Targeted proteoform mapping uncovers specific Neurexin-3 variants required for dendritic inhibition

Abstract

The diversification of cell adhesion molecules by alternative splicing is proposed to underlie molecular codes for neuronal wiring. Transcriptomic approaches mapped detailed cell-type-specific mRNA splicing programs. However, it has been hard to probe the synapse-specific localization and function of the resulting protein splice isoforms, or "proteoforms," in vivo. We here apply a proteoform-centric workflow in mice to test the synapse-specific functions of the splice isoforms of the synaptic adhesion molecule Neurexin-3 (NRXN3). We uncover a major proteoform, NRXN3 AS5, that is highly expressed in GABAergic interneurons and at dendrite-targeting GABAergic terminals. NRXN3 AS5 abundance significantly diverges from Nrxn3 mRNA distribution and is gated by translation-repressive elements. Nrxn3 AS5 isoform deletion results in a selective impairment of dendrite-targeting interneuron synapses in the dentate gyrus without affecting somatic inhibition or glutamatergic perforant-path synapses. This work establishes cell- and synapse-specific functions of a specific neurexin proteoform and highlights the importance of alternative splicing regulation for synapse specification.

Keywords: GABA; RNA; alternative splicing; autism; interneuron; neuronal circuit; proteoform; synaptic adhesion; synaptic specificity; targeted proteomics.

Figure 1

Expression and detection of NRXN3 AS5+ proteoforms in mice (A) Sashimi plots illustrating read distribution and splice junctions arising from mouse Nrxn3 AS5 in ribosome-associated mRNAs isolated from SST interneurons in mouse hippocampus (P28). Exons are depicted as boxes, and introns as dashed lines. Alternative exons and alternative acceptor sites (a,b) are marked in orange and constitutive exons in gray. (B) Amino acids of exon 24 protein coding sequence in Nrxn3-AS5^HA knockin mice. The HA epitopes (green), ω-site (red), and hydrophobic stretch conferring GPI-anchoring are indicated. (C) Schematic illustrating introduction of a translational stop codon in AS5+ (exon 24-containing mRNAs). This results in production of shortened, GPI-anchored NRXN3 proteoforms encoded by mRNAs with a long 3′UTR encoded by exons 25a, 25b, and 25c. AS5- mRNA isoforms encode canonical transmembrane NRXN3 proteins. (D) Western blot of whole neocortex (Cx), cerebellum (Cb), and hippocampal (Hc) extract from P28 wild-type and Nrxn3-AS5^HA/HA knockin mice probed with anti-HA, antineuroligin (NLGN), and anti-beta-actin (β-ACT) antibodies. Position of α- and β-Neurexin proteoforms is indicated. (E) Western blot of hippocampal brain extracts across development (postnatal days 2–60) from Nrxn3-AS5^HA/wt knockin mice probed with anti-HA, antiNeurexin (NRXN), antineuroligin (NLGN), and anti-beta-actin (β-ACT) antibodies. (F) Quantification of protein levels for HA-tagged NRXN3-AS5+ and for PAN-NRXN across development (P2-60), and corresponding mRNA levels assessed by qPCR for exon 24 (AS5+, GPI-anchored proteoform) and exon 25 (AS5-, transmembrane proteoform) of Nrxn3, n = 3 animals per time point. (G) Distribution of endogenous NRXN3-AS5^HA protein in subcellular fractionation of hippocampal cytosolic, membrane, and high-salt (HS) washed membrane fractions (equal percentage of total sample loaded in all lanes). (H) Overexpressed HA-tagged NRXN3-AS5+, NRXN3 AS5-, and placental alkaline phosphatase proteins immunoprecipitated from transfected HEK293 cells after radiolabeling with ³H-ethanolamine. Immunoprecipitates were analyzed by autoradiography (left panel) or probed by western blotting with anti-HA antibodies (right panel). Mean and SEM, one-way ANOVA.

Figure 2

Selective expression of NRXN3 AS5+ proteoforms (A and B) Detection of NRXN3-AS5^HA proteins in 28 day old wild-type (A) and homozygous Nrxn3-AS5^HA mice (B); OB, olfactory bulb, Cx, cortex, St, striatum, Hc, hippocampus, Th, thalamus, Mb, midbrain, Cb, cerebellum, and MO, medulla oblongata. (C) RNA sequencing reads for constitutive (exon 19) and alternative exon 24 of Nrxn3, from ribosome-associated mRNAs isolated from SST^cre, CamKII^cre (CA1), and Grik4^cre (CA3)—defined hippocampal cell populations. Read counts extracted from published data (Furlanis et al., 2019). (D and E) Detection of NRXN3-AS5^HA proteins in hippocampus of wild-type (D) and Nrxn3-AS5^HA/HA mice (E). (F) Fraction of NRXN3-AS5^HA protein expressing cells in the hilus of dentate gyrus coexpressing GAD67, calbindin (CB), or calretinin (CR), N = 4 mice, n = 3–4 brain slices per mouse, P25-30. (G and H) Colocalization analysis of NRXN3-AS5^HA protein (HA, green) with (G) GAD67 (magenta), arrowheads indicate colocalized example cell and (H) calretinin (CR, red) and calbindin (CB, blue) in dentate gyrus in Nrxn3-AS5^HA/HA knockin mice. Note that calretinin is expressed in glutamatergic mossy cells as well as (at higher level) in Cajal-Retzius cells (indicated with arrow head in [H]). Right panel in (H) shows a line scan of HA staining intensity across layers of dentate gyrus (Average of N = 3 animals, n = 2 ROIs per animal). OML, outer molecular layer, MML, middle molecular layer, IML, inner molecular layer, and GCL, granule cell layer. Mean and SEM, two-way ANOVA followed by Bonferroni’s test. Scale bars, 1 mm in (A and B); 200 μm in (D and E); and 50 μm in (G and H).

Figure 3

Presynaptic localization of NRXN3 AS5+ isoforms (A) Overview of NRXN3-AS5^HA protein (HA, green) and somatostatin (SST, blue) labeling in hippocampal CA1, SO, stratum oriens, SP, stratum pyramidale, SR, stratum radiatum, and SLM, stratum lacunosum moleculare. (B) High-magnification view of NRXN3-AS5^HA protein (HA, green) colocalization with VIAAT (red) and somatostatin (SST, blue) in CA1 SLM. (C) High-magnification view of NRXN3-AS5^HA protein (HA, green) colocalization with gephyrin (GEPH, magenta) and PSD-95 (magenta) in CA1 SLM, lower panels magnified view of indicated area. (D) Pre-embedding immunoelectron microscopy on glyoxal-fixed tissue for NRXN3-AS5^HA protein localization at symmetric and asymmetric synapses in SLM of wild-type control and Nrxn3 AS5^HA/HA mice. The edge of the postsynaptic specializations at asymmetrical and symmetrical synapses are each indicated by a pair of white arrowheads. Each immunogold particle is indicated by a black arrowhead, NT, nerve terminal, Den, dendrite, Sp, spine. For overview images, see Figure S3. (E) The density of metal particles detected per 1 μm of synaptic cleft calculated from 15 symmetric and 20 asymmetric synapses from n = 4 sections and N = 2 Nrxn3AS^HA/HA mice, and 17 symmetric and 18 asymmetric synapses from n = 4 sections and N = 2 wild-type mice. Note that there is no statistically significant difference in labeling density of asymmetric synapses as compared with wild-type mice. (F) Vertical distribution of 78 particles from the midline of synaptic cleft across 15 symmetric synapses from n = 4 sections and N = 2 Nrxn3AS^HA/HA mice. (G) Overview of NRXN3-AS5^HA protein (HA, green) and cholecystokinin (CCK, red) labeling in the hilus of the dentate gyrus, OML, outer molecular layer, MML, middle molecular layer, IML, inner molecular layer, and GCL, granule cell layer. (H) High-magnification view of NRXN3-AS5^HA protein (HA, green) colocalization with cannabinoid receptor 1 (CBR1, red) and GABA_A-receptor subunit alpha 1 (GABA_Aα1, blue) in dentate gyrus IML. Scale bars, 50 μm in (A and G); 2 μm in (B, C, and H); and 200 nm in (D). Note that all experiments (except G) were performed on glyoxal-fixed tissue. Mean and SEM, one-way ANOVA followed by Tukey’s multiple comparison.

Figure 4

NRXN3 AS5+ proteoforms recruit specific synaptic ligands (A) Volcano plot of protein abundance (iBAQ, log₂ scale fold change knockin versus WT mice and q value) in anti-HA immunoprecipitates from hippocampi from Nrxn3-AS5^HA/HA and wild-type (negative control) P25-30 mice (N = 5 mice per genotype). Selected known Neurexin ligands detected in the analysis are marked in orange. See Table S1 for detailed data. (B) Heatmap of protein abundance (iBAQ, log₂ scale) of known Neurexin ligands recovered from wild-type and Nrxn3-AS5^HA/HA mice in anti-HA immunoprecipitates and recovered from wild-type mice in control IgG and anti-NRXN immunoprecipitates, respectively. The anti-NRXN antibody is raised against the cytoplasmic tail of NRXN1 nd cross-reacts with all transmembrane NRXNs (Muhammad et al., 2015). Nrxn3 AS5^HA/HA versus wild type: q < 0.001 for NRXN3, NXPH1, FAM19A1, and FAM19A2; q > 0.05 for NRXN1, NLGN3, and DAG1. Anti-NRXN versus control IgG: q < 0.001 for all except C1QL2 (q = 0.0066) and C1QL3 (q = 0.08) (multiple t test with Benjamini, Krieger, and Yekutieli correction). See Table S2 for NRXN immunoprecipitates, and Table S3 for details on selected ligands. Nondetectable proteins depicted as boxes with dashed outline. (C) Fluorescent in situ hybridization for Nrxn3 and ligands Nxph1 and Fam19A2 transcripts in dentate gyrus of P30 mice. (D) Confirmation of differential ligand interactions by western blotting. Input (IN, 1%) and immunoprecipitates with anti-HA (left) or control IgG and anti-NRXN1 antibodies (right) probed with anti-NLGN (top) and anti-HA or anti-NRXN antibodies (bottom). Molecular weight markers indicated in kDa. # indicates heavy IgG-chains. (E) Schematic of NRXN3 domain organization, alternatively spliced segments (blue), and proteotypic peptides (PTPs) of constitutive/common (gray) and proteoform-specific (blue) amino acids indicated, which are quantified for AS3, AS4, and AS6, normalized to recombinant protein expressing all splice isoforms (100% splice site inclusion) and to constitutive exons. Mean and SEM, Scale bars, 100 μm (overview) and 10 μm (high-magnification images).

Figure 5

Translational silencing gates NRXN3 protein expression (A) Quantitative PCRs of major Nrxn transcript isoforms (left panel) and mRNAs containing exon 24, alternative accepters 25a, 25b, and the constitutive exon 25c (right panel, plotted relative to wild type) for wild type and Nrxn3^ΔEx24 mice, normalized to Gapdh, hippocampus, P25-30, N = 4 mice per genotype. (B) Schematic diagram illustrating alternative splicing events in wild-type and Nrxn3^ΔEx24 hippocampus (left) and semiquantitative PCR visualizing Nrxn3 transcript variants arising from alternative splicing at AS5 in wild-type and Nrxn3^ΔEx24 mice (right). Position of primer binding sites on alternative exon segments is illustrated. See STAR Methods for details. (C and D) Detection of AS5 proteoforms by targeted proteomics with heavy peptides targeting alternative accepters 25a, 25b, and constitutive exon 25c (C) or targeting all NRXN1, NRXN2, and NRXN3 proteoforms (D). Ratios of light to heavy peptide detection are displayed in reference to wild-type samples of each peptide. One representative peptide shown, consistent results were obtained for multiple proteotypic peptides for the same proteoform (see Table S6), hippocampus, P25-20, N = 5 mice per genotype. (E and F) Luciferase assay of dual-promoter plasmids expressing firefly (fLuc) and renilla (rLuc) luciferase, left panel: schematic representation of different Nrxn3 exonic sequences fused after the translational stop codon of rLuc, right panel: luciferase activity from renilla luciferase constructs normalized to firefly luciferase activity, N = 2–3 cell cultures, n = 2–3 replicates per culture (E) and mRNA levels determined by RT-qPCR of renilla luciferase constructs normalized to firefly luciferase, N = 3 cell cultures (F) (For sequences of luciferase constructs see Table S4). Mean and SEM, with two-way or one-way ANOVA followed by Bonferroni’s test for both qPCR and proteomic analysis or luciferase assay, respectively.

Figure 6

Impaired GABAergic synaptic transmission in Nrxn3^ΔEx24 mice (A) Representative traces of spontaneous IPSCs from 6- to 8-week-old wild-type and homozygous Nrxn3^ΔEx24 mice recorded in the presence of AP5 and NBQX. (B and C) Cumulative frequency distributions of amplitudes (B) and interevent intervals (C) of dentate gyrus granule cell sIPSCs recorded from wild-type (N = 16 animals, n = 45 cells) and homozygous Nrxn3^ΔEx24 mice (N = 12 animals, n = 35 cells). Insets: average sIPSC amplitudes (B) and frequencies (C) per cell. (D) Representative traces of spontaneous EPSCs from 6- to 8-week-old wild-type and homozygous Nrxn3^ΔEx24 mice recorded in the presence of Picrotoxin. (E and F) Cumulative frequency distributions of amplitudes (E) and interevent intervals (F) of dentate gyrus granule cell sEPSCs recorded from wild-type (N = 4 animals, n = 20 cells) and homozygous Nrxn3^ΔEx24 mice (N = 4 animals, n = 21 cells). Insets: average sEPSC amplitudes (E) and frequencies (F) per cell. (G) Representative traces of miniature IPSCs from 6- to 8-week-old wild-type and homozygous Nrxn3^ΔEx24 mice recorded in the presence of TTX, AP5, and NBQX. (H and I) Cumulative frequency distributions of amplitudes (H) and interevent intervals (I) of dentate gyrus granule cell mIPSCs recorded from wild-type (N = 3 animals, n = 20 cells) and homozygous Nrxn3^ΔEx24 mice (N = 3 animals, n = 19 cells). Insets: average mIPSC amplitudes (H) and frequencies (I) per cell. (J) Representative traces of miniature EPSCs from 6- to 8-week-old wild-type and homozygous Nrxn3^ΔEx24 mice recorded in the presence of TTX and picrotoxin. (K and L) Cumulative frequency distributions of amplitudes (K) and interevent intervals (L) of dentate gyrus granule cell mEPSCs recorded from wild-type (N = 3 animals, n = 17 cells) and homozygous Nrxn3^ΔEx24 mice (N = 3 animals, n = 18 cells). Insets: average mIPSC amplitudes (K) and frequencies (L) per cell. Mean and SEM, analyzed using the Mann-Whitney test.

Figure 7

Reduced dendritic GABAergic inputs onto dentate gyrus granule cells in Nrxn3^ΔEx24 mice (A) Positioning of electrodes to selectively stimulate GABAergic synapses in the inner molecular layer (IML) and granule cell layer (GCL) of the dentate gyrus in the presence of NBQX and AP5. (B and C) Representative traces of evoked IPSCs in response to electrical stimulation (stimulation intensity: 20 μA) in the GCL (B) and IML (C). (D) Representative traces of evoked EPSCs in response to electrical stimulation in the OML (stimulation intensity 20 μA) in the presence of picrotoxin. (E and F) Quantification of evoked IPSC paired-pulse ratios (E) and evoked IPSC amplitudes (F) upon GCL and IML stimulation recorded in dentate gyrus granule cells in wild-type (N = 8 animals, n = 10–12 cells) and homozygous Nrxn3^ΔEx24 mice (N = 7–8 animals, n = 11–12 cells). (G) Quantification of evoked EPSC amplitudes upon stimulation of performant-path inputs in OML in dentate gyrus granule cells of wild-type (N = 4 animals, n = 12 cells) and homozygous Nrxn3^ΔEx24 mice (N = 4 animals, n = 16 cells). (H) Schematic drawing of a dentate gyrus granule cell showing the fields of illumination in the GCL and the OML for optogenetic stimulation of inputs from PV and SST interneurons, respectively. (I and J) Representative traces of optogenetically evoked GABAergic inputs from PV (I) and SST interneurons (J). The example traces show an overlay of the evoked IPSCs to three different laser intensities (1, 2, and 4 mW) for each genotype. (K) Representative traces of glutamatergic inputs evoked by electrical perforant-path stimulation. The example traces show an overlay of the evoked EPSCs to three different stimulation intensities (10, 30, and 100 μA) for each genotype. (L) Dose-response curves showing the mean evoked IPSC amplitudes in response to optogenetic stimulation of PV interneurons recorded in dentate gyrus granule cells from wild-type (N = 7 animals, n = 12 cells) and homozygous Nrxn3^ΔEx24 mice (N = 6 animals, n = 24 cells). (M) Dose-response curves showing the mean evoked IPSC amplitudes in response to optogenetic stimulation of SST interneurons recorded in dentate gyrus granule cells from wild-type (N = 5 animals, n = 24 cells) and homozygous Nrxn3^ΔEx24 mice (N = 6 animals, n = 28 cells). Laser intensities ranging from 1 to 7 mW. (N) Dose-response curve showing the mean evoked EPSCs in response to electrical performant-path stimulation recorded in dentate gyrus granule cells of wild-type (N = 4 animals, n = 8 cells) and homozygous Nrxn3^ΔEx24 (N = 4 animals, n = 8 cells) mice. Stimulation intensities were ranging from 10 to 100 μA. Mean and SEM, analyzed using the Mann-Whitney test.