NG2 glial cells (as from now NG2 cells) are unique in receiving synaptic input from neurons. However, the components regulating formation and maintenance of these neuron–glia synapses remain elusive. The transmembrane protein NG2 has been considered a potential mediator of synapse formation and alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) clustering, because it contains 2 extracellular Laminin G/Neurexin/Sex Hormone-Binding Globulin domains, which in neurons are crucial for formation of transsynaptic neuroligin–neurexin complexes. NG2 is connected via Glutamate Receptor-Interacting Protein with GluA2/3-containing AMPARs, thereby possibly mediating receptor clustering in glial postsynaptic density. To elucidate the role of NG2 in neuron–glia communication, we investigated glutamatergic synaptic transmission in juvenile and aged hippocampal NG2 cells of heterozygous and homozygous NG2 knockout mice. Neuron–NG2 cell synapses readily formed in the absence of NG2. Short-term plasticity, synaptic connectivity, postsynaptic AMPAR current kinetics, and density were not affected by NG2 deletion. During development, an NG2-independent acceleration of AMPAR current kinetics and decreased synaptic connectivity were observed. Our results indicate that the lack of NG2 does not interfere with genesis and basic properties of neuron–glia synapses. In addition, we demonstrate frequent expression of neuroligins 1–3 in juvenile and aged NG2 cells, suggesting a role of these molecules in synapse formation between NG2 glia and neurons.
NG2 cells are a subpopulation of glial cells abundantly distributed throughout the central nervous system. They express a variety of ion channels including different types of voltage-gated K+-, Na+-, and Ca2+ channels (Steinhäuser et al. 1994,; Akopian et al. 1996; Haberlandt et al. 2011). Moreover, they possess functional AMPA receptors (AMPARs) and GABAA receptors and receive a direct synaptic input from neurons (Bergles et al. 2010). The cellular mechanisms underlying formation and maintenance of these neuron–glia synapses and their functional impact are still unclear.
The NG2 protein is composed of a short cytoplasmic C-terminal region, a single transmembrane domain, and a large extracellular N-terminal region (Trotter et al. 2010). Sequence analysis revealed that the extracellular region contains 2 Laminin G/Neurexin/Sex Hormone-Binding Globulin (LNS) domains. In neurexins, these domains facilitate transsynaptic binding to neuroligins and influence formation, specificity, and function of neuronal synapses (Missler et al. 2012). Combined knockout of the 3 neurexin genes led to early postnatal lethality due to impaired synaptic transmission (Missler et al. 2003).
The C-terminus of NG2 carries a Postsynaptic density 95/Discs large/Zonula occludens-1 (PDZ)-binding motif that is recognized by the Glutamate Receptor-Interacting Protein (GRIP), which in turn binds to GluA2/3 subunit containing AMPARs. NG2 and GRIP1 could be coprecipitated from primary oligodendrocyte cultures and developing brain tissue (Stegmüller et al. 2003). Although most gray matter NG2 cells keep their phenotype throughout life, some of them proliferate and differentiate into astrocytes or oligodendrocytes (Dimou et al. 2008; Kang et al. 2010; Huang et al. 2014). Differentiation of NG2 cells is accompanied by the disassembly of neuron–NG2 cell synapses and simultaneous downregulation of NG2 (De Biase et al. 2010; Kukley et al. 2010).
Based on these findings, it was hypothesized that NG2 itself modulates neuron–NG2 cell synapse formation and postsynaptic AMPAR clustering (Trotter et al. 2010). To test this hypothesis, we investigated excitatory postsynaptic currents (EPSCs) in hippocampal NG2 cells of juvenile and aged NG2 knockout mice. Synapse formation was not prevented in the absence of NG2, and surprisingly, NG2 ablation did not alter kinetic properties and density of glial AMPAR responses. We observed the expression of neuroligins-1–3 in the majority of NG2 cells, which might have a role in establishing neuron–NG2 cell synapses.
Materials and Methods
Experiments were performed on NG2-EYFP-knockin mice (Karram et al. 2008) of the genotypes NG2 wild-type (wt), heterozygous NG2 knockout (+/−), and homozygous NG2 knockout (−/−). We compared juvenile (p8–12) and aged (>9 months) mice of either sex. In some experiments, aged C57/Bl6N animals were used as controls, of the same genetic background as wt animals generated from NG2-EYFP-knockin breedings.
Mice were anesthetized with 50% CO2/50% O2 and decapitated. The brain was rapidly removed and cut in horizontal orientation into 200–300 µm thick slices in ice-cold solution containing (in mM): 87 NaCl, 2.5 KCl, 1.25 NaH2PO4, 7 MgCl2, 0.5 CaCl2, 25 NaHCO3, 25 glucose, and 75 sucrose (347 mOsm). After sectioning, slices were incubated for 15 min in the same solution at 35 °C and afterwards stored at room temperature in artificial cerebrospinal fluid (aCSF) consisting of (in mM): 126 NaCl, 3 KCl, 2 MgSO4, 2 CaCl2, 10 glucose, 1.25 NaH2PO4, and 26 NaHCO3. Solutions were equilibrated with 95% O2/5% CO2 to pH of 7.4.
Experiments were performed using an Axioskop FS2 (Zeiss) equipped with IR-DIC and epifluorescence. Slices were perfused with gassed aCSF at room temperature. NG2 cells, located in the stratum radiatum of the hippocampal CA1 region, were identified by EYFP-fluorescence in +/− and −/− animals. Typical morphology and current pattern together with positive post hoc single-cell RT–PCR for platelet-derived growth factor α (PDGFα) receptor identified NG2 cells in wt animals. Whole-cell recordings were obtained using an EPC800 amplifier (HEKA Elektronik). Cells were hold at −80 mV and monitored with the TIDA software (HEKA). Currents were filtered at 1, 3, or 10 kHz and sampled at 6 or 20 kHz. Patch pipettes had resistances between 3 and 6 MΩ when filled with a solution consisting of (in mM): 130 KCl, 2 MgCl2, 0.5 CaCl2, 3 Na2-ATP, 5 1,2-bis(o-aminophenoxy)ethane-N,N,N',N'-tetraacetic acid) is a calcium-specific aminopolycarboxylic acid (BAPTA), and 10 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES). For miniature EPSC (mEPSC) recordings, the following pipette solution was used (in mM): 125 K-gluconate, 20 KCl, 3 NaCl, 2 Na2-ATP, 2 MgCl2, 0.5 EGTA, and 10 HEPES. Voltages were corrected for junction potential. We always added picrotoxin (150 µM) and tetrodotoxin (TTX) (1 µM) for mEPSC recordings. 2,3-dihydroxy-6-nitro-7-sulfamoyl-benzo[f]quinoxaline-2,3-dione (NBQX) (10 µM) was applied via the perfusion system as indicated. Substances were purchased from Sigma, Tocris Biosciences, and Abcam Biochemicals.
Minimal stimulation was achieved by 150 µs biphasic, constant voltage pulses (Model 2100, A-M Systems) to low-resistance (<1 MΩ) aCSF-filled glass pipettes. Focal pressure application was obtained with an Octaflow system (ALA Scientific Instruments). Input resistance (Rin) and membrane capacitance (Cm) were calculated from transients evoked by 10 mV voltage steps. Resting potential (Vrest) was measured within 1 min after establishing the whole-cell configuration using K-gluconate pipette solution. Membrane currents were offline compensated for capacitive artifacts. mEPSCs were detected using the template search of pClamp (Molecular devices). Events <5 pA were discarded. Analysis was carried out applying custom-written IGOR Pro procedures (WaveMetrics).
Single-Cell Reverse Transcription PCR
After recording, the cytoplasm of individual cells was harvested under microscopic control and reverse transcription (RT) was performed using the GoScript reverse transcriptase (Promega). A multiplex two-round single-cell RT–PCR was performed as previously described (Passlick et al. 2013) using the following primers: Neuroligin-1 sense (se)-5′-GACATCCGGAACGCCACTCA, antisense (as)-5′-TTGCCAAGACACTCCCATCATACA (280/340 bp); Neuroligin-1 (nested) se-5′-CAGAATATCATTGATGGCAGATTG, as-5′-ACTCCCATCATACAGATTTCCAGT (231/291 bp); Neuroligin-2 se-5′-ACCGCCAAGCTGCATGCCGACTAC, as-5′-GCACGCGTGGTTTCAAGCCTATGT (370 bp); Neuroligin-2 (nested) se-5′-GCAGAGGGCCGGCCAGAGTG, as-5′-AGCGGTTGGGCTTGGTGTGGAT (229 bp); Neuroligin-3 se-5′-AAGATGGATCCGGCGCTAAGAAAC, as-5′-CCGATGCCAGAGCCAAAGACAGTA (358 bp); Neuroligin-3 (nested) se-5′-AATGACGGGGATGAAGATGAAGAC, as-5′-GCGAAGGGCCTGGATTTGA (243 bp). All se and as primers were located on different exons to prevent amplification of genomic DNA. Neuroligin-1 primer flanked splice site A resulting in 2 discrete PCR products. cDNA fragments were identified by gel electrophoresis using a low-molecular-weight DNA ladder (New England BioLabs) as a length marker.
Data analysis was carried out with the IGOR Pro (WaveMetrics) and Origin (OriginLab) software. Each data set was tested for normal (Gaussian) distribution (Shapiro–Wilk test). For clarity, we depicted Gaussian-distributed data as bar graphs and non-Gaussian-distributed data as box plots. In case of Gaussian distribution, any data point outside 2 times SD was excluded. Gaussian-distributed data were tested for equal variances (Levene test), followed by ANOVA and post hoc Tukey test without or with Welch correction for equal or diverse variances, respectively. Non-Gaussian-distributed data were tested either with Kruskal–Wallis ANOVA with sequential Bonferroni's correction for >2 groups, or with the Mann–Whitney U-test in case of 2 groups. Cumulative probability distributions were tested with the Kolmogorov–Smirnov test. Relative frequencies were compared using Pearson's χ2 test. Significance level was set to P<0.05. Data in bar graphs represent mean ± SD. Box plots indicate median with lower (25) and upper quartile (75), and whiskers denote quartiles 5 and 95.
NG2 Protein Influences Passive Membrane Properties of NG2 Cells
Passive membrane properties were compared in wt, +/−, and −/− mice with whole-cell patch-clamp recordings. NG2 cells displayed the characteristic complex current pattern (Steinhäuser et al. 1994; Fig. 1A). As reported (Kressin et al. 1995; Bordey and Sontheimer 1997; Maldonado et al. 2013), juvenile NG2 cells displayed variable inwardly rectifying K+ (Kir) current amplitudes (Fig. 1A, upper panel), which increased in aged cells (Fig. 1A, lower panel). This developmental change was also reflected by a decrease in Rin (Fig. 1B). In addition, the Rin of aged wt cells was decreased compared with +/− and −/− ones. Vrest was not influenced by NG2 (Fig. 1D). Cm displayed genotype- and age-dependent changes (Fig. 1B). In juvenile mice, Cm of wt cells was smaller than that of +/− and −/− cells. Aged and juvenile wt cells exhibited similar Cm values, whereas +/− and −/− cells showed a decrease in Cm with increasing age. Thus, in juvenile and aged mice, NG2 expression influences cell size.
Loss of the NG2 Protein Does Not Impair NG2 Cell Synapse Formation and Short-Term Synaptic Plasticity
To investigate whether NG2 protein is essential for neuron–NG2 cell synaptic signaling, we investigated evoked EPSCs (eEPSCs). In the presence of the GABAA receptor blocker picrotoxin (150 µM), Schaffer collateral stimulation elicited AMPAR-mediated eEPSCs in NG2 cells of all genotypes tested (Fig. 2A), which were blocked by the AMPA/kainate receptor antagonist NBQX (10 µM; Fig. 2B). The kinetics of AMPAR responses are influenced by various factors, including subunit composition of the receptors, association with auxiliary subunits, and time course of glutamate concentration in the synaptic cleft (Cathala et al. 2005; Jackson and Nicoll 2011). We did not observe NG2-dependent changes in eEPSC kinetics (Fig. 2C). However, during development, there was an acceleration of rise and decay times in +/− and −/− cells.
In neurons, postsynaptic neuroligins alter presynaptic release probability and short-term synaptic plasticity by retrograde transsynaptic signaling (Futai et al. 2007). To test whether NG2 is involved in retrograde signaling at neuron–glia synapses, we performed paired-pulse stimulation experiments (interstimulus interval 20 ms). In agreement with previous studies (Bergles et al. 2000; Ge et al. 2006), hippocampal glutamatergic NG2 cell synapses displayed strong paired-pulse facilitation (Fig. 2D). The paired-pulse ratio (PPR) revealed no differences among genotypes (Fig. 2D). A developmental decrease in the PPR was observed in wt cells.
Thus, NG2 affects neither presynaptic release probability nor assembling and kinetic properties of glutamatergic neuron–NG2 cell synapses.
Frequency of Synaptic Innervation and Postsynaptic AMPAR Density Are Not Regulated by NG2
Contradictory studies postulate that NG2 is either inhibitory or permissive for neurite outgrowth (de Castro et al. 2005; Tan et al. 2005; Trotter et al. 2010). In the case that NG2 binds to a presynaptic partner via its LNS domains, reduced levels of NG2 should affect synapse number. We estimated synaptic connectivity by measuring AMPAR-mediated mEPSCs. Since our paired-pulse stimulation experiments suggested that NG2 has no impact on presynaptic release probability (Fig. 2D), we considered the frequency of mEPSCs as an indirect measure of synaptic connectivity. The frequency of mEPSCs was very low in juvenile mice, and decreased further during development. Importantly, no genotype-dependent changes in mEPSC frequency were detected (Fig. 3A,B).
Since NG2 is part of a complex with GRIP1 and GluA2/3 (Stegmüller et al. 2003), it might be important for trafficking and clustering of AMPARs in the postsynaptic density. To assess whether NG2 influences these processes, we analyzed mEPSC amplitudes in NG2-deficient cells. We increased presynaptic transmitter release by focal application of ionomycin (3 µM, Fig. 3C). Mean mEPSC amplitudes, kinetics, and cumulative probability distribution were indistinguishable between the genotypes (Fig. 3D). In agreement with our eEPSC analysis, all genotypes showed a developmental acceleration of AMPAR gating kinetics (Fig. 3E).
Our data do not support a crucial role for NG2 in regulating synaptic connectivity and postsynaptic AMPAR density. Rather, we report developmental, NG2-independent changes in mEPSC frequency and AMPAR current kinetics, while mEPSC amplitudes remained unchanged during ontogenesis.
NG2 Cells Frequently Express Neuroligins-1–3
Since our electrophysiological data did not support a critical role for the NG2 protein in neuron–NG2 cell synaptic transmission, we asked whether these synapses might be regulated by neuroligin–neurexin interaction as established for neuron–neuron synapses (Missler et al. 2012). Single-cell RT–PCR analysis revealed frequent expression of the 3 isoforms, neuroligins-1–3 in juvenile and aged NG2 cells of +/− and −/− mice (Fig. 4). In aged NG2 cells, significantly reduced neuroligin-3 expression was observed in the absence of NG2 (Fig. 4C). All cells positive for neuroligin-1 expressed the short splice variant lacking the insert at splice site A (Dalva et al. 2007), except for 3 juvenile NG2 cells expressing long and short variants and 2 aged NG2 cells expressing only the long variant.
Transsynaptic adhesion proteins are crucial for proper formation, maturation, and plasticity of neuronal synapses (Dalva et al. 2007), but for neuron–NG2 cell synapses these fundamental processes are unknown. The NG2 protein itself was considered a promising candidate for modulating structural interactions, because the extracellular LNS domains and intracellular PDZ-binding motif of NG2 fit the concept of synapse formation and AMPAR trafficking. Our results show that these unique glutamatergic synaptic junctions form independent of NG2. It is known that neuronal synapses may also form in the absence of neurexins or neuroligins, but these proteins have an additional function as important regulators of presynaptic transmitter release (Futai et al. 2007; Missler et al. 2012). The surprising lack of differences in the PPR between wt and NG2-lacking cells suggests that the NG2 protein also does not influence release probability (Fig. 2D). Since frequency and amplitudes of mEPSCs were indistinguishable between genotypes, our data indicate that the number of synapses and AMPAR trafficking into postsynaptic densities of neuron–NG2 cell synapses do not depend on the expression of NG2 (Fig. 3B,D). The kinetics of glial eEPSCs and mEPSCs also remained unchanged after NG2 ablation (Figs 2C and 3E), suggesting no effect of NG2 on subunit composition, expression of auxiliary subunits, and synapse structure (Cathala et al. 2005; Jackson and Nicoll 2011).
We report that neuron–NG2 cell synapses undergo developmental changes in their physiological properties. While previous studies investigated animals up to an age of 1–2 months (Mangin et al. 2008; De Biase et al. 2010; Kukley et al. 2010), we compared the properties of these synapses in juvenile and aged mice (>9 months). Analyses of eEPSCs and mEPSCs revealed faster AMPAR gating at the mature stage (Figs 2C and 3E). This observation is in line with a previous study on NG2 cells of the dentate gyrus demonstrating the acceleration of synaptic AMPARs during postnatal development (Mangin et al. 2008). The developmental change might be caused by a switch in AMPAR subunit composition (Seifert et al. 1997, 2003; Ge et al. 2006). Alternatively, maturation of the synaptic structure may have influenced the time course of glutamate concentration in the synaptic cleft (Cathala et al. 2005), or gating properties of the receptors were modulated by association with auxiliary subunits (Zonouzi et al. 2011). We noted a higher variability of membrane parameters in the juvenile hippocampus (cf. Figs 1C and 2C), which probably reflects the different age of individual NG2 cells. Actually, several membrane parameters, for example, expression of Kir channels (Kressin et al. 1995) and AMPARs of NG2 cells (Seifert et al. 1997,, 2003) are dynamically regulated during the first days of development, reaching a plateau after a few weeks of age.
Another developmental alteration of neuron–NG2 cell synapses concerned a drastic decrease in mEPSC frequency, indicating lower synaptic connectivity in aged mice (Fig. 3A,B). This extends earlier work showing a relatively low frequency in juvenile mice (Jabs et al. 2005; Kukley et al. 2010), which further decreased in young adult animals (De Biase et al. 2011). Our data confirm that NG2 cells receive synaptic input throughout adulthood, with the unchanged mEPSC amplitudes suggesting a stable AMPAR density throughout development (Fig. 3D).
We noted NG2-dependent changes in passive membrane properties. The increased Rin in aged NG2-lacking animals suggests that the NG2 protein affects the expression and/or function of resting K+ channels in these cells (Fig. 1C). A slight increase in Cm was observed in juvenile NG2-lacking cells, which might reflect morphological changes these cells undergo in the context of their process motility (Haberlandt et al. 2011) or migratory activity. In fact, NG2 cells travel over considerable distances during development or after lesions (Niehaus et al. 1999; Hughes et al. 2013), and NG2 is critically involved in regulating cell polarity and migration (Makagiansar et al. 2007; Biname et al. 2013). Since mEPSC frequency is similar in wt, +/−, and −/− cells, the decreased Cm might indicate that the density of synapses in juvenile −/− cells is lower compared with wt cells. However, this difference disappeared in adult animals.
We demonstrate here that NG2 cells abundantly express neuroligins 1–3, suggesting that neuron–NG2 cell synapse formation is regulated by neuroligin–neurexin interactions, similar to neuron–neuron synapses (Missler et al. 2012). Our data demonstrate an NG2-dependent decrease in the frequency of neuroligin-3 expression in the aged hippocampus, and interestingly a decline of this isoform was recently shown to be involved in tonic endocannabinoid signaling and autism (Foldy et al. 2013). However, since the PPR remained unchanged, it is unlikely that neuroligin-3 is involved in pre- to postsynaptic signaling between neurons and NG2 cells. These findings provide promising targets for interfering with synaptic transmission between neurons and NG2 cells to unravel the functional impact of this unique type of neuron-to-glia signaling.
This work was supported by DFG (SE774/3 and STE552/3) and EC (FP-202167 NeuroGLIA).
We thank T. Erdmann and I. Fiedler for excellent technical assistance and A. Wefers for providing custom-written IGOR macros. Conflict of Interest: The authors declare no competing financial interests.