The ubiquitous presynaptic metabotropic glutamate receptors (mGluRs) are generally believed to primarily inhibit synaptic transmission through blockade of Ca2+ entry. Here, we analyzed how mGluR8 achieves a nearly complete inhibition of glutamate release at hippocampal synapses. Surprisingly, presynaptic Ca2+ imaging and miniature excitatory postsynaptic current recordings showed that mGluR8 acts without affecting Ca2+ entry, diffusion, and buffering. We quantitatively compared the Ca2+ dependence of the inhibition of release by mGluR8 with the inhibition by ω-conotoxin GVIA. These calculations suggest that the inhibition produced by mGluR8 may be explained by a decrease in the apparent Ca2+ affinity of the release sensor and, to a smaller extent, by a reduction of the maximal release rate. Upon activation of mGluR8, phasic transmitter release toward the end of a train of action potentials is greater as compared with presynaptic inhibition induced by blocking Ca2+ entry, which is consistent with the important role of Ca2+ in accelerating the replenishment of released vesicles. The action of mGluR8 was resistant to blockers of classical G-protein transduction pathways including inhibition of adenylate cyclase and may represent a direct effect on the release machinery. In conclusion, our data identify a mode of presynaptic inhibition which allows mGluR8 to profoundly inhibit vesicle fusion while not diminishing vesicle replenishment and which thereby differentially changes the temporal transmission properties of the inhibited synapse.
Metabotropic glutamate receptors (mGluRs) are as widely distributed as their ionotropic counterparts and represent an additional mode of glutamatergic signaling. The most prominent effect of group II and group III mGluRs is to profoundly inhibit transmitter release as auto- and heteroreceptors at many synapses throughout the brain. It is generally believed that these presynaptic mGluRs reduce action potential–driven release primarily by lowering presynaptic Ca2+ entry (Conn and Pin 1997; Wu and Saggau 1997; Anwyl 1999; El Far and Betz 2002), but this has been directly tested at only a few synapses (Takahashi et al. 1996; Cochilla and Alford 1998; Kamiya and Ozawa 1999; Rusakov et al. 2004). In addition to the reduction of Ca2+ entry via inhibition of Ca2+ channels or via activation of K+ channels, various parts of the release machinery could also be directly targeted by the metabotropic signaling. Evidence from the analysis of the modulation of miniature release suggested that certain subtypes of mGluRs also have a secondary effect downstream of Ca2+ entry (e.g., Gereau and Conn 1995; Poncer et al. 1995; Scanziani et al. 1995; Tyler and Lovinger 1995; Manzoni et al. 1997; Schoppa and Westbrook 1997), but its relevance for action potential–driven release is unclear (Wu and Saggau 1997; Anwyl 1999). Furthermore, it remains unknown which functional properties of the release machinery might be modified by this downstream signaling.
The identification of the targets of this downstream signaling by metabotropic receptors is important for several reasons. First, presynaptic Ca2+ triggers a multitude of processes besides vesicle fusion, including various forms of short- and long-term synaptic plasticity (Nicoll and Malenka 1995; Zucker and Regehr 2002) as well as the accelerated replenishment of released synaptic vesicles (Dittman and Regehr 1998; Stevens and Wesseling 1998; Wang and Kaczmarek 1998). Thus, a metabotropic receptor acting on Ca2+ entry is likely to also concomitantly inhibit other Ca2+-dependent processes while a receptor directly modulating the release machinery will bypass such side effects. Second, because transmitter release is a highly nonlinear and saturating process (Heidelberger et al. 1994; Bollmann et al. 2000; Schneggenburger and Neher 2000; Lou et al. 2005), the efficacy of inhibitory mechanisms which lower Ca2+ binding by the release machinery will strongly depend on the amount of Ca2+ entry. As a result, the inhibition by such mechanisms will be significantly weakened by, for example, activity-dependent action potential broadening, while the potency of other mechanisms, which, for example, decrease the maximal fusion rate, will be unaffected. Therefore, the transduction target of presynaptic metabotropic receptors plays a pivotal role for the temporal and dynamic transmission properties of the inhibited synapse.
In the hippocampus, there is a strictly pathway-specific expression of different subtypes of mGluRs (Shigemoto et al. 1997). In particular, the 2 projections from the medial and lateral entorhinal cortex to hippocampal granule cells are decorated exclusively with mGluR2/3 and mGluR8, respectively, both of which potently depress glutamate release (Macek et al. 1996; Dietrich et al. 1997). In the present study, we analyze the transduction target of mGluR8, report that this receptor reduces transmitter release entirely without affecting entry and propose mechanistically which functional properties of the release machinery may be modulated by mGluR8.
Materials and Methods
Male Wistar rats aged 20–40 days were decapitated, and the brains were rapidly removed and sliced in ice-cold oxygenated solution containing (in mM): 90 NaCl, 3 KCl, 2 MgCl2, 2 CaCl2, 1 Na-pyruvate, 10 glucose, 10 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), and 90 sucrose, pH was adjusted to 7.4 or containing (in mM): 87 NaCl, 2.5 KCl, 7 MgCl2, 0.5 CaCl2, 25 NaHCO3, 25 glucose, and 75 sucrose (95% O2, 5% CO2). Afterward, slices were stored at room temperature (22–24 °C) in solution containing the following (in mM): 124 NaCl, 3 KCl, 2 MgCl2, 2 CaCl2, 26 NaHCO3, and 10 glucose (95% O2, 5% CO2). For recording (field excitatory postsynaptic potentials [fEPSPs], patch clamp), slices were transferred to a Haas-type interface chamber and perfused (2–3 mL/min) with artificial cerebrospinal fluid (ACSF) containing (in mM, 35 °C): 125 NaCl, 3 KCl, 2 MgCl2, 2 CaCl2, 20.3 NaHCO3, 10 glucose except when extracellular Ca2+ concentration was altered (see below). For confocal imaging slices were transferred to a submerged chamber and perfused with ACSF containing (in mM, room temperature): 124 NaCl, 3 KCl, 2 MgCl2, 2 CaCl2, 26 NaHCO3, 10 glucose (95% O2, 5% CO2).
Field Excitatory Postsynaptic Potentials
fEPSPs were recorded as described (Dietrich et al. 1997). As criterion for the specificity of fiber tracts stimulated medial molecular layer (MML) versus outer molecular layer (OML), we accepted fEPSPs as originating from the lateral perforant path if they showed paired-pulse facilitation ≥ 120% and as originating from the medial perforant path in case of paired-pulse depression of ∼80% (McNaughton 1980; Dietrich et al. 1997).
Extracellular Solutions Containing Different Ca2+ Concentrations
For Ca2+ concentrations < 4 mM, the Mg2+ concentration of the ACSF was adjusted to maintain the total concentration of these 2 divalent cations at 4 mM. For 4 mM Ca2+, the concentration of Mg2+ was lowered to 0.1 mM and 25 μM D-2-amino-5-phosphonopentanoic acid (d-APV) was added to limit the activation of NMDA receptors.
We employed a HEPES-buffered saline to reliably avoid the partial precipitation of Ca2+ when using 10 and 20 mM Ca2+. Initial experiments were carried out to find the Ca2+ and Mg2+ concentration required in HEPES to produce an equivalent paired-pulse facilitation because the activity of these ions is expected to be higher in the absence of NaH2CO3. We found the following “normal HEPES” saline to be “paired-pulse equivalent” to the above-mentioned ACSF (in mM): 133 NaCl, 3 KCl, 2 CaCl2, 1 MgCl2, 19 glucose, 15 HEPES (pH 7.4, NaOH). The inhibition of fEPSPs by 3 μM 4-phosphonophenylglycine (PPG) was also similar in both solutions: to 60 ± 2.4% (n = 21) and to 58 ± 4.2% (n = 14) in ACSF and normal HEPES, respectively. To reliably identify fibers of the lateral perforant path, each experiment applying 10 or 20 mM Ca2+ was started in normal HEPES to verify paired-pulse facilitation and the inhibition of fEPSPs by PPG before high Ca2+ concentrations were applied to conduct the experiment. Ten millimolar Ca2+ HEPES contained (in mM): 133 NaCl, 3 KCl, 10 CaCl2, 1 MgCl2, 8 glucose, 7 HEPES, 0.025 d-APV (pH 7.4, NaOH). Twenty millimolar Ca2+ HEPES contained (in mM): 120 NaCl, 3 KCl, 20 CaCl2, 1 MgCl2, 5 glucose, 6 HEPES, 0.025 d-APV (pH 7.4, NaOH).
For the experiments involving application of a low (20 μM) concentration of Cd2+ to achieve a rapid and partial block of transmission (Fig. 4B), we also used HEPES saline to prevent complexation of Cd2+ with bicarbonate. These solutions contained (in mM): 133 NaCl, 3 KCl, 4 or 0.85 CaCl2, 1 MgCl2, 13 or 16 glucose, 15 HEPES, 0.025 d-APV (pH 7.4, NaOH).
Miniature Excitatory Postsynaptic Currents
“Blind” whole-cell patch-clamp recordings were obtained from hippocampal granule cells as described previously (Dietrich et al. 2002; Podlogar and Dietrich 2006) with a pipette solution containing (in mM): 125 potassium gluconate, 10 HEPES, 0.5 ethyleneglycol-bis(2-aminoethylether)-N,N,N′,N′-tetra acetic acid (EGTA), 2 MgCl2, 23 KCl, 3 NaCl, pH adjusted to 7.3. Voltages were corrected for a liquid junction potential by offsetting the amplifier to −10 mV before seal formation. Cells were held in voltage clamp at −70 mV, and membrane current was recorded nearly continuously in 10 s long sweeps. Miniature excitatory postsynaptic current (mEPSCs) were recorded from hippocampal granule cells at 35° using the ACSF described above. Series resistance ranged between 15 and 25 MOhm. Signals were filtered at 1 kHz and sampled at 10 kHz. mEPSCs were analyzed by the sliding template algorithm (Clements and Bekkers 1997) provided by NeuroMatic (V1.71) for Igor Pro (V5) with a threshold of 4. For each cell a template was created by averaging ∼50 manually selected very slow currents. All detected events were visually inspected and discarded in case of doubt (<20% in each cell).
Confocal Calcium Imaging
For imaging of intraterminal Ca2+ transients, fibers were loaded with the low-affinity Ca2+ indicator Mg-Green acetoxymethyl ester (Molecular Probes, the Netherlands) as described previously (Dietrich et al. 2003). Alternatively, slices were dye labeled on the stage of the confocal microscope according to the protocol described in Kukley et al. (2007): A glass pipette containing the Ringer-diluted dye was placed in stratum moleculare along the course of the axons for 30–45 min while applying a slight pressure to the back of the pipette. For data acquisition, line scans were obtained perpendicular to the orientation of the molecular layers at 1–3 kHz. To evoke action potentials and Ca2+ entry, fibers were stimulated with a glass electrode as for fEPSPs. Excitation wavelength was 488 nm, and the emitted fluorescence was collected by a 20× objective lens long-pass filtered at 505 nm. The pinhole was set to maximal size. Changes in fluorescence were quantified as ΔF/F. For illustration, the inhomogenity of fluorescence along the line scan was removed by normalizing on an average fluorescence line profile taken before stimulation (cf. Supplementary Figs 1 and 2). Calcium imaging experiments and electrophysiological recordings of synaptic transmission were always executed in different slices and not performed simultaneously to avoid any alterations of transmitter release by the presence of the presynaptic calcium indicator dye.
The following drugs were bath applied: (from Tocris): (R,S)-PPG, (S)-3,4-dicarboxyphenylglycine (DCPG), 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX), DL-APV, (RS)-α-cyclopropyl-4-phosphonophenylglycine (CPPG), (RS)-α-methyl-4-phosphonophenylglycine, (2S,2′R,3′R)-2-(2′,3′-dicarboxycyclopropyl)glycine (DCG), 4-(4-octadecyl)-4-oxobenzenebutenoic acid (OBAA), staurosporine, forskolin, tetrodotoxin, ZD7288, and U73122; (from Alexis): H-7, 1H-[1,2,4]oxadiazolo[4,3-a]quinoxalin-1-one (ODQ), mastoparan, ryanodine, lavendustin A, H-89, 2′,5′-dideoxyadenosine (DDA), and cis-N-(2-phenylcyclopentyl)-azacyclotridec-1-en-2-amine (MDL-12,330A); (from Calbiochem): 9-(tetrahydro-2-furanyl)-9H-purin-6-amine (SQ22536), pertussis toxin, and 2-(2-amino-3-methoxyphenyl)-4H-1-benzopyran-4-one (PD 98059); (from Sigma): ophiobolin A, baclofen, adenosine, 1,3-dipropyl-8-cyclopentylxanthine (DPCPX), calmidazolium, α-[amino-(4-aminophenylthio)methylene]-2-(trifluoromethyl)phenylacetonitrile (SL327), and MRS1191. Substances with intracellular targets were preapplied for at least 30 min to allow for trans-membranous diffusion.
Statistical significance was tested by the 2-tailed students t-test (α = 0.05) unless stated differently. All data are given as mean ± standard error of the mean.
fEPSPs of the lateral entorhinal cortex to hippocampal granule cell projection were evoked by stimulation and recorded in the outer molecular layer of the dentate gyrus. fEPSPs were highly sensitive to the group III mGluR agonist (RS)-PPG. A Hill fit to the concentration response curve revealed that PPG depresses synaptic transmission maximally by 79% with an EC50 (concentration of half-maximal effect) of 0.53 μM and a slope of 0.89 (Fig. 1A,B, n = 6). The action of 3 μM PPG was blocked in a concentration-dependent manner by the group III mGluR–specific antagonist (RS)-α-CPPG with an IC50 of 3.5 μM (Fig. 1B,C). The inhibition of fEPSPs was also seen when applying the mGluR8-specific agonist DCPG (1 μM, Supplementary Fig. 3). These data, together with previous work in mGluR8 knock-out mice, using mGluR8-specific antibodies and pharmacological characterization (Dietrich et al. 1997; Shigemoto et al. 1997; Zhai et al. 2002), strongly suggest that the stimulated synapses in the outer molecular layer contain a high density of presynaptic mGluR8.
Since it is a common scenario that presynaptic neurotransmitter receptors reduce transmitter release by inhibition of voltage-gated Ca2+ channels (Wu and Saggau 1997), we analyzed presynaptic Ca2+ entry by confocal Ca2+ imaging as described (Dietrich et al. 2003) (for details and controls, see Supplementary Figs 1 and 2). Lateral perforant path fibers were loaded with the low-affinity Ca2+ indicator Mg-Green. Fibers were stimulated as for fEPSPs, and fluorescent transients were selectively recorded in the OML in line scan mode (Fig. 2A). The response amplitudes to 3 stimulations at 25 Hz did not diminish (Fig. 2A, 101 ± 2% and 101 ± 2% for the second and third amplitude, respectively, n = 14) suggesting that the indicator dye was far from saturation. Combined application of 10 μM CNQX and 50 μM APV did not affect the fluorescent transient (not shown, n = 3) consistent with the presynaptic origin of the signals. Furthermore, we used patch-clamp and fluorescent recordings from dentate granule cells to verify the absence of a postsynaptic Ca2+ signal (Supplementary Fig. 2).
Contrasting with many other presynaptic receptors, activation of mGluR8 by 3 μM PPG did not affect presynaptic Ca2+ entry (Fig. 2B, n = 6) while the same concentration of PPG significantly reduced transmitter release (Fig. 2C, n = 6). Identical results were obtained with the specific mGluR8 agonist DCPG: At 1 μM, DCPG reduced fEPSPs to 30 ± 2% (n = 4) but did not notably decrease presynaptic Ca2+ entry (n = 4, Supplementary Fig. 3). To verify that we were able to track changes in presynaptic Ca2+ entry, we performed imaging experiments and fEPSP recordings in the MML, where fibers of the medial perforant pathway carrying group II mGluRs are found (Macek et al. 1996; Dietrich et al. 1997; Shigemoto et al. 1997). As shown in Figure 2D, the group II mGluR-specific agonist DCG profoundly reduced presynaptic Ca2+ entry at the medial perforant path-granule cell synapse. With the published value of Ca2+ cooperativity of transmitter release at this synapse (3.5, Qian and Noebels 2001) the reduction of presynaptic Ca2+ entry by DCG almost completely explains the amount of depression of synaptic transmission at medial perforant path synapses (Fig. 2D,E).
To substantiate the conclusion that mGluR8 inhibits transmitter release independently of voltage-gated Ca2+ channels, we tested whether PPG would modulate the frequency of mEPSCs when Ca2+ channels are already blocked by Cd2+. Nearly complete blockade of Ca2+ channels was obtained by perfusion of 100 μM Cd2+ (Fig. 3A,B). In the presence of Cd2+, 1 μM tetrodotoxin and 50 μM picrotoxin mEPSCs were frequently observed in whole-cell recordings from granule cells in hippocampal slices (Fig. 3C). Because only synapses contacting the most distal portion of the dendrites of granule cells carry mGluR8, we aimed at identifying those synaptic currents that originated in this distal area (Macek et al. 1996; Dietrich et al. 1997; Shigemoto et al. 1997). Due to electrotonic filtering, these distal synaptic currents can be assumed to show a particularly slow rise time when compared with currents arising more proximally (Schmidt-Hieber et al. 2007). Therefore, we designed a detection template (see Materials and Methods) to preferentially identify currents originating from lateral perforant path synapses by averaging ∼50 mEPSCs which showed the slowest rise times during an initial screening analysis in each cell (Fig. 3D, rise time10–90% = 2.6 ± 0.1 ms, τdecay = 11 ± 1.9 ms, n = 4). mEPSCs identified with this template were clearly reduced in frequency (to 47 ± 7%, n = 4) but not in amplitude (105 ± 4%, n = 4) by the application of PPG despite the preblock of Ca2+ channels by Cd2+ (Fig. 3D,E). For comparison, we verified in the same cells that the frequency of synaptic currents detected with fast rising and decaying templates (rise time10–90% = 0.7 ± 0.15 ms, τdecay = 3.9 ± 0.6 ms) was not reduced by PPG (114 ± 30% of control, not significant, n = 4).
We further explored the inhibition of action potential–stimulated release by relating the depression of fEPSPs by PPG to that by the well-known blockers of voltage-gated Ca2+ channels ω-conotoxin GVIA (1 μM) and Cd2+ (20 μM, yielding a partial block). In order to compare the dependence of release inhibition on the local intracellular Ca2+ concentration produced by the presynaptic action potential, we varied the amount of Ca2+ entry by altering the extracellular Ca2+ concentration. As can be seen in Figure 4A,B, ω-conotoxin GVIA (n = 6 and 4 for high and low Ca2+) and Cd2+ (n = 7 and 4) are much less effective at inhibiting release in 4 mM extracellular Ca2+ when compared with 1/0.85 mM Ca2+. In contrast, the potency of PPG only weakly, though statistically significantly, depended on the extracellular Ca2+ concentration (Fig. 4C,D, n = 11 and 9). We further analyzed this differential dependence on extracellular Ca2+ more quantitatively.
Mechanistically, there are 3 general categories by which a presynaptic metabotropic receptor can achieve an inhibition of transmitter release. First, it can decrease the local free Ca2+ concentration at the Ca2+ sensor of vesicle fusion, either by directly affecting Ca2+ channels or indirectly, for example, by modulation of action potential properties. Second, it can reduce the Ca2+ sensitivity of the Ca2+ sensor without changing the free Ca2+ concentration. Third, it can reduce the maximal release probability, for example, by reducing the overall number of docked vesicles or by decreasing the maximal fusion willingness of docked vesicles. Whereas the first mechanism seems ruled out based on the above data, the latter 2 are possible modes of action for mGluR8 to depress transmitter release. In the following, we use the fact that these 2 mechanisms show a differential dependence on Ca2+ entry to identify their contribution to the transduction of mGluR8.
The Ca2+ dependence of action potential–induced transmitter release is well described by an n-step, cooperative Ca2+-binding scheme (n ∼ 4 to 5) (Heidelberger et al. 1994; Bollmann et al. 2000; Schneggenburger and Neher 2000; Lou et al. 2005). The release resulting from such binding schemes at various Ca2+ concentrations can be fitted by a Hill function with an apparent cooperativity of napp < n:
Equation (2) is plotted in the lower panel of Figure 5A for 3 different values of a (i.e., different degrees of reduction in local Ca2+) and illustrates that the decrease in transmitter release by blocking Ca2+ channels can be completely overcome provided the initial Ca2+ concentration (before a channel-blocking agent is applied) at the sensor is sufficiently elevated (e.g., to more than 3-fold the control level in this case, Fig. 5A, lower panel).
In analogy to equation (2), we can derive a formalism describing the effect of decreasing the Ca2+ sensitivity of the release machinery (category 2) on transmitter release. The decrease in Ca2+ sensitivity is represented by multiplying KA with the factor b (b ≥ 1):
It is important that equation (3) is formally equivalent to equation (2) if a = 1/b. Therefore, altering Ca2+ sensitivity by multiplying KA with b shows the same dependence on local Ca2+ as reducing the increase in free Ca2+ by a factor (a = 1/b).
In contrast, a reduction in maximal release probability (category 3) cannot be overcome (by definition) and will reduce overall transmitter release to a certain fraction irrespective of the local Ca2+ concentration seen by the sensor. A category 3 mechanism can be modeled by multiplying equation (1) with a positive factor, for example, max ≤ 1. Taken together, constructing an experimental fractional release curve as exemplified in Figure 5A, lower panel, should allow us to decide whether mGluR8 depresses transmitter release by a category 2 or 3 mechanism (category 1 being excluded experimentally by Ca2+ imaging).
We performed a series of experiments with ω-conotoxin GVIA to test the predictions of our model more comprehensively and to verify that the effect of ω-conotoxin GVIA on fEPSPs follows equation (2). We systematically altered the local Ca2+ concentration seen by the release sensor by varying the extracellular Ca2+ concentration. To construct a fractional release curve as shown in Figure 5A, changes in extracellular Ca2+ concentration have to be converted into the respective relative changes of the local intracellular Ca2+ concentration. For this, we first measured how Ca2+ entry depends on extracellular Ca2+. We have previously used whole-cell recordings and axonal Ca2+ imaging to register Ca2+ influx under various concentrations of extracellular Ca2+ (Kukley et al. 2007; Muller et al. 2007). These values as well as that of other investigators in related slice preparations (Mintz et al. 1995; Qian and Noebels 2001) are well fit by a function for a bi-molecular reaction scheme describing Ca2+ channel pore occupancy by extracellular Ca2+ (KD set to 3.3 mM, Akaike et al. 1978; Carbone and Lux 1987; Hille 1992) (Fig. 5B):
The changes in whole-cell Ca2+ currents (Muller et al. 2007) and in global intracellular Ca2+ (Mintz et al. 1995; Qian and Noebels 2001; Kukley et al. 2007) described by equation (4) reflect variations in single channel current (as reasoned in Sinha et al. 1997; Muller et al. 2007). Furthermore, because microdomain Ca2+ is proportional to the single channel Ca2+ current under most conditions (as shown in Supplementary Fig. 4), the output of equation (4) also sufficiently describes the relative changes in local Ca2+ at the release sensor.
We measured the reduction of fEPSPs by ω-conotoxin GVIA at 6 different extracellular Ca2+ concentrations and plotted the fractional fEPSP amplitudes remaining after toxin application versus the relative changes in the local Ca2+ concentration calculated from equation (4). For this, the local Ca2+ concentration achieved under control conditions (2 mM [Ca2+]ex) was set to 1 (Fig. 5C, by scaling the output of eq. 4 by 1.11). As predicted, the elevation of extracellular Ca2+ can almost completely overcome the inhibition by the toxin (red dots). The data points are well fitted by equation (2) with a = 0.81, napp = 3.26, KA = 1.25 (red line). This indicates that transmitter release at the lateral perforant path synapse can be adequately approximated by equation (1) and that equation (2) well models the effect of a category 1 mechanism. The fit parameters indicate that ω-conotoxin GVIA–sensitive (N-type) Ca2+ channels contribute ∼19% (1–0.81) to the total action potential–induced elevation of the local Ca2+ concentration at the sensor. Furthermore, napp and KA provide estimates of the Ca2+ cooperativity and affinity of the release machinery at this synapse.
Next, a similar series of experiments was performed for activation of mGluR8 by PPG. As shown in Figure 5C, PPG, though to a smaller degree, also becomes less effective in depressing transmitter release when Ca2+ influx is strongly elevated (black dots). Since experiments were performed at the same synapse, the fitting procedure with equation (3) was constrained by the values of napp and KA as determined by ω-conotoxin GVIA applications. The best fit only poorly matched the data points (Fig. 5C, gray line) indicating that a pure category 2 mechanism cannot explain the effect of mGluR8 at this synapse. Hence, we introduced an additional parameter into equation (3) representing a Ca2+-independent inhibition of maximal release (max, category 3):
Equation (5) represents a dual action on transmitter release via category 2 and 3 and significantly better approximates the data points of PPG in Figure 5C (black line) despite the additional degree of freedom (P < 0.001, F-test). The best estimates for max and b were 0.73 and 1.27, respectively. This means that the Ca2+ dependence of the PPG effect can be best explained if mGluR8 is assumed to decrease both the overall maximal release probability to 73% and the Ca2+ affinity by increasing KA by 27%. The 2 components of the action of mGluR8 are drawn separately in Figure 5C (dashed lines combine multiplicatively to yield the black line). This graph shows that under most settings of Ca2+ entry, the inhibition of fEPSPs by the effect of mGluR8 on Ca2+ affinity is clearly more pronounced. This situation only changes if Ca2+ entry is increased more than ∼1.6-fold (intersection of curves) as compared with basal levels (2 mM [Ca2+]ext). Importantly, if we do not constrain the parameters napp and KA with the values of the fit of the toxin data, we still obtain very similar values for max and b with 0.78 and 1.29, respectively, by fitting equation (5) (Supplementary Fig. 5, red line). Moreover, in this case, even the estimates for the basal release parameters are well comparable to the ones estimated by ω-conotoxin GVIA application (napp = 3.23, KA = 1.39), showing the robustness of our approach.
To further substantiate this robustness, we performed an additional curve fitting analysis in which we fit both the toxin data and the PPG data with equation (5) simultaneously without constraining any parameters. Also in this case, extending the equation by the parameter max led to a significantly better fit (P < 0.001, F-test), and we derived very similar parameter estimates for ω-conotoxin GVIA: a = 0.82 and for PPG: b = 1.26, max = 0.73 and for the common parameters kA and napp of 1.2 and 3.40, respectively (for details, refer to Supplementary Fig. 6).
The fractional release curves shown in Figure 5C are based on our estimates of the changes in Ca2+ entry upon altering the extracellular Ca2+ concentration. Because the fit used to calibrate the Ca2+ entry strongly extrapolates our data points from 4 mM up to 20 mM of extracellular Ca2+ (for norm. [Ca]local,ctrl = 2.32, compare for Fig. 5B), we asked whether our conclusions regarding the downstream targets of mGluR8 are the same if we assume a different calibration curve. Supplementary Figure 7 shows that the parameters b, max, and KA are quite insensitive to changes in this calibration because the calibration curve affects both the fraction of release curve of PPG and of conotoxin. The only parameter notably affected by a different calibration is napp (Supplementary Fig. 7) which, however, does not influence our conclusions regarding the mechanism of action of mGluR8. Finally, our conclusion is also not affected by the fact that we do not analyze individual synapses and record population synaptic responses (Supplementary Fig. 8).
Action potential–induced Ca2+ entry into synapses is not only important for glutamate release but also for refilling the pool of releasable vesicles in neuronal synapses (Stevens and Wesseling 1998; Wang and Kaczmarek 1998; Hosoi et al. 2007). It has been shown that the accumulation of presynaptic Ca2+ during repetitive synaptic activity can accelerate vesicle recruitment by as much as 10-fold (Hosoi et al. 2007). Therefore, one functional consequence of mGluR8's lack of effect on Ca2+ entry is that synaptic transmission during a train of action potentials for the period of inhibition by mGluR8 should be greater when compared with a similar train of action potentials undergoing depression of release by inhibition of voltage-gated Ca2+ channels. To test this prediction, we compared repetitive synaptic responses when transmission was depressed to the same degree either by PPG or by the Ca2+ antagonist Co2+, which blocks but does not permeate Ca2+ channels. The low-affinity Ca2+ channel blocker Co2+ was chosen for these experiments because it quickly washes in and out of the slice preparation and facilitates the titration of the concentration to achieve an inhibition equal to that of 3 μM PPG. We applied 600 μM Co2+ which depressed fEPSPs to 56 ± 1% (n = 11), very similar to PPG (to 58 ± 2%, n = 9, Fig. 6A,B,D). Once a steady depression of fEPSPs was reached, we stimulated lateral perforant path fibers 35 times at 25 Hz. In the presence of either inhibitor, there was a pronounced initial and transient facilitation of transmission as expected because release probability was decreased (Fig. 6A–C, third pulse: 159 ± 4% and 155 ± 4% for PPG, n = 9, and Co2+, n = 11, respectively). This facilitation declined in both groups presumably because the supply of vesicles became a limiting factor. However, after the first ∼250 ms (5–7 pulses), the responses under Co2+ clearly depressed more strongly while larger fEPSP amplitudes were maintained in the presence of PPG. At the end of the train relative amplitudes amounted to 89 ± 5% and 67 ± 5% under PPG and Co2+, respectively. This difference is consistent with the mGluR8's lack of effect on Ca2+ entry.
It is generally assumed that group II and III mGluRs act through inhibition of adenylate cyclase (Conn and Pin 1997; Anwyl 1999). However, direct evidence that the inhibition of adenylate cyclase is necessary or causative for the reduction of transmitter release is still lacking. To test the involvement of adenylate cyclase in the action of mGluR8 at the lateral perforant path synapse, we inhibited this enzyme by SQ22536 (800 μM) which should occlude the depressing effect of PPG if adenylate cyclase is required. To rule out side effects of SQ22536 on adenosine receptors (Schulte and Fredholm 2002), we preapplied DPCPX (3 μM) as well as MRS1191 (2 μM) to block A1 and A2 receptors, respectively. We verified the effectiveness of inhibition of adenylate cyclase with 3-isobutyl-1-methylxanthine (IBMX, 50 μM). IBMX reduces the breakdown of cAMP and elevates intracellular cAMP levels, if adenylate cyclase is active. Under control conditions (in the presence of A1 and A2 blockers), IBMX increased fEPSPs amplitude as expected by an elevation of presynaptic cAMP levels (Weisskopf et al. 1994; Sakaba and Neher 2001) (Fig. 7A, n = 7). The increase of fEPSPs was effectively prevented by 40 min of SQ22536 application (Fig. 7A, n = 5) showing that adenylate cyclase activity is inhibited. However, SQ22536 did not have an effect on fEPSPs itself nor on the inhibition of fEPSPs by PPG (Fig. 7C, n = 6). Similar results were obtained with another adenylate cyclase inhibitor, DDA (500 μM, n = 4, Fig. 7B) making it unlikely that mGluR8 reduces transmitter release via downregulation of adenylate cyclase.
We next examined a series of other membrane-permeable inhibitors of known G-protein pathways for potential interference with the action of mGluR8 on transmitter release. In each experiment, we first tested the response of fEPSPs to PPG, then applied the inhibitor, typically for 30 min, before we retested the response to PPG (cf. Fig. 8B). As a control, we applied PPG twice (right most bar in Fig. 8A, n = 4). With this approach, we tested for the involvement of protein kinases C, A, and G (1 μM staurosporine, n = 5, 100/30 μM H7/H89, n = 4), Calmodulin signaling (3 μM calmidazolium, n = 5; 25 μM ophiobolin A, n = 4), phospholipases C and A2 (5 μM U73122, n = 7; 5 μM OBAA, n = 4), tyrosine kinase (10 μM lavendustin A, n = 4), nitric oxide-sensitive guanylate cyclase (10 μM ODQ, n = 4), and the mitogen-activated protein kinase pathway (30 μM PD 98059, n = 6, 10 μM SL 327, n = 6) (Fig. 8A). Except lavendustin A, none of the drugs reduced the inhibition of fEPSPs by mGluR8 significantly (paired t-test). The smaller effect of PPG under lavendustin A, albeit statistically significant, appears not to be physiologically relevant as it is still 94 ± 2% of control conditions.
The main finding of the present study is that the inhibition of action potential–driven glutamate release by mGluR8 is exclusively due to a modulation of the release machinery and that it can be best explained, mechanistically, by a ∼30% decrease in the Ca2+ affinity of the release sensor combined with a 25% reduction of maximal release rate. As a result, the inhibition by mGluR8 shares with classical presynaptic inhibition the sensitivity to changes in presynaptic Ca2+ levels. However, it differs from the latter as it cannot be completely overcome by presynaptic Ca2+ increases and it does not interfere with the refilling process of released vesicles. Hence, autoinhibition by mGluR8 can provide an emergency brake against an overdose of glutamate release and converts the inhibited synapse to a high-pass filter when compared with classical presynaptic inhibition via Ca2+ entry.
We were able to mechanistically identify the targets of mGluR8 signaling in typically small and inaccessible native synaptic nerve terminals by an approach which takes advantage of the saturating nature of the Ca2+-dependent transmitter release process. Previous studies of the Ca2+ sensitivity of transmitter release, performed in some of the few preparations where the presynaptic Ca2+ level can be directly manipulated, clearly showed that transmitter release rates saturate upon increasing Ca2+ concentrations (Heidelberger et al. 1994; Bollmann et al. 2000; Schneggenburger and Neher 2000; Lou et al. 2005). Furthermore, and as suggested by the same work, we assume that the local intracellular Ca2+ concentration is necessary and sufficient to determine vesicle fusion. Our model does not include any effect of voltage on the release process as suggested by Parnas and Parnas (2007). This neglect may be justified by the observation that mGluR8 is not activated by ambient glutamate (Bushell et al. 1995; Bushell et al. 1996; Friedl et al. 1999) which is required by the “calcium-voltage” hypothesis (Parnas and Parnas 2007). The basic idea of our approach was to use this saturation to distinguish presynaptic mechanisms which act upstream of Ca2+ binding to the release sensor from downstream mechanisms by the fact that the former but not the latter will be rendered ineffective if the release machinery is brought to near-saturation by elevating presynaptic Ca2+ entry. It was found that the Ca2+ dependence of the inhibition of release by mGluR8 could not adequately be explained by a sole reduction in Ca2+ binding to the sensor. Instead, we had to assume a dual action of mGluR8 on both calcium binding to the sensor and the maximal release rate. Separating the 2 components of mGluR8 signaling algebraically showed that under normal conditions and upon low Ca2+ entry the effect of mGluR8 on Ca2+ binding to the release sensor is primarily responsible for the inhibition of glutamate release.
Our formalism predicts that mGluR8 reduces Ca2+ binding to the release sensor but it does not per se resolve whether this is caused by an effect on the Ca2+-binding properties of the sensor itself, the idea favored by us, or by a change in the very local Ca2+ seen by the sensor. For example, mGluR8 may selectively inhibit only those Ca2+ channels which are directly associated with synaptic vesicles and which may carry only an insignificant fraction of total Ca2+ entry. Alternatively, mGluR8 may, by some means, increase the diffusional distance between Ca2+ channels and the sensor. However, neither mechanism explains on its own why mGluR8 reduces the frequency of mEPSCs when Ca2+ channels are blocked by Cd2+: under these conditions, channel opening and concomitant Ca2+ diffusion from voltage-gated calcium channels to the release machinery does not occur, consequently mGluR8 must reduce mEPSC frequency in a different way. This reduction could be caused by an mGluR8-dependent modulation of spontaneous calcium release events from presynaptic calcium stores. Such events have been demonstrated to be responsible for a significant fraction of mEPSCs at other synapses (Collin et al. 2005). However, an effect of mGluR8 on calcium stores has not been described and would also not explain the inhibition of release triggered by single action potentials. Taken together, the most parsimonious explanation of our experimental data is that mGluR8 directly modulates the release sensor.
It is important to identify the downstream targets of presynaptic modulation because the functional implications differ for a decrease in affinity versus a change in maximal release rate. The degree of inhibition via the former but not the latter mechanism is dependent on the actual amount of Ca2+ entry. Changes in Ca2+ entry are known to occur during physiological settings when a synapse is repetitively activated: broadening of presynaptic action potentials can increase (Jackson et al. 1991; Byrne and Kandel 1996; Geiger and Jonas 2000), while depletion of extracellular Ca2+ (Rusakov and Fine 2003) or inactivation of presynaptic Ca2+ channels (Forsythe et al. 1998) can decrease Ca2+ entry. Thus, the degree of inhibition via a reduction in Ca2+ binding to the release machinery (category 1 and 2) is dependent on previous activity, whereas further downstream mechanisms (category 3) will proportionally scale down synaptic strength independent of the history of activity. The different functional meanings clearly emerge, when considering natural repetitive patterns of neuronal activity, which involve vesicle replenishment, short-term synaptic plasticity and changes in presynaptic Ca2+ entry.
What could be the signaling pathway from mGluR8 to the components of the release machinery? Classically, group III mGluRs, such as mGluR8, negatively couple to adenylate cyclase (Conn and Pin 1997). A reduction in cAMP levels by group III mGluRs has been shown in expression systems and in native slices (Bedingfield et al. 1995; Wu et al. 1998) but it was suggested that the inhibition of spontaneous transmitter release by mGluR8 in mitral cells may be independent of presynaptic cAMP levels (Schoppa and Westbrook 1997). This finding is supported by our study 2-fold: First, adenylate cyclase inhibitors changed neither basal synaptic transmission nor the inhibition of release by mGluR8. We tested various other inhibitors of known G-protein transduction pathways, but none of them prevented the action of mGluR8. A possible alternative explanation may be that activated mGluR8 releases Gβ/γ-subunits which directly interact with the release machinery as shown for 5-HT receptors at a noncortical synapse (Blackmer et al. 2001). Another scenario of a direct interaction of a G-protein–coupled receptor with the release machinery has been reported at the mouse neuromuscular junction: presynaptic muscarinic receptors directly inhibit the release apparatus without an involvement of a G-protein or Gβ/γ-subunits (Parnas and Parnas 2007; Kupchik et al. 2011). However, further experiments are required to provide direct evidence for one of these scenarios.
A decrease in Ca2+ affinity likely reflects a direct or indirect modulation of a component of the release machinery which is responsible for Ca2+ binding and initiating the fusion process. A decreased maximal release probability by mGluR8 could also arise by affecting steps not directly involved in Ca2+-dependent membrane fusion. A recent study indeed reports that noradrenergic presynaptic receptors completely deactivate certain release sites, while others are unaffected (Delaney et al. 2007). This mechanism is, however, unlikely to account for the effect observed in our study because it predicts that paired-pulse facilitation upon activation of the presynaptic receptor remains unchanged (Delaney et al. 2007). In contrast, we observed that in high extracellular Ca2+ concentrations (20 mM), when the effect of mGluR8 is nearly exclusively mediated by the effect on maximal transmitter release (cf. Fig. 5C), there is a strong increase in the paired-pulse index from 76 ± 3% to 103 ± 2% (data not shown, n = 5). At the calyx of Held, a secondary effect of presynaptic γ-aminobutyric acid (GABA)-B receptors on vesicle recruitment during sustained activity has been shown and provides another example of inhibition of transmission independent of changing Ca2+ binding to the release sensor (Sakaba and Neher 2003). However, considering that sustained phasic release during repetitive stimulation is more prominent when compared with Ca2+ channel blockade, an inhibition of vesicle supply by mGluR8 is also not suggested by our data. Thus, our study suggests a modulation of Ca2+-sensing components of the fusion machinery and our functional study may provide useful guidelines to identify appropriate candidates.
Supplementary material can be found at: http://www.cercor.oxfordjournals.org/
Deutsche Forschungsgemeinschaft (DI853/2, DI853/4, SFB TR3 D11, B8, Emmy Noether Programme) and University Clinic Bonn grants (BONFOR).
We thank P. Stausberg, S. Buchholz, and J. Enders for excellent technical assistance. Conflict of Interest : None declared.