Impact of NRSN2 deficiency on memory: Altered excitatory synaptic plasticity associated with reduced expression of NMDA receptor subunits and impaired LTP in the hippocampus

Abstract

Our earlier human studies identified NRSN2 (Neurensin-2), a neuronal-specific vesicular protein, as a candidate gene contributing to 20p13 microdeletion syndrome, yet the functional consequences of NRSN2 deficiency in the nervous system remain poorly understood. To explore the role of Nrsn2 in neurodevelopment and cognitive function, we utilized previously generated homozygous Nrsn2 knockout mice (Nrsn2^-/-) and performed a series of behavioral, morphological, and electrophysiological analyses. Behaviorally, Nrsn2^-/- mice exhibited mild locomotor impairment, as assessed by gait analysis at 4 and 8 weeks of age, as well as significant deficits in spatial learning and memory (Morris water maze) and fear memory (passive avoidance test) at 8 weeks. Morphometric analysis suggested no overt alterations in dendritic complexity or spine density in hippocampal CA1 pyramidal neurons or cerebellar Purkinje cells without developmental malformation. Electrophysiological recordings and immunoblotting analyses may reflect region-specific synaptic alterations. In the hippocampus, expression levels of the NMDA receptor subunits GluN1 and GluN2A were reduced at 4 weeks of age. Consistently, CA1 pyramidal neurons displayed decreased sEPSC frequency with unchanged amplitude under the conditions examined. In addition, an imbalance in hippocampal excitatory/inhibitory transmission was observed, as reflected by altered sEPSC frequency in the absence of changes in sIPSC frequency. In cerebellar Purkinje cells, GluA1-containing AMPA receptors were selectively downregulated, accompanied by reduced frequency and amplitude of sEPSCs and a selective decrease in sIPSC frequency, indicating both excitatory and inhibitory synaptic dysfunction in this region. Collectively, these findings indicate that Nrsn2 deficiency is accompanied by altered excitatory synaptic transmission and reduced long-term potentiation (LTP) at 8 weeks of age, despite preserved dendritic architecture as assessed by Golgi staining. These synaptic and plasticity deficits occur alongside the observed cognitive and motor impairments in Nrsn2^⁻/⁻ mice. This study provides a descriptive phenotypic characterization of Nrsn2 deficiency and offers initial insights into the neurobiological role of NRSN2 and its contribution to neurodevelopment, learning, and memory.

Introduction

20p13 microdeletion syndrome presents a complex genetic disorder characterized by developmental delays, motor delay, intellectual disability, seizure disorder and dysmorphic features [1–7]. Among the implicated genes, NRSN2 has been recognized as a key candidate gene [1, 2].

Neurensin-2 (NRSN2) belongs to the Neurensin gene family, and encodes a 28 kD vesicle membrane protein specially expressed in neural cells of vertebrate brain after birth [8]. It was initially discovered as a homolog of Neurensin-1 (Nrsn1) [8], with differing primarily in its lack of a microtubule-binding domain in the long C-terminus of Neurensin-2 [8]. While Nrsn1 has been associated with neurite extension during nervous development, regeneration and plasticity [9], as well as long-term memory formation [10], the precise role of NRSN2 remains largely unexplored. According to the Whole Cortex & Hippocampus - Smart-seq dataset from the Allen Brain Atlas, Nrsn2 is both expressed in glutamatergic neurons, such as those in the CA1 region and GABAergic neurons. Recent studies indicated that the expression and function of Nrsn2 in GABAergic interneurons are tightly regulated by the chromatin remodeler SMARCA3 [11]. The SMARCA3-Neurensin-2 pathway plays a fundamental role in maintaining synaptic homeostasis in GABAergic neurons under stress condition by modulating AMPA receptor (AMPAR) trafficking and localization. Mass spectrometry analysis has revealed strong interactions between Neurensin-2 and postsynaptic scaffold protein, Homer3, as well as Arpc1b, a component of the actin-related protein 2/3 (Arp2/3) complex. Furthermore, Nrsn2 knockdown has been shown to attenuate excitatory synaptic transmission, as revealed by reductions in both the frequency and amplitude of AMPA-mediated miniature post synaptic currents (mPSC) in Purkinje cells from Nrsn2 KO mice [11]. Interestingly, the transcription levels of Brain-Derived Neurotrophic Factor (BDNF) in the hippocampus were upregulated in Nrsn2 KO mice [11]. The neurotropic hypothesis posits that BDNF plays a crucial role in neuronal synaptic plasticity in the hippocampus and prefrontal cortex [12, 13]. These results suggest a potential link between NRSN2 and synaptic plasticity.

Synaptic plasticity refers to the activity-dependent modifications in the strength or efficiency of synaptic transmission, which are thought to play key roles in memory formation. Increasing evidence suggests that impairments in synaptic plasticity are implicated in several notable neuropsychiatric disorders [14]. These synaptic plasticity mechanisms can be modulated at the presynaptic site through alterations in neurotransmitter release, or at the postsynaptic site through changes in receptor quantity, type, or properties of neurotransmitter receptors [15]. Excitatory synapses contain α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptors (AMPAR) and N-methyl-D-aspartate receptors (NMDAR), both of which are responsible for transmitting signals and initiating long-term changes in synaptic transmission, such as long-term potentiation (LTP) and long-term depression (LTD) [15, 16].

To investigate whether NRSN2 is implicated in 20p13 microdeletion syndrome, we conducted a series of behavioral tests on Nrsn2 homozygous knockout mice (Nrsn2^-/-). Our results revealed that these mice exhibit impaired memory and mild motor incoordination. We further examined glutamate receptor expression and synaptic function to explore the synaptic alterations underlying these behavioral deficits.

Materials and methods

Animals

We used Nrsn2 knockout (Nrsn2 ^-/-) mice recently generated by CRISPR-Cas9 genome editing [17]. Genotypes were verified by PCR and Sanger sequencing. All animal experiments were approved by the Department of Laboratory Animal Science of Fudan University. All efforts were made to minimize the number of animals used. Mice were maintained under a standard 12 h/12 h light/dark cycle and socially housed. Both male and female wild-type (WT) and Nrsn2 knock-out mice (2-12 weeks old, all on a C57BL/6 J background) were used in this study. Mice were included if they maintained normal health status and completed all habituation and experimental procedures. Group size was selected based on previous reports in the same task. No blinding was done for group allocation during the experiments or when assessing the outcome.

Behavioral assays and other assays

Behavioral experiments were performed with age-matched mice (4 weeks and 8-12 weeks) during the light cycle. Mice were gently handled for at least 3 min/day for 7 days prior to testing. When subjected to multiple tests, mice were evaluated in the the following order: open field test (OF), Y maze (YM), rotarod test (RT), gait analysis (GA), Morris water maze (MWM), novel object recognition test (NOR), fear conditioning (FC) and passive avoidance (PA). Each mouse participated in no more than 4 behavioral tests. All tested mice were habituated in the testing room for at least 30 min before the start of experiments.

Electrophysiology

Whole-cell patch clamp recordings were performed on hippocampal CA1 pyramidal neurons and cerebellar Purkinje cells. Cutting solution: 93 mM NMDG, 20 mM HEPES, 25 mM glucose, 30 mM NaHCO₃, 10 mM MgSO₄, 2.5 mM KCl, 1.2 mM NaH₂PO₄, 0.5 mM CaCl₂, 12 mM NAC, 2 mM Thiourea, 5 mM Na-ascorbate and 3 mM Na-pruvate. Recovery solution: 126 mM NaCl, 18 mM NaHCO₃, 1.3 mM MgSO₄, 2.5 mM KCl, 1.2 mM NaH₂PO₄, 1.2 mM CaCl₂ and 11 mM glucose. Slices were subsequently maintained at room temperature until use. Internal solution: 140 mM K-gluconate, 2 mM MgSO₄, 1 mM EGTA₄, 10 mM HEPES, 0.1 mM Na₃-GTP (guanosine 5’-triphosphate), 2 mM CaCl₂ and 2 mM Na₂-ATP (adenosine 5’-triphosphate). The external solution for sIPSCs recordings contained 50 μM 2-amino-5-phosphonopentanoic acid (AP5) and 10 μM 6-cyano-2,3-dihydroxy-7-nitroquinoxaline (CNQX) to block NMDAR and AMPAR receptor synaptic currents, respectively. Exclusion criteria: Exclude cells with unstable seal (membrane resistance < 1 GΩ) or signal fluctuation exceeding 20% during recording. Details of behavior tests, Golgi staining, and Electrophysiology were described in supplementary material.

Western blotting

Western blotting was performed using standard protocols and multiple primary antibodies were used. For detecting postsynaptic proteins, the following antibodies were used: antibodies against β-Tubulin (10094-1-AP), β-actin (HRP-60008), PSD-95 (20665-1-AP), GluA1 (67642-1-Ig), GluA2 (11994-1-AP), GluN2A (28525-1-AP) (all from Proteintech, USA) and GluN1 (A7677) (from ABclonal, China). Protein detection was performed using enhanced chemiluminescence (ECL) on a Tanon 5200 Chemiluminescent Imaging System (Tanon, China). Band intensities were quantified using ImageJ, and the relative expression levels of target proteins were normalized to the grayscale value of β-tubulin or β-actin.

Statistical analysis

Each experiment was repeated at least three times. The analyses were conducted using GraphPad Prism 8, with a significance threshold set at p < 0.05. Data are presented as mean ± SD. For comparisons between two groups, results from the following assays were analyzed with two tailed Student’s t-test: Y maze, open field test, Rotarod test, gait analysis, dendritic spine density and immunoblot analysis. A two-way ANOVA was applied to Morris water maze training, passive avoidance, fear conditioning test and number of neuron intersections. The time spent in the target quadrant was analyzed by one-sample t-tests with the theoretical chance level (25% for quadrant time) as the test value. Differences of time spent in target quadrant between WT and Nrsn2^-/- groups were analyzed using a two tailed Student’s t-test. Mixed-effects Linear Model was applied to the frequency and amplitude of sPSCs. Data from the LTP recording were compared by Tukey’s test following two-way ANOVA. The sample size and p value for each graph was stated in figure legends. No statistical methods were used for sample size estimate.

Results

Nrsn2 Knockout exhibits impaired spatial reference and fear memory

To investigate the effects of Nrsn2 deletion on learning and memory, we performed a series of cognitive tasks using 8-week-old WT and Nrsn2^-/-mice. In the Morris Water Maze test, Nrsn2^-/- mice showed no difference in latency to locate the hidden platform during training, indicating intact spatial memory acquisition (Fig. 1A and B). However, in the probe test, they spent significantly less time in the target quadrant and had fewer platform crossings than WT mice, suggesting impaired long-term spatial memory (Fig. 1C–E). We also performed the Morris Water Maze test with 4-week-old WT and Nrsn2^-/- mice, but no significant difference was observed between two groups (Supplementary Figure 1). The Y-maze test revealed no differences in spontaneous alternation rates or the total number of arm entries (Supplementary Figure 2), suggesting that short-term spatial working memory remained unaffected. To assess fear memory, we performed fear conditioning and passive avoidance tests. We observed no significant differences of freezing time in fear conditioning test between WT and Nrsn2^-/- mice (Fig. 1F–H). Similarly, in the passive avoidance test, both genotypes exhibited comparable latencies to enter the dark box 24 h after training (Fig. 1I). However, after 3 days, Nrsn2^-/- mice displayed significantly reduced latencies to enter the dark box (Fig. 1I) and stayed longer in dark box than WT mice (Fig. 1J), suggesting impaired fear memory. Fear conditioning includes both cued and contextual components. Cued fear conditioning (tone stimulus–foot shock pairing) primarily assesses amygdala-dependent associative memory, with minimal involvement of the hippocampus. In contrast, contextual fear conditioning relies on both the amygdala and hippocampus for spatial encoding and contextual integration [18]. Passive avoidance test inherently requires hippocampus-mediated spatial-contextual memory (associating a specific location (dark chamber) with shock [19]. The discrepancy between these two sets of results suggests that Nrsn2 deficiency is more relevant to hippocampus-dependent memory formation. These results indicate that Nrsn2 deletion specifically impairs long-term spatial and fear memory while leaving short-term fear memory and spatial memory acquisition intact.

Fig. 1. Nrsn2 Knockout exhibits impaired spatial reference and fear memory.

A Representative swim tracks during the probe test. B The latency to reach the hidden platform during a 6-day training period for 8-week-old WT and Nrsn2^-/- mice in the MWM test. V1d indicates “visible platform at Day 1.” H3d indicates “Hidden platform at Day 3.” (WT (male: n = 6; female: n = 5); Nrsn2^-/- (male: n = 7; female: n = 5)). C The time spent in each of the 4 quadrants on day 7 for 8-week-old mice in the MWM test. Target indicates “the Target quadrant where the platform was set.” Opposite indicates “the Opposite quadrant to the target quadrant.” Left indicates “the Left quadrant of the target quadrant.” Right indicates “the Right quadrant of the target quadrant.”. Target: The 95% CI was 28.05–34.18 for WT and 24.15–29.48 for Nrsn2^-/-; Cohen’s d = 0.87. Opposite: The 95% CI was 14.49–17.85 for WT and 18.21–21.92 for Nrsn2^-/-; Cohen’s d = 1.27. D The swimming speed of 8-week-old mice in the MWM test on day 1. E The number of platform- crossing on day 7 for 8-week-old mice in the MWM test. The 95% CI was 1.92–3.24 for WT and 1.25–2.25 for Nrsn2^-/-; Cohen’s d = 0.83. F freezing rate (%) of WT and Nrsn2^-/- mice before training in fear conditioning test (WT (male: n = 9; female: n = 5); Nrsn2^-/- (male: n = 10; female: n = 5)). G Freezing rate (%) of WT and Nrsn2^-/- mice at 24 h, 72 h, and 1-week post-training without tone stimulus (TS) in the fear conditioning test. H freezing rate (%) of WT and Nrsn2^-/- mice at 24 h, 72 h, and 1 week post-training with tone stimulus (TS) in the fear conditioning test. I The latency to enter the dark chamber in the passive avoidance test for WT and Nrsn2^-/- mice at 24 h, 72 h, and 1 week post-training (WT (male: n = 9; female: n = 5); Nrsn2^-/- (male: n = 10; female: n = 5)). 72 h: The 95% CI was 377.92–583.52 for WT and 167.79–406.08 for Nrsn2^-/-; Cohen’s d = 0.89. J The time spent in the dark chamber in the passive avoidance test for WT and Nrsn2^-/- mice at 24 h, 72 h, and 1-week post-training. 72 h: The 95% CI was 1.67–19.58 for WT and 50.52–123.94 for Nrsn2^-/-; Cohen’s d = 1.43. Data are expressed as mean ± SD. Error bars indicate SD. 95% confidence intervals and Cohen’s d were provided as effect-size estimates. *P < 0.05, **P < 0.01, ***P < 0.001. Nrsn2^-/-: Homozygous Nrsn2 knockout mice.

Knockout of Nrsn2 leads to mild Locomotor Impairments

To assess locomotor function, we conducted an open field test, rotarod test and gait analysis on both Nrsn2^-/- and WT mice. In the open field test, 8-week-old Nrsn2^-/- mice showed no significant differences in total distance traveled or time spent in the center compared to WT mice (Fig. 2A, B). Similarly, in rotarod test, both 4-week-old and 8-week-old Nrsn2^-/- mice performed comparably to WT mice, with no significant difference in fall speed (Fig. 2C, D). Gait analysis revealed a mild locomotor impairment. Specifically, 4-week-old Nrsn2^-/- mice exhibited significantly lower maximum contact intensity in their hind feet compared to WT mice (Fig. 2E–H), indicating a subtle gait abnormality, but 8-week-old Nrsn2^-/- mice seemed unchanged (supplementary Figure 3 and 4). Additionally, 8-week-old Nrsn2^-/- mice exhibited a shorter swing duration than WT mice (Fig. 2I–L). These findings suggest that Nrsn2 deletion mildly impairs locomotor ability, particularly in younger mice.

Fig. 2. Knockout of Nrsn2 leads to mildly impaired locomotor ability.

A The total distance traveled by 8-week-old WT and Nrsn2^-/- mice in the Open field test (WT (male: n = 7); Nrsn2^-/- (male: n = 7)). B The time spent in center zone for 8-week-old WT and Nrsn2^-/- mice in the Open field test. C The speed at fall of 4-week-old WT and Nrsn2^-/- mice in the Rotarod test (WT (male: n = 6; female: n = 5); Nrsn2^-/- (male: n = 6; female: n = 6)). D The speed at fall of 8-week-old WT and Nrsn2^-/- mice in the Rotarod test. E The max contact intensity of 4-week-old WT and Nrsn2^-/- mice’s left forefoot (LF) in Gait analysis (WT (male: n = 8; female: n = 5); Nrsn2^-/- (male: n = 8; female: n = 6)). F The max contact intensity of 4-week-old WT and Nrsn2^-/- mice’s right forefoot (RF) in Gait analysis. G The max contact intensity of 4-week-old WT and Nrsn2^-/- mice’s left hind foot (LH) in Gait analysis. The 95% CI was 182.38–193.81 for WT and 171.05–184.25 for Nrsn2^-/-; Cohen’s d = 0.90. H The max contact intensity of 4-week-old WT and Nrsn2^-/- mice’s right hind foot (RH) in Gait analysis. The 95% CI was 179.52–190.83 for WT and 166.57–182.05 for Nrsn2^-/-; Cohen’s d = 0.84. I The swing time of 8-week-old WT and Nrsn2^-/- mice’s left forefoot (LF) in Gait analysis (WT (male: n = 6; female: n = 4); Nrsn2^-/- (male: n = 6; female: n = 6)). J The swing time of 8-week-old WT and Nrsn2^-/- mice’s right forefoot (RF) in Gait analysis. The 95% CI was 0.09–0.10 for WT and 0.08–0.09 for Nrsn2^-/-; Cohen’s d = 1.13. K The swing time of 8-week-old WT and Nrsn2^-/- mice’s left hind foot (LH) in Gait analysis. The 95% CI was 0.09–0.11 for WT and 0.08–0.10 for Nrsn2^-/-; Cohen’s d = 1.01. L The swing time of 8-week-old WT and Nrsn2^-/- mice’s right hind foot (RH) in Gait analysis. The 95% CI was 0.09–0.11 for WT and 0.08–0.10 for Nrsn2^-/-; Cohen’s d = 0.93. Data are expressed as mean ± SD. Error bars indicate SD. 95% confidence intervals and Cohen’s d were provided as effect-size estimates. *P < 0.05, **P < 0.01, ***P < 0.001. Nrsn2^-/-: Homozygous Nrsn2 knockout mice.

Neuron morphology and dendritic spine are unaltered in Nrsn2^-/- mice

To assess the impact of Nrsn2 deletion on neuronal morphology and dendritic spine density, we performed Golgi staining in hippocampal CA1 pyramidal neurons and cerebellar Purkinje cell of 4-week-old mice (Fig. 3A and F). Sholl analysis revealed no significant differences in dendritic complexity, as measured by the number of intersections varying radial distances from the soma center, in either CA1 pyramidal neurons (Fig. 3B, C) or Purkinje cells (Fig. 3G, H). Additionally, Spine density was similar between Nrsn2 ^-/- mice and WT (Fig. 3D and E). These findings indicate that Nrsn2 deletion does not affect neuronal morphology or dendritic spine density.

Fig. 3. Neuron morphology and dendritic spine was normal in Nrsn2^-/- mice.

A Representative image of CA1 pyramidal neuron of 4-week-old WT and Nrsn2^-/- mice after Golgi staining. B Number of intersections from soma in CA1 pyramidal neuron (WT n = 13, Nrsn2^-/- n = 12). C Total intersections within 10 to 100 μm from soma in CA1 pyramidal neuron (WT n = 13, Nrsn2^-/- n = 12). D Representative photomicrograph of dendritic spines of CA1 pyramidal neuron. E Dendritic spine density of CA1 pyramidal neuron in hippocampus of 4-week-old WT and Nrsn2^-/- mice (WT n = 13, Nrsn2^-/- n = 12). F Representative image of cerebellar Purkinje cells of 4-week-old WT and Nrsn2^-/- mice (WT n = 9, Nrsn2^-/- n = 9) after Golgi staining. G Number of intersections from soma in cerebellar Purkinje cells. H Total intersections within 20 to 100 μm from soma in cerebellar Purkinje cells (WT n = 9, Nrsn2^-/- n = 9). Data are expressed as mean ± SD. Error bars indicate SD. 95% confidence intervals and Cohen’s d were provided as effect-size estimates. *P < 0.05, **P < 0.01, ***P < 0.001. Nrsn2^-/-: Homozygous Nrsn2 knockout mice.

Reduced glutamate receptor expression in Nrsn2^-/- mice

To examine the molecular mechanisms underlying memory deficits in Nrsn2^-/- mice, we quantified synaptic protein expression in 4-week-old mice using Western blot (Fig. 4A). In the hippocampus, GluN1 and GluN2A levels were significantly reduced, whereas GluA1, GluA2, GABBR2 and PSD95 showed no difference (Fig. 4B). In the cerebellum, GluA1 expression was significantly decreased in Nrsn2^-/- mice, while GluA2, GluN1, GluN2A, GABBR2, and PSD95 levels remained unchanged (Fig. 4C, D).

Fig. 4. Reduced glutamate receptor expression in Nrsn2^-/- mice.

A Western blot analysis of expression of AMPA receptor (GluA1, GluA2), NMDA receptor (GluN1, GluN2A), GABBR2 and β-tubulin of the hippocampus in wild-type and Nrsn2 ^-/- mice (WT n = 3, Nrsn2^-/- n = 3). B Quantitative analysis of the expression levels of the proteins in A. Relative expression levels of specific proteins were normalized to the grayscale value of β-tubulin or β-actin. GluN1: The 95% CI was 0.53–0.86 for Nrsn2^-/-; Cohen’s d = 2.18. GluN2A: The 95% CI was 0.62–0.67 for Nrsn2^-/-; Cohen’s d = 1.96. C Western blot analysis of expression of AMPA receptor (GluA1, GluA2), NMDA receptor (GluN1, GluN2A), GABBR2 and β-tubulin of the cerebellar cortex in wild-type and Nrsn2 ^-/- mice (WT n = 3, Nrsn2^-/- n = 3). D Quantitative analysis of the expression levels of the proteins in C. GluA1: The 95% CI was 0.60–0.85 for Nrsn2^-/-; Cohen’s d = 1.51. Relative expression levels of specific proteins were normalized to the grayscale value of β-tubulin or β-actin. Data are expressed as mean ± SD. Error bars indicate SD. 95% confidence intervals and Cohen’s d were provided as effect-size estimates. *P < 0.05, **P < 0.01, ***P < 0.001. Nrsn2^-/-: Homozygous Nrsn2 knockout mice.

Knockout of Nrsn2 impairs post-synaptic transmission and long-term potentiation at CA3-CA1 synapses

Electrophysiological analysis of acute brain slices revealed impaired post-synaptic transmission in 4-week-old Nrsn2^-/- mice. Whole-cell patch-clamp recording showed decreased spontaneous excitatory postsynaptic current (sEPSC) frequency by 81.40% (Fig. 5A–C) with unchanged amplitude (Fig. 5D) in CA1 pyramidal neurons of hippocampus. In addition, we did not find any significant difference in sIPSCs in CA1 pyramidal neurons between Nrsn2^-/- mice and WT mice (Fig. 5E–G). In contrast, both sEPSC frequency and amplitude were decreased in cerebellar Purkinje cells by 61.36% and 25.34%, respectively, compared to WT neurons (Fig. 5H–K). Data analysis revealed a significant reduction in the frequency by 29.70%, but not amplitude of sIPSC in cerebellar Purkinje cells of Nrsn2^-/- mice (Fig. 5L–N).

Fig. 5. Knockout of Nrsn2 impairs post-synaptic transmission and long-term potentiation at CA3-CA1 synapses.

A Schematic of hippocampal sEPSC recording. B Representative traces of sEPSC in hippocampal CA1 pyramidal neurons of 4-week-old WT and Nrsn2^-/- mice. Scale of 10 pA and 100 ms is shown. C Statistical comparison of sEPSC frequency recorded in CA1 pyramidal neurons of 4-week-old WT and Nrsn2^-/- mice (WT n = 3 mice, 3–5 cells per mouse, Nrsn2^-/- n = 3 mice, 3–5 cells per mouse). The 95% CI was 1.06-4.93 for WT and 0.27-0.84 for Nrsn2^-/-; Cohen’s d = 0.98. D Statistical comparison of sEPSC amplitude recorded in CA1 pyramidal neurons of 4-week-old WT and Nrsn2^-/- mice (WT n = 3 mice, 3–5 cells per mouse, Nrsn2^-/- n = 3 mice, 3–5 cells per mouse). E Representative traces of sIPSC in hippocampal CA1 pyramidal neurons of 4-week-old WT and Nrsn2^-/- mice. Scale of 10 pA and 100 ms is shown. F Statistical comparison of sIPSC frequency recorded in CA1 pyramidal neurons of 4-week-old WT and Nrsn2^-/- mice (WT n = 3 mice, 2-3 cells per mouse, Nrsn2^-/- n = 3 mice, 2–3 cells per mouse). G Statistical comparison of sIPSC amplitude recorded in CA1 pyramidal neurons of 4-week-old WT and Nrsn2^-/- mice (WT n = 3 mice, 2-3 cells per mouse, Nrsn2^-/- n = 3 mice, 2-3 cells per mouse). H Schematic of cerebellar sEPSC recording. I Representative traces of sEPSC in cerebellar Purkinje cells of 4-week-old WT and Nrsn2^-/- mice. Scale of 10 pA and 100 ms is shown. J Statistical comparison of sEPSC frequency recorded in cerebellar Purkinje cells of 4-week-old WT and Nrsn2^-/- mice (WT n = 3 mice, 2-3 cells per mouse, Nrsn2^-/- n = 3 mice, 2-3 cells per mouse). The 95% CI was 5.74-12.08 for WT and 1.86-7.25 for Nrsn2^-/-; Cohen’s d = 4.59. K Statistical comparison of sEPSC amplitude recorded in CA1 pyramidal neurons of 4-week-old WT and Nrsn2^-/- mice (WT n = 3 mice, 2-3 cells per mouse, Nrsn2^-/- n = 3 mice, 2-3 cells per mouse). The 95% CI was 18.54-33.89 for WT and 14.58-25.20 for Nrsn2^-/-; Cohen’s d = 1.50. L Representative traces of sIPSC in cerebellar Purkinje cells of 4-week-old WT and Nrsn2^-/- mice. Scale of 10 pA and 100 ms is shown. M Statistical comparison of sIPSC frequency recorded in cerebellar Purkinje cells of 4-week-old WT and Nrsn2^-/- mice (WT n = 3 mice, 2-3 cells per mouse, Nrsn2^-/- n = 3 mice, 2-3 cells per mouse). The 95% CI was 15.09-20.41 for WT and 9.76-15.14 for Nrsn2^-/-; Cohen’s d = 1.58. N Statistical comparison of sIPSC amplitude recorded in CA1 pyramidal neurons of 4-week-old WT and Nrsn2^-/- mice (WT n = 3 mice, 2-3 cells per mouse, Nrsn2^-/- n = 3 mice, 2-3 cells per mouse). O Input–output curves constructed from the stimulation intensity vs fEPSP slope in the hippocampus from brain slices of 8-week-old WT and Nrsn2^-/- mice (WT n = 5 mice, 1-2 slices per mouse, Nrsn2^-/- n = 5 mice, 1-2 slices per mouse). P LTP at CA3-CA1 in 8-week-old WT and Nrsn2^-/- mice. Q Normalized fEPSP slope values (as a percentage of the baseline) in pyramidal neurons of 8-week-old WT and Nrsn2^-/- mice (WT n = 5 mice, 1-2 slices per mouse, Nrsn2^-/- n = 5 mice, 1-2 slices per mouse). The 95% CI was 122.81-158.59 for WT and 102.46-113.19 for Nrsn2^-/-; Cohen’s d = 2.18. R Average values of paired-pulse ratio in CA1 from 8-week-old WT and Nrsn2^-/- mice. Data are expressed as mean ± SD. Error bars indicate SD. 95% confidence intervals and Cohen’s d were provided as effect-size estimates. *P < 0.05, **P < 0.01, ***P < 0.001. Nrsn2^-/-: Homozygous Nrsn2 knockout mice. Data were analyzed at the animal level (mean value per mouse).

Further assessment of synaptic function at CA3-CA1 synapses showed impaired long-term potentiation (LTP) in 8-week-old Nrsn2^-/- mice. Input-output curves showed no significant differences between WT and Nrsn2^-/- mice (8 weeks) (Fig. 5O). In 8-week-old WT mice, LTP induction led to a significant increase in field excitatory post-synaptic potential (fEPSP) amplitudes at CA3-CA1 synapses. In contrast, Nrsn2^-/- mice failed to exhibit this increase under the same condition, indicating the impaired LTP (Fig. 5J). The potentiation of fEPSP was significantly reduced in the Nrsn2^-/- mice by 23.41% (Fig. 5P and Q). A paired-pulse ratio (PPR) test in CA1 (8 weeks) showed no significant difference between groups (Fig. 5R). Collectively, these findings demonstrate that Nrsn2 deletion leads to reduced spontaneous excitatory synaptic transmission in hippocampal CA1 pyramidal neurons and cerebellar Purkinje cells, is accompanied by a selective decrease in spontaneous inhibitory synaptic activity in Purkinje cells, and is associated with impaired hippocampal synaptic plasticity.

Discussion

In this study, we found that Nrsn2^-/- mice exhibited several behavioral deficits, including mild impairment in locomotor ability, spatial memory, and fear memory. Importantly, Nrsn2 knockout was associated with decreased glutamate receptor expression, abnormal sEPSC, and abnormal LTP in the hippocampus. These results provided initial insights into the molecular and cellular mechanisms underlying motor delay and intellectual disability in certain population of patients with 20p13 microdeletion syndrome, supporting a role of NRSN2 in synaptic function and cognitive processes.

Nrsn2 encodes a neuronal-specific vesicular protein involved in protein transport [8], with expression restricted to brain regions including cerebral cortex, thalamus, hypothalamus and hippocampus [1]. Although NRSN2 is a key candidate gene for 20p13 microdeletion syndrome, its function in neurodevelopment remains elusive. To address this, we performed behavior tests on Nrsn2 knockout mice, which are related to behaviors in patients with 20p13 microdeletion syndrome. Notably, we observed that Nrsn2 null mice exhibited impaired long-term spatial memory in spatial probe trial of the water maze test and fear memory in passive avoidance test, while other cognitive functions remained unaffected. Especially, Nrsn2 null mice showed normal performance in fear conditioning and the acquisition phase of the water maze test, and short-term spatial working memory in Y maze test. Age-dependent differences were observed in memory related behavior test that 4-week-old mice did not exhibit memory impairment phenotypes, whereas 8-week-old mice showed memory deficits in behavioral tests. This may be associated with the expression pattern of NRSN2 in mice: NRSN2 begins to be expressed after birth, continuously increases, and reaches its peak around 10 weeks of age [8, 20, 21]. The learning and memory impairments observed in Nrsn2-deficient mice are consistent with clinical findings reporting that over 50% of individuals with 20p13 microdeletion exhibit mild to moderate intellectual disability, supporting the notion that NRSN2 dysfunction may play a key role in the cognitive deficits associated with this syndrome.

While haploinsufficiency is a well-established mechanism underlying most microdeletion syndromes, the pathogenesis of 20p13 microdeletion syndrome remains unclear. In our study, heterozygous Nrsn2 knockout (Nrsn2^+/-) mice exhibited no overt behavioral abnormalities in development, open field test, Y maze test and rotarod test (Supplementary Figure 5), whereas homozygous knockout (Nrsn2^-/-) mice showed no overt abnormalities in these tests as well, in addition to mild motor deficits in the gait analysis and memory impairment in the Morris Water Maze test and passive avoidance test. Interestingly, a previous study also reported no significant baseline behavioral alterations in Nrsn2^-/- mice across the open field, social interaction, and sucrose preference tests (SPT) [11]. However, under social defeat stress, these mice displayed increased social avoidance, anhedonia, and anxiety-like behavior, indicating that the behavioral consequences of Nrsn2 loss may be context-dependent.

Although Nrsn2^+/- mice exhibit normal behavioral phenotypes in most assessment, human carriers with a single NRSN2 deletion often present with cognitive or motor impairments. This discrepancy between the heterozygous phenotype in mice and humans may reflect differences in genomic context, dosage sensitivity, or species-specific compensatory mechanisms. First, species differences between humans and mice could play a significant role. There is substantial evidence that heterozygous mice often show no significant phenotypic alterations or a subtle phenotype, whereas null mice exhibit clear deficits [22]. This difference may stem from variations in gene redundancy and compensatory mechanisms, genetic background, developmental plasticity, and species-specific gene functions that distinguish mice from humans. Second, a single-gene Nrsn2 KO in an otherwise intact genome but a megabase-scale 20pter deletion that removes NRSN2 plus dozens of neighboring genes in human patients (including SOX12 and NRSN2-AS). For instance, SOX12, which may be downregulated by NRSN2 [23], and the long non-coding RNA NRSN2-AS1 [24], could contribute to the observed phenotypes. Additionally, the combined effects of these genes may further modulate the behavioral and molecular outcomes for 20p13 microdeletion syndrome. Briefly, our findings indicate that Nrsn2 deficiency is associated with alterations in neurobehavioral phenotypes, and further studies are needed to dissect the contributions of Nrsn2 and its interacting genetic factors to the pathogenesis of 20p13 microdeletion syndrome.

Proper axonal and dendritic development is essential for establishing of synaptic circuitry. Experimental mouse model of cognitive impairment often displayed significant deficits in spine density of the CA1 region of hippocampus, including alterations in density, morphology, and remodeling [25–28]. In previous studies, Nrsn1, the homologous genes of Nrsn2, was reported to play a role in neurite extension during nervous development, regeneration and plasticity [9]. For instance, Purkinje cells transfected with cNrsn1 siRNA exhibited impaired axonal fibers development [29]. To investigate whether Nrsn2 plays a similar role in neural development, we performed Golgi staining on Nrsn2^-/- mice brains. No significant differences in neuronal morphology or spine density in hippocampus CA1 pyramidal neurons or cerebellar Purkinje cell. Unlike NRSN1, NRSN2 protein lacks the microtubule-binding domain found in microtubule-associated proteins such as MAP2 and TAU [8], indicating the potential functional difference in neural development.

Synaptic plasticity underlies learning and memory through mechanisms involving neurotransmitters, receptors, synaptic strength, and spike timing-dependent plasticity. Nrsn2 has been implicated in modulating AMPAR signaling in interneurons and interacts with postsynaptic scaffold proteins such as Homer and Arpc1b [11]. In our study, we observed a reduction in both the frequency and amplitude of spontaneous excitatory postsynaptic currents (sEPSCs) in cerebellar Purkinje cells (Fig. 5H–K), consistent with previous findings [11]. However, for spontaneous inhibitory postsynaptic currents (sIPSCs), we detected a decrease in frequency but not amplitude (Fig. 5L–N), which differs from recent reports [30]. This discrepancy may result from differences in the neuronal subtypes examined, which was measured in dentate gyrus granular cells in their study. To further explore the role of Nrsn2 in excitatory synaptic transmission—a key process linked to learning and memory—we recorded sEPSCs in hippocampal CA1 pyramidal neurons from 4-week-old Nrsn2^-/- mice. We found decreased sEPSC frequency with unchanged amplitude (Fig. 5A–D). A decrease in sEPSC frequency generally suggests reduced presynaptic release probability or fewer functional synapses, whereas unchanged amplitude indicates preserved postsynaptic receptor function [31].

Additionally, we found significantly reduced expression of the glutamate receptor subunits, including AMPA receptors (GluA1) in cerebellar Purkinje cells and NMDA receptors (GluN1 and GluN2A) in the hippocampus of Nrsn2^-/- mice, respectively. Given that NMDA receptors, particularly those in the Schaffer collateral pathway, are critical for long-term potentiation (LTP) [32, 33], a key mechanism of synaptic plasticity underlying learning and memory [34–36], we therefore investigated LTP induction in these mice. Consistent with previous reports showing that GluN2A^-/- mice exhibit reduced hippocampal LTP and spatial learning deficits [37], we observed that theta-burst stimulation (TBS) failed to induce LTP at CA3-CA1 synapses in 8-week-old Nrsn2^-/- mice (Fig. 5P, Q), unlike in wild-type controls. The paired-pulse ratio (PPR) results suggest that the LTP impairment may be more closely associated with postsynaptic plasticity-related molecular mechanisms than with the presynaptic probability of vesicular release. Notably, recent research highlighted that the antidepressant-like effects of broad NMDAR antagonists (e.g., ketamine and MK-801) are mediated through the suppression of GluN2A-type NMDA receptor [38]. Interestingly, Nrsn2 KO mice demonstrated resistance to social defeat stress without depressant-like behaviors [11], a phenotype that may be linked to their reduced GluN2A expression. Furthermore, our prior study suggested enhanced stress tolerance in adult Nrsn2^-/- mice was linked to reduced connectivity in various regions of the default mode network (DMN) [17]. Taken together, our findings show that Nrsn2 deficiency is accompanied by glutamatergic synaptic alterations, including reduced sEPSC frequency, decreased expression of key glutamate receptor subunits (GluA1, GluN1, GluN2A), and impaired LTP. These synaptic phenotypes are present alongside the learning and memory deficits observed in Nrsn2^-/- mice.

Motor delay, affecting 90% of individuals, is the most common manifestation of 20p13 microdeletion syndrome [2]. In our study, we identified only mild alterations in Nrsn2^-/- mice, especially in max contact intensity and swing duration of left hind foot and right hind foot (Fig. 2). However, no significant locomotor impairments were detected by the open field test and rotarod test. Recent studies have increasingly found that the cerebellum is involved in motor learning functions [39, 40] and Purkinje cells (PCs) are the only output neurons in the cerebellar cortex and serve as the integration center of this region, playing a crucial role in motor coordination and motor balance. Recent studies have also reported that Purkinje cell dysfunction was observed in mice with motor deficits [41–43]. The mild motor impairments detected in these mice occur alongside changes in Purkinje cell morphology/function. Our findings indicate that Nrsn2 deficiency is associated with modest alterations in motor function.

This study has several limitations. First, the sample size of mice used in the behavioral experiments was relatively limited in this study. The restricted number of biological replicates limited the extent to which reproducibility and the influence of random variation could be fully assessed. In certain behavioral experiments, we used both female and male mice. Recent research has shown that male and female Nrsn2^-/- mice exhibit differences in body weight changes after 6 weeks of age; however, no particularly significant sex effects have been generally observed in behavioral tests. The sex-specific effects of Nrsn2 may require further investigation in future studies [30]. Second, our molecular analyses centred on glutamate-receptor subunits; only a restricted panel of GABA-receptor subunits was examined by Western blot. Although we observed a reduced frequency of sIPSCs in cerebellar Purkinje cells, a comprehensive assessment of inhibitory synaptic machinery is lacking. Consequently, the data do not fully establish the extent to which NRSN2 haploinsufficiency alone accounts for the 20p13 microdeletion phenotype.

In conclusion, our study highlights an important role of Nrsn2 in learning and memory, underscoring its importance in neurodevelopment. Nrsn2^-/- mice display significant behavioral impairments, particularly in spatial and long-term memory, which are accompanied by synaptic dysfunctions, including altered glutamate receptor expression and impaired long-term potentiation (LTP). These findings suggest that Nrsn2 is associated with synaptic plasticity and excitatory synaptic transmission. Given that NRSN2 is involved in multiple molecular mechanisms, future studies should explore how NRSN2 regulates the trafficking and stability of NMDA and AMPA receptors, which will advance our knowledge of synaptic plasticity regulation. This study provides valuable insights into the molecular mechanisms underlying intellectual disability phenotypes in patients with 20p13 microdeletion.

Supplementary information

41398_2026_4163_MOESM1_ESM.docx ^{(24.9KB, docx)}

Supplemental Materials and Supplementary Results

Supplementary Figure 1 ^{(995.5KB, tif)}

Supplementary Figure 2 ^{(735.1KB, tif)}

Supplementary Figure 3 ^{(2.1MB, tif)}

Supplementary Figure 4 ^{(2.1MB, tif)}

Supplementary Figure 5 ^{(954.3KB, tif)}

Author contributions

Conceived and designed the study: YA, and YD; behavior experiments: JLW, YY; electrophysiological experiment: YKW, YXG; Western Blot: JLW; Mouse KO model: LBW, Analyzed the data: JLW; Prepared the manuscript: JLW; Revised manuscript: YA and YD and YPS. All authors read and approved the final manuscript.

Funding

Project supported by Ningbo science and technology project (2023Z178), Shanghai Municipal Science and Technology Major Project (Grant No. 2017SHZDZX01), and Shanghai 2023 Science and Technology Innovation Action Plan (23JC1401800/23JC1401801/23JC1401802).

Data availability

Data are available from the corresponding authors upon reasonable request.

Competing interests

The authors declare no competing interests.

Ethics approval

This study is approved by the Ethical Review of Animal Experiment Projects at the Department of Laboratory Animal Science of Fudan University. All animal procedures were performed according to the guidelines - Institute of Developmental Biology and Molecular Medicine in China (IDM2021084).

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41398-026-04163-7.