Adult neocortex contains dividing satellite glia population even though their characteristics and functions have still remained unknown. Nestin+/NG2+ cells as major fraction of dividing glial cells express bicuculline-sensitive γ-aminobutyric acid A (GABAA) receptors and receive GABAergic inputs. Due to their high [Cl−]i, GABAergic activation depolarized the cells and then induced Ca2+ influx into them. To assess an effect of this GABAergic excitation, we looked for the expression of neurotrophic factors. Among them, we detected the expression of brain-derived neurotrophic factor (BDNF) on the cells. The level of BDNF expression was elevated after cortical ischemia, and this elevation was blocked by bumetanide, an inhibitor for NKCC1 that blocks the GABAergic depolarization. Furthermore, performing a modified adhesive removal test, we observed that the treatment of bumetanide significantly attenuated the recovery in somatosensory dysfunction. Our results may shed a light on satellite glia population in the cortex and imply their roles in the functional recovery after ischemic injuries.
Adult neocortex contains a large number of dividing glial cells (Rakic 1985; Levison et al. 1999; Kornack and Rakic 2001; Dawson et al. 2003; Koketsu et al. 2003). They never differentiate into neurons (Rakic 1985, 2002a, 2002b, 2006a, 2006b; Kornack and Rakic 2001; Koketsu et al. 2003; Au and Fishell 2006; Bhardwaj et al. 2006; Breunig et al. 2007), but recent studies implies some roles of dividing glial cells in cortical circuitry (Spalding et al. 2005; Bhardwaj et al. 2006; Mandyam et al. 2007). Actually, NG2+ cells as dividing glial cells in adult neocortex possess supportive functions for axon growth (Yang et al. 2006) and cell survival (Stallcup 2002). Therefore, it is important to analyze how adult cortical dividing glial cells communicate with cortical circuitry.
Interestingly, NG2+ cells in the adult mouse hippocampus receive excitatory γ-aminobutyric acidergic (GABAergic) inputs (Lin and Bergles 2004). The concentration of GABA increased drastically at the injured hemisphere after the unilateral focal ischemia (Phillis et al. 1994; Matsumoto et al. 1996). It can be assumed that cortical NG2+ cells may receive excitatory GABAergic inputs and play some role in the functional recovery after ischemic injuries.
From these considerations, we supposed that adult cortical NG2+ glial cells produce some neurotrophic factor upon excitatory GABAergic stimulation. Among them, brain-derived neurotrophic factor (BDNF) is the most favorable molecules because it promotes axonal growth and guidance (Lykissas et al. 2007), neuronal survival, and behavior-related plasticity (Lipsky and Marini 2007). In this study, we focused on a NG2+ cell population that also expresses nestin (Lendahl et al. 1990) and observed that excitatory GABAergic inputs toward the dividing glial cells. In addition, our results may suggest that the increase in BDNF production from cortical dividing cells by excitatory GABAergic stimulation play some roles in the functional recovery from ischemic stroke.
Materials and Methods
Nestin-promoter–green fluorescence protein (GFP) (nestinp-GFP) transgenic mice (Yamaguchi et al. 2000), adult male (≥8 weeks old) C57/BL6 mice (Sankyo Lab Service, Tokyo, Japan), and male Sprague Dawley rats (Sankyo Lab Service) were used in our experiments. Experimental procedures were carried out in accordance with animal experimentation protocols approved by the Animal Care and Use Committee at the University of Tokyo.
5-Bromodeoxyuridine Incorporation Assay
The DNA base analog 5-bromodeoxyuridine (BrdU; Wako, Osaka, Japan) was injected intraperitoneally (100 mg/kg body weight in 0.007 N NaOH in phosphate buffered saline [PBS]) 2 h before perfusion to label fast proliferating cells (short pulse). Alternatively, animals received BrdU through their drinking water (long pulse, 1 mg/ml) for 5, 10, or 30 days.
Immunohistochemistry on free-floating sections was performed as previously described (Koketsu et al. 2003), as was immunocytochemistry on cultured cells (Matsumura et al. 2003). For NG2 staining, mice were perfused with PBS followed by 2% paraformaldehyde (PFA) in PBS. Brains were postfixed with the same solution for 4–6 h and then maintained for one day in 30% sucrose–PBS. The slices were blocked by PBS containing 5% normal goat serum for 30 min and incubated with a primary antibody in the same buffer overnight at 4 °C. For BrdU staining, the cryosections were incubated in 2 N HCl for 30 min at 37 °C. After neutralized by borated buffer (pH8.5) for 10 min, they were rinsed by Tris-buffered saline and blocked and incubated in BrdU antibodies (1:200; rat-IgG; Oxford Biotechnology, Oxford, UK) overnight.
Primary antibodies and dilutions used were as follows: GFP (1:1000, rat-IgG, Nakarai Tesque, Kyoto, Japan, 1:1000, rabbit-IgG, Molecular Probes, Carlsbad, CA); NeuN (1:1000, mouse-IgG, Millipore, Billerica, MA); nestin (1:200, mouse-IgG, BD Transduction Laboratories, Franklin Lakes, NJ; 1:200, mouse-IgG, Millipore; 1:200, rabbit-IgG, a generous gift from Dr Ron Mckay); NG2 (1:500, rabbit-IgG, Millipore); Ki67 (1:500, rabbit-IgG, Novocastra, Newcastle, UK); phospho-Histone H3 (pH3; 1:500, rabbit-IgG, Millipore); parvalbumin (PV; 1:500, mouse-IgG, Sigma, St Louis, MO); vesicular GABA transporter (VGAT; 1:500, rabbit-IgG, Synaptic Systems, Gottingen, Germany); BDNF (1:250; rabbit-IgG, Santa Cruz Biotech, Santa Cruz, CA); nerve growth factor (NGF; 1:500, rabbit-IgG, Millipore); vascular endothelial growth factor (VEGF; 1:200, rabbit-IgG, Santa Cruz Biotech); netrin-1 (1:50, goat-IgG, Santa Cruz Biotech); neurotrophin 3 (NT-3; 1:250, rabbit-IgG, Santa Cruz Biotech); MAP-2 (1:500, mouse-IgG, Sigma); or secondary antibodies: Alexa 488-, Alexa 568-conjugated antibodies (Molecular Probes) were used at a dilution of 1:1000, and Cy5-conjugated antibodies (Jackson ImmunoResearch, West grove, PA) were used at a dilution of 1:250. Cultured cells were counterstained with the nuclear stain 4′,6′-diamino-2-phenylindole (Sigma) at a dilution of 1:1000 in order to count the cell number. Images were taken using a confocal microscope (TCS SP2; Leica, Mannheim, Germany) fitted with individual filter sets for each channel. Image production was then performed using Adobe Photoshop (Adobe System, San Jose, CA).
To check the specificity of the BDNF antibody, we performed 2 control experiments. One was western blotting analysis. We confirmed the antibody only detected ∼14 kDa BDNF protein, not NGF or NT-3 proteins. The other was a control immunostaining experiment of cortical slices using the antibody preincubated with BDNF protein. In this case, we did not detect any fluorescent signal in the slices. From these experiments, we validated the specificity of this anti-BDNF antibody. Similar confirmations of the same antibody have been done by another group previously (Hayashi et al. 2001).
Coronal (400-μm thick) brain slices from adult nestinp-GFP+/+ (homozygote) transgenic mice (a generous gift from Dr Yamaguchi) were prepared as described previously (Fukuda et al. 2003; Tozuka et al. 2005). The ice-cold oxygenated solution containing the following (in millimoles [mM]): 125 NaCl, 2.5 KCl, 2 CaCl2, 1.3 MgCl2, 1.25 NaH2PO4, 0.4 L-ascorbic acid, 25 NaHCO3, and 12.5 D-glucose (95% O2–5% CO2, pH 7.4). Slices were stored in the same solution at 37 °C for 30 min and then transferred into artificial cerebrospinal fluid of the following composition (in mM).
Patch-clamp recordings were made at room temperature from GFP+ cells located in the cerebral neocortex. Targeted patch recordings were performed as described previously (Fukuda et al. 2003). For whole-cell recordings, patch pipettes were filled with a K-gluconate–based solution containing (in mM) 120 K-gluconate, 6 NaCl, 6 CaCl2, 2 MgCl2, 2 MgATP, 0.3 NaGTP, 10 ethyleneglycol-bis(2-aminoethylether)-N,N,N′,N′-tetra acetic acid (EGTA), and 10 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) (pH7.2). Some experiments in voltage clamp mode used an intracellular solution comprised of the following (in mM): 140 CsCl, 1 CaCl2, 2 MgATP, 0.3 NaGTP, 10 EGTA, and 10 HEPES (pH7.2). The pipette solution also contained Alexa Fluor 568 hydrazide sodium salt (1/200 dilution; Molecular Probes). No compensation was made for liquid junction potentials. For perforated patch-clamp experiments, a KCl-based pipette solution containing 5 μg/ml gramicidin (Sigma) was used. Procedures for recording cell electrical activity were conducted as described previously (Yoshida et al. 2004).
Electrical stimulation was performed using a monopolar electrode (Unique Medical, Tokyo, Japan) that was placed in the cerebral neocortex (positioned about 300 μm from the recorded cell). The stimuli used to elicit evoked release varied between 300 and 700 μA and were of 500-μs duration. For these experiments, patch pipettes were filled with a CsCl-based intracellular solution.
For Ca2+ imaging of a particular GFP expressing cell, we recorded a cell with the patch pipette, which was filled with 50 μM Fluo-4 (Molecular Probes) for Ca2+ imaging of the recorded cell in a Ca2+-free solution containing (in mM) 120 K-gluconate, 19 KCl, 10 NaCl, 1 MgCl2, 10 HEPES, and 5 EGTA; pH7.2. In all, 200 μM GABA were applied by using the Y-tube system (Yoshida et al. 2004). This system allowed us to apply the drug onto the targeted cells and avoid the cell movement by pressure when drugs were applied.
In the experiments of neurotransmitter response, GABA, N-methyl-D-asparatic acid (NMDA), and α-amino-3-hydroxy-5-methyl-isoxazole-4-propionic acid (AMPA) were applied by focal application using a PicoPump (PV820; World Precision Instruments, Sarasota, FL). Bicuculline methiodide, 6-cyano-7-nitroquinoxaline-2, 3-dione (CNQX), 4-aminopyridine (4-AP), tetrodotoxin (TTX), kynurenic acid, and CdCl2 were bath applied.
Data were recorded with a CEZ-2400 amplifier (Nihon-Kohden, Tokyo, Japan) and Digidata 1200 interface (Axon Instruments, Foster City, CA) controlled by pCLAMP6 (Axon Instruments) software. Membrane current was sampled online at 4 kHz (PowerLab; AD Instruments, Grand Junction, CO) after filtering at 2 kHz. Data were recorded with Chart 4.0 software (AD Instruments) and stored on the hard disk of a personal computer for later off-line analysis with Igor Pro 4.01 (WaveMetrics, Lake Oswego, OR). For single-cell Ca2+ imaging, the AquaCosmos system (Hamamatsu Photonics, Hamamatsu, Japan) was used as described previously (Imura et al. 2005).
After the recording session, the patch electrode was carefully removed from the cell and the slice was fixed for 4–6 h at 4 °C in 4% PFA in PBS. Immunohistochemistry using antibody against NG2 was performed on free-floating slices as described before.
Mice were perfused with PBS, followed by 4% PFA containing 0.1% glutaraldehyde in PBS at room temperature. Brains were postfixed overnight in 4% PFA in PBS at 4 °C. Fifty-micron-thick vibratome sections of the hippocampus prepared as described above were incubated with rabbit anti-GFP (1:1000; Molecular Probes) for 48 h at 4 °C and then with peroxidase-conjugated goat polyclonal anti-rabbit IgG (1:200; GE Healthcare, Buckinghamshire, UK) for 2 h. The sections were incubated with a 3,3′-diaminobenzidine tetrahydrochloride (DAB; Sigma) solution for 15 min and then with a DAB solution containing 0.01% H2O2 for 15 min. Finally, sections were postfixed with 1% OsO4 in PB, dehydrated, and embedded in Epon 812. Thin sections (70–80 nm) were mounted on uncoated grids, stained with lead citrate, and examined by electron microscopy (EM; H7600, Hitachi, Tokyo, Japan) according to the precedent protocol (Tozuka et al. 2005).
Cell Isolation and Cell Culture
Adult cortical nestin+/NG2+ cells were isolated from adult mouse and rat neocortex by Percoll gradient methods and cultured as described previously (Palmer et al. 1999; Matsumura et al. 2003). We illuminated the brain section including cortical tissue using a white halogen lamp which could distinguish between the gray matter and the white matter very clearly (Matsumura et al. 2003). Hence, we precisely took the neocortex, out of all parts of other brain tissues, including the white matter, the subventricular zone, and the hippocampus. The cerebral neocortex was enzymatically dissociated with papain (2.5 units/ml; Sigma), dispase (1 unit/ml; Invitrogen Corp. Carlsbad, CA), and DNase I (2500 units/ml; Worthington, Freehold, NJ) solution. Whole digested tissue was then suspended in 50% Percoll solution (GE Healthcare) and fractionated by centrifugation for 30 min at 20 000 × g. Fractionated cells between 1.052 g/ml and 1.075 g/ml were plated onto poly-L-ornithine/laminin/fibronectin–coated 24-well cell culture dishes in DMEM/F-12 medium containing 10% fetal bovine serum (FBS) for 24 h. On the next day, the medium was replaced with serum-free growth medium consisting of DMEM/F-12 medium (1:1) with N2 supplement, 20 ng/ml basic fibroblast growth factor (Invitrogen Corp.), and 20 ng/ml epidermal growth factor (Wako) for subsequent 48 h. In this culture system, almost all cells (>90%) were immunoreactive for both nestin and NG2 after 3 days’ culture; but MAP-2+ neurons were less than 1%.
Measurement of the Amount of Secreted BDNF by ELISA
The secretion of BDNF protein from nestin+/NG2+ cells grown in culture medium for 72 h was measured. Cells were transferred to 300 μl HEPES-based buffer containing (in mM) 137 NaCl, 25 glucose, 10 HEPES, 5 KCl, 1 MgCl2, and 3 CaCl2 (pH7.4). Some plates were incubated with GABA (200 μM) and bicuculline methiodide (100 μM) for 9 h. After the process, the amount of BDNF that had been released in the HEPES-based buffer was measured by BDNF sandwich ELISA, using a commercial kit according to the manufacturer instructions (Millipore). After collecting the medium, cells were harvested with trypsin, and then the numbers of cells were counted. BDNF levels were expressed as the release of BDNF/cell.
In pilot experiments, we tested the effect of different doses (50, 200, or 1000 μM) of GABA on BDNF release from nestin+/NG2+ cells. The result was that 200 μM GABA was the most efficient concentration for BDNF secretion (data not shown).
Photothrombotic Ischemic Model
Photothrombotic ischemia at sensory-motor area of adult mouse neocortex was produced according to precedent protocols (Schäbitz et al. 2004, 2007; Lee et al. 2007). Male C57/BL6 mice were first anesthetized by ketamine (50 mg/kg) and xylazine (5 mg/kg). Then they were intravenously administered 20 mg/kg of rose bengal (Sigma) for 5 min. For illumination, a fiber-optic bundle with a 1.0 mm aperture (white-light beam; Luminar Ace LA-150TX, Hayashi Watch Co. Ltd, Tokyo, Japan) was placed stereotaxically onto the skull ±0 mm to the bregma 1.5 mm lateral from the midline on the right hemisphere (Paxinos and Franklin 2001). The mice were illuminated for 20 min to produce reactive oxygen from rose bengal. During the illumination, they were kept on a heating plate in order to keep their temperature at 37.0 ± 1.0 °C. Mice having higher or lower temperature during the illumination were excluded from the experiments, and all experimental instruments were covered to protect from the room light. Sham-operated animals underwent the same experimental procedures without the illumination.
Fluorescence Activated Cell Sorting Analysis
We performed intracellular nestin/BDNF staining and fluorescence activated cell sorting (FACS) analysis according to previous works (Raap et al. 2005; Noga et al. 2008) with some modifications. Adult male C57 BL/6 mice undergone photothrombotic ischemia were perfused by PBS, and their brains were taken out immediately on the day 4–6 after the surgery. In some cases, the mice had been intravenously receiving 10% dimethyl sulfoxide (DMSO) in saline as vehicle or a NKCC1 antagonist, bumetanide at 1 mg/kg/day (Wako) dissolved in the same buffer. Cortical cells were dissociated from ipsilateral hemisphere. Importantly, we carefully cleared off the infarct area which can be recognized by its white color in the ipsilateral hemisphere. The cells were harvested in Hanks’ balanced salt solution with the enzyme as described above for 35 min at 37 °C. The cells were washed by PBS, incubated in 2.5 μg/ml ethidium monoazide bromide (EMA; Molecular Probes) in PBS for 15 min, and lighted for 10 min. They were next fixed by 2% PFA for 20 min. After PBS wash, they were first incubated in PBS containing 0.1% sodium azide (Wako), 0.1% BSA, and 0.1% saponin (Sigma) for 10 min and then blocked in the same buffer with addition of 5% FBS for 45 min. After spun, they were incubated in primary antibodies dissolved in the same buffer with serum for 30 min: mouse anti-nestin (1:1000; BD Pharmingen, San Diego, CA) and rabbit anti-BDNF (1:1000; Santa Cruz Biotech). After washed 2 times, they were incubated in secondary antibodies for 30 min as follows: anti-mouse IgG FITC (1:1000; DAKO, Glostrup, Denmark) and anti-rabbit IgG phycoerythrin (1:2500; Caltag, Carlsbad, CA). The cells were washed 3 times by the buffer without saponin.
We performed FACS analysis according to our previous works (Yoshida et al. 2003; Muramatsu et al. 2005) using FACS Caliber (Becton Dickinson, San Jose, CA). Cellular debris and dead cells were excluded by FSC/SSC scatters and the EMA expression. Cells only incubated with the secondary antibodies were used as control. For compensation, we used cells incubated in either nestin or BDNF antibody. We analyzed about 4000 nestin+ cells from each sample and made histograms of their BDNF intensity.
To acquire control images of BDNF staining, we followed all the method using the buffer not containing saponin. In this experiment, we could not detect any nestin or BDNF signals. Moreover, we also performed the same experiment by using cells kept in a refrigerator for a week after the PFA fixation. Also in this experiment, BDNF signal was not detected.
All behavior tests were conducted by an experimenter blinded to the study code. Adult male C57/BL6 mice were placed on a rotarod cylinder (Model 7600; UGO Basile, Comerio, Italy), and the remaining time on the rod was measured. The trial ended if the mice fell off the rungs or gripped the device and spun around for 2 consecutive revolutions without attempting to walk on the rungs (Schäbitz et al. 2004). The rotation speed was set at steady speed, 25 rpm. All mice were trained to achieve 200-s walking 3 days prior to the surgery. Mice failing this task were excluded from the subsequent experiment. The trained mice were administered sham or ischemic surgery according to the method described above.
For sensory measurement, a modified adhesive removal testing has been done according to the precedent protocol (Komotar et al. 2007) with slight adjustments to apply for mice. Mice first wrapped a 0.7 × 2.0 cm green tape around their cuff. The mice were put into the test box, and then, we waited for 10 s so as that they accommodated this environment. During the next 30 s, mice tried to remove the tape, such as biting the tape or shaking their hand, were measured by a stopwatch. The trails were repeated 4 times about each hand and the averages of the best 2 times were calculated everyday (for details, see Supplementary Videos online).
To analyze the influence of the adult cortical nestin+ cells onto neuronal recovery after the ischemia, ischemia-injured mice were intravenously administered vehicle as 10% DMSO or 1 mg/kg bumetanide (Wako) in 10% DMSO, an inhibitor for NKCC1 transporter (Dzhala et al. 2005; Wang and Kriegstein 2008) everyday.
Data were expressed as mean ± standard error of the mean (SEM). The statistical significance of the difference between means was assessed using 2-sided Student's t-test or Wilcoxon's test, with the level of significance being set at *P < 0.05, **P < 0.01.
Widespread Distribution of Dividing Nestin+/NG2+ Cells in the Adult Mouse Neocortex
The presence of nestin+ cells in the neocortex of the adult mouse brain was characterized by using nestinp-GFP transgenic mice (Yamaguchi et al. 2000) that express GFP under the control of the nestin promoter and its second intron (Zimmerman et al. 1994). This construct limits GFP expression to neural precursors even though nestin protein is also expressed in muscle precursors (Zimmerman et al. 1994). As shown in Figure 1(A–C), nestinp-GFP+ cells could be seen distributed throughout all cortical layers. These cells were typified by numerous, ramified thin processes radiating from a small cell body and were often observed adhering to cell bodies of pyramidal neurons (Fig. 1D), indicating that they are a type of satellite glia population (Rakic 1985; Kornack and Rakic 2001; Koketsu et al. 2003). Following the immunostaining of sections with anti-nestin antibody, all nestinp-GFP+ cells were positively stained with anti-nestin antibody (Fig. 1E). In this area, we did not detect any nestinp-GFP+ cells that also expressed GFAP (data not shown). To further assess whether nestinp-GFP+ cells also expressed NG2, we stained nestinp-GFP+ cells with anti-NG2 antibody. As shown in Figure 1(F), all nestinp-GFP+ cells were immunopositive for NG2. Conversely, more than half of the NG2+ cell population (56%) expressed nestinp-GFP (Fig. 1G,H). The data from triple immunostaining with anti-GFP, anti-nestin, and anti-NG2 were shown in Figure 1(I). These results strongly indicate that the nestin+ cells are a subpopulation of NG2+ cells in the adult neocortex. Interestingly, a part of nestinp-GFP+ cells (2%; 10 out of 500 nestinp-GFP+ cells) weakly but significantly expressed NeuN, even though they showed morphologically multipolar glial phenotype (Fig. 1J).
To assess the dividing activity of nestin+/NG2+ cells, we used 2 intrinsic mitotic marker antibodies, one against Ki67, a cell proliferation-associated nuclear protein expressed during the active cell cycle (G1, S, G2, and M phases) but absent in resting cells (G0 phase), and pH3, an M-phase marker. Small percentages of the total nestinp-GFP+ cell number were immunoreactive for Ki67 (3.83 ± 0.93%) (Fig. 1K) and pH3 (0.50 ± 0.29%) (Fig. 1L), indicating that relatively few nestin+/NG2+ cells are actively cycling in situ. Furthermore, we estimated the proliferative ability of nestin+/NG2+ cells by means of BrdU-pulse labeling experiments. For short-term pulses, in which we labeled proliferating nestin+ cells by injecting the BrdU 2 h before analysis, 1.63 ± 0.26% of nestin+/NG2+ cells in the adult neocortex incorporated BrdU. For long-term pulses, where BrdU was supplied for 5 days, 10 days, or 30 days in the drinking water to label slowly dividing cells, the percentage of BrdU+ cells in the nestin+/NG2+ cell population was 19.98 ± 2.90%, 33.23 ± 1.42%, and 82.33 ± 3.28%, respectively (Fig. 1M,N). Based on these data, we can conclude that the widely distributed nestin+/NG2+ cells in the adult cortex are mitotically active cells.
GABAergic Inputs into Nestin+/NG2+ Cells in the Adult Neocortex
To examine whether any form of communication exists between nestin+/NG2+ cells and surrounding cortical circuits, we performed GFP-guided targeted electrophysiological recordings from nestin+/NG2+ cells in fresh cortical slices to determine possible neural efferent connections. After recordings of nestinp-GFP+ cells with Alexa Fluor 568 in the pipette, we confirmed the expression of NG2 proteins on the surface of all recorded GFP+ cells (n = 15/15; Fig. 2A–D). The basic membrane properties of the recorded nestin+/NG2+ cells showed that they were nonexcitable cells with a graded level of slowly inactivating outward current (Fig. 2E,F), reminiscent of the A current. Their mean input resistance was 157.30 ± 30.51 MΩ (mean ± SEM, n = 48) and their mean membrane potential was –79.07 ± 1.28 mV (n = 48); these characteristics being similar to those of NG2+ cells existing in the postnatal neocortex (Chittajallu et al. 2004) and the postnatal hippocampus (Mangin et al. 2008; Kukley et al. 2008).
To assess whether these nestin+/NG2+ cells actually communicate with surrounding neural circuits, we investigated their responses to either GABA or glutamate. As shown in Figure 2(G), significant inward currents were detected in almost all nestin+/NG2+ cells (n = 34/37) following the focal application of GABA. These currents were completely inhibited (100% of all GABA-responsive nestin+/NG2+ cells; n = 10) by bicuculline, a selective antagonist of the GABAA receptor (GABAA-R) (Fig. 2G). The mean EC50 of GABA was 55.04 μM (Fig. 2H). In addition to GABA, taurine also activated GABAA-R–mediated currents in nestin+/NG2+ cells (Fig. 2G) (n = 8/8). In relation to the activation of glutamate receptor–mediated currents, some nestin+/NG2+ cells (n = 6/13) responded only to an high concentration of AMPA/kainate (0.5 mM; Fig. 2G), whereas none of the recorded cells responded to NMDA (n = 0/11) (Fig. 2G). These results suggest that cortical nestin+/NG2+ cells make GABAergic connections with cortical neurons.
Previous reports have clearly demonstrated that NG2+ cells are contacted by functional GABAergic terminals in developing and adult hippocampus (Lin and Bergles 2004; Kukley et al. 2008; Mangin et al. 2008). Then, we next examined whether adult cortical nestin+/NG2+ cells also receive functional GABAergic inputs from surrounding neural circuits. We next examined whether nestin+/NG2+ cells actively receive functional GABAergic inputs from surrounding cortical neural circuits. Nestin+/NG2+ cells showed very low-frequency, spontaneous postsynaptic currents during prolonged recording times (1–2 events per 10 min of recording) (Fig. 3A–B). We estimated their rise time as being 4.7 ± 1.0 ms (n = 6) and the decay time as being 25.3 ± 5.7 ms (n = 6). This decay time was similar to values of previous studies about GABAergic synaptic inputs into NG2+ cells (Lin and Bergles 2004; Kukley et al. 2007), suggesting that these communications were from GABAergic neurons. To enhance the frequency of synaptic events, 4-AP (200 μM), a K+ channel blocker that augments transmitter release (Hennou et al. 2002) was bath applied. As shown in Figure 3(C–D), the frequency of spontaneous synaptic currents in nestin+/NG2+ cells was increased in the presence of 4-AP and almost completely blocked by bicuculline (n = 5). Next, we found that electrical stimulation of cortical GABAergic neurons elicited GABAergic currents in nestin+/NG2+ cells. These stimuli-evoked currents were completely abolished by either TTX (n = 7) (Fig. 3E,E), or bicuculline (Bic; n = 9) (Fig. 3E,E), whereas kynurenic acid, a broad-spectrum blocker of ionotropic glutamate receptors, had no effect on these inward currents (Kyn; n = 5) (Fig. 3E,E). The reversal potential elicited by stimulation was close to that estimated for Cl− (Fig. 3F). It has demonstrated that NG2+ cells did not connect with surrounding neurons via gap junction coupling or receive tonic release of GABA (Lin and Bergles 2004; Kukley et al. 2008; Mangin et al. 2008). Based on previous reports and our current data, we can conclude that neural inputs on cortical nestin+/NG2+ cells are action potential–dependent GABAergic synaptic events.
To prove synaptic contacts between nestin+/NG2+ cells and GABAergic neurons, we performed a series of experiments in which both EM and confocal microscopy were used. In the EM analysis, we found that presynaptic structures, including vesicles, were closely apposed to the surface of nestin+/NG2+ cells (Fig. 4A–C). We used immunohistochemistry to examine whether GABAergic terminals make contact with the nestin+/NG2+ cells. For GABAergic synapse markers, antibodies against VGAT and gephyrin were used. VGAT+ terminals were observed to be in close apposition to nestin+/NG2+ cells. Many VGAT+ terminals were colabeled with PV, thus giving rise to one hypothesis that PV+ GABAergic interneurons make contact on nestin+/NG2+ cells (Fig. 4D–H). These anatomical results may imply GABAergic terminals on nestin+/NG2+ cells.
To determine the physical nature of the influence of GABAergic inputs on nestin+/NG2+ cells, we evaluated the short-term plasticity of the synaptic junctions they form. Paired-pulse facilitation and paired-pulse depression (PPD) are frequency-dependent short-term changes at GABAergic synapses, which have been suggested to depend on presynaptic neuronal activation. We performed paired-pulse stimulation as previously described (Lin et al. 2005) and found that the GABAergic responses of the nestin+/NG2+ cells showed PPD (Fig. 4I) in a manner similar to evoked inhibitory postsynaptic currents in PV+ fast-spiking neuron synapses (Galarreta and Hestrin 2002). Based on these electrophysiological and anatomical observations, these results strongly suggest that the connections between nestin+/NG2+ cells and PV+ GABAergic neurons may serve important functions in the maintenance and plasticity of the adult cortical neural network.
GABAergic Excitation May Promote the Production of BDNF in Nestin+/NG2+ Cells
In immature neuronal cells, the opening of GABAA-receptor channels often triggers the efflux of Cl− (due to the elevated level of [Cl−]i) and subsequent membrane depolarization (Ben-Ari 2002; Owens and Kriegstein 2002). To assess whether GABAergic activity depolarizes nestin+/NG2+ cells, we performed gramicidin-perforated patch-clamp recordings. The focal application of GABA depolarized nestin+/NG2+ cells held in current clamp mode (Fig. 5A). The evoked currents in nestin+/NG2+ cells reversed at –30.61 ± 3.70 mV (mean ± SEM, n = 10), that is, at a level more depolarized than their resting membrane potential (Fig. 5B). Based on this reversal potential, nestin+/NG2+ cells have a calculated [Cl−]i of 44.67 ± 6.31 mM (mean ± SEM, n = 10), indicating that nestin+/NG2+ cells maintain a higher [Cl−]i than mature cortical neurons as described previously (Yoshida et al. 2004). These data suggest that GABAergic inputs should have excitatory effects on nestin+/NG2+ cells in the adult neocortex.
Next, we tested whether GABAergic activity could trigger the elevation of [Ca2+]i in nestin+/NG2+ cells. In single-cell Ca2+-imaging experiments with whole-cell recordings, patch pipettes were filled with a Ca2+-free solution that contained 45 mM Cl− to mimic the intracellular environment. As shown in Figure 5(C), GABA application resulted in the elevation of [Ca2+]i in the nestin+/NG2+ cell soma, a response that was blocked by bath application of the voltage-gated calcium channel (VGCC) blockers Cd2+ and Ni2+. This transient elevation of Ca2+ was also observed in cell processes (Fig. 5D). It is likely that the GABA-induced, VGCC-mediated influx of intracellular Ca2+ thus modifies an intracellular signaling pathway or triggers the synthesis or release of compound that acts extracellularly.
To assess whether excitatory GABAergic inputs on adult cortical nestin+/NG2+ cells would be beneficial for plasticity and repair in cortical circuits, we performed immunohistochemistry for BDNF, NGF, NT-3, VEGF, and netrin-1. Among these proteins with potentially beneficial effects on the regeneration of neural circuits, the robust expression of BDNF was observed in nestin+/NG2+ cells located over a wide range of sites including the sensory-motor area (Fig. 6A) and entorhinal cortex (Fig. 6B). In contrast, we were unable to detect immunohistochemically any significant expression of the other trophic factors or chemoattractant molecules on nestin+/NG2+ cells.
We next examined the possibility that BDNF secretion from nestin+/NG2+ cells could be induced by excitatory GABAergic stimulation. To quantify the BDNF secretion from nestin+/NG2+ cells, we isolated nestin+/NG2+ cells from adult rat neocortex by means of the Percoll gradient centrifugation method and then measured by ELISA the amount of BDNF secreted (Fig. 6C–D). We evaluated whether GABAergic stimulation induced the elevation of [Ca2+]i and the production of BDNF (Fig. 6E,F). As shown in Figure 6E, the rapid increase in [Ca2+]i was observed in cultured nestin+/NG2+ cells stimulated with GABA. For the evaluation of BDNF production and/or secretion, the culture supernatant from nestin+/NG2+ cells stimulated with GABA was collected and the amount of secreted BDNF was measured by ELISA. We detected the clear increase in BDNF secretion from the culture stimulated with GABA 9 h before in comparison to the control cultures (Fig. 6F). We did not observe the enhanced secretion of BDNF just after GABA stimulation (<30 min) (data not shown). These results may suggest that GABAergic stimulation promotes the production, but not secretion, of BDNF from cultured nestin+/NG2+ cells.
GABAergic Excitation and Increase in the Production of BDNF in Nestin+/NG2+ Cells after Ischemic Stroke
To assess whether the ischemic injuries stimulate the production BDNF in nestin+ cells in the area beside, but apart from the peri-infarct area, we utilized a mouse photothrombosis model at sensory-motor areas (Fig. 6G,H) and performed both immunohistochemical analysis and quantitative FACS analysis. The merit of this lesion model is highly reproducibility in size and location. In a pilot experiment, we detected the elevation of BDNF production in nestin+/NG2+ cells on and after day 4, but not on day 2 after ischemia. In many of nestin+/NG2+ cells located within 50–500 μm apart from the peri-infarct area at day 4, we detected the elevated production of BDNF proteins in cellular process areas as well as cell cytoplasm area (Fig. 6I,J).
To quantify the BDNF levels, we performed FACS analysis after the intracellular staining of BDNF proteins in combination with anti-nestin staining. This method enables us to quantify the amount of BDNF proteins in each cortical cell. The detection of the intracellular BDNF proteins by FACS analysis has been performed elsewhere (Raap et al. 2005; Noga et al. 2008). We carefully dissected the cortical tissue from ipsilateral side of the neocortex, getting rid of the ischemic core and the peri-infarct area where reactive gliosis occurs. Then, we applied stained cells (ca. 500 000–800 000 cells) and evaluated the amount of the intracellular BDNF proteins in nestin+ cortical cells (Fig. 7A). Subsequently, we performed the parallel FACS analysis using both ipsilateral and contralateral cortical cells. As shown in Figure 7(B), fluorescent intensities from anti-BDNF staining were higher in ipsilateral cells than in contralateral cells in the repeated experiment. These results indicated, after ischemic injuries, the increase in the amount of the intracellular BDNF in cortical nestin+ cells as well as nestin− cells that were most likely neurons. The enhanced production of BDNF from neurons besides ischemic region has been reported previously (Kokaia et al. 1998). This experiment would be the first example of quantitative BDNF-FACS analysis of neural cells. From these results, we may observe that cortical ischemic injuries promote the production of BDNF in nestin+ glial cells in the mammalian neocortex.
Next, we assessed whether excitatory GABAergic stimulation would play a role in the promotion of BDNF production in nestin+ cells. To this purpose, we utilized bumetanide, a NKCC1 inhibitor. Because the daily systemic injection of bumetanide blocks the excitatory GABAergic stimulation in brain (Dzhala et al. 2005; Wang and Kriegstein 2008), we prepared the 2 groups of mice underwent cortical ischemia: one with daily bumetanide injection for 4 days and the other with daily vehicle (10% DMSO) injection for 4 days. We purified both ipsilateral and contralateral cortical cells from each group and then applied for BDNF-FACS analysis (Fig. 8). As a result, we observed that the increase in BDNF production after ischemia was completely blocked by this bumetanide treatment. Taken together with all these data, we can stress that GABAergic excitation may promote the production of BDNF from nestin+ cells after the ischemic injuries.
Block of the GABAergic Depolarization Causes the Attenuation of Functional Recovery in Sensory Functions
To assess an effect of excitatory GABAergic stimulation on the functional recovery after ischemic injury, we performed 2 behavior tests: rotarod to evaluate the recovery in general sensory-motor functions and adhesive removal test to evaluate the recovery in sensory function. For the adhesive removal test, we followed a protocol reported by Komotar et al. (2007). The advantages of this new sensory test are as follows. 1) This method can be evaluated by only sensory functions, not including motor functions because it is not necessary to move the impaired hand. 2) Drug influence onto uninjured network can be measured at the same time by analyzing the sensory activities of contralateral side. 3) Pretraining is not necessary, so this experiment is not time consuming. (Komotar et al. 2007; for details, see Supplementary Videos online).
The presence of the microinfarction significantly impaired sensory-motor function in both rotarod and adhesive removal tests (Fig. 9). Next, to analyze whether or not the functional recovery was influenced by GABAergic stimulation, some mice were administered bumetanide. Although, there was a tendency for suppression by the bumetanide treatment on functional recovery as assessed in rotarod test, we did not detect the significant difference between the 2 groups (Fig. 9B). As shown in Figure 9(C), this administration significantly attenuated sensory recovery on the day 6, 7, and 8, as evaluated by the adhesive removal test (Fig. 9C and Supplementary Video 1 online). In addition, the results from the contralateral side indicated that the normal sensory function was not affected by this bumetanide administration (Fig. 9D and Supplementary Video 2 online). All these behavioral data suggest the significant attenuation of functional recovery in sensory functions by the bumetanide treatment.
Adult cortical tissue contains large number of dividing glial cells (Rakic 1985, 2002a, 2002b, 2006a, 2006b; Kornack and Rakic 2001; Koketsu et al. 2003), even though their detailed characteristics and functions have remained unclear. Very recently, it has been reported that in the dorsal root ganglia they affect pain-like sensory behavior (Vit et al. 2008) and neuron–glia communication (Zhang et al. 2007). A series of reports have shown that a particular subpopulation of NG2+ cells form synaptic junctions with GABAergic as well as glutamatergic neurons (Lin and Bergles 2004; Lin et al. 2005; Kukley et al. 2007; Ziskin et al. 2007). In this communication, we report that cortical nestin+ glial cells received GABAergic inputs.
Our histological results clearly demonstrate that PV+ interneurons (basket cells or chandelier cells) extend GABAergic fibers toward nestin+/NG2+ cells (Fig. 4). In addition, a short-term plasticity test showing paired-pulse depression at GABAergic synapses on nestin+/NG2+ cells showed similar characteristics to those shown by PV+ GABAergic synapses on cortical interneurons (Galarreta and Hestrin 2002). These experimental findings clearly indicate that PV+ GABAergic interneurons specifically innervate and make functional GABAergic synapses with nestin+/NG2+ cells.
Interestingly, we also discovered a part of nestinp-GFP+ cells (2%) weakly but significantly expressed a neuronal marker NeuN, even though they belonged to NG2+ cells (Fig. 1J). Gallo et al. have also suggested that some of NG2+ cells expressed NeuN (Belachew et al. 2003; Aguirre et al. 2004) and possessed voltage-gated sodium channels (Chittajallu et al. 2004). We previously detected few BrdU+/NeuN+ cells in neocortex of adult nonhuman primates (Koketsu et al. 2003); however, these cells may belong to these glial cells, thereby the credibility of NeuN to detect neuronal cells remains ambiguous. It can be assumed that a subpopulation of NG2+ glial cells expresses NeuN. In turn, to evaluate neurogenesis in vivo, only double staining with NeuN and BrdU would not be sufficient. Further studies are definitely needed to characterize nestin+/NG2+ cells expressing NeuN.
In this study, we have clearly demonstrated by BDNF-FACS analysis that, upon ischemic injuries, the productions of BDNF from nestin+/NG2+ cells was elevated. This elevation might be due to the excitatory GABAergic stimulation after ischemic injuries (Fig. 8). The quantification of cytokines using FACS was performed mostly in the area of immunology. However, recently, this method has become applicable for quantification of neurotrophins such as BDNF and NT-3 (Raap et al. 2005; Noga et al. 2008). This FACS quantification has various advantages. First, we can detect the presence of intracellular BDNF. Second, we can identify the cell type simultaneously. Third, we can quantify the amount of BDNF precisely. Fourth, we can perform the experiment using fewer cells than other methods as western blotting.
The activity-dependent production of BDNF from neurons had been reported previously, in which calcium signaling promoted the elevation of cyclic adenosine 3′,5′monophosphate and then triggered the production of BDNF (Obrietan et al. 2002). Because nestin+/NG2+ cells are an immature cell type–like nestin+ progenitor cells in adult dentate gyrus (Tozuka et al. 2005), GABA in these cases acts as an excitatory neurotransmitter, initiating the elevation of [Ca2+]i. It is reasonable to postulate that nestin+/NG2+ cells, closely apposed to neighboring neurons, serve as a site of BDNF secretion, supported by the results from Komitova et al. (2006). BDNF promote neuronal survival (Lipsky and Marini 2007), dendrite development (Horch and Katz 2002), synaptic plasticity (Chakravarthy et al. 2006), and neuronal differentiation (Scharfman et al. 2005). However, NG2 knockout mice did not influence on adult neurogenesis in dentate gyrus and hippocampal-dependent behavioral tasks (Thallmair et al. 2006) and they had no difference of development compared with the wild-type mice (Grako et al. 1999). These results suggest that the BDNF expression from adult cortical glial cells may influence on the plasticity rather than the general development of cortical circuitry.
After the injuries, GABA and glutamate increased in the peri-infarct area (Phillis et al. 1994; Matsumoto et al. 1996), and glutamate excitotoxity leads to cellular death (Leach et al. 1993; Lyden and Lonzo 1994; Zipfel et al. 2000). Conversely, GABA (Green et al. 2000; Rudolph and Möhler 2006), GABA agonists (Schwarz-Bloom and Sah 2001), and taurine (Molchanova et al. 2006; Wang et al. 2007) attenuated ischemic damages, suggesting the role of GABAergic network on the protection and repair of neural circuit after ischemic injuries. In addition, recent studies have well suggested that BDNF plays a crucial role in repairing neural circuits. Ectopic expression of BDNF by either the infusion or the viral transfer into the damaged circuit promoted the functional recovery after brain injuries, such as stroke or spinal cord injuries (Schäbitz et al. 2004, 2007; Girard et al. 2005). In addition, the endogenous production of BDNF promoted the functional recovery after brain injuries (Griesbach et al. 2004).
In this study, we also clearly demonstrated that an inhibitor for NKCC1 transporter, bumetanide, suppressed sensory recovery after focal ischemia (Fig. 8C). NKCC1 maintains high [Cl−]i and causes cellular depolarization after the opening of Cl− channels. At early postnatal stage, very recently, the depolarization of cortical neurons by GABAergic stimulation had been reported (Khirug et al. 2008). Others reported the robust expression of NKCC1 in NG2+ cells (Price et al. 2006). [Cl−]i of nestin+/NG2+ cells was extremely high as 45 mM, in which EGABA is around −30 mV. Therefore, it is reasonable to assume that GABAergic depolarization by NKCC1 could occur in such immature nestin+/NG2+ glial cells like neuronal cells during postnatal stage (Wang and Kriegstein 2008).
From the point of satellite glial cells, their activity greatly affects pain-like sensory mechanism of ganglia through secreting cytokines (Scholz and Woolf 2007; Vit et al. 2008). Following inflammation, activation of satellite glial cells promoted the secretion of IL-1 and this factor enhanced excitability of ganglion sensory neurons (Takeda et al. 2007). Not only neuronal cells but also satellite glial cells intervene in the regulation of sensory functions by secreting some neuronal factors, and same phenomenon can arise in the ischemic damaged neocortex. In addition to satellite glial cells, oligodendrocytes, astrocytes, or microglia secrete such growth factors. Actually, we detected few mature glial cells (APC+) expressing BDNF, like a previous study by Dai et al. (2003) in the area 50–500 μm apart from the peri-infarct area (indicated by blue color in Fig. 6H). We focused on this area because we can speculate reorganization of damaged neural circuit. In this area, the density of nestin+/NG2+ cells coexpressing BDNF was extremely higher than that of APC+ cells. In addition, we observed the clear enhancement of BDNF production in nestin+/NG2+ cells in response to the ischemic injury. Therefore, these results imply that nestin+/NG2+ satellite glia may also be involved in the functional recovery after ischemic injuries. Further works must be necessary to elucidate the effect of enhanced BDNF production from adult cortical satellite glial cells on circuit protection and repair. We hope that our results provide additional cues to understand the function and plasticity of adult neocortex and to create a new therapeutic application for ischemic injuries.
Ministry of Education, Science, Sports and Culture of Japan (19208030).
We thank Drs Craig E. Jahr, Vittorio Gallo, Akiko Nishiyama, Christian Steinhäuser, Carlos Vicario-Abejon, and Angelique Bordey for informative discussions and insightful comments. Dr Masahiro Yamaguchi is thanked for the gift of nestinp-GFP transgenic mice and Dr Aiko Hirata for support with the EM analysis. We also thank Dr Masayuki Sekiguchi for the electrophysiological data analysis. Conflict of Interest: None declared.