MIR137 Variants Identified in Psychiatric Patients Affect

ORIGINAL ARTICLE MIR137 variants identiﬁed in psychiatric patients affect synaptogenesis and neuronal transmission gene sets

M Strazisar1,2,10, S Cammaerts1,2,10, K van der Ven1,2,11, DA Forero1,2,3, A-S Lenaerts1,2,12, A Nordin4, L Almeida-Souza2,5, G Genovese6,7,8, V Timmerman2,5, A Liekens1,2, P De Rijk1,2, R Adolfsson4, P Callaerts9 and J Del-Favero1,2,11

Sequence analysis of 13 microRNA (miRNA) genes expressed in the human brain and located in genomic regions associated with schizophrenia and/or bipolar disorder, in a northern Swedish patient/control population, resulted in the discovery of two functional variants in the MIR137 gene. On the basis of their location and the allele frequency differences between patients and controls, we explored the hypothesis that the discovered variants impact the expression of the mature miRNA and consequently influence global mRNA expression affecting normal brain functioning. Using neuronal-like SH-SY5Y cells, we demonstrated significantly reduced mature miR-137 levels in the cells expressing the variant miRNA gene. Subsequent transcriptome analysis showed that the reduction in miR-137 expression led to the deregulation of gene sets involved in synaptogenesis and neuronal transmission, all implicated in psychiatric disorders. Our functional findings add to the growing data, which implicate that miR-137 has an important role in the etiology of psychiatric disorders and emphasizes its involvement in nervous system development and proper synaptic function.

Molecular Psychiatry (2015) 20, 472–481; doi:10.1038/mp.2014.53; published online 3 June 2014

INTRODUCTION association studies and genome scans provided an indication of Schizophrenia (SZ) and bipolar disorder (BD) are two separate additional miRNAs being relevant for both etiologies on the basis – entities, but the presence of coinciding symptoms, frequent of their genomic location.18 21 co-occurrence of both disorders in affected families, similar Variants in the miRNA seed sequence can alter its target epidemiological risks in relatives and overlapping genetic findings spectrum, but the importance of the correct hairpin structure, suggest that they could have at least a partially uniform genetic thermodynamic influences on strand loading and base-pairing base.1,2 The polygenic nature of both disorders, missing reprodu- requirements outside the seed sequence imply that variants cibility in genetics research and often conflicting results originat- located outside the seed region can also influence miRNA ing from a plentitude of meta-analyses, linkage and association biogenesis and/or targeting and therefore are, as such, important studies shows that the search for complex disorder genes is far factors influencing miRNA function.22–24 miRNA profiling using from trivial. The realization that only a low percent of the genome massively parallel sequencing provided an insight into the hetero- codes for proteins and that nonprotein coding genes have an 3,4 geneity in length and sequence of mature miRNAs originating important but poorly understood role in gene regulation, from the same gene (isomiRs).25–27 The genesis of isomiRs can be initiated the interest in the nonprotein coding DNA. MicroRNA caused by different biological phenomena, including imprecise genes (miRNAs) regulate a variety of processes in eukaryotes by cleavage by Drosha or Dicer, which may be influenced by genetic mediating mRNA cleavage, translational repression and transcrip- 28 tional or translational activation.5–9 Strong evidence for the variants. Whether the presence of isomiRs is a regulated process involvement of miRNAs in SZ and/or BD emerged from several and whether isomiRs have a functional role in the regulation is a fi separate observations. First, the identification of brain-expressed matter of a discussion in the scienti c community, but reports miRNAs revealed specific miRNA subsets, involved in normal brain show that at least some of the isomiRs can affect miRNA stability, function.10–13 Second, the studies exploring the importance of target spectra and efficiency of incorporation into the RISC 28–30 miRNA regulation in neurological processes and studies of miRNA complex, when compared with the canonical miRNA. expression signatures in different neuropsychiatric disorders On the basis of expression, association, linkage and functional discerned possible implication of specific miRNAs in SZ and studies, 13 miRNA genes were selected for variant screening that BD.14–17 Third, the combination of the increasing number of impact mature miRNA levels. With this study we provide com- miRNA genes being identified, and data from previously published pelling evidence that variants in the MIR137 gene, observed in BD

1Applied Molecular Genomics Unit, Department of Molecular Genetics, VIB, Antwerp, Belgium; 2University of Antwerp, Antwerp, Belgium; 3Biomedical Sciences Research Group, School of Medicine, Universidad Antonio Nariño, Bogotá, Colombia; 4Department of Clinical Sciences, Division of Psychiatry, Umeå University, Umeå, Sweden; 5Peripheral Neuropathy Group, Department of Molecular Genetics, VIB, Antwerp, Belgium; 6Stanley Center for Psychiatric Research, Broad Institute of MIT and Harvard, Cambridge, MA, USA; 7Program in Medical and Population Genetics, Broad Institute of MIT and Harvard, Cambridge, MA, USA; 8Department of Genetics, Harvard Medical School, Boston, MA, USA and 9VIB Laboratory of Behavioral and Developmental Genetics, KU Leuven, Leuven, Belgium. Correspondence: Professor Dr J Del-Favero, Applied Molecular Genomics Unit, Department of Molecular Genetics, VIB, University of Antwerp, Campus Drie Eiken, Universiteitsplein 1, B-2610 Antwerp, Belgium. E-mail: [email protected] 10These authors contributed equally to this work. 11Multiplicom, Niel, Belgium. 12BRC hiPSC Core Facility, University of Cambridge, Cambridge, UK. Received 7 October 2013; revised 24 April 2014; accepted 28 April 2014; published online 3 June 2014 Reduced miR-137 deregulates synaptic gene sets M Strazisar et al 473 and SZ patients, reduce the levels of the mature miR-137 and catalog number CRL-1573) to test the expression differences in cells known affect gene sets involved in synapse formation and function. not to express detectable levels of miR-137.

MATERIALS AND METHODS RNA isolation Total RNA from untransduced replicates, stably transduced SH-SY5Y and Subjects HEK replicates; for each of the variants, was extracted with the mirVana The association sample, recruited from the Västerbotten region in Northern miRNA Isolation Kit (Life Technologies) according to the manufacturer’s Sweden, included 345 BD patients, 426 SZ patients and 1376 healthy protocol. Aliquots of total RNA were subsequently used for miRNA individuals, all unrelated. The patients were clinically characterized by expression analysis, transduction efficiency normalization and transcrip- trained research nurses and research psychiatrists using register data, data tome profiling. Additional information on RNA isolation is provided in from psychiatric records and semi-structured interviews. Additional Supplementary File S1. information of the selection is provided in Supplementary File S1. The study was approved by the Medical Ethical Committees of the Expression of miR-137 and EGFP universities of Umeå and Antwerp. MiR-137 and EGFP expression was analyzed on ABI Prism 7900 HT Sequence Detection System with either TaqMan microRNA assays or SYBR Gene selection green (Life Technologies). Thirteen miRNA genes were selected on the basis of two criteria: (1) RT–qPCR (quantitative PCR with reverse transcription) reactions were expression in the human brain and (2) location in genomic regions performed on technical triplicates for each biological replicate (wild-type previously associated or linked with BD and/or SZ. As a result of the and variant transduced cells) and for each assay. Reverse transcription for literature mining MIR34A, MIR99B, MIR103A1, MIR103B1, MIR128-1, MIR132, miRNA expression was performed according to the manufacturer’s MIR135B, MIR137, MIR301A, MIR448, MIR764, MIRLET7E and MIRLET7A2 protocol (Life Technologies) with reverse transcription primers for were selected for Sanger sequencing-based variant discovery. The miR-137 (mmu-miR-137 assay, ID 001129) and two endogenous controls references of the published data used for the selection are provided in miR-16 (hsa-miR-16 assay, ID 00391) and RNU24 (RNU24 assay, ID 001001). Supplementary File S1. To determine transduction efficiency we used relative EGFP quantification as transduction efficiency factor. The reaction was done using SYBR Green I mix (Life Technologies) and appropriate primers in final concentrations of Sanger sequencing-based variant discovery 0.3 μM per reaction. The primers used for relative quantification are listed in The DNA sequences containing 500 bases flanking the 3′ and 5′ regions of Supplementary File S1. selected miRNA stem-loop sequences were downloaded from the UCSC The RT–qPCR data were normalized by geometric averaging of multiple genome browser (February 2009, GRCh37/hg19 assembly). The PCR and internal control genes.36 Processing of raw data and calculation of norma- 31 sequencing primers were designed using Primer3 software. lized relative quantities were done with ddCt-method (ΔΔCt) including the The 13 selected miRNA genes were PCR amplified and sequenced from amplification efficiencies of all primer pairs and normalization with several DNA extracted from blood obtained from 346 BD patients, 486 SZ patients endogenous controls to adjust the relative quantification.37 The NRQ of and 514 control subjects. After obtaining primary results of 1346 indivi- EGFP was used to normalize miR-137 expression values for each clone. The duals, we analyzed the MIR137 gene on an additional set of 832 unrelated statistical significance of the results was calculated using two-sample two- control individuals. Detailed information about the amplification protocols, tailed Student's t-test. The generation of SH-SY5Y stable cells transduced primers and sequencing protocols is provided in Supplementary File S1. with MIR137 gene construct for wild-type and variant alleles and the miR-137 expression analysis was independently replicated. Mature miR-137 Fragment analysis levels were also investigated in untransduced SH-SY5Y and cells transduced with an empty EGFP vector (mock transduced), to determine if the For the analysis of the MIR137 VNTRs (including rs66642155, rs58335419, fi fi fl overexpression of miR-137 is signi cantly higher in cells transduced with rs71738863), the target region was ampli ed with FAM uorescent-labeled wild-type MIR137 construct. Relative quantification of the mature miR-137 primers. Fragment analysis was carried out on the 3730XL DNA Analyzer levels between the − 4C>T MIR137 and wild-type MIR137 transduced (Life Technologies, CA, USA). Detailed information about the VNTR typing is HEK293 cells was investigated to assess the effect of the variant in a cell supplied in the Supplementary File S1. line, known not to express detectable levels of MIR137. Two-tailed Fisher’s exact test was used to test − 4C/T allele association between patients and controls. CLUMP v2.432 was used to test the polymorphic VNTR association between patients and controls. CLUMP test Microarray analysis of allelic association reports P-values empirically derived using a Monte Three-microgram aliquots of the total RNA that was initially used for Carlo simulation approach. The value reported in the Result section is the miR-137 expression analysis were also used for transcriptome profiling on significance value using T1 statistics with six degrees of freedom. The the HG-U219 microarray (Affymetrix, CA, USA). Reverse transcription, labeling, number of simulations used for each test was 10 000. quality checks, hybridization, scanning, image analysis, data preprocessing, normalization and annotation of the results was done at the VIB RNA secondary structure prediction Nucleomics Core (http://www.nucleomics.be/, Leuven, Belgium). Normal- 33 ized microarray data were analyzed with SAM, a statistical method The ViennaRNA Package 2.0 was used for prediction of the secondary developed for microarray analysis by Tusher et al.,38 implemented in MeV structures of wild-type and variant miRNAs. Secondary structure predic- v4.8.1.39 The input and detailed information about the analysis are tions were made using three different prediction algorithms; minimum free described in detail in Supplementary File S1. energy, centroid structure and maximum expected accuracy structure. Secondary structure predictions were based on MIR137 stem-loop sequence and 100 bases flanking the 3′ and the 5′ ends of the stem-loop. The Gene set enrichment analysis secondary structures were visualized using VARNA, a visualization applet The gene set enrichment analysis (GSEA) computational method40,41 using used to draw the secondary structure of RNA sequences.34 C5 Gene Ontology gene sets (c5.all.v3.0.symbols.gmt) was adopted to find the gene sets that significantly differ between the variant and wild-type − Generation of SH-SY5Y stable cell lines expressing mature miR-137 MIR137 microarray data sets ( VNTRs or 4C>T; see http://www. broadinstitute.org/gsea/index.jsp for the detailed protocol40,41). Settings Cell culture conditions and the generation of stable human neuroblastoma for the GSEA analysis is provided in Supplementary File S1. SH-SY5Y cells, transduced with lentiviruses encoding an empty vector or MIR137 with or without variants, are described by Almeida-Souza et al.35 and in Supplementary File S1. Media and supplements were purchased Prediction of miR-137 targets from Invitrogen (Life Technologies). The SH-SY5Y cell line was purchased To determine which significant differentially expressed or enriched genes from ATCC (American Type Culture Collection, MA, USA, catalog number (significant genes) could be miR-137 targets, we searched for miR-137 seed CRL-2266). Wild-type MIR137 and − 4C>T MIR137 constructs were also and non-seed matches in the 3′ UTR regions. Detailed information on the transduced in HEK293 cells (American Type Culture Collection, MA, USA, complementarity analysis is provided in Supplementary File S1 and the

© 2015 Macmillan Publishers Limited Molecular Psychiatry (2015), 472 – 481 Reduced miR-137 deregulates synaptic gene sets M Strazisar et al 474 results in Supplementary File S4. To complement the results of the 3′ UTR We obtained rs1625579 and rs1198588 genotypes for 61% analysis, six freely available prediction tools were used (PITA, DIANAmT, controls, 79% BD patients and 96% SZ patients, all with known MIRANDA, MIRWALK, PICTAR and TARGETSCAN). In addition, we explored MIR137 VNTR and − 4C/T genotype. The most frequently observed fi whether the set of signi cant genes is enriched for the predicted miR-137 rs1625579 genotype in patients and controls is the TT genotype fi targets when compared with the remaining (nonsigni cant) genes. The (61.7% in controls and 65.6% in patients) with a T-allele frequency results are presented in Supplementary File S4. of 78.2% in controls and 80.5% in patients, which is comparable with previous reports.20,42–44 The frequency of the rs1198588 RESULTS A-allele, recently found to be associated with SZ,21,45 in our Variant discovery and secondary structure prediction population is also consistent with previously published reports (22.2% in controls and 19.8% in patients). On the basis of these To discover variants in the miRNA genes with potential effects results, the variants are not associated with SZ or BD. Furthermore, on the mature miRNA levels, we performed Sanger sequencing- all individuals, patients and controls carrying the − 4C/T MIR137 based variant analysis of 13 brain-expressed, functional and/or variant, carry the TT rs1625579 and TT rs1198588 genotype and no positional candidate miRNA genes in 771 BD and SZ patients fi fi signi cant association was observed for rs1625579 and rs1198588 and 1376 controls. Sequencing resulted in the identi cation of genotypes when patients and controls were grouped on the basis nine variants in six miRNA genes (Table 1), but only the MIR137 fi of the VNTR genotypes. Additional information on the rs1625579 gene harbored variants that ful lled criteria for further analysis: (1) and rs1198588 genotyping and association analysis is presented change in the predicted secondary structure and (2) higher variant in Supplementary File S2. frequency in the patient population compared with control population. The MIR137 − 4C/T variant (rs185304769) located in the primary Effect of the MIR137 variants on the expression levels of mature sequence was found in three SZ, two BD patients and in three miR-137 control individuals and changes the secondary structure as The MIR137 variants − 4C>T, 4VNTR and 8VNTR were used to study predicted by all three prediction algorithms (Figure 1, structures their effect on the expression levels of the mature miR-137. 1-3B). This variant is more frequent in patients (0.65 versus 0.22% Untransduced SH-SY5Y, mock transduced SH-SY5Y, stable trans- in controls), but not reaching statistical significance (Fisher’s exact duced SH-SY5Y and stable transduced HEK293 cells were grown in test, two-tailed P-value = 0.145) in our cohort. The VNTR triplicates. The transcriptional fusion between the miRNA gene (chr1:98511730-98511794; rs66642155, rs58335419, rs71738863) (with or without the variant) and EGFP (fused at the 3′-terminal of located six bases upstream of the pre-miR-137 is also predicted to the MIR137 sequence), allowed the normalization of the mature substantially change the MIR137 secondary structure, depending miR-137 expression levels based on the transduction efficacy. on the number of repeats (Figure 1, structures 1-3C for 4VNTR and Significantly lower mature miR-137 levels (Student's t-test, 1-3D for 8VNTR). CLUMP analysis shows association between the Po0.05) were observed for all variants when compared with VNTR allele distribution in our patient cohort with an empirical the cells transduced with the wild-type MIR137 construct. Levels P-value (T1 subtest) of P = 0.017 (BD versus controls), P = 0.040 (SZ of miR-137 decreased 0.75-fold (4C>T MIR137), 0.42-fold (4VNTR versus controls) and P = 0.018 (combined patient group versus MIR137) and 0.63-fold (8VNTR MIR137) in mutant MIR137 controls). The highest X2 in the T3 statistics was obtained when transduced cells compared with the wild-type MIR137 transduced 8VNTR repeat allele frequency was tested between patients and cells (Figure 2). controls (P = 0.049), suggesting that the eight repeat allele To substantiate the findings, a completely independent analysis contributes to the significance of the VNTR distribution. In total, starting from construct amplification, cloning and construct trans- 17 control individuals (1.24%) and 20 patients (2.59%) have at duction into the SH-SY5Y wild-type cells, RNA extraction and least one 8VNTR allele. The results of the VNTR fragment analysis RT–qPCR analysis was performed. This independent experiment in the MIR137 gene are shown in Supplementary File S2. on stable transduced SH-SY5Y cells confirmed the lower mature The implication of rs1625579 and rs1198588 in SZ20,21 and miR-137 levels in the cells transduced with the variant constructs, location within or close to MIR137HG, led us to investigate a with even larger expression differences (with miR-137 average fold possible association of these variants in our association sample. changes of 0.1 in the cells transduced with − 4 C>T MIR137, 0.2 for

Table 1. Positions, types and minor allele frequencies of the variants found in the miRNA genes

miRNA gene Location Precursor size miRNA RefSeq ID Variantsa Structural changeb Minor allele frequency

BD (n = 345) SZ (n = 426) CO (n = 514+862)c

MIR137 1p21.3 102 bp NR_029679.1 − 4 C/T Yes T: 0.003 T: 0.004 T: 0.001 − 6 VNTRd Yes 4–9 VNTR: 0.227 4–9 VNTR: 0.278 4–9 VNTR: 0.229 MIR135B 1q32.1 97 bp NR_029893.1 − 60C/T No T: 0.229 T: 0.263 T: 0.251 MIR128-1 2q21.3 82 bp NR_029672.1 +25 G/A No A: 0.000 A: 0.001 A: 0.000 − 114A/G No G: 0.003 G: 0.002 G: 0.006 MIRLET7A2 11q24.1 72 bp NR_029477.1 − 86 G/A No A: 0.004 A: 0.000 A: 0.001 − 41 T/A No A: 0.000 A: 0.000 A: 0.002 MIR99B 19q13.41 70 bp NR_029843.1 − 86 A/G No G: 0.012 G: 0.009 G: 0.006 MIR764 Xq23 85 bp NR_031581.1 +137 G/A No A: 0.009 A: 0.009 A: 0.014 Abbreviations: BD, bipolar disorder; CO, control; SZ, schizophrenia. aThe variant position is given relative to the 5’ end of the precursor miRNA. bStructural change: The column denotes whether the variant introduced the structural change when the sequence with the variant was compared to the wild-type miRNA gene sequence. cThe total number of the controls screened was 514 for all but MIR137, where the number was increased to 1376. dVNTR allele frequencies and genotypes in SZ, BD and control population are presented in detail in Supplementary File S2, Supplementary Table S2_1. No variants were observed in MIR34A, MIR103A1, MIR103B1, MIR301A, MIRLET7E, MIR132 and MIR448 genes.

2a 2b 2c 2d

3a 3b 3c 3d

Figure 1. MFE, centroid and MEA secondary structure predictions of the mir-137 wild-type and variant stem-loop sequences with 100 bp 3′ and 5′ ﬂanking bases. Speciﬁc regions of miRNA stem-loop sequence are highlighted in different colors (light blue—MIR137 precursor hairpin, purple—the mature MIR137 sequence, yellow—the terminal loop sequence, red—the MIR137 seed sequence). The comparison between secondary structure predictions of MIR137 wild-type sequence (that is, 3VNTRs) using MFE (1A), centroid (2A) and MEA prediction algorithm (3A), MIR137 with − 4C>T variant using MFE (1B), centroid (2B) and MEA algorithm (3B); MIR137 with 4VNTR using MFE (1C), centroid (2C) and MEA algorithm (3C) and MIR137 with 8VNTR using MFE(1D), centroid (2D) and MEA algorithm (3D) show that the secondary structure changes due to the variants in the primary sequence. MEA, maximum expected accuracy; MFE, minimum free energy. the cells transduced with 4VNTR MIR137 and 0.1 for the cells higher in the cells transduced with the wild-type MIR137 construct transduced with 8VNTR MIR137 construct; additional results are compared with the mock transduced and untransduced SH-SY5Y presented in Supplementary File S3, Supplementary Figure S3_1). cells. The mature miR-137 levels in the variant MIR137 cells are Even though the SH-SY5Y cells have endogenous miR-137 close to the levels detected in the mock transduced or expression, the mature miR-137 levels are 7.59-fold and 5.44-fold untransduced SH-SY5Y cells, with fold changes between 1.05

Figure 2. Comparison of the mature miRNA normalized relative quantities (NRQs) in the SH-SY5Y cells transduced with MIR137 wild-type and respective variant constructs. The graphs present average miR-137 expression values of cells transduced with wild-type MIR137, 4VNTR or 8VNTR MIR137 constructs (graph A) and cells transduced with wild-type MIR137 and − 4C>T MIR137 construct (graph B). The average fold change between the − 4C>T MIR137 cells and wild-type cells was 0.75 (Student's t-test, P-value = 0.01862). The average fold change between the 4VNTR MIR137 cells and wild-type cells was 0.42 (Student's t-test, P-value = 0.00302) and 0.63 between 8VNTR MIR137 cells and wild-type cells (Student's t-test, P-value = 0.01862). Signiﬁcant differences between mature miR-137 levels in SH-SY5Y cells transduced with variant and wild-type constructs are marked with asterisks (*Po0.05 and **Po0.01). Average fold changes (fc) between the wild-type cells transduced with the wild-type construct and variant cells are presented above asterisks.

Figure 3. Two graphs showing average normalized expression of the significant genes after comparison of the microarray data sets between SH-SY5Y cells transduced with the 4- and 8VNTR MIR137 (a), − 4C>T MIR137 (b) and the wild-type (WT) constructs. White bars represent normalized expression for each significant gene in the cells transduced with the wild-type MIR137 constructs. Black bars represent expression values for significant genes in the cells transduced with the − 4C>T MIR137 constructs (b) or 8VNTR MIR137 constructs (a) and the light gray bars represent the normalized expression values for the significant genes in the cells transduced with the 4VNTR MIR137 constructs (a). The error bars present standard deviation of biological triplicates for each gene.

and 2.8 (Supplementary File S3, Supplementary Table S3_1, cance analysis of microarrays (SAM) was performed to assess the Supplementary Figure S3_2). direct or indirect effects of lowered mature miR-137 levels (due to The mature miR-137 levels were also compared in stable the MIR137 variants) on global gene expression. Transcriptome transduced HEK293 cells to investigate the impact of the variant comparisons revealed only a handful of genes (sixteen) with on the miR-137 levels in cells, known to lack endogenous miR-137 significant differences in expression between cells transduced expression. Cells transduced with − 4C>T MIR137 constructs with the wild-type and variant MIR137 constructs (Figure 3, had significantly lower miR-137 levels when compared with Supplementary File S3, Supplementary Tables S3_2). Of these, the cells transduced with the wild-type construct (Student's t-test, three genes (SPAG4, NPPA and S100A10) had an expression P-value = 0.0018, fold change = 0.89) (Supplementary File S3, difference above the significance threshold when comparing Supplementary Figure S3_3). microarray data between − 4C>T MIR137 and wild-type MIR137 transduced cells (Graph B on Figure 3, Supplementary File S3, Results of microarray transcriptome profiling Supplementary Table S3_2); all three genes were overexpressed in Aliquots of the same total RNA, extracted from the transduced SH- the cells transduced with the variant constructs. When examining SY5Y triplicates, and previously used in the miR-137 expression the effect of number of VNTR repeats, we observed fifteen genes analysis, were used in subsequent microarray experiments. Signifi- with significant differences in expression. Three genes had lower

Table 2. Results of the GSEA showing the top ranked gene sets and the genes contributing to the core enrichment

GSEA analysis C5 enriched Gene set NES FDR NOM CE genes size P-value

mir-137: −4C/T Extracellular structure 30 2.173 0.002 o0.001 ACAN, ACHE, CHRNA1, KLK8, MYH11, NLGN1, PCDHB10, versus wt organization and PCDHB11, PCDHB13, PCDHB16, PCDHB2, PCDHB5, PCDHB6, biogenesis POU4F1 Synaptogenesis 18 2.127 0.003 o0.001 ACHE, NLGN1, PCDHB10, PCDHB13, PCDHB16, PCDHB2, PCDHB5, PCDHB6, POU4F1 Synapse organization 23 2.185 0.003 o0.001 ACHE, CHRNA1, KLK8, NLGN1, PCDHB10, PCDHB11, PCDHB13, and biogenesis PCDHB16, PCDHB2, PCDHB5, PCDHB6, POU4F1

mir-137: 4- and Synapse organization 23 1.834 0.248 o0.001 ACHE, NLGN1, PCDHB10, PCDHB13, PCDHB2, PCDHB3, 8VNTR versus and biogenesis PCDHB5 3VNTR Transmission of nerve 186 1.834 0.159 o0.001 ACCN1, AKAP5, AMIGO1, APOE, BAIAP3, CACNA1B, CALY, impulse CARTPT, CBLN1, CD9, CHRNE, CPLX1, DLG4, DRD2, DRD4, DTNA, GAD1, GLRB, GPR176, GRIA2, GRIK2, GRIN2A, GRM1, HAP1, HRH3, HTR3A, HTR3B, KCNA1, KCNMB1, KCNMB3, KCNMB4, KCNN3, KCNQ3, KCNQ5, KLK8, NLGN1, NPBWR2, NQO1, PCDHB10, PCDHB13, PCDHB2, PCDHB3, PCDHB5, PLP1, PMP22, RAB3A, SCN1B, SLC1A1, SLC1A2, SLC6A5, SST, SYN1, SYN2, SYT5 Synaptic transmission 171 1.834 0.212 o0.001 ACCN1, AKAP5, APOE, BAIAP3, CACNA1B, CALY, CARTPT, CBLN1, CHRNE, CPLX1, DLG4, DRD2, DRD4, DTNA, GAD1, GLRB, GPR176, GRIA2, GRIK2, GRIN2A, GRM1, HAP1, HRH3, HTR3A, HTR3B, KCNA1, KCNMB1, KCNMB4, KCNN3, KCNQ3, KCNQ5, KLK8, NLGN1, NPBWR2, NQO1, PCDHB10, PCDHB13, PCDHB2, PCDHB3, PCDHB5, PLP1, PMP22, RAB3A, SCN1B, SLC1A1, SLC1A2, SLC6A5, SST, SYN1, SYN2, SYT5 Regulation of blood 22 1.828 0.137 0.002 ACE, AGT, CALCA, CARTPT, GCGR, NPPA, PTGS1, REN pressure Abbreviations: C5 enriched, name of the top ranked gene sets when based on C5 gene ontology set collections; CE genes, genes that contributed to the enrichment of the specific gene set; FDR: false discovery rate; Gene set size, number of the genes in the gene set; GSEA, gene set enrichment analysis; NES, normalized enrichment score; NOM P-value, nominal P-value; wt, wild-type. The gene sets met significance threshold of FDRo0.05 and NOM P-valueo0.001 or FDRo0.25 and NOM P-valueo0.01 (when no gene sets met the more stringent threshold). The first column defines the compared microarray data sets. In bold are the genes that contributed to gene set enrichments when comparing microarray results originating from cells transduced with different constructs.

and twelve had higher expression in cells transduced with the transmission gene sets (94% of the genes are present in both gene constructs with increasing number of VNTR repeats (Graph A on sets). Over 50% overlap in genes contributing to the core Figure 3, Supplementary File S3, Supplementary Table S3_3). In enrichments are seen also between − 4 C>T MIR137 and variant total, ten out of sixteen genes are predicted miR-137 targets by at VNTR MIR137 containing cells. The most frequently observed core least one prediction method, revealing a substantial enrichment enrichment genes, contributing to the core enrichment of six with a probability estimation of 0.63. To compare, the BioMart significant gene sets are NLGN1 and four members of the database contains 22 691 human protein coding genes (GRCh37. protocadherin-β cluster (PCDHB2, PCDHB5, PCDHB10, PCDHB13). p11), and the microRNA.org database contains 5487 genes that Three additional members (PCDHB3, PCDHB11 and PCDHB16) are predicted targets of miR-137 (microRNA.org, http://www. contribute to the core enrichment of two or more significant gene microrna.org/), so the probability of a random protein coding sets and in total, four of them are predicted as targets of miR-137 gene is a target of miR-137 is 0.24 (Supplementary File S4). (PCDHB3 is predicted by three, PCDHB11 by four, PCDHB13 by six and PCDHB10 by all prediction algorithms; Supplementary File S3 Results of GSEA and Supplementary File S4). As one miRNA can affect several hundred targets, rigorous fi statistical analysis of the microarray data proved to be insuf cient DISCUSSION to discover genes that are directly affected by decreased miR-137 levels. Therefore we explored the hypothesis that an a priori This study started with Sanger-based sequencing of 13 miRNA defined set of genes may show significant differences in expres- genes in a cohort of patients diagnosed with SZ or BD and sion between the wild-type and variant cells. Significant gene sets controls to detect variants that can affect the maturation and/or 22,23,46 and the genes contributing to the enrichments are presented in function of the respective miRNA. Table 2. Detailed results are presented in Supplementary File S3 Out of thirteen miRNA genes expressed in the brain and located under the GSEA section. in regions associated with SZ or BD, we observed two variants in The gene sets that are enriched in the variant MIR137 contain- MIR137 predicted to alter the secondary structure of the primary ing cells belong to pathways involved in neuronal processes. transcript (Figure 1, Supplementary File S2, Supplementary Table Specifically, they belong to pathways governing synaptogenesis, S2_8) and which were more frequently observed in patients than synapse organization and neuronal/synaptic transmission. The in controls (Table 1, Supplementary Table S2_1). The MIR137 gene, synapse organization and biogenesis gene set is significant for a new ‘hot topic’ gene in psychiatric genetics, is expressed almost both, cells transduced with − 4C>T and VNTR MIR137 constructs. exclusively in the brain,13 is involved in neuronal proliferation and Out of 69 genes belonging to six significant gene sets, 57 differentiation47,48 and regulates neuronal maturation.49 contribute to the core enrichment of at least two gene sets. The A genome-wide meta-analysis from 2011 highlighted largest overlap in genes contributing to the gene-set enrichment rs1625579, located within the primary transcript of MIR137HG, is observed between transmission of nerve impulse and synaptic as the strongest new SZ-associated variant.20 A more recent study,

© 2015 Macmillan Publishers Limited Molecular Psychiatry (2015), 472 – 481 Reduced miR-137 deregulates synaptic gene sets M Strazisar et al 478 an independent multistage genome-wide association study dementia.62 Even with the lack of publications reporting miR-137 comprising over 5000 Swedish SZ cases followed by over 16 000 deregulation in the brain of SZ or BD patients, miR-137 remains an additional SZ cases published by Ripke et al.,21 confirmed the outstanding candidate for psychiatric disorders, due to the results importance of miR-137 and provided additional evidence presented in our study, its involvement in the regulation of genes supporting a possible role of miR-137 in SZ through a second important in brain function and neuronal development (reviewed common variant, located upstream of the MIR137 gene in Wright et al.63) and recently published genome-wide associa- (rs1198588). We explored the rs1625579 and rs1198588 genotypes tion study results.20,21 in our association sample between the VNTR and − 4C/T Differences in the miR-137 levels triggered an interest in and rs1625579 and rs1198588 genotypes (Supplementary File comparing the transcriptomes of the cells transduced with wild- S2, Supplementary Tables S2_3–S2_7). The rs1625579 and type and variant constructs. Using SAM, we could conclude that rs1198588 genotype distributions are in accordance with pre- individual gene expression differences seen on the transcriptional viously published results,21,43,44,50 but failed to show significant levels are insufficient to detect relevant biological function. This association with SZ and/or BD (Supplementary Table S2_4). We observation is not surprising, knowing that an individual miRNA is also failed to find any significant association between genotypes able to downregulate the transcription and/or translation of after clustering patients and controls on the basis of VNTR hundreds of genes but the repression is usually mild.64–66 genotype (Supplementary File S2, Supplementary Tables S2_6 Nevertheless, out of 16 genes that showed different expression and S2_7). The result is not surprising, knowing that several between the cells with lower miR-137 levels, 10 are predicted to publications concluded that the variants are actually not be targeted by miR-137 using different prediction methods associated with the SZ phenotype, but rather with specificSZ (Supplementary File S4) and six of the predicted targets had pre- endophenotypes.51 Rs1625579 specifically has been associated viously been linked to psychiatric disorders via association, linkage with miR-137 expression,52 with an impact on brain function and or expression studies (Supplementary File S4 and Supplementary brain structure,43,44 with neurocognitive performance within File S5). Noise produced by microarray technology can disguise specific psychosis patients groups,53 with severe negative moderate but important transcriptome changes due to divergent symptoms and cognitive deficits50 and with an effect on onset miRNA expression. Furthermore, gene-set effects on the biologi- age in SZ patients.43 To our belief, both variants will less likely cally relevant pathways may be missed when using single gene interfere with the biogenesis of miR-137, simply because of the analysis, due to the modest single gene expression differ- distance to the precursor (rs1625579 is located almost 9 kb ences.41 Gene enrichment methodology allowed us to determine downstream and rs1198588 is located 39 kb upstream of the the effects of the reduced miR-137 levels induced by the variants. precursor MIR137 sequence). Several studies experimentally Enrichment analyses of the genome-wide association study validated SZ-associated genes as miR-137 targets,54,55 elevating results, published by Lips et al.67 and Ripke et al.,21 showed the importance of miR-137 in relation to psychiatric disorders. association between synaptic gene sets and SZ and proposed the The variants characterized and functionally validated in our deregulation of synaptic gene groups as a risk factor for SZ.67 Our study are not novel, but functional and genetic data regarding the study shows that lowering the levels of miR-137, due to variants in SZ/BD phenotypes and the effect on the miR-137 is very limited. the pri-mir-137, cause the deregulation of gene sets involved in To our knowledge, rs185304769 variant has never been investi- the genesis and proper functioning of the synapses. gated in association with miR-137 expression or SZ/BD pheno- To further substantiate the relevance of variant induced types and the results of a VNTR analysis have only recently been miR-137 downregulation, several additional analyses were per- reported for the Japanese SZ population.56 The allele VNTR distri- formed. To detect all possible miR-137 targets among the bution was comparable with the one presented in our study, but significant genes, we used different prediction methods and to the authors failed to identify an association between the identify the genes that relate to SZ, we inspected all significant investigated variants and SZ. hits in SZ gene resource, an online database of SZ-related genetics To know the status of the miR-137 levels, we used an assay studies.68 Finally, we used a search tool for the retrieval of specifically targeting human mature miR-137. Although SH-SY5Y interacting genes/proteins (STRING69) to create association net- cells have a detectable basal expression of miR-137 (Supplemen- works of the significant genes and their known interaction tary File S3, Supplementary Table S3_1, Supplementary Figure partners to discover additional functional evidence of possible S3_2), the basal expression of the mature miR-137 in the mock miR-137 involvement in psychiatric disorders. The results of the transduced and untransduced SH-SY5Y cells was considerably target predictions, cross-referencing the SZ gene resource and lower than in the MIR137 transduced cells. Furthermore, we were using STRING on the top possible miR-137 targets are presented in able to ascertain the same effect of the variant, lower levels of the Supplementary File S4 and Supplementary File S5 and miR-137 in the cells transduced with a variant construct, in an discussed below. independent replication in SH-SY5Y cells and showed the same Eight out of 16 genes that contribute to the enrichment of gene effect using HEK293 cells (Supplementary File S3, Supplementary sets involved in synaptogenesis belong to the protocadherin-β Figure S3_3). SH-SY5Y cells have neuronal-like features, which family of which four (PCDHB10, PCDHB13, PCDHB11 and PCDHB3) allows to assess transcriptome differences between the MIR137 are predicted to be miR-137 targets (Supplementary File S3). Little variant and wild-type construct, providing in-depth information is known about the specific functions of these genes but they are on the effect of miR-137 on genes involved in neuronal processes. predominantly expressed in the nervous system and implicated in The effect of the VNTRs on miR-137 processing has previously correct targeting and function of synapses.70,71 Interestingly, the been described in melanoma cell lines, where an association genes of the protocadherin-β cluster are located at 5q31, which is between higher numbers of VNTRs and lower miR-137 quantity one of 11 genomic regions linked to SZ according to the genome was observed using northern blot.57 We observed the same effect scan meta-analysis results published by Lewis et al. in 2003.18 for − 4C>T and VNTR variants using stable transduced SH-SY5Y The glial glutamate transporters, encoded by SLC1A1 and cells (Figure 2, Supplementary File S2). Downregulation of miR-137 SLC1A2, the latter known as one of the candidate genes in SZ72 was, before our report, implicated mainly in carcinogenesis, and and predicted as miR-137 targets by seven methods, have been miR-137 was described as a tumor suppressor in several types implicated in the pathology of SZ, BD and anxiety disorder of cancer, including neuroblastomas, astrocytomas and glio- (Supplementary File S4).73–75 Other predicted miR-137 targets, blastomas.58,59 Downregulation of miR-137 was also reported such as KCNA1, PLP1, GRIN2A, GRIA2 and CPLX1, are all genes in Alzheimer’s disease patients, where it was proposed as a highly expressed in the brain, relevant for psychiatric disorders potential biomarker for early diagnosis60,61 and in HIV-associated and associated with profound roles in nervous system

Molecular Psychiatry (2015), 472 – 481 © 2015 Macmillan Publishers Limited Reduced miR-137 deregulates synaptic gene sets M Strazisar et al 479 76–80 processes. They seem to be functionally related were recruited, is supported by grants from the Swedish Research Council (grant (Supplementary File S3, Supplementary File S4, Supplementary numbers 345-2003-3883, 315-2004-6977) and the Bank of Sweden Tercentenary Figures S4_1 and S4_2) and involved in SZ genetics Foundation, the Swedish Council for Planning and Coordination of Research, the (Supplementary File S4). The subset of genes (17 out of 84) Swedish Council for Research in the Humanities and Social Sciences and the Swedish predicted as miR-137 targets by at least four different prediction Council for Social Research. methods, have all been implicated in synaptic functions, further supporting the relevance of miR-137 regulation in neurological function (Supplementary Figures S3 and S4). Interestingly, a REFERENCES prominent overlap has been observed between the known functions of the genes identified in our study and the genes 1 Lichtenstein P, Yip BH, Bjork C, Pawitan Y, Cannon TD, Sullivan PF et al. Common fi genetic determinants of schizophrenia and bipolar disorder in Swedish families: a highlighted in a recently published paper, where signi cant population-based study. Lancet 2009; 373:234–239. numbers of predicted and known miR-137 targets were evaluated 2 Purcell SM, Wray NR, Stone JL, Visscher PM, O'Donovan MC, Sullivan PF et al. in relation to their involvement in gene-networks or pathways, Common polygenic variation contributes to risk of schizophrenia and bipolar 63 disturbed in SZ. The overlap between this study and the study disorder. Nature 2009; 460:748–752. by Wright et al.63 includes genes involved in several pathways 3 Mattick JS. Non-coding RNAs: the architects of eukaryotic complexity. EMBO Rep such as glutamate signaling (GRM1, GRIK2, GRIA, GRIN2A) and cell- 2001; 2:986–991. to-cell signaling and interaction (SYN1, SYN2, DLG4, CACNA1B, 4 Perkins DO, Jeffries C, Sullivan P. Expanding the 'central dogma': the regulatory SYT5, KCNA1, SCN1B). role of nonprotein coding genes and implications for the genetic liability to 10 – With the current study we provide strong support for the schizophrenia. Mol Psychiatry 2005; :69 78. 5 Pillai RS. MicroRNA function: multiple mechanisms for a tiny RNA? RNA 2005; 11: involvement of miR-137 in the etiology of psychiatric disorders. – − fl 1753 1761. The mechanism of how the VNTR and 4C>T variants in uence 6 Lagos-Quintana M, Rauhut R, Lendeckel W, Tuschl T. Identification of novel genes miR-137 expression is unknown, but based on our research, coding for small expressed RNAs. Science 2001; 294:853–858. MIR137 biogenesis is affected. Variants may affect precision or 7 Lai EC. Micro RNAs are complementary to 3' UTR sequence motifs that mediate effectiveness of Drosha cleavage, but knowing whether the negative post-transcriptional regulation. Nat Genet 2002; 30:363–364. variants favor the genesis of isomiRs-137, opposed to canonical 8 Place RF, Li LC, Pookot D, Noonan EJ, Dahiya R. MicroRNA-373 induces expression mir-137, can only be established after deep miRNA sequencing. of genes with complementary promoter sequences. Proc Natl Acad Sci USA 2008; The extent to which isomiRs influence the regulation is unknown 105: 1608–1613. and a matter of debate. It seems that isomiRs mainly cooperate 9 Vasudevan S, Tong Y, Steitz JA. Switching from repression to activation: micro- 318 – with canonical miRNAs and target functionally related genes.30 RNAs can up-regulate translation. Science 2007; : 1931 1934. 10 He X, Zhang Q, Liu Y, Pan X. Cloning and identification of novel microRNAs from Hypothetically, − 4C>T and VNTR variants may influence 3′ isomiR 39 – fl rat hippocampus. Acta Biochim Biophys Sin (Shanghai) 2007; :708 714. formation, which is thought not to in uence the seed or the 11 Bak M, Silahtaroglu A, Moller M, Christensen M, Rath MF, Skryabin B et al. targeting. However, if the variants alter Drosha processing and MicroRNA expression in the adult mouse central nervous system. RNA 2008; 14: cause a change in the preferred strand selection, the canonical 432–444. miR-137 levels may decline, as seen in our study. The 5′ isomiRs 12 Liang Y, Ridzon D, Wong L, Chen C. Characterization of microRNA expression are rare, but they may result in seed and targeting alterations and profiles in normal human tissues. BMC Genomics 2007; 8: 166. are supposed to have greater functional impact,28,29,81 as shown 13 Landgraf P, Rusu M, Sheridan R, Sewer A, Iovino N, Aravin A et al. A mammalian for 5′ isomiR-137-1.29 The proposed effect on the targeting, when microRNA expression atlas based on small RNA library sequencing. Cell 2007; 129: – comparing 5′ isomiR-137 putative targets to the canonical 1401 1414. 29 14 Beveridge NJ, Tooney PA, Carroll AP, Tran N, Cairns MJ. Down-regulation of miR-137 targets, is much greater than the observed or subtle, miR-17 family expression in response to retinoic acid induced neuronal differ- but important effect on the global transcriptome, seen in our entiation. Cell Signal 2009; 21: 1837–1845. study. Therefore we believe that the variant induced 5′ isomiR-137 15 Beveridge NJ, Gardiner E, Carroll AP, Tooney PA, Cairns MJ. Schizophrenia is formation is unlikely. associated with an increase in cortical microRNA biogenesis. Mol Psychiatry 2010; Regardless of the mechanism by which VNTR and − 4C>T affect 15: 1176–1189. the mir-137 biogenesis, both variants, found more frequently in SZ 16 Xu B, Hsu PK, Karayiorgou M, Gogos JA. MicroRNA dysregulation in neuro- and BD patients, cause a decline in miR-137 levels and influence psychiatric disorders and cognitive dysfunction. Neurobiol Dis 2012; 46: the regulation of genes involved in synapse-related functions, a 291–301. mechanism known to be involved in pathologies of psychiatric 17 Beveridge NJ, Cairns MJ. MicroRNA dysregulation in schizophrenia. Neurobiol Dis 46 – disorders. 2012; :263 271. 18 Lewis CM, Levinson DF, Wise LH, DeLisi LE, Straub RE, Hovatta I et al. Genome scan meta-analysis of schizophrenia and bipolar disorder, part II: Schizophrenia. Am J CONFLICT OF INTEREST Hum Genet 2003; 73:34–48. 19 Segurado R, Detera-Wadleigh SD, Levinson DF, Lewis CM, Gill M, Nurnberger JI Jr The authors declare no conflict of interest. et al. Genome scan meta-analysis of schizophrenia and bipolar disorder, part III: Bipolar disorder. Am J Hum Genet 2003; 73:49–62. 20 Ripke S, Sanders AR, Kendler KS, Levinson DF, Sklar P, Holmans PA et al. Genome- ACKNOWLEDGMENTS wide association study identifies five new schizophrenia loci. Nat Genet 2011; 43: We are grateful to the patients and control individuals for their cooperation and 969–976. participation in this research study. We acknowledge the contribution of the 21 Ripke S, O'Dushlaine C, Chambert K, Moran JL, Kahler AK, Akterin S et al. Genome- personnel of the VIB Genetic Service Facility (www.vibgeneticservicefacility.be) for wide association analysis identifies 13 new risk loci for schizophrenia. Nat Genet the genetic analyses, Sofie Goethals and Vicky De Winter for their assistance and 2013; 45:1150–1159. expertise in generation of stable cell lines, Bart Aelterman and Jenne Dierckx for their 22 Iwai N, Naraba H. Polymorphisms in human pre-miRNAs. Biochem Biophys Res help with the prediction analyses and Professor Dr Kristel Sleegers for the assistance Commun 2005; 331: 1439–1444. with statistical analysis. We also thank Professor Dr Michel Georges and Dr Haruko 23 Sun G, Yan J, Noltner K, Feng J, Li H, Sarkis DA et al. SNPs in human miRNA genes Takeda from the University of Liege, Faculty of Veterinary Sciences, for their helpful affect biogenesis and function. RNA 2009; 15: 1640–1651. discussion. This research was funded by grants from the Fund for Scientific Research 24 Ma H, Wu Y, Choi JG, Wu H. Lower and upper stem-single-stranded RNA junctions Flanders (FWO-F), the Institute for the Promotion of Innovation through Science and together determine the Drosha cleavage site. Proc Natl Acad Sci USA 2013; 110: Technology in Flanders (IWT-F), VIB International PhD Program and Stanley Medical 20687–20692. Research Institute. The work was also supported by the Swedish Research Council 25 Morin RD, O'Connor MD, Griffith M, Kuchenbauer F, Delaney A, Prabhu AL et al. (grant numbers 2006-4472, 2009-5269, 2009-3413) and the County Councils of Application of massively parallel sequencing to microRNA profiling and discovery Västerbotten and Norrbotten, Sweden. The Betula Study, from which the controls in human embryonic stem cells. Genome Res 2008; 18:610–621.

© 2015 Macmillan Publishers Limited Molecular Psychiatry (2015), 472 – 481 Reduced miR-137 deregulates synaptic gene sets M Strazisar et al 480 26 Marti E, Pantano L, Banez-Coronel M, Llorens F, Minones-Moyano E, Porta S et al. 53 Cummings E, Donohoe G, Hargreaves A, Moore S, Fahey C, Dinan TG et al. A myriad of miRNA variants in control and Huntington's disease brain regions Mood congruent psychotic symptoms and specific cognitive deficits in carriers detected by massively parallel sequencing. Nucleic Acids Res 2010; 38: 7219–7235. of the novel schizophrenia risk variant at MIR-137. Neurosci Lett 2013; 532: 27 Zhou H, Arcila ML, Li Z, Lee EJ, Henzler C, Liu J et al. Deep annotation of mouse 33–38. iso-miR and iso-moR variation. Nucleic Acids Res 2012; 40: 5864–5875. 54 Kwon E, Wang W, Tsai LH. Validation of schizophrenia-associated genes CSMD1, 28 Neilsen CT, Goodall GJ, Bracken CP. IsomiRs--the overlooked repertoire in the C10orf26, CACNA1C and TCF4 as miR-137 targets. Mol Psychiatry 2011; 18:11–12. dynamic microRNAome. Trends Genet 2012; 28:544–549. 55 Kim AH, Parker EK, Williamson V, McMichael GO, Fanous AH, Vladimirov VI. 29 Kozlowska E, Krzyzosiak WJ, Koscianska E. Regulation of huntingtin gene Experimental validation of candidate schizophrenia gene ZNF804A as target for expression by miRNA-137, -214, -148a, and their respective isomiRs. Int J Mol Sci hsa-miR-137. Schizophr Res 2012; 141:60–64. 2013; 14: 16999–17016. 56 Egawa J, Nunokawa A, Shibuya M, Watanabe Y, Kaneko N, Igeta H et al. Rese- 30 Cloonan N, Wani S, Xu Q, Gu J, Lea K, Heater S et al. MicroRNAs and their isomiRs quencing and association analysis of MIR137 with schizophrenia in a Japanese function cooperatively to target common biological pathways. Genome Biol 2011; population. Psychiatry Clin Neurosci 2013; 67: 277–279. 12: R126. 57 Bemis LT, Chen R, Amato CM, Classen EH, Robinson SE, Coffey DG et al. 31 Rozen S, Skaletsky H. Primer3 on the WWW for general users and for biologist MicroRNA-137 targets microphthalmia-associated transcription factor in mela- programmers. Methods Mol Biol 2000; 132:365–386. noma cell lines. Cancer Res 2008; 68: 1362–1368. 32 Sham PC, Curtis D. Monte Carlo tests for associations between disease and alleles 58 Althoff K, Beckers A, Odersky A, Mestdagh P, Koster J, Bray IM et al. MiR-137 at highly polymorphic loci. Ann Hum Genet 1995; 59:97–105. functions as a tumor suppressor in neuroblastoma by downregulating KDM1A. Int 33 Lorenz R, Bernhart SH, Honer Zu Siederdissen C, Tafer H, Flamm C, Stadler PF et al. J Cancer 2013; 133: 1064–1073. ViennaRNA Package 2.0. Algorithms Mol Biol 2011; 6: 26. 59 Silber J, Lim DA, Petritsch C, Persson AI, Maunakea AK, Yu M et al. miR-124 and 34 Darty K, Denise A, Ponty Y. VARNA: Interactive drawing and editing of the RNA miR-137 inhibit proliferation of glioblastoma multiforme cells and induce differ- secondary structure. Bioinformatics 2009; 25: 1974–1975. entiation of brain tumor stem cells. BMC Med 2008; 6: 14. 35 Almeida-Souza L, Goethals S, de Winter V, Dierick I, Gallardo R, Van Durme J et al. 60 Geekiyanage H, Chan C. MicroRNA-137/181c regulates serine palmitoyltransferase Increased monomerization of mutant HSPB1 leads to protein hyperactivity in and in turn amyloid beta, novel targets in sporadic Alzheimer's disease. J Neurosci Charcot-Marie-Tooth neuropathy. J Biol Chem 2010; 285: 12778–12786. 2011; 31: 14820–14830. 36 Vandesompele J, De Preter K, Pattyn F, Poppe B, Van Roy N, De Paepe A et al. 61 Geekiyanage H, Jicha GA, Nelson PT, Chan C. Blood serum miRNA: Non-invasive Accurate normalization of real-time quantitative RT-PCR data by geometric biomarkers for Alzheimer's disease. Exp Neurol 2011; 235: 491–496. averaging of multiple internal control genes. Genome Biol 2002; 3:RESEARCH0034. 62 Zhou L, Pupo GM, Gupta P, Liu B, Tran SL, Rahme R et al. A parallel genome-wide 37 Hellemans J, Mortier G, De Paepe A, Speleman F, Vandesompele J. qBase relative mRNA and microRNA profiling of the frontal cortex of HIV patients with and quantification framework and software for management and automated analysis without HIV-associated dementia shows the role of axon guidance and down- of real-time quantitative PCR data. Genome Biol 2007; 8: R19. stream pathways in HIV-mediated neurodegeneration. BMC Genomics 2012; 13: 38 Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays applied to the 677. ionizing radiation response. Proc Natl Acad Sci USA 2001; 98: 5116–5121. 63 Wright C, Turner J, Calhoun V, Perrone-Bizzozero N. Potential impact of miR-137 39 Saeed AI, Bhagabati NK, Braisted JC, Liang W, Sharov V, Howe EA et al. TM4 and its targets in schizophrenia. Front Genet 2013; 4:58. microarray software suite. Methods Enzymol 2006; 411:134–193. 64 Selbach M, Schwanhausser B, Thierfelder N, Fang Z, Khanin R, Rajewsky N. 40 Mootha VK, Lindgren CM, Eriksson KF, Subramanian A, Sihag S, Lehar J et al. Widespread changes in protein synthesis induced by microRNAs. Nature 2008; PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordi- 455:58–63. nately downregulated in human diabetes. Nat Genet 2003; 34: 267–273. 65 Lim LP, Lau NC, Garrett-Engele P, Grimson A, Schelter JM, Castle J et al. Microarray 41 Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA et al. analysis shows that some microRNAs downregulate large numbers of Gene set enrichment analysis: a knowledge-based approach for interpreting target mRNAs. Nature 2005; 433: 769–773. genome-wide expression profiles. Proc Natl Acad Sci USA 2005; 102: 15545–15550. 66 Baek D, Villen J, Shin C, Camargo FD, Gygi SP, Bartel DP. The impact of microRNAs 42 Green MJ, Cairns MJ, Wu J, Dragovic M, Jablensky A, Tooney PA et al. on protein output. Nature 2008; 455:64–71. Genome-wide supported variant MIR137 and severe negative symptoms predict 67 Lips ES, Cornelisse LN, Toonen RF, Min JL, Hultman CM, International Schizo- membership of an impaired cognitive subtype of schizophrenia. Mol Psychiatry phrenia Consortium et al. Functional gene group analysis identifies synaptic gene 2013; 18: 774–780. groups as risk factor for schizophrenia. Mol Psychiatry 2012; 17: 996–1006. 43 Lett T, Chakavarty M, Felsky D, Brandl E, Tiwari A, Gonçalves V et al. The genome- 68 Jia P, Sun J, Guo A, Zhao Z. SZGR: a comprehensive schizophrenia gene resource. wide supported microRNA-137 variant predicts phenotypic heterogeneity within Mol Psychiatry 2010; 15:453–462. schizophrenia. Mol Psychiatry 2013; 18:443–450. 69 Jensen L, Kuhn M, Stark M, Chaffron S, Creevey C, Muller J et al. STRING 8--a global 44 Whalley HC, Papmeyer M, Romaniuk L, Sprooten E, Johnstone EC, Hall J et al. view on proteins and their functional interactions in 630 organisms. Nucleic Acids Impact of a microRNA MIR137 susceptibility variant on brain function in people at Res 2009; 37:6. high genetic risk of schizophrenia or bipolar disorder. Neuropsychopharmacology 70 Takeichi M. The cadherin superfamily in neuronal connections and interactions. 2012; 37: 2720–2729. Nat Rev Neurosci 2007; 8:11–20. 45 Bergen SE, O'Dushlaine CT, Ripke S, Lee PH, Ruderfer DM, Akterin S et al. Genome- 71 Junghans D, Heidenreich M, Hack I, Taylor V, Frotscher M, Kemler R. Postsynaptic wide association study in a Swedish population yields support for greater CNV and differential localization to neuronal subtypes of protocadherin beta16 in the and MHC involvement in schizophrenia compared with bipolar disorder. Mol mammalian central nervous system. Eur J Neurosci 2008; 27:559–571. Psychiatry 2012; 17:880–886. 72 Sun J, Kuo PH, Riley BP, Kendler KS, Zhao Z. Candidate genes for schizophrenia: a 46 Gong J, Tong Y, Zhang HM, Wang K, Hu T, Shan G et al. Genome-wide identifi- survey of association studies and gene ranking. Am J Med Genet B Neuropsychiatr cation of SNPs in microRNA genes and the SNP effects on microRNA target Genet 2008; 147B: 1173–1181. binding and biogenesis. Hum Mutat 2012; 33: 254–263. 73 Shao L, Vawter MP. Shared gene expression alterations in schizophrenia and 47 Szulwach KE, Li X, Smrt RD, Li Y, Luo Y, Lin L et al. Cross talk between bipolar disorder. Biol Psychiatry 2008; 64:89–97. microRNA and epigenetic regulation in adult neurogenesis. J Cell Biol 2010; 189: 74 Myles-Worsley M, Tiobech J, Browning SR, Korn J, Goodman S, Gentile K et al. 127–141. Deletion at the SLC1A1 glutamate transporter gene co-segregates with schizo- 48 Silber J, Lim D, Petritsch C, Persson A, Maunakea A, Yu M et al. miR-124 and phrenia and bipolar schizoaffective disorder in a 5-generation family. Am J Med miR-137 inhibit proliferation of glioblastoma multiforme cells and induce differ- Genet B Neuropsychiatr Genet 2013; 162B:87–95. entiation of brain tumor stem cells. BMC Med 2008; 6:14. 75 McCullumsmith RE, Meador-Woodruff JH. Striatal excitatory amino acid trans- 49 Smrt RD, Szulwach KE, Pfeiffer RL, Li X, Guo W, Pathania M et al. MicroRNA porter transcript expression in schizophrenia, bipolar disorder, and major miR-137 regulates neuronal maturation by targeting ubiquitin ligase mind bomb- depressive disorder. Neuropsychopharmacology 2002; 26:368–375. 1. Stem Cells 2010; 28: 1060–1070. 76 Lang UE, Puls I, Muller DJ, Strutz-Seebohm N, Gallinat J. Molecular mechanisms of 50 Green M, Cairns M, Wu J, Dragovic M, Jablensky A, Tooney P et al. Genome-wide schizophrenia. Cell Physiol Biochem 2007; 20: 687–702. supported variant MIR137 and severe negative symptoms predict membership of 77 Murphy GG, Fedorov NB, Giese KP, Ohno M, Friedman E, Chen R et al. Increased an impaired cognitive subtype of schizophrenia. Mol Psychiatry 2012; 18:774–780. neuronal excitability, synaptic plasticity, and learning in aged Kvbeta1.1 51 Yin J, Lin J, Luo X, Chen Y, Li Z, Ma G et al. miR-137: A New Player in Schizophrenia. knockout mice. Curr Biol 2004; 14: 1907–1915. Int J Mol Sci 2014; 15: 3262–3271. 78 Knable MB, Barci BM, Webster MJ, Meador-Woodruff J, Torrey EF, Stanley Neu- 52 Guella I, Sequeira A, Rollins B, Morgan L, Torri F, van Erp TG et al. Analysis of ropathology Consortium. Molecular abnormalities of the hippocampus in severe miR-137 expression and rs1625579 in dorsolateral prefrontal cortex. J Psychiatr psychiatric illness: postmortem findings from the Stanley Neuropathology Con- Res 2013; 47: 1215–1221. sortium. Mol Psychiatry 2004; 9:609–620, 544.

Molecular Psychiatry (2015), 472 – 481 © 2015 Macmillan Publishers Limited Reduced miR-137 deregulates synaptic gene sets M Strazisar et al 481 79 Drew CJ, Kyd RJ, Morton AJ. Complexin 1 knockout mice exhibit marked deﬁcits in receptors cause variable neurodevelopmental phenotypes. Nat Genet 2010; 42: social behaviours but appear to be cognitively normal. Hum Mol Genet 2007; 16: 1021–1026. 2288–2305. 81 Xia J, Zhang W. A meta-analysis revealed insights into the sources, conservation 80 Endele S, Rosenberger G, Geider K, Popp B, Tamer C, Stefanova I et al. and impact of microRNA 5'-isoforms in four model species. Nucleic Acids Res 2014; Mutations in GRIN2A and GRIN2B encoding regulatory subunits of NMDA 42: 1427–1441.

Supplementary Information accompanies the paper on the Molecular Psychiatry website (http://www.nature.com/mp)