BMSCs, MG and neurons were isolated from specific pathogen-free grade Wistar rats (Vital River Laboratories, Beijing, China). The rats were maintained under laboratory conditions with free access to standard diet, sterile water and controlled temperature (24°C). The rats were anesthetized by intraperitoneal injection of chloral hydrate (350 mg/kg) and then euthanized by cervical dislocation for sampling.

All the animal experiments performed in this study were approved by the Animal Ethics Committee of Southern Medical University and were conducted in accordance with the instruction of our institute. No minors, persons with disabilities or endangered animal species were involved.

BMSCs were isolated from bilateral femurs and tibias of Wistar rats (aged 4 weeks) based on a previously described method (Huang et al., 2015). Briefly, both ends of the bones were removed to expose the marrow cavity, which was washed in serum-free Dulbecco’s modified Eagle medium (DMEM, Gibco, Carlsbad, CA, USA) and the fluid was collected. Cells were collected by centrifugation at 400× g for 10 min, suspended in DMEM supplemented with 10% fetal bovine serum (FBS) and penicillin-streptomycin (100 U/mL, Gibco) at a density of 2 × 10⁶ cells/mL. Cells were then cultured at 37°C in humidified atmosphere under 5% CO₂. After 3 days, the non-adherent cells were discarded. The culture medium was changed every 3 days and the cells were passaged at approximately 60% confluence.

The immortalized murine MG cell line BV2 (ATCC, Manassas, VA, USA) was cultured in DMEM supplemented with 10% FBS at a density of 1 × 10⁵ cells/mL, and incubated at 37°C in humidified atmosphere under 5% CO₂.

MG were isolated from the brain of newborn (within 24 h) Wistar rats using a previously described method (Hide et al., 2002). The cerebral cortex was sampled under sterile conditions and washed in cold phosphate-buffered saline (PBS). The meninx and blood vessels on the surface were carefully removed, and the remaining cerebral cortex was then cut into cubes (approximately 1 mm³) in serum-free DMEM before digestion in 0.05% trypsin (Gibco) for 30 min at 37°C. The suspension was centrifuged at 400× g for 10 min after termination of the digestion, and the cells were resuspended at a density of 1 × 10⁶ cells/mL in DMEM supplemented with 10% FBS and penicillin-streptomycin (100 U/mL). The cells were then incubated at 37°C in a humidified atmosphere under 5% CO₂ and after approximately 2 weeks, the cells were separated into two layers—the lower astrocytes and the upper putative MG. The cells were digested in 0.05% trypsin and the dishes were gently shaken to resuspend the MG. The digestion was terminated and MG were isolated from the cell suspension by centrifugation, and cultured in complete DMEM at a density of 1 × 10⁶ cells/mL.

Neurons were isolated from the hippocampus of newborn Wistar rats. Microvessels were carefully removed, and the hippocampus was then cut into pieces and digested in 0.125% trypsin for 20 min at 37°C. The digestion was terminated and the cells were filtered and seeded at a density of 1 × 10⁵ cells/mL in DMEM supplemented with 10% FBS and penicillin-streptomycin (100 U/mL). The cells were incubated at 37°C in humidified atmosphere under 5% CO₂.

To analyze GDNF secretion by BMSCs, BV2 and MG cells were subjected to OGD for 4 h, and the CM was collected and centrifuged at 300× g for 10 min. BMSCs were seeded in 96-well plates at a density of 1 × 10⁴ cells/well. After incubation for 24 h, the medium was replaced entirely with 200 μL CM; BMSCs subjected to this procedure acted as the BMSC + BV2/MG + OGD group. The CM of BV2/MG without OGD induction was also collected and added to BMSCs as the BMSC + BV2/MG group. The BMSC group did not receive any treatment. After incubation for 24 h, the BMSCs culture supernatant in each group was collected for determination of the GDNF concentration. Complete culture medium was used as the control group.

To analyze GDNF production by BMSCs following treatment with TNFα, interleukin (IL) 6 or interleukin 1β (IL1β), the levels of these cytokines in the CM of the activated BV2 and MG were detected by ELISA (see “ELISAs” Section). In addition, BMSCs were treated with TNFα (PeproTech, Rocky Hill, NJ, USA) at final concentrations of 0, 1, 5, 10, 20 or 50 ng/mL, and interleukin-6 (IL6) or IL1β (PeproTech) to final concentrations of 0, 1, 5, 10, 50 or 100 ng/mL. The cells were cultured for 24 h, and the medium was then collected to determine the GDNF concentration.

For analysis of the effects of GNDF in repairing OGD injury, neurons were seeded in 96-well plates (1 × 10⁴ cells/well) for 24 h, before the addition of GDNF (500 ng/L, PeproTech), or replacement of the medium with BMSCs culture supernatant, GDNF-silenced BMSCs (see “siRNA Transfection” Section), CM pre-treated BMSCs, or CM pre-treated GDNF-silenced BMSCs for 24 h after OGD exposure. Additionally, neurons were treated with TNFα (5 ng/mL), IL6 (5 ng/mL) or IL1β (5 ng/mL, PeproTech) for 24 h after OGD.

For detection of the protective effects of GDNF against OGD injury, neurons were seeded into 96-well plates (1 × 10⁴ cells/well) or 6-well plates (1 × 10⁵ cells/well) for 24 h. GDNF (500 ng/L) and polyclonal anti-RET (2 μg/mL, the GDNF receptor) antibody (Santa Cruz Biotechnologies, Santa Cruz, CA, USA) were added to neurons for 24 h before OGD induction.

The OGD model was used to activate BV2 and MG cells as well as to injure BMSC and neurons. In brief, the OGD model was established by exposure of cells cultured in glucose-free medium and to a humidified atmosphere of 95% N₂ and 5% CO₂ at 37°C for 4 h (Zhou X. et al., 2014). Cells cultured under normal condition were used as a control group.

For RNA silencing, 150 nmol/L GDNF siRNA (Santa Cruz Biotechnologies) were transfected into BMSCs using Lipofectamine 3000 reagent (Invitrogen) following the manufacturer’s protocol. After 48 h, cells were collected for use in subsequent experiments.

GDNF concentrations in culture supernatants were determined using the rat GDNF ELISA Kit (Cusabio, Wuhan, China) according to the manufacturer’s instructions. Briefly, the culture supernatant was collected by centrifugation at 300× g for 15 min, and 100 μL was added to each well of 96-well plates. After incubation for 2 h at 37°C, the supernatant was removed and the plates were incubated sequentially with the biotin- conjugated primary detection antibody followed by the horseradish peroxidase (HRP)-conjugated secondary antibody for 1 h at 37°C; plates were washed three times between steps. After addition of the chromogenic substrate, the plates were incubated in the dark for 30 min at 37°C. After the reaction was stopped, optical density (OD) was immediately detected at 450 nm using the iMark microplate reader (Bio-Rad, Hercules, CA, USA). Concentrations of TNFα, IL6 and IL1β in BV2 and MG culture supernatants were also detected using the corresponding ELISA kit from PeproTech according to the manufacturer’s instructions.

Neuron cell viability was evaluated after treatment using the MTT Cell Proliferation Assay Kit (ATCC) according to the manufacturer’s instruction. Briefly, neurons (1 × 10⁴ cells) were seeded in each well of 96-well plates for adherence, and treated for 24 h. MTT Reagent was added and the plates were incubated for 4 h before addition of 100 μL Detergent Reagent to dissolve the purple precipitate. The plates were then incubated in the dark at room temperature for 2 h, and the OD was detected at 570 nm using the iMark microplate reader (Bio-Rad).

Cell apoptosis was investigated by flow cytometry after fluorescein isothiocyanate (FITC) and propidium iodide (PI) staining with the Annexin V-FITC Apoptosis Kit (BioVision, Milpitas, CA, USA) according to the manufacturer’s instruction. In brief, 1 × 10⁵ were cells resuspended in 500 μL Binding Buffer for each sample. Annexin V-FITC (5 μL) and PI (5 μL) were added and the cells were incubated at room temperature for 5 min in the dark immediately before flow cytometric analysis (BD Biosciences, San Jose, CA, USA). FITC-positive and PI-negative cells were considered to be apoptotic.

LDH release by neurons was detected using the LDH Cytotoxicity Detection Kit (Roche, Basel, Switzerland) according to the manufacturer’s instructions. Neurons (1 × 10⁴ cells/well) were seeded into 96-well plates. OD was detected at 500 nm by the iMark microplate reader (Bio-Rad).

MMP of neurons was detected by Rhodamin 123 (Sigma-Aldrich) staining. Neurons (1 × 10⁶ cells) were washed twice in artificial cerebral spinal fluid (ACSF, NaCl 124 mM, KCl 3 mM, NaHCO₃ 26 mM, NaH₂PO₄·2H₂O 1.24 mM, MgSO₄·7H₂O 2 mM, CaCl₂ 2 mM and glucose 10 mM) before addition of Rhodamine 123 at a final concentration of 5 μg/mL. The cells were incubated at 37°C for 45 min and washed again in ACSF before the fluorescence intensity was detected by flow cytometry.

The total protein content of neurons was extracted by M-PER Mammalian Protein Extraction Reagent (Thermo Scientific, Carlsbad, CA, USA) according to the manufacturer’s instruction. Protein samples were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to polyvinylidene fluoride membranes (Roche). The membranes were blocked in 5% skimmed milk for 2 h at room temperature and then incubated at 4°C overnight in rabbit polyclonal or monoclonal antibodies for the specific detection of B cell lymphoma 2 (BCL2, ab59348, Abcam, Cambridge, UK), heat shock 60 kDa protein 1 (HSP60, ab46798), mitogen-activated protein kinase (MAPK) kinase 1/2 (MAP2K1/2 alias MEK1/2, ab178876), phospho-MEK1/2 (p-MEK1/2, ab194754), MAPK3/1 (alias ERK1/2, ab17942), p-ERK1/2 (ab76299), phosphoinositide-3-kinase (PI3K p85, ab191606), p-PI3K (ab182651), v-akt murine thymoma (pan-AKT, ab8805), p-AKT (ab38449), MAPK8/9/10 (alias JNK, ab208035), p-JNK, jun proto-oncogene (c-JUN, ab32137), p-c-JUN (ab32385) and Actin (ab3280) which was used as an internal control. After washing in PBS, the membranes were incubated with the goat anti-rabbit IgG (HRP-conjugated, ab7090) secondary antibody at room temperature for 1 h. Positive signals were developed by EasyBlot ECL Kit (Sangon Biotech, Shanghai, China) and analyzed with ImageJ 1.49 (National Institutes of Health, Bethesda, MD, USA).

Data represent the mean ± standard deviation (SD) of five independent experiments. Data were analyzed by SPSS 20 (IBM, New York, NY) using Student’s t test or one-way analysis of variance (ANOVA) as appropriate. P < 0.05 was considered to indicate statistical significance.

GDNF production by BMSC was confirmed by analysis of GDNF concentrations in the culture supernatants (Figures 1A,B). No obvious changes in GDNF production were detected when BMSCs were treated directly with BV2 or MG CM (P > 0.05). However, treatment with the CM of OGD-activated BV2 or MG significantly elevated GDNF levels in BMSCs culture supernatants (P < 0.01). Although BV2 and MG also produced GDNF after OGD induction, the levels were significantly lower than that in the BMSC + BV2 + OGD group (P < 0.001). It can be speculated that OGD-activated BV2 and MG produce factors that further stimulate GDNF production by BMSCs.

Numerous studies have shown that MG produce TNFα, IL6 and IL1β in response to OGD exposure (Chock and Giffard, 2005; Zhou et al., 2016); this was also confirmed in our study. The levels of TNFα, IL6 and IL1β produced by BV2 and MG were increased by OGD (P < 0.001, Figures 2A,B). Next, we investigated GDNF concentrations in BMSC culture supernatants after treatment with TNFα, IL6 or IL1β. Results showed that BMSCs produced significantly more GDNF after treatment with TNFα (P < 0.001, Figure 2C). However, treatment with IL6 and IL1β did not cause an obvious increase in GDNF production (P > 0.05, Figures 2D,E), suggesting that TNFα, but not IL6 and IL1β, is secreted by OGD-activated MG, resulting in the induction of GDNF production by BMSCs.

Next, we performed siRNA-mediated GDNF silencing in BMSCs to assess the functions of GDNF secreted by BMSCs on OGD-injured neurons. As shown in Figure 3A, OGD significantly reduced neuronal viability (P < 0.001). Neurons cultured with GDNF or BMSCs CM alleviated OGD-induced suppression of neuronal viability (P < 0.01), while this effect was partly abolished by GDNF silencing (P < 0.05). Furthermore, CM pre-treated BMSC exhibited a notable increase in neuronal viability (P < 0.05), and this increase was also abolished by GNDF silencing (P < 0.01). These preliminary data indicate that GNDF produced by BMSCs protects neurons from OGD-induced damage.

Given that OGD induces the production of TNFα, IL6 and IL1β by MG, it is possible that these factors in the CM help to promote neuronal viability via a GDNF-independent mechanism. Therefore, we treated OGD-injured neurons with TNFα, IL6 and IL1β to investigate their effects on cell viability. As shown in Figure 3B, TNFα, IL6 and IL1β treatment diminished neuronal viability (P < 0.05), indicating the suppressive role of these pro-inflammatory factors on neuronal viability. Thus, we inferred neuronal viability was promoted mainly by the increased GDNF production by BMSCs, rather than by the TNFα, IL6 and IL1β produced by MG.

Neuronal viability and apoptosis were then investigated following treatment with GDNF and the anti-RET antibody that blocks GDNF binding to its receptor before OGD exposure. GNDF prevented the decrease in neuronal viability and the increase in the rate of apoptotic cells rate induced by OGD (P < 0.05 or P < 0.01, Figures 4A,B). More importantly, these functions of GNDF on neurons were abolished in the presence of the anti-RET antibody (P < 0.05 or P < 0.01). These data indicated that GNDF prevents OGD-induced neuronal injury.

Previous research has emphasized the involvement of the MEK/ERK, PI3K/AKT and JNK/c-JUN signaling pathways in modulation of cell viability and apoptosis (Oufkir et al., 2010; Park et al., 2012; Dillon et al., 2015); thus, key components of these pathways were detected by Western blotting to investigate the capacity of GDNF to regulate these factors. Results showed that GDNF elevated the expression of p-MEK1/2, p-ERK1/2, p-PI3K and p-AKT, and this GDNF-induced elevation was inhibited in the presence of anti-RET (P < 0.001, Figures 5A,B). The total level of these proteins was almost unchanged, suggesting that GDNF promotes their activation. However, p-JNK and p-c-JUN were induced by OGD but were unaffected by GDNF or anti-RET (P > 0.05), implying that GDNF regulates activation of the MEK/ERK and PI3K/AKT pathways but not the JNK/c-JUN pathway.

To investigate the potential mechanism by which GDNF attenuates neuronal injury, we also detected the expression of the apoptosis-related protein BCL2 on mitochondrial membrane (Gross et al., 1999), and the molecular chaperone HSP60, which participates in the folding of mitochondrial precursor proteins (Ostermann et al., 1989). Western blot (Figures 6A–C) analysis showed that both BCL2 and HSP60 were downregulated after OGD induction (P < 0.01), and were upregulated by GNDF administration with or without exposure to OGD conditions (P < 0.05 or P < 0.01).

LDH assays showed that OGD elevated LDH leakage from neurons (P < 0.01, Figure 6D). Although GDNF had no obvious effects on normal neurons (P > 0.05), it markedly decreased LDH leakage in neurons exposed to OGD (P < 0.01). Thus, it can be speculated that GDNF helps to maintain the neuronal membrane integrity after OGD. MMP assays showed similar changes in that GDNF significantly elevated MMP after OGD (P < 0.001, Figure 6E). In contrast, TNFα treatment of normal neurons led to an apparent decrease in MMP (P < 0.01), further suggesting that the effects mediated by TNFα on neurons are opposite to those mediated by GDNF. These results indicate that GDNF maintains mitochondrial stability and suppresses neuronal injury, thus representing a potential mechanism by which GDNF protects against OGD-induced neuronal injury.