MicroRNA-9 and MicroRNA-326 Regulate Human Dopamine D2 Receptor Expression, and the MicroRNA-mediated Expression Regulation Is Altered by a Genetic Variant*

Background: Regulation of dopamine D2 receptor (DRD2) is pathophysiologically and pharmacologically important. Results: miR-9 and miR-326 target to the 3′-UTR of DRD2, and endogenously inhibit DRD2 expression. A functional single nucleotide polymorphism alters such regulation. Conclusion: DRD2 is post-transcriptionally regulated by miR-326 and miR-9. Significance: The study suggests a pathophysiological and pharmacological role of miR-9 and miR-326 in neuropsychiatric disorders. The human dopamine receptor D2 (DRD2) has been implicated in the pathophysiology of schizophrenia and other neuropsychiatric disorders. Most antipsychotic drugs influence dopaminergic transmission through blocking dopamine receptors, primarily DRD2. We report here the post-transcriptional regulation of DRD2 expression by two brain-expressed microRNAs (miRs), miR-326 and miR-9, in an ex vivo mode, and show the relevance of miR-mediated DRD2 expression regulation in human dopaminergic neurons and in developing human brains. Both miRs targeted the 3′-UTR (untranslated region) of DRD2 in NT2 (neuron-committed teratocarcinoma, which endogenously expresses DRD2) and CHO (Chinese hamster ovary) cell lines, decreasing luciferase activity measured by a luciferase reporter gene assay. miR-326 overexpression reduced DRD2 mRNA and DRD2 receptor synthesis. Both antisense miR-326 and antisense miR-9 increased DRD2 protein abundance, suggesting an endogenous repression of DRD2 expression by both miRs. Furthermore, a genetic variant (rs1130354) within the DRD2 3′-UTR miR-targeting site interferes with miR-326-mediated repression of DRD2 expression. Finally, co-expression analysis identified an inverse correlation of DRD2 expression with both miR-326 and miR-9 in differentiating dopaminergic neurons derived from human induced pluripotent stem cells (iPSCs) and in developing human brain regions implicated in schizophrenia. Our study provides empirical evidence suggesting that miR-326 and miR-9 may regulate dopaminergic signaling, and miR-326 and miR-9 may be considered as potential drug targets for the treatment of disorders involving abnormal DRD2 function, such as schizophrenia.

The dopaminergic hypothesis of SZ postulates that dopaminergic overactivity causes psychosis. Underlying this hypothesis are the observations that psychotogenic stimulants (e.g. amphetamines) activate dopamine receptors (1) and that most antipsychotic drugs influence dopaminergic transmission through blocking dopamine (primarily D2) receptors (2)(3)(4)(5)(6)(7)(8)(9). Besides the essential role of DRD2 affinity for antipsychotic effects (10), further support derives from the study of postmortem brains, live brain imaging, and animal models. Postmortem studies of SZ cases consistently found increased density of striatal D2 receptors (8, 11-15). Brain imaging studies have shown that amphetamine-induced increase of dopamine response is correlated with positive symptoms of SZ (16 -18) and elevation of striatal D2 receptors (8,19). Mice overexpressing striatal DRD2 exhibit selective cognitive impairments in working memory tasks (20, 21). Three DRD2 missense variants (Val 96 3 Ala, Pro 310 3 Ser, and Ser 311 3 Cys) impair receptor function (22, 23), and the minor allele of rs1801028 (Ser 311 3 Cys) is associated with SZ 2 in a meta-analysis (24). However, these missense changes are uncommon, and their function does not seem congruent with a hyperdopaminergic hypothesis of SZ. Therefore, it is conceivable that for DRD2, increased expression (rather than structural changes) may play an important role in SZ pathogenesis. DRD2 is also hypothesized to be involved in multiple other neuropsychiatric disorders, e.g. drug addiction (25), neuroticism, and anxiety disorders (26,27), which highlights additional potential importance of studying its regulation.
Multiple DRD2 genetic variants influence DRD2 expression at both the transcriptional and the post-transcriptional levels. At the transcriptional level, the insertion allele of the promoter variant rs1799732 (Ϫ141C Ins/Del) increases in vitro DRD2 transcription and is reported to be nominally associated with SZ in a small Japanese sample (260 cases versus 312 controls) (28). Post-transcriptionally, two intronic SNPs, rs2283265 (intron 5) and rs1076560 (intron 6), in high linkage disequilibrium (LD) with each other, alter DRD2 splicing (29) and are associated with reduced performance in working memory in controls (29) and with cocaine abuse (30,31). We and others have also previously shown that allele C of a functional synonymous variant rs6277 (C957T) increases DRD2 mRNA stability and receptor synthesis in vitro (32) and in vivo (33). Furthermore, rs6277 has been reported to be associated with cognitive performance (34 -36), response to antipsychotic medication (37), and depressive rumination (38). Identification of these functional DRD2 variants highlights the potential importance of noncoding regulatory sequences. miRs have emerged as important post-transcriptional regulators of gene expression. miRs are ϳ22-nucleotide small noncoding RNAs processed from pre-miRs (ϳ100 bp) with a typical folding structure containing a mature miR region and a stem-loop region (39,40). miRs mediate translational repression and/or mRNA decay of a target gene through sequence complementation with the 3Ј-UTRs of their target genes (41,42). There is evidence suggesting that the number of miR target sites at the 3Ј-UTRs of a gene is negatively correlated with RNA stability (43). The 3Ј-UTR of human DRD2 mRNA is predicted (PicTar, TargetScan) to have multiple miR binding sites (Fig. 1,  A and B), including those for miR-326 and miR-9, which are brain-expressed and involved in neurogenesis and neurodifferentiation (44 -46). There are two known SNPs, rs1130354 and rs6274, within a predicted (PicTar) targeting region shared by miR-326 and miR-9 (Fig. 1, B and C). Polymorphisms within the target region may impair the base-pairing between an miR and its target, hence affecting target gene expression. In this study we used a cellular model to examine the effects of miR-326 and of miR-9 on DRD2 receptor expression and whether genetic variation within the 3Ј-UTR of DRD2 mRNA alters the targeting of miRs.

EXPERIMENTAL PROCEDURES
Cell Culture-We purchased CHO-K1 (hereafter referred to as CHO) and human NT2 cell lines from ATCC. We cultured CHO in F-12K and NT2 in Dulbecco's modified Eagle's medium (DMEM; from ATCC) supplemented with 10% fetal bovine serum (FBS) and penicillin/streptomycin 100 units/ml (from ATCC). We maintained cells in 15 ml of growth medium in a 75T flask with a cell density of Ͻ1 ϫ 10 6 /ml in a 37°C incubator with 5% CO 2 , changing cell culture media every other day until the experiment day. We used both NT2 and CHO cell lines for reporter gene assay but used only the NT2 cell line for studying miR-mediated inhibition on endogenous DRD2 expression, because CHO cells do not endogenously express DRD2 (23, 32,47).
Samples and Genotyping-We studied 1870 SZ cases and 2002 controls from the Molecular Genetics of Schizophrenia (MGS) sample of European ancestry (48,49) to evaluate the LD between functional SNPs and for estimating allele frequency of rs1130354. We previously described in detail the recruitment procedures and clinical phenotypes (48,49). We used the Taq-Man Pre-Designed SNP Genotyping Assay (Invitrogen) to genotype rs1130354 in the MGS European ancestry sample. Briefly, we used 4 ng of genomic DNAs in a total of 5 l of a polymerase chain reaction (PCR) reaction in each well on a 384-well plate. We carried out PCR on a 7900HT instrument (Invitrogen) with standard real-time PCR cycling conditions. We made automatic genotype calls with SDS 2.2 software (Invitrogen). The genotyping completion rate was 98 and 96% for cases and controls, respectively. There was no significant deviation from Hardy-Weinberg disequilibrium. The NorthShore University HealthSystem Institutional Review Board approved the study.
DNA Cloning and Sequence Mutagenesis-We used a PCR to amplify from MGS DNAs the DRD2 3Ј-UTR sequences (830 bp; chr11:113,280,381-113,281,210; hg19) spanning the predicted target sites of miR-9 and miR-326. The DNA sequences of the amplified fragments were confirmed by Sanger sequencing. We selected the amplified wild-type (WT) sequence as those with the same sequence as the hg19 reference sequence. We then cloned the wt sequence fragment into Xho1 and Pme1 sites downstream of the firefly luciferase coding region in the psi-CHECK™-2 Vector (Promega). psiCHECK™-2 utilizes firefly luciferase as a reporter gene and Renilla luciferase in the same vector as a control to normalize transfection efficiency and cell number in the reporter gene assay. With the recombined reporter gene vector containing DRD2 3Ј-UTR wt sequence as a backbone vector, we used the QuikChange II Site-Directed Mutagenesis kit (Agilent) to change the nucleotide sequence to construct the mutant alleles of the two known SNPs (rs1130354 and rs6274) within the targeting sites of miR-9 and miR-326. We designed PCR primers (sequences available upon request) for cloning DRD2 3Ј-UTR and for mutagenesis with Primer3 and purchased primer oligos from IDT (Integrated DNA Technologies).
Reporter Gene Assay-We transformed reporter gene DNA constructs containing the DRD2 3Ј-UTR wt sequence (pD2UTR-wt) with rs6274 mutant allele G (pD2UTR-rs6274-G) and with rs1130354 mutant allele G (pD2UTR-rs1130354-G) (Fig. 2) into Escherichia coli DH5␣ and prepared endotoxin-free plasmid DNAs with the StrataPrep EF Plasmid Midiprep kit (Agilent). We next transfected reporter gene plasmid DNAs into CHO or NT2 cells. Briefly, we plated 3.5 ϫ 10 4 cells per well into 24-well plates 24 h before transfection. The next day we transfected ϳ100 ng of reporter gene vector and varying amounts of miR precursors into CHO and NT2 cells at 90% confluence, respectively. We used Lipofectamine 2000 (Invitrogen) for transfection. At 48 h post-transfection, we measured Renilla and firefly luciferase activity in transfected cells using a Dual-Luciferase Reporter Assay system (Promega) in a 96-well plate.
Overexpression of miRs and Anti-miRs in NT2 Cells-We plated NT2 cells at 2.5 ϫ 10 5 cells per well on a 6-well plate 24 h before transfection. At 90% of cell confluency we transfected NT2 cells with Ambion Pre-miR™ miR precursors miR-9 and miR-326 (Invitrogen) at final concentrations of 20, 50, and 100 nM. We also transfected Ambion Anti-miR™ miRNA Inhibitors (Invitrogen) for miR-9 and miR-326. We used DMEM supplemented with 10% FBS without antibiotics to culture the transfected cells for 48 h. We then collected cells for RNA and protein extraction using the mirVana PARIS kit (Invitrogen).
Human iPSCs and Dopaminergic (DA) Neuron Differentiation-We generated three iPSC lines from fibroblasts cells of two individuals (GM01835 and GM03652) (from Coriell) and from skin biopsy of a healthy adult volunteer. The NorthShore University HealthSystem Institutional Review Board approved the study. We used a polycistron lentivirus vector (STEMCCA) containing four standard reprogramming factors (OCT4, SOX2, c-MYC, and KLF4) to infect the fibroblasts for generating iPSCs as previously described (50,51). The iPSC clones were identified by stem cell-like morphology and positive immune fluorescence staining for pluripotent stem cell markers (TRA-1-60, OCT4, NANOG, and SSEA4). The iPSC clones were further evaluated for pluripotency by testing the formation of embryoid bodies that spontaneously differentiate into three germ layers (52). The three germ layers were characterized by immune fluorescence staining of ␣-fetoprotein (endoderm), tubulin ␤-III (ectoderm), and ␣-smooth muscle actin FIGURE 2. Luciferase reporter constructs and reporter gene assay. A, the reporter gene vector ( psiCHECK) containing the 3Ј-UTR of DRD2, showing one putative target site shared by miR-9/miR-326. Gray-shaded letters in each construct represent conserved sequences, and yellow letters indicate the minor alleles of rs6274 and rs1130354. Letters in blue represent the miR nucleotides complementary to the target sequence. miR-9 and miR-326 inhibited the reporter gene (luciferase) expression through targeting the downstream DRD2 3Ј-UTR in transfected CHO cells (B) and NT2 cells (C). Concentration of both miRs had significant effect on reporter gene expression in both CHO cells (ANOVA p Ͻ 0.0001) and in NT2 cells (ANOVA p Ͻ 0.0001 for miR-9 and Ͻ0.0002 for miR-326). We co-transfected cells with pD2UTR-wt and synthetic pre-miRs. We measured luciferase activity at 48 h post-transfection and present data as the average Ϯ S.D. from at least 3 independent experiments, each with 3 measurements for each condition. *** indicates p Ͻ 0.0001 from Student's t test.
(Mesoderm). Human iPSCs were cultured on mouse embryonic fibroblasts (Global Stem) in standard human embryo stem cell media (53). For DA neuron differentiation, we adopted a floor-plate-based midbrain DA differentiation method as previously described (53). In brief, we grew ϳ2.2 million iPSCs on a 10-cm dish pre-coated with Matrigel (BD Biosciences) for 11 days for floor-plate induction by exposing cells to a combination of LDN193189, SB431542, SHH C25II, Purmorphamine, FGF8 and CHIR99021. On day 11 we changed the media to Neurobasal/B27/L-Glut containing medium (NB/B27) (Invitrogen). On day 13 we changed the media to differentiation media (NB/B271, BDNF, ascorbic acid, glial cell-derived neurotrophic factor, dibutyryl-cAMP, TGFb3 and DAPT). On day 20, we dissociated and replated the cells at high density (300 -400 ϫ 10 3 cells/cm 2 ) on 6-well plates and coverslips precoated with poly-D-lysine/laminin in differentiation medium. We kept the neurons in differentiation until the day for a certain experiment. For single neuron gene expression analysis to determine the percentage of DA neurons (THϩ neurons), we manually picked single neurons by serial dilution into wells on 96-well plate and directly lysed cells and carried out reverse transcription using CellsDirect™ One-Step qRT-PCR kit (Invitrogen). Quantitative PCR was performed on Fluidigm BioMark system as described below. We assayed expression of TH, MAP2, and GAPDH, and only samples with all three replicates of a gene assay passing the Ct threshold were considered "expressed." Real-time Quantitative PCR-We reverse-transcribed total RNAs extracted from transfected cells or iPSC/neurons using TaqMan Reverse Transcription Reagents (Invitrogen). We used 350 ng of RNA per 20-l reverse transcription reaction with both random hexamers and oligo(dT) 16 to prime the reaction. To quantify gene expression, we used 2 l of the reversetranscribed cDNA in the subsequent real-time PCR using Taq-Man gene expression assays (Invitrogen) in combination of TaqMan universal PCR Mastermix (Invitrogen). For miR quantification, we used the TaqMan MicroRNA Reverse Transcription kit in the reverse transcription reaction and the Taq-Man MicroRNA Assay (for miR-9, miR-326 and RNU48) in real-time PCR. For expression quantification in NT2 cells, we carried out PCR in a 20-l reaction with 4 replicates on a 384well plate on a 7900HT (Invitrogen) with standard real-time PCR cycling conditions. For expression quantification in iPSCs and DA neurons, we carried out multiplex quantitative PCR on BioMark microfluid Dynamic Gene Expression Array (Fluidigm). We performed pre-amplifications of the reverse-transcribed cDNA products (also known as Specific Target Amplification) separately for miRNA and mRNAs (DRD2 and other genes) following the protocol recommended by Fluidigm. The preamplified products were then diluted (1:4 for miRNA and 1:2.5 for mRNA) and pooled together with equal amounts followed by multiplex PCR on a 48.48 Dynamic Gene Expression Array (Fluidigm) with three technical replicates. We used the standard ⌬-Ct method (Invitrogen) for relative expression quantification. Expression of GAPDH was used as an endogenous control to normalize the mRNA abundances, and the expression of RNU48 was used as an endogenous control to normalize the expression of miRs.
Western Blotting-To quantify the DRD2 expression at the protein level, we plated NT2 cells at 2.5 ϫ 10 5 cells per well on a 6-well plate 24 h before transfection with different miRs or anti-miRs as described above. We lysed cell cultures after 48 h and extracted proteins using the mirVana PARIS kit from each cell culture. We prepared cell lysate with SDS and then ran the cell lysate on a 4 -20% precast linear gradient polyacrylamide Bio-Rad Ready Gel (Bio-Rad). We transferred protein bands onto a nitrocellulose membrane via the semidry transfer method. We then blocked the membrane with 5% nonfat milk (Bio-Rad) in 5% PBS-Tween for 1 h and then incubated sequentially with mouse monoclonal antibody against human DRD2 receptor (sc-5303) and with goat anti-mouse HRP (horseradish peroxidase) (sc-2005) with washing between incubations. We probed the expression of the endogenous control ␤-actin by HRP-conjugated goat polyclonal IgG antibody against actin (C-11) (sc-1615-HRP). We purchased all antibodies from Santa Cruz Biotechnology and used them at the manufacturer's recommended dilution. We exposed the membrane with Super-Signal West Femto Chemiluminescent Substrate (Thermo Scientific). The amounts of DRD2 and ␤-actin on gel images from Western blots were quantified by ImageJ 1.47v (54).
Transcriptome Data of Developing Human Brain Tissues-We obtained the miRNA sequencing data and RNA sequencing (RNA-seq) data of developing brain tissues from the Allen Institute for Brain Science-BrainSpan Atlas of the Developing Human Brain. The full data set includes 1,620 miRNAs of 215 brain samples and 52,376 genes (Gencode v10) of 524 brain samples. These samples are from various postmortem brain regions of developmental stages of different individuals. We extracted 185 samples (for 16 brain regions) with expression data available for both the DRD2 and the two miRNAs, miR-326 and miR-9 (supplemental Table 1). Details of tissue acquisition, RNA-seq and miRNA-seq data processing, and data normalization can be found in the technical white paper BrainSpan Atlas of the Developing Human Brain.
Statistics and Bioinformatics-We used Student's t test to assess the significance of differential expression of reporter gene and endogenously expressed DRD2 between different experimental conditions. Two-tailed p values were reported. In addition, for each miR with Ͼ3 experimental groups, we also performed ANOVA to test for differences across different groups. We extracted genomic sequences from the UCSC genome browser and the miR target sites on the 3Ј-UTRs of DRD2 from the UCSC genome browser tracks annotated by TargetScan (55) and by PicTar (56). We predicted the RNA folding structures between miRs and targeting sequences (i.e. miR/target duplexes) and the minimum folding energy (⌬G) using RNAhybrid (57). We calculated LD between different SNPs of DRD2 gene with Haploview (58).

RESULTS
Both miR-9 and miR-326 Interact with the 3Ј-UTR of DRD2 and Repress Reporter Gene Expression-We used the Dual-Luciferase Reporter Assay System (psiCHECK™-2) to determine whether miR-9 and miR-326 interact with the 3Ј-UTR of DRD2 mRNA, thereby inhibiting the reporter gene (luciferase) expression. In the recombinant reporter gene construct, DRD2 3Ј-UTR sequence spanning the predicted binding sites of miR-9 and miR-326 was positioned downstream of the firefly luciferase open reading frame in the psiCHECK™-2 vector ( Fig. 2A). Transfected pre-miR-9 or pre-miR-326 exhibited a dose-responsive negative effect on reporter gene expression in both NT2 and CHO cell lines. Pre-miR-9 at 1 nM induced ϳ40% reduction of luciferase activity, and pre-miR-326 at 50 nM induced ϳ50% reduction of luciferase activity (Fig. 2, B and C).
miR-9 and miR-326 Inhibit Endogenously Expressed DRD2 in NT2 Cells-A reporter gene assay proved the direct targeting of the two miRs to the 3Ј-UTRs of DRD2. We next examined whether the endogenously expressed DRD2 is subject to the regulation by the two miRs in NT2, a cell line that endogenously expresses miR-9, miR-326, and DRD2 (data not shown) and has been used as a model of dopaminergic neurons for studying dopamine signaling (59 -62). For each miR, we tested whether overexpression of exogenous pre-miRs can inhibit endogenous expression of DRD2 and whether knocking down the endogenous miRs can increase expression of DRD2. Consistent with the reporter gene assay results, overexpression of pre-miR-326 in NT2 cells showed an inhibitory effect on endogenously expressed DRD2 mRNA, with ϳ40% reduction of the DRD2 mRNA level with 50 nM pre-miR-326 (Fig. 3A). At the protein level, overexpression of pre-miR-326 (50 nM) also led to reduced DRD2 receptor expression (Fig. 3B). As expected, transfection of antisense 2Ј-O-methyl-modified oligoribonucleotide (63, 64) for miR-326 (anti-miR-326; at 45 nM) in NT2 cells increased DRD2 protein abundances, presumably through blocking the inhibitory effect of endogenously expressed miR-326 on DRD2 expression (Fig. 3B).
Surprisingly, overexpression of pre-miR-9 in NT2 cells did not reduce endogenously expressed DRD2 either at the mRNA or the protein levels (Fig. 3, A and B). Intuitively, this seemed to contradict to the observed high efficacy of miR-9 in inhibiting reporter gene expression (Fig. 2). However, the high efficacy of exogenous miR-9 as measured in the reporter gene assay (Fig. 2) does not necessarily translate into effective inhibition of endogenously expressed DRD2 by exogenous miR-9 in Fig. 3. It is known that the inhibitory effect of an exogenous miR on its target gene expression may not be detectable in cells where the same miR is endogenously expressed at a high levels, whereas its target gene has relatively low expression (65). In our cellular model miR-9 is expressed at a high level (ϳ8-fold more than miR-326) and DRD2 at a much lower level (data not shown). The endogenous miR-9 is expressed in such a high level that it may have already saturated all the miR-9 targeting site of the lowly expressed DRD2. Consequently, the transfected exogenous miR-9, even at a high concentration (50 nM), did not show any additive effect on the endogenous expression of DRD2. To test for the endogenous miR-9-mediated repression of DRD2, we transfected the cells by antisense 2Ј-O-methyl-modified oligoribonucleotide (63, 64) of miR-9 (anti-miR-9). As predicted, overexpression of anti-miR-9 (45 nM) in NT2 cells led to a substantial increase of DRD2 protein abundances (Fig. 3B), suggesting a strong endogenous inhibitory effect of miR-9 on DRD2 expression in NT2 cells.
Genetic Variation in the Seed Sequence of the DRD2 3Ј-UTR Alters the Effect of miR-326-rs1130354 is within the 7-bp "seed sequence" shared by miR-9 and miR-326, whereas rs6274 is outside the seed sequence (Fig. 1); both SNPs thus could possibly affect miR interactions. For each SNP, we transfected a reporter gene construct containing the DRD2 3Ј-UTR sequence with one of the two alleles together with synthetic precursor miRs into CHO and NT2 cell lines (Fig. 2). As shown in Fig. 4, A  and B, the effect of exogenous miR-9 or miR-326 on reporter gene expression in cells carrying the minor allele G of rs6274 (pD2UTR-rs6274-G) was similar to that in cells transfected with the T allele (pD2UTR-wt), suggesting rs6274 did not disrupt the base pairing between either of the miRs and 3Ј-UTR of DRD2 mRNA. However, the minor allele G of rs1130354 (pD2UTR-rs1130354-G) abolished the inhibitory effect of miR-326. Furthermore, we observed a complete reversal of the miR-326-mediated repression of the reporter gene expression by allele G of rs1130354, indicating that the sequence flanking rs1130354 is likely the only miR-326 target site at the 3Ј-UTR of DRD2 mRNA, consistent with the bioinformatics prediction by TargetScan (Fig. 1B) but not by PicTar (Fig. 1A). The fact that rs1130354 only influences the effect of miR-326 but no miR-9 on DRD2 suggests miR-9 and miR-326 do not share the seed sequence for targeting DRD2 3Ј-UTR, i.e. both RNAs do not compete with each other at the SNP site.
The observed effect of rs1130354 and rs6274 was consistent with in silico analysis (57) of the SNP effect on RNA-folding structures formed by miR326 and its target sequence in DRD2 (Fig. 4C). With the major alleles of both SNPs, the base-pairing between DRD2 3Ј-UTR mRNA and the mature miR-326 results in a typical "stem-loop" structure essential for miR-mediated expression repression. The minor (or "mutant") allele G of rs1130354 disrupts the stem part of this essential structure, and the minimum folding energy (⌬G) of the structure increases from Ϫ25.5 kcal/mol (major allele) to Ϫ23.6 kcal/mol (i.e. a less stable secondary RNA folding structure). rs6274 does not show any allelic effect on the stem-loop folding structure or the minimum folding energy.
Next, we attempted to determine whether rs1130354 functions independently from other known regulatory variants of DRD2, namely, the promoter SNP rs1799732 (Ϫ141C Ins/Del) (28), the missense SNP rs1801028 (Ser 311 3 Cys) (23), the functional synonymous SNP rs6277 (C975T) (32), and the intronic splicing SNPs rs2283265 and rs1076560 (30). We genotyped rs1130354 in our MGS European ancestry sample in which genotype data were already available (48,49) for the other reported functional SNPs listed above. We found that rs1130354 is a rare SNP in our MGS sample (minor allele frequency ϭ 0.4%) and is not in LD with any other known functional variant (Fig. 5). This indicates that the functional allele of rs1130354 is not specifically in the same haplotype background with any other reported functional alleles in an individual and thus functions independently of them.
miRNA Regulation of DRD2 in Differentiating Dopaminergic Neurons and in Human Brains-In human neurons and brain tissues, we further evaluated the physiological relevance of the observed post-transcriptional regulation of DRD2 by miR-326 and miR-9. Given the predominate role of the classical DA hypothesis of SZ pathogenesis, we first examined the dynamic expression changes of DRD2 and the two miRs during DA neuron differentiation. We generated three iPSC lines from different individuals and then differentiated the iPSCs into DA neurons (Fig. 6, A--C). We have achieved high efficiency (80%) of dopaminergic neuron differentiation as previously reported (53). In a single cell gene expression analysis (quantitative PCR on Fluidigm BioMark system) of 36 manually picked single cells from the day 30 DA neurons, 35 are MAP2 positive, of which 28 are THϩ (80%). This allowed us to measure miRs and DRD2 in a relatively homogenous DA neuron population. We found a substantial increase of DRD2 expression in early stage of DA neurons (day 20) compared with iPSCs (Fig. 6D) (ϳ95-fold increase; p ϭ 4.7 ϫ 10 Ϫ5 ); thereafter, the DRD2 expression gradually decreased overtime with a 3.5-fold of decrease in day-45/50 neurons compared with day-20 neurons; p ϭ 0.047). Interestingly, both miR-9 and miR-326 showed an inverse correlation of expression changes with the DRD2 (Spearman's rank correlation r ϭ Ϫ0.4) during the DA neuronal differentiation, suggesting a contribution of post-transcriptional regulation of DRD2 by both miRs.

DISCUSSION
DRD2 plays a central role in dysregulation of dopamine signaling pertinent to SZ (5-9). Here, we show that miR-326 and miR-9 can affect DRD2 expression and that the targeting of miR-326 to DRD2 mRNA is disrupted by a rare functional SNP (rs1130354) within the seed sequence of DRD2 3Ј-UTR. We have demonstrated the miR-mediated post-transcriptional regulatory mechanism of DRD2 in a neuronal cell model (NT2) co-expressing miR-9, miR-326, and DRD2 mRNA/protein. We further showed the pathophysiological relevance of the miRmediated regulation of DRD2 in differentiating human DA neu-rons derived from iPSCs and in developing human brains. Given the central role of DRD2 in classic hyperdopaminergic hypothesis of SZ and its involvement in multiple other neuropsychiatric disorders (2-9), our study suggests a possible role of miR-9 and miR-326 in the pathophysiology of these disorders through modulating the DRD2 expression. Furthermore, as most antipsychotic drugs influence dopaminergic transmission through blocking DRD2 receptors, the demonstrated regulation of DRD2 by the two miRs may provide potential drug targets for treatment of disorders involving abnormal DRD2 function.
With reporter gene assays in two different cell lines, we confirmed the interactions between the 3Ј-UTR of DRD2 mRNA and exogenous miR-9 and miR-326. We also confirm the targeting of endogenous DRD2 mRNA by endogenously expressed miR-9 and miR-326 in NT2 cells. Our results contradict some bioinformatic predictions. For instance, as opposed to the Pic-Tar prediction that there are multiple targeting sites of miR-326 in the 3Ј-UTR of DRD2, this study suggests the existence of a single miR-326 targeting site, because allele G of the functional rs1130354 completely reverses the miR-326-mediated repression of reporter gene expression (Fig. 4, A and B). Furthermore, although both miR-326 and miR-9 are predicted (PicTar) to share the same seed sequence, the seed-disrupting allele G of rs1130354 is found to only affect targeting of miR-326 to DRD2 3Ј-UTR. Thus, our results highlight the limitations of bioinformatic prediction of miR targets and support the previously proposed (65, 66) need for empirical validation of putative regulatory effects of an miR on its important target genes.
To explore the pathophysiological relevance of the observed miR-mediated DRD2 expression, we examined the co-expression of DRD2 and both miRs in differentiating DA neurons and in developing human brains. Co-expression (miRNA and targeted mRNA) analysis has becoming a general paradigm to identify bona fide targets and infer biological roles of miRNAs (67). We identified inverse expression correlations between DRD2 and the two miRs (miR-9 and miR-326) in the course of DA neuron differentiation from iPSC (Fig. 6) and in developing human brains (Fig. 7), suggesting a naturally occurring miR-9and miR-326-mediated post-transcriptional regulation of DRD2 expression. Although miR-326 was expressed in a much lower level than miR-9 across different brain regions (ϳ676fold less in average; Fig. 7 and supplemental Table 1), it showed an overall stronger effect than miR-9 on DRD2 expression variations between different individuals. This was reflected by the observation of more brain regions showing negative expression correlation of DRD2 with miR-326 (n ϭ 10) than with miR-9 (n ϭ 6) and with a bigger magnitude of effect from miR-326 (Fig. 7). The weaker expression correlation of DRD2 and miR-9 could be explained by the unparalleled high level of miR-9 expression than DRD2 expression, which may lead to the saturated binding of miR-9 on DRD2 3Ј-UTR and consequently makes the DRD2 expression less sensitive to the expression changes of miR-9. The less pronouncing effect of miR-9 on DRD2 expression variation in the brain is consistent with our observation from the in vitro cellular model NT2 cells, where overexpression of exogenous miR-9 did not result in significant changes of DRD2 expression (Fig. 3, A and B). It is noteworthy that the strongest negative correlation of DRD2 expression was found with miR-326 (r ϭ Ϫ0.90, p ϭ 0.0003) in the dorsolateral prefrontal cortex, a brain region where disturbances of glutamate, GABA, and dopamine neurotransmissions contribute to the deficits of working memory in schizophrenia (68). These results suggest the pathophysiological relevance of the miR-326-mediated post-transcriptional regulation of DRD2.
The demonstration of DRD2 regulation by miR-9 and miR-326 may also have implications in understanding the biological relevance of the genetic associations recently identified in genome-wide association studies of SZ. It is noteworthy that the associations of DRD2 and both miRs with SZ are nominally significant (10 Ϫ6 Ͻ p Ͻ 10 Ϫ3 ) in the recent meta-analysis of genome-wide association studies of SZ (Psychiatric Genetics Consortium; PGC1) (69) and a combined analysis with Swedish sample (PGC2) (70). The nominally associated SNPs (and their LD proxies) in MIR9 -2 and MIR9 -3 (two of the three MIR9 genes) are within ENCODE-annotated enhancers (histone methylation mark H3K4Me1) immediately upstream the MIR9 genes and thus likely affect MIR9 expression and may subsequently alter DRD2 expression. Furthermore, DRD2 and miR-9 have indirect functional connections to other genome-wide association study-implicated SZ loci. For instance, one of the expression repressors of miR-9, nuclear receptor subfamily 2 group E member 1 (NR2E1; also known as TLX) (71), is a target of miR-137 (72), the latter being one of the most strongly SZassociated loci (69). Moreover, another genome-wide association study-implicated SZ susceptibility gene, ZNF804A, is a target of miR-137 (69,73,74) and also regulates transcription of DRD2 (75). We 3 and others (76) have shown that the SZ risk allele of MIR137 locus likely reduces MIR137 expression. The SZ-risk allele at MIR137 locus thus would increase DRD2 expression through an intracellular regulation cascade of MIR137-TLX-MIR9-DRD2, which is congruent with the classic hyperdopaminergic hypothesis of SZ. Our empirical validation of the post-transcriptional regulation of DRD2 by miR-326 and miR-9 thus suggests that miR-mediated dysregulation of a gene network involving DRD2 may play an important role in SZ pathogenesis.
SNPs in miR target regions are implicated in the pathogenesis of, or susceptibility to some human disorders, e.g. hypertension (77), asthma (78), cancer (79,80), and Tourette syndrome (81). The functional rs1130354 was not found associated with SZ in our MGS European ancestry sample (data not shown). However, rs1130354 (as well as the demonstrated effect of miRs on the expression of DRD2) might be important for other neuropsychiatric disorders and behavioral traits such as social detachment (82), neuroticism (27), and anxiety disorders (26), where dopamine and DRD2 play a central role in their pathophysiology. The demonstration of the functionality of rs1130354 in altering miRNA326 targeting of DRD2 thus warrants further an association test of rs1130354 with other neuropsychiatry disorders and behavioral phenotypes involving dysregulation of dopamine.
We acknowledge the limitation of our simplified cellular model for studying DRD2 regulation. Despite the demonstrated post-transcriptional regulation of DRD2 by both miR-9 and miR-326, transcriptional regulation likely remains a predominate mechanism in shaping the cellular RNA abundance (43). Furthermore, we only examined the post-transcriptional regulatory mechanism mediated by the 3Ј-UTR of DRD2 without considering the possible combined functional effects from other known regulatory variants, such as the functional promoter, splicing, and coding synonymous SNPs (28, 30 -32). Nonetheless, we revealed an miR-based (miR-9 and miR-326) post-transcriptional regulatory mechanism of DRD2 expression. As DRD2 activation in the brain initiates a cascade of intracellular signaling that modulates neurotransmission of FIGURE 6. Dynamic expression changes of DRD2, miR-9 and miR-326 during DA neuron differentiation from human iPSCs. A, human iPSCs were characterized by positive immunofluorescence staining for pluripotent stem cell markers TRA-1-60, OCT4, NANOG, and SSEA4 (4ϫ magnification). Three iPSC lines were generated from three different subjects. B, iPSCs were subjected to DA neuronal differentiation for 30 days (20ϫ magnification) and 45 days (60ϫ magnification). Neurons are ␤III-tubulinϩ, and DA neurons are THϩ and PITX3ϩ. DAPI stains nuclei. C, inverse correlation of DRD2 expression with miR-9 and miR-326 expressions during DA neuronal differentiation. DRD2 expression values were normalized to endogenous control GAPDH. Expression values of the two miRs were normalized to endogenous small RNA control RNU48. The y axis shows the relative expression values in log 2 scale. Multiplex quantitative PCR was used for expression quantification with three technical replicates per iPSC line, and the graphed data are the averages of the three iPSC lines.
importance to the pathogenesis of several neuropsychiatric diseases (for review see Ref. 83), our study may help to understand how DRD2 expression regulation is relevant to the pathogenesis of SZ and other neuropsychiatric disorders as well as to develop more effective clinical treatment of these disorders by modulating with the effect of miRs on DRD2 expression.  . Spearman's rank expression correlation between DRD2 and miR-9 and miR-326 in developing human brains. Listed are brain regions (DFC, dorsolateral prefrontal cortex; ITC, inferolateral temporal cortex; VFC, ventrolateral prefrontal cortex; AMY, amygdaloid complex; CBC, cerebellar cortex; V1C, striate cortex) that showed inverse expression correlation (i.e. negative R) between DRD2 and miRs. In each panel, the x axis shows the expression values of DRD2 (RPKM, reads/kb transcript/million reads; in log 2 scale), and the y axis shows the expression values of miR-9 (red dots) or miR-326 (light blue dots). Each dot represents a unique brain sample with age ranging from 4 months to 23 years old.