prescription of clozapine as this only heightens the risk (Hyde et al., 2015). However, the Tiihonen study has been criticised (De Hert, - 224 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes Correll, & Cohen, 2010) for incomplete reporting and various analytical choices which could explain these findings, and cautions against dismissing the risks associated with clozapine. Despite the disagreement in the literature about the level of risk relative to other antipsychotics and relative to overall benefit, the prevalence of a range of distressing adverse reactions cannot be denied and patient concerns cannot be dismissed. A meta-analysis (de Silva et al., 2016) found that metformin was most effective in preventing antipsychotic induced weight gain in first episode patients, compared to reducing weight in chronic patients who have already gained significant amount of weight and likely developed insulin resistance, highlighting the importance of early intervention and identification of patients at risk. Understanding of these adverse reactions, their mechanisms, who is most at risk and how we may be able to prevent or alleviate them, is therefore necessary. 5.1.1 Mechanisms of adverse reactions to clozapine Most literature regarding predictors of antipsychotic-induced adverse reactions has focused on antipsychotic-induced weight gain and genetic variants. There has been a particular focus on studying the association between the 5-HTR2C receptor and antipsychotic-associated weight gain. Rodent studies highlight its importance in the regulation of food intake in rodents (mice with the HTR2C gene knocked out show chronic hyperphagia and develop obesity and hyperinsulinemia) (Butler & Cone, 2001). Polymorphisms such as C-759T (rs3813929), in the promoter region, have been associated with late-onset diabetes and obesity in non-psychiatric populations, as well as consistently associated with antipsychotic induced weight gain (Zhang & Malhotra, 2013); ten out of 17 studies reporting on the association between this SNP and antipsychotic induced weight gain found the C allele was associated with more weight gain, especially clozapine and olanzapine, and this is confirmed by a meta-analysis reporting that the C allele was associated with over double risk of gaining at least 7% of baseline body weight (clinically significant weight gain) (De Luca, Mueller, de Bartolomeis, & Kennedy, 2007). However, this cannot be the full explanation as a range of antipsychotics with similar moderate 5-HTR2C binding affinity are associated with variable risk of metabolic adverse reactions, and ziprasidone has very minimal metabolic adverse reactions despite having the most potent binding. Unsurprisingly, leptin genes have also been implicated, as leptin is a hormone - released by subcutaneous adipocytes and transported to the hypothalamus - which plays a critical role in regulating food intake and homeostasis. Genetic variants have been associated with obesity elsewhere and studies have associated the functional leptin promoter 2458G/A with antipsychotic induced weight gain, with one finding the polymorphism account for 9% of variation in weight gain - 225 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes (Zhang et al., 2007). Studies find increased levels of leptin in the blood of patients on antipsychotics but observed changes in glucose and lipid metabolism induced by antipsychotics are believed to occur independently of the weight gain (Panariello, Polsinelli, Borlido, Monda, & De Luca, 2012): they occur before the onset of weight gain instead of a consequence afterwards, and the effect sizes between the two patterns of disturbance do not always correlate. As previously discussed, candidate-gene studies are severely limited. Brandl et al., (2015) performed a GWAS for antipsychotic induced weight gain in a subsample of participants from the CATIE study, but did not find any SNPs which reached genome-wide significance. One of the variants with suggestive significance, rs9346455 – which is located 6.6 kb upstream of the opioid growth factor receptor-like 1 gene (OGFRL1) on chromosome 6 -, was replicated in a separate sample for association with antipsychotic induced weight gain though, and in both the minor G-allele was associated with higher weight gain. However, the functional relevance of this polymorphism is currently unclear. Other investigations specifically for clozapine have focused on genetic polymorphisms of CYP1A2, the cytochrome P450 enzyme which is the main metaboliser of clozapine and may affect clozapine clearance and plasma levels; allelic variation has been associated with tardive dyskinesia, and mRNA levels have been associated with adverse reactions in general. In one of the larger studies of 237 long term clozapine users (Viikki et al., 2014), the TT genotype of polymorphisms -1545 (rs2470890) was associated with significantly more severe adverse reactions during clozapine treatment. The sub-analysis did not find significant effects for individual adverse reaction types, but the pattern was similar across all seven types of adverse reactions studies. A review (Shams & Müller, 2014) concludes that the genetic variants with the most consistently replicated findings are in the melanocortin 4 receptor (MC4R), the serotonin 2C receptor (5HT2C), leptin, neuropeptide Y (NPY) and cannabinoid receptor 1 (CNR1) genes. Studies have implicated methylation in metabolic syndrome experienced by patients treated with antipsychotics; for example, Lott et al., (2013) found COMT promoter region methylation, physical activity and metabolic syndrome were significantly associated in Val158Met patients, with genotype and physical activity both playing a role in epigenetic modulation of COMT with apparent implications for metabolic syndrome. One suggestion (Scigliano & Ronchetti, 2013) for the mechanism of antipsychotic induced metabolic effects is that blocking peripheral dopamine receptors increases sympathetic activity, leading to disturbances in autonomic function. Concurrently, the blocking of peripheral muscarinic receptors reduces vagal parasympathetic activity. High levels of adrenaline and noradrenaline - 226 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes oppose the anabolic effects of insulin, promoting glycogen transformation to glucose, inhibiting insulin release and inducing hyperglycaemia, as well as reducing glucose uptake by muscle and adipose tissue promoting insulin resistance, and stimulating adipose lipolysis promoting release of glycerol, free fatty acids, leptin, adiponectin and pro-inflammatory cytokines. Similarly, cortisol increases insulin resistance, promotes abdominal fat accumulation and stimulates intake of carbohydrates and fat. There are widespread peripheral dopamine receptors, present in heart, kidneys, live, pancreas etc. When activated, D2 like receptors inhibit noradrenaline release and adrenaline secretion, indirectly inhibiting sympathetic tone; and when there is intense sympathetic stimulation, dopamine is released alongside noradrenaline and adrenaline, modulating sympathetic discharge. Antipsychotic blockade of D2 receptors therefore is likely to remove inhibition of sympathetic tone and prevent the modulation of sympathetic activity. Support for the hypothesis that hyperglycaemia is caused by increased sympathetic nervous system activity comes from animal research demonstrating that hyperglycaemia following acute antipsychotic administration is blocked by an adrenoceptor antagonist. If it was simply D2 receptor blockade causing metabolic adverse reactions, the most potent D2 blockers would be most strongly associated with these adverse reactions. An explanation for why clozapine and olanzapine are most strongly associated, despite less potent antagonism, is that a) they may already be blocking adrenergic receptors, preventing the effect of increased availability of adrenergic neurotransmitters, and b) they bind at cholinergic M3 receptors where parasympathetic activity would act to counterbalance the increased sympathetic activity. Since the early days of chlorpromazine there has also been a suggestion that the patients who benefit most from treatment are also those who show the most weight gain and metabolic adverse reactions, and while this remains an area of interest this has yet to be unequivocally proven (Sharma, Rao, & Venkatasubramanian, 2014). To further our understanding into these unwanted consequences of clozapine use, I will now present my analysis of DNA methylation trajectories associated with having or not having clozapine- induced increased appetite/weight, drowsiness and cardiovascular symptoms; the latter two have been minimally researched across all research fields, with no existing epigenetic research in humans to the best of my knowledge. 5.2 Statistical method: Adverse reactions to clozapine To explore associations with adverse reactions to clozapine, I compared whether experiencing an adverse reaction or not after twelve weeks of clozapine use was associated with different - 227 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes trajectories of DNA methylation. I analyse the following three adverse reactions – with symptom clusters of the most frequently reported adverse reactions - as predictors: 1) Between-individual classification of experiencing increased appetite/weight (categorical) as a predictor of DNA methylation trajectories over time Increased appetite/weight: experiencing increased appetite/weight was defined as self-report of increased appetite or weight gain more than once in the last week, when assessed at twelve week follow up. 2) Between-individual classification of experiencing drowsiness (categorical) as a predictor of DNA methylation trajectories over time Drowsiness: experiencing drowsiness was defined as self-report of feeling sleepy during the day time or “drugged like a zombie” more than once in the last week, when assessed at twelve week follow up. 3) Between-individual classification of experiencing cardiovascular symptoms (categorical) as a predictor of DNA methylation trajectories over time Cardiovascular symptoms: experiencing cardiovascular symptoms was defined as self-report of feeling chest pain or irregular/fast heart beating more than once in the last week, when assessed at twelve week follow up. To identify DNA methylation changes associated with adverse reactions to clozapine, I again conducted mixed model regression analyses using lmer. For the prediction of different methylation trajectories between individuals with or without each adverse reaction, the interaction between adverse reaction group (experience vs do not experience) and weeks on clozapine was used a predictor of methylation at each CpG sites. These models controlled for baseline age, sex, predicted cell counts (CD8T, CD4T, NK, B, granulocytes and monocytes), smoking at time of sample collection (yes/no), chip ID, previous medications (number of prior atypical antipsychotics, number of prior typical antipsychotics, number of prior benzodiazepines or other mood stabilisers) and current medications (number of current antipsychotics, number of current benzodiazepines or other mood stabilisers, number of current antidepressants), as well as controlling for the repeated measures from individual participants as a random effect. These analyses all produced a model with regression coefficients for each CpG site - representing the percentage change in methylation at the CpG site associated with an additional week on - 228 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes clozapine for those not experiencing the adverse reaction – as well as a p value for the significance of this association. For each model, I report global DNA methylation, differentially methylated positions (DMPs) and pathway analysis results. All of this is conducted as described in previous chapters (Statistical method: Exposure to clozapine). Additionally, the regression coefficients of each model are used for candidate site testing. Candidate CpG site methylation: Based on previous literature on the genetics of clozapine adverse reactions (Shams & Müller, 2014; Viikki et al., 2014), I identified five genes which have been associated with adverse reactions: MC4R, 5HT2C, NPY, and CNR1 and CYP1A2. In total, there were 87 CpG sites for these genes included in the 450k assay which passed QC. The first four genes have exclusively been identified in relationship to weight gain or other metabolic symptoms so were only tested in the increased appetite/weight model. The last gene, CYP1A2, has been associated with numerous adverse reactions so was tested in all models. There were seven CpG sites included in the 450k assay which passed QC. To account for multiple-comparisons in the increased appetite/weight model, the adjusted p value is 0.05/87 = 0.000575. To account for multiple-comparisons in the drowsiness and cardiovascular symptoms models, the adjusted p value is 0.05/7 = 0.00714. The results for each analysis model are reported separately. First I will report results for increased appetite/weight, then drowsiness, then cardiovascular symptoms. - 229 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes 5.3 Results: Increased appetite/weight 5.3.1 Participant characteristics (increased appetite/weight) Twelve weeks after clozapine initiation, thirteen participants were experiencing increased appetite/weight as an adverse reaction to clozapine and nine were not. The characteristics of each group are compared in Table 16, with key details represented in subsequent pie-charts ( ): compared to those who did not experience increased appetite/weight, those who did experience increased appetite/weight were more likely to be black, have comorbidities (especially substance abuse), have not been previously exposed to clozapine, be a responder to clozapine, and experience drowsiness. At twelve weeks, those who experienced increased appetite/weight had greater reductions in total symptom severity and also had lower clozapine levels. However, none of these differences were statistically significant. - 230 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes Table 16: Characteristics of participants with and without increased appetite/weight at 12 weeks Increased No increased appetite/weight appetite/weight Age (years) Mean 36.42 39.78 p=0.55 Minimum 19.94 21.08 Maximum 50.99 61.38 Gender Male 10 7 p=1 Female 3 2 Ethnicity White 5 5 p=0.637 Black 7 3 Asian 1 1 Diagnosis Schizophrenia 11 8 p=1 Schizoaffective Disorder 2 1 Months between Mean 27.90 26.11 p=0.92 diagnosis of Minimum 0 1 treatment- resistance and clozapine start Maximum 111 100 Inpatient Psychiatric Mean 4.08 2.67 p=0.27 Admissions Minimum 0 0 Maximum 10 7 Psychiatric Affective 2 0 p=0.21 Comorbidities Anxiety 1 1 Substance Abuse 3 0 None 7 8 Prior Clozapine Yes 4 1 p=0.57 Exposure No 9 8 Number of prior Mean 0.85 0.67 p=0.59 typical Minimum 0 0 antipsychotics Maximum 2 2 Number of prior Mean 3.15 2.56 p=0.32 atypical Minimum 1 1 antipsychotics Maximum 7 4 Number of prior Mean 0.23 0.22 p=0.97 mood stabilisers Minimum 0 0 Maximum 1 1 Smoker at baseline Yes 2 1 p=1 No 11 8 Yes 4 3 p=1 - 231 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes Smoker at any time- point during study No 9 6 Baseline PANSS Mean 76.46 81 p=0.58 Total Minimum 55 54 Maximum 105 119 Baseline PANSS Mean 19.88 19.11 p=0.80 Positive Minimum 12 9 Maximum 26 36 Baseline PANSS Mean 21.15 22.89 p=0.61 Negative Minimum 11 9 Maximum 30 39 Predominant Positive 4 3 p=1 Baseline Symptoms Negative 9 6 12w PANSS Total Mean 56.82 71.56 p=0.07 Minimum 35 38 Maximum 76 98 6m PANSS Total Mean 58.92 64.63 p=0.53 Minimum 43 37 Mean 77 97 PANSS % Change at Mean -22.34 -10.76 p=0.18 12w Minimum -53.57 -31.03 Maximum 21.82 26.47 PSP % Change at Mean 12.27 57.43 p=0.13 12w Minimum -50 -14.29 Maximum 57.14 162.5 GAF % Change at Mean 39.28 40.60 p=0.96 12w Minimum -10 -25 Maximum 102.86 185.71 Responder at 12w Yes 8 2 p=0.17 No 5 7 Cardiovascular Yes 2 2 p=1 symptoms at 12w No 11 7 Drowsiness at 12w Yes 11 4 p=0.13 No 2 5 Clozapine Levels at Mean 0.36 0.46 p=0.17 12w Minimum 0.21 0.28 Maximum 0.59 0.75 - 232 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes Figure 8: Pie charts comparing key demographics for those with and without increased appetite/weight - 233 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 234 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 235 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 236 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 237 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 238 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 239 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 240 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 241 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 242 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes increased appetite/weight ) mg/l ( Increased appetite/weight No increased appetite/weight ht - 243 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes increased appetite/weight Increased appetite/weight No increased appetite/weight 5.3.2 Global DNA methylation (increased appetite/weight) Not experiencing increased appetite/weight was associated with global demethylation over time. For those without increased appetite/weight, an additional week on clozapine was associated with an average additional decrease in global DNA methylation of 0.00114%. 50.49% of all CpG sites showed an additional increase in methylation associated with time exposed to clozapine if not experiencing increased appetite/weight and 49.51% of all CpG sites showed a decrease. Moreover, when exploring individual time points, split by adverse reaction: at baseline, mean global DNA methylation was 48.316% for those with increased appetite/weight and 48.240% for those without; at six weeks, 48.258% for those with and 48.250% for those without; at twelve weeks, 48.334% for those with s and 48.233% for those without; and at six months, 48.328% for those with and 48.242% for those without. Those who experience increased appetite/weight show higher starting level of global DNA methylation which decreases, increases and then decreases but remains higher than baseline; while those who do not experience these adverse reactions start with - 244 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes lower levels which increase, decrease and increase and end similar to baseline and lower than those who do experience the adverse reactions. increased appetite/weight increased No increased appetite/weight appetite/weight 5.3.3 Differentially methylated positions (increased appetite/weight) Three DMPs reached the threshold for epigenome wide significance (p<1*10-7) and two hundred and fifty one DMPs were trend-level significant (p<5*10-5). Clozapine-associated increased appetite/weight - 245 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes increased appetite/weight - 246 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes Table 17: Top ten DMPs associated with increased appetite/weight (* indicates epigenome wide significance) CpG site Beta P value Gene Location MAPINFO cg10966876 0.195x10-2 6.03E-09* TWIST2 TSS1500 239755461 ch.6.163729680F 0.74 x10-3 4.81E-08* 163809690 0.180 x10- cg21097199 2 5.99E-08* POLQ; 1stExon;5'UTR 121264819 0.268 x10- cg08314457 2 1.51E-07 77507534 cg07957953 0.85 x10-3 1.78E-07 GGT7 TSS1500 33460865 cg20296524 -0.56 x10-3 6.07E-07 TARBP1 1stExon 234614614 -0.182 cg25151834 x10-2 7.31E-07 UNKL Body 1449528 -0.284 cg19155552 x10-2 7.49E-07 TMEM116 3'UTR 112369383 -0.735 cg21167587 x10-2 7.51E-07 RNF38 5'UTR;Body 36397390 0.245 x10- cg09241703 2 7.56E-07 FOXN3 Body - 247 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes The most statistically significant DMP associated with not experiencing increased appetite/weight was cg10966876. For those who did not experience increased appetite/weight, each additional week exposed to clozapine was associated with an additional increase in methylation at this site of 0.195% compared to those who did (p=0.6x10-8). For those who experienced increased appetite/weight, uncorrected means show an average baseline methylation of 10.303%, 9.869% at six weeks, 9.275% at twelve weeks, and 9.548% at six months. For those who did not experience increased appetite/weight, uncorrected means show an average baseline methylation of 9.182%, 10.365% at six weeks, 10.216 % at twelve weeks, and 12.227 % at six months. Those who experience increased appetite/weight show a decrease over time with an increase towards the end of the study period but remain below baseline, while those who do not start at a lower level at baseline but show an increase, slight decrease and then large increase ending at levels higher than both groups at any other point. No increased appetite/weight Increased appetite/weight cg10966876 is located in the TSS1500 promoter region of TWIST2, which encodes twist family BHLH transcription factor 2. This gene is located on chromosome 2 and expressed across numerous - 248 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes tissues with overexpression of mRNA in cervix and secreting glandular tissue and overexpression of protein in serum. Twist family BHLH transcription factor 2 is involved in transcription regulation, repression of pro-inflammatory cytokine expression, glycogen storage and energy metabolism. The second most statistically significant DMP associated with not experiencing increased appetite/weight was located at ch.6.163729680F. For those who did not experience increased appetite/weight, each additional week exposed to clozapine was associated with an additional increase in methylation at this site of 0.0743% compared to those who did (p=0.5x10-7). For those who experienced increased appetite/weight, uncorrected means show an average baseline methylation of 4.525%, 4.304% at six weeks, 4.456 % at twelve weeks, and 4.486% at six months. For those who did not experience increased appetite/weight, uncorrected means show an average baseline methylation of 3.975%, 4.462% at six weeks, 4.712% at twelve weeks, and 5.515 % at six months. Those who experience increased appetite/weight show an initial decrease then increase but remain lower than baseline, while those who do not start at a lower baseline level and increase over time. No increased appetite/weight Increased appetite/weight - 249 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes This CpG site is located on chromosome 6 but is otherwise uncharacterised. The third most statistically significant DMP associated with not experiencing increased appetite/weight was cg21097199. For those who did not experience increased appetite/weight, each additional week exposed to clozapine was associated with an additional increase in methylation at this site of 0.1801% compared to those who did (p=0.6x10-7). For those who experienced increased appetite/weight, uncorrected means show an average baseline methylation of 9.564%, 10.229% at six weeks, 8.689% at twelve weeks, and 10.351% at six months. For those who did not experience increased appetite/weight, uncorrected means show an average baseline methylation of 9.990%, 10.575% at six weeks, 11.867% at twelve weeks, and 13.178% at six months. Those who experience increased appetite/weight show an initial increase, then decrease, then increase and end higher than baseline, while those who do not show increases at each time point. No increased appetite/weight Increased appetite/weight cg21097199 is located in the 1st exon of POLQ, which encodes DNA polymerase theta – this protein repairs double-strand breaks in DNA. This gene is located on chromosome 3 and expressed - 250 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes across numerous tissues, with overexpression of mRNA in oesophagus and testis and overexpression of protein in fetal testis and brain. 5.3.4 Pathway analysis (increased appetite/weight) Pathway analysis indicated that the pathways in Table 18 were the top thirty pathways with an over-representation of DMPs associated with increased appetite/weight. - 251 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes Table 18: Top thirty over-represented pathways - increased appetite/weight Category Name p value over- Number of differentially Number in Ontology representation methylated category GO:0043555 regulation of translation in response to stress 0.00102 5 13 BP GO:0032386 regulation of intracellular transport 0.00141 40 356 BP GO:0033157 regulation of intracellular protein transport 0.00151 25 195 BP GO:0042992 negative regulation of transcription factor import into nucleus 0.00275 8 36 BP GO:0046823 negative regulation of nucleocytoplasmic transport 0.00275 11 60 BP GO:0042308 negative regulation of protein import into nucleus 0.00313 10 51 BP GO:0046822 regulation of nucleocytoplasmic transport 0.00344 22 168 BP GO:0051224 negative regulation of protein transport 0.00375 16 109 BP GO:0033572 transferrin transport 0.00417 7 33 BP GO:0042306 regulation of protein import into nucleus 0.00483 18 133 BP GO:0019722 calcium-mediated signalling 0.00554 15 98 BP GO:0035107 appendage morphogenesis 0.00557 24 139 BP GO:0035108 limb morphogenesis 0.00557 24 139 BP GO:0015682 ferric iron transport 0.00578 7 35 BP GO:0072512 trivalent inorganic cation transport 0.00578 7 35 BP GO:0007004 telomere maintenance via telomerase 0.00583 5 16 BP GO:0048713 regulation of oligodendrocyte differentiation 0.00586 7 25 BP GO:0034504 protein localization to nucleus 0.00685 23 180 BP GO:0043588 skin development 0.00752 36 324 BP GO:0050431 transforming growth factor beta binding 0.00771 5 16 MF GO:0048706 embryonic skeletal system development 0.00773 21 123 BP GO:0090317 negative regulation of intracellular protein transport 0.00791 11 71 BP - 252 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes Category Name p value over- Number of differentially Number in Ontology representation methylated category GO:0048483 autonomic nervous system development 0.00823 10 42 BP GO:1900180 regulation of protein localization to nucleus 0.00837 19 146 BP GO:0021604 cranial nerve structural organization 0.00878 4 10 BP GO:0006730 one-carbon metabolic process 0.00883 6 30 BP GO:0007338 single fertilization 0.00916 12 102 BP GO:0048736 appendage development 0.00938 25 157 BP GO:0060173 limb development 0.00938 25 157 BP GO:0051223 regulation of protein transport 0.00954 34 338 BP - 253 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes 5.3.5 Candidate CpG site methylation (increased appetite/weight) Of six CpG sites located within MC4R with methylation data available, none had a significant association with increased appetite/weight before or after correction for multiple testing. Across all MC4R CpG sites, there was a mean additional increase in methylation of 0.0123% associated with each additional week on clozapine for those who experience increased appetite/weight. Of twenty five CpG sites located within 5HT2C with methylation data available, four had a significant association with increased appetite/weight before but not after correction for multiple testing – cg25885716 showed an additional decrease in methylation (-0.137%, p=0.028), cg05100479 showed an additional increase in methylation (0.0516%, p=0.031), cg18153989 showed a decrease in methylation (-0.211%, p=0.039), and cg10997274 showed a decrease in methylation (-0.097%, p=0.015). Across all 5HT2C CpG sites, there was a mean additional decrease in methylation of 0.0251% associated with each additional week on clozapine for those who experience increased appetite/weight. Of eighteen sites located within NPY with methylation data available, one had a significant association with increased appetite/weight after correction for multiple testing (p<0.575x10-3) – cg11653966 showed an additional increase in methylation for those who did not experience increased appetite/weight (0.1802%, p=0.0002). Six others had a significant association with increased appetite/weight before but not after correction for multiple testing (0.000575 - 254 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes No increased appetite/weight Increased appetite/weight Of twenty six sites located within CNR1 with methylation data available, two had a significant association with increased appetite/weight before but not after correction for multiple testing (0.575x10-3 Of seven sites located within CYP1A2 with methylation data available, none had a significant association with increased appetite/weight before or after correction for multiple testing. Across all CYP1A2 CpG sites, there was a mean additional increase in methylation of 0.00047%. - 255 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes 5.4 Results: Drowsiness 5.4.1 Participant characteristics (drowsiness) Twelve weeks after clozapine initiation, fifteen participants were experiencing drowsiness as an adverse reaction to clozapine and seven were not. The characteristics of each group are compared in Table 19, with key details represented in subsequent pie-charts (Error! Reference source not found.): compared to those who did not experience drowsiness, those who did experience drowsiness were more likely to be male, black, have no comorbidities, have previous exposure to clozapine, have predominantly negative symptoms at baseline, be a responder to clozapine, and experience increased appetite/weight. At twelve weeks, those who experienced drowsiness had lower clozapine levels. However, none of these differences were statistically significant except gender (X=4.348, p=0.037), a variable which was controlled for in analysis. - 256 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes Table 19: Characteristics of participants with and without drowsiness at 12 weeks Drowsiness No drowsiness Age (years) Mean 36.87 39.80 p=0.58 Minimum 19.94 21.08 Maximum 61.38 52.52 Gender Male 14 3 p=0.04* Female 1 4 Ethnicity White 6 4 p=0.533 Black 8 2 Asian 1 1 Diagnosis Paranoid p=1 Schizophrenia 13 6 Schizoaffective 2 1 Months between Mean 29.75 22.43 p=0.69 diagnosis of Minimum 1 0 treatment- resistance and clozapine start Maximum 111 100 Inpatient Psychiatric Mean 3.80 2.86 p=0.40 Admissions Minimum 0 1 Maximum 10 5 Psychiatric Affective 1 1 p=0.85 Comorbidities Anxiety 1 1 Substance Abuse 2 1 None 11 4 Prior Clozapine Yes 4 1 p=0.92 Exposure No 11 6 Number of prior Mean 0.87 0.57 p=0.33 typical Minimum 0 0 antipsychotics Maximum 2 1 Number of prior Mean 2.87 3 p=0.81 atypical Minimum 1 2 antipsychotics Maximum 7 4 Number of prior Mean 0.2 0.29 p=0.7 mood stabilisers Minimum 0 0 Maximum 1 1 Smoker at baseline Yes 3 0 p=0.54 No 12 7 Smoker at any time- Yes 5 2 p=1 point during study No 10 5 - 257 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes Baseline PANSS Mean 77.73 79.57 p=0.85 Total Minimum 55 54 Maximum 105 119 Baseline PANSS Mean 18.77 21.29 p=0.43 Positive Minimum 9 15 Maximum 26 36 Baseline PANSS Mean 22.73 20 p=0.41 Negative Minimum 11 9 Maximum 39 30 Predominant Positive 4 3 p=0.79 Baseline Symptoms Negative 11 4 12w PANSS Total Mean 61.31 67.43 p=0.54 Minimum 35 38 Maximum 79 98 6m PANSS Total Mean 58.93 66.50 p=0.49 Minimum 38 37 Maximum 80 97 PANSS % Change at Mean -18.61 -15.44 p=0.74 12w Minimum -53.57 -39.39 Maximum 21.82 26.47 PSP % Change at Mean 26.77 38.64 p=0.71 12w Minimum -50 -14.29 Maximum 160 162.5 GAF % Change at Mean 46.40 26.73 p=0.35 12w Minimum -10 -25 Maximum 185.71 95.12 Responder at 12w Yes 8 2 p=0.53 No 7 5 Cardiovascular Yes 3 1 p=1 symptoms at 12w No 12 6 Increased Yes 11 2 p=0.13 appetite/weight at 12w No 4 5 Clozapine Levels at Mean 0.37 0.51 p=0.16 12w Minimum 0.21 0.25 Maximum 0.59 0.75 - 258 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes Figure 9: Pie-charts of characteristics in those with and without drowsiness - 259 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 260 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 261 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 262 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 263 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 264 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 265 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 266 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 267 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 268 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes ) mg/l ( - 269 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes 5.4.2 Global DNA methylation (drowsiness) Not experiencing drowsiness was associated with global demethylation. For those without drowsiness, an additional week on clozapine was associated with an average additional decrease in global DNA methylation of 0.00075%. But 51.01% of all CpG sites showed an additional increase in methylation associated with time exposed to clozapine if not experiencing drowsiness and 48.99% of all CpG sites showed a decrease. However, when exploring individual time points, split by response: at baseline, mean global DNA methylation was 48.285% for those with drowsiness and 48.285% for those without; at six weeks, 48.240% for those with and 48.284% for those without; at twelve weeks 48.292% for those with and 48.282% for those without; and at six months 48.295% for those with and 48.291% for those without. Both groups show the same starting level of global DNA methylation and end at a slightly higher level, but those who experience drowsiness show an initial decrease and then an increase beyond those who do not. - 270 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes 5.4.3 Differentially methylated positions (drowsiness) Two DMPs reached the threshold for epigenome wide significance (p<1*10-7) and one hundred and eight nine DMPs were trend-level significant (p<5*10-5). - 271 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 272 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes Table 20: Top ten DMPs associated with drowsiness (* indicates epigenome wide significance) CpG site Beta P value Gene Location MAPINFO 0.161x10- cg12155676 2 3.67E-08* 11970561 cg21399492 0.27 x10-2 9.65E-08* 48055701 -0.305 cg06228491 x10-2 4.17E-07 1578330 cg10864791 0.9 x10-3 4.38E-07 EFHA1 TSS200 22178445 -0.21 x10- cg01406247 2 4.95E-07 34611274 0.189 cg21097199 x10-2 5.67E-07 POLQ 1stExon;5'UTR 121264819 -0.151 cg27299846 x10-2 5.84E-07 MAST3 Body 18235489 0.178 cg02348449 x10-2 9.83E-07 ZSCAN18 TSS1500 58630429 -0.212 cg16402835 x10-2 9.98E-07 PTPN14 3'UTR 214531223 0.227 cg23606925 x10-2 1.04E-06 102471597 - 273 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes The most statistically significant DMP associated with not experiencing drowsiness was cg12155676. For those who did not experience drowsiness, each additional week exposed to clozapine was associated with an additional increase in methylation at this site of 0.161% compared to those who did (p=0.4x10-7). For those who experienced drowsiness, uncorrected means show an average baseline methylation of 9.853%, 9.34% at six weeks, 8.878% at twelve weeks, and 9.194% at six months. For those who did not experience drowsiness, uncorrected means show an average baseline methylation of 9.124%, 8.963% at six weeks, 9.854% at twelve weeks, and 11.296% at six months. Those who experienced drowsiness showed decreases then an increase but remained lower than baseline, whereas those who did not started at a lower baseline level and showed a decrease but then increase and ended higher than baseline. cg12155676 is located on chromosome 2 outside of a characterised gene. The second most statistically significant DMP associated with not experiencing drowsiness was cg21399492. For those who did not experience drowsiness, each additional week exposed to - 274 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes clozapine was associated with an additional increase in methylation at this site of 0.27% compared to those who did (p=0.97x10-7). For those who experienced drowsiness, uncorrected means show an average baseline methylation of 75.871%, 76.545% at six weeks, 76.133% at twelve weeks, and 77.182% at six months. For those who did not experience drowsiness, uncorrected means show an average baseline methylation of 72.780%, 73.063% at six weeks, 76.063% at twelve weeks, and 77.968% at six months. Those who experienced drowsiness showed an increase, decrease and then increase ending higher than baseline, whereas those who did started at a lower baseline level but increase at each time point ending higher than any time point for either group. cg21399492 is located on chromosome 1 outside of any characterised gene. . - 275 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes 5.4.4 Pathway analysis (drowsiness) Pathway analysis indicated that the pathways in Table 21 were the top thirty pathways with an over-representation of DMPs associated with drowsiness. - 276 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes Table 21: Top thirty over-represented pathways - drowsiness Category Name p value over- Number of differentially Number in Ontology representation methylated category GO:00+C3+A2:A2:F21 lamellipodium 0.0004 21 135 CC GO:0016887 axoneme 0.0023 9 61 CC GO:0032648 ciliary cytoplasm 0.0023 9 61 CC GO:0032608 neural tube development 0.0032 20 151 BP GO:0098552 regulation of protein export from nucleus 0.0043 5 23 BP GO:2001222 amino sugar metabolic process 0.006 6 36 BP GO:0032312 neural tube closure 0.0068 13 82 BP GO:0033158 tube closure 0.0072 13 83 BP GO:0010875 regulation of smoothened signalling pathway 0.0076 10 63 BP GO:0061630 cell projection cytoplasm 0.0082 9 73 CC GO:0007127 antioxidant activity 0.0090 8 62 MF GO:0032728 primary neural tube formation 0.0093 13 87 BP GO:0042623 protein homooligomerization 0.0112 20 234 BP GO:0006163 glucosamine-containing compound metabolic process 0.0112 4 23 BP GO:0009203 cochlea morphogenesis 0.0114 5 21 BP GO:0009207 defence response to fungus 0.0125 3 21 BP GO:0009146 N-acetylglucosamine metabolic process 0.0134 3 14 BP GO:0030660 secretion 0.0136 61 856 BP GO:0009143 secretion by cell 0.0136 55 749 BP GO:0007131 positive regulation of protein export from nucleus 0.0152 3 11 BP - 277 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes Category Name p value over- Number of differentially Number in Ontology representation methylated category GO:0035825 neuroblast proliferation 0.0159 8 51 BP GO:0006152 smoothened signalling pathway 0.0160 13 112 BP GO:0046130 phosphate ion homeostasis 0.0167 3 11 BP GO:0046128 trivalent inorganic anion homeostasis 0.0167 3 11 BP GO:0042454 filopodium membrane 0.0169 4 16 CC GO:0009150 recycling endosome membrane 0.0171 5 35 CC GO:0042278 neural tube formation 0.0174 13 99 BP GO:0032373 intraciliary transport particle B 0.0180 3 15 CC GO:0032376 beta-catenin binding 0.0180 9 58 MF GO:0009164 vesicle docking 0.0186 5 30 BP - 278 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes 5.4.5 Candidate CpG site methylation (drowsiness) Of seven sites located within CYP1A2 with methylation data available, two had a significant association with drowsiness before but not after correction for multiple testing: cg24182584 was associated with an additional decrease (-0.037%, p=0.042) and cg14503537 was associated with an increase for those who did not experience drowsiness (0.1002%, p=0.0474). Across all CYP1A2 CpG sites, there was a mean additional increase in methylation of 0.00769%. - 279 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes 5.5 Results: Cardiovascular symptoms 5.5.1 Participant characteristics (cardiovascular symptoms) Twelve weeks after clozapine initiation, four participants were experiencing cardiovascular symptoms as an adverse reaction to clozapine and seven were not. The characteristics of each group are compared in Table 22, with key details represented in subsequent pie charts (Figure 10): compared to those who did not experience cardiovascular symptoms, those who did experience cardiovascular symptoms were more likely to be male, smoke, have prior exposure to clozapine, have predominantly negative symptoms at baseline, experience drowsiness and increased appetite/weight as well, and be a non-responder to clozapine. At twelve weeks, those who experienced cardiovascular symptoms had lower levels of clozapine and had lower reductions in symptom severity (two of four had shown increases in symptoms). However, none of these differences were statistically significant. The only significant difference was that those who experienced cardiovascular symptoms had taken a great number of prior atypical antipsychotic medications (t=-2.387, p=0.03), a variable which was controlled for in analysis. - 280 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes Table 22: Characteristics of participants with and without cardiovascular symptoms at 12 weeks Cardiovascular No cardiovascular symptoms symptoms Age (years) Mean 38.06 37.74 p=0.95 Minimum 26.76 19.94 Maximum 45.30 61.38 Gender Male 4 13 p=0.59 Female 0 5 Ethnicity White 2 8 p=0.783 Black 2 8 Asian 0 2 Diagnosis Paranoid p=1 Schizophrenia 3 15 Schizoaffective 1 3 Months between Mean 58.33 21.19 p=0.36 diagnosis of Minimum 1 0 treatment- resistance and clozapine start Maximum 111 100 Inpatient Psychiatric Mean 5.75 3 p=0.22 Admissions Minimum 2 0 Maximum 10 10 Psychiatric Affective 0 2 p=0.71 Comorbidities Anxiety 0 2 Substance Abuse 1 2 None 3 12 Prior Clozapine Yes 2 3 p=0.44 Exposure No 2 15 Number of prior Mean 1 0.72 p=0.56 typical Minimum 0 0 antipsychotics Maximum 2 2 Number of prior Mean 3.75 2.72 p=0.03* atypical Minimum 3 1 antipsychotics Maximum 4 7 Number of prior Mean 0.25 0.22 p=0.92 mood stabilisers Minimum 0 0 Maximum 1 1 Smoker at baseline Yes 2 1 p=0.12 No 2 17 Yes 3 4 p=0.15 - 281 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes Smoker at any time- point during study No 1 14 Baseline PANSS Mean 74.25 79.22 p=0.62 Total Minimum 55 54 Maximum 94 119 Baseline PANSS Mean 18.25 19.86 p=0.63 Positive Minimum 12 9 Maximum 25 36 Baseline PANSS Mean 22.25 21.78 p=0.84 Negative Minimum 20 9 Maximum 26 39 Predominant Positive 0 7 p=0.36 Baseline Symptoms Negative 4 11 12w PANSS Total Mean 77.33 61 p=0.07 Minimum 67 35 Maximum 86 98 6m PANSS Total Mean 71 58.75 p=0.12 Minimum 60 37 Mean 77.33 97 PANSS % Change at Mean 1.83 -21.92 p=0.17 12w Minimum -25 -53.57 Maximum 26.47 4.11 PSP % Change at Mean 11.34 35.08 p=0.44 12w Minimum -47.73 -50 Maximum 57.14 162.50 GAF % Change at Mean 22.56 43.91 p=0.51 12w Minimum -25 -8.33 Maximum 96.67 185.71 Responder at 12w Yes 1 9 p=0.72 No 3 9 Increased Yes 2 11 p=1 appetite/weight at 12w No 2 7 Drowsiness at 12w Yes 3 12 p=1 No 1 6 Clozapine Levels at Mean 0.315 0.43 p=.09 12w Minimum 0.28 0.21 Maximum 0.35 0.75 - 282 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes Figure 10: Pie charts comparing key demographics for those with and without cardiovascular symptoms - 283 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 284 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 285 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 286 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 287 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 288 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 289 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 290 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 291 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 292 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes ) mg/l ( - 293 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 294 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes 5.5.2 Global DNA methylation (cardiovascular symptoms) Not experiencing cardiovascular symptom was associated with increased global DNA methylation over time. For those without cardiovascular symptoms, an additional week on clozapine was associated with an average additional increase in global DNA methylation of 0.00035%. 51.30% of all CpG sites showed an additional increase in methylation associated with time exposed to clozapine if not experiencing cardiovascular symptoms and 48.70% of all CpG sites showed a decrease. However, when exploring individual time points, split by response: at baseline, mean global DNA methylation was 48.218% for those with cardiovascular symptoms and 48.300% for those without; at six weeks, 48.240% for those with and 48.257% for those without; at twelve weeks 48.259% for those with and 48.294% for those without; and at six months 48.301% for those with and 48.292% for those without. Those who experience cardiovascular symptoms show lower starting level of global DNA methylation which gradually increase; while those who do not experience these adverse reactions start with higher levels which decrease then increase and end similar to baseline but lower than those who do experience the adverse reactions. - 295 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes 5.5.3 Differentially methylated positions (cardiovascular symptoms) No DMPs reached the threshold for epigenome wide significance (p<1*10-7) but eighty two DMPs were trend-level significant (p<5*10-5). - 296 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes - 297 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes Table 23: Top ten DMPs associated with cardiovascular symptoms (* indicates epigenome wide significance) CpG site Beta P value Gene Location MAPINFO cg12590655 0.85x10-3 1.95E-07 164340525 0.155 x10- cg12026992 2 5.37E-07 90004011 cg07374403 -0.48 x10-3 6.65E-07 ATPAF1 TSS200 47134188 cg00692277 0.83 x10-3 1.28E-06 3230847 -0.132 cg23892310 x10-2 2.19E-06 TBCD Body 80816702 -0.218 cg15404971 x10-2 2.82E-06 7392579 0.203 x10- cg06846719 2 3.11E-06 FAM60A 5'UTR 31475291 -0.187 cg18978540 x10-2 3.85E-06 SMAD9 5'UTR 37493431 -0.121 cg14596987 x10-2 4.15E-06 PTCH1; 5'UTR;Body 98249189 cg15718289 0.87 x10-3 4.49E-06 HOXC4;HOXC5;HOXC6 5'UTR;Body;1stExon 54410671 - 298 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes 5.5.4 Pathway analysis (cardiovascular symptoms) Pathway analysis indicated that the pathways in Table 24 were the top thirty pathways with an over-representation of DMPs associated with cardiovascular symptoms. - 299 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes Table 24 Top thirty over-represented pathways - cardiovascular symptoms Category Name p value over- Number of differentially Number in Ontology representation methylated category GO:0009897 external side of plasma membrane 0.00199 11 203 CC GO:0016887 ATPase activity 0.00221 19 352 MF GO:0032648 regulation of interferon-beta production 0.00235 4 31 BP GO:0032608 interferon-beta production 0.00292 4 33 BP GO:0098552 side of membrane 0.00391 14 275 CC GO:2001222 regulation of neuron migration 0.00410 4 23 BP GO:0032312 regulation of ARF GTPase activity 0.00417 5 25 BP GO:0033158 regulation of protein import into nucleus 0.00430 3 16 BP GO:0010875 positive regulation of cholesterol efflux 0.00503 3 13 BP GO:0061630 NA 0.00508 3 15 MF GO:0007127 meiosis I 0.00510 6 70 BP GO:0032728 positive regulation of interferon-beta production 0.00520 3 21 BP GO:0042623 ATPase activity 0.00520 14 243 MF GO:0006163 purine nucleotide metabolic process 0.00525 51 1290 BP GO:0009203 ribonucleoside triphosphate catabolic process 0.00560 42 967 BP GO:0009207 purine ribonucleoside triphosphate catabolic process 0.00560 42 967 BP GO:0009146 purine nucleoside triphosphate catabolic process 0.00574 42 970 BP GO:0030660 Golgi-associated vesicle membrane 0.00578 4 36 CC GO:0009143 nucleoside triphosphate catabolic process 0.00596 42 974 BP GO:0007131 reciprocal meiotic recombination 0.00626 4 33 BP GO:0035825 reciprocal DNA recombination 0.00626 4 33 BP GO:0006152 purine nucleoside catabolic process 0.00627 42 978 BP - 300 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes Category Name p value over- Number of differentially Number in Ontology representation methylated category GO:0046130 purine ribonucleoside catabolic process 0.00627 42 978 BP GO:0046128 purine ribonucleoside metabolic process 0.00641 46 1138 BP GO:0042454 ribonucleoside catabolic process 0.00666 42 988 BP GO:0009150 purine ribonucleotide metabolic process 0.00675 50 1269 BP GO:0042278 purine nucleoside metabolic process 0.00699 46 1140 BP GO:0032373 positive regulation of sterol transport 0.00714 3 15 BP GO:0032376 positive regulation of cholesterol transport 0.00714 3 15 BP GO:0009164 nucleoside catabolic process 0.00756 42 999 BP 5.5.5 Candidate CpG site methylation (cardiovascular symptoms) Of seven sites located within CYP1A2 with methylation data available, none had a significant association with cardiovascular symptoms before or after correction for multiple testing. Across all CYP1A2 CpG sites, there was a mean additional increase in methylation of 0.00035%. - 301 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes 5.6 Discussion: Adverse reactions to clozapine To explore the effects of clozapine on DNA methylation associated with adverse reactions, I used multi-level models to analyse between-individual variation in the experience of increased appetite/weight, drowsiness and cardiovascular symptoms as predictors of DNA methylation trajectories over time. Systems involved in signal transduction, immune function and inflammation, neuronal development, energy metabolism, glycogen, and cholesterol are implicated. 5.6.1 Increased appetite/weight Analysis of the association between epigenome-wide DNA methylation and increased appetite/weight identified three DMPs that were significant at the epigenome-wide threshold (cg10966876, ch.6.163729680F and cg21097199) and two hundred and fifty one other DMPs showing trends towards significance. The majority of the top ten DMPs showed additional decreases in methylation over time associated with not experiencing increased appetite/weight. Those who didn’t experience adverse reactions showed a pretty stable global DNA methylation while those who did started at a higher level which decreased then increased above baseline. Eight of the top ten sites were located within a gene. Interpreting the overlap in biological processes implicated by models of both response and increased appetite/weight can be a challenge. In this sample, while non-significant, there was a trend for participants with a greater response to clozapine to have a higher likelihood of weight gain. It may be that one model is confounded by the other, but it also may be that there is a meaningful association in both models due to a shared underlying mechanism. The first and seventh ranked DMP are located within TWIST2 and UNKL, two genes associated with the immune system, the former involved in cytokine repression. Immune dysfunction has been discussed above in relationship with schizophrenia and clozapine; however inflammation and immune dysfunction are also independently associated with weight gain. Adipose tissue releases cytokines and some have hypothesised that this weight related inflammatory response (particularly increased levels of tumour necrosis factor-α (TNF-α)) leads to metabolic disruption such as insulin resistance (J. Yudkin, 2007). Obesity has been reliably associated with reduced immune response, such as impaired response to flu vaccines (Sheridan et al., 2012), increased infection and reduced host defences such as lower lymphocyte numbers (Milner & Beck, 2012). Similarly, the fifth ranked DMP is within GGT7, a gene associated with glutathione function. Glutathione is an antioxidant which protects against oxidative stress, and oxidative stress has been associated with obesity. Furukawa et al., (2004) report that fat accumulation correlates with systemic oxidative stress in - 302 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes studies of humans and animal models, with increased production of reactive oxygen species and decreased antioxidant enzymes (including glutathione peroxidase), and dysregulation of cytokines such as IL-6. Genetic variation in glutathione s-transferases (GSTs) genes has also been associated with obesity (Almoshabek, Mustafa, Al-Asmari, Alajmi, & Al-Asmari, 2016). Looking at the literature on immune changes and inflammation associated with response to antipsychotics and weight gain, both indicate reduced immune function, which may be a common mechanism underlying therapeutic response to clozapine and clozapine-induced weight gain. TWIST2 is also associated with energy metabolism and glycogen storage, more direct mechanisms for weight gain and metabolic dysfunction. Glycogen is the product of excess carbohydrates becoming stored for a secondary source of energy, after those needed are metabolised to release energy; if there is further excess, they will be converted to fat and eventually adipose tissue. TWIST2, POLQ – in which the third DMP was located – and FOXN3 are all associated with different aspects of gene expression regulation (transcription regulation and repression, DNA polymerase) and pathway analysis also indicated overrepresentation of DMPs involved in translation in response to exogenous stress. As with previous models, this suggests clozapine may have an indirect effect associated with metabolic symptoms via methylation of regulatory genes. Candidate gene testing provided support for an association between increased appetite/weight and methylation within NPY gene, with increase in methylation at cg11653966 for those who didn’t experience increased appetite/weight. Neuropeptide Y, encoded by the NPY gene, is a neurotransmitter which has been shown to stimulate food intake (Pomonis, Levine, & Billington, 1997) and is also involved in storage of energy as fat, so altered regulation of this gene has a clear potential mechanism for weight gain and increased appetite; increased methylation in those who don’t experience increased appetite/weight may be driving reduced expression of this gene and reduced stimulation of food intake. SNPs within NPY have previously been associated with antipsychotic induced weight-gain (In patients primarily treated with clozapine or olanzapine) (Tiwari et al., 2013), suggesting that future work should integrate genetic and epigenetic data on the association between this gene and increased appetite/weight driven by clozapine. There was weak support for any association with CNR1 and MC4R; no associations survived correction for multiple testing. - 303 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes 5.6.2 Drowsiness Analysis of the association between epigenome-wide DNA methylation and drowsiness identified two DMPs that were significant at the epigenome-wide threshold (cg12155676 and cg21399492) and one hundred and eighty nine other sites showing trends towards significance. The majority of the top ten DMPs show additional increases in methylation over time in those without drowsiness. Those who did not experience drowsiness show a relatively stable global DNA methylation with slight increase while those who do show an initial drop and then increase above baseline. Five of the top ten differentially methylated sites are located outside of genes, with five located within genes. The fourth ranked DMP was within EFHA1, a gene associated with calcium ion binding and mitochondrial calcium uptake; disruption in calcium signalling in schizophrenia and the ability of clozapine to normalise this is discussed above. The ninth ranked DMP was located within PTPN14, and this gene, as well as the first ranked pathway, implicates association with cell growth, cell migration and cell adhesion as well as lamellipodium, a cytoskeletal protein projection which facilitates cell migration, propelling cells forward. PTPN14 and other overrepresented pathways also implicate differential methylation in genes associated with antioxidant activity and lymphatic development. The seventh DMP was located in MAST, a gene associated with kinase signalling. Additionally, the sixth and eighth ranked DMPs are located within POLQ and ZSCAN which is associated with regulation of gene expression - DNA and RNA polymerase, nucleic acid binding, and transcription factor activity – as is PTPN14. Unfortunately, there is very minimal research studying causes of individual variation in antipsychotic induced sedation which makes it difficult to place these findings in context; this may be because drowsiness is difficult to define and measure, and differentiate from negative symptoms and issues with motivation, but also because it is considered low risk. However, it is incredibly common and can have a significant impact on quality of life – as Miller, (2004) notes, drowsiness can cause difficulties when reintegrating and interfere with treatment and adherence. These preliminary findings may be a starting point for pursuing a better understanding of why some patients experience greater sedation in response to clozapine. While none of the above mentioned CpG sites were statistically significant, one interesting thing to note is the systems implicated – calcium signalling, cell migration, cell adhesion, inflammation and kinase signalling – were all implicated in analysis of therapeutic response to clozapine also. Candidate gene testing provided only weak support for an association between CYP1A2 methylation and drowsiness; no associations survived correction for multiple testing. - 304 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes 5.6.3 Cardiovascular symptoms Analysis of the association between epigenome-wide DNA methylation and cardiovascular symptoms identified no DMPs that were significant at the epigenome-wide threshold but eighty two trending towards significance. Over the top ten DMPs there was an additional increase in methylation associated with not experiencing cardiovascular symptoms. Those who experience cardiovascular symptoms showed lower starting levels of global DNA methylation which gradually increase, while those who do not experience these adverse reactions start with higher levels which decrease then increase and end at a similar level to baseline levels. Four of the top ten DMPs are located outside of genes, with six are located within genes. The seventh and eight ranked DMPs are located within FAM60A and SMAD9, two genes involved in TGF-beta activity (a cytokine involved in cell growth, proliferation and death), and pathway analysis also indicates overrepresentation of DMPs involved interferon beta production, again implicating the immune system which has been discussed numerous times throughout this thesis. Cardio-vascular problems have been linked to inflammation and immune function in the same way that obesity has (J. S. Yudkin, Kumari, Humphries, & Mohamed-Ali, 2000); in particular, myocarditis – a known adverse reaction to clozapine – is an infection of heart muscle. SMAD9 is also associated with hypertension, often a precursor to cardiovascular damage or itself classified as heart disease. The ninth ranked DMP was located in PTCH1, a gene involved in cholesterol binding, and pathway analysis also indicated overrepresentation of DMPs involved in cholesterol regulation; there is a strong association between cholesterol levels and cardiovascular health, with high levels of low-density lipoprotein cholesterol associated with increased risk of cardiovascular health problems (SCANDINAVIAN SIMVASTATIN SURVIVAL Study Group, 1994) and high levels of high-density lipoprotein cholesterol associated with reduced risk (Gordon et al., n.d.). Antipsychotics have been shown to alter cholesterol levels, so this may be a mechanism via which clozapine modulates risk of cardiovascular symptoms (Hennekens, Hennekens, Hollar, & Casey, 2005). Candidate gene testing found no support for any association between CYP1A2 methylation and cardiovascular symptoms. 5.6.4 Summary There were three epigenome-wide significant hits for increased weight/appetite - cg10966876 in TWIST2 (a transcription factor involved in cytokine expression, glycogen storage and energy metabolism), ch.6.163729680F, and cg21097199 in POLQ (encodes DNA polymerase theta)– and two - 305 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes for drowsiness - cg12155676 and cg21399492. Participants with increased appetite/weight and drowsiness showed an initial decrease in global DNA methylation and then an increase above baseline, whereas participants without remained stable. Participants with cardiovascular symptoms showed a lower baseline global DNA methylation which increased with time, whereas those without started higher and this decreased and then returned to baseline levels. This analysis is limited by the less powered between-participant analysis, especially for the cardiovascular analysis with only four participants who reported cardiovascular symptoms. It is likely that those with the most serious adverse reactions or risk of adverse reactions discontinue clozapine early or are never commenced on clozapine, and are therefore not represented in this sample. There is much less literature on adverse responses to antipsychotics than therapeutic response, however some of these findings coincide with our existing understanding of these adverse reactions, such as an association between increased appetite/weight and DMPs involved in energy metabolism and glycogen function and an association between cardiovascular symptoms and DMPs involved in cholesterol function. Interestingly, there is also significant overlap between systems implicated by my analysis of both therapeutic response to clozapine and adverse reactions to clozapine: immune function, inflammation and antioxidant activity (metabolic, drowsiness and cardiovascular), signal transduction including calcium signalling (metabolic and drowsiness), nervous system development and migration (metabolic and drowsiness), kinase signalling (drowsiness), and ATP activity (cardiovascular). This suggests that mechanism of action of therapeutic response to clozapine may also be related to adverse reactions. For example, clozapine’s regulation of calcium signalling may help normalise aberrant signalling in some systems leading to improved symptoms but create aberrant signalling in other systems leading to drowsiness and metabolic effects. Similarly, the reduction of elevated immune function may have beneficial effects on psychotic symptoms but be directly linked to increased risk of numerous adverse reactions Some methodological caveats should be considered when interpreting the results and the strength of the conclusions that can be drawn. Firstly, even with optimal collection of GASS-C data, the cause of self-reported adverse reactions is undetermined and cannot be definitively attributed to clozapine. Physical sensations such as changes in heart rate or appetite may relate to changes in psychiatric symptoms (e.g. anxiety and depressive symptoms, respectively), changes in lifestyle, or may relate to comorbid illness or adjunctive medications, psychiatric or otherwise. While changes in psychiatric medication were included as variables within the analysis, numerous confounding variables remain. Secondly, while the original intention was to look at change in GASS-C scores so - 306 - Adverse Reactions to Clozapine and Associated DNA Methylation Changes that baseline scores were accounted for, due to significant missing data at baseline the analysis only takes into account cross-sectional report of adverse events at 12 weeks. Therefore, any attribution of these adverse events to clozapine is even more uncertain, as they may have preceded the initiation of clozapine or may even have reduced during clozapine treatment. The analysis simply reflects differences between clozapine patients who report or do not report experiencing these symptoms at 12 weeks; we cannot confidently state that they are experiencing clozapine-induced adverse events. Additionally, as with symptomatic response, the question remains whether these changes in methylation are causal mechanisms or simply reflections of the biological processes underlying adverse reactions. - 307 - General Discussion 6 General Discussion In this chapter, I will first provide an overview of the key findings from the previous results chapters, highlighting areas of overlap. I will then discuss the strengths and limitations of the study and analysis approach, and the subsequent limitations for interpreting the results. Finally, I will put my findings in context of the previous literature on epigenetic studies of antipsychotics and schizophrenia, and schizophrenia aetiology as a whole, before making suggestions for the future research needed to establish the causal role of DNA methylation in clozapine’s mechanism of action. 6.1 Summary of findings This thesis has used Illumina 450k DNA methylation data from twenty-two participants who provided a total of eighty five samples of peripheral blood at four time-points over the first six months of their clozapine initiation. This data has been analysed to address associations between changes in DNA methylation and three broad areas: exposure to clozapine, therapeutic response to clozapine and adverse reactions to clozapine. This is the first research to investigate numerous time- points within six months after clozapine initiation, the first to report on categorical comparison between responder and non-responders, numerous measures of therapeutic response and adverse reactions to clozapine, and the first to have rigorous modelling methods controlling for numerous known confounders and using strict correction for multiple testing. The findings implicate a wide variety of biological systems known to be disrupted in schizophrenia as systems showing differential methylation associated with exposure to and response to clozapine. Exposure to clozapine across models was associated with differential methylation in loci involved in signal transduction, G-coupled protein receptors, AMPAR function, immune function, glucose homeostasis, beta-tubulin and regulation of gene expression, with four epigenome-wide significant DMPs. Candidate gene testing identified an association between clozapine level and methylation at a locus on PVALB, encoding parvalbumin which is involved in calcium signalling at GABAergic interneurons. Change in response to clozapine across models was associated with differential methylation in loci similarly involved in signal transduction, ion channels (especially calcium), G- coupled protein receptors, glutamatergic function, immune function, glucose homeostasis, microtubules, and regulation of gene expression, but additionally neuronal precursors and cell adhesion; there were five epigenome-wide significant DMPs. Symptom severity was similarly associated with differential methylation in loci involved in signal transduction, ion channels, receptor function, immune function, cell adhesion and regulation of expression, but additionally neuronal development and glycine function; there were twenty two epigenome-wide significant DMPs, mostly associated with negative symptom severity. Finally, adverse reactions to clozapine across models - 308 - General Discussion were also associated with differential methylation in loci involved in signal transduction, calcium channels, immune function, cell adhesion, neuronal development and regulation of gene expression –suggesting an association between systems involved in therapeutic response and adverse reaction - but additionally increased appetite/weight gain was associated with energy metabolism and glycogen function and cardiovascular adverse reactions were associated with cholesterol function; there were five epigenome-wide significant DMPs. Candidate gene testing identified an association between increased appetite/weight and methylation at a locus on NPY, a gene involved in regulation of food intake. Signal transduction was a common theme throughout the results, with numerous aspects implicated (ion channels and signalling including potassium and calcium, G-coupled protein receptors, kinase signalling, Wnt signalling, GTP activity, ATP activity, as well as various aspects of neuronal development); regarding specific neurotransmitter systems, AMPAR function was associated with exposure to clozapine and psycho- social functioning, NMDAR and GABA receptor function was associated with response status, and glycine receptor function was associated with symptom severity. Immune function was another common theme, also with numerous aspects implicated – cytokines including interleukins, inflammation response, antioxidant activity, T-cells, interferon gamma signalling and major histocompatibility complex. Of note is that 41% of the top ten DMPs for all models associated with clozapine exposure and response have been previously reported to have a significant correlation between methylation in peripheral blood and brain tissue (70.59% of those in TSS15000 regions and 55.56% of those within 5’UTR regions); while a minority of DMPs, this is a substantial minority and indicates an enrichment of DMPS that have correlated methylation in peripheral blood and brain tissue. 6.2 Strengths and limitations A significant and novel strength of this research is the repeated longitudinal sampling within- individuals over a six month period, as for the analyses of clozapine exposure and within-individual symptom variation this controls for all between-individual variation, increasing the statistical power and simplifying interpretation of the findings by removing numerous potential confounding variables including mQTLs. It also provides information on methylation trajectories, as opposed to a static data point or a simple before and after; this contribution to the literature can be built upon to construct a more complete picture of the consequences of clozapine exposure, and how these relate to therapeutic response and adverse reactions. Another strength is the collection of phenotypic data included in analysis: smoking status at each time point, clozapine levels for a significant proportion of samples, medication history and - 309 - General Discussion medication usage at each time point, and numerous measures of symptom severity, psychosocial functioning and quality of life at each time point. Further phenotypic information on clinical history was also available to help characterise the patient group. Cell count data was also estimated from the methylation data and acted as another covariate which helps limit the potential for confounding due to tissue composition heterogeneity; Jaffe & Irizarry, (2014) found that cell composition correlates strongly with global DNA methylation and that longitudinal changes in cell composition had confounded a number of previous studies investigating age-related changes in DNA methylation, leading to false positives. However, while predicted cell counts do reduce this bias, Shakhbazov et al., (2016) report that they do not fully account for this bias. Unfortunately, cell count data collected by clinical teams was not available to our research team as expected but future studies should seek to obtain observed cell proportions to ensure accurate correction of results. This is one of the first studies to utilise the blood-brain DNA methylation comparison tool to assess Illumina 450k results and report the predicted association between changes in methylation within our peripheral tissue sample and changes in methylation in post mortem blood for each DMP, assisting in interpretation of the findings. While peripheral blood was the only realistic option to enable longitudinal sampling over the period of clozapine initiation, only 41% (37/90) of the top ten DMPs for all models were reported to have a significant correlation between methylation in peripheral blood and brain tissue, and many of these correlations - despite being statistically significant - were weak correlations with low effect sizes. This raises the question of whether the methylation changes reported in this analysis do reflect meaningful changes occurring in the brains of the participants studied. However, 41% of DMPs showing a blood-brain correlations represents a much higher proportion than expected based on previous studies, and is significantly higher than the 7.9% reported by others (Walton et al., 2015), suggesting optimism for the relevance of these findings. Also, while acknowledging the appropriate caveats for peripheral blood, the caveats for other available options cannot be ignored: post-mortem tissue is likely to have accumulated numerous epigenetic changes associated with illness (Pidsley & Mill, 2011) and could not have enabled longitudinal analysis, and animal studies cannot provide complete models of schizophrenia and represent the complex experience of psychotic symptoms in humans. All of these approaches benefit from validation using alternative approaches, to identify consistent findings and build a complete picture of the mechanisms at work. The Illumina 450k array also has limitations. It is not the gold standard of bisulfite genomic sequencing which covers the complete DNA methylome (Dedeurwaerder et al., 2011). While - 310 - General Discussion providing a high through-put method of epigenome-wide methylation data, the 450k array only includes a small proportion of the full range of CpG sites and it is unclear whether we should be focusing on other regions across the epigenome; it is heavily weighted towards promoter regions, but intergenic CpG sites may show more differential methylation and may act as alternate promotors for some genes, particularly those without CpG islands in their promoter regions (Sarda et al., 2017). A number of intergenic CpG sites were identified as top DMPs within my analysis, and developing our understanding of the function of CpG sites such as these is worth pursuing. Some have proposed using methods such as methyl-capture sequencing, which still has significantly lower costs then whole genome sequencing while identifying a greater number of CpG sites then the Illumina 450k array, but it appears to be less sensitive to small effect sizes and thus less suitable for this analysis (Teh et al., 2016). Some of the other issues raised with the Illumina 450k, such as potential errors from probe cross-hybridization (Teh et al., 2016), batch effects (Leek et al., 2010), SNPs contained in probes (Dedeurwaerder et al., 2011) and incomplete bisulfite conversion (D. P. Chen, Lin, & Fann, 2016) can be accounted for, and have in this analysis; cross-hybridising probes, probes containing SNPs, and samples with below 90% bisulfite conversion were removed, and array design and analysis controlled for batch effects. Also, despite the limitations with array-based technologies, they are the most popular - in 2016, (D. P. Chen et al., 2016) reported that over 566 publications have used the Illumina 450k array - and this enables direct comparison between study findings, the use of standardised statistical packages and pipelines, and the use of tools such as the blood-brain correlation tool. Additionally, comparison of methylation profiles obtained using the 450k array and methylation profiles obtained from other arrays and whole-genome bisulfite sequencing suggest high accuracy and reproducibility (Bibikova et al., 2011; Dedeurwaerder et al., 2011). However, the caveat that these microarrays only include a very small fraction of all CpG sites should not be forgotten. Additionally, methylation cannot be differentiated from hydroxymethylation, another chemical modification of cytosines which appears to be most prevalent in adult neuronal cells (Hao Wu & Zhang, 2014) and may have differing epigenetic consequences; high levels of hydroxymethylcytosine are more commonly associated with highly expressed genes (Johnson et al., 2016; Wu et al., 2011). The relatively small sample size is another concern; while every effort was made to identify and recruit all eligible participants due to commence clozapine within our NHS Trust, the time- sensitive nature of the eligibility combined with the requirement for numerous clinical interviews and retention over follow up, in a highly symptomatic patient group, made recruitment challenging. This is in contrast to GWAS in which a single sample and basic phenotypic data are collected and - 311 - General Discussion often at least one of these is collected retrospectively (due to the static nature of DNA sequence), allowing for vast samples of thousands of participants. For the within-participant analysis this is less of a concern because within-participant analysis has increased statistical power, but the between-participant analysis of therapeutic response at twelve weeks and the comparison of patients with and without adverse reactions is less reliable due to the small number in each group; this is particularly true for the adverse reactions analysis in which some of the groups had especially low numbers, whereas the responder vs non- responder categorisation was more evenly split. The small sample also raises issues of generalisability; patterns of DNA methylation in our participants may not be applicable to the broader population of patients with schizophrenia. For example, our sample had a high number of non-responders, potentially due to the clinical teams we recruited through. Also, while gender and smoking were included as covariates, our sample was predominantly male and had a surprisingly low number of smokers. More broadly, while replication of results should always be a strong priority in science to avoid drawing conclusions based on false positives, it becomes even more acutely necessary in a small sample such as this one where the likelihood of spurious findings is increased. Because of the relatively small sample analysed, it was not feasible to do reliable analysis of predictors within the baseline data. Similarly, I could not control for additional covariates such as drug and alcohol use, as this may have led to overfitting the model; therefore, I selected the covariates with the strongest justification for inclusion (such as smoking status and cell counts which are known confounders in methylation studies), and those with most variability over the six months (e.g. I focused on including more sensitive variables of medication use at each time point than history of medication use). The issue of sample size raises the issue of p-values and significance testing. On one hand, epigenome-wide studies run the risk of false positives due to the huge number of CpG sites being assayed, hence the need for a threshold of p<1*10-7 to be considered epigenome wide significant (Lehne et al., 2015). In fact, a very recent analysis (published after the current analysis was conducted) accounts for correlation between neighbouring CpG sites and indicates that a slightly less conservative significance threshold of 2.4 x 10-7 is appropriate for the 450k array (;they also show that a hypothetical genome-wide EWAS array would require a more stringer statistical of 3.6x10-8, highlighting that the literature standard for significance thresholds will need to be dynamic and adjust with developing technology (Safarri et al., 2018). Still, support from broader literature and pathway analysis is of great assistance in determining the likelihood of a false positive: broader - 312 - General Discussion literature can provide plausible mechanisms for a causal relationship as well as give some general support for the involvement of a system in clozapine’s mechanism of action; and it is less likely that a combination of numerous positions from the same pathway are of greater significance by chance than an individual position being of greater significance by chance. The effect sizes for many of the models are also very small, however this should be understood in the context of the measurement scales for other variables; for example, when reporting findings for analysis of time on clozapine, each effect size reported relates to the change predicted for just an additional week on clozapine. While the within-subject analyses holds many variables constant, it does not eliminate the possibility of changes in unmeasured variables such as diet, sleep, social interactions etc.; these are likely to improve with improved symptoms (which would in turn be associated with greater clozapine levels, greater response, greater time exposed to clozapine), and could have consequences for DNA methylation and therefore become a potential confound. Furthermore, a number of the analyses are between-subject; while I have controlled for basic demographic and clinical variables and group comparisons find no significant differences, it is possible that other unmeasured between-group differences including mQTLs could account for the different trajectories associated with therapeutic response or adverse reactions. A particular concern for between-group genetic analysis is the effect of ethnicity and population stratification, in which between-group differences in genetic variables (and therefore epigenetic variables) are driven by differences in the underlying genetic structure of the populations i.e. different proportions of different ethnicities, rather than the group difference of interest. In this sample, there were no significant differences in racial classifications for any of the analyses, but these may have been overly crude categorisations. Also, differences may have had a meaningful impact despite being non-significant; for example, there appears to be a higher proportion of black patients in the group of patients who experienced increase appetite or weight gain, and though this did not meet thresholds for statistical significance it may still have impacted on the results of the analysis. As discussed in the previous chapter, the adverse events analysis is also affected by the lack of baseline data and inability to determine whether adverse events can be attributed to clozapine or other confounding variables. The time-points chosen may not be the most appropriate; the most relevant changes in DNA methylation may occur within hours of the first exposure to clozapine, or may occur after chronic use extending beyond six months of clozapine use. However, as an early study in the area, the time points have clinical relevance for when therapeutic blood levels are reached, when response to clozapine is determined, and after a period of relatively stable long-term use. Many CpG sites appear - 313 - General Discussion to show different directions of change within the first six weeks compared to the rest of the six months; this may indicate an immediate methylation change which is then regulated by subsequent feedback to attenuate the initial reaction, or may indicate that the early direction of effect is a continuation of methylation fluctuation before clozapine was introduced, which then changes once clozapine has been maintained at therapeutic levels for a sustained period. Further studies exploring longer study periods before and after clozapine initiation, and exploring more fine-tuned short term study periods after clozapine initiation are needed. It also may have been valuable to record and analyse norclozapine levels, as a pharmacologically active metabolite of clozapine which has been speculated to underlie some of clozapines unique effects. Our study did not include healthy volunteers as a control group, either on or off clozapine. We would not have been able to include healthy volunteers on clozapine as a control group, to separate out schizophrenia-related changes, due to the significant ethical concerns about prescribing unnecessary long-term antipsychotic treatment with potential adverse reactions with no clinical benefit to volunteers. We did not include healthy volunteers or patients with schizophrenia off clozapine as a control group, which would potentially enable separation of non-clozapine related changes over-time, because in a small non-randomised sample there would likely be vast group differences and different variables changing over time, adding significant noise to any-between participant analysis; therefore, we chose to focus efforts on obtaining data to allow within- participant comparisons and conduct analyses such as the clozapine level and change in symptom severity model, which are unlikely to reflect natural changes over time. However, future analysis could investigate comparisons to datasets of longitudinal DNA methylation changes in healthy controls, and larger scale studies could seek to recruit appropriate control groups to aid interpretation of the results. A significant caveat is that, as Pidsley and Mill (2011) state: “statistical evidence of an interaction reveals little about the molecular pathologic mechanisms underlying psychosis”. While the longitudinal design of the study is an improvement over cross-sectional designs for inferring causality, this limitation remains relevant. This is especially pertinent when methylation across CpG sites is highly correlated, and associations with changes at one probe may reflect a true mechanistic change at another locus. Even if methylation at a specific CpG site is correctly identified as pertinent to clozapine’s mechanism of action, a) this may not be associated with expression of the gene it is located in/nearby and b) it is currently unclear how this might lead to downstream effects on symptoms and adverse reactions in schizophrenia. - 314 - General Discussion Furthermore -as Lappalainen & Greally, (2017) note - while a change in DNA methylation is likely to be reflective of cellular reprogramming, it is not necessarily directly involved or mediating this process; it may simply reflect low transcription factor occupancy, instead of actual cellular changes reflected directly in methylation. In the current analysis, these alternative explanations cannot be distinguished. A better understanding of the functional implications of methylation changes induced by clozapine is critical; firstly, investigating whether these translate to altered gene expression as expected, but then in vitro and animal model work is required to experimentally determine the effect on relevant biological symptoms and subsequent behaviour. Another caveat to the current research is that it studies DNA methylation in isolation, whereas evidence suggests that DNA methylation does not act independently from histone modifications. Incorporating data on histone modifications induced by clozapine could therefore provide a more complete picture; however, currently there is not the same readily available technology for identifying epigenome-wide histone modifications. Another consideration is the heterogeneity of schizophrenia – in treatment-resistance, a key distinction may be between early onset resistance and delayed onset resistance (Demjaha et al., 2017) - with the potential for numerous aetiologies pathways underlying small subsets of the population of schizophrenia. This is one motivator for research into rare genetic variants within schizophrenia, and this may be equally fruitful with epigenetic research. (Dempster et al., 2011) found some large methylation changes specific to one or a few twin pairs, suggesting some rare epigenetic alternations of large effect may be associated with schizophrenia in individuals. Responses to clozapine treatment may be similarly heterogeneous; changes in certain biological pathways may be more relevant in certain individuals than others and these individual differences may get lost when looking for consistent patterns among patients. The apparent outliers evident in some of the analyses may reflect this, though they may also be spurious. The interpretation of this data and the pathway analysis also rests upon the gene annotation databases. All pathway analysis approaches share certain limitations (Khatri, Sirota, Butte, Kim, & Woo, 2012), namely limited information about which transcripts are active in each pathway and the consequences for each pathway, but also incomplete and potentially inaccurate annotation; much is derived from indirect evidence, though this improves with each new version. There are also concerns that pathway analysis is biased towards certain pathways, with some much more heavily annotated due to previous research interest, which may lead to false positive errors (pathways implicated simply due to heavy annotation) or false negative errors (novel pathways or locus dismissed due to their limited annotation) (Haynes, Tomczak, & Khatri, 2017). To reduce concerns of - 315 - General Discussion false positive errors, the statistical package I used in this analysis – GOSeq (M. D. Young, Davidson, Wakefield, Smyth, & Oshlack, 2016) – accounts for pathways which have an existing bias in annotation using probability weighting functions; this makes a significant differences to results when compared to approaches which assume equality of pathways (M. D. Young, Wakefield, Smyth, & Oshlack, 2010). 6.3 Comparison to broader literature Based on the previous studies reporting epigenetic changes induced by clozapine (Aoyama et al., 2013; Dong et al., 2008; Huang et al., 2007; Kinoshita et al., 2017), candidate gene testing was done for all available probes on GAD1, PVALB, RELN, GRIN2A, GRIN2D and GRM7 in association with exposure to clozapine (time or levels). The only DMP which survived correction for multiple testing was cg19136704 within a promoter region of the PVALB gene, where there was a positive association between methylation and clozapine levels; however, the previous study of clozapine had found increased mRNA expression which is not generally associated with increased methylation of promoter regions. Altered parvalbumin activity, particularly dysfunction of parvalbumin-expressing GABAergic interneurons, has been reliably demonstrated in schizophrenia and linked to underlying cognitive impairments (Gonzalez-Burgos et al., 2015; Stedehouder & Kushner, 2016), and understanding the impact of clozapine on parvalbumin is worth pursing further. More broadly, pathway analysis of the categorical response model indicated that there was an over-representation of DMPs involved in regulation of GABA transport associated with response to clozapine, which provides support for the proposal (Guidotti et al., 2012) that clozapine response may be driven by a normalisation GABAergic gene methylation and gene expression. This also coincides with the study by Kinoshita et al., (2017). Broadly speaking, the systems implicated by Kinoshita – GABA, glutamate and cell-adhesion – are also implicated by my analysis; these coinciding results provide support for the involvement of these systems in clozapine’s mechanism of action. However, some of the specific findings receive no support from our analysis: they report demethylation at GAD1, which we tested in our candidate gene analysis and did not find, and they report an association between CREBBP (a binding protein involved in histone acetyltransferase activity and stabilising protein interactions) and changes in symptom severity, which we did not find associated with response. Based on our systematic review of all gene-association studies of clozapine response, candidate gene testing was done for all probes on the gene with independent replication of an association with response to clozapine - DRD3, 5HTR2A, 5HTT/SLC6A4, and GNB3 - in association with categorical response classification to clozapine in our data. None of these survived correction for multiple - 316 - General Discussion testing. This need not dismiss these variants as relevant; loci that have relevant genetic variance may not have relevant epigenetic variance and vice versa. Despite a number of studies suggesting that valproate, a histone deacetylase, enhances the epigenetic effects of clozapine e.g. (Dong et al., 2008) we were not able to conduct statistical analysis on this within the current sample as there were only two instances of patients taking adjunct valproate. Considering sites of differential methylation associated with schizophrenia identified by a review of peripheral blood studies (Teroganova et al., 2016), we found differential methylation trajectories associated with clozapine response (higher baseline levels and further increases in responders) in a gene previously associated with hyper methylation in schizophrenia – DLG4. But more broadly, this review indicated aberrant methylation of genes involved in glutamate, GABA, signal transduction, calcium signalling, potassium channels, microtubule associated proteins, kinase signalling, insulin related genes, glial cells, and immune function, all of which were also implicated by analysis of DMPs associated with exposure and/or response to clozapine. Similarly, considering studies of post- mortem of methylation differences in schizophrenia, these have implicated glutamatergic and GABAergic function, cell proliferation, GCPRs, potassium channels and protein-kinase signalling (Mill et al., 2008; Ruzicka et al., 2015), all of which were systems we found to be associated with exposure and/or response to clozapine. Cumulatively, while a formal comparison to previously identified genetic variants associated with schizophrenia risk was not conducted, my analysis of methylation changes associated with exposure to clozapine and subsequent response implicates systems associated with a wide array of hypotheses regarding schizophrenia. This is not a problem, many of these hypotheses are complementary, focusing on specific dysfunctional systems but acknowledging their interaction with other systems. Drawing ideas from a number of studies (Berridge, 2016; Feigensona et al., 2014; O. Howes, McCutcheon, & Stone, 2015; Landek-Salgado, Faust, & Sawa, 2016; Leza et al., 2015; Lidow, 2003)– we could conceptualise the neurodevelopmental pathway of schizophrenia in the following way: Inflammation and disrupted cytokine levels, caused by exposure to risk factors early in development and/or dysfunction of genes involved in the immune system, lead to altered development of the central nervous system. This inflammation and associated oxidative stress leads to dysregulation of the cytoskeleton – the structural proteins which guide the function and morphology of dendrites and synapses, such as beta-tubulin – perhaps mediated by dysfunctional microglial regulation of these processes. The ubiquitin proteasome pathway, which targets proteins - 317 - General Discussion for degradation and regulates a wide array of functions around the body, is also disrupted. This leads to alterations in neurogenesis, synaptic pruning, ATP metabolism, synapse regulation, dendritic spine density, ion channels, white matter development and connectivity and cell adhesion. This could all be exacerbated by polymorphisms in neuroplasticity-related genes such as RELN and DISC1. Stress in childhood or adolescence may further exacerbate this inflammatory state, perhaps via activation of the HPA axis leading to even higher levels of cytokines, and stress could contribute to increased synaptic pruning through increased subcortical dopamine release. The disrupted development and formation of neurons leads to dysfunction in the signalling of multiple neurotransmitter systems, and the altered immune function also directly leads to this via interactions between microglia and NMDA and T cells and dopamine receptors. Polymorphisms in genes regulating calcium signalling activity may enhance or mediate some of these effects, as disrupted calcium signalling leads to disrupted regulation of: dopamine synaptic uptake, NMDAR inhibition and subsequent GABAergic inhibitory interneuron function, myelination proteins and formation of synaptic connections and cell maintenance. Abnormalities in AMPAR regulation may lead to disrupted depolarisation of post-synaptic membranes, a prerequisite for NMDAR activity at glutamate synapses, therefore leading to altered NMDA signalling. Calmodulin dysfunction may exacerbate the disruption of AMPAR, increasing their sensitivity. NMDAR dysfunction may be further exacerbated by reductions in levels of glycine, which is also required for NMDAR signalling, and by low levels of GABA caused by GABA risk genes. NMDA receptor hypofunction leads to diminished inhibition of the glutamatergic system – in the short term this leads to increased glutamate and hyperactivation of other glutamate receptors, but long term may lead to reductions in glutamate and an upregulation of NMDAR to compensate. This glutamatergic dysfunction may also lead to subsequent mesolimbic dopaminergic hyperactivity and frontal dopaminergic hypoactivity, perhaps through further reductions in GABA synthesis due to inhibition of inhibitory GABA interneurons and/or via inhibition of AMPA receptor desensitisation. In turn, reduced NMDAR signalling reduces the body’s antioxidant response and therefore its protection against inflammation, leading to further damage. Similarly, dopamine activity increases oxidative stress and affects pyramidal spine density. Over time, this may lead to reductions in grey matter and neuronal density. It has already been suggested that DNA methylation could play a role in this chain of events from an early stage with hypermethylation occurring alongside inflammatory processes (Davis, Moylan, Harvey, Maes, & Berk, 2014) or hypermethylation caused by inflammatory processes (Khojasteh- Fard, Tabrizi, & Amoli, 2011). The results of this thesis provide supporting evidence that clozapine - 318 - General Discussion exposure and response is associated with changes in DNA methylation at loci implicated in nearly all aspects of this chain of events. This may be a mechanism for triggering normalisation of these dysfunctional systems, it may be a mediating mechanism in the normalisation, or it could be simply a reflection of normalisation that has occurred. Either way, the vast array of aetiologically relevant systems and processes seemingly affected by clozapine treatment, as indicated by these results, may explain clozapine’s unique effectiveness and ability to address a broader range of the symptoms seen in schizophrenia. Previous literature supports the complex and extensive effects of clozapine, with evidence of clozapine’s ability to: supress immune response (Song et al., 2000), regulate T cells (Fernandez-Egea et al., 2016), influence Wnt signalling (Sutton et al., 2007) and signal transduction cascades (Browning et al., 2005), normalise G-protein dysfunction (Avissar, Barki-Harrington, et al., 2001) and calcium and potassium channel system disruption (Park et al., 2015; Peltola et al., 2016), influence AMPAR function (Schmitt et al., 2003) and NMDAR function (Schwieler, Linderholm, Nilsson-Todd, Erhardt, & Engberg, 2008; Tanahashi et al., 2012), activate glycine receptors (Harvey & Yee, 2013), antagonise GABA receptors (Ohno-Shosaku et al., 2011), increase white matter integrity (Steiner et al., 2014), and normalise disruption of axonal growth processes and dendritic spine deficits (Barros et al., 2009; Critchlow et al., 2006; Park et al., 2015). Kedracka-Krok et al., (2015) report that clozapine induces changes in proteins across all aspects of the cytoskeleton, neuronal and non-neuronal cells, regulatory and synaptic proteins and comment that the robust efficacy of clozapine may be due to the fact it has “one of the most complex pharmacological profiles” with a unique receptor binding profile. Broadly speaking, these results provide optimism that longitudinal psychiatric epigenetic studies can identify within-participant changes in methylation in disease-relevant loci, with enrichment of DMPs previously shown to have correlating methylation in brain and blood. 6.4 Future research There are two broad avenues for future research on DNA methylation and clozapine use. Firstly, research which enables more robust conclusions about the causal and mechanistic processes induced by clozapine and associated with clinical outcomes. An initial step is investigating whether differential methylation trajectories associated with clozapine exposure or response are associated with equivalent changes in gene expression at the same loci or whether different genes are regulated by these CpG sites. Further research should then integrate numerous strands of evidence, to enable a complete picture of the causal relationship between genetic variants, epigenetic modifications including DNA methylation but also histone modifications, gene expression, protein expression, downstream effects and changes in phenotype; for example, providing a - 319 - General Discussion connection between findings of differential methylation in genes associated with GABAergic receptors and findings of altered GABAergic transmission. Experimental work to determine the consequences of methylation at individual CpG sites is particularly warranted. Additional longitudinal studies using different timeframes surrounding clozapine initiation would also be of utility to provide a fine-grained temporal resolution of the methylation trajectories induced by clozapine and seen prior to clozapine initiation. Studies could also look at methylation changes related to levels of norclozapine, the pharmacologically active metabolite of clozapine. In general, research in other larger samples should be used to replicate these findings and different methylation profile techniques and whole genome sequencing should be used to validate these findings and the pathways implicated. Both replication and validation are of utmost importance to ensure the findings are not compromised by the limitations of the small sample size and the caveats regarding Illumina 450k analysis. Larger samples would also allow for more covariates to be included in the analysis (in particular, they should consider population stratification) and generally reduce the chance of spurious findings. This strand of research could enable a greater understanding of the underlying dysfunction of schizophrenia itself, or treatment-resistant schizophrenia specifically, particularly if formal comparisons to datasets indicating genetic variants associated with schizophrenia were conducted. This could highlight the genetic dysfunction which clozapine works to normalise, and which mediates psychotic experiences and improvement in symptoms. Secondly, while this sample was not large enough to identify predictive DMPs at baseline, future studies should aim to explore the potential for DNA methylation in peripheral blood to act as a biomarker for clinical outcomes including both therapeutic response and adverse reactions. Research in this direction will have to consider whether there are consistent predictors and markers across individuals, or whether the most meaningful variation is specific to individuals or sub-groups. Assuming consistent markers can be identified, this research would not be hindered by many of the limitations associated with use of peripheral blood, the Illumina 450k array or bisulfite sequencing in general, such as uncertainty about whether we are actually identifying hydroxymethylation and whether differential methylation is also seen in brain tissue; a biomarker is useful if it is reliable predictor, regardless of the underlying causal relationship or whether it reflects the “true” mechanism. There would be significant immediate benefit to prediction of both favourable and adverse outcomes of clozapine. Reducing the time spent with inadequate or ineffective antipsychotics would reduce both the economic cost of drug treatment, and more importantly the suffering to patients as their symptoms endure and life circumstances potentially worsen while taking a trial and error approach to their medication. Prior knowledge of adverse outcomes would - 320 - Conclusions allow patients to make informed choices about their medications, and could assist clinicians in proactively preventing or treating adverse reactions. 7 Conclusions In conclusion, my analysis of longitudinal samples of peripheral blood taken from patients over the first six months of clozapine treatment identified numerous differentially methylated positions in DNA associated with clozapine exposure, therapeutic response to clozapine and adverse reactions (thirty-three of which were epigenome wide significant). The results suggest that clozapine’s mechanism of action may involve differential methylation at genes involved in signal transduction, GPCRs, glutamatergic systems (including NMDARs and AMPARs), GABAergic systems, glycine receptor function, calcium signalling, glucose homeostasis, immune function, neuronal development, beta tubulin and microtubules, cell adhesion and regulation of transcription; and additionally, that some of these systems may also be involved in adverse reactions to clozapine. The within-participant design, exploration of clinical outcomes, and enrichment of positions which have evidence of correlated methylation in brain and blood provides some confidence that these findings are not spurious and are relevant to clozapine’s efficacy. However, replication is required and DNA methylation is only one piece of the puzzle: the complete mechanistic picture of casual processes leading to clozapine’s remarkable efficacy remains unclear. - 321 - REFERENCES REFERENCES Ajami, A., Abedian, F., Hamzeh Hosseini, S., Akbarian, E., Alizadeh-Navaei, R., & Taghipour, M. (2014). Serum TNF-??, IL-10 and IL-2 in schizophrenic patients before and after treatment with risperidone and clozapine. Iranian Journal of Immunology : IJI, 11(3), 200–209. https://doi.org/IJIv11i3A6 Allen, N. C., Bagade, S., McQueen, M. B., Ioannidis, J. P. A., Kavvoura, F. K., Khoury, M. J., … Bertram, L. (2008). Systematic meta-analyses and field synopsis of genetic association studies in schizophrenia: the SzGene database. Nature Genetics, 40(7), 827–834. https://doi.org/10.1038/ng.171 Allison, D. B., Mentore, J. L., Heo, M., Chandler, L. P., Cappelleri, J. C., Infante, M. C., & Weiden, P. J. (1999). Antipsychotic-induced weight gain: a comprehensive research synthesis. The American Journal of Psychiatry, 156(11), 1686–1696. https://doi.org/10.1176/ajp.156.11.1686 Almoshabek, H. A., Mustafa, M., Al-Asmari, M. M., Alajmi, T. K., & Al-Asmari, A. K. (2016). Association of glutathione S-transferase GSTM1 and GSTT1 deletion polymorphisms with obesity and their relationship with body mass index, lipoprotein and hypertension among young age Saudis. JRSM Cardiovascular Disease, 5, 204800401666964. https://doi.org/10.1177/2048004016669645 Anders, S., & Kinney, D. K. (2015). Abnormal immune system development and function in schizophrenia helps reconcile diverse findings and suggests new treatment and prevention strategies. Brain Research, 1617, 93–112. https://doi.org/10.1016/j.brainres.2015.02.043 Annunziatio, A. (2009). DNA Packaging: Nucleosomes and Chromatin. Retrieved September 27, 2017, from https://www.nature.com/scitable/topicpage/dna-packaging-nucleosomes-and- chromatin-310 Aoyama, Y., Mouri, A., Toriumi, K., Koseki, T., Narusawa, S., Ikawa, N., … Nabeshima, T. (2013). Clozapine ameliorates epigenetic and behavioral abnormalities induced by phencyclidine through activation of dopamine D1 receptor. The International Journal of Neuropsychopharmacology / Official Scientific Journal of the Collegium Internationale Neuropsychopharmacologicum (CINP), 1–15. https://doi.org/10.1017/S1461145713001466 Arango, C., Buchanan, R. W., Breier, A., MacMahon, R., & Carpenter, W. T. (2003). The relationship of clozapine and haloperidol treatment response to prefrontal, hippocampal, and caudate brain volumes. American Journal of Psychiatry, 160(8), 1421–1427. https://doi.org/10.1176/appi.ajp.160.8.1421 - 322 - REFERENCES Arranz, M., Collier, D., Munro, J., & Sham, P. (1996). Analysis of a structural polymorphism in the 5- HT 2A receptor and clinical response to clozapine. Neuroscience …, 217, 177–178. Retrieved from http://www.sciencedirect.com/science/article/pii/0304394096130949 Arranz, M. J., Collier, D. A., Sodhi, M., Ball, D., Roberts, G. W., Sham, P., … Roberts, G. W. (1995). Association between clozapine response and allelic variation in 5-HT2A receptor gene. The Lancet, 346(8970), 281–282. https://doi.org/10.1016/S0140-6736(95)92168-0 Arranz, M. J., Li, T., Munro, J., Liu, X., Murray, R., Collier, D. A., & Kerwin, R. W. (1998). Lack of association between a polymorphism in the promoter region of the dopamine-2 receptor gene and clozapine response. Pharmacogenetics, 8(6), 481–4. https://doi.org/10.1097/00008571- 199812000-00004 Arranz, M. J., Munro, J., Birkett, J., Bolonna, a, Mancama, D., Sodhi, M., … Kerwin, R. W. (2000). Pharmacogenetic prediction of clozapine response. Lancet, 355(9215), 1615–1616. https://doi.org/10.1016/S0920-9964(00)90535-8 Arranz, M., Munro, J., & Owen, M. (1998). Evidence for association between polymorphisms in the promoter and coding regions of the 5-HT2A receptor gene and response to clozapine. Molecular Psychiatry, 3, 61–66. https://doi.org/10.1038/sj.mp.4000348 Avissar, S., Barki-Harrington, L., Nechamkin, Y., Roitman, G., Schreiber, G., Hornykiewics, O., … Mita, T. (2001). Elevated dopamine receptor-coupled G(s) protein measures in mononuclear leukocytes of patients with schizophrenia. Schizophrenia Research, 47(1), 37–47. https://doi.org/10.1016/S0920-9964(00)00038-4 Avissar, S., Roitman, G., & Schreiber, G. (2001). Differential effects of the antipsychotics haloperidol and clozapine on G protein measures in mononuclear leukocytes of patients with schizophrenia. Cellular and Molecular Neurobiology, 21(6), 799–811. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/12043849 Azorin, J. M., Spiegel, R., Remington, G., Vanelle, J. M., Péré, J. J., Giguere, M., & Bourdeix, I. (2001). A double-blind comparative study of clozapine and risperidone in the management of severe chronic schizophrenia. American Journal of Psychiatry, 158(8), 1305–1313. https://doi.org/10.1176/appi.ajp.158.8.1305 Bakshi, V. P., Swerdlow, N. R., & Geyer, M. A. (1994). Clozapine antagonizes phencyclidine-induced deficits in sensorimotor gating of the startle response. The Journal of Pharmacology and Experimental Therapeutics, 271(2), 787–794. Retrieved from http://jpet.aspetjournals.org/content/jpet/271/2/787.full.pdf - 323 - REFERENCES Balu, D. T., & Coyle, J. T. (2015). The NMDA receptor “glycine modulatory site” in schizophrenia: D- serine, glycine, and beyond. Current Opinion in Pharmacology, 20, 109–115. https://doi.org/10.1016/j.coph.2014.12.004 Ban, T. A. (2007). Fifty years chlorpromazine: a historical perspective. Neuropsychiatric Disease and Treatment, 3(4), 495–500. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/19300578 Bannister, A. J., & Kouzarides, T. (2011). Regulation of chromatin by histone modifications. Cell Research, 21(3), 381–95. https://doi.org/10.1038/cr.2011.22 Barbeau, D., Jun Liang, J., ROBITAILLEt, Y., Quirion, R., Srivastava, L. K., & by Dominick Purpura, C. P. (1995). Decreased expression of the embryonic form of the neural cell adhesion molecule in schizophrenic brains. Neurobiology, 92, 2785–2789. Retrieved from http://www.pnas.org/content/92/7/2785.full.pdf Barlas, I. O., Cetin, M., Erdal, M. E., Semiz, U. B., Basoglu, C., Ay, M. E., … Uzun, O. (2009). Lack of association between DRD3 gene polymorphism and response to clozapine in Turkish schizoprenia patients. American Journal of Medical Genetics, Part B: Neuropsychiatric Genetics, 150(1), 56–60. https://doi.org/10.1002/ajmg.b.30770 Barros, C. S., Calabrese, B., Chamero, P., Roberts, A. J., Korzus, E., Lloyd, K., … Müller, U. (2009). Impaired maturation of dendritic spines without disorganization of cortical cell layers in mice lacking NRG1/ErbB signaling in the central nervous system. Proceedings of the National Academy of Sciences of the United States of America, 106(11), 4507–12. https://doi.org/10.1073/pnas.0900355106 Bates, D., Maechler, M., Bolker, B., & Walker, S. (2014). lme4: Linear mixed-effects models using Eigen and S4. Retrieved from https://stats.stackexchange.com/questions/88905/do-i-have-to- cite-lme4-when-using-lmertest/88918 Becker, K. G. (2004). The common variants/multiple disease hypothesis of common complex genetic disorders. Medical Hypotheses, 62(2), 309–317. https://doi.org/10.1016/S0306- 9877(03)00332-3 Beherec, L., Lambrey, S., Quilici, G., Rosier, A., Falissard, B., & Guillin, O. (2011). Retrospective Review of Clozapine in the Treatment of Patients With Autism Spectrum Disorder and Severe Disruptive Behaviors. Journal of Clinical Psychopharmacology, 31(3), 341–344. https://doi.org/10.1097/JCP.0b013e318218f4a1 Benca, R., Obermeyer, W., Thisted, R., & Gillin, J. (1992). Sleep and Psychiatric Disorders: A Meta- analysis. Arch Gen Psychiatry, 49(8), 651–668. Retrieved from - 324 - REFERENCES http://jamanetwork.com/journals/jamapsychiatry/fullarticle/495851 Berridge, M. J. (2014). Calcium signalling and psychiatric disease: bipolar disorder and schizophrenia. Cell Tissue Res, 357(2), 477–492. https://doi.org/10.1007/s00441-014-1806-z Berridge, M. J. (2016). Dysregulation of neural calcium signaling in Alzheimer disease, bipolar disorder and schizophrenia. Prion, 7(1), 2–13. https://doi.org/10.4161/pri.21767 Bibikova, M., Barnes, B., Tsan, C., Ho, V., Klotzle, B., Le, J. M., … Shen, R. (2011). High density DNA methylation array with single CpG site resolution. Genomics, 98(4), 288–295. https://doi.org/10.1016/j.ygeno.2011.07.007 Bishop, J. R., Ellingrod, V. L., Moline, J., & Miller, D. (2006). Pilot study of the G-protein beta3 subunit gene (C825T) polymorphism and clinical response to olanzapine or olanzapine-related weight gain in persons with schizophrenia. Medical Science Monitor, 12(2), BR47-50. Blin, O., & Micallef, J. (2001). Antipsychotic-associated weight gain and clinical outcome parameters. Journal of Clinical Psychiatry. Boks, M. P., de Jong, N. M., Kas, M. J. H., Vinkers, C. H., Fernandes, C., Kahn, R. S., … Ophoff, R. A. (2012). Current status and future prospects for epigenetic psychopharmacology. Epigenetics, 7(1), 20–28. https://doi.org/10.4161/epi.7.1.18688 Bönsch, D., Wunschel, M., Lenz, B., Janssen, G., Weisbrod, M., & Sauer, H. (2012). Methylation matters? Decreased methylation status of genomic DNA in the blood of schizophrenic twins. Psychiatry Research, 198(3), 533–537. https://doi.org/10.1016/j.psychres.2011.09.004 Bousman, C. A., Chana, G., Glatt, S. J., Chandler, S. D., Lucero, G. R., Tatro, E., … Everall, I. P. (2010). Preliminary evidence of ubiquitin proteasome system dysregulation in schizophrenia and bipolar disorder: Convergent pathway analysis findings from two independent samples. American Journal of Medical Genetics Part B: Neuropsychiatric Genetics, 153B(2), 494–502. https://doi.org/10.1002/ajmg.b.31006 Bradshaw, N. J., & Porteous, D. J. (2012). DISC1-binding proteins in neural development, signalling and schizophrenia. Neuropharmacology, 62(3), 1230–1241. https://doi.org/10.1016/j.neuropharm.2010.12.027 Brandl, E., Tiwari, A., Zai, C., Nurmi, E., Chowdhury, N., Arenovich, T., … Müller, D. (2015). Genome- wide association study on antipsychotic-induced weight gain in the CATIE sample. The Pharmacogenomics Journal, 16, 352–356. https://doi.org/10.1038/tpj.2015.59 Breier, A., Buchanan, R. W., Kirkpatrick, B., Davis, O. R., Irish, D., Summerfelt, A., & Carpenter, W. T. - 325 - REFERENCES (1994). Effects of Clozapine on Positive and Negative Symptoms in Outpatients With Schizophrenia. American Journal of Psychiatry, 151, 20–26. Retrieved from http://ajp.psychiatryonline.org/doi/pdf/10.1176/ajp.151.1.20 Breier, A., Buchanan, R. W., Waltrip, R. W., Listwak, S., Holmes, C., & Goldstein, D. S. (1994). The effect of clozapine on plasma norepinephrine: relationship to clinical efficacy. Neuropsychopharmacology : Official Publication of the American College of Neuropsychopharmacology, 10(1), 1–7. https://doi.org/10.1038/npp.1994.1 Brennand, K. J., Simone, A., Jou, J., Gelboin-Burkhart, C., Tran, N., Sangar, S., … Gage, F. H. (2011). Modelling schizophrenia using human induced pluripotent stem cells. Nature, 473(7346), 221– 225. https://doi.org/10.1038/nature09915 Brennand, K., Savas, J. N., Kim, Y., Tran, N., Simone, A., Hashimoto-Torii, K., … Gage, F. H. (2015). Phenotypic differences in hiPSC NPCs derived from patients with schizophrenia. Molecular Psychiatry, 20(3), 361–8. https://doi.org/10.1038/mp.2014.22 Brenner, H. D., Dencker, S. J., Goldstein, M. J., Hubbard, J. W., Keegan, D. L., Kruger, G., … Midha, K. K. (1990). Defining Treatment Refractoriness in Schizophrenia. Schizophrenia Bulletin, 16(4), 551–561. https://doi.org/10.1093/schbul/16.4.551 Browning, J. L., Patel, T., Brandt, P. C., Young, K. A., Holcomb, L. A., & Hicks, P. B. (2005). Clozapine and the mitogen-activated protein kinase signal transduction pathway: Implications for antipsychotic actions. Biological Psychiatry, 57(6), 617–623. https://doi.org/10.1016/j.biopsych.2004.12.002 Brune, G., & Himwich, H. (1962). Effects of methionine loading on the behavior of schizophrenic patients. The Journal of Nervous and Mental Disease, 134(5), 447. Retrieved from http://sfx.kcl.ac.uk/kings?sid=google&auinit=GG&aulast=Brune&atitle=Effects+of+methionine+ loading+on+the+behavior+of+schizophrenic+patients.&title=The+journal+of+nervous+%26+me ntal+disease&volume=134&issue=5&date=1962&spage=447&issn=0022-3018 Buchanan, R. W., Javitt, D. C., Marder, S. R., Schooler, N. R., Gold, J. M., McMahon, R. P., … Carpenter, W. T. (2007). The Cognitive and Negative Symptoms in Schizophrenia Trial (CONSIST): The efficacy of glutamatergic agents for negative symptoms and cognitive impairments. American Journal of Psychiatry, 164(10), 1593–1602. https://doi.org/10.1176/appi.ajp.2007.06081358 Buckley, P., Thompson, P., Way, L., & Meltzer, H. Y. (1994). Substance Abuse Among Patients With Treatment-Resistant Schizophrenia: Characteristics and Implications for Clozapine. American - 326 - REFERENCES Journal of Psychiatry, 151(3), 385–389. Bush, W. S., & Moore, J. H. (2012). Chapter 11: Genome-Wide Association Studies. PLOS Computational Biology, 8(12), 1–11. https://doi.org/10.1371/journal.pcbi.1002822 Butler, A. A., & Cone, R. D. (2001, October 1). Knockout models resulting in the development of obesity. Trends in Genetics. Elsevier Current Trends. https://doi.org/10.1016/S0168- 9525(01)02481-7 Bychkov, E., Ahmed, M. R., & Gurevich, E. V. (2011). Sex differences in the activity of signalling pathways and expression of G-protein-coupled receptor kinases in the neonatal ventral hippocampal lesion model of schizophrenia. Int J Neuropsychopharmacol, 14(1), 1–15. https://doi.org/10.1007/s12020-009-9266-z.A Capannolo, M., Fasciani, I., Romeo, S., Aloisi, G., Rossi, M., Bellio, P., … Maggio, R. (2015). The atypical antipsychotic clozapine selectively inhibits interleukin 8 (IL-8)-induced neutrophil chemotaxis. European Neuropsychopharmacology : The Journal of the European College of Neuropsychopharmacology, 25(3), 413–24. https://doi.org/10.1016/j.euroneuro.2014.12.002 Cardno, A. G., Marshall, E. J., Coid, B., Macdonald, A. M., Ribchester, T. R., Davies, N. J., … Murray, R. M. (1999). Heritability estimates for psychotic disorders: the Maudsley twin psychosis series. Archives of General Psychiatry, 56(2), 162–8. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/10025441 Carroll, L. S., & Owen, M. J. (2009). Genetic overlap between autism, schizophrenia and bipolar disorder. Genome Medicine, 1(10), 102. https://doi.org/10.1186/gm102 Carter, C. J. (2007). EIF2B and oligodendrocyte survival: Where nature and nurture meet in bipolar disorder and schizophrenia? Schizophrenia Bulletin, 33(6), 1343–1353. https://doi.org/10.1093/schbul/sbm007 Catapano, L., & Manji, H. (2007). G Protein-Coupled Receptors in Major Psychiatric Disorders. Biochim Biophys Acta, 1768(4), 976–993. https://doi.org/10.1007/s12020-009-9266-z.A Chen, D. P., Lin, Y. C., & Fann, C. S. J. (2016). Methods for identifying differentially methylated regions for sequence- and array-based data. Briefings in Functional Genomics, 15(6), 485–490. https://doi.org/10.1093/bfgp/elw018 Chen, M. L., Tsai, T. C., Lin, Y. Y., Tsai, Y. M., Wang, L. K., Lee, M. C., & Tsai, F. M. (2011). Antipsychotic drugs suppress the AKT/NF-κB pathway and regulate the differentiation of T-cell subsets. Immunology Letters, 140(1–2), 81–91. https://doi.org/10.1016/j.imlet.2011.06.011 - 327 - REFERENCES Choi, K. H., & Rhim, H. (2010). Inhibition of recombinant Cav3.1 (alpha1G) T-type calcium channels by the antipsychotic drug clozapine. European Journal of Pharmacology, 626(2–3), 123–130. https://doi.org/10.1016/j.ejphar.2009.09.035 Cipriani, A., Boso, M., & Barbui, C. (2009). Clozapine combined with different antipsychotic drugs for treatment resistant schizophrenia. The Cochrane Database of Systematic Reviews, (3), CD006324. https://doi.org/10.1002/14651858.CD006324.pub2 Clancy, S., & Brown, W. (2008). Translation: DNA to mRNA to Protein. Cohrs, S. (2008). Sleep disturbances in patients with schizophrenia: Impact and effect of antipsychotics. CNS Drugs. Springer International Publishing. https://doi.org/10.2165/00023210-200822110-00004 Cole, J., Goldberg, S., & Davis, J. (1966). Drugs in the treatment of psychosis: Controlled Studies. In Psychiatric Drugs. New York: Grune and Stratton. Retrieved from https://scholar.google.com/scholar_lookup?title=Drugs in the treatment of psychosis%3A controlled studies&author=JO. Cole&author=SC. Goldberg&author=JM. Davis&pages=153- 180&publication_year=1966 Collier, D. A., Eastwood, B. J., Malki, K., & Mokrab, Y. (2016). Advances in the genetics of schizophrenia: Toward a network and pathway view for drug discovery. Annals of the New York Academy of Sciences, 1366(1), 61–75. https://doi.org/10.1111/nyas.13066 Conley, R., & Kelly, D. (2001). Management of treatment resistance in schizophrenia. Biological Psychiatry, 50(11), 898–911. https://doi.org/10.1016/S0006-3223(01)01271-9 Conley, R. R., & Kelly, D. L. (2001). Management of treatment resistance in schizophrenia. Biological Psychiatry, 50(1988), 898–911. https://doi.org/10.1016/s0006-3223(01)01271-9 Consortium, T. N. and P. A. S. of the P. G. (2015). Psychiatric genome-wide association study analyses implicate neuronal, immune and histone pathways. Nature Neuroscience, 18(2), 199. https://doi.org/10.1038/nn.3922 Cooper, A. F. (1976). Deafness and psychiatric illness. British Journal of Psychiatry, 129(9), 216–226. Retrieved from http://bjp.rcpsych.org/content/bjprcpsych/129/3/216.full.pdf Corfas, G., Roy, K., & Buxbaum, J. D. (2004). Neuregulin 1-erbB signaling and the molecular/cellular basis of schizophrenia. Nature Neuroscience, 7(6), 575–580. https://doi.org/10.1038/nn1258 Corvin, A., O ’dushlaine, C., Kenny, E., Heron, E., Donohoe, G., Morris, D., & Gill, M. (2010). Molecular Pathways involved in Neuronal Cell Adhesion and Membrane Scaffolding contribute to - 328 - REFERENCES Schizophrenia and Bipolar Disorder Susceptibility. Molecular Psychiatry. Retrieved from https://hal.archives-ouvertes.fr/hal-00509758 Critchlow, H. M., Maycox, P. R., Skepper, J. N., & Krylova, O. (2006). Clozapine and haloperidol differentially regulate dendritic spine formation and synaptogenesis in rat hippocampal neurons. Molecular and Cellular Neuroscience, 32(4), 356–365. https://doi.org/10.1016/j.mcn.2006.05.007 Dalton, S. O., Johansen, C., Poulsen, A. H., Nørgaard, M., Sørensen, H. T., McLaughlin, J. K., … Friis, S. (2006). Cancer risk among users of neuroleptic medication: a population-based cohort study. British Journal of Cancer, 95(7), 934–939. https://doi.org/10.1038/sj.bjc.6603259 Danysz, W. (2002). Positive modulators of AMPA receptors as a potential treatment for schizophrenia. Current Opinion in Investigational Drugs (London, England : 2000), 3(7), 1062–6. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/12186268 Datta, S. R., McQuillin, a, Rizig, M., Blaveri, E., Thirumalai, S., Kalsi, G., … Gurling, H. M. D. (2010). A threonine to isoleucine missense mutation in the pericentriolar material 1 gene is strongly associated with schizophrenia. Molecular Psychiatry, 15(6), 615–628. https://doi.org/10.1038/mp.2008.128 Davis, J., Moylan, S., Harvey, B. H., Maes, M., & Berk, M. (2014). Neuroprogression in schizophrenia: Pathways underpinning clinical staging and therapeutic corollaries. Australian & New Zealand Journal of Psychiatry, 48(6), 512–529. https://doi.org/10.1177/0004867414533012 De Hert, M., Correll, C. U., & Cohen, D. (2010). Do antipsychotic medications reduce or increase mortality in schizophrenia? A critical appraisal of the FIN-11 study. Schizophrenia Research, 117(1), 68–74. https://doi.org/10.1016/j.schres.2009.12.029 De Luca, V., Mueller, D. J., de Bartolomeis, A., & Kennedy, J. L. (2007). Association of the HTR2C gene and antipsychotic induced weight gain: a meta-analysis. The International Journal of Neuropsychopharmacology, 10(5). https://doi.org/10.1017/S1461145707007547 de Silva, V. A., Suraweera, C., Ratnatunga, S. S., Dayabandara, M., Wanniarachchi, N., & Hanwella, R. (2016). Metformin in prevention and treatment of antipsychotic induced weight gain: a systematic review and meta-analysis. BMC Psychiatry, 16(1), 341. https://doi.org/10.1186/s12888-016-1049-5 Deb, S., Thomas, M., & Bright, C. (2001). Mental disorder in adults with intellectual disability. 2: the rate of behaviour disorders among a community‐based population aged between 16 and 64 years. Journal of Intellectual Disability Research, 45(6), 495–505. Retrieved from - 329 - REFERENCES http://ww2.justanswer.es/uploads/vickyb61/2006-10-28_214913_Mental_disorder1.pdf Debnath, M. (2015). Adaptive Immunity in Schizophrenia: Functional Implications of T Cells in the Etiology, Course and Treatment. Journal of Neuroimmune Pharmacology, 10(4), 610–619. https://doi.org/10.1007/s11481-015-9626-9 Dedeurwaerder, S., Defrance, M., Calonne, E., Denis, H., Sotiriou, C., & Fuks, F. (2011). Evaluation of the Infinium Methylation 450K technology. Epigenomics, 3(6), 771–784. https://doi.org/10.2217/epi.11.105 Dedeurwaerdere, S., Boets, S., Janssens, P., Lavreysen, H., & Steckler, T. (2015). In the grey zone between epilepsy and schizophrenia: alterations in group II metabotropic glutamate receptors. Acta Neurologica Belgica, 115(3), 221–232. https://doi.org/10.1007/s13760-014-0407-7 Demjaha, A., Lappin, J. M., Stahl, D., Patel, M. X., MacCabe, J. H., Howes, O. D., … Murray, R. M. (2017). Antipsychotic treatment resistance in first-episode psychosis: prevalence, subtypes and predictors. Psychological Medicine, 47, 1981–1989. https://doi.org/10.1017/S0033291717000435 Dempster, E. L., Pidsley, R., Schalkwyk, L. C., Owens, S., Georgiades, A., Kane, F., … Mill, J. (2011). Disease-associated epigenetic changes in monozygotic twins discordant for schizophrenia and bipolar disorder. Human Molecular Genetics, 20(24), 4786–4796. https://doi.org/10.1093/hmg/ddr416 Deutch, A. Y., & Cameron, D. S. (1992). Pharmacological characterization of dopamine systems in the nucleus accumbens core and shell. Neuroscience, 46(1), 49–56. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/1350665 Dincer, A., Gavin, D. P., Xu, K., Zhang, B., Dudley, J. T., Schadt, E. E., & Akbarian, S. (2015). Deciphering H3K4me3 broad domains associated with gene-regulatory networks and conserved epigenomic landscapes in the human brain. Translational Psychiatry, 5(11), e679. https://doi.org/10.1038/tp.2015.169 Dong, E., Nelson, M., Grayson, D. R., Costa, E., & Guidotti, A. (2008). Clozapine and sulpiride but not haloperidol or olanzapine activate brain DNA demethylation. Proceedings of the National Academy of Sciences of the United States of America, 105(36), 13614–9. https://doi.org/10.1073/pnas.0805493105 Doyle, A. C., & Pollack, M. H. (2003). Establishment of remission criteria for anxiety disorders. The Journal of Clinical Psychiatry, 64 Suppl 15, 40–5. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/14658990 - 330 - REFERENCES Drago, A., Giegling, I., Schäfer, M., Hartmann, A. M., Konte, B., Friedl, M., … Rujescu, D. (2014). Genome-wide association study supports the role of the immunological system and of the neurodevelopmental processes in response to haloperidol treatment. Pharmacogenetics and Genomics, 24(6), 314–319. https://doi.org/10.1097/FPC.0000000000000052 Du, F., Cooper, A. J., Thida, T., Sehovic, S., Lukas, S. E., Cohen, B. M., … Öngür, D. (2014). In Vivo Evidence for Cerebral Bioenergetic Abnormalities in Schizophrenia Measured Using 31 P Magnetization Transfer Spectroscopy. JAMA Psychiatry, 71(1), 19. https://doi.org/10.1001/jamapsychiatry.2013.2287 Duan, X., Chang, J. H., Ge, S., Faulkner, R. L., Kim, J. Y., Kitabatake, Y., … Song, H. (2007). Disrupted-In- Schizophrenia 1 Regulates Integration of Newly Generated Neurons in the Adult Brain. Cell, 130(6), 1146–1158. https://doi.org/10.1016/j.cell.2007.07.010 Duncan, C. E., Chetcuti, A. F., & Schofield, P. R. (2008). Coregulation of genes in the mouse brain following treatment with clozapine, haloperidol, or olanzapine implicates altered potassium channel subunit expression in the mechanism of antipsychotic drug action. Psychiatric Genetics, 18(5), 226–239. https://doi.org/10.1097/YPG.0b013e3283053019 Duncan, L. E., Holmans, P. A., Lee, P. H., O’Dushlaine, C. T., Kirby, A. W., Smoller, J. W., … Cohen, B. M. (2014). Pathway analyses implicate glial cells in schizophrenia. PLoS ONE, 9(2), e89441. https://doi.org/10.1371/journal.pone.0089441 Eack, S. M., Bahorik, A. L., McKnight, S. A. F., Hogarty, S. S., Greenwald, D. P., Newhill, C. E., … Minshew, N. J. (2013). Commonalities in social and non-social cognitive impairments in adults with autism spectrum disorder and schizophrenia. Schizophrenia Research, 148(1–3), 24–28. https://doi.org/10.1016/j.schres.2013.05.013 Elkis, H. (2007). Treatment-Resistant Schizophrenia. Psychiatric Clinics of North America, 30(3), 511– 533. https://doi.org/10.1016/j.psc.2007.04.001 Enomoto, Y., Yamanishi, Y., Izawa, K., Kaitani, A., Takahashi, M., Maehara, A., … Kitaura, J. (2010). Characterization of leukocyte mono-immunoglobulin-like receptor 7 (LMIR7)/CLM-3 as an activating receptor: its similarities to and differences from LMIR4/CLM-5. The Journal of Biological Chemistry, 285(46), 35274–83. https://doi.org/10.1074/jbc.M110.137166 Ertugrul, A., Ucar, G., Basar, K., Demir, B., Yabanoglu, S., & Ulug, B. (2007). Influence of clozapine on platelet serotonin, monoamine oxidase and plasma serotonin levels. Psychiatry Research, 149(1–3), 49–57. https://doi.org/10.1016/j.psychres.2005.12.009 Ertugrul, A., Volkan-Salanci, B., Basar, K., Oguz, K. K., Demir, B., Ergun, E. L., … Ulug, B. (2009). The - 331 - REFERENCES effect of clozapine on regional cerebral blood flow and brain metabolite ratios in schizophrenia: Relationship with treatment response. Psychiatry Research - Neuroimaging, 174(2), 121–129. https://doi.org/10.1016/j.pscychresns.2009.04.007 Essock, S. M., Hargreaves, W. A., Covell, N. H., & Goethe, J. (1996). Clozapine’s effectiveness for patients in state hospitals: results from a randomized trial. Psychopharmacology Bulletin, 32(4), 683–97. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/8993092 Evins, A., Amico, E., Shih, V., & Goff, D. C. (1997). Clozapine treatment increases serum glutamate and aspartate compared to conventional neuroleptics. Journal of Neural Transmission, 104(6– 7), 761–766. Evins, A. E., Fitzgerald, S. M., Wine, L., Rosselli, R., & Goff, D. C. (2000). Placebo-controlled trial of glycine added to clozapine in schizophrenia. American Journal of Psychiatry, 157(5), 826–828. https://doi.org/10.1176/appi.ajp.157.5.826 Farde, L., Nordström, A. L., Wiesel, F. A., Pauli, S., Halldin, C., & Sedvall, G. (1992). Positron emission tomographic analysis of central D1 and D2 dopamine receptor occupancy in patients treated with classical neuroleptics and clozapine. Relation to extrapyramidal side effects. Archives of General Psychiatry, 49(7), 538–44. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/1352677 Farooq, S., Agid, O., Foussias, G., & Remington, G. (2013). Using treatment response to subtype schizophrenia: Proposal for a new paradigm in classification. Schizophrenia Bulletin, 39(6), 1169–1172. https://doi.org/10.1093/schbul/sbt137 Feigensona, K. A., Kusnecovb, A. W., & Silversteina, S. M. (2014). Inflammation and the Two-Hit Hypothesis of Schizophrenia Keith. Neurosci Biobehav Rev, 38, 72–93. https://doi.org/10.1016/j.neubiorev.2013.11.006.Inflammation Fernandez-Egea, E., Vértes, P. E., Flint, S. M., Turner, L., Mustafa, S., Hatton, A., … Bullmore, E. T. (2016). Peripheral immune cell populations associated with cognitive deficits and negative symptoms of treatment-resistant schizophrenia. PLoS ONE, 11(5), 1–16. https://doi.org/10.1371/journal.pone.0155631 Fleischhaker, C., Schulz, E., & Remschmidt, H. (1998). Biogenic amines as predictors of response to clozapine treatment in early-onset schizophrenia. Journal of Psychiatric Research, 32(5), 325– 333. https://doi.org/10.1016/S0022-3956(98)00012-0 Fraga, M. F., Ballestar, E., Paz, M. F., Ropero, S., Setien, F., Ballestar, M. L., … Esteller, M. (2005). From The Cover: Epigenetic differences arise during the lifetime of monozygotic twins. - 332 - REFERENCES Proceedings of the National Academy of Sciences, 102(30), 10604–10609. https://doi.org/10.1073/pnas.0500398102 Frank, J., Lang, M., Witt, S. H., Strohmaier, J., Rujescu, D., Cichon, S., … Rietschel, M. (2015). Identification of increased genetic risk scores for schizophrenia in treatment-resistant patients. Molecular Psychiatry, 20(2), 150–151. https://doi.org/10.1038/mp.2014.56 Fruchter, E., Kapara, O., Reichenberg, A., Yoffe, R., Fono-Yativ, O., Kreiss, Y., … Weiser, M. (2014). Longitudinal association between epilepsy and schizophrenia: A population-based study. Epilepsy and Behavior, 31, 291–294. https://doi.org/10.1016/j.yebeh.2013.10.026 Furukawa, S., Fujita, T., Shimabukuro, M., Iwaki, M., Yamada, Y., Nakajima, Y., … Shimomura, I. (2004). Increased oxidative stress in obesity and its impact on metabolic syndrome. The Journal of Clinical Investigation, 114(12), 1752–61. https://doi.org/10.1172/JCI21625 Gao, L., Li, Z., Chang, S., & Wang, J. (2014). Association of interleukin-10 polymorphisms with schizophrenia: A meta-analysis. PLoS ONE, 9(3), 1–7. https://doi.org/10.1371/journal.pone.0090407 Garcia, G. J., Chagas, M. H., Silva, C. H., Machado-De-Sousa, J. P., Crippa, J. A., & Hallak, J. E. (2015, March). Structural and functional neuroimaging findings associated with the use of clozapine in schizophrenia: A systematic review. Revista Brasileira de Psiquiatria. Associação Brasileira de Psiquiatria (ABP). https://doi.org/10.1590/1516-4446-2014-1387 Gaunt, T. R., Shihab, H. A., Hemani, G., Min, J. L., Woodward, G., Lyttleton, O., … Relton, C. L. (2016). Systematic identification of genetic influences on methylation across the human life course. Genome Biology, 17(1), 61. https://doi.org/10.1186/s13059-016-0926-z Gillespie, A. L., Samanaite, R., Mill, J., Egerton, A., & MacCabe, J. H. (2017). Is treatment-resistant schizophrenia categorically distinct from treatment-responsive schizophrenia? a systematic review. BMC Psychiatry, 17(1), 12. https://doi.org/10.1186/s12888-016-1177-y Goff, D. C., Cather, C., Gottlieb, J. D., Evins, A. E., Walsh, J., Raeke, L., … Green, M. F. (2008). Once- weekly d-cycloserine effects on negative symptoms and cognition in schizophrenia: An exploratory study. Schizophrenia Research, 106(2–3), 320–327. https://doi.org/10.1016/j.schres.2008.08.012 Goldstein, M. E., Anderson, V. M., Pillai, A., Kydd, R. R., & Russell, B. R. (2015). Glutamatergic neurometabolites in clozapine-responsive and -resistant schizophrenia. The International Journal of Neuropsychopharmacology / Official Scientific Journal of the Collegium Internationale Neuropsychopharmacologicum (CINP), 18(6). - 333 - REFERENCES https://doi.org/10.1093/ijnp/pyu117 Gonzalez-Burgos, G., Cho, R. Y., & Lewis, D. A. (2015). Alterations in Cortical Network Oscillations and Parvalbumin Neurons in Schizophrenia. Biological Psychiatry, 77(12), 1031–1040. https://doi.org/10.1016/j.biopsych.2015.03.010 Gordon, D. J., Probstfield, J. L., Garrison, R. J., Neaton, J. D., Castelli, W. P., Knoke, J. D., … Tyroler, H. A. (n.d.). High-Density Lipoprotein Cholesterol and Cardiovascular Disease Four Prospective American Studies. Retrieved from http://circ.ahajournals.org/content/circulationaha/79/1/8.full.pdf Gottesman, I. I., & Shields, J. (1982). Schizophrenia: The epigenetic puzzle. New York: Cambridge University Press. Green, A. I., Alam, M. Y., Sobieraj, J. T., Pappalardo, K. M., Waternaux, C., Salzman, C., … Schildkraut, J. J. (1993). Clozapine response and plasma catecholamines and their metabolites. Psychiatry Research, 46(2), 139–149. https://doi.org/10.1016/0165-1781(93)90016-A Green, A. I., Tohen, M., Patel, J. K., Banov, M., DuRand, C., Berman, I., … Schatzberg, A. F. (2000). Clozapine in the treatment of refractory psychotic mania. American Journal of Psychiatry, 157(6), 982–986. https://doi.org/10.1176/appi.ajp.157.6.982 Greenwood, C. M. T., Husted, J., Bomba, M. D., Hodgkinson, K. A., & Bassett, A. S. (2004). Elevated rates of schizophrenia in a familial sample with mental illness and intellectual disability. Journal of Intellectual Disability Research, 48(6), 531–539. https://doi.org/10.1111/j.1365- 2788.2004.00621.x Grover, S., Hazari, N., Chakrabarti, S., & Avasthi, A. (2015). Delay in initiation of clozapine: A retrospective study from a tertiary care hospital in North India. Psychiatry Research, 226(1), 181–185. https://doi.org/10.1016/j.psychres.2014.12.046 Guidotti, A., Ruzicka, W., Grayson, D. R., Veldic, M., Pinna, G., Davis, J. M., & Costa, E. (2007). S- adenosyl methionine and DNA methyltransferase-1 mRNA overexpression in psychosis. NeuroReport, 18(1), 57–60. https://doi.org/10.1097/WNR.0b013e32800fefd7 Guidotti*, J. Auta, Y. Chen, J.M. Davis, E. Dong, D.P. Gavin, D.R. Grayson, F. M. (2012). Epigenetic GABAergic Targets in Schizophrenia and Bipolar Disorder. Neuropharmacology, 100(2), 130– 134. https://doi.org/10.1016/j.pestbp.2011.02.012.Investigations Guo, J., Walss-Bass, C., & Ludueña, R. F. (2010). The β isotypes of tubulin in neuronal differentiation. Cytoskeleton, 67(7), 431–441. https://doi.org/10.1002/cm.20455 - 334 - REFERENCES Halim, N. D., Weickert, C. S., McClintock, B. W., Weinberger, D. R., & Lipska, B. K. (2004). Effects of chronic haloperidol and clozapine treatment on neurogenesis in the adult rat hippocampus. Neuropsychopharmacology, 29(6), 1063–1069. https://doi.org/10.1038/sj.npp.1300422 Hannon, E., Dempster, E., Viana, J., Burrage, J., Smith, A. R., Macdonald, R., … Shabalin, A. (2016). An integrated genetic-epigenetic analysis of schizophrenia: evidence for co-localization of genetic associations and differential DNA methylation. Genome Biology, 17(1), 176. https://doi.org/10.1186/s13059-016-1041-x Hannon, E., Lunnon, K., Schalkwyk, L., & Mill, J. (2015). Interindividual methylomic variation across blood, cortex, and cerebellum: implications for epigenetic studies of neurological and neuropsychiatric phenotypes. Epigenetics, 10(11), 1024–32. https://doi.org/10.1080/15592294.2015.1100786 Hannon, E., Spiers, H., Viana, J., Pidsley, R., Burrage, J., Murphy, T. M., … Mill, J. (2015). Methylation QTLs in the developing brain and their enrichment in schizophrenia risk loci. Nature Neuroscience, 19(1), 48–54. https://doi.org/10.1038/nn.4182 Harrison, P. J. (2004). The hippocampus in schizophrenia: a review of the neuropathological evidence and its pathophysiological implications. Psychopharmacology, 174(1), 151–162. https://doi.org/10.1007/s00213-003-1761-y Harrison, R. N. S., Murray, R. M., Lee, S. H., Paya Cano, J., Dempster, D., Curtis, C. J., … de Jong, S. (2016). Gene-expression analysis of clozapine treatment in whole blood of patients with psychosis. Psychiatric Genetics, 26(5), 211–217. https://doi.org/10.1097/YPG.0000000000000140 Hartz, S. M., Horton, A. C., Oehlert, M., Carey, C. E., Agrawal, A., Bogdan, R., … Bierut, L. J. (2017). Association Between Substance Use Disorder and Polygenic Liability to Schizophrenia. Biological Psychiatry, 1–7. https://doi.org/10.1016/j.biopsych.2017.04.020 Harvey, R. J., & Yee, B. K. (2013). Glycine transporters as novel therapeutic targets in schizophrenia, alcohol dependence and pain. Nature Reviews Drug Discovery, 12(11), 866–885. https://doi.org/10.1038/nrd3893 Haynes, W., Tomczak, A., & Khatri, P. (2017). Gene annotation bias impedes biomedical research. bioRxiv. Retrieved from http://biorxiv.org/content/early/2017/05/02/133108.abstract Healy, D. J., & Meador-Woodruff, J. H. (1997). Clozapine and haloperidol differentially affect AMPA and kainate receptor subunit mRNA levels in rat cortex and striatum. Molecular Brain Research, 47(1–2), 331–338. https://doi.org/10.1016/S0169-328X(97)00064-8 - 335 - REFERENCES Hefner, G., Shams, M. E. E., Unterecker, S., Falter, T., & Hiemke, C. (2016). Inflammation and psychotropic drugs: The relationship between C-reactive protein and antipsychotic drug levels. Psychopharmacology, 233(9), 1695–1705. https://doi.org/10.1007/s00213-015-3976-0 Hegarty, J., Baldessarini, R., Tohen, M., Waternaux, C., & Oepen, G. (1994). One Hundred Years of Schziophrenia: A Meta-Analysis of the Outcome Literature. American Journal of Psychiatry, 151(10), 1409–1416. Retrieved from http://robertwhitaker.org/robertwhitaker.org/Antipsychotic drugs and chronic illness_files/100years.pdf Heijmans, B. T., & Mill, J. (2012). Commentary: The seven plagues of epigenetic epidemiology. International Journal of Epidemiology, 41(1), 74–78. https://doi.org/10.1093/ije/dyr225 Helgason, L. (1990). Twenty years’ follow-up of first psychiatric presentation for schizophrenia: what could have been prevented? Acta Psychiatrica Scandinavica, 81(3), 231–235. https://doi.org/10.1111/j.1600-0447.1990.tb06486.x Hennekens, C. H., Hennekens, A. R., Hollar, D., & Casey, D. E. (2005). Schizophrenia and increased risks of cardiovascular disease. American Heart Journal, 150(6), 1115–1121. https://doi.org/10.1016/j.ahj.2005.02.007 Heresco-Levy, U., Javitt, D. C., Ebstein, R., Vass, A., Lichtenberg, P., Bar, G., … Ermilov, M. (2005). D- serine efficacy as add-on pharmacotherapy to risperidone and olanzapine for treatment- refractory schizophrenia. Biological Psychiatry, 57(6), 577–585. https://doi.org/10.1016/j.biopsych.2004.12.037 Hertzberg, L., Katsel, P., Roussos, P., Haroutunian, V., & Domany, E. (2015). Integration of gene expression and GWAS results supports involvement of calcium signaling in Schizophrenia. Schizophrenia Research, 164(1), 92–99. https://doi.org/10.1016/j.schres.2015.02.001 Hippius, H. (1989). The history of clozapine. Psychopharmacology, 99(1 Supplement), S3-5. https://doi.org/10.1007/BF00442551 Hodge, K., & Jespersen, S. (2008). Side-effects and treatment with clozapine: A comparison between the views of consumers and their clinicians. International Journal of Mental Health Nursing, 17(1), 2–8. https://doi.org/10.1111/j.1447-0349.2007.00506.x Horvath, S. (2013). DNA methylation age of human tissues and cell types. Genome Biology, 14(10), R115. https://doi.org/10.1186/gb-2013-14-10-r115 Howard, L. M., Barley, E. A., Davies, E., Rigg, A., Lempp, H., Rose, D., … Thornicroft, G. (2010). Cancer - 336 - REFERENCES diagnosis in people with severe mental illness: Practical and ethical issues. The Lancet Oncology. https://doi.org/10.1016/S1470-2045(10)70085-1 Howes, O. D., McCutcheon, R., Agid, O., de Bartolomeis, A., van Beveren, N. J., Birnbaum, M. L., … Correll, C. U. (2016). Treatment-Resistant Schizophrenia: Treatment Response and Resistance in Psychosis (TRRIP) Working Group Consensus Guidelines on Diagnosis and Terminology. American Journal of Psychiatry, 1(9), appi.ajp.2016.1. https://doi.org/10.1176/appi.ajp.2016.16050503 Howes, O., McCutcheon, R., & Stone, J. (2015). Glutamate and dopamine in schizophrenia: An update for the 21 st century. Journal of Psychopharmacology, 29(2), 97–115. https://doi.org/10.1177/0269881114563634 Hu, X., Zhou, H., Zhang, D., Yang, S., Qian, L., Wu, H., … Hong, J.-S. (2012). Clozapine protects dopaminergic neurons from inflammation- induced damage by inhibiting microglial overactivation. Journal of Neuroimmune Pharmacology, 7(1), 187–201. https://doi.org/10.1007/s11481-011-9309-0.Clozapine Huang, H.-S., Matevossian, A., Whittle, C., Kim, S. Y., Schumacher, A., Baker, S. P., & Akbarian, S. (2007). Prefrontal Dysfunction in Schizophrenia Involves Mixed-Lineage Leukemia 1-Regulated Histone Methylation at GABAergic Gene Promoters. Journal of Neuroscience, 27(42), 11254– 11262. https://doi.org/10.1523/JNEUROSCI.3272-07.2007 Hwang, R., Shinkai, T., De Luca, V., Müller, D. J., Ni, X., Macciardi, F., … Kennedy, J. L. (2005). Association study of 12 polymorphisms spanning the dopamine D2 receptor gene and clozapine treatment response in two treatment refractory/intolerant populations. Psychopharmacology, 181(1), 179–187. https://doi.org/10.1007/s00213-005-2223-5 Hwang, R., Shinkai, T., De Luca, V., Ni, X., Potkin, S. G., Lieberman, J. a, … Kennedy, J. L. (2007). Association study of four dopamine D1 receptor gene polymorphisms and clozapine treatment response. Journal of Psychopharmacology (Oxford, England), 21, 718–727. https://doi.org/10.1177/0269881106072341 Hwang, R., Shinkai, T., Deluca, V., Macciardi, F., Potkin, S., Meltzer, H. Y., & Kennedy, J. L. (2006). Dopamine D2 receptor gene variants and quantitative measures of positive and negative symptom response following clozapine treatment. European Neuropsychopharmacology, 16(4), 248–259. https://doi.org/10.1016/j.euroneuro.2005.09.004 Hwang, R., Souza, R. P., Tiwari, A. K., Zai, C. C., Müller, D. J., Potkin, S. G., … Kennedy, J. L. (2011). Gene-gene interaction analyses between NMDA receptor subunit and dopamine receptor gene - 337 - REFERENCES variants and clozapine response. Pharmacogenomics. https://doi.org/10.2217/pgs.10.182 Hwang, R., Zai, C., Tiwari, a, Müller, D. J., Arranz, M. J., Morris, a G., … Kennedy, J. L. (2010). Effect of dopamine D3 receptor gene polymorphisms and clozapine treatment response: exploratory analysis of nine polymorphisms and meta-analysis of the Ser9Gly variant. The Pharmacogenomics Journal, 10(November 2009), 200–218. https://doi.org/10.1038/tpj.2009.65 Hyde, N., Dodd, S., Venugopal, K., Purdie, C., Berk, M., & O’Neil, A. (2015). Prevalence of cardiovascular and metabolic events in patients prescribed clozapine: a retrospective observational, clinical cohort study. Current Drug Safety, 10(2), 125–31. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/25986036 Ibi, D., & González-Maeso, J. (2015). Epigenetic signaling in schizophrenia. Cellular Signalling, 27(10), 2131–6. https://doi.org/10.1016/j.cellsig.2015.06.003 Inestrosa, N. C., Montecinos-Oliva, C., & Fuenzalida, M. (2012). Wnt signaling: Role in Alzheimer disease and schizophrenia. Journal of Neuroimmune Pharmacology, 7(4), 788–807. https://doi.org/10.1007/s11481-012-9417-5 IPAP. (2004). IPAP Schizophrenia Algorithm. Ishikawa, K., Imai, Y., Katsuya, T., Ohkubo, T., Tsuji, I., Nagai, K., … Ogihara, T. (2000). Human G- protein beta3 subunit variant is associated with serum potassium and total cholesterol levels but not with blood pressure. Am J Hypertens, 13(2), 140–145. Retrieved from http://www.ncbi.nlm.nih.gov/htbin- post/Entrez/query?db=m&form=6&dopt=r&uid=0010701813 Ishimoto, T., Ninomiya, K., Inoue, R., Koike, M., Uchiyama, Y., & Mori, H. (2017). Mice lacking BCAS1, a novel myelin-associated protein, display hypomyelination, schizophrenia-like abnormal behaviors, and upregulation of inflammatory genes in the brain. GLIA, 65(5), 727–739. https://doi.org/10.1002/glia.23129 Jaehne, E. J., Ramshaw, H., Xu, X., Saleh, E., Clark, S. R., Schubert, K. O., … Baune, B. T. (2015). In-vivo administration of clozapine affects behaviour but does not reverse dendritic spine deficits in the 14-3-3ζ KO mouse model of schizophrenia-like disorders. Pharmacology Biochemistry and Behavior, 138, 1–8. https://doi.org/10.1016/j.pbb.2015.09.006 Jaffe, A. E., & Irizarry, R. A. (2014). Accounting for cellular heterogeneity is critical in epigenome- wide association studies. Genome Biology, 15(2), R31. https://doi.org/10.1186/gb-2014-15-2- r31 - 338 - REFERENCES Jayakumar, P. N., Gangadhar, B. N., Venkatasubramanian, G., Desai, S., Velayudhan, L., Subbakrishna, D., & Keshavan, M. S. (2010). High energy phosphate abnormalities normalize after antipsychotic treatment in schizophrenia: A longitudinal 31P MRS study of basal ganglia. Psychiatry Research - Neuroimaging, 181(3), 237–240. https://doi.org/10.1016/j.pscychresns.2009.10.010 Jentsch, J. D., & Roth, R. H. (1999). The neuropsychopharmacology of phencyclidine: From NMDA receptor hypofunction to the dopamine hypothesis of schizophrenia. Neuropsychopharmacology, 20(3), 201–225. https://doi.org/10.1016/S0893-133X(98)00060-8 Jin, B., Li, Y., & Robertson, K. D. (2011). DNA methylation: superior or subordinate in the epigenetic hierarchy? Genes & Cancer, 2(6), 607–17. https://doi.org/10.1177/1947601910393957 Johnson, K. C., Houseman, E. A., King, J. E., von Herrmann, K. M., Fadul, C. E., & Christensen, B. C. (2016). 5-Hydroxymethylcytosine localizes to enhancer elements and is associated with survival in glioblastoma patients. Nature Communications, 7, 13177. https://doi.org/10.1038/ncomms13177 Joint Formulary Committee. (2017). Retrieved May 17, 2017, from https://www.evidence.nhs.uk/formulary/bnf/current/4-central-nervous-system/42-drugs- used-in-psychoses-and-related-disorders/421-antipsychotic-drugs/second-generation- antipsychotic-drugs/clozapine Jones, P., & Murray, R. M. (1991). The genetics of schizophrenia is the genetics of neurodevelopment. British Journal of Psychiatry, 158(MAY), 615–623. https://doi.org/10.1192/bjp.158.5.615 Juarez-Reyes, M. G., Shumway, M., Battle, C., Bacchetti, P., Hansen, M. S., & Hargreaves, W. A. (1995). Effects of stringent criteria on eligibility for clozapine among public mental health clients. Psychiatric Services, 46(8), 801–806. https://doi.org/10.1176/ps.46.8.801 Kane, J., Honigfeld, G., Singer, J., & Meltzer, H. (1988). Clozapine for the treatment-resistant schizophrenic: A double-blind comparison with chlorpromazine. Archives of General Psychiatry. https://doi.org/10.1001/archpsyc.1988.01800330013001 Kapur, S., Zipursky, R. B., & Remington, G. (1999). Clinical and theoretical implications of 5-HT2 and D2 receptor occupancy of clozapine, risperidone, and olanzapine in schizophrenia. American Journal of Psychiatry, 156(2), 286–293. https://doi.org/10.1176/ajp.156.2.286 Kay, S. R., Fiszbein, A., & Opler, L. A. (1987). The positive and negative syndrome scale (PANSS) for schizophrenia. Schizophrenia Bulletin, 13(2), 261–76. - 339 - REFERENCES https://doi.org/10.1093/SCHBUL/13.2.261 Kedracka-Krok, S., Swiderska, B., Jankowska, U., Skupien-Rabian, B., Solich, J., Buczak, K., & Dziedzicka-Wasylewska, M. (2015). Clozapine influences cytoskeleton structure and calcium homeostasis in rat cerebral cortex and has a different proteomic profile than risperidone. Journal of Neurochemistry, 132(6), 657–676. https://doi.org/10.1111/jnc.13007 Kelly, A. C., Sheitman, B. B., Hamer, R. M., Rhyne, D. C., Reed, R. M., Graham, K. A., … Jarskog, L. F. (2014). A Naturalistic Comparison of the Long-Term Metabolic Adverse Effects of Clozapine Versus Other Antipsychotics for Patients With Psychotic Illnesses. Journal of Clinical Psychopharmacology, 34(4), 441–445. https://doi.org/10.1097/JCP.0000000000000159 Kelly, D. L., Gale, E. A., & Conley, R. R. (2003). Clozapine Treatment in Patients with Prior Substance Abuse. The Canadian Journal of Psychiatry, 48(2), 111–114. Kennedy, J. L., Altar, C. A., Taylor, D. L., Degtiar, I., & Hornberger, J. C. (2014). The social and economic burden of treatment-resistant schizophrenia. International Clinical Psychopharmacology, 29(2), 63–76. https://doi.org/10.1097/YIC.0b013e32836508e6 Kessler, R. M., Ansari, M. S., Riccardi, P., Li, R., Jayathilake, K., Dawant, B., & Meltzer, H. Y. (2006). Occupancy of striatal and extrastriatal dopamine D2 receptors by clozapine and quetiapine. Neuropsychopharmacology : Official Publication of the American College of Neuropsychopharmacology, 31(9), 1991–2001. https://doi.org/10.1038/sj.npp.1301108 Khatri, P., Sirota, M., Butte, A. J., Kim, S., & Woo, H. (2012). Ten Years of Pathway Analysis: Current Approaches and Outstanding Challenges. PLoS Computational Biology, 8(2), e1002375. https://doi.org/10.1371/journal.pcbi.1002375 Khojasteh-Fard, M., Tabrizi, M., & Amoli, M. M. (2011). Is DNA methylation responsible for immune system dysfunction in schizophrenia? Medical Hypotheses, 77(4), 573–579. https://doi.org/10.1016/j.mehy.2011.06.034 Kim, J. Y., Liu, C. Y., Zhang, F., Duan, X., Wen, Z., Song, J., … Ming, G. (2012). Interplay between DISC1 and GABA Signaling Regulates Neurogenesis in Mice and Risk for Schizophrenia. Cell, 148(5), 1051–64. https://doi.org/10.1016/j.cell.2011.12.037 Kinon, B., Kane, J., Celeste, J., Perovich, R., Ismi, M., Koreen, A., & Weiden, P. (1993). Treatment of neuroleptic-resistance schizophrenic relapse. Psychopharmacology Bulletin, 29(2), 309–314. Retrieved from http://psycnet.apa.org/record/1994-14648-001 Kinoshita, M., Numata, S., Tajima, A., Yamamori, H., Yasuda, Y., Fujimoto, M., … Ohmori, T. (2017). - 340 - REFERENCES Effect of Clozapine on DNA Methylation in Peripheral Leukocytes from Patients with Treatment-Resistant Schizophrenia. International Journal of Molecular Sciences, 18(3). https://doi.org/10.3390/ijms18030632 Kirwan, P., O’Connor, L., Sharma, K., & McDonald, C. (2017). The impact of switching to clozapine on psychiatric hospital admissions: a mirror-image study. Irish Journal of Psychological Medicine, 1–5. https://doi.org/10.1017/ipm.2017.28 Klein, D., & Davis JM. (1969). Review of antipsychotic drug literature. In Diagnosis and Drug Treatment of Psychiatric Disorders (pp. 52–138). Baltimore: Williams and Wilkins. Retrieved from https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=klein+and+davis+1969+antipsych otic&btnG= Klose, R. J., & Bird, A. P. (2006). Genomic DNA methylation: the mark and its mediators. Trends in Biochemical Sciences, 31(2), 89–97. https://doi.org/10.1016/j.tibs.2005.12.008 Kohlrausch, F. B., Salatino-Oliveira, A., Gama, C. S., Lobato, M. I., Belmonte-de-Abreu, P., & Hutz, M. H. (2008). G-protein gene 825C>T polymorphism is associated with response to clozapine in Brazilian schizophrenics. Pharmacogenomics., 9(10), 1429–1436. https://doi.org/10.2217/14622416.9.10.1429 [doi] Kohlrausch, F. B., Salatino-Oliveira, A., Gama, C. S., Lobato, M. I., Belmonte-de-Abreu, P., & Hutz, M. H. (2010). Influence of serotonin transporter gene polymorphisms on clozapine response in Brazilian schizophrenics. Journal of Psychiatric Research, 44(16), 1158–1162. https://doi.org/10.1016/j.jpsychires.2010.04.003 Konopaske, G., Lange, N., Coyle, J. T., & FM, B. (2014). Prefrontal cortical dendritic spine pathology in schizophrenia and bipolar disorder. JAMA Psychiatry, 71(12), 1323–1331. https://doi.org/10.1002/wrna.1178.Alternative Kontkanen, O., & To, P. (2002). Antipsychotic drug treatment induces differential gene expression in the rat cortex. Journal of Neurochemistry, 83, 1043–1053. Kracmarova, A., & Pohanka, M. (2014). The impact of clozapine on regulation of inflammation in murine macrophage cells. Neuro Endocrinology Letters, 35(Suppl 2), 175–179. Kraepelin, E. (1899). Psychiatrie: Ein Lehrbuch fur Studirende und Aerzte. Sechste, vollstandig umgearbeitete Auflage. Leipzig: Verlag von Johann Ambrosius Barth. Krystal Karper, LP., Seibyl, JP., Freeman, GK., Delaney, R., Bremner, JD., Heninger, GR., Bowers, MB - 341 - REFERENCES Jr., Charney, DS., J. H. (1994). Subanesthetic effects of the NMDA antagonist, ketamine, in humans: psychotomimetic, perceptual, cognitive, and neuroendocrine effects. Arch Gen Psychiatry, 51(3), 199–214. https://doi.org/10.1017/CBO9781107415324.004 Kuznetsova, A., Brockhoff, P., & Christensen, R. (2015). Package "lmer test" Retrieved from https://scholar.google.co.uk/citations?user=q8XKpz0AAAAJ&hl=en Lamberti, J. S., Bellnier, T., & Schwarzkopf, S. B. (1992). Weight gain among schizophrenic patients treated with clozapine. The American Journal of Psychiatry, 149(5), 689–690. https://doi.org/10.1176/ajp.149.5.689 Landek-Salgado, M. A., Faust, T. E., & Sawa, A. (2016). Molecular Substrates of Schizophrenia: Homeostatic Signaling to Connectivity. Molecular Psychiatry, 21(1), 10–28. https://doi.org/10.1177/0963721414541462.Self-Control Langosch, J. M., & Trimble, M. R. (2002). Epilepsy, psychosis and clozapine. Human Psychopharmacology, 17(2), 115–119. https://doi.org/10.1002/hup.375 Lappalainen, T., & Greally, J. M. (2017). Associating cellular epigenetic models with human phenotypes. Nature Reviews Genetics, 18(7), 441–451. https://doi.org/10.1038/nrg.2017.32 Lee, B. J., Marchionni, L., Andrews, C. E., Norris, A. L., Nucifora, L. G., Wu, Y. C., … Nucifora, F. C. (2017). Analysis of differential gene expression mediated by clozapine in human postmortem brains. Schizophrenia Research, 185, 58–66. https://doi.org/10.1016/j.schres.2016.12.017 Leek, J. T., Scharpf, R. B., Bravo, H. C., Simcha, D., Langmead, B., Johnson, W. E., … Irizarry, R. A. (2010). Tackling the widespread and critical impact of batch effects in high-throughput data. Nature Reviews. Genetics, 11(10), 733–9. https://doi.org/10.1038/nrg2825 Lehman, A. F., Lieberman, J. A., Dixon, L. B., McGlashan, T. H., Miller, A. L., Perkins, D. O., … Regier, D. (2004). Practice guideline for the treatment of patients with schizophrenia, second edition. The American Journal of Psychiatry, 161(2 Suppl), 1–56. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/15000267 Lehne, B., Drong, A. W., Loh, M., Zhang, W., Scott, W. R., Tan, S.-T., … Chambers, J. C. (2015). A coherent approach for analysis of the Illumina HumanMethylation450 BeadChip improves data quality and performance in epigenome-wide association studies. Genome Biology, 16(1), 37. https://doi.org/10.1186/s13059-015-0600-x Leucht, S., Burkard, T., Henderson, J., Maj, M., Sartorius, N., & Leucht, S. (2007). Physical illness and schizophrenia: a review of the literature. Acta Psychiatrica Scandinavica, 116, 317–333. - 342 - REFERENCES https://doi.org/10.1111/j.1600-0447.2007.01095.x Leucht, S., Cipriani, A., Spineli, L., Mavridis, D., Örey, D., Richter, F., … Davis, J. M. (2013). Comparative efficacy and tolerability of 15 antipsychotic drugs in schizophrenia: A multiple- treatments meta-analysis. The Lancet, 382(9896), 951–962. https://doi.org/10.1016/S0140- 6736(13)60733-3 Lewis, D. A., & Sweet, R. A. (2009). Schizophrenia from a neural circuitry perspective: Advancing toward rational pharmacological therapies. Journal of Clinical Investigation, 119(4), 706–716. https://doi.org/10.1172/JCI37335 Leza, J. C., Bueno, B., Bioque, M., Arango, C., Parellada, M., Do, K., … Bernardo, M. (2015). Inflammation in schizophrenia: A question of balance. Neuroscience and Biobehavioral Reviews, 55, 612–626. https://doi.org/10.1016/j.neubiorev.2015.05.014 Li, E. (2002). Chromatin modification and epigenetic reprogramming in mammalian development. Nat Rev Genet, 3(9), 662–73. https://doi.org/10.1038/nrg887 Liang, Y., Su, Y.-A., Zhao, Z.-G., Gao, N., Huang, J.-Z., Tang, M.-Q., … X., Y. (2015). Acute Effects of Haloperidol, Amisulpride, and Quetiapine on Bone Turnover Markers in Patients with Schizophrenia. Journal of Clinical Psychopharmacology, 35(5), 583–586. https://doi.org/10.1097/JCP.0000000000000379 Liao, Q., Wang, Y., Cheng, J., Dai, D., Zhou, X., Zhang, Y., … Duan, S. (2014). DNA methylation patterns of protein coding genes and long noncoding RNAs in female schizophrenic patients. European Journal of Medical Genetics, (December), 1–10. https://doi.org/10.1016/j.ejmg.2014.12.001 Lichtermann, D., Ekelund, J., Pukkala, E., Tanskanen, A., & Lönnqvist, J. (2001). Incidence of cancer among persons with schizophrenia and their relatives. Archives of General Psychiatry, 58(6), 573–578. https://doi.org/10.1001/archpsyc.58.6.573 Lidow, M. S. (2003). Calcium signaling dysfunction in schizophrenia: a unifying approach. Brain Research Reviews, 43, 70–84. https://doi.org/10.1016/S0165-0173(03)00203-0 Lieberman, J. a, Safferman, a Z., Pollack, S., Szymanski, S., Johns, C., Howard, a, … Kane, J. M. (1994). Clinical effects of clozapine in chronic schizophrenia: response to treatment and predictors of outcome. The American Journal of Psychiatry, 151(January 1995), 1744–1752. https://doi.org/10.1176/ajp.151.12.1744 Lin, C. H., Tsai, S. J., Yu, Y. W., Song, H. L., Tu, P. C., Sim, C. B., … Hong, C. J. (1999). No evidence for association of serotonin-2A receptor variant (102T/C) with schizophrenia or clozapine response - 343 - REFERENCES in a Chinese population. Neuroreport, 10(1), 57–60. https://doi.org/10.1097/00001756- 199901180-00011 Liou, Y.-J., Wang, H.-H., Lee, M.-T. M., Wang, S.-C., Chiang, H.-L., Chen, C.-C., … Wu, J.-Y. (2012). Genome-Wide Association Study of Treatment Refractory Schizophrenia in Han Chinese. PLoS ONE, 7(3), e33598. https://doi.org/10.1371/journal.pone.0033598 Lipina, T., Labrie, V., Weiner, I., & Roder, J. (2005). Modulators of the glycine site on NMDA receptors, d-serine and ALX 5407, display similar beneficial effects to clozapine in mouse models of schizophrenia. Psychopharmacology, 179(1), 54–67. https://doi.org/10.1007/s00213-005-2210-x Lister, R., & Ecker, J. R. (2009). Finding the fifth base: Genome-wide sequencing of cytosine methylation. Genome Research. https://doi.org/10.1101/gr.083451.108 Lott, S. A., Burghardt, P. R., Burghardt, K. J., Bly, M. J., Grove, T. B., & Ellingrod, V. L. (2013). The influence of metabolic syndrome, physical activity and genotype on catechol-O-methyl transferase promoter-region methylation in schizophrenia. The Pharmacogenomics Journal, 13(3), 264–271. https://doi.org/10.1038/tpj.2012.6 Maas, D. A., Vallès, A., & Martens, G. J. M. (2017). Oxidative stress, prefrontal cortex hypomyelination and cognitive symptoms in schizophrenia. Translational Psychiatry, 7(7), e1171. https://doi.org/10.1038/tp.2017.138 Machielsen, M. W. J., & de Haan, L. (2009). Differences in efficacy on substance abuse between risperidone and clozapine supports the importance of differential modulation of dopaminergic neurotransmission. Psychopharmacology Bulletin, 42(4), 40–52. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/20581792 Malhotra, A., Goldman, D., Buchanan, R., Rooney, W., Clifton, A., Kosmidis, M. H., … Pickar, D. (1998). The dopamine D3 receptor (DRD3) Ser9Gly polymorphism and schizophrenia: a haplotype relative risk study and association with clozapine response. Molecular Psychiatry, 3(1), 72–5. https://doi.org/10.1038/sj.mp.4000288 Malhotra, A. K., Goldman, D., Ozaki, N., Breier, A., Buchanan, R., & Pickar, D. (1996). Lack of association between polymorphisms in the 5-HT(2A) receptor gene and the antipsychotic response to clozapine. American Journal of Psychiatry, 153(8), 1092–1094. https://doi.org/10.1176/ajp.153.8.1092 Maric, N. P., & Svrakic, D. M. (2012). Why schizophrenia genetics needs epigenetics: A review. Psychiatria Danubina, 24(1), 2–18. - 344 - REFERENCES Martin, A. K., & Mowry, B. (2016). Increased rare duplication burden genomewide in patients with treatment-resistant schizophrenia. Psychological Medicine, 46(3), 469–476. https://doi.org/10.1017/S0033291715001701 Maschietto, M., Tahira, A. C., Puga, R., Lima, L., Mariani, D., da Silveira Paulsen, B., … Brentani, H. (2015). Co-expression network of neural-differentiation genes shows specific pattern in schizophrenia. BMC Medical Genomics, 8(1), 23. https://doi.org/10.1186/s12920-015-0098-9 Masellis, M., Basile, V., Meltzer, H. Y., Lieberman, J. a, Sevy, S., Macciardi, F. M., … Kennedy, J. L. (1998). Serotonin subtype 2 receptor genes and clinical response to clozapine in schizophrenia patients. Neuropsychopharmacology : Official Publication of the American College of Neuropsychopharmacology, 19(98), 123–132. https://doi.org/10.1016/S0893-133X(98)00007-4 Mccarthy, S. E., Gillis, J., Kramer, M., Lihm, J., Yoon, S., Mistry, M., … Corvin, A. (2014). De novo Mutations in Schizophrenia Implicate Chromatin Remodeling and Support a Genetic Overlap with Autism and Intellectual Disability. Mol Psychiatry, 19(6), 652–658. https://doi.org/10.1038/mp.2014.29.De McClay, J. L., Adkins, D. E., Aberg, K., Stroup, S., Perkins, D. O., Vladimirov, V. I., … van den Oord, E. J. C. G. (2011). Genome-wide pharmacogenomic analysis of response to treatment with antipsychotics. Molecular Psychiatry, 16(1), 76–85. https://doi.org/10.1038/mp.2009.89 McEvoy, J. P., Lieberman, J. A., Stroup, T. S., Davis, S. M., Meltzer, H. Y., Rosenheck, R. A., … Hsiao, J. K. (2006). Effectiveness of clozapine versus olanzapine, quetiapine, and risperidone in patients with chronic schizophrenia who did not respond to prior atypical antipsychotic treatment. American Journal of Psychiatry, 163(4), 600–610. https://doi.org/10.1176/appi.ajp.163.4.600 McRae, A. F., Powell, J. E., Henders, A. K., Bowdler, L., Hemani, G., Shah, S., … Montgomery, G. W. (2014). Contribution of genetic variation to transgenerational inheritance of DNA methylation. Genome Biology, 15(5), R73. https://doi.org/10.1186/gb-2014-15-5-r73 Meador-Woodruff, J. H., King, R. E., Damask, S. P., & Bovenkerk, K. A. (1996). Differential regulation of hippocampal AMPA and kainate receptor subunit expression by haloperidol and clozapine. Molecular Psychiatry, 1(1), 41–53. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/9118313 Meltzer, H. Y. (1990). Commentary: Defining Treatment Refractoriness in Schizophrenia. Schizophrenia Bulletin, 16(4), 563–565. Retrieved from https://watermark.silverchair.com/api/watermark?token=AQECAHi208BE49Ooan9kkhW_Ercy7 Dm3ZL_9Cf3qfKAc485ysgAAAf0wggH5BgkqhkiG9w0BBwagggHqMIIB5gIBADCCAd8GCSqGSIb3D - 345 - REFERENCES QEHATAeBglghkgBZQMEAS4wEQQMgBPJIB0O_Zd6YnzEAgEQgIIBsCAZl-c_ux5crpu- gIbXQDakfQyOlC-6pk-9Wdu9WCq7r Meltzer, H. Y. (1997). Treatment-resistant schizophrenia--the role of clozapine. Current Medical Research and Opinion, 14(1), 1–20. https://doi.org/10.1185/03007999709113338 Meltzer, H. Y. (2012). Clozapine. Clinical Schizophrenia & Related Psychoses, 6(3), 134–144. https://doi.org/10.3371/CSRP.6.3.5 Meltzer, H. Y. (2013). Update on Typical and Atypical Antipsychotic Drugs. Annual Review of Medicine, 64(1), 393–406. https://doi.org/10.1146/annurev-med-050911-161504 Meltzer, H. Y., Bastani, B., Kwon, K. Y., Ramirez, L. F., Burnett, S., & Sharpe, J. (1989). A prospective study of clozapine in treatment-resistant schizophrenic patients I. Preliminary report. Psychopharmacology, 99, S68-72. Retrieved from https://link.springer.com/content/pdf/10.1007/BF00442563.pdf Meltzer, H. Y., & Gudelsky, G. A. (1992). Dopaminergic and serotonergic effects of clozapine. Implications for a unique clinical profile. Arzneimittel-Forschung, 42(2A), 268–72. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/1586397 Meltzer, H. Y., Thompson, P. A., Lee, M. A., & Ranjan, R. (1996). Neuropsychological deficits in schizophrenia Relation to Social Function and Effect of Antipsychotic Drug Treatment. Neuropsychopharmacology, 14, S27-33. Retrieved from https://www.nature.com/npp/journal/v14/n1s/pdf/1380438a.pdf Mesholam-Gately, R. I., Gibson, L. E., Seidman, L. J., & Green, A. I. (2014). Schizophrenia and co- occurring substance use disorder: Reward, olfaction and clozapine. Schizophrenia Research, 155(1–3), 45–51. https://doi.org/10.1016/j.schres.2014.03.002 Mill, J., Dempster, E., Caspi, A., Williams, B., Moffitt, T., & Craig, I. (2006). Evidence for monozygotic twin (MZ) discordance in methylation level at two CpG sites in the promoter region of the catechol-O-methyltransferase (COMT) gene. American Journal of Medical Genetics, Part B: Neuropsychiatric Genetics, 141(4), 421–425. https://doi.org/10.1002/ajmg.b.30316 Mill, J., Tang, T., Kaminsky, Z., Khare, T., Yazdanpanah, S., Bouchard, L., … Petronis, A. (2008). Epigenomic Profiling Reveals DNA-Methylation Changes Associated with Major Psychosis. American Journal of Human Genetics, 82, 696–711. https://doi.org/10.1016/j.ajhg.2008.01.008 Miller, D. D. (2004). Atypical antipsychotics: sleep, sedation, and efficacy. Primary Care Companion to the Journal of Clinical Psychiatry, 6(Suppl 2), 3–7. Retrieved from - 346 - REFERENCES http://www.ncbi.nlm.nih.gov/pubmed/16001094 Milner, J. J., & Beck, M. A. (2012). The impact of obesity on the immune response to infection. The Proceedings of the Nutrition Society, 71(2), 298–306. https://doi.org/10.1017/S0029665112000158 Mistry, H., & Osborn, D. (2011). Underuse of clozapine in treatment-resistant schizophrenia. Advances in Psychiatric Treatment, 17(4), 250–255. https://doi.org/10.1192/apt.bp.110.008128 Mitjans, M., Catalán, R., Vázquez, M., González-Rodríguez, A., Penadés, R., Pons, A., … Arias, B. (2015). Hypothalamic-pituitary-adrenal system, neurotrophic factors and clozapine response: Association with FKBP5 and NTRK2 genes. Pharmacogenetics and Genomics, 25(5), 274–277. https://doi.org/10.1097/FPC.0000000000000132 Moehle, M. S., Luduena, R. F., Haroutunian, V., Meador-Woodruff, J. H., & McCullumsmith, R. E. (2012). Regional differences in expression of β-tubulin isoforms in schizophrenia. Schizophrenia Research, 135(1–3), 181–6. https://doi.org/10.1016/j.schres.2011.12.010 Molina, V., Taboada, D., Aragüés, M., Hernández, J. A., & Sanz-Fuentenebro, J. (2014). Greater clinical and cognitive improvement with clozapine and risperidone associated with a thinner cortex at baseline in first-episode schizophrenia. Schizophrenia Research, 158(1–3), 223–229. https://doi.org/10.1016/j.schres.2014.06.042 Möller, M., Du Preez, J. L., Viljoen, F. P., Berk, M., Emsley, R., & Harvey, B. H. (2013). Social isolation rearing induces mitochondrial, immunological, neurochemical and behavioural deficits in rats, and is reversed by clozapine or N-acetyl cysteine. Brain, Behavior, and Immunity, 30, 156–167. https://doi.org/10.1016/j.bbi.2012.12.011 Morgan, C., & Fisher, H. (2006). Environment and Schizophrenia: Environmental Factors in Schizophrenia: Childhood Trauma--A Critical Review. Schizophrenia Bulletin, 33(1), 3–10. https://doi.org/10.1093/schbul/sbl053 MORGAN, V. A., & Al, E. (2008). Intellectual disability co-occurring with schizophrenia and other psychiatric illness: population-based study. British Journal of Psychiatry, 193(11), 364–372. https://doi.org/10.1192/bjp.bp.107.044461 Mortensen, P. B. (1994). The Occurrence of Cancer in First Admitted Schizophrenic-Patients. Schizophrenia Research, 12(3), 185–194. https://doi.org/10.1016/0920-9964(94)90028-0 Moyer, C., Shelton, M., & Sweet, R. A. (2015). Dendritic Spine Alterations in Schizophrenia. Neuroscience Letters, 601, 46–53. https://doi.org/10.1177/0963721414541462.Self-Control - 347 - REFERENCES Müller, D. J., De Luca, V., Sicard, T., King, N., Hwang, R., Volavka, J., … Kennedy, J. L. (2005). Suggestive association between the C825T polymorphism of the G-protein ??3 subunit gene (GNB3) and clinical improvement with antipsychotics in schizophrenia. European Neuropsychopharmacology, 15(5), 525–531. https://doi.org/10.1016/j.euroneuro.2005.02.001 Müller, N., Weidinger, E., Leitner, B., & Schwarz, M. J. (2015). The role of inflammation in schizophrenia. Frontiers in Neuroscience, 9, 1–9. https://doi.org/10.3389/fnins.2015.00372 Murray, C. J. L., & Lopez, A. D. (1996). The global burden of disease: a comprehensive assessment of mortality and disability from deceases, injuries and risk factors in 1990 and projected to 2010. Harvard University Press, 1, 1–35. https://doi.org/10.1186/1471-2458-13-863 Nasrallah, H. A., & Newcomer, J. W. (2004). Atypical Antipsychotics and Metabolic Dysregulation. Journal of Clinical Psychopharmacology, 24, S7–S14. https://doi.org/10.1097/01.jcp.0000142282.62336.e9 Neale, J. H., Olszewski, R. T., Gehl, L. M., Wroblewska, B., & Bzdega, T. (2005). The neurotransmitter N-acetylaspartylglutamate in models of pain, ALS, diabetic neuropathy, CNS injury and schizophrenia. Trends in Pharmacological Sciences. https://doi.org/10.1016/j.tips.2005.07.004 Need, A. C., Keefe, R. S. E., Ge, D., Grossman, I., Dickson, S., McEvoy, J. P., & Goldstein, D. B. (2009). Pharmacogenetics of antipsychotic response in the CATIE trial: a candidate gene analysis. European Journal of Human Genetics : EJHG, 17(7), 946–57. https://doi.org/10.1038/ejhg.2008.264 NICE. (2009). Psychosis and schizophrenia in adults: prevention and management | Guidance and guidelines | NICE. NICE. Retrieved from https://www.nice.org.uk/guidance/cg178/chapter/1- Recommendations Nielsen, S. M., Toftdahl, N. G., Nordentoft, M., & Hjorthøj, C. (2017). Association between alcohol, cannabis, and other illicit substance abuse and risk of developing schizophrenia: a nationwide population based register study. Psychological Medicine, 47(9), 1668–1677. https://doi.org/10.1017/S0033291717000162 Niwa, M., Jaaro-Peled, H., Tankou, S., Seshadri, S., Hikida, T., Matsumoto, Y., … Sawa, A. (2013). Adolescent stress-induced epigenetic control of dopaminergic neurons via glucocorticoids. Science (New York, N.Y.), 339(6117), 335–9. https://doi.org/10.1126/science.1226931 Nöthen, M. M., Rietschel, M., Erdmann, J., Oberländer, H., Möller, H. J., Nober, D., & Propping, P. (1995). Genetic variation of the 5-HT2A receptor and response to clozapine. Lancet (London, England), 346(8979), 908–9. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/7564704 - 348 - REFERENCES Ohno-Shosaku, T., Sugawara, Y., Muranishi, C., Nagasawa, K., Kubono, K., Aoki, N., … Yoneda, M. (2011). Effects of clozapine and N-desmethylclozapine on synaptic transmission at hippocampal inhibitory and excitatory synapses. Brain Research, 1421, 66–77. https://doi.org/10.1016/j.brainres.2011.08.073 Okruszek, L., Bala, A., Wordecha, M., Jarkiewicz, M., Wysokinski, A., Szczepocka, E., … Marchel, A. (2017). Social cognition in neuropsychiatric populations: a comparison of theory of mind in schizophrenia and mesial temporal lobe epilepsy. Scientific Reports, 7(1), 484. https://doi.org/10.1038/s41598-017-00565-2 Olianas, M. C., Maullu, C., & Onali, P. (1999). Mixed agonist-antagonist properties of clozapine at different human cloned muscarinic receptor subtypes expressed in Chinese hamster ovary cells. Neuropsychopharmacology, 20(3), 263–270. https://doi.org/10.1016/S0893- 133X(98)00048-7 Olney, J. W., Newcomer, J. W., & Farber, N. B. (1999, November). NMDA receptor hypofunction model of schizophrenia. Journal of Psychiatric Research. https://doi.org/10.1016/S0022- 3956(99)00029-1 Owen, M. J., Donovan, M. C. O., Thapar, A., & Craddock, N. (2011). Neurodevelopmental hypothesis of schizophrenia. The British Journal of Psychiatry, 198(3), 173–175. https://doi.org/10.1192/bjp.bp.110.084384 Ozcelik-Eroglu, E., Ertugrul, A., Oguz, K. K., Has, A. C., Karahan, S., & Yazici, M. K. (2014). Effect of clozapine on white matter integrity in patients with schizophrenia: A diffusion tensor imaging study. Psychiatry Research - Neuroimaging, 223(3), 226–235. https://doi.org/10.1016/j.pscychresns.2014.06.001 Pacia, S. V., & Devinsky, O. (1994). Clozapine-related seizures: Experience with 5,629 patients. Neurology, 44(12), 2247–2249. https://doi.org/10.1212/WNL.44.12.2247 Painuly, N., Gupta, N., & Avasthi, A. (2004). Concept and Management of Treatment Resistant Schizophrenia (TRS). Indian Journal of Psychiatry, 46(2), 125–134. Panariello, F., Polsinelli, G., Borlido, C., Monda, M., & De Luca, V. (2012). The role of leptin in antipsychotic-induced weight gain: genetic and non-genetic factors. Journal of Obesity, 2012, 572848. https://doi.org/10.1155/2012/572848 Park, S. J., Jeong, J., Park, Y.-U., Park, K.-S., Lee, H., Lee, N., … Park, S. K. (2015). Disrupted-in- schizophrenia-1 (DISC1) Regulates Endoplasmic Reticulum Calcium Dynamics. Scientific Reports, 5(1), 8694. https://doi.org/10.1038/srep08694 - 349 - REFERENCES Park, Y.-M., Chung, Y.-C., Lee, S.-H., Lee, K.-J., Kim, H., Choi, J.-E., … Lee, H.-J. (2009). G-protein beta3 Subunit Gene 825C/T Polymorphism Is Not Associated with Olanzapine-Induced Weight Gain in Korean Schizophrenic Patients. Psychiatry Investigation, 6(1), 39–43. https://doi.org/10.4306/pi.2009.6.1.39 Pedersen, K. S., Bamlet, W. R., Oberg, A. L., de Andrade, M., Matsumoto, M. E., Tang, H., … Wang, L. (2011). Leukocyte DNA methylation signature differentiates pancreatic cancer patients from healthy controls. PloS One, 6(3), e18223. https://doi.org/10.1371/journal.pone.0018223 Peltola, M. A., Kuja-Panula, J., Liuhanen, J., Võikar, V., Piepponen, P., Hiekkalinna, T., … Rauvala, H. (2016). AMIGO-Kv2.1 Potassium Channel Complex Is Associated with Schizophrenia-Related Phenotypes. Schizophrenia Bulletin, 42(1), 191–201. https://doi.org/10.1093/schbul/sbv105 Perala, J., Suvisaari, J., Saarni, S. I., Kuoppasalmi, K., Isometsa, E., Pirkola, S., … Lonnqvist, J. (2007). Lifetime prevalence of psychotic and bipolar I disorders in a general population. Arch Gen Psychiatry, 64(1), 19–28. https://doi.org/10.1001/archpsyc.64.1.19 Pers, T. H., Timshel, P., Ripke, S., Lent, S., Sullivan, P. F., O’Donovan, M. C., … Hirschhorn, J. N. (2016). Comprehensive analysis of schizophrenia-associated loci highlights ion channel pathways and biologically plausible candidate causal genes. Human Molecular Genetics, 25(6), 1247–1254. https://doi.org/10.1093/hmg/ddw007 Petronijević, N. D., Radonjić, N. V, Ivković, M. D., Marinković, D., Piperski, V. D., Đuričić, B. M., & Paunović, V. R. (2008). Plasma homocysteine levels in young male patients in the exacerbation and remission phase of schizophrenia. Progress in Neuro-Psychopharmacology & Biological Psychiatry, 32, 1921–1926. https://doi.org/10.1016/j.pnpbp.2008.09.009 Petronis, a, Paterson, a D., & Kennedy, J. L. (1999). Schizophrenia: an epigenetic puzzle? Schizophrenia Bulletin, 25(4), 639–655. https://doi.org/10.1708/1069.11712 Peuskens, J. (1999). The evolving definition of treatment resistance. The Journal of Clinical Psychiatry, 12(60), 4–8. Retrieved from http://psycnet.apa.org/record/1999-05687-001 Pickar, D., Owen, R. R., Litman, R. E., Konicki, E., Gutierrez, R., & Rapaport, M. H. (1992). Clinical and biologic response to clozapine in patients with schizophrenia. Crossover comparison with fluphenazine. Archives of General Psychiatry, 49, 345–353. https://doi.org/10.1001/archpsyc.1992.01820050009001 Pidsley, R., & Mill, J. (2011). Epigenetic studies of psychosis: Current findings, methodological approaches, and implications for postmortem research. Biological Psychiatry. https://doi.org/10.1016/j.biopsych.2010.03.029 - 350 - REFERENCES Pina-Camacho, L., Parellada, M., & Kyriakopoulos, M. (2016). Autism spectrum disorder and schizophrenia: boundaries and uncertainties. BJPscyh Advances, 22(5), 316–324. https://doi.org/10.1192/apt.bp.115.014720 Pomonis, J. D., Levine, A. S., & Billington, C. J. (1997). Interaction of the hypothalamic paraventricular nucleus and central nucleus of the amygdala in naloxone blockade of neuropeptide Y-induced feeding revealed by c-fos expression. The Journal of Neuroscience : The Official Journal of the Society for Neuroscience, 17(13), 5175–82. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/9185555 Portela, A., & Esteller, M. (2010). Epigenetic modifications and human disease. Nature Biotechnology, 28(10), 1057–1068. https://doi.org/10.1038/nbt.1685 Potkin, S. G., Basile, V. S., Jin, Y., Masellis, M., Badri, F., Keator, D., … Kennedy, J. L. (2003). D1 receptor alleles predict PET metabolic correlates of clinical response to clozapine. Molecular Psychiatry, 8(1), 109–13. https://doi.org/10.1038/sj.mp.4001191 Pramyothin, P., & Khaodhiar, L. (2010). Metabolic syndrome with the atypical antipsychotics. Current Opinion in Endocrinology, Diabetes and Obesity, 17(5), 460–466. https://doi.org/10.1097/MED.0b013e32833de61c Purcell, S. M., Moran, J. L., Fromer, M., Ruderfer, D., Solovieff, N., Roussos, P., … Mccarroll, S. A. (2014). A polygenic burden of rare dysruptive mutation in schizophrenia. Nature, 506(7487), 185–190. https://doi.org/10.1038/nature12975.A Purkayastha, S., Ford, J., Kanjilal, B., Diallo, S., Inigo, J. D. R., Neuwirth, L., … Banerjee, P. (2012). Clozapine Functions Through the Prefrontal Cortex Serotonin 1A Receptor to Heighten Neuronal Activity via Calmodulin Kinase II- NMDA Receptor Interactions. Journal of Neurochemistry, 120(3), 396–407. https://doi.org/10.1111/j.1747- 0285.2012.01428.x.Identification Qiao, C., Lin, D., & Wang, Y. (2015). The effective diagnosis of schizophrenia by using multi-layer RBMs deep networks. 2015 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), 603–606. https://doi.org/10.1109/BIBM.2015.7359751 Rajkumar, A. P., Poonkuzhali, B., Kuruvilla, A., Srivastava, A., Jacob, M., & Jacob, K. S. (2012). Outcome definitions and clinical predictors influence pharmacogenetic associations between HTR3A gene polymorphisms and response to clozapine in patients with schizophrenia. Psychopharmacology, 224(3), 441–449. https://doi.org/10.1007/s00213-012-2773-2 Ravan, M., Hasey, G., Reilly, J. P., MacCrimmon, D., & Khodayari-Rostamabad, A. (2015). A machine - 351 - REFERENCES learning approach using auditory odd-ball responses to investigate the effect of Clozapine therapy. Clinical Neurophysiology, 126(4), 721–730. https://doi.org/10.1016/j.clinph.2014.07.017 Rees, E., Kendall, K., Pardiñas, A. F., Legge, S. E., Pocklington, A., Escott-Price, V., … Kirov, G. (2016). Analysis of Intellectual Disability Copy Number Variants for Association With Schizophrenia. JAMA Psychiatry, 73(9), 963. https://doi.org/10.1001/jamapsychiatry.2016.1831 Relton, C. L., & Davey Smith, G. (2012). Two-step epigenetic mendelian randomization: A strategy for establishing the causal role of epigenetic processes in pathways to disease. International Journal of Epidemiology, 41(1), 161–176. https://doi.org/10.1093/ije/dyr233 Relton, C. L., & Smith, D. G. (2012). Is epidemiology ready for epigenetics? International Journal of Epidemiology, 41(1), 5–9. https://doi.org/10.1093/ije/dys006 Relton, C. L., & Smith, G. D. (2010). Epigenetic epidemiology of common complex disease: Prospects for prediction, prevention, and treatment. PLoS Medicine, 7(10). https://doi.org/10.1371/journal.pmed.1000356 Reynolds, G. P. (2012). The pharmacogenetics of symptom response to Antipsychotic drugs. Psychiatry Investigation, 9(1), 1–7. https://doi.org/10.4306/pi.2012.9.1.1 Riley, B. (2004). Linkage studies of schizophrenia. Neurotoxicity Research, 6(1), 17–34. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/15184102 Ripke, S., Neale, B. M., Corvin, A., Walters, J. T. R., Farh, K.-H., Holmans, P. a., … O’Donovan, M. C. (2014). Biological insights from 108 schizophrenia-associated genetic loci. Nature, 511, 421– 427. https://doi.org/10.1038/nature13595 Ripke, S., O’Dushlaine, C., Chambert, K., Moran, J. L., Kähler, A. K., Akterin, S., … Sullivan, P. F. (2013). Genome-wide association analysis identifies 13 new risk loci for schizophrenia. Nature Genetics, 45(10), 1150–1159. https://doi.org/10.1038/ng.2742 Robertson, K. (2005). DNA methylation and human disease. Nature Reviews, 6, 597–610. https://doi.org/10.1038/nrg1655 Robicsek, O., Karry, R., Petit, I., Salman-Kesner, N., Müller, F.-J., Klein, E., … Ben-Shachar, D. (2013). Abnormal neuronal differentiation and mitochondrial dysfunction in hair follicle-derived induced pluripotent stem cells of schizophrenia patients. Molecular Psychiatry, 18(10), 1067– 1076. https://doi.org/10.1038/mp.2013.67 Rodriguez, V. M., Andree, R. M., Perez Castejon, M. J., Luisa Catalina Zamora, M., Alvaro, P. C., Luis - 352 - REFERENCES Carreras Delgado, J., & Jose Rubia Vila, F. (1997). Fronto-striato-thalamic perfusion and clozapine response in treatment- refractory schizophrenic patients. A 99mTc-HMPAO study. Psychiatry Research - Neuroimaging, 76(1), 51–61. https://doi.org/10.1016/S0925- 4927(97)00057-7 Rosen, E., & Spiegelman, B. (2006). Adipocytes as regulators of energy balance and glucose homeostasis. Nature, 444(7121), 847–853. https://doi.org/10.1038/nature05483.Adipocytes Rubio, M. D., Wood, K., Haroutunian, V., & Meador-Woodruff, J. H. (2013). Dysfunction of the Ubiquitin Proteasome and Ubiquitin-Like Systems in Schizophrenia. Neuropsychopharmacology, 38(10), 1910–1920. https://doi.org/10.1038/npp.2013.84 Rukova, B., Staneva, R., Hadjidekova, S., Stamenov, G., Milanova, V., & Toncheva, D. (2014). Whole genome methylation analyses of schizophrenia patients before and after treatment. Biotechnology, Biotechnological Equipment, 28(3), 518–524. https://doi.org/10.1080/13102818.2014.933501 Rushlow, W. J., Seah, C., Sutton, L. P., Bjelica, A., & Rajakumar, N. (2009). Antipsychotics affect multiple calcium calmodulin dependent proteins. Neuroscience, 161(3), 877–886. https://doi.org/10.1016/j.neuroscience.2009.03.011 Ruzicka, W. B., Subburaju, S., Benes, F. M., A, M.-S., MA, L., & PS, K. (2015). Circuit- and Diagnosis- Specific DNA Methylation Changes at γ-Aminobutyric Acid–Related Genes in Postmortem Human Hippocampus in Schizophrenia and Bipolar Disorder. JAMA Psychiatry, 72(6), 541. https://doi.org/10.1001/jamapsychiatry.2015.49 Saffari, A., Silver, M. J., Zavattari, P., Moi, L., Columbano, A., Meaburn, E. L., & Dudbridge, F. (2018). Estimation of a significance threshold for epigenome‐wide association studies. Genetic epidemiology, 42(1), 20-33. Saha, S., Chant, D., Welham, J., & McGrath, J. (2005). A systematic review of the prevalence of schizophrenia. PLoS Medicine, 2(5), e141. https://doi.org/10.1371/journal.pmed.0020141 Samanaite, R., Gillespie, A., Sendt, K-V., McQueen, G., MacCabe, J.H. & Egerton, A .(2018) Biological Predictors of Clozapine Response: A Systematic Review. Front. Psychiatry 9:327 Samara, M. T., Dold, M., Gianatsi, M., Nikolakopoulou, A., Helfer, B., Salanti, G., & Leucht, S. (2016). Efficacy, Acceptability, and Tolerability of Antipsychotics in Treatment-Resistant Schizophrenia. JAMA Psychiatry, 73(3), 199–210. https://doi.org/10.1001/jamapsychiatry.2015.2955 Sarda, S., Das, A., Vinson, C., & Hannenhalli, S. (2017). Distal CpG islands can serve as alternative - 353 - REFERENCES promoters to transcribe genes with silenced proximal promoters. Genome Research, 27(4), 553–566. https://doi.org/10.1101/gr.212050.116 SCANDINAVIAN SIMVASTATIN SURVIVAL Study Group. (1994). Randomised trial of cholesterol lowering in 4444 patients with coronary heart disease: the Scandinavian Simvastatin Survival Study (4S). The Lancet, 344(8934). https://doi.org/10.1016/S0140-6736(94)90566-5 Scharfetter, J., Chaudhry, H. R., Hornik, K., Fuchs, K., Sieghart, W., Kasper, S., & Aschauer, H. N. (1999). Dopamine D3 receptor gene polymorphism and response to clozapine in schizophrenic Pakistani patients. European Neuropsychopharmacology, 10(1), 17–20. https://doi.org/10.1016/S0924-977X(99)00044-9 Schmitt, A., Leonardi-Essmann, F., Durrenberger, P. F., Wichert, S. P., Spanagel, R., Arzberger, T., … Gebicke-Haerter, P. J. (2012). Structural synaptic elements are differentially regulated in superior temporal cortex of schizophrenia patients. European Archives of Psychiatry and Clinical Neuroscience, 262(7), 565–577. https://doi.org/10.1007/s00406-012-0306-y Schmitt, A., May, B., Müller, B., Jatzko, A., Petroianu, G., Braus, D. F., & Henn, F. A. (2003). Effects of Chronic Haloperidol and Clozapine Treatment on AMPA and Kainate Receptor Binding in Rat Brain. Pharmacopsychiatry, 36(6), 292–296. https://doi.org/10.1055/s-2003-45116 Schübeler, D. (2015). Function and information content of DNA methylation. Nature, 517(7534), 321–326. https://doi.org/10.1038/nature14192 Schubert, K. O., Föcking, M., Prehn, J. H. M., & Cotter, D. R. (2012). Hypothesis review: are clathrin- mediated endocytosis and clathrin-dependent membrane and protein trafficking core pathophysiological processes in schizophrenia and bipolar disorder? Molecular Psychiatry, 17(7), 669–681. https://doi.org/10.1038/mp.2011.123 Schwieler, L., Engberg, G., & Erhardt, S. (2004). Clozapine Modulates Midbrain Dopamine Neuron Firing Via Interaction with the NMDA Receptor Complex. Synapse, 52(2), 114–122. https://doi.org/10.1002/syn.20008 Schwieler, L., Linderholm, K. R., Nilsson-Todd, L. K., Erhardt, S., & Engberg, G. (2008). Clozapine interacts with the glycine site of the NMDA receptor: Electrophysiological studies of dopamine neurons in the rat ventral tegmental area. Life Sciences, 83(5–6), 170–175. https://doi.org/10.1016/j.lfs.2008.05.014 Scigliano, G., & Ronchetti, G. (2013). Antipsychotic-induced metabolic and cardiovascular side effects in schizophrenia: a novel mechanistic hypothesis. CNS Drugs, 27(4), 249–57. https://doi.org/10.1007/s40263-013-0054-1 - 354 - REFERENCES Seeman, M. V. (2010). Schizophrenia: Women Bear a Disproportionate Toll of Antipsychotic Side Effects. Journal of the American Psychiatric Nurses Association, 16(1), 21–29. https://doi.org/10.1177/1078390309350918 Sekar, A., Bialas, A. R., de Rivera, H., Davis, A., Hammond, T. R., Kamitaki, N., … McCarroll, S. A. (2016). Schizophrenia risk from complex variation of complement component 4. Nature, 530(7589), 177–83. https://doi.org/10.1038/nature16549 Selemon, L. D., & Zecevic, N. (2015). Schizophrenia: a tale of two critical periods for prefrontal cortical development. Translational Psychiatry, 5(8), e623. https://doi.org/10.1038/tp.2015.115 Shaikh, S., Collier, D. A., Sham, P. C., Ball, D., Aitchison, K., Vallada, H., … Kerwin, R. W. (1996). Allelic association between a Ser-9-Gly polymorphism in the dopamine D3 receptor gene and schizophrenia. Human Genetics, 97(6), 714–719. https://doi.org/10.1007/s004390050125 Shakhbazov, K., Powell, J. E., Hemani, G., Henders, A. K., Martin, N. G., Visscher, P. M., … McRae, A. F. (2016). Shared genetic control of expression and methylation in peripheral blood. BMC Genomics, 17(1), 278. https://doi.org/10.1186/s12864-016-2498-4 Sham, P. C., MacLean, C. J., & Kendler, K. S. (1994). A typological model of schizophrenia based on age at onset, sex and familial morbidity. Acta Psychiatrica Scandinavica, 89(2), 135–141. https://doi.org/10.1111/j.1600-0447.1994.tb01501.x Shams, T. A., & Müller, D. J. (2014). Antipsychotic induced weight gain: genetics, epigenetics, and biomarkers reviewed. Current Psychiatry Reports, 16(10), 473. https://doi.org/10.1007/s11920- 014-0473-9 Sharma, E., Rao, N. P., & Venkatasubramanian, G. (2014). Association between antipsychotic- induced metabolic side-effects and clinical improvement: A review on the Evidence for “metabolic threshold.” Asian Journal of Psychiatry, 8(1), 12–21. https://doi.org/10.1016/j.ajp.2013.11.017 Sheridan, P. A., Paich, H. A., Handy, J., Karlsson, E. A., Hudgens, M. G., Sammon, A. B., … Beck, M. A. (2012). Obesity is associated with impaired immune response to influenza vaccination in humans. International Journal of Obesity (2005), 36(8), 1072–7. https://doi.org/10.1038/ijo.2011.208 Sillitoe, R. V, Benson, M. A., Blake, D. J., & Hawkes, R. (2003). Abnormal dysbindin expression in cerebellar mossy fiber synapses in the mdx mouse model of Duchenne muscular dystrophy. The Journal of Neuroscience, 23(16), 6576–85. Retrieved from - 355 - REFERENCES http://www.jneurosci.org/content/jneuro/23/16/6576.full.pdf Silverman, J. M., Mohs, R. C., Davidson, M., Losonczy, M. F., Keefe, R. S., Breitner, J. C., … Davis, K. L. (1987). Familial schizophrenia and treatment response. American Journal of Psychiatry, 144(10), 1271–1276. https://doi.org/10.1176/ajp.144.10.1271 Singh, T., Kurki, M. I., Curtis, D., Purcell, S. M., Crooks, L., McRae, J., … Barrett, J. C. (2016). Rare loss- of-function variants in SETD1A are associated with schizophrenia and developmental disorders. Nature Neuroscience, 19(4), 571. https://doi.org/10.1038/nn.4267 Siskind, D., Mccartney, L., Goldschlager, R., & Kisely, S. (2016). Clozapine v. first-and second- generation antipsychotics in treatment-refractory schizophrenia: systematic review and meta- analysis. The British Journal of Psychiatry, 1–8. https://doi.org/10.1192/bjp.bp.115.177261 Sivagnanasundaram, S., Crossett, B., Dedova, I., Cordwell, S., & Matsumoto, I. (2007). Abnormal pathways in the genu of the corpus callosum in schizophrenia pathogenesis: A proteome study. Proteomics - Clinical Applications, 1(10), 1291–1305. https://doi.org/10.1002/prca.200700230 Small, J. G., Kellams, J. J., Milstein, V., & Moore, J. (1975). A placebo-controlled study of lithium combined with neuroleptics in chronic schizophrenic patients. American Journal of Psychiatry, 132(12), 1315–1317. https://doi.org/10.1176/ajp.132.12.1315 Smith, G. C., Zhang, Z. Y., Mulvey, T., Petersen, N., Lach, S., Xiu, P., … Shepherd, P. R. (2014). Clozapine directly increases insulin and glucagon secretion from islets: Implications for impairment of glucose tolerance. Schizophrenia Research, 157(1–3), 128–133. https://doi.org/10.1016/j.schres.2014.05.003 Sodhi, M. S., Arranz, M. J., Curtis, D., Ball, D. M., Sham, P., Roberts, G. W., … Kerwin, R. W. (1995). Association between clozapine response and allelic variation in the 5-HT2C receptor gene. Neuroreport, 7(1), 169–72. https://doi.org/10.1097/00001756-199512290-00041 Solanki, R. K., Singh, P., & Swami, M. K. (2007). Clozapine: Current perspective. Indian Journal of Psychiatry, 49(4), 271–6. https://doi.org/10.4103/0019-5545.37668 Song, C., Lin, a H., Kenis, G., Bosmans, E., & Maes, M. (2000). Immunosuppressive effects of clozapine and haloperidol: enhanced production of the interleukin-1 receptor antagonist. Schizophrenia Research, 42(2), 157–64. https://doi.org/10.1016/S0920-9964(99)00116-4 Souza, R. P., de Luca, V., Meltzer, H. Y., Lieberman, J. a, & Kennedy, J. L. (2010). Influence of serotonin 3A and 3B receptor genes on clozapine treatment response in schizophrenia. Pharmacogenetics and Genomics, 20(4), 274–6. - 356 - REFERENCES https://doi.org/10.1097/FPC.0b013e328337ce3e Souza, R. P., De Luca, V., Muscettola, G., Rosa, D. V. F., De Bartolomeis, A., Romano Silva, M., & Kennedy, J. L. (2008). Association of antipsychotic induced weight gain and body mass index with GNB3 gene: A meta-analysis. Progress in Neuropsychopharmacology & Biological Psychiatry, 32, 1848–1853. https://doi.org/10.1016/j.pnpbp.2008.08.014 Souza, R. P., Romano-Silva, M. A., Lieberman, J. A., Meltzer, H. Y., MacNeil, L. T., Culotti, J. G., … Wong, A. H. C. (2010). Genetic association of the GDNF alpha-receptor genes with schizophrenia and clozapine response. Journal of Psychiatric Research, 44(11), 700–706. https://doi.org/10.1016/j.jpsychires.2010.01.002 Stedehouder, J., & Kushner, S. (2016). Myelination of parvalbumin interneurons: a parsimonious locus of pathophysiological convergence in schizophrenia. Molecular Psychiatry, 22, 4–12. https://doi.org/10.1038/mp.2016.147 Steiner, J., Jacobs, R., Panteli, B., Brauner, M., Schiltz, K., Bahn, S., … Bogerts, B. (2010). Acute schizophrenia is accompanied by reduced T cell and increased B cell immunity. European Archives of Psychiatry and Clinical Neuroscience, 260(7), 509–518. https://doi.org/10.1007/s00406-010-0098-x Steiner, J., Martins-de-Souza, D., Schiltz, K., Sarnyai, Z., Westphal, S., Isermann, B., … Keilhoff, G. (2014). Clozapine promotes glycolysis and myelin lipid synthesis in cultured oligodendrocytes. Frontiers in Cellular Neuroscience, 8, 384. https://doi.org/10.3389/fncel.2014.00384 Stelzer, G., Rosen, N., Plaschkes, I., Zimmerman, S., Twik, M., Fishilevich, S., … Lancet, D. (2016). The GeneCards Suite: From Gene Data Mining to Disease Genome Sequence Analyses. In Current Protocols in Bioinformatics (p. 1.30.1-1.30.33). Hoboken, NJ, USA: John Wiley & Sons, Inc. https://doi.org/10.1002/cpbi.5 Sutton, L. P., Honardoust, D., Mouyal, J., Rajakumar, N., & Rushlow, W. J. (2007). Activation of the canonical Wnt pathway by the antipsychotics haloperidol and clozapine involves dishevelled-3. Journal of Neurochemistry, 102(1), 153–169. https://doi.org/10.1111/j.1471- 4159.2007.04527.x Suzuki, T., Remington, G., Mulsant, B. H., Uchida, H., Rajji, T. K., Graff-Guerrero, A., … Mamo, D. C. (2012). Defining treatment-resistant schizophrenia and response to antipsychotics: A review and recommendation. Psychiatry Research, 197(1), 1–6. https://doi.org/10.1016/j.psychres.2012.02.013 Suzuki, T., Uchida, H., Watanabe, K., & Kashima, H. (2011). Factors associated with response to - 357 - REFERENCES clozapine in schizophrenia: a review. Psychopharmacology Bulletin, 44(1), 32–60. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/22506438 Swerdlow, N. R., Braff, D. L., & Geyer, M. A. (2016). Sensorimotor gating of the startle reflex: what we said 25 years ago, what has happened since then, and what comes next. Journal of Psychopharmacology, 30(11), 1072–1081. https://doi.org/10.1177/0269881116661075 Swerdlow, N. R., & Geyer, M. A. (1993). Clozapine and haloperidol in an animal model of sensorimotor gating deficits in schizophrenia. Pharmacology, Biochemistry and Behavior, 44(3), 741–744. https://doi.org/10.1016/0091-3057(93)90193-W Szymanski, S., Lieberman, J., Pollack, S., Munne, R., Safferman, A., Kane, J., … Cooper, T. (1993). The dopamine-serotonin relationship in clozapine response. Psychopharmacology, 112(1 Supplement), 85–89. https://doi.org/10.1007/BF02245011 Takata, A., Xu, B., Ionita-Laza, I., Roos, J. L., Gogos, J. A., & Karayiorgou, M. (2014). Loss-of-function variants in schizophrenia risk and SETD1A as a candidate susceptibility gene. Neuron, 82(4), 773–80. https://doi.org/10.1016/j.neuron.2014.04.043 Takeuchi, H., Fervaha, G., & Remington, G. (2016). Consistency between clinician and patient ratings of clozapine-induced side effects. Schizophrenia Research, 174(1–3), 200–201. https://doi.org/10.1016/j.schres.2016.04.005 Tanahashi, S., Yamamura, S., Nakagawa, M., Motomura, E., & Okada, M. (2012). Clozapine, but not haloperidol, enhances glial d -serine and L-glutamate release in rat frontal cortex and primary cultured astrocytes. British Journal of Pharmacology, 165(5), 1543–1555. https://doi.org/10.1111/j.1476-5381.2011.01638.x Taylor, D. M. (2017). Clozapine for Treatment-Resistant Schizophrenia: Still the Gold Standard? CNS Drugs, 31(3), 177–180. https://doi.org/10.1007/s40263-017-0411-6 Teh, A. L., Pan, H., Lin, X., Lim, Y. I., Patro, C. P. K., Cheong, C. Y., … Karnani, N. (2016). Comparison of Methyl-capture Sequencing vs. Infinium 450K methylation array for methylome analysis in clinical samples. Epigenetics, 11(1), 36–48. https://doi.org/10.1080/15592294.2015.1132136 Teroganova, N., Girshkin, L., Suter, C. M., & Green, M. J. (2016). DNA methylation in peripheral tissue of schizophrenia and bipolar disorder: a systematic review. BMC Genetics, 17(1), 17:27. https://doi.org/10.1093/carcin/21.3.461 Thoma, P., & Daum, I. (2013). Comorbid substance use disorder in schizophrenia: A selective overview of neurobiological and cognitive underpinnings. Psychiatry and Clinical Neurosciences, - 358 - REFERENCES 67(6), 367–383. https://doi.org/10.1111/pcn.12072 Thompson, J. V., Clark, J. M., Legge, S. E., Kadra, G., Downs, J., Walters, J. T., … MacCabe, J. H. (2016). Antipsychotic polypharmacy and augmentation strategies prior to clozapine initiation: a historical cohort study of 310 adults with treatment-resistant schizophrenic disorders. Journal of Psychopharmacology. https://doi.org/10.1177/0269881116632376 Tiihonen, J., Lönnqvist, J., Wahlbeck, K., Klaukka, T., Niskanen, L., Tanskanen, A., & Haukka, J. (2009). 11-year follow-up of mortality in patients with schizophrenia: a population-based cohort study (FIN11 study). The Lancet, 374(9690), 620–627. https://doi.org/10.1016/S0140-6736(09)60742- X Tiwari, A. K., Brandl, E. J., Weber, C., Likhodi, O., Zai, C. C., Hahn, M. K., … Müller, D. J. (2013). Association of a Functional Polymorphism in Neuropeptide Y With Antipsychotic-Induced Weight Gain in Schizophrenia Patients. Journal of Clinical Psychopharmacology, 33(1), 11–17. https://doi.org/10.1097/JCP.0b013e31827d145a Tsai, G. E., Yang, P., Chang, Y. C., & Chong, M. Y. (2006). D-alanine added to antipsychotics for the treatment of schizophrenia. Biological Psychiatry, 59(3), 230–234. https://doi.org/10.1016/j.biopsych.2005.06.032 Tsai, G. E., Yang, P., Chung, L. C., Tsai, I. C., Tsai, C. W., & Coyle, J. T. (1999). D-serine added to clozapine for the treatment of schizophrenia. American Journal of Psychiatry, 156(11), 1822– 1825. https://doi.org/10.1176/ajp.156.11.1822 Tucholski, J., Simmons, M. S., Pinner, A. L., Haroutunian, V., McCullumsmith, R. E., & Meador- Woodruff, J. H. (2013). Abnormal N-linked glycosylation of cortical AMPA receptor subunits in schizophrenia. Schizophrenia Research, 146(1–3), 177–183. https://doi.org/10.1016/j.schres.2013.01.031 Tuominen, H. J., Tiihonen, J., & Wahlbeck, K. (2005, January 1). Glutamatergic drugs for schizophrenia: A systematic review and meta-analysis. Schizophrenia Research. https://doi.org/10.1016/j.schres.2004.05.005 Üçok, A., Çkrkçl, U., Karabulut, S., Salaj, A., Öztürk, M., Tabak, Ö., & Durak, R. (2015). Delayed initiation of clozapine may be related to poor response in treatment-resistant schizophrenia. International Clinical Psychopharmacology, 1. https://doi.org/10.1097/YIC.0000000000000086 Van Den Oord, E. J. C. G., Clark, S. L., Xie, L. Y., Shabalin, A. A., Dozmorov, M. G., Kumar, G., … Aberg, K. A. (2016). A whole methylome CpG-SNP association study of psychosis in blood and brain tissue. Schizophrenia Bulletin, 42(4), 1018–1026. https://doi.org/10.1093/schbul/sbv182 - 359 - REFERENCES van Eijk, K. R., de Jong, S., Boks, M. P., Langeveld, T., Colas, F., Veldink, J. H., … Ophoff, R. A. (2012). Genetic analysis of DNA methylation and gene expression levels in whole blood of healthy human subjects. BMC Genomics, 13(1), 636. https://doi.org/10.1186/1471-2164-13-636 van Eijk, K. R., de Jong, S., Strengman, E., Buizer-Voskamp, J. E., Kahn, R. S., Boks, M. P., … Ophoff, R. A. (2015). Identification of schizophrenia-associated loci by combining DNA methylation and gene expression data from whole blood. European Journal of Human Genetics : EJHG, 23(8), 1106–10. https://doi.org/10.1038/ejhg.2014.245 Veerman, S. R. T., Schulte, P. F. J., Begemann, M. J. H., Engelsbel, F., & de Haan, L. (2014). Clozapine augmented with glutamate modulators in refractory schizophrenia: a review and metaanalysis. Pharmacopsychiatry, 47(6), 185–194. https://doi.org/10.1055/s-0034-1383656 Viikki, M., Kampman, O., Seppälä, N., Mononen, N., Lehtimäki, T., & Leinonen, E. (2014). CYP1A2 polymorphism -1545C > T (rs2470890) is associated with increased side effects to clozapine. BMC Psychiatry, 14, 50. https://doi.org/10.1186/1471-244X-14-50 Vonberg, F., & Bigdeli, T. B. (2016). Genetic Correlation Between Schizophrenia and Epilepsy. JAMA Neurology, 73(1), 125–126. https://doi.org/10.1001/jamaneurol.2015.3480.Author Walton, E., Hass, J., Liu, J., Roffman, J. L., Bernardoni, F., Roessner, V., … Ehrlich, S. (2015). Correspondence of DNA Methylation Between Blood and Brain Tissue and its Application to Schizophrenia Research. Schizophrenia Bulletin, 42(2), sbv074-. https://doi.org/10.1093/schbul/sbv074 Ward, S. E., Pennicott, L. E., & Beswick, P. (2015). AMPA receptor-positive allosteric modulators for the treatment of schizophrenia: an overview of recent patent applications. Future Medicinal Chemistry, 7(4), 473–491. https://doi.org/10.4155/fmc.15.4 Weaver, I. C. G., Cervoni, N., Champagne, F. A., D’Alessio, A. C., Sharma, S., Seckl, J. R., … Meaney, M. J. (2004). Epigenetic programming by maternal behavior. Nature Neuroscience, 7(8), 847–854. https://doi.org/10.1038/nn1276 Wimberley, T., MacCabe, J. H., Laursen, T. M., Sørensen, H. J., Astrup, A., Horsdal, H. T., … Støvring, H. (2017). Mortality and Self-Harm in Association With Clozapine in Treatment-Resistant Schizophrenia. American Journal of Psychiatry, appi.ajp.2017.1. https://doi.org/10.1176/appi.ajp.2017.16091097 Woo, T. U. W., & Crowell, A. L. (2005). Targeting synapses and myelin in the prevention of schizophrenia. Schizophrenia Research, 73(2–3), 193–207. https://doi.org/10.1016/j.schres.2004.07.022 - 360 - REFERENCES Wotton, C. J., & Goldacre, M. J. (2012). Coexistence of schizophrenia and epilepsy: Record-linkage studies. Epilepsia, 53(4), 71–74. https://doi.org/10.1111/j.1528-1167.2011.03390.x Wu, H., D’Alessio, A. C., Ito, S., Wang, Z., Cui, K., Zhao, K., … Zhang, Y. (2011). Genome-wide analysis of 5-hydroxymethylcytosine distribution reveals its dual function in transcriptional regulation in mouse embryonic stem cells. Genes & Development, 25(7), 679–684. https://doi.org/10.1101/gad.2036011 Wu, H., & Zhang, Y. (2014). Reversing DNA methylation: mechanisms, genomics, and biological functions. Cell, 156(1–2), 45–68. https://doi.org/10.1016/j.cell.2013.12.019 Wysokiński, A., & Kłoszewska, I. (2014). Blood serum levels of CART peptide in patients with schizophrenia on clozapine monotherapy. Psychiatry Research, 220(1–2), 170–174. https://doi.org/10.1016/j.psychres.2014.08.017 Xu, Q., Wu, X., Li, M., Huang, H., Minica, C., Yi, Z., … Qin, S. (2016). Association studies of genomic variants with treatment response to risperidone, clozapine, quetiapine and chlorpromazine in the Chinese Han population. The Pharmacogenomics Journal, 16, 357–365. https://doi.org/10.1038/tpj.2015.61 Yamamori, H., Hashimoto, R., Fujita, Y., Numata, S., Yasuda, Y., Fujimoto, M., … Takeda, M. (2014). Changes in plasma d-serine, l-serine, and glycine levels in treatment-resistant schizophrenia before and after clozapine treatment. Neuroscience Letters, 582, 93–98. https://doi.org/10.1016/j.neulet.2014.08.052 Yilmaz, Z., Zai, C. C., Hwang, R., Mann, S., Arenovich, T., Remington, G., & Daskalakis, Z. J. (2012). Antipsychotics, dopamine D2 receptor occupancy and clinical improvement in schizophrenia: A meta-analysis. Schizophrenia Research, 140(1–3), 214–220. https://doi.org/10.1016/j.schres.2012.06.027 Young, C. R., Bowers, M. B., & Mazure, C. M. (1998). Management of the adverse effects of clozapine. Schizophrenia Bulletin, 24(3), 381–390. https://doi.org/10.1007/s40266-013-0144-2 Young, M. D., Davidson, N., Wakefield, M. J., Smyth, G. K., & Oshlack, A. (2016). goseq: Gene Ontology testing for RNA-seq datasets. Retrieved from https://bioconductor.org/packages/release/bioc/vignettes/goseq/inst/doc/goseq.pdf Young, M. D., Wakefield, M. J., Smyth, G. K., & Oshlack, A. (2010). Gene ontology analysis for RNA- seq: accounting for selection bias. Genome Biology, 11. Retrieved from http://genomebiology.com/2010/11/2/R14 - 361 - REFERENCES Yudkin, J. (2007). Inflammation, Obesity, and the Metabolic Syndrome. Hormone and Metabolic Research, 39(10), 707–709. https://doi.org/10.1055/s-2007-985898 Yudkin, J. S., Kumari, M., Humphries, S. E., & Mohamed-Ali, V. (2000). Inflammation, obesity, stress and coronary heart disease: is interleukin-6 the link? Atherosclerosis, 148, 209–214. Retrieved from https://s3.amazonaws.com/academia.edu.documents/45796593/Inflammation_obesity_and_c oronary_heart_20160520-23744- 16n3y5s.pdf?AWSAccessKeyId=AKIAIWOWYYGZ2Y53UL3A&Expires=1505151138&Signature=9 PblNdlv7qpK2e6PdR8N2HVIv%2Bo%3D&response-content-disposition=inli Yusufi, B., Mukherjee, S., Flanagan, R., Paton, C., Dunn, G., Page, E., & Barnes, T. R. E. (2007). Prevalence and nature of side effects during clozapine maintenance treatment and the relationship with clozapine dose and plasma concentration. International Clinical Psychopharmacology, 22(4), 238–243. https://doi.org/10.1097/YIC.0b013e32819f8f17 Zai, G. C. M., Zai, C. C. H., Chowdhury, N. I., Tiwari, A. K., Souza, R. P., Lieberman, J. A., … Kennedy, J. L. (2012). The role of brain-derived neurotrophic factor (BDNF) gene variants in antipsychotic response and antipsychotic-induced weight gain. Progress in Neuro-Psychopharmacology and Biological Psychiatry, 39(1), 96–101. https://doi.org/10.1016/j.pnpbp.2012.05.014 Zatz, M., Vallada, H., Melo, M. S., Passos-Bueno, M. R., Vieira, A. H., Vainzof, M., … Gentil, V. (1993). Cosegregation of schizophrenia with Becker muscular dystrophy: susceptibility locus for schizophrenia at Xp21 or an effect of the dystrophin gene in the brain? Journal of Medical Genetics, 30(2), 131–4. https://doi.org/10.1136/JMG.30.2.131 Zhang, J.-P., Gallego, J. A., Robinson, D. G., Malhotra, A. K., Kane, J. M., & Correll, C. U. (2013). Efficacy and safety of individual second-generation vs. first-generation antipsychotics in first- episode psychosis: a systematic review and meta-analysis. The International Journal of Neuropsychopharmacology, 16(6), 1205–1218. https://doi.org/10.1017/S1461145712001277 Zhang, J.-P., Lencz, T., Geisler, S., DeRosse, P., Bromet, E. J., & Malhotra, A. K. (2013). Genetic variation in BDNF is associated with antipsychotic treatment resistance in patients with schizophrenia. Schizophrenia Research, 146(1–3), 285–288. https://doi.org/10.1016/j.schres.2013.01.020 Zhang, J.-P., & Malhotra, A. K. (2011). Pharmacogenetics and antipsychotics: therapeutic efficacy and side effects prediction. Expert Opinion on Drug Metabolism & Toxicology, 7(1), 9–37. https://doi.org/10.1517/17425255.2011.532787 - 362 - REFERENCES Zhang, J.-P., & Malhotra, A. K. (2013). Genetics of schizophrenia: What do we know? Current Psychiatry, 12(3), 24–33. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/24443645 Zhang, X. Y., Tan, Y. L., Zhou, D. F., Haile, C. N., Cao, L. Y., Xu, Q., … Kosten, T. R. (2007). Association of Clozapine-Induced Weight Gain With a Polymorphism in the Leptin Promoter Region in Patients With Chronic Schizophrenia in a Chinese Population. Journal of Clinical Psychopharmacology, 27(3), 246–251. https://doi.org/10.1097/jcp.0b013e3180582412 - 363 - APPENDIX I APPENDIX I Gillespie, A. L., Samanaite, R., Mill, J., Egerton, A., & MacCabe, J. H. (2017). Is treatment- resistant schizophrenia categorically distinct from treatment-responsive schizophrenia? a systematic review. BMC Psychiatry, 17(1), 12. - 364 -