Increases in Intracellular Calcium Triggered by Channelrhodopsin-2 Potentiate the Response of Metabotropic Glutamate Receptor mGluR7*

The metabotropic glutamate receptor 7a (mGluR7a), a heptahelical Gαi/o-coupled protein, has been shown to be important for presynaptic feedback inhibition at central synapses and certain forms of long term potentiation and long term depression. The intracellular C terminus of mGluR7a interacts with calmodulin in a Ca2+-dependent manner, and calmodulin antagonists have been found to abolish presynaptic inhibition of glutamate release in neurons and mGluR7a-induced activation of G-protein-activated inwardly rectifying K+ channel (GIRK) channels in HEK293 cells. Here, we characterized the Ca2+ dependence of mGluR7a signaling in Xenopus oocytes by using channelrhodopsin-2 (ChR2), a Ca2+-permeable, light-activated ion channel for triggering Ca2+ influx, and a GIRK3.1/3.2 concatemer to monitor mGluR7a responses. Application of the agonist (S)-2-amino-4-phosphonobutanoic acid (l-AP4) (1–100 μm) caused a dose-dependent inward current in high K+ solutions due to activation of GIRK channels by G-protein βγ subunits released from mGluR7a. Elevation of intracellular free Ca2+ by light stimulation of ChR2 markedly increased the amplitude of l-AP4 responses, and this effect was attenuated by the calcium chelator 1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid tetrakis (acetoxymethyl ester). l-AP4 responses were potentiated by submembranous [Ca2+] levels within physiological ranges and with a threshold close to resting [Ca2+]i values, as determined by recording the endogenous Xenopus Ca2+-activated chloride conductance. Together, these results show that l-AP4-dependent mGluR7a signaling is potentiated by physiological levels of [Ca2+]i, consistent with a model in which presynaptic mGluR7a acts as a coincidence detector of Ca2+ influx and glutamate release.

Metabotropic glutamate receptors (mGluRs) 4 are G-proteincoupled receptors that play a key role in synaptic transmission. Eight mGluRs have been identified and grouped into three classes (1). mGluR7 is a G␣ i/o protein-linked member of the group III mGluRs that is widely expressed at both glutamatergic and non-glutamatergic synapses and has a crucial role in some forms of hippocampal plasticity (2,3). In addition, mGluR7 Ϫ/Ϫ mice develop spontaneous seizures (4). Therefore, mGluR7 is a current target for the development of novel antiepileptic drugs. mGluR7 has been localized to presynaptic terminals, where it inhibits neurotransmitter release via a G-protein-dependent pathway (5). In cortical presynaptic terminals, mGluR7 colocalizes primarily with and inhibits N-type Ca 2ϩ channels. However, in cerebellar granule cells, mGluR7 inhibits P/Q-type Ca 2ϩ channels via a phospholipase C-dependent pathway (6). mGluR7 may act as a presynaptic feedback inhibitor; elevated synaptic glutamate activates mGluR7, thereby decreasing presynaptic Ca 2ϩ influx and reducing glutamate release (7).
Biochemical studies indicate that mGluR7 signaling is regulated by a variety of intracellular proteins. Specifically, the cytoplasmic C-terminal tail of the predominant splice variant mGluR7a has been shown to bind proteins whose properties are directly or indirectly influenced by changes in intracellular Ca 2ϩ levels, such as calmodulin (CaM) (8), protein kinase C (9), and protein interacting with protein kinase C 1 (PICK-1) (10,11). CaM binds to mGluR7a (and other group III mGluRs) in a Ca 2ϩ -dependent manner, and CaM inhibitors have been shown to prevent presynaptic inhibition at glutamatergic autapses and to disrupt mGluR7a signaling in Xenopus oocytes (8). Moreover, inhibition of P/Q voltage-gated calcium channels by L-AP4 is blocked by BAPTA, suggesting that Ca 2ϩ is required for mGluR7-mediated presynaptic inhibition in cerebellar granule cells (6). These data support the view that presynaptic inhibition by mGluR7 depends on [Ca 2ϩ ] i , likely through regulation of complex protein interactions in the presynaptic nerve terminal (reviews in Refs. 12 and 13).
Here, we tested the dependence of mGluR7a activation on submembranous Ca 2ϩ by using the G-protein-activated inwardly rectifying K ϩ channel (GIRK; Kir3.1/3.2) in a heterologous expression system in which we could temporally control Ca 2ϩ influx with channelrhodopsin-2 (ChR2), a light-activated cation channel (14), monitor submembranous Ca 2ϩ levels with the endogenous Ca 2ϩ -activated chloride conductance (15), and study mGluR7a in relative isolation. mGluR7a, the G␤␥ effector GIRK, and ChR2 were co-expressed in Xenopus oocytes, and membrane currents were measured under two-electrode voltage clamp. Our results show that mGluR7a signaling is potentiated with increasing [Ca 2ϩ ] i .

EXPERIMENTAL PROCEDURES
Chemicals-L-AP4 was obtained from Calbiochem (Darmstadt, Germany), and niflumic acid (NFA) was obtained from Cayman Chemical (Ann Arbor, MI). Stock L-AP4 (100 mM) was dissolved in 100 mM NaOH and stored at Ϫ20°C. L-AP4 was diluted 10 3 -10 5 -fold for experiments (working concentration 1-100 M). NFA was dissolved in DMSO (300 mM stock), kept frozen at Ϫ20°C, and diluted 1000-fold for a working concentration of 300 M. The K i for NFA block of Ca-activated Cl channels in oocytes is 10 M. BAPTA-AM (Sigma product number A1076) was dissolved in DMSO, stored under argon at a stock concentration of 100 -300 mM, and diluted at least 1000-fold for experiments.
Expression Constructs and RNA Synthesis-The mGluR7a cDNA encoding the short isoform of mGluR7 (Ref. 16); (gift from Dr. S. Nakanishi, University of Kyoto, Japan) and the GIRK (Kir3.1/3.2) construct (gift from Dr. Andreas Karschin, University of Würzburg, Germany) have been described previously (8). The ChR2 construct used was a C-terminally truncated version (amino acids 2-315) that lacks most of the C-terminal cytoplasmic domain (14) and contains a single point mutation (H134R), which results in larger currents but does not change ion selectivity (17). In vitro transcription was performed using commercial kits (Stratagene, Agilent Technologies, Santa Clara, CA, or Ambion, Applied Biosystems, Darmstadt, Germany). Synthetic RNAs were stored at Ϫ80°C in aliquots at a concentration of 1 g/l.
Oocyte Expression and Electrophysiological Recordings-Stage V-VI oocytes were removed from anesthetized Xenopus laevis and enzymatically treated to remove follicular cells. The oocytes were incubated at 18 -19°C in oocyte saline (ND-96, in mM: 96 NaCl, 2 KCl, 1 CaCl 2 , 1 MgCl 2 , 5 HEPES, pH 7.4) and injected with in vitro transcribed RNA as follows. For recordings from cells injected with mGluR7a and GIRK RNA, 1 l (25 ng) of mGluR7a RNA was mixed with 1 l of GIRK RNA (ϳ50 pg). Light-activable ChR2 has retinal bound (14). Hence, oocytes were injected on day 1 with a 1:1 mixture of GIRK and mGluR7a RNAs (50 nl/oocyte) and 48 h later with ChR2 RNA (20 ng) followed by incubation in oocyte solution containing trans-retinal (1 M). Oocytes were used for recordings from day 3 to day 7 after the initial injection using two-electrode voltage clamp recording. The volume of the recording chamber was 15 l, and this volume was exchanged with 0.5-1 ml of perfusion solution in ϳ1 s. K ϩ was substituted for Na ϩ in solutions in which K ϩ was elevated. Increases in Ca 2ϩ were achieved by substituting for Na ϩ . Electrodes had resistances of 0.4-2 megaohms and were filled with 3 M KCl in 0.2% (w/v) agar. Data were acquired with an Axon 2B amplifier, Digidata 1200 A/D converter, and pCLAMP 9 software (MDS Inc., Toronto, Canada).
Photoactivation of ChR2-ChR2 was activated in electrodeimpaled oocytes with a mercury arc lamp and a 450 Ϯ 25-nm band filter (5 milliwatt/mm 2 ) as described (14). Neutral density filters with a computer-activated shutter were used to control light intensity.

RESULTS
Control of Submembranous Ca 2ϩ with ChR2-ChR2 is a light-activated eukaryotic cation channel that is permeable to Ca 2ϩ (14). We utilized ChR2 to introduce Ca 2ϩ into the submembranous compartment of Xenopus oocytes to mimic Ca 2ϩ influx through voltage-gated calcium channels into the presynaptic nerve terminal. We took advantage of the well characterized, endogenous Ca 2ϩ -activated chloride current (15,18) to monitor free Ca 2ϩ concentrations beneath the oocyte plasma membrane. Fig. 1A shows the voltage clamp protocol and resulting currents in the presence and absence of extracellular Ca 2ϩ . ChR2 was activated by illuminating the oocyte immediately before recording, and the oocyte remained illuminated during the record. In the absence of extracellular Ca 2ϩ , lightinduced currents at ϩ40 mV were small and stable. In the presence of 2 mM extracellular Ca 2ϩ , a large Cl Ϫ current (I Cl(Ca) ) was evident during the second pulse to ϩ40 mV (note also the Cl Ϫ current during the pulse to Ϫ140 mV). Thus, Ca 2ϩ entry via the ChR2 channel activates I Cl(Ca) .
The dependence of I Cl(Ca) upon activation of ChR2 was determined by varying the light intensity. In the presence of 2 mM [Ca 2ϩ ] o , peak amplitudes of the Cl Ϫ current were measured during the second step to ϩ40 mV (Fig. 1B, asterisk) at increasing light intensities, which activate an increasing number of ChR2 channels. The relationship between I Cl(Ca) and ChR2 current is plotted in Fig. 1C. The relationship was fit by the Hill equation (Fig. 1C, solid line), with k ϭ 0.46 A (value of the ChR2 current at the half-maximal response) and n ϭ 2.13. This relationship is very similar to the relationship between [Ca 2ϩ ] and I Cl(Ca) as determined by Haase and Hartung (18), who applied Ca 2ϩ to the intracellular face of excised inside-out patches of Xenopus oocyte membrane and determined a K d value of 500 Ϯ 20 nM and a Hill coefficient of n ϭ 2.3 Ϯ 0.12 (the average of their data is superimposed on our data in Fig. 1C). We assume that the K d for Ca 2ϩ determined from excised patches is a constant property of the Ca 2ϩ -activated chloride channel. Thus, by equating the value of k for the oocyte shown in Fig. 1C with the K d of I Cl(Ca) for Ca 2ϩ determined by Haase and Hartung (18), we find that I ChR2 ϭ 0.46 A corresponded to a free [Ca 2ϩ ] i of 500 nM. The [Ca 2ϩ ] i produced by smaller and larger currents was assumed to be proportional to the ChR2 current ([Ca 2ϩ ] i ϭ (500/k I ChR2 ) * I ChR2 ). The calibration shown in Fig. 1, B and C, was performed for each oocyte, and each ChR2 current was then converted to free submembranous [Ca 2ϩ ].
Submembranous Ca 2ϩ can be sequestered with Ca 2ϩ chelators without altering Ca 2ϩ flux through ChR2. Fig. 1D demon-strates that extracellular application of the fast, membrane-permeable Ca 2ϩ chelator BAPTA-AM largely inhibited I Cl(Ca) within 5-8 min (69.4 Ϯ 5.7%; average Ϯ S.E., n ϭ 10) and had no effect on ChR2 current (Fig. 1D, squares). The failure to achieve complete block of I Cl(Ca) raises the possibility that some ChR2 and chloride channels are spatially apposed or colocalize in macromolecular complexes. This block of I Cl(Ca) corroborates that I Cl(Ca) reflects [Ca 2ϩ ] i and that BAPTA-AM treatment does not affect ChR2 function and therefore transmembrane Ca 2ϩ flux.
To characterize submembranous Ca 2ϩ kinetics, we designed protocols to monitor the time course of ChR2-induced changes in [Ca 2ϩ ] i . In oocytes, the increase in submembranous [Ca 2ϩ ] depends on the rates of both Ca 2ϩ entry through ChR2 and Ca 2ϩ buffering and removal. Fig. 2A shows the protocol and currents recorded for increasing durations of illumination. The peak Cl Ϫ current is plotted versus light duration for a representative cell in Fig. 2B. The current increased with a time constant of 630 Ϯ 125 ms (average Ϯ S.E., n ϭ 5). Thus, in 1-2 s, the intracellular Ca 2ϩ concentration reaches a steady state. To determine the time required for [Ca 2ϩ ] i to return to the resting level, the light was turned off during the Ϫ140-mV step, and the second step to ϩ40 mV was delayed in 100-ms increments (see the protocol illustrated in Fig. 2C, top). I Cl(Ca) rapidly declined (peak current during the second step to ϩ40 mV, Fig. 2C) following the termination of illumination. The time course for recovery is shown in Fig. 2D. The time constant for the return to resting [Ca 2ϩ ] i levels was 181 Ϯ 14 ms (n ϭ 4). It is possible that the kinetics of changes in I Cl(Ca) are limited by the rates of activation and deactivation of the chloride channel. Evidence from excised patches of oocyte membrane where [Ca 2ϩ ] was increased or decreased rapidly (within a few milliseconds or less) at the cytoplasmic surface have shown that, at the concentrations relevant for our experiments (ϳ0.5 M), the time constant of I Cl(Ca) activation is between 50 (15) and 500 ms (18); the latter value is close to the activation time constant we measured (Fig. 2C). Likewise, the deactivation time constants varied from 30 (15) to 100 -200 ms (18). Again, the latter values are close to ] o . Ca 2ϩ that enters during the voltage step to Ϫ140 mV activates I Cl(Ca) that can be seen as an additional inward current at Ϫ140 mV and as the large transient outward current during the second step to ϩ40 mV. B, dependence of chloride current on light intensity. In the presence of 2 mM [Ca 2ϩ ] o , increasing the intensity of illumination (1, 2, 3, 5, 10, 25, 50, and 100% of maximal intensity) produced correspondingly larger I Cl(Ca) currents during the second step to ϩ40 mV (peak current indicated by asterisk). C, dependence of peak chloride current upon ChR2 current. I ChR2 was measured 50 ms after the transient phase of the ChR2 current during the voltage step to Ϫ140 mV, and I Cl(Ca) was measured at the peak of the response during the second step to ϩ40 mV (panel B, asterisk). Data (black squares) were fit with the Hill equation  Peak currents (during second step to ϩ40 mV) were plotted versus illumination time. Data were fit by a single exponential with ϭ 486 ms. C, delayed voltage step for I Cl(Ca) shows that submembranous Ca 2ϩ is cleared quickly after ChR2 inactivation. Illumination started at the arrow and ceased for all traces at the filled circle. The second step to ϩ40 mV was delayed in increasing steps of 100 ms after the termination of illumination. D, decay of chloride current after cessation of illumination. Peak currents during the second step to ϩ40 mV are plotted versus time after the light was turned off (from records in panel C). Data were fit by a single exponential with ϭ 183 ms.
those measured here for changes in I Cl(Ca) (Fig. 2D). Thus, our ability to measure changes in [Ca 2ϩ ] could be limited by the inherent kinetics of the chloride channel, and changes in submembranous [Ca 2ϩ ] may be faster than the time courses shown in Fig. 2, B and D. We conclude that ChR2 channel activation allows us to increase submembranous [Ca 2ϩ ] in less than 1-2 s to known free Ca 2ϩ levels, and once the light is turned off, [Ca 2ϩ ] i returns to basal levels in much less than a second. Thus, the temporal profile of submembranous [Ca 2ϩ ] can be tightly controlled through illumination.
Potentiation of mGluR7a Agonist Responses by Changes in [Ca 2ϩ ]-Having established a system in which submembranous [Ca 2ϩ ]can be temporally controlled and accurately measured, we next characterized the Ca 2ϩ dependence of mGluR7a signaling. Oocytes co-expressing mGluR7a, GIRK 3.1/3.2, and ChR2 were voltage-clamped, and in each oocyte, the relationship between I ChR2 and I Cl(Ca) was examined as described above. Subsequently, the Ca 2ϩ -activated chloride channel was blocked with NFA (300 M). Following calibration of each oocyte and NFA treatment, the group III-specific agonist L-AP4 was applied. Repeated applications of the same concentration of L-AP4 during increasing levels of illumination, and therefore increasing levels of submembranous [Ca 2ϩ ], were employed to determine changes in mGluR7a function in the presence of Ca 2ϩ , as described in more detail below. Fig. 3A illustrates a typical experiment. Oocytes were bathed in a solution containing 80 mM K ϩ (high potassium) to increase the flux of K ϩ through the inwardly rectifying GIRK channel. Once L-AP4independent GIRK channel currents were stable, L-AP4 was applied in high potassium. This resulted in a characteristic inward current, indicative of L-AP4-dependent G␤␥ activation of GIRK (19). Currents were recorded with [L-AP4] (100 M) and all other conditions constant while varying illumination intensity (and thus submembranous [Ca 2ϩ ]). As expected from Figs. 1 and 2, increasing light intensity resulted in larger ChR2 currents, as indicated by the short latency, fast-desensitizing currents immediately following illumination with a smaller maintained ChR2 current (Fig. 3A). L-AP4 was applied during the plateau phase of the ChR2 current. Most notably, as ChR2 current increased, the responses to L-AP4 were progressively potentiated in comparison with the L-AP4 currents obtained in the absence of illumination. Control experiments showed that L-AP4 responses were blocked by Ba 2ϩ , which inhibits GIRK currents, and that L-AP4 had no effect on ChR2 current (data not shown).
Results similar to those shown in Fig. 3A were observed in seven out of eight oocytes (one oocyte had no increase in agonist-induced GIRK current). The peak L-AP4-induced current amplitude data for five oocytes are plotted in Fig. 3B as a function of [Ca 2ϩ ] i . All responses increased with increasing illumination, and these increases began at very low [Ca 2ϩ ] i (at or below ϳ100 nM). The relationship between the agonist-induced currents and [Ca 2ϩ ] i could be fitted by binding isotherms, with a mean equilibrium dissociation constant of K d ϭ 533 Ϯ 201 nM and a mean maximal current of 333 Ϯ 84 nA (means Ϯ S.E.). However, there was considerable variability between individual oocytes. This variability in the sensitivity of the response to [Ca 2ϩ ] i is likely to reflect differences in the expression ratios of GIRK to mGluR7a between oocytes.
To confirm that the potentiation of mGluR7a-stimulated GIRK currents observed upon light illumination is mediated by Ca 2ϩ , we tested whether BAPTA-AM inhibits light-dependent mGluR7a potentiation (Fig. 3C). BAPTA-AM blocked the ChR2-mediated increase in the mGluR7a response by 65 Ϯ 12% (S.E.; n ϭ 5, p ϭ 0.002). This was similar to the inhibition of I Cl(Ca) (69%, Fig. 1D) seen under the same conditions. These results are consistent with the ChR2-induced potentiation of mGluR7a currents being primarily mediated by an increase in [Ca 2ϩ ] i . To further confirm that the increased L-AP4 response seen in panel A is due to Ca 2ϩ influx, we repeated the stimulation protocol shown in Fig. 3A with a constant light intensity while varying the extracellular concentration of Ca 2ϩ . Fig. 3D shows that the peak amplitude of the L-AP4 response increased with increasing [Ca 2ϩ ] o . The peak L-AP4 response in panel D and additional data are plotted in panel E (Fig. 3). Two additional points are evident from Both modeling and experimental studies suggest that in a restricted space such as the synaptic cleft, extracellular Ca 2ϩ can be transiently reduced during periods of high activity, producing short term synaptic depression (20). In addition, mGluRs belong to a superfamily of proteins that bind extracellular Ca 2ϩ (21). mGluR1 has been reported to be activated by extracellular Ca 2ϩ in the absence of agonist (22), and the agonist responses of the related GABA B receptor are modulated by extracellular Ca 2ϩ (23). Thus, mGluR7a signaling might be regulated not only by changes in [Ca 2ϩ ] i but also by extracellular Ca 2ϩ levels. We therefore examined whether extracellular Ca 2ϩ might activate mGluR7a in the absence of L-AP4 and whether mGluR7a currents elicited by L-AP4 are modulated by changes in the extracellular Ca 2ϩ concentration. We found that extracellular Ca 2ϩ alone was unable to elicit GIRK currents (Fig. 4A, left half, and 4B, filled circles) and that current responses to L-AP4 varied little between 100 M and 2 mM [Ca 2ϩ ] o (Fig. 4A, right half, and Fig. 4B, open diamonds). Although extracellular Ca 2ϩ modulates the activation of mGluR7a by agonist (L-AP4), this effect will be important only if extracellular Ca 2ϩ is reduced to less than 0.1 mM (Fig. 4B), a value well below the changes predicted or measured under physiological conditions at the synapse. Hence, it appears unlikely that physiological changes in extracellular [Ca 2ϩ ] i influence mGluR7a signaling.

DISCUSSION
In this study, we provide direct evidence showing that elevated [Ca 2ϩ ] i modulates agonist activation of mGluR7a. Our results are based on the use of the light-activated cation channel ChR2 (14) for inducing Ca 2ϩ influx and of the endogenous calcium-activated chloride channel of Xenopus oocytes for estimating free intracellular Ca 2ϩ concentrations. The magnitudes of I Cl(Ca) allowed us to determine free [Ca 2ϩ ] i at the inner face L-AP4 responses are superimposed at the beginning of agonist application (note that the responses are small because L-AP4 was 1 M, rather than 100 M in panel A, and the driving force for K ϩ influx was smaller due to differences in the membrane potential, see below). E, the peak L-AP4 response (maximal response for 0.1 mM, and peaks indicated in panel D by asterisks below the 0.3-5 mM traces) are plotted (with others not illustrated in panel D) along with post-BAPTA-AM responses of the same cell (filled circles). Cells were voltage-clamped at Ϫ60 mV (panels C-E) or Ϫ100 mV (panels A and B).
of the membrane, where regulation of mGluR7a activity occurs. Light activation of ChR2 provided rapid, reproducible, and readily reversible changes in submembranous [Ca 2ϩ ]. Importantly, increases in submembranous [Ca 2ϩ ] potentiated the mGluR7a response to the selective agonist L-AP4, as monitored by activation of co-expressed GIRK, and this potentiation could be detected at concentrations close to resting levels of [Ca 2ϩ ] i . A scheme summarizing the experimental system used and our interpretation of the results obtained are presented in Fig. 5. Accordingly, elevation of [Ca 2ϩ ] i by light activation of ChR2 results in Ca 2ϩ binding to CaM or other Ca 2ϩ -sensitive proteins that interact with the C-terminal tail of mGluR7a in the presence of Ca 2ϩ . This would displace prebound G␤␥ from mGluR7a and thereby trigger K ϩ influx via G␤␥ binding to GIRK. Although not indicated in the figure, it is possible that Ca 2ϩ and Ca 2ϩ -binding proteins act directly on GIRK. This alternative explanation seems unlikely because we have found no reports describing potentiation of GIRK currents by increases of intracellular Ca 2ϩ . Conversely, there is mixed evidence for a role of Ca 2ϩ in G-protein-coupled receptor-mediated inhibition of GIRK currents, with some laboratories finding enhanced or accelerated inhibition and others finding no effect of Ca 2ϩ (24 -26). We observed no increase in deactivation or inhibition by Ca 2ϩ . The scheme in Fig. 5 agrees well with our previous model of presynaptic mGluR regulation by Ca 2ϩ -CaM (8,12), which was based on the finding that Ca 2ϩ -CaM and G␤␥ compete for overlapping binding sites in the C-terminal tail of mGluR7a that are conserved in other presynaptic group III mGluRs (27).
What are the implications and predictions of our findings for synaptic physiology? Since mGluR7a is localized in presynaptic terminals of both excitatory and inhibitory synapses and known to inhibit presynaptic Ca 2ϩ channels (6,28), its sensitivity to [Ca 2ϩ ] i will provide a gain control for negative feedback inhibition of neurotransmitter release. Active synapses with elevated presynaptic [Ca 2ϩ ] i will be more potently inhibited upon mGluR7a activation. Such a regulation by Ca 2ϩ implies that inhibition by mGluR7a will be enhanced if the receptor is positioned close to Ca 2ϩ channels. Furthermore, the [Ca 2ϩ ] i required to modulate mGluR7a should lie in a range that is similar to the [Ca 2ϩ ] i , which triggers synaptic vesicle exocytosis. In the literature, the [Ca 2ϩ ] i values reported to initiate neurotransmitter release vary widely (reviewed in Ref. 29). In some excitable cells, exocytosis requires such a high Ca 2ϩ concentration (ϳ100 M) that the fusion machinery is assumed to be within 20 nm of the Ca 2ϩ channels. For other neurons, lower  Heterologously expressed proteins are underlined (GIRK, mGluR7a, and ChR2). The endogenous Ca 2ϩ -activated chloride channel and Ca 2ϩ -sensitive proteins (for example, calmodulin) are included, and G␤␥ subunits are depicted as linked to the plasma membrane. mGluR7a exists as a dimer but is pictured as a monomer for simplicity. Blue light (450 nm) activates ChR2, producing a cation current that results in an influx of Ca 2ϩ . This activates the endogenous Ca 2ϩ -activated chloride channel. The current through this channel (I Cl(Ca) ) enables monitoring and calibration of submembranous [Ca 2ϩ ] as a function of light intensity. Concurrently, Ca 2ϩ activates Ca 2ϩ -sensitive proteins, which interact with mGluR7a and thus potentiate agonist-dependent G␤␥ signaling. G␤␥ signaling is monitored by GIRK activation and influx of K ϩ .
values are reported, and the [Ca 2ϩ ] i necessary for exocytosis appears to reflect a more global concentration estimate within the terminal (neuromuscular junction, 2-4 M (30); calyx of Held, ϳ10 M (31)). Thus, the range of free [Ca 2ϩ ] i over which mGluR7a modulation was observed here is well within or even below that initiating neurotransmitter release from presynaptic terminals. In conclusion, our results are consistent with the concept that presynaptic group III mGluRs act as activity-dependent regulators of neurotransmitter release.
The results described here demonstrate that, at non-saturating agonist concentrations, elevations in [Ca 2ϩ ] i result in increased mGluR7a responses. An unusual feature of mGluR7 is its reportedly low affinity for glutamate (K d ϳ1 mM; (2,32)). Therefore, it has been suggested that mGluR7 is only activated during intense synaptic activity or under extreme circumstances, such as epileptic seizures or ischemia. Indeed, L-AP4 has been found to be neuroprotective under conditions of excitotoxicity, e.g. in the presence of elevated extracellular glutamate (33). The enhanced survival of cultured cerebellar neurons observed in the presence of this agonist does not involve large increases in [Ca 2ϩ ] i and appears to depend on mGluR7 (34). Additional studies also have implicated mGluR7, working by inhibition of glutamate release, in neuroprotection. The increased L-AP4 response seen here at elevated [Ca 2ϩ ] i (Fig. 3) predicts that concentrations of glutamate well below the K d would be effective if intracellular Ca 2ϩ levels were raised. Thus, mGluR7 may be important under physiological conditions, rather than solely providing neuroprotection in situations involving excessive stimulation.
The results reported here provide a new perspective for interpreting the phenotypes of mGluR knock-out mice. Synaptic facilitation depends on the [Ca 2ϩ ] i level in the presynaptic terminal. Potentiation of mGluR7a by increased [Ca 2ϩ ] i , as reported here, should enhance feedback inhibition of Ca 2ϩ channels and thus accelerate recovery from facilitation. Consistent with this prediction, mGluR7-deficient mice show a delayed recovery from synaptic facilitation (4). A train of action potentials can increase presynaptic [Ca 2ϩ ] i , and under these conditions, mGluR7 responses should be enhanced and their speed increased. Thus, mGluR7 could act as a frequency-dependent filter for synaptic transmission by more effectively reducing release at high frequencies. The magnitude of this effect may be modest since there is evidence at the calyx of Held that presynaptic mGluRs contribute only about 10% to synaptic depression (35). However, even a 10% change may be sufficient to produce susceptibility to epilepsy, and indeed, mGluR7 knock-out mice are prone to epileptic seizures, whereas mice deficient in mGluRs 1, 2, 4, 5, 6, or 8 do not show such a phenotype (4).
Increases in [Ca 2ϩ ] i can result not only from the opening of voltage-gated Ca 2ϩ channels but also from the activation of intracellular Ca 2ϩ stores. Thus, Ca 2ϩ release from intracellular stores might provide a molecular link between different mGluR subtypes because activation of group I mGluRs results in the production of inositol 1,4,5-trisphosphate, and hence, the release of stored Ca 2ϩ (36). Since both group I and III receptors have been shown to be presynaptic at the Schaffer collateral-CA1 synapse in the hippocampus (37), release of intracellular Ca 2ϩ could create another pathway for synergy between mGluRs, as postulated for some aspects of long term potentiation and long term depression (38).
Long term depression is the most obvious form of synaptic plasticity in which mGluR7 might be involved since mGluR7 inhibits Ca 2ϩ channels and synaptic transmission. Indeed, at glutamatergic synapses formed by hippocampal mossy fibers on stratum lucidum interneurons, high frequency stimulation has been shown to induce presynaptic long term depression that is mimicked by L-AP4 at concentrations sufficient to activate mGluR7 and is blocked by the group III-specific mGluR antagonist (RS)-␣-methylserine-O-phosphate (3). However, mGluR7 is also present at inhibitory nerve terminals, and inhibition of such terminals should produce excitation due to a reduced release of GABA. Consistent with such a mechanism, activation of group III mGluRs has been found to facilitate glutamate release in some layers of the cortex (39). Also, glutamate spillover resulting from mossy fiber stimulation in cerebellar glomeruli has been shown to suppress GABA release via presynaptic mGluR activation (40). All these results are consistent with an important role of group III mGluRs in bidirectional synaptic modifiability. Furthermore, several lines of evidence indicate that the magnitude of the rise in presynaptic [Ca 2ϩ ] i following stimulation is a major factor in determining whether a synapse shows potentiation or depression (41,42). It therefore is tempting to speculate that the sensitivity of mGluRs to [Ca 2ϩ ] i described here for mGluR7a might play an important role in the selection of long term potentiation versus long term depression.