Neuregulin 1 Type I Overexpression Is Associated with Reduced NMDA Receptor–Mediated Synaptic Signaling in Hippocampal Interneurons Expressing PV or CCK

Abstract Hypofunction of N-methyl-d-aspartate receptors (NMDARs) in inhibitory GABAergic interneurons is implicated in the pathophysiology of schizophrenia (SZ), a heritable disorder with many susceptibility genes. However, it is still unclear how SZ risk genes interfere with NMDAR-mediated synaptic transmission in diverse inhibitory interneuron populations. One putative risk gene is neuregulin 1 (NRG1), which signals via the receptor tyrosine kinase ErbB4, itself a schizophrenia risk gene. The type I isoform of NRG1 shows increased expression in the brain of SZ patients, and ErbB4 is enriched in GABAergic interneurons expressing parvalbumin (PV) or cholecystokinin (CCK). Here, we investigated ErbB4 expression and synaptic transmission in interneuronal populations of the hippocampus of transgenic mice overexpressing NRG1 type I (NRG1tg-type-I mice). Immunohistochemical analyses confirmed that ErbB4 was coexpressed with either PV or CCK in hippocampal interneurons, but we observed a reduced number of ErbB4-immunopositive interneurons in the NRG1tg-type-I mice. NMDAR-mediated currents in interneurons expressing PV (including PV+ basket cells) or CCK were reduced in NRG1tg-type-I mice compared to their littermate controls. We found no difference in AMPA receptor–mediated currents. Optogenetic activation (5 pulses at 20 Hz) of local glutamatergic fibers revealed a decreased NMDAR-mediated contribution to disynaptic GABAergic inhibition of pyramidal cells in the NRG1tg-type-I mice. GABAergic synaptic transmission from either PV+ or CCK+ interneurons, and glutamatergic transmission onto pyramidal cells, did not significantly differ between genotypes. The results indicate that synaptic NMDAR-mediated signaling in hippocampal interneurons is sensitive to chronically elevated NGR1 type I levels. This may contribute to the pathophysiological consequences of increased NRG1 expression in SZ.

NRG1 has several functionally distinct isoforms, of which type I (among others) has been reported to be overexpressed in SZ (Hashimoto et al., 2004;Law et al., 2006). Overexpression of NRG1 type I mRNA, or administration of the protein in early postnatal development, results in pathophysiological changes reminiscent of schizophrenia endophenotype in animal models: alterations in rhythmic gamma-frequency network oscillations (Deakin et al., 2012) and synaptic plasticity (Agarwal et al., 2014), and a behavioral phenotype including age-emergent impairment of hippocampal working memory (Chen et al., 2008;Deakin et al., 2009;Yin et al., 2013;Luo et al., 2014). These findings together suggest that NRG1-ErbB4 signaling may regulate glutamatergic NMDARmediated transmission in interneurons, and that alterations in this mechanism might contribute to the pathophysiology of SZ. To investigate this possibility, we have studied synaptic function in hippocampal interneurons expressing PV or CCK in mice overexpressing NRG1 type 1, using a combination of electrophysiological, optogenetic, and immunohistochemical techniques.

Ethics statement
All animal procedures were performed in accordance with British Home Office regulations and personal and project licenses held by the authors, following local ethics review at the University of Oxford (UK).

Opsin construct transduction
Mice were anesthetized with 2%-4% isoflurane (CHEBI: 6015). AAV2-ChR2-eYFP (in some cases AAV5-ChR2-eYFP) was stereotactically injected via 33-gauge needle attached to a Microlitre Syringe (Hamilton) into midventral CA3 or dorsal CA1 hippocampus. The vector sequence was: pAAV-EF1a-sCreDIO hChR2(H134R)-EYFP-WPRE (Vector Core Services, Gene Therapy Center Virus, University of North Carolina). In each hemisphere, a craniotomy was performed using a micro-torque, and a total volume of 800 nl virus suspension (viral particle suspension titer 4 ϫ 10 12 /mL) was delivered at 80 nl/min by a Micro Syringe Pump Controller (World Precision Instruments). The scalp incision was sutured, and mice were allowed to recover for 10 -21 d. Light exposure of brain tissue during preparation of slices was minimized to avoid photoactivation of ChR2. In experiments, ChR2 was activated by a fixed-spot laser (Laser nominal maximum power 100 mW; Rapp OptoElectronics) light (20-m diameter to evoke IPSCs with minimal stimulation of GABAergic fibers, and 80-m diameter in experiments stimulating glutamatergic fibers with 20-Hz train stimulation) via the microscope objective. Fig. 1 were tagged by the fluorescent marker tdTomato using the crossed mouse line: BAC-CCK-Cre tg with Ai9 mice. In Fig. 2, CCK-expressing interneurons were identified with positive immunoreaction for somatic pro-CCK or by positive immunoreaction for axonal CB1R when the soma recovery was compromised. In Figs. 1, 2, and 3, the PV-expressing cells were identified by genetic fluorescence marker in PV-Cre mice crossed with Ai9 mice. Recorded cells were filled with neurobiotin (0.3% w/v) and visualized, and some were anatomically identified as basket cells by their characteristic predominant axon distribution in str. pyramidale and the lack of axo-axonic cell axon terminal cartridges (Klausberger and Somogyi, 2008). In addition, the basket cells in Fig. 4 were confirmed immunonegative for axonal CB1R (Katona et al., 1999;Tsou et al., 1999;Bodor et al., 2005;Armstrong and Soltesz, 2012). Pyramidal cells (PCs) were identified by their somatodendritic structure with mushroom spines along the dendrites.
Extracellular electrical stimuli were applied via a bipolar electrode (50 -100 s, 50 -400 A) in stratum oriens and current isolator (CBAPC75PL1, FHC) every 15 s. Synaptic currents were post hoc lowpass filtered at 1 KHz. Pharmacologically isolated AMPA receptor (AMPAR)-mediated EPSC peak amplitude was recorded at -60 mV, and the NMDAR-mediated EPSC amplitude was measured in the presence of the AMPA/kainate receptor blocker NBQX at a membrane potential 40 mV positive to their measured reversal potential estimated by a linear fitting curve of the New Research Figure 1. ErbB4 expression in PV ϩ and PVinterneurons and the ErbB4 expression levels in hippocampus of WT and the NRG1 tg-type-I mice. A, Immunostaining for ErbB4, the NRG1 receptor, in the ventral hippocampus CA3 area neurons using highly specific rabbit anti-ErbB4 (polyclonal anti-antiserum 5941; Neddens and Buonanno, 2010). A1, Double immunolabeling for PV (Cy3) and ErbB4 (Alexa488). Merged image shows double-labeled neurons (arrowhead) and ErbB4 ϩ interneurons immunonegative for PV (arrows). s.r, stratum radiatum. A2, In mice with genetic fluorescence marker (tdTomato) in CCK cells (tdTom-CCK), ErbB4 immunostaining with Alexa Fluor 488 shows the expression in many CCK ϩ neurons in s.r. and stratum pyramidale (s.p.). Cre-dependent tdTomato signal is strong in putative CA3 interneurons (soma in s.r.) and weaker in s.p., where the majority of pyramidal cell somata are located (contrast adjustment in the image). In merged image, arrowheads point at interneuron somata with both fluorescent signals. Scale bars, 50 m. Confocal microscope images. B-D, Cell density analysis of hippocampal interneurons immunopositive for ErbB4 in the WT and NRG1 type I-overexpressing mice (NRG1 tg-type-I mice). B1, ErbB4 immunoreaction (20-m stack image) in sample hippocampal sections of WT (left) and NRG1 tg-type-I mice (right). Scale bar, 100 m. B2, Box plots show ErbB4 ϩ cell soma density current-voltage relation for at least 20 evoked NMDAR EPSCs measured between -20 and 65 mV (Deleuze and Huguenard, 2016). In cells where no NMDAR EPSC was detected, the current was defined as 0.mEPSC recordings (2 min for AMPAR and 2 min for NMDAR mEPSCs) were acquired at 20 kHz and bandpass filtered offline (cutoff frequencies 4 Hz to 5 or 6 kHz at -65 mV, 2-500 Hz at 40 mV) for analysis. Events were detected with an amplitude threshold-crossing algorithm in pClamp (Molecular Devices, SCR_011323). Criteria for threshold detection for NMDAR mEPSCs (at 40 mV) were amplitude threshold 7 pA, duration 0.8 -200 ms, with noise rejection 0.8 ms. For the AMPAR mEPSCs (at -65 mV) the amplitude threshold was 5 pA, duration 0.5-100 ms, with noise rejection 0.5 ms) evaluated after blockade of AMPARs with NBQX (25 M). The same detection criteria were employed for all cells. Number of AMPAR mEPSCs investigated in the analyses were as follows: in wild-type (WT) basket cells (median and interquartile range), 424 and 279 -680 events (7 cells); in NRG1 tg-type-I basket cells, 394 and 301-470 events (6 cells); in WT pyramidal cells, 134 and 128 -205 events (7 cells); in NRG1 tg-type-I pyramidal cells, 95 and 64 -150 events (10 cells). The numbers of NMDAR mEPSCs measured in similar time window were as follows: in WT basket cells (median and interquartile range), 513 and 178 -792 events (6 cells); in NRG1 tg-type-I basket cells, 348 and 280 -520 events (6 cells). mEPSC frequency was calculated from the 2-min time window as the event occurrence in Hz. Average mEPSC amplitude was calculated in each cell from all events occurring in the 2-min time window.
In experiments using optogenetic stimulation of GABAergic fibers, the monosynaptic IPSCs were measured at 0 to -10 mV. Optogenetic stimulation of the glutamatergic fibers (5 pulses at 20 Hz) was applied every 30 s while the disynaptic IPSCs were recorded (on average at 11 mV; see Results) in postsynaptic pyramidal cells. The optogenetically evoked postsynaptic currents were lowpass filtered offline at 1 kHz, and the evoked postsynaptic current charge was analyzed with pClamp10.2 (Molecular Devices, SCR_011323).

Statistics
A t test was used for data that were normally distributed (Shapiro-Wilk test) and with n Ն 10 in tested groups. Otherwise, a Mann-Whitney U test or rank sum test was used.
Sections for immunoreactions were washed in 50 mM TBS-Tx, blocked in 20% normal horse serum (NHS, Vector Laboratories) in TBS-Tx for at least 1 h at room temperature (20 -24°C), and incubated in primary antibodies for 48 h at 4°C in TBS-Tx with 1% NHS. Fluorochromeconjugated secondary antibodies were applied overnight at 4°C in TBS-Tx with 1% NHS. Mounted sections in continued (measured up to 20-m depth from the section surface) in WT (blue, n ϭ 9 sections in 3 mice) and NRG1 tg-type-I (red, 12 sections in 3 mice) mice hippocampi. The plot shows median and interquartile range. Fewer ErbB4 ϩ somata were detected in the NRG1 tg-type-I mice compared to the WT mice in all hippocampal areas. From the left: whole hippocampus including areas CA1, CA2, and CA3; area CA1-2 restricted to alveus, stratum oriens, and stratum pyramidale; area CA1-2 restricted to stratum radiatum and lacunosummoleculare; area CA3 containing alveus with strata oriens and pyramidale; and area CA3 with strata lucidum and radiatum and lacunosum-moleculare. p-values compare data between genotypes (Mann-Whitney U test). C1, Immunoreaction for PV in the same sections as in B1. C2, Cell density analyses show no difference in the observed PV ϩ cell somata between the two genotypes as indicated by p values (Mann-Whitney U test). Box plots as in B2. D1, Merged ErbB4 and PV immunolabeling in the sample sections above. D2, Box plots show proportion of the double-labeled cells (co-immunoreactive for ErbB4 and PV) in the PV ϩ cell population in WT and NRG1 tg-type-I mice. The analyses show unaltered proportion in the whole hippocampus and in most subregions compared separately. The significant p value is bolded. E, Immunoblot analysis of ErbB4 expression levels in WT and NRG1 tg-type-I mice using hippocampal extracts. E1, The antibody against ErbB4 detects a band of the predicted protein size (ϳ150 kDa) in hippocampal protein extracts. Left lane, no nonspecific bands were detected in the secondary-only antibody control (right lane). E2, Hippocampal extracts from 6 WT and 6 NRG1 tg-type-I mice of both genders (3 males and 3 females in each genotype in scrambled order) tested for ErbB4 expression. GAPDH was used as a loading control. E3, Box plot shows (mean and interquartile range) densitometry analysis comparison of the ErbB4 levels normalized by the GAPDH in the 12 hippocampal extracts (6 in both genotypes including 3 males and 3 females). The results indicate a general trend to lower ErbB4 levels in NRG1 tg-type-I mice, but with no significant difference between the genotypes (Mann-Whitney U test).

Figure 2.
Reduced synaptic NMDAR-mediated currents in hippocampal interneurons expressing PV or CCK in the NRG1 tg-type-I mice. A, Interneurons expressing PV or CCK in the CA3 area. A1, Sample image of a recorded PV interneuron identified by PV expression-dependent fluorescent genetic marker tdTomato (tdTom-PV). Recorded cells were also visualized with filled neurobiotin (nb, Alexa Fluor 488). A2, Recorded cells not showing tdTomato signal were identified as CCK ϩ interneurons post hoc with positive somatic immunoreaction for pro-CCK (left; Cy5, arrowhead) or in the absence of recovered soma and dendrites (right) by positive reaction for axonal cannabinoid receptor type 1 (CB1R, Cy3). Scale bars from left: 10, 20, and 10 m, respectively. B, Reduced NMDAR-versus AMPAR-mediated EPSCs ratio (N/A ratio) in glutamatergic synaptic input to interneurons expressing PV. Electrical stimulation was applied in CA3 stratum oriens aiming to activate associative/commissural pathways. AMPAR-mediated EPSCs were recorded at -60 mV (in PiTX, 100 M) and blocked by NBQX (25 M) to record NMDAR-mediated EPSCs (at 40 mV from their reversal potential). B1, Averaged EPSCs (10 traces) in sample PV ϩ interneurons in WT and NRG1 tg-type-I mouse (black, AMPAR EPSCs; green, NMDAR EPSCs in the presence of NBQX; gray, following application of NMDAR blocker DL-AP5). Scale bars, 100 pA, 25 ms. B2, Cumulative histograms of the N/A amplitude ratios in all studied PV ϩ interneurons (WT, blue line; NRG1 tg-type-I , red line). p indicates difference between the genotypes (Mann-Whitney U test). C, Reduced N/A ratio in glutamatergic synaptic input to the CCK ϩ interneurons. C1, Averaged EPSCs (10) in sample cells in the WT and in the NRG1 tg-type-I mouse with scaling as above. C2, Cumulative histogram quantifying the N/A ratios in CCK ϩ interneurons with p indicating significant difference between the genotypes (Mann-Whitney U test). D, The N/A ratio is unaltered between the genotypes in the CA3 pyramidal cells. D1, Averaged EPSCs (10 traces) in sample pyramidal cells with scaling as above. D2, Cumulative histograms of the N/A ratios.
Vectashield were evaluated at Ն40ϫ magnification using confocal laser-scanning microscopy (LSM710, Carl Zeiss) with Zen2008 software. Digital micrographs were constructed from z-stacks with ImageJ software (SCR:003070). Micrographs were only adjusted for brightness and contrast. The primary antibodies used were rabbit anti-ErbB4
tions were incubated overnight with Alexa Fluor 488conjugated and Alexa Fluor 647-conjugated secondary antibodies both raised in donkey, respectively, in TBS-Tx with 1% NHS. Sections containing mid-ventral hippocampus from both hemispheres were scanned using an epifluorescence microscope (AxioImager M2; Zeiss) equipped with Stereoinvestigator software (MBF Bioscience). Optical sections of 1 m were acquired using a 20ϫ objective at a final depth of 20 m from the section surface, while the first 1 m from the section surface was defined as a guard zone and not scanned (Bocchio et al., 2015). Brightness and contrast acquiring settings were adjusted for each section, to achieve good visualization of all positive cells for a specific neuromarker across all section areas. Cell counting was performed offline. Distinct hippocampal regions were visually delineated and analyzed as individual anatomically defined subregions as follows: CA1-2 alveus (alv)/stratum oriens (s.o)/stratum pyramidale (s.p), CA1-2 stratum radiatum (s.r)/stratum lacunosum-moleculare (s.l-m), CA3 alv/s.o/s.p, and CA3 stratum lucidum (s.l)/s.r/s.l-m. Cells were counted when the cell somata or nuclei came into focus with the optical dissector.
These results show that NRG1 type I overexpression does not produce significant changes in the coexpression of ErbB4 and PV in most hippocampal areas or in the spatial distribution of PV ϩ neurons in the hippocampus. Yet, these data suggest that NRG1 overexpression leads to altered ErbB4 ϩ cell soma count of interneurons other than those expressing PV. This could either emerge from changes in the migration, survival, and proliferation of these cells during neurodevelopment (Flames et al., 2004;Li et al., 2012a)  We found no significant difference in ErbB4 protein levels between the two genotypes using Western blot analysis of whole-hippocampus extracts (n ϭ 6 including 3 males and 3 females in both genotypes, p ϭ 0.310, Mann-Whitney U test; Fig. 1E1-E3). This discrepancy continued interquartile range) show data from all the PCs studied. C2, The evoked IPSC amplitudes. C3, The IPSC half decay. C4, The IPSC rise time. C5, The IPSC paired-pulse (50 ms) ratio (2nd/1st IPSC amplitude). p values with Mann-Whitney test. D, The IPSCs from CCK-fibers do not show significant difference between the genotypes. D1, Sample trace. D2-D5, The IPSC amplitude, IPSC half decay, rise time, and paired-pulse ratio, respectively (Mann-Whitney test). may be attributed partly to the fact that the cell density analysis focused on cells in specific hippocampal subregions, whereas the lysates in the immunoblots comprised the entire hippocampus, possibly masking subregionspecific differences (see Discussion).
In conclusion, the above results suggest that expression level or pattern of the ErbB4 in some hippocampal CA1 and CA3 cells is altered in response to NRG1 type I genomic overexpression (see Discussion). In addition, the analyses confirm earlier findings that ErbB4 is present in the hippocampal interneurons expressing PV or CCK (Vullhorst et al., 2009), and that both PV ϩ and the PVinterneuron subpopulations expressing the receptor ErbB4 are present in the NRG1 tg-type-I mouse hippocampus (see Fig. 1B2).

Hippocampal interneurons expressing PV or CCK have reduced synaptic NMDAR-mediated currents in the mice overexpressing NRG1 type I
Next, we studied synaptic AMPAR-and NMDARmediated glutamatergic EPSCs in three neuron subpopulations in the CA3 area of acute hippocampal slices; PV ϩ interneurons (Fig. 2A1), CCK ϩ interneurons (Fig. 2A2), which both commonly express the ErbB4 (see Fig. 1), and pyramidal cells in which the receptor is absent (Vullhorst et al., 2009). All cells were studied in the whole-cell voltage clamp mode in hippocampal slices from mice expressing fluorescent marker (tdTomato) in PV-interneurons (see Materials and methods). The CCK ϩ GABAergic interneurons were identified post hoc by positive immunoreaction for cytoplasmic pro-CCK (tested when cell soma was recovered, n ϭ 3 in WT control and n ϭ 4 in NRG1 tg-type-I ) or axonal CB1R (tested when only interneuron axon was recovered, n ϭ 7 and n ϭ 7 respectively; Fig. 2A2; Katona et al., 1999). We applied electrical microelectrode stimulation in the CA3 stratum oriens aiming to activate predominantly associative-commissural fibers. Blockers for GABA A and GABA B receptors (picrotoxin, 100 M, and CGP55845, 1 M) were present in all experiments. We found that the NMDAR-mediated EPSCs in PV ϩ interneurons of the NRG1 tg-type-I mice were smaller, in comparison to the AMPAR EPSCs, than in their WT littermate controls (measuring a ratio of the NMDAR-EPSC and the AMPAR-EPSC amplitude, N/A ratio; Fig. 2B1). The evoked average glutamatergic EPSCs in the NRG1 tg-type-I mice were (median, interquartile range): NMDA EPSC, 19.8 pA, 10.4 -45.5 pA; AMPAR EPSC, 110.7 pA, 79.1-136.0 pA. Correspondingly, the N/A ratio in the NRG1 tg-type-I mice was 0.18, 0.08 -0.29 (n ϭ 29). In the WT control mice, the NMDA EPSC amplitude was 47.6 pA (median, interquartile range 29.1-60.8 pA), and the AMPAR EPSC amplitude 127.8 pA, 79.6 -214.7 pA. Hence the N/A ratio in WT was 0.28, 0.19 -0.42 (n ϭ 38). The N/A ratios in PV ϩ cells of the two genotypes were different (p ϭ 0.010, Mann-Whitney U test). Fig. 2B2 shows cumulative histograms of the N/A ratios measured in the PV ϩ interneurons of the two genotypes.
Because both interneuron populations comprise various specialized cell types (Klausberger and Somogyi, 2008;, and glutamatergic synapse features may vary between individual interneuron types (Papp et al., 2013), we visualized and anatomically examined the recorded interneurons (filled with neurobiotin) to identify basket cells (PVBCs; Fig. 3A) in the PV ϩ subpopulation (see Fig. 2B). We confirmed 22 PVBCs (12 in the WT mice and 10 in the NRG1 tg-type-I mice). Interestingly, the PVBC group in both genotypes showed parametric distribution of the N/A values (in the NRG1 tg-type-I mice W ϭ 0.91, p ϭ 0.270; in the WT mice, W ϭ 0.96, p ϭ 0.780; Shapiro-Wilk test) showing that the N/A values have less variation in an identified PV ϩ cell type subpopulation than in the entire PV ϩ cell population in general. The PVBC data showed smaller N/A EPSC ratio in the NRG1 tg-type-I mice (0.14 Ϯ 0.04, n ϭ 10) than in the WT control mice (0.31 Ϯ 0.04, n ϭ 12; p ϭ 0.006, mean Ϯ SEM, t test; Fig. 3B1,B2). In addition to the basket cells, we identified two axo-axonic cells (Nissen et al., 2010) in the NRG1 tg-type-I mice (their average N/A ratios 0.09 and 0.21) and one in the WT control littermates (N/A ratio ϭ 0.18). Because of their low number, these cells were not separately compared between the genotypes (but the cells were included in the PV ϩ cell pool in Fig. 2).
The findings of unchanged AMPAR mEPSCs in the PVBCs and PCs (and the moderate reduction of the NMDAR mEPSC amplitude specifically in the PVBCs in the NRG1 tg-type-I mice) indicate that the altered N/A ratio observed (Fig. 3) was caused by reduced postsynaptic NMDAR currents in the NRG1 tg-type-I mice PVBCs.

GABAergic inhibitory currents from parvalbumin-or cholecystokinin-expressing CA3 interneurons are not altered in NRG1 tg-type-I mice
Given that alterations in NRG1 levels can acutely change inhibitory synapses and modify them long term (Okada and Corfas, 2004;Yin et al., 2013;Agarwal et al., 2014), we studied whether GABAergic synaptic output from interneurons expressing either PV or CCK is also altered in the NRG1 tgtype-I mice. To selectively stimulate axons from these interneurons, we prepared slices from NRG1 tg-type-I and WT mice expressing Cre-protein either in PV ϩ cells or CCK ϩ interneurons and transduced with a Cre-dependent adenoassociated virus (AAV)-channelrhodopsin-2 (ChR2)-eYFP construct (see Materials and methods). Expression of the construct in the two types of GABAergic fibers is illustrated in Fig. 5A1,A2. GABAergic IPSCs were elicited in the CA3 area pyramidal cells stimulating the interneuron axons locally with brief laser light pulses (3 ms, 473 nm) focused in stratum pyramidale. Stimulation intensity was set to use minimal laser power required for stable IPSCs (Fig. 5B). In all experiments, the postsynaptic pyramidal cells (voltage clamped at 0 to 10 mV) were recorded in the presence of glutamate receptor blockers NBQX (25 M) and DL-AP5 (100 M). The optically evoked IPSCs were blocked with picrotoxin (100 M) in all experiments tested (n ϭ 8 of 8 in IPSCs from PV ϩ fibers, and n ϭ 3 of 3 from CCK ϩ fibers).

Reduced NMDAR-driven recurrent inhibition in the hippocampus in NRG1 tg-type-I mice
Finally, we investigated whether the reduced synaptic NMDAR-mediated transmission in these two common recurrent inhibition interneuron subpopulations had consequences for the GABAergic inhibition evoked by repetitive firing of the hippocampal glutamatergic neurons. To study this, we optogenetically stimulated glutamatergic fibers, focusing the laser light pulses in stratum pyramidale and stratum oriens, aiming to activate the recurrent disynaptic GABAergic pathway. We did the experiments in the CA1 area to avoid polysynaptic glutamatergic discharge generated in the CA3 recurrent glutamatergic circuits (Maccaferri and McBain, 1995). We used hippocampal slices of the NRG1 tg-type-I ϩ/Ϫ mice and their littermate WT controls both crossed with CaMKII-CreϮ and transduced with the AAV-ChR2-eYFP construct in the hippocampus (Fig. 6A). We made a translaminar surgical cut in the slices from alveus to stratum lacunosum-moleculare in the CA1-CA2 area border to exclude the CA3 area recurrent excitatory loop and polysynaptic discharges (Maccaferri and McBain, 1995).
We applied bursts of five pulses of stimuli at 20 Hz every 60 s to generate disynaptic IPSCs in the CA1 area pyramidal cells. The IPSCs were recorded at a reversal potential of the EPSCs (11.1 Ϯ 0.7 mV, mean Ϯ SEM) elicited in the same cells (n ϭ 13 comprising 7 cells in the NRG1 tg-type-I mice and 6 cells in the WT controls; Fig.  6B1,B2). Long-term plasticity blockers KN-62 (3 M) and MCPG (200 M) were present in all experiments for longterm stability of the disynaptic IPSCs (Perez et al., 2001;Kullmann and Lamsa, 2011;Campanac et al., 2013). After a stable baseline (at least 5 min), NMDAR blocker DL-AP5 (100 M) was washed in (Fig. 6C1,C2). This suppressed the evoked recurrent GABAergic IPSC in the WT mice to 0.66 of baseline (charge median, interquartile range 0.61-0.71, p ϭ 0.031 vs. baseline, n ϭ 6 cells), and in the NRG1 tg-type-I mice to 0.74 (charge median, interquartile range 0.66 -0.83, p ϭ 0.026 vs. baseline, n ϭ 7 cells) compared to the baseline (Mann-Whitney rank sum test). The IPSC charge was compared in each experiment between the last 3 min in baseline, and in an equal time window in the presence of DL-AP5 (at 5-8 min after DL-AP5 application). The suppression of the disynaptic IPSCs by the NMDAR blocker was larger in the WT than in the NRG1 mice (p ϭ 0.014, Mann-Whitney U test). The IPSCs were fully blocked at the end by NBQX (25 M) in all experiments tested to verify their disynaptic origin (n ϭ 4 in the WT controls, and n ϭ 4 in the NRG1 tg-type-I mice).
The results, summarized in Fig. 6D, indicate smaller NMDAR-mediated excitatory drive of hippocampal GABAergic interneurons in the NRG1 tg-type-I mice compared to their WT littermates.

Discussion
Our results show that transgenic overexpression of NRG1 type I, an isoform of NRG1 that has elevated levels in some patients with SZ (Hashimoto et al., 2004;Law et al., 2006;Chong et al., 2008;see also Boer et al., 2009;Parlapani et al., 2010;Hahn, 2011), is associated with a hypofunction of NMDAR-mediated synaptic signaling in two major GABAergic interneuron populations in mouse hippocampus.
The reduced ratio of NMDAR-to AMPAR-mediated synaptic currents was observed in the hippocampal GABAergic interneuron populations expressing either PV or CCK, but not in PCs. This finding on cell type specificity is in line with the cortical ErbB4 expression pattern: various studies have demonstrated that ErbB4 expression is predominant in GABAergic interneurons, whereas it is absent in PCs (Vullhorst et al., 2009;Fazzari et al., 2010;Neddens and Buonanno, 2010;Abe et al., 2011;Pitcher et al., 2011;Del Pino et al., 2017). As illustrated in Fig. 1, we confirmed here the ErbB4 expression in both PV ϩ and CCK ϩ interneurons, as has been previously reported (Vullhorst et al., 2009). It should be noted that because of contrast adjustment, Fig. 1A2 shows low CCK-Cre-dependent fluorophore intensity in the CA3 pyramidal cells compared to interneurons, although CCK is expressed in both cell populations (Burgunder and Young, 1990;Geibel et al., 2014;Rombo et al., 2015).
ErbB4 ϩ interneurons expressing either PV or CCK were found in the NRG1 tg-type-I mice, but the cell soma counting analysis indicated that the density of ErbB4 ϩ neurons not coexpressing PV is reduced in the hippocampus of NRG1 type I-overexpressing mice. This suggests that in some interneurons, either the ErbB4 receptor abundance has changed or detectable ErbB4 immunoreactivity has decreased (e.g. due to an altered subcellular localization or a change in epitope accessibility). Surprisingly, Western blot did not detect the reduction of ErbB4 expression, although the detectable ErbB4 ϩ neuron soma number was reduced. We offer two possible explanations for this. The counting of immunohistochemically revealed ErbB4 ϩ cell somata is a nonquantitative method (giving cells clearly ErbB4 ϩ or cells not confirmed positive). If the antibodylabeled fluorescence signal in the soma is low, it becomes increasingly challenging to confirm it as immunopositive compared to background. This could happen in the NRG1 type I-overexpressing mice without a significant change in total hippocampal ErbB4 protein level, if subcellular location of the ErbB4 changed (decreased in soma) or the ErbB4 protein is internalized in some interneurons , Longart et al., 2007, making its detection by the antibody less evident. The discrepancy may also be attributed to the fact that the cell density analysis focused on cells in specific hippocampal subregions, whereas the lysates in the immunoblots comprised the entire hippocampus, possibly masking subregion-specific differences.
We found that synaptic NMDAR currents were reduced in interneurons expressing PV or CCK, but not in pyramidal cells in the NRG1 tg-type-I mice. Furthermore, we show that not only is the NMDAR-mediated synaptic component reduced in comparison to the AMPAR currents in the CCK ϩ cells or PV ϩ cells, but a similar significant change is also seen in anatomically identified PV ϩ basket cells. The analyses of the quantal miniature currents in identified PV ϩ basket cells indicate that the reduced NMDAR-to AMPAR-mediated synaptic responses are due to smaller postsynaptic NMDAR currents, rather than increased AMPAR EPSCs. Finally, we show reduced NMDAR-dependent excitatory drive of recurrent GABAergic inhibition in the hippocampus of the NRG1 type I-overexpressing mice using optogenetically driven selective stimulation of hippocampal pyramidal cells.
Of note, in this transgenic mouse line, the overexpression of NRG1 type I is under the Thy-1.2 promoter, which is not equally expressed in all hippocampal pyramidal cells (Dobbins et al., 2018). This raises a possibility that NRG1 release in the hippocampus is not homogeneous, having variable effects on ErbB4-positive cells. This might at least partially explain the N/A ratio variation in PV ϩ cells of the NRG1-overexpressing mice illustrated in Fig. 2. However, the N/A ratio variation may also emerge from lack of the NRG1 receptor in some PV ϩ cells and CCK ϩ interneurons (Bean et al., 2014).
The results suggest that NMDAR-signaling abnormalities in these two major GABAergic interneuron populations may contribute to the hippocampal pathophysiology thought to occur in SZ (Harrison et al., 2003;Gonzalez-Burgos et al., 2011;Curley and Lewis, 2012). In this respect, our results bring together three theories of SZ pathophysiology; genetic heritability, inhibitory circuit dysfunction, and NMDAR hypofunction affecting GABAergic inhibitory interneurons such as PV ϩ basket cells (Zylberman et al., 1995;Lisman et al., 2008;Belforte et al., 2010;Korotkova et al., 2010;Gonzalez-Burgos and Lewis, 2012;Volk and Lewis, 2014;Banerjee et al., 2015;Krystal, 2015). Malfunction of PV ϩ basket cells has been commonly suggested to underlie aberrant coordinated network activities, in particular the gamma frequency oscillations, is associated with cognitive dysfunction in animal models , and is hypothesized to do so as well in SZ patients Uhlhaas and Singer, 2010;Gonzalez-Burgos and Lewis, 2012;Harrison et al., 2012;Marín, 2012;. Interestingly, the specific alterations of gamma oscillation features that were observed in hippocampal slice preparations from the NRG1 tg-type-I mice (Deakin et al., 2012) differed from findings of in vivo studies in which NMDARs were selectively knocked out in PV-expressing interneurons (Korotkova et al., 2010;Carlén et al., 2012). In fact, it has been proposed that NMDAR hypofunction in PV ϩ cells renders the brain networks more prone to exhibit the schizophrenia-associated behavioral and electrophysiological alterations, and that the actual phenotypes develop when NMDAR hypofunction simultaneously coexists in other neuron types (Bygrave et al., 2016). Importantly, our results here show postsynaptic suppression of the NMDAR signaling in interneurons expressing CCK. In physiologic conditions, these hippocampal interneurons have large synaptic NMDAR-mediated currents (Fricker and Miles, 2000;Maccaferri and Dingledine, 2002;Matta et al., 2013). Thus, it is likely that the alterations observed in the NRG1 tg-type-I mouse hippocampal network activity and hippocampus-dependent behavior (Deakin et al., 2009(Deakin et al., , 2012 emerge at least partially from NMDAR hypofunction in the PV ϩ and CCK ϩ interneuron subpopulations. Although we failed to detect changes in AMPAR-mediated glutamatergic currents or the function of GABAergic synapses, it is possible that these can be subject to changes at a later stage of the phenotype progression also in the NRG1 type I mutant mice Fazzari et al., 2010;Abe et al., 2011;Ting et al., 2011).
In summary, our results indicate that synaptic NMDARmediated signaling in hippocampal interneurons is sensitive to chronically elevated NRG1 type I levels. Further studies will be required to determine the mechanism by which NRG1 type I overexpression results in the observed NMDAR hypofunction, and to what extent these alterations are sufficient to explain the previously reported phenotypes in these mice (Michailov et al., 2004;Deakin et al., 2009Deakin et al., , 2012. Possible cellular mechanisms underlying the NMDAR hypofunction include altered receptor subunit phosphorylation (Hahn et al., 2006;Bjarnadottir et al., 2007;Pitcher et al., 2011;Banerjee et al., 2015) or modulation of the trafficking and expression of NMDAR subunits (Ozaki et al., 1997;Gu et al., 2005;Abe et al., 2011;Luo et al., 2014). Importantly, it has been shown that neuregulin 2 (NRG2), which also signals via ErbB4, facilitates the physical interaction of ErbB4 with the NMDAR GluN2B subunit, leading to internalization of the subunit and hence NMDAR hypofunction (Vullhorst et al., 2015). Finally, the changes in NMDAR-mediated synaptic transmission observed in transgenic NRG1 type I mice could in part mirror what takes place in SZ, given the elevated NRG1 type I expression seen in the brain in the disease. Further studies are needed to explore this possibility and the potential role of therapeutic interventions targeting the NRG1 signaling pathway.