microRNA‐125b and its downstream Smurf1/KLF2/ATF2 axis as important promoters on neurological function recovery in rats with spinal cord injury

Abstract The purpose of this study is to investigate the role of microRNA‐125b (miR‐125b) and its mechanism in spinal cord injury (SCI) by targeting Smurf1. After loss‐ and gain‐function approaches were conducted in SCI rat models and neural stem cells (NSCs) isolated from foetal rats, the Basso‐Beattie‐Bresnahan (BBB) score was calculated, and related protein expression was determined by Western blot analysis and cell apoptosis by TUNEL staining. NSC viability was detected by CCK‐8, migration abilities by Transwell assay and apoptosis by flow cytometry. The relationship between miR‐125b, Smurf1 and KLF2 was evaluated by dual‐luciferase reporter gene experiments, Co‐IP and in vivo ubiquitin modification assays. Inhibition of miR‐125b and KLF2 and the up‐regulation of Smurf1 and ATF2 were observed in SCI rats. BBB scores were elevated, the expression of Nestin, NeuN, GFAP, NF‐200 and Bcl‐2 protein was enhanced but that of Bax protein was reduced, and cell apoptosis was inhibited in SCI rats after up‐regulating miR‐125b or silencing ATF2. Smurf1 was a target gene of miR‐125b, which promoted KLF2 degradation through its E3 ubiquitin ligase function, and KLF2 repressed the expression of ATF2 in NSCs. The results in vivo were replicated in vitro. miR‐125b overexpression promotes neurological function recovery after SCI.

the human orthologue of the miRNA lin-4, is highly expressed in human cells and tissues, including the brain, thyroid glands and pituitary gland. 9 Previously reported data have demonstrated that miR-125b could promote neurological recovery after SCI in rats. 10 Bioinformatics analysis also confirmed that Smad ubiquitylation regulatory factor-1 (Smurf1) was targeted by miR-125b. Smurf1 is a HECT-type E3 ubiquitin ligase, which has E3 ligase-dependent and -independent activities in a variety of cells. 11 Thus, Smurf1 has been illustrated to induce neuronal necroptosis after lipopolysaccharide (LPS)-induced neuroinflammation. 12 Furthermore, Smurf1 could induce the ubiquitination and degradation of Krüppel-like factor 2 (KLF2) by serving as an E3 ligase. 13 KLF2 was reported to exert neuroprotective effects on ischaemic stroke by modulating the blood-brain barrier function. 14 KLF2 could down-regulate the activating transcription factor-2 (ATF2) to repress the constitutive pro-inflammatory transcription. 15 Taken these findings into account, we therefore suggested that miR-125 may have a pivotal role in SCI via Smurf1/KLF2/ATF2 axis. Thus, we established an SCI rat model and isolated neural stem cells (NSCs) from foetal rats to verify this hypothesis and suggested novel targets for SCI treatment.

| Compliance with ethical standards
The experiments involving animals followed the principles outlined in the National Institute of Health Guide for the Care and Use of Laboratory animals with the ratification of Animal Experiments in School of Public Health, Jilin University.

| Bioinformatics analysis
The downstream target genes of miR-125b in rats were predicted by the following bioinformatics websites: miRDB (http://www.mirdb. org/), TargetScan (http://www.targe tscan.org/vert_72/) and miR-Walk (http://mirwa lk.umm.uni-heide lberg.de/). Because the three websites adopted different scoring mechanisms for binding sites, the jvenn tool (http://jvenn.toulo use.inra.fr/app/examp le.html) was used to obtain the intersection of the three prediction results. The STRING online website (https://strin g-db.org/) was employed to conduct interaction analysis of genes to further screen the target genes of miR-125b, and the analysis results was visualized by the Cytoscape 3.5.1 software. To predict the downstream regulators of genes, the GeneCards database (https://www.genec ards.org/) was employed to find the genes that encode for the interaction of proteins and SCI-related genes. Afterwards, the intersection was obtained by the jvenn tool to screen out the interacting regulatory genes associated with SCI. The coding proteins of genes were classified by Panther website (http://www.panth erdb.org/), and transcription factors were selected for subsequent predictions. To further retrieve the regulatory factors, the co-expression relationship of transcription factors was searched in the Coexpedia website (http://www.coexp edia.org/). Finally, binding sites of transcription factor on gene promoters were predicted through the GeneCards database.

| SCI model in rats
A total of 165 adult male Sprague Dawley (SD) rats (provided by Center of Laboratory Animals, Jilin University, China (license No. SCXK(Ji)2008-0005)), weighing 180-220 g, were collected. In addition, 15 rats were grouped as the normal control, 15 rats received sham operation (only removal of T9-T11 lamina and spinous process), and the remaining rats were adopted for induction of SCI model using the modified Allen method. 16 Rats were kept in a special cage designated for laboratory rats and were allowed to move freely with access to water and food, prior to the experimental procedures. The rats were anaesthetized by intraperitoneal injection of 20 g/L pentobarbital sodium (40 mg/kg). The posterior midline incision was made to remove the T9 spinous process, part of the T10 spinous process and part of the lamina, with the dura mater exposed. A 10-g hammer was allowed to fall freely from a height of 25 mm and strike the dural sac with impact energy of 25 mm × 10 g and a damage diameter of 2.5 mm. The bottom of the colliding rod, in contact with the spinal cord, presented an arc-shaped hollow with a diameter of 2.5 mm, which was identical to the surface of the spinal cord. The successful sign of the strike was the spasmodic swing of the tail, the retracted flapping of the lower limbs and body, and the flaccid paralysis of the lower limbs. After the procedure, the rats were carefully nursed, fed, and the urine was allowed to pass three times a day until the reflex bladder emptying was established. Rats that did not produce the above findings in the modelled group were removed, and the needed rats were supplemented.

| Basso-Beattie-Bresnahan (BBB) scoring
The motor function of hind limbs of modelled and control rats was observed at different time-points, that is at day 3, 7, 14, 21 and 28 of the surgical procedure. All rats were allowed to roam freely on a 3-meterdiameter circular platform, whereas the BBB motor function score was recorded and calculated by two professional trained experimenters. Rats were then placed on the experimental platform to adapt to the environment for 10 minutes, whereas the observation time of each rat was 5 minutes. The experimenters recorded and observed 10 different movements of each rat, including forelimb, elbow joint, trunk, ankle joint, adjacent small joint and tail. The scores recorded by the two different experimenters were averaged and were regarded as the final BBB score of each rat. The score was divided into 3 categories according to the 21 scoring standard: the first category evaluated animals that could not stand (1-7 points) by assessing the activities of each joint of the hind limbs; the second category evaluated the gait and coordination function of the hind limbs (8-13 points); the third category evaluated the fine movements of the claws in the movement (14-21 points). 17,18

| Grouping of SCI rats
The SCI rats were transfected with negative control (NC)-agomir oligonucleotides, miR-125b agomir oligonucleotides, overexpression (oe)-NC lentivirus, short hairpin RNA (sh)-NC lentivirus, sh-ATF2 lentivirus, oe-Smurf1 lentivirus, and oe-ATF2 lentivirus alone or in combination. After the model was successfully developed, oligonucleotides and lentivirus were diluted in PBS and then intrathecally injected into rats. The rats in each group were protectively injected with 3 μL phosphate buffer saline (PBS) mixture containing 0.5 nmol/L transfection, with strict accordance to the instructions; the rats in the control group were injected with 3 μL PBS for 3 consecutive days. 19 All oligonucleotides and plasmids were synthesized by Shanghai Genechem Co., Ltd. The PBS was infused into the heart of rats slowly until the limbs and body of rats were completely stiff and straight.
After perfusion, the spinal canal of the rats was carefully cut (without damaging of the spinal cord) and the spinal cord tissues in the injured area were collected and placed in a 10% neutral formalin solution and the number of samples was recorded.

| Haematoxylin-eosin (HE) staining
After the tissues were fixed and dehydrated, they were embedded using a paraffin embedding machine (SPCC-6D, Dongguan Spectral Standard Experimental Equipment Technology Co., Ltd.) and sectioned into pieces at 2-4 μmol/L. After conventional dewaxing, the sections were hydrated with gradient alcohol, left to stain with haematoxylin for 5 minutes, differentiated with 1% hydrochloric acid alcohol and blued with 5% ammonia water. The sections were then allowed to stain with 0.5% eosin for 1 minutes and sealed with neutral gum after dehydration and clearing. The pathological changes of the injured spinal cord were observed under a 400 × high power optical microscope. The ratio of inflammatory cell infiltration and necrosis area to the whole visual field area in each visual field was calculated by taking 5 random visual fields from each section: 0 point for no lesion, 1 point for < 25%, 2 points for 25%-50%, 3 points for 50%-75% and 4 points for > 75%.

| RNA isolation and quantification
The total RNA content in tissue or cell samples was extracted with the TRIzol reagent (Invitrogen) which was pre-cooled at 4℃.
Complementary DNA was synthesized using a PrimerScript reverse transcription Kit or a cDNA reverse transcription Kit (K1622, Beijing Reanta Biotechnology Co., Ltd.). Afterwards, the product was placed in a refrigerator at −80℃ for PCR reaction to occur and to avoid repeated freezing and thawing. Reverse transcription-quantitative polymerase chain reaction (RT-qPCR) was operated on an ABI7500 Real-Time PCR system (Applied Biosystems) using SYBR ® Premix Ex TaqTM II (Takara). The relative expression level of mRNA or miR was standardized by glyceraldehyde-3-phosphate dehydrogenase (GAPDH) or U6 expression. These values were then raised to the power of 2 (2 −ΔΔCt ) to yield fold expression relative to the reference point. The primers are depicted in Table 1.

| Isolation, culture and identification of NSCs
Foetal SD rats with the gestation age of 14.5-16.5 days were collected in this study. The foetal rats were aseptically obtained and factor (bFGF), 20 ng/mL epidermal growth factor (EGF), 100 U/mL penicillin and 100 μg/mL streptomycin, and seeded into a 25 cm 2 culture flask at the density of 5 × 10 5 cells/mL, followed by incubation at 37℃ with 5% CO 2 and saturated humidity. Half of the medium was changed once every 3 days, and cells were passaged once every 7 days. After subsequent subculture, purified cells were obtained.
The suspended neurosphere with stable growth at the third passage was harvested for subsequent experiments.

| Cell transfection
The cells were seeded onto a 6-well plate for 24 hours prior to trans-

| Co-immunoprecipitation (CO-IP)
The cells were washed twice with pre-cooled PBS to remove the residual medium and serum. Following the complete absorbance of PBS, 1 mL of fresh PBS was added and the cells were transferred into a 1.5 mL centrifuge tube and centrifuged for 5 min at 37℃ and 3000 r/min. The supernatant was subsequently discarded. A Protease inhibitor cocktail (Roche Diagnostics GmbH) was supplemented into the E1A lysis buffer, [250 mmol/L NaCl, 50 mmol/L 4-(2-hydroxyeth yl)-1-piperazineëthanesulphonic acid (HEPES; PH 7.5) and 0.1% NP-40, 5 mmol/L ethylene diamine tetra-acetic acid (EDTA)] which was lysed by a lysis buffer on ice and sonicated with 3% power for 3 minutes, followed by 15-min centrifugation at 4℃ and 12 000 r/min.
Following that, 40 μL of lysate was taken as the control of IB whereas the remaining 650 μL lysate was added with antibodies and mixed for 3 hours at 4℃. The samples were supplemented with 40 μL of protein A/G-Agarose, mixed at 4℃ for more than 8 h and washed with 1 mL of E1A lysis for 10 minutes for 3 times. Furthermore, the samples supplemented with E1A lysis and 2 × sample buffer solutions were boiled at 100℃ for 15 minutes. The samples were then subjected to Western blot analysis with the same antibodies used above.

| Detection of ubiquitination modification in vivo
When the cell density was approximately 80%, the cells were trans- and mixed with protein A/G-Agarose for more than 8 hours at 4℃.
The cells were then washed at least three times with RIPA lysis and allowed to denature with 2 × sample buffer at 100℃ for 15 minutes.
Thereafter, the cells were tested by Western blot analysis (the same antibodies used above) or stored at −20℃.
The cells at the logarithmic growth phase were cultured in a 96-well plate with 5 × 10 3 cells/well for 3 days. The cells were detected from 0 hour after seeding. Afterwards, the cells in each well were cultured with 10 μL of CCK-8 solution for 2 hours. Then, the optical density (OD) value at 450 nm was measured by a microplate reader (Bio-Rad 680, Bio-Rad, Hercules) at 24 hours, 48 hours, 72 hours and 96 hours, respectively, followed by plotting of the cell proliferation curve.

| Flow cytometry
After cells were transfected for 48-hours, they were treated with 0.25% trypsin without EDTA, collected into a centrifuge tube, cen- Cells at a density of 1 × 10 6 were suspended by 100 μL staining solution, shaken and mixed well followed by being incubated at room temperature for 15 minutes. Afterwards, one ml of HEPES buffer was added into cells which were then shaken and mixed completely.
A flow cytometer (Bio-Rad ZE5, Bio-Rad) was adopted to determine cell apoptosis. The maximum absorption wavelength of FITC was at 488 nm, and the excitation wavelength was at 525 nm. The maximum absorption and emission wavelength of the PI-DNA complex were at 535 nm and 615 nm, respectively.

| Statistical analysis
All measurement data were depicted as mean ± standard deviation and analysed by SPSS 21.0 software (IBM) with P < .05 as a level of statistical significance. Data between the two groups were compared by an unpaired t test, and comparisons among multiple groups were performed with one-way analysis of variance (ANOVA) whereas pairwise comparison within groups was conducted by post hoc test. Cell viability at different time-points was compared using twoway ANOVA, and scores at different time-points were compared by repeated-measures ANOVA, followed by Bonferroni post hoc test.

| Overexpression of miR-125b inhibited SCI and promoted the recovery of nerve function in rats
SCI models were initially generated on SD rats, and the BBB score was utilized to evaluate the motor function of rats. As depicted in Figure 1A, the BBB score of SCI rats was decreased compared to that of normal rats and sham-operated rats, which suggested the successful establishment of the SCI rat model. On the other hand, RT-qPCR results showed that the expression of miR-125b was lowered in SCI rats than that in normal rats and sham-operated rats ( Figure 1B). HE staining revealed that SCI rats showed severe deformation of spinal cord, disorder of tissue structure, unclear boundary of grey-white matter, disappearance of Nissl body, cell swelling, vacuole degeneration, pyknosis, fragmentation and dissolution of some nerve nuclei, inflammatory cell infiltration, phagocyte aggregation, glial cell proliferation and formation of scar tissue, and vacuole in damage zone, compared with normal rats and sham-operated rats, which indicated the significant increase in degree of pathological injury ( Figure 1C). According to our TUNEL assay results, apoptosis was considered positive when the nuclei were stained in brown. In contrast to normal rats and sham-operated rats, apoptosis cells in SCI rats were significantly increased (P < .05; Figure 1D). Western blot analysis was performed to detect the protein expressions of neural stem cell marker factor (Nestin), neuron factor (NeuN), GFAP, NF-200 and apoptosis-related factors (Bax and Bcl-2). The expression of Nestin, NeuN, GFAP, NF-200 and Bcl-2 proteins was remarkably lowered, whereas the expression of Bax protein was higher in SCI rats than that in normal rats and sham-operated rats (P < .05; Figure 1E,F), indicating that the proliferation and differentiation of neurons were inhibited whereas apoptosis was increased. In conclusion, we found that miR-125b had a lower expression in SCI rats.
F I G U R E 1 SCI development is repressed and promoted the recovery of nerve function by overexpressing miR-125b in rats. Normal rats and sham-operated rats were used as controls, and SCI rats were treated or not treated with NC-agomir and miR-125b. Furthermore, transfected NC-agomir and miR-125b agomir oligonucleotides were transfected into the SCI modelled rats. The BBB scores were calculated and demonstrated that the BBB score of miR-125b agomir-treated SCI rats was significantly elevated ( Figure 1A).
RT-qPCR results indicated that the expression of miR-125b was enhanced in SCI rats with miR-125b agomir ( Figure 1B). Moreover, HE staining showed that the pathological injury degree of SCI rats was decreased with miR-125b agomir ( Figure 1C). TUNEL assay results documented that treatment with miR-125b agomir potently reduced apoptotic cells in SCI rats ( Figure 1D). The results of Western blot analysis in SCI rats revealed that the expression of Nestin, NeuN, GFAP, NF-200 and Bcl-2 protein was significantly increased; however, the expression of Bax protein was lowered after treatment with miR-125b agomir ( Figure 1E,F). Therefore, our findings revealed that the overexpression of miR-125b could prevent the pathological process of spinal cord injury, promote the expression of NSC-and neuron cell-related factors, inhibit their apoptosis and restore neural function.

| miR-125b targeted down-regulated Smurf1
Following the identification on the critical role of miR-125b in SCI, we then aimed to explore the downstream mechanism of miR-125b in SCI. Bioinformatics websites including miRDB, TargetScan and miRWalk were utilized to predict the downstream target genes of miR-125b, and 55 genes were collected from the intersection of the three prediction results (Figure 2A). After further analysis of gene interaction through the STRING online website and visualization of analysis results using Cytoscape 3.5.1 software, it was found that Smurf1 gene was at the core of the interaction network ( Figure 2B).
Our data from the Targetscan and miRanda websites illustrated that the miR-125b and Smurf1 3'UTR in rats and human had a targeted binding site ( Figure 2C). Dual-luciferase reporter gene assay depicted that the luciferase activity of Smurf1-WT was significantly decreased after the transfection of miR-125b mimic (P < .05), whereas smurf1-MUT exhibited no significant difference (P > .05), indicating that Smurf1 was the target gene of miR-125b ( Figure 2D).
Furthermore, RT-qPCR and Western blot analysis showed that the mRNA ( Figure 2E) and protein ( Figure 2F) expression of Smurf1 in SCI rats was significantly increased, compared with that of normal rats and sham-operated rats (P < .05).
Furthermore, NSCs were selected for the subsequent experiments because of their high differentiation and migration abilities.
NSCs are notably characterized for their continuous differentiation into neurons in the process of normal neuroregulation, to supplement dead or damaged neurons to maintain the normal process of neuroregulation. Thus, NSCs were later transfected with miR-125b mimic, miR-125b inhibitor and their NCs. As shown in the results of Figure 2G-I, the expression of miR-125b was enhanced whereas the expression of Smurf1 declined in miR-125b mimic-transfected NSCs, which was contrary to NSCs transfected with miR-125b inhibitor. Our data showed that Smurf1 was highly expressed in SCI tissues, whereas miR-125b targeted and inhibited the Smurf1 in NSCs.

| Overexpression of miR-125b targeted Smurf1 to promote proliferation and migration but inhibit apoptosis in NSC
To further investigate the effect of miR-125b on the proliferation, migration and apoptosis of NSCs by regulating Smurf1, NSCs were transfected with miR-125b mimic and oe-Smurf1. The results of RT-qPCR and Western blot analysis showed that the expression of miR-125b in NSCs transfected with miR-125b mimic was evidently increased, whereas that of Smurf1 was down-regulated. Further investigation revealed that oe-Smurf1 treatment did not alter the expression of miR-125b; however, the expression of Smurf1 was up-regulated in NSCs. NSCs transfected with NC mimic + oe-Smurf1 exhibited significantly elevated expression of miR-125b in comparison with NSCs transfected with miR-125b mimic + oe-Smurf1, illustrating the lowered expression of Smurf1 ( Figure 3A-C). Figure 3D), Transwell assay ( Figure 3E) and flow cytometry ( Figure 3F) revealed that miR-125b mimic enhanced the viability and migration abilities of NSCs although significantly reducing apoptosis in NSCs; however, the outcome was reversed after Smurf1 was overexpressed. In addition, the effects of oe-Smurf1 on NSC viability, migration and apoptosis were negated by the additional treatment with miR-125b mimic. Hence, these results demonstrated that miR-125b could induce proliferation and migration, although repressing the apoptosis in NSCs by targeting Smurf1.

| Smurf1 promoted the degradation of KLF2 through its E3 ubiquitin ligase function
We aimed to explore the effects of KLF2 on the expression of Smurf1 during SCI. For this purpose, we used the GeneCards database to find 1527 interacting proteins with Smurf1 ( Figure 4A) and 1803 genes related to SCI; 238 interacting genes related to SCI were obtained from the intersection ( Figure 4B). Twenty-seven transcription factors were identified by classifying the encoded proteins of the genes on the Panther website ( Figure 4C). We then obtained a coexpression network of transcription factors through the Coexpedia database, with KLF2 at a key position ( Figure 4D). RT-qPCR and Western blot analysis were conducted to detect the expression of KLF2 in SCI rats. Our results demonstrated that the expression of KLF2 in SCI rats was significantly lowered compared with that of normal rats and sham-operated rats (P < .05; Figure 4E,F). Pearson's correlation analysis depicted a negative correlation between the expression of KLF2 and Smurf1 in SCI rats (n = 15) (P < .05; Figure 4G).
Intriguingly, the results of Co-IP experiments suggested that overexpressed IP Smurf1 was specifically bound to KLF2 protein, whereas overexpressed IP KLF2 was specifically bound to the Smurf1 protein, indicating that Smurf1 and KLF2 could interact with each other ( Figure 4H). Next, we overexpressed Smurf1 in NSCs and added either the proteasome inhibitor MG132 or DMSO. According to Figure 4I,J, Smurf1 mRNA and protein expression was noticeably increased; however, KLF2 was markedly lowered in NSCs treated with DMSO + oe-Smurf1 (P < .05) compared with that of NSCs treated with oe-NC + DMSO; there was no significant difference in the expression of Smurf1 and KLF2 mRNA in NSCs treated with oe-NC + MG132 (P > .05), whereas the expression of Smurf1 and KLF2 protein was markedly enhanced and the level of KLF2 protein was more significantly elevated (P < .05); the expression of Smurf1 mRNA and protein was remarkably increased in NSCs treated with oe-Smurf1 + MG132, whereas that of KLF2 mRNA was decreased but its protein level was increased (P < .05). In contrast to NSCs treated with oe-NC + MG132, the expression of Smurf1 mRNA and protein in NSCs treated with oe-Smurf1 + MG132 was significantly elevated, whereas the expression of KLF2 mRNA and protein was decreased (P < .05). Moreover, NSCs with overexpressed levels of Smurf1 exhibited a large number of Ub molecules that were transferred by KLF2, thus producing a classical ubiquitin UB molecular distribution band; however, in cells treated with MG132, there were no Ub molecules transferred by KLF2 (P < .05; Figure 4K), indicating that Smurf1 promoted the degradation of KLF2 through its E3 ubiquitin ligase function. In summary, our data suggested that the expression of KLF2 in SCI rats was lowered, whereas Smurf1 could promote the degradation of KLF2 through its E3 ubiquitin ligase function.

| Smurf1 inhibited proliferation and migration and promoted apoptosis of NSCs by promoting the degradation of KLF2
KLF2 and Smurf1 were overexpressed in NSCs to investigate whether Smurf1 could promote the degradation of KLF2 to regulate the proliferation, migration and apoptosis abilities of NSCs. Our results from the RT-qPCR and Western blot analysis revealed that oe-Smurf1 treatment enhanced the expression of Smurf1, but the expression of KLF2 was declined in NSCs; oe-KLF2 treatment did not influence the expression of Smurf1, but KLF2 expression was increased in NSCs. Transfection with oe-Smurf1 + oe-KLF2 increased the expression of Smurf1, whereas the expression of KLF2 was diminished in NSCs, compared with transfection with oe-NC + oe-KLF2 ( Figure 5A,B). As demonstrated in Figure 5C-E, the overexpression of Smurf1 decreased the viability and migration properties of NSC but increased NSC apoptosis, which was contrary after the overexpression of KLF2. Moreover, the effects of oe-KLF2 on the viability, migration and apoptosis properties of NSC were nullified by additional treatment with oe-Smurf1. Therefore, these results indicated that Smurf1 induced apoptosis but repressed the proliferation and migration abilities in NSCs by degrading KLF2.

| KLF2 inhibited the expression of ATF2 to promote proliferation and migration but suppress apoptosis in NSCs
The downstream mechanism of KLF2 in SCI was investigated. ATF2 was found to be a transcription factor with the binding site on the KLF2 gene promoter, according to the GeneCards database. RT-qPCR and Western blot analysis demonstrated that the expression of ATF2 in SCI rats was significantly higher than that in normal rats and sham-operated rats (P < .05; Figure 6A,B). Pearson's correlation analysis showed that the KLF2 level was negatively correlated with the expression of ATF2 in SCI rats (15 rats) (P < .05; Figure 6C).
To verify whether KLF2 could inhibit the expression of ATF2, KLF2 and ATF2 were overexpressed in NSCs cells, followed by conducting RT-qPCR and Western blot analysis. Our results indicated that after up-regulating KLF2, KLF2 expression was elevated; however,

F I G U R E 3
The overexpression of miR-125b or silencing of Smurf1 induces proliferation and migration but represses apoptosis in NSCs. NSCs were transfected with NC mimic + oe-NC, miR-125b mimic + oe-NC, NC mimic + oe-Smurf1 or miR-125b mimic + oe-Smurf1. A, Expression of miR-125b in NSCs measured by RT-qPCR normalized to U6. B, Smurf1 mRNA expression in NSCs measured by RT-qPCR normalized to GAPDH. C, Expression of Smurf1 protein in NSCs detected by Western blot analysis normalized to GAPDH. D, NSC viability measured by CCK-8. E, NSC migration evaluated by Transwell assay. F, NSC apoptosis assessed by flow cytometry. * P < .05 vs NSCs transfected with NC mimic + oe-NC, # P < .05 vs NSCs transfected with NC mimic + oe-Smurf1. Comparisons among multiple groups were performed with one-way analysis of variance (ANOVA), and cell viability at different time-points was compared using two-way ANOVA. The cell experiment was repeated three times

| Smurf1 inhibits the recovery of neurological function after SCI in rats by promoting the degradation of KLF2 and up-regulating ATF2
The aforementioned experiments showed that Smurf1 could promote the degradation of KLF2, whereas KLF2 decreased the expression of ATF2, speculating that Smurf1 could up-regulate ATF2 by promoting the degradation of KLF2. To study the mechanism through which the Smurf1/KLF2/ATF2 axis affects SCI and neurological function recovery in vivo, we first performed RT-qPCR to verify the silencing  Figure 7A). Thus, sh-ATF2#1 was chosen for the subsequent experiments. Smurf1 was subsequently overexpressed, and ATF2 was silenced in SCI rats. Silencing ATF2 increased the BBB scores, which was reversed by the overexpression of Smurf1 ( Figure 7B). RT-qPCR and Western blot analysis indicated no significant difference observed in the expression of Smurf1 and KLF2, in comparison with that of the treatment with oe-NC + sh-NC; however, the expression of ATF2 was declined after treatment with oe-NC + sh-ATF2.
Moreover, the treatment with oe-Smurf1 + sh-ATF2 elevated the expression of Smurf1 but reduced the expressions of KLF2 and ATF2 in SCI rats. In contrast to SCI rats treated with oe-NC + sh-ATF2, the expressions of Smurf1 and ATF2 in SCI rats treated was decreased in ATF2-silenced SCI rats, which was nullified in the presence of additional of oe-Smurf1 ( Figure 7G,H). These results suggested that the overexpression of Smurf1 promoted KLF2 degradation to up-regulate ATF2, thus promoting SCI and inhibiting the recovery of neural function in SCI rats.

| Up-regulation of miR-125b promoted the recovery of neurological function in SCI rats via Smurf1/KLF2/ATF2 axis
We further aimed to investigate whether the up-regulation of miR-125b regulated the Smurf1/KLF2/ATF2 axis to inhibit SCI and promote the recovery of neural function in vivo. For this specific purpose, miR-125b and ATF2 were overexpressed in SCI model rats. As shown in Figure 8A, miR-125b agomir treatment significantly enhanced the BBB scores. However, the overexpression of ATF2 reduced BBB scores but its results were reversed by the addition of miR-125b agomir. Our results of RT-qPCR and Western blot analysis showed that the expression of miR-125b and KLF2 in SCI rats treated with miR-125b agomir + oe-A-NC was significantly enhanced compared with that of SCI rats treated with NC-agomir + oe-A-NC, but the expression of Smurf1 and ATF2 was noticeably diminished (P < .05). On the contrary, no significant difference was observed in the levels of miR-125b, Smurf1 and KLF2 in SCI rats treated with NC-agomir + oe-ATF2 (P > .05), whereas the expression of ATF2 was substantially elevated (P < .05); the expression of miR-125b, KLF2 and ATF2 in SCI rats treated with miR-125b agomir + oe-ATF2 was markedly increased, whereas the expression of Smurf1 was decreased (P < .05). In comparison with SCI rats treated with NC-agomir + oe-ATF2, miR-125b and KLF2 expression were increased whereas the expression of Smurf1 and ATF2 was decreased in SCI rats treated with miR-125b agomir + oe-ATF2 (P < .05; Figure 8B-D). HE staining demonstrated that the degree of pathological injury was reduced by miR-125b agomir but was elevated by oe-ATF2, which was rescued by additional miR-125b agomir ( Figure 8E). Moreover, miR-125b agomir-treated SCI cells exhibited a reduction of apoptotic cells, whereas oe-ATF2-treated SCI rats showed an increased activity of cell apoptosis; however, the result was nullified by the additional treatment with miR-125b agomir ( Figure 8F)  Bcl-2 was reversed by additional treatment with miR-125b agomir ( Figure 8G,H). Collectively, the overexpression of miR-125b mediated the Smurf1/KLF2/ATF2 axis, thus inhibiting SCI and promoting the recovery of neural function in SCI rats.

| D ISCUSS I ON
Spinal cord injury is characterized by neurological deficit, and the major focus of its cure is often emphasized on the regeneration of axons in the central nervous system. 20 Recently, miRs have been reported to mediate various neurobiological processes, like cell proliferation, growth, differentiation and neural activity, along with the pathogenic processes of SCI, including apoptosis, demyelination, oxidation and inflammation. 21 miR-125b is also known to promote neurological recovery in rats with SCI. 10 Although each miRNA is governed by many potential targets, it still remains a challenge to understand the roles of miR in SCI that depend on the identification of bona fide molecular targets. Based on the above, we intended to elucidate the mechanism by which the biological function of miR-125b influences the neurological recovery after SCI.
Our data revealed that miR-125b up-regulated KLF2 by targeting Smurf1, thus repressing ATF2 and promoting neurological recovery in SCI rats. Initially, our study uncovered the down-regulation of miR-125b and KLF2, as well as the up-regulation of Smurf1 and ATF2 in SCI rats. However, after the enforced expression of miR-125b or KLF2 or silencing of Smurf1 or ATF2, neurological function recovery was promoted in SCI rats. It is also well-known that miRs play a critical role in the progression of neurological disease by regulating neuronal communication in a tight balance. 7 Mounting researches have demonstrated that miR-125b promotes neurological function recovery in neurological diseases. For instance, a study conducted by Quiroz et al uncovered that miR-125b promoted the neurological function recovery in SCI rats. 10 In addition, miR-125b mimic might prevent neuroinflammation and aberrant p53 network activation-induced apoptosis during ischaemia-reperfusion injury by down-regulating TP53INP1. 22 Also, miR-125b plays a crucial role in orchestration of cell proliferation, differentiation and migration in neural stem/progenitor cells by targeting Nestin. 23 Another research unravelled the depressive role of miR-125b in cerebral ischaemiareperfusion injury by blocking Bax/Cytochrome C/Caspase-3 apoptotic pathway. 8 Additionally, another study elucidated that the up-regulation of KLF2 was a promoter of neurological function recovery in rats with subarachnoid haemorrhage. 24 In consent with our results, it has been documented that high expression of Smurf1 was observed after SCI in rats. 25 The up-regulation of ATF2 expression has also been reported in injured L5/6 dorsal root ganglia and spinal cord after nerve injury. 26 Our results from the dual-luciferase reporter gene assay depicted that Smurf1 was a putative target gene of miR-125b. In consistent with previously reported data, our study confirmed that Smurf1 promotes the degradation of KLF2 via its E3 ubiquitin ligase function. 13 It was also reported that KLF2 repressed the expression of ATF2, 15 which was in line with our results.
Our results exhibited that overexpressing miR-125b or KLF2 and silencing Smurf1 or ATF2 promoted the proliferation and migration in NSCs, but repressed the activity of apoptosis in NSCs, respectively. A recent study has shown that the overexpression of miR-125b significantly promoted the differentiation and migration abilities of NSC. 23 Of note, the up-regulation of miR-125b attenuated the activity of cardiomyocyte apoptosis induced by hypoxia in vitro; a decrease in activity of cardiomyocytes apoptosis was also confirmed in vivo after acute myocardial infarction. 27 Furthermore, it has been demonstrated that the ectopic expression of Smurf1 could enhance the neuronal necroptosis to promote LPS-induced neuroinflammation. 12 Ling et al observed that the overexpression of KLF2 could promote pancreatic acinar cell proliferation, but cell apoptosis was repressed in acute pancreatitis following the overexpression of KLF2. 28 Taken together, these reported findings speculated that miR-125b had promoted proliferation and migration, whereas apoptosis was inhibited in NSCs via Smurf1/KLF2/ATF2 axis, thus alleviating SCI.
Our data indicated that miR-125b served as a potential therapeu- warrants further experiments about SCI on spinal cord neurons.

ACK N OWLED G EM ENTS
We acknowledge and appreciate our colleagues for their valuable efforts and comments on this paper.

CO N FLI C T O F I NTE R E S T
The authors declare no competing interest.