Circular RNA, CDR1as, modulates spinal brosis via the miR-7a-5p/TGF-βR2 axis and the Smads signaling pathway

Damage to the spinal cord is the most serious complication of spinal injury. The method to reduce brogenesis after spinal cord injury, to facilitate repair, is an ongoing hurdle despite advancements in research. Non-coding RNA plays an important role in the progression of many diseases, but the study of its role in the progression of spinal brosis is still emerging. Here, we investigated the function of circular RNAs, specically CDR1as, in spinal brosis and characterized its molecular mechanism and pathophysiology. The presence of CDR1as in the spinal cord was veried by sequencing and polymerase chain reaction assays. Further, gene and protein expression of miR-7a-5p and TGF-βR2 were measured to evaluate their predicted interactions. The combination miR-7a-5p/TGF-βR2 was predicted by bioinformatics and validated using a luciferase reporter assay. The regulatory effects and activation pathways of miR-7a-5p and CDR1as on spinal brosis were subsequently veried by Western blot. miR-7a-5p inhibitor and siCDR1as were transfected and inhibited the effect of siCDR1as. These results indicate that CDR1as/miR-7a-5p/TGF-βR2 interactions may exert important functions and suggest potential therapeutic targets for treating spinal brotic diseases.


Introduction
Over 4,500 years ago, military physicians recognized that a crushed or transected spinal cord could not regenerate; today, despite a long history of conceptual advances, the goal of practical, reproducible, functionally meaningful, and clinically translational spinal cord regeneration has yet to be achieved [1].
War injuries, car accidents, and falls all contribute to the incidence of spinal cord injury SCI [2,3]. Worryingly, as the population ages, the number of recorded cases of traumatic SCIs due to falls has risen from 16 to 30.5% since 2012 [1].The clinical presentation of SCI depends on the severity and location of the injury. Currently, most acute treatments are aimed at stabilizing the spinal cord and restoring homeostasis, while long-term treatments mainly manage the symptoms of maladaptive and secondary complications [3,4]. Regardless of etiology, severity, or complexity, spinal cord injuries exhibit three lesion sites: the non-nerve (interstitial) lesion core, the scar, and the reactive surviving nerve tissue. The compartments are made up of very different cell types that have completely different roles in repair and regeneration [1,5]. The particular cellular biology and molecular mechanisms in each chamber affect axon growth or regeneration in different ways. Understanding these differences is critical to understanding the requirements to achieve or improve regeneration and to design rational, mechanistically targeted interventions [2,3,6]. Based on the theory that an injured axon cannot grow beyond obstructive glial and connective tissue, it is widely believed that after central nervous system injury, scars will form and axonal regeneration will be blocked [7,8]. Fibroblasts are the main connective tissue cells in the body, and these cells provide the structural framework throughout the extracellular matrix and only invade or occur in the central nervous system after injury [9][10][11].
Circular RNAs (circRNAs) are a class of primitive non-coding RNA (ncRNAs) that were rst discovered by electron microscopy in the late 1970s, in human cell lines [12,13]. Initially, these circRNAs were considered to be potentially pathogenic by-products of abnormal splicing or "transcription/splicing noise", and thus they received little attention. However, with the development and improvement of transcriptomics and bioinformatics, doubts about the existence and signi cance of circRNAs have gradually disappeared [14]. Unlike linear RNAs, circRNAs are characterized by a remarkable continuous closed-loop structure, formed by a "back-splicing" event wherein a covalent bond is formed between the splice donor and the splice acceptor [13]. This unique loop structure prevents circRNAs from exhibiting typical mRNA-processing characteristics, such as a poly(A)-tail, which makes them structurally stable and highly resistant to exonuclease degradation [15]. circRNAs have been reported to act as microRNA(miRNA) sponges to competitively modulate miRNA expression levels and interact with RNAbinding proteins (RBPs) and regulate post-transcriptional gene expression, thereby suppressing miRNA function, [14]. Most strikingly, circRNA expression is associated with the occurrence and development of many diseases, including a variety of cancers, endocrinium, cardiovascular disease, skeletal musculature disease, and neurological diseases [15]. However, whether circRNAs play a regulatory role in scar formation after SCI has rarely been reported. As such, there is a need to identify differentially expressed circRNAs and to clarify their general mechanism in spinal brosis.
CDR1as, also called CiRS-7 and circRNA0001878, is one of the earliest discovered and most studied circRNAs [7]. CDR1as and sponging miR-7a-5p have been associated with nervous system development, repair, and diseases [13]. In a previous study, we employed SCI mouse models and performed highthroughput sequencing of the tissue within the lesion epicenter during the acute phase [16]. In the present study, we analyzed the sequencing data and found that both CDR1as and miR-7a-5p were signi cant differentially expressed after SCI. Whether CDR1as/miR-7a-5p interactions have functional effects on spinal brosis remains to be determined. To this end, we constructed a CDR1as/miR-7a-5p /TGF-βR2 targeted regulatory network, based on the sequencing results. Using Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis, we looked for potential functional downstream pathway cascade chain proteins. Furthermore, we tracked the changes of miR-7a-5p and TGF-βR2 expression of the interacting axis in vivo within 14 days of SCI, and then veri ed the regulatory axis and downstream pathways through in vitro experiments. This study adds knowledge to the repertoire of ncRNAs activity during the pathophysiological process of SCI, and provides novel insight for future therapeutic strategies for spinal brosis.

Materials And Methods
Genome Mapping, KEGG Pathways, and ncRNA Regulatory Network Analysis The sequencing data of SCI tissues were derived from previous studies [16], The samples were divided into an SCI group and a Sham group, with three replicates each. The tissue samples were derived from the spinal lesion epicenter of modi ed standard Allen's drop mouse model, 3 days after SCI. Sample reads were aligned to the reference genome from Ensembl or National Center for Biotechnology Information databases using the Tophat 2 [17] package, which initially removed a portion of the reads based on the quality control information accompanying each read and then mapped the reads to the reference genome. Tophat allows multiple alignments per read, builds a database of potential splice junctions, and compares the previously unmapped reads against a database of putative junctions. The aligned read les were processed by in-house scripts, which use the normalized RNA-seq fragment counts to measure the relative abundances of the transcripts. The unit of measurement is fragment per kilobase of exon model per million (FPKM) reads mapped. CIRCExplorer was used for denovo assembly of the mapped reads to circRNAs [18]. The assembled circRNAs from all samples were used to identify unique circRNAs; circRNAs were identi ed if they had a statistical p-value of < 0.05.
For pathway signi cance-enrichment analysis, we used a major public database, KEGG (https://www.kegg.jp/) [19]. To determine the most important biochemical, metabolic, and signal transduction pathways, a hypergeometric test was applied to isolate the signi cantly enriched pathways involving genes that were differentially expressed, compared with the whole-genome background.
An ncRNA regulatory network was constructed to pro le the interactions and functional links among dysregulated messenger RNAs (mRNAs), miRNAs, and circRNAs in the acute stage of SCI. The targets of miRNAs were predicted by TargetScan (http://www.targetscan.org/) by adopting the default parameters, as previously described [20]. CircRNA/miRNA/mRNA interaction networks were constructed using Cytoscape (San Diego, CA, USA).

Construction of mouse SCI model
Eighteen (nine for the SCI group and nine for the Sham group), healthy, male C57BL/6 mice were used for our animal models, following a previously used approach [16]. In brief, in the SCI group, the exposed spinal cord was struck using a 6 g weight, which was dropped from a height of 6 cm (Allen's drop). The sham group had the T8-T10 segments of the thoracic vertebra exposed without an Allen's drop. At each experimental time point (3,7, and 14 days after surgery), three mice from each group were randomly selected. Tissue was then extracted for quantitative reverse-transcription PCR (qRT-PCR) and Western Blot.

Primary cell extraction, Culture, Cytological identi cation and Transfection
Tissue from twenty C57BL/6 mice was used for primary spinal broblast extraction within 3 days of birth. After the isolated spinal cord was cut into pieces, 0.125% trypsin of 10 times volume was added and digested for 15 min in a magnetic mixer at 37℃. Medium was added for neutralization and the mixture was centrifuged at 1,000rpm for 5 min; then, the supernatant was discarded. The pellet was cultured in t-25 culture bottles, placed in an incubator at 37℃ with 5% CO 2 , and cultured in modi ed Eagle's medium (Life Technologies, Carlsbad, CA, USA) containing 10 % fetal bovine serum (Gibco, Brisbane, Australia) and 100 IU/mL penicillin-streptomycin (Gibco). The medium was renewed after 24 hours. The cell density reached about 90% after a week of continuous cell culture. The cells were digested normally with 0.25% trypsin, placed in an incubator for 30-40 min after placing the culture asks back inside the incubator, the culture medium was aspirated, and the non-adhered cells were discarded. Newly made culture medium was added and the cells were placed in an incubator at 37℃ with 5% CO2. The medium was changed 2-3 days later. Fibroblasts were identi ed by immuno uorescence, as previously described [20]. In brief, after washing, permeating, and blocking, cells were incubated with Type I collagen antibody (Abcam, Cambridge, UK) at 4℃, overnight. After incubating with a secondary antibody (Invitrogen, CA, USA) and staining with 4',6-diamidino-2-phenylindole (DAPI; Invitrogen), immuno uorescence was analyzed under a uorescence microscope (Olympus Corporation, Tokyo, Japan). Transfection of siRNA-CDR1as (siB2009231003539885, RiboBio, Guangzhou, China, GCCGTATCCAGGGTTTCCA), miR-7a-5p mimics, miR-7a-5p inhibitor, and their negative controls (RiboBio) was initiated when the cells were 30%-50% con uent. After 48 hours, the cells were stimulated with TGFβ (10 ng/ml, ACRO, Beijing, China) for 48 h.

RNA preparation and PCR validation
All qRT-PCR assays were carried out as previously described [20]. In brief, total RNA was isolated and then veri ed/quanti ed by the OD260/280 absorbance ratio, then reverse-transcribed into cDNA. qRT-PCR was performed using the Applied Biosystems (Wilmington, DE, USA) 7500 RT-PCR system and repeated three times. GAPDH was used as an internal control to normalize relative circRNA and mRNA expression levels.
miRNA expression levels were normalized by U6. The 2 -ΔΔCT method was used for comparative quantitation. The speci c primers used for each target were as follows: mmu-CDR1as-Forward: AATCTATGCCTTCCACAAATG; mmu-CDR1as-Reverse: GACCTTGACAGTGTTGGA; mmu-miR-7a-5p- Protein isolation and Western blot Total proteins were extracted then separated by electrophoresis on sodium dodecyl sulphatepolyacrylamide gels (Solarbio, Beijing, China) and subsequently transferred to 0.22-mm polyvinylidene di uoride membranes (Solarbio). The membranes were blocked with 5% skim milk at room temperature for 1 hour. Then, the membranes were incubated with primary antibodies overnight at 4℃. Incubation with secondary antibodies (Solarbio) was performed the following day for 2 hours at room temperature. The antibodies (anti-Fibronectin, anti-Smad7, anti-p-Smad2, anti-p-Smad3, anti-Smad2/3, and anti-GAPDH antibody) used in this study were formulated as previously described [20,21]. The protein levels were visualized using the West Pico ECL Substrate (Solarbio), and GAPDH was used as an endogenous control for normalization.
Luciferase reporter assay and uorescence in situ hybridization 293T cells were inoculated into 96-well plates at a con uence of 70%. The plasmids pMIR-REPORT-CDR1as (WT) (LC Science, Houston, TX) and pMIR-REPORT-CDR1as (MT1+MT2) (LC Science) were transfected 24 hours later. 0.2g concentration of transfection reagent was used, the nal concentration of microRNA was 100 nM. The DNA, miRNA, transfection reagent, and incubate were diluted and set at room temperature for 5 minutes. The diluted DNA and miRNA were mixed with the transfection reagent (Invitrogen), respectively, and incubated at room temperature for 20 minutes. Approximately 50 μL of culture medium was discarded from each well and 25 μL DNA transfection mixture and 25μL miRNA transfection mixture were added to each well. Data were measured after 2S. Following the addition of 50μL of pre-mixed Stop&Glo reagent to each well. Fluorescence was detected using the Opera Phenix HCS system (PerkinElmer, MA, USA).

Statistical Analysis
Statistical analyses were carried out using GraphPad Prism software, and data are expressed as mean ± standard deviation (SD). Statistical signi cance between the two groups was assessed using independent-samples t-test, while analysis of variance (ANOVA) with post-hoc Dunnett's corrections was performed for comparison between two or more groups. P < 0.05 was considered statistically signi cant.

Identi cation and Validation of CDR1as expression
Information and chromosomal distribution of the differentially expressed circRNAs, Our analysis revealed that CDR1as is down-regulated after SCI, compared with the Sham group (log2 fold-change = -1.61, p 0.05; Figure 1 A-B). CDR1as is transcribed from the antisense strand of the CDR1as gene (ChX:61173982-61195535) ( Figure 1C). CircRNAs are single-stranded, closed-loop RNAs, generally formed by transcribed linear RNA through back-splicing processing. Due to the diversi ed variable shearing during RNA processing, it is often necessary to identify circRNAs by high-throughput sequencing and bioinformatics analysis, including the identi cation of splice junctions and the identi cation of circRNA sequence composition. Forward Primer-1 and Reverse Primer-1 can be combined for PCR to detect the conventional circRNA expression levels and subsequent sequencing of the ampli ed product can identify the exact sequence of splice junction. The splice junction sites of CDR1as were validated by Sanger sequencing (Figure 1D). We designed two sets of primers for CDR1as, to verify the head-to-tail splicing of CDR1as. cDNA and genomic DNA(gDNA) ampli ed results of CDR1as (divergent) and linear transcripts (convergent) primers were compared. However, the divergent primers could only amplify CDR1as using cDNA as templates, and no ampli cation product was observed on using gDNA ( Figure  1E). The expression of CDR1as in the lesion epicenter decreased signi cantly within 3 days after SCI compared with the level of expression in the Sham group, as evidenced by qRT-PCR, (Figure 1F), **p < 0.01.

Prediction of CDR1as/miR-7a-5p/TGF-βR2 network and the downstream Smads signaling pathways
After interaction network analysis, we obtained the predicted targeted binding relationship results of all the differentially expressed circRNAs. due to the space constraints, we only selected CDR1as and six other randomly selected differentially expressed circRNAs (ciRNA948, circRNA14958, circRNA002087, circRNA0005489, circRNA722, and circRNA14885) ( Figure S1). CDR1as/miR-7a-5p/TGF-βR2 is marked with a red-dotted line in the partial plot of the interaction network ( Figure 2). KEGG signaling pathway analysis showed that the TGF-β/Smads signaling pathway-related proteins were signi cantly differentially expressed after SCI. Stringtie enrichment analysis was carried out, showing that there are complex regulatory relationships among the pathway-related proteins. The key proteins are highlighted in red (Figure 3).

Construction and con rmation of miR-7a-5p/TGF-βR2 paired relationship
The relative expression of miR-7a-5p was measured by qRT-PCR on days 3,7, and 14 post-SCI and compared with the expression levels in the sham group. The expression of miR-7a-5p decreased signi cantly 3 days after SCI, and then increased slightly until day 7 post-SCI, then increased slowly until 14 days after SCI. However, expression levels could only recover to one-third of the expression in the sham group on day 14 post-SCI ( Figure 4A). The relative expression of TGF-βR2 was also measured by qRT-PCR and revealed that TGF-βR2 expression was signi cantly higher than that of the Sham group, 3 days after SCI. The level reached its peak at day 7 post-SCI, although the expression level on day 14 was lower than that on day 7. However, expression was still signi cantly higher than that of the sham group ( Figure 4B). miR-7a-5p/TGF-βR2 axis was predicted and constructed using TargetScan ( Figure 4C). The miR-135b-5p speci c target site in the 3'UTR of TGF-βR2 was predicted as GTCTTCC and the binding sites were veri ed by uorescent reporters ( Figure 4D). The expression of TGF-βR2 protein was measured by Western Blot and revealed that the trend of protein expression was consistent with that of RNA ( Figure  4E). **P<0.01, ***P<0.001.
CDR1as/miR-7a-5p/TGF-βR2 axis regulated the spinal broblast brosis Immuno uorescence staining of Collagen I was used to verify primary spinal broblasts. Type I collagen in the cytoplasm of broblasts uoresces green and the nucleus uoresces blue after DAPI staining. After merging, the overlap of the cytoplasm and nucleus of the broblasts could be clearly seen, con rming that the extracted primary broblasts were pure and that there was no contamination with other cell types ( Figure 5A-B). The expression levels of brosis-associated proteins, bronectin and collagen I, were measured by Western Blot after treatment with miR-7a-5p mimics and siCDR1as with or without TGF-β1 transfection, respectively. The results showed that miR-7a-5p mimics signi cantly suppressed the expression of bronectin and collagen I ( Figure 7C). siCDR1as also effectively inhibited the expression of the brosis-associated proteins ( Figure 5D). After the co-transfection of miR-7a-5p inhibitor and siCDR1as in primary spinal broblasts, the results of the Western Blot showed that the effect of miR-7a-5p inhibitor and siCDR1as on brosis had an antagonistic effect, and the use of siCDR1as followed by the use of miR-7a-5p inhibitor weakened the effect of siCDR1as ( Figure 5E).
CDR1as/miR-7a-5p/TGF-βR2 axis regulated the activation of Smads signaling pathway The effects of miR-7a-5p mimics on Smads signaling were detected in primary spinal broblasts treated with or without TGF-β1, and the expression of TGF-βR2 were detected by Western blot. Additionally, we assessed the activity of the downstream protein Smad2/3 and phosphorylated-Smad2/3. The results of Western Blot showed that miR-7a-5p mimics inhibited the expression of TGF-βR2. Total expression levels of Smad2 and Smad3 were unchanged; however, the expression of phosphorylated Smad2 and Smad3 were inhibited. This con rmed that miR-7a-5p mimics inhibited the phosphorylation of Smad2 and Smad3 ( Figure 6A). Simultaneously, functional effects of siCDR1as were veri ed. SiCDR1as had the same effect as miR-7a-5p mimics on the phosphorylation levels of Smad2 and Smad3 ( Figure 6B). The results of Western Blot showed that co-transfection of the miR-7a-5p inhibitor and siCDR1as attenuated the effect of siCDR1as on inhibiting the Smad pathway ( Figure 5E). Overall, the results indicated that CDR1as promotes brosis through the miR-7a-5p/TGF-βR2 axis via activating the Smads signaling pathway.

Discussion
Damage to the spinal cord is the most serious complication of spinal injury, and recovery of neurological function or motor function after SCI is a worldwide problem [1]. Studies have shown that axonal regeneration after SCI is the basis of nerve-cell repair and the key to functional recovery. The main obstacle of axonal regeneration is closely related to factors such as scar formation and the weakened intrinsic regeneration power of neurons. After SCI, it is common to see brous scarring and glial scar formation. These scars appear at different time points and fuse into a scar mass at about 2 weeks after injury. The inner layer is a dense brous scar and the outer layer is a glial scar, and the entire scar structure is known as the non-neural lesion core [6,8] .The traditional thought is that the glial scar inhibits the axon regeneration process, and previous research from our lab as well others, has con rmed that the glial scar occurs gradually during a certain period of the recovery process. The site of the axon regeneration and connection plays an important role, while brous scarring causes a hypertrophic physical barrier to form. The secretion of NG2 proteoglycan molecules inhibits the formation of a chemical barrier after prolonged SCI periods, thus playing a role in inhibition, severely hampering functional axon regeneration and neural recovery [9,11,22]. miRNA are evolutionary conserved, 18-22-nt long noncoding RNAs [23]. miRNA has diverse targets in different organisms, and there are many complementary sequences that can interact with various proteins [24]. The function of miRNAs are regulated by many factors, among which circRNAs act as a kind of sponge to e ciently target and alter the activity of microRNAs. Thus, they have an important in uence on the activity and function on miRNAs [15]. CircRNAs are a class of covalently bonded non-coding RNA molecules with a loop structure [12]. They are structurally stable, functionally diverse, widely distributed, and are signi cantly expressed in the central nervous system, regulated the storage, sorting, and localization of miRNAs, also act as competitive RNAs for miRNAs in neurite growth and neuron migration [25]. One of the earliest discovered and most well-known circRNA is CDR1as, also termed ciRS-7(functions as miR-7 sponge) or circRNA0001878(circBase) [26]. For example, CDR1as was rst discovered because of its bonding relationship with miR-671 as CDR1as contains a binding site for miR-671. This complementary combination induced Argonaute (AGO)-mediated cleavage of CDR1as. However, if none of the miR-7a-5p binding sites for CDR1as are complementary, for more than 12 nucleotides, AGO-mediated cleavage upon miR-7a-5p binding does not occur [27][28][29]. Interestingly, CDR1as harbors more than 70 conventional binding sites for miR-7 [30]. miR-7a-5p is found in zebra sh, drosophila melanogaster, mouse, rat, and human, suggesting that it has conserved stability across species. Interestingly, miR-7a-5p is transcribed from three loci in the human genome and one locus in the mouse genome [31]. miR-7-targeted transcripts, such as α-synuclein EGFR, RAF1, KLF4, PARP, SP1, and PI3K, play a regulatory role in pathophysiological processes of miR-7, such as in nerve development, nerve injury, central nervous system tumors, and Parkinson's disease [32]. Although some studies have demonstrated that the CDR1as/miR-7a-5p axis plays a potential regulatory role in various systems [33], there have been no studies on the potential role of CDR1as-mediated brosis in spinal broblasts.
Our previous RNA-sequencing results showed that CDR1as was signi cantly down-regulated in the lesion epicenter of the spinal cord. In this study, we rst veri ed the existence of CDR1as in spinal cord tissue, and then used qRT-PCR to prove that the change of CDR1as expression after SCI was consistent with the sequencing results. Next, we constructed a differential expression interaction network based on the sequencing results. We predict the existence of a CDR1as/miR-7a-5p/TGF-βR2 interaction axis. We chose this interaction axis is not only due to the large differences in the expression of interaction axis members before and after SCI, but also based on previous research. We have previously con rmed that TGF-β initiates the process of brosis [21]. We have also successfully used miR-21 to attenuate the phosphorylation of Smads proteins downstream of TGF-βR1/2 [20]. However, the function of miR-21 is to promote the progression of brosis, so in this study, miR-7a-5p, which can competitively bind TGF-βR2, was taken into consideration. TGFBR2 has been involved in various diseases associated with brosis, and studies suggest that its inhibition could reduce the levels of p-Smad2/3 and thereby suppress brosis [7]. We performed KEGG analysis on the sequencing results and found that activation of the Smads pathway was changed differently before and after injury. Furthermore, we conducted enrichment analysis on this pathway and con rmed that there were numerous links between downstream proteins. After validation of the targeted relationship, the functions and pathways of CDR1as/miR-7a-5p/TGF-βR2 were included in the scope of validation. In vitro experiments were done to mimic the conditions observed in in the acute stage of injury, and TGF-β1-stimulated broblasts were used to simulate the physiological state after SCI [21]. miR-7a-5p mimic and siCDR1as were further used separately to stimulate the activated broblasts. Consistent with our prediction, both of them had inhibitory effects on brosis. When we co-transfected miR-7a-5p inhibitor with siCDR1as, the effect of siCDR1as was inhibited. These results con rmed the role of the CDR1as/miR-7a-5P/TGF-βR2 as a regulatory axis.
However, the present study has some limitations. Since there are one-to-many and many-to-one relationships between circRNAs and miRNAs, the targeted binding relationships veri ed in this study may not be unique. Because of the long sequences of circRNAs, a single circRNA can simultaneously combine with multiple miRNAs. For example, circHIPK3 can simultaneously bind nine miRNAs, including miRNA-152 and miRNA-193a [34]. It has been previously reported that CDR1as itself can also sponge miR-135 and inhibit its function [35]. Similarly, a miRNA can target multiple mRNAs simultaneously [14,36]. For example, KLF4, which plays an important role in the neural repair process, can also be targeted and regulated by miR-7a-5p [37].The regulation process of ncRNAs may be a network rather than a single thread. Multiple miRNAs are bound by the same circRNA, and multiple targeted mRNAs are combined with the same miRNA. These interactions form a network, which may have synergistic or antagonistic effects through multiple pathways. In the future, we will screen, and verify, the downstream regulatory networks more accurately and fully elucidate the targeted interactions by combining a variety of circRNAs and miRNA mimics or inhibitors. In conclusion, the present study veri ed the structure and function of the CDR1as/miR-7a-5p/TGF-βR2 interaction axis in spinal brosis. Over-expression of miR-7a-5p could inhibit brosis by targeting TGF-βR2. siCDR1as was found to have the same effect as miR-7a-5p mimics and acted as an inhibitor of brosis molecules, through sponging miR-7a-5p. These ndings provide new insight for the treatment of movement disorders caused by brosis after SCI. Declarations