The Perils of Navigating Activity-Dependent Alternative Splicing of Neurexins

Neurexins are presynaptic cell-adhesion molecules essential for synaptic function that are expressed in thousands of alternatively spliced isoforms. Recent studies suggested that alternative splicing at splice site 4 (SS4) of Nrxn1 is tightly regulated by an activity-dependent mechanism. Given that Nrxn1 alternative splicing at SS4 controls NMDA-receptor-mediated synaptic responses, activity-dependent SS4 alternative splicing would suggest a new synaptic plasticity mechanism. However, conflicting results confound the assessment of neurexin alternative splicing, prompting us to re-evaluate this issue. We find that in cortical cultures, membrane depolarization by elevated extracellular K+-concentrations produced an apparent shift in Nrxn1-SS4 alternative splicing by inducing neuronal but not astroglial cell death, resulting in persistent astroglial Nrxn1-SS4+ expression and decreased neuronal Nrxn1-SS4– expression. in vivo, systemic kainate-induced activation of neurons in the hippocampus produced no changes in Nrxn1-SS4 alternative splicing. Moreover, focal kainate injections into the mouse cerebellum induced small changes in Nrxn1-SS4 alternative splicing that, however, were associated with large decreases in Nrxn1 expression and widespread DNA damage. Our results suggest that although Nrxn1-SS4 alternative splicing may represent a mechanism of activity-dependent synaptic plasticity, common procedures for testing this hypothesis are prone to artifacts, and more sophisticated approaches will be necessary to test this important question.

Several studies reported that alternative splicing of neurexins is activity-dependent (Gorecki et al., 1999;Rozic-Kotliroff and Zisapel, 2007;Iijima et al., 2011;Rozic et al., 2011;Ding et al., 2017). Gorecki et al., were the first to investigate the regulation of neurexin alternative splicing using kainic acid (KA) and pentylenetetrazole (PTZ) stimulation, but concluded that alternative splicing of Nrxn2 at SS1-SS4 was not activitydependent (Gorecki et al., 1999). Subsequently, Rozic-Kotliroff and Zisapel (2007) and Rozic et al. (2011) showed that Nrxn2 and Nrxn3 alternative splicing at SS3 is modulated by neuronal depolarization induced by high KCl (50 mM). Rozic et al. (2011) also showed that altnernative splicing of all three neurexins at SS4 is modulated in the rat hippocampus by fear conditioning (Rozic-Kotliroff and Zisapel, 2007;Rozic et al., 2011). Furthermore, Iijima et al. (2011) used mild KCl depolarization (25 mM), KA (50 µM), or electrical field stimulation in cultured mouse cerebellar granular cells to demonstrate an increase in Nrxn1 and Nrxn2 SS4-mRNA as a function of activity, whereas Nrxn3 mRNA was unaffected. Iijima et al. (2011) also found a significant increase in Nrxn1-SS4-mRNA abundance when a cerebral hemisphere was exposed to KA through focal injection. In a more extensive recent study, finally, Ding et al. (2017) showed that neuronal activity induces histone modifications that in turn regulate Nrxn1 SS4 alternative splicing. In this astonishing study using KCl-induced depolarization, optogenetic stimulation, and fear conditioning, Ding et al. (2017) uncovered robust changes in Nrxn1 SS4 alternative splicing in mice by various forms of neuronal activity, and confirmed these changes using RNAseq experiments.
Altogether, especially the more recent papers paint a compelling picture of activity-dependent regulation of neurexin alternative splicing at SS4, the only site of neurexin alternative splicing that has been shown to be physiologically relevant. These results are of great potential significance because they suggest a novel mechanism of synaptic plasticity. However, the various papers report quite variable results and use diverse treatment conditions. Moreover, KCl and KA treatment are known to be neurotoxic (Sun et al., 1992;Cheng and Sun, 1994;Pollard et al., 1994;Kasof et al., 1995;Simonian et al., 1996;Cheung et al., 1998;Takahashi et al., 1999;Gluck et al., 2000;Milatovic et al., 2002;Le Duigou et al., 2005;Wang et al., 2005;Rienecker et al., 2020), thus raising the questions of validity. To address these issues, we here re-evaluated the activity-mediated regulation of Nrxn1 SS4 alternative splicing using simple in vitro and in vivo paradigms that replicate those employed by Iijima et al. (2011) and Ding et al. (2017), and we extended these experiments to Nrxn2 and Nrxn3 SS4 splice forms. Our data suggest that most, if not all, apparent activity-dependent alternative splicing of Nrxn1 at SS4 induced by high KCl or by KA is produced by the loss of Nrxn1 from neurons that are damaged by the unphysiological KCl or KA stimulation. Since astrocytic Nrxn1-SS4+ remains when neuronal Nrxn1-SS4-is lost, the neuronal cell death creates an apparent shift in Nrxn1 SS4 alternative splicing, thus eliciting the impression of regulated alternative splicing. Although our data do not exclude the possibility of activity-dependent Nrxn1 alternative splicing under other conditions, they suggest that previous reports of such events may have been confounded by the neurotoxic effects of the stimulation paradigms that were employed.

Mice
All animal experiments performed in this study were conducted in accordance with protocols approved by the Administrative Panel on Laboratory Animal Care at Stanford University. Wild type mice used in this study were outbred CD-1 or C57BL/6J [Crl: CD1(ICR) and C57BL/6J mice obtained from Charles River]. Baf53b-Cre [also known as Actl6b-Cre; Tg(Actl6b-Cre)4092Jiwu/J; Stock No: 027826], and RiboTag [B6J.129(Cg)-Rpl22 tm1.1Psam /SjJ, Stock No: 029977] mice were obtained from the Jackson Laboratory. Female homozygous Baf53b-Cre mice were used to breed with RiboTag males in order to avoid male germline recombination of RiboTag allele. Genotyping was performed using primers prescribed by the animal supplier.
For KCl-mediated depolarization, the final depolarization buffer consisted 31% depolarization buffer (170 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM HEPES), 50% old neuronal conditioned medium and 19% fresh neuronal medium. For 10 to 50 mM KCl treatment experiments, KCl concentration in the depolarization buffer was adjusted accordingly. DIV11 neuronal cultures were treated with the final depolarization solution or NaCl buffer (170 mM NaCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM HEPES) for 10 min and washed and replaced with old neuronal medium. The culture plate was kept in an incubator and samples were collected at various time points. For chronic KCl exposure, depolarization buffer was left continuously until the samples were collected at different time points.

Pure Cortical Glial Culture
Embryos at E16.5 day from WT CD-1 mice used for culturing pure cortical glial cells. Cortical cells were subjected to harsh trituation to ensure low number of viable neurons (McCarthy and De Vellis, 1980). Cells were plated with Dulbecco's Modified Eagle Medium (DMEM) with 10% FBS. Media was changed after 24 h. At day 5, when glial cultures were ∼80% confluency, they further purified by trypsnizing with 0.05% trypsin for 5 min, centrifugation for 5 min at 1,200 RPM, resuspending and plating. This is considered passage 1 where residual neuronal cells were eliminated. On day 10, the cells were subjected to passage 2 and collected at day 14 for subsequent analysis.

Systemic and Stereotactic Administration of Kainic Acid
For systemic administration of KA, P30-45 WT CD-1 or Baf53b x RiboTag mice were used. 20 mg/kg of Kainic Acid (KA) monohydrate (Sigma, K0250) dissolved in PBS was delivered via intraperitoneal injection. Mice were monitored for seizures throughout the time course. Mice were euthanized using isoflourane and decapitated to harvest the brain at indicated time points. For IHC, brain were embedded in Optimal Cutting Temperature (OCT) solution on dry ice and stored in −80 • C until further usage. For RNA extraction and immunoprecipitation of RiboTag-mRNA Cortex and hippocampus were dissected out and snap-frozen in liquid nitrogen and stored in −80 • C until processing.
Focal administration of KA using sterotactic surgery was performed as described previously (Iijima et al., 2011). Briefly, for anesthetizing WT CD-1 mice (P30-P60), the stock solution of tribromoethanol was made by dissolving 5 g into 5 mL Tamyl alcohol. The working solution was made by diluting to 80-folds into PBS. 0.2 ml working solution per 10 g body weight of mouse was used for anesthesia before mounting the mouse in the stereotax surgery station. A glass micropipette attached to a 10-µL Hamilton syringe was used to deliver KA solution into one hemisphere of the cerebral cortex via a hole made in the occipital bone. Two microliters of 50 µM KA in PBS containing bromphenol blue dye (0.5 mg/ml) was delivered at the rate of 0.4 µl/min. After recovery, mice were monitored for seizures and brain was dissected at indicated time points for RT-PCR, RNA-ISH, and TUNEL assays.

RNA Extraction
RNA was extracted using Trizol reagent (Thermo Fisher, 15596026) following the manufacturer instructions.
PCR amplicons were separated on agrose gel with GelRed gel dye (Biotium) and imaged using the ChemiDoc Gel Imaging station (Bio-Rad). Band intensities were quantified using Image Lab (Bio-Rad) software. To negate the intensity differences that may arise from increased dye incorporation with amplicon size, the band intensity values were normalized to the size of amplicons when the splicing ratio was calculated. Relative expression levels of neurexins in RT-PCR was calculated by combining the intensity of SS4+ and SS4-bands and normalized to Actb. Single Molecule RNA Fluorescent in-situ Hybridization (smRNA-FISH) Focal KA-injected mice were euthanized with isoflurane and decapitated. Brains were quickly removed and embedded in OCT solution on dry ice. Brain tissue was sectioned at 12µm thickness using a Leica cryostat (CM3050-S) and mounted directly onto Superfrost Plus histological slides and stored at −80 • C until further use. Single-molecule RNA-FISH for Nrxn1, Nrxn2, Nrxn3, and Fos mRNA (Advanced Cell Diagnostics, probe cat# 461511-C3, 533531-C2, 505431, and 316921) was performed using the multiplex RNAscope platform (Advanced Cell Diagnostics, 323100) according to the manufacturer instructions for fresh-frozen sections. Samples were mounted using Prolong Gold antifade mounting medium (ThermoFisher, Cat# P36930).

Immunohistochemistry (IHC)
Mouse brain section and slide were prepared as described in above section. After removing slides from −80 • C, tissue were let thaw at RT for 5 min. Tissue were then fixed using ice cold 4% PFA for 30 min at RT. Tissue were washed three times with PBS and incubated with blocking solution (5% normal goat serum and 0.3% TritonX-100 in PBS) for 1 h at RT. Tissue sections were then incubated with primary antibodies diluted in blocking solution overnight at 4 • C. Next day, sections were washed three times with PBS and incubated with secondary antibodies (Alexa Fluor conjugated, 1:1,000) diluted in blocking solution for 2 h at RT. Sections were washed three times with PBS and once with water. After drying, sections were mounted using DAPI Fluormount-G (Southern Biotech). Rabbit anti-cFos polyclonal antibody (Millipore, PC05; 1:500) was used in IHC sections.

Immunocytochemistry (ICC)
For immunocytochemistry, cells were gently washed with warm PBS and fixed for 10 min at RT with 4% PFA in PBS. Following fixation, cells were washed three times with PBS and permeabilized and blocked for 1 h at RT using 0.3% Triton-X 100 and 5% normal goat serum diluted in PBS (blocking solution). Following this step, cells were incubated with primary antibodies diluted in the blocking solution for overnight shacking at 4 • C. Next day, cells were washed three times with PBS and then incubated with secondary antibodies (Alexa Fluor conjugated, 1:1,000) diluted in blocking solution for 1 h at RT. Cells were then washed three times with PBS and the coverslips were mounted on glass microscope slides with DAPI-Fluoromount-G mounting solution (Southern Biotech). Following antibodies and dilutions were used; chicken anti-MAP2 (EnCor Biotechnology, CPCA-MAP2; 1:1,000) and mouse anti-GFAP antibody (Neuromab, 75-240; 1:1,000),

TUNEL Assay
Tissue sections (10-16 micron thickness) from focal KA-injected brain were used for TUNEL assay. Click-iT Plus TUNEL Assay for in situ Apoptosis Detection with Alexa Fluor 488 dye (Thermo Fisher, C10617) kit was used following the manufacturer instruction.

Immunoprecipitation of RiboTag-mRNA
Immunoprecipitation of polyribosome-bound (RiboTag) mRNA was carried out as described previously (Sanz et al., 2009) with minor modifications. Frozen hippocampus or cortex samples were thawed in fresh homogenization buffer at 10% weight/volume and ground using Dounce homogenizer. Homogenates were then purified by centrifugation and 10% of the supernatant was used as input. The remaining supernatant lysate was incubated with pre-washed anti-HA magnetic beads (Thermo) in a rotating mixer overnight at 4 • C. The beads were washed three times with a high-salt buffer and mRNA was eluted with RLT lysis buffer (Qiagen) containing 2-mercaptoethanol. RNA was extracted from input and IP samples using RNeasy Micro kit (Qiagen) and the concentration was measured using NanoDrop spectrophotometer and stored at −80 • C until further qPCR and RT-PCR analysis.

Image Acquisition
RNA-ISH, IHC, ICC, and TUNEL samples were imaged using VS120 automated slide scanner (Olympus). For ICC of neuronalastroglial co-cultures, whole coverslips were imaged and the mean fluorescence intensity (MFI) was calculated using Image J software. RNA-seq data analysis, gene set enrichment analysis (GSEA) and differential splicing analysis were analyzed by workflows on Basepair software (https://www.basepairtech.com/) with a pipeline that consisted following steps: (a) Alignment of reads to the transcriptome derived from UCSC genome assembly using STAR with default parameters (Dobin et al., 2013). (b) Measuring of read counts for each transcript was performed using feature counts (Liao et al., 2014). (c) Differential gene expression (DE) was determined using DESeq2 pipeline (Love et al., 2014). For pathway analysis, cut off parameters of read count >10, p < 0.05 and p-adjusted (FDR, corrected for multiple hypotheses testing) < 0.1 were used. Differential splicing analysis was performed using LeafCutter tool by incorporating parameters described in Li et al. (2018).

High KCl Depolarization Induces Neuronal Cell Death, Obscuring Neurexin SS4-Alternative Splicing Patterns
To quantify the depolarization-induced neurexin SS4 splicing ratio, we exposed cortical cultures derived from E16.5 mouse embryos to 10, 20, 30, 40, and 50 mM KCl or NaCl at DIV11 (days in-vitro) for 10 min (Ding et al., 2017). We analyzed the cells by RT-PCR after 6 h and by immunocytochemistry after 24 h ( Figure 1A). Semi-quantitative RT-PCR of Nrxn1 SS4 splicing revealed a significant change in the Nrxn1-SS4 splicing ratio (defined as Nrxn1-SS4+/Nrxn1-SS4-), with an increasing preponderance of the SS4+ form associated with increasing KCl concentrations (Figures 1B,C). These data confirm previous results (Ding et al., 2017). Total expression of Nrxn1 and other key synaptic genes (Syn1, Syp, and CamK2a), however, decreased with increasing KCl concentrations, with the suppression of most synaptic gene expression becoming significant at higher KCl concentrations ( Figure 1D). Importantly, KCl concentrations above 20 mM also produced a marked increase in the cell death marker Cycs (Boehning et al., 2003), whereas the levels of Casp3 and Casp9 did not change (Figures 1B,D). Addition of NaCl at equivalent concentrations had no effect on any of these parameters.
These results suggest that elevated KCl causes both a change in Nrxn1-SS4 alternative splicing and a possible induction of cell death. To validate these results, we treated the cortical cultures with 50 mM KCl (55 mM final concentration) for 10 min, and harvested mRNAs from the cells at various times between 30 min to 48 h after the treatment ( Figure 1E) (Ding et al., 2017). RT-PCR confirmed that K + -depolarization induced a significant shift in the Nrxn1-SS4 splicing ratio after 6 h, whereas the equivalent Nrxn3-SS4 splicing ratio was not changed (Figures 1F,G). In parallel with this shift, however, the Nrxn1 and Nrxn3 total mRNA levels exhibited a decrease ( Figure 1H). In the same samples, we also measured the mRNA levels of cFos and of Casp3 and Casp9 ( Figure 1H). cFos mRNA levels increased transiently at 30 min, and then decreased again to control levels at 6 h as expected. Unexpectedly, however, cFos levels thereafter increased again to high levels at the 24 and 48 h time points (Figure 1H). The second, but not the initial, cFos expression increase was associated with an equivalent large increase in Casp3 and Casp9 expression ( Figure 1H). We did not observe such an increase in the experiments of Figure 1D because in that experiment we only examined the 6 h time point (Figure 1H). We next measured the level of neurexin SS4 splicing in neurons that were chronically exposed to 50 mM KCl, and collected RNA at 0.5, 1.0, and 1.5 h time points (Supplementary Figure 1A). We found no significant difference in the SS4 ratio or total neurexin levels (Supplementary Figures 1B-D). Viewed together, these experiments suggest that the K + -depolarization induced a biphasic response in cortical cultures: An initial phase of immediate-early gene expression (e.g., cFos) that is not accompanied by a change in Nrxn1-SS4 alternative splicing or increase in cell death, and a later phase of cell death that also features increased cFos expression and produces an apparent change in Nrxn1-SS4 alternative splicing.
To test this conclusion, we treated cortical cultures for 10 min with increasing concentrations of KCl and analyzed them by immunocytochemistry for the dendritic marker MAP2 and the astrocyte marker GFAP. Transient exposures (10 min) of cortical cultures to 30-50 mM KCl decreased MAP2 staining (∼40-75% decline), but had no significant effect on GFAP levels (Figures 2A-C). The K + -depolarization strikingly altered the neuronal cell morphology, with a degeneration of dendrites and swelling of cell bodies (Figure 2A). We then treated cortical cultures with 50 mM KCl for 10 min, and measured the MAP2 and GFAP signals at 30 min to 48 h afterwards as described in Figure 1E. K + -depolarization caused a large decrease (∼50%) in MAP2 staining after 6 h that persisted for the remaining time of the experiment. Although we also observed a lesser gradual decrease in the GFAP signal after K + -depolarization (∼20% after 6 h), this change was not significant. Again, neurons exhibited large morphological changes, with fragmented dendrites lacking MAP2 in chronic KCl exposure, apparent from the 0.5 h timepoint onwards (Supplementary Figures 1E-G).
In all experiments, we used treatments of cortical cultures with NaCl at equivalent concentrations as a negative control, confirming specificity and ruling out osmotic artifacts (Figure 2). Viewed together, these data indicate that the K + -depolarization both causes neuronal cell death and induces a shift in Nrxn1-SS4 alternative splicing, suggesting a causal relationship. Thus, we set out to explore the neuronal cell death might create the impression of a shift in neurexin SS4 alternative splicing.
To further test this hypothesis, we treated cortical cultures with cytosine arabinoside (Ara-C), which blocks astrocyte proliferation (Price and Brewer, 2001). Ara-C was added to the cultures at DIV3, DIV6, or DIV9, and cells were exposed to 50 mM KCl or NaCl at DIV11 for 10 min (Figure 3E). RT-PCR analyses, performed 6 h afterwards, showed that AraC prevented the apparent shift in Nrxn1-SS4 alternative splicing (Figures 3F,G). Immunocytochemistry of the cultures revealed that AraC, when given to the cultures earlier (at DIV3 or DIV6), impaired dendritic development, decreased overall GFAP expression, and occluded the cytotoxic effect of high KCl on neurons (Supplementary Figures 2A,B). When AraC was given late (at DIV9), it no longer abolished the cytotoxic effect of KCl on neurons as assessed by MAP2 staining (Supplementary Figure 2C). These results indicate that when K + -depolarization induces neuronal cell death in cortical cultures, the persistent expression of astroglial Nrxn1-SS4+ produces the impression of an activity-dependent shift in Nrxn1-SS4 alternative splicing even though Nrxn1-SS4 alternative splicing does not actually change.

RNAseq Analyses Confirm Expression of Cell-Death Markers Induced by K + -Depolarization
Our results suggest that the interpretation of previous papers concluding that Nrxn1-SS4 alternative splicing is highly activitydependent (Iijima et al., 2011;Ding et al., 2017) may have been confounded by neuronal cell death. To further examine this potential confound, we embarked on a direct analysis of the RNAseq results of Ding et al. (2017), who graciously deposited their data in the public domain (https://www.ncbi.nlm.nih.gov/ geo/query/acc.cgi?acc=GSE93682). Ding et al. (2017) obtained these RNAseq data from neurons under experimental conditions identical to those used here (see Figure 1E vs. Figure 4A).
Differential gene expression analysis of the RNAseq data revealed 1,781 downregulated and 1,246 upregulated genes ( Figure 4B, Supplementary Table 1). Among downregulated genes, neuronal genes were prominent, and synaptic vesicle proteins and proteins involved in synaptic transmission were generally downregulated more than two-fold (Figures 4C,D,  Supplementary Figure 3A). Neurexins and their ligands were similarly decreased in expression, with Nrxn1 levels declining two-fold (Figures 4D,E, Supplementary Figure 3A). The majority of upregulated genes, conversely, was related to inflammatory microglial and astrocytic reactions (Figures 4F-H,  Supplementary Figure 3B). Expression of caspases and other cell-death-inducing genes was also increased-in the case of Fas, eight-fold ( Figure 4H, Supplementary Figures 3C,D). Viewed together, this pattern of gene expression changes is diagnostic of a neuronal cell-death scenario with an astroglial inflammatory reaction. These results suggest that in the experiments of Ding et al. (2017), the KCl treatment also induced cell death with a loss of neuronal gene expression, including that of Nrxn1, and an appearance of regulated Nrxn1-SS4 alternative splicing produced by a change in the relative abundance of neuronal vs. astrocytic Nrxn1.
To deepen our analysis of the effect of sustained KClmediated depolarization on neuronal transcription, we analyzed two additional RNAseq datasets from recent studies that involved chronic KCl exposure of cultured neurons (Ataman et al., 2016;Quesnel-Vallieres et al., 2016).   Table 5). No significant de-enrichment of synaptic or neuronal markers was present, suggesting that no prominent neuronal cell death occurred. Although no enrichment for "Regulation of RNA Splicing" was observed, differential splicing analysis identified 82 genes with regulated splicing events (Supplementary Figure 6B, Supplementary Table 6). However, neurexins were not present among the 82 genes with differentially regulate alternative splicing, suggesting that even in these experiments in which the neurons appear to be healthy, massive neuronal stimulation did not change alternative splicing of neurexins (Supplementary Figures 5B, 6A). In addition, 60 genes among the 183 upregulated genes were categorized as "Regulation of Neuron Apoptotic Process" (Supplementary Figure 6A, Supplementary Table 7), indicating that even with neuronal silencing, extensive stimulation of neurons triggers a pro-apoptotic gene expression program.

Kainic Acid-Induced Neuronal Activation Does Not Alter Nrxn1-SS4 Splicing in vivo
Systemic administration of the glutamate receptor agonist KA induces massive synchronous activation of neurons in vivo (Wang et al., 2005). To test the effect of such activation on Nrxn1-SS4 alternative splicing, we injected adult mice with KA (20 mg/kg) intraperitoneally, and analyzed them at 0.5-24 h after injections. Immunohistochemistry showed prominent cFos expression in the hippocampus at the 1, 2, and 4 h time points (Figure 5A). RT-PCR revealed a 2-to 4-fold elevated cFos and Arc expression both in the hippocampus and the cortex between 0.5 and 4 h after injections, with a peak at 1 h after injections, indicating that KA efficiently stimulated neuronal activity throughout the brain (Figures 5B,C). However, the Nrxn1-SS4 splicing ratio was not altered by KA-induced neuronal stimulation in vivo in either the hippocampus or the cortex (Figure 5D). Moreover, total Nrxn1 levels did not significantly change (Figure 5E). These results are consistent with a recent study that surveyed alternative exon usage upon systemic administration of KA using RNAseq, and found no evidence for differential neurexin alternative splicing (Denkena et al., 2020).
Because the Nrxn1-SS4 splice pattern varies between neurons and astroglial cells (Figures 3A-C), it is possible that the lack of a KA-induced change in the Nrxn1-SS4 splice ratio was occluded by differential changes in neurons and astrocytes. To address this possibility, we employed a neuron-specific genetic RiboTag strategy to isolate translating neuronal mRNAs ( Figure 6A). We injected KA into Baf53b-Cre mice [which exhibit panneuronal Cre expression (Zhan et al., 2015) that were crossed with RiboTag mice (Sanz et al., 2009)]. We then harvested the hippocampus at 2-24 h after injections, immunoprecipitated polyribosome-bound mRNAs, and analyzed them by qRT-PCR. Quality control measurements showed an enrichment for neuronal genes and a loss of glial markers in the isolated mRNA ( Figure 6B). Immunohistochemistry revealed induction of cFos at 2 h after the KA administration that correlated with mRNA levels (Figures 6C-E), demonstrating that the KA injection stimulated neurons. However, measurements of the SS4 splicing ratio of all three neurexins failed to uncover any stimulationinduced changes in both the neuronal and the total input mRNA (Figure 6F), indicating that neuronal activity induced by systemic KA administration did not alter neurexin SS4 alternative splicing. Moreover, total neurexin levels were also not changed ( Figure 6G). Thus, neuronal activation in the hippocampus by KA does not regulate neurexin SS4 alternative splicing.

Focal KA Injection Into Cerebellum May Alter Neurexin SS4 Alternative Splicing but Suppresses Neurexin Expression
Direct microinjection of KA into the cerebellum was shown to cause activity-dependent alternative splicing of Nrxn1 at SS4 (Iijima et al., 2011). To explore whether neurexin SS4 alternative splicing might be activity-dependent in the cerebellum, different from the hippocampus or cortex, we stereotactically injected KA into one cerebellar hemisphere (ipsilateral). We examined the injected cerebellum after 5 h, using the uninjected contralateral hemisphere as a control ( Figure 7A). Immunohistochemistry, single molecule RNA fluorescent in-situ hybridization (smRNA-FISH), and RT-PCR analysis revealed a prominent induction of cFos at the injection site, consistent with KA-induced neuronal activation (Figures 7B-D). Using micro-dissected cerebellar tissue and RT-PCR analyses, we then compared the SS4 splice ratio between the ipsi-and contra-lateral sites for all neurexins. Strikingly, the SS4+/SS4-ratio decreased for all three neurexins.  (E) Experimental paradigm. Cultured cortical neurons isolated from E16.5 mouse embryos were treated with Ara-C from indicated DIV. At DIV11, cells were exposed to high-KCl or NaCl for 10 min and samples were collected after 6 h for RT-PCR and ICC. (F) RT-PCR shows Nrxn SS4 splicing in various Ara-C treatment regimens with NaCl and KCl treatments.
(G) Quantification of splicing ratio shows dynamic change in Nrxn2 and Nrxn3 SS4 splicing between early and late AraC treated control (NaCl) groups. No change in splicing ratio between NaCl and KCl treated cells within specific Ara-C treatment regimens (* = p 0.0357, ns = not significant). (H) Quantification of relative expression show dynamics of Nrxn1 and Nrxn2, but not Nrxn3 between early and late Ara-C treated control (NaCl) groups. No change in expression levels between NaCl and KCl treated cells within specific Ara-C treatment regimens (* = p 0.0153, ** = p 0.0054, Nrxn1; * = p 0.039, DIV3 vs. DIV6; p 0.0204, DIV3 vs. DIV9). All numerical data are represented as means ± SEM. Statistical significance was calculated by two-way ANOVA using Sidak's multiple comparison test.
KA caused a relative decrease in the Nrxn1-, Nrxn2-, and Nrxn3-SS4+ levels, which is opposite to what we observed in KCl-treated cells (Figures 7C,E). The changes were most pronounced for Nrxn3 (∼40% decrease), whereas they were modest for Nrxn1 (∼15% decrease). The expression levels of neuronal and astroglial marker genes, however, did not change except for a small decrease (∼15%) in synaptophysin (Syp) (Figures 7F,G). Neurexin mRNA levels were modestly decreased, but this decrease was not statistically significant ( Figure 7H). These experiments confirm the results of Iijima et al. (2011), suggesting that KA-induced neuronal activation alters neurexin SS4 alternative splicing. However, they are based on dissecting out injected brain tissue, which could be prone to errors. To more directly analyze the injected area of the cerebellum, we analyzed sections by smRNA-FISH for neurexins (Figure 8). Strikingly, smRNA-ISH uncovered a dramatic loss of Nrxn1, Nrxn2, and Nrxn3 expression in the KA-injected area but not in the surrounding tissue (Figures 8A,B). This result suggests that the dissection of injected tissue was too imprecise to capture the stimulated neurons, and that the KA injection did cause a global change in total neurexin gene expression in the stimulated neurons. To elucidate whether this change was associated with a neurotoxic effect of KA that is known to induce DNA damage and cell death (Kasof et al., 1995;Simonian et al., 1996), we tested for DNA damage at the site of KA exposure. We labeled the KA injection sites 5, 24 and 48 h after treatments using terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining. The KA injected sites were strongly positive for TUNEL signals at all-time points (Figures 8C,E), consistent with highly fragmented DNA induced by KA-driven excitotoxicity and apoptosis. At 5 h, TUNEL-positive cells were mostly granule layer cells, whereas at 24 and 48 h, both granule and molecular layer cells were positive for TUNEL signals (Figure 8D). The sustained DNA damage even after 48 h of KA injection indicates longlasting cell damage that may have a profound impact on the physiology, and confounds the interpretation of the neurexin SS4 alternative splicing observed by RT-PCR.

DISCUSSION
This study was initiated to explore the mechanistic basis for the activity-dependent alternative splicing of neurexins at SS4, which is of interest because it regulates the glutamate receptor composition of synapses (Aoto et al., 2013;Dai et al., 2019). In pursuing this goal, we first aimed to establish optimal conditions for studying activity-dependent neurexin SS4 alternative splicing, based on pioneering papers that were recently published (Iijima et al., 2011;Ding et al., 2017). However, as we analyzed neurons treated as described in these papers, either in culture or in vivo, we noticed that the apparent stimulation-induced change in neurexin SS4 alternative splicing was associated with neuronal cell death. Given the importance of neurexin alternative splicing, we therefore examined in detail the effect of various conditions that are thought to regulate neurexin SS4 alternative splicing on the overall expression of marker genes and on neuronal viability. Our data show that most reported conditions that promote activity-dependent SS4 alternative splicing of neurexins predispose to neuronal cell death, suggesting that the interpretation of such studies is not straightforward. Thus, it is at present unclear whether activity-dependent alternative splicing of neurexins at SS4 is a physiological event.
Application of elevated levels of KCl (55 mM) to cultured neurons has been employed in many studies to trigger activitydependent gene expression programs and to stimulate alternative splicing of neuronal mRNAs, including neurexins (Martinowich et al., 2003;Greer and Greenberg, 2008;Schor et al., 2009;Kim et al., 2010;Iijima et al., 2011;Rozic et al., 2011Rozic et al., , 2013Ding et al., 2017). This approach is thought to mimic the pattern of gene activation that occurs in response to physiological stimuli in brain (Greer and Greenberg, 2008). However, although elevated extracellular KCl induces membrane depolarization, it may not always actually enhance neuronal activity. Treatment of cultured neurons with 8 mM KCl retains spontaneous neuronal activity, but shifts the pattern of that activity from burst to tonic spike firing (Golbs et al., 2011). Treatment of cultured neurons with KCl above 10 mM conversely diminishes spontaneous activity (Grubb and Burrone, 2010;Rienecker et al., 2020), suggesting that persistent neuronal depolarization does not necessarily induce neuronal stimulation. Furthermore, persistent neuronal depolarization produces sustained increases in intracellular calcium, a key second messenger for numerous cellular processes, including transcription and alternative splicing (Collins et al., 1991;Grubb and Burrone, 2010;Rienecker et al., 2020). In addition, neurons die at concentrations of >50 mm KCl (Collins and Lile, 1989). Na + influx upon KCl-mediated depolarization enhances cell death and induces autophagy (Takahashi et al., 1999;Shehata et al., 2012). In the light of this relationship of KCl to neuronal survival, the utility of the persistent depolarization paradigm as an "activity-inducing system" to study alternative splicing may need re-evaluation. While there are clear merits of using elevated extracellular KCl (extensively reviewed in Rienecker et al., 2020), the inherent confounds and limitations of this approach in studying complex and heterogeneous splice isoforms could outweigh its benefits.
The results of our KCl treatment experiments and RNAseq analyses exemplify and highlight potential disadvantage of this approach. Historically, the glutamate receptor agonist KA has been used as a neurotoxin to induce cell death to study neurodegeneration in rodents (Pollard et al., 1994;Ankarcrona et al., 1995;Simonian et al., 1996;Wang et al., 2005;Kuzhandaivel et al., 2010). KA is 30-fold more neurotoxic than glutamate and activates kainitetype ionotropic glutamate receptors (Bleakman and Lodge, 1998). Kainate receptor activation by KA increases intracellular Ca 2+ , elevates reactive oxygen species, and stimulates multiple other biochemical reactions, resulting in neuronal cell death (Sun FIGURE 6 | SS4 splicing ratio shows marked difference between IP and input mRNA but no significant difference within IP or input samples across different time points (G) Quantification of Nrxn relative expression levels show general difference in IP and input mRNA, except for Nrxn1 but no significant difference with in IP or input samples across different time points. All numerical data are represented as means ± SEM. Statistical significance was calculated by two-way ANOVA using Sidaks's multiple comparison test. Scale bar 100 µm.  , 1992;Cheng and Sun, 1994;Gluck et al., 2000;Milatovic et al., 2002). Similar to the KCl-based assays, KA-induced neuronal activity has also been used to study activity-dependent transcription and alternative splicing of neuronal mRNAs (Iijima et al., 2011;Denkena et al., 2020). Systemic administration of KA may have lesser effects on cell death than focal microinjection of KA into brain regions (Le Duigou et al., 2005). However, systemic KA did not lead to any significant change in neurexin SS4 splicing, although the neurons were massively activated as revealed by induction of cFos. While we observed a small shift in the SS4 alternative splicing of neurexins in areas of the cerebellum that were injected with KA, the strong induction of sustained DNA damage and the suppression of neurexin mRNA levels suggests apoptosis-induced cell death. Moreover, in cerebellar granule cells, KA-induced apoptosis correlates with activation of c-Jun, an immediate-early gene (Cheung et al., 1998). Studies of alternative splicing after KA-induced neuronal activity often do not examine neuronal survival despite the fact that KA is a strong neurotoxic agent (Iijima et al., 2011). The correlation of neurexin SS4 splice changes and immediate-early gene activation with significant DNA damage muddles the interpretation of the alternative splicing results.
While it is possible that activity-dependent neurexin alternative splicing at SS4 occurs in a physiological context, this complex question needs to be tackled using assays that truly mimic neuronal activity in order to address the biological significance. It is likely that the use of massive stimulation paradigms or "hammering" approaches to elicit strong gene expression and alternative splicing changes does more harm than good. Moreover, any major conclusion arising from studies that utilize such approaches may call into question the actual mechanistic underpinning of activity-dependent alternative splicing.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the methods section.

ETHICS STATEMENT
The animal study was reviewed and approved by Administrative Panel on Laboratory Animal Care (APLAC) at Stanford University.

AUTHOR CONTRIBUTIONS
KL-A and TCS conceived the study. KL-A performed the experiments. KL-A and TCS analyzed the data and wrote the manuscript. All authors contributed to the article and approved the submitted version.

ACKNOWLEDGMENTS
We thank K. Arendt, L. Chen, J. Polepalli, K. Raju, A. Sclip, and J. Trotter and all Sudhof, Chen and Malenka lab members, and A. Ahmed (King's College London) for helpful discussions. We thank X. Ding and J. Guan (Tsinghua University) for providing a detailed protocol of embryonic primary neuronal culture. Schema in Figures 5A, 6A, and 7A are created with BioRender.com.