The pathogenic factor of ZC4H2-associated rare disorder is a postsynaptic regulator for synaptic activity and cognitive function

Significance

Intellectual disability represents a core clinical manifestation of ZC4H2-associated rare disorder (ZARD), yet the underlying synaptic etiology remains elusive. In this study, with a conditional knockout mouse model depleting its pathogenic factor ZC4H2 in forebrain pyramidal neurons, we find that ZC4H2 is a postsynaptic regulator for AMPA receptors stability and excitatory synaptic activity, thereby playing essential roles in brain cognitive function. Notably, administration of perampanel, a clinical approved AMPA receptor-specific antagonist, successfully rescues synaptic dysregulation and cognitive malfunction of this disease model. These findings determine a physiological role of ZC4H2 in regulating excitatory synaptic activity and uncover synaptic dysfunction as a key mechanism driving ZARD-related intellectual disability, offering a targeted therapeutic avenue for cognitive impairment.

Abstract

The maintenance of excitatory synaptic activity is crucial for cognitive function and genetic mutations are responsible for the pathogenesis of related brain disorders. However, the roles of these pathogenic factors in synaptic dysregulation and cognitive malfunction are still poorly understood. In this study, a conditional knockout mouse model lacking ZC4H2—an X-linked gene implicated in ZC4H2-associated rare disorder (ZARD) —in forebrain excitatory neurons is generated and these mice exhibit cognitive malfunction, recapitulating the intellectual disability manifestation of ZARD. Mechanistically, ZC4H2 harbors a protein interaction network with key excitatory synaptic regulators and ZC4H2 interacts directly with AMPA receptors (AMPARs) and regulates their ubiquitination at the postsynaptic sites, thereby maintaining AMPARs protein stability and synaptic expression. ZC4H2 deficiency specifically and aberrantly increases AMPAR-mediated excitatory synaptic transmission and impairs synaptic plasticity of long-term potentiation. More importantly, pharmacological treatment with perampanel, an AMPAR-specific antagonist, successfully restores the excitatory synaptic activity and cognitive function of ZC4H2-deficient mice. Together, we establish that ZC4H2 is a postsynaptic regulator for AMPARs and excitatory synaptic activity and highlight that the dysregulation of these biological processes is a crucial etiology underlying ZARD-associated intellectual disability.

Synapses are fundamental units of neural circuits that transmit information and exert pivotal roles for brain physiological function. The postsynaptic membrane, a structurally specialized site, orchestrates the spatial organization of neurotransmitter receptors, receives presynaptic input, and converts chemical signals into electrical and biochemical responses to modulate postsynaptic neuronal activity. The postsynaptic density (PSD) constitutes a morphologically distinct and protein-rich subdomain composed of neurotransmitter receptors, adhesion molecules, abundant scaffolding proteins, cytoskeletal elements, and diverse signaling enzymes (1–3). Proteomic studies have identified over 1,000 proteins within mammalian excitatory PSD (4, 5), however, the functional roles of most in synaptic regulation and brain function remain uncharacterized.

Genetic studies have identified various pathogenic variants in genes encoding PSD proteins in individuals with neuropsychiatric and neurodevelopmental disorders. Mounting evidence has implicated synaptic dysfunction as a primary driver of these conditions (6, 7). Intellectual disability, a prevalent neurodevelopmental disorder affecting 1 to 3% of the population, is defined by substantial impairments in cognitive function and adaptive behavior. It presents with considerable genetic and clinical heterogeneity and is frequently comorbid with other neuropsychiatric disorders. Severe intellectual disability is often monogenetic, with over 2,500 pathogenic genes identified to date. Notably, more than half of these genes encode proteins present in the presynaptic or postsynaptic compartments, where they contribute to synaptic transmission and plasticity (8–10). Despite these advances, our understanding of the physiological functions of many intellectual disability-associated pathogenic factors is still limited, and further efforts are needed to clarify their roles in synaptic pathogenesis.

Fast excitatory synaptic transmission in the central nervous system is predominantly mediated by AMPA receptors (AMPARs), whose activity critically shapes synaptic strength and plasticity. Pathogenic mutations in AMPAR subunits, including GluA2 (11) and GluA3 (12), have been identified in individuals with intellectual disability, implicating disrupted AMPAR function in disease pathogenesis. Notably, bidirectional dysregulation of AMPAR-mediated synaptic activity has been linked to distinct intellectual disability-related genetic lesions. For example, mouse models deficient in TARP-γ2 and FRRS1l exhibit reduced AMPARs synaptic expression and diminished excitatory synaptic activity (13, 14). In contrast, models involving SynGAP (15, 16) or RNF220 (17) show aberrant enhancement of AMPAR-mediated synaptic transmission and impaired synaptic plasticity. Clinically both AMPARs positive allosteric modulators and antagonists have been developed for the treatment of neuropsychiatric disorders (18, 19). However, their application in addressing intellectual disability-related symptoms remains unclear, despite growing interest in AMPARs as viable translational targets.

ZC4H2-associated rare disorder (ZARD) is a rare and X-linked neurodevelopmental disorder characterized by a broad range of clinical symptoms involving both the central and peripheral nervous systems. Pathogenic mutations of ZC4H2, which encodes a C4H2-type zinc finger protein, are implicated in this disease (20–22). Previous studies, including our own, have demonstrated that ZC4H2 plays diverse and region-specific roles in neural development, contributing to the proliferation of cerebellar granule neuron progenitors (23), neurogenesis of GABAergic interneurons in the midbrain (24), development of V2 interneurons and motor neurons in the spinal cord (20, 24) and noradrenergic neurons in the locus coeruleus (25), and the spinal cord patterning (26–28). Although intellectual disability represents a core clinical manifestation of ZARD, the underlying etiology remains elusive. It has been reported that ZC4H2 has the capability to localize at the postsynaptic site of excitatory synapses and its pathogenic variants affect dendritic spine density in primary cultured neurons (20). However, whether ZC4H2 modulates synaptic activity and cognitive function has not yet been clarified.

In this study, we found that selective knockout of ZC4H2 in forebrain excitatory neurons impairs cognitive performance in mice, and these behavioral deficits are accompanied by increased excitatory synaptic transmission and disrupted long-term potentiation (LTP). Moreover, ZC4H2 harbors a protein interactome network with several key PSD proteins including AMPARs, and ZC4H2 depletion results in increased postsynaptic expression of GluA1 and GluA2 receptors. Of note, pharmacological inhibition of AMPARs using perampanel, a clinically approved noncompetitive antagonist of AMPARs, successfully restores the dysregulated synaptic activity and impaired cognitive function in ZC4H2-deficient mice.

Results

ZC4H2 Deficiency in Forebrain Excitatory Neurons Impairs Cognitive Function in Mice.

Given that intellectual disability is a core clinical symptom of ZARD patients (20), the contribution of ZC4H2 to cognitive function was examined in a targeted genetic model. Complete loss of ZC4H2, known to play a critical role in central nervous system development (23, 25, 28), leads to neonatal lethality in mice (28). As such, a conditional knockout strategy was employed in the current study. Specifically, ZC4H2^fl/fl and Emx1-Cre mice were crossed to achieve selective deletion of ZC4H2 in forebrain excitatory neurons. Male ZC4H2^wt/y;Emx1-Cre^+/− (Control) and ZC4H2^fl/y;Emx1-Cre^+/− (ZC4H2 cKO) mice were used for the subsequent analyses (Fig. 1A). At postnatal day 21 (P21), both ZC4H2 mRNA and protein levels were significantly reduced in cortical and hippocampal lysates from ZC4H2 cKO mice (SI Appendix, Fig. S1 A and B). Immunostaining analyses further confirmed robust depletion of ZC4H2 in neurons of both regions (SI Appendix, Fig. S1 C and D). Morphological assessments revealed no significant difference in body and brain sizes, cortical thickness, or dentate gyrus area between ZC4H2 cKO and Control mice (SI Appendix, Fig. S1 E and F). Hematoxylin and eosin (H&E) staining of coronal brain sections showed no gross structural abnormalities (SI Appendix, Fig. S1G), and cortical patterning appeared intact based on normal expression of layer 2/3 marker Satb2 and layer 5/6 marker Ctip2 (SI Appendix, Fig. S1H). Together, these results suggest that conditional loss of ZC4H2 in forebrain excitatory neurons does not impair gross brain development in mice, enabling investigation of its role in postnatal synaptic function and behavior.

Fig. 1.

ZC4H2 knockout in forebrain excitatory neurons impairs mice cognitive functions. (A) Schematic cartoon of the simplified crossing strategy for generation of ZC4H2^wt/y;Emx1-Cre^+/− (Control) and ZC4H2^fl/y;Emx1-Cre^+/− (ZC4H2 cKO) male mice. (B–I) Behavioral tests of open field (B), novel object recognition (C), three-chamber social test (D and E), Morris water maze (F and G), and contextual fear conditioning (H and I) for P60 Control (n = 15) and ZC4H2 cKO (n = 15) mice. (B) Bar graphs (mean ± SEM), overlaid with the actual data points, show the total travel distance, the average velocity, and the time spending in the center area. (C) Bar graphs (mean ± SEM), overlaid with the actual data points, show the total exploring time to old and novel objects, and the logarithm of discrimination index. (D and E) Bar graphs (mean ± SEM), overlaid with the actual data points, show the percentage of spending time close to inanimate ball and animated stranger mouse and the logarithm of social ratio (D), and the percentage of spending time close to new and familiar animated strangers and the logarithm of social ratio (E). (F and G) The escape latencies (mean ± SD) to find the platform throughout the 8-d learning trials (F), and spatial memory retrieval of these mice used in (F) was examined when the platform was removed, and bar graphs (mean ± SEM), overlaid with the actual data points, show the total movement distance, the time and distance in target quadrant, and the number of crossing target zone (G). (H and I) The percentage of freezing behavior (mean ± SD) during the fear conditioning period (H), and fear memory of these mice used in (H) was tested by exposure to the environment only, and bar graphs (mean ± SD), overlaid with the actual data points, show the percentage of freezing behavior at 30 min, 1 d, and 7 d after the contextual fear conditioning (I). *P < 0.05; **P < 0.01; ***P < 0.001, and two-tailed Student’s t test for (B–E, G, and I), and two-way ANOVA with Bonferroni correction analysis for (F and H). Str, stranger.

A series of behavioral tests were subsequently conducted to assess the impact of ZC4H2 deficiency on cognitive performance (Fig. 1). In the open field test, no significant differences were observed in total distance traveled, average velocity, or time spent in the center zone between Control and ZC4H2 cKO mice (Fig. 1B), suggesting that forebrain-specific loss of ZC4H2 did not affect general locomotor activity or anxiety-like behavior. However, in the novel object recognition test, which evaluates nonspatial memory under nonaversive conditions, ZC4H2 cKO mice displayed a significantly reduced preference for the novel object despite comparable total exploration time (Fig. 1C), suggesting impaired nonspatial memory. Social behavior was further assessed using the three-chamber paradigm. Both groups demonstrated normal sociability, showing greater interaction with an unfamiliar conspecific than with inanimate object (Fig. 1D). In contrast, social memory was selectively impaired in ZC4H2 cKO mice, which showed a significantly lower preference for a novel stranger over a familiar one compared to the control littermates (Fig. 1E).

Spatial learning and memory were evaluated using the Morris water maze, which showed that ZC4H2 cKO mice required more time to locate the hidden platform across days 5 to 8 of training (Fig. 1F), suggesting a higher escape latency and impairment in spatial learning. In the probe trial conducted 24 h after training, no obvious differences were found in the total swimming distance (Fig. 1G), suggesting preserved motor function. However, ZC4H2 cKO mice spent significantly less time in the target quadrant, a shorter distance within it, and made fewer crossings over the target zone compared with the control littermates (Fig. 1G), demonstrating deficits in spatial memory retention.

Finally, fear conditioning was used to test short- and long-term memory. Notably, ZC4H2 cKO mice showed attenuated freezing responses during the postshock period of trials 3 and 4 in the training session (Fig. 1H). In subsequent memory recall tests, freezing behavior remained significantly reduced in ZC4H2 cKO mice at both 30 min and 7 d postconditioning (Fig. 1I), suggesting impairments of both recent and remote memory.

Collectively, these findings demonstrate that ZC4H2 loss in forebrain excitatory neurons disrupts multiple cognitive domains, including recognition, social, spatial, and associative memory, consistent with an intellectual disability-like phenotype.

ZC4H2 Deficiency Dysregulates Excitatory Synaptic Transmission and Plasticity.

Given the strong association between synaptic activity dysfunction and intellectual disability (8, 9), the impact of ZC4H2 deficiency on synaptic activity was systematically investigated using the whole-cell patch-clamp recording on hippocampal CA1 pyramidal neurons from ZC4H2 cKO and Control mice (Fig. 2A). ZC4H2-deficient neurons exhibited a significant increase in both amplitude and frequency of miniature excitatory postsynaptic currents (mEPSCs) while decay kinetic remained unchanged (Fig. 2B). In contrast, miniature inhibitory postsynaptic currents (mIPSCs) showed no significant differences in amplitude, frequency, or decay time between ZC4H2 cKO and Control mice (Fig. 2C). These results suggest that ZC4H2 specifically regulates excitatory synaptic transmission.

Fig. 2.

ZC4H2 deficiency causes dysfunction of excitatory synaptic activity. Electrophysiological recording analyses of mEPSCs and mIPSCs parameters of hippocampal CA1 neurons from Control and ZC4H2 cKO mice. (A) Schematic cartoon of the experimental workflows. (B) Bar graphs (mean ± SEM) overlaid with the actual data points and cumulative distribution plots show the mEPSCs amplitude, frequency, and decay kinetics of Control (n = 12, black) and ZC4H2 cKO (n = 12, green) neurons. The scale bars for the representative mEPSC sample traces are 20 pA/500 ms. (C) Bar graphs (mean ± SEM) overlaid with the actual data points and cumulative distribution plots show the mIPSCs amplitude, frequency, and decay kinetics of Control (n = 11, black) and ZC4H2 cKO (n = 12, green) neurons. The scale bars for the representative mIPSC sample traces are 30 pA/500 ms. Cumulative distribution functions show no irregularities. (D–G) Simultaneous electrophysiological recording analyses of AMPAR and NMDAR-mediated eEPSCs on hippocampal CA1 neurons. (D) Schematic cartoon of the experimental workflows of dual whole-cell recording analyzing AMPAR and NMDAR-mediated eEPSCs. (E and G) Open and filled circles represent amplitudes for single pairs and mean ± SEM, respectively, and insets show sample current traces from Control (black) and ZC4H2 cKO (green) neurons. Bar graphs (mean ± SEM), overlaid with the actual data points, show the normalized AMPAR (E, n = 12) and NMDAR (G, n = 11) eEPSC amplitudes presented in scatter plots. (Scale bars, 50 pA/20 ms.) (F) Bar graph (mean ± SEM), overlaid with the actual data points, shows the paired-pulse ratios from Control (n = 12) and ZC4H2 cKO neurons (n = 12). (H and I) Whole-cell electrophysiological recording analyses of paired-LTP on hippocampal CA1 neurons. (H) Schematic cartoon of the experimental workflows. (I) Chart line graph (mean ± SEM) shows the percentage of respective baseline of Control (black, n = 8) and ZC4H2 cKO (green, n = 9) neurons before LTP induction, and bar graph (mean ± SEM), overlaid with the actual data points, shows the normalized eEPSCs amplitude at 45 min after LTP induction. Sample traces show EPSCs before (1) and 45 min after LTP (2). The scale bars are 25 pA/25 ms. (J) Transmission electron microscopy analyses of PSD thickness and width from the hippocampal CA1 areas of Control (black, n = 3) and ZC4H2 cKO (green, n = 3) mice, and violin plot graph overlaid with box and whisker plot of the actual data distribution. (Scale bar, 0.5 μm.) (K) Image analyses of the spine density and neuronal morphology of hippocampal CA1 pyramidal neurons from Control (black, n = 4) and ZC4H2 cKO (green, n = 4) mice. Schematic cartoon of the experimental workflows of virus stereotactic injection (Left). Violin plot graphs overlaid with box and whisker plot of the actual data distribution show the spine density, the major dendrite length and the total intersection, and line chart graph (mean ± SEM) shows the number of intersections along the dendritic trees at all distances from the soma. [Scale bars, 50 μm and 5 μm (Enlarged Inset).] *P < 0.05; **P < 0.01; ***P < 0.001, and Mann–Whitney U test with Bonferroni correction analysis for (B, C, and F), two-tailed Wilcoxon signed-rank sum test for (E and G), and two-tailed Student’s t test for (I–K).

Next, an adeno-associated virus (AAV) expressing Cre recombinase and EGFP was stereotactically delivered into the hippocampal CA1 area of ZC4H2^fl/y mice at P0 to sparsely knockout ZC4H2. Dual whole-cell recordings were conducted approximately three weeks later to examine AMPAR- and NMDA receptor (NMDAR)-mediated evoked EPSCs (eEPSCs) from an EGFP-positive pyramidal neuron and an adjacent uninfected control (Fig. 2D). AMPAR-mediated eEPSCs were significantly elevated in ZC4H2-deficient neurons (Fig. 2E), while NMDAR-mediated responses remained unchanged (Fig. 2G), suggesting that ZC4H2 specifically regulates AMPAR-mediate excitatory synaptic transmission. The paired-pulse ratio was found unaffected in ZC4H2 cKO mice (Fig. 2F), supporting a postsynaptic regulatory mechanism. To further assess synaptic plasticity, LTP was induced in CA1 pyramidal neurons from acute hippocampal slices (Fig. 2H), and it was found that ZC4H2 cKO mice displayed a marked reduction in the potentiation of AMPAR-mediated eEPSCs (Fig. 2I), indicating impaired synaptic plasticity of LTP. These findings collectively demonstrate that ZC4H2 is required for proper AMPAR-dependent synaptic transmission and plasticity.

PSD ultrastructure of hippocampal neurons was further examined via transmission electron microscopy. No significant differences were observed in either PSD thickness or the PSD length between ZC4H2 cKO and Control mice (Fig. 2J). However, morphological analysis of hippocampal CA1 pyramidal neurons following sparse labeling with AAV-hSyn-Flp:FDIO-EGFP revealed notable structural changes. ZC4H2-deficient neurons exhibited increased dendritic spine density, enhanced dendritic branching, and elongated apical dendrites (Fig. 2K), implicating ZC4H2 in the regulation of neuronal architecture and excitatory synaptogenesis.

Identifying the ZC4H2 Protein Interactome in Hippocampal CA1 Neurons by TurboID-Based Proximity Labeling.

Although ZC4H2 has been preliminarily found to localize at PSD (20), its molecular function in mature neurons remains largely unknown, To elucidate the regulatory network associated with ZC4H2 in excitatory synapse, TurboID-based proximity labeling combined with mass spectrometry was employed to profile its protein interactome in vivo. AAV-DJ-hSyn-TurboID or AAV-DJ-hSyn-ZC4H2-TurboID expression virus was stereotactically injected into the hippocampal CA1 area of adult C57BL/6 mice. Three weeks postinjection, biotin was administered subcutaneously for 7 consecutive days to enable biotinylation of ZC4H2-proximal proteins. Labeled proteins were isolated from hippocampal lysates using magnetic streptavidin beads and identified by quantitative liquid chromatography–tandem mass spectrometry (Fig. 3A). Western blot analysis confirmed that the expression level of ZC4H2-TurboID was lower than the endogenous ZC4H2 and verified its enrichment at postsynaptic sites (SI Appendix, Fig. S2A). Principal component analysis (PCA) of three biological replicates revealed clear segregation between ZC4H2-TurboID and TurboID-only samples (SI Appendix, Fig. S2B). Proteins present in at least two replicates were retained and statistical cut-offs (P-value < 0.01 and fold change > 2) were applied to define the potential ZC4H2 proximal interactome. This approach yielded 256 proteins that were either uniquely present or significantly enriched in ZC4H2-TurboID samples (Dataset S1). Among them, RNF220, a previously validated ZC4H2-interacting protein (28), was only found in the experimental group (Fig. 3B and Dataset S1), confirming both the specificity and reliability of the strategy. Gene ontology (GO) analysis revealed strong enrichment for components of the PSD and glutamatergic synapses (Fig. 3C) and biological processes related to excitatory synaptic transmission, learning, and memory (Fig. 3D). Notably, ZC4H2-proximal proteins include multiple key synaptic regulators, such as scaffolding proteins Shank2 and Homer3, glutamate receptor subunits GluA1, GluA2, GluA3, GluN1, GluN2A, and GluK2, and AMPAR auxiliary subunits TARP-γ2, CKAMP44, and CKAMP52 (Fig. 3B and Dataset S1). Moreover, overexpression of ZC4H2-TurboID did not alter the endogenous expression levels of AMPARs, NMDARs, or PSD marker PSD-95 (SI Appendix, Fig. S2C), suggesting that ZC4H2 overexpression may have a limited effect on neuronal functions. To further examine functional associations, protein–protein interaction networks were constructed using the STRING database (29). The resulting interactome clustered into four functional modules, including glutamatergic synapse, RNA binding and metabolism, ribosome and protein synthesis, and cytoskeleton organization and kinase (Fig. 3E), suggesting that ZC4H2 might be involved in diverse biological processes to exert its regulatory functions in mature neurons.

Fig. 3.

TurboID-based proximity labeling analyses of the protein interactome of ZC4H2. (A) Schematic cartoon of the experimental workflows of TurboID-based proximity labeling and LC–MS/MS analyses. (B) The proteins significantly enriched in ZC4H2-TurboID samples (n = 3) compared to TurboID samples (n = 3) with the statistical cut-offs of P-value < 0.01 and fold change > 2. (C and D) GO enrichment analyses of the enriched proteins identified in (B). (E) STRING analysis and functional clustering of the ZC4H2 proximal proteins identified in (B).

ZC4H2 Directly Interacts with AMPARs and Regulates Their Polyubiquitination and Stability.

To investigate whether ZC4H2 influences the composition of postsynaptic protein complex, PSD fractions were isolated from the cortex and hippocampus of ZC4H2 cKO and Control mice, and the expression levels of synaptic proteins identified in the ZC4H2 interactome were examined. Interestingly, the protein levels of GluA1 and GluA2 receptors were markedly elevated in both cortical and hippocampal PSD fractions of ZC4H2 cKO mice, whereas the expression levels of AMPAR auxiliary subunits and other associated proteins remained unchanged (Fig. 4A and SI Appendix, Fig. S3 A and C). As previous work has implicated ZC4H2 in regulating the ubiquitin-dependent depredation of interacting proteins (27, 28), the increases in GluA1 and GluA2 expression suggest possible disruption in AMPARs turnover. Immunoprecipitation of GluA1 or GluA2 from cortical and hippocampal PSD fractions revealed a marked reduction in polyubiquitination in ZC4H2-deficient mice, accompanied by a significant accumulation of both receptors in input and precipitated factions (Fig. 4B and SI Appendix, Fig. S3 B and D). To directly assess the effect of ZC4H2 on AMPARs turnover, cycloheximide chase assays were performed in N2A cells. CRISPR/Cas9-mediated depletion of ZC4H2 significantly prolonged the half-life of both GluA1 and GluA2 (Fig. 4C), whereas ZC4H2 overexpression accelerated their degradation (Fig. 4D). Although TurboID-based proximity labeling may capture indirect associations within protein complexes, coimmunoprecipitation experiments demonstrated that ZC4H2 physically binds GluA1 and GluA2. In HEK293 cells, Flag-tagged ZC4H2 coprecipitated with HA-tagged GluA1 and GluA2 (Fig. 4E), and reciprocal coimmunoprecipitation confirmed the interaction (Fig. 4F). Furthermore, endogenous ZC4H2 robustly coimmunoprecipitated with both GluA1 and GluA2 from cortical and hippocampal lysates, and vice versa (Fig. 4G and SI Appendix, Fig. S3E). Together, these findings indicate that ZC4H2 directly associates with AMPARs and regulates their polyubiquitination and stability, thereby modulating AMPARs expression at postsynaptic sites in mature neurons.

Fig. 4.

ZC4H2 is a direct interacting regulator of AMPARs for polyubiquitination and protein stability. (A) Western blot analyses of the expression levels of indicated synaptic proteins in the PSD fractions from the hippocampus of Control (n = 3) and ZC4H2 cKO (n = 3) mice. α-Tubulin was used as the internal control. (B) Polyubiquitination analyses of endogenous GluA1 and GluA2 protein in the hippocampus PSD fractions of Control (n = 3) and ZC4H2 cKO (n = 3) mice. (C) Cycloheximide chase analyses of protein half-lives of endogenous GluA1 and GluA2 in N2A of ZC4H2 knockout by CRIPSR/Cas9, and chart line graph (mean ± SD) show normalized levels against indicated proteins expression from the respective control without cycloheximide treatment (n = 4). (D) Cycloheximide chase analyses of protein half-lives of endogenous GluA1 and GluA2 in N2A overexpressing ZC4H2, and chart line graph (mean ± SD) show normalized levels against indicated proteins expression from the respective control without cycloheximide treatment (n = 4). (E and F) Coimmunoprecipitation analyses of ZC4H2 interaction with GluA1 or GluA2 with indicated HA or Flag antibody after overexpressed in HEK 293 cells. (G) Coimmunoprecipitation analyses of the interaction between endogenous ZC4H2 and GluA1 or GluA2 in the hippocampus with the indicated antibodies. CTL: control; cKO, conditional knockout; IB: immunoblot; IP: immunoprecipitation; WCL: whole cell lysate. *P < 0.05; **P < 0.01, and two-way ANOVA with Bonferroni correction analysis for (C and D).

AMPAR Antagonist Perampanel Reverses ZARD-Related Intellectual Disability Deficits.

The cognitive impairments observed in ZC4H2 cKO mice recapitulate key aspects of ZARD-associated intellectual disability, providing a tractable model for mechanistic and therapeutic investigation. Given that ZC4H2 loss leads to aberrant up-regulation of postsynaptic AMPARs expression and enhanced excitatory synaptic transmission, we hypothesized that excessive AMPARs activity may underlie the cognitive deficits. To test this possibility, perampanel—a clinically approved, noncompetitive AMPAR antagonist used in adolescent epilepsy treatment (30–34)—was selected for intervention. To first validate the efficacy and selectivity of perampanel, whole-cell recordings were performed in hippocampal acute slices from wild-type mice, and it was found that perampanel significantly reduced AMPAR-mediated eEPSC responses without affecting NMDAR-mediated responses (SI Appendix, Fig. S4 A and B). ZC4H2 cKO and Control mice were then administered perampanel or vehicle via daily intragastric gavage for seven consecutive days, followed by preparation of hippocampal acute slices for electrophysiological recordings analysis of mEPSCs (Fig. 5A). In control neurons, perampanel treatment induced a modest and significant reduction in mEPSC amplitude and frequency (Fig. 5B). Notably, in ZC4H2-deficient neurons, perampanel fully restored both amplitude and frequency of mEPSCs to control levels, without altering decay kinetics (Fig. 5B). Despite this robust functional rescue, western blot analysis showed that perampanel did not reduce the aberrantly increased expression levels of GluA1 and GluA2 in PSD fractions from the hippocampus of ZC4H2 cKO mice (SI Appendix, Fig. S4C). These findings demonstrate that pharmacological inhibition of AMPARs effectively restores synaptic transmission in ZC4H2-deficient neurons and support perampanel as a candidate therapeutic for ZARD-related cognitive dysfunction.

Fig. 5.

Perampanel restores the ID-related deficits of ZC4H2 cKO mice. (A and B) Electrophysiological recording analyses of mEPSCs parameters of hippocampal CA1 neurons from Control and ZC4H2 cKO mice treated with vehicle or perampanel. Schematic cartoon of the experimental workflows (A). Bar graphs (mean ± SEM), overlaid with the actual data points, and cumulative distribution plots show mEPSCs amplitude, frequency, and decay kinetics of Control-Vehicle (n = 12, black), Control-Perampanel (n = 16, brown), ZC4H2 cKO-Vehicle (n = 12, green), and ZC4H2 cKO-Perampanel (n = 12, blue) neurons (B). Cumulative distribution functions show no irregularities. (C–H) Behavioral tests of novel object recognition (D), three-chamber social test (E and F), and Morris water maze (G and H) for P60 mice of Control-Vehicle (n = 13), Control-Perampanel (n = 13), ZC4H2 cKO-Vehicle (n = 13), and ZC4H2 cKO-Perampanel (n = 13). Schematic cartoon of the experimental workflows (C). Bar graphs (mean ± SEM), overlaid with the actual data points, show the percentage of spending time close to new object and logarithm of discrimination index (D). Bar graphs (mean ± SEM), overlaid with the actual data points, show the percentage of spending time close to inanimate ball and animated stranger mouse and logarithm of social ratio, and the percentage of spending time close to new and familiar animated strangers and logarithm of social ratio (E and F). Escape latencies (mean ± SD) to find the platform throughout the 8-d learning trials (G), and spatial memory retrieval of these mice used in (G) was examined when the platform was removed, and bar graphs (mean ± SD), overlaid with the actual data points, show total movement distance, time, and distance in target quadrant, and number of crossing target zone (H). Str, stranger. (I) Whole-cell electrophysiological recording analyses of paired-LTP on hippocampal CA1 neurons. Chart line graph (mean ± SEM) shows the percentage of respective baseline of Control-Vehicle (black, n = 9), Control-Perampanel (red, n = 9), ZC4H2 cKO-Vehicle (green, n = 8), and ZC4H2 cKO-Perampanel (blue, n = 9) neurons before LTP induction, and bar graph (mean ± SEM), overlaid with the actual data points, shows the normalized eEPSCs amplitude at 45 min after LTP induction. Sample traces show EPSCs before (1) and 45 min after LTP (2). (Scale bars, 25 pA/25 ms.) *P < 0.05; **P < 0.01, and two-way ANOVA with Bonferroni correction analysis for (B and I), two-tailed Student’s t test for (D–H).

To assess the therapeutic potential of AMPAR inhibition on cognitive performance, 8-wk-old age ZC4H2 cKO and Control mice were treated with perampanel using the same dosing strategy, followed by a series of behavioral assays to evaluate their cognitive abilities (Fig. 5C). In the novel object recognition test, perampanel had no effect on total exploration time but partially restored recognition memory in ZC4H2 cKO mice, while having no measurable impact on Control mice (Fig. 5D). In the three-chamber test, perampanel treatment had no effect on social interaction across genotypes but fully rescued social memory deficits in ZC4H2 cKO mice (Fig. 5 E and F). Spatial learning and memory were examined using the Morris water maze test. Results showed that total swimming distance was comparable among groups, confirming that perampanel had no effect on motor function (Fig. 5H). During the training period, perampanel-treated ZC4H2 cKO mice exhibited a shortened escape latency by day 6 relative to vehicle-treated counterparts (Fig. 5G), suggesting partial rescue of learning deficits. In the probe trial, perampanel-treated ZC4H2 cKO mice showed performance indistinguishable from either treatment group of Control mice, including comparable time spent, distance traveled, and number of platform crossings within the target quadrant (Fig. 5H), demonstrating full restoration of spatial memory. Despite these cognitive improvements, perampanel failed to restore synaptic plasticity, with LTP remaining impaired in ZC4H2 cKO mice treated with perampanel and showing no changes in Control mice (Fig. 5I). Consistently, BS3 cross-linking-based immunoprecipitation analysis revealed that the elevated surface expression of GluA1 and GluA2 in ZC4H2-deficient neurons was unaffected by perampanel (SI Appendix, Fig. S4D), suggesting that AMPARs trafficking was not altered. Collectively, these results demonstrate that pharmacological inhibition of AMPARs by perampanel effectively reverse cognitive malfunction associated with ZC4H2 deficiency by suppressing excessive AMPAR-mediated synaptic transmission, although it did not correct persistent deficits in AMPARs trafficking and synaptic plasticity of LTP.

Discussion

This study identifies ZC4H2, the pathogenic factor in ZARD, as a postsynaptic regulator that orchestrates excitatory synaptic activity through direct association with core PSD molecules, including AMPARs. Loss of ZC4H2 disrupts this regulatory network, resulting in aberrant accumulation of postsynaptic AMPARs and increases AMPAR-mediated synaptic transmission, thereby leading to compromised synaptic plasticity and cognitive deficits. Notably, pharmacological inhibition of AMPARs with the noncompetitive inhibitor perampanel effectively restores excitatory synaptic function and reverses cognitive impairments, underscoring the therapeutic potential of targeting AMPAR hyperactivity in ZC4H2-associated intellectual disability.

Excitatory postsynaptic proteins fulfill diverse and essential roles in synaptic architecture and function, including the anchoring of postsynaptic receptors and transduction of synaptic activity into biochemical signal cascades (1, 2). Genetic disruption of these proteins frequently results in impaired excitatory synaptic transmission and plasticity, a pathological mechanism broadly implicated in cognitive malfunction across neurodevelopmental and neuropsychiatric disorders (6, 9). Expanding the mechanistic understanding of these pathogenic PSD proteins is therefore critical for elucidating the synaptic etiology of brain pathologies. Intellectual disability is a prominent clinical feature of ZARD. This research reveals that ZC4H2 is a key modulator of excitatory synaptic function and cognitive capacity and forms a specific protein interactome enriched in glutamatergic synaptic components in hippocampal neurons. Notably, several proteins within this network, including GluA2 (11), GluA3 (12), GluN1 (35), GluN2A (36), and GluK2 (37), carry mutations linked to intellectual disability and altered receptor function. ZC4H2 also interacts with postsynaptic scaffolding and regulatory proteins, such as Shank2 (38, 39) and TARP-γ2 (13), which have intellectual disability-associated genetic mutations. These findings suggest that disruption of the ZC4H2-centered synaptic protein network constitutes a central pathogenic mechanism underlying ZARD-related intellectual disability. In addition to its role in synaptic activity, ZC4H2 depletion also causes pronounced abnormalities in neuronal morphology and synaptogenesis, implicating it in the structural development of neural circuits. Proteomic analyses further reveal that ZC4H2 interacts with several key proteins involved in RNA metabolism and translational control. These include YTHDF3, a N6-methyladenosine (m⁶A) reader that regulates mRNA translation and stability (40), as well as the RNA-binding proteins G3BP2 and FUS, both of which participate in diverse RNA processing pathways and are linked to neurodegenerative disease (41, 42). These findings suggest that ZC4H2 may coordinate neuronal function through dual regulation of synaptic activity and RNA homeostasis. Defining the specific aspects of RNA metabolism under ZC4H2 control, and their physiological and pathological relevance to neuronal function, represents a critical direction for future investigation.

Neuropsychiatry disorders represent a heterogeneous spectrum of complex diseases, often characterized by overlapping clinical phenotypes and multifactorial genetic underpinnings. Intellectual disability is a core manifestation in ZARD, yet patients frequently present with additional comorbidities such as seizures and motor impairments (20). Interestingly, mice with conditional deletion of ZC4H2 in forebrain excitatory neurons exhibited profound cognitive deficits without spontaneous seizures or motor abnormalities, suggesting that ZC4H2 likely exerts critical functions in other neural cell types or brain regions beyond the excitatory forebrain neurons. In addition, similar cognitive phenotypes can arise from distinct genetic etiologies. More than 2,500 genes have been implicated in intellectual disability, and a growing body of studies evidence suggests that many converge on a common pathological mechanism dysregulating excitatory synaptic activity (10). In particular, enhanced postsynaptic AMPARs expression and resultant dysregulation of AMPAR-mediated synaptic transmission and plasticity have been linked to multiple pathogenic factors, including RNF220 (17) and SynGAP (15, 16). The present study identifies a similar pathological mechanism in ZC4H2 deficiency, wherein aberrantly elevated AMPARs expression impaired synaptic activity and cognitive function. Notably, pharmacological inhibition of AMPARs with perampanel, a selective noncompetitive antagonist, restored synaptic transmission and rescued cognitive performance in ZC4H2 cKO mice. These findings support AMPAR normalization as a potentially generalizable therapeutic strategy for treating cognitive dysfunction associated with genetic forms of intellectual disability. Future work should examine whether perampanel confers similar benefits in other genetic models, particularly those with shared synaptic pathologies. It is also important to address current limitations: All experiments were conducted in male mice due to the X-linked nature of ZC4H2, yet female patients with ZARD-related intellectual disability have been reported. Determining whether these mechanistic findings extend to females is critical for translational relevance. Furthermore, over 30 missense mutations in ZC4H2 have been identified in affected individuals. Clarifying whether these variants represent loss-of-function alleles that disrupt AMPAR regulation will be essential for refining patient stratification and guiding clinical application of AMPAR-targeted therapies such as perampanel.

Materials and Methods

All mice were maintained with C57BL/6 background and kept in individual ventilated cages (IVC) in specific pathogen free (SPF) environment, and handled according to guidelines approved by the Animal Care and Use Committee of Kunming Institute of Zoology, Chinese Academy of Sciences (IACUC-PA-2023-08-023). Cells were cultured in DMEM supplemented with 10% FBS and 1% penicillin and streptomycin. Full details on the materials and methods are described in SI Appendix. The statistics of the graphs in the figures are included in Dataset S2.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information.