Long-Term Potentiation of Neuron-Glia Synapses Mediated by Ca2+-Permeable AMPA Receptors

See allHide authors and affiliations

Science  09 Jun 2006:
Vol. 312, Issue 5779, pp. 1533-1537
DOI: 10.1126/science.1124669


Interactions between neurons and glial cells in the brain may serve important functions in the development, maintenance, and plasticity of neural circuits. Fast neuron-glia synaptic transmission has been found between hippocampal neurons and NG2 cells, a distinct population of macroglia-like cells widely distributed in the brain. We report that these neuron-glia synapses undergo activity-dependent modifications analogous to long-term potentiation (LTP) at excitatory synapses, a hallmark of neuronal plasticity. However, unlike the induction of LTP at many neuron-neuron synapses, both induction and expression of LTP at neuron-NG2 synapses involve Ca2+-permeable AMPA receptors on NG2 cells.

Glial cells in the central nervous system (CNS) constitute a heterogeneous population of cell types. Macroglia-like NG2 cells express the chondriotin sulfate proteoglycan NG2 and have been described as oligodendrocyte precursor cells (OPCs) or given other names (13). NG2 cells in the CA1 area of the hippocampus receive direct glutamatergic and γ-aminobutyric acid (GABA)–ergic synaptic inputs from neurons (4, 5). The structure of the neuron-glia synapses found in NG2 cells differs from that of neuronal synapses by having a less well-defined postsynaptic density and smaller presynaptic boutons that contain fewer vesicles (4). With the exception of developing neuromuscular junctions (6), long-term potentiation (LTP) has been observed only at synapses between neurons. Thus it is of interest to examine whether the neuron-NG2 cell synapses have adequate expression and localization of components required for both the induction and expression of LTP.

Whole-cell recordings were made from NG2 cells in the CA1 region of rat hippocampal slices. Glial cell membrane currents induced by the stimulation of Schaffer collaterals (SCs) were used to monitor rapid neuron-glial signaling. The identity of astrocytes and NG2 cells was determined by both electrophysiological recording and immunostaining. Staining with antibodies to glial fibrillary acidic protein (GFAP) and NG2 revealed two distinct nonoverlapping cell populations (fig. S1) (4). NG2 cells were identified by having a relatively high input resistance (131.7 ± 5.4 megohms, n = 160 cells), large transient A-type currents (IA) and delayed rectifier K+ currents (IK), and small tetrodotoxin (TTX)–sensitive Na+ currents that failed to generate typical action potentials (Fig. 1 and fig. S1D) (4, 7). We found no detectable voltage-dependent Ca2+ currents in these cells (Fig. 1, F and G). When cells with the above characteristics were marked by intracellular loading with biocytin or lucifer yellow and subsequently immunostained with an antibody against NG2, all of the cells (14 out of 14) were found to be NG2-positive (fig. S1).

Fig. 1.

Electrophysiological properties of NG2 cells. (A) Three-dimensional reconstruction of a typical NG2 cell identified electrophysiologically. The cell was loaded with Oregon-green 488 BAPTA-1 in the CA1 region of a living hippocampal slice. Scale bar, 10 μm. (B) and (C) Outward currents recorded from an NG2 cell evoked by step voltages (100 ms, 10 mV) from –80 to +50 mV with a prepulse of –120 mV (B) or –30 mV (C) for 300 ms. In (B), the black circle indicates + the position of peak currents as measured in (J). (D) Isolated IA reached by subtracting (C) from (B). (E) Transient inward currents recorded from an NG2 cell induced by the same voltage pulses as in (B) in the presence of K+ channel blockers with or without TTX. (F) and (G) Voltage-dependent Ca2+ currents were recorded in neurons but not in NG2 cells. In (G), error bars for neurons (solid circles, n = 8 cells) and for NG2 cells (open circles, n = 7 cells) indicate SEM. (H) Morphology of a typical astrocyte identified electrophysiologically. The cell was loaded and imaged as in (A). White arrow indicates the end-feet of the astrocyte enwrapping a microvessel. Red asterisk indicates the dye-containing pipette. Scale bar, 10 μm. (I) Current responses from an astrocyte evoked by step voltages (100 ms, 20 mV) from –160 to +20 mV. Red circle indicates the position of peak currents as measured in (J). (J) Example current-voltage relation curves constructed from recordings as shown in (B) (peak current) and (I).

Electrical stimulation of SCs, in the presence of the GABA type A (GABAA) receptor antagonist bicuculline, elicited inward currents in NG2 cells with rapid kinetics (Fig. 2A; rise time, 0.71 ± 0.06 ms; decay time, 2.13 ± 0.17 ms; n = 11), which was consistent with the existence of direct monosynaptic inputs made by SCs on NG2 cells (4). The excitatory postsynaptic currents (EPSCs) exhibited short-term plasticity similar to that induced at SC-CA1 pyramidal cell synapses, as suggested by the paired pulse ratio (PPR) of these two types of synapses over a broad range of paired pulse intervals (Fig. 2B). The EPSCs were reversibly abolished by the application of kynurenic acid (Kyn) (Fig. 2D and fig. S1E), a broad-spectrum blocker of ionotropic glutamate receptors. These EPSCs were due to the activation of non-N-methyl-d-aspartate receptors (NMDARs), because the antagonist 6,7-dinitroquinoxaline-2,3-dione (DNQX) completely abolished the responses, whereas the NMDARs antagonist d,l-2-amino-5-phosphonovaleric acid (APV) had no effect on these currents recorded at +40 mV to remove the Mg2+ block of NMDAR (Fig. 2, C and D).

Fig. 2.

Properties of EPSCs and glutamate-induced responses in NG2 cells. (A) Example traces showing SC stimulation-induced fast response in an NG2 cell (bottom) and slow response in an astrocyte (top). (B) PPRs in NG2 cells and CA1 pyramidal neurons induced by SC stimulations at various interstimulus intervals (n = 7 pairs). Insets represent example recordings from an NG2 cell. (C) Inhibition of EPSCs in an NG2 cell by DNQX (10 μM) but not by APV (100 μM). Insets from left to right represent sample traces (average of five traces) from control, perfusion with APV, and APV plus DNQX, respectively. (D) Summary data from experiments as shown in (C). Each column represents averaged data of EPSCs recorded during 5-min perfusion (or washout) of antagonists. Data are normalized with the control (averaged EPSC amplitude during initial 5-min recording before drug application). *, P < 0.05. (E) Averaged rectification of SC-NG2 EPSCs from hippocampal slices at different development stages. Insets show example EPSCs recorded at +40 mV (red) and –60 mV (black). * (P < 0.01) indicates significance as compared with P8–10 group. (F) Averaged EPSCs in the presence of NAS. Data are normalized with control EPSCs (before NAS application). Significance as compared with the control group (**P < 0.001) or between the two groups (*P < 0.05) is indicated. (G) and (H) Imaging of glutamate-induced Ca2+ elevation in NG2 cells. Left panels depict NG2 cells filled with Rhod-2 through a patch pipette (red asterisks). Right panels provide example traces for glutamate-induced Ca2+ signal (color traces) and current recording (black traces) in the absence or presence of DNQX (G) and NAS (H). Traces in different colors in the right panel correspond to the imaged regions in the left panel marked by squares of the same color. Black bars indicate perfusion of 500 μM glutamate (Glu). (I) Summarized data of glutamate-induced current and Ca2+ fluorescent change (ΔF/F; see supporting online material) and current as shown in (G) and (H). * (P < 0.01) and ** (P < 0.001) indicate significance as compared with the control group.

Most AMPARs in the CNS are Ca2+-impermeable because of the presence of the GluR2 subunit, although GluR2-lacking AMPARs have been observed in some developing and mature neurons as well as in glial cells (4, 8, 9). Consistent with the previous finding that NG2 cells contain Ca2+-permeable AMPA receptors (CaPARs) (4), we found that EPSCs observed in NG2 cells were significantly reduced by philanthotoxin-433 (PhTx), a toxin that specifically blocks the CaPARs (10) (Fig. 2D). Currents mediated by CaPARs display inward rectification, exhibiting a reduction of outward currents at depolarizing membrane potentials because of a voltage-dependent block by intracellular polyamines (8, 9). To estimate the proportion of the EPSC component mediated by CaPARs, we examined the rectification of EPSCs by monitoring the ratio of EPSCs recorded at –60 mV to that at +40 mV (I–60/I+40). The rectification of EPSCs was stronger in 1-week-old rats [(postnatal day 8 (P8) to P10] than that in rats at P13 to P15 and P19 to P21. Furthermore, the proportion of EPSCs inhibited by 30 to 50 μM 1-naphthylacetyl spermine (NAS), a synthetic analog of Joro spider toxin that selectively blocks CaPARs (11), was also larger in young animals, suggesting a developmental reduction in the EPSC mediated by CaPARs (Fig. 2, E and F). To directly estimate the glutamate-induced Ca2+ influx through CaPARs, NG2 cells were loaded with the Ca2+ dye Rhod-2, and the hippocampal slices were perfused with glutamate in the presence of cyclothiazide, which blocks the desensitization of AMPARs. We found that glutamate-induced currents and Ca2+ elevations were largely abolished by DNQX and significantly reduced by NAS. Thus, a large proportion of glutamate-induced Ca2+ elevation in NG2 cells is mediated by CaPARs (Fig. 2, G to I). The residual Ca2+ elevation in the presence of NAS may be caused indirectly by the activation of (i) Ca2+-impermeable AMPARs that may induce Ca2+ elevation through a Na+-Ca2+ exchanger and (ii) metabotropic glutamate receptors that mobilize intracellular Ca2+ stores. Because the perfusion of glutamate activates both synaptic and extrasynaptic AMPARs, the finding that NAS caused stronger inhibition of glutamate-induced current (Fig. 2I) than of SC-induced EPSCs (Fig. 2F) suggests that CaPARs are more prevalent in extrasynaptic AMPARs. In some experiments, Ca2+ elevation in NG2 cells induced by the stimulation of SCs was also detected by line scanning with a high temporal resolution (fig. S2).

Evoked EPSCs in NG2 cells were compared before and after tetanic or theta burst stimulation (TBS) of the SCs. Although tetanic stimulation failed to induce an apparent change in EPSCs (Fig. 3H and fig. S3C), TBS resulted in a persistent increase in the EPSC amplitude in NG2 cells (Fig. 3, A and B), analogous to LTP found at neuronal synapses. The effectiveness of TBS in LTP induction at NG2 synapses may be attributed to the fact that these EPSCs exhibited substantial amplitude reduction during tetanic stimulation but not during TBS (fig. S3), which may result in a higher amount of Ca2+ influx during TBS than during tetanic stimulation. Bath application of APV, which largely abolishes CA1 neuronal LTP (1214), did not affect TBS-induced “glial LTP” (gLTP) of SC-NG2 cell synapses (Fig. 3, C, G, and H). In contrast, application of TBS during a brief period of perfusion with Kyn resulted in no change in EPSCs in NG2 cells (Fig. 3, E, G, and H), suggesting that TBS-induced gLTP requires the activation of nonNMDA receptors.

Fig. 3.

Persistent enhancement of EPSCs in NG2 cells induced by TBS of SCs. Example (A) and summarized (B) data showing the TBS (arrows)–induced gLTP at NG2 cell synapses are shown. Insets in (A) are sample traces before and after TBS, respectively. (C to F) Summary data showing the time course of EPSCs before and after TBS in the presence of 100 μM APV (C), 0.5 mM Kyn (E) applied during the period indicated by the bar, 4 μM PhTx (F), or 10 mM intracellular BAPTA in NG2 cells (D). (G) Cumulative distribution of EPSC amplitudes after TBS as shown in (B) to (F). Data are normalized with the average EPSC amplitude (vertical dotted line) before TBS for each recording. (H) Summary of results showing averaged percent changes in EPSC amplitudes estimated during the period from 15 to 20 min after TBS or tetanic stimulation. * (P < 0.05) and ** (P < 0.01) indicate significant changes in EPSC amplitude after induction.

Intracellular Ca2+ elevation is critical for LTP induction at many synapses (13). We found that TBS failed to induce gLTP when NG2 cells were loaded with a fast Ca2+ chelator [1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid (BAPTA)]. A persistent reduction in the EPSC amplitude was observed instead (Fig. 3, D, G, and H). This reduction may result from an incomplete Ca2+ buffering in the fine processes of NG2 cells where SC-NG2 synapses are located, suggesting potential bidirectional Ca2+-dependent plasticity (13) at these synapses. Given the lack of voltage-gated Ca2+ current in these NG2 cells (Fig. 1, F and G), we further examined whether CaPARs are responsible for the Ca2+ influx required for gLTP induction. Application of PhTx completely prevented the induction of gLTP by TBS (Fig. 3, F to H). This gLTP is likely to be expressed as an increased glutamate response in NG2 cells, because PPRs were not significantly changed after gLTP induction (Fig. 4, A and B), suggesting no presynaptic change in glutamate release probability (13).

Fig. 4.

Comparing EPSC properties in an NG2 cell before and after TBS. (A) Example EPSC recordings from an NG2 cell showing similar PPR (10 Hz) recorded before and after TBS application. (B) Summary of averaged PPR at 10 Hz (n = 5 cells) and 20 Hz (n = 4 cells) before and after TBS induction. (C and D) Summary of the rectification of EPSCs examined before and after TBS induction in NG2 cells loaded with [(D) n = 7] or without [(C) n = 7] 0.1 mM spermine. Insets, example EPSCs from an NG2 cell recorded at +40 and –60 mV, respectively, before and after TBS induction. * (P < 0.05) and ** (P < 0.01) indicate significance between the control and post-TBS groups.

The trafficking of GluR1-containing AMPARs plays a crucial role in the expression of NMDAR-dependent LTP (15, 16). Induction of NMDAR-dependent LTP in CA1 neurons, which do not express CaPARs, did not change the rectification of AMPAR-mediated EPSCs unless neurons were overexpressed with green fluorescent protein-GluR1 (16); this finding suggests that, after LTP induction, the endogenous GluR1 subunits are trafficked to synapses in heteromeric forms containing GluR2, so that the increased AMPARs are also Ca2+-impermeable. Activation of CaPARs by tetanic stimulation at cerebellar stellate cell synapses results in a prolonged decrease in the rectification of AMPAR-mediated EPSCs due to the replacement of CaPARs with Ca2+-impermeable receptors (17, 18), leading to a decreased Ca2+ influx and EPSC amplitude recorded at the physiological membrane potential (–60 mV), a plasticity that may resemble the long-term depression and represent a protective mechanism. Synaptic strengthening induced by the activation of the CaPARs has been reported in some dorsal horn neurons (19) and amygdala interneurons (20), but it is not clear whether the proportion of the CaPAR-mediated EPSCs is changed after LTP induction in these synapses. We found that TBS induced an increased rectification of EPSCs (Fig. 4C), suggesting an increased function or proportion of CaPARs at these synapses. Loading NG2 cells with spermine caused more rectification of SC-induced EPSCs (Fig. 4, compare C with D), suggesting that endogenous polyamines are sufficient (21) but not saturating in inducing inward rectification of AMPAR currents. Furthermore, the induction of gLTP in NG2 cells loaded with spermine also resulted in an increased rectification of EPSCs (Fig. 4D). Such synaptic modification will increase not only the amplitude of synaptic responses but also Ca2+ influx and subsequent activation of Ca2+-dependent intracellular signaling. The intracellular C terminal of the proteoglycan NG2 has been reported to bind with the PDZ domain of GRIP1 (22), an adaptor protein important for AMPAR trafficking and synaptic plasticity (18). It is possible that the interaction of NG2 proteoglycan and GRIP1 in these glial cells played a role in the trafficking of CaPARs found in the present study.

The conversion of CaPARs into Ca2+-impermeable receptors at Bergmann glial cells, by transfection with the GluR2 subunit, results in the retraction of glial processes that ensheath synapses and multiple innervations of Purkinje cells by climbing fibers (23). Pathological insults such as ischemia also down-regulate GluR2 expression in OPCs and neurons (24), leading to the expression of CaPARs and enhanced glutamate toxicity after ischemia. Thus, an adequate level of activity by CaPARs is important for maintaining normal synaptic structure and function. Glutamate receptor activation is linked to the proliferation and differentiation of OPCs (25), and NG2 cells can differentiate into neurons both in vitro and in vivo (26, 27). The expression of gLTP and the elevation of CaPARs in SC-NG2 synapses may thus contribute to activity-dependent neuronal regulation of NG2 cell differentiation. Like astrocytes (12, 28, 29), NG2 cells may also secrete neuroactive factors to regulate neuronal functions. Rapid neuron-NG2 cell signaling may allow rapid feedback regulation of neuronal functions by Ca2+-dependent secretion of neuroactive factors, and the strength of such feedback regulation will increase after the induction of gLTP.

Supporting Online Material

Materials and Methods

Figs. S1 to S3

References and Notes

Stay Connected to Science

Navigate This Article