Upregulated NMDAR-mediated GABAergic transmission underlies autistic-like deficits in Htr3a knockout mice.

Mutations in serotonin pathway genes, especially the serotonergic receptor subunit gene HTR3A, are associated with autism. However, the association of HTR3A deficiency with autism and the underlying mechanisms remain unknown.
Methods: The Htr3a knockout (KO) mice were generated using transcription activator-like effector nuclease technology. Various behavior tests, including social interaction, social approach task, olfactory habituation/dishabituation, self-grooming, novel object recognition, contextual fear conditioning, elevated plus maze, open field and seizure susceptibility, were performed to assess the phenotypes. Transcriptome sequencing was carried out to search for molecular network and pathways underlying the phenotypes. Electrophysiological recordings, immunoblotting, immunofluorescence staining, immunoprecipitation, and quantitative real-time PCR were performed to verify the potential mechanisms. The N-methyl-D-aspartate receptor (NMDAR) antagonist memantine was used to treat the KO mice for rescuing the phenotypes.
Results: The Htr3a KO mouse model showed three phenotypic domains: autistic-like behaviors (including impaired social behavior, cognitive deficits, and increased repetitive self-grooming), impaired memory, and attenuated susceptibility to pentylenetetrazol-induced seizures. We observed enhanced action potential-driven γ-aminobutyric acid-ergic (GABAergic) transmission in pyramidal neurons and decreased excitatory/inhibitory (E/I) ratio using the patch-clamp recording. Transcriptome sequencing on the hippocampus revealed the converged pathways of the dysregulated molecular networks underlying three phenotypic domains with upregulation of NMDAR. We speculated that Htr3a KO promotes an increase in GABA release through NMDAR upregulation. The electrophysiological recordings on hippocampal parvalbumin-positive (PV+) interneuron revealed increased NMDAR current and NMDAR-dependent excitability. The NMDAR antagonist memantine could rescue GABAergic transmission in the hippocampus and ameliorate autistic-like behaviors of the KO mice.
Conclusion: Our data indicated that upregulation of the NMDAR in PV+ interneurons may play a critical role in regulating GABAergic input to pyramidal neurons and maybe involve in the pathogenesis of autism associated with HTR3A deficiency. Therefore, we suggest that the NMDAR system could be considered potential therapeutic target for autism. © The author(s).

Introduction

Autism spectrum disorder (ASD), prevalent in 0.75% to 1.1% of the population, is a class of neurodevelopmental disorders with core features of impaired social interaction and communication difficulties, along with stereotyped behaviors and restricted interests 1, 2. Apart from these core features, most of individuals with ASD show a lower level of intelligence quotient (IQ) 3 and comorbidity with epilepsy (EP) 2.

The serotonergic, γ-aminobutyric acid-ergic (GABAergic) and glutamatergic systems are reported to be involved in ASD 4, 5. Multiple lines of evidence indicate that ASD is associated with mutations in many genes that affect the ratio between neuronal excitation and inhibition 6-9. For instance, mutations underlying tuberous sclerosis 10, fragile X syndrome 11 and Angelman syndrome 9, targeting proteins critical for synaptic functions, have been associated with unbalanced excitation and inhibition. Both upregulation 12 and downregulation 13 in the ratio of excitation to inhibition were observed in the brains of autistic mouse models. The serotonergic system is shown to play a critical role in maintaining the balance of excitatory/inhibitory (E/I) transmission to keep proper functions of neuronal networks in the brain 14-17.

Human brain development undergoes high serotonin synthesis during childhood, and dysregulation of the developmental process is reported in autistic children 18, 19. Elevation of serotonin levels in the whole blood and platelets is detected in nearly 30% of individuals with ASD 20, 21. The serotonin transporter gene (Sert) knockout mice show hyperserotonemia and autistic-like behaviors 21. The serotonin transporter gene SLC6A4 has also been implicated in ASD 22, 23. The TPH2 gene, which encodes the rate-limiting enzymes that control serotonin biosynthesis and is associated with autism 24-26. Mutations within genes in the serotonin pathways, especially the serotonergic receptor subunit gene HTR3A, have been reported to associate with autism 27, 28.

The 5-HT3 receptor consists of the five subunits 5-HT3A-E, of which the subunits 5-HT3A-C are expressed in the brain 29. The 5-HT3A subunit is essential for all the functional 5-HT3 receptors, mainly expressed in the interneurons of the cerebral limbic system, including the hippocampus, cortex, and amygdala 30, 31. Previous studies have reported impaired social behaviors 32, decreased anxiety 33, and impaired fear memory extinction 34 in Htr3a knockout mice. These studies on Htr3a knockout mouse models focused on different behavior aspects. Since previous studies indicated that HTR3A might be involved in human autism, a systematic evaluation of autistic-like behaviors of mice with the knockout of this gene is needed, and elucidation of the underlying molecular pathways would shed light on the pathogenesis of autism involving HTR3A mutations.

Here, we generated Htr3a knockout (KO) mice, displaying autistic-like behaviors, impaired learning/memory (LM) and attenuated susceptibility to seizures. Electrophysiological recordings revealed a decreased excitatory/inhibitory (E/I) ratio caused by the enhancement of action-potential-driven GABAergic input to pyramidal neurons. We carried out transcriptome sequencing on the hippocampus and systematically searched for molecular pathways by integrating the transcriptomics and protein interaction data. The converged pathways associated with different behavioral phenotypes indicated upregulation of N-methyl-D-aspartate receptor (NMDAR) system in the mutants. Immunostaining of GluN2B and parvalbumin revealed an upregulation of NMDAR in the parvalbumin (PV+) positive cells. The electrophysiological recording showed enhanced NMDAR current and excitability in PV+ interneurons, and the NMDAR antagonist D-APV reduced their excitability. Intraperitoneal (i.p.) injection with memantine (an NMDAR inhibitor) reversed the abnormal behaviors and spontaneous inhibitory postsynaptic currents (sIPSCs) of the Htr3a KO mice. These results suggested that NMDAR upregulation in PV+ interneurons increased excitability and GABAergic output, possibly contributing to the imbalance of excitation and inhibition and leading to autistic-like behaviors.

Materials and Methods

Animals

The generation of Htr3a KO C57BL/6N mice by transcription activator-like effector nuclease (TALEN) technology were generated by Cyagen Biosciences (China). Briefly, 11 bases (GGGGAAGgtaa) in exon 1 of the mouse Htr3a gene (Ensemble ID: ENSMUSG00000032269; Gene ID: 15561) were chosen as TALEN target sites. The 11-base deletion across the junction of the first exon and intron disrupted the reading frame after the 16th codon in all the isoforms based on the UCSC genome browser (https://genome.ucsc.edu/). TALEN mRNAs transcribed in vitro and injected into fertilized eggs for KO mouse production. The genomic region spanning the deletion was PCR-amplified using forward primer 5'-AGTTGGAAAAGCAGTCTGCCTGG-3' and reverse primer 5'-TTGACCCCACACCTCAGAATCCT-3', followed by Sanger sequencing for genotyping. Age-matched (8-12 weeks) male or female mice and the wild-type (WT) littermates of the KO mice were used for behavioral tests, and male mice were used for all other tests.

Behavior tests

Procedures for behavior tests are described in Supplementary Information, including the animal housing and handling, home cage social interaction test, social approach task, olfactory habituation/dishabituation test, self-grooming test, novel object recognition, contextual fear conditioning, elevated plus maze test, open field test, seizure susceptibility test and memantine rescue experiments. All experimental procedures on animals followed the guidelines and recommendations of the Institutional Animal Care and Use Committee (IACUC) and were approved by the Southern Medical University Experimental Animal Ethics Committee.

RNA-seq and differential expression analysis

Each RNA sample was extracted from the hippocampi of adult mice according to the manufacturer's protocol (RNAeasy Mini Kit, Qiagen, USA). The quality and yield of the isolated RNAs were assessed using a NanoDrop Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) and Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). Differentially expressed genes (DEGs) were the intersection part of DESeq2 (adj.p < 0.05) and edgeR (p < 0.01).

Immunoblotting and immunoprecipitation

The hippocampal tissue from 8-week mice was lysed in lysis buffer with 1 mM PMSF and centrifugated at 14,000 g for 15 min. For immunoprecipitation, the supernatant lysate was incubated with rabbit anti-GluN1 (1:50, 5704S, Cell Signaling Technology, USA) or IgG antibody (1:500, 2729S, Cell Signaling Technology, USA) at 4 °C overnight. Then the mixture was incubated with Protein A/G plus-Agarose (SC-2003, Santa Cruz, USA) for 3 h at 4 °C. After centrifugation at 1000 g for 5 min at 4 °C, the immunoprecipitates were washed three times with lysis buffer, and then boiled in protein loading buffer for 5 min. Western blotting was performed to detect the proteins using the following primary antibodies at 1:1000 dilution: rabbit anti-GluN1 (5704S, Cell Signaling Technology, USA), rabbit anti-GluN2A (ab124913, Abcam, UK), rabbit anti-GluN2B (14544S, Cell Signaling Technology, USA), rabbit anti-PSEN1 (6543S, Cell Signaling Technology, USA), anti-AMPAR1 (ab109450, Abcam, UK), rabbit anti-AMPAR2 (ab206293, Abcam, UK).

Immunostaining

Mice were anesthetized with isoflurane and then perfused with ice-cold saline, followed by 4% paraformaldehyde (PFA) in 0.1 M phosphate-buffered saline (PBS), pH 7.4. After post-fixation overnight in 4% PFA at 4 °C, brains were transferred to 30% sucrose in 0.1 M PBS, pH 7.4. Coronal sections (40 µm) of the hippocampus were obtained on a cryostat (Leica CM3050 S) and rinsed in 1% TritonX-100 in PBS. The hippocampus sections were incubated in blocking buffer (containing 1% TritonX-100, 1% bovine serum albumin and 10% goat serum in PBS) for 2 h and then with following primary antibodies in blocking buffer for 48 h at 4 °C: rabbit anti-GluN2B (ab65783, Abcam, UK; 1:200) and mouse anti-PV (PV 235, Swant, Switzerland; 1:3,000). After washing three times, the sections were incubated with appropriate secondary antibodies Alexa Flour 594-conjugated anti-mouse IgG (1:1,000; Thermo Fisher, USA) or Alexa Fluor 488-conjugated anti-rabbit IgG (1:1,000; Thermo Fisher, USA) for 2 h. After three more washes in PBS, sections were mounted with the Pro-Long anti-fade medium (Invitrogen). Fluorescent images were collected using a confocal microscope (Nikon A1) and analyzed using the Image J software.

qRT-PCR

Total mRNAs from hippocampal tissues were extracted using standard column purification according to the manufacturer's protocol (RNAeasy Mini Kit, Qiagen, USA), and reverse-transcribed into cDNAs using the PrimeScriptTM RT reagent Kit following the manufacturer's protocol (Takara, code No. RR037A). The qRT-PCRs were carried out using Applied Biosystems ABI 7500. The housekeeping gene Gapdh was used as the internal control. The relative expression levels of the genes were calculated using the 2-∆∆CT method. The primers used for mRNA quantification are listed in Supplementary Information.

Slice preparation

Mice were anesthetized with isoflurane and then decapitated. The brain was quickly isolated and transferred to ice-cold oxygenated artificial cerebrospinal fluid (ACSF) containing 220 mM sucrose, 2.5 mM KCl, 1.3 mM CaCl2, 2.5 mM MgSO4, 1 mM NaH2PO4, 26 mM NaHCO3, and 10 mM glucose. The coronal hippocampal slices (300 μm) were prepared using VT-1200S vibratome (Leica, Germany). The hippocampal slices were incubated in ACSF containing 126 mM NaCl, 3 mM KCl, 1.25 mM NaH2PO4, 1.0 mM MgSO4, 2.0 mM CaCl2, 26 mM NaHCO3, and 10 mM Glucose at 34 °C for 30 min. Then slices were put at room temperature (25 ± 1 °C) for 2 to 8 h. All extracellular solutions were constantly carbonated (95% O2, 5% CO2).

Electrophysiology

Whole-cell patch-clamp recordings of the postsynaptic currents and excitability on hippocampal neurons were carried out as previously described 35. During the procedure, the recording chamber was continuously perfused with ACSF (2 mL/min) saturated with 95% O2/5% CO2 at 32-34 °C. The postsynaptic currents and excitability of the hippocampal neurons were recorded using MultiClamp 700B amplifier and 1440A digitizer (Molecular Devices) under IR-DIC visualization (Zeiss, Axioskop 2). The glass pipettes (resistance of 3-6 MΩ) were pulled with a micropipette puller (P-97, Sutter Instrument). Spontaneous excitatory postsynaptic currents (sEPSCs) were recorded at the reversal potential of GABAA receptors (-70 mV) in the presence of bicuculline (20 μM). Glass pipettes were filled with the intracellular solution containing 130 mM potassium gluconate, 20 mM KCl, 10 mM HEPES, 4 mM Mg-ATP, 0.3 mM Na-GTP, 10 mM disodium phosphocreatine, and 0.2 mM EGTA, pH 7.2, 295 mOsm. For the neuronal excitability recording, the pipette solution was the same as sEPSCs recording. Neurons were held at -70 mV, and the spike discharges were induced by injection of hyperpolarizing and depolarizing current steps. Spontaneous inhibitory postsynaptic currents (sIPSCs) were recorded at the reversal potential of ionotropic glutamate receptors at a holding potential of 0 mV. Glass pipettes were filled with the intracellular solution containing 110 mM Cs2SO4, 0.5 mM CaCl2, 2 mM MgCl2, 5 mM EGTA, 5 mM HEPES, 5 mM tetraethylammonium, and 5 mM ATP-Mg, pH 7.2, 292 mOsm. For miniature postsynaptic current recording, tetrodotoxin (1 μM) was included in the perfusion solution. For evoking EPSCs, the stimulating electrode was placed in the DG fiber path, approximately 0.2 mm away from the recorded cell bodies in the CA1. The internal solution contained 115 mM CsMeSO4, 20 mM CsCl, 10 mM HEPES, 2.5 mM MgCl2, 10 mM sodium phosphocreatine, 5 mM QX-314, 4 mM Na-ATP, 0.4 mM Na-GTP, and 0.6 mM EGTA, pH 7.3, 285 mOsm. AMPAR-mediated currents were recorded at -70 mV, and NMDAR-mediated currents were voltage-clamped at +40 mV and quantified by measuring the amplitude current at 50 ms after stimulation. The E/I ratio was calculated by recording sEPSCs and sIPSCs on the same cell at -70 mV and 0 mV, respectively, with the internal solution containing: 130 mM CsMeSO4, 10 mM NaCl, 10 mM EGTA, 4 mM Mg-ATP, 0.3 mM Na-GTP, 10 mM HEPES, 290 mOsm, adjusted to pH 7.4 with CsOH. The E/I ratio was calculated as the frequency of sEPSC divided by the frequency of sIPSC per cell. Currents were filtered at 3 kHz with a low-pass filter, and data were digitized at 10 kHz and acquired using the pCLAMP 10 software. Series resistance was continuously monitored for each neuron. The neuron recordings were discarded if the variation of series resistance was over 20%. Data analyses were conducted by the Mini Analysis software (Synaptosoft).

Statistical analysis

Statistical comparisons were conducted in GraphPad Prism 8. The homoscedasticity and normality of the distributions of data were determined using GraphPad Prism 8 before assigning specific statistical tests. Where normality and equal variance between samples groups were achieved, Student's t test, one-way analysis of variance (ANOVA) or two-way (ANOVA) were used. Where normality or equal variance of samples failed, Mann Whitney test or Kruskal-Wallis one-way ANOVA was performed. Most data were presented using box and scatterplots depicting the median, 5%-95% confidence intervals and individual values. Some data were expressed as the means ± standard errors of the means (SEM). Significance was conventionally set as **** p < 0.0001, ***p < 0.001, **p < 0.01 and *p < 0.05.

Results

Autistic-like behaviors, impaired memory and attenuated susceptibility to seizures in Htr3a KO mice

We generated an 11-base (GGGGAAGgtaa) frame-shift deletion in the first exon of Htr3a in C57BL/6N mice using the TALEN technology (), disrupting the reading frame after the 16th codon in all the isoforms based on the UCSC genome browser (https://genome.ucsc.edu/). The ablation of Htr3a transcripts in the brains of knock out mice was determined by performing qRT-PCR on total RNA from the hippocampus. The forward primer was located at the deletion site in the first exon (3' end in the deletion region) and the reverse prime was in the common region of the isoform transcripts. The Htr3a transcripts were present in the hippocampus from WT, but were not detected in the tissue from Htr3a mice (Student's t test, p = 0.0014) ().

In the three-chamber social approach task test, both Htr3a and WT mice had a significant preference for sniffing or exploring a cage containing a stranger mouse (stranger 1) rather than an empty cage (For male mice, Mann Whitney test, p = 0.5254, Figure ; for female mice, Student's t test, p = 0.2426, ). However, compared with the WT controls, Htr3a mice showed less preference for sniffing or exploring the stranger 2 mouse rather than the familiar stranger 1 mouse (For male mice, Student's t test, p = 0.0448, Figure ; for female mice, Student's t test, p = 0.0198, ). In the home cage social interaction test, Htr3a mice spent shorter time engaged in exploring a stranger mouse, an active social interaction, than WT littermates (For male mice, Student's t test, p = 0.0125, Figure ; for female mice, Student's t test, p = 0.0116, ). In the olfactory habituation/dishabituation test, both WT and Htr3a mice demonstrated normal ability to habituate or dishabituate to odors. However, the Htr3amice spent less time sniffing the social odors than WT mice (For male mice, two-way ANOVA, Bonferroni's multiple comparisons test, for cage 1 (1st), p < 0.0001, for cage 2 (1st) p = 0.0023, Figure ; for female mice, two-way ANOVA, Bonferroni's multiple comparisons test, for cage 1 (1st), p < 0.0001, for cage 2 (1st) p < 0.0001, ). These observations indicated that the Htr3amice may have less interest in social odors. The male Htr3a mice displayed longer total duration time in repetitive self-grooming (Mann Whitney test, p = 0.0047, Figure ), while the female Htr3a mice displayed normal self-grooming duration (Mann Whitney test, p = 0.1128, ). In the elevated plus-maze test, we found no difference in the time spent in the open arms between the WT and Htr3a mice (For male mice, Student's t test, p = 0.8526, for female mice, Student's t test, p = 0.8742, ). In the open field test, the Htr3a mice traveled a similar total distance compared with the WT controls (For male mice, Student's t test, p = 0.6452, ; for female mice, Student's t test, p = 0.4041, ), suggesting that the Htr3a mice had normal locomotion.

The novel object recognition test was performed to assess the cognitive function. During the familiarization session, both Htr3a and WT mice showed no preference for the object (For male mice, Student's t test, p = 0.9353, ; for female mice, Student's t test, p = 0.4579, ). During the test session, the Htr3a mice spent less time sniffing a novel object than the familiar one (For male mice, Student's t test, p = 0.0122, Figure ; for female mice, Student's t test, p = 0.0060, ), suggesting that the Htr3a mice had impaired recognition memory. In the contextual fear conditioning test, the mutant mice showed a significantly reduced fear memory compared to the WT controls (For male mice, Student's t test, p = 0.0269, Figure ; for female mice, Student's t test, p = 0.0035, ).

Since EP shows comorbidity with ASD 2, we assessed the susceptibility to seizures of the mutant mice by injecting of pentylenetetrazol (PTZ) at a dose of 60 mg/kg body weight. Of the 20 WT mice, 11 (55%) displayed clonus, 7 (35%) displayed clonic seizure. Of the 19 the Htr3a mice, 9 (47.37%) displayed clonus and 2 (10.53%) displayed clonic seizure, demonstrating that the Htr3a mice could have significantly less severity of PTZ-induced seizure than WT mice (Mann Whitney test, p = 0.0115) (Figure ). However, no difference was observed between the mutant and WT female mice in the PTZ-induced seizure test (Mann Whitney test, p = 0.2445) ().

Enhanced action potential-dependent GABAergic transmission

The hippocampus in human brain is crucial for memory 36, associated with autism 37, 38 and EP 39, suggesting that it may play an essential role in the behavioral phenotypes of the Htr3a mice. The serotonergic system is involved in balancing E/I transmission 14-17. Therefore, we recorded the sEPSCs (Figure ) and sIPSCs (Figure ) on the pyramidal neurons in the hippocampal CA1 region by whole-cell voltage clamp. We observed no significant change in both frequency (Mann Whitney test, p = 0.7577) and average amplitude (Student's t test, p = 0.1108) of sEPSCs in the Htr3a mice (Figure ). There was a higher frequency (Student's t test, p = 0.0422) of sIPSCs in Htr3amice, but no significant changes in amplitude (Student's t test, p = 0.1866) (Figure ). To assess the E/I balance of the same cell, we recorded the sEPSCs and sIPSCs on the pyramidal neurons at the holding voltage -70 mV and 0 mV, respectively (). Similar results for sEPSCs and sIPSCs were observed (, ). The E/I ratio () between the frequency of sEPSCs and sIPSCs was decreased, suggesting an enhanced inhibitor transmission (Student's t test, p = 0.0328). We blocked neuronal firing with the presence of 1 mM tetrodotoxin (TTX) and recorded miniature postsynaptic currents (Figure ). We found that in pyramidal cells of the hippocampal slices both mEPSCs and mIPSCs were not changed in Htr3a KO mice compared with WT (For mEPSC frequency, Student's t test, p = 0.6440; for mEPSC amplitude, Student's t test, p = 0.8525; for mIPSC frequency, Mann Whitney test, p = 0.4234; for mIPSC amplitude, Student's t test, p = 0.2205) in pyramidal cells of the hippocampal slices. These results suggested that the enhancement of sIPSCs frequency was dependent on the elevated excitability of the input GABAergic neurons.

Differentially expressed genes in the hippocampus are enriched with genes involved in ASD, learning/memory and epilepsy

To further explore the underlying molecular mechanisms, we performed transcriptome profiling on 3 pools of hippocampi from 12 Htr3a KO mice and 3 pools from 12 control littermates. The samples of Htr3a-/- mice were well separated from those of the littermate WT mice in the principal component analysis (PCA) (). Out of 16,435 expressed genes (), 2,092 were identified as differentially expressed genes (DEGs) (FDR ≤ 0.05) (Figure ), including 1,010 up-regulated and 1,082 down-regulated genes (). The Gene Ontology (GO) (Figure , ) biological processes, known to be associated with E/I balance and the regulation of neuronal excitability, such as axon development, axonogenesis, regulation of synapse structure or activity, synapse organization, and regulation of membrane potential, were found to be enriched. Many enriched KEGG (Kyoto Encyclopedia of Genes and Genomes) pathways (Figure ; ) were also involved in the synaptic transmission and neuronal excitability, such as axon guidance, calcium signaling pathway, glutamatergic synapse, retrograde endocannabinoid signaling.

We constructed a mouse hippocampal interactome () by integrating the hippocampal expressed genes with protein interaction data (See Methods). We then mapped the 2,092 differentially expressed genes (DEGs) to the mouse hippocampal interactome to retrieve the DEG Network (245 nodes and 231 edges) (), including the DEGs and their first interacting neighbors with co-expression. The top major components of the DEG Network included four modules with distinct functions (named as ion channel regulation, neuronal transcription regulation, synaptic transmission and circadian rhythm regulation) (), which are known to be related to autism 40-43.

Both DEGs and the DEG Network were enriched with ASD-candidate genes, LM-related genes, and EP-candidate genes, suggesting that DEG Network consisted of signaling pathways underlying the autistic-like phenotypes, impaired memory and reduced susceptibility to seizure. Compared with all DEGs, the DEG Network exhibited further enrichment with ASD candidate genes () (two-tailed Fisher's exact test, ratio = 0.2041, p = 1.65E-08) (Figure ), EP candidate genes () (two-tailed Fisher's exact test, ratio = 0.0980, p = 4.34E-05) (Figure ), and LM-related genes () (two-tailed Fisher's exact test, ratio = 0.1469, p = 5.19E-08) (Figure ). The DEG Network was also enriched in signaling pathways involved in ASD, LM and EP (Figure , ), such as circadian rhythm 43, 44, cAMP pathway 45, 46, long-term potentiation 47, 48 and thyroid hormone signaling pathway 36, 49. Interestingly, many genes of the NMDAR system (Grin1, Grin2a, Grin2b, Grin2c) were involved in these pathways (Figure ).

Upregulation of NMDAR pathways underlying ASD, learning/memory and epilepsy

To further search for converged molecular pathways underlying ASD-, LM- and EP-related phenotypes, we mapped ASD candidate genes, LM-related genes or EP candidate genes to the mouse hippocampal interactome network respectively, and retrieved three subnetworks (ASD Network, EP Network and LM Network). The ASD Network (Figure , ), LM Network (Figure ) and EP Network (Figure ) included their candidate genes and their first co-expressed protein-protein interaction (PPI) neighbors. Of the genes in the ASD-, LM- and EP networks, 26.38% (62/235), 16.99% (35/206), 24.79% (30/121) were DEGs.

The upregulation of NMDAR subunit genes Grin2b and Grin1 were among the top hub genes in the ASD- (), EP- () and LM () Networks. Of the 143 enriched KEGG pathways of the ASD-, EP- and LM Networks ( ), 16 pathways contain NMDAR. Interestingly, 87.5% (14/16) of the NMDAR-containing pathways were shared by all three networks (-S14). The serotonergic synapse, glutamatergic synapse, GABAergic synapse pathways were shared by the ASD-, EP- and LM Networks. These results were consistent with previous findings that the serotonergic, GABAergic and glutamatergic systems are commonly involved in autism 4, 5.

The ASD, EP, and LM networks with 21 common genes in a shared network module included proteins encoded by Grin1, Grin2a, and Grin2b and their interactors (Figure ), suggesting the involvement of the common NMDAR system in the three phenotypic domains in Htr3a mice. We performed immunoprecipitation of the endogenous proteins in the mouse hippocampus to confirm GluN1-GluN2A, GluN1-GluN2B and GluN1-PSEN1 interactions (Figure ). The expression of NMDAR and AMPAR was quantified by qRT-PCR (Figure ) and Western blotting (Figure ) to verify the upregulated expression of the gene products. The expression of GluN1 and GluN2B subunits was higher in the hippocampus of the KO mice. Since the action potential-driven GABAergic input in pyramidal neurons was enhanced and glutamatergic and GABAergic synapse pathways were shared by ASD, EP, and LM networks (), we speculated that the elevated expression of NMDAR genes might upregulate the excitability of GABAergic neurons.

Increased NMDAR function in PV+ interneurons

Both cholecystokinin positive (CCK+)-pyramidal neuron inhibitory and CCK+-PV+-pyramidal neuron disinhibitory connections 50 control the GABAergic input of pyramidal neurons in the hippocampus. In the GABAergic synapse pathway shared by the ASD-, EP- and LM Networks, Cacna1a and Nsf were upregulated (). The Cacna1a gene encodes the P/Q calcium channel, which is specifically expressed in the PV+ interneuron and controls the GABA release in the hippocampus 50. NSF (N-ethylmaleimide sensitive factor), an ATPase associated with various cellular activities protein, is required for intracellular membrane fusion. Notably, it is reported that NSF is required for the NMDAR-potentiated inhibitory transmission 51. Since Cacna1a is uniquely expressed in PV+ interneurons, the upregulation of Cacna1a indicated that it might be involved in the increased GABA release in PV+ interneurons. We applied P/Q calcium channel antagonist, ω-agatoxin-TK, in the sIPSCs recording. ω-agatoxin-TK (0.25 μM) decreased both the frequency and amplitude of sIPSC, indicating that the P/Q calcium channel-dependent GABA release may be the source of enhanced GABAergic input in the Htr3a mice ().

Next, we immune-stained hippocampal slices for GluN2B and PV and observed GluN2B upregulation in CA1 PV+ interneurons in Htr3a mice (Mann Whitney test, p = 0.0005, Figure ). We crossed PV-Cre:Ai14 mice and Htr3a mice to mark PV+ cells in Htr3a mice, and recorded the NMDAR and AMPAR input-output curve on PV+ interneurons (Figure ). The NMDAR current of PV+ interneurons was larger in the PV-Cre:Ai14:Htr3a mice than the PV-Cre:Ai14 mice (two-way ANOVA, p < 0.0001), while the AMPAR current was unchanged (two-way ANOVA, p = 0.1782) (Figure ). We also found the NMDAR current to be larger in hippocampal pyramidal neurons in KO mice than the controls (two-way ANOVA, p < 0.0001), while the AMPAR current was normal in pyramidal neurons in the Htr3a mice (two-way ANOVA, p = 0.6975) ().

To determine whether the enhanced of NMDAR current regulated the excitability of PV+ interneurons, we evaluate the spike number of PV+ interneurons by the injection of depolarizing current steps (500 ms) to characterized the firing properties (Figure ). Compared to the controls, the excitability of hippocampal PV+ interneurons was increased in PV-Cre:Ai14:Htr3a mice (two-way ANOVA, p < 0.0001) (Figure ). These results were consistent with the enhanced presynaptic action potential-driven GABAergic input of pyramidal neurons in the Htr3a mice (Figure ). Moreover, the D-APV, an antagonist of NMDAR, decreased the firing frequency of PV+ interneurons in both WT (two-way ANOVA, p = 0.0152) and KO mice (two-way ANOVA, p = 0.0077) (Figure ). Most importantly, D-APV application reduced the excitability in the KO mice to a level similar to that of WT mice (two-way ANOVA, p = 0.5734), suggesting that the enhanced excitability of PV+ interneurons in the KO mice was dependent on the upregulation of NMDAR, and contributed to the enhanced inhibitory GABAergic transmission and decreased E/I ratio in pyramidal neurons.

Autistic-like deficits in the Htr3a-/- mice rescued by NMDAR blockade

To verify the upregulated NMDAR in autistic-like behaviors, memantine (5 mg/kg) was i.p. administrated to mice to inhibit NMDAR. The saline-treated Htr3a-/- mice displayed the same deficits as described above (Figure ). The injected memantine in Htr3a and WT mice did not affect their social ability (Figure ). In the social novelty phase, memantine-treated Htr3a mice spent significantly more time with the stranger 2 mouse than the familiar one (stranger 1) compared to saline-treated Htr3amice (two-way ANOVA, p = 0.0153, Figure ), to a level comparable to the wild type control (two-way ANOVA, p = 0.5541, Figure ). In the home-cage social interaction test, memantine could significantly increase the time in active interactions, including allo-grooming, following and mounting, in the Htr3a mice (two-way ANOVA, p = 0.0412, Figure ), while memantine-treated KO mice showed a similar level of active interactions as saline-treated WT mice (two-way ANOVA, p = 0.4847, Figure ). Acute memantine administration also significantly rescued repetitive self-grooming in the Htr3a mice (two-way ANOVA, p = 0.0007, Figure ), to a level similar to saline-treated WT mice (two-way ANOVA, p > 0.9999) (Figure ). We also observed that memantine administration could significantly increase the exploration time of the novel object in the Htr3a-/- mice (two-way ANOVA, p = 0.0371, Figure ), to a level comparable to wild type mice (two-way ANOVA, between saline-treated WT and memantine-treated KO mice, p = 0.5385, Figure ). In contextual fear conditioning test, we found that memantine treatment significantly increased fear memory of Htr3a-/- mice (two-way ANOVA, p = 0.0434, Figure ), to a level comparable to the WT control animals (two-way ANOVA, between saline-WT and memantine-treated KO mice, p = 0.6754, Figure ). Also, the Htr3a-/- mice treated with memantine showed increased PTZ-induced seizure scores (two-way ANOVA, between saline- and memantine-treated KO mice, p = 0.0022; between saline-WT and memantine-treated KO mice, p = 0.0629) (Figure ). However, memantine did not affect the behaviors of WT mice (Figure ). These results demonstrated that inhibition of NMDAR with memantine could rescue these behavioral phenotypes of the Htr3a-/- mice.

Since we observed increased frequency of sIPSCs in the Htr3a-/- mice (Figure ), suggesting enhanced inhibitory drive to pyramidal neurons, we further assessed memantine effects on the inhibitory input to pyramidal neurons in the Htr3a-/- mice. We used 1 μM memantine, which is within the therapeutic brain concentration range (~0.5-1 μM) 52. Memantine significantly reduced sIPSC frequency but not amplitude in the Htr3a-/- mice (paired t test, p = 0.0013 for frequency in KO mice bath with or without memantine; p = 0.1230 for amplitude in KO mice bath with or without memantine), while it did not affect sIPSCs in WT mice (Figure ). These results of memantine-rescue experiments, together with the findings on upregulation of NMDAR and enhanced excitability in the PV+ interneurons (Figure ), provided evidence that increasing the NMDAR signaling in PV+ interneurons enhanced GABAergic transmission and thereby caused the imbalance of E/I leading to the autistic-like behaviors.

Discussion

The Htr3amice displayed autistic-like behaviors (Figure , ), including less active social interactions, impaired social novelty, decreased social odor interest and increased repetitive selfgrooming. In a previous study, Htr3a knockout mice displayed impaired sociability but with normal social interactions 32, while the Htr3amice showed normal sociability (Figure ) but less active interactions in the home cage social interaction test (Figure ). Such discrepancy could be due to experimental settings: the previous study used new cages in the social interaction tests 32, while we used home cages to minimize environmental stress. The social interaction test, which allows direct contact with stranger animals within the home cages, may be more sensitive to the detection of social deficits in rodents 53, 54.

The olfactory habituation/dishabituation test suggested that the Htr3amice showed intact olfaction, but a decreased interest in the social stimulus odors (Figure ), which may be interpreted as decreased interest in social interactions. Apart from the core autistic behaviors, the Htr3a mice showed impaired memory (Figure ), consistent with the reduced IQ in most of the autistic children 3. The Htr3a mice also showed attenuated susceptibility to seizures (Figure ), which is the antithesis of the comorbidity between ASD and EP in humans 55. The potential mechanism of attenuated susceptibility to seizures in the Htr3a KO mice might be due to the decreased E/I ratio by NMDAR upregulation in the PV+ interneurons. Previous studies performed in ASD patients and mouse models indicated that the E/I imbalance toward hyperexcitability might contribute to ASD-EP comorbidity 55. Both increased 12 and decreased 13 the E/I ratio were observed in the brains of autistic mouse models, suggesting that the E/I imbalance, either decreased or increased E/I ratio, may underly autism 56.

The hippocampal DEGs are enriched with ASD candidate genes, LM-related genes and EP candidate genes. It is of note that the DEG Network showed further enrichment with these groups of genes (Figure ), suggesting that the DEGs are involved in the major pathways underlying the three phenotypes. The major components of the DEG Network included subnetworks enriched for functions such as ion channel regulation, transcription regulation, synaptic transmission and circadian rhythm regulation (module 1-4 in ). We searched the literature for every node in the synaptic transmission module (module 3) (), and found that 50% (14/28) of the genes were involved in ASD, 28.5% (8/28) of the genes are involved in EP, and 21.4% (6/28) of the genes are involved in both ASD and EP (). In this network module, Grin1, Grin2a and Grin2b, encoding for subunits of the NMDAR, were both hub and up-regulated genes. NMDAR are known to be the key regulators in the molecular pathways in learning and memory 57. Previous studies have suggested that GRIN1, GRIN2A, and GRIN2B were associated with autism 58 and epileptic disorders 59, 60. These results led us to further search for converged pathways involved in the three phenotypic domains.

The ASD Network, LM Network and EP Network contain up-regulated Grin1 and Grin2b, which are also hub genes of these networks (Figure ), suggesting that NMDAR-related pathways might be the common pathways underlying autistic-like behaviors, impaired learning/memory and attenuated susceptibility to seizure. These results are consistent with previous findings of the NMDAR system. Some mutations in GRIN1 were associated with intellectual disability, behavioral abnormalities, and stereotypical movements 61, while others in GRIN2A correlated with a spectrum of neurodevelopmental disorders with speech delay, apraxia and EP 62; similarly, some mutations in GRIN2B were associated with neurodevelopmental disorders characterized by mild to profound developmental delay/intellectual disability, often with EP and ASD behaviors 63.

Some functions (GO terms and KEGG pathways) associated with E/I balance and the regulation of action potential were enriched in DEGs and the DEG Network (Figure ). The GABAergic input to pyramidal neurons mainly comes from CCK+ and PV+ neurons in the hippocampus 64. We found upregulation of Cacna1a (encoding the P/Q-type calcium channel), which is one of key shared genes in the converged pathways (, ). It has been reported that the P/Q-type calcium channel is specifically expressed in PV+ neurons, and thus exclusively mediates GABA release from hippocampal PV+ interneurons 50. We found that the P/Q-type calcium channel antagonist, ω-agatoxin-TK, rescued enhanced sIPSCs in the KO mice, suggesting the key role of PV+ interneurons in the GABAergic input to pyramidal neurons. We observed increased NMDAR expression and enhanced NMDAR-dependent excitability in the hippocampal PV+ neurons in the Htr3a-/- mice (Figure ), consistent with the increased action potential-dependent GABAergic transmission detected in pyramidal neurons (Figure ). The NMDAR antagonist D-APV rescued the excitability of PV+ neurons and memantine rescued sIPSCs, indicating that the NMDAR system in PV+ neurons may play a critical role in GABAergic transmission and E/I balance. The NMDAR antagonist memantine rescued autistic-like features, impaired memory, attenuated seizure susceptibility in Htr3amice (Figure ), and that memantine suppressed the GABAergic transmission of pyramidal neurons in the Htr3amice (Figure ). These results also suggest that the NMDAR system can be targeted in designing therapy for autism. Since the NMDAR antagonist memantine is used in adjunct therapy for autism 65, 66, other proteins in the NMDAR pathways can be potential therapeutic targets.

Taken together, the Htr3a KO mice exhibited autistic-like behaviors, impaired learning/memory and attenuated PTZ-induced seizures, suggesting that HTR3A as a potential causal gene of autism. The transcriptome sequencing of the mouse hippocampus revealed a landscape of dysregulated genes with significant enrichment of ASD, EP candidate genes and LM related genes. Analysis of the converged molecular networks suggested a key role for NMDAR in PV+ interneurons in determining the phenotypes, which was further confirmed by the rescue experiments in the KO mice using the NMDAR antagonist memantine.

Supplementary methods, figures and table legends.

Click here for additional data file.

Supplementary tables.

Click here for additional data file.