Permissive role of Cdc2 activity induced from astrocytes in neurite outgrowth


Address correspondence and reprint requests to Uk Namgung, PhD, Department of Oriental Medicine, Daejeon University, Daejeon 300-716, Korea. E-mail:


Following spinal cord injury, glial cells are recognized as major environmental factors hampering axon's regenerative responses. However, recent studies suggested that, in certain circumstances, reactive astrocytes may have a permissive role for axonal regeneration and functional recovery. Here, we report that Cdc2 activation in astrocytes is positively linked to axon growth. Cdc2 was strongly, but transiently, induced from reactive astrocytes within and around the injury cavity. Cdc2 levels in primary, non-neuronal cells prepared from injured spinal cord were up-regulated by extending the pre-injury period. Cdc2-mediated vimentin phosphorylation was strongly induced in astrocytes after long-term culture (7 days, LTC) as compared with short-term culture (3 days, STC). Induction levels of phospho-vimentin in LTC astrocytes were positively associated with increased neurite outgrowth in co-cultured dorsal root ganglion neurons. β3 integrin mRNA was induced in LTC astrocytes and activation of β3 integrin was regulated by Cdc2 activity. Furthermore, genetic depletion and pharmacological blockade experiments demonstrate that activation of Cdc2 and β3 integrin in LTC astrocytes is required for neurite outgrowth. Our data suggest that the Cdc2 pathway may play an important role in determining phenotypic expression of astrocytes such that astrocytes provide permissive environments for axonal regeneration following spinal cord injury.

Abbreviations used

chondroitin sulfate proteoglycan


dorsal root ganglion


extracellular matrix


long-term culture


spinal cord injury


short-term culture

Regenerative responses of CNS axons after injury reflect intrinsic, neuronal properties and intricate environmental influence as activated glial cells participate in inflammatory and repair processes and form glial scars, which are a prominent pathological feature (Silver and Miller 2004; Donnelly and Popovich 2008). Reactive astrocytes express and secrete chondroitin sulfate proteoglycans known to inhibit axonal growth by physically blocking growth cone elongation or transmitting inhibitory signals to axons (Monnier et al. 2003; Silver and Miller 2004; Dickendesher et al. 2012). Moreover, reactive astrocytes express diverse, molecular triggers and modulators, which may induce phenotypic heterogeneity and lead to a dual response in axonal regeneration (Ridet et al. 1997; Rolls et al. 2009; Sofroniew 2009; White et al. 2011). For example, secretion and accumulation of different types of chondroitin sulfate proteoglycans in and around the glial scar and production of growth factors and cytokines and their interaction with the extracellular matrix proteins and neuritic counterpart molecules are important for determining whether reactive astrocytes act as either beneficial or detrimental factors for axonal regeneration (Jones et al. 2003; Silver and Miller 2004; Busch et al. 2010).

Conditioning effects of pre-injury given to the peripheral nerve can be observed a few days later via the enhancement of axonal elongation into the spinal cord beyond the dorsal root entry zone in vivo and the neurite outgrowth of cultured neurons (Richardson and Issa 1984; Smith and Skene 1997). Recently, we found that enhanced axonal growth is mediated by activated Schwann cells prepared from pre-injured sciatic nerve (Chang et al. 2012). Considering the phenotypic diversity of reactive astrocytes, it is worthwhile to explore whether or not reactive astrocytes, which are prepared acutely after spinal cord injury (SCI), are activated in vitro to favor axon growth. Upon scratch wounding, integrin receptor activation was reported to mediate migration of primary astrocytes via CDC42 and PKCζ pathway (Etienne-Manneville and Hall 2001), and astrocytes from newborn animals were supportive of neurite outgrowth of co-cultured neurons through intercellular communication among N-cadherin, integrins, and counterpart, adaptor molecules (Tomaselli et al. 1988). Also, genetic ablation of reactive astrocytes caused disruption of protective neuropathological responses after SCI, suggesting a role of reactive astrocytes for spinal axon growth (Faulkner et al. 2004).

Here, we investigated the effects of astrocytes expressing Cdc2 on axonal regeneration. Cdc2 is a prototypical serine/threonine protein kinase playing a critical role in cell cycle progression (Santamaría et al. 2007), and Cdc2-mediated Schwann migration is involved in peripheral nerve regeneration (Han et al. 2007). As reactive astrocytes around the injury cavity show active migration and proliferation, we hypothesized that Cdc2 activation, if any, in astrocytes may play a significant role in regenerative processes following SCI. We showed that Cdc2-mediated vimentin phosphorylation and β3 integrin activation induced from long-term culture (LTC) astrocytes are involved in neurite outgrowth of co-cultured dorsal root ganglion (DRG) neurons.

Materials and methods

Experimental animals

Sprague–Dawley rats (male, 200–250 g, DaeHan Biolink, Seoul, Korea) were maintained in an animal room with regulated temperature (22°C), 60% humidity, and a 12-h light and 12-h dark cycle. Animals were fed commercial chow (DaeHan Biolink) and allowed to drink water ad libitum. All protocols involving live and postoperative animal care were approved by the Daejeon University Institutional Animal Use and Care Committee and in accordance with the Animal-Use Statement and Ethics Committee Approval Statement for Animal Experiments provided by Daejeon University (Daejeon, Korea).

Spinal cord surgery and immunohistochemistry

For spinal cord surgeries, animals were anesthetized with a mixed dosage of ketamine (80 mg/kg) and xylazine (5 mg/kg). Using aseptic technique, a laminectomy was performed to expose the spinal cord at thoracic levels 9–10. A contusion injury was induced using the NYU compactor by dropping a 10 g impactor (a cylindrical metal tapered with 2 mm of tip diameter) from 5 mm height onto the exposed dura, which, according to our previous study, produces a moderate level of injury that can be functionally recovered to a certain level by treadmill training (Oh et al. 2009). Following spinal cord surgery, rats recovered for 3–14 days, and spinal cord tissues at the injury area were dissected, frozen immediately at −75°C, and embedded into OCT medium. Pre-treatment and primary and secondary antibody incubation of sections (20 μm) were performed as described previously (Han et al. 2007). Primary antibodies used were anti-glial fibrillary acidic protein (GFAP) (rabbit polyclonal, diluted 1 : 1000, Dako, Glostrup, Denmark), anti-Cdc2 (mouse monoclonal, 1 : 2000, Santa Cruz Biotechnology, Santa Cruz, CA, USA), anti-APC mouse (CC1, mouse monoclonal, 1 : 200, Calbiochem, Darmstadt, Germany), anti-CD11 (1 : 200, BD Biosciences, San Jose, CA, USA), and anti-phospho-vimentin antibody (1 : 400, MBL, Nagoya, Japan). Secondary antibodies used were fluorescein-goat anti-mouse (1 : 400, Molecular Probes, Eugene, OR, USA) and rhodamine-goat anti-rabbit (1 : 400, Molecular Probes) antibodies. Cellular nuclei were stained with 2.5 μg/mL of Hoechst dye 33258 (bis-benzimide, Sigma, St. Louis, MO, USA) for 10 min prior to final wash with 0.1% Triton X-100 in phosphate-buffered saline. Sections were then mounted to slides with a gelatin medium. Samples were viewed with a Nikon fluorescence microscope, and images were captured using a Nikon camera (Nikon, Tokyo, Japan). Merged images were produced using layer-blending options of Adobe Photoshop.

Astrocyte culture and co-culture with DRG neurons

For astrocyte primary culture, intact or injured spinal cord tissues at the lower thoracic levels were dissected from adult rats, minced, and dissociated with 0.5 mg/mL type XI collagenase (Sigma) for 30 min at 37°C in a 5% CO2 incubator. Dissociated tissues were centrifuged for 1 min at 574 g, and supernatants were removed. Cells were re-suspended in 500 μL Dulbecco's modified Eagle's medium containing type SII trypsin (Sigma; 0.5 mg/mL) for 15 min and treated with EDTA (0.5 mM), soybean trypsin inhibitor (50 μg/mL), and DNase I (20 μg/mL) during the last 5 min. Cells were washed twice by re-suspension and centrifugation. After re-suspending with Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, cells (5 × 105 cells) were plated on 60-mm dishes pre-coated with poly-l-ornithine (0.1 mg/mL; Sigma) and laminin (0.02 mg/mL, BD Bioscience). Cells were cultured for 12 h, changed to BME containing 10% serum (5% fetal bovine serum plus 5% horse serum), 2 mM glutamine, and 1% penicillin–streptomycin and cultured as long as 7 days before transfer. In these culture conditions, active proliferation of glial cells from injured spinal cord in adult rats was induced 3–4 days in vitro (DIV) and reached a full confluency at 7 DIV. Enriched astrocytes were prepared as described previously (Codeluppi et al. 2009). Briefly, glial cells in 60-mm dish, which had been cultured for 3 or 7 days, were placed onto the orbital shaker (200 rpm) (Vision Scientific, Bucheon, Korea) for 2 h at 37°C, and suspended cells were removed. The astrocytes that remained attached were incubated under the preconditioned medium and used for further analysis.

For DRG neuron single culture or co-culture with astrocytes, DRG at lumbar levels 4–5 in adult, male rats (= 3 per experiment, 7–8 weeks old) were prepared, dissociated, and incubated 3 days following sciatic nerve injury. We used 3 day pre-injury condition since, with this time window, neurite outgrowth of DRG neurons was optimally induced as preconditioning responses and initial molecular events in relation to axonal regeneration such as Cdc2-mediated signaling pathway may be investigated appropriately. Primary astrocytes were cultured on coverslips (1 × 104 cells per 12 mm) pre-coated with poly-l-ornithine (0.1 mg/mL) or with poly-l-ornithine plus laminin (0.02 mg/mL) for 24 h prior to addition of DRG sensory neurons (1.5 × 102 cells). The co-culture or DRG single culture was maintained in 500 μL of BME supplemented with 10% serum in the presence of Cdc2 inhibitor purvalanol A or equivalent volume of dimethylsulfoxide vehicle for 24 h before cell harvest. In some experiments, cells were treated with Mn2+, c-Src inhibitor SU6656 (Santa Cruz Biotechnology), anti-β3 integrin antibody (Santa Cruz Biotechnology), FAK inhibitor (FAK inhibitor 14, Santa Cruz Biotechnology), or an integrin inhibitor RGD peptide (Santa Cruz Biotechnology). Fixed cells were analyzed by immunofluorescence staining as described previously (Namgung and Xia 2000) using primary antibodies such as anti-GFAP antibody (1 : 1000, DAKO), anti-Cdc2 antibody (1 : 400, Santa Cruz Biotechnology), anti-vimentin antibody (1 : 400, Sigma), anti-phospho-vimentin antibody (1 : 400, MBL, Japan), anti-β3 integrin antibody (1 : 200, Santa Cruz Biotechnology), and pY397 focal adhesion kinase (p-FAK; 1 : 400, Santa Cruz Biotechnology) antibodies and secondary antibodies such as fluorescein-goat anti-mouse (1 : 400, Molecular Probes) and rhodamine-goat anti-rabbit (1 : 400, Invitrogen, Carlsbad, CA, USA) antibodies. For analysis of neurite outgrowth, DRG neurons were subjected to immunofluorescence staining with anti-neurofilament-200 (NF-200) antibody (1 : 400, Sigma), and digital images of neuronal process were captured and transferred to Adobe Photoshop. The length of neurite processes that exhibited clear outgrowth (longer than the cell diameter) from the cell body was analyzed with i-Solution software (Image and Microscope Technology, Goleta, GA, USA). The number of cells displaying fluorescent signals, which were captured from three randomly selected microscopic fields per coverslip was counted, and values from 3–4 independent experiments were averaged.

Western blot analysis

The dorsal half of the spinal cord covering 5–7 mm of rostral and caudal spinal cord at the injury area was dissected from adult male rats (= 3–4 per data set, 7–8 weeks), washed with ice-cold phosphate-buffered saline, and sonicated under 400-600 μL of triton lysis buffer (20 mM Tris, pH 7.4, 137 mM NaCl, 25 mM β-glycerophosphate, pH 7.14, 2 mM sodium pyrophosphate, 2 mM EDTA, 1 mM Na3VO4, 1% Triton X-100, 10% glycerol, 5 μg/mL leupeptin, 5 μg/mL aprotinin, 3 μM benzamidine, 0.5 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride). Cultured astrocytes were harvested and lysed as mentioned previously (Namgung and Xia 2000). Protein (15 μg) from in vivo spinal cord tissue or cultured cells was used for western blot analysis using anti-Cdc2 (1 :2000, Santa Cruz), anti-vimentin (1 : 15 000, Sigma), anti-phospho-vimentin (1 : 2000, MBL), and anti-actin (1 : 100: ICN Biomedicals, Cleveland, OH, USA) antibodies, and horseradish peroxidase-conjugated secondary antibodies (1 : 10 000, goat anti-rabbit; Santa Cruz Biotechnology, or sheep anti-mouse; Amersham Biosciences, Buckinghamshire, UK).

Real-time quantitative RT-PCR

Total RNA was extracted from spinal cord tissue or cultured astrocytes using Easy Blue reagent (Intron Biotechnology, Sungnam, Korea), and 3 μg was used for a reverse transcriptase reaction in 35 μL (MMLV RT, Promega, Madison, WI, USA) for 2 h at 37°C. Synthesized cDNA (2 μL aliquot from 35 μL of RT reaction) was denatured at 50°C for 2 min and at 94°C for 10 min. Forty cycles of real-time PCR were performed with the cycling profile of 95°C for 10 s and 60°C for 1 min using power SYBR Green RNA to-C1™1-step kit. Relative quantitation (RQ) was measured using G3PDH as an internal standard. The primer sequences and amplified DNA sizes for individual integrin were as follows: αv integrin (Forward, 5′ TCG ACT GGA TAG AGG GAA GAG 3′; Reverse, 5′ GCG GAT GAC TTC AGG GAA TAG 3′), β1 integrin (Forward, 5′ CGA TAG GTC CAA CGG CTT AAT 3′; Reverse, 5′ CTG CCA GTG TAG TTG GGA TAG 3′), β3 integrin (Forward, 5′ GCT GTC CTG TAT GTG GTA GAA G 3′; Reverse, 5′ CAG AGT AGC AAG GCC AAT GA 3′).

Plasmid constructs, siRNA, and transfection

Plasmids expressing siRNAs of mouse Cdc2 and rat β3 integrin simultaneously with control siRNA were obtained from Santa Cruz Biotechnology. Transfection of primary astrocytes was performed as recommended by the manufacturer. Briefly, the solution containing siRNA or plasmid DNA was mixed with the transfection medium and reagents (Santa Cruz Biotechnology) and added to cells (5 × 104 cells per 24-well plate, 60–80% confluency). Cells were washed 6 h later, replaced with fresh medium, and cultured for an additional 48 h prior to cell fixation for immunofluorescence staining. In co-culture experiments, transfected astrocytes were mixed with DRG neurons (1 × 102 cells) and cultured for another 48 h. For detection analysis of transfected cells, pmaxGFP (1 μg) was added to each transfection reaction.

Statistical analysis

Data were presented as mean ± SEM. The mean of individual groups were compared by one-way anova followed by the Tukey test (SPSS computer software version 12.0). Values of *p < 0.05 were considered statistically significant.


In vivo analysis of Cdc2 induction following SCI

We investigated Cdc2 induction within the injury area of spinal cord. Cdc2 was rarely detected in the intact spinal cord, but strongly induced between 3 and 7 days following injury and down-regulated 14 days later (Fig 1a). Cdc2 signals in the injury area were observed in GFAP-positive astrocytes, showing peak levels at 7 days after injury followed by decreased levels 14 days later (Fig 1b). Cdc2 signals were also observed in CC1-labeled-oligodendrocytes and CD11-labeled macrophages, but overall levels of Cdc2 in both CC1-positive cells and CD11-positive cells were lower than those in astrocytes (Fig 1c).

Figure 1.

Cdc2 induction in glial cells after spinal cord injury. (a) Western blot analysis of Cdc2 protein in the injured spinal cord. Actin western blotting for loading control. (b) Immunofluorescence staining of Cdc2 signals in the injured spinal cord. Representative merged images for Cdc2 and glial fibrillary acidic protein (GFAP) signals in the spinal cord 0–14 days post-injury (dpi) (top). Co-localization of two protein signals is indicated by arrows (yellow). Quantification of% Cdc2-positive cells in GFAP-positive astrocytes. Cdc2 signals in GFAP-labeled astrocytes were analyzed from 3–5 sections from each experiment. (= 3 independent experiments **p < 0.01, ***p < 0.001 vs. 0 (c) Merged images of CC1-labeled oligodendrocytes and CD11-labeled macrophages in the spinal cords after lesion. Scale bars in (b) and (c) = 100 μm.

Cdc2 function in LTC astrocytes

As Cdc2 induction was clearest in astrocytes after SCI, we investigated Cdc2 function in primary astrocytes. Cdc2 levels in primary, non-neuronal cells prepared from spinal cord tissue were increased by extending the pre-injury period of spinal cord as long as 5 days (Fig 2a). GFAP-positive astrocytes are the major cell type in a culture 2 DIV, in which Cdc2 signals were detected clearly (Fig 2b). The number of Cdc2-positive astrocytes was significantly higher in cells prepared from pre-injured animals as compared to the non-injury control (% Cdc2 positive cells in astrocytes; 74 ± 5.5% in 5 day pre-injury group versus 17 ± 5.2% in non-injury group, p < 0.001, = 4). Moreover, subcellular distribution of Cdc2 signals within astrocytes spread out more widely in the pre-injury group when compared with the control group (Fig 2b). Thus, increases in both Cdc2 expression per cell and percentage of Cdc2-positive cells may contribute to enhanced Cdc2 levels.

Figure 2.

Cdc2 induction in primary astrocytes. (a) Western blot analysis of Cdc2 in cultured glial cells. Spinal cord tissues were prepared from intact animals (dpi; 0), and animals 1 and 5 days post-injury (dpi), and dissociated cells were cultured for 2 days. Actin western blotting for loading control. (b) Confocal fluorescent view of Cdc2 signals in glial fibrillary acidic protein (GFAP)-labeled primary astrocytes. Glial cells were prepared from intact spinal cord (non-injured) or 5 days after spinal cord injury (pre-injured), and cultured for 2 days prior to immunofluorescence staining. Cdc2 signals were co-localized with GFAP-positive astrocytes. Scale bar = 20 μm.

Recently, we reported that vimentin phosphorylation by Cdc2 is functionally involved in Schwann cell activation and peripheral nerve regeneration (Chang et al. 2012). Here, we investigated phospho-vimentin as a mediator of Cdc2 activity in primary astrocytes. Although vimentin is constitutively expressed in the intact spinal cord and further increased after SCI, Cdc2-specific phospho-vimentin was weakly detected in the spinal cord 7 days after injury (Fig 3a). Phospho-vimentin signals were detected within the injury cavity as well as the surrounding glial scar area and were largely co-localized with GFAP-positive astrocytes (Fig 3b). Within the injury cavity, some GFAP-labeled astrocytes were positive to Cdc2 and phospho-vimentin, while other cells were positive only to Cdc2 or phospho-vimentin (% astrocytes positive for Cdc2 or phospho-vimentin: 36 ± 4.7% and 16 ± 3.8% respectively, = 4). As most of the phospho-vimentin signals were merged with GFAP and Cdc2 signals (Fig 3b), it was estimated that approximately 16% of astrocytes were dual-positive to Cdc2 and phospho-vimentin. To explore the possibility that prolonged culture of astrocytes alters its physiological properties, we prepared two types of astrocytes, which were enriched after 3 and 7 days of the initial culture (short-term culture, STC versus long-term culture, LTC). Phospho-vimentin levels were higher in LTC astrocytes than those in STC astrocytes. In STC astrocytes, phospho-vimentin was up-regulated by treatment of Mn2+ known to activate integrin (Takagi et al. 2002) (Fig 3c). Phospho-vimentin induction in both LTC and STC was largely inhibited by purvalanol A. Overall, levels of total vimentin in LTC were increased by Mn2+ treatment and unaltered by purvalanol A treatment. In accordance with previous studies, purvalanol A inhibited vimentin phosphorylation effectively with 10 μM in primary astrocytes (Goga et al. 2007). Other cyclin-dependent kinases require doses much higher for inhibition by purvalanol A (Gray et al. 1998), indicating that the purvalanol A concentration used in this study was selective for Cdc2 inhibition among Cdk family proteins. Although all phospho-vimentin-positive astrocytes in LTC were Cdc2-positive, some Cdc2-positive cells did not exhibit phospho-vimentin signals (% of Cdc2-positive and phospho-vimentin-negative astrocytes: 19.0 ± 3.6%, = 3), which indicates that all astrocytes expressing Cdc2 are not active necessarily for vimentin phosphorylation (Fig 3d). Furthermore, a few of phospho-vimentin-positive cells were GFAP-negative (white arrow in Fig 3e), and many of those cells showed dividing nuclear morphology (green arrow in Fig 3e).

Figure 3.

Induction of vimentin phosphorylation by Cdc2 in astrocytes after long-term culture (LTC). (a) Western blot analysis of vimentin (Vim) and Cdc2-specific phospho-vimentin (p-vim) in intact (dpi: 0) or injured spinal cord of 3, 7, and 14 days post-injury (dpi). (b) Distribution of phospho-vimentin, Cdc2, and glial fibrillary acidic protein (GFAP) signals in the spinal cord, which were prepared 7 days after lesion. Merged images demonstrate co-localization of Cdc2 and phospho-vimentin signals in GFAP-positive astrocytes. Arrowheads indicate co-localized signals (yellow), and vertical arrows in Cdc2/GFAP merged image indicate Cdc2 signals from other cell types. (c) Induction of phospho-vimentin and total vimentin in primary astrocytes. Dissociated glial cells from pre-injured spinal cord were cultured for 3 and 7 days [short-term culture (STC) and LTC, respectively], and enriched astrocytes were cultured for 24 h, treated with Mn2+ (500 μM, 30 min), purvalanol A (10 μM, 24 h), or DMSO vehicle, and incubated for an additional 24 h. Vimentin protein was detected as multiple bands at a 57–60 kDa. (d) Localization of Cdc2 and phospho-vimentin signals in LTC astrocytes. (e) Localization of phospho-vimentin and GFAP signals in LTC astrocytes. The Hoechst 33258-stained nuclei that are phospho-vimentin-positive, but GFAP-negative, are indicated (white and green arrows with enlarged nuclear morphology in rectangles). Scale bars = 100 μm (b, e); 20 μm (d). Actin western blotting in (a) and (c) for loading control.

Integrin receptor activation can promote axon growth via intercellular communication (Lemons and Condic 2008). For instance, activation of the two-receptor system, including N-Cad and integrin, facilitates neurite outgrowth of retinal ganglion cells cultured on an astrocyte layer (Tomaselli et al. 1988). To investigate whether the Cdc2-phospho-vimentin signaling system in astrocytes is linked to intercellular communication via integrin activation, we compared the effects of STC and LTC astrocytes on neurite outgrowth. Enriched astrocytes in STC and LTC were clearly detected by immunofluorescence staining of vimentin. Confocal images illustrated that astrocytes were in close contact with neurite processes of co-cultured DRG neurons (arrowheads in Fig 4a) and some of the neurites extended along the astrocytic processes (arrows). In the culture condition of dual addition, dual removal of Mn2+ and laminin, or the removal of laminin alone, neurite outgrowth was significantly facilitated by co-cultures with LTC astrocytes when compared with single cultures (Fig 4b, c). Neurite outgrowth of DRG neurons in single cultures and co-cultures was significantly increased with co-treatment of Mn2+ and laminin. In the culture with Mn2+ and laminin, purvalanol A treatment significantly decreased neurite outgrowth when co-cultured with STC and LTC astrocytes (Fig 4d, e). Cdc2 protein was not detected in DRG neurons prepared from rats, which underwent preconditioned injury on the sciatic nerve for 0–7 days, when analyzed by western blot analysis (Fig 4f) and also by immunofluorescence staining (data not shown). Thus, these data indicate that the Cdc2 activity, which mediates neurite outgrowth in co-culture, is derived from co-cultured astrocytes. A recent study reported that purvalanol A can inhibit c-Src (Hikita et al. 2010), which is known to interact with integrin and FAK (Giancotti and Ruoslahti 1999). To determine potential inhibitory effects of purvalanol A on neurite outgrowth via c-Src pathway, cells were treated with SU6656, a specific inhibitor of c-Src. In LTC devoid of Mn2+, neurite outgrowth was not affected by SU6656. However, in Mn2+-containing culture conditions, SU6656 significantly decreased neurite outgrowth though it was not as effective as purvalanol A, raising the possible involvement of c-Src partly in β3 integrin activation for neurite outgrowth (Fig 4g). We also examined effects of potential soluble factors that may be released from co-cultured astrocytes on neurite outgrowth. Incubation of DRG neurons with conditioned medium of astrocytes cultures did not change neurite extension when compared to single cultures (Fig 4h).

Figure 4.

Enhanced neurite outgrowth of dorsal root ganglion (DRG) neurons by co-culture with long-term culture (LTC) astrocytes. (a) Representative confocal images of DRG neurons and LTC astrocytes in co-culture. (b) Effects of Mn2+ and astrocyte co-culture on neurite outgrowth. DRG neurons were cultured for 24 h prior to co-culture with short-term culture (STC) or LTC astrocytes that had been grown for 24 h on coverslips coated with both laminin (0.02 mg/mL) and polyornithin (0.1 mg/mL) (laminin+) or polyornithin only (laminin-). Single-cultured or co-cultured cells were incubated for 24 h, treated with Mn2+ (500 μM, 30 min), and incubated for an additional 24 h. ***p < 0.001 versus corresponding single-cultured neurons (= 4); p < 0.05 vs. corresponding single culture and co-culture groups in Mn2+- and laminin-depleted conditions (= 4). (c) Representative images on data in (b). (d) Inhibition of neurite outgrowth by purvalanol A. Cells were cultured and treated with purvalanol A (10 μM) or DMSO vehicle in the same culture conditions as (b). (e) Representative images on data in (d). (f) No induction of Cdc2 from DRG tissues. Sciatic nerves were given injury for 0–7 days, and protein lysates from the sciatic nerve and DRG lumbar levels 4–5 were used for Cdc2 western blot analysis. (g) Effects of c-Src inhibition by SU6656 on neurite outgrowth of DRG neurons in co-culture with astrocytes. After 24 h cultures, cells were treated with Mn2+ (500 μM, 30 min) (±), purvalanol A (10 μM, 24 h), DMSO vehicle, and SU6656 (5 μM, 24 h). *p < 0.05, ***p < 0.001. (h) Addition of LTC medium to DRG culture did not affect on neurite outgrowth. Culture medium of DRG neurons was replaced with LTC supernatant which was collected after 24 h cultivation. Neurons were harvested 24 h later for neurite analysis. ***p < 0.001 versus to LTC cell group. In (b), (d), (g), and (h), neurite length was determined by analyzing more than 20 neurons in 5–10 random microscopic fields, averaged over four independent experiments, and compared between experimental groups. Scale bar = 20 μm (a), 100 μm (c, e).

Activation of β3 integrin of LTC astrocytes for neurite outgrowth

To determine which type(s) of integrin in LTC astrocytes are involved in mediating neurite outgrowth, we examined integrin expression in spinal cord tissue and cultured astrocytes by quantitative RT-PCR. mRNA expression levels of αv, β1, and β3 integrins were elevated in Mn2+-treated LTC astrocytes as compared to intact or injured spinal cord tissues (Fig 5a). Yet, when comparing αv, β1, and β3 integrins, the induction level of β3 integrin mRNA in LTC astrocytes was most significantly increased by Mn2+ treatment. Thus, we focused on exploring the possible involvement of β3 integrin of Mn2+-treated LTC astrocytes in neurite outgrowth of co-cultured neurons. Among LTC cells, 95% of Hoechst-stained nuclei were GFAP-positive astrocytes, and a certain portion of astrocytes was Cdc2 and/or β3 integrin positive (Fig 5b). Specifically, the number of β3 integrin-positive astrocytes was lower than Cdc2 positive cells (% Cdc2-positive cells vs. β3 integrin-positive cells; 77 ± 6.1% vs. 50 ± 3.0%, = 4), and most of β3 integrin-positive cells were Cdc2 positive as well (% dual-positive astrocytes; 41 ± 3.2%). To examine the functional relationship between Cdc2 and β3 integrin, LTC astrocytes were treated with Cdc2 and β3 siRNAs. Cdc2 and β3 siRNAs were effective in knocking down their own proteins (Fig 5c, d). Phospho-FAK is known to interact with the active form of integrin (Parsons, 2003). Thus, we analyzed phospho-vimentin and phospho-FAK as downstream markers of Cdc2 and β3 integrin activities in LTC astrocytes, respectively. In cells showing dual-positive signals of phospho-FAK and phospho-vimentin, phospho-FAK signals were detected not only in the cell periphery where phospho-vimentin signals were high but also in sub-membranous, non-nuclear zone (arrow in Fig 5e). To determine whether β3 integrin activation is subjected to Cdc2 activity or vice versa, phospho-vimentin and phospho-FAK were analyzed in LTC astrocytes after siRNA transfection. Cdc2 siRNA transfection, which blocks phospho-vimentin, significantly suppressed phospho-FAK signals (Fig 5f). However, transfection of β3 integrin siRNA was effective only in blocking phospho-FAK production, which suggests that Cdc2 activity may act upstream of the β3 integrin activation. A quantitative comparison of phospho-vimentin and phospho-FAK after different siRNA treatments confirmed subordinate regulation of FAK phosphorylation by Cdc2 (Fig 5g).

Figure 5.

Induction and regulation of β3 integrin in long-term culture (LTC) astrocytes. (a) Quantitative RT-PCR analysis of integrin expression in intact (SC-intact) and injured spinal cord (SCI-injury; 5 day pre-injury), and LTC astrocytes with or without Mn2+. Data were normalized with G3PDH, and values for intact spinal cord were arbitrarily assigned as 1. *p < 0.05 and **p < 0.01, ***< 0.001 versus Mn2+-treated LTC astrocytes. (= 4 independent experiments). (b) Cdc2, phospho-vimentin, and glial fibrillary acidic protein (GFAP) signals in 2-day cultured LTC astrocytes. A representative image showing GFAP signals above Hoechst-stained nuclei demonstrate abundance of astrocytes in LTC cells. Another set of images in astrocyte culture shows co-localization pattern of Cdc2 and β3 integrin. (c) Representative images on blockade of Cdc2 and β3 integrin expression in LTC astrocytes by siRNA transfection. (d) Quantification of Cdc2 or phospho-vimentin positive astrocytes after siRNA transfection. The number of labeled cells from four to six randomly selected microscopic fields was averaged over four independent experiments and compared by experimental groups. ***p < 0.001. (e) Fluorescent view of phospho-vimentin and phospho-FAK in LTC astrocytes. Phospho-FAK signals were seen over nuclear periphery besides cell surface area (arrow). (f) Regulation of phospho-vimentin and phospho-FAK signals in LTC astrocytes transfected with different siRNA constructs. (g) Quantitative comparison of phospho-vimentin and phospho-FAK signals in cells transfected with different siRNA. Protein signals per cell in (g) were represented as the ratio of summated signal intensity to total number of cells (identified by Hoechst 33258 nuclear staining) in the microscopic field. Mean signal intensity from four to six randomly selected microscopic fields was averaged over four independent experiments and compared by experimental groups *p < 0.05 and **< 0.01. In (c–g), cells grown on laminin (0.02 mg/mL) and polyornithine (0.1 mg/mL) were harvested 48 h for immunofluorescence staining. Scale bars = 100 μm (b, c, f), 20 μm (e).

To determine whether β3 integrin is involved in neurite outgrowth of co-cultured neurons, primary astrocytes were transfected with Cdc2 and β3 integrin siRNAs, and the neurite length was compared with cells transfected with control siRNA. Many transfected cells, as identified by co-transfected green fluorescence protein (GFP), were in close contact with neurite processes (Fig 6a). When we measured the length of neurites that were in direct contact with GFP-expressing astrocytes, the mean length of groups treated with Cdc2 and β3 integrin siRNA were significantly shorter than the group treated with control siRNA (Fig 6b). Since we found that levels of phospho-FAK were down-regulated by Cdc2 siRNA treatment, we investigated the effects of direct pharmacological blockade of FAK on neurite outgrowth. As shown in Fig 6C, neurite outgrowth of DRG neurons co-cultured with astrocytes was significantly inhibited by FAK inhibitor. Similarly, blockade of β3 integrin by antibody or RGD peptide effectively decreased neurite outgrowth.

Figure 6.

Effect of β3 integrin and phospho-FAK blockade in long-term culture (LTC) astrocytes on neurite outgrowth. (a, b) LTC astrocytes were co-transfected with pmaxGFP plus individual siRNA expression plasmids. Cells grown on laminin (0.02 mg/mL) and poly-ornithine (0.1 mg/mL) were harvested 48 h later for neurite measurements. Representative images for neurite outgrowth of dorsal root ganglion neurons in co-culture (red; NF-200 stained) (a). Astrocytes expressing siRNA were identified by co-transfected pGFP signals (green). Quantification of neurite length in siRNA transfected cells (b). Neurite length was determined by analyzing 6–10 dorsal root ganglion neurons per coverslip whose neurite branches directly contacted transfected astrocytes (= 4 independent experiments). (c) Effects of pharmacological blockade of FAK and β3 integrin on neurite outgrowth. LTC astrocytes on Mn2+ culture were treated with purvalanol A (10 μM, 24 h), anti-β3 integrin antibody (20 μg/mL, 10 min), FAK inhibitor (FAKI; 5 μM, 1 h), and RGD peptide (1 mg/mL, 1 h). Neurite lengths were measured and averaged from more than 20 neurons in 5–10 random microscopic fields (= 4 independent experiments). In (b) and (c), **p < 0.01, ***p < 0.001. Scale bar = 100 μm.


This study demonstrates that primary astrocytes, after prolonged culture, induce Cdc2-mediated vimentin phosphorylation and subordinated activation of β3 integrin and that they are associated with enhanced neurite outgrowth of co-cultured DRG neurons. Cdc2 induction patterns in reactive astrocytes were robust but transient, which was similarly observed in Schwann cells from the regenerating sciatic nerve following injury (Han et al. 2007; Marinelli et al. 2010). Our recent study shows that some reactive astrocytes that expressed Cdc2 were found within the injury cavity, indicating their migration from the surrounding scarring zone (Seo et al. 2013), as the migration of Schwann cells toward the distal portion of regenerating peripheral nerve has been noted (Torigoe et al. 1996; Han et al. 2007). Thus, by understanding pathophysiological function of Cdc2 in reactive astrocytes, one may gain insights into how reactive astrocytes are involved in regenerative processes after SCI. Among the Cdc2 substrates known to be involved in cell cycle progression (Ubersax et al. 2003), the presence of caldesmone in the cytoplasm mediates cell migration (Manes et al. 2003; Han et al. 2007). Having not observed caldesmone phosphorylation in relation to Cdc2 after SCI (TS and UN, unpublished observation), we examined vimentin as a potential Cdc2 substrate in association with astrocyte function for axon regeneration.

Vimentin is highly expressed in developing neural tissues, down-regulated postnatally, and produced in the epicenter following SCI (Dahl et al. 1981; Lazarides 1982; Ridet et al. 1997; Busch et al. 2010). Experiments using GFAP and vimentin knockout mice suggest that both intermediate filaments act as inhibitory factors for axonal arborization (Menet et al. 2003). Vimentin is present in the macro-architecture assembly of the cell, but, once phosphorylated, it can mediate translocation of integrin-containing endosomes into the plasma membrane and promote cell motility (Ivaska et al. 2005). Vimentin phosphorylation by Cdc2 was shown to release filamentous assembly (Chou et al. 1990; Chang and Goldman 2004). Moreover, an immunohistochemical study localized Cdc2-specific phospho-vimentin in the radial glial cells undergoing active migration in developing neural tissue (Kamei et al. 1998). Here, we identified phospho-vimentin signals in GFAP-positive astrocytes within the injury cavity. This raises the possibility that Cdc2-mediated vimentin phosphorylation in reactive astrocytes that migrated into the injury cavity may be involved in regenerative responses of injured spinal axons.

When primary astrocytes, prepared from pre-injured spinal cord, were subjected to a prolonged culture (longer than 7 days), their proliferation rate was greatly increased at 5–6 days following initial induction phase. Phospho-vimentin signals in the LTC astrocytes were higher than those in the STC cells and in vivo spinal cord tissue. Then, Mn2+ treatment up-regulated phospho-vimentin levels in STC, suggesting functional link between Cdc2 and β3 integrin activation (see below). Integrin, a heterodimeric transmembrane protein composed of α and β subunits, interacts with dozens of molecules at extracellular and intracellular domains and mediates bidirectional signaling. It was reported that integrin activation in the growth cone was involved in regulating axon growth (Lemons and Condic 2008). Likewise, several α and β integrins were identified in astrocytes, and their functional involvement in cell adhesion and migration, modulation of neurite outgrowth, and microvessel integrity following ischemia have been suggested (Tomaselli et al. 1988; Etienne-Manneville and Hall 2001; Tagaya et al. 2001). Notably, β1 integrin was implicated in the induction of non-proliferative reactive gliosis and functional recovery following SCI (Robel et al. 2009; Renault-Mihara et al. 2011). Integrin activation can be regulated at the gene expression level as well as post-translational, conformational change in response to external stimulation (Condic and Letourneau 1997; Takagi et al. 2002).

We found that β3 integrin mRNA was strongly induced from Mn2+-treated LTC astrocytes. Furthermore, analysis of phospho-vimentin and p-FAK signals as effectors for Cdc2 and β3 integrin activation, respectively, revealed that two proteins were co-localized in many, if not all, LTC astrocytes and siRNA inhibition of Cdc2 expression down-regulated FAK phosphorylation. These results suggest that Cdc2-induced vimentin phosphorylation may act as an upstream event that activates β3 integrin. Interestingly however, Mn2+ treatment in STC astrocytes elevated phospho-vimentin levels, suggesting that β3 integrin activation by Mn2+ may induce phospho-vimentin via Cdc2 expression. Indeed, a previous report showed the induction of Cdc2 mRNA by αvβ3 integrin expression in cancer cells (Manes et al. 2003). The heterogeneity in the interaction of β3 integrin with the α integrin counterpart and several adaptor molecules at intracellular and extracellular domains could affect diverse signaling cascades among different cell types. A mechanistic basis on the regulation of Cdc2 and β3 mRNA expression may be explored, for instance, by using endothelial cells, in which vimentin filament is associated with αvβ3 in focal contact and mediates cell adhesion (Bhattacharya et al. 2009).

Our data show that neurite outgrowth of DRG neurons was enhanced by co-culture with LTC astrocytes and further increased by pre-treatment of LTC astrocytes with laminin and Mn2+. Genetic depletion of β3 integrin as well as Cdc2 mRNA attenuated neurite outgrowth. Furthermore, blockade of FAK in astrocytes decreased neurite outgrowth. Although our data suggest that Cdc2 activity is associated with β3 integrin activation in astrocyte-mediated neurite outgrowth, involvement of other signaling pathways such as c-Src in β3 integrin activation cannot be ruled out. In summary, we demonstrated that reactive astrocytes can attain permissiveness for regenerative processes via the activation of Cdc2 pathway. Further studies on role of β3 integrin in the reactive astrocytes and its interaction with extracellular molecules may be important to develop strategy for spinal axon regeneration via the regulation of glial cells.


This study was supported by Basic Science Research Program through the National Research Foundation (NRF) funded by the Ministry of Education, Science and Technology, Korea [2010-0023869] to Uk Namgung. We thank Dana Toy for a critical reading of the manuscript and Seung-Hyung Kim for real-time PCR experiment.

Disclosure of potential conflicts of interest

The authors declare they have no potential conflicts of interest.