Application of an optogenetic byway for perturbing neuronal activity via glial photostimulation

Abstract

Dynamic activity of glia has repeatedly been demonstrated, but if such activity is independent from neuronal activity, glia would not have any role in the information processing in the brain or in the generation of animal behavior. Evidence for neurons communicating with glia is solid, but the signaling pathway leading back from glial-to-neuronal activity was often difficult to study. Here, we introduced a transgenic mouse line in which channelrhodopsin-2, a light-gated cation channel, was expressed in astrocytes. Selective photostimulation of these astrocytes in vivo triggered neuronal activation. Using slice preparations, we show that glial photostimulation leads to release of glutamate, which was sufficient to activate AMPA receptors on Purkinje cells and to induce long-term depression of parallel fiber-to-Purkinje cell synapses through activation of metabotropic glutamate receptors. In contrast to neuronal synaptic vesicular release, glial activation likely causes preferential activation of extrasynaptic receptors that appose glial membrane. Finally, we show that neuronal activation by glial stimulation can lead to perturbation of cerebellar modulated motor behavior. These findings demonstrate that glia can modulate the tone of neuronal activity and behavior. This animal model is expected to be a potentially powerful approach to study the role of glia in brain function.

Much of the brain function is often perceived as being achieved solely by neuronal activity in contemporary science. However, the brain is actually occupied mostly by another type of cells, the glia. Their role was classically considered to be restricted to the maintenance of neuronal survival and functioning, such as trophic support, uptake of glutamate, and removal of K⁺ from the extracellular space (1), but rapid and dynamic activity of glia correlated with behavior and cognitive functions has recently been shown (2, 3). However, demonstration of the causal link between glial activity and neuronal activity is required to accept the possibility that glia can indeed participate in the information processing in the brain.

In the cerebellum, there are direct and rapid mechanisms for neurons to communicate with glia. Stimulation of the parallel fibers (PFs), the axons of granule cell neurons (GCs), triggers glutamate release from synaptic sites and additional ectopic sites, evoking an inward current in the adjacent major cerebellar astrocytes, the Bergmann glial cells (BGs), with components due to both activation of Ca²⁺-permeable AMPA receptors (AMPARs) and electrogenic uptake of glutamate (4–6). Computational modeling has suggested that such synaptically induced activity would cause strikingly large depolarization (up to 40 mV) of the BG membrane processes local to the activated sites (7). Furthermore, PF burst activity elicits transient and local Ca²⁺ increases in BG processes via AMPA and ATP receptor-mediated mechanisms (2, 8–11). Despite the accumulating evidence of neuron-to-glia communication in the cerebellum, if the communication between these cells is a one-way relationship, the glia would only be a “listener” and it would not be able to actively participate in a layer of information processing. The close anatomical apposition of BGs to neighboring Purkinje cells (PCs) suggests the possibility that BGs could “speak up” and interact functionally with PCs (Fig. S1). Release of transmitter from glia (gliotransmitter) has been demonstrated in several brain areas (12–16) and we sought to address whether such gliotransmitter could mediate glia-to-neuron communication in the cerebellum.

To deal with this question, we need a method to selectively activate the glia. Conventional electrophysiological techniques, such as extracellular electrical stimulation, inevitably stimulate both neuron and glia. Thus, gliotransmitter release could have been unintentionally evoked in previous studies but its effect overlooked. Direct intracellular stimulation of glia is also difficult as the input resistance of these cells is low and the depolarization generated by the somatic electrode will not propagate through the complex glial structures. We used optogenetics to circumvent this problem (17). An important advantage of the optogenetic tools, such as channelrhodopsin-2 (ChR2), a light-gated nonselective cation channel, is its cell type specificity. Although glia-specific expression of ChR2 has been achieved using viral vectors in previous studies (12, 18), we took an approach to generate a transgenic mouse line in which highly-sensitive mutant [C128S (19)] of ChR2 was selectively expressed in astrocytes including BGs using the tetracycline transactivator (tTA)–tet operator (tetO) system (20). In contrast to viral vectors, the transgenic lines allow stable and reproducible expression of transgenes in animals. In this study, we demonstrate the presence of glia-to-neuron communication in the transgenic animals both in vitro and in vivo. These transgenic tools would be indispensable for investigating intercellular signaling in the brain.

Results

Optogenetic Stimulation of Bergmann Glial Cells Drives Neuron Activity.

To selectively stimulate the glia, we generated two lines of mice (20); one in which the expression of tetracycline-controlled tTA is driven by megalencephalic leukoencephalopathy with subcortical cysts 1 (Mlc1)-promoter, which has been proven to be astrocyte specific (21), and the other in which tetO is connected to the expression of C128S mutant of ChR2 (19) fused with enhanced yellow fluorescent protein (EYFP). By obtaining bigenic [Mlc1-tTA::tetO-ChR2(C128S)-EYFP] mice, strong EYFP expression was found in BGs in the cerebellum (Fig. 1A). A point mutation (C128S) of the ChR2 causes prominent increases in photosensitivity as well as alteration in the photocycle, enabling a step-like control of the photocurrent by blue and yellow light (19). To validate the selective expression of ChR2(C128S) in glial cells, whole-cell patch-clamp recordings were performed in acute cerebellar slices in the presence of a mixture of pharmacological agents that silence neuronal activity (Fig. 1B). Current responses to photostimulation were detected only in BGs (n = 24), but not in PCs (n = 10), GCs (n = 10), and stellate/basket cells (SC/BCs) (n = 11). Furthermore, simultaneous in situ hybridization for tTA and immunohistochemistry for brain lipid-binding protein (BLBP) confirmed the specificity of transgene expression in BGs (Fig. S2A) with modest expression in the protoplasmic astrocytes in the GC layer (GL) (Fig. S2B). These results demonstrate that, in our transgenic mice, ChR2 expression is restricted to glial cells.

Fig. 1.

Glia photostimulation in vivo leads to neuronal activity. (A) Crossing two lines of mice [Mlc1-tTA and tetO-ChR2(C128S)-EYFP] yielded mice in which ChR2(C128S)-EYFP was selectively expressed in Mlc1-positive astrocytes including cerebellar BGs. The lower panels show a differential interference contrast (DIC) image and a confocal stack image of EYFP in an identical region of the parasagittal section of the cerebellum. (B) Typical current traces in response to photostimulation recorded from a BG, PC, GC, and SC/BC (similar results from 24, 10, 10, and 11 cells, respectively). Recordings were performed in the presence of TTX, Cd²⁺, PIC, and NBQX to silence neuronal activity and extract ChR2 activity. (C and D) Sequence of light applied in vivo via optical fiber placed on top of the skull. (E) c-fos mRNA (10 min after the photostimulation) was increased in PCs (arrowheads), GCs, and SC/BCs (n = 4 animals).

In BGs, photostimulation with a blue light produced an inwardly developing current that was terminated partially by a flash of yellow light; however, the remaining current required >30 s to return to baseline (Fig. 1B and Fig. S3A). We were uncertain about the cause of this slowly developing current; therefore, we generated parvalbumin (PV)-tTA::tetO-ChR2(C128S)-EYFP mice in which EYFP expression was confirmed in PCs (Fig. S3B). In this mouse line, delivery of the blue light produced a constant inward current in PCs, which was rapidly terminated with the yellow light pulse. We assumed that the cause of the additional current in BGs is the high conductance of BGs to K⁺ making them sensitive to changes in extracellular K⁺ concentrations (22). As expected, 5 mM Ba²⁺ and 5 mM 4-aminopyridine (4-AP), K⁺ channel blockers, inhibited a large fraction of the slow component, whereas the fast component remained intact (Fig. S3 C and D). These results indicate that the fast component is likely attributable to ChR2 opening itself and the slow component to extracellular K⁺ change. The size of both of these components developed in an age-dependent manner (Fig. S3 E–G).

Our aim was to assess whether glia-to-neuron communication can be elicited by photostimulation of glia. To evaluate this in vivo, an optic fiber was placed just on the skull above the cerebellum of the transgenic mouse (Fig. 1C). While the mouse moved freely in a home cage, a single photostimulation (200-ms blue and 200-ms yellow with an interval of 500 ms) was applied (Fig. 1D). After killing the mouse, in situ hybridization for c-fos mRNA, an immediate-early gene as a marker of neuronal activation, was performed (Fig. 1E). At the region directly beneath the optic fiber, strong c-fos expression was observed in ∼50% of the PCs and SC/BCs, and a subset of GCs (Fig. S2C). Expression of c-fos was not observed in nonphotostimulated ChR2(+) mice or in photostimulated ChR2(−) mice (Fig. S2 D and E). The microglial cell distribution and morphology appeared normal (Fig. S2F).

Identity of the Gliotransmitter and Mechanism of Release.

Even though only the glia was photostimulated, in the absence of pharmacological agents, c-fos was elevated in neurons and a burst of firing activity could be elicited in PCs in slices (Fig. S4A). Although elevated K⁺ could be one of the mechanisms underlying neuronal excitation, we examined whether any other substance is released following glia photostimulation. The receptors expressed on PC membrane were used as biosensors to detect the released substance in acute cerebellar slices. Neuronal transmitter release was inhibited by applying 0.5 µM tetrodotoxin (TTX) and 100 µM Cd²⁺, to block action potentials and voltage-gated Ca²⁺ channels, respectively, and 100 µM pictrotoxin (PIC) was applied to block GABA_A receptors. Glia photostimulation in slices from Mlc1-tTA::tetO-ChR2(C128S) mice produced an inward current in the recorded PC (Fig. 2A), suggesting that the receptors on PCs were activated. In control condition, we noticed that glia-photostimulated PC currents declined with time during repetitive photostimulation (Fig. S5C). The direct rundown of ChR2 activation was also observed in PCs from PV-tTA::tetO-ChR2(C128S) mice and in BGs from Mlc1-tTA::tetO-ChR2(C128S) mice (Fig. S5 A and B). These results imply that the rundown of the glia-photostimulated PC currents is, at least in part, due to the rundown of ChR2 activation. The following data were normalized to the amplitudes of currents in PCs recorded in the presence of TTX and Cd²⁺ at each corresponding period after establishment of the whole-cell recordings.

Fig. 2.

Glia photostimulation triggers glutamate release. Glia-photostimulated PC currents (Left) and time course of the amplitude change (Right) before and after bath application of drugs (drug application from time = 0). Amplitudes of the current are plotted against time with the amplitude normalized to account for the rundown as described in Fig. S5C. (A) The PC current was completely eliminated by GYKI53655 (n = 5; P < 0.001 0–2 min before versus 10–14 min after drug application, paired t test). (B) Inhibition of glutamate transporter by TBOA augmented the PC current (n = 5). (C) Inhibition of anion channels with DIDS abolished the PC current (Upper; n = 6, P < 0.001), whereas the BG photocurrent per se was not significantly altered by DIDS (Lower; n = 6, P > 0.05).

We next sought to identify the transmitter. Because glia-photostimulated PC currents remained intact in the presence of PIC (Fig. S4B), GABA is not likely to be the major transmitter (23). Application of AMPAR antagonist, 10 µM 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide (NBQX) (Fig. S4C) or 100 µM GYKI53655 (P < 0.001, n = 6; Fig. 2A), entirely abolished the glia-photostimulated PC current, whereas 20 µM cyclothiazide (CTZ), an AMPAR desensitization inhibitor, effectively amplified the current (Fig. S4D). If the glutamate concentration rise is slow, the AMPARs would become desensitized, whereas CTZ would be able to cancel this desensitizing effect. The kinetics of the slowly developing PC current was mimicked by local pressure injection (puff) of exogenous glutamate (Fig. S4E). Addition of 100 µM dl-threo-β-benzyloxyaspartic acid (TBOA), glutamate transporter blocker, caused a significant increase in the amplitude of the glia-photostimulated PC current (n = 5; Fig. 2B). Glutamate transporter blockade likely increased the effective concentration of glutamate reaching the AMPARs. Although BG-AMPARs are known to rapidly desensitize (6), in the presence of CTZ and TBOA, the released glutamate acting onto BG-AMPARs could be detected (Fig. S4F). Taken together, these lines of experiments demonstrate that glia photostimulation causes releases of glutamate, and the released glutamate reaches the low-affinity PC AMPARs [EC₅₀ = 432 µM (24)] at a sufficient concentration.

The above pathway is the straightforward explanation of our results; however, as K⁺ is also likely to be released upon glia photostimulation (Fig. S3 C and D), alternative pathways leading from glia photostimulation to PC current could be speculated. First, it is possible that [K⁺]_o increase causes depolarization of PCs and results in release of glutamate from PCs (25), even though the recordings were made in the presence of TTX and Cd²⁺. Direct depolarization of PCs can be accomplished by ChR2-mediated excitation using PV-tTA::tetO-ChR2(C128S) mice (Fig. S6A). Although multiple neighboring PCs were likely to be depolarized by PC photostimulation, currents evoked in the recorded PC were unaffected by NBQX, suggesting that depolarization of neighboring PCs does not cause glutamate release detectable by PC AMPARs. Second, it is possible that glia is releasing glutamate but the released glutamate acts on neighboring cells other than the PCs and the depolarization of these neighboring cells causes [K⁺]_o increase, which is detected as currents in the recorded PCs. It is known that Ca²⁺-permeable (GluR2-lacking) AMPARs are not expressed on PCs but are expressed on BGs and SC/BCs and these receptors could be selectively blocked by 1-naphthylacetyl spermine (NASPM). The glia-photostimulated PC current was insensitive to 100 µM NASPM (Fig. S6B); thus, this alternative pathway mediated by Ca²⁺-permeable AMPARs is unlikely.

Several pathways for glutamate release from glial cells are plausible (26): (i) reversal of uptake by Na⁺-dependent glutamate transporters, (ii) functional unpaired connexin hemichannels, (iii) volume-sensitive anion channel opening, and (iv) Ca²⁺-dependent exocytosis. It has been suggested that the reversal of glutamate transporter can be blocked by extracellular TBOA (27); however, TBOA failed to inhibit glia-photostimulated PC current but rather enhanced it (Fig. 2B). In the presence of 100 µM carbenoxolone, one of the most widely used gap junction/hemichannel blockers, the current response remained normal (Fig. S7A), ruling out the contribution of hemichannel opening. Blocking swelling activated anion channels with 100 µM 5-nitro-2-(3-phenylpropyl amino)benzoic acid (NPPB) did not alter the glia-photostimulated PC current (23) (Fig. S7B). 4,4'-Diisothiocyanostilbene- 2,2'-disulfonic acid (DIDS) is another type of anion channel blocker, which has been shown to block glutamate release from astrocytes (28). DIDS (1 mM) abolished the glia-photostimulated PC current (P < 0.001, n = 6; Fig. 2C, Upper), whereas the BG photocurrent remained intact (P > 0.05, n = 5; Fig. 2C, Lower). No change in the current responses of PCs to exogenous puff application of glutamate was observed in the presence of DIDS (Fig. S7C). These results suggest that glutamate release by glia photostimulation is likely via anion channels not sensitive to NPPB but sensitive to DIDS (29).

Glial Activation Induces Long-Term Depression of PF–PC Synapses.

As BG processes are closely situated to the PFs in the molecular layer (ML), it is possible that glia photostimulation induces activation of the PFs. In the absence of TTX and Cd²⁺, such interaction could lead to neuronal glutamate release from PF synapses, which could lead to the amplification of the glia-photostimulated PC current. To examine this possibility, photostimulation was restricted to the ML to selectively activate the BG processes (Fig. 3A, Left). The PC currents produced by the ML photostimulation were unaffected by TTX application (P > 0.05, n = 5; Fig. 3A, Upper). This result indicates that the activation of BG processes alone is sufficient to trigger glia-photostimulated PC current and it is unlikely that neuronal glutamate release is triggered. In contrast, photostimulation of the region including both ML and GL evoked PC current, which was significantly reduced by TTX treatment (P < 0.01, n = 5; Fig. 3A, Lower). Because weak EYFP (Fig. 1A) and tTA (Fig. S2B) signals were detected in the GL, protoplasmic astrocytes in the GL could also be photostimulated as well as the BGs, although the EYFP signal per unit area was 3.3-fold higher in the ML compared with the GL. This result suggests that wide-field photostimulation not only causes glutamate release from BGs but also causes GC firing via activation of GL astrocytes in the absence of TTX and Cd²⁺. This result accounts for the observation that not only PCs but also GCs were positive for c-fos after the photostimulation in vivo (Fig. 1E).

Fig. 3.

Synaptic plasticity induced by glia photostimulation. (A) (Upper) When photostimulation was aimed specifically to the ML, no significant change in the amplitudes of glia-photostimulated PC current was observed after application of TTX (n = 5 cells; P > 0.05, paired t test). (Lower) PC currents evoked by photostimulation of a region including both ML and GL were significantly reduced by TTX (n = 5 cells; P < 0.001, paired t test). EYFP was bleached after the recording to confirm the location of photostimulation. (B) Electrically stimulated PF–PC EPSCs in response to paired-pulse protocol (100-ms interval) before and after a single ML+GL photostimulation. LTD of the PF–PC EPSC was observed. (C) Time course of the first PF–PC EPSC amplitude (ML+GL stim, n = 5; LY367385, n = 5; *P < 0.05, Student t test). Photostimulation was applied at time 0. (D and E) Summary of the changes in the normalized EPSC amplitude (D) and the mean paired-pulse ratio (E) before and 20–25 min after the glia photostimulation. Photostimulation of ML alone was also sufficient to produce LTD (n = 7). All recordings were done in the presence of PIC.

The released glutamate can activate not only AMPARs but also metabotropic glutamate receptors type 1 (mGluR1), which are expressed on the peripheral of the PC spines and dendrites (30), but have not been detected on cerebellar glial cell membrane (31). Activation of mGluR1 is essential for inducing synaptic plasticity at PF–PC synapses and for coordinated motor behavior (32). BG processes are situated immediately adjacent to the mGluR1 on PCs, which raises a possibility that glutamate released from glia photostimulation could induce changes in neurotransmission from PFs to PCs. To test this, we examined the effect of the glia photostimulation on PF–PC excitatory postsynaptic currents (EPSCs) (Fig. 3B). A single 10-s photostimulation to the ML+GL region induced long-term depression (LTD) of PF–PC EPSCs (P < 0.05, n = 5; Fig. 3 C and D). Application of an mGluR1 antagonist, 100 µM LY367385, blocked the LTD (P > 0.05, n = 5; Fig. 3D). No change in the input resistance of PCs was detected after photostimulation (before, 100.6 ± 10.5 MΩ; after, 105.6 ± 8.9 MΩ; P > 0.05, paired t test). The photostimulation did not significantly alter the EPSC paired-pulse ratio (P > 0.05, n = 5; Fig. 3E) nor the number of fibers activated by the electrical stimulation, as examined by fiber volley recordings (Fig. S8A), implying that presynaptic components were not affected by the photostimulation. As we have shown that wide-field ML+GL photostimulation likely activates GCs via activation of GL astrocytes, it is possible that GC firing subsequent to glia photostimulation alone was sufficient to induce LTD. To exclude this possibility, photostimulation was restricted to the ML. As shown above, neuronal glutamate release from PFs was not likely to be induced by this photostimulation (Fig. 3A). LTD was induced to a similar level as with the ML+GL photostimulation (Fig. 3D). This suggests that glutamate release from glia was the underlying mechanism of glia-photostimulated LTD. However, it still remained possible that glia photostimulation caused depolarization of the recorded and the surrounding PCs, which could somehow lead to LTD. To test this possibility, direct depolarization of PC was induced by using PV-tTA::tetO-ChR2(C128S) mice. No LTD was triggered by the selective PC photostimulation (Fig. S8B). The above results demonstrate that glial release of glutamate not only can activate PC AMPARs but also can activate mGluR1 and initiate pathways leading to LTD.

Brief Photostimulation Triggered Glutamate Release.

Photostimulation of BGs for 10 s resulted in the depolarization of the BGs by 10.0 ± 0.7 mV (Fig. S9A), which was comparable to that induced by strong electrical stimulation in the MLs (100 V × 20; Fig. S9D). Using more moderate stimulation (60 V × 10), typically used in acute slice experiment for studying burst activity of PF–PC synapses, BGs were depolarized by 3.6 ± 0.5 mV (n = 6; Fig. 4A). Photostimulation with short light pulses (200-ms blue and 200-ms yellow with an interval of 500 ms) evoked comparable depolarization of the BGs by 2.7 ± 0.5 mV (n = 6; Fig. 4A). This same light pattern was used in the in vivo experiments (Fig. 1D).

Fig. 4.

Brief photostimulation was sufficient in inducing glutamate release. (A) A typical electrical stimulation used in acute slice experiments depolarized BGs. Short photostimulation (200-ms blue and 200-ms yellow light with an interval of 500 ms in the absence of any drugs) produced similar depolarization. (B) Glia-photostimulated PC currents in response to various durations of blue and fixed duration (200 ms) of yellow light pulses. Glutamate release was detected with pulses of blue light as short as 100 ms. These recordings were made under the condition that detection sensitivity of AMPARs on PCs was maximally amplified by coapplication of CTZ and TBOA (plus TTX, Cd²⁺). (C) Summary of the amplitudes of the PC currents plotted as a function of duration of blue light (control, filled circles, n = 5; CTZ and TBOA, open circles, n = 5). All recordings were done in the presence of PIC.

To evaluate the minimum amount of glia photostimulation required for detecting glutamate release, CTZ and TBOA were coapplied to maximally increase the sensitivity of PC AMPARs to glutamate. In this condition, a 100-ms blue pulse was sufficient to evoke detectable current response in PCs (Fig. 4B). Such short light pulse evoked depolarization of BGs by 1.4 ± 0.2 mV (Fig. S9A). This depolarization was comparable to the depolarization evoked by paired-pulse electrical stimulation of PFs (60 V×2; Fig. S9D). This implies that such weak stimuli are sufficient to evoke glutamate release from glia. The amplitude of the glia-photostimulated PC current increased proportionally with the duration of blue light application (Fig. 4C).

Glia Activation Drives Behavioral Changes.

We next aimed to evaluate whether signals initiated from the glia could sufficiently propagate through the neuronal network and ultimately drive behavioral responses in vivo. For this purpose, we used the moderate light sequence that produced depolarization of BGs to comparable levels as with electrical PF burst stimulation (Fig. 4A). Photostimulation was applied through optical fibers inserted close to the cerebellar flocculus region (Fig. 5A), and an induction of c-fos mRNA in a subset of PCs, GCs, and SC/BCs was observed by the glia photostimulation (Fig. 5B). According to slice experiments, the mechanisms underlying the glia-photostimulated neural activity are likely due to K⁺ efflux and glutamate release. In this condition, we assessed the effect of glia photostimulation on the eye movement (Fig. 5C). Amplitude increase of the horizontal optokinetic reflex (HOKR) upon repetitive visual stimuli has been shown to be controlled by the activity of the flocculus (33, 34). Glia photostimulation was done while this HOKR was induced and perturbation of the smooth eye pursuit of the visual stimuli was evoked nearly every time the photostimulation was applied (Fig. 5D, Movie S1). In addition, pupil dilation was also evoked (Fig. 5 E and F), which is in agreement with the fact that the cerebellum and the deep cerebellar nuclei modulate pupil size (35). After the pupil area came back close to the basal value (Fig. 5F), the HOKR amplitude was measured and an increase compared with the baseline was observed (Fig. 5 E and G). This effect lasted for a couple of minutes. Despite the perturbation of eye movement as well as the transient increase in the HOKR amplitude, HOKR amplitude increase after 1-h continuous visual training was observed in most mice as in the ChR2(−) mice (Fig. S10).

Fig. 5.

Glia-driven behavior. (A) Schematic of the optical fiber insertion location in the brain atlas (6.0 mm posterior to the bregma). (B) Expression of c-fos mRNA was detected in the PC layer and ML/GL after a single incident of photostimulation (200-ms blue and 200-ms yellow light with an interval of 500 ms) delivered through the optical fibers. (C) A photo of the HOKR experiment apparatus. (D) HOKR was induced in head-fixed unanesthetized mice. The absolute difference was taken between the average horizontal eye movement (gray; 5 min before glia photostimulation) and the eye movement after glia photostimulation (black; same durations as above) via optic fibers inserted into the cerebellum and was integrated for 25 s [n = 6 and 5 animals for ChR2(−) and ChR2(+), respectively; *P < 0.01, Student t test]. (E) (Middle Right) Pupil area plotted against time. (Bottom Right) Two-minute average of the horizontal eye movement before the glia photostimulation (gray) and after the pupil area came back within 1 SD from the baseline (black). (F) Peak increase in the pupil area and time for the cessation of pupil dilation were summarized (n = 5 animals). The time it took for the pupil to return within 1 SD from the baseline was defined as the recovery time. (G) The average gain of horizontal eye movement at 0–2 min and 2–4 min after cessation of pupil dilation were normalized to the gain before the glia photostimulation for ChR2(+) mice. As pupil dilation did not occur in ChR2(−) mice, eye movements were measured after 0–2 min and 2–4 min after 45 s from photostimulation. Results were summarized for ChR2(−) and ChR2(+) (n = 6 and 5; *P < 0.05, Student t test).

Discussion

Using optogenetics, we showed that selective glia photostimulation drives neuronal activity, synaptic plasticity, and behavioral response, and that the glia-to-neuron signaling is mediated by glial release of glutamate through DIDS-sensitive anion channels.

Mechanism Underlying Glia-Driven Neuronal Activity.

Rapid glia-to-neuron communication could be mediated by extracellular glutamate and/or K⁺ (36). Glia photostimulation likely caused changes in extracellular K⁺, as K⁺ channel-mediated currents in addition to ChR2-mediated currents were observed in BGs (Fig. S3D). In addition to the accumulation of extracellular K⁺, we suggest PCs are excited by glutamate released from glia photostimulation. Evidence supporting AMPAR and mGluR1 activation following glia photostimulation was shown (Figs. 2 and 3, and Fig. S4) and the AMPAR-mediated PC currents were blocked by DIDS, which blocks transporters and channels that allow the movement of negatively charged molecules, including glutamate. We propose that these DIDS-sensitive channels not only control the ambient glutamate level (29) but also regulate the amount of glutamate released in response to rapid and transient activity of glia. Glial depolarization alone or movement of ions associated with ChR2(C128S) likely triggers glutamate release through the DIDS-sensitive anion channels; however, the mechanism leading from photoactivation to the glutamate release need to be evaluated further. An important issue is whether the same glutamate release mechanisms operate under physiological conditions. In our experiments, 10-s glia photostimulation resulted in depolarization of up to 10 mV in BGs (Fig. S9A). Glia photostimulation likely induces depolarization homogeneously throughout the cells, as ChR2 is expressed throughout the entire processes. When PFs were electrically stimulated repetitive times, BGs responded with a comparable amplitude of depolarization (Fig. S9 B–D). However, because of the low input resistance, much of the depolarization produced locally is likely to be attenuated and underestimated. A computational model has estimated that depolarization as much as 40 mV is likely produced at the ensheathing BG membrane in response to synaptic stimuli (7); therefore, similar or even larger depolarization compared with our BG photostimulation protocol could be produced physiologically, although such a profound depolarization has not been substantiated experimentally due to inaccessibility of glial fine processes by recording electrodes. We also showed that glia photostimulation for 100 ms was sufficient to trigger depolarization of BGs (Fig. S9A) and release of glutamate was detectable with PC AMPARs (Fig. 4C). Such depolarization of BGs can also be evoked by a couple of electrical stimulation typically used for studying neuronal synaptic transmission (Fig. S9 B–D); therefore, glial release of glutamate may have unintentionally been evoked in previous studies, which could complicate the interpretation of those studies.

Plasticity Induced by Glia Photostimulation.

We demonstrated that glia photostimulation could trigger long-term plasticity of PF–PC synaptic transmission. In a previous report, stimulation of sparsely distributed PFs failed to evoke LTD, whereas stimulation of PF bundles in the ML readily evoked LTD (37). The latter stimulation protocol inevitably activates the BGs as well (9, 10). We did not evaluate whether glial release of glutamate alone is sufficient in producing LTD with weaker glia photostimulation than the 10-s protocol. Because mGluR1 is expressed at a location that is heavily protected by glutamate transporters, summation of glutamate concentration from neuronal spillover and glial release may be required to overcome the protection and activate the mGluR1. Glial glutamate release could thus be considered as an amplification mechanism for detection of close by activity and induction of plasticity. However, it still remains unresolved whether our stimulation protocol was physiologically relevant. Further studies are required to reveal participation of glial cells in long-term synaptic plasticity under physiological conditions.

Conclusion and Future Perspective.

We generally assume that our mind is dictated solely by the activity of neurons. Contrary to this prevailing notion, recent evidence proposed a potential role of glia in actively participating in the neural information processing (1, 12–15). In this study, we directly demonstrate that signals initiated from glial cells can drive neuronal activity and animal behavior. We expect that our model can be a useful tool to assess how the “debate” between neurons and glia can set the tone of our mind.

Materials and Methods

Full details of all methods are available in SI Materials and Methods. All animal procedures were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals (38) and approved by the Animal Research Committee of the National Institute for Physiological Sciences. For slice electrophysiology, parasagittal cerebellar slices were prepared from young mice [postnatal day 17 (P17) to P24]. For in vivo optical stimulation, blue and yellow light were applied through a plastic optical fiber (0.75 or 0.5 mm in diameter) to the cerebellum. For HOKR experiment, a sheet of paper with checkered pattern was placed semicircularly (radius, 32 cm) surrounding the mouse and oscillated horizontally and sinusoidally.