NG2 glial cells integrate synaptic input in global and dendritic calcium signals

NG2-expressing glial cells (NG2 cells) are a large population of self-renewing cells present in substantial numbers in all white and grey matter regions of the CNS throughout development and adulthood (Nishiyama et al., 1999; Butt et al., 2005; Trotter et al., 2010). NG2 cells are precursor cells to all oligodendrocytes, which exclusively provide isolating myelin to axons and are thereby essential for rapid conduction velocity while maintaining small axon diameter in compact brains. Recent evidence suggests that myelination is plastic and activity-dependent and that synaptic transmission to NG2 cells plays a key role in detecting patterns of activity. However, the mechanisms by which neuronal activity is integrated by NG2 cells and transformed into precisely tuned axonal conduction velocities via myelination is not known.

NG2 cells stand out from other glial cells as they receive direct vesicular synaptic input from neighboring axons (Bergles et al., 2000; Lin and Bergles, 2004; Jabs et al., 2005; Ge et al., 2006; Kukley et al., 2007; Ziskin et al., 2007; Karadottir et al., 2008; Kukley et al., 2008; Velez-Fort et al., 2010), and based on these findings, it has recently been proposed that release of neurotransmitter onto NG2 cells may be the key signal mediating activity-dependent myelination and instructing the cells about the need to generate more oligodendrocytes and myelin (Wake et al., 2011; Hines et al., 2015; Mensch et al., 2015; Koudelka et al., 2016). NG2 cells are also unique amongst glial cells as they express a large number of voltage-gated ion channels (VGCs) (Chittajallu et al., 2004; Jabs et al., 2005; Karadottir et al., 2008; De Biase et al., 2010; Kukley et al., 2010; Clarke et al., 2012). However, while it is textbook knowledge that neurons use VGCs to integrate synaptic input and to transform it into a pattern of action potentials and associated calcium signals, the functional significance of VGCs and synaptic input in NG2 cells has so far remained obscure. NG2 cells are not capable of firing action potentials (Bergles et al., 2000; Lin et al., 2005; Tong et al., 2009; De Biase et al., 2010) and their postsynaptic potentials (PSPs) are so small (Jabs et al., 2005; Haberlandt et al., 2011) that it has been questioned whether synaptic input can even achieve a relevant depolarization to recruit VGCs. Furthermore, previous attempts to elicit calcium signals in NG2 cells by synaptic stimulation failed and led the authors to conclude that the synaptic input was too weak (Velez-Fort et al., 2010; Haberlandt et al., 2011). Thus, at present, we do not know how NG2 cells could transform synaptic input into activity-dependent myelination.

On the other hand, the density of synapses might be massively underestimated, as the low frequency of spontaneous synaptic events may be due to the smaller surface area of NG2 cells when compared to that of neurons (Kukley et al., 2007; Kukley et al., 2008). Furthermore, the high-input impedance of the delicately thin dendritic branches could result in strong local depolarization even if the response arriving at the soma is small (Sun and Dietrich, 2013).

Deciphering how NG2 cells can integrate and process synaptic input is of fundamental importance for understanding brain development and for improving re-myelination of damaged white matter. Therefore, in this study, we investigated whether synaptic depolarizations recruit VGCs in the somatic or dendritic compartment of NG2 cells. We report that NG2 cells possess a dedicated signal integration machinery, different from that of neurons and other types of glial cells, and generate local dendritic or global Ca²⁺ responses depending on the pattern and type of incoming activity.

We first set out to obtain a measure of the maximal level of synaptic depolarization of NG2 cells. To this end, we recorded DsRed-expressing NG2 cells in CA1 stratum radiatum from an NG2-DsRed mouse line in current-clamp mode (Figure 1A) and optimized the electrical stimulation of Schaffer collaterals to estimate the maximally attainable postsynaptic depolarization. In four NG2 cells resting at −85 mV, we recorded a substantial peak depolarization of PSPs up to −35.4 ± 8.6 mV (n=4, Figure 1B).

Figure 1

Download asset Open asset

To further explore whether PSPs recruit VGCs and to obtain precise control of the degree and timing of intracellular depolarization, we induced mock PSPs by current injection. Realistic mock PSPs were generated by deriving the current injection waveform from miniature excitatory postsynaptic currents (EPSCs) recorded in NG2 cells representing the average conductance change in response to release of a single glutamate-filled vesicle. Upon increasing the amplitude of the injected current waveform (Figure 1C, grey shaded area) the response increased concomitantly, and we observed an obvious shortening of the voltage response (black line) reminiscent of the activation of VGCs (Figure 1C). To analyze the impact of VGCs, we compared these mock PSP responses to tiny hyperpolarizations, which we produced by injecting very small negative mock currents designed to avoid activation of VGCs. We arithmetically scaled up such small voltage responses by the ratio of the injected current strengths (see Materials and methods for details) to create a hypothetical, large response unaffected by VGCs (dashed grey line, Figure 1C).

To dissect the contribution of different types of VGCs to the shortening of mock PSPs, we bath-applied established blockers of VGCs (Figure 2A,B). The A-type K⁺-channel blocker 4-aminopyridine (4-AP) potently reversed the shortening of the mock PSP and significantly increased the half-width from 2.4 ± 0.2 ms to 3.8 ± 0.5 ms (n=4, Paired t-test). In addition, 4-AP increased the amplitude of the mock PSP from 65.8 ± 2.1 mV to 78.7 ± 4.3 mV (significantly different, s.d., Paired t-test). By contrast, blocking Na⁺ channels by using tetrodotoxin (TTX) significantly reduced the amplitude from 66.1 ± 0.9 mV to 57.9 ± 1.1 mV (n=7, Paired t-test), and also slightly increased the half width of PSPs (from 2.3 ± 0.3 ms to 3.0 ± 0.2 ms, s.d., Paired t-test). The latter is likely secondary to the reduced amplitude, which results in reduced recruitment of A-type channels. Delayed rectifier K⁺ channels are not recruited by individual mock PSPs because tetraethylammonium (TEA) affected neither their amplitude nor their half-width (n = 6, Paired t-test).

Figure 2

Download asset Open asset

In order to identify the membrane potential range in which VGCs modulate synaptic potentials, we varied the amplitude of mock current injections in the absence or presence of ion-channel blockers. We quantified VGC-induced changes of mock PSPs by dividing their amplitude and half-width by the amplitude and half-width of scaled passive responses (negative current injection as above) and plotted the resulting half-width- and amplitude-ratio against the membrane potential reached at the peak of the recorded mock PSPs (Figure 2C–E). These plots clearly showed that in the absence of channel blockers, the durations of mock PSPs were increasingly shortened when the membrane depolarization exceeded −60 mV (Figure 2D, black line), consistent with the low-voltage activation range of A-type channels. Near the maximally attainable average depolarization of −16.5 ± 1.0 mV, the half-width ratio of mock PSPs was substantially shortened to 47.1 ± 3.4% (Figure 2D). This shortening is almost exclusively due to the action of A-type K⁺ channels, whereas blocking Na⁺ and delayed-rectifier K⁺ channels did not detectably change the duration of mock PSPs (Figure 2C,D). In contrast to the half-width, the amplitude of mock PSPs was only slightly reduced, as indicated by an amplitude ratio near 1 even for large PSPs (Figure 2E, black line). This linear scaling of the amplitude of mock PSPs in control conditions is a result of a balanced interplay of fast Na⁺ and A-type K⁺ channels (Figure 2C,E): if A-type or Na⁺ channels are blocked, the amplitude of mock PSPs is increased or decreased by ~15% and ~30%, respectively. Taken together, these results demonstrate a prominent role of A-type K⁺ channels in shaping synaptic input and suggest that even medium-sized synaptic input (∆ 20–30 mV) is accelerated by A-type K⁺ channels, while the amplitude of synaptic depolarizations faithfully encodes the input strength due to mutual antagonism of A-type K⁺ and fast Na⁺ channels.

In neurons, voltage-gated K⁺ and Na⁺ channels are responsible for generating an action potential waveform upon integration of synaptic input. Action potentials are potent openers of voltage-gated calcium channels (VGCCs) and elicit prominent calcium signals. Since NG2 cells do not fire action potentials (Bergles et al., 2000; Lin et al., 2005; Tong et al., 2009; De Biase et al., 2010), we wondered whether PSPs may directly trigger intracellular calcium signals, and therefore combined 2-photon Ca²⁺ imaging with our mock PSP stimulation paradigm. During injection of a robust stimulus (10 mock PSPs at 100 Hz), we monitored the fluorescence of the calcium indicator dye Fluo-4 (200 µM) with line-scans across proximal dendrites and the soma of NG2 cells (Figure 3A–D). Indeed, we detected clear calcium transients in the soma and proximal dendrites (3.9 ± 1.3% and 12.7 ± 1.4%, Figure 3E). These Ca²⁺ signals required the activation of VGCCs, as they were highly sensitive to a cocktail of VGCCs blockers (Figure 3F,G). Further, the high sensitivity of Ca²⁺ signals to combined application of SNX-482 (1 µM, R-type VGCCs blocker, Newcomb et al., 1998) and TTA-P2 (30 µM, T-type VGCCs blocker, Choe et al., 2011) suggested that they are mediated by low-voltage-activated Ca²⁺ channels (Figure 4A and B). In contrast, Na⁺/Ca²⁺ exchangers, which were previously shown to mediate pharmacologically induced Ca²⁺ signals, are not involved in mock EPSP-induced Ca²⁺ signals as blocking these exchangers with KB-R7943 (10 µM, Iwamoto et al., 1996) did not decrease their amplitude (Figure 4A and B). Ca²⁺ signals decayed back to baseline in an exponential manner (τ_decay=3.16 ± 0.5 s, n=9, see Materials and methods, Figure 4C) and could be reproducibly elicited over a period of more than 12 min (Figure 4D, n=5).

Figure 3

Download asset Open asset

Figure 4

Download asset Open asset

Because somatic current injection regularly produced a calcium response in proximal dendrites, we analyzed the spatiotemporal distribution of calcium in NG2 cells in more detail by combining 2-photon microscopy and fast frame acquisition (30 Hz). Scanning areas and focus planes were selected to almost fully contain at least one primary dendrite which extended up to ~35 µm from the soma (Kukley et al., 2010, Figure 5A). Injection of 10 mock PSPs as above generated an almost immediate increase in fluorescence throughout all scanned dendrites, which decayed on the order of several seconds (Figure 5B and D). We quantified this signal within regions of interest (ROIs) covering successive 5 µm segments along dendrites (Figure 5C). This analysis showed that the Ca²⁺ signal amplitude remained constant throughout the dendritic tree, or even increased towards the periphery (Figure 5D and E), and that the signal started almost simultaneously, irrespective of the distance from the soma (Figure 5F). In particular, the spread of the Ca²⁺ signal was much faster than expected from pure diffusional propagation from the soma into dendrites (Figure 5F). Together with the undiminished amplitude and steep rise of the dendritic Ca²⁺ signal, the data suggest that the membrane depolarization of the mock PSPs rapidly spreads throughout the dendritic tree and opens local dendritic VGCCs. The minimal delay and the rapid rise of dendritic Ca²⁺ signals in NG2 cells is comparable to the fast dendritic Ca²⁺ responses of CA1 pyramidal neurons recorded and stimulated under identical conditions (Figure 5G and H), suggesting that NG2 cells are capable of similar dendritic signal processing.

Figure 5

Download asset Open asset

A train of 10 large PSPs may not happen very frequently under physiological conditions because it requires a large number of presynaptic neurons to fire in a highly synchronized manner. We therefore stimulated NG2 cells with a more realistic pattern of synaptic input. We allowed firing of presynaptic neurons to deviate from perfect synchrony and modeled a temporal distribution of presynaptic activity using a Gaussian distribution with a standard deviation of 25 ms (Figure 6A). We progressively increased the number of injected Gaussian-distributed quanta (Q) and monitored intracellular calcium levels with 2-photon microscopy in line-scan mode. In the control group, Ca²⁺ responses were seen only rarely (Figure 6B). By contrast, when blocking A-type K⁺ channels, we observed larger depolarizations often carrying prominent spikelets (in five out of seven cells, average amplitude 40.2 ± 6.7 mV), which were accompanied by substantial intracellular Ca²⁺ elevations (Figure 6C). When plotting the peak depolarization against the current injection strength (being similar across groups, Figure 7A), voltage responses in the control and 4-AP groups were indistinguishable as long as they did not exceed ~−45 mV, consistent with the threshold for recruitment of A-type channels as determined above. As expected, larger current injections (injections equivalent to >600 glutamate quanta, see Materials and methods) much more effectively depolarized NG2 cells in the presence of 4-AP (Figure 7A). To discriminate Ca²⁺ responses from non-responding trials, we analyzed the line-scan profiles with a template-fitting algorithm. This analysis revealed (Figure 7A) that for both groups Ca²⁺ responses (filled symbols) are much more frequent if NG2 cells are depolarized beyond approximately −45 mV, suggesting that blocking A-type channels facilitates Ca²⁺ signals by allowing for stronger depolarizations. Ca²⁺ signals were not only significantly more frequent in the presence of 4-AP (control: 9/40; 4-AP: 22/43; Fisher’s exact test, Figure 7B), but the responding trials also showed a significantly larger amplitude: 10.1 ± 0.7% and 14.1 ± 1.5% in the control and 4-AP groups, respectively (Student’s t-test, Figure 7C). Further analysis showed that the fraction of trials that responded with a Ca²⁺ signal steadily increased towards unity with the quantal content of the stimulus in both groups. However, 4-AP treatment left-shifted this curve, lowered the 50% threshold from ~1200 to ~600 quanta, and thereby strongly increased the propensity of NG2 glial cells to generate Ca²⁺ signals (Figure 7D). Interestingly, re-plotting the fraction of responding trials against the peak depolarization minimizes the difference between the two groups seen when plotting the fraction against injection strength (Figure 7E), further indicating that 4-AP does not change the voltage dependence of the Ca²⁺ signal but does change the recruitment of Ca²⁺ channels. The strength of the template-fitting algorithm lies in its ability to identify clear responses unambiguously, but as a result, 'non-responding' trials do not enter the analysis. To characterize the whole population of NG2 cell dendrites, we therefore undertook an additional analysis including all dendrite recordings. We binned trials according to stimulus strength, as illustrated in Figure 7D, but calculated the average ∆F/F ratios across all trials. This analysis revealed a small but statistically significant Ca²⁺ signal in the control group and also showed that dendritic Ca²⁺ levels linearly increased with input strength when A-type K⁺ channels were blocked (Figure 7F). In summary, these results indicate that A-type K⁺ channels tightly gate the generation of Ca²⁺ signals in response to synaptic input to NG2 cells.

Figure 6

Download asset Open asset

Figure 7

Download asset Open asset

Dendrites are likely to receive the majority of synapses as they provide the major part of the cell's surface area (Kukley et al., 2008; Sun and Dietrich, 2013). Thus the largest fraction of synaptic currents originates in dendrites and its integration is difficult to study with somatic current injection. To explore dendritic integration of synaptic input by NG2 cells, we employed 2-photon-based MNI-glutamate uncaging. This technique allowed us to produce very localized, diffraction-limited, and brief (0.65 ms, at 720 nm) photo-release of glutamate onto dendritic segments, mimicking the rapid liberation of glutamate from presynaptic vesicles. By pointing the scanner at segments in the outer two thirds of the NG2 cell dendritic tree (see scheme in Figure 8A, grey shaded area and Figure 8B), we produced uncaging EPSPs (uEPSPs, Figure 8C), which compared well in amplitude and kinetics to the small spontaneously occurring fast EPSPs in NG2 cells (Figure 8D,E). Generating six uEPSPs within the same dendritic subsegment almost simultaneously produced a compound uEPSP with an amplitude of 3.2 ± 0.9 mV, whereas arithmetic summation of the six sequentially acquired responses from the same individual positions amounted to 4.4 ± 1.3 mV (Figure 8C bottom panel, F, G).

Figure 8

Download asset Open asset

We next asked whether dendrites may generate Ca²⁺ signals more readily in response to local synaptic input and, in particular, whether they might respond to the additional activity of a few synapses if many other synapses were already active throughout the whole dendritic tree. To address this question, we employed mock PSPs to simulate a large compound synaptic response, sized to remain below the threshold of Ca²⁺ responses, and then added six active synapses, co-localized within one dendritic subsegment, by local 2-photon glutamate uncaging (>10 µm from soma, Figure 9A). Under control conditions, the addition of six uEPSPs to the mock PSP hardly increased the depolarization (−40.8 ± 1.4 mV vs −39.6 ± 1.3 mV, Figure 9B), as expected from the small amplitude of the uEPSP, and also did not generate an increased calcium response in the stimulated dendrite (6.0 ± 1.2% vs 8.8 ± 2.5% ∆F/F, one-way ANOVA for correlated samples, Tukey HSD, Figure 9B and D). However, the situation was strikingly different if we inhibited A-type K⁺ channels by application of 4-AP. The simulated global synaptic input of the same strength applied alone or together with the six uEPSPs produced a comparable depolarization of the soma (−36.5 ± 1.4 mV vs −34.0 ± 1.7 mV, Figure 9C). In addition, the mock PSP on its own was not associated with a clear Ca²⁺ response. However, adding local uEPSPs significantly boosted the Ca²⁺ signal in the stimulated dendrite (6.7 ± 1.3% vs 14.3 ± 3.3% ∆F/F, one-way ANOVA for correlated samples, Tukey HSD, Figure 9C,D). Taken together, these data show that dendritic intracellular Ca²⁺ levels in NG2 cells indicate the activity of local synaptic input despite the small amplitude of the locally induced synaptic potential.

Figure 9

Download asset Open asset

Considering the pivotal and powerful role of the activity of A-type K⁺ channels in allowing calcium signals to be triggered in NG2 cells, we asked whether synaptic input itself may bring a relevant fraction of A-type channels into inactivation. When plotting the half-width of the mock PSP voltage responses in a train (10 PSPs at 100 Hz), we found a clear increase of 26.3 ± 3.3% (n=7, Figure 10A,B). This progressive broadening of PSPs is very likely to reflect an increasing fraction of inactivated A-type K⁺ channels as this broadening was not observed in the 4-AP-treated group (n=8, Figure 10A,B). Such an increase in half-width is very likely to enhance Ca²⁺ entry via an increased tail current during the repolarization phase (Ca²⁺ channels opened, prolonged period with large and increasing driving force for Ca²⁺). Indeed, with broadened PSPs in the 4-AP treated group, we observed a significantly increased Ca²⁺ response (Figure 10C and D, 12.7 ± 1.4% vs 24.2 ± 4.2% in the control and 4-AP groups, respectively, Student’s t-test). Thus, the data suggest that use-dependent inactivation of A-type channels renders a high frequency train of synaptic input that is effective in causing Ca²⁺ responses in NG2 glial cells. Amplification of mock PSP-induced Ca²⁺ signals was also seen when voltage-gated K⁺ channels were blocked by the inclusion of Cs⁺ in the pipette solution (n = 10, Figure 10C and D, 12.7 ± 1.4% vs 30.5 ± 9.3% in control and Cs⁺ groups, respectively, Wilcoxon-Mann-Whitney test). Furthermore, the pipette solution also contained 1 µM of the SERCA blocker thapsigargin and thus showed that Ca²⁺ stores are not important for mock PSP-induced Ca²⁺ signals in NG2 cells.

Figure 10

Download asset Open asset

While the above experiments demonstrated that synaptic input can lead to large depolarization of the soma of NG2 cells (Figure 1B), all Ca²⁺ signals reported so far were evoked by direct current injection. Therefore, we designed an experiment to show that synaptic transmitter release can trigger local Ca²⁺ signals in NG2 cell dendrites. We placed a small glass electrode in the immediate vicinity of a small dendritic branch (Figure 10E) and elicited a brief train of EPSPs (five or six stimuli at 100 HZ, Figure 10E). As seen with local glutamate uncaging, these dendritically induced EPSPs were recorded with small amplitudes by the electrode attached to the soma (7.1 ± 1.1 mV, n=4). Nevertheless, in the nearby dendrite, these EPSPs were accompanied by a clear Ca²⁺ signal with kinetics and size similar to those shown above for Ca²⁺ signals induced by current injection or glutamate uncaging (21.1 ± 3.4%, n=4, Figure 10E, 1 µM thapsigargin in the pipette). Furthermore, as argued above, this observation also suggested that the local depolarization in the dendrite is much larger than that observed at the soma and that there is strong voltage attenuation along a dendrite. More importantly, this experiment substantiates the physiological relevance of Ca²⁺ signals in NG2 cells.

The key finding of the present study is that NG2 cells, which were previously believed to be electrically unexcitable, are effective integrators of synaptic activity. Moreover, they possess two elaborate integration modes governed by A-type K⁺ channels: local compartmentalized integration in individual dendritic subunits, versus global integration leading to massive Ca²⁺ influx throughout the entire dendritic arborization of NG2 cells.

It has been known for a long time that NG2 cells, formerly also called 'complex astrocytes' (Steinhauser et al., 1994a, 1994b; Sun and Dietrich, 2013), express considerable levels of VGCs but their physiological role has not yet been identified. Further, it has been questioned whether the typically small synaptic events in NG2 cells ever recruit VGCs. Earlier studies even proposed that some voltage-gated K⁺ channels (delayed-rectifier- but not A-type K⁺ channels) may serve functions other than modifying membrane potential and the authors proposed their involvement in controlling cell cycle timing (Gallo et al., 1996; Knutson et al., 1997; Ghiani et al., 1999a, 1999b).

Mock PSPs mimicking glutamatergic input, the most widespread neurotransmitter released onto NG2 cells in the CNS, allowed us to systematically study how different types of VGCs shape synaptic input. Given that NG2 cells show voltage-activated Na⁺ currents (De Biase et al., 2010; Kukley et al., 2010) and that some reports even found regenerative electrical activity when stimulating NG2 cells with rectangular current injections (Chittajallu et al., 2004; Karadottir et al., 2008; Ge et al., 2009), it could be assumed that these fast channels prominently amplify PSPs. By contrast, we observed that the amplitude of mock PSPs increases linearly with input strength (Figure 1C, Figure 2E) due to a fine balance of the fast but low amplitude Na⁺ currents and the much larger but slightly slower A-type K⁺ currents, with the latter counteracting the amplifying effect of the Na⁺ channels. 

While neurons possess a variety of signaling routes, they primarily integrate synaptic input by summation of electrical signals and generation of an action potential when the summed input crosses the threshold. Single action potentials regularly open VGCCs and are associated with Ca²⁺ transients of constant size. Therefore, intracellular Ca²⁺ levels in neurons are proportional to the number of action potentials in a brief burst or to the frequency of action potentials in a long train (Helmchen et al., 1996). NG2 cells also sum incoming electrical synaptic input but, in contrast to neurons and in the absence of action potentials, there is no threshold function for synaptic potentials. The above-described interplay of Na⁺ and K⁺ channels allows the size of PSPs to grow proportionally with input strength (Figure 2E). Similarly, the amplitude of Ca²⁺ signals in NG2 cells gradually increases with input strength (Figure 7F). Thus, in contrast to neurons, NG2 cells use Ca²⁺ levels to gradually encode the level of synaptic activity itself. Both locally induced and global Ca²⁺ signals in NG2 cells decay with a time constant on the order of seconds, while the Ca²⁺ levels in neurons usually return to baseline within several hundred milliseconds. Their low incremental amplitude, their slow return to baseline, and their reliable occurrence even in distal dendrites are characteristics displayed by Ca²⁺ signals in NG2 cells that are key to effectively integrating synaptic input of a wide range of frequencies across their dendritic tree.

Even though the presence of VGCCs in NG2 cells has previously been shown (Haberlandt et al., 2011; Cheli et al., 2014; Larson et al., 2016), their recruitment by synaptic activity has so far remained obscure. Our results demonstrate for the first time that synaptic-like (mock) depolarizations trigger robust and rapid Ca²⁺ signals in the soma and dendrites of NG2 cells. We furthermore provide evidence that the observed Ca²⁺ signals are mediated by low-threshold VGCCs: (1) they were triggered solely by current injection; (2) they started to appear when cells were depolarized beyond −50 to −40 mV; (3) they were blocked by nickel/cadmium and by a cocktail of the R- and T-type channel blockers SNX-482 and TTA-P2; (4) they were potentiated by blocking K⁺ channels and showed a rapid rise, implying that the Ca²⁺ source was active for only a few milliseconds; (5) they were resistant to blocking Na⁺/Ca²⁺ exchangers or Ca²⁺ stores by 10 µM KB-R7943 and thapsigargin, respectively. Thus, the Ca²⁺ signals we recorded in NG2 cells are quite distinct from the store-mediated Ca²⁺ elevations seen in astrocytes, which occur and decay over many seconds. Haberlandt et al. reported that depolarization-induced Ca²⁺ signals in NG2 cells are very sensitive to thapsigargin, seemingly contradicting our findings (Haberlandt et al., 2011). However, the aforementioned study used a much stronger stimulus and depolarized NG2 cells for 100 ms to +20 mV (voltage clamp), and the resulting Ca²⁺ signal showed kinetics an order of magnitude slower than those reported here. Thus, it is conceivable that NG2 cells generally can generate store-dependent Ca²⁺ signals but that store-dependent signaling is not readily recruited by synaptic activity studied here.

The kinetics and magnitude of Ca²⁺ signals in NG2 cell dendrites, induced by local synaptic stimulation or local glutamate uncaging, match those evoked by somatic current injection very well, suggesting that they are also caused by recruitment of VGCCs. Under the conditions used here, mimicking the activity of a very small number of synapses, Ca²⁺ entry through glutamate receptors appears not to be involved because 4-AP boosted the response and because executing the glutamate uncaging protocol alone never triggered a Ca²⁺ response (unpublished observations). However, this does not imply that glutamate receptors cannot be a source of Ca²⁺ entry if more widespread opening of receptors is achieved (see Ge et al., 2006).

Ca²⁺ signals in NG2 cells are restrictively gated by A-type K⁺ channels. This may be one of the reasons why previous publications concluded that synaptically driven Ca²⁺ signals in NG2 cells are unlikely to occur (Velez-Fort et al., 2010; Haberlandt et al., 2011). By which mechanisms do A-type K⁺ channels gate Ca²⁺ signals? Since the voltage-dependence of the Ca²⁺ signal itself remained unchanged after blocking A-type K⁺ channels (Figure 7E), it is very likely that A-type channels under control conditions render synaptic input less effective in causing Ca²⁺ signals by shortening and decreasing the associated synaptic depolarization. In consequence, the impact of synaptic transmission on Ca²⁺ signaling in NG2 cells is dependent on the activity of A-type K⁺ channels. Considering that Ca²⁺ levels have been implicated in the developmental behavior and gene expression of NG2 cells, A-type K⁺ channels are likely to play a key regulatory role in whether and how neuronal transmitter release affects oligodendrogenesis and myelination (Pende et al., 1994; Paez et al., 2009, 2010). The fact that NG2 cells are most responsive to synaptic input from neurons when A-type channels are suppressed raises the question of under which conditions such suppression is achieved. Activity of A-type channels can temporarily be reduced when depolarizations bring them into a state of inactivation. Thus, those patterns of synaptic input that maximize inactivation of A-type channels enhance the impact of axonal activity on NG2 cell behavior. Based on previous reports on the kinetics of A-type K⁺ channels (Steinhauser et al., 1994b), our data allows us to predict that brief trains of action potentials (50–100 ms) should be very effective at inactivating A-type channels and will open the 'calcium-gate' in NG2 cells. Apart from functionally and temporarily reducing A-type currents by inactivation, transcriptional regulation may be another way to control A-type current amplitude in NG2 cells. Indeed, there is evidence for developmental and reactive regulation of K⁺ channels in response to ischemia/injury (Lytle et al., 2009; Pivonkova et al., 2010). Such regulated expression of A-type K⁺ channels may allow NG2 cells to permanently tune their responsiveness to incoming neuronal activity. This adaptation is probably required to match the responsiveness of NG2 cells to patterns of activity, which vary across regional neuronal circuits throughout the brain (Sun and Dietrich, 2013).

When performing glutamate uncaging along NG2 cell dendrites, we regularly observed depolarizations with similar size and kinetics to miniature EPSPs. For the first time, this directly demonstrates that there are many synapse-like clusters of glutamate receptors along many dendrites of NG2 cells. Such widespread responsiveness to glutamate throughout the dendritic tree indicates that NG2 cells may be capable of dendritic computation. Dendritic computation in NG2 cells is further suggested by the fact that local glutamate uncaging boosted the dendritic Ca²⁺ response despite the small amplitude of the associated somatic voltage responses. In particular, the tiny incremental depolarization caused by adding dendritic uncaging (∆ 2.5 mV) is far too small to significantly enhance the recruitment of VGCCs at the soma (Figure 9C). This implies that the dendritic voltage excursion caused by glutamate uncaging is actually substantially larger than that measured at the soma, and that this more pronounced local depolarization recruits Ca²⁺ channels in the dendrite. Considering the thin diameter of NG2 cell dendrites, such severe attenuation of dendritic voltage is not surprising (Chan et al., 2013; Sun and Dietrich, 2013). A larger dendritic voltage excursion is also suggested by the level of the absolute depolarization of the soma (as opposed to the uncaging-induced increment in depolarization considered above), which is clearly below the voltage range typically required to reliably induce Ca²⁺ signals (Figure 7). Further, all Ca²⁺ signals recorded in dendrites – regardless of whether induced by synaptic stimulation, by a train of mock PSPs, by gaussian distributed mock PSPs, or by glutamate uncaging – display a rapid rise, implying that the Ca²⁺ source must be very close to the optical recording site, the dendrite. Finally, Ca²⁺ signals induced in the soma rapidly spread into all dendrites and reached even the distal ends with undiminished amplitudes. This can only be achieved if VGCCs are present all along the dendrites to regenerate the signal. As we did not detect any dendrites that did not generate Ca²⁺ responses, the data indicate that most, if not all, dendrites of NG2 cells are equipped with VGCCs.

There is a similar line of argument for a dendritic localization of A-type channels in NG2 cells. Blocking these channels with 4-AP allowed glutamate uncaging to boost the Ca²⁺ response, i.e. to recruit more Ca²⁺ channels by a stronger depolarization compared to control conditions. However, the resulting depolarization measured at the soma was negligibly increased by the K⁺ channel blocker when compared to control conditions and thus cannot explain additional recruitment of Ca²⁺ channels (Figure 9). Therefore, application of 4-AP must have allowed for a larger depolarization of the local membrane in the dendrite in response to uncaging, which was not visible in the somatic recording. The most parsimonious explanation is that 4-AP was also acting on dendritic A-type K⁺ channels, which renders dendritic glutamate uncaging more effective in generating a larger local depolarization and recruiting dendritic VGCCs. Taken together, our data show that NG2 cells express voltage-dependent Ca²⁺ and K⁺ channels in their dendrites and that these channels actively shape synaptic input as independent processing units.

It has been put forward that synaptic transmission from unmyelinated axons to NG2 cells may play a pivotal role in triggering these responses of NG2 cells (Wake et al., 2011; Hines et al., 2015; Mensch et al., 2015), also reviewed in (Almeida and Lyons, 2014; Petersen and Monk, 2015). Based on our present findings, we propose that synaptically induced Ca²⁺ signaling is well-suited to initiate actions of NG2 cells in the context of activity-dependent myelination. First, intracellular Ca²⁺ levels in NG2 cells allow for optimal temporal and spatial summation of incoming neuronal input and their dependence on the functional state of A-type channels provides certain types of plasticity for synchronized activity. Second, NG2 cells seem to generate Ca²⁺ signals in at least two different spatial domains, which may serve distinct roles in the context of activity-dependent myelination. Dendritic input, possibly even induced by an individual axon, can trigger local Ca²⁺ signals. Such local Ca²⁺ signals could be involved in stabilizing or enhancing contact sites with individual axons (Hines et al., 2015; Mensch et al., 2015) and could enable axon-specific integration and computation. On the other hand, global Ca²⁺ signals, which are induced by a larger number of active synapses throughout the dendritic tree, spread within the NG2 cell and involve the soma, are particularly qualified for triggering global cellular actions, such as division, differentiation, survival or motility of NG2 cells in response to neuronal activity. Further, our finding of local processing of synaptic input raises the question of how local dendritic computation in NG2 cells is relevant for myelination and whether the spatial arrangement of synaptic junctions on NG2 cells affects myelination of the connected axon.

In conclusion, our study shows for the first time that NG2 cells possess a dedicated synaptic integration repertoire, distinct from that of neurons and other types of glial cells, by encoding the number of active synapses in a graded fashion. This integration builds on Ca²⁺ signals in different spatial domains and appears as a specialized type of Ca²⁺ excitability of NG2 cells. Voltage-gated K⁺ channels dampen Ca²⁺ signals but this can be overcome by temporally and spatially coincident synaptic depolarizations. Thus, specific patterns of axonal activity may promote myelination by instructing NG2 cells to interact with axons or generate new oligodendrocytes.

1. 1. Allbritton NL
   2. Meyer T
   3. Stryer L

   (1992) Range of messenger action of calcium ion and inositol 1,4,5-trisphosphate

   Science 258:1812–1815.

   https://doi.org/10.1126/science.1465619

   • Google Scholar
2. 1. Almeida RG
   2. Lyons DA

   (2014) On the resemblance of synapse formation and CNS myelination

   Neuroscience 276:98–108.

   https://doi.org/10.1016/j.neuroscience.2013.08.062

   • Google Scholar
3. 1. Bergles DE
   2. Roberts JD
   3. Somogyi P
   4. Jahr CE

   (2000) Glutamatergic synapses on oligodendrocyte precursor cells in the hippocampus

   Nature 405:187–191.

   https://doi.org/10.1038/35012083

   • Google Scholar
4. 1. Butt AM
   2. Hamilton N
   3. Hubbard P
   4. Pugh M
   5. Ibrahim M

   (2005) Synantocytes: the fifth element

   Journal of Anatomy 207:695–706.

   https://doi.org/10.1111/j.1469-7580.2005.00458.x

   • Google Scholar
5. 1. Chan CF
   2. Kuo TW
   3. Weng JY
   4. Lin YC
   5. Chen TY
   6. Cheng JK
   7. Lien CC

   (2013) Ba2+- and bupivacaine-sensitive background K+ conductances mediate rapid EPSP attenuation in oligodendrocyte precursor cells

   The Journal of Physiology 591:4843–4858.

   https://doi.org/10.1113/jphysiol.2013.257113

   • Google Scholar
6. 1. Cheli VT
   2. Santiago González DA
   3. Spreuer V
   4. Paez PM

   (2015) Voltage-gated Ca2+ entry promotes oligodendrocyte progenitor cell maturation and myelination in vitro

   Experimental Neurology 265:69–83.

   https://doi.org/10.1016/j.expneurol.2014.12.012

   • Google Scholar
7. 1. Chittajallu R
   2. Aguirre A
   3. Gallo V

   (2004) NG2-positive cells in the mouse white and grey matter display distinct physiological properties

   The Journal of Physiology 561:109–122.

   https://doi.org/10.1113/jphysiol.2004.074252

   • Google Scholar
8. 1. Choe W
   2. Messinger RB
   3. Leach E
   4. Eckle VS
   5. Obradovic A
   6. Salajegheh R
   7. Jevtovic-Todorovic V
   8. Todorovic SM

   (2011) TTA-P2 is a potent and selective blocker of T-type calcium channels in rat sensory neurons and a novel antinociceptive agent

   Molecular Pharmacology 80:900–910.

   https://doi.org/10.1124/mol.111.073205

   • Google Scholar
9. 1. Clarke LE
   2. Young KM
   3. Hamilton NB
   4. Li H
   5. Richardson WD
   6. Attwell D

   (2012) Properties and fate of oligodendrocyte progenitor cells in the corpus callosum, motor cortex, and piriform cortex of the mouse

   Journal of Neuroscience 32:8173–8185.

   https://doi.org/10.1523/JNEUROSCI.0928-12.2012

   • Google Scholar
10. 1. Clements JD
    2. Bekkers JM

    (1997) Detection of spontaneous synaptic events with an optimally scaled template

    Biophysical Journal 73:220–229.

    https://doi.org/10.1016/S0006-3495(97)78062-7

    • Google Scholar
11. 1. De Biase LM
    2. Nishiyama A
    3. Bergles DE

    (2010) Excitability and synaptic communication within the oligodendrocyte lineage

    Journal of Neuroscience 30:3600–3611.

    https://doi.org/10.1523/JNEUROSCI.6000-09.2010

    • Google Scholar
12. 1. Gabso M
    2. Neher E
    3. Spira ME

    (1997) Low mobility of the Ca2+ buffers in axons of cultured Aplysia neurons

    Neuron 18:473–481.

    https://doi.org/10.1016/S0896-6273(00)81247-7

    • Google Scholar
13. 1. Gallo V
    2. Zhou JM
    3. McBain CJ
    4. Wright P
    5. Knutson PL
    6. Armstrong RC

    (1996) Oligodendrocyte progenitor cell proliferation and lineage progression are regulated by glutamate receptor-mediated K+ channel block

    Journal of Neuroscience 16:2659–2670.

    https://doi.org/10.1016/0736-5748(96)80263-2

    • Google Scholar
14. 1. Ge WP
    2. Yang XJ
    3. Zhang Z
    4. Wang HK
    5. Shen W
    6. Deng QD
    7. Duan S

    (2006) Long-term potentiation of neuron-glia synapses mediated by Ca2+-permeable AMPA receptors

    Science 312:1533–1537.

    https://doi.org/10.1126/science.1124669

    • Google Scholar
15. 1. Ge WP
    2. Zhou W
    3. Luo Q
    4. Jan LY
    5. Jan YN

    (2009) Dividing glial cells maintain differentiated properties including complex morphology and functional synapses

    PNAS 106:328–333.

    https://doi.org/10.1073/pnas.0811353106

    • Google Scholar
16. 1. Ghiani CA
    2. Eisen AM
    3. Yuan X
    4. DePinho RA
    5. McBain CJ
    6. Gallo V

    (1999a)

    Neurotransmitter receptor activation triggers p27(Kip1 )and p21(CIP1) accumulation and G1 cell cycle arrest in oligodendrocyte progenitors

    Development 126:1077–1090.

    • Google Scholar
17. 1. Ghiani CA
    2. Yuan X
    3. Eisen AM
    4. Knutson PL
    5. DePinho RA
    6. McBain CJ
    7. Gallo V

    (1999b)

    Voltage-activated K+ channels and membrane depolarization regulate accumulation of the cyclin-dependent kinase inhibitors p27(Kip1) and p21(CIP1) in glial progenitor cells

    Journal of Neuroscience 19:5380–5392.

    • Google Scholar
18. 1. Haberlandt C
    2. Derouiche A
    3. Wyczynski A
    4. Haseleu J
    5. Pohle J
    6. Karram K
    7. Trotter J
    8. Seifert G
    9. Frotscher M
    10. Steinhäuser C
    11. Jabs R

    (2011) Gray matter NG2 cells display multiple Ca2+-signaling pathways and highly motile processes

    PLoS One 6:e17575.

    https://doi.org/10.1371/journal.pone.0017575

    • Google Scholar
19. 1. Helmchen F
    2. Imoto K
    3. Sakmann B

    (1996) Ca2+ buffering and action potential-evoked Ca2+ signaling in dendrites of pyramidal neurons

    Biophysical Journal 70:1069–1081.

    https://doi.org/10.1016/S0006-3495(96)79653-4

    • Google Scholar
20. 1. Hines JH
    2. Ravanelli AM
    3. Schwindt R
    4. Scott EK
    5. Appel B

    (2015) Neuronal activity biases axon selection for myelination in vivo

    Nature Neuroscience 18:683–689.

    https://doi.org/10.1038/nn.3992

    • Google Scholar
21. 1. Iwamoto T
    2. Watano T
    3. Shigekawa M

    (1996) A novel isothiourea derivative selectively inhibits the reverse mode of Na+/Ca2+ exchange in cells expressing NCX1

    The Journal of Biological Chemistry 271:22391–22397.

    https://doi.org/10.1074/jbc.271.37.22391

    • Google Scholar
22. 1. Jabs R
    2. Pivneva T
    3. Hüttmann K
    4. Wyczynski A
    5. Nolte C
    6. Kettenmann H
    7. Steinhäuser C

    (2005) Synaptic transmission onto hippocampal glial cells with hGFAP promoter activity

    Journal of Cell Science 118:3791–3803.

    https://doi.org/10.1242/jcs.02515

    • Google Scholar
23. 1. Káradóttir R
    2. Hamilton NB
    3. Bakiri Y
    4. Attwell D

    (2008) Spiking and nonspiking classes of oligodendrocyte precursor glia in CNS white matter

    Nature Neuroscience 11:450–456.

    https://doi.org/10.1038/nn2060

    • Google Scholar
24. 1. Knutson P
    2. Ghiani CA
    3. Zhou JM
    4. Gallo V
    5. McBain CJ

    (1997)

    K+ channel expression and cell proliferation are regulated by intracellular sodium and membrane depolarization in oligodendrocyte progenitor cells

    Journal of Neuroscience 17:2669–2682.

    • Google Scholar
25. 1. Koudelka S
    2. Voas MG
    3. Almeida RG
    4. Baraban M
    5. Soetaert J
    6. Meyer MP
    7. Talbot WS
    8. Lyons DA

    (2016) Individual neuronal subtypes exhibit diversity in CNS myelination mediated by synaptic vesicle release

    Current Biology 26:1447–1455.

    https://doi.org/10.1016/j.cub.2016.03.070

    • Google Scholar
26. 1. Kukley M
    2. Capetillo-Zarate E
    3. Dietrich D

    (2007) Vesicular glutamate release from axons in white matter

    Nature Neuroscience 10:311–320.

    https://doi.org/10.1038/nn1850

    • Google Scholar
27. 1. Kukley M
    2. Kiladze M
    3. Tognatta R
    4. Hans M
    5. Swandulla D
    6. Schramm J
    7. Dietrich D

    (2008) Glial cells are born with synapses

    The FASEB Journal 22:2957–2969.

    https://doi.org/10.1096/fj.07-090985

    • Google Scholar
28. 1. Kukley M
    2. Nishiyama A
    3. Dietrich D

    (2010) The fate of synaptic input to NG2 glial cells: neurons specifically downregulate transmitter release onto differentiating oligodendroglial cells

    Journal of Neuroscience 30:8320–8331.

    https://doi.org/10.1523/JNEUROSCI.0854-10.2010

    • Google Scholar
29. 1. Larson VA
    2. Zhang Y
    3. Bergles DE

    (2016) Electrophysiological properties of NG2(+) cells: Matching physiological studies with gene expression profiles

    Brain Research 1638:138–160.

    https://doi.org/10.1016/j.brainres.2015.09.010

    • Google Scholar
30. 1. Lin SC
    2. Bergles DE

    (2004) Synaptic signaling between GABAergic interneurons and oligodendrocyte precursor cells in the hippocampus

    Nature Neuroscience 7:24–32.

    https://doi.org/10.1038/nn1162

    • Google Scholar
31. 1. Lin SC
    2. Huck JH
    3. Roberts JD
    4. Macklin WB
    5. Somogyi P
    6. Bergles DE

    (2005) Climbing fiber innervation of NG2-expressing glia in the mammalian cerebellum

    Neuron 46:773–785.

    https://doi.org/10.1016/j.neuron.2005.04.025

    • Google Scholar
32. 1. Lytle JM
    2. Chittajallu R
    3. Wrathall JR
    4. Gallo V

    (2009) NG2 cell response in the CNP-EGFP mouse after contusive spinal cord injury

    Glia 57:270–285.

    https://doi.org/10.1002/glia.20755

    • Google Scholar
33. 1. Matthews EA
    2. Schoch S
    3. Dietrich D

    (2013) Tuning local calcium availability: cell-type-specific immobile calcium buffer capacity in hippocampal neurons

    Journal of Neuroscience 33:14431–14445.

    https://doi.org/10.1523/JNEUROSCI.4118-12.2013

    • Google Scholar
34. 1. Mensch S
    2. Baraban M
    3. Almeida R
    4. Czopka T
    5. Ausborn J
    6. El Manira A
    7. Lyons DA

    (2015) Synaptic vesicle release regulates myelin sheath number of individual oligodendrocytes in vivo

    Nature Neuroscience 18:628–630.

    https://doi.org/10.1038/nn.3991

    • Google Scholar
35. 1. Newcomb R
    2. Szoke B
    3. Palma A
    4. Wang G
    5. Chen X
    6. Hopkins W
    7. Cong R
    8. Miller J
    9. Urge L
    10. Tarczy-Hornoch K
    11. Loo JA
    12. Dooley DJ
    13. Nadasdi L
    14. Tsien RW
    15. Lemos J
    16. Miljanich G

    (1998) Selective peptide antagonist of the class E calcium channel from the venom of the tarantula Hysterocrates gigas

    Biochemistry 37:15353–15362.

    https://doi.org/10.1021/bi981255g

    • Google Scholar
36. 1. Nishiyama A
    2. Chang A
    3. Trapp BD

    (1999) NG2+ glial cells: a novel glial cell population in the adult brain

    Journal of Neuropathology and Experimental Neurology 58:1113–1124.

    https://doi.org/10.1097/00005072-199911000-00001

    • Google Scholar
37. 1. Paez PM
    2. Fulton D
    3. Colwell CS
    4. Campagnoni AT

    (2009) Voltage-operated Ca(2+) and Na(+) channels in the oligodendrocyte lineage

    Journal of Neuroscience Research 87:3259–3266.

    https://doi.org/10.1002/jnr.21938

    • Google Scholar
38. 1. Paez PM
    2. Fulton DJ
    3. Spreur V
    4. Handley V
    5. Campagnoni AT

    (2010) Multiple kinase pathways regulate voltage-dependent Ca2+ influx and migration in oligodendrocyte precursor cells

    Journal of Neuroscience 30:6422–6433.

    https://doi.org/10.1523/JNEUROSCI.5086-09.2010

    • Google Scholar
39. 1. Passlick S
    2. Trotter J
    3. Seifert G
    4. Steinhäuser C
    5. Jabs R

    (2016) The NG2 protein is not required for glutamatergic neuron-NG2 cell synaptic signaling

    Cerebral Cortex 26:51–57.

    https://doi.org/10.1093/cercor/bhu171

    • Google Scholar
40. 1. Pende M
    2. Holtzclaw LA
    3. Curtis JL
    4. Russell JT
    5. Gallo V

    (1994) Glutamate regulates intracellular calcium and gene expression in oligodendrocyte progenitors through the activation of DL-alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors

    PNAS 91:3215–3219.

    https://doi.org/10.1073/pnas.91.8.3215

    • Google Scholar
41. 1. Petersen SC
    2. Monk KR

    (2015) Neurobiology: Myelin goes where the action is

    Current Biology 25:R562–565.

    https://doi.org/10.1016/j.cub.2015.04.054

    • Google Scholar
42. 1. Pivonkova H
    2. Benesova J
    3. Butenko O
    4. Chvatal A
    5. Anderova M

    (2010) Impact of global cerebral ischemia on K+ channel expression and membrane properties of glial cells in the rat hippocampus

    Neurochemistry International 57:783–794.

    https://doi.org/10.1016/j.neuint.2010.08.016

    • Google Scholar
43. 1. Steinhäser C
    2. Jabs R
    3. Kettenmann H

    (1994a) Properties of GABA and glutamate responses in identified glial cells of the mouse hippocampal slice

    Hippocampus 4:19–35.

    https://doi.org/10.1002/hipo.450040105

    • Google Scholar
44. 1. Steinhäuser C
    2. Kressin K
    3. Kuprijanova E
    4. Weber M
    5. Seifert G

    (1994b) Properties of voltage-activated Na+ and K+ currents in mouse hippocampal glial cells in situ and after acute isolation from tissue slices

    Pflügers Archiv European Journal of Physiology 428:610–620.

    https://doi.org/10.1007/BF00374585

    • Google Scholar
45. 1. Sun W
    2. Dietrich D

    (2013) Synaptic integration by NG2 cells

    Frontiers in Cellular Neuroscience 7:255.

    https://doi.org/10.3389/fncel.2013.00255

    • Google Scholar
46. 1. Tong XP
    2. Li XY
    3. Zhou B
    4. Shen W
    5. Zhang ZJ
    6. Xu TL
    7. Duan S

    (2009) Ca(2+) signaling evoked by activation of Na(+) channels and Na(+)/Ca(2+) exchangers is required for GABA-induced NG2 cell migration

    The Journal of Cell Biology 186:113–128.

    https://doi.org/10.1083/jcb.200811071

    • Google Scholar
47. 1. Trotter J
    2. Karram K
    3. Nishiyama A

    (2010) NG2 cells: Properties, progeny and origin

    Brain Research Reviews 63:72–82.

    https://doi.org/10.1016/j.brainresrev.2009.12.006

    • Google Scholar
48. 1. Vélez-Fort M
    2. Maldonado PP
    3. Butt AM
    4. Audinat E
    5. Angulo MC

    (2010) Postnatal switch from synaptic to extrasynaptic transmission between interneurons and NG2 cells

    Journal of Neuroscience 30:6921–6929.

    https://doi.org/10.1523/JNEUROSCI.0238-10.2010

    • Google Scholar
49. 1. Wagner J
    2. Keizer J

    (1994) Effects of rapid buffers on Ca2+ diffusion and Ca2+ oscillations

    Biophysical Journal 67:447–456.

    https://doi.org/10.1016/S0006-3495(94)80500-4

    • Google Scholar
50. 1. Wake H
    2. Lee PR
    3. Fields RD

    (2011) Control of local protein synthesis and initial events in myelination by action potentials

    Science 333:1647–1651.

    https://doi.org/10.1126/science.1206998

    • Google Scholar
51. 1. Zhou Z
    2. Neher E

    (1993) Mobile and immobile calcium buffers in bovine adrenal chromaffin cells

    The Journal of Physiology 469:245–273.

    https://doi.org/10.1113/jphysiol.1993.sp019813

    • Google Scholar
52. 1. Zhu X
    2. Bergles DE
    3. Nishiyama A

    (2008) NG2 cells generate both oligodendrocytes and gray matter astrocytes

    Development 135:145–157.

    https://doi.org/10.1242/dev.004895

    • Google Scholar
53. 1. Ziskin JL
    2. Nishiyama A
    3. Rubio M
    4. Fukaya M
    5. Bergles DE

    (2007) Vesicular release of glutamate from unmyelinated axons in white matter

    Nature Neuroscience 10:321–330.

    https://doi.org/10.1038/nn1854

    • Google Scholar

Author details

1. Wenjing Sun

   Department of Neurosurgery, University Clinic Bonn, Bonn, Germany

   Contribution

   WS, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article

   For correspondence

   wsun@uni-bonn.de

   Competing interests

   The authors declare that no competing interests exist.

   "This ORCID iD identifies the author of this article:" 0000-0002-0905-7420

2. Elizabeth A Matthews

   Department of Neurosurgery, University Clinic Bonn, Bonn, Germany

   Contribution

   EAM, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

3. Vicky Nicolas

   Department of Neurosurgery, University Clinic Bonn, Bonn, Germany

   Contribution

   VN, Acquisition of data, Analysis and interpretation of data

   Competing interests

   The authors declare that no competing interests exist.

4. Susanne Schoch

   Department of Neuropathology, University Clinic Bonn, Bonn, Germany

   Contribution

   SS, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

5. Dirk Dietrich

   Department of Neurosurgery, University Clinic Bonn, Bonn, Germany

   Contribution

   DD, Conception and design, Analysis and interpretation of data, Drafting or revising the article

   For correspondence

   dirk.dietrich@uni-bonn.de

   Competing interests

   The authors declare that no competing interests exist.

   "This ORCID iD identifies the author of this article:" 0000-0002-4307-2448

• 2,261

  views

• 489

  downloads

• 46

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.