Is neuronal communication with NG2 cells synaptic or extrasynaptic?

Abstract

NG2-expressing glial cells (NG2 cells) represent a major pool of progenitors able to generate myelinating oligodendrocytes, and perhaps astrocytes and neurones, in the postnatal brain. In the last decade, it has been demonstrated that NG2 cells receive functional glutamatergic and GABAergic synapses mediating fast synaptic transmission in different brain regions. However, several controversies exist in this field. While two classes of NG2 cells have been defined by the presence or absence of Na⁺ channels, action potential firing and neuronal input, other studies suggest that all NG2 cells possess Na⁺ conductances and are the target of quantal neuronal release, but are unable to trigger action potential firing. Here we bring new evidence supporting the idea that the level of expression of Na⁺ conductances is not a criterion to discriminate NG2 cell subpopulations in the somatosensory cortex. Surprisingly, recent reports demonstrated that NG2 cells detect quantal glutamate release from unmyelinated axons in white matter regions. Yet, it is difficult from these studies to establish whether axonal vesicular release in white matter occurs at genuine synaptic junctions or at ectopic release sites. In addition, we recently reported a new mode of extrasynaptic communication between neurones and NG2 cells that relies on pure GABA spillover and does not require GABAergic synaptic input. This review discusses the properties of quantal neuronal release onto NG2 cells and gives an extended overview of potential extrasynaptic modes of transmission, from ectopic to diffuse volume transmission, between neurones and NG2 cells in the brain.

Introduction

Glial cells express a large set of metabotropic and ionotropic receptors for neurotransmitters that allow them to respond to neuronal activity. Neurotransmitter receptor activation, extensively studied in astrocytes, leads to complex information processing in the form of calcium signals (Pasti et al. 1997; Perea & Araque, 2005; Wang et al. 2006; Schummers et al. 2008). In the absence of evidence for anatomical synaptic contacts between neurones and astrocytes, the extrasynaptic action of neurotransmitters spilling out of synaptic cleft is likely the major neurone-to-astrocyte signalling mechanism. Another mode of extrasynaptic communication exists between climbing fibres and Bergmann glia in the cerebellum. Non-synaptic quantal events mediated by α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPA) receptors in Bergmann glia result from an ectopic release of glutamate from climbing fibres outside active zones (Matsui & Jahr, 2003; Matsui et al. 2005). Spillover and ectopic release are thus two modes of communication between neurones and glia implicating extrasynaptic receptors.

Although chemical synapses have long been regarded as a signalling mechanism exclusive to neurones in the brain, Bergles et al. (2000) demonstrated the presence of functional and morphological synaptic contacts between neurones and NG2-expressing glial cells (NG2 cells) in the postnatal hippocampus. Since this pioneer work, fast glutamatergic and/or GABAergic synaptic currents have been observed in different regions of grey matter, including neocortex (Chittajallu et al. 2004; Kukley et al. 2008; Ge et al. 2009; Velez-Fort et al. 2010), hippocampus (Lin & Bergles, 2004; Jabs et al. 2005; Ge et al. 2006), cerebellum (Lin et al. 2005) and medial nucleus of the trapezoid body (Muller et al. 2009). Surprisingly, it has also been established that NG2 cells detect quantal neurotransmitter release from unmyelinated axons in white matter regions such as corpus callosum (Kukley et al. 2007; Ziskin et al. 2007), optic nerve (Kukley et al. 2007) and white matter cerebellum (Karadottir et al. 2005, 2008). Whether vesicular release in white matter occurs at genuine synaptic junctions or at ectopic release sites is still controversial.

These cells constitute the main endogenous pool of progenitors in the postnatal brain, which serves as a major source of myelinating oligodendrocytes during postnatal development, and perhaps as progenitors endowed with the ability to generate astrocytes and neurones. In addition, they play a crucial role during remyelination which occurs as a repair process in demyelinating lesions (Nishiyama et al. 2009). A tempting hypothesis on the role of synaptic transmission in NG2 cells is that this point-to-point communication allows neurones to control NG2 cell activity precisely, influencing oligodendrogenesis and thus myelination of the brain. In support of this hypothesis, recent findings have shown that the transition of NG2 cells towards a premyelinating stage is accompanied by a rapid loss of glutamatergic synaptic inputs under normal conditions (De Biase et al. 2010; Kukley et al. 2010) and after demyelinating lesions (Etxeberria et al. 2010). In addition, the frequency of spontaneous GABAergic synaptic activity dramatically decreases in NG2 cells of the barrel cortex after the second postnatal week (Velez-Fort et al. 2010), when cortical oligodendrocyte differentiation reaches a peak (Baracskay et al. 2002). However, the millisecond time scale of fast synaptic signals and the low frequency of synaptic events in NG2 cells suggest that this synaptic communication is more suitable for other cellular functions requiring precise spatiotemporal signals. Among these functions, recognition of specific axonal partners, process motility and promotion of migration are good candidates.

Although AMPA and GABA_A receptors of NG2 cells are believed to operate mainly through conventional synaptic transmission, these and other receptors expressed in the membrane of these cells may also be activated by extrasynaptic neurotransmitters. In tissue preparations, NG2 cells express different functional receptors such as NMDA, kainate, nicotinic and purinergic receptors that are probably activated extrasynaptically and may be powerful transducers of neuronal signals (Karadottir et al. 2005; Kukley & Dietrich, 2009; Velez-Fort et al. 2009; Hamilton et al. 2010). Also, we recently demonstrated that direct quantal release of GABA on NG2 cells is lost early in postnatal cortical development and replaced by an unusual extrasynaptic mode of communication, mediated solely by GABA spillover (Velez-Fort et al. 2010). In this review we discuss the evidence for synaptic neurone-to-NG2 cell signalling and evaluate the possible existence of multiple modes of extrasynaptic communication from ectopic to diffuse volume transmission between neurones and NG2 cells.

Synaptic transmission in grey matter NG2 cells

Evidence for direct synaptic contacts on NG2 cells has been the subject of recent reviews and we will thus succinctly describe the physiological and morphological properties of these synapses in this section (see Gallo et al. 2008; Nishiyama et al. 2009; Bergles et al. 2010). Patch-clamp recordings in NG2 cells of the hippocampus revealed the presence of spontaneous rapid AMPA and GABA_A receptor-mediated currents (Bergles et al. 2000; Lin & Bergles, 2004; Jabs et al. 2005). In grey matter regions, these currents exhibit the typical characteristics of neuronal synaptic events, i.e. submillisecond rise times and decays with time constants of few milliseconds (Bergles et al. 2000). We should take into account, however, that spontaneous synaptic events occur at frequencies 30–97% lower than those recorded in neurones of the same region (Mangin et al. 2008; Muller et al. 2009). In the presence of the Na⁺ channel blocker tetrodotoxin (TTX), miniature glutamatergic and GABAergic events are also detected at very low frequencies of between < 0.01 and 0.03 Hz (Bergles et al. 2000; Velez-Fort et al. 2010). These frequencies can be enhanced when potent secretagogues such as α-latrotoxin and ruthenium red are added in the bath, demonstrating that miniature currents result from the spontaneous fusion of transmitter-filled vesicles (Bergles et al. 2000; Lin & Bergles, 2004). In addition, neuronal fibre stimulation evoked glutamatergic responses characterized by short onset latencies, variable amplitudes and fast kinetics, as expected for a synchronized vesicular release. It is noteworthy, however, that evoked GABAergic currents do not always have kinetics compatible with a synaptic release, as discussed below (Tanaka et al. 2009; Velez-Fort et al. 2010).

Physiological data have been corroborated by ultrastructural analyses revealing the presence of morphological specialized junctions between presynaptic neuronal terminals and NG2 cells (Fig. 1A; Bergles et al. 2000; Lin & Bergles, 2004; Lin et al. 2005; Jabs et al. 2005; Muller et al. 2009; Kukley et al. 2008). Indeed, vesicle-containing terminals are apposed to NG2 cell processes and, although not always detectable in biocytin-stained cells, postsynaptic densities are also clearly recognized by their electron-dense material in immunogold experiments (Bergles et al. 2000). In electron micrographs, a single neuronal bouton often appears to contact both a postsynaptic spine and an NG2 cell (Fig. 1A, right; Bergles et al. 2000; Lin et al. 2005; Muller et al. 2009). This morphological feature correlates with the occurrence of highly synchronized spontaneous glutamatergic synaptic currents between neurones and NG2 cells of the hippocampus and of the medial nucleus of trapezoid body (Mangin et al. 2008; Muller et al. 2009). Whether NG2 cells and neurones show synchronous GABAergic activity is unknown at present. Only a restricted description of functional and morphological GABAergic synaptic properties of NG2 cells exists (Lin & Bergles, 2004; Jabs et al. 2005; Karadottir et al. 2008; Velez-Fort et al. 2010). The GABAergic system of NG2 cells probably constitutes an interesting subject for future investigation.

Fig. 1.

Examples of different potential modes of transmission between neurones and NG2 cells. (A) Synaptic transmission. Vesicle-containing presynaptic compartments directly contact an NG2 cell process forming specialized synaptic junctions similar to those described in neurones (left). Released neurotransmitters diffuse across a narrow cleft to activate rapidly high densities of postsynaptic receptors. A single presynaptic bouton can also innervate simultaneously a neuronal spine and an NG2 cell process (right). (B) Ectopic transmission. Neurotransmitters could be found in small ‘synaptic-like’ vesicles located at varicosities along axons and released at non-synaptic sites. High densities of extrasynaptic receptors would be activated close to these release sites to induce fast rising quantal events in NG2 cells. These events, however, would have smaller amplitudes than those recorded at a genuine synapse. (C) Local spillover transmission. NG2 cell processes could enwrap or pass very close to neuronal synapses, allowing high densities of extrasynaptic receptors to sense locally neurotransmitters spilling out of the synaptic cleft. No miniature events would be detectable in this case. (D) Diffuse volume transmission. Diffuse transmission may occur at non-synaptic sites and, compared to other types of transmission modes, released neurotransmitters would have to cross large distances (typically > 1 μm) before reaching a targeted cell. Other modes of transmission are not excluded.

Quantal events in white matter NG2 cells: synaptic or ectopic?

In white matter regions, quantal release of glutamate on NG2 cells has been described recently during postnatal development and after lysolecithin-induced demyelination in adult animals (Karadottir et al. 2005, 2008; Kukley et al. 2007; Ziskin et al. 2007; Etxeberria et al. 2010). This unexpected discovery supports the idea that white matter in the brain is not only a passive region where bundles of axons connect different brain regions, but a place where local information is precisely transferred at specialized sites from neurones to glia. Even so, a certain number of discrepancies exist regarding the physiological and morphological properties of axonal–NG2 cell junctions.

As for classical synaptic events, whole-cell currents of white matter NG2 cells are characterized by rise times in the submillisecond range and decay time constants of few milliseconds. Ca²⁺-impermeable AMPA receptors mediate axonal–NG2 cell currents in the first postnatal weeks (Kukley et al. 2007; Ziskin et al. 2007), whereas Ca²⁺-permeable receptors predominate in adult mice (Ziskin et al. 2007). Therefore, modifications of the subunit composition of AMPA receptors expressed at the membrane of NG2 cells probably occur during development. After lesion, immunostainings against GluR2/3 subunits confirmed the expression of AMPA receptors in migrating NG2 cells from the subventricular zone (Etxeberria et al. 2010). However, it was not possible, on the basis of these experiments, to establish the presence or absence of GluR2 subunits and thus to predict the Ca²⁺-permeability of receptors during a demyelination/remyelination process. As Ca²⁺ could be an important intracellular transducer of neuronal signals in NG2 cells, it would be interesting to determine which AMPA receptor subtype predominates during remyelination of callosal axons.

Unmyelinated axons constitute the major source of axonal projections to white matter NG2 cells and their mechanisms of glutamate release exhibit many hallmarks of classical synaptic transmission, i.e. the presence of voltage-gated calcium channels as key mediators of Ca²⁺ entry into the axon, the existence of discrete Ca²⁺ microdomains, a highly sustained synchronous release of transmitter during intense stimulation and the presence of at least part of the machinery for synaptic vesicle exocytosis (Kukley et al. 2007; Ziskin et al. 2007; see also Alix et al. 2008 for the machinery). However, whether glutamate release in white matter regions occurs at true synaptic sites is still unclear. Ultrastructural analysis reported by Ziskin et al. (2007) in corpus callosum showed NG2 cell processes opposed to nerve terminals containing VGLUT1⁺ vesicles. The presence of discrete junctions characterized by a rigid parallel apposition of membranes and electron-dense material strongly suggests the existence of genuine synaptic contacts between axons and callosal NG2 cells (Fig. 1A, left; Ziskin et al. 2007). In contrast, in Kukley et al. (2007), electron microscopy analysis in corpus callosum and optic nerve revealed the presence of axonal varicosities in close contact with an NG2 cell process containing very few to a large number of small vesicles, showing clefts with irregular and relatively wide sizes and lacking electron-dense material. In addition, most putative vesicle release sites were not opposed to an NG2 cell membrane, suggesting that these structures are not specifically associated with NG2 cells. Another substantial discrepancy exists regarding the amplitudes of miniature currents and the estimated number of axonal fibres that contact a single callosal NG2 cell: −18 pA at a holding potential of −90 mV and ∼ 10 contacts in Ziskin et al. (2007) vs. −4 pA at a holding potential of −80 mV and hundreds of contacts in Kukley et al. (2007). Whereas data reported by Ziskin et al. (2007) are in favour of real glutamatergic synapses between axons and NG2 cells (Fig. 1A), those by Kukley et al. (2007) suggest the existence of an ectopic release of glutamate in white matter (Fig. 1B). Ectopic quantal release of glutamate outside active zones has already been described between neurones and Bergman glia of the cerebellum and thus exists between neurones and glial cells (Matsui & Jahr, 2003). If ectopic quantal release is the main mechanism governing axonal–NG2 cell communication in white matter, we can expect that its properties differ from those described for synapses of NG2 cells and neurones in grey matter. In keeping with this assumption, the inhibition of the amplitude of evoked currents by the low-affinity competitive antagonist of AMPA receptors γ-d-glutamylglycine (γ-DGG) was significantly greater in callosal NG2 cells than in CA1 pyramidal neurones as expected for a lower concentration of glutamate reaching NG2 cell receptors (Kukley et al. 2007). A similar observation was reported for ectopic release onto Bergman glia (Matsui & Jahr, 2003). It is noteworthy, however, that the effect of γ-DGG may be different between callosal NG2 cells and pyramidal neurones because of a different subunit AMPA receptor composition.

Quantal neurotransmitter release directly on NG2 cells seems to be a usual signalling pathway in different regions of the brain. However, quantal transmission on these cells may exist in two distinct forms according to the brain region: classical synaptic transmission in grey matter (Fig. 1A) and ectopic transmission in white matter (Fig. 1B). Substantial differences between ectopic and synaptic release mechanisms have been described in Bergman glia of the cerebellum and may be used, at least partly, for a comparative evaluation (Matsui & Jahr, 2004). Further detailed and comparative physiological analyses are thus needed to clarify the synaptic or extrasynaptic nature of neurotransmitter release from callosal axons. A revised axonal–NG2 and extended analysis of axonal–NG2 cell contact morphology should also help to resolve the controversy in white matter.

Which classes of NG2 cells respond to quantal vesicular release?

Most studies have established that virtually all NG2 cells in mice and rats receive glutamatergic input after the first postnatal week and in demyelinating lesions, suggesting that the presence of this input on NG2 cells is a general rule (Bergles et al. 2010; Etxeberria et al. 2010). In grey matter, the acquisition of synapses occurs during cell division, thus very early in NG2 cell development when cells are actively proliferating in the brain (Kukley et al. 2008; Ge et al. 2009). However, Karadottir et al. (2008) reported the existence of two classes of NG2 cells in the cerebellar white matter, defined by the presence or absence of Na⁺ channels; only those expressing high levels of these channels are supposed to trigger action potential discharges and receive axonal input. Yet, the existence of two distinct subpopulations of NG2 cells, recognized by the levels of Na⁺ channel expression, was contradicted in recent reports arguing that Karadottir et al. (2008) recorded from pre-oligodendrocytes, which progressively, but rapidly, lose both their functional Na⁺ channels and neuronal inputs (De Biase et al. 2010; Kukley et al. 2010). The capacity of NG2 cells to elicit action potential discharges is also under intense debate.

If the question of which NG2 cells receive glutamatergic inputs in the brain is controversial, even less is known about NG2 cells contacted by GABAergic synapses. We recently reported that layer V NG2 cells in the barrel cortex receive GABAergic synaptic contacts in the second, but not in the fourth, postnatal week (Velez-Fort et al. 2010). To evaluate whether two distinct cortical NG2 cells can be identified during the period of loss of synaptic GABAergic activity, we studied the level of expression of Na⁺ conductances at these two developmental stages. Whole-cell recordings of DsRed⁺ NG2 cells were performed at a holding potential of −90 mV with a CsCl-based intracellular solution containing 4AP and tetraethylammonium (TEA) to inhibit voltage-gated K⁺ channels and to better isolate Na⁺ conductances (Fig. 2). Current responses elicited by voltage steps from −60 to +40 mV revealed the presence of an Na⁺ conductance in 95% of recorded cells in the second postnatal week, but only in 40% in the fourth postnatal week (n = 57 and n = 67, respectively; Fig. 2A,C). However, we noticed that the loss of Na⁺ currents in the fourth postnatal week was concomitant with a decrease of cell membrane resistance and an increase of a passive conductance expression that may mask Na⁺ currents (Fig. 2A; Velez-Fort et al. 2010). An effective way to increase the membrane resistance in these cells is by adding extracellular Ba²⁺ in the perfusate (Velez-Fort et al. 2009). The presence of extracellular Ba²⁺ unmasked a Na⁺ current in almost all tested cells (Fig. 2B,C). In addition, no significant changes in the Na⁺ current density were observed between the second and fourth postnatal week in the presence of extracellular Ba²⁺, confirming that Na⁺ channel expression did not decrease with age (Fig. 2D). Only two of 17 cells recorded during the fourth postnatal week did not express an Na⁺ current (Fig. 2C). To discard the possibility that DsRed is preferentially expressed in a subpopulation of cortical NG2 cells having Na⁺ currents, we performed immunostainings against NG2 in perfused brains. We observed that virtually all NG2⁺ cells of the cortex express DsRed, whereas most but not all of DsRed⁺ cells were NG2⁺ (100% and 73.4 ± 3.5%, respectively; n = 2 animals; Fig. 3A). It is also noteworthy that DsRed⁺/NG2⁻ cells expressed low levels of fluorescence (Fig. 3A, inset), suggesting that these cells are in a more mature state, as reported by Ziskin et al. (2007). It is thus likely that the two cells of our sample lacking Na⁺ currents in Ba²⁺ were in a transition state towards a more mature phenotype (Fig. 2C). To further test this hypothesis, we performed post-recording immunostainings against NG2 in DsRed⁺ cells stained with biocytin during the fourth postnatal week. In these experiments, no Ba²⁺ was applied in the bath to preserve the quality of the slice. We observed that three of four DsRed⁺ cells were positive for NG2 (not shown). However, one DsRed⁺ cell with no detectable Na⁺ currents after leak subtraction was NG2^– (Fig. 3B). Taken together, our results indicate that the large majority, if not all, of NG2 cells of the somatosensory cortex express voltage-gated Na⁺ channels regardless the age of the animal and the presence of GABAergic synaptic inputs. However, we cannot completely exclude that a small proportion of NG2 cells lacking Na⁺ currents exist in other brain regions.

Fig. 2.

NG2 cells of the barrel cortex express high levels of voltage-gated Na⁺ channels, but do not elicit action potential discharges. (A) Currents induced by depolarizing steps from −60 to +40 mV in layer V cortical NG2 cells held at −90 mV and recorded in a CsCl-based intracellular solution containing TEA (10 mm) and 4AP (4 mm) at PN6 (A1, left) and PN25 (A2, left). Corresponding I–V curves (right) were obtained after leak subtraction. Note that in our recording conditions the reversal potential of Na⁺ was +97 mV. Inward sodium currents () at PN6 were observed at negative potentials and unmasked at positive potentials by leak subtraction (A1). Conversely, was not apparent using this procedure at PN25 (A2). (B) Recordings from a cortical NG2 cell during a voltage step from −90 to −20 mV at PN25, before (black trace) and after bath applications of 1 mm Ba²⁺ (grey trace). Note that extracellular Ba²⁺ revealed . (C) Histograms showing the percentage of cortical NG2 cells displaying voltage-gated Na⁺ currents in the second (black) and fourth (grey) PN weeks, in control conditions and after bath application of Ba²⁺ (100 μm–1 mm). Note the large increase of the percentage of NG2 cells showing in the fourth postnatal week. (D) Comparison of I–V relationships of Na⁺ conductances recorded at different voltage steps from −60 to +40 mV in the second (black) and fourth (grey) postnatal weeks. Note that the amplitudes of are similar between both postnatal stages, except at −40 mV (Mann–Whitney test, *P < 0,01). (E) Whole-cell current clamp recordings from cortical NG2 cells held at −70 mV at PN11 (E1) and PN25 (E2) with a KCl-based intracellular solution, during successive current injections from 200 pA. Less immature spikes were elicited in the fourth postnatal week (E3).

Fig. 3.

Immunoreactivity of NG2 proteoglycan in NG2-DsRed transgenic mice in the fourth postnatal week. (A) Confocal images of NG2 in the somatosensory cortex of a perfused PN26 NG2-DsRed mouse (stack of seven z sections, each 2 μm). Note that all NG2⁺ cells (green) are DsRed⁺ (red) but that not all DsRed⁺ cells are NG2⁺. A DsRed⁺/NG2⁺ cell and a DsRed⁺/NG2⁻ cell are outlined in the white frame. These two cells are shown at higher magnification in the insets (lower panels). (B) Post-recording immunostaining of a DsRed⁺ cell stained with biocytin at PN26. (B1) Currents induced by depolarizing steps from −60 to +40 mV in a DsRed⁺ cell held at −90 mV and recorded in a CsCl-based intracellular solution containing 5.4 mm biocytin. Note the lack of Na⁺ currents (this is also the case after leak subtraction). (B2) Confocal images of the same DsRed⁺ cell after fixing the slice in 4% paraformaldehyde and immunostaining with both streptavidin-conjugated-Alexa488 (green) and NG2 revealed with a secondary antibody coupled to Alexa633 (red; stacks of 21 z sections, each 1 μm). In our confocal recording conditions, the DsRed is not visible. Note the lack of NG2 expression of the cell (arrowheads). Arrows show NG2⁺ cells in the same slice.

To test for the ability of NG2 cells to fire action potentials, we recorded the response to injections of depolarizing currents in a KCl-based intracellular solution. A single immature spike was elicited in a large proportion of NG2 cells; no discharge of action potentials in these cells was observed (Fig. 2E). When present, spikes were rudimentary in nature and resembled those described by others in NG2 cells of the neocortex (Chittajallu et al. 2004; Ge et al. 2009), i.e. they had a depolarized threshold potential, most often grew in size as the depolarization increased, and showed a small amplitude (Fig. 2E1, inset). In agreement with the apparent (but false) decrease of Na⁺ currents during development, fewer cells showed immature spikes in the fourth than in the second postnatal week (n = 11 and n = 24 recorded cells, respectively, Fig. 2E). We can thus conclude that different classes of NG2 cells cannot be distinguished by the level of expression of Na⁺ channels in the somatosensory cortex and that, even if Na⁺ current amplitudes can reach 1 nA (mean: −600 ± 62 pA at −20 mV; from −133 pA to 1.17 nA; n = 25), NG2 cells are not able to elicit single or trains of genuine action potentials. This is in agreement with recent reports in other brain regions and using different transgenic mouse strains (De Biase et al. 2010; Etxeberria et al. 2010; Kukley et al. 2010). We could still argue that different electrophysiological properties reported for NG2 cells might be caused by the fact that observations come from different experimental models, i.e. transgenic mice vs. non-transgenic rats. Indeed, Clarke et al. (2010) recently showed that the mean size of this current in NG2 cells of rats is −0.9 nA and about 4.5-fold larger than in mouse NG2 cells. Yet, Kukley et al. (2007) reported in the same species, a mean Na⁺ current amplitude of −0.3 nA. In addition, Chittajallu et al. (2004) showed in CNP-eGFP transgenic mice that the mean Na⁺ current amplitude of cells eliciting immature spikes in the neocortex is around −0.9 nA (in our case, one-third of cortical NG2 cells also reaches this amplitude regardless of the age of the animal in NG2-DsRed mice). Therefore, although different animals and brain regions may influence in some extent the size of Na⁺ currents in NG2 cells, discrepancies regarding its amplitude and the ability of these cells to generate genuine action potentials cannot be explained solely by differences in the experimental model. Further experiments in which post-recording immunostainings against NG2 or PDGFαR are performed in each recorded cell are probably needed. Nevertheless, it has to be considered that NG2 cells do not have the same electrophysiological properties in young and adult mice (Kressin et al. 1995; Zhou et al. 2006; Velez-Fort et al. 2010), which strengthens the idea that distinct types of NG2 cells, probably playing different roles, emerge during postnatal development. Indeed, it is more and more likely that NG2 cells constitute a heterogeneous group of cells comprising distinct subpopulations that vary according to the postnatal developmental stage and brain region (Butt et al. 2002; Mallon et al. 2002; Nishiyama et al. 2009). The electrophysiological criteria that allow for a reliable identification of different NG2 cell classes are still unclear.

Different modes of spillover transmission between neurones and NG2 cells

In the last years, most of the studies on neurone-NG2 cell interactions in the brain have been focused on understanding the direct quantal transmission of glutamate on NG2 cells, leaving aside other neuronal signalling mechanisms that might be important or even predominant. In that sense, spillover transmission not only is a prevailed mode of communication between neurones and astrocytes (see Introduction) but also operates among neurones influencing neuronal network functioning (Szapiro & Barbour, 2009). Here we evaluate the existence of different modes of spillover transmission between neurones and NG2 cells.

Recently, we studied the GABAergic activity of NG2 cells in the barrel cortex during the first postnatal month (Velez-Fort et al. 2010), a period in which physiological myelination occurs in deep cortical layers. By using whole-cell recordings and classical calcium imaging at different postnatal stages, we demonstrated that more than 90% of fast spontaneous synaptic currents recorded in cortical NG2 cells are GABAergic and sensitive to GABA_A receptor antagonists (Velez-Fort et al. 2010). Surprisingly, these functional synaptic inputs are completely lost after the second postnatal week, as revealed by the absence of spontaneous and miniature events. Despite the loss of functional GABAergic synapses, transient GABA_A-mediated currents are still detected in the adult upon low frequency stimulation of neuronal fibres. These evoked currents show very slow kinetics compared to those recorded in young animals, suggesting that they are not synaptic in the adult. Tanaka et al. (2009) recently reported similar slow transient GABAergic currents in nestin⁺/NG2⁺ cells of the adult neocortex, but the authors attributed these currents to genuine synaptic responses. Pharmacological experiments allowed us to establish that the concentration of GABA reaching GABA_A receptors in NG2 cells of older mice is lower than that reaching synaptic receptors. By using the voltage jump technique, we also showed that slow kinetics of GABAergic responses in NG2 cells do not result from signal filtering, confirming that evoked currents are purely extrasynaptic. GABAergic transmission in the adult thus relies on pure GABA spillover (Fig. 1C). Because of its unique properties, this form of spillover transmission between interneurones and NG2 cells can be considered a novel mode of neurone-glia communication that allows for a transient and relatively fast communication compared to other forms of volume transmission. Interestingly, recent reports have also described the existence of a similar mode of communication in neurones of cerebellum, neocortex and hippocampus (Szapiro & Barbour, 2007; Markwardt et al. 2009; Olah et al. 2009). Therefore, this form of communication is not exclusive for glia, but also exists between different neuronal types and brain regions. This spillover transmission, which does not require direct synaptic junctions, implies that GABA diffuses over relatively short distances from neuronal synapses. The anatomical modifications occurring at interneurone–NG2 cell junctions during postnatal cortical development remain unresolved. These changes could be complex if they involve a transition from a synaptic (Fig. 1A) to an extrasynaptic junction (Fig. 1C) and may implicate a profound reorganization of NG2 cell processes in the extracellular environment. Confocal and ultrastructural analysis will thus be necessary to unravel developmental morphological changes.

In addition to AMPA and GABA_A receptors, NG2 cells in tissue preparations are known to express other functional receptors on their membranes, such as NMDA, kainate, nicotinic and purinergic receptors (Karadottir et al. 2005; Kukley & Dietrich, 2009; Velez-Fort et al. 2009; Hamilton et al. 2010). As quantal currents in NG2 cells are mediated solely by either AMPA or GABA_A receptors, these receptors are most probably activated at a location distant from direct quantal release sites. A relevant example of extrasynaptic receptors is that of NMDA receptors in white matter regions reported to be activated by high extrasynaptic levels of glutamate, as those reached during ischaemic conditions (Karadottir et al. 2005). NMDA receptors of NG2 cells have an unusual subunit composition containing NR1, NR2C and NR3 that is weakly blocked by extracellular Mg²⁺ and probably allows glutamate to evoke NMDA receptor-mediated currents at resting potentials, contributing to tissue damage during ischaemia (Karadottir et al. 2005). Therefore, these receptors must be specialized to discern changes in the ambient extracellular glutamate concentration, especially when the uptake glutamate system is overwhelmed and pathologically high levels of glutamate are accumulated in the extracellular space. Kainate receptors are another case of glutamate receptors that are most likely expressed at extrasynaptic sites, as demonstrated in a recent report showing the functional expression of non-GluR5 kainate receptors in NG2 cells of the hippocampus (Kukley & Dietrich, 2009). NMDA and kainate receptors are therefore two glutamate receptor subtypes potentially activated by widespread diffusion of transmitters. However, the activation mode and the excitatory action of these receptors in NG2 cells under physiological conditions remain to be established.

Little is known about the expression and mode of activation of receptors for neurotransmitters other than glutamate and GABA in NG2 cells, but some examples exist in the literature. We recently demonstrated the functional expression of highly Ca²⁺-permeable α7 nicotinic receptors in NG2 cells of the hippocampus, suggesting the existence of cholinergic neurone-to-NG2 cell signalling in the grey matter (Velez-Fort et al. 2009). However, the stimulation of neuronal hippocampal cholinergic afferents, in the presence of the anticholinesterase agent eserine, failed to induce any synaptic nicotinic receptor-mediated current in NG2 cells. Although the desensitization or the low density of receptors may prevent the detection of synaptic currents, the most plausible activation mode of nicotinic receptors in NG2 cells is by an extrasynaptic release of transmitter (Fig. 1D). Indeed, it is believed that a majority of hippocampal cholinergic release sites are non-synaptic and contribute to diffuse volume transmission in neurones (Dani & Bertrand, 2007). In addition, the immunolabelling of acetylcholinetransferase, the enzyme responsible for the production of ACh, reveals that varicosities containing ACh in the hippocampus may be in front of glial cell processes (Descarries et al. 1997). NG2 cells in an isolated optic nerve preparation respond to adenosine triphosphate (ATP) by elevations of the intracellular Ca²⁺ concentration mediated by P2Y1 and P2X7 purinergic receptors (Wigley et al. 2007; Hamilton et al. 2010). The release of ATP is triggered by nerve stimulation and the amplitude and duration of Ca²⁺ signals in NG2 cells increase in parallel with increased stimulus strength, following neuronal activity (Hamilton et al. 2010). It is possible that nerve axons release ATP by a non-synaptic mechanism at sites where axonal membranes are in close contact with NG2 cell processes (Fig. 1D; Hamilton et al. 2010). However, as axonal action potentials stimulate the release of ATP from astrocytes and NG2 cells enwrap short segments of astrocytic processes in the optic nerve (Hamilton et al. 2008, 2010), astrocytes constitute another potential source of ATP generating NG2 cell signals. This leaves open the interesting possibility that astrocytes are implicated in a tripartite communication system with neurones and NG2 cells in the brain. Indeed, ATP, glutamate and GABA are three major gliotransmitters released by astrocytes (Angulo et al. 2008; Halassa & Haydon, 2010) that could activate extrasynaptic receptors of NG2 cells and influence their cell function. NMDA and kainate receptors are potential targets of glutamate released by astrocytes, as these glial cells have been reported to maintain the ambient extracellular glutamate concentration (Jabaudon et al. 1999; Cavelier & Attwell, 2005; Le Meur et al. 2007).

In addition to functional receptors reported in NG2 cells, the expression of mGluR5 metabotropic glutamate receptors in O4-positive cells of the rat corpus callosum has been demonstrated by immunocytochemical analysis (Luyt et al. 2003). Considering that almost 80% of NG2 cells are labelled by O4 in this region (Kukley et al. 2007), we can infer that NG2 cells express the mGluR5 protein. The α1A and α1B adrenergic receptors also co-localize with NG2-positive cells in the mouse cerebral cortex (Papay et al. 2004, 2006). α1-Adrenergic and mGluR5 receptors are likely to be activated extrasynaptically by neurotransmitters spreading from neuronal synapses or from non-synaptic release sites. Many other receptors have been observed in cultured oligodendrocyte precursors (for instance, Belachew & Gallo, 2004), but their expression in situ has not been demonstrated as yet. It would be interesting to establish to what extent these receptors are functionally expressed in NG2 cells in brain slices and their mechanisms of activation.

Concluding remarks

Since the exciting discovery of functional synaptic contacts in NG2 cells, many efforts have been focused on understanding the functioning of these unique synapses of the brain. This research has motivated the study of signalling mechanisms between neurones and NG2 cells in pathophysiological conditions and increased the interest of physiologists working in neurone–glia interactions for this glial cell type. However, our present knowledge of neurone-to-NG2 cell communication in the brain remains relatively limited and the reasons why neurones signal to NG2 cells are still enigmatic. This communication appears as complex as for neurones with a variety of possible signalling mechanisms, from very local (synaptic) to a widespread diffusion of neurotransmitters (volume transmission). The answer to the question ‘is neuronal communication with NG2 cells synaptic or extrasynaptic?’ is that it is probably both. However, the predominance of synaptic over extrasynaptic transmission in distinct physiological situations and brain regions as well as the roles of different communication modes remain elusive.