Reduced SNAP25 protein fragmentation contributes to SNARE complex dysregulation in schizophrenia postmortem brain

Abstract

Recent studies associated schizophrenia with enhanced functionality of the presynaptic SNARE (soluble N-ethylmaleimide-sensitive factor attachment protein receptor) complex. Altered degradation pathways of the three core SNARE proteins: synaptosomal-associated protein 25 (SNAP25), syntaxin-1 and vesicle-associated membrane protein (VAMP) could contribute to enhanced complex function. To investigate these pathways, we first identified a 15-kDa SNAP25 fragment (f-S25) in human and rat brains, highly enriched in synaptosomal extractions, and mainly attached to cytosolic membranes with low hydrophobicity. The presence of f-S25 is consistent with reports of calpain-mediated SNAP25 cleavage. Co-immunoprecipitation assays showed that f-S25 retains the ability to bind syntaxin-1, which might prevent VAMP and/or Munc18–1 assembly into the complex. Quantitative analyses in postmortem human orbitofrontal cortex (OFC) revealed that schizophrenia (n = 35), but not major depression (n = 15), is associated with lower amounts of f-S25 (–37%, P = 0.027), and greater SNARE protein-protein interactions (35%, P < 0.001), compared with healthy matched controls (n = 28). Enhanced SNARE complex formation was strongly correlated with lower SNAP25 fragmentation rates (R = 0.563, P < 0.001). Statistical mediation analyses supported the hypothesis that reduced f-S25 density could upregulate SNARE fusion events in schizophrenia. Cortical calpain activity in schizophrenia did not differ from controls. f-S25 levels did not correlate with total calpain activity, indicating that if present, schizophrenia-related calpain dysfunction might occur locally at the presynaptic terminals. Overall, the present findings suggest the existence of an endogenous SNARE complex inhibitor related to SNAP25 proteolysis, associated with enhanced SNARE activity in schizophrenia.

INTRODUCTION

Synaptosomal-associated protein 25 (SNAP25) is a central molecule in the cycle of vesicle trafficking and neurotransmitter release. Along with its binding partners, syntaxin-1 and vesicle-associated membrane protein (VAMP, or synaptobrevin), SNAP25 forms the core structure of the soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE), the heterotrimeric complex driving presynaptic vesicles through the docking, priming and fusion steps, culminating in neurotransmitter release (Sudhof and Rothman, 2009; Rizo and Sudhof, 2012; Sudhof, 2013).

Variation in the SNAP25 gene and/or altered protein function are associated with a number of psychiatric illnesses, including schizophrenia. Multiple studies report an association between autism, as well as attention deficit hyperactivity disorder (ADHD), and genetic variation in the SNAP25 gene across different populations (Mill et al., 2005; Braida et al., 2015; Safari et al., 2017). While the association between SNAP25 polymorphisms and the risk of schizophrenia is inconsistent (Kawashima et al., 2008; Fanous et al., 2010; Lochman et al., 2013; Cupertino et al., 2016; Kang et al., 2017; Ortega-Alonso et al., 2017), syntaxin-1 genetic variation may contribute to this illness (Wong et al., 2004). Mutations affecting SNAP25 function, rather than protein expression levels, might be required to recapitulate neuronal dysfunctions contributing to schizophrenia. Support for this concept comes from the so-called ‘blind-drunk’ murine strain. These animals carry a mutation in the SNAP25 gene which abnormally enhances the affinity of SNARE protein-protein interactions (PPIs) (Jeans et al., 2007), and show a schizophrenia-like endophenotype sensitive to antipsychotic treatment (Oliver and Davies, 2009). In contrast, mice lacking one copy of the SNAP25 gene (e.g. ‘coloboma’ mice) display hyperactivity, impaired social abilities, poor learning and memory performance, and attention deficits (Hess et al., 1992; Heyser et al., 1995; Corradini et al., 2014; Braida et al., 2015, 2016), but no psychosis-like behavioral manifestations.

Comparative postmortem brain studies using schizophrenia cases and controls showed similar discrepancies when reporting relative brain amounts of individual components of the SNARE machinery (Gabriel et al., 1997; Honer et al., 1997, 2002; Thompson et al., 1998; Sokolov et al., 2000; Halim et al., 2003; Barakauskas et al., 2010, 2016; Castillo et al., 2010; Gil-Pisa et al., 2012). While region-dependent alterations of SNAP25, syntaxin-1 or VAMP protein expression levels may occur in schizophrenia brains, these could be causative or secondary (i.e. compensatory mechanisms) to dysregulated synaptic activity. Indeed, altered SNARE functionality (with or without altered expression of protein amounts) could be part of the pathological substrate of schizophrenia (Johnson et al., 2008; Katrancha and Koleske, 2015; Ramos-Miguel et al., 2015a; Cupertino et al., 2016; Egbujo et al., 2016). Several reports documented dysregulated SNARE protein-protein interactions (PPIs) and/or complex assembly alterations compatible with abnormally enhanced synaptic activity in cortical and striatal brain regions (Barakauskas et al., 2010; Castillo et al., 2010; Ramos-Miguel et al., 2015b). Notably, greater overall SNARE function is consistent with neuroimaging studies reporting aberrant global brain signal in both patients with schizophrenia, and healthy volunteers experiencing psychotic episodes following ketamine administration (Driesen et al., 2013; Yang et al., 2014).

Among the multiple mechanisms that could enhance SNARE (and thereby synaptic) function, for the present study we focused on SNARE protein degradation mechanisms, which could potentially yield active breakdown products. For example, previous studies identified a stable SNAP25 fragment resulting from calpain proteolytic activity in various types of neuronal cultures (Ando et al., 2005; Grumelli et al., 2008), as well as short SNAP25 N-terminal peptides increased in the cerebrospinal fluid of patients with Alzheimer’s disease (Brinkmalm et al., 2014a). The presence of similar fragments from syntaxin-1 or VAMP was only documented in epithelial type II cells (Zimmerman et al., 1999), and the possible physiological role of these SNARE byproducts in neurosecretory activities was not investigated. Interestingly, both SNARE-regulated exocytosis and calpain proteolytic activity are calcium-dependent processes (Campbell and Davies, 2012; Sudhof, 2013). Since the same second messenger triggers both cellular events, a crosstalk mechanism between them with functional implications in schizophrenia brain appears plausible. Remarkably, altered expression of other calpain substrates and calpain-cleaved fragments was also previously reported in postmortem brains of subjects with schizophrenia (Engmann et al., 2011; Ramos-Miguel et al., 2013).

The present study tests the hypothesis that dysregulated proteolysis of the SNARE machinery could underlie the abnormally enhanced SNARE functionality observed in previous studies in schizophrenia brains (Barakauskas et al., 2010; Ramos-Miguel et al., 2015b). We first screened for stably expressed SNARE protein fragments potentially existing in human and rodent cortical samples. Following characterization experiments indicating a possible physiological role of the identified SNAP25 fragment, we quantified this peptide in postmortem orbitofrontal cortices (OFC) of subjects with schizophrenia and controls, obtained from two different brain banks. To address the illness specificity of the reported findings, OFC samples from subjects with major depression were also included in the study. Next, using data from previous studies where upregulated SNARE functionality was demonstrated in the same brain samples, we constructed mathematical models showing a pathway-wise association across neurochemical findings to clinical diagnoses. Finally, we assessed calpain activity in the same brain samples to evaluate whether the overall tissue activity of this enzyme might explain the neurochemical findings of the study.

EXPERIMENTAL PROCEDURES

Human postmortem brains

Brain samples were collected from two independent postmortem human tissue banks, the Macedonian/New York State Psychiatric Institute Brain Collection, and the Stanley Medical Research Institute, Neuropathology Consortium (Appendix Table S1). All procedures involving human postmortem sample use were reviewed and approved by the UBC Clinical Research Ethics Board, and next-of-kin consent was obtained prior to tissue collection. Demographic and toxicological characteristics of subjects with schizophrenia (n = 35), major depressive disorder (MDD, n = 15), and healthy matched controls (n = 28) were extensively detailed in previous studies (Johnston et al., 1997; Barakauskas et al., 2010), and briefly summarized in Appendix Table S1. Antemortem histories of psychiatric illnesses were confirmed following DSM-IV criteria for postmortem diagnosis (American Psychiatric Association, 1994). Standard histopathologic examinations were performed to exclude subjects displaying neuropathologies that may potentially confound the present results. Grey matter samples from the OFC, approximating Brodmann’s area (BA) 10/47 were dissected using a standard atlas (Mai et al., 1997), and homogenized in Tris-buffered saline (TBS) supplemented with 1% of protease inhibitors (Sigma, St. Louis, MO, USA). The OFC was selected for its contribution to psychotic and cognitive symptoms in schizophrenia and its key role in overall cognitive function (Dolan et al., 1995; Carter et al., 1997; Camchong et al., 2008; Goghari, 2011). Brain homogenates were immediately aliquoted and stored at –80º until use, and the protein concentration was determined using the DC Assay (Bio-Rad, Hercules, CA, USA). Before quantitative experiments, protein concentrations were equalized in all samples by adding appropriate volumes of homogenization buffer.

To model the potential effects of PMI on the synaptic targets of the study, we used an independent cohort of human postmortem brain samples (healthy controls) obtained at the National Specimen Brain Bank (Los Angeles, CA, USA), including n = 9 males with a mean age of 66±5 (range 40–86) and 15±2 h of PMI in average (range 7–25 h) (Ramos-Miguel et al., 2015b). In order to investigate PMI as the sole source of variability, cases with previous history of serious diseases (including mental illnesses), or presenting with positive blood toxicology for psychotropic drugs or ethanol were excluded. Only subjects experiencing sudden and unexpected deaths (i.e. motor-vehicle accidents, gunshot murders) were allowed in this cohort. Grey matter samples were available from the inferior temporal gyrus (IT; approximating BA 20), and were carefully dissected, homogenized and stored as above. Protein concentrations were equalized prior to quantitative assays.

Animals and drug treatments

All experiments using laboratory animals were in compliance with the Guidelines of the Canadian Council on Animal Care, and previously approved by the Animal Care Committee at the University of British Columbia. Adult male Sprague-Dawley rats were purchased at Charles-River (Montreal, QC, Canada), and housed under standard conditions of temperature (22±1ºC), humidity (70%), and light/dark (12/12-h cycle), with unlimited access to standard rat chow pellets and drinking water. Haloperidol hydrochloride (1 mg/kg/day) and clozapine (20 mg/kg/day) were both from Tocris (Bristol, UK), and dissolved in 0.9% NaCl saline solution adjusted to pH 5.5. Animals (n = 10 per group) received one daily intraperitoneal (i.p.) injection of pH-adjusted saline or the corresponding antipsychotic drug for 28 consecutive days, as previously described (Barakauskas et al., 2010). Twenty-four hours after the last injection, all rats were sacrificed by decapitation, and brains were immediately removed and rinsed in ice-cold artificial cerebrospinal fluid. Frontal cortices were quickly dissected, frozen on dry ice, and stored at –80ºC. Tissue homogenization was performed as described above.

Cell cultures

Human embryonic kidney HEK293 cells and murine neuroblastoma N2A cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% of FBS, 1 mM of sodium pyruvate, 2 mM of L-glutamine, 50 units of Penicillin and 50 μg of Streptomycin, as previously described (Zhang et al., 2017). Rat pheochromocytoma PC12 cells were cultured in RPMI 1640 containing 10% of FBS, 5% HS, 1 mM of sodium pyruvate, 2 mM of L-glutamine, 50 units of Penicillin and 50 μg of Streptomycin. E18 hippocampal neurons derived from pregnant rats were grown in neurobasal medium with B27 in poly-D-lysine coated dishes, following standard procedures (Brewer et al., 1993). All cells were maintained at 37°C in an incubator containing 5% of CO2.

SDS-PAGE and quantitative immunoblotting

Brain amounts of multiple protein targets were resolved by sodium dodecyl sulfate (SDS) polyacrylamide gel electrophoresis (PAGE) followed by quantitative (q) immunoblotting (IB), essentially as previously reported (Ramos-Miguel et al., 2015c). Briefly, brain samples were combined with equal volumes of a 2× loading buffer (100 mM Tris, pH 6.8, 20% glycerol, 4% SDS, 200 mM β-mercaptoethanol, 0.02% bromophenol blue), and boiled for 5 minutes, unless otherwise indicated. Ten-µg protein aliquots were then loaded onto 12% polyacrylamide minigels (Bio-Rad, Hercules, CA, USA), along with a prestained protein ladder (Bio-Rad). For quantitative purposes, all gels were also loaded in triplicate with a standard sample, prepared by pooling equal amounts of all control samples. Proteins were transferred to polyvinylidene difluoride (PVDF) membranes (Bio-Rad), following manufacturer instructions. After incubation for 1 h in a blocking solution containing 5% skim milk in TBS supplemented with 0.1% Tween-20 (TBST), membranes were incubated overnight at 4ºC with the corresponding primary antibody diluted in the blocking solution (see Appendix Table S2). Following thorough washes in TBST, membranes were exposed to the appropriate peroxidase-conjugated anti-mouse, anti-rabbit, or anti-goat IgG/M secondary antibodies (Jackson ImmunoResearch, West Grove, PA, USA), diluted 1:5,000 in the blocking solution. Chemiluminescence was then enhanced with commercial reagents (ECL Plus; PerkinElmer, Waltham, MA, USA), and the immunoreactive signal was digitalized in a LAS-3000 imaging system (Fujifilm, Tokyo, Japan). In quantitative studies, bound antibodies were stripped out from the membranes, which were subsequently reprobed with anti-β-actin antibody, in order to adjust for protein load.

The integrated optical density (IOD) of the immunoreactive bands was quantitated in ImageGauge v.4.22 (Fujifilm). In a given qIB membrane, the IOD of each immunoreactive band was first divided by its corresponding β-actin IOD value. When appropriate, the ratio of the breakdown product of a particular protein to its full-length species was estimated before β-actin normalization. The average of the three IOD values from the in-gel standard triplicates was then obtained. The IOD values from all other samples in the same gel were then calculated in percent to the standard average. Each brain sample was loaded in at least three different gels, and the mean across these values was taken as a final estimate.

Crude synaptosomal and lipid raft extractions

Purification of crude synaptosome- and lipid raft-enriched fractions from human cortical specimens was achieved following standard procedures (Ramos-Miguel et al., 2015c). Briefly, to extract crude brain synaptosomes, cortical samples were first homogenized in 4 mM HEPES buffer, pH 7.4, containing 0.32 M sucrose and 1% protease inhibitors. Nuclei, along with cell and tissue debris, were removed by centrifugation (1,000×g, 10 min), and supernatants were further centrifuged (17,000×g, 15 min) to separate cytosol (supernatant) from total membranes (pellet). The latter fraction was carefully loaded on top of a three-layer discontinuous sucrose gradient (0.32, 0.8, and 1.2 M), and ultracentrifuged (96,000×g, 12 h). The resulting interfaces between the 0.8–1.2 M (synaptosomes), and the 0.32–0.8 M (myelin-containing fragments) sucrose solutions were collected and resolved by IB alongside the prior fractions.

The lipid raft enrichment method was based on the particular solubility features of these membrane microdomains (i.e. TritonX-insoluble, SDS-soluble) (Ohrfelt et al., 2011). Cortical tissue was initially ground in ice-cold Tris buffer (10 mM Tris pH 6.8, 1% protease inhibitors), and fractions were sequentially collected from supernatants following five consecutive centrifugation (31,000×g, 1 h)-resuspension cycles, yielding (1) cytosolic, (2) 0.5% TritonX-soluble, (3) 2% TritonX-soluble, (4) 0.1% SDS-soluble (rafts), and (5) detergent insoluble proteins.

Immunoprecipitation

Sheep anti-mouse IgG coated magnetic Dynabeads (Thermo Fisher Scientific, Waltham, MA, USA) were used for immunoprecipitation (IP) assays, as previously described (Ramos-Miguel et al., 2015b). Briefly, beads were conjugated with anti-SNAP25 (SP12), anti-synataxin-1 (SP7), or anti-VAMP (SP11) antibodies (Honer et al., 1993) for 2 h at room temperature in TBS supplemented with 0.1% bovine serum albumin (BSA) and 2 mM EDTA, and then blocked for 1 h in the same buffer additionally containing 3% BSA and 0.1% Triton X-100. In parallel, human or rat brain homogenates were solubilized for 1 h at 4ºC in TBS containing 1 mM EDTA, 1% protease inhibitor cocktail (Sigma), and 0.5% Triton X-100. Tissue debris were pelleted by centrifugation at 16,000×g for 30 min at 4ºC. Supernatants were then precleared with unconjugated magnetic beads to remove those proteins that may bind to the beads by electromagnetic interactions, regardless of antibody-antigen reactions. IP was performed overnight at 4ºC by mixing 0.75 mg of solubilized brain protein with 0.15 mg of conjugated Dynabeads per reaction. Negative controls were run in parallel, combining the same amounts of brain sample and unconjugated, anti-mouse IgG-coated magnetic beads. Elution of IP products was accomplished by incubation of the beads at room temperature in SDS-PAGE loading buffer for 5 minutes. After boiling for another 5 min, IP and co-IP products were resolved by SDS-PAGE and standard IB.

Enzymatic dephosphorylation of brain proteins

To assess the impact of full protein dephosphorylation on syntaxin-1-SNAP25 protein-protein interactions. pooled rat cortical homogenates were incubated in triplicate in the absence or presence (1 U/µl) of active calf alkaline phosphatase (Millipore, Billerica, MA, USA), at 37ºC for 1 h (Ramos-Miguel et al., 2009). Reactions were stopped with 50 mM sodium pyrophosphate, and chilled on ice. Samples were then prepared for IP followed by IB as indicated above.

Calpain assay and in silico prediction analysis of SNAP25 cleavage site

Calpain proteolytic activity in cortical homogenates was assayed using a commercially available kit (Abcam Cat# ab65308, Cambridge, MA, USA), following manufacturer’s recommendations. Due to limitations of tissue availability, only samples from the Macedonian/NY State Collection were assayed. Briefly, brain samples were solubilized in a 25-mM Tris-HCl buffer, pH 7.2, supplemented with 0.5% Triton X-100, and 1% of EDTA-free protease inhibitor cocktail (Pierce Biotechnology, Rockford, IL, USA). After centrifugation at 16,000×g for 30 min at 4ºC, supernatants were aliquoted and stored at −80ºC until use, and protein concentrations were estimated by DC assay. In quantitative assays, 96-well black polystyrene microplates with an optical flat bottom (Corning Inc., Corning, NY, USA) were loaded in triplicate with either 100 µg of total brain protein or active calpain-1 (range 0.125–2 ng). Additionally, other wells contained solubilization buffer alone (blank), or 1 ng of calpain-1 plus 1 ng of a calpain inhibitor (Z-LLY-FMK), as negative controls. Appropriate amounts of 10× reaction buffer and fluorescent substrate (Ac-LLY-AFC) were then added to all wells, and reactions were allowed for 1 h at 37ºC with orbital shaking. Fluorescence was then quantified at 400/505 nm excitation/emission, allowing 10 nm of bandwidth, in a Safire2 microplate reader (Tecan, Männedorf, Switzerland). Finally, background signal was subtracted, and results were expressed in relative fluorescence units (RFU) per mg of total protein.

To investigate the potential cleavage site(s) of calpain-mediated SNAP25 fragmentation, we used a publicly available, in silico tool at www.calpain.org (duVerle et al., 2010). This approach uses a multiple kernel learning algorithm coupled to a large database of known calpain substrates (duVerle et al., 2010; DuVerle et al., 2011).

Data analysis and statistics

Normal distribution of all datasets was initially confirmed with the Kolmogorov-Smirnov normality test. Multivariate analyses were carried out to explore possible associations between the neurochemical targets and potential confounding variables, including sex, age at death, PMI, or brain pH. We then analyzed differences in brain synaptic protein levels across diagnostic groups by analysis of covariance (ANCOVA), followed by Tukey’s HSD post-hoc test, controlling models for all variables with significant associations with presynaptic proteins. Pearson’s correlations were also used to inspect putative associations between the synaptic targets. To address the hypothesis that SNAP25 cleavage may contribute to schizophrenia by enhancing SNARE complex formation, we conducted a logistic regression mediation model (with schizophrenia diagnosis as the outcome), using a bias-corrected bootstrap confidence interval (from 10,000 samples) for the indirect effect (MacKinnon et al., 2007; Hayes, 2017). In animal studies, the effects of antipsychotic drugs on brain proteins were addressed with one-way analysis of variance (ANOVA), followed by Tukey’s test. Student’s t-test was used to analyze the effect of alkaline phosphatase on brain protein-protein interactions. All tests were two-tailed, with statistical significance set to p < 0.05. All datasets were analyzed with JMP, version 13.0.2 (SAS Institute, Cary, NC, USA), and plots were built in GraphPad Prism 6.0a (GraphPad Software, La Jolla, CA, USA). The PROCESS macro for SPSS version 24.0 was used to construct 95% confidence intervals for the indirect effect in the mediation analysis (Hayes, 2017).

RESULTS

Characterization of SNARE protein fragments

While SNAP25 cleavage was described earlier in neuroblastoma cell lines, as well as in primary hippocampal and cerebellar cultures (Ando et al., 2005; Grumelli et al., 2008), the potential existence of stable breakdown products from this and other SNARE proteins (i.e. syntaxin-1 and VAMP) in human brain tissue has not been reported. Western blot analyses of both human and rat cortical samples revealed the presence of 15–20-kDa bands in both species using 4 different antibodies against SNAP25 (Figure 1A), fully matching those N- and C-terminal SNAP25 fragments previously reported in cultured cells (Grumelli et al., 2008). While all three antibodies directed against the N-terminal domain of SNAP25 demonstrated an identical fragmentation pattern, with one major band slightly above 15-kDa, the C-terminal antibody reacted with two sharp species at ~15 and ~20 kDa, neither of which overlapped with the N-terminal fragments (Figure 1A). Additionally, another S25 polyclonal antibody raised against central/C-terminal residues of the molecule did not react with any SNAP25 fragment (data not shown), suggesting that the cleavage site(s) of S25 molecule overlaps with the epitope detected by the latter antisera. Although the features studied below involving the SNAP25 breakdown products could also be present in the C-terminal fragments, for practical reasons we decided to focus on the ~15-kDa N-terminal fragment only (hereon referred to as f-S25).

Figure 1.

Identification and localization of SNARE protein fragments. (A) Non-boiled and boiled samples from postmortem human (H; 59-year-old male with 7 h of PMI) inferior temporal cortex (IT, BA 20) and rat (R) frontal cortex were resolved by standard SDS-PAGE followed by immunoblotting with the indicated antibodies against N-terminal (N-t) and C-terminal (C-t) residues of SNAP25 (S25), syntaxin-1 (STX1) and VAMP. The identified SDS-resistant and boiling-sensitive SNARE complexes, full-length (FL) monomeric SNARE proteins, and several SNAP25 breakdown products, including the ~15-kDa N-terminal fragment (f-S25), were labeled with grey, solid black, or clear arrowheads, respectively. (B) FL- and f-S25 were also immunodetected (SP12 antibody) in primary hippocampal cultures from two different rats (PHN-½), as well as in neuroblastoma (N2A) and pheochromocytoma (PC12) cell lines, but not in human embryonic kidney (HEK) cell cultures. (C) Sequential fractionation of subcellular compartments in postmortem human IT samples (three different subjects were tested with similar results) revealed the relative enrichment of f-S25 in crude synaptosomal extractions, compared to FL-S25. The purity of the obtained fractions was tested in previous studies using specific markers (Ramos-Miguel et al., 2015c, 2017) (D) Sequential fractionation of membrane-related microdomains in postmortem human IT samples (three different subjects were tested with similar results) revealed the relative enrichment of f-S25 in the 0.5% Triton X-100 soluble (matching least hydrophobic cytoplasmic membranes), while barely present in Triton X-100-insoluble SDS-soluble (mainly containing lipid rafts) extractions, compared to FL-S25. (A–D) Molecular masses (in kDa) of prestained standards are shown on the left. (B–D) All samples were boiled for 5 min prior to loading into the gels.

Notably, all neuron-derived cultures and cell lines tested, but not HEK cells, expressed f-S25 (Figure 1B), indicating that SNAP25 cleavage is not a mere consequence of brain tissue postmortem degradation. The f-S25 band intensity was significantly weaker compared to that of monomeric, full-length SNAP25 (FL-S25). Thus, minimal (and perhaps undetectable) induction of FL-S25 cleavage may yield large increases in f-S25 accumulation, as we observed for other breakdown products (Ramos-Miguel and Garcia-Sevilla, 2012; Ramos-Miguel et al., 2013). Neither anti-syntaxin-1 (SP6 and SP7) nor anti-VAMP (SP10 and SP11) antibodies immunodetected bands below the molecular weights of the corresponding monomeric bands (Figure 1A), and the existence of syntaxin-1 and VAMP breakdown products was considered irrelevant for the purpose of the study.

Since SNARE protein-protein interactions (PPIs) are SDS resistant (Hayashi et al., 1994), immunoblots of non-boiled samples show multiple bands (~65–200 kDa) putatively corresponding to SNARE complexes with variable stoichiometry (see SNAP25 and syntaxin-1 panels in Figure 1A). Because our VAMP monoclonal antibodies recognize additional spurious bands, the presence of SNARE complexes in anti-VAMP immunoblots may be obscured. In any case, boiling human and rat brain homogenates prior to SDS-PAGE fully dissociated SNARE complexes, and a concomitant increase in all SNARE monomers (including VAMP) was observed. Remarkably, f-S25 signal also increased upon full SNARE dissociation (Figure 1A), indicating that this cleaved species can form SDS-resistant complexes with potential physiological activity at synapses. In fact, f-S25 was highly enriched in crude synaptosomal extractions from postmortem human brain, even when compared to FL-S25 (Figure 1C) and other presynaptic proteins (not shown; see for comparison (Ramos-Miguel et al., 2017). Furthermore, f-S25 appears to preserve the palmitoylated cysteine residues present in the FL-S25 central domain, as this breakdown product only precipitated along with membrane attached proteins (Figure 1D). Unlike FL-S25, f-S25 was barely present in lipid raft microdomains. In turn, f-S25 precipitated along with low-density membrane-related domains, where most vesicle fusion activity occurs (Aoyagi et al., 2005).

f-S25–syntaxin-1 interaction is sensitive to SNARE (de)phosphorylation

To further explore potentially functional aspects of f-S25, we next addressed the question of whether this species retains the ability to bind other SNARE proteins, as suggested when comparing immunoblots from boiled and non-boiled brain homogenates above. To avoid inter-individual variability, we pooled human cortical homogenates from n = 9 healthy subjects, which were utilized as inputs in IP assays with antibodies against SNAP25 (SP12), syntaxin-1 (SP7), and VAMP (SP11). For comparative purposes, images displaying FL- and f-S25 come from the same immunoblots at short (20 s) or long (3 min) exposure time, respectively. Regardless of the SNARE protein targeted, all IP reactions yielded comparable amounts of FL-S25, syntaxin-1 and VAMP (Figure 2A), suggesting that these molecules are near to equimolarity in human brain tissue, and among them, the great majority might be found preassembled under the present IP assay conditions. Strikingly, f-S25 was abundantly immunoprecipitated with anti-SNAP25 and anti-syntaxin-1 antibodies, but barely detected in VAMP IP reactions. This observation may not only indicate that the f-S25–VAMP interaction is weak (or perhaps nonexistent), but also that f-S25 bound to syntaxin-1 may hinder VAMP recruitment to the complex. The fact that Munc18–1 (a syntaxin-binding protein) was also found in comparable amounts in all three IP reactions (Figure 2A) further supports that f-S25 is not required (and possibly hinders) full SNARE assembly.

Figure 2.

Functional characterization of the 15-kDa SNAP25 fragment (f-S25) protein-protein interactions (PPIs). (A) Solubilized human brain proteins from the inferior temporal cortex (IT, BA 20; two different subjects were tested with similar results) were used as inputs for immunoprecipitation (IP) assays using magnetic beads conjugated with antibodies against mouse IgG (IgG; negative control), SNAP25 (SP12), syntaxin-1 (STX1) (SP7), or VAMP (SP11). The resulting IP products, and the input sample (diluted 1:10; i.e. 0.75 µg of total brain protein loaded), were boiled for 5 min and resolved by SDS-PAGE followed by IB with specific antibodies against SNAP25 (SP14), syntaxin-1 (SP6), VAMP (SP10), and Munc18–1 (see Appendix Table S2). For all tested antibodies, no immunoreactivity was detected in the IgG IP reactions, reflecting a high purity of the extraction procedure and the specificity of the protein-protein interactions. The relatively large enrichment of f-S25 is denoted by the apparent absence of signal in the input, compared to that of FL-S25. (B) Solubilized proteins from rat brain samples (pool of 20 different rat frontal cortices) were incubated (in triplicate) in the absence (AP–) or presence (AP+) of alkaline phosphatase (AP; 1 U/µl, 1 h, 37ºC). AP– and AP+ brain homogenates were used as inputs for IP assays with anti-syntaxin-1 antibody (SP7), and the final products were resolved by SDS-PAGE followed by immunoblotting, as above. (C) Quantification of the immunoreactive bands in (B) revealed that AP treatment significantly reduced f-S25–syntaxin-1, while increasing Munc18–1–syntaxin-1, PPIs; ***P < 0.001 (Student’s t-test). (A and B) Molecular mass (in kDa) of the nearest prestained standard marker is shown on the left.

Phosphorylation of SNARE proteins and partners is known to regulate the complex formation, and hence neurotransmitter release in vivo (Snyder et al., 2006). To further understand the putative physiological role of SNAP25 fragmentation in presynaptic terminals, we assessed the impact of full dephosphorylation of brain proteins on syntaxin-1–f-S25 binding affinity in vitro. For these assays, we used rat, rather than human cortical homogenates to avoid the large impact of postmortem degradation on phosphoprotein detection (Ramos-Miguel et al., 2009). Samples were initially incubated in the presence or absence of concentrations of alkaline phosphatase known to fully dephosphorylate SNARE proteins in brain homogenates (Ramos-Miguel et al., 2015b), and then used as inputs in syntaxin-1 IP extractions. While alkaline phosphatase treatment had no significant impact on core SNARE PPIs, dephosphorylated syntaxin-1 displayed much lower efficacy (–86%, P < 0.001) to pull down f-S25 (Figure 2B,C). On the other hand, enzymatic dephosphorylation enhanced 2-fold (+115%, P < 0.001) syntaxin-1–Munc18–1 interaction, further suggesting that f-S25 attachment to syntaxin-1 may prevent full SNARE constitution.

Inspection of datasets and correlations with potentially confounding variables

In the present study, we used standard qIB to measure the relative amounts of FL- and f-S25 in postmortem OFC samples from n = 78 subjects, and also utilized data obtained in a previous work (in the same samples) assessing the immunodensity of the 150-kDa SNARE complex, characterized as the major SNARE species in a blue-native (BN)-PAGE setting (Ramos-Miguel et al., 2015b, 2018). All data series obtained in these neurochemical assessments were normally distributed, according to Kolmogorov-Smirnov test. Quantitative data for f-S25 showed greater between-subject variability (interquartile range [IQR]: 44.7–109.3) than that of FL-S25 (IQR: 89.0–109.2), 150-kDa SNARE complex (IQR: 93.9–119.6), or β-actin (IQR: 93.9–119.6). Subsequent β-actin normalization of the datasets did not alter these observations, and had no substantial impact on the overall results of the study. Therefore only β-actin-normalized data is presented hereon to keep consistency with previous studies (Ramos-Miguel et al., 2015b, 2015c). Similarly, f-S25 correction by FL-S25 did not significantly alter the results, and data analyses were performed using f-S25 alone and/or using the ratio of f-S25 to FL-S25, as considered most appropriate in each case.

Multivariate analyses did not reveal relevant associations between potentially confounding variables (age at death, gender, brain pH, presence of antipsychotics, benzodiazepines, and/or ethanol, and/or smoking habit) and FL-S25 or SNARE complex measures. However, some of these variables showed significant associations with f-S25 (and with the ratio of f-S25 to FL-S25; Figure 3). Importantly, when subjects were segregated by diagnostic groups, lower amounts of cortical f-S25 were observed in males (versus females) with schizophrenia (–34%, P = 0.007), and in MDD cases who committed suicide compared to those who did not (–67%, P = 0.029) (Figure 3A). We also observed a non-significant difference in f-S25 cortical levels across cohorts (–16%, P = 0.233). Similarly, larger PMI (R = 0.433; P < 0.001), and perhaps higher brain pH (R = –0.212; P = 0.062) values, were associated with greater f-S25 cortical immunodensities (Figure 3B). Note that these figures represent the ratio of f-S25 to FL-S25 (rather than f-S25 alone) to emphasize that controlling for total protein amounts did not correct f-S25 data. These observations remained unchanged when uncorrected f-S25 data was used instead (not shown).

Figure 3.

(A–B) Associations between multiple potentially confounding variables of the study and the relative amounts of the SNAP25 fragment (f-S25) to full-length SNAP25 (FL-S25) ratio in the orbitofrontal cortex (OFC) of subjects with major depression (MDD, blue), schizophrenia (SCZ, red), and healthy controls (CTL, green). (A) For potentially confounding discrete variables, such as cohort source [i.e. Stanley Medical Research Institute Neuropathology Consortium (SNC), and the Macedonian/NY State Psychiatric Institute Collection (MNC)], sex, suicide as the cause of death, or antipsychotic use, comparisons were made using multiple t-tests within clinical diagnosis groups (i.e. CTL/MDD/SCZ). Bars represent group mean ± SEM values of the f-S25/FL-S25 ratios in percentage to an in-gel standard sample (pool of controls). *P < 0.05, and **P < 0.01 (Student’s t-test) (B) Scatterplots depicting the associations between age at death, postmortem interval (PMI), or brain pH, and the f-S25/FL-S25 ratio in all subjects from both cohorts combined. For each analysis, the corresponding Pearson’s coefficients and P-values are provided. (C) Effect of postmortem degradation on the cortical (inferior temporal gyri) immunodensities of full-length SNAP25 (FL-S25) and the SNAP25 fragment (f-S25). Cortical samples were collected from 9 previously healthy men suffering sudden and unexpected deaths, and the immunodensities of the SNAP25 species were quantified by standard SDS-PAGE followed by immunoblotting with anti-SNAP25 (SP12) antibody. β-actin content was used as a loading control. Individual densities (calculated as a percentage of all subjects) are represented versus the corresponding postmortem interval value (PMI; range: 7–25 h). Scatterplots representing linear decay models do not show a significant PMI effect on either neurochemical measure (Pearson’s coefficients and P-values are provided at the top-right corners within each plot). Bottom: representative immunoblots of FL-S25, f-S25, and β-actin. (D) Effect of one-month of antipsychotic treatments on rat cortical immunodensities of FL-S25 and f-S25. Animals were treated with saline-vehicle (Sal; n = 10), haloperidol (Hal, 1 mg/kg; n = 10) or clozapine (Clo, 20 mg/kg; n = 10) for 28 days. Cortical densities of the SNAP25 species were quantified by standard SDS-PAGE, followed by immunoblotting anti-SNAP25 (SP12) antibody. β-actin content was used as a loading control. Columns represent mean ± SEM values in percent to saline-treated animals. For each quantified species, representative immunoblots are shown below the plots.

To further explore the magnitude of the impact of postmortem protein degradation on brain SNAP25 immunoreactive brands, we quantified FL-S25 and f-S25 in a separate human postmortem brain series where PMI was isolated as the major apparent source of variability (n = 9; PMI range: 7–25 h). In this cohort, neither FL-S25 (R = 0.138; P = 0.724) nor f-S25 (R = 0.136; P = 0.727) showed significant associations with PMI (Figure 3C), indicating that intrinsic between-subject variability in humans has greater impact on FL-S25 and f-S25 cortical levels than PMI itself.

While the presence of antipsychotic drugs did not appear to be a potential confounding factor in our psychiatric postmortem brain cohort (Figure 3A), several studies reported (inconsistent) effects on brain SNARE proteins in rodents (recently summarized in (Ramos-Miguel et al., 2015d)). Therefore, we modeled the potential effects of one-month treatment with haloperidol or clozapine (as representatives of typical and atypical antipsychotics) in rats. Consistent with the above observations in human postmortem brain, exposure to either of these drugs did not significantly alter FL-S25 or f-S25 cortical immunodensities (Figure 3D), and possible antipsychotic-driven effects were no longer considered.

In summary, brain bank source, sex, PMI and brain pH were characterized as potential confounding factors in the present study, and therefore were included as covariates in all the following statistical analyses.

Downregulation of SNAP25 cleavage associates with enhanced SNARE assembly in schizophrenia OFC

Quantitative IB analyses in the OFC of cases and controls revealed that the immunoreactivity of the 15-kDa f-S25 band (–37%, P = 0.027), but not that of FL-S25 (+3%, P = 0.825), was significantly lower in subjects with schizophrenia, when compared to healthy controls (Figure 4A,B and Table 1). In contrast, in MDD, cortical amounts of both FL-S25 and f-S25 did not differ from control subjects, which may confer illness specificity to schizophrenia brain findings. Although f-S25 is a breakdown product from FL-S25 (and an inverse correlation between them might be expected), higher levels of the cleaved species were strongly associated with greater FL-S25 immunodensities (R = 0.390; P < 0.001) (Figure 4C), indicating that overall SNAP25 synthesis remains the major driving force for f-S25 brain availability.

Figure 4.

Quantitative analyses of presynaptic SNARE species in the orbitofrontal cortex (OFC, BA 10/47) of subjects with schizophrenia (SCZ), major depression cases (MDD), and healthy matched controls (CTL). (A) Immunodensities of full-length SNAP25 (FL-S25) and the 15-kDa fragment (f-S25) were quantified by immunoblotting with anti-SNAP25 (SP12) antibody following standard SDS-PAGE, and data was normalized with the corresponding β-actin immunodensities. SNARE complex formation (SNARE; 150 kDa band) was quantified in a previous study (Ramos-Miguel et al., 2015b) by immunoblotting with anti-SNAP25 (SP12), anti-syntaxin-1 (SP7), and anti-VAMP (SP10) antibodies following blue native (BN)-PAGE. Quantitative SNARE data was very similar for all three antibodies, and the average across them was used as the final estimate. Individual values (points), as well as the mean ± SEM (horizontal bars), were represented as percentage values relative to a standard sample (pool of controls) loaded in triplicate in all gels. ANCOVA detected between-group differences for f-S25 and SNARE (see Table 1). *P < 0.05, ***P < 0.001 (ANCOVA followed by Tukey’s HSD test). (B) Representative immunoblots of the SNAP25 species and β-actin (obtained by SDS-PAGE), and the 150-kDa SNARE complexes (obtained by BN-PAGE). Molecular mass (in kDa) of the nearest prestained standard marker is shown on the left. (C) Scatterplots showing the associations between the different SNAP25 species and 150-kDa SNARE complexes. The corresponding Pearson’s correlation test results are shown on the top of each plot.

Table 1.

Summary of the ANCOVA test results analyzing the differences across diagnostic groups on the neurochemical measures in cortical (OFC) samples.

Source	df	FL-S25				f-S25				SNARE complex
		SS	F-ratio	P-value		SS	F-ratio	P-value		SS	F-ratio	P-value
Cohort	1	309	1.03	0.314		6176	2.64	0.109		473	0.47	0.497
Sex	1	56	0.19	0.668		6600	2.82	0.097		673	0.66	0.418
PMI	1	579	1.93	0.169		22803	9.75	0.003*		1300	1.28	0.261
Brain pH	1	307	1.02	0.315		274	0.12	0.733		1001	0.99	0.324
Diagnosisa	2	108	0.18	0.836		17639	3.77	0.028*		19355	9.54	<.001*
Model	6	1156	0.64	0.697		49181	3.50	0.004*		20880	3.43	0.005*
Error	71	21322				166117				72005

Abbreviations: ANCOVA, analysis of covariance; df, degrees of freedom; f-S25, 15-kDa SNAP25 fragment; FL-S25, full-length SNAP25; OFC, orbitofrontal cortex; SNARE, soluble N-ethylmaleimide-sensitive factor attachment protein receptor; SS, sum of squares.

^aDiagnostic groups compared included healthy controls, subjects with major depression, and subjects with schizophrenia

^*Statistically significant

On the other hand, the intact SNARE complex, previously characterized and quantified as a 150-kDa band by BN-PAGE (Ramos-Miguel et al., 2015b, 2018), was robustly increased in OFC postmortem samples from schizophrenia cases (+35%, P < 0.001), compared with controls (Figure 4A,B). In subjects with MDD, cortical levels of SNARE complex also appeared slightly, but not significantly, higher than those in control OFC samples (+22%, P = 0.118). Notably, cortical immunodensities of FL-S25 (R = 0.166; P = 0.146) and f-S25 (R = −0.251; P = 0.026) correlated with SNARE complex assembly in opposite directions. These contrasting relationships became evident when analyzing the association between SNARE complex levels and the f-S25/FL-S25 ratio (R = −0.563; P < 0.001), suggesting that the balance between these two SNAP25 species could importantly determine SNARE complex assembly.

Reduced f-S25 could contribute to the pathophysiology of schizophrenia via SNARE complex formation

The above findings suggest a pathophysiologic mechanism where lower f-S25 availability (e.g. inhibition of FL-S25 cleavage) could facilitate SNARE complex formation, and eventually result in enhanced likelihood of schizophrenia. This hypothesis implies a mechanism where SNARE complex formation acts as mediating step in the association between f-S25 and schizophrenia, which can be tested statistically. The properties of the specific linear and logistic regression models constructed to test the hypothesis (schematized in Figure 5) were in agreement with the conventional postulates and assumptions for this type of modeling approach (Hayes, 2017). The analyses predict that cortical f-S25 amounts ~50% less than the average increased by 2.6-fold the likelihood of schizophrenia (odds ratio [OR] = 0.38, P = 0.013) (Table 2, Model 2). Similarly, a ~30% increase in SNARE complex levels was associated with a 3-fold greater risk of schizophrenia (OR = 2.99, P < 0.001). Remarkably, when both synaptic measures were included in the model (Table 2, Model 3), only SNARE complex (OR = 2.53, P = 0.017), but not f-S25 (OR = 0.71, P = 0.433), remained significantly associated with schizophrenia diagnosis, validating the hypothesis of a full mediation of SNARE complex in the f-S25–schizophrenia risk association. In this modeling approach, the indirect effect (a×b = −0.617) accounted for 65% of the total effects (c = −0.946) of f-S25 on schizophrenia diagnosis, which exceeded by far the 35% contribution estimated for the direct effects (c’ = −0.339) (Table 2).

Figure 5.

Schematic representation and mathematical expression of the statistical approach to support the mediation relationship between the SNAP25 fragment (f-S25) and SNARE complex formation in the orbitofrontal cortex, increasing the risk of schizophrenia (SCZ). Within the formulae: β, covariate coefficient; cov, covariate; err, model error; int, model intercept; n, number of covariates

Table 2.

Series of linear (Model 1) and logistic (Models 2 and 3) regression models testing (as schematically represented in Figure 5) the relationship between schizophrenia diagnosis (dependent variable, Y) and the OFC levels of f-S25 (independent variable, X) and SNARE complex (mediator, M), controlling for FL-S25 amounts and other potentially confounding variables

Model # Method	Model outcome	Term	Term effects
			Coeff ID‡	Std. β	S.E.	OR‡‡ (95% CI)	Wald test	P-value
Model 1 Ordinary least squares	SNARE complex	Intercept		0.005	0.095	–	–	0.962
		Cohort		−0.019	0.131	–	–	0.885
		Sex		0.063	0.101	–	–	0.531
		PMI		0.099	0.134	–	–	0.462
		Brain pH		−0.114	0.096	–	–	0.239
		FL-S25		0.027	0.106	–	–	0.798
		f-S25 (X)	a	−0.663	0.111	–	–	<.001*
Model 2 Nominal logistic regression	Schizophrenia diagnosis	Intercept		−0.149	0.258	–	0.33	0.565
		Cohort		−0.889	0.372	5.91 (1.37–25.3)	5.72	0.017*
		Sex		0.259	0.280	0.59 (0.19–1.78)	0.85	0.356
		PMI		0.539	0.367	1.71 (0.83–3.51)	2.16	0.142
		Brain pH		0.146	0.256	1.15 (0.70–1.91)	0.33	0.567
		FL-S25		0.218	0.291	1.24 (0.70–2.20)	0.56	0.455
		f-S25 (X)	c	−0.946	0.383	0.38 (0.18–0.82)	6.11	0.013*
Model 3 Nominal logistic regression	Schizophrenia diagnosis	Intercept		−0.113	0.269	–	0.18	0.674
		Cohort		−0.965	0.394	6.88 (1.46–32.2)	5.99	0.014*
		Sex		0.246	0.295	0.61 (0.19–1.94)	0.70	0.404
		PMI		0.486	0.375	1.62 (0.77–3.39)	1.68	0.195
		Brain pH		0.301	0.274	1.35 (0.78–2.31)	1.21	0.272
		FL-S25		0.219	0.303	1.24 (0.68–2.25)	0.52	0.471
		f-S25 (X)	c’	−0.339	0.431	0.71 (0.30–1.66)	0.62	0.433
		SNARE complex	b	0.930	0.389	2.53 (1.18–5.43)	5.72	0.017*

Abbreviations: CI, confidence intervals; f-S25, 15-kDa SNAP25 fragment; FL-S25, full-length SNAP25; OFC, orbitofrontal cortex; OR, odds ratio; PMI, postmortem interval; SNARE, soluble N-ethylmaleimide-sensitive factor attachment protein receptor complex; S.E., standard error; Std. β, standardized beta-coefficient.

^‡Coefficient identifier (ID) as represented in Fig. 5:

^atotal effect X➔M

^btotal effect M➔Y

^ctotal effect X➔Y

^c’direct effect X➔Y; the indirect effect X➔Y is estimated as the a×b product (i.e. a×b = −0.617; S.E. = 0.439; 95% CI = [−1.50]–[−0.01])

^‡‡Odd ratios per unit change in regressor

^*Statistically significant

Overall calpain activity is not altered in schizophrenia OFC

Since calpain is known to cleave FL-S25 (Ando et al., 2005; Grumelli et al., 2008), we addressed the possibility that downregulated calpain activity would be responsible for the lower amounts of f-S25 in schizophrenia OFC. The quantitative assay showed high accuracy and a broad linear range, demonstrating the presence of active calpain (at least 0.125–2 ng of calpain range) (Figure 6A). Similarly, signal intensity of calpain activity was log-linear in relation to total brain protein loaded in the assay wells (linearity range: 10–300 µg of protein) (data not shown). In the quantitative assays of postmortem human OFC samples (note that the SFNC collection was not analyzed here due to availability limitations), the overall tissue proteolytic activity of calpain did not differ between schizophrenia cases and controls (Figure 6B). Furthermore, total calpain activity was not associated with the SNAP25 cleavage in the presynaptic terminals, as represented by the f-S25 to FL-S25 ratio (Figure 6C), or f-S25 immunodensity alone (not shown).

Figure 6.

Quantitative analyses of calpain activity using a commercial kit in ELISA microplates. (A) Standard curve depicting the linear association between the amount of active calpain (range 0.125–2.0 ng) and the relative fluorescence units (RFU) obtained when 5 µl of the fluorogenic substrate (Ac-LLY-AFC) were present. Background signal (calculated from wells where all components excluding active calpain were added) was subtracted before data representation. Substrate cleavage was fully prevented when 1 ng of calpain inhibitor (Z-LLY-FMK) was added into the wells prior to adding active calpain. (B) Quantitative calapin assay in postmortem OFC samples from schizophrenia cases (SCZ) and controls (CTL) of the Macedonian/NY State Psychiatric Institute collection. (C) Scatterplots depicting the lack of association between total cortical calpain activity and the ratio of 15-kDa SNAP25 fragment (f-S25) to full-length SNAP25 (FL-S25) in the same SCZ/CTL samples. Similar negative results were used when uncorrected f-S25 datasets were used. Pearson’s correlation test result is shown on the top.

In silico characterization of calpain-mediated SNAP25 fragmentation

The expected cleavage site(s) of SNAP25 was investigated in silico using an online available algorithm exploring calpain-1 consensus sequences in putative calpain substrates (duVerle et al., 2010). Overall, prediction studies identified ten residues where calpain-1 could potentially proteolyze both SNAP25 splice variants, SNAP25A and SNAP25B (Figure 7A–C). Among them, only four sites (Gly¹¹¹, Ala¹¹⁴, Met¹²⁷, Arg¹³⁵) might yield N-terminal fragments of a molecular size similar to that of f-S25 (15–18 kDa). Since multiple N- and C-terminal fragments are usually recognized in the immunoblots of human and rodent brain samples, it is likely that several of these sites are actually targeted by calpain-1.

Figure 7.

(A–C) In silico prediction of the cleavage sites in (A) SNAP25A and (B) SNAP25B peptide sequences potentially targeted by calpain-1, and (C) scores and probabilities obtained with the prediction algorithm (available at www.calpain.rog) for all significant cleavage sites. Red boxes contain strongest candidate sites considering the molecular size observed for the SNAP25 fragment (f-S25). (D) Diagram summarizing the molecular pathway and the potential physiological role of SNAP25 (S25) fragmentation.

DISCUSSION

We explored neurochemical alterations that could underlie the abnormally enhanced SNARE PPIs occurring in schizophrenia OFC. The results suggest a pathophysiologic mechanism whereby a cleavage product of SNAP25 contributes to these abnormalities.

As a first step, we used total brain homogenates to identify stable breakdown products from all three SNARE proteins that could be detected by IB with multiple, specific antibodies, following standard SDS-PAGE separation. While no breakdown products from syntaxin-1 or VAMP were detected, three different monoclonal antibodies selectively recognizing the N-terminal domain of SNAP25 (SP12, SP14, and SMI-81) identified a 15–18-kDa antigen, herein called f-S25. A previous study documented the presence of a similar N-terminal fragment with the same apparent molecular size, generated by calpain-mediated SNAP25 proteolytic cleavage, in hippocampal primary cultures and other neuronal cell lines (Grumelli et al., 2008). Of note, SMI-81 is known to react only against the acetylated N-terminus of the SNAP25 molecule (Connell et al., 2009), whereas SP12 was able to immunoprecipitate small SNAP25 N-terminal peptides following trypsin digestion, and analysis by mass spectrometry (Brinkmalm et al., 2014b). Although the SP14 epitope has not been described in detail, both SMI-81 (not shown) and SP14 (Figure 2A,B) could recognize SP12-immunoprecipitated f-S25 from human and rat brains. These findings are consistent with f-S25 being present in the N-terminal portion of the SNAP25 molecule, possibly resulting from calpain proteolytic activity. In addition, an antibody raised against the C-terminal sequence of SNAP25 detected non-overlapping fragments sized at ~15 and ~20 kDa, respectively. Similar results were observed in another study reporting that calpain proteolytic activity yielded a ~20-kDa SNAP25 fragment in cultured cerebellar granule cells, recognized with antibodies raised against the C-terminal portion of SNAP25 (Ando et al., 2005). Future analyses using mass spectrometry will be required to verify the exact sequences of these truncated N/C-terminal SNAP25 peptides, as well as the precise cleavage site(s). Our in silico assay identified four residues in the SNAP25 central domain targetable by calpain-1, and which could yield N-terminal fragments compatible with f-S25 molecular size. Notably, the presence of multiple cleavage sites is consistent with the fact that at multiple bands were detected in the immunoblots with anti-N/C-terminal SNAP25 antibodies within the range of 15–20 kDa.

Interestingly, several findings of the present study strongly suggest that, rather than being a byproduct of SNAP25 degradation, f-S25 might actually serve functional purposes in mammalian nerve endings. First, compared with FL-S25, f-S25 is highly enriched in those synaptosomal membrane microdomains with low hydrophobicity (i.e. 0.1% TritonX soluble) where the vast majority of presynaptic vesicle fusion events occur (Aoyagi et al., 2005). Second, f-S25 selectively binds syntaxin-1, and displays very low (or perhaps no) affinity for VAMP. In fact, co-IP assays may indicate that f-S25 could serve as an endogenous inhibitor for full SNARE assembly (Figure 7D summarizes a proposed mechanism). Supporting this possibility, treatment of f-S25 with alkaline phosphatase resulted in greater amounts of syntaxin-1 PPIs with other SNARE partners, such as Munc18–1, complexin and synaptotagmin (present results and (Ramos-Miguel et al., 2015b)). Another plausible mechanism is proteolytic processing of SNAP25 serving to control, upon excessive neurotransmission, the readily releasable vesicle pool by detaching preassembled SNARE complexes from the presynaptic membrane. Note that vesicle fusion is triggered by Ca²⁺ penetration into the presynaptic terminal, and sustained micromolar concentrations of this ion are required for calpain activation (Campbell and Davies, 2012). Therefore, continuous firing events may keep elevated Ca²⁺ amounts in the terminal, which may in turn activate calpain locally and trigger SNAP25 cleavage. Notably, previous studies reported that sustained stimulation of ionotropic glutamate receptor channels (e.g. NMDA and kainate receptors) in cultured neurons robustly activated calpain-dependent SNAP25 fragmentation (Ando et al., 2005; Grumelli et al., 2008), which had a negative impact on the efficiency of further neurotransmitter release. However, both studies assumed that this deleterious effect was a consequence of reducing FL-S25 availability at synapses: the potential active role of the resulting truncated peptides was not examined. In this context, the ability of f-S25 to bind syntaxin-1, and potentially hinder full SNARE assembly, is a novel relevant finding of the present study. Either way, SNAP25 fragmentation, and the concomitant upregulation of f-S25, appears to provide a negative feedback tool tuning down neurotransmitter release in the presynaptic terminals. This model is of course speculative; the potential physiological activity of f-S25 is not demonstrated in the present series of experiments. Future studies overexpressing and/or knocking down fully characterized f-S25 will be needed to determine the synaptic role (if any) of this SNAP25 species.

Importantly, in postmortem OFC samples from subjects with schizophrenia, f-S25 expression levels were consistently lower than those in controls. In marked contrast, the ability of SNARE proteins to form complexes was enhanced in the same schizophrenia brain extracts. Furthermore, statistical modeling of the obtained datasets supported the hypothesis that the relatively lower amounts of f-S25 were an integral part of the pathway to elevated SNARE assembly in schizophrenia. These modeling approaches were controlled for multiple possibly relevant factors, such as sex or PMI, that could impact on protein expression and/or degradation rates. However, human postmortem studies are subjected to numerous, unexpected, and potentially confounding factors, which may represent limitations that must be acknowledged.

Consistent with the present data, up to three different murine models displaying features of schizophrenia display alterations leading to exacerbated neurotransmitter release. The blind-drunk mouse carries a spontaneous mutation resulting in a single amino acid substitution (I67E) in the SNAP25 peptide sequence (Jeans et al., 2007). This mutant SNAP25 protein displays enhanced efficiency to form SNARE complexes, associated with a schizophrenia-like endophenotype in the mouse (Jeans et al., 2007; Oliver and Davies, 2009). Another murine model overexpressing the SNARE partner Munc18–1 showed hypersensitivity to amphetamine, with abnormally elevated dopamine release compared to littermate controls (Uriguen et al., 2013). Finally, conditional knockouts for SNAP25 in glutamatergic neurons were found to mimic key features in patients with schizophrenia (Yang et al., 2017). Strikingly, SNAP25 deficiency in these mice led to upregulated levels of syntaxin-1 and VAMP2 at the presynaptic membrane, resulting in enhanced glutamate release. Altogether, these data suggest that qualitative and quantitative alterations in protein expression (in rodents) or proteolytic deficits (in human postmortem brains) favoring SNARE formation and/or overall neurotransmitter release, may result in features associated with schizophrenia. In agreement, neuroimaging studies reported greater global brain signal in patients with schizophrenia, a measure potentially related to the overall neurotransmitter release (Yang et al., 2014).

While calpain was previously shown to cleave SNAP25, the proteolytic activity of this enzyme, quantified in whole brain tissue homogenates, did not differ between schizophrenia subjects and controls. Calpain is a ubiquitous enzyme expressed in a wide variety of cell types, and most cellular compartments. Specific, presynaptic terminal calpain proteolytic activity (responsible for SNAP25 cleavage) was not assessed in the present assay, and may represent an insignificant fraction of the total calpain activity in whole tissue homogenates. Consistent with these observations, a previous study reported a robust downregulation of synaptic calpain activity associated with schizophrenia, as quantified by the immunodensity of another calpain substrate: the cyclin-dependent kinase 5 cofactor p35 (Ramos-Miguel et al., 2013). In turn, the lack of association between synaptic f-S25 immunodensity and total tissue calpain activity is consistent with the observed differences in SNAP25 fragmentations between schizophrenia cases and controls being unrelated to postmortem artifacts, or spurious manipulation of brain tissues during the assays, both of which would be driven by overall, rather than just synaptic, calpain activity. To verify whether or not downregulated synaptic calpain activity is responsible for the lower amounts of f-S25 in schizophrenia OFC, future studies could evaluate calpain activity in synaptosomal extractions, rather than total brain homogenates. Unfortunately, in the present work (as in most human postmortem brain studies), the limited amount of tissue available was a barrier to carrying out such experiments, due to the large quantities of starting material required.

CONCLUSIONS

In summary, the present study showed that altered processing resulting in lower than expected amounts of a SNAP25 fragment might be a relevant pathophysiological mechanism in schizophrenia brain. While the physiological role of f-S25 remains to be fully described, cleavage of FL-S25 might be a novel negative feedback mechanism to modulate excessive neurotransmission. Downregulation of this modulatory mechanism could contribute to excessive neurotransmission in schizophrenia. From a pharmacological perspective, these findings may provide an interesting target for future drug development, as calpain is the sole SNAP25 protease described. A therapeutic strategy selectively boosting presynaptic calpain activity could be a focus for future schizophrenia drug research.

Abbreviations:

ANCOVA

analysis of covariance

BA

Brodmann’s area

BN

PAGE blue native PAGE

BSA

bovine serum albumin

CTL

control

DC

detergent compatible

EDTA

ethylenediaminetetraacetic acid

ELISA

enzyme-linked immunoadsorbant assay

f-S25

SNAP25 fragment

FL-S25

full-length SNAP25

IB

immunoblotting

IOD

integrated optical density

IP

immunoprecipitation

IT

inferior temporal cortex

MDD

major depressive disorder

OFC

orbitofrontal cortex

PAGE

polyacrylamide gel electrophoresis

PBS

phosphate-buffered saline

PMI

postmortem interval

PPIs

protein-protein interactions

PVDF

polyvinylidene difluoride

qIB

quantitative IB

SCZ

schizophrenia

SDS

sodium dodecyl sulfate

SNAP25 (or S25)

synaptosomal-associated protein 25

SNARE

soluble N-ethylmaleimide-sensitive factor attachment protein receptor

TBS

Tris-buffered saline

VAMP

vesicle-associated membrane protein