Retromer is an evolutionarily conserved endosomal trafficking complex that mediates the retrieval of cargo proteins from a degradative pathway for sorting back to the cell surface. To promote cargo recycling, the core retromer trimer of VPS (vacuolar protein sorting)26, VPS29 and VPS35 recognises cargo either directly, or through an adaptor protein, the most well characterised of which is the PDZ [postsynaptic density 95 (PSD95), disk large, zona occludens] domain-containing sorting nexin SNX27. Neuroligins (NLGs) are postsynaptic trans-synaptic scaffold proteins that function in the clustering of postsynaptic proteins to maintain synaptic stability. Here, we show that each of the NLGs (NLG1–3) bind to SNX27 in a direct PDZ ligand-dependent manner. Depletion of SNX27 from neurons leads to a decrease in levels of each NLG protein and, for NLG2, this occurs as a result of enhanced lysosomal degradation. Notably, while depletion of the core retromer component VPS35 leads to a decrease in NLG1 and NLG3 levels, NLG2 is unaffected, suggesting that, for this cargo, SNX27 acts independently of retromer. Consistent with loss of SNX27 leading to enhanced lysosomal degradation of NLG2, knockdown of SNX27 results in fewer NLG2 clusters in cultured neurons, and loss of SNX27 or VPS35 reduces the size and number of gephyrin clusters. Together, these data indicate that NLGs are SNX27–retromer cargoes and suggest that SNX27–retromer controls inhibitory synapse number, at least in part through trafficking of NLG2.
Retromer is a highly conserved endosomal sorting complex involved in the retrieval and retrograde transport of endocytosed transmembrane proteins (cargoes) to the trans-Golgi network or cell surface [1,2]. The core cargo recognition complex of retromer comprises three vacuolar protein sorting (VPS) proteins, VPS26, VPS29 and VPS35, which, in association with multiple other proteins, including members of the sorting nexin (SNX) family [3,4], concentrate cargoes on the endosomal surface for subsequent sorting. For many cargo proteins, the core retromer complex is sufficient for sequence-specific cargo recognition. Additionally, however, cargoes can be coupled to retromer through adaptor proteins, the best-characterised of which is SNX27.
SNX27 is the only SNX to contain a PDZ [postsynaptic density 95 (PSD95), disk large, zona occludens] domain and binds to type 1 PDZ-binding motifs of interacting proteins . To simultaneously interact with a cargo protein and the retromer complex, the PDZ domain of SNX27 has an extended core fold allowing it to bind VPS26 non-canonically . Through this mechanism, SNX27 links PDZ ligand-containing cargo proteins to the endosomal trafficking machinery, rescuing them from lysosomal degradation and promoting recycling to the plasma membrane [4,7,8].
The first validated transmembrane cargo of SNX27–retromer was the β2-adrenergic receptor . Subsequent proteomics and structural analyses have identified over 400 potential SNX27 transmembrane protein ligands [9,10], suggesting that SNX27 acts as a major sorting determinant for a wide array of cargo proteins. Furthermore, while many of these SNX27 cargoes require retromer for their retrieval, many are insensitive to retromer depletion, suggesting that, for some cargo proteins, SNX27 can promote recycling independently of retromer .
The role and cargoes of retromer and SNX27 are of particular interest in neurons because of the highly complex and stringently regulated trafficking of membrane proteins required to maintain neuronal excitability and support synaptic transmission. SNX27–retromer cargoes identified in neurons include α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)- and N-methyl-d-aspartate (NMDA)-type glutamate receptors, which support the vast majority of excitatory synaptic transmission [11–16] and Kir3 potassium channels [17,18]. Furthermore, consistent with an important role in neuronal function, SNX27 heterozygous knockout mice have diminished numbers of AMPA and NMDA receptors and exhibit learning and memory deficits . These workers also reported that SNX27 is reduced in human Down's syndrome brains and up-regulating SNX27 rescues defects in Down's syndrome model mice . Moreover, point mutations in the retromer components VPS26 and VPS35 lead to familial Parkinson's disease , and levels of retromer are reduced in the brains of humans with Alzheimer's disease , suggesting that defective retromer-dependent trafficking pathways may directly lead to various neurodegenerative disease states.
While the actions of SNX27–retromer at excitatory synapses is the subject of extensive research, potential roles at inhibitory synapses have been less widely explored. However, a recent proteomics study identified the inhibitory synapse-specific protein neuroligin 2 (NLG2) as a candidate SNX27–retromer cargo , and it has subsequently been reported that the core retromer component VPS35 associates with NLG2 in cultured hippocampal neurons .
NLGs are postsynaptic cell adhesion molecules that act as ligands for presynaptic neurexins to form trans-synaptic complexes [22,23]. NLG2 is well established to play key roles in the maturation and function of inhibitory synapses [24,25]. Importantly, NLG2 interacts with the Rho GDP/GTP-exchange factor collybistin to recruit and cluster the inhibitory postsynaptic scaffolding protein gephyrin [26–28], which is required for the synaptic localisation of glycine receptors (GlyRs) and major GABAA receptor (GABAAR) subtypes [29,30]. Correspondingly, mice lacking NLG2 display anxiety phenotypes, have dysfunctional inhibitory postsynaptic structures  and show a decrease in both postsynaptic gephyrin and GABAAR cluster density in hippocampal neurons . Moreover, mutations in NLG2 are implicated in a range of psychiatric disorders including autism spectrum disorder and schizophrenia [33–36].
In this study, we show that SNX27 interacts with each of the NLG proteins and rescues NLG2 from lysosomal degradation. Moreover, loss of SNX27, or of the core retromer component VPS35, perturbs the clustering of NLG2 and, correspondingly, reduces inhibitory synapse number as assessed by staining for the inhibitory synapse marker protein gephyrin. These data highlight SNX27 as a novel regulator of NLG trafficking, and SNX27–retromer as an important determinant of inhibitory synapse number, at least in part, through regulation of NLG2.
The following antibodies were used in the present study: anti-β-actin (mouse monoclonal, Sigma–Aldrich, clone AC-15, A5441, 1:10 000), anti-epidermal growth factor receptor (EGFR) (rabbit monoclonal, Abcam, ab52894, 1:1000), anti-Gephyrin (mouse monoclonal, Synaptic Systems, clone mAb7a, 147 021, 1:1000), anti-GFP (chicken polyclonal, Abcam, ab13970, 1:600), anti-GFP (rat monoclonal, ChromoTek, 3H9, 1:5000), anti-GluA2 (rabbit polyclonal, Synaptic Systems, 182 103, 1:1000), anti-GST (goat polyclonal, GE Healthcare, 27-4577-01, 1:10,000), anti-HA (mouse monoclonal, Sigma–Aldrich, clone HA-7, H3663, 1:1000), anti-HA (rabbit polyclonal, Sigma–Aldrich, H6908, 1:100), anti-LAMP1 (rabbit polyclonal, Abcam, ab24170, 1:1000), anti-NLG1 (mouse monoclonal, Neuromab, N97A/31, 1:500), anti-NLG2 (mouse monoclonal, Synaptic Systems, clone 5E6, 129 511, 1:2000 WB, 1:600 ICC), anti-NLG3 (mouse monoclonal, Neuromab, N110/29, 1:500), anti-GluN1 (rabbit monoclonal, Cell Signaling, D65B7 5704, 1:2000), anti-SNX27 (mouse monoclonal, Abcam, clone 1C6, ab77799, 1:1000), anti-SNX27 (rabbit polyclonal, a kind gift from Dr Martin Playford (NIH, U.S.A., 1:2000) and anti-VPS35 (rabbit polyclonal, Abcam, ab97545, 1:2000).
DNA constructs and shRNA
For Sindbis virus production, human SNX27 was cloned into the vector pSinRep5(nsP2S) . The intracellular C-terminal (CT) of NLG1 (amino acids 719–843), NLG2 (amino acids 700–836) and NLG3 (amino acids 731–848) were cloned from rat whole brain cDNA into the vector pEGFP-C3. The PDZ-AAA mutants of NLG-CT were created by site-directed mutagenesis of the NLG1–3 CT pEGFP-C3 constructs. For the purification of recombinant protein, human SNX27 was subcloned into the vector pET28a and rat NLG2-CT-WT and -PDZ-AAA into pGEX-4T-1. For lentivirus production, shRNAs driven by an H1 promoter were produced for the knockdown of rat SNX27 (target sequence 5′-aagaacagcaccacagaccaa-3′), rat VPS35 (target sequence 5′-aacagtggagatattcaataa-3′ ) and rat NLG2 (target sequence 5′-gcaagttcaacagcaaggaaa-3′), and oligonucleotides were cloned into a modified pXLG3-GFP viral vector . To ensure specificity, shRNA target sequences were BLAST searched against the rat genome before use. For rescue experiments, GFP-tagged rat SNX27 or VPS35 were silently point-mutated at three consecutive codons in the shRNA target sequence to make them insensitive to the shRNA and cloned into the corresponding shRNA construct in place of GFP. The fidelity of all constructs was confirmed by DNA sequencing.
Cell culture and transfection
Primary neuronal cultures were prepared from embryonic day (E) 18 Wistar rat brains. Dissociated cortical cells were grown in 35 mm plastic dishes (5.5 × 105 cells per well), and hippocampal cells were grown on 25 mm glass coverslips (1.6 × 105 cells per coverslip) in plating medium (neurobasal medium, 5% horse serum, 2% B27 and 1% Glutamax), which was exchanged for feeding medium 16 h after plating (neurobasal medium, 2% B27 and 0.4% Glutamax). HEK293T cells were cultured in a T75 flask containing Complete DMEM (DMEM, 10% foetal bovine serum [FBS], 1% l-glutamine and 1% penicillin/streptomycin) and regularly passaged. For transfection, HEK293T cells were grown in 35 mm plastic dishes in complete DMEM which was replaced prior to transfection with transfection DMEM (DMEM, 10% FBS and 1% l-glutamine). All cells were cultured in a humidified incubator at 37°C and 5% CO2. Hippocampal neurons (for all imaging experiments) and HEK293T cells were transfected using Lipofectamine 2000 (ThermoFisher).
Lentivirus and Sindbis virus production
Cell lysing and immunoprecipitation
All solutions were pre-chilled to 4°C and all steps were carried out on ice. Cells were washed with 1× phosphate-buffered saline (PBS) and lysed in lysis buffer [50 mM Tris, pH 7.4, 150 mM NaCl, 0.5% Triton X-100 and cOmplete protease inhibitors (Roche) in ddH2O]. Lysates were briefly vortexed, kept on ice for 20 min to allow for protein solubilisation and then clarified by centrifuging at 20 000×g for 20 min at 4°C. Lysate supernatants were either used for Western blotting or for immunoprecipitation, in which case a 25 μl input sample was retained. GFP-trap beads (ChromoTek) were washed twice with lysis buffer by centrifuging at 1000×g for 2 min at 4°C. Lysate was then added to the beads and incubated for 1 h at 4°C after which the beads were washed three times with lysis buffer. Samples were prepared for Western blotting by being resuspended in 2× Laemmli sample buffer and boiled for 5 min at 95°C.
Surface biotinylation experiments were carried out as described recently . Briefly, neurons were washed in ice-cold PBS, followed by labeling with 0.3 mg/ml Sulfo-NHS-SS-Biotin (ThermoFisher). After extensive washes with PBS, and a single wash with 50 mM ammonium chloride to quench any unreacted biotin, cells were lysed and cleared lysates were prepared as described above. A lysate sample was then taken, and equal amounts of lysate were incubated with streptavidin-agarose (Sigma–Aldrich) on a wheel for 1 h at 4°C. Beads were then washed four times with lysis buffer, and proteins were eluted by heating in 2× Laemmli buffer for 10 min at 95°C.
Samples for Western blotting were resolved by sodium dodecyl sulfate–polyacrylamide gel electrophoresis and separated proteins were transferred to Immobilon-PVDF (polyvinylidene difluoride) 0.45 µm membrane (Merck Millipore). After protein transfer, the membrane was briefly washed with TBS-T (1× Tris-buffered saline, 0.1% Tween), blocked in 5% non-fat milk powder and incubated with primary antibody diluted in 5% non-fat milk powder for 1 h at RT. Following incubation, the membrane was briefly washed three times with TBS-T and incubated with HRP-conjugated secondary antibody (Sigma–Aldrich, 1:10 000) diluted in 5% non-fat milk powder for 45 min at RT. Afterwards, the membrane was washed three times for 5 min in TBS-T at RT and incubated for 1 min at RT with an enhanced chemiluminescence substrate. Protein bands were visualised using an Odyssey Fc (LI-COR) detection system or in a dark room by exposing X-ray film (Kodak) to the membrane and developing the film with an automatic medical X-ray film processor (Konica).
Recombinant protein purification and in vitro interaction assay
SNX27 in the vector pET28a and NLG2-CT-WT and -PDZ-AAA in pGEX-4T-1 were transformed into competent BL21 Escherichia coli using standard procedures. Single colonies were grown in ampicillin-containing 2×YT broth at 37°C and protein production was induced with 25 µM isopropyl β-d-thiogalactopyranoside for 15 h at 18°C. E. coli were then centrifuged at 3000×g for 20 min at 4°C to pellet cells and resuspended in lysis buffer (25 mM Tris, pH 7.5, 150 mM NaCl, 10% glycerol and cOmplete protease inhibitors in ddH2O + 10 mM imidazole for His-fusion protein only) and supplemented with Triton X-100 to a final concentration of 1%. Cells were lysed by sonication, left on ice for 15 min to allow for solubilisation and clarified by centrifugation at 50 000×g for 30 min at 4°C. Glutathione or Ni-NTA Sepharose 4 Fast Flow beads were washed twice with wash buffer (25 mM Tris, pH 7.5, 150 mM NaCl, 10% glycerol and 1% Triton X-100 in ddH2O), and bacterial lysate was incubated on washed beads for 2 h at 4°C. Following incubation, the beads were washed four times with wash buffer and fusion proteins were eluted with three 5-min incubations at RT with elution buffer (wash buffer supplemented with 10 mM glutathione for GST-fusion proteins or wash buffer without glycerol and Triton X-100 supplemented with 250 mM imidazole for His-fusion protein). To test for interactions, Glutathione Sepharose 4 Fast Flow beads were washed twice with bead wash buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1% Triton X-100 and cOmplete protease inhibitors in ddH2O). About 1 µg GST or GST-NLG2-CT in binding buffer [wash buffer containing 0.1% bovine serum albumin (BSA)] was incubated with the washed beads for 1 h at 4°C. Following immobilisation, beads were washed four times with bead wash buffer, then incubated with 50 ng of His-SNX27 in binding buffer for 2 h at 4°C and subsequently washed three times with bead wash buffer. Approximately 8 ng of His-SNX27 in binding buffer was used as an input and all samples were prepared for Western blotting by resuspension in 2× Laemmli sample buffer and boiled for 5 min at 95°C.
Hippocampal neurons on glass coverslips were transfected with DNA constructs on DIV 16, and fixed and immunostained for confocal microscopy on DIV 21. Neurons were fixed in warm 4% paraformaldehyde (PFA) in 1× PBS for 15 min and then washed three times with 1× PBS. The PFA was quenched with 1 ml of 100 mM glycine in 1× PBS after which the neurons were washed three times with 1× PBS. Neurons were then permeabilised and blocked with TBP (0.1% Triton X-100 and 2% BSA in 1× PBS) for 20 min. After blocking, the neurons were incubated with primary antibodies diluted in 3% BSA in 1× PBS for 45 min and washed three times with 1× PBS. The neurons were then incubated with Cy2-, Cy3- or Cy5-conjugated secondary antibodies (Sigma–Aldrich, 1:600) diluted in 3% BSA in 1× PBS for 45 min. After four washes with 1× PBS, the coverslips were mounted on glass microscope slides using Fluoromount-G with DAPI (ThermoFisher) and stored at 4°C until imaged.
Image acquisition and quantification
Confocal images were captured using a 63× HCX PL APO CS oil-immersion objective on a Leica SP5-II confocal laser scanning microscope attached to a Leica DMI 6000 inverted epifluorescence microscope. All parameters, including excitation and gain, were kept constant across all neurons and coverslips imaged in an experiment to generate a complete dataset. A full dataset was acquired by imaging eight neurons from three different coverslips from three independent dissections per experimental group unless otherwise stated. Image analysis was performed using ImageJ software (NIH) in combination with the Fiji image processing package. For each neuron, the Z-stack was projected at maximum intensity and three regions of interest (ROIs) were drawn free-hand around secondary dendrites. For cluster analysis, the channel was thresholded to create a binary image and the cluster size (area in µm2) and density per ROI were measured using the Analyse Particles Tool. For quantification of colocalisation, the Pearson's correlation for each ROI was measured using the Colocalisation coloc2 Tool. Numerical values obtained in ImageJ were processed in Microsoft Excel and exported to GraphPad Prism (version 7.0) for statistical analysis.
For statistical analysis of cluster size, a cumulative frequency distribution analysis showed a non-parametric distribution and values were therefore compared for statistical significance using a Kolmogorov–Smirnov or Kruskal–Wallis test. Statistical analysis of cluster density and Pearson's correlation was performed using a Student's t-test or ANOVA. Statistical analysis of Western blots was performed using one-sample t-test followed by Bonferroni post hoc test to account for multiple comparisons. Data are presented as mean ± SEM and parameters for significance are *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001.
SNX27 interacts with NLG1, 2 and 3
We first investigated if SNX27 binds to NLGs by expressing GFP-SNX27 in cultured cortical neurons using Sindbis virus. Neurons were infected with virus at DIV 17, and GFP-SNX27-interacting proteins were isolated by GFP-trap assays 20- to 22-h post-infection, followed by Western blotting for NLG1, 2 and 3. As shown in Figure 1A, all of the endogenous NLG subtypes co-immunoprecipitated with GFP-SNX27. To determine whether this interaction was mediated by the intracellular CT domains of the NLGs, we next expressed the CT domains of each NLG fused to GFP in HEK293T cells and isolated interacting proteins by GFP-trap followed by Western blotting for endogenous SNX27 (Figure 1B). Again, NLG1–3 CT interacted with SNX27, but the affinity for SNX27 was highest in the case of NLG2.
SNX27 binds NLG1–3 through a PDZ interaction.
To define if NLG binding to SNX27 was mediated by PDZ interactions, we mutated the CT PDZ-ligand in the GFP-NLG1–3 CT constructs (Threonine−2, Arginine−1, Valine0) to three alanines and expressed these GFP-tagged PDZ-AAA mutants in HEK293T cells. In all cases, the SNX27–NLG interaction was completely lost in the GFP-NLG-CT-PDZ-AAA mutants (Figure 1C), consistent with NLG1–3 interacting with SNX27 through their CT PDZ ligands.
SNX27 binding to NLG2 is directly mediated by a type 1 PDZ interaction
Given that NLG2 bound most robustly to SNX27 in our assays, and the fact that it is a key component of inhibitory synapses, we focused on the properties and potential roles of SNX27–retromer interaction with NLG2. To exclude any contribution of an indirect association and confirm that NLG2 and SNX27 bind directly via a PDZ interaction, we performed in vitro binding assays using purified proteins. Free GST, GST-NLG2-CT-WT, GST-NLG2-CT-PDZ-AAA and His-SNX27 were expressed and purified from E. coli. The purified GST-fusion proteins were incubated with His-SNX27 and isolated on glutathione beads followed by Western blotting. As shown in Figure 2, recombinant His-SNX27 bound to GST-NLG2-CT-WT but not to GST-NLG2-CT-PDZ-AAA, confirming a direct PDZ-mediated interaction between SNX27 and NLG2.
SNX27 and NLG2 bind directly through a PDZ interaction.
SNX27 knockdown decreases total levels of NLG1–3 and known synaptic SNX27–retromer cargoes
Knockdown of SNX27, or of the core retromer component VPS35, decreases total levels of cargo proteins in clonal cell lines by preventing their retrieval from lysosomal degradation . We therefore examined the effects of knocking down SNX27 or VPS35 in cultured rat cortical neurons. Neurons were infected at DIV 9–11 with lentiviral constructs that express shRNAs against SNX27 (shSNX27), VPS35 (shVPS35) or a control non-targeting shRNA. Seven-days post-infection, neurons were lysed and protein levels were analysed by Western blotting. shSNX27 and shVPS35 were very efficient and significantly decreased levels of SNX27 and VPS35 to 13.1 ± 3.26 and 15.6 ± 3.49%, respectively, of control cells (Figure 3A,C).
Loss of SNX27–retromer components decreases total levels of NLGs and of known SNX27–retomer cargoes in cortical neurons.
Consistent with SNX27 playing a key role in NLG recycling, SNX27 knockdown dramatically reduced levels of NLG1–3, indicating that these proteins are rescued from degradation in a manner dependent on their interaction with SNX27 (Figure 3B,D). In addition, we tested the levels of subunits of AMPARs and NMDARs, which have both been previously reported as SNX27–retromer cargoes [11,13–16]. Consistently, shSNX27 significantly reduced total levels of the AMPAR subunit GluA2 and NMDAR subunit GluN1. Total levels of EGFR, which is not a retromer cargo , were unaffected by SNX27 or VPS35 knockdown (Figure 3B,E). Importantly, the reduced levels of NLG2 following SNX27 knockdown were rescued to control levels by concomitant re-expression of a GFP-tagged shRNA-resistant SNX27 (Supplementary Figure S1A), confirming that the effects of the knockdown were specific to loss of SNX27.
SNX27 traffics NLG2 independently of retromer
Interestingly, knockdown of the core retromer component VPS35 had a different profile to SNX27 knockdown. NLG1 and NLG3 levels were significantly reduced by shVPS35, albeit not to the same extent as for shSNX27, suggesting that their trafficking is partly dependent on the core retromer trimer (Figure 3B,D). Levels of GluA2 were also reduced to a similar extent by shSNX27 and shVPS35. Again, we validated the specificity of these effects by confirming that the loss of NLG1 and NLG3 upon VPS35 knockdown was rescued by expression of a GFP-tagged shRNA-resistant VPS35 (Supplementary Figure S1B).
In stark contrast, however, total levels of NLG2 and of the NMDAR subunit GluN1, which were reduced by >50% by shSNX27, were unaffected by shVPS35. Intriguingly, these results suggest that SNX27 selectively acts via a mechanism independent of the core retromer VPS trimer in the recycling of NLG2 and GluN1 (Figure 3B,E).
SNX27 knockdown reduces NLG2 surface expression
To determine whether the loss of total NLG2 resulting from SNX27 knockdown also led to a corresponding loss of NLG2 from the cell surface, we performed surface biotinylation experiments. Neurons were infected with control, shSNX27 or shVPS35 viruses at DIV 9–11, and surface proteins were labeled with biotin 7 days later. Cells were then lysed, and surface proteins were isolated on streptavidin beads. SNX27 knockdown decreased total NLG2 levels with a corresponding decrease in surface NLG2 levels (Supplementary Figure S2). VPS35 knockdown, on the other hand, did not alter either total NLG2 or surface NLG2 levels (Supplementary Figure S2).
SNX27 knockdown increases lysosomal degradation of NLG2
To investigate the mechanism underlying SNX27 regulation of total NLG2 levels, we analysed colocalisation of NLG2 with the lysosomal marker protein LAMP1 following SNX27 or VPS35 knockdown. Neurons were transfected with shRNA constructs at DIV 16 and then fixed and stained for NLG2 and LAMP1 5 days later. As expected from knockdown of SNX27 leading to a loss of total NLG2 protein, shSNX27 resulted in a significant increase in NLG2/LAMP1 colocalisation (Figure 4A,B). However, consistent with the lack of effect of VPS35 knockdown on NLG2 (Figure 3B,D), shVPS35 did not increase NLG2/LAMP1 colocalisation (Figure 4A,B). Importantly, knockdown of SNX27 or VPS35 had no effect on LAMP1 cluster density or cluster size (Supplementary Figure S3), indicating that loss of these proteins was not altering global levels of lysosomal degradation. These data indicate that SNX27, but not VPS35/retromer, is involved in retrieving NLG2 from lysosomal degradation.
Knockdown of SNX27 increases lysosomal degradation of NLG2.
SNX27 knockdown decreases the density but not the size of NLG2 clusters
NLG2 is predominantly localised at inhibitory synapses, so we next investigated how SNX27 or VPS35 knockdown affects NLG2 distribution. DIV 16 hippocampal neurons were transfected with GFP-tagged shSNX27 or shVPS35, fixed at DIV 21, immunostained for NLG2, and imaged by confocal microscopy (Figure 5). For these experiments, we chose to analyse NLG2 clusters in secondary dendrites, defined as dendritic branches emerging from primary dendrites which, in turn, emerge from the soma. This approach allowed us to unambiguously identify single NLG2 clusters away from the high density of clusters around the soma. These analyses revealed a significant decrease in the number of NLG2 puncta in both SNX27 and VPS35 knockdown neurons (Figure 5A,B). Furthermore, there was a significant decrease in the size of NLG2 puncta in the VPS35 knockdown cells. Surprisingly, however, the size of NLG2 puncta was not significantly reduced by SNX27 knockdown (Figure 5A,C).
Loss of SNX27–retromer function decreases NLG2 cluster size and density in hippocampal neurons.
SNX27 or VPS35 knockdown decreases both the density and size of gephyrin clusters
NLG2 has been reported to recruit and directly bind gephyrin, a scaffolding protein that is a central component of inhibitory synapses , so we next tested the effects of SNX27 or VPS35 knockdown on gephyrin clustering. Consistent with their effects on NLG2, ablation of either SNX27 or VPS35 significantly decreased the size and density of gephyrin clusters in hippocampal neurons, demonstrating that both SNX27 and retromer play key roles in regulating the number of inhibitory synapses (Figure 6).
Loss of SNX27–retromer function decreases gephyrin cluster size and density in hippocampal neurons.
NLG2 knockdown significantly decreases the size and density of gephyrin clusters
Finally, to determine if the effects of SNX27 or VPS35 knockdown on gephyrin clustering are mediated by NLG2, we directly knocked down NLG2. Our shNLG2 caused a ∼50% decrease in NLG2 levels (Figure 7A,B), similar to the effect of shSNX27 on total NLG2 levels (Figure 3). Interestingly, ∼50 loss of NLG2, directly mimicking the effect of SNX27 loss, resulted in a reduction in the density and size of gephyrin clusters, similar to loss of SNX27 or VPS35 (Figure 7C–E). Taken together, our data are consistent with a model whereby loss of SNX27 or VPS35 affects inhibitory synapse number via altered stability or clustering of NLG2, and highlight SNX27 and retromer as novel determinants of inhibitory synapse number, at least in part, through regulation of NLG2.
NLG2 knockdown decreases gephyrin cluster size and density in hippocampal neurons.
It is well established that the regulation and trafficking of NLGs is an important determinant of synapse formation and stability [22,23]. However, relatively little is known about how NLG trafficking is mediated, and few direct NLG-interacting proteins have been identified.
Here, we identify NLG1–3 as novel SNX27 PDZ-binding proteins and show that loss of SNX27 in cultured neurons decreases NLG1–3 levels. Since NLG2, an NLG primarily involved in the maintenance of inhibitory synapses [26–28], was the highest affinity SNX27 interactor, we focused on this interaction. We show that loss of SNX27 increases lysosomal degradation of NLG2 and reduces the number of NLG2 clusters in cultured hippocampal neurons. These data demonstrate that SNX27 plays a key role in rescuing endocytosed NLG2 from lysosomal degradation to allow recycling back to the plasma membrane.
Interestingly, knockdown of the core retromer component VPS35 neither decreased total levels of NLG2 nor increased localisation at lysosomes. However, VPS35 knockdown did decrease both NLG2 cluster size and density in secondary dendrites. The most straightforward explanation for these results is that increased degradation of another, as yet unidentified, retromer cargo plays a role in NLG2 clustering. Indeed, it has been reported that NLG2 can heterodimerise with NLG3 at inhibitory synapses . Since NLG3 levels were reduced by VPS35 knockdown, it remains possible that loss of VPS35 results in reduced NLG2 clustering through loss of NLG3. Nonetheless, together with knockdown of SNX27, these data indicate that both SNX27 and VPS35 control the number or size of NLG2 clusters through either direct trafficking of NLG2 (via SNX27) or indirect mechanisms (VPS35).
It is notable that, while knockdown of SNX27 reduced levels of all NLG proteins, knockdown of VPS35 did not reduce total levels of NLG2, and reduced levels of NLG1 and NLG3 more modestly than knockdown of SNX27. Thus, SNX27 appears to retrieve NLG2 from degradation in a manner distinct from its more well-characterised role of acting as a PDZ adaptor protein for retromer. While the mechanisms underpinning this retromer-independent trafficking pathway remain to be defined, we observed a similar profile for the NMDA receptor subunit GluN1, suggesting that it may represent an important additional regulatory pathway for synaptic membrane proteins. Indeed, a previous proteomic study identified proteins in HeLa cells that are reduced or lost following knockdown of SNX27 but not of VPS35, suggesting that this may represent a generalised SNX27-dependent trafficking pathway .
NLG2 plays a key role in regulating inhibitory synapse number through controlling the clustering of the postsynaptic scaffold protein gephyrin , but there have been no reports for the involvement of SNX27–retromer in this process. Here, consistent with the effects of knockdown of SNX27 or VPS35 on NLG2 cluster size and density, we show that knockdown of SNX27 or VPS35 reduced gephyrin cluster size and density and thus inhibitory synapse number. Moreover, these effects on gephyrin clustering are similar to directly knocking down NLG2, suggesting that the effects of SNX27 or VPS35 on inhibitory synapse number are mediated, at least in part, through controlling the abundance and targeting of NLG2.
SNX27 dysfunction has been shown to underlie excitatory synaptic defects associated with Down's syndrome and epilepsy [15,44]. Moreover, retromer dysfunction has been reported in a number of neurodegenerative disorders, including Alzheimer's and Parkinson's diseases [19,20,45]. Our results indicate that retromer and/or SNX27 dysfunction likely contributes to pathogenesis through altered degradation or mistargeting of the NLG proteins. In particular, our data demonstrate that SNX27 and retromer control inhibitory synapse number, at least in part through trafficking of NLG2, suggesting that retromer dysfunction could play a role in disorders characterised by aberrant inhibitory synaptic transmission or network function, such as epilepsy and autism. Further studies will be required to investigate this exciting possibility.
All experiments in the present study were performed in accordance with UK Home Office Schedule 1 guidelines. Animals were killed by cervical dislocation using procedures approved by the Home Office Licensing Team at the University of Bristol (UIN UB/18/004).
bovine serum albumin
epidermal growth factor receptor
foetal bovine serum
postsynaptic density 95 (PSD95), disk large, zona occludens
vacuolar protein sorting
C.S.B. performed all imaging and the majority of biochemistry and molecular biology. Y.N. performed VPS35 rescue experiments. J.M.H. provided supervision, mentoring and project direction. K.A.W. conceived the project, performed molecular biology and biochemistry, and supervised the project. All authors provided intellectual input into project direction and contributed to writing the manuscript.
We are grateful to the Medical Research Council (K.A.W. and J.M.H.), Biotechnology and Biological Sciences Research Council (K.A.W. and J.M.H.), Parkinson's UK (K.A.W. and J.M.H.), Alzheimer's Research UK (K.A.W.) and British Heart Foundation (J.M.H.) for funding. C.S.B. was funded by a Bristol Research into Alzheimer's disease and Care of the Elderly PhD studentship.
We thank Dr Martin Playford (NIH, U.S.A.) for the generous gift of the rabbit SNX27 antibody used in the present study and Dr Kirsty McMillan for critical reading of the manuscript. We also thank the Wolfson Bioimaging Facility (University of Bristol).
The Authors declare that there are no competing interests associated with the manuscript.