Adult NSCs were isolated from the spinal cords of six patients as previously described (Mothe and Tator, 2015). All procedures have been approved by the ethics committee of Shanghai Changzheng Hospital, The Second Military Medical University and were in compliance with the research protocols involving human subjects. The periventricular spinal cord tissues were obtained from organ transplant donors with written informed consent. In brief, tissues were aseptically harvested and minced into pieces followed by digestion with Papain (20 U/ml) containing DNase I, 1 mM L-cysteine and 0.5 mM EDTA. Cell suspensions were washed once and centrifuged through a discontinuous density gradient to remove membrane fragments. The cell pellet at the bottom were resuspended and filtered through cell strainer before plating in pre-coated culture plates in serum-free Neurobasal-A medium (Thermo Fisher Scientific, Waltham, MA, United States) containing 20 ng/ml EGF, 20 ng/ml bFGF, 2 mM L-glutamine, 100 μg/ml penicillin-streptomycin, 2% B27, and 2 μg/ml heparin. The cells were cultured at 37°C with 5% humidified CO₂, in the presence of vehicle controls or 10 μM FA (Sigma–Aldrich, St. Louis, MO, United States) for 24 h (Lin et al., 2010). Hypoxic conditions were established by treating cells with 100 μM CoCl₂ for 24 h. After treatments, adherent NSCs were harvested by trypsinization and resuspended in PBS containing 4% fetal bovine serum.

Male Sprague Dawley rats, aged 6–8 weeks and weighing 200–220 g, were used in the study with approval from the animal research review committee of Shanghai Changzheng Hospital, The Second Military Medical University. All methods were performed in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. In a spinal cord contusion injury model, the rats were anesthetized using 3% isoflurane through inhaling, and subsequently received dorsal laminectomy at the level of the ninth thoracic vertebra (T9) followed by contusion injury using the NYU impactor (Gruner, 1992), in which a 10 g weight impact rod was dropped from a height of 15 cm. The exposed spinal cord was then intrathecally administered through a 30-gauge needle with 4 μl of cultured NSCs (5 × 10⁵ cells/4 μl) at the site of the injury. Skin was closed by sutures and the animals were allowed to recover on a heating pad. Postoperative treatments included saline for rehydration and Baytril to prevent urinary tract infection. Following transplant the rats were orally administered with 100 mg/mL enrofloxacin (Bayer Animal Health) in drinking water. Rats were assigned in the following experimental groups: (1) sham (n = 12), rats underwent sham operations; (2) SCI (n = 12), rats underwent SCI operations; (3) SCI+NSC (n = 12), rats underwent SCI operations and injected with PBS-treated NSCs; (4) SCI+FA-NSC (n = 12), rats underwent SCI operations and injected with FA pre-treated NSCs; (5) SCI+FA-NSC-control (n = 12), rats underwent SCI operations and injected with FA pre-treated NSC stably expressing control miR; (6) SCI+FA-NSC-miR-590 (n = 12), rats underwent SCI operations and injected with FA pre-treated NSC stably expressing miR-590.

Extent of SCI was evaluated at indicated days post SCI, using Basso, Beattie and Bresnahan (BBB) scores. The locomotor function of rats was determined based on the BBB scale as described (Buller et al., 2010). In brief, locomotor activity was examined over a period of 4 min by independent observers in a blinded manner.

Spinal cord edema was assessed by the determination of spinal cord water content, which was calculated as follows: spinal cord water % = (wet weight - dry weight)/wet weight × 100%. The dry weight was measured after the injured spinal cords were dried for 24 h at 110°C using the Havppvel method.

Spinal cord tissues were harvested and weighed at indicated days post SCI. The tissues were processed by MPO (myeloperoxidase) assay kits (Sigma–Aldrich, St. Louis, MO, United States) for determination of the extent of polymorphonuclear leukocyte (PMN) infiltration. The MPO activity was measured by the rate of change in absorbance at 460 nm and expressed as units of MPO (the quantity of MPO enzyme to degrade 1 μmol of peroxide per minute) per gram of wet tissues.

Expression of miR-590 was measured with mature miRNA assay kit (478168_mir; ThermoFisher Scientific, Waltham, MA, United States) according to manufacturer’s instructions. The MISSION Lenti miR-590 mimic (HLMIR0815) was purchased from Sigma–Aldrich, and were packaged for transduction to create stable cell lines according to manufacturer’s instructions.

Total isolated RNA from cell cultures was reverse transcribed to complementary cDNAs using Superscript II according to manufacturer’s instructions (Bio-Rad, Hercules, CA, United States). SYBR Green dye-based detection method was used by using the SYBR Green PCR Master Mix assay (Applied Biosystems, Waltham, MA, United States). A series of dilutions of control cDNA were used to generate the standard curves and validate the melting curves for each primer set. Triplicated PCR reactions were carried out for each sample. GAPDH was used as a housekeeping gene for normalization. Primers used in the current study were: VEGF forward 5′-GAG ATG AGC TTC CTA CAG CAC-3′, reverse 5′-TCA CCG CCT CGG CTT GTC ACA-3′; GAPDH forward 5′-AGG GCT GCT TTT AAC TCT GGT-3′, reverse 5′-CCC CAC TTG ATT TTG GAG GGA-3′.

For the evaluation of miR-590 targeting potential, NSCs were transfected with pGL3 luciferase reporter constructs harboring the miR-590 target sequence (wild type or mutant) for the 3′-UTR of VEGF mRNA. For transcription reporter assay of miR-590, the putative promoter region was selected based on the upstream sequence of miR-590 gene. The promoter region of miR-590 contains one potential HRE, and has been cloned into pGL3 luciferase reporter plasmid at the upstream of Luc open reading frame. Mutagenesis was done by using a QuikChange site-directed mutagenesis kit (Stratagene), and confirmed by DNA sequencing. All transfections were performed by Nucleofector system (Lonza, Basel, Switzerland), and the activities of firefly luciferase and renilla luciferase (as internal control) in the cell lysates were measured with the Dual-Luciferase Assay System (Promega, Madison, WI, United States).

Western blot was performed in cultured cells following various treatments. The protein lysates (1% NP40, 50 mM Tris, 5 mM EDTA, 1% SDS, 1% sodium deoxycholate, 1% Triton X-100, 1 mM PMSF, 10 mg/ml aprotinin, 1 mg/ml leupeptin, and pH = 7.5) were measured by Bradford assay and the same amount of proteins was resolved on SDS-PAGE followed by an electric transfer to a PVDF membrane. The blots were blocked by 5% non-fat milk and incubated with primary antibodies, including HIF-1α (ab51608, 1:1000; Abcam, Cambridge, MA, United States), VEGF (ab51745, 1:2000; Abcam, United States) and GAPDH (ab9484, 1:1000; Abcam, United States). The blots were then incubated by appropriate HRP conjugated secondary antibodies, and signals were visualized by an enhanced ECL-based imaging system.

Interaction of HIF-1α to the HREs at the promoter of miR-590 gene was examined by ChIP assay. Briefly, cells were cross-linked with 1% formaldehyde, sheared to an average size of 400 bp, and subsequently immunoprecipitated with antibodies against HIF-1α (Abcam, ab2185) or control IgG. Immunoprecipitated DNA was then isolated for ChIP-PCR using primers for the HRE sites in the promoter region of miR-590. A positive control antibody (RNA polymerase II) and a negative control non-immune IgG were used to demonstrate the efficacy of the kit reagents (Epigentek Group Inc, P-2025-48). Binding efficiency (Bound/RPII%) was calculated by using a ratio of amplification efficiency of the ChIP sample over Polymerase RNA II as follows: 2(RPII CT – Sample CT) × 100%.

All in vitro data were generated from at least three independent experiments, and all in vivo data were generated from n = 12 rats each experimental group. Values were expressed as means ± SD. Two-tailed Student t-test or one-way ANOVA with Tukey’s post hoc test were used when appropriate to evaluate the data for statistical comparison, and P < 0.05 indicates statistically significant differences.

Human NSCs used in the study were isolated from human patients, and expressions of NSC specific markers CD44, CD271, CD24, and CD184 were analyzed. As shown in Figure 1A, the isolated NSCs were indeed negative for CD44 and CD271, and positive for CD24 and CD184, confirming their NSC identity (Yuan et al., 2011). Immunofluorescent staining with an NSC-specific marker protein nestin further confirmed the identity of the isolated NSCs (Figure 1B).

Next, a rat model of SCI was established (Figure 2A, and Materials and Methods), and SCI-operated rats received transplant of NSCs with or without FA pre-treatment. We then evaluated the function recovery by monitoring assessing BBB behavioral scores in all four experimental groups of rats. BBB score of SCI-operated rats was significantly reduced throughout the 21 days evaluation period, indicated the successful establishment of SCI rat model (Figure 2B). SCI-operated rats also received transplant of untreated NSCs (SCI+NSC) exhibited gradual recovery of BBB scores starting from day 7 to day 21 after SCI operation, compared to that of SCI rats, indicated the effectiveness of NSC transplant therapy in reparing SCI damage. Importantly, SCI rats receiving FA pre-treated NSCs (SCI+FA-NSC) experienced a much more rapid recovery of BBB score. By day 21, BBB score of SCI+FA-NSC rats was significantly higher than SCI+NSC rats, which suggested that FA pre-treatment further enhanced the effect of NSC transplant therapy in reparing SCI.

Water content and myeloperoxidase (MPO) are two indicators of SCI, both of which were reported to significantly increase after SCI (Mbori et al., 2016). Therefore, we also examined levels of these two parameters in the spinal cord (Figures 2C,D), and found both parameters exhibited the same trend as that of BBB scores: SCI rats exhibited significantly increased water content and MPO level than sham rats, while SCI+NSC rats displayed recovery to some extent, and SCI+FA-NSC rats showed even more significant recovery compared to SCI+NSC.

Next, we aimed to uncover the molecular mechanism underlying FA actions in improving SCI recovery following NSC transplant therapy. As a recent study implicated a potential role of hypoxia-induced VEGF in SCI (Long et al., 2015), we therefore examined the effect of FA treatment on expression of VEGF and HIF-1α in the isolated NSC culture. As shown in Figures 3A,B, expressions of VEGF mRNA and protein, as well as HIF-1α protein, were significantly increased by FA treatment. We further reasoned that, since microRNA (miRNA, miR) functions to inhibit target gene expression (Bartel, 2009), they might also be involved in FA action and regulate VEGF expression. We therefore performed miRNA microarray analysis in isolated NSCs in the presence or absence of FA treatment, and have found 11 miRNAs downregulated by FA treatment (Figure 3C). Among these downregualted miRNAs, miR-590 was particularly interesting because its expression was reported to be negative correlated with VEGF (Zhou et al., 2016). We next verified the microarray data on miR-590 expression with miRNA assay, and confirmed that FA treatment indeed could significantly downregulate miR-590 expression in isolated NSCs (Figure 3D).

Given our above results that FA treatment could increase VEGF and decrease miR-590, we next investigated the relationship between the two. By using TargetScanHuman in silicon analysis (Lewis et al., 2005), we found a putative miR-590 targeting site on the 3′-UTR of VEGF mRNA (Figure 4A). Luciferase reporter assay was then employed to test whether this predicted sequence was responsible for the negative correlation between miR-590 and VEGF, where wild type (LF-wt-VEGF) or mutated (LF-mut-VEGF) sequences from VEGF mRNA 3′-UTR were cloned to the 3′-prime of luciferase reporter gene (Figure 4B). Stable expression of miR-590 was also introducted into the isolated NSCs (Supplementary Figure S1). As shown in Figure 4C, luciferase activity of LF-wt-VEGF was significantly reduced by miR-590 expression in NSCs, whereas that of LF-mut-VEGF was unaffected, indicating that miR-590 was able to recognize the predicted targeting site on 3′-UTR of VEGF mRNA. Direct examination by Western blot analysis ialso confirmed that VEGF protein was indeed greatly inhibited by miR-590 expression in isolated NSCs (Figure 4D).

Besides between miR-590 and VEGF, another negative correlation also existed between miR-590 and hypoxia/HIF-1α. Indeed we found a putative consensus HRE site (Wenger and Gassmann, 1997) in the promoter region of miR-590 gene (Figure 5A). Subsequently, a ChIP assay was performed to test the direct binding of HIF-1α to the HRE in miR-590 promoter (Figure 5B). Under normoxia condition, a slightly enhanced binding of HIF-1α, compared to control IgG, to the HRE region could be observed. While under hypoxia condition, the binding between HIF-1α and the HRE was markedly increased compared to normoxia. To verify whether expression of miR-590 itself was regulated by hypoxia via this HRE, the predicted wild type HRE sequence (Wt-HRE-LF) and its mutated version (Mut-HRE-LF) were cloned to the 5′-prime of luciferase reporter gene (Figure 5A). As shown in Figure 5C, luciferase activity of Wt-HRE-LF was significantly reduced by hypoxia in NSCs, whereas Mut-HRE-LF was unchanged. MiRNA assay also confirmed that miR-590 expression was indeed significantly repressed hypoxia (Figure 5D). The above results indicated that in isolated NSCs, hypoxia could repress miR-590 via the HRE in its promoter region.

Next, we induced hypoxia in NSCs stably expressing miR-590, and examined the effect on VEGF expression (Figures 6A,B). Compared with NSCs expressing control miR under the same conditions, hypoxia failed to elevate either miRNA or protein expressions of VEGF in NSCs expressing miR-590, indicating that hypoxia regulation on VEGF was indirect and was mediated by miR-590 repression. In addition, expression of miR-590 was also able to inhibit FA-induced VEGF upregulation (Figures 6C,D), further confirming that miR-590 was required for the effect of FA in inducing VEGF.

To further test whether the above discovered mechanism was responsible for the FA action in improving NSC transplanted therapy, FA pre-treated NSCs expressing either control or miR-590 were transplanted into SCI rats (Figure 7). As expected, BBB scores of SCI-FA-NSC-control rats was significantly improved compared to SCI rats receiving no transplant (Figure 7A). However, BBB scores of SCI-FA-NSC-miR-590 rats was significantly lower than the SCI-FA-NSC-control group throughout the 21 days evaluation period. Similarly, spinal cord water content and MPO level also exhibited the same trend as BBB scores (Figures 7B,C). At the end of day 21, histological analysis of the spinal cord at the injury sites also indicated similar comparisons among the four experimental groups of rats (Supplementary Figure S2). Altogether, these results clearly indicated that miR-590 expression comproised the FA-enhanced recovery process after the transplant.