Synapses between NG2 glia and neurons


Dominik Sakry, Molecular Cell Biology, Department of Biology, Johannes Gutenberg University of Mainz, Bentzelweg 3, 55128 Mainz, Germany. E:


NG2-expressing glia are precursors to oligodendrocytes and subpopulations of astrocytes. They are unique among glial cells in that they enter into synaptic specialisations with neurons throughout all areas of grey and white matter and at all ages. To date, the NG2 cells appear to represent a postsynaptic compartment, and synapses are formed with axons. With differentiation to oligodendrocytes, NG2 is downregulated and myelin antigens upregulated: this coincides with a loss of the synaptic contacts between neurons and NG2 glial cells. The functional roles of this glial–neuron synapse in regulation of differentiation into myelinating oligodendrocytes or additionally responding to and modulating neuronal network activity remain to be elucidated.

NG2+ cells

Cells expressing the NG2 chondroitin sulphate proteoglycan (NG2+ glia) can be seen as a separate group of vertebrate glia distinct from astrocytes, myelinating oligodendrocytes and microglia (Peters, 2004; reviewed in Nishiyama et al. 2009; Trotter et al. 2010). NG2+ glia make up at least 5% of total glia in the mature CNS (Gallo et al. 2008), and are evenly distributed throughout various brain regions (Fig. 1) in both white and grey matter. They are migratory and proliferate even in the adult brain (Psachoulia et al. 2009). In the light of their differentiation into oligodendrocytes (see below) they are often equated with oligodendrocyte precursor cells (OPC). NG2+ glia divide and migrate to occupy defined areas where the cells are evenly spaced, contacting one another via their processes: ablation of a small area of OPCs in zebrafish using a laser demonstrated repopulation of the area with new and evenly spaced progenitors (Kirby et al. 2006).

Figure 1.

 Immunofluorescent staining of an adult heterozygous NG2/eYFP knock-in mouse brain (Karram et al. 2008). (A–C) Red channel represents bassoon staining (Bsn), green channel eYFP under the NG2 promoter and blue phosphorylated heavy neurofilament (NF-H/SMI31). (A) Coronal section showing the above markers/channels and an overlay of all three channels (merge). NG2+ cells show an equal distribution over the whole slice (scale bar 350 μm). (B) Enlarged view of the area surrounded by the square in (A) (scale bar 100 μm). (C) Standard deviation projection of a selected dividing NG2+ cell (green) in the hippocampus with surrounding staining for Bsn (red) and NF-H (blue). Image was acquired with a CLSM (scale bar 20 μm, z-stack 5 μm). (D) Maximum projection of a CLSM stack in the cortex. The red channel represents stained neuronal cell bodies (NeuN). Intimate contact between the cell body of an NG2+ glial cell and a NeuN+ neuron is shown (scale bar 20 μm, z-stack 5 μm).

NG2+ glia can differentiate into myelinating oligodendrocytes, astrocytes (Dimou et al. 2008; Zhu et al. 2008; Leoni et al. 2009), as well as possibly pyramidal neurons in the cortex (Guo et al. 2010), although a fraction of the cell population remains at the precursor stage under normal conditions at all ages (Rivers et al. 2008). Recent publications suggest that their developmental plasticity may be higher during early development (Zhu et al. 2011), whereas in the adult animal it is restricted to oligodendrocytes (Kang et al. 2010). The ability of NG2+ glia to give rise to myelinating oligodendrocytes and to remyelinate in the mature brain is important for demyelinating disorders such as multiple sclerosis or leukodystrophies. On account of their high proliferation rate (Psachoulia et al. 2009), OPCs are susceptible to mutations and therefore discussed as a source for brain tumours such as gliomas (Stallcup & Huang, 2008). Indeed, many gliomas express the NG2+ proteoglycan (see below).

Oligodendrocyte precursor cells within the CNS express in addition to NG2 distinct lineage markers such as platelet-d derived growth factor receptor alpha (PDGFR-α). OPCs are therefore often referred to as NG2+ cells or polydendrocytes (Nishiyama et al. 2009; Trotter et al. 2010). A subpopulation of pericytes can also express NG2; however, pericytes express PDGFR-β and not PDGFR-α, allowing them to be clearly distinguished from OPCs (Karram et al. 2008).

Distribution of the neuron–glia synapse

In 2000 it was first described that neurons from the CA3 region of the mouse hippocampus make synaptic contact with NG2+ cells in the hippocampal CA1 region (Bergles et al. 2000). Similar observations were later published using mice in which two populations of glial cells could be identified, those expressing receptors for glutamate (GluR+) and those expressing transporters for glutamate (GluRT+); the first group were described to receive synaptic input from neurons and appear to be NG2+ cells (Matthias et al. 2003; Jabs et al. 2005; Bergles et al. 2010). These studies utilised electrophysiology to show that neuronal activity resulted in measurable currents in the NG2+ glial cells.

These type of neuronal–glial synapses were subsequently described in the cerebellum (Lin et al. 2005), corpus callosum (Kukley et al. 2007; Ziskin et al. 2007) and cortex. Remarkably, the cortical NG2+ cells appear to maintain synaptic contact during cell division (Kukley et al. 2008; Ge et al. 2009; Fig. 1C). As NG2+ cells in all brain regions appear to form such synapses, this seems to be a characteristic feature of this glial cell type. Most strikingly, the synapses disappear with differentiation of the NG2+ cell to proteolipid protein expressing (PLP+) oligodendrocytes, in spite of retention of expression of the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and γ-aminobutyric acid (GABA)A receptors on the cells (De Biase et al. 2010; Kukley et al. 2010). In white matter tracks, most synapses are between unmyelinated areas of axons and processes of NG2+ cells. Unmyelinated areas of axons have recently been reported to release glutamate (Kukley et al. 2007; Ziskin et al. 2007). The total number of synapses per NG2+ cell and their distribution on the cell is still largely unelucidated, although indirect estimates suggest this could be up to 70 on one NG2+ cell from one cerebellar climbing fibre (Lin et al. 2005).

Since NG2+ cells are a persistent cell population in the adult brain, future studies aim at revealing the influence and importance of this synaptic connection under normal conditions in the mature brain as well as during development and in disease.

Structure of the neuron–glia synapse

The first receptors reported on OPCs (NG2+ glia) were voltage-gated ion channels (Sontheimer et al. 1989; Barres et al. 1990), whose discovery was seen in a new light when glutamatergic AMPA receptors (Bergles et al. 2000) and GABAA (Williamson et al. 1998; Lin & Bergles, 2004; Jabs et al. 2005) receptors were reported to be expressed by NG2+ glia in synaptic contact with neurons. Release of GABA and glutamate from the neurons forming the synaptic contacts resulted in activation of the glial GABAA and AMPA receptors. Other receptor types such as N-methyl-d-aspartate (NMDA) or kainate receptors have been reported to be expressed by NG2+ cells, but little is known about their functional role (Kukley & Dietrich, 2009; Bergles et al. 2010; Hamilton et al. 2010). Apart from the expression of the neurotransmitter receptors, there is a paucity of knowledge regarding the ultrastructure of these neuron–glia synapses. The initial electron microscopic pictures suggest an electron dense postsynaptic density (PSD) at the synaptic contact site in the NG2+ cell and an active zone with characteristic synaptic vesicles on the neuronal side (Bergles et al. 2000; Lin et al. 2005; Kukley et al. 2007). The protein components of these pre- and postsynaptic structures are not yet elucidated. However, it is known that NG2+ cells express the scaffolding protein GRIP1, which can simultaneously bind the NG2 protein via the GRIP 7th PDZ domain binding to the C-terminal PDZ-binding motif of NG2, and the GluR2/3 subunits of the AMPA receptor (Stegmüller et al. 2003).

Transcriptome data from isolated NG2+ cells (K. Karram and J. Trotter, unpublished communication) (Cahoy et al. 2008) demonstrate that the cells have mRNA for some classic PSD proteins like PSD95, but also for presynaptic markers such as synaptophysin or bassoon. The translation of these mRNAs into proteins and their contribution to the synapse has to be proven at the ultrastructural level, to reveal parallels or differences to neuron–neuron synapses. To date, the neuron–NG2+ cell synapse has been reported to be unidirectional from neuron to NG2+ cell, with the NG2+ cells representing a postsynaptic compartment and the neuronal axons forming a presynaptic compartment with the NG2+ glia. Confocal laser-scanning microscope (CLSM) images (Hamilton et al. 2010) reveal presynaptic markers co-localising on NG2+ cells but, due to the resolution limit of this technique and the enormous numbers of conventional synapses in vivo, it is difficult to be really sure if these presynaptic markers originate from the NG2+ cells or the innervating neuron.

Function of the neuron–glia synapse

The demonstration of synapses on NG2+ cells was a break with the paradigm that this specialised cell–cell interaction is a sole property of neurons within the CNS. Even though astrocytes participate in the establishment and function of neuron–neuron synapses, via the so-called tripartite synapse (Halassa et al. 2007; Eroglu & Barres, 2010; Faissner et al. 2010), a direct and exclusive synaptic contact between a neuron and a glial cell is unique to NG2+ glia. One obvious potential functional role for the synaptic contact between glia and neurons would be to regulate the availability and differentiation of myelinating oligodendrocytes. Myelination is known to be under electrical control (Barres & Raff, 1993; Zalc & Fields, 2000), and the expression of some axonal adhesion molecules such as L1, which can contribute to the regulation of CNS myelination (White et al. 2008), is also under electrical control (discussed in Lee & Fields, 2009).

Glutamate added to cerebellar cultures appeared to block differentiation and proliferation of NG2+ OPCs (Yuan et al. 1998), and it could be argued that the NG2+ cells are maintained as progenitors at the neuron–glial synapses with unmyelinated axons or in grey matter via glutamate release from the neurons. A change in the signal emanating from neurons could lead to differentiation. This would be the case in normal development, or in disease where recent work using genetically modified mice has confirmed previous assumptions that NG2+ glia are recruited to form remyelinating oligodendrocytes when required (Tripathi et al. 2010). Initial studies show increased synaptic contacts of NG2+ cells with demyelinated axons prior to remyelination, and a reduction of these synaptic contacts with synthesis of new myelin (Etxeberria et al. 2010). A functional regulatory link between the neuron–NG2+ cell synapse and myelination would have implications for all white matter diseases.

The ability of neurons to modulate the strength of synaptic signal transduction in long-term potentiation (LTP) is one of the hallmarks of synaptic plasticity. The finding that NG2+ cells can exhibit LTP (Ge et al. 2006) integrates them deeper into the neuronal network. LTP is thought to be the underlying cellular mechanism for learning and memory, and is associated with reorganisation of synaptic structures. If a similar mechanism operates in OPCs the question arises as to how these cells react to altered neuronal signals and in the end contribute to higher brain functions.

Two publications have reported generation of action potentials by NG2+ glia, this seems to be at variance with many other reports where no action potentials were measured in NG2+ cells under physiological conditions (Káradóttir et al. 2008; Ge et al. 2009). As a note of caution, the NG2 ectodomain is readily cleaved and deposited in the extracellular matrix: thus, use of antibodies against the extracellular domain of NG2 may in some instances localise deposited protein rather than the cells themselves. In this respect, it is interesting that the very first report of antibodies against NG2 stated that NG2 was expressed by interneurons (Stallcup et al. 1983).

Nevertheless, it has been clearly shown that NG2+ cells are able to sense neuronal activity and react with a depolarisation of the membrane potential. This is likely to result in a calcium increase in NG2 cells after synaptic stimulation (Hamilton et al. 2010). This could subsequently alter gene expression in the glial processes or cell body, similar to the chain of events that is set in motion by calcium increases in neurons (Greer & Greenberg, 2008). Do NG2+ cells release neuroactive substances such as growth factors or substances that could modulate the neuronal network? This would imply a bidirectional function for the neuron–glia synapse, which until now has only been reported to function as a uni-directional connection, and suggests that NG2 glia perceive neuronal stimulation in their capacity as myelinating oligodendrocytes but also as modulators of neuronal function. To date there is evidence that only a minority of NG2+ cells are directly coupled by gap junctions (Maglione et al. 2010). This would mean that every NG2+ cell responds more or less independently to the neuronal activity sensed over the neuron–glia synapses. NG2+ glia react fast and strongly to CNS damage including demyelination, stab wounds and viral infections (Dawson et al. 2000; Reynolds et al. 2002). It has recently been shown that NG2+ glia expand strongly in a mouse model for amyotrophic lateral sclerosis in which the superoxide dismutase is mutated in neurons and glial cells (Kang et al. 2010). It is conceivable that NG2+ glia sense the ‘state of health’ of their partner neurons over the neuron–glial synapse and respond accordingly: by proliferation and possibly release of neuroprotective substances.

A classic neuron–neuron synapse exhibits a localised depot of specific mRNAs in a translationally silenced state in dendritic or axonal synaptic compartments (Dahm et al. 2007). Upon reception of a stimulus translation of the RNAs ensues (Hüttelmaier et al. 2005). This initiates a fast reaction and bypasses the requirement to first ship protein newly synthesised in the cell body to the distant synapse. Oligodendrocytes similarly transport the mRNA for myelin basic protein in a translationally silenced state to the ends of their processes (Barbarese et al. 1999), where the translational inhibition is relieved by activation of fyn kinase in the glial cells (White et al. 2008). It will be interesting to discern whether specific mRNAs are selectively located at the neuronal or glial side of neuron–glial synapses.

The ability to conditionally modulate these synaptic contacts, for instance with the use of NG2+ cell-specific conditional knockout of AMPA and GABAA receptors and gene expression profiling after stimulation, would provide an ideal framework to answer many of these questions.

Materials and methods

Antibodies and immunofluorescence staining

Primary antibodies: bassoon (rabbit) kind gift of Prof. Gundelfinger (Magdeburg), SMI 31 (mouse) Covance, GFP (chicken) Abcam, Neun (mouse) Millipore.

Secondary antibodies: Alexa488 goat anti-chicken Invitrogen, Cy3 goat anti-rabbit Dianova, Cy5 goat anti-mouse Dianova.

Brains of heterozygous NG2/eYFP knock-in (Karram et al. 2008) mice were fixed for 2 h in 4% paraformaldehyde at 4 °C. Coronal sections with a thickness of 50 μm were cut with a Leica VT 1000S vibratome. Slices were permeablised and blocked for 2 h at RT [phosphate-buffered saline (PBS), 0.4% TX-100, 10% normal goat serum (NGS)]. Primary antibodies were applied for 24 h at 4 °C (PBS, 10% NGS). PBS washing was applied three times for 15 min. Secondary antibodies were applied for 2 h at RT (PBS, 10% NGS), followed by washing in PBS. Slices were mounted in Moviol on poly-l-lysine-coated slides.

Image acquisition and processing

Low-power images were taken with a Leica DM6000 fluorescence microscope and joined together with metamorph imaging software.

Confocal laser-scanning microscope images were taken with a Leica SP5 and the Leica imaging software. Each channel was recorded independently and the z-movement increment was 150 nm.

Further image processing was done with the ImageJ software.


The authors’ work is supported by the Deutsche Forschungsgemeinschaft (SPP1172).