Presynaptic NMDA receptors act as local high-gain glutamate detector in developing cerebellar molecular layer interneurons


  • Bénédicte Rossi,

    1. Laboratoire de Physiologie Cérébrale, CNRS-UMR 8118, Université Paris Descartes, Université Paris Diderot, Paris, France
    Search for more papers by this author
  • Thibault Collin

    Corresponding author
    • Laboratoire de Physiologie Cérébrale, CNRS-UMR 8118, Université Paris Descartes, Université Paris Diderot, Paris, France
    Search for more papers by this author

Address correspondence and reprint requests to Thibault Collin, Laboratoire de Physiologie Cérébrale, CNRS-UMR 8118, Université Paris Descartes, Université Paris Diderot, 45 rue des Saints Pères, 75006 Paris, France. E-mail:


In the classical view, NMDA receptors (NMDARs) are located postsynaptically and play a pivotal role in excitatory transmission and synaptic plasticity. In developing cerebellar molecular layer interneurons (MLIs) however, NMDARs are known to be solely extra- or presynaptic and somewhat poorly expressed. Somatodendritic NMDARs are exclusively activated by glutamate spillover from adjacent synapses, but the mode of activation of axonal NMDARs remains unclear. Our data suggest that a volume transmission is likely to stimulate presynaptic NMDARs (preNMDARs) since NMDA puffs directed to the axon led to inward currents and Ca2+ transients restricted to axonal varicosities. Using local glutamate photoliberation, we show that pre- and post-synaptic NMDARs share the same voltage dependence indicating their containing NR2A/B subunits. Ca2+ transients elicited by NMDA puffs are eventually followed by delayed events reminding of the spontaneous Ca2+ transients (ScaTs) described at the basket cell/Purkinje cell terminals. Moreover, the presence of Ca2+ transients at varicosities located more than 5 μm away from the uncaging site indicates that the activation of preNMDARs sensitizes the Ca2+ stores in adjacent varicosities, a process that is abolished in the presence of a high concentration of ryanodine. Altogether, the data demonstrate that preNMDARs act as high-gain glutamate detectors.

Abbreviations used



action potential


cyclopiazonic acid


molecular layer interneuron








Oregon Green BAPTA 1


parallel fiber


presynaptic NMDA receptor



NMDA receptors (NMDARs) are nonspecific cationic ionotropic glutamatergic receptors that are located in both synaptic and extrasynaptic compartments. There is also growing evidence for the existence of presynaptic NMDARs (preNMDARs) in the cortex (Berretta and Jones 1996; Sjöström et al. 2003), brainstem (Bach and Smith 2012), neocortex (Buchanan et al. 2012) and cerebellum (Glitsch and Marty 1999; Casado et al. 2002; Duguid and Smart 2004; Duguid et al. 2007; Rossi et al. 2012). Bath-applied NMDA has been shown to enhance action potential (AP)-independent GABA release onto Purkinje cells and molecular layer interneurons (MLIs) through preNMDAR located on MLIs axons (Glitsch and Marty 1999; Rossi et al. 2012). Huang and Bordey (2004) showed that application of the glutamate transporter antagonist TBOA increased the frequency of mIPSC recorded on Purkinje neurons, an increase that was reversed in the presence of AP-5. These authors concluded that this effect results from the activation of NMDARs situated on GABAergic terminals or on the axons and their data indicated that synaptically released glutamate could enhance GABA release through preNMDAR opening. Alternatively, a retrograde glutamate signaling from Purkinje cells has also been proposed as a mean to activate preNMDARs in the context of depolarization-induced potentiation of inhibition (DPI; Duguid and Smart 2004). At the parallel fiber (PF) to MLI synapse, only AMPARs are recruited with small PF activity and it is only with intense stimulation that NMDARs get involved in the postsynaptic response (Carter and Regehr 2000; Clark and Cull-Candy 2002). Therefore, the glutamate released from few PF terminals does not activate NMDARs because of their extrasynaptic localization (Clark and Cull-Candy 2002). However, a high frequency stimulation of PF which has been shown to generate a glutamate spillover sufficient to activate dendritic NMDARs in MLIs (Carter and Regehr 2000; Clark and Cull-Candy 2002) was suggested to promote a Ca2+ entry into the presynaptic terminals through preNMDARs leading to long term potentiation of GABAergic synapses (Liu and Lachamp 2006). In concrete terms, these data have indirectly shown that preNMDARs behave as glutamate sniffers allowing the cell to detect a spillover from adjacent excitatory synapses as well as a retrograde glutamate release (see Duguid and Smart 2004). Nonetheless, no direct evidence has been brought up so far that preNMDARs can be activated through volumic transmission. Another feature of MLIs NMDARs is their rather poor expression, the NMDAR-related component of an EPSC being always of minor importance compared to its AMPAR-related counterpart (Carter and Regehr 2000; Rossi et al. 2008). However, bath application of NMDA leads to a large increase in mIPSC rate (Glitsch and Marty 1999) which appears disproportionate to the amplitude of the NMDARs currents. Accordingly, several studies have shown that an amplification pathway based upon the recruitment of ryanodine receptors is necessary to observe the effect of preNMDARs activation (Duguid and Smart 2004; Huang and Bordey 2004; Rossi et al. 2012). Nonetheless, the involvement of ryanodine receptor in these processes has always been inferred from an indirect parameter such as the frequence of the mIPSC or the amplitude of evoked IPSCs.

In a previous report, we have shown that NMDA triggers axonal Ca2+ rises under somatic voltage clamp and we have detected discrete spots of high sensitivity to NMDA within the axon using uncaging of NMDAR agonists (Rossi et al. 2012). In this study, we set up a protocol to simulate glutamate spillover by a puff of NMDA directed on the axonal compartment of voltage-clamped MLIs with the aim to directly tackle the glutamate sensing ability of preNMDARs. Using this paradigm, we were able to trigger Ca2+ transients along the axon and to examine their dynamic properties. Interestingly delayed Ca2+ signals, reminiscent of the spontaneous Ca2+ transients (ScaTs; Conti et al. 2004), were frequently observed in axons responding to NMDA. Moreover, the presence of ryanodine reduced the amplitude of the Ca2+ transients triggered by local uncaging of 4-methoxy-7-nitroindolinyl-caged-l-glutamate (MNI-glutamate) performed in the presence of 2,3-dihydroxy-6-nitro-7-sulfamoyl-benzo[f]quinoxaline-2,3-dione (NBQX) without affecting the inward currents. Altogether, our results directly demonstrate that preNMDARs act as high-gain glutamate sensors since they can convert the presence of NMDA into a Ca2+ entry further amplified by mobilizing the internal Ca2+ stores through a ryanodine-sensitive pathway. Such a synergistic signalization is likely to support an efficient spillover detector function. Our data also suggest that a few NMDARs distributed along the axon can support the large increase in mIPSCs frequency observed with NMDA in the bath.



Experiments were carried out on Sprague–Dawley rat of both sexes that were provided by Elevage Janvier (Saint Berthevin, Elevage Janvier, France) and kept at the animal house of the Centre Universitaire des Saints Pères which has been approved by the ‘Préfecture de Police’ following inspection by Veterinary Services and the French Ministry of Research and the Ministry for Health (European Directive 86/609/EEC/approval number A-750607). Postnatal day 12–16 rats were anaesthetized by inhalation of isoflurane and killed by decapitation. Parasagittal (200 μm) cerebellar slices were prepared using a vibroslicer (Leica VT1200S; Leica Biosystems, Wetzlar, Germany) in ice-cold bicarbonate buffered saline (BBS) solution containing the following (in mM): 125 NaCl, 2.5 KCl, 2 CaCl2, 1 MgCl2, 1.25 NaH2PO4, 26 NaHCO3, and 10 glucose (saturated with 95% O2 – 5% CO2), pH 7.3. The slices were incubated at ~ 34°C for 30 min and were then stored at 20–23°C. During recordings, the slices were superfused with the above solution (1–1.5 mL/min) at 20–23°C except for uncaging experiments. Cerebellar slices were observed using an upright microscope (Zeiss, Oberkochen, Germany) with DIC Nomarski optics and a 60X water immersion objective (Olympus, Tokyo, Japan). Stellate and basket cells were first identified by their location in the molecular layer and collectively referred to as MLIs.

Calcium imaging

For these experiments, WCR pipettes were filled with (in mM) 155 Cs Gluconate, 6 KCl, 4.6 MgCl2, 10 HEPES, 0.4 Na-GTP, and 4 Na-ATP pH 7.3 with CsOH supplemented with 0.05 Oregon Green 488 BAPTA-1 (OG1; Invitrogen, Grand Island, NY, USA). Alexa 488 (20 μM) was eventually added to resolve long axonal stretches. Tight-seal WCRs were obtained with borosilicate pipettes (4–6 MΩ) from superficial somata using an EPC-9 amplifier (HEKA Electronik, Darmstadt, Germany). Series resistance values ranged from 15 to 25 MΩ and were compensated for by 60%. Currents were filtered at 1.3 kHz and sampled at a rate of 250 μs/point. Digital fluorescence images were obtained using an excitation-acquisition system from T.I.L.L. Photonics (Munich, Germany). Briefly, to excite fluorescence of the Ca2+ dye OG1, light from a 75 W Xe lamp was focused on a scanning monochromator set at 488 nm and coupled, by a quartz fiber and a lens, to the microscope, equipped with a dichroic mirror and a high-pass emission filter centered at 505 and 507 nm, respectively. Images were acquired by a Peltier-cooled CCD camera (IMAGO QE; 1376 × 1040 pixels; pixel size: 244 nm after 53X magnification and 2 × 2 binning). To induce axonal [Ca2+]i rises, four APs at 20 ms intervals were produced by depolarizing the cell for 3 ms to 0 mV from a holding value of −70 mV which induces a propagated action potential (Tan and Llano, 1999). For focal application of NMDA, a patch pipette filled with a solution containing NMDA (50 μM) was coupled to a Picospritzer II (General Valve Corp., Fairfield, NJ, USA) system. It was placed above the slice surface, at roughly 10 μm from the recorded MLI axon. The duration of NMDA application was 0.5 s. In every hand-drawn regions of interest (ROIs) the average fluorescence was determined as a function of time as previously reported (Collin et al. 2005).

Local uncaging of MNI-glutamate

The single mode optical fiber output of a 405 nm diode laser (Point Source, Iflex 2000, Qioptiq, Luxembourg) was expanded with a 40 mm focal length positive lens in a dual LED lamphouse (OptoLED; Cairn Research, Kent, UK) and reflected with a 45° dichroic mirror (425 DCXR; Chroma, Bellows Falls, VT, USA) into the epifluorescence condenser of the microscope as described previously (Rossi et al. 2008). This optical system yielded a spot of approximately 1 μm diameter at the focal plane in a 100 μM pyranine solution as measured with the CCD camera. For MNI-glutamate uncaging, the perfusion needed to be turned off to minimize consumption of the cages. A HEPES-buffered solution supplemented with NaHCO3 to control internal pH was used (composition in mM: 135 NaCl, 4 KCl, 2 NaHCO3, 25 glucose, 3 CaCl2, 0.2 MgCl2, and 10 HEPES, pH 7.4 with NaOH). MgCl2 was raised up to 1 mM for the ‘physiological Mg2+’ experiments.


The following drugs applied in the bath were used when appropriate: NMDA, tetrodotoxin (TTX), Mibefradil, NBQX, ryanodine. All drugs were purchased at Ascent. Other chemicals were purchased at Sigma (Saint Louis, MO, USA). MNI-Glutamate was kindly provided by Tocris Bioscience (Boston, MA, USA).

Statistical analysis

Comparison between two experimental groups was performed using the paired or the unpaired Student's t-test. < 0.05 was considered significant.


Stimuli mimicking glutamate spillover activate preNMDARs

Previous work on preNMDARs in MLIs has priviledged very short stimuli to obtain highly local receptor activation (Rossi et al. 2012). These experimental conditions were selected to demonstrate the axonal nature of the receptors, but they do not necessarily reflect the conditions of activation of the receptors by physiologically relevant stimuli. An intense stimulation of parallel fibers (PFs) has been shown to activate iGluRs-mediated currents and Ca2+ transients in somato-dendritic and axonal compartments of MLIs following glutamate spillover (Carter and Regehr 2000; Rossi et al. 2008). With normal extracellular solution, in the presence of 2 mM Mg2+, and when cells were maintained at the physiological holding potential of −70 mV, most of the response was due to activation of AMPA receptors (AMPARs), and the NMDAR-mediated component was minor. To reveal potential axonal NMDAR activation, we included glycine (20 μM) in the external solution, we reduced the external Mg2+ concentration to 200 μM, and we tested various holding potentials. Mibefradil (20 μM) was added to the superfusate to ensure that T-type Ca2+ channels could not be activated by axonal depolarization as suggested (Christie and Jahr 2008; Myoga et al. 2009). We mimicked glutamate spillover by applying short (500 ms) puffs of NMDA (50 μM) (see 'Methods'). With internal fluorescent dyes, axons and dendrites were easily resolved in MLIs (Christie and Jahr 2008) allowing an adequate positioning of the puffing pipette (Fig. 1a). When the axon was morphologically identified, four propagated action potentials (APs) were elicited by 50 Hz somatic depolarizations (duration: 3 ms) from −70 to 0 mV (Fig. 1b). Since AP-induced Ca2+ transients are markedly larger in the axon than in dendrites (Llano et al. 1997), this test confirmed the axonal nature of the neurite. Following axon identification, TTX (0.2 μM) was added to the bath and local axonal Ca2+ signaling was examined in response to NMDA puffs. In five cells, the NMDA puff led to inward currents synchronized with Ca2+ transients (Fig. 1c). Current displayed amplitudes of 29 ± 2 pA (n = 5 cells) at a holding potential of −70 mV while their average amplitude was of 34 ± 4 pA (n = 5 cells) at −40 mV. The corresponding Ca2+ transients peaked with ΔF/Fo = 32 ± 5% at −70 mV and ΔF/Fo = 28 ± 4% at −40 mV (n = 5 cells). These data suggest that preNMDARs and extrasynaptic NMDARs are both susceptible to be activated by spillover of glutamate from adjacent synapses. Since no synaptic NMDARs have been found in MLIs, the volumic transmission appears as being the sole way to open NMDARs in these cells.

Figure 1.

preNMDARs are activated by a short puff of agonist. (a) An Oregon Green BAPTA 1 (OGB-1) filled molecular layer interneuron (MLI) is submitted to a puff of NMDA directed to its axonal compartment. The white square delimits the region of interest selected to illustrate Ca2+ signals. (b) Time course of ΔF/Fo for the axonal region of interest and simultaneous current recording obtained with four propagated APs. (c) Time course of ΔF/Fo for the axonal region of interest and simultaneous current recording obtained in response to a 500 ms axonal puff of NMDA performed at a holding potential of −70 mV (upper panel) or −30 mV (lower panel). The black arrowheads indicate the beginning of the stimulations.

Pre- and post-synaptic NMDARs share the same voltage dependence

The results depicted in Fig. 1c confirm the voltage dependence of preNMDARs in the presence of Mg2+. Moreover, the effects of bath NMDA on the frequency of mIPSCs have been shown to be inhibited by Zn2+ and ifenprodil (Rossi et al. 2012). Taken together these data indicate that preNMDARs contain either NR2A or NR2B subunits but also raise the question of putative differences between NMDARs of post- and pre-synaptic compartments. To address this issue we compared the current to voltage relationship of the receptors in axonal and somato-dendritic compartments. We used the laser-driven MNI-glutamate photoactivation method to selectively activate NMDARs in a given compartment. As previously shown, uncaging experiments provide a very precise control of iGluR activation in terms of time and space (Trigo et al. 2009). In this series of experiments, slices were incubated in a low Mg2+ (0.2 mM) HEPES-buffered saline supplemented with MNI-glutamate (final concentration, 0.9 mM) and NBQX (5 μM) for at least 0.5 h before the beginning of the experiment. MLIs were dialyzed with a Cs+ gluconate -based intracellular solution containing Oregon Green BAPTA 1 (OGB-1) and Alexa 488. The use of a Cs+-based internal solution allows for a better control of the voltage especially for highly depolarizing potentials and the use of short laser pulses allowed to repeat the stimulation without damaging the cells. The I-V curves recorded in the presence of Mg2+, are near linear above −30 mV and reverse close to 0 mV (Fig. 2). At more negative potentials, both I-V relationships exhibit a negative slope reflecting Mg2+ block (Nowak et al. 1984). Aside from amplitude scaling, no difference could be found between the I-V curves recorded in the pre- or post-synaptic compartments (Fig. 2aii and bii). The strong Mg2+-related voltage dependence observed in both cellular compartments suggests that in MLIs, both axonal and somatodendritic NMDARs contain either NR2A or NR2B subunits.

Figure 2.

Current to voltage relationship of pre- and post-synaptic NMDARs. (a) current to voltage relationship performed using presynaptic uncaging of 4-methoxy-7-nitroindolinyl-caged-l-glutamate (MNI-glutamate). (i) Typical traces obtained at the voltage indicated on the left in response to 300 μs laser pulse. (ii) Curve averaged from four distinct experiments. (b) Current to voltage relationship performed using somatodendritic uncaging of MNI-glutamate. (i) Typical traces obtained at the voltage indicated on the left in response to 300 μs laser pulse. (ii) Curve averaged from four distinct experiments. (c) MNI-glutamate was uncaged on axonal varicosities by a 300 μs laser pulse in the presence of a physiological Mg2+ concentration (2 mM) and 2,3-dihydroxy-6-nitro-7-sulfamoyl-benzo[f]quinoxaline-2,3-dione (NBQX) (5 μM). (i) Typical time courses of the Ca2+ signals obtained at −70 mV (black trace) and at −40 mV (red trace). (ii) Matching current traces at both holding voltage values (−70 mV, black trace; −40 mV, red trace). (d) summary of the results obtained under physiological Mg2+ conditions in terms of Ca2+ transient maximal amplitude (upper panel) and inward current peak value (lower panel). The same symbol (+, #, *) was used for the same cell. Average values ± SEM are displayed in red within the graphs and the holding voltage value is indicated under the lower panel.

We next used the same methodological paradigm to challenge the ability of preNMDARs to be activated under a physiological Mg2+ concentration (1 mM). Accordingly, 300 μs laser pulses were performed on axonal varicosities in MLIs filled with OGB1 and a Cs-based intracellular solution to eventually change the holding potential to a depolarized value (−40 mV). At a holding potential of −70 mV, current responses eventually associated with Ca2+ transients were hardly detectable. In a series of 24 cells, we were able to detect uncaging-evoked signals in only five varicosities distributed among three cells. The current values were very small (5 ± 2 pA; n = 5; N = 3; Fig. 2cii, d lower panel) and the Ca2+ transients displayed an average amplitude of 7 ± 3% (n = 5; N = 3; Fig. 2ci, d upper panel). The same stimulation was then applied when the holding voltage was changed to −40 mV eliciting a significant increase in both the peak current (16 ± 4 pA; n = 5; N = 3; Fig. 2cii, d lower panel) and the amplitude of the Ca2+ transient (19 ± 3%; n = 5; N = 3; Fig. 2ci, d upper panel). These data indicate that preNMDARs can be opened under physiological Mg2+ as initially proposed by Duguid and Smart (2004) and confirm their voltage dependence.

PreNMDARs drive SCaT-like Ca2+ signals

Puff delivery of NMDA revealed atypical forms of Ca2+ signaling. Whereas the axonal Ca2+ response to propagated APs was highly synchronized in all varicosities (Fig. 3b), the response to NMDA puffs (in the presence of TTX) displayed a significant amount of delay in some varicosities (between 0 and 3.5 s; 1.3 s on average, n = 4; Fig. 3c). A second delayed response which did not coincide with any current component (middle panel on Fig. 3c) appeared approximately 30 s after the initial Ca2+ rise (peak to peak interval: 28 ± 6 s, n = 3). These two features of Ca2+ signaling, delay in the initial response and/or appearance of a second delayed component, were observed in three of the five responding cells. The ΔF/Fo of the delayed event was 20 ± 7% (n = 3) whereas the main transient displayed a ΔF/Fo of 25 ± 8% (n = 3). The average ratio between transients elicited by four APs versus delayed events was 4.3. This suggests that delayed events have an amplitude comparable to that elicited by a single AP, and that these signals are large enough for driving neurotransmitter release. These presynaptic transients are reminiscent of localized events recorded in basket cell terminals called SCaTs (for ‘Spontaneous Calcium Transients’: Llano et al. 2000; Conti et al. 2004; for review see Ross 2012). However, given the very low rate of SCaTs in unstimulated conditions, the late Ca2+ rises illustrated in Fig. 3 are clearly causally linked to NMDA application. We therefore propose that activation of presynaptic NMDARs could amplify and prolong the response to spilled-over glutamate by facilitating the occurrence of SCaTs. These complex Ca2+ signals might be responsible for the strength of mIPSCs frequency increase obtained with bath NMDA (Glitsch and Marty 1999; Rossi et al. 2012).

Figure 3.

Puffing NMDA on the presynaptic compartment triggers complex Ca2+ signaling. (a) A 500 ms puff of NMDA directed on the axon on a molecular layer interneuron (MLI) filled with Alexa 488 (20 μM) and Oregon Green BAPTA 1 (OGB-1) (50 μM) is applied from the point indicated by the pipette drawing. The white squares (numbered 1, 2, and 3) delimit the regions of interest selected to measure the Ca2+ signals. Scale bar: 5 μm. (b) The cell was initially stimulated with four propagated APs. Upper panel: spatiotemporal profiles of the ΔF/Fo signals along the line drawn in (a). The regions of interest corresponding to (a) are indicated by numbers and arrowheads. Middle panel: time course of the ΔF/Fo transients measured at the locations indicated by the arrows in the upper panel; arrow color corresponds to the color used to plot each trace. Lower panel: Simultaneous current recording. The onset of the stimulus is indicated by a dashed line. (c) Axonal response to a 500 ms puff of NMDA. Upper panel: spatiotemporal profiles of the ΔF/Fo signals along the line drawn in (a). The regions of interest are identical to (b). Middle panels: time course of the ΔF/Fo transients measured at the locations indicated by the arrows in the upper panel; arrow color corresponds to the color used to plot each trace. Lower panel: Simultaneous current recording. The onset of the stimulus is indicated by a dashed line.

Local photorelease of glutamate triggers complex Ca2+ signals in MLIs axons

Using wide field uncaging of MNI-glutamate in the presence of AMPAR blockers, we have shown that most axons respond to NMDA with Ca2+ transients, but that this sensitivity is restricted to a small number of varicosities discretely distributed along the axons (Rossi et al. 2012). To increase our rate of success in recording axonal NMDA responses, we selected cells located at the surface of the slice. Under these conditions, we were able to routinely detect current and Ca2+ responses for light pulses as short as 100 μs, a procedure that is likely to substantially limit the spatial diffusion of the agonist. This technique avoids the caveats of the puffing method since a brief laser pulse is likely to deliver glutamate in a really confined area and therefore limits the diffusion of the agonist from its uncaging point. In that respect, we chose the longest stretches of axonal structures that could be entirely comprised in the same plane of focus. Figure 4a illustrates such an axon that was successively stimulated using 100 μs and 300 μs-long 405 nm laser pulses. Interestingly, Ca2+ transients were not restricted to the uncaging spot (spot 2, Fig. 4b and c) and moreover, the largest response did not always occur at the uncaging spot (Fig. 4b and c). In the illustrated instance, the largest transient consistently appeared in a varicosity located at a distance of approximately 5 μm from the uncaging spot (Fig. 4b and c, spot 3 and corresponding traces). Two other spots (spots 1 and 4, Fig. 4b and c) situated on both sides of the uncaging point also gave Ca2+ transients (Fig. 4b and c, traces 1 and 4). Finally, the remote Ca2+ responses observed in spots 1 and 4 (Fig. 4b and c) peaked ~ 150–200 ms after the one recorded in spot 1 (uncaging point). Spot 4 is of particular interest because it is located ~ 13 μm away from the uncaging spot. We next asked whether these remote responses result from the spatial diffusion of the photoreleased glutamate or if they reflect an intra-axonal Ca2+ release mechanism. The point spread function of our uncaging system in the horizontal plane could be fitted by a gaussian function with a half width of 0.74 μm (Fig. 4eii). Moreover, MNI-glutamate photoactivation by a 300 μs pulse failed to elicit any response if the target structure was moved farther than 4 μm from the laser spot (Fig. 4ei). Therefore, Ca2+ signals appearing at distances of 4 μm or more from the uncaging site cannot result from the diffusion of photoreleased glutamate. Out of six cells which displayed a preNMDAR-activated Ca2+ signal in a focusing plane containing several varicosities, five showed complex Ca2+ signals similar to those described in Fig. 4b and c. The results obtained in the five cells are summarized in Fig. 4d and show that the distance from the uncaging point is correlated with the delay between the peak of the Ca2+ transient recorded at the uncaging site and the peak of the signal at the remote site (R = 0.53). In light of these results, we propose that an intracellular process leading to Ca2+ mobilization is recruited by preNMDAR activation.

Figure 4.

Local activation of preNMDARs stimulates propagating Ca2+ signals. Typical experiment: the molecular layer interneuron (MLI) was filled with Alexa 488 (20 μM) and Oregon Green BAPTA 1 (OGB-1) (50 μM) and bathed in the presence of 4-methoxy-7-nitroindolinyl-caged-l-glutamate (MNI-Glutamate) (0.9 mM). (a) An axonal stretch of approximately 25 μm has been chosen for displaying four varicosities. (b) The axon is submitted to a 100 μs laser pulse. Upper panel: spatiotemporal profiles of the ΔF/Fo signals along the line drawn in (a). The regions of interest corresponding to (a) are indicated by numbers and arrowheads. Middle panel: time course of the ΔF/Fo transients measured at the locations indicated by the arrows in the upper panel; arrow color corresponds to the color used to plot each trace. Lower panel: Simultaneous current recording. The onset of the stimulus is indicated by a dashed line. (c) Same paradigm as in (b) with a 300 μs laser pulse. (d) Summary of the results obtained in five cells. The distance from the uncaging point is plotted versus the delay between the peak of the Ca2+ transient recorded at the uncaging site and the peak of the signal at the remote site. (e) (ii) The size of the actual uncaging spot was measured by drawing a line across its image in fluorescence and fitting the projection of the spot on the line by a Gaussian curve (AU: fluorescence arbitrary units; full width at half maximum = 0.74 μm). (i) Spatial resolution of axonal uncaging using MNI-glutamate in the absence of NBQX to optimize the detectability of the signal. To construct this curve, the relative location of the laser and of the preparation was moved in the direction orthogonal to the neurite, in six separate experiments. The data were fitted to an exponential decay with a space constant of 1.43 μm.

Involvement of presynaptic Ca2+ stores

To explore further the proposal that Ca2+ stores could be involved in the NMDAR mediated Ca2+ response, we compared the results obtained by uncaging MNI-glutamate on axons Ca2+ two different laser pulse durations (100 and 300 μs). On average, 100 μs pulses yielded currents of 8 ± 1 pA (n = 4, Fig. 5ai and ii) and Ca2+ transients with ΔF/Fo of 43 ± 14% (n = 4, Fig. 5ai and ii) whereas 300 μs pulses gave currents of 21 ± 3 pA (n = 6, Fig. 5ai and ii) and Ca2+ transients displaying ΔF/Fo of 24 ± 6% (n = 6, Fig. 5ai and ii). Thus, whereas the current amplitude grows with the pulse duration, as expected, the amplitudes of the Ca2+ transients fail to display a significant increase (Fig. 5aii; Student's t-test); such a behavior could reflect a non linear amplification for the latter signal. Moreover, preNMDAR activation in MLIs has been shown to promote the spontaneous Ca2+ release from presynaptic ryanodine-sensitive Ca2+ stores (Duguid and Smart 2004) and presynaptic spontaneous Ca2+ transients (SCaTs) are suppressed by high concentrations of ryanodine (100 μM: Llano et al. 2000; Conti et al. 2004). Accordingly, we investigated the effect of 100 μM ryanodine on the uncaging of MNI- glutamate in MLIs axons. For these experiments, ryanodine was preincubated together with MNI-glutamate in the HEPES-buffered medium (see 'Methods'). As a result, control and ryanodine experiments were performed in two distinct groups. Axonal identification was as usual verified by recording the Ca2+ signal elicited by four propagated APs in control and ryanodine. As shown in Fig. 5bi, the response was significantly smaller in the ‘ryanodine’ group (ΔF/Fo = 47 ± 9%, n = 7) than in the ‘control group’ (ΔF/Fo = 86 ± 13%, n = 8). This result confirms the effects of ryanodine on axonal AP-evoked Ca2+ transients reported by Llano et al. (2000). Glutamate uncaging experiments are summarized in Fig. 5bii. In the presence of ryanodine, axonal Ca2+ transients displayed a much smaller ΔF/Fo than in control conditions (6.8 ± 3.4%, n = 5 vs. 43 ± 14%, n = 6; Student's t-test) although the inward current was not significantly different (13 ± 4 pA, n = 5 vs. 14 ± 4 pA, n = 6; Student's t-test). Altogether, these results clearly demonstrate the importance the ryanodine-sensitive Ca2+ pools in shaping NMDAR-mediated presynaptic Ca2+ signals.

Figure 5.

preNMDAR induced Ca2+ signals are linked to Ca2+ stores. (a) (i) Left: Time course of ΔF/Fo (upper panel) and simultaneous current recording (lower panel) obtained with 4-methoxy-7-nitroindolinyl-caged-l-glutamate (MNI-Glutamate) photorelease in the presence of 2,3-dihydroxy-6-nitro-7-sulfamoyl-benzo[f]quinoxaline-2,3-dione (NBQX) by a 100 μs laser pulse (indicated by an arrow) focused at the axonal region of interest. Right: same paradigm with a 300 μs pulse. (ii) Summary data for current recording (upper panel) and simultaneous ΔF/Fo (lower panel). (b) (i) Axonal ΔF/Fo transients elicited by four APs measured in control conditions (ctl) and in the presence of 100 μM ryanodine (rya). (ii) Axonal ΔF/Fo transients elicited by MNI-glutamate uncaging in the presence of NBQX (100 μs laser pulse) measured in control conditions (ctl) and in the presence of 100 μM ryanodine (rya). The number of cells contributing to the mean is indicated on the baregraphs.


In this study, we investigated the functional relevance of preNMDARs of MLIs by exploring their potential role as high-gain glutamate detectors. Salient results are that (i) preNMDARs can be activated by a brief NMDA puff; (ii) pre- and post-synaptic NMDARs display the same current to voltage relationship; (iii) preNMDARs promote a Ca2+ signaling amplification process relying on ryanodine-sensitive Ca2+ stores; (iv) they are directly distributed along the axon, but their ability to recruit ryanodine receptors allows them to greatly amplify the initial Ca2+ rise.

Glutamate released during small PF activity is highly constrained to the post-synaptic density and does not reach NMDARs which are considered as being solely extrasynaptic in MLIs (Clark and Cull-Candy 2002). However, high frequency bursts of PF activity, as recorded in vivo (Chadderton et al. 2004), are known to produce a substantial amount of spillover to nearby MLIs (Liu and Lachamp 2006). Such a mechanism has been proposed to activate extrasynaptic NMDARs at the dendritic (Carter and Regehr 2000) or axonal (Liu and Lachamp 2006) levels. The presence of functional dendritic NMDARs has been initially shown by iontophoresis of NMDA (Clark and Cull-Candy 2002) or aspartate (Christie and Jahr 2008) and has been confirmed by glutamate uncaging in the presence of an appropriate set of blockers (Pugh and Jahr 2011; Rossi et al. 2012). On the other hand, the very existence of preNMDARs in MLIs is controversial. The first electrophysiological evidence for these receptors was brought up by Glitsch and Marty at the basket cell/Purkinje cell synapse and at the MLI/MLI synapse (Glitsch and Marty 1999) and the earliest NMDA-evoked presynaptic Ca2+ signals were reported at stellate cell terminals (Shin and Linden 2005). More recently, the group of Jahr refuted the existence of preNMDARs and proposed that the elevation of presynaptic Ca2+ leading to GABA release was because of post-synaptically located receptors opening Ca2+ channels through propagated depolarization (Christie and Jahr 2008; Christie et al. 2011; Pugh and Jahr 2011). Accordingly, a previous report of ours mentioned a high number of failures in recording preNMDARs and concluded that a functional identification of preNMDARs in MLIs is complicated (Rossi et al. 2012; see Duguid and Smart 2009; Duguid 2013 for review). This failure rate was further increased when preNMDARs were challenged in the presence of a physiological Mg2+ concentration. However, a previous depolarization is likely to facilitate the opening of preNMDARs since a change in the holding potential clearly led to an increase in the amplitude of both the inward current and Ca2+ transient amplitudes (see Fig. 2c and d). This article details experiments aimed to increase the rate of success in recording such events. Indeed, the axonal Ca2+ transients that were actually elicited by NMDA puffs confirm that clamping the somatic potential at a negative value does not prevent preNMDARs activation. These results are in keeping with those obtained with long-lasting bath application of the agonist that led to large and sustained Ca2+ elevations (Rossi et al. 2012) and suggest that preNMDARs constitute glutamate sniffers that can be activated by either spillover or retrograde glutamate release.

From the current to voltage relationships presented hereby, we can propose that MLI's NMDARs contain the NR2A and B subunits, a proposal that matches our previous pharmacological study (Rossi et al. 2012). It has been suggested that NR2A-containing NMDARs are most likely synaptic whereas NR2B-containing receptors are mainly extrasynaptic (Groc et al. 2006). This possibility, however, is still controversial and the discrepancies among authors might result from developmental or local issues regarding the actual location of NMDARs subunits (see Yashiro and Philpot 2008 for review). Besides NR2B-containing NMDARs have been reported to carry a greater Ca2+ charge than their NR2A-containing counterpart (Sobczyk et al. 2005). Nonetheless, the open probability of NR1/NR2B channels is significantly lower than that of NR1/NR2A channels (Erreger et al. 2005). Altogether, the amount of Ca2+ entering the presynaptic terminals upon NMDARs opening only might not be sufficient to obtain a detectable Ca2+ transient unless the intracellular stores get enrolled in the process. In particular, the use of cyclopiazonic acid (CPA, a SERCA inhibitor) in the experiments of Christie and Jahr (2008) could have prevented the onset of preNMDAR-mediated Ca2+ elevations making therefore the events undetectable. The delayed Ca2+ events that arise a few second after the initial Ca2+ rise are reminiscent of the SCaTs observed at the basket cell/Purkinje cell terminals (Llano et al. 2000; Conti et al. 2004). Although spontaneous by nature, SCaTs appear with a frequency that has been shown to be increased more than fourfold by an elevated external Ca2+ concentration and the presence of a low (5–10 μM) concentration of ryanodine. Indeed, the opening of preNMDARs triggers an increase in axonal Ca2+ that subsequently mobilizes ryanodine-sensitive Ca2+ stores (Duguid and Smart 2004; Rossi et al. 2012): these two concomitant actions could very well constitute a plausible initiation process for SCaTs and definitely support the bases of an amplification process.

Local uncaging of either NNI-Glutamate (in the presence of NBQX) or MNI-NMDA has been shown to elicit Ca2+ transients and current when adequately performed on presynaptic varicosities (Rossi et al. 2012). We next sought to clarify the reasons why we were so often failing in recording preNMDARs responses. Since light scattering at 405 nm was estimated as 50% at 18 μm depth in rat cerebellum (Trigo et al. 2009), the local uncaging efficiency quickly declines with depth. Therefore, we tried to concentrate on cells whose soma were located at the surface of the slice. As a consequence, we substantially increased the number of axonal structures available for efficient local uncaging over the previous study. However, this method still suffers from one main limitation: superficial MLIs are rarely healthy. Approximately, the two thirds of the cells would die within a few minutes after break in and therefore, only cells that were able to give satisfactory axonal Ca2+ transients to propagated somatic action potentials were included in the study. Altogether, the use of superficial cells allowed us to obtain responses with a very short laser pulse (100 μs) and to compare them with those obtained with a longer pulse (300 μs). Interestingly, although the currents obtained with 300 μs were significantly larger than with 100 μs (p < 0.01, Student's t-test; see 'Preparation' for values), the Ca2+ transients amplitudes ranked in the opposite direction (see 'Preparation') but were not significantly different. Such a ‘non-linear’ behavior suggest the involvement of the Ca2+ stores in the development of the transients. The implication of the stores in the genesis of the preNMDARs-induced Ca2+ transients was further confirmed by the presence of complex signals appearing with local uncaging of glutamate. More precisely, the occurence of Ca2+ transients at varicosities located more than 5 μm away from the laser spot indicates that the Ca2+ entering the axon through activation of preNMDARs is able to sensitize the Ca2+ stores in adjacent varicosities. It is complicated to differentiate between diffusion of glutamate around the neurite and development of an intra-axonal Ca2+ wave. There is no doubt that a glutamate spillover will generate a certain amount of glutamate diffusion and accordingly, the Ca2+ signals obtained with glutamate puffs will at least partially result from this diffusion. However, the existence of Ca2+ transients appearing more than 10 s after the puff clearly indicates that an intracellular process is stimulated by preNMDARs. With uncaging experiments, diffusion of glutamate along the axon becomes much less likely. A refined calibration of our laser-driven uncaging system showed a spot bearing a half width < 1 μm. Using this very spot with short laser pulses (100 or 300 μs), the spatio-temporal properties of MNI-glutamate (0.9 mM) photoactivation are such that the glutamate does not spread farther than 4 μm (see Fig 4e). Under these conditions, the most likely hypothesis is that the local Ca2+ entry resulting from preNMDAR opening is relayed by a Ca2+ release process that spreads along the adjacent varicosities. Moreover, the fact that the occurrence of this spreading is inhibited by high concentration ryanodine confirms the crucial involvement of ryanodine-sensitive Ca2+ stores in the preNMDAR-trigger Ca2+ signaling process (see Duguid and Smart 2004). Such a mechanism explains why the preNMDAR-mediated Ca2+ transients are difficult to observe since cytoslic dialysis resulting from the whole-cell patch clamp configuration is known to greatly alter the functional integrity of the stores (Conti et al. 2004). Another issue commonly associated with whole-cell patch clamp is the dialysis of the intracellular compartment and especially an alteration of the internal Ca2+ buffers concentrations. MLIs are known to be mostly expressing parvalbumin (PV) as main Ca2+ buffer the influence of which has been tackled on synaptic transmission in a previous study of ours (Collin et al. 2005). The question now arises if the dialysis of PV though the pipette is likely to favor the occurrence of preNMDAR-mediated Ca2+ transients. In other words, could the physiological PV concentration prevent the appearance of glutamate uncaging-evoked Ca2+ signals and limit their spreading? In this context, we performed a series of simulations that led to the conclusions that (i) the Ca2+ signal amplitude is not affected by PV; (ii) the kinetic of the signal becomes more and more bi-exponential when the PV concentration is increased and (iii) an increased PV concentration is not likely to affect the occurrence of distal Ca2+ transients (see fig. S1 and Table S1 in the supplementary file). Altogether, the same mechanisms that activate dendritic NMDARs can simultaneously activate preNMDARs and the synergistic functional interaction between preNMDARs and Ca2+ pools is likely to contribute a sustained release of GABA when it is triggered in MLI's presynaptic compartment through glutamate spillover.


This study was supported by Agence Nationale de la Recherche (contract BLAN08-2_31083) and Fondation pour la Recherche Médicale (FRM Team, A. M.). We thank Alain Marty for valuable discussions and comments on the manuscript.

B.R. and T.C. designed, performed and analyzed the experiments. T.C. wrote the article.

None of the authors has to declare any conflict of interest.