miR-936 is Increased in Schizophrenia and Inhibits Neural Development and AMPA Receptor-Mediated Synaptic Transmission

Abstract

MicroRNAs (miRNAs) are non-coding RNAs that regulate gene expression and play important roles in the development and function of synapses. miR-936 is a primate-specific miRNA increased in the dorsolateral prefrontal cortex (DLPFC) of individuals with schizophrenia. The significance of miR-936 increase to schizophrenia is unknown. Here, we show that miR-936 in the human DLPFC is enriched in cortical layer 2/3 and expressed in glutamatergic and GABAergic neurons. miR-936 is increased from layers 2 to 6 of the DLPFC in schizophrenia samples. In neurons derived from human induced pluripotent stem cells (iNs), miR-936 reduces the number of excitatory synapses, inhibits AMPA receptor-mediated synaptic transmission, and increases intrinsic excitability. These effects are mediated by its target gene TMOD2. These results indicate that miR-936 restricts the number of synapses and the strength of glutamatergic synaptic transmission by inhibiting TMOD2 expression. miR-936 upregulation in the DLPFC, therefore, can reduce glutamatergic synapses and weaken excitatory synaptic transmission, which underlie the synaptic pathology and hypofrontality in schizophrenia.

Introduction

Schizophrenia is a serious psychiatric disorder with such symptoms as hallucinations, delusions, and cognitive impairments, which typically emerge in late adolescence and early adulthood.¹ The cognitive impairment in schizophrenia is associated with reduced activation of the dorsolateral prefrontal cortex (DLPFC) during cognitive processing.² The DLPFC of people with schizophrenia is reduced in volume and thickness due to a decrease in neuropil.³ DLPFC has a long maturational trajectory as the maturation of its synaptic connectivity, functionality in working memory and emotional regulation, and the stabilization of its global microRNA (miRNA) expression last into late adolescence.^4–7 This delayed maturation prolongs experience-dependent brain plasticity but makes the DLPFC vulnerable to disturbances during brain development.⁸

Dysregulation of gene expression in the brain, has been implicated in schizophrenia.⁹ miRNAs are small non-coding RNAs that regulate gene expression by binding to mRNAs, usually at the 3′ untranslated region, to suppress translation or destabilize mRNAs.¹⁰ miRNAs are involved in the development and activity-induced modification of synapses and have been implicated in schizophrenia.^11–13 We previously showed that miR-936 is increased in the DLPFC of individuals with schizophrenia using next-generation sequencing.¹⁴ This increase is likely caused by the disease rather than antipsychotics as there is no correlation between miR-936 expression and psychiatric medications.¹⁴ miR-936 is a primate-specific miRNA with a distinct developmental profile in the human DLPFC, which is low in toddlers and high in adolescents.¹⁴ No genetic association with schizophrenia has been reported for miR-936. This is not surprising given the poor overlap of published results from genetic and expression studies of schizophrenia.^15,16

Here, we validated miR-936’s expression change in an independent cohort, examined the laminar and cell-type distribution of miR-936 in the human DLPFC, and explored its function in human induced pluripotent stem cell (iPSC)-derived glutamatergic neurons.

Methods

Human Brain Samples

Human brain tissues were provided by the NIMH Human Brain Collection Core. Informed consent was obtained from the surviving next-of-kin.

Differentiation of iPSCs

WTC11i iPSCs were differentiated with a 2-step protocol (pre-differentiation and maturation).¹⁷ For pre-differentiation, iPSCs were seeded on plates coated with Matrigel and cultured for 3 days in knockout DMEM medium containing N2 supplement, non-essential amino acids, mouse laminin, brain-derived neurotrophic factor, neurotrophin-3, Y-27632, SB431542, XAV939, LDN-193189, and doxycycline. For maturation, pre-differentiated cells were co-cultured with rat glial cells on coverslips coated with polyethylenimine and in medium containing 50% DMEM/F12, 50% Neurobasal-A medium, 0.5× B27, 0.5× N2, GlutaMax, NEAA, mouse laminin, BDNF, and NT3.

RNA Extraction and Quantitative Polymerase Chain Reaction

Total RNAs were isolated with Trizol reagent and treated with RNase-free DNase I to eliminate genomic DNAs. cDNAs were synthesized using the TaqMan microRNA Reverse Transcription Kit and analyzed with the miR-936 TaqMan probe.

Immunolabeling

Cell cultures were fixed with 4% paraformaldehyde in PBS, blocked with 5% normal goat serum in 0.1% Triton X-100, followed by incubation with primary and secondary antibodies. Formaldehyde fixed human DLPFC samples were cryoprotected with 30% sucrose and cryo-sectioned into 40-µm-thick sections. After permeabilization with 0.5% Triton X-100, sections were acetylated in 0.25% acetic anhydride/100 mM triethanolamine (pH 8) and hybridized with LNA probes for miR-936 or a mixture of previously reported 4 oligoprobes for vGlut1 at 55°C overnight.¹⁸ Sections were washed at 65°C with SSC and blocked with 5% Roche blocking reagent in 1X SSC. The probe was detected with HRP conjugated antibodies and tyramide. Immunolabeling was conducted after fluorescence in situ hybridization (FISH).

Image Acquisition and Image Analysis

Images were acquired in the NIMH Systems Neuroscience Image Resource and the NINDS Neuroscience Light Imaging Facility. Brain sections were imaged on an Axio Scan.Z1 slide scanner (Zeiss) at 10× magnification. iNs were imaged on a confocal microscope (LSM800) with a 63× (NA 1.4) oil immersion objective. Neurolucida 360 software was used for image analysis. All image acquisition and image analysis were done blindly to treatments.

Electrophysiology

Recording micropipettes (3–6 MΩ) were filled with internal solution composed of (in mM): 130 K-gluconate, 0.1 EGTA, 1 MgCl₂, 2 MgATP, 0.3 NaGTP, 10 HEPES, 5 NaCl, 11 KCl, and 5 Na₂-phosphocreatine (pH 7.4, 290–295 mOsM/kg). Electrical signals were recorded at 30°C using a MultiClamp-700B amplifier and analyzed with Clampfit v10.7 and OriginPro 2015G software. Cells with access resistance ≥25 MΩ were excluded from further analysis.

Statistical Analysis

The GraphPad Prism 6 software was used for statistical analysis. Two-tailed Student’s t-test, 1-way ANOVA and 2-way ANOVA were used. Bonferroni test was used for post-hoc analysis. P < .05 was considered significant.

Results

The Laminar and Cell-Type Distribution of miR-936 in the Human DLPFC

To validate our earlier finding of increased miR-936 expression in the DLPFC in schizophrenia,¹⁴ we examined the postmortem brains of a different human cohort (supplementary table 1) with FISH. We examined the specificity of miR-936 FISH in mouse brains which lacks the miR-936 gene. Cortical layers were identified by immunohistochemistry (IHC) for cut like homeobox 1 (Cux1, enriched in layer 2/3) and Ctip2 (enriched in layer 5/6)¹⁹ (supplementary figure 1A). No fluorescence signals were detected in the mouse cortex (supplementary figure 1A), indicating the specificity of miR-936 FISH.

We applied the miR-936 probe to human DLPFC sections. Cortical layers were identified by IHC for Cux1 and microtubule-associated protein 1B (MAP1B, enriched in layer 5/6).²⁰ miR-936 was nonuniformly distributed across the cortical layers with the strongest signal in layer 2 (figures 1A–D). Consistent with our previous findings,¹⁴ miR-936 increased in the DLPFC of schizophrenia samples from layer 2 to 6 (figures 1C and 1D).

Fig. 1.

miR-936 expression in the human dorsolateral prefrontal cortex (DLPFC). Human DLPFC sections (A–D) were used for FISH for miR-936 and immunostaining for cortical layer markers. (A–C) Representative images of combined FISH for miR-936 and immunostaining for Cux1 (A and C) and Map1b (B) in human DLPFC from control (A–C) and schizophrenia (C) samples. The numbers to the left of the images in A–C indicate cortical layers. (D) Quantification of miR-936 expression across the DLPFC; the integrated miR-936 fluorescence intensity in the soma of cells identified by DAPI staining is plotted against the distance from the pia mater; F_{(11, 99)} = 7.4, P < .0001, 2-way ANOVA. (E) Quantification of Manders’ correlation coefficient (MCC) for colocalization of vGlut1 with miR-936; P = .03, 2-tailed Student’s t-test. (F) MCC for colocalization of GAD-65/67 with miR-936; P = .0002, 2-tailed Student’s t-test. (G) miR-936 expression level in vGlut1⁺ cells; F_{(11, 174)} = 2, P < .03, 2-way ANOVA. (H) miR-936 expression level in GAD-65/67⁺ cells; F_{(11, 165)} = 0.8, P = .8, 2-way ANOVA. n = number of brains in D–H. (I–K) Representative images of double FISH (E) or combined FISH and immunostaining in control human DLPFC sections. Data are represented as mean ± SEM; *P < .05, ** P < .01, ***P < .001. Scale bars: 200 µm for low-magnification images, 20 µm for high magnification images.

We further examined the cell types expressing miR-936 with double FISH for miR-936 and vGluT1 (a glutamatergic neuron marker), and FISH combined with IHC for miR-936 and GAD-65/67 (a GABAergic neuron marker). We measured the Manders’ colocalization coefficient (MCC), which indicates the fraction of one protein co-localizing with the other,²¹ for cells in layer 2/3 as they had relatively high miR-936 expression and cell density. miR-936 was detected in both vGlut1⁺ and GAD-65/67⁺ cells, and SCZ samples had more vGlut1⁺ cells and fewer GAD-65/67⁺ cells that expressed miR-936 (figures 1E, 1F, 1I, and 1L). In SCZ samples, individual vGlut1⁺ cells expressed more miR-936, while miR-936 expression in GAD-65/67⁺ cells was unchanged (figures 1G and 1H). No miR-936 was detected in GFAP⁺ glial cells (figure 1K).

miR-936 Increases Intrinsic Excitability and AP Threshold in iNs

To assess the effect of miR-936 increase on glutamatergic neurons, we used neurons derived from the human iPSC line WTC11i. WTC11i cells have stably integrated neurogenin-2 gene under control of a doxycycline-inducible promoter that enables differentiation into glutamatergic cortical neurons.¹⁷ We used a differentiation protocol that combines doxycycline-induced neurogenin-2 expression, dual-SMAD inhibition of BMP and TGF-β signaling pathways, and inhibition of Wnt signaling (supplementary figure 2A).²² This approach enables iNs to exhibit AMPAR and NMDAR-mediated currents within 60 days of culture.²² Rat glial cells were added to the iPSC culture in a 2:1 ratio (2 iPSCs, 1 glia) to provide structural and metabolic support to developing iPSC-derived neurons (iNs). By 60–65 days in culture, iNs expressed the neuron-specific tubulin Tuj1 and vGlut1, and grew processes containing the dendritic marker MAP2, the presynaptic active zone protein Bassoon, and the postsynaptic density protein PSD-95 (supplementary figures 2B and 2C).

To test if our iNs possess the electrophysiological properties of neurons, we conducted whole-cell patch-clamp recording. Older iNs had lower input resistance (R_in,), larger cell capacitance (Cp), and a more hyperpolarized resting membrane potential (RMP) (supplementary figures 2D–F) than younger iNs. The threshold and half-width of action potentials (AP) were decreased, while rheobase (the minimal current that can trigger AP), AP amplitude, and the number of AP induced by current injection increased from day 14 to day 56 (supplementary figures 2G–M). Rheobase and R_in are inversely correlated.²³ A high R_in and low rheobase correlate with high intrinsic excitability.²⁴ Hence, older iNs had lower intrinsic excitability. These alterations in electrophysiological properties with age are similar to those observed in iNs and during the postnatal development of glutamatergic neurons in rodents.^25–30 Our morphological and electrophysiological analyses indicate that the iPSCs had differentiated into glutamatergic neurons. Next, we tested if miR-936 is expressed in iNs using quantitative reverse transcription polymerase chain reaction (qRT-PCR). While miR-936 was negligible in iPSCs, day-3 iNs, and day-15 iNs, it was detectable in day-60 iNs (figure 2A).

Fig. 2.

The effect of miR-936 on the passive membrane properties and AP in iNs. iNs were transduced with lentivirus expressing miR-936, miR-936Δ (miR-936 mutant), miR-936-sponge (miR-936-SPG) or scrambled miR-936-sponge (Scr-SPG) on day-3 and used for qRT-PCR at designated time points or electrophysiology on day-60. (A) The expression change of miR-936 relative to day-3 iNs; n = number of cultures. (B) Sample APs evoked by current injection (10 ms, 50 pA). (C) Quantification of AP threshold in day-60 iNs transduced with designated virus; F_{(4, 171)} = 36.2, P < .0001, 1-way ANOVA. (D) Quantification of the number of APs induced by current injection; F_{(4, 76)} = 7.5, P < .0001, 2-way ANOVA. (E) Sample traces of membrane potentials induced by injection of 500-ms currents. (F) Quantification of R_input in day-60 iNs; F_{(4, 115)} = 145, P < .0001, 1-way ANOVA. (G) Quantification of rheobase in day-60 iNs; F_{(4, 72)} = 120.04, P < .0001, 1-way ANOVA. Data are presented as mean ± SEM, n = the number of cells from 3 independent cultures in C, D, F, and G. Bonferroni test was used for post hoc analysis; ** P < .01, ***P < .001.

To test for the function of miR-936 in iNs, we transduced day-3 iNs with lentivirus expressing GFP along with miR-936, miR-936Δ (a miR-936 mutant carrying 3-point mutations in the seed region and incapable of binding to miR-936 targets; supplementary table 2), or miR-936-sponge that binds to miR-936 in the seed region for loss-of-function studies.³¹ The miR-936 and miR-936Δ lentivirus increased miR-936 and miR-936Δ, respectively, while miR-936-sponge lentivirus reduced miR-936 in both day-15 iNs and day-60 iNs (figure 2A). Scrambled-sponge had no significant effect (figure 2A). The reduction of miRNA by sponges has been reported in previous studies.^32,33 This is possibly due to either miRNA degradation and/or inhibition of miRNA detection by sponges that bind to miRNAs. In both cases, sponges cause less miR-936 binding to its natural target mRNAs.

RMP, Cm, AP amplitude, and AP half-width were unchanged by virus expressing miR-936 or miR-936-sponge (supplementary table 3). However, miR-936 increased AP threshold and decreased the number of AP induced by current injection (figures 2B–E). Additionally, R_in increased and rheobase decreased in iNs overexpressing miR-936 (figures 2F and 2G)_. miR-936Δ did not affect the above electrophysical properties (figures 2B–G). Hence, miR-936 increases the intrinsic excitability of iNs. miR-936-sponge had opposite effects on AP and intrinsic excitability as miR-936. Since a high AP threshold and high intrinsic excitability are characteristic of immature neurons, these findings suggest that miR-936 inhibits the electrophysiological development of glutamatergic neurons.

miR-936 Inhibits both Spontaneous and Evoked Excitatory Postsynaptic Currents

We next tested if miR-936 affects synaptic transmission by recording miniature excitatory postsynaptic currents (mEPSCs) in the presence of TTX (300 nM) in day 60–65 iNs. mEPSCs were mediated by AMPA receptors as they were blocked by the AMPA receptor antagonist NBQX (20 µM) (figure 3A). While miR-936 decreased both the frequency and amplitude of mEPSCs, miR-936-sponge increased them and miR-936Δ had no effect (figures 3A–C).

Fig. 3.

miR-936 reduces excitatory postsynaptic synaptic responses and restricts synapse number. (A–C) iNs were transduced with lentivirus on day-3 and used for electrophysiology on day 60–65. (A) Sample current traces of mEPSCs in day-60 iNs; NBQX (20 mM) was added to the bath. (B and C) Quantification of frequency and amplitude of mEPSCs; mEPSC frequency: F_{(4, 127)} = 34, P < .0001; mEPSC amplitude: F_{(4, 103)} = 22, P < .0001; 1-way ANOVA. (D–G) iNs transduced with lentivirus expressing GFP along with designated sequences were co-cultured with ChR2-expressing iNs. ChR2^- and GFP⁺ iNs were recorded on day-60. (D) Schematic drawing of recording in cultures. (E) Sample traces of light-evoked EPSC_AMPA recorded at a holding potential of −70 mV and EPSC_NMDA at +40 mV and 60–70 ms after stimulation. (F and G) Quantification of light-evoked EPSC_AMPA and EPSC_NMDA. EPSC_AMPA: F_{(4, 102)} = 58, P < .0001, 1-way ANOVA. (H–J) ChR2-iNs were transduced with designated lentivirus and co-cultured with unmodified, untransduced iNs; light-evoked EPSCs were recorded in untransduced iNs. (H) Schematic drawing of recording in cultures. (I) Sample current traces induced by paired stimuli. (J) Quantification of paired-pulse ratio (PPR). (K) Representative images of day-60 iNs; scale bars, 20 µm for low-magnification images and 10 µm for high-magnification images. (L) Quantification of synapses as indicated by puncta doubly positive for PSD-95 and Bassoon; n = number of neurons from 3 independent experiments, F_{(4, 95)} = 57.9, P < .0001, 1-way ANOVA with Bonferroni test for post-hoc comparison. Data are presented as mean ± SEM. *P < .05, **P < .01, ***P < .001.

As mEPSC frequency correlates with the probability of synaptic vesicle release and synapse number, and mEPSC amplitude is dependent on the abundance of postsynaptic AMPA receptors,^34,35 we examined these synaptic properties by evoking synaptic responses with optical stimulation. To effectively elicit synaptic transmission between iNs, we generated an iPSC line expressing channelrhodopsin (ChR2) for light-induced action potentials and tdTomato for cell visualization. iPSCs with and without ChR2 were co-cultured. In ChR2⁺ cells, 473-nm light illumination (0.5 mW/mm²) elicited APs (supplementary figures 3A and 3B). In iNs without ChR2, the frequency of spontaneous EPSCs was increased by light stimulation (supplementary figure 3C), presumably due to synaptic transmission from synaptically connected ChR2⁺ iNs. These results indicate that optical stimulation can induce AP and synaptic transmission between iNs.

The ChR2-iPSCs were co-cultured with iPSCs transduced with lentivirus expressing GFP alone or along with miR-936, miR-936Δ, miR936-sponge or Scr-sponge. The mixed culture was stimulated with 473-nm light pulses (5 ms in pulse duration, 0.5 mW/mm²) to evoke EPSCs. miR-936 decreased AMPA receptor-mediated currents (EPSC_AMPA), miR-936-sponge increased EPSC_AMPA, and they had no effect on NMDA receptor-mediated currents (EPSC_NMDA) (figures 3E–G). miR-936Δ and Scr-Sponge had no effect on EPSC_AMPA or EPSC_NMDA (figures 3E–G).

To assess presynaptic release probability, we transduced ChR2-iPSCs with lentivirus expressing GFP alone or along with miR-936, miR-936Δ, miR936-sponge or Scr-sponge, then co-cultured them with unmodified, untransduced iPSCs. Optically evoked EPSCs were recorded in non-fluorescent iNs (figure 3H). Paired-pulse ratio (PPR), the ratio of the second response to the first response induced by a pair of stimuli, was comparable in all groups (figure 3J), excluding presynaptic release probability as the cause of reduced mEPSC frequency in miR-936 overexpressing cells. Moreover, synapse number assessed from double staining for PSD-95 and Bassoon was decreased by miR-936, increased by miR-936-sponge, and unchanged by miR-936Δ or Scr-Sponge (figures 3K and 3L).

Taken together, these results indicate that miR-936 inhibits both spontaneous and evoked AMPA receptor-mediated synaptic transmission and reduces synapse number.

miR-936 Regulates Synaptic Transmission and Synapse Number Through TMOD2

To investigate the mechanism by which miR-936 regulates synapses, we examined potential miR-936 targets identified by TargetScan.³⁶ TMOD2 is a predicted target with multiple miR-936 binding sites. It is an actin-binding protein involved in the morphological remodeling of dendritic spines during long-term synaptic depression.^37,38 TMOD2 overexpression increases the number of dendritic spines in rat hippocampal neurons.³⁹ TMOD2 knockout mice have enhanced long-term synaptic potentiation (LTP).⁴⁰ Given the multiple miR-936 binding sites in TMOD2 which make it more likely to be modulated by miR-936 and the connection of TMOD2 to synapses, we chose TMOD2 for further testing. We confirmed that TMOD2 is a miR-936 target in iNs as miR-936 reduced TMOD2 and miR-936-sponge increased TMOD2 (supplementary figures 4A and 4B).

We transduced iNs with lentivirus expressing TMOD2 or TMOD2 siRNAs (supplementary figures 4C and 4D). TMOD2 siRNA’s efficacy and specificity were previously validated.³⁸ TMOD2 and TMOD2 siRNA did not affect RMP, Cm, AP amplitude or AP half-width (supplementary table 3). TMOD2 decreased R_in, AP threshold, and it increased rheobase, the number of AP induced by current injection (figures 4A–D; supplementary figures 5A and 5B). The amplitude and frequency of mEPSCs, EPSC_AMPA, and synapse number were increased by TMOD2 (figures 4E–L). TMOD2 siRNAs had opposite effects on these measurements (figures 4E–L). EPSC_NMDA was comparable in all groups (figures 4H–J). In addition, the effects of miR-936 and miR-936-sponge were blocked by TMOD2 or TMOD2-siRNAs (figures 5A–G; supplementary figures 5A and 5D). These results indicate that TMOD2 has opposite effects on the electrophysiological properties and synapse number in iNs as miR-936 and mediate, at least partially, the effect of miR-936 on synapses.

Fig. 4.

The effect of TMOD2 on the electrophysiological properties and synapse number in iNs. Day-60 iNs transduced with the designated virus were patched for electrophysiology (A–J) or immunostaining (K and L). (A–D) Quantification for R_input (F_(4,80) = 287, P < .0001, 1-way ANOVA), rheobase (F_(4,863) = 43, P < .0001, 1-way ANOVA), AP threshold (F_{(4, 167)} = 64, P < .0001, 1-way ANOVA), and the number of AP induced by current injection (F_{(4, 76)} =2.6, P < .0001, 2-way ANOVA). (E) Sample traces of mEPSCs. (F) Quantification of mEPSC frequency; F_{(4, 127)} = 53, P < .0001, 1-way ANOVA. (G) Quantification of mEPSC amplitude; F_{(4, 102)} = 30, P < .0001, 1-way ANOVA. (H) Sample traces of light-evoked EPSCs. (I) Quantification of light-evoked EPSC_AMPA; F_{(4, 98)} = 3.3, P = .015, 1-way ANOVA. (J) Quantification of light-evoked EPSC_NMDA. (K) Representative images of immunolabeling for PSD-95 and Bassoon; scale bars, 20 µm for low-magnification images and 10 µm for high-magnification images. (L) Quantification of the number of synapses as indicated by puncta doubly positive for Bassoon and PSD-95; n = number of cells from 3 independent experiments; F_{(4, 95)} = 48.72, P < .0001, 1-way ANOVA with Bonferroni test for post-hoc comparison. Data are presented as mean ± SEM. *P < .05, **P < .01, ***P < .001.

Fig. 5.

TMOD2 reverses the effect of miR-936 on the electrophysiological properties and synapse number. Day-60 iNs transduced with the designated virus were used for electrophysiology (A–I) or immunostaining (J and K). (A–D) Quantification for R_input (F_(3,144) = 31.7, P < .0001, 1-way ANOVA), rheobase (F_{(3, 63)} = 9.2, P < .0001, 1-way ANOVA), AP threshold (F_{(3, 142)} = 13.29, P < .0001, 1-way ANOVA), and the number of AP induced by current injection (F_{(3, 57)} = 2.3, P < .0001, 2-way ANOVA). (E) Sample traces of mEPSCs. (F) Quantification of mEPSC frequency (F_{(3, 70)} = 45, P < .0001, 1-way ANOVA). (G) Quantification of mEPSC amplitude (F_{(3, 70)} = 6.8, P = .0005, 1-way ANOVA). (H) Sample traces of light-evoked EPSCs. (I) Quantification of paired-pulse ratio. (J) Representative images of immunolabeling for PSD-95 and Bassoon; scale bars, 20 µm for low-magnification images and 10 µm for high-magnification images. (K) Quantification of the number of synapses as indicated by puncta doubly positive for PSD-95 and Bassoon. n = number of cells from 3 independent experiments; F_{(3, 64)} = 29.3, P < .0001, 1-way ANOVA with Bonferroni test for post-hoc analysis. Data are presented as mean ± SEM. *P < .05, **P < .01, ***P < .001.

To test if miR-936 overexpression affects synaptic physiology in older neurons as well, we transduced day-50 iNs with virus that alters miR-936 and TMOD2 and recorded from them on day 60. RMP, Cm, R_input, AP amplitude, and half-width were unchanged (supplementary table 4). However, miR-936 increased AP rheobase and decreased AP-threshold, AP firing rate, and the amplitude and frequency of mEPSCs, while miR-936-sponge had opposite effects (supplementary figures 6A–E). The effects of miR-936 and miR-936-sponge on the passive membrane properties and mEPSCs were abolished by TMOD2 and TMOD2 siRNA, respectively (supplementary figures 6A–E). These results indicate that miR-936 elevation in older neurons can also influence the electrophysiological properties.

Discussion

miRNAs have pleiotropic effects on cellular activities. Altered miRNA expression has been detected in the brains of individuals with schizophrenia.⁴¹ Here, we examined the function of miR-936, a miRNA increased in the DLPFC of schizophrenia cases. Our study shows that miR-936 is expressed in neurons and that increased miR-936 expression alters synapse number, AMPA receptor-mediated currents, and intrinsic excitability in glutamatergic iNs through, at least in part, TMOD2.

The miR-936 gene is primate-specific and it is located on chromosome 10 in humans. Our earlier study using next-generation sequencing showed that miR-936 is increased by ~2-folds in the DLPFC of people with schizophrenia.¹⁴ We replicated this finding here by using FISH in an independent set of postmortem tissues. FISH also revealed that miR-936 has a non-uniform distribution pattern across the cortical layers of DLPFC with the highest expression in layer 2/3. This is in line with the high degree of heterogeneity in gene expression through cortical layers.^42,43 In the DLPFC of schizophrenia cases, miR-936 in glutamatergic neurons is increased throughout layers 2 to 6. Despite such limitations common to postmortem studies as relatively small number of brains due to limited availability of human brains and biological variations of individual brains, they provide valuable information on gene expression in specific brain regions, cortical layers, and cell types that are inaccessible from genomic sequences. The postmortem work could be complemented by future examination of miR-936 expression in iPSCs derived from schizophrenia samples. This is an interesting experiment beyond the scope of this study.

We used human iNs because miR-936 is only present in primates. While miR-936 overexpression did not affect passive membrane properties, it increased AP threshold and intrinsic excitability and decreased both spontaneous and evoked synaptic transmission. miR-936 also decreased synapse number. Higher AP threshold and intrinsic excitability, weaker synaptic transmission, and fewer synapses are characteristics of immature neurons.^27,44–46 These findings suggest that miR-936 inhibits neural development. This function in combination with the developmental profile of miR-936 in the DLPFC, which is low in toddlers and high in adolescents,¹⁴ suggests that miR-936 is well-suited to regulate brain development in that the low level of miR-936 in toddlers allows neural development, while the miR-936 increase during adolescence facilitates the refinement of synaptic connectivity. As adolescence precedes the typical age of onset of schizophrenia, miR-936 dysregulation during this period may increase the risk of schizophrenia.

There is a reduction of dendritic spine density in the DLPFC of people with schizophrenia, particularly in layer 3.^47,48 The mechanism of spine loss in schizophrenia appears to be unrelated to antipsychotic medications.³ Our findings of reduced synaptic transmission and synapse number in glutamatergic neurons with elevated miR-936 raise an intriguing possibility that miR-936 may contribute to the synaptic pathology of schizophrenia.

We confirmed that TMOD2 is a target gene of miR-936 in iNs. TMOD2 is an actin-binding protein belonging to the tropomodulin family and expressed in the central nervous system.⁴⁹ It caps the minus end of actin filaments, thereby blocking actin monomer exchange, inhibiting actin depolymerization, and stabilizing actin filaments.³⁷ Consistent with the reported effect of TMOD2 on dendritic spines, TMOD2 increased synapse number. TMOD2 abolished the effect of miR-936 on intrinsic properties, synaptic transmission, and synapse number, indicating that TMOD2 is essential for the effect of miR-936.

In summary, this study shows that miR-936 upregulation affects synapse number and synaptic function, therefore, may contribute to the synaptic pathology in schizophrenia.

Funding

This work was supported by the Intramural Research Program of National Institute of Mental Health, National Institutes of Health (ZIA MH002882).