SNX6-mediated subunit-specific secretory trafficking of AMPA receptors regulates synaptic function and plasticity Authors: Jiaqi Zhai, Deng Pan, Chao-Hua Jiang, Jiang Chen, Dou Wang, Ranran Mao, Zhenzhen Guo, Huali Geng, Jiawei Nie, Lei Chen, Yun Stone Shi, Jia-Jia Liu Published: 2025 ABSTRACT The AMPA-type glutamate receptors (AMPARs) mediate the majority of fast excitatory synaptic transmission in the brain and are central to synaptic plasticity in excitatory neurons. Functional AMPARs are tetramers composed of different combinations of the channel-forming GluA1-GluA4 subunits, with GluA2 critically influencing receptor biophysics and trafficking. However, the mechanisms governing AMPAR subunit-specific trafficking and assembly in resting and activated neurons remain incompletely understood. Using hippocampal neurons and conditional knockout mice, we demonstrate that sorting nexin 6 (SNX6), a membrane trafficking regulator, selectively mediates sorting of newly synthesized GluA2 into the post-Golgi secretory pathway before its assembly with GluA1, a process essential for AMPAR delivery to the plasma membrane. Loss of SNX6 diverts GluA2 to lysosomal degradation, leading to decreases in constitutive and activity-dependent surface expression of AMPARs, impairment of AMPAR-mediated synaptic transmission and the NMDA-type glutamate receptor (NMDAR)-dependent long-term potentiation, and deficits in learning and memory. Collectively, our findings reveal a novel, subunit-specific mechanism for AMPAR biogenesis and trafficking, and underscore the importance of regulatory steps within the biosynthetic secretory pathway in synaptic receptor delivery. Introduction The α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)-type ionotropic glutamate receptors (AMPARs) mediate fast excitatory synaptic transmission in the central nervous system and are key determinants in synaptic plasticity. These tetrameric ion channels are composed of the GluA1-4 subunits, with GluA2 playing a pivotal role in determining receptor biophysics. RNA editing of GluA2 transcripts converts glutamine at position 607 to the positively charged arginine, generating a Ca2+ selectivity filter in the channel pore of the heteromeric receptor complex1. The Ca2+ permeability of the GluA2-containing AMPARs is regulated by the auxiliary subunits transmembrane AMPAR regulatory proteins (TARPs) and cornichon (CNIH)2. In hippocampal CA1 pyramidal neurons, GluA1-GluA2 heterotetramers represent the dominant AMPAR subtype3. In ectopic expression systems, GluA1, 3, and 4 preferentially co-assemble with GluA2, but form homomeric complexes in its absence4–6. The carboxyl terminal domains of AMPAR subunits interact with different trafficking proteins and govern their mode of synaptic delivery. Early studies revealed that GluA2-containing receptors undergo constitutive synaptic delivery, whereas GluA1 mediates activity-dependent synaptic trafficking during long-term potentiation (LTP)7,8. However, emerging evidence indicates that both subunits contribute to the dynamic receptor turnover via distinct pathways. For example, sorting nexin 27 (SNX27) interacts with GluA1 to facilitate plasma membrane (PM) trafficking during LTP9. Glutamate receptor interacting protein 1 (GRIP1) binds GluA2 and is essential for both synaptic targeting during LTP10 and endocytic removal during long-term depression (LTD)11. Protein interacting with C-kinase 1 (PICK1) directly modulates GluA2, mediating subunit switching during LTP12 and activity-dependent internalization during LTD13. Despite extensive characterization of AMPAR structure and function, and elucidation of mechanisms for their insertion and removal at the PM, our understanding of the regulatory mechanisms governing receptor assembly and early secretory trafficking is still limited14. In eukaryotic cells, newly synthesized proteins are sorted in the trans-Golgi network (TGN) and trafficked to diverse subcellular destinations, including the PM, endosomes, and lysosomes15. In neurons, the endoplasmic reticulum (ER) network is contiguous throughout the axonal and somatodendritic compartments. Studies in rat central nervous system (CNS) neurons demonstrate that GluA1 and GluA2 are synthesized in both somata and dendrites16. While most of the somatically synthesized GluA1 follows the canonical Golgi-dependent secretory pathway17, in dendrites, locally synthesized GluA1 traffics through the ER-Golgi intermediary compartment (ERGIC) into a Golgi-independent pathway, accumulating in Rab11-positive recycling endosomes prior to PM insertion18. Although the KIF5 motor drives transport of GluA2 vesicles into dendrites via the adapter protein GRIP119,20, whether newly synthesized GluA2 and GluA1 assemble exclusively in the ER membrane before PM trafficking remains unclear. SNX6, a conserved member of the sorting nexin family, regulates protein sorting and trafficking. It contains an N-terminal Phox (PX) domain with dual protein/phosphoinositide-binding capacity and a C-terminal Bin/Amphiphysin/Rvs (BAR) domain enabling dimerization with other SNX BAR domains21. In mammalian cells, SNX6 acts as a cargo adapter for the dynein–dynactin molecular motor, mediating vesicular transport of the transmembrane cargo CI-MPR from endosomes to the TGN22–24. In mouse hippocampal neurons, SNX6 mediates dynein-driven endosomal transport of the postsynaptic scaffold protein Homer1b/c from the soma to distal dendrites, maintaining the structural and functional integrity of dendritic spines25. However, its role in sorting and trafficking of other neuronal cargoes remains unexplored. In this study, we further investigate the synaptic transmission and plasticity phenotypes in Snx6 knockout (KO) neurons. Unexpectedly, we find that while SNX6 is not involved in active vesicular transport of GluA1 and GluA2 in dendrites, it selectively mediates sorting of newly synthesized GluA2 into the post-Golgi secretory pathway for AMPAR surface expression. Loss of SNX6 results in mis-sorting of GluA2 into the lysosomal degradation pathway. Lack of GluA2 for AMPAR assembly causes a decrease in surface expression of the GluA1-containing receptor complex, reducing synaptic levels of AMPARs, impairing AMPAR-mediated transmission and LTP, and disrupting learning/memory processes. Together, these findings contribute to our understanding of mechanisms underlying subunit-specific secretory trafficking of AMPARs that is critical for synaptic transmission and plasticity. Results The AMPARs at the cell surface lack GluA2 in Snx6 KO neurons Previously, we found that genetic ablation of SNX6 in hippocampal pyramidal cells leads to reduced cell surface expression of GluA1 and GluA2 and impairment of AMPAR-mediated synaptic transmission25. Mechanistically, we uncovered that SNX6 mediates vesicular transport of Homer1b/c to the postsynaptic site, thereby maintaining AMPAR surface levels in dendritic spines through regulation of receptor endocytosis and recycling25. Intriguingly, electrophysiological recordings of Snx6 KO neurons in the hippocampal CA1 region of acute brain slices revealed not only reduced evoked excitatory postsynaptic current (EPSC) amplitudes, but also a marked inward rectification of AMPAR (Fig. 1A, B). In the hippocampus, approximately 80% of hippocampal AMPARs are tetrameric assemblies of GluA1 and the Q/R edited GluA2 subunits3,26, the Ca2+ permeability of which is regulated by TARPS and CNIH2. Since GluA2-containing AMPARs typically exhibit linear or near-linear current/voltage (I/V) relationships27, the observed biophysical changes in Snx6 KO neurons strongly suggest GluA2 subunit loss from the receptor complex. Consistent with these functional findings, immunofluorescence staining and confocal microscopy confirmed significant reductions in surface expression of GluA1 and GluA2 in both somata and dendrites of KO neurons (Fig. 1C, D)25. To investigate potential mechanisms underlying these subunit deficits, we compared transcriptional and translational profiles of GluA1 and GluA2 in hippocampal tissues from wild-type and Snx6 KO mice. While qRT-PCR analysis showed no significant differences in their mRNA levels, Western blot quantification revealed a selective reduction in GluA2 protein abundance without altering GluA1 expression (Fig. 1E, F). Collectively, these results indicate that SNX6 regulates GluA2 protein levels and AMPAR surface expression in hippocampal neurons.Fig. 1Ablation of SNX6 causes impairment of AMPAR-mediated synaptic transmission and loss of GluA2-containing AMPARs.A Electrophysiological analysis of hippocampal CA1 pyramidal neurons in acute brain slices. Scatterplots show amplitudes of AMPAR-mediated evoked excitatory postsynaptic currents (eEPSCs) for single pairs (open circles) of control (WT) and Cre virus-infected (KO) Snx6fl/fl neurons and mean ± SEM (filled circle). The inset displays sample traces from a pair of control (black) and infected neurons (green). The bar graph shows the mean ± SEM of AMPAR-eEPSCs. n = 29 pairs; 4 mice. Statistical significance was determined by Wilcoxon matched-pairs signed rank test. Scale bars, 100 pA, 20 ms. B Left: normalized I-V curves show AMPAR-eEPSCs measured at the holding potential of −70 mV, 0 mV, and 40 mV, respectively. Center: sample traces from a pair of control (black) and infected neurons (green). Right: the bar graph shows the mean ± SEM of rectification index (RI). n = 25 pairs; 4 mice. Statistical significance was determined by Wilcoxon matched-pairs signed rank test. Scale bars, 100 pA, 20 ms. C Cultured Snx6fl/fl hippocampal neurons were transfected with construct expressing EGFP (WT) or EGFP-2A-Cre (KO) on DIV11, fixed on DIV15, and immunostained for surface GluA1. EGFP serves as cell fill. Shown are representative confocal images (left) and quantification of fluorescent signals (right) in the soma and dendrites. Data represent mean ± SEM, n = 10-18 neurons from three independent experiments, 2-3 dendrites/neurons. Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 5 μm. D Same as (C) except that cells were immunostained for surface GluA2. Data represent mean ± SEM, n = 17-23 neurons from three independent experiments, 2-3 dendrites/neurons. Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 5 μm. E RNA was extracted from the hippocampi of 8-week-old Snx6fl/fl (WT) and Nestin-Cre;Snx6fl/fl (KO) mice, and the relative expression levels of AMPAR were examined using quantitative PCR. Data represent mean ± SEM, n = 3 mice. Statistical significance was determined by two-tailed Student’s t-test. F GluA1, GluA2, and SNX6 were assessed by immunoblotting of protein extracts from the hippocampi of 8-week-old Snx6fl/fl (WT) and Nestin-Cre;Snx6fl/fl (KO) mice. The scatterplots show the levels of GluA1 and GluA2 relative to GAPDH. Data represent mean ± SEM, n = 3 mice. Statistical significance was determined by two-tailed Student’s t-test. Ablation of SNX6 causes impairment of AMPAR-mediated synaptic transmission and loss of GluA2-containing AMPARs. A Electrophysiological analysis of hippocampal CA1 pyramidal neurons in acute brain slices. Scatterplots show amplitudes of AMPAR-mediated evoked excitatory postsynaptic currents (eEPSCs) for single pairs (open circles) of control (WT) and Cre virus-infected (KO) Snx6fl/fl neurons and mean ± SEM (filled circle). The inset displays sample traces from a pair of control (black) and infected neurons (green). The bar graph shows the mean ± SEM of AMPAR-eEPSCs. n = 29 pairs; 4 mice. Statistical significance was determined by Wilcoxon matched-pairs signed rank test. Scale bars, 100 pA, 20 ms. B Left: normalized I-V curves show AMPAR-eEPSCs measured at the holding potential of −70 mV, 0 mV, and 40 mV, respectively. Center: sample traces from a pair of control (black) and infected neurons (green). Right: the bar graph shows the mean ± SEM of rectification index (RI). n = 25 pairs; 4 mice. Statistical significance was determined by Wilcoxon matched-pairs signed rank test. Scale bars, 100 pA, 20 ms. C Cultured Snx6fl/fl hippocampal neurons were transfected with construct expressing EGFP (WT) or EGFP-2A-Cre (KO) on DIV11, fixed on DIV15, and immunostained for surface GluA1. EGFP serves as cell fill. Shown are representative confocal images (left) and quantification of fluorescent signals (right) in the soma and dendrites. Data represent mean ± SEM, n = 10-18 neurons from three independent experiments, 2-3 dendrites/neurons. Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 5 μm. D Same as (C) except that cells were immunostained for surface GluA2. Data represent mean ± SEM, n = 17-23 neurons from three independent experiments, 2-3 dendrites/neurons. Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 5 μm. E RNA was extracted from the hippocampi of 8-week-old Snx6fl/fl (WT) and Nestin-Cre;Snx6fl/fl (KO) mice, and the relative expression levels of AMPAR were examined using quantitative PCR. Data represent mean ± SEM, n = 3 mice. Statistical significance was determined by two-tailed Student’s t-test. F GluA1, GluA2, and SNX6 were assessed by immunoblotting of protein extracts from the hippocampi of 8-week-old Snx6fl/fl (WT) and Nestin-Cre;Snx6fl/fl (KO) mice. The scatterplots show the levels of GluA1 and GluA2 relative to GAPDH. Data represent mean ± SEM, n = 3 mice. Statistical significance was determined by two-tailed Student’s t-test. GluA2 is mis-sorted into the lysosomal degradation pathway in Snx6 KO neurons To elucidate the mechanism by which SNX6 regulates GluA2 levels and AMPAR trafficking, we first characterized the subcellular distribution of GluA1 and GluA2 in Snx6 KO neurons by immunofluorescence staining and confocal microscopy. Strikingly, while total GluA1 immunoreactivity remained unchanged in KO neurons compared with the wild-type, GluA2 signals were significantly reduced in both somata and dendritic processes (Fig. 2A, B). Given SNX6’s established role in protein sorting and trafficking22,23,25,28, we hypothesized that GluA2 might undergo accelerated degradation in the absence of SNX6. To test this possibility, we treated Snx6 KO neurons with MG132 (a proteasome inhibitor), chloroquine, or leupeptin (both lysosome inhibitors). Remarkably, pharmacological inhibition of lysosomal activity with chloroquine or leupeptin restored GluA2 protein levels in KO neurons (Fig. 2C). Of note, in hippocampal neurons, most lysosomes with degradative activity are distributed in the soma29. Indeed, while colocalization between GluA2 and the lysosomal marker LAMP1 increased in Snx6 KO neurons following lysosome inhibitor treatment (Fig. 2C), lysosomal inhibition failed to rescue surface GluA2 expression in KO neurons (Fig. 2D), indicating mis-sorting of GluA2 into degradative lysosomes. To further investigate trafficking defects, we analyzed the subcellular distribution of GluA1 and GluA2 in the Snx6 KO neurons by co-immunostaining with markers for the secretory, endocytic and recycling pathways. Image analysis revealed a specific reduction in GluA2 colocalization with Rab8 (a marker for secretory endosomes)30–33, whereas the localization of GluA1 remained unaltered (Fig. 2E, and Fig. S1). Collectively, these data suggest that SNX6 is critical for sorting GluA2 into the secretory pathway, thereby preventing its mis-targeting to lysosomal degradation machinery. Moreover, together with the finding that total GluA1 was unchanged (Fig. 1E, F), these results indicate that SNX6 exerts a direct effect on GluA2, which in turn indirectly reduces the abundance of surface GluA1-GluA2 heteromeric complexes.Fig. 2Loss of SNX6 causes lysosomal degradation of GluA2.A Cultured Snx6fl/fl hippocampal neurons were transfected with construct expressing EGFP (WT) or EGFP-2A-Cre (KO) on DIV11, fixed on DIV15, and immunostained for total GluA1. Shown are representative confocal images (left) and quantification of fluorescent signals (right) in the soma and proximal (Pro, 0–30 μm from the cell body), middle (Mid, 30-60 μm), and distal (Dis, 60–90 μm) dendrites. Data represent mean ± SEM, n = 14-15 neurons from three independent experiments, 2-3 dendrites/neuron. Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 5 μm. B Same as (A) except that cells were immunostained for total GluA2. Data represent mean ± SEM, n = 20 neurons/group from three independent experiments, 2-3 dendrites/neuron. Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 5 μm. C Same as (A) except that neurons were treated with chloroquine (100 μM), leupeptin (100 μg/mL) or MG132 (10 μM) for 4 h, fixed and immunostained for LAMP1 and GluA2. Shown are representative confocal images (upper panels) and quantification of total GluA2 and colocalization of GluA2 with LAMP1 in the soma (lower panels). Data represent mean ± SEM, n > 20 neurons from two independent experiments. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. Scale bar, 5 μm. D Same as (A) except that neurons were treated with leupeptin (100 μg/mL) or DMSO for 4 h, fixed and immunostained for surface GluA2. Shown are representative confocal images (left) and quantification of fluorescent signals (right) in the soma and dendrites. Data represent mean ± SEM, n = 16-20 neurons from two independent experiments, 2-3 dendrites/neuron. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. Scale bars, 5 μm. E Same as (A) except that neurons were fixed and immunostained for intracellular Rab8 and GluA2. Shown are representative confocal images (left) and quantification of GluA2 colocalization with Rab8 (right) in the soma and dendrites. Data represent mean ± SEM, n = 16-26 neurons from two independent experiments, 2-3 dendrites/neuron. Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 5 μm. Loss of SNX6 causes lysosomal degradation of GluA2. A Cultured Snx6fl/fl hippocampal neurons were transfected with construct expressing EGFP (WT) or EGFP-2A-Cre (KO) on DIV11, fixed on DIV15, and immunostained for total GluA1. Shown are representative confocal images (left) and quantification of fluorescent signals (right) in the soma and proximal (Pro, 0–30 μm from the cell body), middle (Mid, 30-60 μm), and distal (Dis, 60–90 μm) dendrites. Data represent mean ± SEM, n = 14-15 neurons from three independent experiments, 2-3 dendrites/neuron. Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 5 μm. B Same as (A) except that cells were immunostained for total GluA2. Data represent mean ± SEM, n = 20 neurons/group from three independent experiments, 2-3 dendrites/neuron. Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 5 μm. C Same as (A) except that neurons were treated with chloroquine (100 μM), leupeptin (100 μg/mL) or MG132 (10 μM) for 4 h, fixed and immunostained for LAMP1 and GluA2. Shown are representative confocal images (upper panels) and quantification of total GluA2 and colocalization of GluA2 with LAMP1 in the soma (lower panels). Data represent mean ± SEM, n > 20 neurons from two independent experiments. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. Scale bar, 5 μm. D Same as (A) except that neurons were treated with leupeptin (100 μg/mL) or DMSO for 4 h, fixed and immunostained for surface GluA2. Shown are representative confocal images (left) and quantification of fluorescent signals (right) in the soma and dendrites. Data represent mean ± SEM, n = 16-20 neurons from two independent experiments, 2-3 dendrites/neuron. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. Scale bars, 5 μm. E Same as (A) except that neurons were fixed and immunostained for intracellular Rab8 and GluA2. Shown are representative confocal images (left) and quantification of GluA2 colocalization with Rab8 (right) in the soma and dendrites. Data represent mean ± SEM, n = 16-26 neurons from two independent experiments, 2-3 dendrites/neuron. Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 5 μm. SNX6 directly interacts with GluA2 and regulates its sorting into the secretory pathway for surface expression The observation that GluA2 undergoes lysosomal degradation in Snx6 KO neurons implies a critical role for SNX6 in directing GluA2 trafficking to the secretory pathway for cell surface expression. To test this possibility, first we determined whether SNX6 interacts with GluA2. Co-immunoprecipitation (coIP) of endogenous proteins from mouse brain lysates detected substantial interaction between SNX6 and GluA2, but not with GluA1 or the NMDAR subunit NR2B (Fig. 3A). Consistently, immunofluorescence staining and confocal microscopy demonstrated partial colocalization of SNX6 with GluA2 in the neuronal somata and dendritic shafts (Fig. 3B). Moreover, coIP from HEK293T cells heterologously expressing GluA2 indicated that the intracellular C-terminal domain of GluA2 is essential for SNX6 binding (Fig. 3C, D). Notably, no interaction was detected between GluA2 and SNX5, a close homolog of SNX6 (Fig. 3E). Further, in vitro pull-down assay verified direct interaction between the C-terminus of GluA2 and both the N-terminal PX and the C-terminal BAR domains of SNX6 (Fig. 3F). Together, these data establish a specific and direct interaction between SNX6 and GluA2.Fig. 3SNX6 mediates secretory trafficking of GluA2 for surface expression.A Co-immunoprecipitation was performed on hippocampal lysates from 8-week-old Snx6fl/fl (WT) or Nestin-Cre;Snx6fl/fl (KO) mice using anti-SNX6 antibodies, followed by immunoblotting with the indicated antibodies. Shown are representative western blots (left) and quantification of interaction strength (right). The SNX6 band is indicated with a red asterisk. Data represent mean ± SEM, N = 4. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. B Cultured Snx6fl/fl hippocampal neurons were transfected with construct expressing EGFP, fixed on DIV15, and immunostained for SNX6 and GluA2. Shown are representative confocal images (left) and quantification of SNX6-GluA2 colocalization (right) in the soma (upper) and dendrites (lower). Data represent mean ± SEM, n = 13–14 neurons from two independent experiments, 2-3 dendrites/neuron. Scale bars, 5 μm. C Schematic representation of the domain structures of SNX6 and GluA2, as well as the fragments of SNX6 and GluA2 used in this study. D HEK293T cells co-transfected with constructs expressing Flag-SNX6 and SEP-GluA2 or SEP-GluA2ΔC were lysed for immunoprecipitation with anti-Flag antibody conjugated-agarose beads. The Flag vector served as negative control. Input and bound proteins were analyzed by SDS-PAGE and immunoblotting with antibodies against Flag and GFP. Shown are representative western blots (left) and quantification of interaction strength (right). Data represent mean ± SEM, N = 3. Statistical significance was determined by two-tailed Student’s t-test. E HEK293T cells co-transfected with constructs expressing SEP-GluA2 and Flag-SNX5 or Flag-SNX6 were lysed for immunoprecipitation with anti-Flag. Input and bound proteins were analyzed by SDS-PAGE and immunoblotting with antibodies against Flag and GFP. Shown are representative western blots (left) and quantification of interaction strength (right). Data represent mean ± SEM, N = 3. Statistical significance was determined by two-tailed Student’s t-test. F GST-tagged SNX6 or its fragments conjugated to glutathione-Sepharose were incubated with His-GluA2-C. Input and bound proteins were analyzed by SDS-PAGE and immunoblotting with antibodies against His. Shown are representative western blot and coomassie blue staining (left), and quantification of interaction strength (right). Data represent mean ± SEM, N = 3. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. G Model of the SNX6-GluA2-C complex predicted by AlphaFold 2. Bottom panel: interactions at the dimeric interface of SNX6-GluA2. H HEK293T cells co-transfected with constructs expressing Flag-GluA2 and EGFP-SNX6, EGFP-SNX6 S37A/S41A (EGFP-SNX6mut) or EGFP-SNX6 K138A (EGFP-SNX6KA) were lysed for immunoprecipitation with anti-Flag. Input and bound proteins were analyzed by SDS-PAGE and immunoblotting with antibodies against Flag and GFP. Shown are representative western blots (left) and quantification of interaction strength (right). Data represent mean ± SEM, N = 4. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. I HEK293T cells co-transfected with constructs expressing mCherry-Homer1c and EGFP, EGFP-SNX6 or EGFP-SNX6mut were lysed for immunoprecipitation with anti-GFP. Input and bound proteins were analyzed by SDS-PAGE and immunoblotting with antibodies against mCherry and GFP. Shown are representative western blots (left) and quantification of interaction strength (right). Data represent mean ± SEM, N = 3. Statistical significance was determined by two-tailed Student’s t-test. J Cultured Snx6fl/fl hippocampal neurons were co-transfected with constructs expressing mCherry (WT) and EGFP, or mCherry-2A-Cre (KO) and EGFP, EGFP-SNX6 or EGFP-SNX6mut on DIV11, fixed on DIV15, and immunostained for total GluA2. Shown are representative confocal images (upper) and quantification of fluorescent signals (lower) in the soma and dendrites. Data represent mean ± SEM, n = 14-17 neurons from two independent experiments, 2-3 dendrites/neuron. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. Scale bars, 5 μm. K Same as (J) except that cells were immunostained for surface GluA2. Data represent mean ± SEM, n = 10-14 neurons from two independent experiments, 2-3 dendrites/neuron. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. Scale bars, 5 μm. SNX6 mediates secretory trafficking of GluA2 for surface expression. A Co-immunoprecipitation was performed on hippocampal lysates from 8-week-old Snx6fl/fl (WT) or Nestin-Cre;Snx6fl/fl (KO) mice using anti-SNX6 antibodies, followed by immunoblotting with the indicated antibodies. Shown are representative western blots (left) and quantification of interaction strength (right). The SNX6 band is indicated with a red asterisk. Data represent mean ± SEM, N = 4. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. B Cultured Snx6fl/fl hippocampal neurons were transfected with construct expressing EGFP, fixed on DIV15, and immunostained for SNX6 and GluA2. Shown are representative confocal images (left) and quantification of SNX6-GluA2 colocalization (right) in the soma (upper) and dendrites (lower). Data represent mean ± SEM, n = 13–14 neurons from two independent experiments, 2-3 dendrites/neuron. Scale bars, 5 μm. C Schematic representation of the domain structures of SNX6 and GluA2, as well as the fragments of SNX6 and GluA2 used in this study. D HEK293T cells co-transfected with constructs expressing Flag-SNX6 and SEP-GluA2 or SEP-GluA2ΔC were lysed for immunoprecipitation with anti-Flag antibody conjugated-agarose beads. The Flag vector served as negative control. Input and bound proteins were analyzed by SDS-PAGE and immunoblotting with antibodies against Flag and GFP. Shown are representative western blots (left) and quantification of interaction strength (right). Data represent mean ± SEM, N = 3. Statistical significance was determined by two-tailed Student’s t-test. E HEK293T cells co-transfected with constructs expressing SEP-GluA2 and Flag-SNX5 or Flag-SNX6 were lysed for immunoprecipitation with anti-Flag. Input and bound proteins were analyzed by SDS-PAGE and immunoblotting with antibodies against Flag and GFP. Shown are representative western blots (left) and quantification of interaction strength (right). Data represent mean ± SEM, N = 3. Statistical significance was determined by two-tailed Student’s t-test. F GST-tagged SNX6 or its fragments conjugated to glutathione-Sepharose were incubated with His-GluA2-C. Input and bound proteins were analyzed by SDS-PAGE and immunoblotting with antibodies against His. Shown are representative western blot and coomassie blue staining (left), and quantification of interaction strength (right). Data represent mean ± SEM, N = 3. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. G Model of the SNX6-GluA2-C complex predicted by AlphaFold 2. Bottom panel: interactions at the dimeric interface of SNX6-GluA2. H HEK293T cells co-transfected with constructs expressing Flag-GluA2 and EGFP-SNX6, EGFP-SNX6 S37A/S41A (EGFP-SNX6mut) or EGFP-SNX6 K138A (EGFP-SNX6KA) were lysed for immunoprecipitation with anti-Flag. Input and bound proteins were analyzed by SDS-PAGE and immunoblotting with antibodies against Flag and GFP. Shown are representative western blots (left) and quantification of interaction strength (right). Data represent mean ± SEM, N = 4. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. I HEK293T cells co-transfected with constructs expressing mCherry-Homer1c and EGFP, EGFP-SNX6 or EGFP-SNX6mut were lysed for immunoprecipitation with anti-GFP. Input and bound proteins were analyzed by SDS-PAGE and immunoblotting with antibodies against mCherry and GFP. Shown are representative western blots (left) and quantification of interaction strength (right). Data represent mean ± SEM, N = 3. Statistical significance was determined by two-tailed Student’s t-test. J Cultured Snx6fl/fl hippocampal neurons were co-transfected with constructs expressing mCherry (WT) and EGFP, or mCherry-2A-Cre (KO) and EGFP, EGFP-SNX6 or EGFP-SNX6mut on DIV11, fixed on DIV15, and immunostained for total GluA2. Shown are representative confocal images (upper) and quantification of fluorescent signals (lower) in the soma and dendrites. Data represent mean ± SEM, n = 14-17 neurons from two independent experiments, 2-3 dendrites/neuron. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. Scale bars, 5 μm. K Same as (J) except that cells were immunostained for surface GluA2. Data represent mean ± SEM, n = 10-14 neurons from two independent experiments, 2-3 dendrites/neuron. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. Scale bars, 5 μm. To map the GluA2 binding site on SNX6, we integrated structural predictions from DeepMind AlphaFold2 (Fig. 3G) with mutagenesis studies in HEK293T cells. Mutations of the top-ranked residues Ser37 and Ser41 (S37A/S41A, referred to as SNX6mut) but not K138 (K138A, referred to as SNX6KA), amino acids predicted to mediate the interaction with GluA2, significantly impaired coIP efficiency (Fig. 3H). Importantly, the mutations did not disrupt SNX6 binding to the postsynaptic scaffold protein Homer1c (Fig. 3I). Further, functional validation in hippocampal neurons revealed that wild-type SNX6, but not the GluA2-binding deficient mutant or SNX5, rescued both total cellular levels and cell surface expression of GluA2 in Snx6 KO cells (Fig. 3J, K, and Fig. S2). Notably, both the wild-type and the GluA2-binding deficient mutant of SNX6 rescued the dendritic transport defect of Homer1b/c in Snx6 KO neurons (Fig. S3). All together, these results confirm that SNX6 is required for GluA2 sorting into the secretory pathway for surface expression. Given SNX6’s established role as dynein adapter for endosomal transport22,23,25, next we investigated its involvement in dendritic trafficking of GluA2-containing AMPARs. Live imaging of fluorescently tagged SNX6 and AMPAR subunits in cultured hippocampal neurons did not detect comovement of SNX6 with either GluA1 or GluA2 (Fig. 4A, B, Video 1). Further, analysis of GluA1/GluA2 dynamics in wild-type and Snx6 KO neurons revealed no significant differences in the speed and run length of their anterograde or retrograde transport (Fig. 4C, D, Video 2 and 3). Collectively, these findings exclude a role for SNX6 in vesicular transport of GluA1 and GluA2 in dendritic shafts.Fig. 4SNX6 is not involved in vesicular transport of GluA1 and GluA2 in the dendritic shaft.A Cultured Snx6fl/fl hippocampal neurons were co-transfected with constructs expressing EGFP-SNX6 and mScarlet-GluA1 on DIV11. On DIV15, cells were imaged live to record the movement of EGFP-SNX6 and mScarlet-GluA1. Shown are representative kymographs. Scale bar, 5 μm. B Same as (A) except that neurons were co-transfected with constructs expressing EGFP-SNX6 and mScarlet-GluA2. Scale bar, 5 μm. C Cultured Snx6fl/fl hippocampal neurons were co-transfected with constructs expressing EGFP (WT) or EGFP-2A-Cre (KO) and mScarlet-GluA1 on DIV11. On DIV15, cells were imaged live to record the movement of mScarlet-GluA1. Shown are representative kymographs and quantification of speed and run length of mScarlet-GluA1. Scale bar, 5 μm. D Same as (C) except that neurons were co-transfected with constructs expressing EGFP (WT) or EGFP-2A-Cre (KO) and mScarlet-GluA2. Scale bar, 5 μm. All data represent mean ± SEM, n = 16-18 neurons from three independent experiments. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. SNX6 is not involved in vesicular transport of GluA1 and GluA2 in the dendritic shaft. A Cultured Snx6fl/fl hippocampal neurons were co-transfected with constructs expressing EGFP-SNX6 and mScarlet-GluA1 on DIV11. On DIV15, cells were imaged live to record the movement of EGFP-SNX6 and mScarlet-GluA1. Shown are representative kymographs. Scale bar, 5 μm. B Same as (A) except that neurons were co-transfected with constructs expressing EGFP-SNX6 and mScarlet-GluA2. Scale bar, 5 μm. C Cultured Snx6fl/fl hippocampal neurons were co-transfected with constructs expressing EGFP (WT) or EGFP-2A-Cre (KO) and mScarlet-GluA1 on DIV11. On DIV15, cells were imaged live to record the movement of mScarlet-GluA1. Shown are representative kymographs and quantification of speed and run length of mScarlet-GluA1. Scale bar, 5 μm. D Same as (C) except that neurons were co-transfected with constructs expressing EGFP (WT) or EGFP-2A-Cre (KO) and mScarlet-GluA2. Scale bar, 5 μm. All data represent mean ± SEM, n = 16-18 neurons from three independent experiments. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. GluA1 and GluA2 from different post-ER trafficking pathways assemble to form the AMPAR receptor complex for surface expression In dendrites, whilst numerous synaptic transmembrane proteins travel through the ER-ERGIC-Golgi satellites (GS) microsecretory system en route to the PM34, newly synthesized GluA1 traverses the ERGIC, bypasses the Golgi apparatus and localizes to Rab11-labeled recycling endosomes for subsequent delivery to the cell surface18,35. Given the finding that SNX6 mediates the sorting of GluA2, but not GluA1, into the secretory pathway, we hypothesized that newly-synthesized GluA1 and GluA2 enter distinct post-ER trafficking routes and assemble into the receptor complex at a common subcellular location before being transported to the PM. To explore this hypothesis, we first examined the subcellular distribution of GluA1 and GluA2 in the ER and synaptosomes isolated by differential centrifugation and fractionation with sucrose step gradients (Fig. 5A, B). Immunoblotting indicated that, both AMPAR subunits, along with the NMDAR subunit NR2A, were enriched in synaptosomes at the 34–43% interface of the step gradients. However, they were also present in light membrane fractions (membranes collected from the 11–25% interface after step gradient fractionation of S2, the supernatant after 12,000 × g centrifugation) enriched with the ER (Fig. 5C). Subsequently, we conducted coIP analysis of the GluA1-GluA2 complex in these two fractions. Intriguingly, compared to the robust association of GluA2 with GluA1 in the synaptosomes, significantly less GluA2 was detected in the GluA1 immunoprecipitates from the ER fraction (Fig. 5D). These results suggest that the ER membrane is not the sole site for the assembly of GluA1-GluA2 AMPARs.Fig. 5GluA1-GluA2 AMPARs are not assembled exclusively at the ER.A Schematic diagram of the subcellular fractionation steps used to generate the indicated fractions. B Distinct layers obtained from sucrose gradient centrifugation. C Samples form (A) and (B) were analyzed using SDS-PAGE and immunoblotting with antibodies against NR2A, GluA1, GluA2, and the ER marker Calreticulin. D The ER and synaptosome (Syn) fractions were immunoprecipitated with antibodies to GluA1 coupled to Dynabeads Protein G. Input and bound proteins were analyzed by SDS-PAGE and immunoblotting with antibodies against GluA1 and GluA2. Shown are representative western blot and quantification of GluA1 immunoprecipitated (normalized to input) and GluA2 co-immunoprecipitated (normalized to GluA1). Data represent mean ± SEM, N = 4. Statistical significance was determined by two-tailed Student’s t-test. E Cultured Snx6fl/fl hippocampal neurons were transfected with construct expressing EGFP (WT) or EGFP-2A-Cre (KO) on DIV11. On DIV15, neurons were treated with leupeptin (100 μg/mL) for 4 h, fixed and immunostained for intracellular GluA1 and GluA2. Shown are representative confocal images (left) and quantification of GluA1-GluA2 colocalization (right) in the soma and dendrites. The colocalization was analyzed by Mander’s colocalization coefficient which indicates the fraction of GluA2 overlaying with GluA1 and vice versa. Data represent mean ± SEM, n = 13–16 neurons, 2-3 dendrites/neuron. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. Scale bars, 5 μm. GluA1-GluA2 AMPARs are not assembled exclusively at the ER. A Schematic diagram of the subcellular fractionation steps used to generate the indicated fractions. B Distinct layers obtained from sucrose gradient centrifugation. C Samples form (A) and (B) were analyzed using SDS-PAGE and immunoblotting with antibodies against NR2A, GluA1, GluA2, and the ER marker Calreticulin. D The ER and synaptosome (Syn) fractions were immunoprecipitated with antibodies to GluA1 coupled to Dynabeads Protein G. Input and bound proteins were analyzed by SDS-PAGE and immunoblotting with antibodies against GluA1 and GluA2. Shown are representative western blot and quantification of GluA1 immunoprecipitated (normalized to input) and GluA2 co-immunoprecipitated (normalized to GluA1). Data represent mean ± SEM, N = 4. Statistical significance was determined by two-tailed Student’s t-test. E Cultured Snx6fl/fl hippocampal neurons were transfected with construct expressing EGFP (WT) or EGFP-2A-Cre (KO) on DIV11. On DIV15, neurons were treated with leupeptin (100 μg/mL) for 4 h, fixed and immunostained for intracellular GluA1 and GluA2. Shown are representative confocal images (left) and quantification of GluA1-GluA2 colocalization (right) in the soma and dendrites. The colocalization was analyzed by Mander’s colocalization coefficient which indicates the fraction of GluA2 overlaying with GluA1 and vice versa. Data represent mean ± SEM, n = 13–16 neurons, 2-3 dendrites/neuron. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. Scale bars, 5 μm. To test the possibility that GluA1 and GluA2 assemble after converging from different biosynthetic secretory pathways, we analyzed their colocalization in leupeptin-treated Snx6 KO neurons. We reasoned that if they assemble after GluA2 is sorted into the post-Golgi secretory pathway, then even when lysosomal degradation is inhibited in KO cells, their colocalization would not increase because GluA2 alone would be trapped in endolysosomes. Indeed, leupeptin treatment did not restore the colocalization between GluA1 and GluA2 in KO neurons (Fig. 5E), indicating that mis-sorting of GluA2 into the degradation pathway disrupted the assembly of GluA1-A2 AMPARs. These data are also consistent with the reduced distribution of GluA2, but not GluA1, to Rab8-labeled secretory endosomes in Snx6 KO neurons (Fig. 2E), suggesting that substantial portions of GluA1 and GluA2 enter separate post-ER trafficking pathways. To further investigate the secretory trafficking and assembly of AMPAR subunits, we employed the FK506-binding protein (FKBP)-based ER retention system of secretory cargo36 to study the trafficking itineraries of newly synthesized GluA1 and GluA2. To this end, we generated ER-retained versions of mCherry-GluA1 and EGFP-GluA2 by inserting FKBP into the first extracellular loop after their signal peptide, respectively (Fig. 6A). Dimerization of FKBP causes aggregation and retention of the receptor subunits in the ER, which could be dissolved rapidly upon the addition of SLF, a membrane permeable synthetic ligand of FKBP36,37. After adding SLF to release the subunits from the ER and simultaneously inhibiting protein synthesis with cycloheximide, we monitored their distribution in both the soma and dendrites over time. Of note, SLF-treated cells were kept at 20 °C for 2 hours to synchronize secretory cargoes in post-ER compartments before being shifted back to 37 °C18. By immunofluorescence staining and confocal microscopy, we tracked the progression of newly synthesized GluA1 and GluA2 through the secretory network. While the vast majority of GluA1 followed the ER-ERGIC-RE route for surface delivery18, a significant portion of GluA2 trafficked through sialyltransferase (St3)-labeled Golgi structures (the Golgi apparatus in the soma and Golgi satellites in dendrites)34 and Rab8-labeled secretory endosomes32,33 in both somata and dendrites (Fig. 6B–E, and Video 4). Notably, 40–50% of the ER-released GluA2 colocalized with SNX6 (Fig. 6F). Moreover, the colocalization of ER-released GluA2 with Rab8-labeled vesicles but not St3-labeled Golgi structures was lower in Snx6 KO neurons (Fig. 6C, D), and this was rescued by expressing wild-type but not the GluA2-binding deficient mutant of SNX6 (Fig. 6G), indicating that SNX6 mediates the sorting of GluA2 from the Golgi compartment to Rab8-labeled secretory endosomes.Fig. 6Significant fractions of GluA1 and GluA2 are differentially trafficked from their site of biosynthesis prior to assembly into the AMPAR receptor complex for surface expression.A Schematic diagram of the inducible-release constructs. FM: FKBP-F36M. B Cultured Snx6fl/fl hippocampal neurons co-transfected with constructs expressing GFP-ERGIC53, 3×FM-mCherry-GluA1 (mCherry-GluA1) and 3×FM-Flag-GluA2 (Flag-GluA2) were treated with SLF on DIV14, fixed 1 h after release from the 20 °C temperature block and immunostained for GFP, mCherry, and Flag. Shown are representative confocal images (left) and quantification of GluA1/GluA2 colocalization with GFP-ERGIC53 (right) in the soma and dendrites. Data represent mean ± SEM, n = 15 neurons, 1-3 dendrites/neuron. Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 5 μm. C Cultured Snx6fl/fl (WT) or Nestin-Cre;Snx6fl/fl (KO) hippocampal neurons co-expressing St3-mCherry, 3×FM-Flag-GluA1 (Flag-GluA1) and 3×FM-EGFP-GluA2 (EGFP-GluA2) were treated with SLF, fixed 1 h or 2 h after release from the temperature block and immunostained for EGFP, mCherry, and Flag. Shown are representative confocal images at the 2 h time point (left), quantification of GluA1/GluA2 colocalization with St3-mCherry in WT neurons, and GluA2 colocalization with St3-mCherry in WT and KO neurons (right). Data represent mean ± SEM, n = 12 neurons, 1-3 dendrites/neuron. Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 5 μm. D Cultured Snx6fl/fl (WT) or Nestin-Cre;Snx6fl/fl (KO) hippocampal neurons co-expressing 3×FM-mCherry-GluA1 (mCherry-GluA1) and 3×FM-EGFP-GluA2 (EGFP-GluA2) were treated with SLF, fixed 2 h or 4 h after release from the temperature block and immunostained for Rab8, mCherry, and EGFP. Shown are representative confocal images of 4 h (left), and quantification of colocalization between GluA1/ GluA2 and Rab8 in WT (upper right) and GluA2 colocalization with Rab8 in WT and KO (lower right). Data represent mean ± SEM, n = 16 neurons, 1-3 dendrites/neuron. Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 5 μm. E Cultured Snx6fl/fl hippocampal neurons co-expressing 3×FM-mCherry-GluA1 (mCherry-GluA1) and 3×FM-EGFP-GluA2 (EGFP-GluA2) were treated with SLF, fixed 2 h or 4 h after release from the temperature block and immunostained for Rab11, mCherry, and EGFP. Shown are representative confocal images of 4 h (left) and quantification of colocalization between GluA1/GluA2 and Rab11 (right) in the soma and dendrites. Data represent mean ± SEM, n = 12 neurons, 1-3 dendrites/neuron. Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 5 μm. F Cultured Snx6fl/fl hippocampal neurons co-expressing 3×FM-mCherry-GluA1 (mCherry-GluA1) and 3×FM-EGFP-GluA2 (EGFP-GluA2) were treated with SLF, fixed 2 h after release from the temperature block and immunostained for SNX6, mCherry and EGFP. Shown are representative confocal images (left) and quantification of colocalization between GluA1/GluA2 and SNX6 (right) in the soma and dendrites. Data represent mean ± SEM, n = 13 neurons, 1-3 dendrites/neuron. Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 5 μm. G Cultured Snx6fl/fl hippocampal neurons co-expressing mCherry (WT), 3×FM-EGFP-GluA2 (EGFP-GluA2) and Flag, or mCherry-2A-Cre (KO), EGFP-GluA2 and Flag, Flag-SNX6 or Flag-SNX6mut were treated with SLF, fixed 4 h after release from the temperature block and immunostained for Rab8, mCherry, and EGFP. Shown are representative confocal images (left) and quantification of GluA2-Rab8 colocalization (right) in the soma and dendrites. Data represent mean ± SEM, n = 12–17 neurons, 1–3 dendrites/neuron. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. Scale bars, 5 μm. (H) Cultured Snx6fl/fl (WT) or Nestin-Cre;Snx6fl/fl (KO) hippocampal neurons co-expressing 3×FM-mCherry-GluA1 (mCherry-GluA1) and 3×FM-EGFP-GluA2 (EGFP-GluA2) were treated with SLF, fixed after release from the temperature block and surface stained for mCherry-GluA1 with antibodies to mCherry, then permeabilized and immunostained for intracellular mCherry and EGFP. Shown are representative confocal images (left) and quantification of the surface/cytoplasm ratio of GluA1, as well as changes in surface GluA1 at 2 h or 4 h relative to 0 h in WT and KO neurons (right). Data represent mean ± SEM, n = 10-17 neurons from two independent experiments, 1-3 dendrites/neuron. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. Scale bars, 5 μm. Significant fractions of GluA1 and GluA2 are differentially trafficked from their site of biosynthesis prior to assembly into the AMPAR receptor complex for surface expression. A Schematic diagram of the inducible-release constructs. FM: FKBP-F36M. B Cultured Snx6fl/fl hippocampal neurons co-transfected with constructs expressing GFP-ERGIC53, 3×FM-mCherry-GluA1 (mCherry-GluA1) and 3×FM-Flag-GluA2 (Flag-GluA2) were treated with SLF on DIV14, fixed 1 h after release from the 20 °C temperature block and immunostained for GFP, mCherry, and Flag. Shown are representative confocal images (left) and quantification of GluA1/GluA2 colocalization with GFP-ERGIC53 (right) in the soma and dendrites. Data represent mean ± SEM, n = 15 neurons, 1-3 dendrites/neuron. Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 5 μm. C Cultured Snx6fl/fl (WT) or Nestin-Cre;Snx6fl/fl (KO) hippocampal neurons co-expressing St3-mCherry, 3×FM-Flag-GluA1 (Flag-GluA1) and 3×FM-EGFP-GluA2 (EGFP-GluA2) were treated with SLF, fixed 1 h or 2 h after release from the temperature block and immunostained for EGFP, mCherry, and Flag. Shown are representative confocal images at the 2 h time point (left), quantification of GluA1/GluA2 colocalization with St3-mCherry in WT neurons, and GluA2 colocalization with St3-mCherry in WT and KO neurons (right). Data represent mean ± SEM, n = 12 neurons, 1-3 dendrites/neuron. Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 5 μm. D Cultured Snx6fl/fl (WT) or Nestin-Cre;Snx6fl/fl (KO) hippocampal neurons co-expressing 3×FM-mCherry-GluA1 (mCherry-GluA1) and 3×FM-EGFP-GluA2 (EGFP-GluA2) were treated with SLF, fixed 2 h or 4 h after release from the temperature block and immunostained for Rab8, mCherry, and EGFP. Shown are representative confocal images of 4 h (left), and quantification of colocalization between GluA1/ GluA2 and Rab8 in WT (upper right) and GluA2 colocalization with Rab8 in WT and KO (lower right). Data represent mean ± SEM, n = 16 neurons, 1-3 dendrites/neuron. Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 5 μm. E Cultured Snx6fl/fl hippocampal neurons co-expressing 3×FM-mCherry-GluA1 (mCherry-GluA1) and 3×FM-EGFP-GluA2 (EGFP-GluA2) were treated with SLF, fixed 2 h or 4 h after release from the temperature block and immunostained for Rab11, mCherry, and EGFP. Shown are representative confocal images of 4 h (left) and quantification of colocalization between GluA1/GluA2 and Rab11 (right) in the soma and dendrites. Data represent mean ± SEM, n = 12 neurons, 1-3 dendrites/neuron. Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 5 μm. F Cultured Snx6fl/fl hippocampal neurons co-expressing 3×FM-mCherry-GluA1 (mCherry-GluA1) and 3×FM-EGFP-GluA2 (EGFP-GluA2) were treated with SLF, fixed 2 h after release from the temperature block and immunostained for SNX6, mCherry and EGFP. Shown are representative confocal images (left) and quantification of colocalization between GluA1/GluA2 and SNX6 (right) in the soma and dendrites. Data represent mean ± SEM, n = 13 neurons, 1-3 dendrites/neuron. Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 5 μm. G Cultured Snx6fl/fl hippocampal neurons co-expressing mCherry (WT), 3×FM-EGFP-GluA2 (EGFP-GluA2) and Flag, or mCherry-2A-Cre (KO), EGFP-GluA2 and Flag, Flag-SNX6 or Flag-SNX6mut were treated with SLF, fixed 4 h after release from the temperature block and immunostained for Rab8, mCherry, and EGFP. Shown are representative confocal images (left) and quantification of GluA2-Rab8 colocalization (right) in the soma and dendrites. Data represent mean ± SEM, n = 12–17 neurons, 1–3 dendrites/neuron. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. Scale bars, 5 μm. (H) Cultured Snx6fl/fl (WT) or Nestin-Cre;Snx6fl/fl (KO) hippocampal neurons co-expressing 3×FM-mCherry-GluA1 (mCherry-GluA1) and 3×FM-EGFP-GluA2 (EGFP-GluA2) were treated with SLF, fixed after release from the temperature block and surface stained for mCherry-GluA1 with antibodies to mCherry, then permeabilized and immunostained for intracellular mCherry and EGFP. Shown are representative confocal images (left) and quantification of the surface/cytoplasm ratio of GluA1, as well as changes in surface GluA1 at 2 h or 4 h relative to 0 h in WT and KO neurons (right). Data represent mean ± SEM, n = 10-17 neurons from two independent experiments, 1-3 dendrites/neuron. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. Scale bars, 5 μm. Intriguingly, both GluA1 and GluA2 reached the PM as early as 2 hours after the cells were released from the 20 °C temperature block (Fig. 6H and Fig. S4). This raised the question: are they transported to the cell surface separately, or do they assemble into the AMPAR complex in a post-ER compartment before being trafficked to the PM? To answer this question, we capitalized on the mis-sorting of GluA2 caused by the loss of SNX6 function, and compared the surface expression of ER-released GluA1 and GluA2 in wild-type and Snx6 KO neurons. Confocal microscopy showed that in KO neurons, not only the surface expression of ER-released GluA2 was inhibited (Fig. S4), the increase in the surface levels of newly synthesized GluA1 over time following ER release was also significantly inhibited even in the presence of newly synthesized GluA2 (Fig. 6H). This suggests that PM delivery of GluA1 requires GluA2 from the post-Golgi secretory pathway. Notably, these results are consistent with the decreased surface expression of endogenous GluA1 in Snx6 KO neurons (Fig. 1C). SNX6-regulated sorting and trafficking of AMPARs is required for synaptic transmission and plasticity GluA2 exerts critical regulatory functions not only in determining the calcium permeability but also in governing receptor assembly and trafficking38. Previous studies have demonstrated that AMPARs lacking GluA2 are less efficiently expressed at the synapse38, while neurons deficient in GluA2 are impaired in LTP expression39. Based on our findings that SNX6 ablation leads to decreased surface expression of GluA1 (Fig. 1C), and that Nestin-Cre;Snx6fl/fl mice exhibit deficits in hippocampal-dependent spatial learning and memory25, we reasoned that reduced availability of GluA2 for GluA1-GluA2 AMPAR complex formation might compromise both constitutive and activity-dependent synaptic delivery of GluA1-containing receptors. Indeed, in Snx6 KO neurons, we observed significant reductions in resting-state synaptic levels of both GluA1 and GluA2 compared to wild-type cells. Moreover, neuronal activity-induced increases in synaptic AMPARs were also attenuated in these KO cells (Fig. 7A, B). Rescue experiments revealed that wild-type SNX6, but not the GluA2-binding deficient mutant, restored both basal and activity-dependent PM trafficking, as well as the synaptic expression of AMPARs (Fig. 7A, B, and Fig. S5). Collectively, these results establish that SNX6-mediated sorting of GluA2 into the secretory pathway is essential for maintaining proper constitutive and activity-regulated synaptic expression of AMPARs.Fig. 7SNX6-mediated sorting of GluA2 into the secretory pathway is required for synaptic transmission and LTP expression.A Cultured Snx6fl/fl hippocampal neurons were co-transfected with constructs expressing mCherry (WT) and EGFP, or mCherry-2A-Cre (KO) and EGFP, EGFP-SNX6 or EGFP-SNX6mut on DIV11. On DIV15, neurons were chemically induced LTP with glycine for 5 min, fixed 30 min after glycine washout, and immunostained for EGFP and surface GluA1. Shown are representative confocal images (upper) and quantification of surface GluA1 in spines, as well as changes in synaptic GluA1 normalized to resting state (lower). Data represent mean ± SEM, n = 11-16 neurons, 330-450 spines. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. Scale bars, 5 μm. B Same as (A) except that cells were immunostained for surface GluA2. Data represent mean ± SEM, n = 12-13 neurons, 380–440 spines. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. Scale bars, 5 μm. C AAV viruses expressing Cre or Cre and SNX6 wild-type (WT) or mutant were stereotaxically injected into the CA1 regions of Snx6fl/fl mouse brain at P0. Acute hippocampal slices were prepared on P14-P21 for electrophysiological recordings. Shown are comparisons of AMPAR-eEPSCs in non-infected (WT) and infected CA1 pyramidal neurons. Representative traces are displayed in the insets. The bar graphs show mean ± SEM (WT vs KO, 4 mice; WT vs KO + SNX6, 4 mice; WT vs KO + SNX6mut, 3 mice). Statistical significance was determined by two-tailed Student’s t-test. Scale bars: 50 pA, 50 ms. D Rectification index of AMPARs in control (WT), Cre expressing (KO), Cre and SNX6 WT (KO + SNX6) or SNX6 mutant (KO + SNX6mut) expressing CA1 pyramidal neurons. Representative traces of AMPAR-eEPSCs recorded at −70mV, 0 mV and +40 mV are displayed in the insets. Data are presented as mean ± SEM (WT vs KO, 4 mice; WT vs KO + SNX6, 4 mice; WT vs KO + SNX6mut, 3 mice). Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 50 pA, 50 ms. E Rescue of the LTP impairment phenotype in Snx6 KO neurons by SNX6 WT but not the GluA2-binding deficient mutant. Shown are relative amplitudes of AMPAR-eEPSCs (mean ± SEM) in non-infected (WT) and virus-infected neurons before and after a whole-cell LTP-pairing protocol. Data were normalized to baseline AMPAR-eEPSC amplitude prior to LTP induction. Sample traces before and after LTP induction are displayed on the top. Bar graphs on the right show percentage of baseline 35–45 min after pairing. Data are presented as mean ± SEM (WT vs KO: WT, n = 5; KO, n = 5; 5 mice; WT vs KO + SNX6: WT, n = 5; KO + SNX6, n = 5; 6 mice; WT vs KO + SNX6mut: WT, n = 5; KO + SNX6mut, n = 5; 6 mice). Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 50 pA, 50 ms. F 8-week-old Snx6fl/fl mice were stereotaxically injected with AAV expressing Cre or Cre and SNX6 wild-type (WT) or mutant in the CA1 regions of brain, and subjected to the Morris water maze test 2 weeks later. Data represent mean ± SEM (number of animals: WT, 10; KO, 11; KO + SNX6, 11; KO + SNX6mut, 9 from two independent experiments). Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. For path length, P = 0.0387 when comparing WT with KO (*), P = 0.9991 when comparing WT with KO + SNX6, P = 0.0333 when comparing WT with KO + SNX6mut (#), P = 0.0453 when comparing KO with KO + SNX6 (&). For escape latency, P = 0.0357 when comparing WT with KO (*), P = 0.9991 when comparing WT with KO + SNX6, P = 0.0467 when comparing WT with KO + SNX6mut (#), P = 0.0214 when comparing KO with KO + SNX6 (&). G Same as (F) except that mice were subjected to contextual fear conditioning. Shown are levels of freezing behavior 1, 2, 3 and 4 days after contextual fear training. Data represent mean ± SEM (number of animals: WT, 10; KO, 11; KO + SNX6, 11; KO + SNX6mut, 9 from two independent experiments). Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. P = 0.0015 when comparing WT with KO (**), P = 0.8076 when comparing WT with KO + SNX6, P = 0.0071 when comparing WT with KO + SNX6mut (##), P = 0.0043 when comparing KO with KO + SNX6 (& &). SNX6-mediated sorting of GluA2 into the secretory pathway is required for synaptic transmission and LTP expression. A Cultured Snx6fl/fl hippocampal neurons were co-transfected with constructs expressing mCherry (WT) and EGFP, or mCherry-2A-Cre (KO) and EGFP, EGFP-SNX6 or EGFP-SNX6mut on DIV11. On DIV15, neurons were chemically induced LTP with glycine for 5 min, fixed 30 min after glycine washout, and immunostained for EGFP and surface GluA1. Shown are representative confocal images (upper) and quantification of surface GluA1 in spines, as well as changes in synaptic GluA1 normalized to resting state (lower). Data represent mean ± SEM, n = 11-16 neurons, 330-450 spines. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. Scale bars, 5 μm. B Same as (A) except that cells were immunostained for surface GluA2. Data represent mean ± SEM, n = 12-13 neurons, 380–440 spines. Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. Scale bars, 5 μm. C AAV viruses expressing Cre or Cre and SNX6 wild-type (WT) or mutant were stereotaxically injected into the CA1 regions of Snx6fl/fl mouse brain at P0. Acute hippocampal slices were prepared on P14-P21 for electrophysiological recordings. Shown are comparisons of AMPAR-eEPSCs in non-infected (WT) and infected CA1 pyramidal neurons. Representative traces are displayed in the insets. The bar graphs show mean ± SEM (WT vs KO, 4 mice; WT vs KO + SNX6, 4 mice; WT vs KO + SNX6mut, 3 mice). Statistical significance was determined by two-tailed Student’s t-test. Scale bars: 50 pA, 50 ms. D Rectification index of AMPARs in control (WT), Cre expressing (KO), Cre and SNX6 WT (KO + SNX6) or SNX6 mutant (KO + SNX6mut) expressing CA1 pyramidal neurons. Representative traces of AMPAR-eEPSCs recorded at −70mV, 0 mV and +40 mV are displayed in the insets. Data are presented as mean ± SEM (WT vs KO, 4 mice; WT vs KO + SNX6, 4 mice; WT vs KO + SNX6mut, 3 mice). Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 50 pA, 50 ms. E Rescue of the LTP impairment phenotype in Snx6 KO neurons by SNX6 WT but not the GluA2-binding deficient mutant. Shown are relative amplitudes of AMPAR-eEPSCs (mean ± SEM) in non-infected (WT) and virus-infected neurons before and after a whole-cell LTP-pairing protocol. Data were normalized to baseline AMPAR-eEPSC amplitude prior to LTP induction. Sample traces before and after LTP induction are displayed on the top. Bar graphs on the right show percentage of baseline 35–45 min after pairing. Data are presented as mean ± SEM (WT vs KO: WT, n = 5; KO, n = 5; 5 mice; WT vs KO + SNX6: WT, n = 5; KO + SNX6, n = 5; 6 mice; WT vs KO + SNX6mut: WT, n = 5; KO + SNX6mut, n = 5; 6 mice). Statistical significance was determined by two-tailed Student’s t-test. Scale bars, 50 pA, 50 ms. F 8-week-old Snx6fl/fl mice were stereotaxically injected with AAV expressing Cre or Cre and SNX6 wild-type (WT) or mutant in the CA1 regions of brain, and subjected to the Morris water maze test 2 weeks later. Data represent mean ± SEM (number of animals: WT, 10; KO, 11; KO + SNX6, 11; KO + SNX6mut, 9 from two independent experiments). Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. For path length, P = 0.0387 when comparing WT with KO (*), P = 0.9991 when comparing WT with KO + SNX6, P = 0.0333 when comparing WT with KO + SNX6mut (#), P = 0.0453 when comparing KO with KO + SNX6 (&). For escape latency, P = 0.0357 when comparing WT with KO (*), P = 0.9991 when comparing WT with KO + SNX6, P = 0.0467 when comparing WT with KO + SNX6mut (#), P = 0.0214 when comparing KO with KO + SNX6 (&). G Same as (F) except that mice were subjected to contextual fear conditioning. Shown are levels of freezing behavior 1, 2, 3 and 4 days after contextual fear training. Data represent mean ± SEM (number of animals: WT, 10; KO, 11; KO + SNX6, 11; KO + SNX6mut, 9 from two independent experiments). Statistical significance was determined by one-way analysis of variance with a Tukey post hoc test. P = 0.0015 when comparing WT with KO (**), P = 0.8076 when comparing WT with KO + SNX6, P = 0.0071 when comparing WT with KO + SNX6mut (##), P = 0.0043 when comparing KO with KO + SNX6 (& &). Given the central role of AMPARs in synaptic plasticity, we hypothesized that LTP would be compromised in Snx6 KO neurons. To determine the functional significance of SNX6-GluA2 interaction in synaptic activity, we conducted molecular replacement in a small subset of neurons by stereotaxic co-injection of adeno-associated viruses (AAV) expressing the Cre recombinase and SNX6 into the hippocampal CA1 region of Snx6fl/fl mice at postnatal day 0 (P0) (Fig. S6A). Electrophysiological recordings of CA1 pyramidal neurons in acute brain slices from virus-injected animals at P14-21 revealed that, first, wild-type but not the GluA2-binding deficient mutant of SNX6 restored both AMPAR-mediated synaptic currents and their linear current-voltage relationship in Snx6 KO neurons (Fig. 7C, D). Second, analysis of LTP induction in the Schaffer collateral-CA1 pathway demonstrated that, Snx6 KO neurons exhibited reduced LTP amplitude compared to controls. Critically, this LTP deficit was rescued by expressing wild-type but not the GluA2-binding deficient mutant of SNX6 (Fig. 7E). These results together establish that the SNX6-GluA2 interaction is required for AMPAR-mediated synaptic transmission and LTP expression. Finally, as synaptic plasticity is the cellular correlate of learning and memory, we sought to determine the physiological relevance of SNX6-mediated GluA2 secretory trafficking. To this end, we performed knockout and rescue experiment by stereotaxically injecting Cre recombinase- and SNX6-expressing AAV into the hippocampal CA1 region of adult Snx6fl/fl mice (Fig. S6B). In line with our previous studies25, behavioral assessments using the Morris water maze and contextual fear conditioning, paradigms known to depend on NMDAR-mediated synaptic plasticity in the hippocampus40–43, revealed that Cre-mediated knockout of Snx6 caused significant deficits in spatial learning and long-term memory retention (Fig. 7F, G). Importantly, these behavioral impairments were rescued by re-expression of wild-type SNX6, but not the GluA2-binding deficient mutant (Fig. 7F, G). Altogether, these data indicate that SNX6-dependent sorting of GluA2 into the secretory pathway is essential for synaptic plasticity and mnemonic processes. Discussion In this study, we have identified a novel function of SNX6 in CNS neurons. Specifically, SNX6 plays a role in sorting of GluA2 into the secretory pathway and regulating the surface expression of GluA1-GluA2 AMPARs in hippocampal excitatory neurons (model, Fig. S7). Previously, we discovered that SNX6 serves as cargo adapter for dynein-driven endosomal transport of the postsynaptic scaffold protein Homer1b/c from the cell body to dendrites25. Whilst SNX6 interacts with Homer1b/c via the PX domain, both its PX and BAR domains interact with the C-terminus of GluA2 (Fig. 3). Moreover, the GluA2-binding deficient mutant of SNX6 could rescue the localization of Homer1b/c to distal dendrites in Snx6 KO neurons (Fig. S3). These findings reveal that in excitatory neurons, SNX6 regulates the synaptic localization of Homer1b/c and AMPARs through distinct mechanisms. Importantly, we have demonstrated that a significant portion of GluA1 and GluA2 synthesized in the ER enter separate trafficking pathways before they assemble into the AMPAR complex for cell surface delivery. Previous studies reported stepwise AMPAR assembly in the ER44. Intriguingly, our electrophysiological analysis of Snx6 KO neurons revealed selective loss of GluA2 from AMPARs on the cell surface. Subsequent cell biological experiments traced this deficit to misrouting of GluA2 into the lysosomal degradation pathway. To investigate the underlying trafficking itinerary, we analyzed the assembly of GluA1-GluA2 AMPARs in the ER and synaptosomes by coIP, followed by the ER retention and inducible release assays to track newly synthesized AMPAR subunits in parallel. Although both assays were semi-quantitative, they indicated that a fraction of ER-derived GluA1 and GluA2 follow distinct routes prior to assembly. Immunofluorescence staining and colocalization analysis further showed that, whereas the vast majority of GluA1 bypasses the Golgi apparatus and proceeds via the ER-ERGIC-RE route, a substantial fraction of GluA2 traverses the Golgi en route to the PM (this study, Hanus et al., and Bowen et al.)18,45. Moreover, relative to GluA1, GluA2 is preferentially sorted into the Golgi-Rab8 secretory pathway (Fig. 6C–E). Importantly, via direct protein-protein interaction, SNX6 mediates incorporation of GluA2 into Rab8 carriers to prevent its mis-sorting into the lysosomal pathway (Fig. 6F, G). It should be noted that the fluorescence microscopy-based ER retention and inducible release assay we used to track GluA1 and GluA2 offers only limited spatiotemporal resolution. Consequently, the subtle differences in subcellular distribution and the mild reduction of GluA2 localization to Rab8-positive vesicles observed in Snx6 KO neurons (Fig. 6B–E) likely represent a substantial underestimate of both the divergent trafficking routes of the two subunits and the extent to which GluA2 secretory trafficking is impaired upon SNX6 loss. Such underestimation notwithstanding, the disruption is sufficient to produce the pronounced deficits in synaptic transmission and plasticity (Fig. 7). Nevertheless, how the post-ER trafficking of GluA1 and GluA2 are coordinated for their subsequent assembly, and where they come together to form the receptor complex prior to PM delivery, remains an open question. Among the four AMPAR subunits, GluA2 is particularly important in determining the single-channel conductance and calcium permeability of the receptor1,2. Alteration in the proportion of GluA2-containing AMPARs, and consequently, the ratio of CP- to CI-AMPARs, represent one of the key regulatory mechanisms in various forms of synaptic plasticity46. We have shown that loss of SNX6 leads to the lysosomal degradation of a significant fraction of GluA2, decreases in both the steady-state surface expression and activity-dependent synaptic delivery of GluA1-containing AMPARs, and impairment of LTP. These findings suggest that the post-ER assembly of GluA1 with GluA2 not only is important for efficient assembly and surface expression of the AMPAR complex for basal synaptic transmission, but also regulates activity-dependent plasma membrane AMPAR trafficking. Interestingly, in GluA2 null mutant mice, NMDAR-independent LTP in the hippocampal CA1 region is enhanced, even as AMPAR-mediated synaptic transmission is impaired47. This observation suggests two key points: first, GluA2 is required for the efficient assembly of AMPAR complexes to enable surface expression; second, Ca2+ influx through GluA2-deficient AMPARs is sufficient to drive LTP induction47. Of note, a clear contrast emerges between phenotypes: partial GluA2 loss in Snx6 KO neurons impairs NMDAR-mediated LTP, whereas complete GluA2 loss in null mutant mice enhances NMDAR-independent LTP. This contrast underscores GluA2’s critical regulatory function in both the constitutive and activity-induced PM trafficking of AMPARs, and the expression of AMPAR-mediated LTP. Further research is needed to understand how GluA2 regulates the assembly and trafficking of the receptor complex in both resting and activated neurons. Several members of the SNX protein family are involved in the trafficking of neurotransmitter receptors and synaptic function. For example, SNX1 regulates the recycling of the AMPAR subunit GLR-1 in C. elegans neurons48. In mammalian CNS neurons, SNX27 is implicated in the endocytic trafficking of the NR2C subunit of NMDARs, and both the basal and activity-dependent synaptic delivery of AMPARs9,49–52. In rat hippocampal neurons, SNX27 mediates endocytic recycling of the cell adhesion molecule β1-integrin and is critical for not only the maintenance of excitatory synapses, but also the structural and functional plasticity of dendritic spines53. In the present study, we have identified SNX6 as a sorting factor for the secretory trafficking of GluA2. The distinct roles of SNXs in intracellular trafficking in neurons might be achieved through their interaction with specific cargo proteins. The mechanism by which SNX6 mediates the sorting of GluA2 from the Golgi complex to secretory endosomes warrants further investigation. Moreover, since neurons rely heavily on the trafficking machinery to regulate their structure and function, elucidating the distinct roles of SNX family members in trafficking of synaptic components could provide valuable insights into the regulatory mechanisms governing synaptic function. Mutations in the GluA2 gene, which have been identified in patients with intellectual disability, autism spectrum disorder, and epilepsy, have been found to cause defects in AMPAR-mediated synaptic transmission and voltage rectification54. Notably, a clinical study of chromosome 14q11-q22 deletion syndrome (OMIM 613457) has identified Snx6 as a candidate gene associated with neurodevelopmental disorders such as intellectual disability, hypotonia, epilepsy and corpus callosum abnormalities55. To advance our understanding of the mechanisms underlying synaptic dysfunction-related diseases for the purposes of prevention and treatment, the pathophysiological functions of SNX6-mediated sorting and trafficking of neurotransmitter receptors and other synaptic proteins necessitate further investigation. Methods Animals All animal experiments were performed in accordance with the guidelines established by the Animal Care and Use Committee of the Institute of Genetics and Developmental Biology, Chinese Academy of Sciences (Approval codes AP2019026 and AP2022070). The Nestin-Cre transgenic C57BL/6 J mice were obtained from the Nanjing Biomedical Institution of Nanjing University56 and bred at the Institute of Genetics and Developmental Biology, Chinese Academy of Sciences. All animals were housed in standard mouse cages at 22–24 °C on a 12 h light/dark cycle with access to food and water freely. Generation of Snx6fl/fl and Nestin-Cre;Snx6fl/fl mice on the C57BL/6 J background were as previously described25. Genotyping of mouse lines was performed using genomic PCR on tail DNA from offspring, utilizing the following primer pairs: loxPF/loxPR, 5’-AGAGTCACTATCAGAGCCTCTTCAG-3’ and 5’-GAGATGCTCAGCACACTTTCTAC-3’ (a PCR product of 366 base pairs in Snx6fl/fl mice and Nestin-Cre;Snx6fl/fl) ; FRTtF/loxtR: 5’-TTTGGCATTATCAAAACGTTGTTGTA-3’ and 5’-GAGATGCTCAGCACACTTTCTCTAC-3’ (a PCR product of 295 base pairs in Nestin-Cre;Snx6fl/fl and none in Snx6fl/fl mice) ; CreF/CreR,5’-TGCCACGACCAAGTGACAGCAATG-3’ and 5’-ACCAGAGACGGAAATCCATCGCTC-3’ (a PCR product of 388 base pairs in Nestin-Cre;Snx6fl/fl mice and none in Snx6fl/fl mice) . Constructs The pAOV-CaMKIIα-EGFP-2A-3Flag, pAOV-CaMKIIα-mCherry-2A-3Flag, pAOV-CaMKIIα-GFP-2A-Cre and pAAV-hSyn-MCS-3Flag-tWPA constructs were purchased from OBiO Technology (Shanghai, China). The pCAG-SEP-GluA1 construct was a gift from Dr. Yong Zhang (Peking University)57. The pCI-SEP-GluA1 (Catalog#24000), pCI-SEP-GluA2 (Catalog#24001)58 and mCherry-FKBP-MTM1 (Catalog#51614) constructs were purchased from Addgene. The mScarlet-GluA1 construct was a gift from Dr. Pingyong Xu (Institute of Biophysics, Chinese Academy of Sciences). The pCMV-Flag-SNX6 construct was a gift from Dr. Hiroyoshi Ariga (Hokkaido University)59. The pCMV-GFP-C1-mSNX6, pCMV-Flag-SNX5, GST-SNX6-FL, GST-SNX6-N (aa 1-181) and GST-SNX6-C (aa 182-406) constructs were described previously22,60. The GFP-ERGIC53 construct was gift from Dr. Liang Ge (Tsinghua University). The pCAG-St3-mCherry construct was generated by replacing SEP-GluA1 in pCAG-SEP-GluA1 with St3 amplified from mouse brain cDNA34 and mCherry. The mScarlet-GluA2 construct was generated by replacing SEP in pCI-SEP-GluA2 with mScarlet61. The pCI-SEP-GluA1ΔC (aa 1-826) was derived from pCI-SEP-GluA1 by truncation of aa 827–907. The pCI-SEP-GluA2ΔC (aa 1-833) was derived from pCI-SEP-GluA2 by truncation of aa 834-883. EGFP-SNX6-mut (S37A/S41A) and EGFP-SNX6-K138A mutants were generated by site-directed mutagenesis using hSNX6 as template. The pAAV-hSyn-EGFP-2A-SNX6-3Flag-tWPA and pAAV-hSyn-EGFP-2A-SNX6mut-3Flag-tWPA viral constructs were generated via a two-step cloning strategy. First, SNX6 or SNX6mut (S37A/S41A) was PCR amplified using the EGFP-SNX6 or EGFP-SNX6mut plasmid DNA as template, respectively, and cloned into the pAAV-CaMKIIα-EGFP-2A-MCS-3Flag vector (OBiO Technology, Shanghai, China) to generate EGFP-2A-SNX6. Because the constructs were too large for viral packaging, the EGFP-2A-SNX6 and EGFP-2A-SNX6mut fragments were PCR amplified and subcloned into the pAAV-hSyn-MCS-3Flag-tWPA vector (OBiO Technology, Shanghai, China) to generate the AAV constructs for in vivo rescue experiments. The 3xFM-mCherry-GluA1 construct was generated by replacing SEP in pCAG-SEP-GluA1 with 3xFM repeats and mCherry. The FM (F36M) mutant was created by site-directed mutagenesis using mCherry-FKBP-MTM1 as template. The 3xFM-EGFP-GluA2 construct was generated by replacing GluA1 coding sequence in 3xFM-mCherry-GluA1 with that of GluA2. The 3xFM-3xFlag-GluA1 and 3xFM-3xFlag-GluA2 constructs were generated by replacing the coding sequence of mCherry or EGFP in 3xFM-mCherry-GluA1 or 3xFM-EGFP-GluA2 with 3xFlag tag. The bacterial expression construct for His-GluA2-C (aa 834-883) was generated by PCR amplification and insertion into pET28a(+). Adeno-associated virus (AAV) viral particles carrying pAOV-CaMKIIα-mCherry-2A-Cre, pAOV-CaMKIIα-EGFP-2A-Cre, pAAV-hSyn-EGFP-2A-SNX6-3Flag-tWPA and pAAV-hSyn-EGFP-2A-SNX6mut-3Flag-tWPA were purchased from OBiO Technology. Antibodies The antibodies used in this study are as follows: mouse anti-GluA1 (Millipore, MAB2263; 1:100 for immunofluorescence (IF) and 1:1000 for Western blotting (WB)); mouse anti-GluA2 (Invitrogen, 32-0300; 1:100 for IF and 1:1000 for WB), rabbit anti-GluA1 (Cell Signaling Technology, 13185; 1:100 for IF), mouse anti-His (Cell Signaling Technology, 2366; 1:2000 for WB), rabbit anti-NMDAR2A (Rockland, 612-401-D89; 1:1000 for WB), rabbit anti-SNX6 (Novus Biologicals, NBP2-13361; 1:100 for IF and 1:1000 for WB), rabbit anti-LAMP1 (Abcam, ab24170; 1:300 for IF), rabbit anti-Rab8 (Sigma-Aldrich, R5530; 1:100 for IF), rabbit anti-Flag (Sigma-Aldrich, F7425; 1:3000 for WB), rabbit anti-Rab11 (Thermo Fisher, 700184; 1:100 for IF), rabbit anti-Calreticulin (Thermo Fisher, MA532131; 1:1000 for WB), mouse anti-GFP (MBL, M048-3; 1:500 for IF and 1:2000 for WB), rabbit anti-GFP (MBL, 598; 1:500 for IF and 1:2000 for WB), rabbit anti-RFP (MBL, PM005; 1:500 for IF and 1:2000 for WB), mouse anti-RFP (MBL, M165-3; 1:500 for IF and 1:2000 for WB), rat anti-GFP (Santa Cruz Biotechnology, sc-101536; 1:3000 for WB), rabbit anti-Rab5B (Santa Cruz Biotechnology, sc-598; 1:100 for IF). Rabbit anti-SNX6 (1:500 for WB) was described previously22. Alexa Fluor dye-conjugated secondary antibodies (1:1500 for IF) were obtained from Molecular Probes (Invitrogen, CA, USA). Cell culture, transfection, and drug treatment Primary mouse hippocampal neurons were cultured as previously described60. Briefly, hippocampi were dissected from postnatal day 0 (P0) C57BL/6 J mice, dissociated using 0.125% trypsin in Hank’s Balanced Salt Solution without Ca²⁺ and Mg²⁺ at 37 °C for 17 min, and triturated in DMEM supplemented with 10% FBS and 10% F12 (GIBCO). Hippocampal neurons were plated on coverslips in 24-well plates that were pre-coated with poly-D-lysine at a density of 2.5–3.0 × 104 cells per well. Four hours after plating, the medium was replaced with serum-free Neurobasal A (NB-A) media supplemented with GlutaMAX, 2% B27, and 0.3% glucose. Half of the medium was replaced every three days until use. For immunofluorescence staining, transient transfections were carried out using LipofectamineTM LTX according to the manufacturer’s instructions (Invitrogen) on the 11th day in vitro (DIV) after plating. Briefly, DNA (0.5 μg per well) was mixed with 0.5 μL of PLUS reagent in 50 μL of Neurobasal A (NB-A) medium, and then mixed with 1.0 μL of Lipofectamine LTX in 50 μL of NB-A medium. After incubating for 20 min, the mixture was added to the neurons and incubate at 37 °C for 1 h. Neurons were then cultured in the original medium at 37 °C in 5% CO2 for 4–5 days. For drug treatment, primary neurons cultured on coverslips were treated with chloroquine (100 μM, C6628, Sigma-Aldrich), leupeptin (100 μg/mL, L2884, Sigma-Aldrich) or MG132 (10 μM, 474790, Sigma-Aldrich) for 4 h. HEK293T cells (Clontech #632180) were cultured as previously described62. Briefly, HEK293T cells were cultured at 37 °C with 5% CO2 in DMEM (C11995500BT, Gibco) supplemented with 10% (vol/vol) FBS (F0193, Sigma-Aldrich). Transfections with plasmid DNA were conducted using LipofectamineTM 2000 (11668019, Invitrogen) when the cell density reached 60%–70%, in accordance with the manufacturer’s instructions. Chemically induced LTP (cLTP) Chemical LTP was performed as previously described (6). Briefly, DIV16 mouse hippocampal neurons were treated with Mg2+-free extracellular solution (in mM: 2.5 KCl, 125 NaCl, 2 CaCl2, 5 HEPES, 33 glucose, 0.2 glycine, 0.02 bicuculline, and 0.003 strychnine, pH 7.4) for 5 min to induce LTP. Cells were then incubated with extracellular solution without glycine for 5 min or 30 min. Immunofluorescence staining, confocal image acquisition and analysis The procedures for immunofluorescence staining were performed essentially as previously described (2). Neurons were fixed with 4% paraformaldehyde (PFA)/4% sucrose in PBS at room temperature (RT) for 10 min at DIV16. Subsequently, they were blocked in PBS containing 2% bovine serum albumin (BSA) and 0.4% Triton X-100 for 1 h at room temperature, followed by incubation with primary antibodies overnight at 4 °C. Appropriate secondary antibodies conjugated with Alexa Fluor 488, Alexa Fluor 555, or Alexa Fluor 647 were then applied for detection. Staining of surface GluA1 and GluA2 was performed as previously described (6) with minor modifications. Briefly, after fixation, neurons were blocked with 5% normal goat serum in PBS for 45 min, and incubated with mouse anti-GluA1 or mouse anti-GluA2 antibodies in PBS with 5% normal goat serum overnight at 4 °C, followed with appropriate fluorescence-conjugated secondary antibodies. For staining of LAMP1, PFA-fixed neurons were blocked in PBS containing 2% BSA and 0.4% Saponin for 30 min at room temperature, and incubated with the primary antibody in PBS with 2% BSA overnight at 4 °C. For staining of Rab8 and Rab11, the cytosolic proteins were extracted from the neurons before fixation. Neurons were rinsed with KHM buffer (2 mM Mg(CH3COO)2, 110 mM CH3CO2K and 20 mM HEPES (pH 7.4)) and extracted with KHM buffer containing 25 mg/mL digitonin for 5 min on ice. Neurons were then rinsed with KHM buffer and fixed with 4% PFA/4% sucrose in PBS for 10 min at RT, blocked with PBS containing 2% BSA and 25 mg/mL digitonin for 15 min, followed by overnight incubation with primary antibodies at 4 °C. For surface staining of mCherry-GluA1 in the inducible-release experiment, after fixation, neurons were blocked with 5% normal goat serum in PBS for 45 min, and incubated with mouse anti-RFP antibodies in PBS with 5% normal goat serum overnight at 4 °C, followed with AlexaFluorTM 647-conjugated secondary antibodies. Neurons were then permeabilized and blocked in PBS containing 2% bovine serum albumin (BSA) and 0.4% Triton X-100 for 1 h at room temperature, followed by incubation with rabbit anti-RFP and rat anti-GFP for 1 h at room temperature, followed with appropriate secondary antibodies for detection. Images were collected using confocal microscopy (Nikon EZ-A1, 100× oil immersion, 4× optical zoom) and analyzed with the NIS-Elements AR software (Nikon, Japan) or Fiji ImageJ (National Institutes of Health, USA). Unless otherwise stated, only segments of the secondary and tertiary dendrites were imaged (distance from the cell body: 30–120 μm; length per segment: 30–40 μm). In each independent experiment, 10 to 15 neurons per group were analyzed, with 2 to 3 segments of dendrites analyzed for each neuron. To quantify total or surface GluA1 or GluA2, dendrites or spines labeled with GFP or mCherry (cell fill) were outlined manually, and the mean intensity of fluorescent signals per unit area was calculated using NIS-Elements AR. For colocalization analysis, Mander’s colocalization coefficient (MCC) was measured using the colocalization module of NIS-Elements AR. Image analyses were performed blinded to genotype and experimental conditions. Live imaging acquisition and analysis For live imaging, neurons were cultured on chambered coverglass (NuncTM Lab-TekTM, Thermo Fisher Scientific). Cells were transfected at DIV11 and imaged live at DIV15 to monitor the motility of GluA1, GluA2 and SNX6. Images were acquired continuously at 0.5 frame/sec using a Nikon TE2000-E equipped with an autofocusing system (Perfect Focus Unit, Nikon), 488 and 561 nm solid laser, 473/543 filter, 60× 1.49 oil immersion lens (Nikon CFI Apochromat TIRF), and an EMCCD camera (iXon Ultra 897, ANDOR) at 37 °C with 5% CO₂, controlled by the μManager software. Inducible-release trafficking Neurons were transfected on DIV11. On DIV14, neurons were pretreated with cycloheximide (10 μM, 239764, Sigma-Aldrich) for 2 h, after which SLF (75 μM, HY-114872, Sigma-Aldrich) was added. Cells were immediately transferred to an incubator at 20 °C for 2 h before being moved back to 37 °C for the indicated duration. Neurons were then fixed and immunostained as described previously. Protein expression and purification His-GluA2-C (aa 834-883), GST, GST-SNX6-FL, GST-SNX6-N (aa 1-181) or GST-SNX6-C (aa 182-406) was expressed in Escherichia coli BL21 (DE3). Bacteria were induced at OD600 of ∼0.6 with 400 μM IPTG for 18 h at 16 °C or 4 h at 37 °C, and collected by centrifugation at 3,000 × g. For His-tagged proteins, bacteria were resuspended in lysis buffer (50 mM NaH2PO4 pH 8.0, 300 mM NaCl, and 10 mM imidazole) supplemented with 1 mM PMSF, and lysed by sonication. Supernatants were collected by centrifugation at 12,000 × g for 20 min at 4 °C and incubated with Ni-NTA resin (70666-4; Millipore) overnight at 4 °C. After washing the Ni-NTA resin five times with wash buffer (50 mM NaH₂PO₄, pH 8.0, 300 mM NaCl, 20 mM imidazole), proteins were eluted with elution buffer (50 mM NaH₂PO₄, pH 8.0, 300 mM NaCl, 250 mM imidazole). For GST-fusion protein, bacteria were resuspended in PBS. For GST-SNX6-FL, GST-SNX6-N (aa 1-181) or GST-SNX6-C (aa 182-406), bacteria were resuspended in PBS supplemented with 0.2% N-lauroylsarcosine and 1 mM PMSF, and lysed by sonication. Supernatants were collected by centrifugation at 12,000 × g for 20 min at 4 °C and incubated with glutathione Sepharose 4B (GE17-0756-01; Sigma-Aldrich) overnight at 4 °C. The glutathione Sepharose 4B resins were washed five times with PBS before GST-pull down experiment. Co-immunoprecipitation (IP) and GST-pull down Co-IP was performed as previously described (7). Briefly, HEK293T cells were rinsed with ice-cold PBS and lysed using lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 5 mM EDTA, 1% NP-40) containing the following protease inhibitors: 10 μM leupeptin, 1 mM benzamide, 1 μg/mL pepstatin A and 1 mM PMSF. Supernatants were collected by centrifugation at 12,000 × g for 20 min at 4 °C. For the GFP or Flag co-IP assay, supernatants were incubated with agarose beads conjugated to anti-GFP antibody (SA070005, Smart-Lifescience, China) or anti-Flag antibody (A2220, Sigma) overnight at 4 °C, followed by five washes with lysis buffer containing 0.1% NP-40. For endogenous SNX6 co-immunoprecipitation (co-IP), mouse hippocampi were homogenized in lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA, 1% NP-40) supplemented with protease inhibitors. The lysates were then centrifuged at 12,000 × g for 20 min at 4 °C to remove cell debris. The supernatant was first incubated with anti-SNX6 antibodies at 4 °C for 2 h, followed by incubation with Dynabeads Protein G (Invitrogen) overnight at 4 °C. Finally, the beads were washed five times with lysis buffer containing 0.1% NP-40. For coIP from subcellular fractions, the ER and synaptosomes fractions were incubated with anti-GluA1 antibodies at 4 °C for 2 h, followed by incubation with Dynabeads Protein G overnight at 4 °C. The beads were washed five times with lysis buffer containing 0.1% NP-40. Immunoprecipitates were boiled in SDS loading buffer (250 mM Tris-HCl pH 6.8, 0.5% bromophenol blue (w/v), 10% SDS (w/v), 50% (vol/vol) glycerol, 5% (vol/vol) 2-mercaptoethanol) at 100 °C, and analyzed by SDS-PAGE and immunoblotting. Subcellular fractionation Mouse brain was homogenized in lysis buffer 1 (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA, 0.32 M sucrose) supplemented with 10 μM leupeptin, 1 mM benzamide, 1 μg/mL pepstatin A and 1 mM PMSF. The homogenate was centrifuged at 1,000 × g for 10 min at 4 °C and the supernatant (S1) was collected. The S1 was centrifuged at 12,000 × g for 20 min to obtain S2 and Pellet 2 (P2). The P2 was resuspended with a syringe (15–20 strokes) in lysis buffer 1. S2 and P2 were loaded onto step density gradients of 25/30/34/43% (w/w) sucrose in lysis buffer and centrifuged at 100,000 × g (MLA-55 rotor, Optima MAX Ultracentrifuge; Beckman Coulter) for 1 h. The interfaces were collected and diluted with sucrose-free lysis buffer, and then centrifuged at 100,000 × g for 30 min. The resulting pellets were resuspended in lysis buffer 1 for further experimentation. RT-PCR and quantitative PCR Total RNA from wild-type (WT) or Snx6 knockout (KO) mouse hippocampi was extracted using Trizol (15 596-026, Invitrogen). First-strand cDNA was synthesized from 1 μg of total RNA using the HiScript III RT SuperMix Kit (R323-01, Vazyme), following the manufacturer’s instructions. Quantitative PCR (qPCR) was conducted using Taq Pro Universal SYBR qPCR Master Mix (Q712-02, Vazyme) following the manufacturer’s guidelines on a CFX96 instrument (Bio-Rad). The primer pairs for GAPDH are 5′-CATCACTGCCACCCAGAAGCTG-3′ and 5′-ATGCCAGTGAGCTTCCCGTTCAG-3′, the primer pairs for GluA1 are 5′- GTGAACATCAGCGACAGCTT-3′ and 5′-GGAGTGATGAAGCACACATGG-3′, the primer pairs for GluA2 are 5′-GCCGAGGCGAAACGAATGA-3′ and 5′-CACTCTCGATGCCATATACGTTG-3′ Electrophysiology To achieve brain region-specific Snx6 knockout, neonatal Snx6fl/fl mice received stereotaxic microinjection of high-titer AAV preparations carrying pAOV-hSyn-mCherry-2A-Cre (AAV-mCherry-2A-Cre) (1 ~ 5 × 1013 IU/ml) targeting the hippocampal CA1 region within 24 h postpartum. Pups were anesthetized via hypothermic analgesia for 4–5 min on ice and positioned in a customized ceramic stereotaxic frame mounted using lambda as the anatomical reference (X, Y) = (0, 0). Viral delivery (10 nL/site) was performed at seven sites (coordinates in mm, [X, Y, Z] = [1.2, 1.2, 1.4/1.0/0.6] and [1.5, 1.0, 1.7/1.3/0.9/0.5]) targeting the hippocampus in each cerebral hemisphere using a microsyringe (Sutter Instrument, CA, USA) with beveled-glass micropipettes. For the rescue experiments, P0 Snx6fl/fl littermates received co-injections of AAV expressing mCherry-2A-Cre and EGFP, EGFP-2A-SNX6-3Flag or EGFP-2A-SNX6mut-3Flag. Approximately 20 nL of the viral solution was injected at each of the seven sites. Post-procedural recovery involved immediate return to maternal cages, with electrophysiological assessments conducted two weeks after injection. To prepare acute brain slices for electrophysiological analyses, mice were anesthetized with tribromoethanol (330–390 mg/kg, T48402, Sigma-Aldrich) and underwent rapid decapitation followed by craniectomy. Brains were immersion-cooled in oxygenated (95% O2/5% CO2) low-Ca2+/high-Mg2+ high-sucrose cutting solution (in mM: KCl 2.5, NaHCO3 25, NaH2PO4 1.25, glucose 10, sucrose 210, Na-Ascorbate 1.3, CaCl2 0.5, MgSO4 7). Coronal 300 μm slices containing hippocampus were prepared using a vibrating tissue slicer (Leica VT1200S; Leica Biosystems, Wetzlar, Deutschland). Slices were immediately transferred to standard artificial cerebrospinal fluid (ACSF) (in mM: NaCl 119, KCl 2.5, MgSO4 1.3, CaCl2 2.5, glucose 11, NaHCO3 26.2, NaH2PO4 1, pH 7.3–7.4) at 33 °C and continuously bubbled with 95% O2/5% CO2. After 30 min of incubation, slices were transferred to a recording chamber with the same extracellular buffer at room temperature (RT, 22 ~ 25 °C). For whole-cell patch clamp recordings, brain slices were continuously perfused with standard ACSF at RT under visual control with infrared-differential interference contrast visualization (BX51WI, Olympus, Tokyo, Japan) with a 40× water-immersion objective. Fluorescently labeled CA1 pyramidal neurons were identified via epifluorescence microscopy (pE-300, CoolLED Ltd, Andover, UK). Whole-cell recordings were performed on an AAV-infected neuron and an adjacent uninfected neuron. Patch clamp data were obtained using a Multiclamp 700B amplifier (Molecular Devices, CA, USA), and signals were low-pass filtered at 2 kHz and digitized at 10 kHz using an Axon Digidata 1550B analog-to-digital board (Molecular Devices, CA, USA). All electrophysiological recordings were carried out blindly and all numbers in the data figures indicate the number of recorded neurons. Chemical reagents were purchased from Sigma-Aldrich unless specified. Data were collected and analyzed using the pClamp 10.6 software (Molecular Devices, CA, USA). To analyze Schaffer collateral-evoked AMPAR excitatory postsynaptic currents (AMPAR-eEPSCs) and rectification index, cells were recorded with 3–7 MΩ borosilicate electrodes filled with the intracellular solution (in mM: CsMeSO3 135, HEPES 10, EGTA 0.3, QX-314 5, MgATP 4, NaGTP 0.3, and NaCl 8, Spermine-Cl4 0.1, pH 7.3), following stimulation of Schaffer collaterals with a bipolar stimulation electrode placed in the stratum radiatum of CA1 region. AMPAR-eEPSCs were measured at a holding potential of −70 mV, and rectification index was calculated as the ratio of the slopes of the lines connecting AMPAR-eEPSC values from 0 mV to 40 mV and from -70mV to 0 mV in the presence of 50 μM D-AP5(D-APV, HY-100714A, MedChemExpress, NJ, USA) to block NMDARs, and 10 μM Bicuculline (505875, Sigma-Aldrich) to block GABAA receptors. The calculation was as follows:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${RI}=\frac{7({I}_{40}-{I}_{0})}{4({I}_{0}-{I}_{-70})}$$\end{document}RI=7(I40−I0)4(I0−I−70)where IX represents the EPSCs recorded at a holding potential of x mV. Rectification index values were measured using the subtracted currents at −70 mV, 0 mV and +40 mV and calculated as described above. LTP of AMPAR eEPSCs was induced with a protocol in which Schaffer collateral axons were stimulated with theta burst at 2 Hz for 90 s, while neurons were depolarized to 0 mV, after 3- to 5-min steady baseline recording, but no more than 7 min when achieved whole-cell configuration to avoid “wash-out” of LTP. Synaptic responses were monitored for 45 min post-induction at -70 mV. Stereotaxic injection and behavioral tests Stereotaxic injection of adult mice was performed as previously described (6). Briefly, eight-week-old naive male and female Snx6fl/fl mice were anesthetized with isoflurane (1–2% mixed with oxygen) and placed in a microinjection system (World Precision Instruments, FL, USA). After sterilized with iodophors and 75% (vol/vol) alcohol, a midline incision was made on the scalp of the mouse, between the ears. Holes were drilled into both sides of the skull for needle insertion. The viral particles expressing mCherry-2A-Cre and EGFP-2A-FLAG, EGFP-2A-SNX6-3Flag or EGFP-2A-SNX6mut-3Flag (1 μL, 2.0 × 1012 viral genomes/mL) were injected into the hippocampal CA1 region, using the following coordinates relative to bregma: posterior, 2.0 mm; lateral, 1.8 mm; ventral, 1.4 mm. The injection rate was set at 0.125 mL/min, and the needle was maintained in place for 5 min before withdrawal, after which the skin was sutured using surgical thread. The animal behavior experiments were conducted two weeks after the viral injection, with procedures previously described (6). For the Morris water maze experiment, when the flag was visible on the platform, the mice were exposed to the visible trial. Once the flag was removed, the platform, which was submerged 1 cm below the water surface, remained fixed in position. The mice underwent training for 7 days, during which latency to reach the platform and path length were recorded. If a mouse failed to locate the platform within 60 seconds, it was placed on the platform for 20 seconds. After training, the platform was removed, and the number of times each mouse crossed the 1.5× platform circle area within 60 seconds was quantified. The swimming trajectories in the pool were monitored and analyzed using an automated system (Smart 3.0, Panlab SMART video tracking system, Barcelona, Spain). The fear conditioning test was conducted using a standard fear conditioning apparatus (Harvard Apparatus, MA, USA). Mice were placed in a fear conditioning chamber and allowed to explore freely for 2 minutes, followed by a 30-second auditory conditioned stimulus (CS). Immediately after CS offset, a 2-second, 0.9 mA foot shock was delivered. Following a 60-second inter-trial interval (ITI), a second 30-second auditory CS was presented. On each of the subsequent four days, mice were re-exposed to the fear conditioning chamber and allowed to explore freely for 2 minutes. A single 30-second CS was then delivered, and freezing behavior during the CS was quantified as the percentage of time spent freezing. The chamber was cleaned with 70% ethanol between mice to eliminate olfactory cues. The freezing behavior was automatically scored and analyzed using FREEZING software (Harvard Apparatus, MA, USA). Statistics and reproducibility All data are presented as means ± standard error of the mean (SEM) from at least three independent biological replicates. Statistical analyses were performed using GraphPad Prism 5 (GraphPad Software). For comparisons between two samples, unpaired Student’s t-test was used unless otherwise noted. One-way analysis of variance (ANOVA) with a Dunnett’s or Tukey’s multiple comparisons were used to evaluate statistical significance for three or more groups of samples. A P value of less than 0.05 was considered statistically significant. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Transparent Peer Review file Supplemental information Description of Additional Supplementary Files Supplementary Data 1 Video 1 Video 2 Video 3 Video 4 Reporting Summary Transparent Peer Review file Supplemental information Description of Additional Supplementary Files Supplementary Data 1 Video 1 Video 2 Video 3 Video 4 Reporting Summary