Journal of Neurophysiology

Silent Synapses in Developing Cerebellar Granule Neurons

Gabriele Losi, Kate Prybylowski, ZhanYan Fu, Jian Hong Luo, Stefano Vicini


Silent synapses are excitatory synapses endowed exclusively withN-methyl-d-aspartate (NMDA) responses that have been proposed to acquire α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) responses during development and after long-term potentiation (LTP). These synapses are functionally silent because of the Mg2+ block of NMDA receptors at resting potentials. Here we provide evidence for the presence of silent synapses in developing cerebellar granule cells. Using the patch-clamp technique in the whole-cell configuration, we recorded the spontaneous excitatory postsynaptic currents (sEPSCs) from rat cerebellar granule cells in culture and in slices at physiological concentration of Mg2+ (1 mM). A holding potential of +60 mV removes Mg2+ block of NMDA channels, allowing us to record NMDA-sEPSCs. We thus compared the frequency of AMPA-sEPSCs, recorded at −60 mV, with that of NMDA-sEPSCs, recorded at +60 mV. NMDA-sEPSCs occurred at higher frequency than the AMPA-sEPSCs in most cells recorded in slices from rats at postnatal day (P) <13 and in culture at 6–8 days after plating (DIV6–8). In a few cells from young rats (P6–9) and in most neurons in culture at DIV6 we recorded exclusively NMDA-sEPSCs, supporting the hypothesis of existence of functional synapses with NMDA and without AMPA receptors. Increasing glutamate release in the slice with cyclothiazide and temperature increased AMPA and NMDA-sEPSCs frequencies but failed to alter the relative ratio of frequency of occurrence. Frequency ratio of NMDA versus AMPA-sEPSCs in slices was correlated with the weighted time constant of decay (τw) of NMDA-sEPSCs and decreased with development along the reported decrease of τw. We suggest that the prevalence of synaptic NR2A subunits that confer faster kinetics is paralleled by the disappearance of silent synapses early in cerebellar development.


The role ofN-methyl-d-aspartate (NMDA) receptors in the development of the cerebellum has been largely described (Komuro and Rakic 1993; Rabacchi et al. 1992;Rossi and Slater 1993). Rat cerebellar granule cells in slice are therefore a good model to study the role of these receptors in synaptogenesis and in neuronal development. Cerebellar granule cells are localized after birth in the outer layer and migrate with particular regular timing to the inner layer until postnatal day 21 (P21) (Altman 1982; Ito 1984). This migration is paralleled by the formation and maturation of synaptic contacts with l-glutamate-releasing mossy fibers (Hamori and Somogyi 1983). The small size of cerebellar granule cells and the relatively small number of dendrites and synapses per cell (from 3 to 4; Palay and Chan-Palay 1974;Palkovits et al. 1972) make it possible to record spontaneous excitatory postsynaptic currents (sEPSCs) with good resolution and ideal voltage-clamp conditions (Silver et al. 1992). Additional interest in the mossy fiber-granule cells relay comes from the recent discovery of its ability to undergo long-term potentiation (LTP; D'Angelo et al. 1999). Previous works revealed that mossy fiber–granule cells synaptic transmission is mediated by NMDA and non-NMDA receptors (Garthwaite and Brodbelt 1990) but in immature granule cells the NMDA component is predominant (D'Angelo et al. 1993). This raises the possibility of postsynaptic sites with only NMDA receptors termed as “silent synapses” (Liao et al. 1995), since these NMDA-only synapses would normally be blocked by magnesium at physiological resting potentials (Mayer et al. 1984; Nowak et al. 1984).

NMDA receptors are formed by the assembly of multiple NR1 subunits and at least one of NR2 type (see Cull-Candy et al. 2001 for review). In cerebellar granule neurons, the NR1 NMDA receptor subunit is ubiquitously present during development. In contrast, NR2B subunit protein expression begins in late embryonic stages and levels decrease during the second postnatal week (Takahashi et al. 1996;Wang et al. 1995; Watanabe et al. 1994). At the stage when NR2B levels decrease, NR2A subunit expression increases (Wang et al. 1995; Watanabe et al. 1994). Recombinant NMDA receptors comprising NR1/NR2B subunits are characterized by slow deactivation kinetics while insertion of the NR2A subunit confers faster kinetics (Monyer et al. 1994; Vicini et al. 1998). In fact the developmental change of NMDA-EPSCs kinetics (Cathala et al. 2000; Rumbaugh and Vicini 1999) is prevented in NR2A knock-out mice (Takahashi et al. 1996).

Recently the presence of NR2B subunits has been correlated with the critical period to induce LTP in thalamocortical synapses (Barth and Malenka 2001) and with the high probability of release of immature hippocampal synapse (Chavis and Westbrook 2001). Therefore the role of the NR2B NMDA receptor subunit in early synaptic physiology seems to be of fundamental importance. In this study we investigate the existence of silent synapses in developing cerebellar granule neurons in relationship to the decay time of NMDA receptor-mediated EPSCs, an indicator of the relative abundance of expression of NR2B subunit at synaptic sites (Cathala et al. 2000; Rumbaugh and Vicini 1999).

To study the relative contribution of NMDA receptors during synaptic formation and maturation, we took advantage of the voltage-dependent magnesium blockade of the NMDA receptor. Recordings at positive holding potentials were compared with those obtained at negative potentials in cultured neurons and slices from rats at different postnatal ages (P6–15). Our study will extend the evidence for a predominant contribution of NMDA receptors in early synapses to cerebellar granule cells and correlate these data with the presence of specific NMDA receptor subtypes.


Cerebellar granule cells culture

Primary cultures of rat cerebellar granule neurons were prepared from 7-day-old Sprague-Dawley rat cerebella. Cells were dispersed with trypsin (0.25 mg/ml, Sigma, St. Louis, MO) and plated at a density of 0.8–1 × 106 on 35-mm Nunc dishes, coated with poly-l-lysine (10 μg/ml; Sigma). Cells were cultured in basal Eagle's medium supplemented with 10% bovine calf serum, 2 mM glutamine, 100 μg/ml gentamycin (all from Invitrogen, Carlsbad, CA) and maintained at 37°C in 6% CO2. The final concentration of KCl in the culture medium was adjusted to 25 mM (high K+).

To achieve functional synapse formation, 4 days after plating (DIV4) media was replaced with low (5 mM) potassium media (MEM supplemented with 5 mg/ml glucose, 0.1 mg/ml transferrin, 0.025 mg/ml insulin, 2 mM glutamine, and 20 μg/ml gentamycin; Invitrogen) as previously described in Chen et al. (2000). At DIV4 cytosine arabinofuranoside (10 μM; Sigma) was added to all cultures to inhibit glial proliferation. Recordings were made from DIV6–8 neurons in culture.


Sagittal slices of cerebellum (150–200 μM) were prepared from 6- to 15-day-old (P6–15) Sprague-Dawley rats as previously described (Puia et al. 1994). Whole-cell recordings from cerebellar granule neurons were obtained under visual control with a Nikon Eclipse 600FN microscope (Nikon, Japan) equipped with Nomarski IR optics and an electrically insulated water immersion 60× objective with a long working distance and high numerical aperture.


The recording chamber was continuously perfused at 5 ml/min with an extracellular medium composed, for cultured neurons, of the following (in mM): 145 NaCl, 5 KCl, 1 MgCl2, 1 CaCl2, 5 HEPES, 5 glucose, 25 sucrose, 0.25 phenol red, and 20 μM d-serine (all from Sigma). For slices the perfusion solution was composed of the following (in mM): 120 NaCl, 3.1 KCl, 1 MgCl2, 1.25 K2HPO4, 26 NaHCO3, 2.0 CaCl2, 5 glucose, 25 sucrose, and 0.005 d-serine (all from Sigma). This solution was maintained at pH 7.4 by bubbling with 5% CO2-95% O2. In some experiments the extracellular solution was heated at 33°C by means of a custom-made in-line heater. All other experiments were performed at room temperature (24–26°C). NMDA receptor-mediated synaptic responses were pharmacologically isolated by bicuculline methiodide (BMI, 50 μM; Sigma) and 2,3-dihydro-6-nitro-7-sulfamoyl-benzo(F)quinoxaline (NBQX, 5 μM; Tocris, Ballwin, MO). In some experiments 3-[(±)-2-carboxypiperazin-4-yl]-propyl-1-phosphonic acid (CPP, 10–20 μM; Tocris), cyclothiazide (50 μM; Tocris), and tetrodotoxin (TTX, 1 μM; Sigma) were also included. All drugs were superfused through parallel inputs to the perfusion chamber or locally perfused by means of a Y tube (Murase et al. 1989).

Electrodes were pulled in two stages on a vertical pipette puller from borosilicate glass capillaries (Wiretrol II, Drummond, Broomall, PA). Typical pipette resistance was 5–7 MΩ. Intracellular (patch pipette) solutions contained the following (in mM): 145 Cs-methane sulfonate, 10 bis-(o-aminophenoxy)-N,N,N′,N′-tetraacetic acid (BAPTA), 50 MgCl2, 5.0 ATP-Na, 0.2 guanosine 5′-triphosphate (GTP)-Na, and 10 HEPES, adjusted to pH 7.2 with CsOH. Whole-cell recordings were performed with a patch-clamp amplifier (Axopatch 200, Axon Instrument, Foster City, CA).

Data collection and analysis

Currents were filtered at 1 kHz with an eight-pole low-pass Bessel filter (Frequency Devices, Haverhill, MA), digitized at 5–10 kHz using an IBM-compatible microcomputer equipped with Digidata 1200 data acquisition board and Pclamp 8 software (both from Axon Instruments). Off-line data analysis, curve fitting, and figure preparation were performed with Clampfit 8 (Axon Instruments), Origin 4.1 (Microcal, Northampton, MA), and Minianalysis (Synaptosoft, Decatur, GA) software. Fitting of the decay phase of currents recorded from granule cells in culture and slice was performed using a simplex algorithm for least-squares exponential fitting routines. Decay times of averaged currents derived from fitting to double exponential equations of the form I(t) =If exp(−tf) +Is exp(−ts), whereIf and Is are the amplitudes of the fast and slow decay components, and τf and τs are their respective decay time constants used to fit the data. To compare decay times between different subunit combinations, we used a weighted mean decay time constant τw = [If /(If +Is )] ∗ tf + [Is /(If +Is )] ∗t s. Data values are expressed as mean ± SE unless otherwise indicated. P values represent the results of independent t-tests, with prior analysis of variance (ANOVA).

The absence of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)-sEPSCs in cells recorded at −60 mV was determined by the lack of observable spontaneous inward currents during observation times of ≥5 min. AMPA-sEPSCs frequency was not significantly different in seven cells held at −60 and −90 mV holding potentials (0.08 ± 0.02 and 0.06 ± 0.01 Hz).


Evidence of silent synapses in cerebellar granule neurons in culture

The occurrence of functional synapse formation was tested with whole-cell voltage-clamp recordings of spontaneous postsynaptic currents in cerebellar granule neurons in primary culture. As previously reported (Chen et al. 2000; Mellor et al. 1998), culturing for several days in a growth media with physiological potassium concentration was necessary to detect synaptic currents, presumably because this reversed the depolarization of the cell caused by high potassium and allowed assembly of a pool of presynaptic neurotransmitter vesicles (Mellor et al. 1998). Figure 1 Aillustrates an example of whole-cell recording from a granule neuron at DIV8. An intracellular solution with cesium allowed us to clamp the potential of the recorded cell at positive values with effective potassium channel blockade. At a holding potential of +60 mV there was a high frequency of occurrence of large outward currents (spontaneous inhibitory postsynaptic currents, ISPCs) that were abolished by BMI 50 μM (Fig. 1 A2). The remaining outward currents were characterized by smaller amplitude and slower decay (Fig.1 A3) and were sensitive to the NMDA receptor-antagonist CPP (NMDA-sEPSCs, Fig. 1 A4). In the combined presence of BMI and CPP fast outward currents similar to those recorded were still observed at −60 mV and could be antagonized by NBQX (AMPA-sEPSCs, not shown).

Fig. 1.

Evidence of silent synapses in cerebellar granule cells in culture. Whole-cell voltage-clamp recordings of spontaneous excitatory postsynaptic currents (sEPSCs). A: in a granule neuron held at −60 mV in primary culture at 8 days after plating (DIV8) (1); changing the holding potential to +60 mV (2) revealed the high-frequency occurrence of large outward currents that were abolished by BMI (3). The remaining outward currents were characterized by lower frequency of occurrence, smaller amplitude, and slower decay and were sensitive to the N-methyl-d-aspartate (NMDA) receptor antagonist 3-[(±)-2-carboxypiperazin-4-yl]-propyl-1-phosphonic acid (CPP) (4). In the combined presence of bicuculline methiodide (BMI) and CPP fast outward currents were still observed being antagonized by 2,3-dihydro-6-nitro-7-sulfamoyl-benzo(F)quinoxaline (NBQX; not shown).B: sEPSCs recorded from a neuron at DIV6. NMDA-sEPSCs occurred exclusively at +60 mV holding potential.

In five granule neurons at DIV8 we compared the frequency of occurrence of AMPA-sEPSCs recorded at −60 mV with that of NMDA-sEPSCs recorded at +60 mV in the presence of BMI 50 μM. AMPA-sEPSCs occurred in all cells tested; events were fast decaying (τfast4.3 ± 2 ms) and occurred at 0.13 ± 0.05 Hz. NMDA-sEPSCs were characterized by slow decay kinetics (τw104 ± 21 ms) and occurred at 0.19 ± 0.03 Hz. The average ratio of NMDA-sEPSCs to AMPA-sEPSCs frequency calculated cell by cell was 1.9 ± 0.5, indicating a higher frequency of occurrence of NMDA-sEPSCs. In five of seven cells investigated at a younger age (DIV6) exclusively NMDA-sEPSCs were recorded (Fig. 1 B). These results suggest that excitatory synapses in primary culture of cerebellar granule neurons initially express only NMDA receptors and later acquire AMPA receptors.

Evidence of silent synapses in young cerebellar granule neurons in slice

Cerebellar granule cells in vivo receive input only from mossy fibers (excitatory) and Golgi cells (inhibitory) (Palay and Chan-Palay 1974). In primary culture cerebellar granule cells receive excitatory contacts from other granule cells. Thus the formation of silent synapses in cultured granule neurons might have been an artifact of the culture condition. To rule out this possibility and to verify the relevance of these findings in vivo, we studied cerebellar granule neurons in slice. Spontaneous excitatory mossy fibers synaptic currents (sEPSCs) were recorded in slices from P6–15 rats in presence of γ-aminobutyric acid-A (GABAA) receptor blocker BMI (50 μM) andd-serine (5 μM) to allow NMDA receptor activation. Miniature EPSCs (mEPSCs) recorded in six cells at P8–10 in the presence of TTX (1 μM) had the similar frequencies at positive potentials before and after TTX application of 0.21 ± 0.04 and 0.19 ± 0.05 Hz, respectively. Therefore this drug was omitted in the recording solution in the following experiments. sEPSCs were compared in the same cell at negative (−60 mV) and positive (+60 mV) holding potentials. Not all cells studied had observable sEPSCs at either positive or negative potential, particularly in slices from older rats (P>14–15). This could be related to the severing of granule cell dendrites by the slicing procedures or to the reported decreased probability of release of mature synapses (Chavis and Westbrook 2001). At negative potentials (−60 mV) we observed fast sEPSCs (Fig. 2, A andB, upper traces) with decay time characterized by a large fast exponential component (τf4.5 ± 0.7; n = 5, P11) followed by a small slow decaying component, as previously reported (D'Angelo et al. 1993; Silver et al. 1992). sEPSCs at −60 mV were abolished by NBQX (5 μM, not shown) and thus identified as AMPA-mediated events (AMPA-sEPSCs). Frequency of occurrence of AMPA-sEPSCs in 36 cells in rats at P6–8 (Table1) showed high variability from cell to cell and was not significantly different from that measured in 24 cells from P14–15 rats (Table 1). sEPSCs recorded at positive potentials (+60 mV) in rats at P6–8 had a slow decay and were abolished by the selective NMDA antagonist CPP (20 μM, n = 3, P8). The remaining fast sEPSCs were not investigated because of the possibility of rectifying AMPA receptors at cerebellar synapses (Liu and Cull-Candy 2000). sEPSCs at +60 mV in rats at P6–8 occurred at a higher frequency than at −60 mV (Fig. 2, A andE) in 35 of 43 cells studied (equivalent to 81.4%). In contrast, in more developed rats (P14–15) sEPSCs frequency of occurrence was smaller at positive than at negative potentials in 18 of 24 cells investigated (Fig. 2 E). Average of frequency of NMDA-sEPSCs significantly decreased during development (Table 1). The NMDA/AMPA ratio of frequency of sEPSCs calculated for each individual cell (Fig. 2 E) at P6–8 was 2.96 ± 0.48 (n = 36), while the ratio measured in cells from rats at P14–15 was 0.67 ± 0.09 (n = 24). In 7 cells of 43 at P6–8 we recorded exclusively NMDA-sEPSCs. It is possible that the frequency of NMDA-sEPSCs could be affected by the positive holding potential used to record these events. This was not the case since in eight granule cells in rats at P8 the frequency of NMDA-sEPSCs was similar when events were recorded at +60 mV in 1 mM Mg2+ or in Mg2+-free solution at −60 mV (0.13 ± 0.04 and 0.13 ± 0.06 Hz, respectively, Fig. 2 F).

Fig. 2.

Evidence of silent synapses in cerebellar granule cells in slice.A and B: sEPSCs recorded from granule cells in slices from rats at different ages: postnatal day (P) 8 inA and P15 in B (BMI 50 μM andd-serine 5 μM are present). In both neurons relief of magnesium block of NMDA receptors by recording at positive holding potential (+60 mV) reveals outward sEPSCs. At P8 the occurrence of sEPSCs at holding potential of +60 mV is more frequent than that observed at −60 mV. At P15, a low frequency of sEPSCs occurrence is observed at both holding potentials. C: average of 14 sEPSCs recorded at +60 mV from the same 2 neurons of Aand B. sEPSCs recorded from older neurons are faster and smaller. Averages are normalized, superimposed, and fitted inD. E: the ratio of frequency of sEPSCs at +60 mV versus −60 mV calculated for all neurons studied during development where α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)sEPSCs were recorded. F: sEPSCs recorded from a neuron held at −60 mV in the presence (1) or absence (2) of Mg2+ (1 mM) and in the presence of Mg2+ at +60 mV (3). Relief of NMDA receptor Mg2+ block at positive holding potential or in the Mg2+-free solution allows more frequent occurrence of sEPSCs in a similar proportion.

View this table:
Table 1.

Properties of spontaneous EPSCs in developing cerebellar granule neuron in slice

If a small amount of agonist is released this may be sufficient to activate only NMDA receptors (Choi et al. 2000;Gasparini et al. 2000; Renger et al. 2001) because of their higher affinity for glutamate compared with AMPA receptors (Patneau and Mayer 1990). A determination of low AMPA-sEPSCs frequency could therefore be due to the failure of receptor activation and/or an inability to detect very fast or small events lost in the background noise. To rule out this possibility, we recorded AMPA-sEPSCs in the presence of cyclothiazide (50 μM), a compound that enhances glutamate release and removes AMPA receptor desensitization (Diamond and Jahr 1995;Ishikawa and Takahashi 2001; Yamada and Tang 1993). At P7 cyclothiazide increased the frequency of AMPA and NMDA-sEPSCs by 97 ± 45 and 99 ± 33%, respectively (n = 8, Fig. 3,A and B), consistent with the reported action of this compound on presynaptic ion channels (Diamond and Jahr 1995; Ishikawa and Takahashi 2001). The NMDA/AMPA-sEPSCs frequency ratio was not affected as 2.5 ± 0.9 without and 2.2 ± 0.6 with cyclothiazide. The weighted time constant (τw) of AMPA-sEPSCs decay in the same cells increased from 6.2 ± 1.1 to 16.7 ± 2.1 ms in the presence of cyclothiazide (Fig. 3 C; P < 0.01, paired t-test). Amplitude and kinetics of NMDA-sEPSCs in the presence of cyclothiazide were not studied because of the reported potentiation of the AMPA component (Yamada and Tang 1993) and inhibition of the NMDA component (Losi et al. 2000). To increase the release probability, we recorded sEPSCs from neurons (P8–10) at 33°C and compared the sEPSCs to that occurred at 25°C (Fig. 4, Aand B). At 33°C both AMPA and NMDA-sEPSCs increased in amplitude (Fig. 4 D1). At 33°C NMDA-sEPSCs were faster decaying than at room temperature (Fig. 4, C andD2). Both AMPA and NMDA-sEPSCs frequencies doubled when the temperature increased from 25 to 33°C (Fig. 4 D3). However, the ratio of frequency of NMDA versus AMPA-sEPSCs was not affected by the temperature increase (Fig. 4 D4). Thus increasing the release probability and quantal release failed to uncover latent AMPA responses in granule neurons from young rats.

Fig. 3.

Effect of cyclothiazide on sEPSCs. A andB: sEPSCs recorded from a granule neuron in slice (P7) in the absence (A) and the presence (B) of cyclothiazide 50 μM (BMI 50 μM and d-serine 5 μM are present). In both conditions the occurrence of sEPSCs is more frequent at a holding potential of +60 mV than at −60 mV. In the presence of cyclothiazide sEPSCs occur at a higher frequency. Calibration bars apply to both panels. C: averages of sEPSCs recorded at −60 mV from the same neuron of A andB in control conditions (left) and with cyclothiazide (right) with an indication of the τw resulting from a double exponential fit.

Fig. 4.

Effect of temperature on sEPSCs. A and B: sEPSCs recorded from a granule neuron in slice (P10) at 2 different temperatures: 25°C in A and 33°C in B(BMI 50 μM and d-serine 5 μM are present). At both temperatures the occurrence of sEPSCs is more frequent at a holding potential of +60 mV than at −60 mV. At 33°C NMDA-sEPSCs show higher frequency and amplitude and faster kinetics than at 25°C. Calibration bars apply to both panels. C: averages of sEPSCs recorded at +60 mV from the same neuron of A andB at 25°C (left) and at 33°C (right). D: histogram showing the average effect of temperature (25°C, white columns; 33°C, gray columns) on sEPSCs peak amplitude (D1), NMDA-sEPSCs decay (D2), frequencies (D3), NMDA/AMPA-sEPSCs frequency ratio (D4); (n), *P < 0.05, **P < 0.01 pairedt-test.

Change in NMDA-sEPSCs kinetics is paralleled by the disappearance of silent synapses

The average of 14 sEPSCs recorded at +60 mV from two neurons of different ages are normalized, superimposed, and fitted in Fig.2 D. The analysis of decay kinetics of NMDA-sEPSCs recorded at +60 mV revealed a decrease of the τw (Fig.5 A) during development: from 234 ± 23 ms at P6 (n = 15) to 83 ± 24 ms at P15 (n = 9) in cerebellar granule neurons in slices, as previously reported for evoked NMDA EPSCs (Cathala et al. 2000; Rumbaugh and Vicini 1999), suggesting the synaptic insertion of NR2A subunits and the decrease of NR2B subunits. A decrease in the average peak amplitude was also observed in granule neurons from rats at P6–8 and P14–15 (Table 1).

Fig. 5.

Change in NMDA-sEPSCs kinetics parallels the disappearance of silent synapses. A: scatter plot and superimposed linear fitting of the weighted time constant of the exponential fit of the decay of NMDA-sEPSCs averaged for all the cells studied at each postnatal day against the age of the rat investigated.B: the weighted time constant of decay of sEPSCs recorded at positive potential is correlated with the ratio of frequency of occurrence of sEPSCs at +60 mV versus −60 mV. Linear fitting is superimposed: R = 0.31,P < 0.01. Each data point is from a different granule neuron. Cells with only NMDA-sEPSCs were excluded.

We then compared the decay kinetics of NMDA-sEPSCs with the ratio of frequency of occurrence at opposite holding potentials. In Fig.5 B we illustrate the τw of NMDA-sEPSCs recorded at positive potential plotted against the frequency ratio of sEPSCs at +60 mV versus −60 mV for every individual neuron studied. The correlation coefficient R from the linear fitting was 0.31 and was statistically significant.


Silent synapses contain only NMDA receptors that are not functional at resting potential due to the voltage-dependent Mg2+ blockade (Isaac et al. 1995;Liao et al. 1995). We used sEPSCs frequencies at positive- and negative-holding potentials to compare the occurrence of NMDA versus AMPA receptors at synapses in cerebellar granule cells developing in culture. Our results indicate that cerebellar granule neurons in primary culture can form functional excitatory synapses containing only NMDA receptors, as has been reported for hippocampal neurons (Isaac et al. 1995; Liao et al. 1995). In cultured neurons natural excitatory inputs are replaced by connections between the cerebellar granule cells themselves. However, similar pure NMDA synapses are observed in granule neurons in slices from developing rats between postnatal day 6 and 13. This suggests the relevance of these findings for the development of excitatory synapses to cerebellar granule neurons in vivo.

The NMDA-sEPSCs frequencies were significantly higher than AMPA-sEPSCs frequencies when considered by individual cell in slices from young rats, giving NMDA/AMPA frequency ratio values higher than 1. This would indicate the presence of postsynaptic sites endowed exclusively with NMDA receptors. Further support for the evidence of silent synapses in developing granule neurons is that some neurons both in slices and in primary cultures showed only NMDA-sEPSCs and no AMPA-sEPSCs. Indeed a dominant role of NMDA-mediated currents in EPSCs amplitude in early developing synapses was previously reported (D'Angelo et al. 1993, but see Cathala et al. 2000). An alternative explanation for our data could be a higher frequency of presynaptic release induced by the postsynaptic depolarization, since AMPA and NMDA-sEPSCs frequencies were measured at different holding potentials. We suggest that this hypothesis is less likely since NMDA-sEPSCs frequencies were similar when measured at positive potential with Mg2+or at negative potential in a Mg2+-free solution. We did not study the AMPA-sEPSCs frequency at positive potential nor the possibility of rectifying AMPA receptors at cerebellar synapses (Liu and Cull-Candy 2000).

AMPA receptors have lower affinity for glutamate than NMDA receptors, thus it has been suggested that silent synapses could instead be “whispering” synapses (Choi et al. 2000;Gasparini et al. 2000; Renger et al. 2001). These whispering synapses would occur if the amount of agonist released was sufficient to activate only NMDA and not AMPA receptors. Low AMPA-sEPSCs frequency could thus be due to the failure of receptor activation by presynaptic glutamate release. Cyclothiazide application and increasing recording temperature in slices from young rats increased the probability and quantal release of glutamate and the frequency of both AMPA and NMDA-sEPSCs but did not affect their relative ratio. This suggests the presence of true silent synapses. The increased amplitude of NMDA-sEPSCs with temperature or cyclothiazide indicates a possible lack of saturation of postsynaptic NMDA receptors (Chen et al. 2001).

Taken together these results support the hypothesis of existence of functional synapses with NMDA and without AMPA receptors at young ages that are silent at negative potentials because of the magnesium block.Silver et al. (1992) were not able to distinguish convincingly the occurrence of pure NMDA-sEPSCs in cerebellar slices from background channel activation. We do not know the source of discrepancy with our study. One possibility is that in our experiments NMDA-sEPSCs were recorded in the presence of a physiological magnesium concentration that potentiates NMDA responses at positive potentials (Paoletti et al. 1995). Alternatively the absence of TTX might have allowed us to measure larger size NMDA-sEPSCs due to release from more than one site. However, our data from a limited comparison of currents recorded in Mg2+-free solution and in the presence of TTX seem to argue against these possibilities.

Our results show a correlation between the slow kinetics of NMDA-sEPSCs and the occurrence of silent synapses (Fig. 5 B). This suggests that during the early steps of mossy fiber–granule cells synaptic formation there are pure NMDA synapses containing primarily the NR2B subunits. The prevalence of the NR2B subunit at this age had been demonstrated by the sensitivity of the NMDA-EPSCs to NR2B selective blockers, although a contribution of NR2A subunits to synaptic NMDA receptors has also been suggested (Cathala et al. 2000; Rumbaugh and Vicini 1999). The developmental decrease of sensitivity to NR2B blockers and fast NMDA-EPSCs decay kinetics indicate the prevalence of the NR2A subunit in more mature synapses (Cathala et al. 2000;Rumbaugh and Vicini 1999). It remains to be proven, however, that the presence of NR2B subunits is required for the formation of silent synapses. It is likely that when NR2A subunit becomes dominant in the postsynaptic receptor pool, later in development, silent synapses disappear. We suggest that the prevalence of synaptic NR2A subunits that confer faster kinetics parallels the disappearance of silent synapses early during cerebellar development.

The decrease of NMDA-sEPSCs frequency and amplitude that we observed during development could be explained with a reduction of postsynaptic NMDA receptors as well as a reduction in the probability of release at mature presynaptic sites. A recent work demonstrated a higher probability of release in immature hippocampal synapses containing NR2B NMDA receptors versus mature synapses comprising NR2A subunits (Chavis and Westbrook 2001). A coordinated signaling link was established between pre- and postsynaptic sites and the authors speculated that the reduced size of mature versus immature active zones could account for the reduction of probability of release. The reduction of NMDA-sEPSCs frequency could also be due to a decrease in the total number of synapses. A reduction of the presynaptic size or a reduction of the number of synapses should lead to a reduction in frequency of AMPA-sEPSCs, which was not observed. However, a maintained AMPA-sEPSCs frequency during development could be due to the synaptic insertion of AMPA receptors that would compensate for the reduction of release probability or the reduction of total number of synapses.

A hypothesis to explain LTP in hippocampus and cortex involves silent synapses (Faber et al. 1991). During LTP induction, silent synapses may become functional by recruiting AMPA receptors (Durand et al. 1996; Isaac et al. 1995;Liao et al. 1995). This in turn produces an initial depolarization of the synaptic membrane and relieves the magnesium block of the NMDA receptor when the presynaptic terminal releases glutamate. A correlation was previously seen between the decay kinetics of NMDA-EPSCs, LTP, and the critical period of developmental plasticity (see Fox et al. 1999 for review). More recently, however, it has been shown that the presence of NR2B subunit rather than the slow decay of NMDA-EPSCs was correlated with the critical period to induce LTP in thalamocortical synapses (Barth and Malenka 2001). Our results on the occurrence of silent synapses correlated with the presence of NR1/NR2B NMDA receptors could provide a mechanistic explanation for the relationship between the presence of NR2B subunits and the occurrence of LTP. It remains to be seen if the reported capability to induce LTP at mossy fiber–granule cell relay (D'Angelo et al. 1999) has a developmental time course that parallels the decrease in silent synapses.


We thank Drs. Giulia Puia and Lorenzo Corsi for critically reading this work.

This work was supported by National Institute of Mental Health Grants MH-58946 and MH-01680.


  • Address for reprint requests: S. Vicini, Dept. of Physiology and Biophysics, Basic Science Building, Rm. 225, Georgetown University Medical School, 3900 Reservoir Rd., Washington, DC 20007 (E-mail: svicin01{at}


View Abstract