The deubiquitinase USP6 affects memory and synaptic plasticity through modulating NMDA receptor stability

Ubiquitin-specific protease (USP) 6 is a hominoid deubiquitinating enzyme previously implicated in intellectual disability and autism spectrum disorder. Although these findings link USP6 to higher brain function, potential roles for USP6 in cognition have not been investigated. Here, we report that USP6 is highly expressed in induced human neurons and that neuron-specific expression of USP6 enhances learning and memory in a transgenic mouse model. Similarly, USP6 expression regulates N-methyl-D-aspartate-type glutamate receptor (NMDAR)-dependent long-term potentiation and long-term depression in USP6 transgenic mouse hippocampi. Proteomic characterization of transgenic USP6 mouse cortex reveals attenuated NMDAR ubiquitination, with concomitant elevation in NMDAR expression, stability, and cell surface distribution with USP6 overexpression. USP6 positively modulates GluN1 expression in transfected cells, and USP6 down-regulation impedes focal GluN1 distribution at postsynaptic densities and impairs synaptic function in neurons derived from human embryonic stem cells. Together, these results indicate that USP6 enhances NMDAR stability to promote synaptic function and cognition.


Introduction
Humans have evolved an advanced cognitive capacity over other mammalian species, featuring quantitatively enhanced functional abilities, such as learning and memory, as well as qualitatively new abilities such as speech. Human intelligence has been attributed to expanded regions in the cerebral cortex and enhanced complexity of neuronal connectivity in the human central nervous system (CNS). Genetic factors are thought to be fundamental to the evolution of the human brain. Several genes, including the hominoid-specific gene TBC1 domain family member 3 (TBC1D3) [1], the primate-specific gene Transmembrane protein 14B (TMEM14B) [2], and the human-specific genes Ras homolog (Rho) GTPase activating protein 11B 2 (ARH-GAP11B) [3] and Notch 2 N-terminal like (NOTCH2NL) [4,5], have been found to be important for cortical development. In addition, synaptic wiring within the complex neural network contributes to advanced cognition and social function. The human neocortex contains approximately 1.5 × 10 14 synaptic connections [6]. Pyramidal neurons in the human cortex form twice as many synapses as those in other primates, such as marmosets and macaques [7]. Several genes have been implicated in cognitive function through their involvement in synapse formation. SLIT-ROBO Rho GTPase activating protein 2C (SRGAP2C), a splice variant of SRGAP2, promotes synapse maturation by inhibiting the function of the SRGAP2 gene [8]. Dysfunction of Sushi repeat containing protein X-linked 2 (SRPX2), which regulates synaptogenesis in cortical neurons, is associated with intellectual disability and language disorder [9,10].
Proper physiological function and the regulation of excitatory synapses in the CNS strongly correlate with cognitive ability [11]. Glutamate is the primary neurotransmitter in excitatory glutamatergic neurons in the mammalian brain [12]. Ionotropic α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and N-methyl-D-aspartate (NMDA) glutamate receptors are the two major receptor types found within the postsynaptic density (PSD) of glutamatergic synapses, which transmit presynaptic signals to postsynaptic neurons [13]. NMDAtype glutamate receptor (NMDAR) stimulation triggers neuronal calcium influx through membrane-bound ion channels and activates intracellular calcium/calmodulin dependent protein kinase II alpha (CamK2a) or phosphatase 1 (PP1) signaling pathways. Stimulation of NMDAR-dependent calcium signaling pathways leads to the redistribution of AMPA receptors (AMPARs), resulting in synaptic changes associated with plasticity, such as long-term potentiation (LTP) or long-term depression (LTD), which form the physiological basis of learning and memory [14]. Modulation of GluN1 [15], GluN2B [16], and GluA1 [17] function through mutation or expression is associated with intellectual impairment. Distribution and regulation of NMDARs and AMPARs at PSDs is crucial to physiological synaptic function and cognition. Both NMDARs and AMPARs are degraded through the ubiquitin (Ub)-proteasome system (UPS) [18][19][20][21][22]. However, mechanisms underlying Ub targeting and the homeostatic maintenance of glutamate receptors in synaptic function remain largely unknown.
Ub-specific protease (USP) 6 is a hominoid-specific protein deubiquitinase (DUB) containing Tre-2/USP6, BUB2, and Cdc16 (TBC) and USP domains, and the USP6 gene is found specifically in humans and orangutans [23]. USP6 TBC and USP domains are highly homologous to TBC1D3 and USP32 [23], respectively. Perturbation of USP6 expression through aberrant chromosomal translocation is associated with mental retardation [24] and Asperger syndrome [25]. In addition, a recent study revealed that TBC1D3 may play a crucial role in human intelligence by enhancing neural progenitor cell generation and cortical folding during development [1]. These findings suggest that the USP6 gene may also be involved in the evolution of human intelligence.
Here, we generated a transgenic USP6-overexpression mouse model in which human USP6 is specifically expressed in excitatory neurons within the cerebral cortex and hippocampus. USP6 transgenic mice exhibited enhanced learning and memory behavior in Morris water maze (MWM) and novel object recognition (NOR) tests. Moreover, transgenic USP6 expression increases synaptic function and NMDAR expression in USP6 transgenic mouse brains. Consistent with cognitive enhancement associated with USP6 gain of function, short hairpin RNA (shRNA)-mediated USP6 depletion in human embryonic stem cell (ESC)-derived neurons reduces NMDAR expression and function. We also found that USP6 binds to and deubiquitinates NMDARs, thereby enhancing NMDAR stability. In summary, we identified USP6 as a novel hominoid-specific regulator of NMDARs; USP6 enhances NMDAR stability through posttranslational homeostasis mechanisms, and its modulation can impact synaptic activity and memory formation.

USP6 enhances learning and memory behavior in USP6 transgenic mice
Previous studies have shown that USP6 is expressed in multiple human tissues, and the highest expression is found in testes [23]. Using quantitative reverse transcription PCR (qRT-PCR) analysis, we found that USP6 expression was higher in the adult human cortex compared with fetal cortex (gestational week [GW] 22 to 27) ( Fig 1A). To further characterize USP6 expression in different cell types in the CNS, we examined USP6 transcripts in H9 human ESCs, astrocytes, induced human excitatory neurons, and interneurons. USP6 mRNA levels are high in differentiated excitatory neurons and interneurons, whereas some expression was detectable in ESCs and astrocytes (Fig 1B).
To investigate the potential effects associated with USP6 gain of function in mouse brain, we generated a transgenic mouse model expressing C-terminal hemagglutinin (HA)-tagged human USP6 under the regulation of the CamK2a promoter (CAM-USP6) (Fig 1C). Six independent founder lines were analyzed to detect USP6 protein expression, and two lines featuring elevated USP6 expression levels-i.e., USP6#1 (line 1) and USP6#2 (line 2)-were selected for breeding and analysis. Consistent with the CamK2a expression pattern previously characterized in mouse brain [26], we observed transgenic USP6 expression in the cortex and hippocampus, whereas its expression was undetectable in the cerebellum (Fig 1D). The body weight of CAM-USP6 mice (lines 1 and 2) was comparable to that of wild-type (WT) littermates (S1A Fig). Moreover, no significant difference in locomotor activity or anxiety was observed in open field tests, and CAM-USP6 transgenic animals did not differ from WT littermates in cumulative distance traveled (S1B Fig Given that USP6 is a hominoid-specific gene and disruption of USP6 gene expression by chromosomal translocation is associated with mental retardation [24], USP6 may play a role in higher cognitive functions in humans. To test this hypothesis, we firstly evaluated learning and memory in CAM-USP6 animals and their WT littermates using the NOR test and observed that CAM-USP6 transgenic animals spent more time in the zone containing novel objects and feature a higher discrimination index compared with WT controls (Fig 1E). In addition, we next examined the mice for spatial reference memory and reversal learning in the MWM task ( Fig 1F) and found that CAM-USP6 transgenic animals required less time to reach the hidden platform during training in acquisition and reversal phase of MWM tests (Fig 1G). During the probe tests of both MWM and reversal learning, CAM-USP6 mice spent significantly more time in the target quadrant than WT mice (Fig 1H), indicating that USP6 overexpression enhances spatial memory and cognitive flexibility in CAM-USP6 mice. Because we observed comparable MWM trends in an additional USP6 transgenic line (line 2) (S2 Fig), the effects observed were unlikely to be due to gene disruption events from random insertion of the human USP6 cDNA.

USP6 enhances social interaction and ultrasonic vocalization in vivo
Aberrant genetic disruption of the USP6 locus is also associated with social behavior disorders, such as Asperger syndrome [25]. To determine whether transgenic USP6 expression can affect social behavior in mice, we performed a three-chamber test (Fig 2A) and found that the CAM-USP6 mice spent more time approaching a "stranger" mouse compared with an empty cage during sociability test assays ( Fig 2B). In social novelty tests, CAM-USP6 mice and littermate controls showed similar preference to a stranger mouse compared with a familiar animal ( Fig 2C). We further characterized the communication behavior of CAM-USP6 pups through ultrasonic vocalization (USV) recordings. Pups separated from their mothers typically emit USV signatures that initiate a "search-and-retrieve" response in nearby mothers [9]. Isolationinduced pup USV behavior is an assay widely used to characterize mouse models of human diseases that involve deficits in language and social behavior [9,27,28]. CAM-USP6 pups vocalized more frequently than WT controls on postnatal day (P)7 (Fig 2D and 2E), indicating that transgenic USP6 expression can enhance communication between pups and dams. Together, these results indicate that USP6 up-regulation enhances spatial memory behavior and aspects of language and social interaction.

USP6 expression does not affect cortical development in USP6 transgenic mice
Expression of the USP6 paralog TBC1D3 has been previously observed to induce cortical folding during brain development by increasing the production of neural progenitor cells. To investigate the potential role of USP6 in embryonic neural development, we generated transgenic mice expressing USP6 using a Nestin promoter (Nestin-USP6) (S3A Fig) and  To determine whether USP6 expression affects synaptic density and morphology, we performed transmission electron microscopy (TEM) and Golgi staining and observed increased dendritic spine density in cortical layer II and V pyramidal neurons from CAM-USP6 mouse brain (S5A and S5B Fig). Although dendritic spine density of hippocampal CA1 pyramidal neurons in CAM-USP6 mice was found to be higher than in WT littermates, these trends were not found to be statistically significant (S5A and S5B Fig). WT: n = 9 mice, CAM-USP6: n = 9 mice. � P < 0.05 as determined by Student t test. The underlying data for this figure can be found in S1 Data. CAM, CamK2a; NS, not significant; P, postnatal day; USP, ubiquitin-specific protease; USV, ultrasonic vocalization; WT, wild-type.

CAM-USP6 mice display enhanced synaptic function
Thus far, we established that USP6 plays a role in memory/social behavior. To determine whether transgenic USP6 expression can enhance NMDAR-dependent synaptic function in vivo, we performed electrophysiological recordings using acute hippocampal slices from WT and CAM-USP6 mice. We observed comparable input-output responses ( Fig 3A) and pairedpulse ratios (Fig 3B) between CAM-USP6 mice and WT littermate controls, suggesting that USP6 expression had minimal to no effect on basal synaptic transmission. To determine whether USP6 expression influences synaptic plasticity in CAM-USP6 mice, we examined NMDAR-dependent LTP and LTD responses in CAM-USP6 mice. We found that hippocampal CA1 LTP (100 Hz for 1 second) was enhanced ( Fig 3C) and LTD (1 Hz for 900 seconds) was attenuated (Fig 3D) in CAM-USP6 compared with WT controls. As both AMPARs and NMDARs are involved in synaptic plasticity, we first examined miniature excitatory postsynaptic currents (mEPSCs) in hippocampal CA1 neurons from CAM-USP6 and WT mice and found that both mEPSC amplitude and frequency were unchanged between the two groups (S6 Fig). Given that AMPAR mEPSC recordings showed normal responses (S6 Fig), AMPAR function appears to be intact in both CAM-USP6 and WT mice. To characterize the potential differences in NMDAR function in the CAM-USP6 mice, we measured NMDA/AMPA ratios and evoked NMDA current (evoked NMDA receptor-mediated excitatory postsynaptic current [eEPSC]) in CAM-USP6 mice and WT controls; we observed increased NMDA/AMPA ratios and NMDA-eEPSC amplitudes in hippocampal CAM-USP6 CA1 neurons (Fig 3E and  3F). Together, these results indicate that USP6 expression enhances NMDAR-dependent synaptic function in CAM-USP6 mice.

USP6 stabilizes NMDARs through its deubiquitinating activity
Thus far, mechanisms underlying USP6-mediated memory enhancement and social behavior appear to be independent of developmental events and changes in brain physiology. However, we observed USP6 distribution in synaptosome and PSD-enriched fractions using biochemical fractionation (S7 Fig), suggesting that USP6 may affect synaptic function through the deubiquitination of synaptic proteins. Therefore, we immunopurified trypsin-digested proteins extracted from mouse cortical tissue for peptides labeled with a di-Gly Ub-labeling signature and compared the relative abundance of these di-Gly peptides in cortical tissues from CAM-USP6 mice and WT controls ( Fig 4A). Using this method, we identified 175 proteins (150 up-regulated and 25 down-regulated) that were differentially conjugated to Ub between CAM-USP6 and WT mouse brains (S1 Table).
Interestingly, we observed enrichment of di-Gly-labeled proteins involved in pathways related to synaptic function and plasticity, including LTP in WT versus CAM-USP6 transgenic mouse brain (Fig 4B and 4C). Differential Ub signatures were observed in the small Rho GTPase Rac family small GTPase 1 (Rac1), previously shown to be enriched in the hippocampus and contribute to LTP function [29], and synaptotagmin (Syt)1, which has been previously shown to work cooperatively with Syt7 in mediating LTP [30] (Fig 4D). Interestingly, we also observed differences in CAM-USP6 versus WT Ub signatures in postsynaptic components directly related to PSD function, including PSD95 and the NMDAR subunit GluN2B (Fig 4D). These results indicate that CAM-USP6 may drive changes in synaptic function to modulate/ enhance memory, cognition, and social behavior.
To determine whether the expression of key targets identified in the Ub-associated proteome is altered in CAM-USP6 brain, we performed immunoblot analyses of proteins from CAM-USP6 and WT mouse brain. We observed up-regulation of GluN1, GluN2A, and GluN2B NMDAR subunits in CAM-USP6 mouse cortex and hippocampus, but little or no difference was observed in GluA1 and GluA2 AMPAR levels ( Fig 5A and 5B). We observed little or no change in glutamate receptor mRNA expression by qRT-PCR analysis of the CAM-USP6 mouse cortex, indicating that the differential NMDAR expression was due to modulation at the protein level as opposed to alterations in transcription (S8 Fig). To examine whether USP6 up-regulation can affect the distribution of NMDAR subunits to the cell surface, we subjected primary neurons from WT and CAM-USP6 mouse embryos to cell surface biotin-labeling assays. We observed significant up-regulation of cell surface GluN1, but no change was observed in the AMPAR subunit GluA1 or transferrin receptor (TfR), indicating that CAM-USP6 plays a potential role in enhancing cell surface NMDAR distribution ( Fig 5C). Moreover, compared with WT controls, increased GluN1, GluN2A, and GluN2B levels were observed in both synaptosome and PSD-enriched fractions from CAM-USP6 mouse brains ( Fig 5D). Given that these results suggest a role for USP6 in mediating physiological GluN1 homeostasis and cell surface distribution, we examined the effects of USP6 depletion in a human HEK293T cell line expressing GluN1. We found that shRNA-mediated USP6 depletion markedly down-regulated GluN1 protein expression and increased levels of ubiquitinated GluN1 (S9 Fig and Fig 5E). In contrast, GluN1 levels were not perturbed by shRNA-mediated TBC1D3 or USP32 depletion (S10 Fig), suggesting that USP6 mediates specific effects in targeting GluN1 for deubiquitination.

USP6 interacts with NMDARs to attenuate UPS-mediated NMDAR turnover
After establishing a role for USP6 in modulating NMDAR ubiquitination and expression, we determined whether USP6 interacts with NMDAR subunits. Immunoprecipitation (IP) of USP6 from CAM-USP6 mouse brain lysates successfully demonstrates coprecipitation of USP6 with NMDAR subunits (GluN1 and GluN2B); no coimmunoprecipitation (co-IP) interaction was detected with AMPAR subunits (GluA1 and GluA2) and amyloid precursor protein (APP) (Fig 6A). We further characterized NMDAR binding to TBC or USP domain USP6 homologs (i.e., TBC1D3 and USP32); we found that both USP6 and TBC1D3 coimmunoprecipitated with GluN1/GluN2B, but USP32 failed to interact with NMDAR ( Fig 6B and 6C). These results indicate that the N-terminal TBC domain is required for NMDAR interaction.
To determine whether USP6 influences GluN1 turnover, we performed cycloheximide chase assays in HEK293T cells overexpressing GluN1; we found that GluN1 degradation was markedly accelerated by shRNA-mediated USP6 down-regulation ( Fig 7D). Together, these results indicate that USP6 can interact with NMDAR subunits and promotes their stabilization by reducing NMDAR ubiquitination.

USP6 depletion reduces NMDAR expression and function in ESC-derived human excitatory neurons
To investigate the effect of USP6 in a human neuronal model system, we differentiated human excitatory neurons from H9 ESCs (Fig 8A). To confirm successful differentiation of ESCs to   (Fig 8B). We also confirmed maturation by characterizing action potential response in differentiated neurons through patch-clamp recording in response to an increasing stimulus (Fig 8C).
To further determine whether USP6 loss of function negatively regulates GluN1 expression in ESC-derived human neurons, we used lentiviruses to transduce USP6-targeted shRNAs in differentiated neurons (Fig 8D). USP6 knockdown markedly diminished the formation of total dendritic GluN1 puncta and colocalized GluN1/PSD95 puncta (Fig 8E). In addition, shRNA-mediated down-regulation of USP6 diminished mEPSC frequency but had a minimal effect on mEPSCs in induced human excitatory neurons (Fig 8F). Together, these results indicate that USP6 down-regulation can reduce GluN1 clustering/distribution to PSD and impair mEPSC frequency in an ESC-derived human neuronal model.

Discussion
USP6 is a hominoid-specific gene that appeared during the evolution of the hominoid genome approximately 12-16 million years ago. The genomic USP6 locus is prone to frequent chromosomal breakage and translocation events in which genomic rearrangement and consequent down-regulation of USP6 expression likely leads to intellectual impairment and aberrations in social behavior, such as mental retardation and autism spectrum disorder [24,25]. Together, these findings suggest that USP6 plays an important role in the evolution of human intelligence.
Protein homeostasis and regulatory mechanisms underlying protein synthesis and degradation are crucial for normal CNS function. Specific regulation of functional components involved in synaptic transmission is also subject to modulation by protein degradation mechanisms. Recent studies have highlighted the function of UPS in synaptic plasticity and various processes related to learning and memory [18,31,32]. However, how synaptic plasticity is coupled to the regulation of specific Ub proteases and the ubiquitination of specific proteins remains largely unclear. Although previous studies have indicated that numerous Ub ligases, including ubiquitin protein ligase E3A (UBE3A) [33][34][35][36], Neural precursor cell expressed developmentally down-regulated protein 4 (NEDD4) [37,38], and Cadherin 1 (Cdh1) [39,40], play a fundamental role in synaptic remodeling and function, opposing roles for DUB enzymes in regulating synaptic function have not been well characterized. Within the hominoid genome, USP6 and USP41 represent the sole hominoid-specific DUB enzyme species, suggesting that potential Ub-dependent regulatory mechanisms involving DUBs play a role in neurodegeneration and/or intelligence in hominoids. Our research reveals the pivotal role of USP6 in regulating NMDAR ubiquitination and stability. Activation of NMDARs is required for neurites from 15 neurons); quantification of PSD95 puncta: control shRNA (n = 61 neurites from 44 neurons), USP6 shRNA1 (n = 29 neurites from 23 neurons), and USP6 shRNA2 (n = 21 neurites from 15 neurons). The data represent means ± SEM. � P < 0.05, �� P < 0.01, ��� P < 0.001 as determined by one-way ANOVA with Tukey's post hoc analysis. (F) Representative NMDA mEPSC recordings in induced human excitatory neurons infected with USP6 shRNA lentivirus. Data represent means ± SEM. n = 3. � P < 0.05 as determined by one-way ANOVA with Tukey's post hoc analysis. The underlying data for this figure can be found in S1 Data. B27, B-27 serum-free supplement; BDNF, brainderived neurotrophic factor; EGF, epidermal growth factor receptor; ESC, embryonic stem cell; FGF2, fibroblast growth factor 2; GDNF, glial cell-derived neurotrophic factor; Glu, glutamate ionotropic receptor; MAP2, microtubuleassociated protein 2; mEPSC, miniature excitatory postsynaptic current; N2aa, DMEM-F12 medium with N2-supplement and ascorbic acid; NMDA, N-methyl-D-aspartate; NPC, neural progenitor cell; PSD, postsynaptic density; qRT-PCR, quantitative reverse transcription PCR; shRNA, short hairpin RNA; USP, ubiquitin-specific protease; vGluT1, vesicular glutamate transporter. https://doi.org/10.1371/journal.pbio.3000525.g008 LTP and LTD at hippocampal CA1 synapses [41,42], implicating UPS-mediated synaptic remodeling in hominoid cognition. In addition to NMDARs, we identified other synaptic proteins as USP6 substrates, which can potentially contribute to enhancements in synaptic function. These substrates include the small Rho GTPase Rac1 and the synaptotagmin Syt1, which are components previously shown to be enriched in the hippocampus and important for LTP function [29,30]. In future studies, it will be of interest to determine whether mechanisms underlying synaptic plasticity are coupled to the activation or distribution of USP6 and its targets in learning and memory.
The UPS is a fundamental protein degradation system, and aberrant UPS function has been observed in several neurodegenerative diseases, such as Alzheimer disease, Parkinson disease, and Huntington disease [43]. Many neurodegenerative diseases feature the pathological appearance of intracellular Ub-positive inclusions that coincide with the proteostatic dysfunction of other aggregate-prone proteins in neurodegeneration [44]. Together, this suggests that UPS dysfunction in neurodegenerative disorders contributes to the accumulation of neurotoxic proteins that affect neurological decline and neurodegenerative onset. Based on our findings here, it will be interesting to determine the specific role of USP6-mediated protein turnover in neurodegenerative disease.
During human evolution, several primate-or human-specific genes have evolved, which may be fundamentally important to accommodate enhanced aspects of human cognition. Given that the USP6 gene locus is located on a breakage-prone region within chromosome 17, USP6 likely evolved through genomic translocation events resulting in the fusion of TBC1D3 and USP32. Given that mutations in NMDAR subunits, such as GluN1 and GluN2B, have been previously linked to intellectual impairment [15,16], it appears reasonable that hominoid-derived factors may have evolved to enhance downstream effectors associated with synaptic function and cognition/memory. USP6 has evolved a unique capacity to enhance NMDAR function; although TBC1D3 interacts with NMDARs, TBC1D3 lacks a domain with deubiquitinating protease activity. Although USP32 is a DUB, USP32 requires a TBC domain to interact with NMDARs. Therefore, the combinatorial evolution of USP6 may confer a unique function driving the enhancements in hominoid-specific synaptic function in the CNS.
Here, we also present a model system for studying hominoid-specific genes and characterizing their function in cognition; utilizing gain-of-function analysis in mouse models combined with shRNA-targeted gene depletion in human ESC-derived neurons, we established a systematic pipeline to determine how hominoid genes can affect cognition and intelligence. Furthermore, this system may enable subsequent studies investigating mutations in hominoid-specific genes manifesting disorders related to intelligence and social behavior. This system can also be used to recapitulate differential gene composition observed in various hominoid species, which may ultimately define how cognitive function evolved in humans.
In summary, we identified USP6 as a novel hominoid component that enhances NMDAR stability and function (Fig 9), in which USP6 plays a pivotal role in the evolution of intelligence. Given that NMDARs and their dysfunction have been implicated in human intelligence and cognition, enhancing USP6 function may be a potential option in future therapeutic studies.

Ethics statement
All the protocols and procedures for the mice studies were approved by the School of Medicine, Xiamen University. The animal care and use protocol adhered to guidelines of Institutional Animal Care and Use Committee of Xiamen University. The human studies were approved with informed consent by the ethical review board at the School of Medicine, Xiamen University (Project# XDYX2019011).

Human brain collection
Human fetal brain tissue was collected and curated by the Women and Children's Hospital, School of Medicine, Xiamen University (Table 1). Patients provided informed consent in accordance with the legal and institutional ethical guidelines defined by the hospital. Donors comprised pregnant women discontinuing their pregnancies due to congenital heart defects in the fetuses. All donors provided the fetal brains voluntarily and gave informed consent. Karyotype analysis confirmed that all fetal samples used had normal karyotypes. Adult brain tissue was collected from Xiamen Humanity Hospital following trauma surgery with informed patient consent according to the legal and institutional ethical requirements of the hospital ( Table 1). All brain tissues were stored on dry ice and transported to the laboratory for subsequent examination and processing.

Real-time PCR analysis
Tissue or cell samples were homogenized, and total RNA was isolated using TRIzol reagent (Invitrogen) according to the manufacturer's instructions. cDNA was synthesized using Super-Script III Reverse Transcriptase (TOYOBO) with random hexamers and OligdT primers. qRT-PCR was performed using a ROCHE 480 real-time LightCycler system and SYBR green reagent (ABI); primer sequence information is in Table 3. At least three independent experiments were performed for all experiments described, and all results presented were calculated from CT values derived from the qRT-PCR reactions. All values were normalized to β-actin.

Histology, immunofluorescence, and confocal imaging
Male mice were anesthetized with isoflurane followed by serial intracardial perfusion with PBS and 4% PFA [46]. Whole brains were rapidly dissected and postfixed in 4% PFA at 4˚C overnight, dehydrated in PBS containing 30% sucrose at 4˚C, and embedded and frozen in OCT at −80˚C. Brain tissues were sectioned into 15-μm slices using a Leica microtome. For histological analysis, frozen sections were stained with 1% cresyl violet (Beyotime) for Nissl staining. For immunofluorescence staining, brain sections were permeabilized using 0.1% Triton X-100 in PBS, blocked in 5% normal bovine serum in PBS, and incubated with primary antibodies at 4˚C overnight. Fluorochrome-conjugated secondary antibodies were then used to detect primary antibody signals, and stained sections were imaged using a Leica SP8 confocal microscope.

Target gene siRNA sequences
Control siRNA

Interneuron differentiation from human ESCs
The interneuron induction was performed as previously described [48]. Human H9 ESCs were trypsinized and cultured as floating spheres in low adherent flasks in KSR medium Table 3. Primer sequences.

TEM
TEM was used to detect synaptic ultrastructure as previously described [49]. Male mice were anesthetized with isoflurane followed by intracardial perfusion with PBS, followed by fixative containing 4% PFA and 2.5% glutaraldehyde. Tissue of frontal cortex and hippocampus were rapidly dissected, postfixed in fixative at 4˚C overnight, and then postfixed by immersion in 1% osmium tetroxide at 4˚C for 2 hours. The specimens were dehydrated in an ethanol gradient from 30% to 100% and embedded in Spurr's resin. After dehydration and embedment, serial ultrathin sections (70-nm in thickness) were prepared and stained with lead citrate and uranyl acetate, and TEM images were captured using a Hitachi HT-7800 transmission electron microscope. The number of synaptic connections was quantified using NIH ImageJ software by a person who was blinded to the animal genotype.

Golgi staining
Golgi staining was performed in 2-month-old CAM-USP6 male transgenic mice using the FD Rapid Golgi Stain Kit (FD NeuroTechonologies) according to the manufacturer's instructions as previously described [50]. Dendritic spine density was imaged using a NIKON microscope in the CA1 region of the hippocampus and quantified using NIH ImageJ software [51]. Student t tests were used to determine statistical significance between WT and USP6 transgenic neurons.

Ub-modified proteome
Ubiquitinated proteins were detected using an IP-MS/MS strategy as previously described [52,53]. Briefly, 2-month-old male WT and CAM-USP6 mouse cortices were collected and frozen in liquid nitrogen. Frozen samples were pooled and ground into a powder (1 g per sample) for total protein extraction using TCA-acetone. Proteins were resuspended in UA buffer (8 M urea, 150 mM Tris-HCl, pH 8.0), and lysates were sonicated, centrifuged, and filtered. Supernatants were collected and quantified by Bradford assay (Thermo Scientific Multiskan). Protein (20 mg) was reduced using 1.25 mM dithiothreitol (DTT) for 30 minutes at 55˚C, and resulting free cysteines were alkylated with 10 mM iodoacetamide for 15 minutes at room temperature in the dark. Approximately 20 mg of DTT and iodoacetamide-treated protein was digested overnight at 37˚C with TPCK trypsin (Worthington) at an enzyme-to-substrate ratio of 1:50 after a 4-fold dilution in 20 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) (pH 8.0). Trypsin digestion was stopped by the addition of trifluoroacetic acid to a final concentration of 1%. After precipitates were removed by centrifugation for 20 minutes at 10,000g, supernatants were desalted using Sep-Pak Classic C18 Cartridges (Waters) followed by lyophilization.
Lyophilized peptides were dissolved in immunoaffinity purification (IAP) buffer (50 mM MOPS−NaOH, pH 7.2, 10 mM Na 2 HPO 4 , and 50 mM NaCl) and centrifuged at 10,000g at 4˚C for 10 minutes. For each sample, 250 μg of di-Gly-Lys antibody cross-linked on agarose beads (PTMScan Ub remnant motif K-ε-GG kit, Cell Signaling Technology) was used to immunopurify cleaved di-Gly Ub signatures, and di-Gly-Lys-containing peptides were enriched as previously described.
Fractions were individually injected for nanoLC-MS/MS analysis, and MS data were analyzed using MaxQuant software version 1.5.3.17 (Max Planck Institute of Biochemistry in Martinsried, Germany). Gene Ontology (GO) and KEGG pathway analysis was performed to characterize categorical gene function in the components identified.

Preparation of synaptosome and PSD fractions
Hippocampal tissues from male mice were dissected and homogenized on ice in 10 volumes of cold sucrose buffer (0.32 M sucrose and 25 mM HEPES, pH 7.4) with protease inhibitors. Homogenates were centrifuged at 710g for 10 minutes at 4˚C to isolate supernatant (S1) from large debris and nuclei. The S1 fraction was centrifuged at 10,000g for 15 minutes at 4˚C. Supernatants were retained (S2; light membrane and cytosolic fraction), and pellets were washed twice with cold sucrose buffer and resuspended in cold HEPES-buffered saline (HBS) containing 25 mM HEPES and 150 mM NaCl (pH 7.4) to obtain synaptosome fractions. PSDenriched fractions were prepared by solubilizing synaptosomes in 1% Triton HBS at 4˚C for 30 minutes and, subsequently, centrifuging at 10,000g for 20 minutes. Pellets were resuspended in 3% sodium dodecyl sulfate (SDS) in HBS to yield PSD-enriched fractions [50,54].

co-IP assays
CAM-USP6 male mouse brain or transfected HEK293T cells were lysed in 0.5% TNEN lysis buffer supplemented with protease inhibitors. For all experiments, total protein concentration was determined by BCA assay (Thermo Fisher). Lysates were immunoprecipitated using anti-Myc, HA, or Flag antibodies in the presence of Protein G Dynabeads (Thermo Fisher), followed by immunoblot analysis.

Field potential recordings
Female mice were anesthetized with isoflurane, and brains were rapidly removed and placed in ice-cold, high-sucrose cutting solution containing 120 mM sucrose, 64 mM NaCl, 26 mM NaHCO 3 , 10 mM glucose, 2.5 mM KCl, 1.25 mM NaH 2 PO 4 , 10 mM MgSO 4 , and 0.5 mM CaCl 2 . Slices were sectioned on a Leica vibratome in high-sucrose cutting solution and immediately transferred to an incubation chamber with artificial cerebrospinal fluid (ACSF) containing 120 mM NaCl, 26 mM NaHCO 3 , 10 mM glucose, 3.5 mM KCl, 1.25 mM NaH 2 PO 4 , 1.3 mM MgSO 4 , and 2.5 mM CaCl 2 . Slices were allowed to recover at 32˚C for 1 hour before equilibration at room temperature for 1 hour. During recordings, slices were placed in a recording chamber perfused with ACSF and continuously aerated with 95% O 2 /5% CO 2 .
For LTP field recordings, the Shaffer collateral was stimulated with a concentric bipolar electrode placed in the CA3 stratum radiatum. Field potentials in the CA1 stratum radiatum were recorded using a micropipette filled with ACSF. Baseline responses were obtained every 20 seconds with a stimulation intensity that yielded a 30%-40% maximal amplitude (mV) response. LTP was induced by four trains of high-frequency stimulation (100 Hz, 1 second) separated by 30-second intervals. For LTD field recordings, concentric bipolar and patch electrodes were placed as described in the LTP protocol above. Baseline responses were obtained every 20 seconds with a stimulation intensity that yielded a 50%-60% maximal amplitude (mV) response. LTD stimulation was induced at a frequency of 1 Hz for 900 seconds. For input-output recordings, electrical stimulation in the CA3 stratum radiatum was applied sequentially in a gradient ranging from 0 to 0.8 mA in 0.1-mA steps. For paired-pulse ratio recordings, two pulses with stimulation intensities that yielded a 40% maximal amplitude (mV) response spaced at determined time intervals (20,50,100,200, and 500 milliseconds) were administered.

Whole-cell patch-clamp recordings
Acute hippocampal slices from female mice were processed and maintained as described above. To obtain NMDAR-to AMPAR-EPSC ratios, a concentric bipolar electrode was placed in the stratum radiatum to evoke EPSCs in CA1 pyramidal cells. CA1 pyramidal cells were affixed to patch electrodes containing 140 mM CsCH 3  Mixed AMPAR-EPSC and NMDAR-EPSC outputs were then recorded at +40 mV using the same stimulation pulse (100 μA). Peak NMDAR-EPSC was calculated at 50 milliseconds from the initial mixed EPSC output.
Action potentials in induced human excitatory neurons were recorded in current-clamp mode, in which recording pipettes were filled with intracellular solution containing (in mM): 122 K-gluconate, 5 NaCl, 0.3 CaCl 2 , 2 MgCl 2 , 1 egtazic acid (EGTA), 10 HEPES, 5 Na 2 -ATP, and 0.4 Na 3 -GTP; pH was adjusted to 7.2-7.3 with KOH, and the osmolarity was adjusted to 280 mOsm/kg with sucrose. Action potentials were generated by direct intracellular current injections (500 milliseconds) of increasing magnitude (in 10-pA steps) from −30 pA to 120 pA. In current-clamp mode, NMDAR-mediated mEPSCs (NMDA-mEPSCs) were recorded in the presence of 1 μM TTX, 20 μM CNQX, and 100 μM PTX with a Cs-based internal solution at a holding potential of −70 mV.

Open field test
In open field tests, the behavior of each mouse was characterized in an open field box (60 × 60 cm, height 60 cm), and total distance traveled was monitored to assess motor behavior. Percentage of the time spent in the center zone was recorded to determine anxiety-like behavior. To exclude estrogenic effects, only male mice were used in behavior tests.

MWM and reversal learning
MWM were performed as previously described [55,56]. The tests were conducted in a lightblue circular pool (diameter 1.2 m, height 0.5 m) filled with opaque water maintained at 22˚C. Four spatial markers comprising different shapes visible to the animals were affixed on the walls of the pool. A round fixed platform (diameter 10 cm) was submerged approximately 1-2 cm below the water level. On training days, the mouse was forced to swim in the water and allowed to search for 60 seconds to find the hidden platform, where the mouse remained for 10 seconds. For cases in which the mouse was unable to find the platform, the mouse was guided to the hidden platform and allowed to remain on the platform for 10 seconds. Each mouse was trained for 6 days with four trials per day from four different quadrants used to enter the pool. Mouse behavior was monitored and analyzed using Smart 3.0 software (Panlab), and escape latency was scored for each trial. On the last day, a probe test was performed without the platform, and mice were allowed to swim for 60 seconds; the probe test was performed with two trials; time spent in each quadrant and the target platform quadrant were subsequently analyzed [55]. Reversal learning was performed based on the MWM, with the following adjustments: each mouse was trained for 6 days with two trials per day from two different quadrants used to enter the pool, and the probe test was performed with one trial to reduce swimming intensity. After the probe test, the platform was moved to the opposite quadrant, and each mouse was trained for 3 days with 2 trials per day from two different quadrants used to enter the pool. On the last day, a second probe test was performed without the platform, and mice were allowed to swim for 60 seconds; time spent in each quadrant and the target platform quadrant were subsequently analyzed [56].

NOR
NOR was used to detect learning and memory according to a previous method [57]. At 24 hours before testing, each mouse was habituated for 5 minutes in a chamber (40 × 40 × 40 cm). During testing, the mouse was exposed to a set of three identical objects in the chamber for 10 minutes. The mouse was then removed from the chamber for 2 minutes while the chamber was cleaned with 70% ethanol and one of the objects was replaced with a novel object. The mouse was then returned to the chamber, and the interaction time with the two familiar or novel objects was recorded. Total interaction time was determined as the sum of interaction times (familiar 1 + familiar 2 + novel object). The discrimination index (%) was defined as time spent exploring the novel object/total interaction time × 100. An interaction was defined as active investigation of the object while the mouse was oriented toward and within 1 cm of the object. Mice with a total interaction time of less than 3 seconds were excluded from analysis.

Three-chamber test
Social communication behavior was performed using a three-chamber test as previously described [58]. The apparatus consisted of a plexiglass chamber comprising three equally sized compartments, where the mouse was allowed to explore three chambers using two sliding doors. Smart 3.0 software (Panlab) was used to record and analyze mouse behavior. Two wire cages were placed in the left and right chamber, respectively. The test consisted of three phases: (1) During habituation, the test mouse was placed in the center chamber and allowed to freely explore chambers for 5 minutes. (2) Sociability tests were performed in which an unfamiliar mouse (stranger 1) was introduced into a wire cage in one side of the chamber as the stimulus mouse. Location of stranger 1 in the left-or right-side chamber was evenly balanced across subjects. The test mouse was allowed to freely explore the chambers for 10 minutes. The time sniffing the wire cage with stranger 1 or the empty cage was recorded. (3) In the social novelty test phase, a new unfamiliar mouse (stranger 2) was placed into the empty cage, and then the test mouse was allowed to freely explore each chamber for 10 minutes. The time sniffing the wire cage with stranger 1 or stranger 2 was recorded.

USV recording
Isolation-induced USV in infant pups was assessed as previously described [58] using hardware and software provided by Avisoft Bioacoustics. At 10 minutes prior to initiating the experiment, the dam was separated from its litter to create a stable pretest baseline. P7 pups were isolated from their housing cages and individually placed in a recording box situated in an anechoic chamber [8,9]. An ultrasonic microphone was positioned approximately 5 cm above the pup and set to record vocalizations of 50-100 kHz with a 40-db cutoff over a 5-minute recording period.

Statistical analysis
Statistical analyses were performed with GraphPad Prism. The data distribution was assessed by a Kolmogorov-Smirnov nonparametric test of equality. Differences between two means were assessed by Student t test. Differences among multiple means were assessed, as indicated, by one-way, two-way, or repeated-measures ANOVA followed by Bonferroni's, Dunnett's, or Tukey's post hoc analysis. Error bars represent the SEM. Null hypotheses were rejected at P > 0.05 (P < 0.05 was considered to be statistically significant).