SAM68 regulates neuronal activity-dependent alternative splicing of neurexin-1

Abstract

The assembly of synapses and neuronal circuits relies on an array of molecular recognition events and their modification by neuronal activity. Neurexins are a highly polymorphic family of synaptic receptors diversified by extensive alternative splicing. Neurexin variants exhibit distinct isoform-specific biochemical interactions and synapse assembly functions, but the mechanisms governing splice isoform choice are not understood. We demonstrate that Nrxn1 alternative splicing is temporally and spatially controlled in the mouse brain. Neuronal activity triggers a shift in Nrxn1 splice isoform choice via calcium/calmodulin-dependent kinase IV signaling. Activity-dependent alternative splicing of Nrxn1 requires the KH-domain RNA-binding protein SAM68 that associates with RNA response elements in the Nrxn1 pre-mRNA. Our findings uncover SAM68 as a key regulator of dynamic control of Nrxn1 molecular diversity and activity-dependent alternative splicing in the central nervous system.

Figure 1. Depolarization-dependent alternative splicing of Neurexins in cerebellar neurons

(A) Schematic diagram outlining organization of Neurexin alpha and neurexin beta protein variants. The positions of alternatively spliced segments (AS1,2,3,4,5) are indicated by arrows. LamininG-domains are shown as ovals, EGF-domains as rectangles, black line is the trans-membrane domain. The cartoon on the right illustrates the 90bp cassette exon (20) at AS4. Position of primers used for analysis indicated by arrows. (B) Developmental regulation of Nrxns at AS4 in mouse cerebellum (n=3). (C) Dissociated cultures of mouse cerebellar neurons maintained in vitro for 14 days were depolarized by addition of 25 mM KCl for 6 hours. Semi-quantitative PCR analysis with primers for AS4, c-Fos or Gapdh. (D) Quantitative real-time PCR for total Nrxn1, Nrxn1 4(+) and Nrxn1 4(−). Nrxn mRNA / Gapdh values in untreated control cells were set to 1.0 and compared to cells stimulated (25 mM KCl) for 6 hours (n=4). (E) Analysis of semi-quantitative PCR data from panel C. The fraction of Nrxn 4(−) compared to total Nrxn is plotted (n=4). (F) The glutamate receptor agonist kainic acid (50 µM) was applied for 6 hours in presence or absence of the AMPA/kainate receptor antagonist CNQX (50 µM) (n=4–5 independent cultures). Electrical field stimulation was applied in three trains (3 min of 200Hz) spaced by 7 mins and RNA was isolated 5.5 hrs after stimulation (n=6–8 independent cultures per condition). Mean and SEM. See also Figure S1 and S2.

Figure 2. Depolarization-dependent alternative splicing of neurexin requires CaMKIV and controls trans-synaptic signaling

(A) Dissociated cultures of mouse cerebellar neurons (DIV14 days) were depolarized (25 mM KCl) for 6 hours and CdCl₂ (20 µM) or nifedipine (10 µM) were added as indicated. Nrxn 4(−) levels were quantified by semi-quantitative PCR (n=4). (B) CaMK signaling in depolarization-dependent alternative splicing of Nrxn1 was inhibited by co-application of pharmacological inhibitors or RNAi knockdown of CaMKIV (n=4–7). c-Fos activation was probed by RT-PCR to confirm immediate-early gene induction. (C) Schematic diagram illustrating binding preferences of neurexin splice isoforms. Beta-NRX1 4(−) interacts similarly with the NL1A and NL1B splice variants, the 4(+) insertion in the beta-NRX1 4(+) variant (red insert) prevents strong adhesive interactions with NL1B but still allows for interactions with NL1A. The Cbln1/GluD2 complex specifically interacts with the 4(+) insertion in beta-NRX1 4(+). (D) HEK293 cells co-expressing GFP and NL1B (left) or GluD2 (right) were added to cerebellar neurons at DIV9 and analyzed 2 days later. For depolarizing conditions, 25mM KCl was added to the culture medium at DIV 6. Presynaptic differentiation assessed by vGluT1 staining. (E) HEK293 cells co-expressing GFP and NL1B (left) or NL1A (right) were added at DIV7 under depolarizing or control conditions. Pharmacological inhibitors (5 µM STO609 or 10 µM FK605) were applied at DIV 6. (F) Morphometric analysis of vGluT1 staining at neuron-HEK293 contact sites analyzing the fraction of the vGluT1-positive area of the GFP-positive neuroligin-expressing cell (n>30 cells per condition, 3 independent experiments). Mean and SEM. Scale bars are 5µm. See also Figure S2.

Figure 3. SAM68-dependent alternative splicing of Neurexins in vitro

(A) Schematic diagram of Nrxn1–4 splice reporter construct containing AS4 with constitutive exons (orange), alternative exon 20 (red), and introns shown as lines. Two AU-rich sequence stretches are highlighted with arrows. Intron 19 (>40 kbp) was truncated leaving >550 bp adjacent to the splice donor and the acceptor sites intact. (B) Splice reporter expression vector was co-transfected into HEK293 cells with negative control DNA or epitope-tagged expression constructs for GFP-SAM68, T7-Nova1, T7-Nova2, Xpress-nPTB, YFP-hnRNPA1, YFP-hnRNPH1, GFP-SLM1, GFPSLM2. Alternative splice isoform choice was measured by semi-quantitative RT-PCR with primers flanking the alternative splice site (n=3). Expression of all transfected RNA-binding proteins was confirmed by Western blotting with anti-GFP, -T7 or -Xpress antibodies. Molecular weights in kDa. (C) Co-transfection of increasing amounts of a SAM68 expression vector with the Nrxn1–4 splice reporter. Alternative splicing was assayed by semi-quantitative RT-PCR and HA-tagged SAM68 detected by immunoblot. (D) GFP-tagged point mutants in SAM68 KH-domain and nuclear localization signal were co-expressed with the Nrxn1-4 splice reporter and alternative splicing was analyzed by semi-quantitative RT-PCR. Protein expression levels were compared using western blotting. (E) Co-transfection of plasmids encoding constitutively active Fyn-kinase and Nrxn1–4 reporter. Mean and SEM.

Figure 4. Mapping of SAM68-response elements in Nrxn1

(A) Intronic sequences preceding (19-a, 19-b, 19-c) or following exon 20 (20-a to 20-d) were deleted from Nrxn1–4. Intron 19-c deletion was combined with deletions 20-a to 20-d in intron 20. Panels show semi-quantitative PCR results from HEK293 cells without and with co-expression of SAM68 and Nrxn1–4, respectively. Quantitation was performed for products in presence of SAM68 (n=4). (B) Biotinylated RNA-oligonucleotides probes covering intronic sequence between 20-b and 20-c deletions used in pull-down experiments with mouse adult brain extracts. Probe A* contains two nucleotide changes (red). Bound proteins were analyzed by western blotting with anti-SAM68, anti-SLM1, anti-SLM2, anti-Fox3/NeuN, and anti-SF2/ASF antibodies. SAM68 binding was quantified by densitometric scanning of Western blot signals (n=4). (C) Splice reporters were delivered into cortical pyramidal cell precursors of E14.5 embryos by in utero electroporation and reporter processing was examined at P6. Reporter expression levels were examined with primers specific for constitutive reporter sequence elements and RNA input levels were monitored by semi-quantitative PCR for endogenous Gapdh (n=3–6). Mean and SEM. See also Figure S3.

Figure 5. Cell-type specific STAR protein expression in the cerebellum

(A) Specificity of anti-STAR protein antibodies confirmed by probing lysates from HEK293 cells overexpressing GFP-tagged SAM68, SLM1, or SLM2. (B) Immunoblot of adult cortical, hippocampal, and cerebellar tissue extracts probed with STAR protein antibodies. (C) Developmental regulation of STAR protein expression in the cerebellum from postnatal day 0 (P0) to adult (>P60). (D,E) Para-sagittal sections through adult cerebellum triple stained with anti-NeuN, anti-RORα, and anti-SAM68, anti-SLM1, or anti-SLM2 antibodies. In E, P7 cerebellum was analyzed using anti-SAM68 antibodies. ML, molecular layer; PCL, Purkinje cell layer; IGL, internal granular layer; EGL, external granular layer. Scale bars are 50 µm. See also Figure S4.

Figure 6. Cerebellar alterations in SAM68 KO mice

(A) Low magnification view of para-sagittal section of brain stem and cerebellum reveals anti-SAM68 immune-reactivity which is lost in SAM68 KO (−/−) tissue. (B) Motor performance of male control mice and SAM68 KO littermates were tested on an accelerating rotarod. The latency to fall is plotted for days 1,3, and 7 of training (n=5–8 mice per genotype). (C) Immunoreactivity for synaptic vesicle proteins VGluT1 (magenta) and VAMP2 (green) in the internal granular layer (IGL) of adult wild-type and SAM68 KO cerebellar cortex. The right panel shows single rosettes at high magnification. Scale bars are 20µm in (left) and 5µm in (right). (D) Representative line scans from wild-type and SAM68 KO cerebellum displaying immunoreactivity for vGluT1 in mossy fiber rosettes in C. Cumulative probability of VGluT1 and VAMP2 average intensities in cerebellar glomeruli compared between wild-type (blue line) and SAM68 KO mice (orange) (n> 438 rosettes from 3 animals per genotype, Komolgorov-Smirnov test). (E) Quantitative PCR performed on cortical, cerebellar, brainstem cDNAs from control (+/+, black) and SAM68 KO mice (−/−, white; n=3 animals per genotype). Abundance of total Nrxn1 mRNA, Nrxn1 4(+) and Nrxn1 4(−) transcripts is expressed compared to Gapdh (values for control tissue were set to 1.0). Mean and SEM. See also Figure S5.

Figure 7. SAM68 is essential for regulation of activity-dependent splicing of Nrxn1

(A) Relative abundances of phosphorylated and non-phosphorylated SAM68 were assessed by liquid-chromatography-tandem mass spectrometry following SAM68-IP from control and depolarized cerebellar granule cells in culture. (B) The CaMK consensus motif in the amino acid sequence surrounding ser 20 is highlighted in bold. Ratios of the normalized peptide abundances (depolarized / control) are shown (total SAM68 (depolarized/control) levels (phospho-serine 20: 8.39 ± 0.90, p < 0.01). (C) Heterozygous control (+/−) or homozygous SAM68 KO neurons (−/−) were incubated in depolarizing medium for 6 hours and exon 20 inclusion was monitored by semi-quantitative PCR with flanking primers for Nrxn1, Nrxn2, or Nrxn3 (n=4). Loss of Sam68 expression was confirmed with oligonucleotides amplifying the Sam68 cDNA. (D) Depolarization-dependent alternative splicing of the calcium-response element (CaRRe) containing pre-mRNAs Grin1 exon 5 and Kcnma1 STREX exon in control and SAM68 KO neurons (n=4). (E) Neuronal activity-dependent alternative splicing in vivo examined by stereotaxic injection of 50 µM kainic acid (or saline as control) into the mouse cerebellum. Animals were sacrificed 5 hours after injection. Neuronal stimulation was confirmed by the increase in the expression of c-Fos mRNA in the injected ipsilateral versus non-injected contralateral side by semi-quantitative (middle panel) and quantitative PCR (right panel, Rq(c-Fos/Gapdh) for contralateral side set to 1.0, log scale, n=3 animals). (F) Inclusion of exon 20 in Nrxn1 mRNA probed using quantitative PCR on cDNA derived from the injected cerebella (n=5 animals per genotype). Rq(Nrxn/Gapdh) values for the ipsilateral (injected) hemisphere (white columns) are expressed relative to the Rq(Nrxn/Gapdh) values obtained for the uninjected contralateral side from the same animals (black columns). Mean and SEM. See also Figure S6.