Ppp4r3a deficiency leads to depression-like behaviors in mice by modulating the synthesis of synaptic proteins

Major depressive disorder (MDD) is one of the most common and disabling mental disorders, affecting an estimated 350 million people globally and causing enormous personal suffering and societal economic burdens (Whiteford et al., 2013; Otte et al., 2016). MDD can be a lethal illness because of the increased risk of suicide, cardiac disease and cerebrovascular disease (Walker et al., 2015). Current pharmacotherapies [selective serotonin reuptake inhibitors (SSRIs)] for MDD require a prolonged time (weeks, if not months) to produce clinical improvement (Rush et al., 2006). The high incidence of partial or no response, along with the delayed onset of antidepressant effects, leaves many patients with inadequate treatment (Rush et al., 2006). A single subanesthetic dose of ketamine can induce rapid (within hours) and robust (up to 7 days) antidepressant responses in patients (Berman et al., 2000). However, the risks of neurotoxicity, cystitis, dissociative side effects and abuse of ketamine make it difficult for this treatment to gain widespread acceptance for clinical use (Bonaventura et al., 2021). Thus, identifying genetic changes related to depressive disorders and finding novel antidepressant targets are prospective directions for MDD therapy.

Consistent evidence from studies on the etiology and treatment of MDD are suggestive of the central role of homeostatic control of synaptic connections (Duman and Aghajanian, 2012; Duman et al., 2016). Previous postmortem studies demonstrated decreased expression of synapse-related genes, reduced number of synapses and dendritic complexity, associated with neuronal atrophy and decreased brain volume in MDD patients (Drevets, 2000; MacQueen et al., 2008; Kang et al., 2012). Chronic unpredictable mild stress (CUMS) in rodents is a putative model of depression (Willner et al., 1987; Kallarackal et al., 2013). Studies on rodents have revealed decreased levels of synaptic proteins, reduced synaptic density, neuronal atrophy and cell loss in depression-like mice under chronic stress (Radley et al., 2006; Kang et al., 2012). Recent reports suggest that the rapid antidepressant actions of ketamine depend on fast induction of synaptogenesis, which can persist for a long time and can, therefore, reverse the synaptic deficits and anhedonia caused by chronic stress (Berman et al., 2000; Zarate et al., 2006). Regulation of synaptic protein synthesis is the cornerstone for homeostatic control of synaptic connections; it is effected by complex signaling molecules, including neurotransmitters, cytokines, growth and neurotrophic factors, energy and metabolic factors, sex steroids and the hypothalamic–pituitary–adrenal axis (Duman and Aghajanian, 2012). The mammalian target of rapamycin complex 1 (mTORC1) cascade is the convergent downstream pathway that directly drives the translation of synaptic proteins (Sonenberg and Hinnebusch, 2009). However, the upstream mediator involved in sensing of complex signaling changes and in modulating the synthesis of synaptic proteins under chronic stress remains unclear.

Protein phosphatase 4 (PP4), which belongs to the PP2A-type phosphatases (including PP2AC, PP4C and PP6C), is a protein complex composed of a catalytic subunit PP4C and regulatory subunits (Gingras et al., 2005). PP2A was reported to participate in various neurodegenerative diseases, and, in particular, inhibition of PP2A activity was found in sporadic Alzheimer's disease (AD), focal cerebral ischemia and traumatic brain injury (Koh, 2011; Sontag and Sontag, 2014; Shultz et al., 2015). Notably, PP2A inhibition attenuated depression-like symptoms in a learned helplessness model of depression in mice (Lecca et al., 2016). Protein phosphatase 4 regulatory subunit 3A (PPP4R3A), also known as suppressor of MEK null 1 (SMEK1), is a regulatory subunit of PP4 enzyme. In Dictyostelium, smek is necessary for defects in polarity, chemotaxis speed and directionality in mek1-null and erk1-null cells, which accounts for the name smek (Mendoza et al., 2005). PPP4R3A has been reported to play an important role in the nervous system. PPP4R3A deficiency prevented neural stem cells from differentiating, leading to neurogenesis deficits during mouse cortical development (Lyu et al., 2013; Chang et al., 2017; Moon et al., 2017). A protective PPP4R3A variant, rs2273647-T, was identified in AD patients that helps to slow the onset and progression of AD pathology via reduced glucose uptake by the brain (Christopher et al., 2017). In addition, PPP4R3A plays an important role in the maintenance of embryonic stem cell pluripotency, hepatic gluconeogenesis, glucose metabolism, tumor suppression, tumor angiogenesis, microRNA biogenesis and transcription initiation (Yoon et al., 2010; Lyu et al., 2011; Byun et al., 2012; Dong et al., 2012; Ma et al., 2014; Su et al., 2017; Sen et al., 2020; Wang et al., 2021). Some neurogenesis-related or neurodegenerative disorder-related genes have been proven to be involved in depressive disorders (Choi et al., 2018; Ntim et al., 2020). However, the role of PPP4R3A in emotional regulation is poorly understood.

In the present study, we noticed reduced expression of PPP4R3A in CUMS-exposed depression-like mice. To determine whether Ppp4r3a is involved in depression and is related to the synthesis of synaptic proteins, we analyzed the behavioral effects and synaptic basis in Ppp4r3a-knockout and -overexpressing (OE) mice. Importantly, we endeavored to reveal the role of Ppp4r3a in chronic stress-induced depression and to find the bridge by which Ppp4r3a regulates the synthesis of synaptic proteins. In summary, we concentrated on the regulatory mechanism of synaptic synthesis in depression and investigated the molecular basis of depression to find possible novel therapeutic targets.

To study the role of PPP4R3A in the central nervous system, we analyzed its expression profile in the brains of wild-type mice. Specific cohorts of mice used for analysis are shown in Fig. S1. In adult wild-type (WT) mice, PPP4R3A was widely expressed in regions of brain, including the olfactory bulb, PFC, hippocampus, striatum, substantia nigra, hypothalamus, medulla oblongata and cerebellum (Fig. S2A). Next, we detected PPP4R3A expression in tissues from WT mice at postnatal day (P)0, P3, P5, P7, P10, P14, P21 and P28, and observed that expression in the hippocampus progressively decreased during development and became stable after P14 (Fig. S2B). Nuclear localization of PPP4R3A was confirmed by co-immunostaining of PPP4R3A with NeuN (also known as RBFOX3), GFAP and IBA1 in brain slices of adult WT mice, revealing PPP4R3A immunoreactivity in neurons, astrocytes and microglia, respectively (Fig. S2C).

In rodent models of CUMS, exposure to unpredictable stress for several weeks results in anhedonia and synapse defects, which are core symptoms of depression (Willner et al., 1987; Kallarackal et al., 2013). We exposed WT male C57BL/6 mice to 28 days of CUMS and validated the effects on depression using behavioral tests, including the sucrose preference test (SPT), forced swim test (FST), tail suspension test (TST), open field test (OFT) and elevated plus maze (EPM) test (Fig. 1A). To ensure the successful depression modeling of CUMS, all mice were subjected to the behavioral tasks. Detailed experimental procedures for behavioral tests are provided in the Supplementary Materials and Methods. As expected, CUMS-exposed mice showed reduced preference for sucrose in the SPT trials, which assessed anhedonia of mice (Fig. 1B). In the FST and TST trials, assessing behavioral despair in an escapable and struggling environment (indicating depression-like behavior), CUMS-exposed mice showed increased immobile time (Fig. 1C,D). In OFT trials, CUMS-exposed mice showed reduced time spent in the center zone, indicating anxiety-like behavior (Fig. 1E). In EPM trials, CUMS-exposed mice showed decreased exploratory time spent in the open arms, also a clue of anxiety-like behavior (Fig. 1F). These behavioral changes confirmed successful depression modeling in mice after CUMS exposure. Although carryover effects might emerge in mice after a series of behavioral tasks, we exposed CUMS and non-CUMS mice to identical procedures, thus minimizing the potential confounding effects for subsequent analysis.

Next, we isolated the PFC and hippocampus from CUMS-exposed and control mice to examine the expression levels of synaptic proteins. The results suggested reduced levels of GluA1 (also known as GRIA1) and PSD95 (also known as DLG4), which are needed for the formation and maturation of synapses, in the PFC and hippocampus (Fig. 1G,H). Importantly, we found that the PPP4R3A level was significantly reduced in the PFC and hippocampus in CUMS-exposed mice using western blot and immunostaining analyses (Fig. 1I-K). Thus, our data demonstrate reduced expression of PPP4R3A in CUMS-exposed mice, suggesting its possible involvement in the development of depression-like behaviors in mice.

To explore the effects of Ppp4r3a in the cortex and hippocampus, we specifically deleted the gene in the cortex and hippocampus by crossbreeding Emx1-IRES-Cre mice with Ppp4r3a^flox/flox mice (Fig. 2A; Fig. S3A). Effective knockdown of PPP4R3A in the cortex and hippocampus of mice was evident in western blot analyses (Fig. 2B,C). The Emx1 conditionally knockout (EcKO) mice had normal birth rate, survival rate and body weight compared to those of WT mice. The whole brain size and weight of EcKO mice showed no apparent differences from those of the WT mice (Fig. S3B,C). Additionally, no appreciable differences in cortical thickness or hippocampal size or number of neurons were observed by Nissl staining (Fig. S3D).

Next, we performed a global analysis of depression-like, anxiety-like, locomotion, spatial learning and memory behaviors of mice. Given the basal motional endurance differences between female and male mice in behavioral tests, we obtained results for female and male EcKO mice, and similar trends were shown for both sexes in all the behavioral tests. Adult 8-week-old EcKO mice spontaneously showed diminished sucrose preference during the SPT trials, and increased immobility time in the FST and TST trials, indicating depression-like behavior (Fig. 2D-F). In addition, EcKO mice showed reduced interest in exploring the open arms in EPM trials and decreased time spent in the center zone in OFT, indicating anxiety-like behavior (Fig. 2G,H). Normal locomotion activity of EcKO mice was evident from the similar movement distance in OFT trials compared with that of control mice and normal motor endurance in rotarod tests (Fig. 2H,I). By Barnes maze tests with adult 8-week-old EcKO and control mice, normal spatial learning and memory skills were observed in finding targets in EcKO mice (Fig. 2J). Fear conditioning is a form of learning through which individuals learn the relationships between aversive stimuli and other conditions, and is implicated in the pathogenesis of anxiety. In particular, anxious individuals are hypothesized to show stronger fear learning that is more resistant to extinction (Uys et al., 2003; Indovina et al., 2011). Thus, we performed context- and cue-associated fear conditioning tests on adult 8-week-old EcKO and control mice to further assess their anxiety-like traits. During the training process, EcKO mice showed increased reactivity to the first electric shock and displayed similar levels of freezing after three repeats of the training. During the testing process, EcKO mice showed similar memories for context-induced fear and strengthened memories for cue-induced fear (in new compartment) (Fig. 2K). Thus, the increased reactivity for sudden fear stimuli and stronger learning memory for fear suggested an anxiety-like trait in the EcKO mice. In addition, hot plate tests were performed to assess the sensory difficulties of mice. The results showed no differences in foot withdrawal latency in adult 8-week-old EcKO mice compared with control mice, indicating normal sensory functions (Fig. S4G). In summary, our results demonstrate that loss of PPP4R3A in the cortex and hippocampus led to elevated anxiety- and depression-like behaviors in mice.

To explore the synaptic basis in the hippocampus underlying aberrant depression- and anxiety-like behaviors in EcKO mice, we recorded the miniature excitatory postsynaptic currents (mEPSCs) in the hippocampal CA1 pyramidal neurons (Fig. 3A,B). The results demonstrated reduced amplitude of mEPSCs in EcKO mice, suggesting decreased expression of postsynaptic receptors (Fig. 3C). The frequency of mEPSCs in EcKO mice was reduced, indicating decreased release of the presynaptic excitatory neurotransmitter, glutamate (Fig. 3D). Next, we investigated whether the impairment of excitatory synaptic transmission was associated with alterations in the number of synapses and expression of glutamate receptors. Golgi staining showed reduced dendritic spine density in the hippocampus in EcKO mice (Fig. 3E). Notably, decreased expression of several NMDA receptor and AMPA receptor (AMPAR) subunits [GluA1, GluN1 (also known as GRIN1) and GluN2A (also known as GRIN2A)] and several excitatory synapse-related proteins [vGlut1 (also known as SLC17A7), synaptophysin and PSD95] were observed in KO mice (Fig. 3F). No changes in expression of markers of inhibitory synapses (vGAT1, officially known as SLC32A1) and excitatory neurons (Camk2a) were observed (Fig. 3F). The decreased GluA1 and PSD95 levels in EcKO mice were further confirmed by immunostaining of the hippocampus sections (Fig. 3G). In summary, Ppp4r3a deficiency leads to synaptic deficits in the hippocampus, characterized by reduced expression of synapse-related proteins, decreased dendritic density and impairment of excitatory glutamate transmission.

The alteration of synaptic protein synthesis by regulating cap-dependent mRNA translation is a crucial pathway for rapid antidepressant response (Sonenberg and Hinnebusch, 2009; Aguilar-Valles et al., 2021). Thus, we examined the global protein synthesis in the brain of WT and EcKO mice using a technique called surface sensing of translation (SUnSET) (Schmidt et al., 2009). We injected puromycin into the lateral ventricle of mice to label newly synthesized polypeptides in vivo (Fig. 4A). The levels of puromycin-labeled proteins in the hippocampus were significantly decreased in EcKO mice (Fig. 4B,C). Thus, Ppp4r3a deficiency led to lower protein synthesis, consistent with the decrease in GluA1 and PSD95 levels in the hippocampus of EcKO mice. To further verify that the diminished expression of synaptic proteins was because of reduced protein synthesis and not excessive protein degradation, we constructed PPP4R3A-OE and PPP4R3A-deficient SH-SY5Y cell lines to analyze the relationship of PPP4R3A with GluA1 and PSD95. Notably, upon treatment with MG132 (an inhibitor of proteasome-mediated protein degradation), the positive regulatory effect of PPP4R3A on GluA1 and PSD95 remained unchanged (Fig. S5A). Upon treatment with cycloheximide (an inhibitor of protein synthesis), protein degradation remained unchanged (Fig. S5B). Therefore, Ppp4r3a positively drives synaptic protein synthesis by triggering mRNA translation in the hippocampus.

Activation of mTORC1 is critical for the initiation of cap-dependent mRNA translation via the phosphorylation of two key effectors, eIF4E-binding proteins (4E-BPs) and p70S6 kinase 1 (S6K1) (Hoeffer and Klann, 2010) (Fig. 4D). Diminished mTORC1 signaling in the hippocampus of Ppp4r3a-knockout mice was indicated by reduced levels of phosphorylated mTOR, S6K1, eIF4EBP1 and eEF2 (Fig. 4E,F). The phosphorylation level of eIF2α, an inhibitor of mRNA translation initiation through an mTORC1-independent pathway, was not changed (Fig. 4E,F). In addition, the regulatory effect of PPP4R3A on mTORC1 was further verified in PPP4R3A-OE or shPPP4R3A SH-SY5Y cells (Fig. S5C). ERK (also known as MAPK) and PKB/AKT signaling was also detected in the PPP4R3A-OE SH-SY5Y cells, which are known to be critical for the activation of mTORC1 and need further analysis in the future (Fig. S5D). In summary, PPP4R3A drives the synthesis of synaptic proteins by activating the mTORC1 cascade.

To investigate whether Ppp4r3a overexpression could rescue depression-like behavioral defects and synaptic deficits in CUMS-induced mice, we overexpressed Ppp4r3a in the cortex and hippocampus of mice by crossbreeding ROSA26-Ppp4r3a^flox/flox mice with Emx1-IRES-Cre mice (Fig. 5A; Fig. S2E). Effective overexpression of PPP4R3A in the cortex and hippocampus of Ppp4r3a-OE mice was observed using western blot and immunostaining analyses (Fig. 5B,C). After 28 days of CUMS, Ppp4r3a-OE mice showed resistance to CUMS-induced depression-like behaviors compared with control mice, as confirmed by normal sucrose preference in the SPT and normal immobile time in the FST or TST (Fig. 5D-F). Next, we focused on the reduced mTORC1 signaling and synaptic proteins in CUMS-induced mice. As expected, Ppp4r3a-OE mice exhibited attenuated CUMS-induced downregulation of mTORC1 signaling and disrupted synaptic protein synthesis in the hippocampus, consistent with the protective effects in behavioral tests (Fig. 5G; Fig. S6A). Thus, replenishing PPP4R3A in the cortex and hippocampus was sufficient for correcting defective depression-like behaviors and disrupted synthesis of synaptic proteins caused by CUMS.

Using cultured primary hippocampal neurons from the Ppp4r3a-OE mice, we further validated the positive regulatory effect of Ppp4r3a on synaptic protein synthesis via mTORC1-mediated mRNA translation (Fig. 5H,I). Rapamycin (an inhibitor of mTORC1) treatment of Ppp4r3a-OE hippocampal neurons blocked the positive regulation of synaptic protein synthesis by Ppp4r3a through the direct blocking of mTORC1 (Fig. 5J; Fig. S6B). In addition, rapamycin treatment of PPP4R3A-OE SH-SY5Y cells also restored increased synthesis of GluA1 and PSD95 by inhibiting mTORC1 directly (Fig. S6C). Together, our findings suggest a significant role of PPP4R3A in driving synaptic protein synthesis and in resisting depression-like behaviors in chronic stress by activating the mTORC1 cascade.

Chronic stress is a potential trigger for MDD. In our study, in CUMS-exposed depressive mice, diminished expression of PPP4R3A was observed in the PFC and hippocampus, suggesting the involvement of Ppp4r3a in the development of depression. Furthermore, mice with Ppp4r3a deficiency in the cortex and hippocampus mimicked CUMS-induced depression- and anxiety-like behaviors. The overexpression of Ppp4r3a in the cortex and hippocampus of mice rescued CUMS-induced depression-like behaviors, indicating that Ppp4r3a is a protective factor under chronic stress. Together, these findings suggest prospects for Ppp4r3a in depression therapy.

The significant role of increased synaptic protein synthesis in antidepressant therapy has been previously noted. Many factors have been related to the disrupted expression of synapse-related proteins, which increases the susceptibility to depression. For identifying novel targets for antidepressant therapy, it is essential to clarify the regulatory mechanism of synaptic protein synthesis. The mTORC1 cascade is the known downstream convergence pathway crucial for synthesis of new proteins and synaptogenesis, and it regulates the initiation and elongation of mRNA translation (Sonenberg and Hinnebusch, 2009; Hoeffer and Klann, 2010). Previous postmortem studies in MDD patients indicated an association between deficits in mTOR-dependent mRNA translation and deficits in synaptic protein synthesis in the PFC (Feyissa et al., 2009; Jernigan et al., 2011). Preclinical and clinical studies have shown that the molecular basis for the rapid and robust antidepressant effects of ketamine is the acute activation of mTORC1, which leads to persistent synaptogenesis and behavioral relief (Li et al., 2010). Another study showed that 4E-BP1 and 4E-BP2, the key downstream effectors of mTORC1, are central to the antidepressant activity of ketamine (Aguilar-Valles et al., 2021). In addition, mTORC1 activation is the major effect of chronic treatment with conventional antidepressant drugs and SSRIs (Dagestad et al., 2006; Liu et al., 2015). Therefore, clarifying the role of an endogenous genetic inducer of the mTORC1 cascade in modulating synaptic protein synthesis is meaningful for antidepressant therapy.

Herein, we provide direct evidence that Ppp4r3a deficiency impairs synaptic protein synthesis and synaptogenesis and consequently results in depression-like behaviors by inhibiting the mTORC1 cascade. Examination of the synaptic basis in Ppp4r3a-deficient mice revealed the dysfunction of hippocampal synapses, including disrupted synthesis of synaptic proteins, reduced dendritic spine density and diminished excitatory glutamate synaptic transmission. Ppp4r3a deficiency in mice decreased mTORC1 signaling and reduced synaptic protein synthesis in the hippocampus, in accordance with impairment of hippocampal excitatory synaptic function and depression-like behaviors. The overexpression of Ppp4r3a in mice attenuated downregulation of mTORC1 and disrupted synaptic protein synthesis, thus leading to relief in depression-like behaviors. Rapamycin treatment of Ppp4r3a-OE hippocampal neurons blocked the positive regulatory effect of Ppp4r3a on synaptic protein synthesis by directly inhibiting mTORC1. Overall, our findings show that Ppp4r3a is a genetic inducer of mTORC1 and synaptic protein synthesis.

The activation of ERK and PKB/AKT is related to the activation of mTORC1 (Bockaert and Marin, 2015). Ketamine administration activates ERK, PKB/AKT and mTORC1 signaling in mice, and treatment with an ERK inhibitor (U0126), a PI3K inhibitor (LY294003) or rapamycin blocks ketamine-induced increases in mTORC1 phosphorylation and synaptogenesis (Li et al., 2010). Studies on the role of mTORC1 in the antidepressant effects of ketamine and its two enantiomers have shown that (S)-ketamine requires mTORC1, but (R)-ketamine requires ERK, activation (Yang et al., 2018). In in vitro experiments, we identified activation of PKB/AKT and ERK signaling in PPP4R3A-OE cells. The specific mechanism of triggering of mTORC1 signaling by PPP4R3A to produce rapid antidepressant responses needs to be further investigated in the future.

Intriguingly, we noticed the positive regulatory effect of PPP4R3A on GluA1 and PSD95, which are needed for the maturation of existing synapses and for the formation of new ones. GluA1, a very important subunit of AMPARs, induces the trafficking and integration of AMPARs (Zhang and Abdullah, 2013). Previous studies have shown that adult hippocampal long-term potentiation was absent in GluA1-deficient mice, but this synapse modification in CA1 was not required for spatial learning (Zamanillo et al., 1999; Mack et al., 2001). Our findings that Ppp4r3a-deficient mice demonstrate synapse dysfunction but display normal spatial learning memory further validates this viewpoint. Increased expression and phosphorylation of GluA1 are common therapeutic targets for the treatment of depression, schizophrenia, chronic drug addiction and AD (Eastwood et al., 1997; Duman et al., 2019; Qu et al., 2021). Moreover, increased GluA1 expression causes pain and is involved in epilepsy (Su et al., 2015; Joshi et al., 2018). PSD95, the principal scaffold protein in the postsynaptic density, regulates the anchorage, trafficking and localization of glutamate receptors, thus playing a role in MDD (Ehrlich and Malinow, 2004). Other studies have shown that the aberrant expression of PSD95 is associated with schizophrenia, AD, Huntington's disease, autism spectrum disorders and pain disorder (Arbuckle et al., 2010; de Bartolomeis et al., 2014; Savioz et al., 2014; Zhang et al., 2014). Thus, the positive regulatory effect of PPP4R3A on GluA1 and PSD95 indicates that PPP4R3A might play a more extensive role in psychiatric or neurodegenerative disorders and needs to be explored further.

In summary, PPP4R3A is a positive genetic regulator of synaptic protein synthesis, which activates the mTORC1 cascade, thereby playing an essential role in hippocampal synaptogenesis, maintenance of dendritic spine density and excitatory glutamate synaptic transmission. PPP4R3A is involved in chronic stress-induced depression and is a promising biomarker and therapeutic target for depressive disorder.

All animal experiments were conducted according to standard research protocols approved by the Animal Care and Use Committee, School of Basic Medical Sciences, Shandong University. Animals were group housed and maintained on a 12 h light/dark cycle at a constant temperature of 22°C, with free access to food and water. To analyze the expression pattern of PPP4R3A in mice, WT C57BL6/J mice were used for western blotting and immunostaining. To demonstrate the role of PPP4R3A in CUMS-induced depression, adult litter/age-matched male C57BL/6J mice (8 weeks old) were used for CUMS modeling and subsequent behavioral tests, immunostaining and western blot analysis. Then, to evaluate the impact of PPP4R3A deficiency in vivo, Ppp4r3a knockout mice were used for behavioral analysis, mEPSCs trials, Golgi staining and protein synthesis analysis. To further verify the role of PPP4R3A under CUMS pressure, Ppp4r3a-OE mice were used for behavioral analysis and protein synthesis analysis under CUMS modeling. Detailed cohorts of mice used for specific experimental analysis are shown in Fig. S1. All behavioral studies were conducted by an experimenter blinded to the genotypes of the mice.

We used a modified CUMS procedure described by Jaggar et al. (2017) and Nasca et al. (2013). C57BL/6 WT male mice (8 weeks old) were divided into control (n=10) and CUMS (n=15) groups. Mice in the CUMS group were maintained in individual cages and continuously received the following stress for 28 days: (1) 12 h water deprivation and 2 h exposure to an empty bottle; (2) 12 h food deprivation; (3) 12 h exposure to a wet cage (21°C water, filled 2 cm above the cage bottom); (4) 2 h of physical restraint in stainless steel retainers; (5) 12 h of cage tilt (45°); (6) 24 h of light/dark inversion; (7) 5 min of forced swimming at 4-8°C; (8) cold exposure at 4°C for 45 min; (9) overcrowding without bedding for 3 h (five mice in a 1 l beaker, 10 cm×13.5 cm); (10) cage exchange for 3 h (empty cage of other mice). The detailed procedure for 28 days is presented in Table S1. Mice were subjected to one or two stressors per day with an adequate interval for recovery. The above stressors were carried out randomly every week, and the behavioral tests were conducted 24 h after the last stressor to assess the depressive-like behaviors. The timeline of the behavioral tests is shown in Fig. 1A.

Ppp4r3a selectively knockout mice, EcKO mice, were generated following a standard Cre-LoxP recombination strategy. Ppp4r3a^flox/flox mice were generated by Cyagen Biosciences Inc. (Guangzhou, China). The targeting vector included a Rox-flanked Neo cassette, two homology arms and two loxP loci. The linearized vector was delivered to embryonic stem cells through electroporation, followed by drug selection, PCR screening and Southern blot verification. Positive F0 chimera mice were crossed with WT mice to generate F1 heterozygous Ppp4r3a^flox/+ mice. Emx1-IRES-Cre (The Jackson Laboratory, Stock No. 005628) mice have an Emx1 promoter driving Cre recombinase expression in ∼88% of the neurons of the neocortex and hippocampus. Thus, we crossbred Ppp4r3a^flox/flox mice with Exm1-IRES-Cre mice to selectively knock out Ppp4r3a in the cortex and hippocampus.

Ppp4r3a-OE mice (C57BL/6) were generated using the standard Cre-LoxP system. ROSA26-Ppp4r3a^flox/flox mice were generated by Biocytogen Corporation (Beijing, China) using the Biocytogen Extreme Genome Editing system. The targeting vector included a CAG promoter, a loxP-Stop-loxP transcriptional stop element and a Ppp4r3a CDS-HA tag-WPRE-polyA element. The single-guide RNA (sgRNA) was designed near the ROSA26 insertion site. Cas9/sgRNA and the targeting vector were microinjected into the fertilized eggs of mice to obtain chimeric F0 mice. F1 mice were obtained by crossing positive F0 mice with WT mice. ROSA26-Ppp4r3a^flox/flox mice were mated with Emx1-IRES-Cre mice to overexpress Ppp4r3a in the cortex and hippocampus of mice.

The SPT is the gold standard test to assess depression-like behavior in rodents by assessing anhedonia (core symptom of depression). The FST and TST assess despair behavior by measuring immobile time in an inescapable environment, allowing a simple assessment of potential antidepressant activity. The OFT and EPM test are routinely used to assess anxiety-like behaviors in rodents, and the OFT is also used to provide a qualitative and quantitative measurement of exploratory and locomotor activity. The Barnes maze is a dry-land maze that is commonly used for measuring spatial learning and memory in rodents. The rotarod test is used for assessing general locomotion. Fear conditioning is a form of learning that is implicated in the pathogenesis of anxiety disorders, thus stronger fear learning and memory in mice indicate increased degree of anxiety. The hot plate test is used for assessing basal sensory–motor disturbance in mice. The behavioral tasks of mice were performed in the order represented in the flowchart in Fig. S1. Detailed experimental procedures are provided in the Supplementary Materials and Methods.

Postnatal 8-week-old EcKO mouse and their control littermates were prepared for mEPSC tests. In brief, hippocampal slices were prepared by coronal sectioning (300 μm thick) using a vibrating slicer (Leica, VT 1000 S, Germany) in oxygenated ice-cold cutting solution. The prepared slices were immediately transferred to artificial cerebrospinal fluid (ACSF) at 32°C and incubated for 30 min. Whole-cell patch-clamp recordings were performed in hippocampal CA1 cells. For recording, slices were continuously perfused with ACSF at a flow rate of 6 ml/min with a temperature at 31±1°C. Cells were visualized with a microscope (BX51-WI, Olympus, Tokyo, Japan). After adding 1 μM tetrodotoxin (TTX) and 10 μM bicuculline in the ACSF, mEPSCs were recorded at a holding clamping voltage at −70 mV. The experimental data were recorded using a MultiClamp 700B amplifier and Clampfit 10.6 software (both Molecular Devices, San Jose, CA, USA) and analyzed by Prism 8.0.2 (GraphPad Software, San Diego, CA, USA) to calculate the current amplitude and discharge frequency, reflecting the presynaptic glutamate release and expression of postsynaptic receptors.

Each mouse was dissected immediately after transcardial perfusion with 4% paraformaldehyde (PFA). The brain was isolated and post-fixed in 4% PFA overnight. Fixed tissues were serially sectioned (4 μm) after paraffin embedding. The sections were dewaxed with gradient xylene and anhydrous ethanol, antigen repaired with high temperature, blocked with 10% goat or donkey serum, and immunostained with specific antibodies. After immunostaining, the sections were panoramically scanned with a scanning microscope and were analyzed with CaseViewer 2.0 software (3D HISTECH, Budapest, Hungary). Some representative images were obtained through a fluorescence microscope (Olympus).

SH-SY5Y cells were obtained from our laboratory storage and were cultured in Dulbecco's modified Eagle medium containing 1% glutamate, 10% fetal bovine serum and maintained at 37°C with 5% CO₂ in a humidified atmosphere.

Data analysis was performed using GraphPad Prism 8.0. Data are displayed as mean±s.e.m. Statistical differences between two groups (non-CUMS versus CUMS or control versus EcKO or control versus PPP4R3A-OE) were generally analyzed using an unpaired Student's t-test. P<0.05 was considered statistically significant. Two-way ANOVA followed by Bonferroni post hoc test was applied for analyzing behavioral tests of Ppp4r3a-OE mice and control mice with or without CUMS and was applied for analyzing social behaviors. The Kolmogorov–Smirnov test was used to analyze the cumulative probability of mEPSC amplitude and interevent intervals.