Inhibition of Crmp1 Phosphorylation at Ser522 Ameliorates Motor Function and Neuronal Pathology in Amyotrophic Lateral Sclerosis Model Mice

Abstract Amyotrophic lateral sclerosis (ALS) is a rapidly progressive and fatal neurodegenerative disorder that affects upper and lower motor neurons; however, its pathomechanism has not been fully elucidated. Using a comprehensive phosphoproteomic approach, we have identified elevated phosphorylation of Collapsin response mediator protein 1 (Crmp1) at serine 522 in the lumbar spinal cord of ALS model mice overexpressing a human superoxide dismutase mutant (SOD1G93A). We investigated the effects of Crmp1 phosphorylation and depletion in SOD1G93A mice using Crmp1S522A (Ser522→Ala) knock-in (Crmp1ki/ki) mice in which the S522 phosphorylation site was abolished and Crmp1 knock-out (Crmp1−/−) mice, respectively. Crmp1ki/ki/SOD1G93A mice showed longer latency to fall in a rotarod test while Crmp1−/−/SOD1G93A mice showed shorter latency compared with SOD1G93A mice. Survival was prolonged in Crmp1ki/ki/SOD1G93A mice but not in Crmp1−/−/SOD1G93A mice. In agreement with these phenotypic findings, residual motor neurons and innervated neuromuscular junctions (NMJs) were comparatively well-preserved in Crmp1ki/ki/SOD1G93A mice without affecting microglial and astroglial pathology. Pathway analysis of proteome alterations showed that the sirtuin signaling pathway had opposite effects in Crmp1ki/ki/SOD1G93A and Crmp1−/−/SOD1G93A mice. Our study indicates that modifying CRMP1 phosphorylation is a potential therapeutic strategy for ALS.


Introduction
Amyotrophic lateral sclerosis (ALS) is a rapidly progressive and fatal neurodegenerative disorder that mainly affects motor neurons in the brain and spinal cord (Taylor et al., 2016). Only a few treatments with limited efficacy are currently available (Kim and Taylor, 2017). Since mutations in SOD1, which encodes superoxide dismutase 1 (SOD1), were first discovered to cause ALS in 1993 (Rosen et al., 1993), at least 27 genes have been found to be definitively associated with familial ALS, including transactive response DNA binding protein (TARDBP), fused in sarcoma (FUS) and chromosome 9 open reading frame 72 (C9ORF72; Chia et al., 2018).
Some of the pathogenic proteins responsible for neurodegenerative diseases have been shown to regulate and modulate diverse protein functions and intracellular pathways through posttranslational modifications such as phosphorylation, acetylation, and methylation. In spinocerebellar ataxia type 1 (SCA1), phosphorylation of ataxin-1, the causative protein for SCA1, plays a critical role in ataxin-1 aggregation (Emamian et al., 2003). Similarly, phosphorylation or acetylation of huntingtin alters its aggregation properties and neuronal toxicity in Huntington's disease (HD; Arbez et al., 2017). In ALS, cytoplasmic aggregates of RNA-binding proteins such as phosphorylated TDP-43 (encoded by TARDBP) or FUS constitute a well-known, major pathologic marker for ALS. Acetylation of TDP-43 inhibits RNA binding and promotes the aggregation of phosphorylated TDP-43 (Cohen et al., 2015). By contrast, phosphorylation of FUS inhibits FUS aggregation and ameliorates FUS-related cytotoxicity, while loss of arginine methylation of FUS promotes FUS aggregation (Hofweber et al., 2018). Beyond the context of these proteins that directly cause disease, the importance of posttranslational modifications is evidenced by the finding that removing phosphorylation sites of neurofilament (NF) delays disease onset and prolongs survival in ALS model mice (Lobsiger et al., 2005). However, there is currently a limited understanding of how protein phosphorylation contributes to ALS pathogenesis. We therefore conducted phosphoproteomic analysis in an ALS model using the SOD1 G93A mouse line, which is the most extensively used type of mouse in the study of ALS.
Through this analysis, we identified Collapsin response mediator protein 1 (Crmp1), previously not known to be highly phosphorylated in ALS model mice. CRMP1 belongs to a family of neuronal phosphoproteins (the CRMPs) and was originally identified as an intracellular protein that mediates semaphorin 3A (Sema3A) signaling (Goshima et al., 1995). CRMPs have been correlated with neurologic disorders such as Alzheimer's disease (Uchida et al., 2005;Petratos et al., 2008;Ikezu et al., 2020), HD (Stroedicke et al., 2015), and schizophrenia (Yamashita et al., 2013;Nakamura et al., 2018;Nomoto et al., 2021). Additionally, Sema3A-CRMPs signaling has been suggested to be involved in ALS pathogenesis because Sema3A is upregulated in the motor cortex of ALS patients and the terminal Schwann cell adjacent to neuromuscular junctions (NMJs) in SOD1 G93A mice (De Winter et al., 2006;Körner et al., 2016). Moreover, elevation of both Crmp4 and Crmp4-dynein complex leads to neuronal death in ALS model mice (Duplan et al., 2010;Maimon et al., 2021) while the inhibition of Crmp2 phosphorylation ameliorates the motor phenotype of SOD1 G93A mice (Numata-Uematsu et al., 2019). However, it remains uncertain whether CRMP1 is involved in the pathomechanism of ALS. The only available data on CRMP1 in ALS is that Crmp1 and Crmp4 are highly abundant in the interactome of M337V mutant compared with wild-type (WT) TDP-43 (Feneberg et al., 2020). CRMP1 regulates neuronal cell migration, dendritic spine development, and synaptic plasticity through CRMP1 phosphorylation. CRMP1 and CRMP2 are phosphorylated by cyclin-dependent kinase 5 (Cdk5) at Ser522 (Cole et al., 2006;Yamashita et al., 2007). This phosphorylation is essential for mediating intracellular Sema3A signaling and primes the subsequent phosphorylation of Thr509, Thr514, and Ser518 residues by glycogen synthase kinase 3b (GSK3b or Gsk3b in mouse; Uchida et al., 2005;Cole et al., 2006). CRMP1 is also phosphorylated by Fyn at Tyr504 but not Ser522 (Kawashima et al., 2021). In this study, we investigated the effects of Crmp1 on disease progression in an ALS mouse model, and found that Crmp1 phosphorylation at Ser522 is a key event in motor impairment in ALS.

Ethics statement
This study was conducted in strict accordance with the Yokohama City University Guide for the Care and Use of Laboratory Animals (permission numbers F-A-19-030 and F-A-16-069), and experimental protocols were approved by the Independent Review Boards of Yokohama City University (permission numbers F-D-21-49 and F-D-18-70).

Animal
C57BL/6N mice for producing zygotes and MCH-ICR mice to act as recipient and foster mothers were purchased from CLEA Japan.
sgRNA was purified by phenol-chloroform extraction and ethanol precipitation, and resuspended in OPTI-MEM (Thermo Fisher Scientific). Chemically synthesized single-stranded DNA with the following sequence was used as donor DNA (Nihon Gene Research Laboratories): 59-GCCAGCTACACCCAAACATGCTGCTCCTGCTCCTT CTGCCAAATCGGCGCCTTCTAAACACCAACCCCCAC CCATCCGGAACCTCCACCAGTCC-39. The mutant sequence to be introduced is underlined. A mixture of sgRNA, Cas9 protein (#632641; Clontech), and donor oligonucleotide was introduced via electroporation into pronuclear stage zygotes generated by IVF. Electroporated embryos were cultured overnight and transferred into the oviducts of 0.5 dpc (days post cotium) pseudopregnant females.

Behavioral analysis
Body weight measurements and rotarod test were performed weekly. For Crmp1 À/À /SOD1 G93A (n = 20, 10 males and 10 females) and Crmp1 ki/ki /SOD1 G93A mice (n = 20, 10 males and 10 females), behavioral analysis began at six weeks of age (6 w) and continued until 26 w. An accelerating rotarod test was performed, using rotation speeds of 5-40 rpm for 300 s. Mice underwent two trials with an intertrial interval of .20 min and measurements of the time elapsed to fall from the rotating cylinder were averaged and recorded (Numata-Uematsu et al., 2019). Survival was defined based on the age at which mice could no longer roll over within 30 s after being placed on their back.
Immunostained cells were analyzed in four random fields/sections using a deconvolution fluorescence microscope system (BZ-X800; Keyence).
The percentage of neuromuscular innervation was measured at 42-96 randomly selected synaptic sites per mouse (n = 3 in each group, all males). Endplate occupancy was determined by assessing the extent of overlap of axon terminal signal (labeled by SV2/2H3) with endplate signal (labeled by a-BTX). The degree of denervation was determined as previously described (Numata-Uematsu et al., 2019). Endplates were scored as "denervated" when ,5% of the endplate was deemed occupied by the axon terminal; "fully innervated" means .95% occupancy; "partially innervated" means intermediate occupancy.

Proteomics
Lumbar spinal cords were isolated from 20-week-old mice that were first sedated by anesthesia with M/M/ B:0.3/4/5 and killed thereafter by rapid decapitation. Dissected spinal cords were stored at À80°C until use. Lumbar spinal cords were sonicated on ice at 20 kHz for 30 s a total of four times in lysis buffer [50 mM NH 4 HCO 3 , 8 M urea, 4% sodium deoxycholate, 1% phosphatase inhibitor cocktail 2 (#P5726; Sigma-Aldrich), 1% phosphatase inhibitor cocktail 3 (#P0044; Sigma-Aldrich), 1% protease inhibitor mix (#03969-21; Nacalai Tesque)] using a Branson cell disruptor. Cleared spinal lysate was obtained by centrifugation at 15,000 Â g for 15 min at 4°C. Proteins were precipitated with four volumes of cold acetone and resuspended in 200 ml of lysis buffer. A total of 600-mg protein extracted from each spinal sample was reduced with 10 mM dithiothreitol and alkylated with 12.5 mM iodoacetamide. Proteins were diluted with three volumes of 50 mM NH 4 HCO 3 before digestion with trypsin (Promega) at an enzyme:substrate ratio of 1:20 overnight at 37°C. Sodium deoxycholate was removed from samples using the phase-transfer surfactant method using ethyl acetate (Masuda et al., 2008). After desalting using an OASISHLB 1 ml (Waters), phosphopeptide enrichment was performed with a homemade TiO 2 -C8 tip column using solution compositions specified by the Titansphere Phos-TiO kit (GL Sciences). Our homemade TiO 2 -C8 tipcolumn was made from a 200-ml pipette tip (D200; Gilson) by layering 3 mg of TiO 2 particles (GL Sciences) on top of C8 disk filters (Empore C8; 3M Corporation). After drying, the peptides obtained were dissolved in 0.1% formic acid and 2% acetonitrile (ACN) and analyzed using a Q-Exactive mass spectrometer (Thermo Fisher Scientific) equipped with an UltiMate 3000 LC system (Thermo Fisher Scientific). Peptides were loaded on a trap column (100 mm Â 20 mm, C18, 5 mm, 100 Å; Thermo Fisher Scientific) and subsequently separated on a Nano HPLC capillary column (75 mm Â 180 mm, C18, 3 mm; Nikkyo Technos). Buffer A was 0.1% formic acid in 2% ACN while buffer B was 0.1% formic acid in 95% ACN. Peptides were eluted with a linear gradient of 2-33% buffer B for 120 min.
Label-free quantitative analysis was conducted using the software Progenesis QI for proteomics (Nonlinear Dynamics). For protein and peptide identification, peak lists were created using the software Progenesis QI for proteomics and searched against mouse protein sequences in the UniProtKB/Swiss-Prot database http:// www.uniprot.org/ using the MASCOT software (Matrix Science). Basic search parameters were as follows: trypsin digestion with two missed cleavages permitted; peptide mass tolerance, 65 ppm; fragment mass tolerance, 60.05 Da; usual variable modifications, methionine oxidation, cysteine carbamidomethylation, protein N-terminal acetylation, and N-terminal carbamylation. For phosphoproteomic analysis, additional variable modification parameters for analyzing MS data were phosphorylation of serine, threonine, and tyrosine. Identifications were filtered at a 1% false discovery rate and significance peptide score !30. Protein interaction analysis was conducted with the online tool STRING (https://string-db.org, default settings; PMID 15608232). Ingenuity pathway analysis (IPA; content version: 60467501/62089861, release date: 2020-11-19/2021-02-17; QIAGEN) was used for pathway analysis. Females and males were used for phosphoproteomics and proteomics, respectively.

Pathway analysis
Genes mapped from significantly upregulated or downregulated peptides and phosphorylated peptides were used to identify cellular and molecular processes, pathways and clusters using STRING and IPA software. Activation z scores were calculated using IPA's z score algorithm to predict the overall activation or inhibition of the functional cellular processes/pathways and upstream regulators identified. A positive z score implies an overall predicted activation of the process/pathway/upstream regulator, whereas a negative z score implies an overall predicted inhibition or downregulation of the pathway/ process/upstream regulator. IPA considers z scores of !2 indicative of significant activation while z scores À2 are indicative of inhibition. Cellular processes/upstream regulators with no z scores imply that IPA cannot generate prediction states for these functionalities.

Statistical analysis
Statistical evaluation of behavioral analysis, immunohistochemistry, and Western blotting results was performed using Prism 8 (GraphPad Software). For Rotarod test and body weight data, a two-way ANOVA with Fisher's LSD test was used at each age. The Kaplan-Meier method was used to analyze survival and onset in each SOD1 G93A mouse strain. Immunohistochemistry and Western blotting results were analyzed using either ImageJ or ImageQuant software (GE Healthcare).

Results
Phosphoproteomic analysis of the spinal cord of SOD1 G93A mice at 20 weeks of age To characterize posttranslational changes associated with ALS, we determined the phosphoproteomic profile of the spinal cord of SOD1 G93A (n = 5) mice and compared it with that of WT (n = 4) mice at 20 weeks of age. Canonical pathway analysis using IPA identified semaphorin neuronal repulsive signaling pathway as one of the major pathways affected in the SOD1 G93A mouse spinal cord (Fig.  1A). Thirteen upregulated and six downregulated phosphorylation sites were specified in 11 and 4 proteins that make up this pathway, respectively (Fig. 1B). Signaling proteins downstream of Sema3A included Crmp1, Crmp2, Crmp5, Gsk3b , and Farp1. For Crmp1 in particular, phosphorylation was significantly higher at both Ser8 and Ser522 (Fig. 1B). The interactome of upregulated phosphopeptides (fold change .1.5, p , 0.05) in SOD1 G93A mice was visualized by STRING: functional protein association networks (https://stringdb.org/; Fig. 1C). We again found changes in phosphoproteins associated with axon guidance, including Crmp1, Crmp5, GSK3b , Cfl1, and Rock1. Sema3A signaling has a well-established relationship with Crmp1 phosphorylation at Ser522, but not at Ser8. Therefore, in this study, we focused on the role of CRMP1 Ser522 phosphorylation in ALS pathogenesis.

The effects of total depletion of Crmp1 and of blocking Crmp1 phosphorylation at Ser522 on phenotypes of ALS mice
To determine the roles of total CRMP1 and CRMP1 phosphorylation at Ser522 in ALS pathogenesis, we employed Crmp1 knock-out (Crmp1 À/À ) mice (Cole et al., 2006) and Crmp1 knock-in (Crmp1 ki/ki ) mice, the latter of which we newly established by introducing the S522A mutation to block Crmp1 phosphorylation at Ser522 using the CRISPR/Cas9 system (Extended Data Fig. 2-1A). Sanger sequencing confirmed successful introduction of the mutation (Extended Data Fig. 2-1B,C), and Western blotting failed to detect Crmp1-Ser522 phosphorylation in Crmp1 ki/ki mice (Extended Data Fig. 2-1D). Increased Crmp2-Ser522 phosphorylation in Crmp1 ki/1 and Crmp1 ki/ki mice compared with WT mice might be attributed to the compensatory effect of Crmp2.
Next, we analyzed Sema3A response in cultured DRG neurons from Crmp1 ki/ki mice. The Sema3A-induced growth cone collapse response of Crmp1 ki/ki DRG neurons at E14-15 was significantly lower than that of WT neurons measured at 1 and 3 U/ml Sema3A (F (1,56) = 20.67, 0.5 U/ml: p = 0.067, 1 U/ml: p = 0.008, 3 U/ml: p = 0.006, two-way repeated measures ANOVA followed by Bonferroni's multiple-comparisons test; Extended Data Fig. 2-1E,F). This result indicates that blocking Crmp1 phosphorylation at Ser522 attenuates the Sema3A signal in mouse DRG neurons. Furthermore, we evaluated the motor function of Crmp1 ki/ki mice and no significant differences were observed between WT and Crmp1 ki/ki mice in a rotarod test (Extended Data Fig. 2-1G). Crmp1 À/À mice have been shown to have normal motor function (Yamashita et al., 2013).
Total depletion of Crmp1 and blocking Crmp1 phosphorylation at Ser522 differentially regulated the sirtuin signaling pathway in SOD1 G93A mice Finally, we investigated the molecular basis for differential clinical and pathologic phenotypes observed between Crmp1 À/À /SOD1 G93A and Crmp1 ki/ki /SOD1 G93A mice. For this purpose, we performed the proteomic analysis of the lumbar spinal cord of Crmp1 À/À /SOD1 G93A and Crmp1 ki/ki /SOD1 G93A mice in comparison with SOD1 G93A mice. Using IPA, we visualized the canonical pathways affected in each SOD1 G93A strain. The top 10 canonical continued Figure 1. Phosphoproteomic analysis of spinal cords from WT and SOD1 G93A mice at 20 weeks of age. A, Top 10 canonical pathways identified based on the molecules that were differentially expressed (max fold change .1.5, ANOVA p , 0.05) between WT and SOD1 G93A mice in phosphoproteomics. B, A list of phosphoproteins in semaphorin neuronal repulsive signaling pathway with significant expression changes in the lumbar spinal cords of SOD1 G93A mice compared with WT mice (max fold change 1.5; p , 0.05). Phosphorylation sites, fold-change levels, and p-value are also shown. C, Protein-protein interaction (PPI) network of differentially upregulated proteins in SOD1 G93A mice visualized by STRING: https://string-db.org/. Red nodes are phosphoproteins associated with axon guidance in reactome pathway. The red arrow indicates Crmp1. pathways affected in Crmp1 À/À /SOD1 G93A and Crmp1 ki/ki / SOD1 G93A mice are shown in Extended Data Figure 3-1A,B, respectively. Extended Data Figure 3-1C shows the results of clustering analysis for SOD1 G93A versus Crmp1 À/À /SOD1 G93A mice and SOD1 G93A versus Crmp1 ki/ki / SOD1 G93A mice. Intriguingly, the sirtuin signaling pathway was differently affected in Crmp1 À/À /SOD1 G93A (downregulated) and Crmp1 ki/ki /SOD1 G93A mice (upregulated).

Discussion
The pathogenesis of ALS involves diverse pathways, including oxidative damage, disruption of protein clearance, mitochondrial dysfunction, apoptosis, axonal transport defects, growth factor deficiency, glial cell pathology, glutamate excitotoxicity, and disruptions in RNA metabolism (Rothstein, 2009;Taylor et al., 2016).
Because phospho-antibody against Ser522 has an identical phosphorylation consensus motif for both Crmp1 and Crmp2, it can discriminate these two Crmps by Western blotting (Extended Data Fig. 2-2A) but not by immunohistochemistry. Despite this limitation, we characterized the types of cells with Crmp1 phosphorylation using immunofluorescence analysis involving anti-phospho-Ser522 Crmp1/2 antibody. As shown in Extended Data Figure 2-2D,E, Crmp1/2 in the ventral horn of the lumbar spinal cord was localized in neurons but not in astrocytes, which indicates that elevated phosphorylation of Crmp1 Ser522 may largely occur in neurons. This is also compatible with the previous finding that glial cells do not express CRMP1 protein (Bretin et al., 2005;Luo et al., 2012).
Research Article: New Research Sema3A signaling hyperactivates the complex of Cdk5 and p25, an activator of Cdk5, in the brain and spinal cord in a mouse model of ALS as well as in ALS patients (Nguyen et al., 2001;Klinman and Holzbaur, 2015;Bk et al., 2019). Excessive Cdk5 activity is associated with induction of neuronal loss (Cheung and Ip, 2012) and Cdk5 inhibition in the motor neurons prevents motor neuronal death in ALS model mice (Bk et al., 2019). Moreover, primary cultured DRG neurons from Crmp1 ki/ki mice expressing the Crmp1 S522A mutant were less sensitive to Sema3A stimulation than those from WT mice (Extended Data Fig. 2-1E,F), in contrast to a previous finding that ectopic expression of the Crmp1 S522D phosphomimetic mutant in DRG neurons potentiated the Sema3A-induced growth cone collapse response (Nakamura et al., 2014).
In addition to the differential Crmp1 phosphorylation between WT and SOD1 G93A mice, expression of total Crmp1 was significantly reduced in the spinal cord of SOD1 G93A mice (Extended Data Fig. 2-2A,C). Therefore, to investigate the effect of the total Crmp1 amount, we analyzed Crmp1 À/À /SOD1 G93A mice and found that deleting Crmp1 in SOD1 G93A mice caused motor function to deteriorate slightly but did not affect survival or body weight. Crmp1 knock-down was previously reported to reduce the number of spinal cord neurons in vitro (Kurnellas et al., 2010). In addition, CRMP1 protein levels were reduced in the brains of HD patients, and those of Crmp1 were decreased in a mouse model (Stroedicke et al., 2015). Moreover, CRMP1 knock-down by siRNA has been reported to enhance misfolding and toxicity of mutant huntingtin in an HD cell model, whereas CRMP1 overexpression shows the opposite effect (Stroedicke et al., 2015). These findings indicate that CRMP1 may have beneficial effects against neurodegenerative diseases, including HD and ALS, although the molecular mechanism of the decrease in CRMP1/Crmp1 in patients and mouse models is currently unknown. Unexpected mild deterioration of clinical (Fig. 2F) and pathologic (Fig. 3A,D) phenotypes in Crmp1 À/À / SOD1 G93A mice may reflect functional redundancy of other CRMP family proteins such as CRMP2.
We performed proteomics analysis followed by IPA analysis to investigate the underlying mechanism responsible for the phenotypic differences between Crmp1 À/À / SOD1 G93A and Crmp1 ki/ki /SOD1 G93A mice (Extended Data Fig. 3-1). The sirtuin signaling pathway was upregulated in Crmp1 ki/ki /SOD1 G93A mice but downregulated in Crmp1 À/À /SOD1 G93A mice. Sirtuin signaling regulates cell survival, energy expenditure, and metabolic control through its energy-sensing and redox-sensing functions, and it is associated with lifespan extension (Tang, 2017). In ALS, the sirtuin signaling pathway has been reported to confer beneficial effects on motor neuron survival, such as promoting autophagy and mitophagy and suppressing protein misfolding and aggregate formation (Watanabe et al., 2014;Tang, 2017). These previous findings are consistent with the fact that the sirtuin signaling pathway reduces SOD1 G93A toxicity in the context of Crmp1 ki/ki but increases it with Crmp1 À/À . However, the mechanism of the relationship between Crmp1 alteration and the sirtuin signaling pathway in ALS model mice remains to be determined.
When considering therapeutic strategies, treatment effects may be enhanced by targeting not only the phosphorylation of CRMP1 but also that of CRMP2, because the inhibition of Crmp2 phosphorylation also ameliorates the motor phenotype of SOD1 G93A mice (Numata-Uematsu et al., 2019). In humans, it is necessary to identify small-molecule inhibitors that selectively block Ser522 phosphorylation of CRMP1 and ideally also that of CRMP2. Alternatively, cell-permeable peptides that compete with CRMP1/2-Ser522 may be effective if used with appropriate cell-targeting vehicles to avoid the development of autoimmunity.
To conclude, we have comprehensively investigated proteins that are specifically phosphorylated in ALS model mice, and found evidence that Crmp1 phosphorylation at Ser522 is likely involved in ALS pathogenesis. Blocking Crmp1 phosphorylation at Ser522 led to improvements in the clinical and pathologic phenotypes of ALS model mice. These improvements were associated with alteration of the sirtuin signaling pathway. In humans, simultaneously suppressing Ser522 phosphorylation of both CRMP1 and CRMP2 may be a potential therapeutic strategy for ALS.