Alternative Splicing of Presynaptic Neurexins Differentially Controls Postsynaptic NMDA- and AMPA-Receptor Responses

Abstract

AMPA- and NMDA-type glutamate receptors mediate distinct postsynaptic signals that differ characteristically among synapses. How postsynaptic AMPA- and NMDA-receptor levels are regulated, however, remains unclear. Using newly generated conditional knockin mice that enable genetic control of neurexin alternative splicing, we show that in hippocampal synapses, alternative splicing of presynaptic neurexin-1 at splice-site #4 (SS4) dramatically enhanced postsynaptic NMDA- but not AMPA-receptor-mediated synaptic responses without altering synapse density. In contrast, alternative splicing of neurexin-3 at SS4 suppressed AMPA- but not NMDA-receptor-mediated synaptic responses, while alternative splicing of neurexin-2 at SS4 had no effect on NMDA-receptor- or AMPA-receptor-mediated responses. Presynaptic overexpression of the neurexin-1β and neurexin-3β SS4+ splice variants, but not of their SS4− splice variants, replicated the respective SS4+ knockin phenotypes. Thus, different neurexins perform distinct non-overlapping functions at hippocampal synapses that are independently regulated by alternative splicing. These functions trans-synaptically control NMDA- and AMPA-receptors, thereby mediating presynaptic control of postsynaptic responses.

INTRODUCTION

Neurexins are presynaptic cell-adhesion molecules expressed from three homologous genes (Nrxn1, Nrxn2, and Nrxn3 in mice), each of which encodes longer α- and shorter β- neurexins (reviewed in Reissner et al., 2013; Kasem et al., 2018; Südhof, 2017). In addition, the Nrxn1 gene, but not the Nrxn2 or Nrxn3 genes, encodes a smaller Nrxnγ-isoform (Sterky et al., 2017). Neurexins are highly homologous multidomain proteins that contain several extracellular protein-interaction domains, a single transmembrane region, and a short cytoplasmic sequence. Although neurexins are well studied, even simple questions about their functions remain unanswered, such as whether different neurexins mediate similar or different functions, or whether neurexins perform primarily cell-autonomous presynaptic or non-cell-autonomous trans-synaptic roles (Südhof, 2017). In human patients, mutations in neurexin genes have been repeatedly associated with neuropsychiatric disorders (Kasem et al., 2018; Doherty et al., 2012; Tam et al., 2009). NRXN1 mutations have been particularly frequently observed. Hundreds of patients with copy-number variations that affect only NRXN1 expression have been described, rendering NRXN1 mutations relatively common - albeit still rare - single-gene mutations associated with schizophrenia, Tourette syndrome, epilepsy, and autism (Sanders et al., 2011; Rees et al., 2014; Møller et al., 2013; Huang et al., 2017; Marshall et al., 2017). Although NRXN1 mutations are known to impair synaptic transmission in human neurons (Pak et al., 2015), how precisely NRXN1 mutations predispose to schizophrenia or autism is unclear.

Neurexin mRNAs are alternatively spliced at 6 canonical, highly homologous sites that produce thousands of isoforms (Ullrich et al., 1995; Schreiner et al., 2014; Treutlein et al., 2014). Among the sites of neurexin alternative splicing, splice-site #4 (SS4) has been studied most intensely. Alternative splicing at SS4 is shared by all α- and β-neurexins and involves inclusion or exclusion of a highly conserved 90 bp exon, producing SS4− and SS4+ variants of neurexins (Tabuchi and Südhof, 2002). Alternative splicing at SS4 is temporally and spatially regulated, suggesting a major role in synaptic regulation (Ullrich et al., 1995). In a compelling study, fear learning was shown to acutely regulate alternative splicing of Nrxn1 at SS4 in the hippocampus of mice via a mechanism involving histone modifications, suggesting an epigenetic control of neurexin-dependent synaptic properties even during a simple one-trial learning paradigm (Ding et al., 2017).

Multifarious ligands that bind to all neurexins have been identified, including neuroligins, cerebellins, and LRRTMs (reviewed in Yuzaki, 2017; Roppongi et al., 2017; Südhof, 2017). Alternative splicing of neurexins at SS4 regulates their binding to several of these ligands. LRRTMs primarily bind to neurexins lacking an insert in SS4 (Ko et al., 2009; Siddique et al., 2010), cerebellins to neurexins containing an insert in SS4 (Uemura et al., 2010; Joo et al., 2011; Matsuda and Yuzaki, 2011), and neuroligins to either SS4− or SS4+ neurexins but with distinct affinities (Ichtchenko et al., 1995; Comoletti et al., 2006; Boucard et al., 2005; Chih et al., 2006; Elegheert et al., 2017). However, despite a large amount of studies on neurexins and abundant information about the dynamics and biochemistry of neurexin alternative splicing, little is known about the physiological significance of neurexin alternative splicing. Knockin mice in which the alternatively spliced SS4 sequence of Nrxn3 was rendered constitutively spliced-in exhibited a decrease in AMPA-receptor- (AMPAR-) mediated synaptic responses in hippocampal synapses (Aoto et al., 2013). This decrease was reversed by presynaptic constitutive excision of SS4, but it remained unclear whether NMDA-receptor- (NMDAR-) mediated responses were also affected. Importantly, the possibly similar function of alternative splicing in other neurexins at SS4, a site that is highly homologous among neurexins, has not been examined - in fact, it is still unclear whether different neurexins are functionally similar or distinct.

Here, we investigated the role of neurexins and of their alternative splicing at SS4 using genetic manipulations of alternative splicing of endogenous neurexins. Specifically, we made conditional knockin (cKI) mice for Nrxn1 and Nrxn2 similar to those we previously generated for Nrxn3 (Aoto et al., 2013). In these SS4+ cKI mice, the alternatively spliced SS4 exon is constitutively included in all mRNAs (i.e., neurexins are no longer alternatively spliced, but expressed as SS4+ variants), but can be constitutively excised by Cre-recombinase (i.e., neurexins are converted into SS4− variants). Using the three neurexin SS4+ cKI mouse lines, we then asked whether different mammalian neurexins perform similar functions that are coordinately regulated by alternative splicing, or whether different neurexins mediate distinct functions that are independently regulated by alternative splicing. Surprisingly, our data reveal that Nrxn1^SS4+ selectively enhanced NMDAR-mediated synaptic responses without altering AMPAR responses, whereas Nrxn2^SS4+ had no effect on either NMDAR- or AMPAR-mediated responses, and Nrxn3^SS4+ selectively suppressed AMPAR-mediated responses without changing NMDAR-mediated responses. Thus, different neurexins perform distinct functions in hippocampal synapses by independently controlling postsynaptic NMDAR- or AMPAR-mediated responses, enabling dynamic control by presynaptic neurons of postsynaptic glutamate receptor responses.

RESULTS

Constitutive expression of Nrxn1^SS4+ increases NMDAR-mediated responses

Using the same strategy that we previously employed for the control of SS4 alternative splicing of Nrxn3 (Aoto et al., 2013), we generated Nrxn1-SS4+ cKI mice. In these mice, alternative splicing of Nrxn1 mRNAs at SS4 is abolished, leading to constitutive expression of Nrxn1-SS4+ that can be converted by Cre-recombinase into constitutive expression of Nrxn1-SS4− (Figure 1A). Specifically, we used homologous recombination in murine ES cells to mutate the non-canonical splice acceptor sequence of the exon encoding SS4, and produced mice from these ES cells. In addition, we inserted LoxP sites into the introns flanking the alternatively spliced SS4 exon. Thereby we converted the normally non-canonical splice-acceptor sequence of the SS4 exon in the endogenous Nrxn1 gene into a canonical splice-acceptor sequence, causing constitutive inclusion of the SS4 exon. As a result, Nrxn1-SS4+ cKI mice constitutively express Nrxn1^SS4+ mRNAs in the absence of Cre-recombinase, but Nrxn^SS4− mRNAs in the presence of Cre-recombinase (Figure 1A). mRNA measurements in neurons cultured from Nrxn1-SS4+ cKI mice confirmed constitutive expression of Nrxn1^SS4+ and Nrxn^SS4− mRNAs in the absence and presence of Cre-recombinase, respectively (Figure 1B, 1C).

Figure 1: Alternative splicing of presynaptic neurexin-1 (Nrxn1) at splice site #4 (SS4) regulates postsynaptic NMDAR- but not AMPAR-mediated synaptic transmission in cultured hippocampal neurons.

A. Genetic strategy for Nrxn1-SS4+ conditional knockin (cKI) mice. The non-canonical splice acceptor (SA) sequence of alternatively spliced exon 21 (E21) of the Nrxn1 gene was converted into a canonical splice acceptor sequence by homologous recombination in ES cells (red letters = mutated residues), and exon 21 was flanked by LoxP sites (single letters = restriction enzyme sites [B, BamH1; E, EcoR5; P, Pst1]). Nrxn1-SS4+ cKI neurons constitutively express Nrxn1^SS4+, but are rendered Nrxn1^SS4− by Cre-recombinase.

B & C. Validation of Nrxn1-SS4+ conditional knockin (cKI) mice. B, Summary graphs showing total Nrxn1 mRNA levels (left) and the Nrxn1 SS4 alternative splicing ratio (Nrxn1^SS4−/(Nrxn1^SS4+ + Nrxn1^SS4−); right), as measured by quantitative RT-PCR of hippocampal neurons cultured from WT and Nrxn1^SS4+ mutant mice. C, Representative gel exhibiting Nrxn1 SS4 alternative splice variants. Nrxn1^SS4+ neurons were infected at DIV4–5 with lentiviruses expressing ΔCre-eGFP (retains Nrxn1^SS4+ genotype) or Cre-eGFP (produces Nrxn1^SS4− genotype). Neurons were analyzed at DIV14–16. Note that wild-type cultured hippocampal neurons predominantly but not exclusively express the SS4− variant of Nrxn1 (see also Fuccillo et al., 2015).

D. Nrxn1-SS4 alternative splicing controls NMDAR- but not AMPAR-EPSCs in cultured hippocampal neurons. Evoked NMDAR- and AMPAR-EPSCs were recorded from cultured WT neurons or Nrxn1-SS4+ neurons infected with lentiviruses expressing ΔCre-eGFP (to retain Nrxn1^SS4+ genotype) or Cre-eGFP (to produce Nrxn1^SS4− genotype).

E. Nrxn1-SS4 alternative splicing does not affect presynaptic release probability as monitored by paired-pulse ratio (PPR) measurements for NMDAR-mediated responses in Nrxn1^SS4+ and Nrxn1^SS4− hippocampal neurons obtained as described in A (left, sample traces; right, summary graphs).

F. Representative images of WT, Nrxn1^SS4+ and Nrxn1^SS4− hippocampal neurons that were obtained as described in A and stained by immunocytochemistry for MAP2 and the inhibitory and excitatory synapse markers GAD65 and vGlut1, respectively, as indicated.

G & H. Nrxn1-SS4 alternative splicing does not affect the density or size of excitatory or inhibitory synapses as monitored by immunocytochemistry (summary graphs of the density (left), staining intensity (middle) and size (right) of vGlut1-positive excitatory (G) and GAD65-positive inhibitory synaptic puncta (H) quantified by immunocytochemical staining of WT, Nrxn1^SS4+ and Nrxn1^SS4− hippocampal neurons as described in F).

Numerical data are means ± SEM; numbers in bars present independent experiments or number of total cells/ independent experiments; statistical analyses were performed by one-way ANOVA (* = p <0.05; non-significant comparisons are not noted).

As a first test of the physiological role of Nrxn1-SS4 alternative splicing, we infected hippocampal neurons cultured from Nrxn1-SS4+ cKI mice with lentiviruses encoding inactive (ΔCre) or active Cre-recombinase (Cre) to produce Nrxn1^SS4+ and Nrxn1^SS4− neurons, respectively, using hippocampal neurons from WT mice as a control. We then measured evoked synaptic responses by whole-cell patch-clamp recordings, and observed a dramatic increase (~50%) in NMDAR-EPSCs in Nrxn1^SS4+ neurons compared to Nrxn1^SS4− or control neurons, without a change in AMPAR-EPSCs (Figure 1D). The increase in NMDAR-EPSCs was not associated with a change in paired-pulse ratio, mini frequency, or excitatory and inhibitory synapse numbers, suggesting that postsynaptic NMDARs are selectively increased by constitutive expression of Nrxn1^SS4+ (Figure 1E–1H, S1). These results were unexpected given that similar manipulations of the homologous Nrxn3 gene caused a decrease in AMPAR-mediated responses (Aoto et al., 2013). However, these results were intriguing because in human patients, NRXN1 mutations are linked to neuropsychiatric disorders, especially schizophrenia (Rees et al., 2014; Marshall et al., 2017), which in turn may involve impairments in NMDAR-mediated synaptic transmission (Hu et al., 2016; Dalmau, 2016; Yan et al., 2016).

Cultured neurons do not always maintain physiologically relevant synaptic functions, prompting us to examine the role of Nrxn1-SS4 alternative splicing in vivo. For this purpose, we chose the hippocampal CA1 region in which Nrxn1-SS4 alternative splicing is dynamically regulated by behavioral activity (Ding et al., 2017). We stereotactically infected the CA1 region of Nrxn1-SS4+ cKI mice at postnatal day 21 (P21) with adeno-associated viruses (AAVs) encoding ΔCre (as a control; retains Nrxn1^SS4+ genotype) or Cre (produces Nrxn1^SS4− genotype). We then analyzed the injected mice at P35–42, using WT mice of the age as a further control. In the analysis, we focused on the CA1➔subiculum synapse as the major output pathway of CA1 neurons (Figure 2A; Bohm et al., 2018; Cembrowski et al., 2018). This experimental approach permits selective analysis of the effect of presynaptic Nrxn1-SS4 alternative splicing in CA1 neurons on synaptic transmission in postsynaptic subiculum neurons that were not virally manipulated (Xu et al., 2011).

Figure 2: Alternative splicing of presynaptic Nrxn1-SS4 regulates postsynaptic NMDAR- but not AMPAR-mediated synaptic transmission in vivo at CA1-subiculum synapses.

A. Experimental strategy for in vivo manipulations. Left, diagram of stereotactic injections; right, illustration of electrophysiological recordings from acute slices. Hippocampal CA1 regions of Nrxn1-SS4+ cKI mice were bilaterally infected by stereotactic injections of AAVs expressing ΔCre-eGFP or Cre-eGFP at P21, and subiculum synapses were analyzed at P35-P42. Expression of ΔCre-eGFP retains presynaptic Nrxn1^SS4+ genotype, whereas expression of Cre-eGFP induces a presynaptic Nrxn1^SS4− genotype in a mouse that is otherwise Nrxn1^SS4+ (green fluorescence = nuclear Cre-eGFP).

B. Nrxn1-SS4 alternative splicing has no effect on excitatory synapse density in the subiculum (left, representative sections stained for vGluT1 and MAP2; right, summary graph of synapse density assessed by vGluT1 signal (a.u. = arbitrary units)).

C. Nrxn1-SS4 alternative splicing has no detectable effects on spontaneous miniature mEPSCs monitored in the presence of TTX at a holding potential of −70 mV (left, representative traces; right, summary graphs of the mEPSC frequency, amplitude, and rise times.

D. Nrxn1-SS4+ decreases the AMPAR/NMDAR ratio in subiculum neurons compared to WT controls; this phenotype is reversed by presynaptic Cre-mediated excision of Nrxn1-SS4 that renders CA1 neurons Nrxn1^SS4− (left, sample traces of AMPAR/NMDAR EPSCs from WT and Nrxn1-SS4 cKI mice after CA1 expression of ΔCre-eGFP (retains Nrxn1^SS4+ genotype) or Cre-eGFP (produces Nrxn1^SS4− genotype); right, summary graphs of the AMPAR/NMDAR EPSCs ratio).

E & F. Nrxn1^SS4+ enhances NMDAR- but not AMPAR-EPSC amplitudes in subiculum burst-spiking neurons; presynaptic conversion of Nrxn1^SS4+ to Nrxn1^SS4− reverses NMDAR-EPSC enhancement (left, sample traces; middle, input-output plots; right, summary graphs of input-output relations).

G & H. Same as E & F, but recorded from regular-spiking subiculum neurons.

All data are means ± SEM. Number of sections/mice (B) and neurons/mice (C-H) are indicated in bars. Statistical significance was assessed by unpaired two-tailed t-test or one-way and two-way ANOVA (*P≤0.05, **P≤0.01, and ***P≤0.001).

Immunocytochemistry showed that constitutive presynaptic expression of Nrxn1^SS4+ or Nrxn1^SS4− in the CA1 region had no apparent effect on excitatory synapse density in the subiculum (Figure 2B, S2A). Consistent with this conclusion, the frequency and amplitude of mEPSCs were unaffected in subiculum neurons by Nrxn1-SS4 alternative splicing, as monitored whole-cell patch-clamp recordings in acute slices at a holding potential of −70 mV to detect AMPAR-mediated responses (Figure 2C). Nrxn1-SS4 alternative splicing also produced no change in the frequency of mEPSCs as monitored at at a holding potential of +60 mV to capture NMDAR-mediated contributions to mEPSCs (Figure S2B). Presynaptic conversion of Nrxn1^SS4+ into Nrxn^SS4− did, however, induce a small but significant decrease in the charge of presumptive NMDAR-mediated mEPSCs, consistent with an increased NMDAR-response in Nrxn1^SS4+ synapses observed in cultured neurons (Figure S2B). Overall, these observations demonstrate that presynaptic Nrxn1-SS4 alternative splicing in CA1 neurons has no effect on the density of the output synapses of these neurons in the subiculum.

We next recorded evoked AMPAR- and NMDAR-EPSCs in CA1➔subiculum synapses (Figure 2A). We first measured the ratio of AMPAR- to NMDAR-mediated EPSCs by monitoring in the same neurons first AMPAR-mediated responses (quantified as the peak EPSC) at a holding potential of −70 mV, and then NMDAR-mediated responses (quantified at 50 ms after the peak EPSC) at a holding potential of +40 mV (Figure 2D). Presynaptic Nrxn1^SS4+ produced a significant decrease in the AMPAR/NMDAR ratio that was reversed by presynaptic conversion of Nrxn1^SS4+ to Nrxn1^SS4− (Figure 2D), again in agreement with the observation of an increase in NMDAR-mediated responses in cultured neurons.

To directly test whether NMDAR-mediated responses are increased in Nrxn1^SS4+ neurons in vivo, we examined isolated NMDAR- and AMPAR-mediated responses in acute slices. In these and all subsequent subiculum recordings, we separately analyzed regular- and burst-spiking neurons, which are the two principal types of pyramidal neurons in the subiculum that can be distinguished by their intrinsic electrical properties (Figure S2C; Bohm et al., 2018). We elicited EPSCs by extracellular stimulation, monitored AMPAR- and NMDAR-mediated EPSCs separately, and used input/output measurements to control for possible differences in stimulus strength (Figure 2E–2H). Consistent with the results obtained in cultured neurons, we detected no differences in AMPAR-mediated synaptic responses between WT, Nrxn1^SS4+, and Nrxn1^SS4− neurons (Figure 2E, 2G), but observed a ~50–60% increase in NMDAR-mediated synaptic responses in Nrxn1^SS4+ neurons compared to Nrxn1^SS4− and WT control neurons (Figure 2F, 2H). This increase was detected in both burst- and regular-spiking neurons. Presynaptic Nrxn1-SS4 alternative splicing had no effect on paired-pulse ratios of AMPAR- or NMDAR-EPSCs, suggesting that the increase in NMDAR-mediated synaptic responses was caused by a postsynaptic mechanism (Figure S2D–2G).

Burst- and regular-spiking neurons of the subiculum exhibit distinct types of LTP, with burst-spiking neurons producing presynaptic LTP involving changes in release probability, while regular-spiking neurons exhibit postsynaptic NMDAR-dependent LTP involving changes in AMPAR levels (Wozny et al., 2008). As for Nrxn3-SS4 alternative splicing, Nrxn1-SS4 alternative splicing had no effect on presynaptic LTP in burst-spiking neurons (Figure 3A–3D). However, different from Nrxn3-SS4 alternative splicing in which constitutive Nrxn3^SS4+ expression blocked NMDAR-dependent postsynaptic LTP (Aoto et al., 2013), constitutive expression of Nrxn1^SS4+ enhanced postsynaptic NMDAR-dependent LTP in regular-spiking subiculum neurons (Figure 3E–3H). Again, this enhancement was reversed by presynaptic conversion of Nrxn1^SS4+ to Nrxn1^SS4− (Figure 3E–3H). This enhanced LTP was likely a consequence of the increased NMDAR-response in Nrxn1^SS4+ neurons, and probably did not reflect a change in LTP itself (Nicoll, 2017).

Figure 3: Nrxn1-SS4 alternative splicing does not have a major effect on LTP in the subiculum or the CA1 region.

A & B. Presynaptic LTP induced in burst-spiking neurons by four 100 Hz/1 s stimulus trains with 10 s intervals under voltage-clamp mode is unchanged by Nrxn1-SS4 alternative splicing (A, sample traces; B, summary plot). LTP was recorded in subiculum slices from WT control mice and from Nrxn1-SS4+ cKI mice after presynaptic CA1 region infection with AAVs expressing ΔCre-eGFP (retains Nrxn1 genotype) or Cre-eGFP (produces Nrxn1 genotype).

C & D. Summary graphs of the LTP magnitude (C, normalized average EPSC amplitudes during the last 5 mins after LTP induction) and of the paired-pulse ratios before and after LTP (D, as a measure of the release probability).

E-H. Presynaptic Nrxn1-SS4+ enhances NMDAR-dependent LTP in the subiculum. Same as (A-D) but recorded from regular-spiking neurons in which the stimulus train induces postsynaptic NMDAR-dependent LTP without changing the PPR.

I & J. Nrxn1-SS4+ has no effect on NMDAR-dependent LTP in the CA1 region induced by thetaburst stimulation. LTP was recorded in acute hippocampal slices from Nrxn1-SS4+ and WT mice in the presence of picrotoxin (I, example traces (left top) and summary plot (left bottom) of EPSCs before and after theta-burst stimulation of Schaffer collaterals; J, summary graph of the normalized LTP magnitude during the last 5 mins of recordings).

K & L. Nrxn1-SS4+ also has no effect on NMDAR-dependent LTP in the CA1 region induced by two consecutive 100 Hz stimulus trains applied for 1 s with a 10 s interval. Data were obtained and are presented as in panels I & J.

Data are means ± SEM. Number of neurons/mice analyzed are shown in the graphs. Statistical analyses were performed by one-way (C, G) or unpaired two-tailed t-test (D, H, J. L), with * = p ≤ 0.05.

Thus, Nrxn1 alternative splicing at SS4 does not impair NMDAR-mediated LTP in subiculum neurons. However, this conclusion varies from that of a previous study on neurexin alternative splicing at SS4 in the hippocampal CA1 region (Traunmüller et al., 2016). Traunmüller et al. (2016) showed that deletion of the RNA splicing factor SLM2 caused increased inclusion of SS4 inserts in all neurexin mRNAs and also blocked NMDAR-dependent LTP. Strikingly, LTP was rescued in SLM2 KO mice by constitutive heterozygous expression of Nrxn1^SS4−, suggesting that the SLM2 KO may block LTP by increasing the levels of Nrxn1^SS4+ (Traunmüller et al., 2016). To test directly whether Nrxn1^SS4+ blocks LTP in CA1 region neurons, we examined NMDAR-dependent CA1-region LTP induced by standard induction protocols (theta-burst stimulation as used by Traunmüller et al. (2016), and two 100 Hz, 1 sec tetani). We observed no effect of Nrxn1^SS4+ on LTP in either of these experiments, suggesting that Nrxn1^SS4+ does not affect NMDAR-mediated LTP (Figure 3I–3L). Thus, the block of LTP observed in the SLM2 KO mice may have been due to other SLM2-dependent activities, possibly Nrxn3^SS4+ which is known to abolish NMDAR-dependent LTP (Aoto et al., 2013).

Nrxn1-SS4 alternative splicing in hippocampal CA1 region controls spatial memory

The hippocampus is central to contextual memory (Morris, 2013; Kitamura et al., 2015). Thus, selective changes in NMDAR-mediated signaling at output synapses from the hippocampal CA1-region might be expected to induce alterations in contextual learning and memory. To test this hypothesis, we examined contextual memory formation in Nrxn1- SS4+ cKI mice after injection of AAVs encoding ΔCre (retains Nrxn1^SS4+ expression) or Cre (induces selective Nrxn1^SS4− expression in CA1 neurons in an otherwise Nrxn1^SS4+ background). In this experiment, we are comparing mice in which all neurons express exclusively Nrxn1^SS4+ to mice in which only CA1 neurons express Nrxn1^SS4− but all other neurons still express exclusively Nrxn1^SS4+. Strikingly, selective conversion of Nrxn1^SS4+ to Nrxn1^SS4− in CA1 neurons significantly decreased contextual memory, as measured by two tests, passive avoidance and the Barnes maze (Figure 4, S3A–S3C). These two tests employ different contexts (a marked chamber vs. a circle with escape holes) and different behavioral motivations (electrical shock vs. escape from an exposed, brightly illuminated place; see Figure S3), and thus broadly monitor contextual memory. The dependence of a mouse’s performance in these tests on Nrxn1 SS4 alternative splicing suggests that the postsynaptic NMDAR-signal strength at CA1 output synapses is important for contextual memory, consistent with earlier studies (Morris, 2013).

Figure 4: Nrxn1-SS4 alternative splicing in hippocampal CA1 region controls contextual memory.

The CA1 region of Nrxn1-SS4+ cKI mice was infected using stereotactic injections at P21 with AAVs expressing ΔCre-eGFP (retains Nrxn1^SS4+ genotype) or Cre-eGFP (produces presynaptic Nrxn1^SS4− genotype), and mice were analyzed at P42–67.

A. Nrxn1-SS4 alternative splicing in the hippocampal CA1 region does not regulate open field behavior of mice.

B. Nrxn1-SS4 alternative splicing in the hippocampal CA1 region does not affect training in the passive avoidance test (left, summary plots and graphs of chamber entries; right, summary plots and graphs for the movement activity [average values in the ‘safe’ chamber during the last training trial]).

C. Nrxn1-SS4 alternative splicing controls spatial memory in mice as measured by conditioned passive avoidance. Entry latency, number of entries, and activity of mice on day 1 and day 7 after training were measured.

D. Cartoon of Barnes spatial maze test. In the circular Barnes maze, mice learn the escape location of the single open hole by remembering visual cues shown in color.

E. Presynaptic Nrxn1-SS4 alternative splicing controls learning in the Barnes maze test. Summary graphs show the latency, distance traveled and primary errors of Nrxn1^SS4+ and Nrxn1^SS4− mice that were tested for 90 s at 24 hours after training.

F. Summary plot of the target-hole preference in a probe trial with all holes closed after training and testing reveals that Nrxn1^SS4+ mice remember the target hole location better than Nrxn1^SS4− mice (in which subiculum NMDAR responses are decreased).

Data are means ± SEM (number of mice shown in graphs apply to all mice in a series). Statistical analyses were performed using an unpaired one-tailed t-test (B, C) or two-tailed t-test (A, E, F), with * = p ≤ 0.05.

Nrxn3-SS4 alternative splicing does not impact NMDAR-EPSCs

The selective function of presynaptic Nrxn1-SS4 alternative splicing in regulating postsynaptic NMDAR-mediated responses is surprising given the earlier finding that SS4 alternative splicing of Nrxn3 controls postsynaptic AMPAR-responses (Aoto et al., 2013). However, these earlier studies did not test a possible role for Nrxn3-SS4 alternative splicing in regulating NMDAR EPSCs. Therefore, we examined whether Nrxn3-SS4+ cKI mice display an NMDAR phenotype in addition to the AMPAR changes. We measured NMDAR-EPSCs in Nrxn3-SS4+ cKI mice after presynaptic expression of either ΔCre or Cre, but failed to observe a difference (Figure S3D–S3G). Thus, Nrxn3-SS4 alternative splicing selectively regulates AMPAR-EPSCs, whereas Nrxn1-SS4 alternative splicing selectively controls NMDAR-EPSCs, suggesting that these two homologous neurexins perform distinct trans-synaptic functions.

Nrxn2-SS4 alternative splicing has no effect on either AMPAR- or NMDAR-EPSCs

The finding that alternative splicing of Nrxn1 and Nrxn3 at SS4 independently regulates NMDAR- and AMPAR-EPSCs, respectively, raises the question whether the third neurexin, Nrxn2, performs an analogous function in regulating glutamate receptors. To address this question, we generated Nrxn2-SS4+ cKI mice using the same approach as for Nrxn1- SS4+ and Nrxn3-SS4+ cKI mice (Figure 5A). We again validated the genetic approach using mRNA measurements in cultured neurons (Figure 5B, 5C). We then tested the effect of Nrxn2 alternative splicing at SS4 on NMDAR- or AMPAR-mediated synaptic responses, but uncovered no difference between Nrxn2^SS4+ and Nrxn2^SS4− states in either NMDAR- or AMPAR-responses (Figure 5D–5G, S4A, S4B). Thus, Nrxn2 alternative splicing at SS4 is not required for maintaining normal NMDAR- or AMPAR-mediated synaptic responses.

Figure 5: Generation and characterization of Nrxn2-SS4+ cKI mice containing constitutive insertion (Nrxn2^SS4+) or exclusion (Nrxn2^SS4−) of SS4 in Nrxn2.

A. Strategy for generating Nrxn2-SS4+ cKI mice, identical to that employed for Nrxn1-SS4+ cKI (Figure 1A) and Nrxn3-SS4+ cKI mice (Aoto et al., 2013). The splice acceptor (SA) sequence of the alternatively spliced SS4 exon 20 (E20) was mutated by homologous recombination to convert a non-canonical into a canonical SA sequence, thus ablating alternative splicing (mutated nucleotes are shown in red). Simultaneously, loxP sites were introduced into the introns flanking the exon 20 to enable Cre-dependent excision of the alternatively spliced exon. Single letters indicate locations of restriction enzyme sites (S, Sfc1; B, BamH1; P, Pst1).

B & C. Validation of Nrxn2-SS4+ cKI mice by mRNA measurements. B, Summary graph showing total Nrxn2^SS4+ (blue bars) and Nrxn2^SS4− (green bars) mRNA levels (normalized for b-actin levels) as measured by quantitative RT-PCR of hippocampal neurons cultured from WT and Nrxn2-SS4+ cKI mice that had been infected at DIV4–5 with lentiviruses expressing mutant (ΔCre, retains Nrxn2^SS4+ genotype) or active Cre-recombinase (Cre, produces Nrxn2^SS4− genotype). C, Representative gel of RT-PCR products exhibiting Nrxn2 SS4 alternative splice variants from neurons obtained as in B.

D. Nrxn2-SS4 alternative splicing does not alter NMDAR-mediated EPSCs monitored in burst-spiking subiculum neurons. Acute slices from WT mice and Nrxn2-SS4+ cKI mice whose CA1 region neurons had been stereotactically infected at P21 with AAVs expressing ΔCre-eGFP or Cre-eGFP were analyzed by electrophysiological recordings with a holding potential of +40 mV to monitor NMDAR-mediated EPSCs in the presence of CNQX and picrotoxin as described in Figure 2 (left, sample traces; middle summary plot, input/output curves; right summary graph, slopes of input/output curves).

E. Nrxn2-SS4 alternative splicing has no effect on AMPAR-mediated EPSCs monitored in burst- spiking subiculum neurons. Analyses and panels are as described for D, except that recordings were made in the absence of CNQX but with picrotoxin at a holding potential of −70 mV, and that only the Nrxn2^SS4+ and Nrxn2^SS4− conditions were compared.

F & G. Same as D & E, but in regular-spiking neurons.

Data are means ± SEM (number of experiments shown in bars in B and of cells/mice shown in graphs D-G apply to all mice in a series). Statistical analyses were performed using one-way and two-way ANOVA or an unpaired two-tailed t-test.

Nrxn1^SS4+-induced increases in NMDAR-EPSCs and Nrxn3^SS4+-induced decreases in AMPAR-ESPCs are directly mediated by postsynaptic receptor responses

To further test the conclusions derived from the input/output measurements described above, we asked whether presynaptic alternative splicing of Nrxn1 or Nrxn3 at SS4 alters postsynaptic currents that are induced in subiculum neurons by direct application of AMPA, NMDA, or GABA. In these experiments, we did not test Nrxn2 since it had no NMDAR or AMPAR phenotype (Figure 5). We stereotactically infected the CA1 region of Nrxn1-SS4+ or Nrxn3-SS4+ cKI mice at P21 with AAVs encoding Cre or ΔCre, and sectioned acute slices from these mice and from control WT mice at P35-P42. We then recorded postsynaptic currents from subiculum neurons in response to direct applications of AMPA, NMDA, or GABA.

We found that AMPA-induced EPSCs in the subiculum were similar in WT and Nrxn1-SS4+ cKI mice, independent of whether the Nrxn1-SS4+ cKI mice had been injected with Cre or ΔCre (Figure 6A). In Nrxn3-SS4+ cKI mice, however, AMPA-induced EPSCs were dramatically decreased when the CA1 region had been infected with ΔCre; this decrease was reversed by CA1-region expression of Cre that converts presynaptic Nrxn1^SS4+ to Nrxn1^SS4− (Figure 6A). NMDA-induced EPSCs, conversely, were strongly enhanced in Nrxn1-SS4+ cKI mice that had been infected with ΔCre, and this phenotype again was reversed by presynaptic expression of Cre (Figure 6B). NMDA-induced EPSCs were not altered by Nrxn3-SS4 manipulations (Figure 6B), and neither the Nrxn1 nor the Nrxn3 manipulations altered IPSCs in response to GABA applications (Figure 6C). Since in these experiments we are monitoring isolated postsynaptic receptor responses as a function of presynaptic alternative SS4 splicing, these results confirm that Nrxn1 and Nrxn3 independently control postsynaptic NMDAR and AMPAR levels, respectively, by a trans-synaptic signaling mechanism.

Figure 6: Nrxn1 and Nrxn3 alternative splicing at SS4 in CA1 region neurons controls NMDAR- and AMPAR-mediated responses, but not GABAR-mediated responses, in subiculum neurons in response to puffed AMPA, NMDA and GABA.

For all experiments, the CA1 region of Nrxn1-SS4+ or Nrxn3-SS4+ cKI mice was injected at P21 with AAVs encoding ΔCre-eGFP (retains SS4+ genotype) or Cre-eGFP (produces SS4− genotype). Acute slices were sectioned from injected mice or WT control mice at P35-P42. Postsynaptic currents were recorded from patched subiculum pyramidal neurons in the presence of TTX in response to puffed AMPA (50 μM, in the presence of picrotoxin and D-APV), NMDA (50 μM, in the presence of picrotoxin and CNQX) or GABA (50 μM, in the presence of D-APV and CNQX). NMDA, AMPA or GABA were applied with a picospritzer via a puffing pipette positioned about 2 cell distances from the recorded neuron.

A. Presynaptic Nrxn3-SS4+ but not Nrxn1-SS4+ selectively suppresses AMPAR-mediated responses (top, example traces of postsynaptic responses to puffed AMPA delivered with increasing application times and monitored at a −70 mV holding potential; bottom left, summary plots of the dose-response curve to puffed AMPA and summary graph of the slope of that curve measured in WT and Nrxn1 and Nrxn1 neurons; bottom right, same as bottom left, except that Nrxn3^SS4+ and Nrxn3^SS4− neurons were analyzed.

B. Presynaptic Nrxn1-SS4+ but not Nrxn3-SS4+ selectively enhances NMDAR-mediated responses. Experiments are the same as in A, but monitored at a +40 mV holding potential in response to applications of NMDA.

C. Presynaptic Nrxn1 and Nrxn3 SS4-splice variants have no effect on GABA-receptor mediated responses. Experiments were performed as in A.

Numerical data are means ± SEMs. Numbers in bars or labels present number of cells/mice analyzed; statistical significance was assessed by one-way and two-way ANOVA (* = p < 0.05; ** = p < 0.01; *** = p < 0.001).

Combined constitutive insertion or exclusion of SS4 in all three neurexins produces the sum of the Nrxn1-SS4 and Nrxn3-SS4 phenotypes

Our results show that Nrxn1 alternative splicing at SS4 selectively regulates NMDAR-ESPCs, that Nrxn3 alternative splicing at SS4 selectively regulates AMPAR-EPSCs, and that Nrxn2 alternative splicing at SS4 has no effect on either NMDAR- or AMPAR-EPSCs (Figure S5A). However, it is possible that partial redundancy between neurexins may have occluded a redundant function of neurexins in these synaptic responses. To examine this possibility, we tested the total effect of alternative splicing at SS4 of all neurexins. We generated and validated triple Nrxn123-SS4+ cKI mice in which SS4 of all three neurexin genes is rendered constitutively spliced-in, but is constitutively excised by Cre-recombinase (Figure S5B, S5C).

We injected triple Nrxn123-SS4+ cKI mice with Cre- or ΔCre-expressing AAVs in the CA1 region at P21, and analyzed their subiculum synapses electrophysiologically at P40–42. Measurements of mEPSCs and mIPSCs uncovered a significant decrease in both the mEPSC frequency and amplitude without a change in mIPSC frequency or amplitude (Figure S6A–S6D). Analysis of AMPAR- and NMDAR-mediated synaptic responses in the CA1➔subiculum synapses of these mice uncovered a combination of the separate Nrxn1 and Nrxn3 SS4 phenotypes without evidence of super-additive effects (Figure 7A, 7B). Specifically, the Nrxn123-SS4+ condition caused a large increase in NMDAR-mediated and a correspondingly large decrease in AMPAR-mediated responses, both of which were reversed upon presynaptic conversion of SS4+ into SS4− (Figure 7). We observed no changes in the paired-pulse ratio of evoked AMPAR- or NMDAR-EPSCs (Figure S6E, S6F). These results suggest that there may be redundancy in the functions of SS4 alternative splicing of the three neurexins as revealed by the mEPSC measurements, but that the combined phenotype in evoked responses induced by simultaneous manipulation of SS4 alternative splicing in all three neurexins is similar to the sum of their individual phenotypes.

Figure 7: Genetic control of presynaptic SS4-alternative splicing of all neurexins (Nrxn1, Nrxn2, and Nrxn3) using triple Nrxn123-SS4+ cKI mice reveals opposite effects of SS4+ splice variants on AMPAR- and NMDAR-mediated synaptic transmission.

A & B. Coordinate insertion of SS4 in all neurexins using triple Nrxn123^SS4+ cKI mice enhances NMDAR- but decreases AMPAR-EPSC amplitudes; presynaptic conversion of Nrxn123^SS4+ to Nrxn123^SS4− reverses these phenotypes. Input-output relations were recorded from acute slices from control (WT) and Nrxn123-SS4⁺ triple cKI mice; the latter had been infected in the CA1 region by stereotactic injections of AAVs encoding ΔCre-eGFP (to retain Nrxn123^SS4+ genotype) or Cre-eGFP (to induce Nrxn123^SS4− genotype). Numerical data are means ± SEMs. Numbers in bars or labels present number of cells/mice analyzed; statistical significance was assessed by one-way and two-way ANOVA (* = p < 0.05; ** = p < 0.01).

Forced presynaptic expression of SS4+ and SS4– neurexins replicates knockin phenotypes

Viewed together, our data indicate that Nrxn1, Nrxn2, and Nrxn3, despite their homology, perform surprisingly different presynaptic functions that are regulated by alternative splicing at SS4 and that collectively control the postsynaptic receptor composition of output synapses. However, since Nrxn1^SS4+ and Nrxn3^SS4+ are constitutively expressed in Nrxn1-SS4+ and Nrxn3-SS4+ cKI mice throughout development, a potential concern of these experiments is that SS4+ cKI manipulations may have affected synapse specification during development. To address this concern, we tested whether exogenous neurexins introduced into the presynaptic CA1 region of juvenile WT mice dominantly control the postsynaptic receptor composition of subiculum synapses in a manner regulated by alternative splicing. Specifically, we used stereotactic injections of AAVs to express Nrxn1β^SS4+, Nrxn1β^SS4−, Nrxn3β^SS4+, or Nrxn3β^SS4− in the CA1 region of WT mice at P21, and analyzed AMPAR- and NMDAR-mediated synaptic transmission at CA1➔subiculum synapses at P35–42 (Figure 8, S7).

Figure 8: Presynaptic overexpression of SS4+ splice variants of Nrxn1β or Nrxn3β in the CA1 region replicates the Nrxn1-SS4+ and Nrxn3-SS4+ cKI phenotype, demonstrating that Nrxn1 and Nrxn3 exhibit intrinsically different functional properties.

A & B. AAV-mediated presynaptic overexpression of Nrxn1β^SS4+ but not of Nrxn1β^SS4− in hippocampal CA1 neurons enhances NMDAR- but not AMPAR-EPSCs, whereas presynaptic overexpression of Nrxn3β^SS4+ but not of Nrxn3β^SS4− has no effect on NMDAR-ESPCs but suppresses AMPAR-EPSCs (C &D) elicited by CA1 output neurons in the subiculum (left, sample traces; middle, input-output curves; right, summary graphs of the slopes of input-output curves).

C & D. Same as A & B, but for presynaptic expression of Nrxn3β^SS4+ or Nrxn3β^SS4−.

Data are means ± SEM. Number of neurons/mice are indicated in bars. Statistical significance was assessed by one-way and two-way ANOVA (* = p<0.05; ** = p < 0.01; *** = p < 0.001).

Strikingly, presynaptic overexpression of Nrxn1β^SS4+ induced a ~100% increase in NMDAR-mediated synaptic strength without significantly affecting AMPAR-mediated responses, whereas presynaptic overexpression of Nrxn1β^SS4− had no effect on either NMDAR- or AMPAR-mediated responses (Figure 8A, 8B). Overexpression of Nrxn3β^SS4+, conversely, caused a ~50% decrease in AMPAR-mediated synaptic strength without significantly altering NMDAR-mediated responses; again, overexpression of Nrxn3β^SS4− had no effect on either NMDAR- or AMPAR-mediated responses (Figure 8C, 8D). All neurexin variants were expressed at similarly high levels as validated by quantitative RT-PCR measurements (Figure S7A, S7B). None of the β-neurexins had an effect on paired-pulse ratios, suggesting that they did not change presynaptic neurotransmitter release (Figure S8A–S8D). Thus, presynaptic neurexin alternative splicing at SS4 controls postsynaptic glutamate receptors in a dominant fashion, independent of the developmental context, with Nrxn1 and Nrxn3 exhibiting distinct intrinsic activities.

DISCUSSION

Recent work revealed that neurexins are central regulators of synapse properties (Südhof, 2017). Despite extensive studies, however, most fundamental features of neurexins remain incompletely understood. For example, it is unknown whether different neurexins perform similar or distinct functions in synapses, or whether neurexin alternative splicing is generally of physiological significance. Here, we have asked two basic questions: First, do different neurexins mediate uniform and redundant functions at a defined synapse, or do they perform distinct and independent actions? Second, are these functions of neurexins coordinately regulated by alternative splicing at SS4? We addressed these questions at one particular synapse, the excitatory output synapse of CA1 pyramidal neurons on subiculum neurons, which represents the major efferent pathway of the hippocampus. In our experiments, we used a rigorous conditional genetic approach that enables control over alternative splicing of endogenous neurexins, and coupled this approach to in vivo manipulations and acute slice physiology.

Our results suggest two major conclusions that may change the way we think about neurexins. First, although neurexins exhibit a high degree of sequence homology, the three neurexin genes and their alternatively spliced products perform profoundly different functions. Given the similarity especially between Nrxn1 and Nrxn3 (Treutlein et al., 2014), this result was unexpected. It indicates that the functions of neurexins are more diverse than envisioned, that their non-coordinated alternative splicing (Fuccillo et al., 2015) can independently regulate these diverse functions, and that neurexins must be intrinsically different despite similar ligand-binding properties. As a result, it may no longer be appropriate to talk about ‘neurexins’ in a generic term, but more fitting to always refer to individual neurexins.

Second, alternative splicing of presynaptic Nrxn1 and Nrxn3 robustly regulates postsynaptic glutamate receptor composition, but acts on distinct receptor types in opposite directions: Whereas Nrxn1^SS4+ selectively enhances NMDAR-signaling, Nrxn3^SS4+ equally selectively suppresses AMPAR-signaling. Our results thus suggest that different from previous ideas (Aoto et al., 2013), the decrease in AMPAR EPSCs in Nrxn3 SS4+ cKI mice is not due to a loss of Nrxn3^SS4− function, but rather to a gain of Nrxn3^SS4+ function. Since the epigenetic state of the presynaptic neuron determines neurexin alternative splicing (Ding et al., 2017), this state controls the glutamate neurotransmitter response of the postsynaptic neuron. In other words, activity-dependent epigenetic changes in a presynaptic neuron can regulate a postsynaptic receptor response, representing a non-canonical plasticity mechanism. Because alternative splicing of different neurexins is not coordinately regulated in a neuron (Fuccillo et al., 2015), Nrxn1-dependent regulation of NMDARs and Nrxn3-dependent regulation of AMPARs operate independent of each other, expanding the long-term plasticity of synaptic signaling. This finding is particularly relevant because activation of NMDARs and AMPARs produces qualitatively and quantitatively different signals, and their relative activity during synaptic transmission is a major determinant of how a synapse, and thereby a circuit, processes information (Silver, 2010).

Our observations also raise a number of questions. How is it possible that the same splice variants of two different presynaptic neurexins, Nrxn1^SS4+ and Nrxn3^SS4+, produce distinct effects on different postsynaptic glutamate receptors? Presumably Nrxn1^SS4+ and Nrxn3^SS4+ act by differential binding to postsynaptic ligands. However, deciphering which ligands are involved in regulating NMDARs and AMPARs will be a major challenge. Neuroligins and cerebellins are the only known ligands for SS4+ splice variants of neurexins (Comoletti et al., 2006; Elegheert et al., 2017; Joo et al., 2011), but neuroligins do not appear to influence AMPAR-EPSCs (Jiang et al., 2017). Cerebellins, in turn, are poorly expressed in the hippocampus (Seigneur and Südhof, 2017) and may act more in synapse formation than in the regulation of glutamate receptors (Yuzaki, 2017). Thus, possibly unknown neurexin ligands may be involved.

Another major question regards the mechanism of neurexin alternative splicing. It is puzzling that although compelling data implicate the STAR family of splicing factors in the alternative splicing of all neurexins (Ehrmann et al., 2013; Traunmüller et al., 2016), different neurexins are not alternatively spliced coordinately at the same site in a given cell (Fuccillo et al., 2015). Thus, how neurexin alternative splicing is mediated mechanistically, and how it is dynamically regulated by activity (Ding et al., 2017), remains a major question. A third question concerns the functions of neurexins in other synapses, and the generality of the specific function we described here for hippocampal CA1➔subiculum synapses. Abundant evidence indicates that α- and β-neurexins perform distinct roles (e.g., see Missler et al., 2001; Anderson et al., 2015; Chen et al., 2016). Clearly the function we described here for SS4 alternative splicing represents only one particular role of neurexins in a specific synapse, as already evidenced by our studies on the role of Nrxn3 in olfactory bulb synapses (Aoto et al., 2015) and by our experiments showing that triple Nrxn123 cKOs caused a broad and heterogeneous range of phenotypes (Chen et al., 2016). What other functions neurexins perform in other synapses, and how these might be regulated is yet another major issue that will be a focus of future experiments.

STAR* METHODS

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Thomas C. Südhof (tcs1@stanford.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Mouse Generation and Husbandry

Nrxn1-SS4+ and Nrxn2-SS4+ conditional knockin (cKI) mice for Nrxn1 and Nrxn2 were generated using the approach we previously described for Nrxn3 (Aoto et al., 2013). Briefly, the exons encoding the alternatively spliced sequences in the Nrxn1 and Nrxn2 genes contain non-canonical splice acceptor sequences (Tabuchi and Südhof, 2002) that were converted to canonical splice acceptor sequences by homologous recombination in ES cells, rendering these SS4 exons constitutively spliced in the Nrxn1 or Nrxn2 mRNAs. At the same time, the alternatively spliced exon was flanked with loxP recombination sequences to allow Cre-dependent excision of the exon, which renders SS4 in Nrxn1 or Nrxn2 mRNAs constitutively spliced out (Figure 1A, 5A). Triple Nrxn123 SS4+ cKI mice (Nrxn123-SS4+ cKI mice) were generated by crossing the Nrxn1-, Nrxn2-, and Nrxn3-SS4+ cKI lines (Figure S5B). All mice were maintained on a mixed C57BL/6/SV129 (wild type) background. Primers (IDT) were used for genotyping as follows: Nrxn1-SS4+, forward: 5’-AGACAGACCCGAACAACCAA-3’, reverse: 5’-TGCTAGGCCTATTTCAGATGCT-3’; Nrxn2-SS4+, forward: 5’-ACCCTTGGGGGTGAGAGTAA-3’, reverse: 5’-TCTTTGGGATGGTGAGGAAG-3’; Nrxn3-SS4+, forward: 5’-CTCCAACCTGTCATTCAAGGG-3’, reverse: 5’-CTACGGGCCGGTTATATTTG-3’. All mouse studies were performed according to protocols approved by the Stanford University Administrative Panel on Laboratory Animal Care. All procedures conformed to NIH Guidelines for the Care and Use of Laboratory Animals and were approved by the Stanford University Administrative Panel on Laboratory Animal Care. All mice were housed in the Stanford animal facility under supervision of the Stanford animal care unit; all mice were healthy and not kept in a sterile facility. In all studies, we examined littermate male or female mice, except for the comparisons of WT mice with SS4+ vs. SS4− variants of the various neurexins in which no WT littermate mice could possibly be obtained, and an age- and sex-matched control was used. All mice had a mixed hybrid genetic background containing Sv129, C57/Bl6, and CD1 components.

Primary Cultures

Hippocampal neurons were cultured from newborn male and female WT (C57BL/6J) or Nrxn1-SS4+ cKI mice.

METHOD DETAILS

Experimental design

Viral manipulations were carried out on randomized sex- and age-matched mice or on cultures obtained from dissociated brain tissue that was obtained from both male and female mice. Since no prior information on effect sizes existed, no power analyses for estimating sample sizes were possible and standard approaches of testing for phenotypes were used as described (Aoto et al., 2013). All experiments were performed on anonymized samples, and no data were excluded.

mRNA Measurements

mRNA was prepared from hippocampal cultures that had been infected as indicated with lentiviruses at DIV4–5 and harvested at DIV14–16, or from brain tissue directed from the CA1 region of P35–42 mice. RNA extraction was taken by using Qiagen RNeasy kit according to manufacturer’s protocol (QIAGEN) and quantified using an ND-1000 spectrophotometer (NanoDrop, ThermoScientific). Quantitative RT-PCR was performed using the VeriQuest Probe One-Step qRT-PCR Master Mix (Affymetrix) based on the manufacturer’s instructions, and reactions were carried out and quantified using a 7900HT Fast RT-PCR instrument (Applied Biosystems). Expression levels were normalized to GAPDH (Figure 1B) or β-actin (All other figures) as endogenous internal control. The following PrimeTime qPCR Assays (IDT) were used (shown as gene, primerl, probe, primer2): Nrxn1β, TGGCCCTGATCTGGATAGTC, ACCACAT CCACCATTT CCAT, AAT CTGTCCACCACC TTTGC; Nrxn1-SS4+, GTTGATGAATGGCTACTCGACAA, CACAATCTTCAATAGCCAAG CAACCATAA, G CCATATT CAGAACTTT CAAG CC; Nrxn1-SS4−, TACCCTGCAGGGCGTCA, CACAAT CTT CAATAG CCAAGCAACCAT AA, G CCAT ATT CAGAACTTT CAAG CC; Nrxn2-SS4+, GGAAACTTTGATAACGAGCGC, AGACAGAGAATCCCCTACCGGCT, CCCCTATCTTGATGG CAGC; Nrxn2-SS4−, TACCCGGCAGGACGCCA, CCAGGTGTCCGGCCTCTACT, CCAGTACCTT GAGCCCATT G; Nrxn3β, CACCACTCTGTGCCTATTT C, TCTATCGCTCCCC TGTTTCC, GGCCAGGTAT AG AG GAT GA; Nrxn3-SS4+, TCCCCTTCAAATATAACCGGC, AGCTAACCATCTTCAACACCCAGGC, ACGTCCTTTGTCCTTTCCTC; Nrxn3-SS4−, AGAGAACT CCTGTCAAT GAT G, CAAATACCACGTTGTGCG CTT CACCAG GAAT G, TTAGCTGCCGGCCTGTA. For semi-quantitative RT-PCR measurements of neurexin SS4 alternative splicing, total RNA was extracted using TRIzol and cDNA was synthesized using the SuperScript III First-Strand Synthesis System (Invitrogen) according to the manufacturer’s instructions. PCR primers to detect Nrxn-SS4 isoforms (Forward, reverse): Nrxn1SS4, CTGGCCAGTTATCGAACGCT, GCGATGTTGGCATCGTTCTC; Nrxn2SS4, CAACGAGAGGTACCCGGC, TACTAGCCGTAGGTGGCCTT; Nrxn3SS4, ACACTTCAGGTGGACAACTG, AGTTGACCTTG GAAGAGACG; β-actin, TTGTTACCAACTGGGACGACA, TCGAAGTCTAGAGCAACATAGC.

Immunoblotting

It was performed as described previously (Seigneur et al, 2018; Zhang et al, 2015). Briefly, DIV14 hippocampus culture were collected with pre-chilled RIPA buffer (10 mM Tris- Cl, pH 8.0, 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS, 140 mM NaCl) containing 1 mM PMSF protease inhibitor cocktail (Roche Applied Science) for 1h at 4°C. Samples were then centrifuged at 20,000 × g for 10 min at 4°C to get the supernatants. Protein were loaded into a Criterion TGX 4–20% Tris-Glycine precast gel (Bio-Rad) and separated via SDS-PAGE at 100 V for ~1.3h and then transferred onto membrane using the Trans-Blot Turbo transfer system (Bio-Rad). Membranes were then blocked with 5% milk in TBS containing 0.1% Tween 20 (TBST) for 1h at RT, and then incubated in primary antibody overnight at 4°C. Membranes were washed 3X with TBST, and then incubated in fluorescent labeled secondary antibodies (donkey anti-rabbit IR dye 680/800CW, 1:10000; donkey anti-mouse IR dye 680/800CW, 1:10000; and donkey antiguinea pig IR dye 680RD, 1:10000; LI-COR Bioscience). Membranes were scanned using an Odyssey Infrared Imager and analyzed with Odyssey software (LI-COR Biosciences). Intensity value for each protein was first normalized to actin and then normalized to the control sample. The antibodies were used as follow: Neuroligin-1 (1:500, mouse monoclonal antibody 4F9), monoclonal mouse anti-β-actin (1:10000; A1978 Sigma-Aldrich; RRID: AB_476692), polyclonal rabbit anti- PSD95 (1:500; L667 Sudhof lab), polyclonal rabbit anti-Synapsin (1:1000; E028 Sudhof lab), CASK (1:1000mouse monoclonal antibody 610782, BD Transduction Laboratories™), Neurexin (1:500, rabbit polyclonal antibody, G394), GAD65 (1:500, mouse monoclonal antibody, DSHB), Synaptotagmin 1(1:1000, mouse monoclonal antibody CL41.1), polyclonal rabbit anti-vGAT (1:1000; 131003 Synaptic Systems; RRID: AB_887869), polyclonal guinea pig anti-vGluT1 (1:1000; AB5905 Millipore; RRID: AB_2301751), polyclonal guinea pig anti-vGluT2 (1:1000; AB2251 Millipore; RRID: AB_1587626).

DNA constructs and Viruses

Cre-eGFP and ΔCre-eGFP constructs (Aoto et al., 2013) and AAV-DJ vector (Xu et al., 2012) were described previously. Expression vectors for full-length Nrxn1β^SS4+, Nrxn1β^SS4−, Nrxn3β^SS4+, and Nrxn3β^SS4− were previously described (Aoto et al., 2013), and recloned into AAV-DJ vector for in vivo overexpression as a fused mRNA composed of a 5’ eGFP containing a nuclear localization sequence followed by a p2A sequence and the various β-neurexin sequences. Viral titers were determined by qPCR, and the expression levels mediated by the viruses were quantified in microdissected brain tissue (Figure S7).

Electrophysiology

For electrophysiological recordings from cultured neurons, hippocampal neurons were cultured from WT or Nrxn1-SS4+ cKI mice, infected on DIV4–5 with lentiviruses encoding Cre-eGFP (to induce the Nrxn1^SS4− genotype) or ΔCre-eGFP (to retain the Nrxn1^SS4+ genotype), and analyzed at DIV14–16 as described (Aoto et al., 2013). Cultures were continuously superfused with ACSF solution (in mM): 120 NaCl, 2.5 KCl, 1 NaH₂PO₄, 26.2 NaHCO₃, 2.5 CaCl₂, 1.3 MgSO₄-7 H₂O, 11 D-Glucose, ~290 mOsm. mEPSCs and EPSCs were recorded with an internal solution containing (in mM): 117 Cs-methanesulfonate, 15 CsCl, 8 NaCl, 10 TEA-Cl, 0.2 EGTA, 4 Na₂-ATP, 0.3 Na₂-GTP, 10 HEPES, pH 7.3 with CsOH (~300 mOsm). Evoked EPSCs were recorded in 100 μM picrotoxin and 50 μM D-(−)-2-amino-5-phosphonopentanoic acid (D-APV) for AMPAR, 100 μM picrotoxin and 10 μM 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) for NMDAR. Paired-pulse ratios of NMDAR EPSCs were monitored using interstimulus intervals of 20–2000 ms. mIPSCs were recorded with an internal solution containing (in mM): 145 CsCl, 5 NaCl, 10 HEPES-CsOH pH 7.3, 10 EGTA, 4 Na2-ATP, and 0.3 Na₂GTP (~300 mOsm). mEPSCs and mIPSCs were recorded in 0.5 μM tetrodotoxin (TTX) plus either 100 μM picrotoxin for mEPSCs or 10 μM CNQX and 50 μM D-APV for mIPSCs, respectively. 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 bipolar stimulating electrode (A-M Systems, Carlsborg, WA) placed 100–150 μm from the soma of recorded neurons, and controlled by a Model 2100 Isolated Pulse Stimulator (A-M Systems, Inc.) synchronized with the Clampfit 10 data acquisition software (Molecular Devices). Synaptic currents were monitored with a Multiclamp 700B amplifier (Molecular Devices). Data were collected at 10 kHz and filtered with a lowpass filter at 2 kHz. For all experiments, the experimenter was blind to the recording condition.

Electrophysiological recordings from the subiculum in acute hippocampal slices were essentially performed as described (Aoto et al., 2013; Xu et al., 2012). In brief, slices were prepared from WT and mutant mice at P35–42, which is 2–3 weeks after stereotactic infections with AAVs encoding Cre, ΔCre, or various β-neurexins. Horizontal brain slices (300 μm thickness) were cut in a high sucrose cutting solution containing (in mM) 85 NaCl, 75 sucrose, 2.5 KCl, 1.3 NaH₂PO₄, 24 NaHCO₃, 0.5 CaCl₂, 4 MgCl₂ and 25 D-glucose. Slices were equilibrated in ACSF at 31.5 °C for 30 min, followed by an hour at room temperature. Slices were then transferred to a recording chamber containing ACSF as described above, but maintained at 30.5°C. To induce evoked synaptic responses in the subiculum, a nichrome stimulating electrode was placed at the most distal portion of hippocampal CA1 region as illustrated in Figure 2A. The firing type of subiculum neurons (burst- spiking vs. regular-spiking) was identified by injecting a depolarizing current immediately after breaking in and monitoring action potential patterns in current-clamp mode (Figure S2C). AMPAR- EPSC input/output curves were measured in whole-cell voltage-clamp mode (holding potential = −70 mV) by using an internal solution containing (in mM): 137 K-gluconate, 5 KCl, 10 HEPES, 4 ATP-Mg₂, 0.5 GTP-Na₂, 10 phosphocreatine, 0.2 EGTA, pH 7.2 with KOH. AMPAR/NMDAR ratios were measured using an internal solution containing (in mM) 117 Cs-methanesulfonate, 15 CsCl, 8 NaCl, 10 TEA-Cl, 0.2 EGTA, 4 Na2-ATP, 0.3 Na2-GTP, 10 HEPES, pH 7.3 with CsOH (~300 mOsm). NMDAR input/output curves, LTP measurement and mEPSCs (holding potentials = −70 mV for AMPAR-EPSCs, +40 mV for NMDAR-EPSCs, and +60 mV NMDAR-mEPSCs) were recorded with an internal solution containing (in mM): 135 Csmethanesulfonate, 8 NaCl, 2 MgCl₂, 0.25 EGTA, 4 Mg-ATP, 0.3 Na2-GTP, 10 HEPES, 5 Na₂-Phosphocreatine, pH 7.3 with CsOH (~300 mOsm). mIPSCs were measured by using 140 CsCl, 0.5 EGTA, 4 Mg-ATP, 0.3 Na2-GTP, 10 HEPES, 5 Na₂-Phosphocreatine, 2 QX-314, pH 7.3 with CsOH (~300 mOsm) with a holding potential = −70 mV. For all evoked NMDAR-EPSCs current were clamped at +40 mV. All recordings in the hippocampus were performed in the presence of 100 μM picrotoxin for AMPAR-EPSCs and LTP, 100 μM picrotoxin and 10 μM CNQX for NMDAR-EPSCs, 100 μM picrotoxin and 0.5 μM TTX for mEPSCs, and 10 μM CNQX, 50 μM D-APV and 1 μM Strychnine for mIPSCs. Paired-pulse ratios were monitored with interstimulus intervals of 20–2000 ms. LTP was induced by four tetani of 100 Hz stimulus trains applied for 1 s with 10 s intervals under voltage-clamp mode (holding potential = 0 mV). Pre-LTP (averaging last 5 mins as baseline) and post-LTP (averaging the last 5 mins) were recorded at 0.1 Hz. Paired-pulse ratios were measured with 40 ms interstimulus intervals before and after LTP. Measurements of the AMPAR/NMDAR were performed in 100 μM picrotoxin at holding potentials of −70 mV (AMPAR-EPSCs) or +40 mV (NMDAR-EPSCs, quantified at 50 ms after the stimulus). All data were analyzed with Igor software (WaveMetrics). Miniature events were handpicked with a threshold of 5 pA by using the Igor program (Dai et al., 2015).

Puffing applications of AMPA (R-S AMPA hydrobromide, Tocris Bioscience, in the presence of picrotoxin, D-APV and TTX), NMDA (Tocris Bioscience, in the presence of picrotoxin, CNQX and TTX) and GABA (Tocris Bioscience, in the presence of D-APV, CNQX and TTX) were performed with 10 psi for 5–200 ms by using Picospritzer III (Parker Instrumentation). The total charge was calculated within 10 s, 20 s, 40 s from puff application. The slope was calculated from the data obtained between the pre-puffing baseline to 100 ms for AMPA and to 50 ms for NMDA and GABA.

LTP measurements in CA1 pyramidal neurons were also performed in acute hippocampal slices obtained as described above using whole-cell patch-clamp recordings with the same extracellular and intrapipette solutions that included extracellular picrotoxin (100 μM). Schaffer-collateral axons were stimulated extracellularly, and LTP was induced either by theta-burst stimulation (four trains were comprised of ten bursts (each burst included four stimuli at 100 Hz) at 5 Hz with 10s interval) or by two applications of 100 Hz stimuli separated by 10 s.

Stereotactic Injections

Stereotactic injections of AAVs into mice at P21 were performed essentially as described (Aoto et al., 2013 and 2015). Mice were anesthetized with ketamine-medetomidine, and AAVs were injected using a stereotaxia instrument (David Kopf) and a syringe pump (Harvard Apparatus) with ~0.85 pl (CA1 region) or 0.4 pl (subiculum) of concentrated virus solution (10⁸⁻⁹ TU) at a slow rate (0.1ul/min) into the CA1 region of the intermediate hippocampus (Bregma coordinates (mm): AP: −3.1, ML: ± 3.4, DV: −2.5) or into the subiculum (Bregma coordinates (mm): AP: 3.3, ML: ± 3.3, DV: 2.5). After 2–3 weeks, AAV-mediated expression was confirmed by the presence of eGFP. Images (Figure 2A) were taken using the Olympus Fluoview Imaging software (Olympus). The expressed fluorescence of all slices for physiology was confirmed under a fluorescence microscope (Olympus).

Immunocytochemistry

For hippocampal cultures, cells were fixed with 4% paraformaldehyde (PFA) + 4% sucrose for 15 mins, then washed 3x times 1x phosphate buffered saline (PBS) and incubated in blocking buffer (0.3 % Triton X-100 and 5% goat serum in PBS) for 1 h at RT (Aoto et al., 2013). Cover slips were incubated overnight at 4 °C with primary antibodies diluted in blocking buffer (anti-vGluT1, 1:500, guinea pig, Millipore; anti-GAD65, 1:1000, mouse, DSHB; anti-MAP2, 1:500, rabbit, Millipore), followed by treating with secondary antibodies (1:1000, Alexa 488, 545, 633, Invitrogen) at room temperature for 1 hour after washing 3x times with PBS. In the end, cover slips were washed 3x times again and mounted with Fluoromount-G (SouthernBiotech; Birmingham, AL). Image acquisition and quantification were performed as described (Aoto et al., 2013). In brief, cells were chosen from three or more independent cultures. Images were taken from at least three coverslips per experiment. Fluorescent images were acquired with FV1000 BX61WI laser-scanning confocal microscope at room temperature by using an Olympus (Tokyo, Japan) Plan Apochromat 60x oil objective (NA: 1.42; WD: 0.15) set to 3x zoom with sequential acquisition, 3 frames averaging, and 1024×1024 pixel resolution. Within the same experiment, the same settings were always used. 7–12 optical sections (0.5 μm) were used and maximum pixel intensity projections were created. For quantifications, images were thresholded by intensity to exclude background signals and the number of puncta (size ranged from 0.1–4.0 μm²) was quantified. For each experiment, at least 15 cells per condition were analyzed to obtain the mean and S.E.M. Data shown in Figure 1F and 1G represent the average of the mean values from at least 3 independent experiments. Conditions of image-capture and quantification were all blind to researcher.

For hippocampal cryosections, experiments were performed essentially as described (Zhang et al., 2015). Mice were anesthetized with isoflurane, perfused with 10 ml PBS and followed by 30 ml 4% PFA in 1x PBS via a perfusion pump (2 ml/min). Whole brains were dissected out and kept in PFA for 6 hours, then post-fixed in 30% sucrose (in 1× PBS) for 24 h-48 h at 4°C. Horizontal brain sections (30 μm) were collected at −20C with a cryostat (Leica CM1050). Sections were washed with PBS and incubated in blocking buffer (0.3 % Triton X-100 and 5% goat serum in PBS) for 1 h at RT, and incubated for overnight at 4 °C with primary antibodies diluted in blocking buffer (anti-vGluT1, 1:500, guinea pig, Millipore and anti-MAP2, 1:500, rabbit, Millipore). Sections were washed 10 mins for 3 times in 1x PBS, followed by treating with secondary antibodies (1:1000, Alexa 405, 545, 633, Invitrogen) at 4 °C overnight, and washed 10 mins for 3 times with 1x PBS. All of procedures were under agitation. All sections were then mounted on superfrost slides and covered with Fluoromount-G as described. Serial confocal z-stack images (0.5 μm step for 10 μm at 1024 × 1024 pixel resolution) were acquired using a Nikon confocal microscope (A1Rsi) with a 60x oil objective (PlanApo, NA1.4). All acquisition parameters were kept constant among different conditions within experiments. For data analyze, MaxProjection were performed for each image, and averaged intensity (mean ± S.E.M) of vGlut1 from whole area of subiculum (object size range 0.06–0.12 mm²). 24–25 sections were used for each condition (n≥3 animals per condition) and were analyzed with Nikon analysis software.

Mouse behavior tests

Cohorts of Nrxn1-SS4+ cKI mice were injected stereotactically into the CA1 region at P21 with AAVs encoding Cre or ΔCre. Mice were handled daily for 5 days prior to behavioral experiments starting at 4 weeks after AAV infections. Mice were maintained with a normal 12/12 hr daylight cycle, and analyzed in the assay sequence and at the time shown in Figure S3A. Open field test. Each injected mouse was placed in an open white box to run 5 mins and analyzed with Viewer III software (Biobserve). Barnes maze tests were performed as described (Patil et al., 2009; Sunyer et al., 2007; Anderson et al., 2017). In brief, visual cues were inserted in a Barnes maze as indicated in Figure 4D. Mice were placed in the middle of maze and trained by performing strong light to trigger them to find and jump into target hole with 4 trails/15mins inter-trail interval daily for 4 days, and tested 24 hours later in 90 s trials (Figure S3C). Between each trail, we cleaned the maze using a solution with 1% Virkon-S Disinfectant to avoid olfactory cues. Primary errors were calculated as entering the wrong hole before reaching the target hole. The latency was the time needed to reach the target hole. Track length was the total length traveled until jumping into target hole during training or during test in 90 s. The time spent in each hole area (20 holes) was analyzed as indicated in Figure 4D. Passive chamber avoidance test. We modified the passive avoidance test (Cimadevilla et al., 2001; Ambrogi et al., 1984; Qiao et al., 2014) by introducing visual cues in the two chambers to test spatial memory. The Protocol was performed as: Two chambers (A and B) were designed with different visual cues (Figure S3B) under dim light with a gate between them. Chamber B has foot shock with electronic current (intensity: 0.15 mA, duration: 2s). Mice can explore A and B freely. At the training day, mice were put in chamber A. Once they went to chamber B, they got foot shock after 2 s delay. In this case, they returned to chamber A immediately. This is one trial of learning. It might come as another trail, once they visited chamber B again, then they would get foot shock again. This training process was completed until mice were able to stay in chamber-A more than 2 mins. Between each mouse, chambers were cleaned by 70% ethanol solution. During training, mouse learned to stay at chamber-A, thus latency was getting to be 120s and activity was getting down due to the shock fear memory. After 1 day and 7 days, they were tested by putting back into chamber-A to record latency to chamber-B, activity in chamber-A and entry times from A to B in 2 mins. Upon this approach, two groups of AAV-Cre and AAV-ΔCre injected mice were tested. All behavior assays were carried out and analyzed by researchers blindly.

QUANTIFICATION AND STATISTICAL ANALYSIS

All data are shown as means ± SEM, numbers in bars or labels present number of cells/mice or cells/cultures analyzed, with statistical significance (* = p<0.05, ** = p<0.01 and *** = p<0.001) determined by Student’s t-test or single-factor analysis of variance (ANOVA). Non-significant results (p>0.05) are not specifically identified.

DATA AND SOFTWARE AVAILABILITY

All data reported here are published with this paper as a supplementary excel file (Supplemental Table S1). All software used is listed in the Key Resources table. Primer sequences are also included in Supplementary Table S2.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
anti-vGluT1	Millipore	Cat. No. AB5905
anti-vGlutT2	Millipore	Cat. No. AB2251
anti-GAD65	DSHB	Cat. No. mGAD6-a
anti-MAP2	Millipore	Cat. No. AB5622
anti-Synaptotagmin 1	Südhof lab	CL41.1
anti-Neurexin	Südhof lab	G394
anti-CASK	BD Transduction Laboratories	Cat. No. 610782
anti-PSD95	Südhof lab	L667
anti-Synapsin	Südhof lab	E028
anti-vGAT	Synaptic Systems	Cat. No. 131003
anti-Neuroligin-1	Südhof lab	4F9
anti-β-actin	Sigma	Cat. No. A1978
Bacterial and Virus Strains
Lenti-hSyn-Cre-eGFP	Aoto et al., 2013	N/A
Lenti-hSyn-eGFP	Aoto et al., 2013	N/A
pAAV-hSyn-Cre-eGFP	Aoto et al., 2015	N/A
pAAV-hSyn-eGFP	Aoto et al., 2015	N/A
pAAV-hSyn-eGFP-p2A-Nrxn1 βSS4+/−	This paper	N/A
pAAV-hSyn-eGFP-p2A-Nrxn3βSS4+/−	This paper	N/A
Chemicals, Peptides, and Recombinant Proteins
CNQX	Tocris	Cat. No. 0190
Picrotoxin	Tocris	Cat. No. 1128
D-APV	Tocris	Cat. No. 0106
TTX	Fisher Scientific	Cat. No. 50–753-2807
(RS)-AMPA hydrobromide	Tocris	Cat. No. 1074
NMDA	Tocris	Cat. No. 0114
GABA	Tocris	Cat. No. 0344
Strychnine hydrochloride	Sigma	Cat.No.S8753
QX-314 bromide	Tocris	Cat.No.1014
Experimental Models: Organisms/Strains
Mouse: C57BL/6J wildtype	The Jackson Laboratory	Jax Stock no: 000664
Mouse: Nrxn1-SS4+, Nrxn2-SS4+ or Nrxn123-SS4+ cKI	This paper	N/A
Mouse: Nrxn3 SS4+ cKI	Aoto et al., 2013	N/A
Oligonucleotides
Primers for all genotyping, see Supplemental Table S2	This paper	N/A
Primers for all qRT-PCR, see Supplemental Table S2	This paper	N/A
GAPDH	IDT	Cat# 4352932–0809025
b-actin	IDT	Mm.PT.56a. 33540333
Software and Algorithms
Clampfit 10	Molecular Devices	https://www.moleculardevices.com/products/axon-patch-clamp-system/acquisition-and-analysis-software/pclamp-software-suite
Igor software	Wavemetrics	https://www.wavemetrics.com/downloads
Viewer III	Bioserve	http://www.biobserve.com/behavioralresearch/products/viewer/
SigmaPlot	Systat Software	https://systatsoftware.com/sp/download.html

Supplementary Material

2

Supplemental Table S1: Data for all figures, Related to Figures 1–8 and Supplemental Figures 1–8.

3

Supplemental Table S2: Primers for all genotyping and qRT-PCR, Related to STAR Methods.