A SLM2 Feedback Pathway Controls Cortical Network Activity and Mouse Behavior

Summary The brain is made up of trillions of synaptic connections that together form neural networks needed for normal brain function and behavior. SLM2 is a member of a conserved family of RNA binding proteins, including Sam68 and SLM1, that control splicing of Neurexin1-3 pre-mRNAs. Whether SLM2 affects neural network activity is unknown. Here, we find that SLM2 levels are maintained by a homeostatic feedback control pathway that predates the divergence of SLM2 and Sam68. SLM2 also controls the splicing of Tomosyn2, LysoPLD/ATX, Dgkb, Kif21a, and Cask, each of which are important for synapse function. Cortical neural network activity dependent on synaptic connections between SLM2-expressing-pyramidal neurons and interneurons is decreased in Slm2-null mice. Additionally, these mice are anxious and have a decreased ability to recognize novel objects. Our data reveal a pathway of SLM2 homeostatic auto-regulation controlling brain network activity and behavior.

, related to Figure 2. Downstream noncoding exons are conserved between the human and chicken SLM2 genes. (A) UCSC browser screenshot (Rosenbloom et al., 2014) showing the 3′ end of the human SLM2 gene with RNAseq reads from Human Body Map 2.0 (GSE30611), in the tissues indicated. (B) Violin plots displaying MAJIQ quantification for use of SLM2 alternative end 2 across RNAseq data from human tissues (Illumina Human Body Map 2.0, GSE30611), with higher levels of inclusion within the brain and testis which express higher levels of SLM2 protein (C) Screenshot showing RNAseq reads from the 3' end of the chicken SLM2 gene, measured in testis, brain and kidney. (D & E) Violin plots of chicken RNAseq expression data from 2 different datasets (D, GSE41637; E, GSE41338) indicate high levels of SLM2 alternative end 2 in the brain compared with other tissues.  (Rosenbloom et al., 2014) showing the 3′ end of the human SAM68 gene. (B) Expression analysis of SAM68 within a stable HEK293 cell line expressing a tetracycline-inducible SAM68 gene. Agarose gel showing that endogenous human SAM68 alternative end 2 is induced by tetracycline and stabilized by cycloheximide treatment. This experiment was performed with three independent sets of biological samples, and one complete experiment is shown here. (C) Agarose gel showing 3′ end selection creating SAM68 alternative isoforms 1 and 2 within the HEK293 stable cell line that over-expresses Sam68 protein. Patterns of 3' end formation are shown after treatment with either a control siRNA or an siRNA directed against UPF1, indicating stabilisation of isoform with alternative end 2. The siRNA treatment was carried out after tetracycline induction. (D) Quantitation of percentage selection of SAM68 alternative isoform 2 following siRNA-mediated depletion of UPF1 or a control depletion, using triplicate biological samples as in part (C), quantitated using capillary gel electrophoresis. Error bars are SEM. (E) Levels of UPF1 mRNA measured using qPCR in the HEK293 stable cell line before and after siRNA-mediated depletion. Error bars are SEM. (F) mRNA isoform levels from the endogenous U2AF35 gene, which is a known target for NMD (isoform C is stabilised as predicted by UPF1 siRNA treatment). (G) Levels of endogenous Sam68 protein decrease after tetracycline induction of ectopic Sam68 protein levels. Protein samples were purified from the control cell line (left panels) and the Sam68-FLAG expressing cell line (right panels) at 0 and 24 hours after tetracycline addition. Patterns of protein expression were measured by Western blotting using anti-Sam68 antisera or anti-actin antibody. (A) Agarose gel showing detection of alternative 3′ ends in Slm1 mRNAs from hippocampus and total brain RNA isolated from wild type and Slm2 null mouse backgrounds: notice the decreased selection of alternative end 2 in the knockout background. (B) Percentage of Slm1 mRNAs that terminate with alternative 3′ end 2 in different brain structures dissected from 3 wild type and 3 Slm2 null mice. Each bar represents the mean percentage value, and the error bar is standard error of the mean. Probability (P) values were calculated using an independent two-sample t-test between heterozygote and knockout mice compared to wild type. Significant values are highlighted by * P<0.05, *** P<0.001.Statistical analyses (t tests) were carried out using Graphpad, using RT-PCR data collected from capillary gel electrophoretic analysis of at least three independent replicates in each case. Error bars represent SEM. (C) Pattern of Sam68 3′ end formation in wild type and Slm2 knockout mouse hippocampus. (D) Agarose gel showing detection of alternative 3′ ends in Slm2 mRNAs from brain structure RNAs isolated from wild type and Sam68 null mouse backgrounds. (E) Percentage of Slm2 mRNAs that terminate with alternative 3′ end 2 in different brain structures dissected from 3 wild type and 3 Sam68 null mice. Each bar represents the mean percentage value, and the error bar is standard error of the mean. Probability (P) values were calculated using an independent two-sample t-test between knockout mice compared to wild type. Statistical analyses (t tests) were carried out using Graphpad, using RT-PCR data collected from capillary gel electrophoretic analysis of three independent replicates in each case. Error bars represent SEM. (A) Example long time course local field potential traces showing the emergence of divergent network activity during the bath application of kainate. Note in the KO slice (blue trace) and associated spectrogram there is a slow build up of γ frequency activity whereas in the WT (red trace) and associated spectrogram the activity is dominated by high amplitude burst discharges. The lower trace shows this activity from a wild type mouse as selected from dashed box. Intermittent burst discharges are co-existent with on-going γ oscillations. (B) Wild type CA3 traces (WT, shown in red) show alternating γ oscillations and epileptiform burst activity, while Tstar knockout traces (KO, shown in blue) show background γ oscillations but no high amplitude epileptiform activity. Measurements were made from 25 slices made from 5 individual wild type mice, and 21 slices made from 4 individual Slm2 null mice. (Ciii) Dot plot showing individual data points for peak frequency and area power of γ oscillations in EC in WT (red) and KO (blue). Each dot represents a recording from an individual slice (9 slices analysed from 4 SLM2 null mice, and 9 slices analysed from 5 wild type mice). The horizontal bars represent group averages. Area power and peak frequency are changed significantly (*=p<0.05) when slices from WT and KO mice are compared. Peak frequency and power values were obtained from power spectra generated with Fourier analysis in the Axograph X software package (Kagi, Berkeley, CA). Power for a given frequency band was determined as the area under the peak in the power spectra between 20 and 80 Hz for γ frequency oscillations. All values are given as the mean ± SE where distributed normally; otherwise, data are expressed as the median (interquartile range). Power spectra were constructed off-line from digitized data (digitization frequency, 10 kHz), using a 60 s epoch of recorded activity. Analysis of the data was performed by the individual who conducted the experiment but who was blind to the origin of the slices. (A and B) Slm2 KO male and female mice show normal behaviour in PPI/acoustic startle tests compared to wild type (n=12 male Slm2 null mice, n=12 wild type male mice, n=12 female Slm2 null mice, n=12 female wild type mice). As expected, the degree of inhibition increased with increased pre-pulse tone ( In humans linked with autism and schizophrenia. Functional redundancy between Neurexin1, Neurexin2 and Neurexin3 mutants in mouse knockout experiments (Missler et al., 2003). In mouse knockin models, alternative splicing of Neurexin3 AS4 affects synaptic activity via AMPA receptor.

Neurexin2
As above As above As above.

Neurexin3
As above As above As above. In mouse knockin models, alternative splicing of Neurexin3 AS4 affects synaptic activity via AMPA receptor.

Stxbp5l
(tomosyn2) 57 amino acid peptide cassette exon May play a role in vesicle trafficking and exocytosis, and neurotransmitter release at synapse.

Statement on biological and technical replicates
Biological replicates used in this study were from individual animals or cells. Technical replicates were multiple tests performed on the same samples.

Statistical methods
All statistical methods are described at relevant points in the text and supplemental information. Briefly, percentage splicing inclusions are shown as averages +/-standard error of the mean, and t-tests were used to analyse the significance of pairwise comparisons, using Graphpad Prism. Electrophysiological data was analysed to generate averages, and t-tests were used to analyse the significance of pairwise comparisons. Mouse behaviour within open fields and in relation to novel and familiar objects was statistically analysed using one way ANOVAs, and is presented as an average plus or minus the standard error of the mean, and using the STATISTICA analysis package. Sample size was determined by setting the probability of a Type I error and power at 0.05 and 0.80, respectively. Due to multiple testing p-values for the rotarod and acoustic startle tests, phenotyping data were corrected for each group of mice tested using Bonferroni correction to determine significance level. Males and females were analysed separately with significance level for males set at p<0.005 and females p<0.0042. The rotarod data were analysed using Welch's t-test. For PPI the data was normalised and analysed with repeated measures ANOVA.

RNAseq analysis
To avoid detection of transcriptome differences that might arise from strain or sex differences (Su et al., 2008), we back crossed our Slm2 KO allele onto the C57Bl26 background for 8 generations, and used backcrossed adult male mice for subsequent analysis. Since Slm2 is highly expressed in the hippocampal fields CA1-CA3 but not in the dentate gyrus of the hippocampus, we dissected CA1-CA3 separately from the dentate gyrus. RNA was extracted from cells using RNeasy Plus Mini Kit (Qiagen) following manufacturer's instructions and re-suspended in nuclease-free water. All RNA samples were DNase treated using DNA-free kit (Ambion) and stored at −80°C prior to RNA quality control check using 2100 Agilent Bioanalyser and mRNA library prep using TruSeq mRNA library kit (Illumina). Paired-end sequencing was done in total for six samples (three biological replicates of wild type and Slm2 knockout CA1-CA3 regions). Sequencing was on an Illumina Hiseq 2000 machine as previously described (Best et al., 2014).
RNA-seq data were processed and analyzed to identify differentially expressed genes and exons which have differential usages among transcripts of a gene. The quality of sequencing reads was firstly checked with FastQC (Andrews). Poly-N tails were trimmed off from reads with an in house perl script. The 14 bp on the left ends of all reads were clipped off with Seqtk (Han et al., 2013) to remove biased sequencing reads caused by random hexamer priming (Hansen et al., 2010). Low quality bases (Q < 30) and standard Illumina (Illumina, Inc. California, U.S.) paired-end sequencing adaptors on 3' ends of reads were trimmed off using Trim-galore (http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/) and only those that were at least 20bp in length after trimming were kept. The high quality reads were then mapped to the mouse reference genome mm10 with Tophat2 (Kim et al, 2013) and STAR (Dobin et al., 2013). Alternative splicing events were assessed using MAJIQ and VOILA software packages (Vaquero-Garcia et al., 2016). Briefly, uniquely mapped, junction-spanning reads were used by MAJIQ to construct splice graphs for transcripts from a custom Ensembl transcriptome annotation and to quantify PSI (within conditions) and ΔPSI (between conditions) for all local splicing variations (LSVs). The captured LSVs include classical alternative splicing events (e.g. cassette exons, alternative 5' splice sites, etc.) as well as more complex variations. LSVs with an expected change of greater than 10% were then visualized using VOILA to produce splice graphs, violin plots representing PSI and ΔPSI quantifications, and interactive HTML outputs for changes between wild type and Slm2 KO CA1-CA3 regions (http://paros.pcbi.upenn.edu/collab/Ehrmann_et_al/voila/dpsi_WTvKO/). The splicing changes corresponding to these violin plots were then examined visually on the UCSC genome browser (Fujita et al., 2011), and 26 candidate alternative exons were further tested by RT-PCR. The relatively low confirmation rate by RT-PCR (10/26 = 38%) compared to previous reports for MAJIQ's analysis (Vaquero et al 2016) may be attributed to the difficulty in dissecting the CA1-CA3 regions separately from the dentate gyrus, which introduces variability between samples (see above). Thus, it is plausible that the relatively small list of differentially spliced exons reported is a conservative estimate to the regulatory effects of Slm2.
ms and the pre-pulses at 55, 65, 70 and 75 dB for 10 ms. The pre-pulses preceeded the pulse by 50 ms. Intertrial interval was set at 20-30 seconds.
Due to multiple testing p-values for the rotarod and acoustic startle tests, phenotyping data were corrected for each group of mice tested using Bonferroni correction to determine significance level. Males and females were analysed separately with significance level for males set at p<0.005 and females p<0.0042. The rotarod data were analysed using Welch's t-test. For PPI the data was normalised and analysed with repeated measures ANOVA.
The novel object recognition test was performed as described (Antunes and Biala, 2012) under Italian Board of Health Approval (Authorisation n. 6/2015/PR). Mice were first transferred to the experimental room and left undisturbed in their home cage for 30-min acclimation in the new environment. During the first habituation session, each mouse was placed for 10-min in the testing arena (empty cubic box 50x50x30 cm made of white opaque plastic material) and then returned to the home cage for a 10-min interval. Then, each mouse was placed in the testing arena for the sample trial, which consisted in the exposition of two identical objects for 10-min period. Objects were either two colored plastic cubes (5x5x5cm) or two glass cylinders (8 cm high and 5 cm diameter) and were presented according to a random schedule. The objects were cleaned with 10% ethanol before the third session. The interest for the objects by the mice was measured as exploration, which was defined as time mice spent sniffing or touching the objects with nose and/or forepaws.
At the end of the sample trial, mice were placed back in their home cage and were left undisturbed for a 60-min inter trial interval. During the following test trial, each mouse was placed back in the testing arena where one of the two objects remained unchanged (familiar object FO) while the other one was replaced with a different one (novel object NO). In this session, object exploration was measured as above and the interest for the NO was inferred by calculating the preference index (NO/FO+NO ratio). A preference index above 50% indicates that the NO was preferred to FO, while a preference index of 50% indicates that mice spent the same amount of time in exploration of the two objects.
Mouse behavior was video recorded by a video camera positioned above the testing arena. An experimenter blind to experimental conditions manually assessed mouse exploratory behavior toward the objects. General exploratory and locomotory activities were assessed through Noldus Ethovision system (The Netherlands). Experimental groups included 10 mice each.

Localisation of SLM2 within the mouse hippocampus and entorhinal cortex
Whole brains from a wild type mouse were fixed with buffered 4% paraformaldehyde either by transcardial perfusion (whole brains) or by immersion fixation (slices). Tissue was then cryoprotected by immersion in 30% sucrose in phosphate buffered saline overnight, and 40 μm sections cut on a freezing microtome. Sections were collected in Tris buffered saline (TBS) and processed for double label immunofluorescence histochemistry. Primary antibodies used for staining were: rabbit α-SLM2 at 1:250 (Ehrmann et al., 2013); mouse α-Parvalbumin (PV) at 1:5000 (Sigma Aldrich); mouse α-non-phosphorylated neurofilament (NPNF) at 1:1000 (Covance, monoclonal SMI-32). Sections were incubated free floating with these primary reagents, diluted in TBS with 0.3% Triton X-100 (TBST) and 3% of the appropriate normal serum, overnight at 4°C. Following washing in TBS, some sections were incubated with biotinylated secondary antibody diluted 1:200 in TBST for 2 hours, washed and then incubated with avidin-Texas Red (1:200 dilution in TBST; Vector Labs) for 2 hours. Sections incubated with avidin-Texas Red were simultaneously incubated with Alexa-Fluor (488) goat anti-mouse secondary antibody (1:200 dilution in TBST; AbCam, Cambridge, UK). Double-label immunofluorescence sections were mounted in Vectashield with DAPI (Vector Labs) and viewed on a Nikon A1R confocal microscope. Areas of CA1-3 hippocampus and EC were examined for evidence of double labelling of neurons containing PV or NPNF with SLM2. PV/SLM2 double labelling was quantified by sampling PV cells from 5 sections: 63 PV cells from hippocampus, and 50 from EC, examining single planes of focus at all wavelengths to accurately assign a nucleus to each PV cell profile, and subjectively score the level of SLM2 immunopositivity of each nucleus; strongly labelled (easily visible) weakly labelled (requires checking) or not visible above background.
In vitro brain slice electrophysiology Slices (400 μm) containing hippocampus and EC were prepared from young adult male Slm2 knock-out mice (Slm2 −/− ) and wild-type (WT) litter mates. All procedures were performed according to the requirements of the United Kingdom Animals Scientific Procedures Act (1986). Animals were anesthetized with inhaled isoflurane, immediately followed by an intramuscular injection of ketamine (≥ 100 mg/kg) and xylazine (≥ 10 mg/kg). Animals were perfused intracardially with 50 ml of modified artificial CSF (ACSF), which was composed of the following (in mm): 252 sucrose, 3 KCl, 1.25 NaH 2 PO 4 , 24 NaHCO 3 , 2 MgSO 4 , 2 CaCl 2 , and 10 glucose. All salts were obtained from BDH Chemicals (Poole, UK). The brain was removed and submerged in cold (4-5°C) ACSF during dissection. Horizontal slices were cut and transferred to a recording chamber maintained at 34°C at the interface between ACSF [containing the following (in Mm): 126 NaCl, 3 KCl, 1.25 NaH 2 PO 4 , 24 NaHCO 3 , 1 MgSO 4 , 1.2 CaCl 2 , and 10 glucose] and warm, moist carbogen gas (95% O 2 /5% CO 2 ). Slices were permitted to equilibrate for 45 min before any recordings commenced. Slices were prepared from a knock-out mice and wild-type on the same experimental day with the experimentalist blinded to the origin of the slices. Subsequent decoding of the origin of slices was revealed for purposes of analysis.