Region-Specific Phosphorylation Determines Neuroligin-3 Localization to Excitatory Versus Inhibitory Synapses

Abstract

Background: Neuroligin-3 is a postsynaptic adhesion molecule involved in synapse development and function. It is implicated in rare, monogenic forms of autism, and its shedding is critical to the tumor microenvironment of gliomas. While other members of the neuroligin family exhibit synapse-type specificity in localization and function through distinct interactions with postsynaptic scaffold proteins, the specificity of neuroligin-3 synaptic localization remains largely unknown.

Methods: We investigated the synaptic localization of neuroligin-3 across regions in mouse and human brain samples after validating antibody specificity in knockout animals. We raised a phospho-specific neuroligin antibody and used phosphoproteomics, cell-based assays, and in utero CRISPR/Cas9 (clustered regularly interspaced short palindromic repeats/Cas9) knockout and gene replacement to identify mechanisms that regulate neuroligin-3 localization to distinct synapse types.

Results: Neuroligin-3 exhibits region-dependent synapse specificity, largely localizing to excitatory synapses in cortical regions and inhibitory synapses in subcortical regions of the brain in both mice and humans. We identified specific phosphorylation of cortical neuroligin-3 at a key binding site for recruitment to inhibitory synapses, while subcortical neuroligin-3 remained unphosphorylated. In vitro, phosphomimetic mutation of that site disrupted neuroligin-3 association with the inhibitory postsynaptic scaffolding protein gephyrin. In vivo, phosphomimetic mutants of neuroligin-3 localized to excitatory postsynapses, while phospho-null mutants localized to inhibitory postsynapses.

Conclusions: These data reveal an unexpected region-specific pattern of neuroligin-3 synapse specificity, as well as a phosphorylation-dependent mechanism that regulates its recruitment to either excitatory or inhibitory synapses. These findings add to our understanding of how neuroligin-3 is involved in conditions that may affect the balance of excitation and inhibition.

Keywords: Autism; Excitatory synapse; Inhibitory synapse; Neuroligin; Phosphorylation; Scaffolding protein.

Figure 1.. Endogenous NL3 localizes to excitatory synapses in cortical, and to inhibitory synapses in subcortical regions in mice and humans

(A) Validation of NL3 antibody immunolabeling in WT versus Nlgn3 KO mouse brain shows NL3-specific punctate signal. Cortex displays small puncta and thalamus displays large puncta. Scale bar 45 μm. (B) NL3 (magenta) in mouse colocalizes with subsets of PSD95 (green) puncta in cerebral cortex, hippocampus CA1 stratum oriens, and cerebellum molecular layer (scale bar, 3 μm) but not in brainstem. Scale bar 14 μm. (C) Quantification of NL3 percentage overlapping with PSD95 puncta versus randomized overlap in cerebral cortex (Cx), hippocampus CA1 stratum radiatum (CA1 SR), stratum oriens (CA1 SO), and cerebellum molecular layer (CB ML) in the mouse. (D) NL3 (magenta) in mouse significantly colocalizes with GPHN (green) in brainstem, thalamus centrolateral nucleus (CL), and globus pallidus, but not in cerebral cortex. Scale bar 7 μm. (E) Quantification of the percentage GPHN puncta that overlap with NL3 in thalamus CL and brainstem. NL3 is fully localized to a subset of inhibitory postsynapses in thalamus and brainstem in mouse. (F) Labeling in perfused human brain of NL3 (magenta) associated with bona fide synaptic puncta, identified by presynaptic marker Synaptophysin (SYP, blue) in apposition to excitatory postsynaptic marker PSD95 (green, left panel set) and inhibitory postsynaptic marker GPHN (green, right panel set). Each image set shows one image from cortex, thalamus, and brainstem displayed in three vertical panels with different combinations of co-labeled markers as indicated. Scale bar 2 μm. (G) Quantification of excitatory synapse NL3-PSD95 and inhibitory synapse NL3-GPHN puncta densities in cortex, thalamus, and brainstem from one human brain.

Figure 2.. Identification of endogenous serine phosphorylation within the GPHN-binding site of NL3

(A) Alignment of C-terminal intracellular domains of human NL1-NL4. Conserved and conservative residues are in red. PDZ- and GPHN-binding sites are shaded blue. The serine within the GPHN-binding site corresponding to NL3-pS799 is highlighted in green. (B) Coomassie stained SDS-PAGE of anti-NL3 immunopurified eluate from rat brain extract. The extracted NL3 band as identified by mass spectrometry is boxed. Light (IgG₂₅) and heavy (IgG₅₀) immunoglobulin chains from the antibody are indicated. (C) Fragment ion mass spectrum of the doubly charged precursor of the monophosphorylated peptide NL3 (791–800). The inset shows the mass spectrum of the parent phosphopeptide (M_obs = 1300.6354, M_calc = 1300.6176, relative mass error = 13.7 ppm). Although a contaminating parent peptide ([M+2H+]²⁺ = 650.294) was co-isolated for fragmentation with the target parent peptide ([M+2H+]²⁺ = 651.325), conclusive N-terminal b-ion (b₂, b₃, b₇ in their non-phosphorylated forms) and C-terminal y-ion (y₅-y₈ in their phosphorylated forms) series, together with the neutral loss of the phospho moiety from y₃, clearly indicates phosphorylation on S799 while excluding the other potential phosphorylation sites of the peptide DYTLTLRRSP. (D) Spot blots assessing the phosphospecificity of the 6808 antibody-peptide mix, targeting the epitope containing phosphorylated S799 on NL3. Phosphorylated and corresponding unphosphorylated peptides spotted on nitrocellulose at increasing amounts (50 ng to 2 μg). Immunoreactivity was measured by infrared fluorescence of dye-conjugated secondary antibody. (E) Western blots of lysates from cerebral cortex (Cx), hippocampus (Hp), striatum (St), thalamus (Th), and cerebellum (Cb), probed for actin (loading control, bottom), NL3 (middle), and phosphorylated at S799 NL3 (NL3-pS, top). Antibody 6808 recognizes a band at the size of NL3 that is absent in Nlgn3 KO lysates, indicating specific recognition of native phosphorylated NL3. While NL3 was detected in all regions examined, bands corresponding to NL3-pS799 were specifically detected as a major band in hippocampus, and minor bands in cortex and cerebellum.

Figure 3.. Association of GPHN with NLs is affected by phosphomimetic mutation of serine at the GPHN-binding site in three exogenous systems.

(A) Yeast-two-hybrid assays using NL3 intracellular domains as bait, with WT S799 (NL3^ICD), phospho-null (~S799A), or phosphomimetic (~S799D) variants used against prey constructs of empty vector control (∅), GPHN (Geph), or S-SCAM fragment encompassing PDZ domains 1–3 (PDZ). Left: base plate showing comparable growth of transformed yeast cultures across all plate segments. Right: colorimetric β-galactosidase reaction on colony-lift replicate membrane. Positive interaction between NL3 mutants and postsynaptic scaffolds is indicated by cyan color reaction. (B) Yeast-two-hybrid assays with bait intracellular domains from NLs 1–4 in WT (ICD) and corresponding GPHN-binding site phospho-null (~S->A) and phosphomimetic (~S->D) mutations versus full-length GPHN prey constructs. Phosphomimetic mutation at the GPHN-binding serine disrupts interaction with GPHN. (C) Membrane co-clustering assays in COS7 cells transfected with GFP-GPHN (green) and myc-CB2^SH3- to induce GPHN membrane microaggregates. HA-NL3 (magenta) or its corresponding phospho-null (HA-NL3^S799A) or phosphomimetic (HA-NL3^S799D) mutants were transfected to assess co-clustering with GPHN membrane microaggregates (scale bar corresponds to 10 μm for main panel and 3.6 μm for inset). (D) Pearson’s correlation coefficients between GFP-GPHN microaggregates and surface-stained HA-NL3, HA-NL3^S799A, or HA-NL3^S799D; p < 0.01, one-way ANOVA with post-hoc Bonferroni from 3 independent experiments. (E) Expression of HA-NL3, phospho-null (HA-NL3^S799A), and phosphomimetic (HA-NL3^S799D) mutants in DIV14 cultured neurons, immunolabeled for endogenous inhibitory postsynaptic marker GPHN (green) and inhibitory presynaptic marker VIAAT (magenta), merged in bottom row. Scale bar 10 μm. (F) Quantification of dendritic (p < 0.001, one-way ANOVA with post-hoc Bonferroni, n ≥ 60 cells per condition from 5 independent experiments) and perisomatic (p < 0.01, one-way ANOVA with post-hoc Bonferroni, n ≥ 35 cells per condition from 3 independent experiments) GPHN clusters in neurons transfected with HA-NL3 or corresponding mutants. Phosphomimetic mutation at the GPHN-binding serine specifically abolishes NL3 interaction with GPHN in yeast, hinders NL3-GPHN co-clustering in cell lines, and NL3-mediated GPHN recruitment in cultured neurons.

Figure 4.. Construction and validation of CRISPR knockout plasmids for NL3.

(A) Structure of the Nlgn3 gene. Exons are numbered with annotation of ATG and open reading frame in dark grey. Closeup of Nlgn3 nucleotide sequence and translation of the exon 4 region targeted by Nlgn3 gRNA indicated by black bar. (B) Split GFP assay to visualize cleavage of the Nlgn3 gRNA target sequence. Neuro-2A cells imaged 48 h post transfection with three plasmids as schematized: i) gRNA plasmid containing either control (∅) or gRNA targeting Nlgn3, ii) plasmid expressing Cas9 and RFP (magenta), and iii) cleavage reporter plasmid expressing BFP (grey) and containing GFP sequence split by Nlgn3 gRNA target sequence. Successful cleavage of the Nlgn3 gRNA target sequence by Cas9+gRNA results in recombination of the reporter plasmid producing GFP fluorescence (cyan). Scale bar 10 μm. Graph shows the percentage of BFP and RFP cells that show cleavage reporter GFP expression in control gRNA 17.9% (± 7.7% S.E.M.) and Nlgn3 gRNA 89.3% (± 2.5% S.E.M.) transfections. Averages from N=3 biological replicates, each averaged from n=9 imaged fields. Scale bar 100 μm. (C) Genomic DNA sequencing of Nlgn3 exon 4 locus to assess and confirm deleterious insertion-deletion (indel) production. Mouse Neuro-2A cells co-transfected with Cas9 plasmid, GFP reporter plasmid (cyan), and either control or Nlgn3 gRNA plasmid assayed for cleavage in B. Scale bar 100 μm. Bulk genomic DNA was isolated 7 days post-transfection and used as template for PCR of Nlgn3 exon 4. (D) Chromatographs of mixed amplicon Sanger sequencing from control gRNA::∅ (top) or gRNA::Nlgn3 (bottom) transfected Neuro-2A cells from C. Control chromatographs were used as reference for unmixing of indel fitting of the mixed signal in the gRNA::Nlgn3 sample. Histogram shows assignment of indel frame frequency. Bins of deleterious frame-shift indels labeled in red, together make up 31.3 % of total reads. In-frame indels or wild type sequence bins are marked in grey. They comprised 62.4% of reads, which is close to the range anticipated for untransfected cells.

Figure 5.. Phosphomimetic mutation of the GPHN-binding site on NL3 determines excitatory versus inhibitory synapse specificity In vivo.

(A) Schematic of in utero electroporation approach to determine synapse localization of tagged NL3 variants in a knockout background mouse cortex. Plasmid cocktails were used to test each HA-tagged NL3 variant (WT, S799A phospho-null, and S799D phosphomimetic) consisting of Cas9 and gRNA::Nlgn3 to knock out endogenous NL3, GPHN.FingR-GFP (cyan) to label inhibitory postsynapses, and PSD95-RFP (magenta) to label excitatory postsynapses. Plasmid cocktails were electroporated into E14.5 mouse embryos and brains imaged at P7. Example images are shown of electroporation site in two magnifications. Scale bars 500 μm (left), and 100 μm (right). (B) Specificity of excitatory versus Inhibitory postsynapse localization was quantified in three brains for each HA-NL3 variant on a Nlgn3 CRISPR KO background, as well on a WT background (see Supplementary Fig. S9). Localization was quantified based on cross-correlation analysis of the HA signal to inhibitory (I) and excitatory (E) synapse marker signals. Top plot shows average cross-correlation values (P) in paired E-I analyses from each brain section in each of three brains per variant. Bottom plot shows paired mean differences of excitatory versus inhibitory correlations per brain in Cumming estimation plots. Effect size is shown with sampling distributions and vertical bars indicating 95% confidence interval. Positive values indicate specificity of HA-NL3 variants for excitatory postsynapses (blue), while negative values indicate specificity for inhibitory postsynapses (red). In the absence of endogenous NL3, cortically expressed HA-NL3 showed excitatory synapse specificity for WT (0.31 ± 0.053) and phosphomimetic S799D mutant (0.47 ± 0.045) variants, while phospho-null S799A mutant HA-NL3 showed inhibitory synapse specificity (−0.27 ± 0.066). Values are E-I mean difference from n=6 to 13 sections (depending on electroporation field size) from each brain, with deviations in SEM for N=3 brains per group. (C) Representative images in electroporation fields of cortex after Nlgn3 knockout (CRISPR KO::NL3) showing merged (top) and individual localization of HA-NL3 variants in white, inhibitory postsynapse marker GPHN.FingR-GFP in cyan, and excitatory postsynapse marker PSD95-RFP in magenta. Yellow arrows indicate bona fide synaptic puncta with HA colocalizing with excitatory (WT and S799D columns) or inhibitory (S799A column) postsynapse markers. Scale bar 5 μm. WT and phosphomimetic S799D mutant HA-NL3 localizes to excitatory synapses in cortex, while phospho-null S799A mutant HA-NL3 inappropriately localizes to inhibitory synapses in cortex.

Figure 6.. Model of NL3 region-specific, phosphorylation-dependent synapse localization.

Schema of the proposed model of NL postsynaptic localization synthesizing the current NL3 findings with previous findings (14, 19, 25, 53, 63). In cortical regions of the brain (shaded blue in coronal brain schema), including the cerebral cortex and hippocampus, NL3 is phosphorylated at its GPHN-binding site leading to association with PSD95 and localization to excitatory postsynapses. In subcortical regions of the brain (shaded red), including basal ganglia, parts of thalamus, hypothalamus, and brainstem, NL3 remains unphosphorylated at its GPHN-binding site leading to association with GPHN and localization to inhibitory postsynapses. Synapse-type specificities of NL3 are schematized together with other NLs in dimer states as previously described. GPHN-binding sites are phosphorylated in NL dimers comprised of NL3-pS799 (pS) and NL1-pY782 (pY) at excitatory postsynapses. NL3 dimers are S799 unphosphorylated together with NL2 dimers at inhibitory synapses.