The effect of clozapine on DNA methylation in patients with schizophrenia and differences in DNA methylation status for responders compared to non-responders

(1)

4

J. van Schie

11197552

Bachelor’s thesis

Biomedical Sciences

Universiteit van Amsterdam, Amsterdam

Supervisor: M.Z. van der Horst

Project manager: dr. J.J. Luykx

Examiner: dr. T.R. Werkman

July 1st, 2020

The effect of clozapine on DNA methylation in patients with schizophrenia and

(2)

2

Abbreviations

ACTN1 alpha-actinin-1 CGI Clinical Global Impression CpGI CpG Island CLZ clozapine DISC1 Disrupted-In-SCZ-1 DMP differentially methylated positions DMR differentially methylated regions DRD2 dopamine receptor D2 DSM-V Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition DNAm DNA methylation EPS extrapyramidal side effects FDR false discovery rate GO Gene Ontology HNRNPH1 heterogeneous nuclear ribonucleoprotein H PANSS Positive and Negative Symptoms Scale SAD schizoaffective disorder SCZ schizophrenia TRS treatment-resistant schizophrenia UGCG UDP-glucose ceramide glucosyltransferase PPP4R2 protein phosphatase 4 regulatory subunit 2

Abstract

IMPORTANCE DNA methylation might play an important role in pathophysiology of

schizophrenia or might be an important biomarker of risk. OBJECTIVE The aim of this study was to gain more insight into the molecular mechanisms of clozapine by examining epigenetic changes in patients with schizophrenia. Therefore the DNA methylation status of patients diagnosed with schizophrenia was examined before and after initiation of clozapine treatment. This study also examined differences in DNA methylation for responders compared to non-responders. METHODS This study conducted a genome-wide DNA methylation profiling in blood samples (802.579 CpG sites) from patients with schizophrenia or schizoaffective disorder treated with clozapine (n = 11) for six months. DNA methylation was determined at three moments: before initiation of clozapine treatment, four-twelve weeks after initiation of clozapine treatment and six months after initiation of clozapine treatment. Results revealed significantly different DNA methylation for one CpG site after initiation of clozapine treatment (four-twelve weeks) in all participants. RESULTS The results also showed one significantly different CpG site between responders and non-responders which is consistent for all three moments of DNA methylation measurement and five CpG sites which also show different DNA methylation. CONCLUSIONS The genes annotated to the found CpG sites were either associated directly with schizophrenia or had involvement in pathways which include quantitative risk factors for schizophrenia. The findings of this study could have implications for understanding the working mechanisms of antipsychotics such as clozapine and understanding the phenomenon of treatment-resistant schizophrenia.

(3)

3

Introduction

Schizophrenia (SCZ) is a psychotic disorder which is characterized by major symptoms such as hallucinations, delusions, disorganized speech and abnormal psychomotor behaviour, also referred to as “positive” symptoms [1]. Other symptoms are social withdrawal, decreased emotion expression and lack of motivation, also referred to as “negative” symptoms. SCZ has related disorders which have some common symptoms, such as schizoaffective disorder (SAD). SAD patients have experiences of psychotic and mood symptoms, characterised by depressive and manic episodes, alternated with schizophrenic symptoms such as delusions and hallucinations [2]. Approximately, 30% of the SCZ patients is diagnosed with treatment-resistant SCZ (TRS) [3], about 20% of all SCZ patients are complete non-responders [4].

Treatment-resistant SCZ is generally diagnosed when patients fail to show clinical improvement to at least two adequate antipsychotic trials over a period of six weeks [5], [6]. The clinical and psychotic improvement of a patient shows if a patient is responsive or non-responsive to a particular antipsychotic agent. In the case of SCZ, responsiveness can be determined using the Positive and Negative Symptoms Scale (PANSS), Clinical Global Impression (CGI), or other psychiatric scales [5]. When using the PANSS scale the score must be at least (>)50% [7] from a baseline cut-off for an adequate response. This applies to patients who have not already been diagnosed with TRS. For patients diagnosed with TRS a smaller reduction could indicate response to antipsychotic agent, in general this is true for a reduction cut-off of at least (>)20-25% [3], [7]. So, patients diagnosed with TRS show low reduction scores for at least two trials with

different antipsychotic agents. But, there is also a group of patients within this TRS population who keep responding insufficiently to multiple antipsychotic agents (two or more trials), these patients are called complete non-responders.

In the case of TRS CLZ has demonstrated superior efficacy in clinical response by relieving negative and positive symptoms [8]. Another important reason is its minimal cause of extrapyramidal side effects (EPS), such as dystonias, bradykinesia, akathisia, and dyskinesias [9]. Despite the minimal EPS, CLZ has several common side effects and some serious side effects. Common side effects of CLZ are sedation, hypersalivation, tachycardia, hypo-/hypertension, and weight gain [10]. It is evident that CLZ users have an increased risk of developing obesity, diabetes mellitus, and cardiovascular disease [11], [12]. Although these common side effects can have a significant impact on the lives of patients, they are generally tolerated by the patient [13]. Uncommon but more serious side effects of CLZ are agranulocytosis (~1% incidence), cardiovascular and respiratory arrest (2-4% incidence) and seizures (2-5% incidence) [10]. CLZ has the highest risk of seizures compared to other antipsychotic agents. The risk of seizures is actually dose-dependent and can be about 5% for a daily dosage of more than 600 mg. However, by lowering the daily dosage and prescription of additional antiepileptic agents, this risk can be significantly reduced, in most cases enabling the patient to continue the treatment [14], [15]. Agranulocytosis is a serious life-threatening side effect of CLZ, which generally occurs within the first few months of treatment (50% within twelve weeks, 75% within six months) [9], [16]. CLZ-induced agranulocytosis has a suggested

(5)

5 cumulative mortality of 0.1-0.3% with white

blood cell monitoring [17]. Because CLZ-induced agranulocytosis mostly occurs within the first six months, it is suggested that only blood monitoring for the first six months could be enough to minimalize the risk of fatality to negligible [18].

Although CLZ can have serious side effects, the clinical efficacy of CLZ is found evident in several studies [19]–[21]. It is also clear that the incidence of most side effects can be significantly reduced by mandatory blood monitoring and adjusting daily dosage [10]. Despite recommendations and evidence delivered by researchers, including CLZ effectiveness and low risk of side effects, the prescription of CLZ seems to be low [22]. In general, CLZ is prescribed when at least 2 trials with any other antipsychotic (either typical or atypical) agents do not have sufficient clinical improvement [6], [23]. If the response and possible side effects could be estimated on forehand, CLZ could be prescribed more adequately which could lead to a higher overall efficacy of this antipsychotic agent. This might be achieved by examining epigenetic markers for side effects of and response to CLZ, as this area is of growing interest to recent studies.

DNA methylation (DNAm) is one of the best studied epigenetic modifications of the genome and it has profound roles in many biological processes, especially associated with development (e.g., embryotic development, genomic imprinting and chromosome stability) [24]. Likewise DNAm has an important role in the development of the brain [24], [25]. However, DNAm is not only important for development of the brain in early life but also mediates plasticity and neurogenesis in the adult brain [26]–[28]. DNAm is a covalent modification on a cytosine located on a CG dinucleotide, called

a CpG site. This modification is catalysed by an enzyme family of DNA methyltransferases (DNMT), which transfers a methyl group to the fifth carbon residue of cytosine forming 5-methyl-cytosine (5mC) [29], [30]. About 10% of CpG sites are located in a high density region (~2kb), called CpG islands (CpGI). These CpGIs are mostly found at transcription start sites (promoters) and are predominantly hypomethylated [31], [32]. In addition to CpGIs, CpG shores (2kb regions upstream or downstream from CpGI) [33], shelfs (2kb regions upstream or downstream from CpG shores) [33] and “open seas” (CpG sites located outside previous regions) [35], [36] are classified regions for CpG sites. See Figure 1 for illustration. Locations of aberrant DNAm in CpGIs or in other regions of CpG sites, are called differentially methylated positions (DMPs) or differentially methylated regions (DMRs) when multiple DMPs are found close to each other in the same CpG region.

DNAm status can be measured in the form of DMPs and DMRs by analysing data obtained with advanced microarray techniques. These DMPs and DMRs are investigated and associated with human diseases in the so called Epigenome-Wide Association Studies (EWAS) [37]. This mapping of DNAm data in diseased humans and large groups of healthy humans (controls) could result in finding biomarkers useful for specific disease diagnostics [38]. A growing number of human diseases are therefore already associated with aberrant DNAm [39]. Likewise, a growing number of neurological diseases like neurodegenerative diseases [40] and mental disorders, such as SCZ, are being associated with aberrant DNAm in different cell tissues [41]–[46].

EWAS studies are relatively new, but Genome-Wide Association Studies (GWAS)

(6)

6 are being performed for over more than a

decade. GWAS studies already showed many genetic risk loci associated with SCZ. A big GWAS meta-analysis from 2014 executed by the Psychiatric Genomics Consortium (PGC) showed 108 risk loci associated with SCZ [47]. The most recent GWAS meta-analysis from 2018 identified 145 risk loci, of which 93 were also found in the PGC study [48]. Multiple studies show that some of these genetic risk loci are located in glutamate and GABA genes, which might impair functioning of glutamatergic and GABAergic neurotransmission and therefore may contribute to the pathophysiology of SCZ [49], [50]. For some antipsychotics, including CLZ, a few animals studies even showed that the pathological effect of CLZ might be due to DNA demethylation in associated glutamatergic and GABAergic genes [51], [52]. All these findings have led to a growing interest in DNAm in relation to antipsychotics, and increasing number of studies suggest that the clinical effects of some atypical antipsychotics might be due to alterations of DNAm of SCZ associated genes [53]–[55].

A very recent study showed epigenetic aging biomarkers for SCZ using biological clocks including chronological time, mortality, mitotic division, and telomere length, this study also found a unique effect of CLZ. They found that CLZ, at least in men, had a deaccelerating effect on several epigenetic aging clock [56]. Another recent study on SCZ and biological biomarkers derived from DNAm data showed differences between treatment-resistant patients and treatment-responsive patients for blood composition estimates and DNAm derived smoking scores. They also found 93 DMPs associated with psychosis and 1048 DMPs associated with SCZ, and seven DMPs

associated specifically with TRS [57]. Both studies demonstrate that using DNAm status and biomarkers derived from this DNAm can give more insight into the epigenetic profiles of patients with SCZ and might help to predict treatment response, including clinical significance and side effects. In respect to CLZ, a recent EWAS study (admittedly low-powered) found altered DNAm status of specific sites within SCZ related gene regions. Genome wide, they found that 13.052 CpG sites were increased in DNAm and 16.082 were decreased in DNAm status after initiation of CLZ treatment [58].

The aim of this study is to gain more insight in the possible epigenetic mechanisms of CLZ which might explain the beneficial clinical effects of CLZ for SCZ patients. Therefore this study examined the DNAm status of SCZ patients before and after initiation of CLZ treatment. DMPs were then compared to earlier findings and Gene Ontology (GO) was used to investigate their role in biological processes in order to say something about their possible involvement in the beneficial effects of CLZ. This study also examined DMPs between responders and non-responders to CLZ over time. The DMPs found in these analyses were also compared to earlier findings and GO was used to give insight in biological processes of the DMP related genes. This has been done to investigate whether different epigenetic findings could explain the differences in phenotype and clinical response. Thus, the main goal of this study was to examine if DNAm status could explain the relationship between CLZ treatment and its clinical effects in SCZ.

(7)

7

Materials & Methods

Data obtainment and quality control

The blood samples and additional patient information originating from questionnaires were obtained from trainees and researches participating in the CLOZIN study group in the Netherlands. DNA extraction from the blood samples and microarrays were executed by Rob Flynn from The University of Exeter Medical School in England. Quality control of the DNAm data was executed by Grant Neilson from The University of Exeter Medical School in England.

Participants

Eleven patients with SCZ were recruited via pharmacies in the Netherlands (via UPPER: facilitates research groups operating in the field of Pharmacology and belongs to the department of Pharmacoepidemiology and Clinical Pharmacology from University of Utrecht), directly via UMC Utrecht (department of Psychiatry) or voluntarily signed up to participate in the CLOZIN study.

Of all participants (mean age: 36.6 +/- 14.4 years) nine were male and two were female. Most participants had origins in North-West Europe, three participants had origins elsewhere (Burundi, Curacao, and Turkey). Eight participants had been diagnosed with SCZ and three participants had been diagnosed with SAD. The diagnosis of SCZ and SAD was determined following DSM-V criteria by at least one psychiatrist. The participants which were included had insufficient clinical response to a minimal of one trial with any given antipsychotic agent other than CLZ, or had unwanted side effects. Most patients had used several different antipsychotic agents before starting with CLZ. Peripheral blood was collected from the participants at three different moments in

time: before introduction to CLZ, after four to twelve weeks and after six months of treatment with CLZ. The patients had a mean prescribed dose of CLZ of 200.8 +/- 117.6 mg/day with a duration of six months. Psychotic symptoms of the participants were evaluated using PANSS scores, these scores were determined at the time of blood collection. The mean PANSS score was 79.3 at baseline, 64.6 at the first measurement and 64.6 at the second measurement.

Defining responders and non-responders

Response to CLZ treatment was determined using the PANSS scores. The reduction cut-off threshold for response was set at 20% reduction of the PANSS score. The reduction was determined using the average score of the first and second PANSS score after the start of treatment (T1 and T2), this average score was compared to the baseline score (before treatment or T0). Thus, patients with a reduction of more than 20% were considered responders, patients with a reduction of less than 20% were considered non-responders. Five patients were classified as responders and six patients were classified as non-responders to CLZ treatment.

DNA methylation Methods

Genomic DNA was extracted from the blood using the QIAamp DNA blood mini Kit (Qiagen). Bisulfate conversion of 500 ng genomic DNA was performed using the EZ DNAm kit (Zymo Research). The converted DNA was then analysed using the Infinium Methylation EPIC array (Illumina), which provides great coverage, reliability, and reproducibility [59], [60]. The methylation signals of more than 850.000 CpG sites were measured.

Data processing

(8)

8 the ratio of methylation signals from each

probe compared to the sum of both methylated and unmethylated (respectively, the highest and lowest fluorescence intensity) probe signals. Giving each probe a DNAm status or “beta-value” between 0 (unmethylated) and 1 (methylated).

Quality control for the selection of CpG sites for data analysis was executed with the “wateRmelon” package in R [63] using the following steps and criteria: (1) methylated and unmethylated signal intensities were checked (0 samples failed), (2) bisulphite conversion statistics were determined (>80% conversion, 0 samples failed), (3) sex check (predicted sex matches reported sex, 0 samples failed), (4) genotype check and genetic correlation (DNAm data matches genotype and patients are unrelated, 0 samples failed), (5) filtering probes on beadcounts and detection values (samples removed if; detection p-value > 0.05 in 1% of sites, 0 samples failed, sites removed if; beadcounts < 3 in 5% of samples and/or detection p-value > 0.05 in 1% of samples, 11.414 probes failed/removed), (6) outliers were filtered (0 samples failed), (7) the beta-values of the remaining probes were then normalised (samples with high normalisation violence were removed, 0 samples failed), (8) at last probes containing common SNPs and probes with non-specific binding were removed from the samples to prevent artifacts. The final dataset of the normalised beta values included 36 samples with each 802.579 sites.

Statistical analyses

Statistical analyses were performed following a workflow inspired by a previous paper from Maksimovic et al. [63]. Beta-values were converted into “M-values” using logistic transformation, which were used for

statistical analysis. It is shown that M-values are more statistically valid for analysis of differential DNAm status, beta-values are more suitable for biological interpretation and reporting results of DNAm [61]. Linear regression was performed using the “Limma” package [64] in R studio (version 1.2.5042)to assess DNAm data and CLZ treatment. An false discovery rate (FDR) correction was done for multiple comparison, a FDR cut-off less than 0.1 was considered significant. DNA annotation of the CpG sites was done using the “IlluminaHumanMethylationEPICmanifest” package in R studio (version 1.2.5042) and used for GO analysis. This GO analysis was performed using the functional profiler tool provided by g:Profiler (version: e99_eg46_p14_f929183), only ontology terms of GO biological processes and Reactome pathways were used as recourse [65].

Results

Two statistical analyses were executed, differential DNAm was investigated for all participants over time and differential DNAm between responders and non-responders over time.

DNA methylation changes after initiation

of CLZ treatment

Differential methylation analysis of different timepoints during CLZ treatment only showed one significant DMP (Q-value: 0.08), for this DMP DNAm had decreased at the first measurement after starting CLZ treatment (T1) compared to the baseline measurement before the start of CLZ treatment (T0). The location of this CpG site relative to the CpG content of the gene was classified as being a CpG Island (CpGI) shore. This DMP was annotated to the gene UGCG and was located within the promoter region (see Table 1). DNAm status (beta-value) of

(9)

9 this CpG site for each moment of

measurement is shown in Figure 2. No significant changes were found between other timepoints for CLZ treatment (i.e., baseline compared to the second measurement (T0 vs. T2) and the first compared to the second measurement (T1 vs. T2)).

DNA methylation differences between

responders and non-responders

Statistical analysis of differential DNAm changes for responders compared to non-responders over time did not reveal any significant results. Analysis of DNAm changes within the non-responders and responders group also did not reveal any significantly different changes during and after CLZ treatment.

However, analysing DNAm differences for responders compared to non-responders at all three moments of measurement individually did reveal significant DMPs for each moment. Comparison of non-responders with responders at baseline measurement (T0) found six significant DMPs, two CpG sites with higher and four with lower average methylation status for non-responders. Locations relative to genes were classified as within promoter for two CpGs, as 5’UTR for two CpGs, as 3’UTR for one CpG, and one CpG was unclassified. Classification of the CpG locations relative to the CpG content in the genes showed that four CpGs were located in a CpGI, one in a CpGI shore, and one in an open sea. A top table of DMPs found at baseline measurement is shown in Table 2 (DMPs with a FDR of less than 0.2 are shown).

At the first measurement after initiation of CLZ treatment, two significant DMPs were found between responders and

non-responders, both with higher average DNAm status for non-responders. One gene was located in a promoter and one CpG was located in a 3’UTR region. Classification of the two CpG locations relative to the CpG content in the genes showed that one was located in a CpGI and one in a CpGI shore. A top table of DMPs found in this analysis are shown in Table 3 (DMPs with a FDR less than 0.2 are shown).

At second measurement after initiation of CLZ treatment one significant DMP was found between responders and non-responders, this CpG had a higher DNAm status for non-responders. This CpG was found in the promoter region relative to its related genes. Classification of the CpG locations relative to the CpG content in the gene showed that it was located in a CpGI Shore. A top table of DMPs found in this analysis is shown in Table 4 (DMPs with a FDR less than 0.2 are shown). DNAm (beta-values) of the most significant CpG sites found in this analysis are shown in Figure 3.

Interestingly, there were multiple top ranked DMPs found which were annotated to the same gene. Comparison of responders and non-responders at baseline measurement revealed three CpG sites related to the RP5-1029F21.3 gene (Q-values were respectively: 0.06, 0.16 and 0.16), two sites related to the HNRNPH1 gene (Q-values were respectively: 0.02 and 0.09) and two sites related to the ACTN1 gene (Q-values were respectively: 0.06 and 0.16). Comparison at the second measurement revealed three CpG sites that were also related to the RP5-1029F21.3 gene (Q-values were respectively: 0.12, 0.19 and 0.19). Two of these CpG sites were found in both baseline and second measurement comparison between responders and non-responders. The most interesting finding

(10)

10 was a DMP which was found significantly

different between responders and non-responders for all three moments of measurement. For this CpG site (cg03847932) a strongly significant difference was found at all three moments of measurement (Q-values were respectively: 0.002, 0.005 and 0.001), this CpG site was annotated to the PPP4R2 gene.

Gene ontology of top ranked DMPs

The UGCG or ceramide glucosyltransferase gene encodes for a transferase protein that belongs to the protein class of glycosyltransferases. Glycosyltransferases are enzymes that catalyse the transfer process of monosaccharides from a sugar nucleotide derivatives to a sugar or amino acid receptor.

Most GO terms resulting from GO analysis for UGCG (the only annotated gene resulting from the first statistical analysis) were related to “metabolic processes”, such as “digestion” and “water homeostasis”. Pathway analysis (GO/Reactome terms) also revealed terms related to the “leptin-mediated signalling pathway”, “metabolism of glycosphingolipids” and “metabolism of sphingolipids” (FDR: Q-value < 0.05).

The PPP4R2 gene encodes for a serine/threonine-protein phosphatase 4 regulatory subunit 2 and has a function in protein dephosphorylation, removing one or more phosphoric residues from a protein. The other gene associated with one of the top ranked DMPs was the HNRNPH1 gene, which encodes for heterogeneous nuclear ribonucleoprotein H. This protein belongs to the protein class of RNA splicing factors. This protein forms complexes with other HNRNP proteins which provide substrate pre-mRNA processing into translational mRNA. The ACTN1 gene encodes for alpha actinin, which is a crosslinker between actin filaments, it

acts directly to connect the actin cytoskeleton to integrins on the plasma membrane. The top ranked RP5-1029F21.3 seemed to be a non-coding gene, the gene was therefore not found in the GO database.

For the top ranked genes resulting from the second statistical analysis, the following GO terms were found. PPP4R2 and HNRNPH1 both showed GO terms related to “RNA processing” and “mRNA processing” (FDR: Q-value < 0.05). More specifically, PPP4R2 seems to have a role in “DNA Double Strand Break Repair” and “Homology Directed Repair” (Reactome terms), and HNRNPH1 has a role in “mRNA splicing” and “FGFR2 alternative splicing (and signalling)” (Reactome terms). GO analysis also showed GO terms for ACTN1 related to “regulation of cytoskeletal remodelling”, “nephrin family interactions” and “Cell-Cell communication” (Reactome terms).

Discussion

This study tried to gain more insight into the working mechanisms of CLZ. Therefore, DNAm data of SCZ patients using CLZ over a time span of 6 months were analysed. The results only showed one significant DMP in all participants after initiation of CLZ treatment. The UGCG protein is involved in the metabolism of sphingolipids, responsible for the conversion of ceramide into glucosylceramide. Glucosylceramide is a precursor of the globosides and gangliosides, which are both forms of sphingolipids. Disruption in the metabolism of these sphingolipids has been associated with SCZ and it might have a major role in the pathology of SCZ [66]. Disruption of sphingolipid metabolism is also linked to other human diseases, mostly degradation disorders [67]. An animal study showed that mice with UGCG knock-out specifically in

(11)

11 neural cells showed less dendritic complexity

and early pruning, they also found down-regulation of the expression of genes associated with brain development [68]. Alteration in expression of the UGCG gene is not directly associated with SCZ, but several other genes involved in sphingolipid metabolism are [68]. However, in respect to the results, it might be possible that the alteration of UGCG gene initiated by CLZ treatment could partly counteract the disruption in sphingolipid metabolism and thereby contribute to the clinical effects. The results of GO analysis show that the annotated gene UGCG is associated with some important metabolic processes, including the reaction to leptin. It has also been found that sphingolipid metabolism plays a role in several metabolic diseases, such as obesity and insulin resistance [69], [70]. Presumably, altered expression of the UGCG gene initiated by CLZ treatment changes sphingolipid metabolism in such a way that it can lead to clinical improvement of patients with SCZ on the one hand, but causes some metabolic side effects on the other.

Since CLZ is mostly prescribed to SCZ patients who have an insufficient response to other antipsychotics most of the patients can be classified as non-responsive or even as TRS. Therefore, this study also distinguished these types of patients and examined the possible different effects of CLZ on DNAm for these groups. No result was found which indicates that CLZ has a different effect on DNAm for responsive patients compared to non-responsive patients. The results did reveal differences in DNAm status between responsive and non-responsive patients at certain timepoints during the CLZ treatment. For one CpG site the different DNAm status was even consistent for all three moments of

measurement. The annotated gene to this DMP was PPP4R2. In one study, PPP4R2 was shown to be relevant to the DISC1 (Disrupted-In-SCZ-1) pathway, this study also found genetic evidence supporting that the DISC1 pathway has a role in a spectrum of major mental illnesses, including SCZ and BPD [71]. The DISC1 gene is involved in different processes which are associated with neural development and brain maturation. Interestingly, the ACTN1 gene, which was annotated to multiple top ranked DMPs, also seems to interact with the DISC1 pathway [72]. Another study demonstrated that the molecular basis for SCZ changes from early to chronic stage, in their results the PPP4R2 gene was found differentially expressed for the intermediate stage of SCZ (based on duration of illness) [73]. In respect to the results of this study, differences in DNAm for these genes might be important to the pathophysiology of SCZ since they interacts with SCZ associated genes and pathways.

The top table also revealed two DMPs annotated to the HNRNPH1 gene, one was significant (Q-value < 0.05), at baseline measurement. A study from 2015 showed that HNRNPH1 is a quantitative trait gene for methamphetamine sensitivity, showing a contribution to neurobehavioral dysfunction

associated to dopaminergic

neurotransmission [74]. No GWAS studies have found an association of HNRNPH1 variants to any disease. However, some significant SNPs are found to be associated with SCZ, bipolar disease, and major depression, which can modulate splicing and binding affinity of HNRNPH1 [75]. The altered functioning of HNRNPH1 by these SNPs can lead to alternative splicing of genes such as DRD2 (encoding for dopamine receptor D2), the main antagonized target of effective antipsychotics. In addition to DRD2, there are

(12)

12 some suspectable genes for SCZ that have

been found to show abnormal splicing patterns, including the above mentioned DISC1 gene [76]. In respect to the results, a difference in DNAm of the HNRNPH1 gene between responders and non-responders could explain the variance in clinical response to antipsychotic agents such as CLZ. This could even have implications for understanding the epigenetic basis of treatment-resistant SCZ.

Limitations

This study has several limitations that have to be noted. (1) This study had a very small sample size, so replication studies with larger sample sizes are needed. (2) Participants who were selected for this study were treated with several antipsychotics before CLZ treatment, this might give a disrupted view about the real effect of CLZ on DNAm. (3) The daily prescribed dosage of CLZ was not uniform for all participants and the true CLZ administration was unknown. Meaning that the effect of CLZ on DNAm could strongly differ between participants. (4) Classification of treatment responsiveness was determined by only using the PANSS scores, classification based on more than one scale is preferred. (5) The DNAm changes were not normalised,

because control data of healthy subjects was absent. This means that DNAm differences unrelated to CLZ or SCZ could have been measured. (6) The relationship between DNAm and gene expression was not examined. DNAm differences do not directly indicate changes in molecular mechanisms and further examination is required. (7) DNA was extracted from the blood, therefore results only give an indication of DNAm in other cell tissues. Thus, similar studies using brain tissue are needed for further interpretation of the results.

Conclusion

The results of this study indicate that CLZ might affect DNAm of genes associated with sphingolipid metabolism. In addition, they indicate that the epigenetic profiles of responsive and non-responsive SCZ patients might be different. The findings of this study could have implications for understanding the working mechanisms of antipsychotics such as CLZ and for the development of novel therapeutics for treating SCZ and SCZ-like disorders. Further functional studies are required to clarify the molecular mechanisms of CLZ and the epigenetic differences between SCZ patients.

(13)

13

Tables and Figures

Figure 1. A: Illustration of functional regions relative to gene. TSS200 (promoter) is the region from transcription start site (TSS) to 200 nucleotides upstream of TSS; TSS1500 (promoter) is the region from 200 to 1500 nucleotides upstream of TSS; 5’UTR, 1st_{exon, gene body and 3’UTR are} also functional regions. B: Illustration of CpG regions. The 2 kb regions directly upstream and downstream of CpG island are “CPG island shores” and the 2kb regions upstream and

downstream of CpG island shores are called “CpG island shelves”. CpG sites located outside one of these regions were classified as “open seas”. This figure was adopted from the paper “High density DNAm array with single CpG site resolution” by Bibikova et al. (2011) .

(14)

14 Figure 3: DNAm status (beta values) of the most significantly differentially methylated CpGs between responders and non-responders at corresponding time points. On the horizontal axis “T” indicates the moment of

measurement (before CLZ treatment (0), 4-12 weeks after initiation CLZ treatment (1) and 6 months after initiation CLZ treatment (2)) and “R” indicates responsiveness to CLZ (non-responsive (0) and responsive (1)). A: CpG in the promoter of PPP4R2 gene which is found a significant DMP between responders and non-responders for all three time points. B: CpG in the 5’UTR site of HNRNPH1 gene which is found a significant DMP between responders and non-responders only at baseline measurement.

Figure 2: DNAm status (beta values) of the only significantly changed CpG site after initiation of CLZ treatment. On the horizontal axis “T” indicates the moment of

measurement (before CLZ treatment (0), 4-12 weeks after initiation CLZ treatment (1) and 6 months after initiation CLZ treatment (2)). CpG site in the promoter (TSS1500) of UGCG gene which is significantly different after CLZ treatment (T1) in all patients.

(15)

15

Probe Chromosome Position Relation to Island Relative to gene region Illumina Gene Annotation LogFC P-value FDR (Q-value)

cg27510203 9 114658327 N-shore TSS1500 UGCG -0.63 9.59E-08 0.08

Table 1: DMPs between baseline (T=0) and 4-12 weeks after initiation of CLZ treatment (T=1). LogFC is the log fold change of the M-values. FDR is the false discovery rate (or Q-value), FDR values that are shown in bold are considered significant.

(16)

16

Probe Chromosome Position Relation to Island Relative to gene regions Illumina Gene Annotation logFC P-value FDR (Q-value) cg03847932 3 73045556 N-shore TSS1500 PPP4R2 -1.68 2.21E-09 0.002

cg24585418 5 179059426 Island 5’UTR HNRNPH1 -1.34 4.94E-08 0.02

cg04850148 17 34539744 Open sea NA NA 2.29 3.86E-07 0.06

cg01707795 14 69341738 Island 3’ UTR ACTN1 -2.07 3.93E-07 0.06

cg04370829 17 406249 Island TSS200 RP5-1029F21.3 1.20 3.94E-07 0.06

cg26135573 5 179059668 Island 5’UTR HNRNPH1 -1.08 6.73E-07 0.09

cg05845997 17 76657659 Open sea NA NA -0.69 1.82E-06 0.16

cg08861434 13 112062652 Island NA NA -1.82 2.21E-06 0.16

cg05256999 16 88902466 Island 3’UTR GALNS -0.71 2.24E-06 0.16

cg20547929 17 4036387 Island TSS200 RP5-1029F21.3 1.44 2.24E-06 0.16

cg18057887 11 133800913 Island NA NA -1.96 2.48E-06 0.16

cg18727120 17 406137 N-shore TSS1500 RP5-1029F21.3 1.52 2.53E-06 0.16

Cg14065121 9 77643271 Island 5’UTR C9orf41 0.84 2.82E-06 0.16

Cg07176949 2 132093109 Open sea NA NA -0.74 2.83E-06 0.16

Cg27036347 14 69341603 Island 3’UTR ACTN1 -1.72 2.92E-06 0.16

Cg19586483 6 1105152070 Open sea NA NA -2.06 3.20E-06 0.16

Cg24969716 11 45407537 Open sea 5’UTR RP11-430H10.2 -1.98 3.54E-06 0.17

Cg26580609 2 18766140 Island TSS1500;

3’UTR

U6; NT5C1B -0.86 3.78E-06 0.17

Cg14788749 22 35937309 Island 5’UTR RASD2 0.95 4.52E-06 0.19

Table 2: DMPs between responders and non-responders at baseline measurement (T=0). LogFC is the log fold change of the M-values. FDR is the false discovery rate (or Q-value), FDR values that are shown in bold are considered significant.

(17)

17

Probe Chromosome Position Relation to Island Relative to gene region Illumina Gene Annotation logFC P-value FDR (Q-value) cg03847932 3 73045556 N-shore TSS1500 PPP4R2 -1.59 7.28E-09 0.005

cg01707795 14 69341738 Island 3’UTR ACTN1 -2.18 1.55E-07 0.06

Table 3: DMPs between responders and non-responders at 4-12 weeks after initiation of CLZ treatment (T=1). LogFC is the log fold change of the M-values. FDR is the false discovery rate (or Q-value), FDR values that are shown in bold are considered significant.

(18)

18

Probe Chromosome Position Relation to Island Relative to gene region Illumina Gene Annotation LogFC P-value FDR (Q-value) cg03847932 3 73045556 N-shore TSS1500 PPP4R2 -1.71 1.62E-09 0.001 cg04370829 17 406249 Island TSS200 RP5-1029F21.3 1.18 5.49E-07 0.12

cg08306218 15 56299382 Open sea TSS200 CNOT6LP1 -1.33 5.90E-07 0.12

cg01707795 14 69341738 Island 3’UTR NA -2.02 6.13E-07 0.12

cg24585418 5 179059426 Island 5’UTR NA -1.14 9.37E-07 0.15

cg27036347 14 69341603 Island 3’UTR NA -1.81 1.26E-06 0.16

cg04850148 17 34539744 Open sea NA NA 2.13 1.43E-06 0.16

cg15158859 10 133978304 Open sea 5’UTR JAKMIP3 -0.67 1.89E-06 0.18

cg02873991 12 25151263 Open sea TSS1500 C12orf77 -0.84 2.28E-06 0.19

cg07176949 2 132093109 Open sea NA NA -0.74 2.67E-06 0.19

cg19586483 6 110512070 Open sea NA NA -2.08 2.77E-06 0.19

cg18727120 17 406137 N-Shore TSS1500 RP5-1029F21.3 1.51 3.05E-06 0.19

cg01214346 17 406501 Island 5’UTR RP5-1029F21.3 1.47 3.20E-06 0.19

Table 4: DMPs between responders and non-responders at 6 months after initiation of CLZ treatment (T2). LogFC is the log fold change of the M-values. FDR is the false discovery rate (or Q-value), FDR values that are shown in bold are considered significant.

(19)

19

References

[1] W. Gaebel and J. Zielasek, “SCZ in 2020: Trends in diagnosis and therapy,” Psychiatry and Clinical

Neurosciences. 2015, doi: 10.1111/pcn.12322.

[2] D. Malaspina et al., “Schizoaffective Disorder in the DSM-5,” Schizophr. Res., 2013, doi:

10.1016/j.schres.2013.04.026.

[3] H. Y. Meltzer, “Treatment-resistant SCZ - The role of CLZ,” Current Medical Research and Opinion.

1997, doi: 10.1185/03007999709113338.

[4] M. T. Samara, A. Nikolakopoulou, G. Salanti, and S. Leucht, “How many patients with SCZ do not

respond to antipsychotic drugs in the short term? An analysis based on individual patient data from

randomized controlled trials,” Schizophr. Bull., 2019, doi: 10.1093/schbul/sby095.

[5] T. Suzuki et al., “Defining treatment-resistant SCZ and response to antipsychotics: A review and

recommendation,” Psychiatry Res., 2012, doi: 10.1016/j.psychres.2012.02.013.

[6] A. F. Lehman et al., “Practice Guideline for the Treatment of Partients with SCZ, Second Edition,”

American Journal of Psychiatry. 2004, doi: 10.1176/appi.books.9780890423363.45859.

[7] S. Leucht, J. M. Kane, W. Kissling, J. Hamann, E. Etschel, and R. R. Engel, “What does the PANSS

mean?,” Schizophr. Res., 2005, doi: 10.1016/j.schres.2005.04.008.

[8] S. Leucht et al., “Comparative efficacy and tolerability of 15 antipsychotic drugs in schizophrenia: A

multiple-treatments meta-analysis,” Lancet, 2013, doi: 10.1016/S0140-6736(13)60733-3.

[9] L. J. Friedman et al., “CLZ — a novel antipsychotic agent,” New England Journal of Medicine. 1991,

doi: 10.1056/NEJM199108153250716.

[10] D. D. Miller, “Review and management of CLZ side effects,” Journal of Clinical Psychiatry. 2000.

[11] D. C. Henderson et al., “CLZ, diabetes mellitus, weight gain, and lipid abnormalities: A five-year

naturalistic study,” Am. J. Psychiatry, 2000, doi: 10.1176/appi.ajp.157.6.975.

[12] D. C. Henderson et al., “CLZ, diabetes mellitus, hyperlipidemia, and cardiovascular risks and mortality:

Results of a 10-year naturalistic study,” J. Clin. Psychiatry, 2005, doi: 10.4088/JCP.v66n0905.

[13] D. Taylor, L. Shapland, G. Laverick, J. Bond, and J. Munro, “CLZ - A survey of patient perceptions,”

Psychiatr. Bull., 2000, doi: 10.1192/pb.24.12.450.

[14] O. Devinsky, G. Honigfeld, and J. Patin, “CLZ-related seizures,” Neurology, 1991, doi:

10.1212/wnl.41.3.369.

[15] S. V. Pacia and O. Devinsky, “CLZ–related seizures: Experience with 5, 629 patients,” Neurology,

1994, doi: 10.1212/wnl.44.12.2247.

[16] J. M. J. Alvir, J. A. Lieberman, A. Z. Safferman, J. L. Schwimmer, and J. A. Schaaf, “CLZ-Induced

Agranulocytosis – Incidence and Risk Factors in the United States,” N. Engl. J. Med., 1993, doi:

10.1056/NEJM199307153290303.

[17] D. Cohen, J. P. A. M. Bogers, D. Van Dijk, B. Bakker, and P. F. J. Schulte, “Beyond white blood cell

(20)

20

10.4088/JCP.11r06977.

[18] P. F. J. Schulte, “Risk of CLZ-associated agranulocytosis and mandatory white blood cell monitoring,”

Annals of Pharmacotherapy. 2006, doi: 10.1345/aph.1G396.

[19] J. Kane, “CLZ for the Treatment-Resistant Schizophrenic,” Arch. Gen. Psychiatry, 1988, doi:

10.1001/archpsyc.1988.01800330013001.

[20] K. Wahlbeck, M. Cheine, A. Essali, and C. Adams, “Evidence of CLZ’s effectiveness in SCZ: A

systematic review and meta-analysis of randomized trials,” American Journal of Psychiatry. 1999, doi:

10.1176/ajp.156.7.990.

[21] J. McEvoy, “Effectiveness of CLZ Versus Olanzapine, Quetiapine, and Risperidone in Patients With

Chronic SCZ Who Did Not Respond to Prior Atypical Antipsychotic Treatment,” Am. J. Psychiatry,

2006, doi: 10.1176/appi.ajp.163.4.600.

[22] J. P. A. M. Bogers, P. F. J. Schulte, D. Van Dijk, B. Bakker, and D. Cohen, “CLZ underutilization in the

treatment of SCZ: How can CLZ prescription rates be improved?,” Journal of Clinical

Psychopharmacology. 2016, doi: 10.1097/JCP.0000000000000478.

[23] H. Y. Meltzer, “CLZ: Balancing safety with superior antipsychotic efficacy,” Clinical SCZ and Related

Psychoses. 2012, doi: 10.3371/CSRP.6.3.5.

[24] Z. D. Smith and A. Meissner, “DNA methylation: Roles in mammalian development,” Nature Reviews

Genetics. 2013, doi: 10.1038/nrg3354.

[25] R. Lister et al., “Global epigenomic reconfiguration during mammalian brain development,” Science

(80-. )., 2013, doi: 10.1126/science.1237905.

[26] Z. Wang, B. Tang, Y. He, and P. Jin, “DNAm dynamics in neurogenesis,” Epigenomics. 2016, doi:

10.2217/epi.15.119.

[27] E. M. Jobe and X. Zhao, “DNA methylation and Adult Neurogenesis,” Brain Plast., vol. 3, no. 1, pp. 5–

26, 2016, doi: 10.3233/bpl-160034.

[28] P. Tognini, D. Napoli, and T. Pizzorusso, “Dynamic DNA methylation in the brain: A new epigenetic

mark for experience-dependent plasticity,” Front. Cell. Neurosci., vol. 9, no. AUGUST, pp. 1–11, 2015,

doi: 10.3389/fncel.2015.00331.

[29] M. Erhlich et al., “Amount and distribution of 5-methylcytosine in human DNA from different types of

tissues or cells,” Nucleic Acids Res., 1982, doi: 10.1093/nar/10.8.2709.

[30] A. Bird, “DNA methylation patterns and epigenetic memory,” Genes and Development. 2002, doi:

10.1101/gad.947102.

[31] M. Gardiner-Garden and M. Frommer, “CpG Islands in vertebrate genomes,” J. Mol. Biol., 1987, doi:

10.1016/0022-2836(87)90689-9.

[32] A. M. Deaton and A. Bird, “CpG islands and the regulation of transcription,” Genes Dev., 2011, doi:

10.1101/gad.2037511.

[33] R. A. Irizarry et al., “The human colon cancer methylome shows similar hypo- and hypermethylation

(21)

21

[34] M. Bibikova et al., “High density DNA methylation array with single CpG site resolution,” Genomics,

2011, doi: 10.1016/j.ygeno.2011.07.007.

[35] J. Sandoval et al., “Validation of a DNA methylation microarray for 450,000 CpG sites in the human

genome,” Epigenetics, 2011, doi: 10.4161/epi.6.6.16196.

[36] R. Pidsley et al., “Critical evaluation of the Illumina MethylationEPIC BeadChip microarray for

whole-genome DNA methylation profiling,” Genome Biol., 2016, doi: 10.1186/s13059-016-1066-1.

[37] V. K. Rakyan, T. A. Down, D. J. Balding, and S. Beck, “Epigenome-wide association studies for

common human diseases,” Nature Reviews Genetics. 2011, doi: 10.1038/nrg3000.

[38] P. W. Laird, “The power and the promise of DNA methylation markers,” Nature Reviews Cancer.

2003, doi: 10.1038/nrc1045.

[39] K. D. Robertson, “DNA methylation and human disease,” Nature Reviews Genetics. 2005, doi:

10.1038/nrg1655.

[40] A. Berson, R. Nativio, S. L. Berger, and N. M. Bonini, “Epigenetic Regulation in Neurodegenerative

Diseases,” Trends Neurosci., vol. 41, no. 9, pp. 587–598, 2018, doi: 10.1016/j.tins.2018.05.005.

[41] M. Kinoshita et al., “DNA methylation signatures of peripheral leukocytes in SCZ,” NeuroMolecular

Med., vol. 15, no. 1, pp. 95–101, 2013, doi: 10.1007/s12017-012-8198-6.

[42] L. F. Wockner et al., “Genome-wide DNA methylation analysis of human brain tissue from SCZ

patients,” Transl. Psychiatry, vol. 4, no. September 2013, pp. 1–8, 2014, doi: 10.1038/tp.2013.111.

[43] M. Kinoshita et al., “Aberrant DNA methylation of Blood in SCZ by Adjusting for Estimated Cellular

Proportions,” NeuroMolecular Med., vol. 16, no. 4, pp. 697–703, 2014, doi:

10.1007/s12017-014-8319-5.

[44] N. Teroganova, L. Girshkin, C. M. Suter, and M. J. Green, “DNA methylation in peripheral tissue of

SCZ and bipolar disorder: A systematic review,” BMC Genet., 2016, doi:

10.1186/s12863-016-0332-2.

[45] C. Montano et al., “Association of DNA methylation differences with SCZ in an epigenome-wide

association study,” JAMA Psychiatry, vol. 73, no. 5, pp. 506–514, 2016, doi:

10.1001/jamapsychiatry.2016.0144.

[46] E. Dong, M. Nelson, D. R. Grayson, E. Costa, and A. Guidotti, “CLZ and sulpiride but not haloperidol

or olanzapine activate brain DNA demethylation,” Proc. Natl. Acad. Sci. U. S. A., vol. 105, no. 36, pp.

13614–13619, 2008, doi: 10.1073/pnas.0805493105.

[47] S. Ripke et al., “Biological insights from 108 SCZ-associated genetic loci,” Nature, 2014, doi:

10.1038/nature13595.

[48] A. F. Pardiñas et al., “Common SCZ alleles are enriched in mutation-intolerant genes and in regions

under strong background selection,” Nat. Genet., 2018, doi: 10.1038/s41588-018-0059-2.

[49] J. T. Coyle, “Glutamate and SCZ: beyond the dopamine hypothesis.,” Cellular and molecular

neurobiology. 2006, doi: 10.1007/s10571-006-9062-8.

(22)

22

glutamate, GABA and related genes in SCZ and bipolar disorder: A decade of advance,”

Neuroscience and Biobehavioral Reviews. 2010, doi: 10.1016/j.neubiorev.2010.01.002.

[51] E. Dong, M. Nelson, D. R. Grayson, E. Costa, and A. Guidotti, “CLZ and sulpiride but not haloperidol

or olanzapine activate brain DNA demethylation,” Proc. Natl. Acad. Sci. U. S. A., 2008, doi:

10.1073/pnas.0805493105.

[52] E. Dong, P. Tueting, F. Matrisciano, D. R. Grayson, and A. Guidotti, “Behavioral and molecular

neuroepigenetic alterations in prenatally stressed mice: relevance for the study of chromatin

remodeling properties of antipsychotic drugs,” Transl. Psychiatry, 2016, doi: 10.1038/tp.2015.191.

[53] Y. Murata, M. Nishioka, M. Bundo, F. Sunaga, K. Kasai, and K. Iwamoto, “Comprehensive DNA

methylation analysis of human neuroblastoma cells treated with blonanserin,” Neurosci. Lett., 2014,

doi: 10.1016/j.neulet.2014.01.038.

[54] M. G. Melka et al., “The effects of olanzapine on genome-wide DNA methylation in the hippocampus

and cerebellum,” Clin. Epigenetics, 2014, doi: 10.1186/1868-7083-6-1.

[55] M. Kinoshita et al., “Effect of CLZ on DNA methylation in peripheral leukocytes from patients with

treatment-resistant SCZ,” Int. J. Mol. Sci., vol. 18, no. 3, 2017, doi: 10.3390/ijms18030632.

[56] A. T. Higgins-Chen, M. P. Boks, C. H. Vinkers, R. S. Kahn, and M. E. Levine, “SCZ and Epigenetic

Aging Biomarkers: Increased Mortality, Reduced Cancer Risk, and Unique CLZ Effects,” Biol.

Psychiatry, 2020, doi: 10.1016/j.biopsych.2020.01.025.

[57] E. Hannon et al., “Large-scale analysis of DNA methylation identifies cellular alterations in blood from

psychosis patients and molecular biomarkers of treatment-resistant SCZ,” bioRxiv, p.

2020.04.05.026211, Jan. 2020, doi: 10.1101/2020.04.05.026211.

[58] M. Kinoshita et al., “Effect of CLZ on DNA methylation in peripheral leukocytes from patients with

treatment-resistant SCZ,” Int. J. Mol. Sci., 2017, doi: 10.3390/ijms18030632.

[59] S. Moran, C. Arribas, and M. Esteller, “Validation of a DNA methylation microarray for 850,000 CpG

sites of the human genome enriched in enhancer sequences,” Epigenomics, 2016, doi:

10.2217/epi.15.114.

[60] R. Pidsley et al., “Critical evaluation of the Illumina MethylationEPIC BeadChip microarray for

whole-genome DNA methylation profiling,” Genome Biol., vol. 17, no. 1, pp. 1–17, 2016, doi:

10.1186/s13059-016-1066-1.

[61] P. Du et al., “Comparison of Beta-value and M-value methods for quantifying methylation levels by

microarray analysis,” BMC Bioinformatics, 2010, doi: 10.1186/1471-2105-11-587.

[62] R. Pidsley, C. C. Y Wong, M. Volta, K. Lunnon, J. Mill, and L. C. Schalkwyk, “A data-driven approach to

preprocessing Illumina 450K methylation array data,” BMC Genomics, 2013, doi:

10.1186/1471-2164-14-293.

[63] J. Maksimovic, B. Phipson, and A. Oshlack, “A cross-package Bioconductor workflow for analysing

methylation array data,” F1000Research, 2016, doi: 10.12688/f1000research.8839.1.

[64] R. C. Gentleman et al., “Bioconductor: open software development for computational biology and

(23)

23

[65] J. Reimand et al., “g:Profiler-a web server for functional interpretation of gene lists (2016 update),”

Nucleic Acids Res., 2016, doi: 10.1093/nar/gkw199.

[66] S. Narayan, S. R. Head, T. J. Gilmartin, B. Dean, and E. A. Thomas, “Evidence for disruption of

sphingolipid metabolism in SCZ,” J. Neurosci. Res., 2009, doi: 10.1002/jnr.21822.

[67] T. Kolter and K. Sandhoff, “Sphingolipid metabolism diseases,” Biochimica et Biophysica Acta -

Biomembranes. 2006, doi: 10.1016/j.bbamem.2006.05.027.

[68] R. Jennemann et al., “Cell-specific deletion of glucosylceramide synthase in brain leads to severe

neural defects after birth,” Proc. Natl. Acad. Sci. U. S. A., 2005, doi: 10.1073/pnas.0500893102.

[69] K. M. Boini, C. Zhang, M. Xia, J. L. Poklis, and P. L. Li, “Role of sphingolipid mediator ceramide in

obesity and renal injury in mice fed a high-fat diet,” J. Pharmacol. Exp. Ther., 2010, doi:

10.1124/jpet.110.168815.

[70] B. Chaurasia and S. A. Summers, “Ceramides - Lipotoxic Inducers of Metabolic Disorders,” Trends in

Endocrinology and Metabolism. 2015, doi: 10.1016/j.tem.2015.07.006.

[71] W. Hennah and D. Porteous, “The DISC1 pathway modulates expression of neurodevelopmental,

synaptogenic and sensory perception genes,” PLoS One, 2009, doi: 10.1371/journal.pone.0004906.

[72] J. K. Millar, S. Christie, and D. J. Porteous, “Yeast two-hybrid screens implicate DISC1 in brain

development and function,” Biochem. Biophys. Res. Commun., 2003, doi:

10.1016/j.bbrc.2003.10.101.

[73] S. Narayan et al., “Molecular profiles of SCZ in the CNS at different stages of illness,” Brain Res., 2008,

doi: 10.1016/j.brainres.2008.08.023.

[74] N. Yazdani et al., “Hnrnph1 Is A Quantitative Trait Gene for Methamphetamine Sensitivity,” PLoS

Genet., 2015, doi: 10.1371/journal.pgen.1005713.

[75] S. J. Glatt, O. S. Cohen, S. V. Faraone, and M. T. Tsuang, “Dysfunctional gene splicing as a potential

contributor to neuropsychiatric disorders,” Am. J. Med. Genet. Part B Neuropsychiatr. Genet., 2011,

doi: 10.1002/ajmg.b.31181.

[76] K. Nakata et al., “DISC1 splice variants are upregulated in SCZ and associated with risk

(24)