Network Analysis of Gene Expression in Peripheral Blood Identifies Mtor and NF-&Kappa

ORIGINAL ARTICLE Network analysis of gene expression in peripheral blood identiﬁes mTOR and NF-κB pathways involved in antipsychotic-induced extrapyramidal symptoms

S Mas1,2,3, P Gassó1,2, E Parellada1,2,3,4, M Bernardo2,3,4,5 and A Lafuente1,2,3

To identify the candidate genes for pharmacogenetic studies of antipsychotic (AP)-induced extrapyramidal symptoms (EPS), we propose a systems biology analytical approach, based on protein–protein interaction network construction and functional annotation analysis, of changes in gene expression (Human Genome U219 Array Plate) induced by treatment with risperidone or paliperidone in peripheral blood. 12 AP-naïve patients with first-episode psychosis participated in the present study. Our analysis revealed that, in response to AP treatment, constructed networks were enriched for different biological processes in patients without EPS (ubiquitination, protein folding and adenosine triphosphate (ATP) metabolism) compared with those presenting EPS (insulin receptor signaling, lipid modification, regulation of autophagy and immune response). Moreover, the observed differences also involved specific pathways, such as anaphase promoting complex /cdc20, prefoldin/CCT/triC and ATP synthesis in no-EPS patients, and mammalian target of rapamycin and NF-κB kinases in patients with EPS. Our results showing different patterns of gene expression in EPS patients, offer new and valuable markers for pharmacogenetic studies.

The Pharmacogenomics Journal (2015) 15, 452–460; doi:10.1038/tpj.2014.84; published online 27 January 2015

INTRODUCTION intermediate phenotype that is regulated by a combination of 8 Because patients with schizophrenia differ substantially in terms of genetic, epigenetic and environmental factors. Several authors the side effects they experience, the severity of such effects, and have documented the potential utility of blood-based transcrip- also in their clinical response, treatment with antipsychotics (AP) tomic profiling as a source of biomarkers for schizophrenia. needs to be individually tailored to each patient. Genetic factors Although acknowledging that the blood-based approach might are generally assumed to contribute to variable treatment share some of the disadvantages inherent to the use of post- response, and this has led to a number of pharmacogenetic mortem brain tissue, as well as the fact that gene expression is studies being performed. Although treatment response has been only moderately correlated between the two sample types, several observed to correlate with a number of genetic variants, such as authors have reported that blood could act as a ‘surrogate’ for 9 dopamine D2 and D3 receptor variants1,2, there are as yet no underlying pathophysiology in psychiatric disorders. Moreover, definitive predictors of response.3 peripheral blood cells synthesize and express dopamine and also Pharmacogenetics has been driven by a candidate gene serotonin receptors and transporters on their plasma membrane, approach.4 The disadvantage of this approach is that the objects suggesting that they may constitute a cellular tool with which to of study are circumscribed by our current understanding of the monitor the effects of pharmacological treatments.10,11 Although mechanisms by which drugs act, and therefore this method gene association and expression studies have implicated many cannot identify hitherto unsuspected predictor genes. Apart from genes in the pharmacological effect of AP, most results have not their receptor-binding profile, little is known about the underlying been replicated or supported by the meta-analysis. Therefore, one molecular mechanism by which AP act. Since inhibition of research avenue involves the search for dysregulated molecular dopamine D2 receptors (which has been correlated with AP pathways. The advantage of this type of approach is that it has the efficacy and side effects such as extrapyramidal symptoms (EPS)) potential to reconcile poor biomarker reproducibility across should be achieved instantaneously, the beneficial therapeutic studies by identifying common biological pathways or functional effect cannot be limited to a straightforward interaction between modules.12 Moreover, in addition to the success of these methods such receptors and the drug. for the biological interpretation of genomic data, information is One approach that could help to elucidate the molecular also now available on relationships between gene products, signatures of AP treatments involves integrating pharmacoge- especially protein–protein interactions (PPI), and this can be used nomic data with other sources of data, for example, from the to define other types of modules. The use of the interactome studies of gene expression.5–7 Gene expression can bridge the gap allows the identification of sub-networks of interacting proteins between genetic variation and side effect susceptibility as an associated with genomic experiments, and these sub-networks

1Department Pathological Anatomy, Pharmacology and Microbiology, University of Barcelona, Barcelona, Spain; 2Institut d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Barcelona, Spain; 3Centro de Investigación Biomédica en Red de Salud Mental (CIBERSAM), Barcelona, Spain; 4Clinic Schizophrenia program, Psychiatry service, Hospital Clínic de Barcelona, Barcelona, Spain and 5Department Psychiatry and Clinical Psychobiology, University of Barcelona, Barcelona, Spain. Correspondence: Dr A Lafuente, Department Anatomic Pathology, Pharmacology and Microbiology, University of Barcelona, IDIBAPS, CIBERSAM (Centro de Investigación Biomédica en Red de Salud Mental, Spanish Ministry of Health, Instituto Carlos III), Casanova 143, Barcelona E-08036, Spain. E-mail: [email protected] Received 9 July 2014; revised 22 September 2014; accepted 5 November 2014; published online 27 January 2015 mTOR and NF-κB as candidate genes for AP-induced EPS S Mas et al 453 can be considered functional modules.13 Therefore, gene net- Table 1. Demographic and pharmacological characteristics of study works inferred from genomic data could be considered a higher participants level of structural functional modules operating in the cell. Here, we propose a systems biology analytical approach, based All Patients not Patients on PPI network construction and functional annotation analysis, of presenting presenting changes in gene expression induced by treatment with risper- EPS EPS ﬁ idone or paliperidone in peripheral blood of drug-naive rst N1266 psychotic-episode schizophrenia patients, to examine the relation- Gender (male/female) 6/6 3/3 3/3 ship between these changes and the appearance of EPS. Identiﬁed Age (years ± s.d.) 26.08 ± 3.96 26.33 ± 4.27 25.83 ± 4.02 functional modules will provide potential candidate genes for Tobacco use N (%) 8 (66.66) 4 (66.66) 4 (66.66) Cannabis use N (%) 6 (50.00) 3 (50.00) 3 (50.00) future pharmacogenetic studies. Cocaine use N (%) 1 (8.33) 0 (0.00) 1 (16.6)

Antipsychotic MATERIALS AND METHODS Risperidone (N) 633 Dosage (mg day − 1 ± s.d.) 5.83 ± 0.44 5.66 ± 0.57 6.00 ± 0.00 Subjects Paliperidone (N) 633 For the present study, we recruited 18 AP-naïve patients with first-episode Dosage (mg day − 1 ± s.d.) 12.85 ± 2.85 13.00 ± 4.58 12.75 ± 1.50 psychosis, all of them treated with risperidone or paliperidone (the active metabolite of risperidone). Subjects were diagnosed according to DSM-IV Abbreviation: EPS, extrapyramidal symptoms. criteria. For each patient two blood samples were collected in PAXgene Blood RNA tubes (PreAnalytiX Gmbh, Hombrechtikon, Switzerland); one upon admission to the psychiatric unit (prior to the initiation of AP treatment), (untreated status), and a second after 9 days (±1 day) of discovery rate. We also calculated the gene expression fold change. The continuous treatment with risperidone or paliperidone (treated status), or different comparisons made are summarized in Figure 1. as soon as any EPS appeared and prior to starting treatment with anti- parkinsonian drugs. No other concomitant treatments were administered during this time. All the participants were living in Catalonia at the time of PPI Network construction and evaluation the study, and only Caucasians were included. This ethnicity was Genes showing differential expression in response to AP treat- determined by self-reported ancestries; patients reported the ethnicity of ment were used (false discovery rate corrected P-value o0.01, fold each grandparent and we excluded subjects who mentioned non-European change42). The SNOW program, implemented in the Babelomics 4.3 suite, ancestry. Written informed consent was obtained from each subject. The was used for the analysis. SNOW contains a database of PPI from different 14 study was approved by the Ethics Committee of the Hospital Clínic. public repositories, and these were used to generate an interactome. Acute EPS induced by AP medication were assessed using the Simpson– Given a set of gene products, the sub-network defined by them can easily Angus Scale, and were followed up for four weeks. Eight participants be determined by mapping all members onto the interactome. The presented EPS, including acute dystonia (N = 5), and drug-induced minimum connected network (MCN), defined as the shortest network that parkinsonism (muscle rigidity and bradykinesia) (N = 3). Two participants connects all the interacting nodes within a gene list, can be deduced using 15 presenting EPS were removed from the study as they received anti- the Dijkstra algorithm. parkinsonian treatment before the collection of the second blood sample. In order to identify the MCN among our list of gene products we used Of the ten patients without EPS, we selected six participants for matching by the curated interactome available in SNOW, which is formed by the age, sex, smoking, cannabis use, AP and dosage with the remaining six interactome based on PPI validated by at least two independent methods. patients presenting EPS (three acute dystonia and three parkinsonism). We also allowed the inclusion of extra nodes, not included in our list, that Finally, participants in the microarray analysis were 12 AP-naïve patients connect two or more nodes in the list. with first-episode psychosis, six with EPS and six without EPS matched by Network enrichment analysis was performed to test whether the age, sex, smoking, cannabis use, AP and dosage (Table 1). For each patient parameters that described a network were beyond their random expecta- 14 two samples were analyzed with microarray; one corresponding to the tions or not. The parameters analyzed were connection degree (which untreated condition (sample collected prior to the initiation of AP accounts for the number of partners of direct interactions a particular node treatment), and the other corresponding to the treated condition (sample has), clustering coefficient (which accounts for the connectivity of a given collected after 9 days (±1 day of continuous treatment with risperidone or node and also for the connectivity of the neighborhood to which this node paliperidone). is connected) and betweenness centrality (which is related to the existence of hubs connecting different parts of the network).14 The distribution of random expectations of the network parameters of a Sample collection, RNA isolation, and microarray hybridization MCN from N nodes can be constructed by repeating 10 000 times the Total RNA was isolated in accordance with the manufacturer's protocol following steps: N proteins are randomly sampled from the interactome, (PAXgene Blood RNA kit, PreAnalytiX Gmbh). The purity and integrity of and are then mapped to create a MCNrandom whose network parameters RNA was assessed using an Agilent 2100 Bioanalyzer (Agilent Technolo- are calculated. The Kolmogorov–Smirnov test can then be used to gies, Palo Alto, CA, USA), and any degraded samples were discarded. compare the parameter distribution obtained from the problem list with A total of 1 μg of purified RNA from each of the samples was submitted to the corresponding random expectations.15 the Kompetenzzentrum für Fluoreszente Bioanalytik Microarray Technol- ogy (KFB, BioPark Regensburg GmbH, Regensburg, Germany) for labeling and hybridization to Human Genome U219 Array Plate (Affymetrix, Santa Gene set enrichment analysis and visualization Clara, CA, USA), following the manufacturer’s protocols. The Human The MCN constructed with SNOW for the genes differentially expressed in Genome U219 Array Plate comprises more than 530 000 probes covering response to AP treatment (treated vs untreated conditions) for patients over 36 000 transcripts and variants, which in turn represent more than with and without EPS were uploaded jointly into Cytoscape 3.0.2. We then 16 20 000 genes mapped through Unigene (www.ncbi.nlm.nih.gov/unigene) used ClueGO v2.1, a Cytoscape plug-in, to perform a gene set enrichment or via Refseq annotation (http://www.ncbi.nlm.nih.gov/RefSeq/). analysis and to identify the functional modules that were common to both patients with and without EPS, or specific to just one of them. Two databases of functional modules were selected for the enrichment analysis: Microarray data analysis the unstructured terms of biological processes from Gene Ontology Microarray analysis was performed using the Babelomics 4.3 suite (http:// (GO),17 and the Reactome pathway database.18 Only functional modules www.babelomics.org/). The data were standardized using robust multichip (GO terms or reactome pathways) with an adjusted P-value o0.05 and analysis. Multiple probes mapping to the same gene were merged using experimental evidence were selected for analysis. GO terms were restricted the average as the summary of the hybridization values. Differential gene to levels 3–9. No restrictions were applied to the number or percentage of expression was detected using the Limma package from Babelomics. To genes in each category. The background set of genes used to perform the account for multiple testing effects, P-values were corrected using the false functional enrichment was the whole proteome, and the GO terms and the

Figure 1. Summary of the different comparisons made for differential gene expression analysis by using Limma statistics. (a) Volcano plot of the comparison of untreated extrapyramidal symptoms vs untreated no-extrapyramidal symptoms. (b) Volcano plot of the comparison of treated extrapyramidal symptoms vs treated no-extrapyramidal symptoms. (c) Volcano plot of the comparison of treated extrapyramidal symptoms vs untreated extrapyramidal symptoms. (d) Volcano plot of the comparison of treated no-extrapyramidal symptoms vs untreated no-extrapyramidal symptoms. (e) Venn diagram showing the gene expression overlap between extrapyramidal symptoms and no- extrapyramidal symptoms in response to antipsychotic treatment (untreated vs treated). Volcano plots represented the logarithmic values of the P-value and the fold change.

Reactome pathways were those included in the reference set of CLueGO. were compared according to untreated (Figure 1a) or treated Genes involved in each MCN were mapped to their enriched GO term or status (Figure 1b). By contrast, several genes were downregulated Reactome pathway based on the hypergeometric test (two side), with the in both patients not presenting EPS (Figure 1c) and patients – P-value being corrected by the Benjamini Hochberg method. ClueGO presenting EPS (Figure 1d) when comparing untreated and created two functional module networks (one for GO terms and another treated status. In response to AP treatment, 121 genes were for Reactome pathways) in which the different functional modules were clustered according to the association strength between terms calculated differentially expressed in patients not presenting EPS and 81 using chance-corrected kappa statistics.16 In these functional module genes were downregulated in EPS. Fifty genes were down- networks the node represented the significant modules (the size of the regulated in both patients presenting EPS and patients not node reflects the module’s enrichment significance), the edges between presenting EPS, as can be observed in the Venn diagram nodes were defined based on their shared genes, and the different colors (Figure 1e). A complete list of genes with their corresponding refer to the different clusters.16 statistical test and P-values can be found in the Supplementary fi Finally, we de ned an enriched functional module (GO terms or Material (Supplementary Tables S1 and S2). Reactome pathways) specific to patients with EPS or patients without EPS if over 66% of the genes in the functional module were from one of the groups. PPI Network construction Genes that changed their expression significantly (false discovery rate corrected P-value o0.05, fold change42) in response to RESULTS treatment were used to construct PPI networks for both patients Differential gene expression presenting EPS and patients not presenting EPS. Figure 1 shows the different comparisons made in the present The MCN constructed for patients not presenting EPS com- study, and summarizes the results obtained using volcano plots prised 431 gene products (Supplementary Material Supplemen- (representing the logarithmic values of the P-value and the fold tary Figure S1a) and included 101 of the 121 differentially expressed change). It can be seen that no significant results were obtained genes (83.47%). External nodes represented 76.56% of all nodes in when patients presenting EPS and patients not presenting EPS the MCN. The node of the obtained PPI network showed more

Figure 2. Functional networks obtained from case and control protein–protein interaction networks according to ClueGO: (a) Biological processes network. Each node represents a gene ontology biological process. The node size represents the enriched P-value corrected by the Benjamini–Hochberg method. Edge between nodes based on their kappa score level; and (b) Reactome pathway network. Each node represents a Reactome pathway. The node size represents the enriched P-value corrected by the Benjamini–Hochberg method. Edge between nodes based on their kappa score level.

connections (connectivity degree P-value = 9 × 10 − 4), higher con- different distributions between patients not presenting EPS and nectivity (clustering coefficient P-value = 2 × 10 − 4), and more hub patients presenting EPS (Table 2). Patients not presenting EPS nodes (betweenness centrality P-value = 2 × 10 − 4) than random were related to cluster 1, which includes GO terms associated with expectations. ubiquitination processes, and to cluster 7, which includes terms The MCN constructed for patients presenting EPS comprised associated with mitochondrial adenosine triphosphate (ATP) 231 gene products (Supplementary Material Supplementary synthesis and transport. By contrast, patients presenting EPS were Figure S1b) and included 64 of the 81 differentially expressed related to cluster 2 and cluster 4, both of which comprised genes (79.01%). External nodes represented 72.29% of all nodes in terms involved in insulin receptor signaling and lipid metabolism. the MCN. The node of the obtained PPI network showed more Although clusters 3 and 5 (related to regulation of autophagy connections (connectivity degree P-value = 3 × 10 − 3), higher con- and macroautophagy processes) and cluster 6 (related to immune nectivity (clustering coefficient P-value = 2 × 10 − 3), and more hub response) were identified in both patients not presenting EPS nodes (betweenness centrality P-value = 2 × 10 − 3) than random and patients presenting EPS, patients presenting EPS presented expectations. more GO terms in these clusters than did patients not presenting These results indicate that in both patients presenting EPS and EPS, which may indicate that these biological processes have a patients not presenting EPS, genes downregulated in response to greater role in patients presenting EPS than in patients not ‘ ’ AP treatment encode directly interacting proteins beyond the presenting EPS. The GO term transcription elongation was only fi level expected by chance, suggesting that AP perturb common identi ed in patients presenting EPS but it was not included in any molecular networks. cluster. Figure 2b shows the ClueGO network using reactome pathways. Different pathways were clustered in five functionally related Gene set enrichment analysis groups. In this analysis, the differences between patients present- To provide a functional interpretation of the generated MCN, and ing EPS and patients not presenting EPS (Table 3) became more to identify differences between patients presenting EPS and evident, and were in agreement with the results obtained with GO patients not presenting EPS, overrepresented groups of function- terms. Patients not presenting EPS were enriched in pathways ally related genes (GO biological process groups and Reactome related to protein folding (prefoldin and CCT/triC pathways, pathways) were identified using ClueGO. cluster 1), ATP synthesis (cluster 2), and ubiquitination (anaphase Figure 2a displays the ClueGO network showing the significant promoting complex (APC)/Cdc20 pathways, cluster 3). The NOTCH GO biological process of the genes predicted to be associated pathway was also identified only in patients not presenting EPS, with risperidone treatment. It can be seen that the different GO but it could not be included in any cluster. By contrast, patients terms were clustered in seven different groups, which showed presenting EPS were enriched in pathways involving NF-κB

Table 2. Gene set enrichment analysis of biological processes from GO obtained from constructed PPI networks of patients presenting EPS and patients not presenting EPS

GO cluster GO identiﬁer GO name Network % Genes patients % Genes patients speciﬁcity not presenting EPS presenting EPS

Cluster 1 GO:0007050 Cell cycle arrest no-EPS 72.31 36.15 Cluster 1 GO:0006633 Fatty acid biosynthetic process EPS 45.15 67.73 Cluster 1 GO:0031397 Negative regulation of protein ubiquitination no-EPS 96.02 9.60 Cluster 1 GO:0006511 Gbiquitin-dependent protein catabolic process no-EPS 100.00 0.00 Cluster 1 GO:0051437 Positive regulation of ubiquitin-protein ligase no-EPS 100.00 0.00 activity involved in mitotic cell cycle Cluster 1 GO:0051248 Negative regulation of protein metabolic process no-EPS 87.58 27.34 Cluster 1 GO:0070647 Protein modification by small protein no-EPS 90.87 19.27 conjugation or removal Cluster 1 GO:0000209 Protein polyubiquitination no-EPS 87.72 17.54 Cluster 1 GO:0031145 anaphase promoting complex-dependent proteasomal no-EPS 100.00 0.00 ubiquitin-dependent protein catabolic process Cluster 1 GO:0007093 Mitotic cell cycle checkpoint no-EPS 90.90 9.09 Cluster 1 GO:0071156 Regulation of cell cycle arrest no-EPS 80.57 14.28 Cluster 1 GO:0043161 Proteasomal ubiquitin-dependent protein catabolic process no-EPS 100.00 0.00 Cluster 1 GO:0032446 Protein modification by small protein conjugation no-EPS 90.85 22.71 Cluster 1 GO:0016567 Protein ubiquitination no-EPS 92.05 16.17 Cluster 1 GO:0032269 Negative regulation of cellular protein metabolic process no-EPS 91.97 24.52 Cluster 1 GO:0031396 Regulation of protein ubiquitination no-EPS 94.71 14.57 Cluster 1 GO:0031398 Positive regulation of protein ubiquitination no-EPS 100.00 0.00 Cluster 2 GO:0048015 Phosphatidylinositol-mediated signaling EPS 12.50 87.50 Cluster 2 GO:0036092 Phosphatidylinositol-3-phosphate biosynthetic process EPS 0.00 100.00 Cluster 2 GO:0070555 Response to interleukin-1 EPS 43.06 64.60 Cluster 2 GO:0046890 Regulation of lipid biosynthetic process EPS 43.08 71.80 Cluster 2 GO:0006839 Mitochondrial transport None 59.74 59.74 Cluster 2 GO:0015909 Long-chain fatty acid transport EPS 43.06 64.60 Cluster 2 GO:0010565 Regulation of cellular ketone metabolic process EPS 33.86 79.02 Cluster 2 GO:0006109 Regulation of carbohydrate metabolic process EPS 17.96 89.82 Cluster 3 GO:0016241 Regulation of macroautophagy None 50.00 50.00 Cluster 3 GO:0042149 Cellular response to glucose starvation EPS 33.33 66.66 Cluster 3 GO:0006944 Cellular membrane fusion no-EPS 77.81 51.86 Cluster 3 GO:0010508 Positive regulation of autophagy EPS 45.50 68.26 Cluster 3 GO:0010506 Regulation of autophagy EPS 51.87 77.81 Cluster 3 GO:0048284 Organelle fusion no-EPS 77.37 38.68 Cluster 4 GO:0043303 Mast cell degranulation EPS 26.41 79.24 Cluster 4 GO:0030258 Lipid modification EPS 30.54 81.45 Cluster 4 GO:0008286 Insulin receptor signaling pathway EPS 15.69 94.15 Cluster 5 GO:0042594 Response to starvation EPS 6.62 62.38 Cluster 5 GO:0007033 Vacuole organization None 67.08 67.08 Cluster 5 GO:0016236 Macroautophagy EPS 6.73 62.13 Cluster 5 GO:0031669 Cellular response to nutrient levels EPS 6.45 60.27 Cluster 6 GO:0002726 Positive regulation of T-cell cytokine production EPS 31.54 63.09 Cluster 6 GO:0070423 Nucleotide-binding oligomerization domain containing EPS 33.33 66.66 signaling pathway Cluster 6 GO:0032481 Positive regulation of type I interferon production None 52.83 52.83 Cluster 6 GO:0031293 Membrane protein intracellular domain proteolysis EPS 26.41 79.24 Cluster 6 GO:0032479 Regulation of type I interferon production no-EPS 71.08 43.08 Cluster 6 GO:0032743 Positive regulation of interleukin-2 production EPS 31.54 63.09 Cluster 6 GO:0035872 Nucleotide-binding domain leucine rich repeat containing no-EPS 60.00 40.00 receptor signaling pathway Cluster 6 GO:0038061 NIK/NF-kappaB cascade None 52.83 52.83 Cluster 6 GO:0032655 Regulation of interleukin-12 production EPS 26.41 79.24 Cluster 6 GO:0035666 TRIF-dependent toll-like receptor signaling pathway None 53.89 53.89 Cluster 7 GO:0006754 ATP biosynthetic process no-EPS 63.25 50.60 Cluster 7 GO:0042776 Mitochondrial ATP synthesis coupled proton transport no-EPS 71.80 43.08 Unclustered GO:0006368 Transcription elongation from RNA polymerase II promoter EPS 0.00 100.00 Abbreviations: ATP, adenosine triphosphate; EPS, extrapyramidal symptoms; GO, gene ontology; PPI, protein–protein interaction; TRIF, Toll/IL-1 receptor (TIR) domain-containing adaptor. The table shows the GO terms identified, their cluster distribution according to ClueGO, their specificity for patients presenting EPS or patients not presenting EPS, and the percentage of genes involved in each group of patients.

activation (cluster 4) and mammalian target of rapamycin (mTOR)/ DISCUSSION AMP-activated protein kinase (AMPK) (cluster 5), which were To identify new candidate genes for pharmacogenetic studies involved in several GO terms identiﬁed previously in patients of AP-induced EPS we performed a PPI network analysis of presenting EPS, such as insulin receptor pathway, lipid metabolism differential gene expression induced by AP treatment in and autophagy. peripheral blood of AP-naïve patients with ﬁrst-episode psychosis.

Table 3. Gene set enrichment analysis of reactome pathways obtained from constructed PPI networks of EPS and NO-EPS

Reactome Reactome identiﬁer Reactome name Network % Genes % Genes cluster speciﬁcity patients not patients presenting EPS presenting EPS

Cluster 1 REACTOME:16936 Prefoldin-mediated transfer of substrate to CCT/TriC no-EPS 100.00 0.00 Cluster 1 REACTOME:17050 Folding of actin by CCT/TriC no-EPS 100.00 0.00 Cluster 1 REACTOME:17029 Cooperation of prefoldin and TriC/CCT in actin and no-EPS 100.00 0.00 tubulin folding Cluster 2 REACTOME:6759 Formation of ATP by chemiosmotic coupling no-EPS 71.80 43.08 Cluster 2 REACTOME:6305 Respiratory electron transport, ATP synthesis by chemiosmotic no-EPS 63.25 50.60 coupling and heat production by uncoupling proteins. Cluster 3 REACTOME:2137 Mitotic spindle checkpoint no-EPS 100.00 0.00 Cluster 3 REACTOME:6871 APC/C:Cdc20-mediated degradation of securin no-EPS 100.00 0.00 Cluster 3 REACTOME:6837 Regulation of APC/C activators between G1/S no-EPS 100.00 0.00 and early anaphase Cluster 3 REACTOME:6904 Phosphorylation of the APC/C no-EPS 100.00 0.00 Cluster 3 REACTOME:1538 Cell cycle checkpoints no-EPS 90.90 9.09 Cluster 3 REACTOME:6761 APC/C:Cdh1-mediated degradation of Cdc20 and other no-EPS 100.00 0.00 APC/C:Cdh1-targeted proteins in late mitosis/early G1 Cluster 3 REACTOME:6954 Activation of APC/C- and APC/C:Cdc20-mediated no-EPS 100.00 0.00 degradation of mitotic proteins Cluster 3 REACTOME:75820 no-EPS 90.56 16.98 Cluster 3 REACTOME:6820 APC/C:Cdc20-mediated degradation of cyclin B no-EPS 100.00 0.00 Cluster 3 REACTOME:6828 APC/C-mediated degradation of cell cycle proteins no-EPS 100.00 0.00 Cluster 3 REACTOME:6785 Autodegradation of Cdh1 by Cdh1:APC/C no-EPS 100.00 0.00 Cluster 3 REACTOME:21279 Regulation of mitotic cell cycle no-EPS 100.00 0.00 Cluster 3 REACTOME:1041 Inhibition of the proteolytic activity of APC/C required for the no-EPS 100.00 0.00 onset of anaphase by mitotic spindle checkpoint components Cluster 3 REACTOME:75842 Antigen processing: Ubiquitination & no-EPS 95.58 12.74 Cluster 3 REACTOME:1072 Inactivation of APC/C via direct inhibition of the APC/C complex no-EPS 100.00 0.00 Cluster 3 REACTOME:6781 APC/C:Cdc20-mediated degradation of mitotic proteins no-EPS 100.00 0.00 Cluster 3 REACTOME:6867 Conversion from APC/C:Cdc20 to APC/C:Cdh1 in late anaphase no-EPS 100.00 0.00 Cluster 4 REACTOME:118823 Cytosolic sensors of pathogen-associated DNA EPS 43.06 64.60 Cluster 4 REACTOME:6783 Toll-like receptor 3 cascade EPS 46.03 61.38 Cluster 4 REACTOME:118638 Downstream signaling events of B-cell receptor None 47.31 59.14 Cluster 4 REACTOME:6809 MyD88-independent cascade EPS 46.03 61.38 Cluster 4 REACTOME:118563 RIP-mediated NF-κB activation via DAI EPS 43.06 64.60 Cluster 4 REACTOME:24969 TRAF6 mediated NF-κB activation EPS 43.08 71.80 Cluster 4 REACTOME:116008 PI3K events in ERBB2 signaling EPS 40.00 60.00 Cluster 4 REACTOME:13443 Regulated proteolysis of p75NTR EPS 26.41 79.24 Cluster 4 REACTOME:12578 GAB1 signalosome EPS 40.00 60.00 Cluster 4 REACTOME:12555 Downstream TCR signaling EPS 35.92 71.85 Cluster 4 REACTOME:21281 TAK1 activates NF-κB by phosphorylation and activation EPS 35.92 71.85 of IKKs complex Cluster 4 REACTOME:75829 PIP3 activates AKT signaling EPS 25.00 75.00 Cluster 4 REACTOME:13537 p75NTR signals via NF-κB None 53.89 53.89 Cluster 4 REACTOME:13696 NF-kB is activated and signals survival None 53.89 53.89 Cluster 4 REACTOME:25359 RIG-I/MDA5 mediated induction of IFN-alpha/beta pathways None 58.94 58.94 Cluster 4 REACTOME:118764 DAI mediated induction of type I IFNs EPS 43.06 64.60 Cluster 5 REACTOME:762 IRS-related events EPS 20.36 91.63 Cluster 5 REACTOME:21387 Energy dependent regulation of mTOR by LKB1-AMPK EPS 28.72 86.16 Cluster 5 REACTOME:6838 mTOR signaling EPS 25.30 88.56 Cluster 5 REACTOME:21393 Regulation of Rheb GTPase activity by AMPK EPS 33.08 82.70 Cluster 5 REACTOME:1988 AMPK inhibits chREBP transcriptional activation activity EPS 21.53 86.13 Cluster 5 REACTOME:11082 Import of palmitoyl-CoA into the mitochondrial matrix EPS 26.41 79.24 Cluster 5 REACTOME:976 PI3K cascade EPS 20.36 91.63 Cluster 5 REACTOME:498 Signaling by insulin receptor EPS 20.36 91.63 Cluster 5 REACTOME:21285 Regulation of AMPK activity via LKB1 EPS 28.71 86.16 Cluster 5 REACTOME:332 IRS-mediated signaling EPS 20.36 91.63 Cluster 5 REACTOME:456 PKB-mediated events EPS 25.30 88.56 Cluster 5 REACTOME:1195 Insulin receptor signaling cascade EPS 20.36 91.63 Unclustered REACTOME:299 Signaling by NOTCH no-EPS 66.84 40.10 Abbreviations: AKT, v-akt murine thymoma viral oncogene homolog 1; AMPK, AMP-activated protein kinase; APC, anaphase promoting complex; ATP, adenosine triphosphate; DAI, Z-DNA binding protein 1; EPS, extrapyramidal symptoms; IFN, interferon, alpha 1; NF-κB, Nuclear factor κ beta; PPI, protein– protein interaction; PKB, protein kinase B; RIP, receptor (TNFRSF)-interacting serine-threonine kinase 1; TCR, T-cell antigen receptor. The table shows the pathways identiﬁed, their cluster distribution according to ClueGO, their speciﬁcity for EPS or no-EPS and the percentage of genes involved in each group of patients.

Our analysis revealed that constructed networks were enriched for prefoldin/CCT/triC and ATP synthesis in no-EPS patients, and different biological processes in patients without EPS (ubiquitina- mTOR and NF-κB kinases in patients with EPS. Interestingly, most tion, protein folding and ATP metabolism) compared with those of these processes and pathways are downstream effectors of presenting EPS (insulin receptor signaling, lipid modification, AKT, a kinase related to schizophrenia19 and AP response.20 regulation of autophagy and immune response). Moreover, the Moreover, the processes identified in peripheral blood in our observed differences also involved specific pathways directly study are functionally related to important processes in brain related to these biological processes, such as APC/cdc20, tissues and could be related to schizophrenia and AP response.

© 2015 Macmillan Publishers Limited The Pharmacogenomics Journal (2015), 452 – 460 mTOR and NF-κB as candidate genes for AP-induced EPS S Mas et al 458 In our sample we were unable to find differences in gene The mTOR has emerged over the last decade as a central expression when comparing patients with without EPS, in either regulator of cell growth and metabolism.51 The energy status of the untreated or the treated condition. The lack of significant the cell is signaled to mTOR through AMPK, a master sensor of differences in gene expression signature in the untreated condi- intracellular energy depletion.52 In addition, several growth tion makes it impossible to identify peripheral predictors of EPS factors, such as insulin and insulin-like growth factor 1, regulate susceptibility prior to the establishment of an AP treatment. This the mTOR pathway through stimulation of both the PI3K/Akt and may be because EPS susceptibility is due to subtle differences that Ras/ERK pathways.53 The canonical Wnt pathway also activates are only relevant during or after AP administration, but could mTOR, in this case through inhibition of GSKB.54 not be observed before. In this regard, the lack of significant mTOR regulates lipid synthesis through positive regulation of differences in the treated condition also reinforces this hypothesis, both sterol regulatory element-binding protein 155 and peroxi- and is in agreement with the results presented here, demonstrat- some proliferator-activated receptor gamma.56 mTOR also ing that the susceptibility to EPS and EPS per se are related not to modulates glucose metabolite through positive regulation of specific genes but to a more complex response involving different HIF1alfa.57 The multiple aspects of cell metabolism that are pathways and complex networks involving several genes. regulated by mTOR, and the distinct roles that it plays in different In patients not presenting EPS treatment with AP regulated tissues, means that mTOR plays a pivotal role in metabolic processes such as ubiquitin proteosome and protein folding. disorders. Specifically, it regulates: food intake in the hypothala- These processes are involved in maintaining a delicate balance mus; adipogenesis in adipose tissue; protein metabolism and between the conservation of functional proteins and the refolding muscle mass in muscle; lipogenesis, ketogenesis, and gluconeo- of misfolded proteins or their degradation, which are important genesis in the liver; and beta-cell mass and insulin secretion in mechanisms for cellular homeostasis. Dysregulation of the ubi- pancreas.51 The identification of mTOR as a downstream effector quitination system in schizophrenia has been identified in post- of AP in peripheral tissues could help to explain the metabolic side mortem brain tissue.21–24 Misfolded proteins form aggregates effects related to second-generation AP58, and may provide a new (also called the aggresome) that could impair different cellular source of candidate genes for pharmacogenetic studies. process. Several studies have reported evidence of protein folding In the induction stages of autophagy, many signals converge at aberrations in schizophrenia.25–27 the level of mTOR.51 Autophagy is an intracellular protein Related to these processes, two pathways were enriched in degradation mechanism that engulfs cytoplasmatic constituents patients not presenting EPS: the APC/cdc20 pathway and the Tric/ and organelles within vesicles and delivers these to lysosomes.59 CCT pathway. APC is an E3 ubiquitin ligase28 that has emerged as Autophagy is important for proper development and neuronal a key regulator of diverse developmental processes in neurons.29–31 signaling, which ensures the formation of appropriate neuronal The regulation of APC/cdc20 seems to be related to HDAC6, which connections and their function.60 Recent studies have demon- is one target of the Akt/GSK3B pathway.31,32 Tric/CCT (TCP-1 ring strated a crucial role for autophagy-related genes in the develop- complex/chaperonin containing TCP-1) is a chaperonin complex ment and maturation of axons, dendrites, and synapses.61 In that has been related to cell death prevention through response addition, research has recently indicated a possible role for to protein misfolding and aggregation.33 autophagy in schizophrenia pathophysiology, since differences in Patients not presenting EPS were also enriched for ATP autophagy-related genes could be observed in post-mortem biosynthesis and the ATP synthesis pathway. Several studies have brains of schizophrenia patients versus healthy subjects.62 A reported reduced levels of ATP, decreased volume density, and further point of note is that autophagy has been shown to be deformation of mitochondria in schizophrenia.34,35 Interestingly, affected by different AP, including FGA such as haloperidol63 or DISC1 (disrupted in schizophrenia 1), a candidate gene for chlorpromazine,64 and SGA such as clozapine63 and sertindole.65 schizophrenia involved in the Akt/GSK3B pathway, could repre- Recently, different groups of researchers have demonstrated sent a link between protein folding regulation26 and mitochon- that the mTOR pathway may be modulated by several AP such as drial dysfunction.36 chlorpromazine,64 olanzapine,58 haloperidol66 and risperidone.67 Our analysis revealed that the constructed network in patients Moreover, four SNPs in four genes involved in the mTOR pathway presenting EPS was enriched in different biological processes have been used to develop a predictor of EPS that has been related to the NF-κB pathway (inflammatory response) and the recently patented (European Patent EP13382027.4).68 This phar- mTOR pathway (lipid biosynthesis, insulin signaling and autop- macogenetic predictor, including the SNPs rs1130214 (AKT1), hagy). NF-κB is constitutively activated in neurons, regulating the rs456998 (FCHSD1), rs7211818 (Raptor) and rs1053639 (DDIT4), expression of an increasingly recognized number of genes correctly predicted 85% of EPS appearance in three different involved in cell survival, synaptic plasticity and memory in the cohorts.68 adult brain.37 There is growing evidence in support of a role for Our results showing different patterns of gene expression in EPS NF-κB in schizophrenia. First, NF-κB is activated by cytokines, patients, patterns linked moreover to different biological pro- neurotrophic factors such as BDNF and glutamate, which have cesses and functional pathways, offer new and valuable markers been strongly associated with schizophrenia.38–42 Second, first- for pharmacogenetic studies. In particular, the present study episode, drug-naive schizophrenic patients show higher levels of identifies the NF-κB and mTOR pathways as playing a keyrole in cytokines in serum, and NF-κB activation (increased RELA gene AP-induced EPS. The integration of pharmacogenomic data from expression and p65 nuclear activity) in peripheral blood mono- peripheral blood in humans (as in the case of the results presented nuclear cells.43 Third, pathway analysis identified NF-κB as a hub, here) with data from microarrays of differentially expressed genes where multiple, diverse signal transduction pathways, enriched for from in vitro models69 and animal models is the final goal of the schizophrenia genetic risk factors, converge.44,45 present study, and represents an example of the convergent NF-kB has been identified as a target regulated by AP.46,47 functional genomic approach.6 These results highlight the Several studies suggest that increased levels of NF-κB p65 subunit importance of integrating pharmacogenomic data of gene in striatum may be involved in haloperidol-induced cellular expression experiments with systems biology analysis to increase changes and may increase the vulnerability of the dopaminergic our understanding of the molecular mechanisms of action of AP neurons to a possible neurotoxic effect in haloperidol-treated drugs. In agreement with recent suggestions of the US National rats.48,49 Furthermore, NF-kB has been positively correlated with Institute of Mental Health,70 we focus in the understanding of the haloperidol-induced dopamine supersensitivity and orofacial core mechanisms involved in drug response, in order to identify dyskinetic behaviors.48 It has also been related to treatment- networks and pathways that could be affected, instead of the resistant schizophrenia.50 analysis of independent genes with small effects in drug response.

The Pharmacogenomics Journal (2015), 452 – 460 © 2015 Macmillan Publishers Limited mTOR and NF-κB as candidate genes for AP-induced EPS S Mas et al 459 The ability to study combinations of variants in genes involved in 18 Joshi-Tope G, Gillespie M, Vastrik I, D’Eustachio P, Schmidt E, de Bono B et al. these identified networks and pathways is more likely than Reactome: a knowledgebase of biological pathways. Nucleic Acids Res 2005; 33: studying individual genes to yield insight into AP response D428–D432. mechanisms and suggest drug targets.71 19 Zheng W, Wang H, Zeng Z, Lin J, Little PJ, Srivastava LK et al. The possible role of the Akt signaling pathway in schizophrenia. Brain Res 2012; 1470:145–158. 20 Freyberg Z, Ferrando SJ, Javitch JA. Roles of the Akt/GSK-3 and Wnt signaling CONFLICT OF INTEREST pathways in schizophrenia and antipsychotic drug action. Am J Psychiatry 2010; 167:388–396. The authors declare no conflict of interest. 21 Altar CA, Jurata LW, Charles V, Lemire A, Liu P, Bukhman Y et al. Deficient hippocampal neuron expression of proteasome, ubiquitin, and mitochondrial genes ACKNOWLEDGMENTS in multiple schizophrenia cohorts. Biol Psychiatry 2005; 58:85–96. 22 Rubio MD, Wood K, Haroutunian V, Meador-Woodruff JH. Dysfunction of the This study was supported by the Spanish Ministry of Health, Instituto de Salud Carlos ubiquitin proteasome and ubiquitin-like systems in schizophrenia. Neuropsycho- III (FIS, Fondo de Investigacion Sanitaria PI10/02430) and the Catalan Innovation, pharmacology 2013; 38: 1910–1920. Universities and Enterprise Authority (Grants DURSI 2009SGR1295, 2009SGR1501); 23 Bousman CA, Chana G, Glatt SJ, Chandler SD, Lucero GR, Tatro E et al. Preliminary ‘ ’ and Sara Borrell contract from the Spanish Ministry of Health, Instituto de Salud evidence of ubiquitin proteasome system dysregulation in schizophrenia and Carlos III (FIS, Fondo de Investigación Sanitaria) (Grant CD09/00296) (to P.G.). The bipolar disorder: convergent pathway analysis findings from two independent authors thank the Language Advisory Service of the University of Barcelona, Spain for samples. Am J Med Genet B Neuropsychiatr Genet 2010; 153B:494–502. manuscript revision. The authors also thank Ana Meseguer for sample collection 24 Bousman CA, Chana G, Glatt SJ, Chandler SD, May T, Lohr J et al. Positive assistance. symptoms of psychosis correlate with expression of ubiquitin proteasome genes in peripheral blood. Am J Med Genet B Neuropsychiatr Genet 2010; 153B: – REFERENCES 1336 1341. 25 Atkin T, Kittler J. DISC1 and the aggresome: a disruption to cellular function? 1 Lafuente A, Bernardo M, Mas S, Crescenti A, Aparici M, Gassó P et al. Poly- Autophagy 2012; 8:851–852. morphism of dopamine D2 receptor (TaqIA, TaqIB, and-141C Ins/Del) and dopa- 26 Atkin TA, Brandon NJ, Kittler JT. Disrupted in Schizophrenia 1 forms pathological mine degradation enzyme (COMT G158A, A-278G) genes and extrapyramidal aggresomes that disrupt its function in intracellular transport. Hum Mol Genet symptoms in patients with schizophrenia and bipolar disorders. Psychiatry Res 2012; 21:2017–2028. 2008; 161: 131–141. 27 Korth C. Aggregated proteins in schizophrenia and other chronic mental diseases: 2 Gassó P, Mas S, Bernardo M, Alvarez S, Parellada E, Lafuente A. A common variant DISC1opathies. Prion 2012; 6:134–141. in DRD3 gene is associated with risperidone-induced extrapyramidal symptoms. 28 Peters JM. The anaphase promoting complex/cyclosome: a machine designed to 9 – Pharmacogenomics J 2009; :404 410. destroy. Nat Rev Mol Cell Biol 2006; 7:644–656. 3 Llerena A, Berecz R, Peñas-Lledó E, Süveges A, Fariñas H. Pharmacogenetics of 29 García-Higuera I, Manchado E, Dubus P, Cañamero M, Méndez J, Moreno S et al. 14 – clinical response to risperidone. Pharmacogenomics 2013; :177 194. Genomic stability and tumour suppression by the APC/C cofactor Cdh1. Nat Cell 4 Mas S, Llerena A, Saíz J, Bernardo M, Lafuente A. Strengths and weaknesses of Biol 2008; 10: 802–811. pharmacogenetic studies of antipsychotic drugs: the potential value of the 30 Yang Y, Kim AH, Bonni A. The dynamic ubiquitin ligase duo: Cdh1-APC and Cdc20- 13 – PEPs study. Pharmacogenomics 2012; : 1773 1782. APC regulate neuronal morphogenesis and connectivity. Curr Opin Neurobiol fi 5 Ikeda M, Tomita Y, Mouri A, Koga M, Okochi T, Yoshimura R et al. Identi cation of 2010; 20:92–99. novel candidate genes for treatment response to risperidone and susceptibility for 31 Kim AH, Puram SV, Bilimoria PM, Ikeuchi Y, Keough S, Wong M et al. Acen- schizophrenia: integrated analysis among pharmacogenomics, mouse expression, trosomal Cdc20-APC pathway NO-EPS dendrite morphogenesis in postmitotic and genetic case-control association approaches. Biol Psychiatry 2010; 67:263–269. neurons. Cell 2009; 136:322–336. 6 Niculescu AB, Le-Niculescu H. Convergent Functional Genomics: what we have 32 Chen S, Owens GC, Makarenkova H, Edelman DB. HDAC6 regulates mitochondrial learned and can learn about genes, pathways, and mechanisms. Neuropsycho- transport in hippocampal neurons. PLoS One 2010; 5: e10848. pharmacology 2010; 35: 355–356. 33 Swinnen E, Büttner S, Outeiro TF, Galas MC, Madeo F, Winderickx J et al. Aggre- 7 Ayalew M, Le-Niculescu H, Levey DF, Jain N, Changala B, Patel SD et al. Con- some formation and segregation of inclusions influence toxicity of α-synuclein vergent functional genomics of schizophrenia: from comprehensive under- and synphilin-1 in yeast. Biochem Soc Trans 2011; 39: 1476–1481. standing to genetic risk prediction. Mol Psychiatry 2012; 17:887–905. 34 Rezin GT, Amboni G, Zugno AI, Quevedo J, Streck EL. Mitochondrial dysfunction 8 Kuzman MR, Medved V, Terzic J, Krainc D. Genome-wide expression analysis of and psychiatric disorders. Neurochem Res 2009; 34: 1021–1029. peripheral blood identifies candidate biomarkers for schizophrenia. J Psychiatr Res 35 Dror N, Klein E, Karry R, Sheinkman A, Kirsh Z, Mazor M et al. State-dependent 2009; 43: 1073–1077. alterations in mitochondrial complex I activity in platelets: a potential peripheral 9 Sullivan PF, Fan C, Perou CM. Evaluating the comparability of gene expression in marker for schizophrenia. Mol Psychiatry 2002; 7: 995–1001. blood and brain. Am J Med Genet B Neuropsychiatr Genet 2006; 141B:261–268. 36 Park YU, Jeong J, Lee H, Mun JY, Kim JH, Lee JS et al. Disrupted-in-schizophrenia 1 10 Buttarelli FR, Fanciulli A, Pellicano C, Pontieri FE. The dopaminergic system in (DISC1) plays essential roles in mitochondria in collaboration with Mitofilin. Proc peripheral blood lymphocytes: from physiology to pharmacology and potential 107 – applications to neuropsychiatric disorders. Curr Neuropharmacol 2011; 9:278–288. Natl Acad Sci USA 2010; : 17785 17790. 11 Rivera-Baltanas T, Olivares JM, Martinez-Villamarin JR, Fenton EY, Kalynchuk LE, 37 Gutierrez H, Davies AM. Regulation of neural process growth, elaboration and κ 34 – Caruncho HJ. Serotonin 2A receptor clustering in peripheral lymphocytes is structural plasticity by NF- B. Trends Neurosci 2011; : 316 325. fi altered in major depression and may be a biomarker of therapeutic efficacy. 38 Buckley PF, Pillai A, Howell KR. Brain-derived neurotrophic factor: ndings in 24 – J Affect Disord 2014; 163:47–55. schizophrenia. Curr Opin Psychiatry 2011; : 122 127. fl 12 Minguez P, Dopazo J. Functional genomics and networks: new approaches in the 39 Dean B. Understanding the role of in ammatory-related pathways in the extraction of complex gene modules. Expert Rev Proteomics 2010; 7:55–63. pathophysiology and treatment of psychiatric disorders: evidence from human 14 13 García-Alonso L, Alonso R, Vidal E, Amadoz A, de María A, Minguez P et al. Dis- peripheral studies and CNS studies. Int J Neuropsychopharmacol 2011; : covering the hidden sub-network component in a ranked list of genes or proteins 997–1012. derived from genomic experiments. Nucleic Acids Res 2012; 40: e158. 40 Javitt DC, Schoepp D, Kalivas PW, Volkow ND, Zarate C, Merchant K et al. Trans- 14 Minguez P, Götz S, Montaner D, Al-Shahrour F, Dopazo J. SNOW, a web-based tool lating glutamate: from pathophysiology to treatment. Sci Transl Med 2011; 3: for the statistical analysis of protein-protein interaction networks. Nucleic Acids Res 102mr2. 2009; 37: W109–W114. 41 Watanabe Y, Someya T, Nawa H. Cytokine hypothesis of schizophrenia patho- 15 Minguez P, Dopazo J. Assessing the biological significance of gene expression genesis: evidence from human studies and animal models. Psychiatry Clin Neurosci signatures and co-expression modules by studying their network properties. PLoS 2010; 64:217–230. One 2011; 6: e17474. 42 Willard SS, Koochekpour S. Glutamate, glutamate receptors, and downstream 16 Bindea G, Mlecnik B, Hackl H, Charoentong P, Tosolini M, Kirilovsky A et al. signaling pathways. Int J Biol Sci 2013; 9:948–959. ClueGO: a Cytoscape plug-in to decipher functionally grouped gene ontology and 43 Song XQ, Lv LX, Li WQ, Hao YH, Zhao JP. The interaction of nuclear factor-kappa B pathway annotation networks. Bioinformatics 2009; 25: 1091–1093. and cytokines is associated with schizophrenia. Biol Psychiatry 2009; 65:481–488. 17 Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM et al. Gene 44 Sun J, Jia P, Fanous AH, van den Oord E, Chen X, Riley BP et al. Schizophrenia gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat networks and pathways and their applications for novel candidate gene selection. Genet 2000; 25:25–29. PLoS One 2010; 5: e11351.

© 2015 Macmillan Publishers Limited The Pharmacogenomics Journal (2015), 452 – 460 mTOR and NF-κB as candidate genes for AP-induced EPS S Mas et al 460 45 Roussos P, Katsel P, Davis KL, Giakoumaki SG, Lencz T, Malhotra AK et al. Con- metabolic dysregulation with atypical antipsychotic drugs. J Pharmacol Exp Ther vergent ﬁndings for abnormalities of the NF-κB signaling pathway in schizo- 2013; 347:126–135. phrenia. Neuropsychopharmacology 2013; 38:533–539. 59 Levine B, Klionsky DJ. Development by self-digestion: molecular mechanisms and 46 Chen ML, Tsai TC, Lin YY, Tsai YM, Wang LK, Lee MC et al. Antipsychotic drugs biological functions of autophagy. Dev Cell 2004; 6:463–477. suppress the AKT/NF-κB pathway and regulate the differentiation of T-cell sub- 60 Lee KM, Hwang SK, Lee JA. Neuronal Autophagy and Neurodevelopmental Dis- sets. Immunol Lett 2011; 140:81–91. orders. Exp Neurobiol 2013; 22:133–142. 47 Rizig MA, McQuillin A, Ng A, Robinson M, Harrison A, Zvelebil M et al. A gene 61 Inoue K, Rispoli J, Yang L, Macleod D, Beal MF, Klann E et al. Coordinate regulation expression and systems pathway analysis of the effects of clozapine compared to of mature dopaminergic axon morphology by macroautophagy and the PTEN haloperidol in the mouse brain implicates susceptibility genes for schizophrenia. signaling pathway. PLoS Genet 2013; 9: e1003845. J Psychopharmacol 2012; 26: 1218–1230. 62 Merenlender-Wagner A, Malishkevich A, Shemer Z, Udawela M, Gibbons A, Scarr E 48 Bishnoi M, Chopra K, Kulkarni SK. Differential striatal levels of TNF-alpha, NFkap- et al. Autophagy has a key role in the pathophysiology of schizophrenia. Mol paB p65 subunit and dopamine with chronic typical and atypical neuroleptic Psychiatry 2013; e-pub ahead of print 2013doi:10.1038/mp.2013.174. treatment: role in orofacial dyskinesia. Prog Neuropsychopharmacol Biol Psychiatry 63 Park J, Chung S, An H, Kim J, Seo J, Kim DH et al. Haloperidol and clozapine block 2008; 32: 1473–1478. formation of autophagolysosomes in rat primary neurons. Neuroscience 2012; 49 Saldaña M, Bonastre M, Aguilar E, Marin C. Role of nigral NFkappaB p50 and p65 209:64–73. subunit expression in haloperidol-induced neurotoxicity and stereotyped beha- 64 Shin SY, Lee KS, Choi YK, Lim HJ, Lee HG, Lim Y et al. The antipsychotic agent vior in rats. Eur Neuropsychopharmacol 2006; 16:491–497. chlorpromazine induces autophagic cell death by inhibiting the Akt/mTOR 50 Liou YJ, Wang HH, Lee MT, Wang SC, Chiang HL, Chen CC et al. Genome-wide pathway in human U-87MG glioma cells. Carcinogenesis 2013; 34: association study of treatment refractory schizophrenia in Han Chinese. PLoS One 2080–2089. 2012; 7: e33598. 65 Shin JH, Park SJ, Kim ES, Jo YK, Hong J, Cho DH. Sertindole a potent anta-

51 Laplante M, Sabatini DM. mTOR signaling in growth control and disease. Cell 2012; gonist at dopamine D2 receptors, induces autophagy by increasing reactive 149: 274–293. oxygen species in SH-SY5Y neuroblastoma cells. Biol Pharm Bull 2012; 35: 52 Hardie DG. Role of AMP-activated protein kinase in the metabolic syndrome and 1069–1075. in heart disease. FEBS Lett 2008; 582:81–89. 66 Bonito-Oliva A, Pallottino S, Bertran-Gonzalez J, Girault JA, Valjent E, Fisone G. 53 Laplante M, Sabatini DM. mTOR signaling at a glance. J Cell Sci 2009; 122: Haloperidol promotes mTORC1-dependent phosphorylation of ribosomal protein 3589–3594. S6 via dopamine- and cAMP-regulated phosphoprotein of 32 kDa and inhibition 54 Inoki K, Ouyang H, Zhu T, Lindvall C, Wang Y, Zhang X et al. TSC2 integrates Wnt of protein phosphatase-1. Neuropharmacology 2013; 72:197–203. and energy signals via a coordinated phosphorylation by AMPK and GSK3 to 67 Korostynski M, Piechota M, Dzbek J, Mlynarski W, Szklarczyk K, Ziolkowska B et al. regulate cell growth. Cell 2006; 126: 955–968. Novel drug-regulated transcriptional networks in brain reveal pharmacological 55 Porstmann T, Santos CR, Grifﬁths B, Cully M, Wu M, Leevers S et al. SREBP activity properties of psychotropic drugs. BMC Genomics 2013; 14: 606. is regulated by mTORC1 and contributes to Akt-dependent cell growth. Cell 68 Mas S, Gassó P, Ritter MA, Malagelada C, Bernardo M, Lafuente A Pharmacoge- Metab 2008; 8: 224–236. netic predictor of extrapyramidal symptoms induced by antipsychotics: Multi- 56 Kim JE, Chen J. Regulation of peroxisome proliferator-activated receptor-gamma locus interaction in the mTOR pathway. Eur Neuropsychopharmacol 2014; e-pub activity by mammalian target of rapamycin and amino acids in adipogenesis. ahead of print; doi: 10.1016/j.euroneuro.2014.11.011. Diabetes 2004; 53: 2748–2756. 69 Mas S, Gassó P, Bernardo M, Lafuente A. Functional analysis of gene expression in 57 Düvel K, Yecies JL, Menon S, Raman P, Lipovsky AI, Souza AL et al. Activation of a risperidone treated cells provide new insights in molecular mechanism and new metabolic gene regulatory network downstream of mTOR complex 1. Mol Cell candidate genes for pharmacogenetic studies. Eur Neuropsychopharmacol 2013; 2010; 39: 171–183. 23:329–337. 58 Schmidt RH, Jokinen JD, Massey VL, Falkner KC, Shi X, Yin X et al. Olanzapine 70 Marder SR. Perspective: retreat from the radical. Nature 2014; 508: S18. activates hepatic mammalian target of rapamycin: new mechanistic insight into 71 Hyman SE. Perspective: revealing molecular secrets. Nature 2014; 508: S20.

Supplementary Information accompanies the paper on the The Pharmacogenomics Journal website (http://www.nature.com/tpj).