Abstract
Presynaptic α-neurexins are highly expressed and more frequently linked to neuropsychiatric and neurodevelopmental disorders than β-neurexins. However, how extracellular sequences specific to α-neurexins enable synaptic transmission is poorly understood. We identified a mutation in an extracellular region of neurexin-3α (A687T), located in a region conserved among α-neurexins and throughout vertebrate evolution, in a patient diagnosed with profound intellectual disability and epilepsy. We systematically interrogated this mutation using a knockdown-replacement approach, and discovered that the A687T mutation enhanced presynaptic morphology and increased two critical presynaptic parameters: (1) presynaptic release probability, and (2) the size of the readily releasable pool exclusively at excitatory synapses in mixed sex primary mouse hippocampal cultures. Introduction of the mutation in vivo and subsequent analysis in ex vivo brain slices made from male and female mice revealed a significant increase in excitatory presynaptic neurotransmission that occluded presynaptic but not postsynaptic LTP. Mechanistically, neurexin-3αA687T enhanced binding to LRRTM2 without altering binding to postsynaptic neuroligin-1. Thus, neurexin-3αA687T unexpectedly produced the first neurexin presynaptic gain-of-function phenotype and revealed unanticipated novel insights into how α-neurexin extracellular sequences govern both transsynaptic adhesion and presynaptic neurotransmitter release.
SIGNIFICANCE STATEMENT Despite decades of scientific scrutiny, how precise α-neurexin extracellular sequences control synapse function remains enigmatic. One largely unpursued avenue to identify the role of precise extracellular sequences is the interrogation of naturally occurring missense mutations. Here, we identified a neurexin-3α missense mutation in a compound heterozygous patient diagnosed with profound intellectual disability and epilepsy and systematically interrogated this mutation. Using in vitro and in vivo molecular replacement, electrophysiology, electron microscopy, and structure–function analyses, we reveal a novel role for neurexin-3α, unanticipated based on α-neurexin knock-out models, in controlling presynaptic morphology and neurotransmitter release at excitatory synapses. Our findings represent the first neurexin gain-of-function phenotype and provide new fundamentally important insight into the synaptic biology of α-neurexins.
Introduction
Neurexins (Nrxns) are a family of essential presynaptic cell adhesion molecules that are thought to play pleiotropic roles in synapse formation, maintenance, and plasticity in a brain region, cell-type-, and synapse-specific manner, however, the exact roles each neurexin performs at the synapse are still poorly defined (Aoto et al., 2015; Südhof, 2017; Dai et al., 2019). Nrxns are encoded by three evolutionarily conserved genes (Nrxn1–3) that give rise to longer α and shorter β neurexin gene products from two independent promoters (Südhof, 2017). α-Nrxns and β-Nrxns share transmembrane and cytoplasmic sequences, which promote presynaptic function by interacting with synaptotagmin-1, CASK, Mint-1, and Veli (Reissner et al., 2013), but differ in the length and complexity of their extracellular sequences. α-Nrxn extracellular sequences are comprised of six laminin-neurexin-sex hormone (LNS1–6) domains interspersed with three EGF-like repeats (EGF1–3). β-Nrxns are essentially truncated α-Nrxns that contain only LNS6 (Fig. 1A). The extracellular region of neurexins is subject to extensive alternative splicing. Although poorly understood, alternative splicing is thought to govern neurotransmission by modulating transsynaptic interactions with postsynaptic ligands.
α-Nrxns are alternatively spliced at six conserved sites (SS1–6) that generate over 1000 splice isoforms (Fig. 1A; Ullrich et al., 1995), and of the six conserved splice sites, SS4 has received the most scrutiny. Splice-site 4 is located in LNS6 and is thus shared by α- and β-Nrxns. The inclusion or exclusion of the alternative SS4 exon controls transsynaptic interactions with almost all known neurexin ligands (Südhof, 2017). Although largely untested, Nrxn1α crystal structures suggest that α-Nrxns function as extracellular scaffolds by simultaneously interacting with multiple postsynaptic ligands via LNS1–6 to enable synapse function (Miller et al., 2011; Chen et al., 2017). Despite this proposed role, a functional interrogation of extracellular sequences unique to α-Nrxns has represented a daunting task due in large part to its length (∼1470 aa) and complexity (6 LNS domains, 3 EGF-like repeats and 6 alternative splice sites).
Given the length and complexity of α-Nrxn extracellular sequences, our functional understanding of these molecules has almost entirely relied on α-Nrxn constitutive knock-out (KO) mice. The constitutive ablation of all α-Nrxns impaired presynaptic release at excitatory and inhibitory synapses and caused a reduction of somatic calcium currents (Missler et al., 2003; Zhang et al., 2005). Interestingly, the loss of individual α-Nrxns resulted in distinct brain region and cell-type-specific phenotypes. Nrxn1α KO mice exhibited impaired postsynaptic excitatory synaptic strength in hippocampus (Etherton et al., 2009). By contrast, Nrxn3α KO mice did not have impaired excitatory synaptic strength in hippocampus but instead exhibited a significant reduction in presynaptic release probability at inhibitory synapses in olfactory bulb (Aoto et al., 2015). Together, these findings support the notion that individual α-Nrxns participate in essential overlapping and non-overlapping functions at the synapse. Notably, although informative, the reported phenotypes are a result of the complete loss of all transsynaptic interactions (e.g., the extracellular scaffold) and intracellular signaling. While these studies demonstrated the functional importance of α-Nrxns, the identification of the precise sequences necessary to enable these synaptic properties remains unknown.
One largely unpursued avenue to characterize the role of α-Nrxn extracellular sequences is the interrogation of naturally occurring missense mutations. Mutations in exons that encode for α-neurexin-specific extracellular sequences are linked to neuropsychiatric and neurodevelopmental disorders, strongly suggesting that these poorly understood α-Nrxn extracellular sequences play a significant role in sculpting synapse function (Südhof, 2008; Reissner et al., 2013). Moreover, these extracellular mutations might serve as valuable tools to identify and address outstanding questions regarding how α-neurexin extracellular sequences govern synapse function to ultimately provide novel functional insight that might otherwise be obscured in KO models. To this end, we identified a Nrxn3α compound heterozygous patient diagnosed with profound intellectual disability and epilepsy and systematically interrogated a missense mutation that revealed a novel role for neurexin-3α in controlling presynaptic morphology and neurotransmitter release at excitatory synapses.
Materials and Methods
Experimental model and subject details
Mouse generation and husbandry.
All mouse work was approved by University of Colorado Denver Anschutz Medical Campus animal use committees. For all experiments, wild-type (WT) male and female mice (culture: P0, in vivo stereotactic injections: 3 weeks old, ex vivo slice: 5–6 weeks old) were used. Mice were kept on a 12 h light/dark cycle.
Patient genomic sequences were provided by GeneDx and use of the sequences was approved by University of Colorado School of Medicine (COMIRB: 16-0152) to Dr. Scott Demarest.
Analysis of culture hippocampal neurons.
Hippocampi from P0 C57BL/6J were cultured on Matrigel-coated glass coverslips (Carolina Biological, 633009) as previously described (Aoto et al., 2013), infected with lentiviruses at 4 d in vitro (DIV) and analyzed at 13–15 DIV.
Virus preparations.
AAVs and lentiviruses were engineered using the human synapsin promoter to drive expression. shRNA resistant mouse cDNAs were used for rescue constructs expressed from separate lentiviruses. Lentiviruses were made in HEK293T cells as previously described (Aoto et al., 2013). Adeno-associated viruses (AAVs) were prepared as previously described (Aoto et al., 2013).
Plasmids.
The A687T mutation was introduced into mouse Nrxn3α SS4+ and SS4− cDNAs (XM_006515560.2) previously described (Aoto et al., 2015) by Gibson Assembly: F/R: ACCTGCGAAAGAGAGACATCTATCCTGAGCT/AGCTCAGGATAGATGTCTCTCTTTCGCA gGT. Gibson Assembly was used to insert mutant Nrxn3α into FSW: F/R: TCGTGCCTGAGAGCGCAGTCTAGAGAATTCGCCACCATGAGCTTTAC/GATAAGCTTGAT ATCGAATTGTTAACGGATCCTTACACATAATACTCCT and into AAV hSYN: F/R TGAGAGCGCAGTCGAGAAGGTACCGGATCCGCCACCATGAGCTTTACCCTCCACTCA/T TGATTATCGATAAGCTTGATATCGAATTCTTACACATAATACTCCTTGTCCTTG.
HEK aggregation assays were performed with cDNAs in pcDNA3.1 (Invitrogen) or pCMV. Gibson assembly was used to insert Nrxn3α and NL1 and 3 into pcDNA3.1. Nrxn3α F/R: TTTAAACTTAAGCTTGGTACCGAGCTCGGATCCGCCACCATGAGCTTTACCCTCCACTC/GAGCGGCCGCCACTGTGCTGGATATCTGCAGAATTCTTACACATAATACTCCTTGTCCTT; NL1 F/R: TAAACTTAAGCTTGGTACCGAGCTCGGCCACCATGGCACTTCCCAGATGCA TGTGG/CGCCACTGTGCTGGATATCTGCAGAATTCTATACCCTGGTTGTTGAATGTGAA; NL3 was inserted into pcDNA3.1 at HindIII/XbaI using F/R primers: AAGCTTGCCACCATGTG GCTGCAGCCCTCGCTGT/TCTAGACTATACACGGGTAGTGGAGTGTGAGT.
Immunocytochemistry.
Live surface immunolabeling of HA-tagged Nrxn3α WT and A687T was performed as previously described (Aoto et al., 2013). Briefly, live neurons were washed with prewarmed PBS 0.5 mm CaCl2 1 mm MgCl2 and stained with mouse anti-HA (1:500; BioLegend) diluted in blocking buffer (2% normal goat serum; (Sigma-Aldrich) in PBS) for 5 min at 37°C. Cells were washed three times with PBS for 5 min on ice before fixing with 4% PFA + 4% sucrose. Donkey anti-mouse Cy3-conjugated secondary antibody (1:200; Jackson ImmunoResearch; RRID:AB_2315777), diluted in blocking buffer, was applied for 45 min at room temperature.
Fixed staining of neurons was performed as previously described (Aoto et al., 2013). Briefly, neurons were fixed with 4% PFA and 4% sucrose, washed 3× with PBS before permeabilization with PBS+0.2% Triton X-100. Neurons were then incubated for 45 min in blocking buffer (2% normal goat serum + 0.2% Triton X-100 in PBS). Primary antibodies were applied to detect presynaptic proteins vGluT1 (1:1000; Synaptic Systems; RRID:AB_887878) and vGAT (1:500; Synaptic Systems; RRID:AB_887869), or postsynaptic markers: PSD95 (1:200; ThermoFisher; RRID:AB_325399) and Gephyrin (1:500; Synaptic Systems; RRID:AB_2619834). Secondary Cy2, Cy3 and Cy5-conjugated secondary antibodies (1:200; Jackson ImmunoResearch; RRID:AB_2307443; RRID:AB_2340813; RRID:AB_2492288) were applied then mounted on glass slides with Fluoromount-G (Southern Biotech) followed by imaging using spinning disk confocal microscopy at room temperature using a Zeiss Axio Observer microscope with a 63× oil-immersion objective Plan-Apo/1.4 NA objective, using 488, 594, and 647 laser excitation and a CSU-XI spinning-disk confocal scan head (Yokogawa) coupled to an Evolve 512 EM-CCD camera (Photometrics) controlled by SlideBook software. At least three coverslips per condition were imaged per session, and ∼15–25 cells were acquired per condition. The same settings for laser power and photomultiplier were used to account for negligible bleed-through. During analysis, images were thresholded by intensity to exclude background noise and puncta were counted in the range of 0.25 μm2.
qRT-PCR of Nrxn3α mRNA in primary hippocampal cultures.
Quantitative RT-PCR with β-actin as an internal control was used for mRNA measurements. mRNA was isolated using Quick-RNA Micro-Prep mRNA isolation kit (Zymo Research). qScript XLT 1-Step RT-qPCR ToughMix Kit (Quantabio) was used for RT. CFX384 Real-Time System RT-PCR instrument (Bio-Rad) was used to run and analyze reaction. TempPlate 384-Well Full-Skirt PCR, Single Notch polypropylene plates were used (USA Scientific). β-actin: F: GACTCATCGTACTCCTGCTTG, R: GATTACTGCTCTGGCT CCTAG, Probe: CTGGCCTCACTGTCCACCTTCC. Nrxn3α: F: TCCAGTTCAAGACCA CTTCAG, R: GTCACTGTTGCCTTTG ATCAC, Probe: TGGGACCATTGCCGAGATCA AACA. Nrxn3β: F: ACCACTCTGTGCCTATTTCTATC, R: TGTGCTGGGTCTGTCATTT G, Probe: TCGCTCCCCTGTTTCCCTTCG.
Culture electrophysiology.
Whole-cell voltage-clamp recordings were made from pyramidal neurons, identified by the presence of dendritic spines (GFP fill), in all conditions. Recording pipettes were pulled to 3–4 mΩ and series resistance was constantly monitored. Cells that experienced series resistance changes >20% were discarded. Cultures were continuously superfused with room temperature ACSF (in mm): 126 NaCl, 2.5 KCl, 1 NaH2PO4, 26.2 NaHCO3, 2.5 CaCl2, 1.3 MgSO4•7H2O, 11 d-glucose, ∼290 mOsm. The internal solution consisted of the following (in mm): 115 Cs-methanesulfonate, 15 CsCl, 8 NaCl, 0.2 EGTA, 10 HEPES, 4 Mg-ATP, 0.3 Na-GTP, 10 TEA-Cl, 10 Na2-phosphocreatine, 1 MgCl2, ∼290 mOsm. Cells were held at −70 mV and mIPSCs (0.5 μm tetrodotoxin, 0.5 μm NBQX, and 50 μm d-AP5) and mEPSCs (0.5 μm tetrodotoxin and 100 μm picrotoxin) were recorded. Miniature events were handpicked and analyzed in Clampfit 10 (Molecular Devices) using template matching and a threshold of 5 pA. Evoked synaptic currents were elicited with a homemade nichrome bipolar stimulating electrode placed 100–150 μm from the soma of recorded neurons, using a Model 2100 Isolated Pulse Stimulator (A-M Systems) controlled by Clampex 10 data acquisition software (Molecular Devices). After formation of the whole-cell configuration and equilibration of the internal solution, the series resistance was compensated to 5–8 MΩ. For AMPAR responses, cells were held at −70 mV and recorded in the presence of picrotoxin (100 μm) d-AP5 (50 μm). For evoked NMDAR, cells were held at +40 mV and recorded in NBQX (5 μm). Evoked IPSCs were measured at −70 mV in the presence of d-APV (50 μm) and NBQX (5 μm). Presynaptic release probability was measured during paired pulse recordings at interstimulus intervals (ISIs) ranging from 20 to 1000 ms. Cells that did not exhibit short-term plasticity at all short ISIs (20–100) were discarded from further analysis. For all evoked measurements, QX-314 (5 mm) was added to the internal to prevent firing. For all experiments, the experimenter was blind to the recording condition.
Stereotactic injections.
Stereotactic injections were performed as previously described (Liu et al., 2019). Briefly, P21 WT C57BL/6J mice were deeply anesthetized with tribromoethanol (200 mg/kg). Using a syringe pump (WPI), 1 μl AAV was injected bilaterally into CA1 at a rate of 14 ml/h. The pulled glass micropipette (WPI) was held in place for 5 min to prevent backflow before being slowly withdrawn. Coordinates for CA1 (from bregma): AP: −3.2, ML: ±3.45, DV: −2.5 (relative to pia). Efficiency and localization of infection was assessed mRuby fluorescence in CA1 on day P35. Stitched images were acquired using an Olympus VS120 slide-scanning microscope with Olympus VS-ASW software.
Acute slice electrophysiology
Whole-cell recording in subiculum was performed on 300 μm thick acute slices from WT mice coinfected with AAV encoding shRNA and mRuby or shRNA + Nrxn3αWT/A687T SS4− into CA1 were performed in vivo by stereotactic injection on P21 mice.
Slices were made in cold cutting solution containing the following (in mm): 85 NaCl, 75 sucrose, 2.5 KCl, 1.3 NaH2PO4, 24 NaHCO3, 0.5 CaCl2, 4 MgCl•6H2O, and 25 d-glucose. Slices were allowed to recover at 31.5°C for 30 min then at room temperature for 1 h in oxygenated ACSF (same as culture electrophysiology). Acute slices where transferred to a recording chamber continuously superfused with oxygenated ACSF (1.5 ml/min) maintained at 29.5°C. A potassium gluconate internal was used to allow for the depolarization and spiking identification of regular versus burst spiking cells in subiculum (in mm): 140 K-gluconate, 5 KCl, 10 HEPES, 4 ATP-Mg2+, 0.5 GTP-Na+, 10 phosphocreatine, 0.2 EGTA. Series resistance was constantly monitored and cells that experienced changes of >20% were discarded. Cells that did not exhibit an AMPAR-mediated EPSC I/O relationship were excluded. AMPAR-mediated EPSCs were recorded at −70 mV and evoked using a homemade nichrome electrode controlled by an A-M Systems Isolated Pulse Stimulator in the presence of d-AP5 (50 μm) and PTX (100 μm). Paired-pulse ratios (PPRs) were measured with ISIs of 40 ms. LTP was induced in regular and burst spiking cells in current-clamp mode by stimulating 4 times at 100 Hz with 10 s intertrain intervals and recorded in voltage-clamp in the presence of PTX (100 μm). EPSCs during baseline and post-LTP induction were sampled at 0.1 Hz and binned per minute. The magnitude of LTP was determined by comparing the average EPSC amplitude of 50–60 min post-induction to the average amplitude of 10′ baseline in each experiment. PPRs were assessed before LTP induction and 60 min post-induction.
All synaptic currents were monitored with a MultiClamp 700B amplifier and digitized with a Digidata 1440A or 1550B (Molecular Devices). Data were collected at 10 kHz and filtered with a low-pass filter at 2 kHz. For all experiments, the experimenter was blind to the recording condition.
Scanning electron microscopy
Cells were fixed on the coverslips in 2.5% glutaraldehyde in 0.1 m sodium cacodylate buffer. Cells were first rinsed three times in 0.1 m sodium cacodylate buffer, pH 7.4, then postfixed in 1% osmium tetroxide and 1.5% potassium ferrocyanide for 30 min. After three washes in water, the cells were immersed in 1% uranyl acetate in water for 30 min. Following two rinses in water, they were dehydrated through a graded ethanol series (50, 70, 95, and 100% ×2) for 5 min each. The coverslips were quickly blotted and a drop of Epon/Araldite resin was applied to the coverslips before prefilled BEEM capsules were flipped over on top of them. Samples were polymerized for 24 h at 60°C in an oven. The coverslips were then removed by dipping the block in liquid nitrogen. Ultrathin sections (60 nm) were cut on a Reichert Ultracut S from a small trapezoid positioned over the cells and were picked up on Formvar-coated slot grids (EMS). Sections were imaged on a FEI Tecnai G2 transmission electron microscope with an AMT digital camera. Vesicles were counted on electron micrographs using Reconstruct software. A vesicle was considered “docked” if it made direct contact with the presynaptic membrane as described by Bacaj et al. (2015). Vesicles within 100 nm distance from the presynaptic membrane were considered to be in the “putative active zone”. Length of active zone (AZ) was measured as the length of the presynaptic membrane opposite to the putative postsynaptic density (PSD). The length of the putative PSD was determined by enhancing contrast on Reconstruct to identify boundary of PSD seen as increased cytoplasmic opacity and thickness bordering the postsynaptic membrane. Three independent cultures were analyzed by a blinded experimenter.
HEK cell aggregation study
HEK293T cells expressing either GFP alone or together with LRRTM2 or neuroligin-1 were mixed in a 1:1 ratio with cells expressing mCherry alone or together with neurexin-3αWT/A687T with or without a splice insert in SS4, and incubated at room temperature for 30–60 min in DMEM-containing the following (in mm): 50 HEPES-NaOH, pH 7.4, 10% FBS, 10 MgCl2, and 10 CaCl2. Cells were imaged by fluorescence microscopy using a wide-field Zeiss Axio Vert 200M with an Olympus 5× air objective (NA: 0.55; WD: 26 mm). Extent of aggregation was quantified using MATLAB scripts that calculated the percentage overlap of green (GFP) on red (mCherry) puncta.
Quantification and statistical analyses
Statistical analysis was performed using GraphPad Prism 7 and Microsoft Excel. Distribution of data was evaluated for normality using Shapiro–Wilkes test. Additional analyses included independent unpaired t tests, and ANOVAs. Correction for multiple comparisons was also used where appropriate. Kolmogorov–Smirnov test was used to analyze the statistical significance of cumulative probabilities. Outliers were identified based on a predefined criteria using the ROUT method of Prism, where Q = 1%. Outlier analysis was performed on immunocytochemistry experiments to identify synaptic clusters comprised of multiple overlapping puncta, which if included in the analysis would artificially skew the data toward larger puncta sizes. All bar graphs represent the mean from at least three independent experiments except where noted. Data values reported are given as mean ± SEM. All bar graphs represent the mean from at least three independent experiments except where noted.
Contact for reagent and resource sharing
Further information and requests for resources and reagents should be directed to and will be fulfilled by the corresponding author.
Results
The A687T mutation does not impair surface trafficking of Nrxn3α
Whole exome sequencing revealed a truncating mutation in the paternal allele (p1076dupC, NM_001272020.1) which is similar to a nonfunctional secreted isoform previously tested (Aoto et al., 2015), and a missense mutation in the maternal allele (pA310T, NM_004796.5) in a sequence located between EGF2 and LNS4 of Nrxn3α that is conserved between α neurexins and throughout vertebrate evolution (Fig. 1A). In mice, the orthologous mutation is A687T (Nrxn3αA687T). Intriguingly, mutations in this linker region of other neurexins, including a rare missense mutation in Nrxn1α linked to autism spectrum disorder located immediately upstream of the residue studied here (Fig. 1A, highlighted in blue; Yan et al., 2008), suggests that this unstudied sequence of residues is important for the synaptic function of neurexins.
We first asked whether the A687T mutation alters the surface availability of Nrxn3α. Although unlikely due to the location of the mutation in a linker between two structured domains, the alanine-to-threonine mutation might adversely affect trafficking of Nrxn3αA687T, which would effectively recapitulate a KO allele. Reliable antibodies that detect Nrxn3α do not exist, therefore, we designed lentiviruses that encoded for extracellular HA-tagged Nrxn3α to quantify surface Nrxn3α levels (Fig. 1B). We positioned the extracellular HA tag in the 5′ end of exon 24, the last coding exon for Nrxn3, using a strategy previously reported for Nrxn1 (Pak et al., 2015 their p. 1). The introduction of the HA tag in this locus does not alter surface trafficking of endogenous protein (Pak et al., 2015). We transduced primary hippocampal neurons at 4 DIV with lentivirus expressing HA-tagged Nrxn3αWT or Nrxn3αA687T with or without the splice-site 4 exon (SS4+ or SS4−), and then performed live anti-HA surface labeling at 14 DIV. We manipulated SS4 because the presence or absence of the alternative splice-site 4 exon regulates binding to almost all known neurexin ligands and is differentially expressed in excitatory and inhibitory synapses (Aoto et al., 2015; Fuccillo et al., 2015; Nguyen et al., 2016). We found that the mutation did not alter surface expression relative to WT for either splice-site 4 isoform (integrated fluorescence intensities, WT SS4−: 8062 ± 1353.7 n = 18; A687T SS4−: 8846 ± 1986.1 n = 14, p = 0.946; WT SS4+: 8779 ± 1239.7 n = 18; A687T SS4+: 9466 ± 1571.2 n = 18, p = 0.733; one-way ANOVA multiple comparisons; Fig. 1C). Importantly, we observed little surface HA immunoreactivity for WT and mutant Nrxn3α in dendrites, but observed significant immunofluorescence in axons and in axon varicosities opposed to postsynaptic spines, indicating that the mutation did not impair axon trafficking (Fig. 1C, arrowheads). Importantly, transduction of neurons with neurexin-3α expressing lentivirus rescued synaptic function in Nrxn3 mutant mice (Aoto et al., 2013), which strongly suggests that a sufficient fraction of exogenous Nrxn3α localizes to presynaptic AZs to govern function. We next asked whether this mutation impacts synapse morphology and/or function.
Generation and functional characterization of the molecular replacement approach
To mitigate potential confounds caused by overexpression artifacts and by the presence of endogenous Nrxn3α, we used a molecular replacement approach that consisted of lentivirus-mediated shRNA knockdown of endogenous Nrxn3α and replacement with exogenous WT or A687T Nrxn3α (Fig. 2A). Lentivirus-mediated transduction is highly efficient in primary neuron cultures and routinely results in the infection of >95% of all neurons, which is essential when studying the functional role of presynaptic molecules (Aoto et al., 2013). We tested three Nrxn3α shRNAs that targeted distinct regions of mRNA unique to Nrxn3α and found that each shRNA knocked down endogenous Nrxn3α mRNA by 70–80% (left; Nrxn3α relative to GFP control: shRNA1: 0.88 ± 0.055, n = 3, p = 0.0001; shRNA2: 0.73 ± 0.096, n = 3, p = 0.0002; shRNA3: 0.66 ± 0.080, n = 3, p = 0.0004; one-way ANOVA multiple comparisons; Fig. 2B). Of the three shRNAs tested, we chose to use shRNA1 because it displayed the greatest knockdown of endogenous Nrxn3α mRNA. Replacement with shRNA-resistant Nrxn3αWT or Nrxn3αA687T SS4+ or SS4− resulted in a ∼fourfold increase in mRNA levels (shRNA: 0.2912 ± 0.04317 n = 4; shRNA/WT SS4 + 4.643 ± 0.2135 n = 4; shRNA/A687T SS4 + 3.265 ± 0.8536, n = 4; shRNA/WT SS4− 3.668 ± 0.8963, n = 4; shRNA/A687T SS4− 3.337 ± 1.066, n = 4; shRNA vs shRNA/WT SS4−: p = 0.0035; shRNA vs shRNA/A687T SS4−: p = 0.0084; shRNA vs shRNA/WT SS4+: p = 0.0003; shRNA vs shRNA/A687T SS4+: p = 0.0102; one-way ANOVA multiple comparisons; Fig. 2B). Importantly, none of the Nrxn3α shRNAs tested altered mRNA levels of Nrxn3β (GFP control: 1 ± 0 fold-change, n = 4; shRNA: 1.179 ± 0.1585 fold-change, n = 4; shRNA/WT SS4+: 1.183 ± 0.2973 fold-change, n = 4; shRNA/A687T SS4+: 1.15 ± 0.1863 fold-change, n = 4; shRNA/WT SS4−: 1.163 ± 0.2299 fold-change, n = 4; shRNA/A687T SS4−: 1.276 ± 0.1551 fold-change, n = 4, p = 0.9522; one-way ANOVA multiple comparisons; Fig. 2B, right).
In contrast to the Nrxn3α/β cKO, it has been previously demonstrated in primary hippocampal cultures that constitutive ablation of Nrxn3α does not alter excitatory or inhibitory synapse function (Aoto et al., 2015). To test for off-target functional effects produced by shRNA, we virally transduced primary hippocampal neurons with shRNA1 or control lentivirus and systematically measured inhibitory and excitatory synaptic transmission parameters using whole-cell voltage-clamp electrophysiology. Membrane capacitance (Cm) and membrane resistance (Rm), monitored during IPSC recordings were unchanged following knockdown of Nrxn3α (Cm: GFP control: 115.4 ± 6.869 pF, n = 29; shRNA: 130.6 ± 8.547 pF, n = 28; p = 0.1661; Rm: GFP control: 235 ± 18.32 MΩ, n = 29; shRNA: 258.1 ± 19.31 MΩ, n = 28; p = 0.3926; Student's t test; Fig. 2C) or EPSC recordings (Cm: GFP control: 157.9 ± 5.674 pF, n = 31; shRNA: 144.8 ± 6.21 pF, n = 31; p = 0.1283; Rm: GFP control: 160.3 ± 7.422 MΩ, n = 31; shRNA: 171.1 ± 8.869 MΩ, n = 31; p = 0.3518; Student's t test; Fig. 2G). Additionally, the frequency and amplitude of miniature IPSCs [(mIPSCs) frequency, GFP control: 1.593 ± 0.1998 Hz, n = 29; shRNA: 1.181 ± 0.1783 Hz; p = 0.1307; amplitude, GFP control: 14.94 ± 0.6115 pA, n = 29; shRNA: 14.16 ± 0.6757 pA, n = 28; p = 0.3929] and miniature EPSCs [(mEPSCs) frequency, GFP control: 0.793 ± 0.1948 Hz, n = 41; shRNA: 0.6178 ± 0.21 Hz, n = 43; p = 0.5736; amplitude, GFP control: 15.39 ± 1.136 pA, n = 41; shRNA 14.26 ± 0.9563 pA, n = 43; p = 0.4751; Student's t test) of shRNA1 infected neurons were similar to the mock infected control condition (Fig. 2D,H), which is consistent with the Nrxn3α KO mouse (Aoto et al., 2015).
Recent studies suggest that spontaneous and action potential evoked release may represent mechanistically distinct processes (Ramirez and Kavalali, 2011; Crawford et al., 2017; Chanaday and Kavalali, 2018). Thus, we tested whether the acute knockdown of Nrxn3α alters action-potential evoked inhibitory or excitatory synaptic transmission. We observed no overt differences in electrically evoked IPSC amplitudes (GFP control: 1.774 ± 0.07477 nA, n = 30; shRNA: 1.299 ± 0.2054 nA, n = 24; p = 0.0955; Student's t test) or IPSC PPRs (40 ms ISI: GFP control: 0.5507 ± 0.02025, n = 30; shRNA: 0.6012 ± 0.04813, n = 24; p = 0.3886; Student's t test) between control and shRNA1 conditions (Fig. 2E,F). The only PPR ISI that exhibited a significant difference was at 50 ms (GFP control: 0.5505 ± 0.02213, n = 30; shRNA: 0.6314 ± 0.0126, n = 24; p = 0.0337 Student's t test), however, we failed to detect a similar change measuring PPRs at shorter ISIs. Additionally, reduced IPSC PPR at a 50 ms ISI, indicative of increased release probability, is incongruent with the reported mIPSC frequency and the slightly reduced evoked IPSC amplitudes observed for shRNA1. Similarly, AMPAR-mediated (GFP control: 0.7181 ± 0.05714 nA, n = 50; shRNA: 0.6066 ± 0.06402 nA, n = 32; p = 0.2635; Student's t test) and NMDAR-mediated evoked EPSC amplitudes (GFP control: 0.5026 ± 0.06404 nA, n = 17; shRNA: 0.5081 ± 0.09263 nA, n = 15; p = 0.9602; Student's t test) and NMDAR PPRs (40 ms ISI: GFP control: 0.572 ± 0.05819, n = 17; shRNA: 0.6344 ± 0.0931, n = 15; p = 0.5691; Student's t test) were unaltered by shRNA1 (Fig. 2I–K). Together, these data suggest that knockdown of Nrxn3α by shRNA1 does not alter spontaneous or action potential-evoked inhibitory or excitatory neurotransmission. Thus, for all molecular replacement experiments in vitro and in vivo, we used shRNA1 alone as an infection control.
Molecular replacement with Nrxn3αA687T SS4− selectively increases presynaptic vGluT1 cluster size in vitro
To examine whether the A687T mutation alters synaptic morphology, we fixed and stained 14 DIV primary hippocampal cultures following lentivirus-mediated molecular replacement with both SS4 isoforms of Nrxn3αWT or Nrxn3αA687T at 4 DIV. To investigate excitatory and inhibitory postsynaptic morphology, we immunostained primary hippocampal neurons with antibodies to visualize gephyrin and PSD-95, which are postsynaptic markers for inhibitory and excitatory synapses, respectively (Fig. 3A). We found that the density, intensity, and size of the gephyrin and PSD95-immunoreactive clusters were unaltered by replacement with either SS4 isoform of Nrxn3αA687T (gephyrin density, shRNA: 6.7 ± 1.3, n = 64/3; shRNA/WT SS4−: 6.2 ± 1.5, n = 63/3; shRNA/A687T SS4−: 6.2 ± 1.4, n = 67/3; shRNA/WT SS4+: 6.2 ± 1.0, n = 67/3; shRNA/ A687T SS4+: 5.6 ± 0.5, n = 57/3; intensity, shRNA: 42.3 ± 5.3 AU; shRNA/WT SS4−: 40.9 ± 5.0 AU; shRNA/ A687T SS4−: 43.1 ± 6.1 AU; shRNA/WT SS4+: 44.2 ± 6.5; shRNA/A687T SS4+: 45.6 ± 6.3; size, shRNA: 0.78 ± 0.06 μm2; shRNA/WT SS4−: 0.70 ± 0.06 μm2; shRNA/A687T SS4−: 0.8 ± 0.13 μm2; shRNA/WT SS4+: 0.80 ± 0.07 μm2; shRNA/A687T SS4+: 0.78 ± 0.12 μm2; p = 0.99; one-way ANOVA, multiple comparisons; Fig. 3B,C). Next, we quantified presynaptic morphology using antibodies to detect the inhibitory marker vGAT and the excitatory marker vGluT1, following molecular replacement with Nrxn3αWT or Nrxn3αA687T (Fig. 3D). Neither SS4 isoform of Nrxn3αA687T altered vGAT cluster density, quantified as the number of clusters per unit length of dendrite, intensity or size (density, shRNA: 3.1 ± 0.18, n = 77/3; shRNA/WT SS4−: 2.75 ± 0.42, n = 72/3; shRNA/A687T SS4−: 3.7 ± 0.02, n = 81/3; shRNA/WT SS4+: 2.8 ± 0.19, n = 61/3; shRNA/A687T SS4+: 2.8 ± 0.35, n = 59/3; intensity, shRNA: 43.3 ± 9.1 AU; shRNA/WT SS4−: 44.6 ± 9.8 AU; shRNA/A687T SS4−: 42.5 ± 7.6 AU; shRNA/WT SS4+: 41.4 ± 8.2 AU; shRNA/A687T SS4+: 41.7 ± 7.0 AU; size, shRNA: 0.63 ± 0.033 μm2; shRNA/WT SS4−: 0.63 ± 0.030 μm2; shRNA/A687T SS4−: 0.72 ± 0.16 μm2; shRNA/WT SS4+: 0.63 ± 0.040 μm2; shRNA/A687T SS4+: 0.53 ± 0.033 μm2; p > 0.05; one-way ANOVA, multiple comparisons; Fig. 3E). However, we observed a striking ∼50% increase in the size of vGluT1 clusters after molecular replacement with Nrxn3αA687T SS4− without alterations in vGluT1 cluster density and intensity relative to controls (density, shRNA: 4.7 ± 0.24, n = 77/3; shRNA/WT SS4−: 4.54 ± 0.50, n = 72/3; shRNA/A687T SS4−: 4.98 ± 0.23, n-81/3; shRNA/WT SS4+: 5.23 ± 0.46, n = 61/3; shRNA/A687T SS4+: 4.48 ± 0.20, n = 59/3; p > 0.05; intensity, shRNA: 52.3 ± 11.5 AU; shRNA/WT SS4−: 53.1 ± 13.8 AU; shRNA/A687T SS4−: 49.0 ± 11.8 AU; shRNA/WT SS4+: 52.0 ± 13.9 AU; shRNA/A687T SS4+: 50.3 ± 12.7 AU; p > 0.05; size, shRNA: 0.84 ± 0.090 μm2; shRNA/WT SS4−: 0.90 ± 0.040 μm2; shRNA/A687T SS4−: 1.29 ± 0.066 μm2; shRNA/WT SS4+: 0.81 ± 0.072 μm2; shRNA/A687T SS4+: 0.89 ± 0.069 μm2; shRNA vs shRNA/A687T SS4−: p = 0.0085; shRNA/WT SS4− vs shRNA/A687T SS4−: p = 0.02; shRNA/WT SS4+ vs shRNA/A687T SS4−: p = 0.005; shRNA/A687T SS4+ vs shRNA/A687T SS4−: p = 0.018; one-way ANOVA, multiple comparisons; Fig. 3F). This intriguing increase in vGluT1 cluster size suggests that in excitatory boutons there are more vesicles of neurotransmitter and/or more vGluT1 molecules per vesicle. By contrast, the inclusion of the SS4 exon (SS4+) did not alter the density, intensity or size of vGluT1-immunoreactive clusters (Fig. 3F). Unexpectedly, these data suggest that the A687T mutation does not impact inhibitory synapse morphology but exhibits a selective splice-site 4-dependent increase in the abundance of vGluT1 in excitatory presynaptic terminals, a phenotype distinct from those previously observed in the Nrxn3α/β cKO. More broadly, our data surprisingly suggest that the linker sequence between EGF2 and LNS4 participates in organizing synapse morphology.
Although knockdown of Nrxn3α did not elicit a functional synaptic phenotype (Fig. 2), we wanted to address whether the striking Nrxn3αA687T SS4− dependent increase in vGluT1 puncta size (Fig. 3H) could be simply explained as the ability of Nrxn3αA687T SS4−, but not Nrxn3αWT SS4−, to rescue a morphological phenotype induced by shRNA1. To address this potential explanation, we infected neurons with control shRNA or shRNA1 and quantified immunoreactive vGAT and vGluT1 cluster parameters. We found no differences in the density, intensity or size of vGAT or vGluT1 clusters (vGAT, density: Control: 3.26 ± 0.22; n = 44/3; shRNA: 3.05 ± 0.20; n = 46/3; p = 0.526; intensity, Control: 51.1 ± 9.96 AU; shRNA: 57.0 ± 6.73 AU; p = 0.646; size, Control: 0.710 ± 0.009 μm2; shRNA: 0.680 ± 0.043 μm2; p = 0.649; vGluT1: density, Control: 4.16 ± 0.074; n = 44/3; shRNA: 4.16 ± 0.23; n = 46/3; p = 0.990; intensity, Control: 55.5 ± 10.4 AU; shRNA: 55.8 ± 8.87 AU; p = 0.987; size, Control: 0.994 ± 0.089 μm2; shRNA: 0.966 ± 0.095 μm2; p = 0.839; Student's t test; Fig. 4A–C). This is consistent with the notion that acute knockdown of endogenous Nrxn3α alone does not alter synaptic properties (Fig. 2C–K) and reinforces the idea that the SS4− isoform of Nrxn3αA687T selectively acts to robustly enhance vGluT1 cluster size.
Molecular replacement with Nrxn3αA687T SS4− selectively enhances excitatory presynaptic function in vitro
Given the surprising morphological phenotype, we next asked whether this mutation alters basal inhibitory or excitatory synaptic transmission in primary hippocampal cultures. To study inhibitory transmission, we focused on the Nrxn3αA687T SS4+ isoform because this isoform is highly expressed in subpopulations of GABAergic interneurons and displays the highest binding affinity to neuroligin-2, a key inhibitory postsynaptic neurexin ligand (Koehnke et al., 2010; Nguyen et al., 2016). We coinfected neurons with shRNA and Nrxn3αA687T SS4+ expressing lentiviruses at 4 DIV and performed whole-cell electrophysiology at 13–14 DIV. Molecular replacement with WT or A687T Nrxn3α did not alter passive membrane properties as membrane capacitance and resistance were unchanged (Cm: shRNA: 145.8 ± 7.8369 pF, n = 45; shRNA/WT SS4+: 153.1 ± 7.598 pF, n = 59; shRNA/A687T SS4+: 134.8 ± 9.627 pF, n = 49, p = 0.3358; Rm: shRNA: 362.9 ± 52.56 MΩ, n = 45; shRNA/WT SS4+: 346.1 ± 25.11 MΩ, n = 59; shRNA/A687T SS4+: 381.8 ± 35.66 MΩ, n = 49, p = 0.8286; one-way ANOVA, multiple comparisons; Fig. 5A). Additionally, Nrxn3αA687T SS4+ did not induce changes in the frequency or amplitude of spontaneous mIPSCs (frequency, shRNA: 1.817 ± 0.3052 Hz, n = 45; shRNA/WT SS4+: 1.657 ± 0.3976 Hz, n = 59; shRNA/A687T SS4+: 1.648 ± 0.202 Hz, n = 49, p = 0.9124; amplitude, shRNA: 15.49 ± 0.7477 pA, n = 45; shRNA/WT SS4+: 16.45 ± 1.737 pA, n = 59; shRNA/A687T SS4+: 16.53 ± 1.5131 pA, n = 49, p = 0.8465; one-way ANOVA, multiple comparisons' Fig. 5B). We next assessed action-potential evoked IPSCs and found that the A687T mutation did not alter peak evoked IPSC amplitudes (shRNA: 1.806 ± 0.3682 nA, n = 29; shRNA/WT SS4+: 1.493 ± 0.06414 nA, n = 26; shRNA/A687T SS4+: 1.641 ± 0.1688 nA, n = 23, p = 0.6580; one-way ANOVA, multiple comparisons; Fig. 5C). IPSC PPRs, an indirect measurement of presynaptic release, were also unchanged (40 ms ISI: shRNA: 0.5519 ± 0.09024, n = 29; shRNA/WT SS4+: 0.4872 ± 0.009555, n = 26; shRNA/A687T SS4+: 0.5103 ± 0.08722, n = 23, p = 0.8535; one-way ANOVA, multiple comparisons; Fig. 5D). Our electrophysiological analyses of inhibitory synaptic transmission are consistent with our morphological analyses and argue that the A687T mutation does not alter inhibitory synapse function in primary hippocampal neurons.
We next measured excitatory synaptic transmission following molecular replacement with Nrxn3αA687T SS4−. We focused on the Nrxn3αA687T SS4− isoform because Nrxn3αA687T SS4− but not Nrxn3αA687T SS4+ selectively increased the size of presynaptic vGluT1 cluster size. In addition, Nrxn3 SS4− is highly expressed in excitatory neurons in hippocampus and plays a dominant role at excitatory hippocampal synapses to engage in functionally instructive transsynaptic interactions (Aoto et al., 2013, 2015; Fuccillo et al., 2015; Nguyen et al., 2016). We found that molecular replacement with Nrxn3αA687T SS4− did not alter membrane capacitance or resistance (Fig. 5E). We first monitored spontaneous mEPSCs and observed a Nrxn3αA687T SS4− dependent increase in mEPSC frequency (shRNA: 1.08 ± 0.11 Hz, n = 23/3; shRNA/WT SS4−: 1.01 ± 0.04 Hz, n = 28/3; shRNA/A687T SS4−: 1.92 ± 0.28 Hz, n = 31/3; shRNA vs shRNA/MUT SS4−: p = 0.033; shRNA/WT SS4− vs shRNA/A687T SS4−: p = 0.236; one-way ANOVA multiple comparisons) but not amplitude (shRNA: 11.7 ± 0.93 pA, n = 23/3; shRNA/WT SS4−: 10.8 ± 0.48 pA, n = 28/3; shRNA/A687T SS4−: 11.2 ± 0.85 pA, n = 31/3; shRNA vs shRNA/A687T SS4−: p = 0.90; shRNA/WT SS4− vs shRNA/A687T SS4−: p = 0.91; one-way ANOVA, multiple comparisons; Fig. 5F). mEPSC frequency is commonly thought to reflect changes in synapse numbers and/or changes in presynaptic release probability while mEPSC amplitudes typically correlate to changes in postsynaptic strength. Combined with our findings that molecular replacement with Nrxn3αA687T SS4− does not alter excitatory synapse numbers, but rather vGluT1 cluster size (Fig. 3H), our functional data suggest that presynaptic properties are enhanced by molecular replacement with mutant Nrxn3α. To further test the consequence of Nrxn3αA687T SS4− on synaptic transmission, we assessed electrically evoked AMPAR- and NMDAR-mediated excitatory postsynaptic synaptic currents. Consistent with the robust increase in mEPSC frequency (Fig. 5F), molecular replacement with Nrxn3αA687T SS4− induced a ∼50% increase in the amplitude of AMPAR- and NMDAR-mediated EPSCs (AMPAR, shRNA: 0.714 ± 0.089 nA, n = 36/3; shRNA/WT SS4−: 0.672 ± 0.101 nA, n = 38/3; shRNA/A687T SS4−: 1.464 ± 0.058 nA, n = 33/3; shRNA vs shRNA/A687T SS4−: p = 0.002; shRNA/WT SS4− vs shRNA/A687T SS4−: p = 0.001; NMDAR, shRNA: 0.488 ± 0.076 nA, n = 38/3; shRNA/WT SS4−: 0.411 ± 0.109 nA, n = 42/3; shRNA/A687T SS4−: 0.909 ± 0.127 pA, n = 39/3; shRNA vs shRNA/A687T SS4−: p = 0.047; shRNA/WT SS4− vs shRNA/A687T SS4−: p = 0.0375; one-way ANOVA, multiple comparisons; Fig. 5G,H). To quantify presynaptic release, we monitored NMDAR-mediated evoked EPSC PPRs at increasing ISIs. PPRs provide sensitive, but indirect measurements of presynaptic release. We found that replacement with the splice-site 4 lacking isoform of Nrxn3αA687T significantly reduced the NMDAR PPR, which is consistent with an increase in presynaptic release probability (40 ms ISI, shRNA: 0.721 ± 0.055, n = 24; shRNA/WT SS4+: 0.760 ± 0.062, n = 23; shRNA/A687T SS4+: 0.573 ± 0.0618, n = 28; shRNA vs shRNA/A687T SS4−: p = 0.284; shRNA/WT SS4− vs shRNA/A687T SS4−: p = 0.015; one-way ANOVA, multiple comparisons; Fig. 5I).
Given the striking SS4− dependent effect on presynaptic vGluT1 cluster size and the unexpected increase in mEPSC frequency, AMPAR- and NMDAR-mediated evoked EPSC amplitudes, and presynaptic release probability, we next asked whether Nrxn3αA687T SS4+ could induce an identical synaptic phenotype. Similar to molecular replacement with Nrxn3αA687T SS4−, Nrxn3αA687T SS4+ did not alter membrane capacitance or resistance (Cm, shRNA: 164.2 ± 8.96 pF, n = 19; shRNA/WT SS4+: 161.0 ± 8.9 pF, n = 22; shRNA/A687T SS4+: 166.1 ± 6.47 pF, n = 23; shRNA vs shRNA/A687T SS4+: p = 0.987; shRNA/WT SS4+ vs shRNA/A687T SS4+: p = 0.900; Rm, shRNA: 180.7 ± 12.9 MΩ, n = 17; shRNA/WT SS4+: 191.7 ± 10.4 MΩ, n = 20; shRNA/A687T SS4+: 193.8 ± 12.2 MΩ, n = 17; shRNA vs shRNA/A687T SS4+: p = 0.723; shRNA/WT SS4+ vs shRNA/A687T SS4+: p = 0.991; one-way ANOVA, multiple comparisons; Fig. 6A). However, unlike molecular replacement with Nrxn3αA687T SS4−, Nrxn3αA687T SS4+ did not impact mEPSC frequency (shRNA: 2.06 ± 0.73 Hz, n = 19/3; shRNA/WT SS4+: 2.44 ± 0.63 Hz, n = 22/3; shRNA/A687T SS4+: 1.99 ± 0.38 Hz, n = 23/3; shRNA vs shRNA/A687T SS4+: p = 0.996; shRNA/WT SS4+ vs shRNA/A687T SS4+: p = 0.863; one-way ANOVA, multiple comparisons) or amplitude (shRNA: 15.3 ± 0.35 pA, n = 19/3; shRNA/WT SS4+: 15.2 ± 1.78 pA, n = 22/3; shRNA/A687T SS4+: 15.0 ± 1.72 pA, n = 23/3; shRNA vs shRNA/A687T SS4+: p = 0.984; shRNA/WT SS4+ vs shRNA/A687T SS4+: p = 0.999; one-way ANOVA, multiple comparisons), evoked AMPAR-mediated (shRNA: 0.4955 ± 0.0607 nA, n = 32, p = 0.7994; shRNA/WT SS4+: 0.5394 ± 0.08129 nA, n = 29; shRNA/A687T SS4+: 0.4519 ± 0.1204 nA, n = 30), NMDAR-mediated EPSC amplitudes (shRNA: 0.5346 ± 0.0586 nA, n = 21, p = 0.7497; shRNA/WT SS4+: 0.4659 ± 0.3058 nA, n = 26; shRNA/A687T SS4+: 0.5204 ± 0.06491 nA, n = 26), or NMDAR PPRs (40 ms ISI, shRNA: 0.7205 ± 0.03718, n = 21;shRNA/WT SS4+: 0.707 ± 0.04925, n = 26; shRNA/A687T SS4+: 0.7501 ± 0.08145, n = 26, p = 0.8733; one-way ANOVA, multiple comparisons; Fig. 6B–E). Together, these data suggest that the A687T mutation induces a SS4-dependent gain-of-function phenotype by selectively increasing presynaptic efficacy at excitatory synapses, which provides crucial insight into this extracellular sequence.
Nrxn3αA687T SS4− increases readily releasable pool size at excitatory synapses and the number of docked vesicles
Presynaptically, action potential-evoked excitatory transmission is a function of the likelihood that a single vesicle of neurotransmitter will be exocytosed following an action potential and the number of vesicles ready to release (Kaeser and Regehr, 2017). Therefore, in addition to the decreased PPRs following molecular replacement with Nrxn3αA687T SS4−, changes in the size of the readily releasable pool (RRP) might contribute to the excitatory transmission phenotype (Fig. 5F) and electrically evoked AMPAR- and NMDAR-mediated EPSC amplitudes (Figs. 5G,H). We analyzed the size of the RRP using two distinct approaches. We functionally measured the size of the RRP using 0.5 m hypertonic sucrose and performed electron microscopy (EM) on cultured neurons to quantify the number of docked vesicles.
First, we monitored IPSCs and AMPAR-mediated EPSCs evoked by the acute (30 s) application of 0.5 m hypertonic sucrose (Kaeser and Regehr, 2017). We found that the size of the inhibitory RRP was not altered by Nrxn3αA687T relative to control and Nrxn3αWT SS4− conditions (shRNA: 12.863 ± 3.212 nC, n = 29; shRNA/WT SS4+: 11.573 ± 3.322 nC, n = 27; shRNA/A687T SS4+: 12.895 ± 5.188 nC, n = 24, p = 0.9654), which is consistent with the findings that Nrxn3αA687T does not alter inhibitory morphology or synaptic function (Fig. 7A). By contrast, molecular replacement with Nrxn3αA687T SS4− increased EPSC charge transfer, which is indicative of an increase in the size of the RRP (shRNA: 4.51 ± 0.29 nC, n = 24/3; shRNA/WT SS4−: 4.31 ± 0.45 nC, n = 29/3; shRNA/A687T SS4−: 7.25 ± 0.62 nC, n = 38/3; shRNA vs shRNA/A687T SS4−: p = 0.0148; shRNA/WT SS4− vs shRNA/A687T SS4−: p = 0.0107; one-way ANOVA multiple comparisons; Fig. 7B). Together, the surprising increase in presynaptic efficacy induced by Nrxn3αA687T SS4− is a consequence of an increase in both presynaptic release probability and size of the functional RRP.
Second, we quantified the impact of Nrxn3αA687T on the ultrastructure of hippocampal synapses in culture using EM. Similar to our electrophysiological analyses, we coinfected primary hippocampal cultures at 4 DIV with lentiviruses to perform molecular replacement with SS4 isoforms of Nrxn3α and processed the neurons at 14 DIV for EM. We measured AZ length, the number of docked vesicles ≤100 nm from the AZ and the number of vesicles per AZ area from all analyzable synapses. In primary culture, the majority of neurons are excitatory (Benson et al., 1994) and excitatory synapses represent 65–75% of all dendritic synapses (Benson and Cohen, 1996), thus, the majority of synapses analyzed likely represent excitatory synapses. Consistent with this and in agreement with our electrophysiological assessment, we found that although the length of the AZ did not change, molecular replacement with Nrxn3αA687T SS4− significantly increased the number of docked vesicles per AZ length and the number of vesicles per AZ area (putative AZ length, shRNA: 0.2667 ± 0.01022 μm, n = 81; shRNA/WT SS4−: 0.2886 ± 0.01148 μm, n = 69; shRNA/A687T SS4−: 0.2492 ± 0.01245 μm, n = 56, p = 0.0633; docked vesicles/100 nm of putative AZ, shRNA: 1.466 ± 0.07305, n = 81; shRNA/WT SS4−: 1.522 ± 0.08064, n = 70; shRNA/A687T SS4−: 1.869 ± 0.094, n = 59,; shRNA vs shRNA/A687T SS4−: p = 0.0019; shRNA/WT SS4− vs shRNA/A687T SS4−: p = 0.0123; vesicles per putative AZ area, shRNA: 0.4024 ± 0.0139, n = 82; shRNA/WT SS4−: 0.3941 ± 0.01833, n = 70; shRNA/A687T SS4−: 0.4559 ± 0.01821, n = 59; shRNA vs shRNA/A687T SS4−: p = 0.0481; shRNA/WT SS4− vs shRNA/A687T SS4−: p = 0.0234; one-way ANOVA multiple comparisons; Fig. 7C). Our ultrastructural examination of excitatory and inhibitory synapses validates our electrophysiological analyses, indicating that one role for the sequences between EGF2 and LNS4 is to regulate RRP size at excitatory synapses. We next asked whether the Nrxn3αA687T SS4− gain-of-function presynaptic phenotype identified in developing primary hippocampal neurons is also present in more mature neurons within an intact circuit.
In vivo presynaptic molecular replacement with Nrxn3αA687T increases basal excitatory presynaptic properties and impairs presynaptic LTP in ex vivo slices
We next performed molecular replacement in vivo and measured synaptic transmission in ex vivo acute hippocampal slices at CA1-subiculum synapses. We stereotactically coinjected P21 mice in CA1 with an AAV encoding Nrxn3α shRNA and mRuby, and an AAV encoding WT or A687T Nrxn3α SS4− (Fig. 8A). Three to 4 weeks after injection, we prepared 300 μm horizontal slices and measured excitatory synaptic transmission at CA1–subiculum synapses (Aoto et al., 2013). We focused on this circuit because we can largely restrict infection to presynaptic CA1 neurons and monitor excitatory synaptic strength from uninfected postsynaptic subicular principle neurons to unambiguously study the presynaptic contribution of Nrxn3α (Fig. 8B,C; Aoto et al., 2013, 2015; Anderson et al., 2015; Dai et al., 2019). To stimulate CA1 afferents, we placed an extracellular stimulating electrode in the stratum oriens/alveus and recorded from two major types of principle neurons in subiculum, identified by action-potential spiking patterns as regular-spiking (RS) or burst-spiking (BS; Fig. 8C). Presynaptic Nrxn3αA687T SS4− dramatically enhanced the AMPAR input/output relationship (I/O slope RS, shRNA 1.692 ± 0.2417, n = 28; shRNA/WT SS4− 1.444 ± 0.2257, n = 18, p = 0.0059; shRNA/ A687T SS4− 3.288 ± 0.5766, n = 19; shRNA vs shRNA/A687T SS4−: p = 0.0059; shRNA/WT SS4− vs shRNA/A687T SS4−: p = 0.0039, Fig. 8D; I/O slope BS, shRNA 1.515 ± 0.2386, n = 35; shRNA/WT SS4− 1.525 ± 0.3146, n = 21; shRNA/A687T SS4− 3.27 ± 0.4437, n = 21; shRNA vs shRNA/A687T SS4−: p = 0.0005; shRNA/WT SS4− vs shRNA/A687T SS4−: p = 0.0022; one-way ANOVA, multiple comparisons; Fig. 8D,E). Additionally we observed decreased AMPAR-mediated PPRs at CA1-RS and CA1-BS synapses (R2/R1 RS, shRNA 1.748 ± 0.07283; n = 36; shRNA/WT -SS4 1.834 ± 0.0904, n = 30; shRNA/A687T -SS4, n = 39; shRNA ns shRNA/A687T SS4−: p = 0.0458; shRNA/WT SS4− vs shRNA/A687T SS4−: p = 0.0081; Fig. 8F; R2/R1 BS, shRNA 1.81 ± 0.06886, n = 36; shRNA/WT SS4− 1.973 ± 0.1092, n = 30; shRNA/A687T SS4− 1.525 ± 0.06085, n = 26; shRNA vs shRNA/A687T SS4−: p = 0.0466; shRNA/WT SS4− vs shRNA/A687T SS4−: p = 0.0015; one-way ANOVA, multiple comparisons; Fig. 8F,G). These results from acute hippocampal slices are congruent with the presynaptic phenotype observed in primary neurons and support the idea that the A687T mutation in Nrxn3α SS4− causes a selective increase in basal presynaptic efficacy at excitatory synapses. We next asked whether presynaptic Nrxn3αA687T SS4− alters activity-dependent synaptic plasticity.
The CA1-subiculum circuit is advantageous to study the function of presynaptic molecules for two major reasons. First, as discussed above, it is easy to selectively target presynaptic CA1 neurons with minimal spillover to subiculum to specifically study presynaptic function. Second, CA1 afferents onto regular and burst spiking neurons exhibit distinct forms of long-term potentiation (LTP). Although LTP at CA1–RS and CA1–BS cell synapses are both dependent on NMDAR activity, LTP is expressed postsynaptically at CA1-RS synapses but LTP manifests presynaptically as an increase in release probability at BS synapses (Wozny et al., 2008). To induce LTP at CA1-subiculum synapses, we administered a 4 × 100 Hz electrical stimulus to CA1 afferents for 1 s with 10 s intervals while in whole-cell current-clamp mode at resting membrane potential. Synaptic responses pre- and post-LTP induction were acquired while holding the postsynaptic cell at −70 mV in whole-cell voltage-clamp mode. Molecular replacement with presynaptic Nrxn3αA687T SS4− did not alter the magnitude of RS cell LTP, as all conditions exhibited a ∼150–160% potentiation (shRNA: mean = 1.628 ± 0.04216, p = 0.9380; shRNA/WT SS4− 1.605 ± 0.05447, p = 0.7723; shRNA/A687T SS4− 1.675 ± 0.04666, p = 0.583; Fig. 9A). In stark contrast, presynaptic molecular replacement with Nrxn3αA687T SS4−, but not Nrxn3αWT SS4−, resulted in impaired LTP at CA1–BS cell synapses (shRNA 1.793 ± 0.02224, n = 11; shRNA/WT SS4− 1.802 ± 0.02901, n = 11; shRNA/A687T SS4− 0.995 ± 0.02366, n = 11, shRNA vs shRNA/A687T SS4−: p ≤ 0.0001; shRNA/WT SS4− vs shRNA/A687T SS4−: p ≤ 0.0001; one-way ANOVA, multiple comparisons; Fig. 9B). The impairment of LTP specifically at CA1-burst spiking cell synapses was surprising, yet our functional and morphological characterization in primary culture and hippocampal slices strongly support the notion that presynaptic Nrxn3αA687T SS4− enhances excitatory presynaptic efficacy. Thus, we hypothesized that the Nrxn3αA687T SS4− mediated increase in basal presynaptic efficacy occluded presynaptic LTP at CA1–BS synapses.
To test whether the increased presynaptic release probability measured in Nrxn3αA687T SS4− infected slices occluded presynaptic LTP at CA1–BS cell synapses, we measured PPRs 10 min before and 60 min after LTP induction for control (shRNA alone) and molecular replacement (Nrxn3αWT SS4− and Nrxn3αA687T SS4−) conditions. Consistent with a postsynaptic locus of LTP at CA1-RS synapses that was not affected by Nrxn3αA687T SS4−, LTP induction at these synapses did not increase presynaptic release probabilities (shRNA, pre: 1.90 ± 0.273, post: 1.77 ± 0.218, n = 9/6, p = 0.379; shRNA/WT SS4−, pre: 1.72 ± 0.105, post: 1.82 ± 0.274, n = 11/9, p = 0.682; shRNA/A687T SS4−, pre: 1.674 ± 0.122, post: 2.04 ± 0.150, n = 14/10, p = 0.0289; one-way ANOVA, multiple comparisons; Fig. 9C). Interestingly, we observed a small but statistically significant increase in the PPR for the Nrxn3αA687T SS4− condition, which suggests that even though presynaptic release probability was decreased, we still observed robust LTP at CA1–RS synapses. This further supports the notion that LTP is expressed postsynaptically at these synapses. In contrast, at CA1-BS synapses, presynaptic release probability after LTP induction was significantly increased in control and Nrxn3αWT SS4− conditions, however, release probability was not further enhanced when Nrxn3αA687T SS4− was expressed in presynaptic CA1 neurons. (shRNA, pre: 2.183 ± 0.1317, n = 5, post: 1.293 ± 0.2964, n = 3, p = 0.0188; shRNA/WT SS4− pre: 1.852 ± 0.1424, n = 6, post: 1.346 ± 0.08974, n = 5, p = 0.0189; shRNA/A687T SS4−, pre: 1.697 ± 0.0996, post: 1.736 ± 0.09715, n = 6, p = 0.7965; one-way ANOVA, multiple comparisons; Fig. 9D). Together, these data indicate that the basal Nrxn3αA687T SS4− induced enhancement of presynaptic release probability at CA1-BS cell synapses occludes additional potentiation following LTP induction further suggesting that this linker region in Neurexin-3α between LNS4 and EGF2 is important for regulating presynaptic efficacy and the manifestation of LTP at these synapses.
Nrxn3αA687T selectively enhances aggregation with the excitatory ligand, LRRTM2
The presynaptic gain-of-function phenotype is restricted to the splice-site 4 lacking Nrxn3αA687T isoform, suggesting that SS4-dependent transsynaptic interactions are required. We posited that the A687T extracellular mutation might alter binding to ligands which exclusively localize to excitatory synapses and preferentially bind to SS4-lacking neurexins, such as neuroligin-1 (NL1) and leucine rich repeat transmembrane protein 2 LRRTM2 (Südhof, 2008; Ko et al., 2009; de Wit et al., 2009). NL1 and LRRTM2 have redundant roles in excitatory synapse formation; however, NL1 and LRRTM2 also have nonredundant roles at excitatory synapses. Whereas NL1 controls NMDAR synaptic strength, LRRTM2 controls AMPAR synaptic strength (Soler-Llavina et al., 2011; Budreck et al., 2013; Wu et al., 2019). Intriguingly, similar to the phenotype observed following molecular replacement with Nrxn3αA687T SS4−, overexpression of postsynaptic LRRTM2 robustly increases presynaptic vGluT1 cluster size (de Wit et al., 2009). Moreover, Nrxn3 SS4− has previously been shown to be a functionally important interaction partner with LRRTM2 at excitatory synapses in hippocampus (Aoto et al., 2013). Therefore, interactions of α-Nrxns with NL1 and LRRTM2 can encode for distinct aspects of synapse function.
To test whether A687T alters interactions with NL1–3 and LRRTM2 in trans, we used a HEK cell aggregation assay. In one population of HEK293T cells, NL1, NL2, NL3, or LRRTM2 were cotransfected with GFP while in a second population of HEK293T cells, SS4 isoforms of WT or A687T Nrxn3α were cotransfected with mCherry. These two populations of cells were incubated in suspension before aggregation was quantified by the colocalization of GFP with mCherry (Fig. 10A). Cell aggregation in conditions where GFP or mCherry were expressed alone was not observed (Fig. 10). The A687T mutation did not alter aggregation with any neuroligin compared with WT (p = 0.1452; Fig. 10B,C, F–I). Unexpectedly, relative to Nrxn3αWT conditions, we observed that Nrxn3αA687T SS4− induced a striking increase in aggregation with LRRTM2 (WT SS4−: 28 ± 5.4%, n = 3; A687T SS4−: 45 ± 1.7%, n = 3, p = 0.0136; one-way ANOVA, multiple comparisons; Fig. 10E). Consistent with the idea that LRRTM2 only binds to SS4-lacking neurexins, we failed to observe aggregation when LRRTM2 was co-incubated with either WT or A687T Nrxn3α SS4+ (Fig. 10D). Highlighting the importance of this unstudied linker region, our data suggest that at excitatory synapses, Nrxn3αA687T SS4− preferentially interacts with LRRTM2, which changes synaptic signaling and enhances presynaptic function.
Discussion
Here, we identified a novel missense mutation in a neurexin-3α compound heterozygous patient diagnosed with profound intellectual disability and epilepsy and systematically studied this mutation using molecular replacement in primary cultures and ex vivo acute slices. The mutation, located in a completely unstudied but evolutionarily conserved region, provided an avenue to directly probe the functional role of this extracellular sequence of Nrxn3α. Our results reveal that this sequence is a critical mediator in controlling trans interactions with postsynaptic ligands, presynaptic morphology, and release probability specifically at excitatory synapses. Perhaps it is not surprising that a mutation in Nrxn3α results in a presynaptic phenotype, given that neurexins are essential presynaptic proteins, however, it is remarkable that the phenotypes reported here are not observed in the Nrxn3α KO. Moreover, to the best of our knowledge, we identify the first manipulation of any neurexin that results in a potent but selective gain-of-function presynaptic phenotype at excitatory synapses. In addition to identifying a novel and unpredicted role for Nrxn3α extracellular sequences, our results highlight the value of studying these naturally occurring missense mutations because they have significant potential to provide novel functional insight that might otherwise be obscured in KO models.
In primary culture, molecular replacement with Nrxn3αA687T SS4− but not Nrxn3αA687T SS4+ significantly increased vGluT1 cluster size independent of changes in excitatory synapse density or inhibitory synaptic morphology (Fig. 3G, H) and enhanced mEPSC frequency, increased evoked AMPAR- and NMDAR-mediated synaptic transmission (Fig. 5F–I). Additionally, our data suggest that the striking presynaptic gain-of-function phenotype could be explained by changes in two presynaptic properties: (1) increased presynaptic release probability (Fig. 5I), and (2) enlarged the size of the excitatory RRP (Fig. 7B,C). In acute hippocampal slices, we found that presynaptic molecular replacement with Nrxn3αA687T SS4− at synapses formed by afferents from CA1 onto postsynaptic subicular RS and BS neurons significantly enhanced AMPAR-mediated synaptic strength and increased excitatory presynaptic release probability (Fig. 8D–G). Surprisingly, increased neurotransmitter release induced by Nrxn3αA687T SS4− occluded presynaptic, but not postsynaptic, LTP at CA1–subiculum synapses (Fig. 9).
We used cell aggregation assays to gain mechanistic insight into how the A687T mutation alters Nrxn3α function. We surprisingly found that the A687T mutation selectively enhanced trans interactions with LRRTM2 without altering interactions with any neuroligins (Fig. 10). LRRTM2 and NLs bind to the same surface in LNS6 (Siddiqui et al., 2010), which suggests that the mutation does not alter the direct binding properties of LNS6 but independently exerts its effect. How might enhanced Nrxn3α binding with LRRTM2 affect synaptic transmission? Elegant single molecule tracking experiments demonstrated that NL1 is uniformly distributed at excitatory post-synapses, whereas LRRTM2 forms tight domains near the core of the postsynaptic density (Chamma et al., 2016). An attractive explanation, based on single molecule tracking and our results presented here, is that the A687T mutation in Nrxn3α SS4− may cause the presynaptic rearrangement of the molecular architecture. Under normal circumstances, Nrxn3αWT SS4− interacts with NL1 and LRRTM2, and is evenly distributed in the presynaptic bouton, however, the preferential interaction of Nrxn3αA687T SS4− with LRRTM2 may cause a redistribution of Nrxn3αA687T SS4− and its intracellular signaling components more centrally in the AZ, ultimately leading to the observed presynaptic gain-of-function phenotype. Unfortunately, how Nrxn3αA687T SS4− functions to increase excitatory RRP size and presynaptic release probability via intracellular signaling is unknown because our understanding of the intracellular signaling downstream of neurexins is incomplete. Moreover, the answer to the fundamental question: “do individual neurexins engage distinct or redundant presynaptic intracellular signaling pathways?” is unknown. These data emphasize the growing importance of the transsynaptic adhesion code, i.e., how precise cell-type-specific neurexin isoforms govern interactions at the synapse to promote neurotransmission, and suggest that the presynaptic localization of α-neurexin interactions is profoundly important to function.
Given that Nrxn3α is highly expressed in many GABAergic interneuron subtypes in hippocampus, it is surprising that the A687T mutation did not affect inhibitory synapse morphology or function (Ullrich et al., 1995; Nguyen et al., 2016; Lukacsovich et al., 2019). Indeed, genetic deletion of Nrxn3α significantly impaired presynaptic release at inhibitory synapses in olfactory bulb neurons and deletion of both Nrxn3α and Nrxn3β in anterior cingulate cortex impaired neurotransmitter release from somatostatin interneurons (Aoto et al., 2015; Keum et al., 2018). A technical limitation to this study is the use of Nrxn3α replacement cDNAs, which are not subject to alternative splicing. Here, we manipulated the inclusion of the SS4 exon without testing the impact of the other splice sites, which may be crucial for GABAergic interneuron function (Fuccillo et al., 2015; Nguyen et al., 2016; Lukacsovich et al., 2019). To gain a more comprehensive understanding, it will be critical to study Nrxn3αA687T and other α-Nrxn missense mutations in an environment where splicing is controlled by endogenous mechanisms unique to each cell type.
Our study represents the first systematic functional interrogation of a neurexin-3α human mutation. Interestingly, a recent study identified a naturally occurring extracellular mutation in neurexin-3α (R498W) in a 129S1/SvlmJ mouse strain (Keum et al., 2018). The R498W mutation is located in LNS3, which is upstream of the mutation studied here (Fig. 1A; linker between EGF2 and LNS4). Keum et al. (2018) demonstrated that the R498W mutation specifically enhanced fear empathy in 129S1/SvlmJ mice relative to other commonly used mouse lines that contain an arginine at position 498 of Nrxn3α. They posited that the mutation in Nrxn3α impaired presynaptic release selectively at synapses made by somatostatin interneurons in anterior cingulate cortex (ACC), thus, in ACC Nrxn3α extracellular sequences appear to necessary for the release of inhibitory neurotransmitter at somatostatin synapses, however, the mutation itself was not directly functionally assayed at those synapses. By contrast, here we demonstrate that a completely different extracellular region of Nrxn3α is responsible for controlling presynaptic release and the size of the RRP at excitatory synapses in hippocampus. Together, these data suggest that different extracellular sequences specific for α neurexins may have the capacity to govern synaptic transmission in a brain-region, cell-type-, and synapse-specific manner.
To the best of our knowledge, our findings represent the first interrogation of a human neurexin mutation and the first reported gain-of-function manipulation of any neurexin, which significantly strengthens our understanding of the synapse-specific roles of α-Nrxn extracellular sequences. The discovery of the Nrxn3αA687T phenotype, unpredicted based on neurexin KO models, highlights the utility of studying patient mutations in genes that encode neurexins, increasing our knowledge of neurexin biology and providing insight into how these mutations alter synapse function that underlie neuropsychiatric and neurodevelopmental disorders.
Footnotes
This work was supported by Grants from the National Institute of Mental Health (R00MH103531 and R01MH116901 to J.A.), a predoctoral training Grant from the National Institute of General Medicine (T32GM007635 to S.R.), and a Lyda Hill Gilliam Fellowship for Advanced Study (GT11021 to S.R.). We thank the affected individual and their family for generously sharing genome sequencing data, members of the Aoto laboratory for helpful discussions, Dr. Samantha Schwartz for MATLAB programming, Dr. Joshua Black for q-PCR use, Dr. Jennifer Bourne in the Electron Microscopy Center at the University of Colorado School of Medicine, Dr. Scott Demarest for the COMIRB protocol, and Thomas Südhof (Stanford University) for the NL2 and LRRTM2 plasmids.
The authors declare no competing financial interests.
- Correspondence should be addressed to Jason Aoto at jason.aoto{at}ucdenver.edu