Activity-Dependent Change in AMPA Receptor Properties in Cerebellar Stellate Cells

Synaptic currents display a wide range of rectification values

We first examined the possibility that synaptic activity modulates the expression of GluR2-containing AMPARs under normal conditions. If the targeting of the AMPAR subtype is a dynamic process (and does not occur in an all-or-none manner), then differences would be expected in the degree of rectification displayed by different synapses. To determine whether such variation occurred between stellate cells in our slices, we examined evoked EPSCs over a range of holding potentials in cells from 46 rats. In addition, we attempted to determine whether the AMPAR subtype expressed was related to the level of spontaneous synaptic activity. In cerebellar stellate cells, the spontaneous synaptic currents and evoked EPSCs arise from a single type of afferent input: the parallel fibers (granule cell axons). We reasoned that in cells in which spontaneous EPSCs (sEPSCs) occurred at a high rate, a given synapse would be more likely to have experienced spontaneous presynaptic activity than in cells in which spontaneous EPSC frequency was low. Thus, if the activity-induced targeting of GluR2-containing AMPARs occurred under normal conditions, the expression of GluR2 at the synapse would be expected to correlate with the rate of spontaneous EPSCs in the cell and to depend on the previous history of the synapse.

Stellate cells generally displayed a low rate of spontaneous synaptic activity (Fig. 1E); the average spontaneous EPSC frequency was 0.72 ± 0.28 Hz (n = 65 cells). In 57 cells, the frequency of spontaneous EPSCs was <1.1 Hz, with a mean frequency of 0.18 ± 0.03 Hz. In such cells, the amplitude of the synaptic current was reduced at depolarized potentials, and the I–V relationship of evoked EPSCs displayed clear inward rectification, indicative of the presence of synaptic GluR2-lacking AMPARs. An example of this is depicted in Figure 1A,B(EPSCs at +40 and −60 mV; spontaneous EPSC frequency, 0.62 Hz). Among the 65 cells examined, 8 displayed a rate of spontaneous EPSC activity that was markedly higher, with a mean frequency of 4.6 ± 1.8 Hz (n = 8; p < 0.0001). Figure1C,D illustrates data from a cell in which the spontaneous EPSCs occurred at 15.7 Hz. The synaptic currents in these cells gave linear I–V plots, indicating the absence of polyamine block at depolarized potentials and suggesting the presence of GluR2-containing (Ca²⁺-impermeable) AMPARs. The ratio of EPSC amplitude at +40 versus −60 mV, R_+40/−60, was subsequently used to determine the relative rectification behavior of synaptic currents. The value for R_+40/−60 from cells that displayed a higher rate of spontaneous EPSC activity (sEPSC frequency of >1.2 Hz; R_+40/−60 = 0.37 ± 0.03; n= 8) was significantly greater than that from those cells in which sEPSC frequency was low (R_+40/−60 = 0.28 ± 0.01; n = 57; p < 0.005) (Fig.1F). Thus, synaptic currents in cells that displayed a high rate of spontaneous EPSCs appeared to be mediated largely by GluR2-containing receptors, consistent with the idea that increased spontaneous synaptic activity is related to the expression of GluR2 at these synapses.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Synaptic currents in cerebellar stellate cells displayed various degrees of rectification. A, B, Data from a cell that displayed a spontaneous EPSC rate of 0.62 Hz.A, Mean evoked EPSC traces at +40 and −60 mV (100 mm spermine present in the pipette solution).B, The synaptic current displayed an inwardly rectifyingI–V relationship. The solid line is a linear regression line fitted to the data points at hyperpolarized potentials, and the dashed line connects the data points at depolarized potentials. The EPSC amplitude at depolarized potentials fell below the solid line, indicative of inward rectification. C, D, Data from a stellate cell in which spontaneous EPSCs occurred at 15.7 Hz. C, Mean evoked EPSCs at +40 and −60 mV. D, The I–Vrelationship of synaptic currents shown in C. The data points at depolarized potentials fell along the solid line, indicating a linear I–V relationship.E, Histogram of sEPSC frequency (n = 65 cells; mean, 0.72 ± 0.28 Hz). F, Ratio of EPSC amplitudes at +40 versus −60 mV (R_+40/−60) determined from cells displaying high sEPSC frequency (open triangles) and low sEPSC frequency (open circles). Cells with a high rate of sEPSCs had a higher mean R_+40/−60 value (filled triangle,n = 8) than cells with low sEPSC frequency (filled circle; n = 27;p < 0.005). G, Histogram of the ratio of the EPSC amplitudes at +40 versus −60 mV (n = 65 cells; mean, 0.29 ± 0.01).H, Relationship between sEPSC frequency and the ratio of EPSC amplitudes at +40 versus −60 mV. Inset, Average sEPSC frequency in cells in which evoked EPSCs gave R_+40/−60 values of <0.3 (n = 38) and >0.3 (n = 27) (p < 0.05).

We obtained a wide range of values for the ratio of EPSC amplitude at +40 versus −60 mV in stellate cells. As can be seen from Figure1G, these ranged from 0.11 to 0.51, with an average value of 0.29 ± 0.01 (n = 65 cells). Synaptic currents that displayed inward rectification (R_+40/−60 < 0.30; mean R_+40/−60 = 0.23 ± 0.01;n = 38) were obtained from cells in which the rate of sEPSCs was low (sEPSC frequency of <1.1 Hz for 37 cells and 1.67 Hz for 1 cell; mean sEPSC frequency, 0.25 ± 0.05 Hz) (Fig.1H). In contrast, cells from which synaptic currents gave a high R_+40/−60 value (R_+40/−60 > 0.30; mean R_+40/−60 = 0.37 ± 0.01; n= 27) exhibited a wide range of spontaneous EPSC frequencies (sEPSC frequency of <0.6 Hz for 20 cells and 1.2–16 Hz for 7 cells), with a higher mean sEPSC frequency value (1.38 ± 0.65 Hz;p < 0.05) (Fig. 1F,H). This would be consistent with the notion that the rectification property of synaptic currents was related not only to the spontaneous synaptic activity at the time of measurements but also to the previous history of the synapse. These observations are in keeping with the idea that surface expression of GluR2-containing receptors varies between stellate cell synapses and is related to the level of intrinsic synaptic activity.

Decreased Ca²⁺ permeability accompanies the change in rectification of extrasynaptic AMPARs

The activity-dependent change in the rectification properties of the synaptic current (from inwardly rectifying to linear) implies a decrease in the Ca²⁺ permeability of the synaptic AMPARs (Liu and Cull-Candy, 2000). However, it is difficult to test for this directly by examining EPSCs in various levels of external Ca²⁺, because this procedure greatly alters transmitter release. While synaptic activity triggers the targeting of GluR2-containing receptors to the synapses, spontaneous action potential activity facilitates their expression in the soma (the majority of stellate cells fire action potentials spontaneously in cerebellar slices in the absence of excitatory synaptic inputs [Hausser and Clark, 1997]). Our previous experiments have suggested that Ca²⁺ entry via N-type calcium channels increases the expression of extrasynaptic GluR2-containing AMPA receptors in the soma (Liu and Cull-Candy, 2000). We therefore examined the AMPAR channels in outside-out patches excised from the soma of stellate cells to determine whether their change in rectification properties was accompanied by decreased permeability to Ca²⁺.

As shown in Figure 2A, bath application of the AMPAR agonist kainate (100 μm) to outside-out patches evoked steady-state currents with outwardly rectifying I–V relationships. A similar outwardly rectifying I–V relationship was described for steady-state currents in patches from chick nMAG neurons and in cells expressing recombinant GluR2-containing receptors (Verdoorn et al., 1991; Raman and Trussell, 1995). In 26 patches examined, the average rectification index value was 1.1 ± 0.1 (n = 26) (Fig. 2C; Table1). This suggests that GluR2-containing AMPARs are present in the soma. After TTX treatment (1 μm) for >2 hr, the agonist-evoked currents displayed an inwardly rectifying I–V relationship (Fig.2B), and the rectification index was 0.52 ± 0.08 (n = 8) (Fig. 2D; Table 1).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Rectification properties of kainate-evoked currents in outside-out patches excised from the soma of stellate cells. A, Outwardly rectifying I–Vrelationship of kainate-evoked currents in an outside-out patch from a control cell. Spermine (100 μm) was included in the pipette solution. B, The agonist-evoked current in an outside-out patch excised from a TTX-treated stellate cell exhibited an inwardly rectifying I–V relationship. The slice was treated with 1 μm TTX for >2 hr and subsequently washed before the recording. C, Histogram of the rectification index of kainate-evoked currents in outside-out patches from control cells (defined as the ratio of current amplitude at +40 mV vs the predicted linear value at +40 mV). The rectification index was 1.1 ± 0.1 (n = 26 patches). D, Histogram of the rectification index of agonist-evoked currents in outside-out patches from TTX-treated cells (n = 8 patches) and cells treated with ω-conotoxin (ω-CTX) (500 nm ω-conotoxin GVIA for >2 hr before recording; n = 7 patches). A current trace was shown in our previous publication (Liu and Cull-Candy, 2000).

To test whether an increased Ca²⁺permeability accompanied this change, we compared agonist-evoked currents in patches from control and TTX-treated cells in Na⁺-rich (135 mm NaCl) and Ca²⁺-rich (30 mmCaCl₂) external solutions (see Materials and Methods). The relative Ca²⁺ permeability was determined from reversal potentials of these currents in Na⁺-rich and Ca²⁺-rich solutions. Because the presence of intracellular spermine does not alter the reversal potential of agonist-evoked currents regardless of whether it is mediated by Ca²⁺-permeable or Ca²⁺-impermeable AMPARs (Rozov and Burnashev, 1999), spermine was not included in the pipette solution so as to minimize voltage-dependent block of Ca²⁺-permeable receptors. As shown in Figure 3A, the current in control patches reversed at more hyperpolarized potentials in Ca²⁺-rich (−51.6 ± 8.8 mV;n = 4) than in Na⁺-rich (2.2 ± 3.2 mV) solution. In contrast, patches from TTX-treated cells displayed reversal potentials of −14.8 ± 6.1 mV (n = 5; p < 0.01 compared with theV_revCa value in control patches) in Ca²⁺-rich solution (Fig. 3B), quite close to the value obtained in Na⁺-rich solution (−1.1 ± 3.6 mV) in these same patches. The inwardly rectifying AMPARs (in TTX-treated cells) therefore displayed a greater degree of Ca²⁺ permeability.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

High Ca²⁺ permeability of AMPARs in outside-out patches from TTX-treated cells. TheI–V relationships of agonist-evoked currents in an outside-out patch were measured in Na⁺-rich solution and in Ca²⁺-rich solution. The reversal potentials of these currents, V_revNa andV_revCa, were determined from theirI–V plots. A, I–Vrelationship of a response from an outside-out patch from control cells. The currents reversed at a more hyperpolarized potential in Ca²⁺-rich solution (30 mmCaCl₂; dashed line) than in Na⁺-rich solution (135 mm NaCl;solid line), indicating a lower permeability to Ca²⁺ than to Na⁺.Bottom, The same I–V relationship as shown at the top, but on an expanded scale. Spermine was not included in the pipette solution. B,I–V relationship of a response from a patch from TTX-treated cells. The reversal potential of the agonist-evoked current in Ca²⁺-rich solution is close to the reversal potential in Na⁺-rich solution. C, The Ca²⁺ permeability of AMPARs in outside-out patches from control cells (n = 4) was significantly lower than that from TTX-treated cells (n = 5; p < 0.04). The Ca²⁺ permeability, P_Ca/P_Na, was calculated from the reversal potentials in Na⁺-rich and Ca²⁺-rich solutions (as described in Materials and Methods).

The reversal potentials were determined from I–Vrelationships and corrected for the junction potential of 9.2 mV in Na⁺-rich solution and 6.3 mV in Ca²⁺-rich solution. From these experiments, we estimated the relative Ca²⁺ permeability, P_Ca²⁺/P_Na⁺, of the AMPA receptors (see Materials and Methods). In patches from TTX-treated cells, our estimate for Ca²⁺ permeability (P_Ca/P_Na = 1.64 ± 0.52; n = 5) was at least six times higher than in patches from control cells (P_Ca/P_Na = 0.25 ± 0.11; n = 4; p < 0.04) (Fig.3C). It is of note that the very low (but nonzero) permeability of NMDG via AMPA receptors (P_NMDG/P_Cs = 0.01–0.02) (Burnashev et al., 1996) may slightly alter the calculated permeability ratios. Thus, the activity-dependent increase in the rectification index was indeed accompanied by a decreased Ca²⁺ permeability of the somatic AMPAR channels (Table 1).

Time course of synaptic currents

Having found that the expression of GluR2-containing AMPARs at the synapse was related to higher spontaneous synaptic activity, we examined whether the presence of GluR2 modified the time course of the EPSC at these synapses. The EPSC decay time was fitted either with a single exponential or by the sum of two exponential functions. The weighted decay time constant of synaptic currents at −60 mV, in cells that displayed a high rate of spontaneous synaptic activity (4.6 ± 0.03 Hz; R_+40/−60 = 0.37 ± 0.03; n = 8), was 0.72 ± 0.06 msec. The cells in which both the rate of spontaneous EPSCs and the R_+40/−60 value were low (0.18 ± 0.07 Hz,p < 0.03; R_+40/−60 = 0.23 ± 0.01, p < 0.001; n = 8) displayed EPSCs with a weighted decay time constant of 0.77 ± 0.06 msec (p = 0.59). This suggests that EPSCs displaying different rectification properties have a similar decay time course in these cells.

Because the decay time constant varied among stellate cells (ranging from 0.50 to 1.1 msec), we examined the effects of GluR2 on the time course of the EPSC at the same synapse by making use of the fact that spermine selectively blocks GluR2-lacking AMPARs at depolarized potentials. This allowed us to compare EPSCs at +40 mV, which are expected to be mediated only by GluR2-containing receptors, with currents at −60 mV mediated by a mixed AMPAR population. As a control, we first tested whether EPSCs mediated primarily by GluR2-containing receptors displayed different decay kinetics when measured at these two potentials. In cells that displayed a high rate of spontaneous EPSCs, the weighted decay time constant of EPSCs at +40 mV was 0.84 ± 0.13 msec (n = 8), not significantly different from that at −60 mV (p = 0.38 by paired ttest).

We subsequently examined the kinetic properties of EPSCs recorded from cells in which spontaneous EPSC frequency occurred at a low rate (0.15 ± 0.05 Hz; n = 14). An example is illustrated in Figure4A. The time course of evoked EPSCs at both positive and negative potentials was fitted either with a single exponential or by the sum of two exponential functions. In 5 of the 14 cells, the EPSCs at −60 mV were well fitted with a single exponential decay. In the other nine cells, the decay time course was best fitted by the sum of two components. The fast component made up 95 ± 2% of the peak amplitude and decayed with a time constant of 0.67 ± 0.03 msec; the slower component decayed with a time constant of 2.42 ± 0.31 msec (Fig. 4). The weighted mean decay time constant was 0.72 ± 0.04 msec (n = 14), and the average 10–90% rise time was 0.16 ± 0.01 msec (Fig. 4C,D).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

EPSCs mediated by GluR2-containing and by GluR2-lacking AMPARs have similar kinetic properties. A, At −60 mV, EPSCs were mediated by a mixed population of AMPARs; the 10–90% rise time was 0.20 msec. The decay time course (fitted with the single exponential function) gave a time constant of 0.86 msec. At +40 mV, EPSCs were mediated by GluR2-containing AMPARs. The 10–90% rise time of the synaptic current was 0.18 msec. The decay time course was best fitted with the two exponential functions. The fast component (96% of peak amplitude) had a decay time constant of 0.81 msec; the slow component was 3.28 msec. The weighted decay time constant was 0.90 msec. B, Synaptic currents at −60 and +40 mV normalized to their peak current amplitude to allow comparison of decay time. In this example, the decay time was marginally slower at +40 mV.C, The decay time constant of synaptic currents at −60 mV was not significantly different from that at +40 mV (p = 0.12 by paired t test;n = 14 cells). Open circles andopen triangles are individual data, and filled circle and filled triangle are average values ± SE. D, Summary of the 10–90% rise time of synaptic currents at −60 and +40 mV (p = 0.37 by paired t test;n = 14 cells).

At +40 mV, in 9 of the 14 cells, the decay time course was fitted by two exponential functions: a fast component with a decay time constant of 0.66 ± 0.08 msec (96 ± 2% of the peak amplitude) and a slower component with a decay time constant of 3.38 ± 0.47 msec (Fig. 4A). The decay of the synaptic current in the other five cells could be well fitted by a single exponential function. The mean weighted decay time constant for all cells was 0.87 ± 0.08 msec (n = 14); this was not significantly different from the decay time constant at −60 mV (p = 0.12 by paired t test) (Fig.4B,C). The small difference depicted in Figure4B was not significant in relation to the data as a whole. The average 10–90% rise time was 0.17 ± 0.02 msec (Fig.4D). These observations suggest that targeting of GluR2-containing AMPARs to the synapse did not significantly alter the decay time course of the synaptic current in stellate cells.

Effects of cyclothiazide on inward and outward synaptic currents

Cyclothiazide suppresses the desensitization and deactivation of AMPA receptors and increases their apparent affinity for glutamate (Patneau et al., 1993; Trussell et al., 1993; Partin et al., 1996;Yamada and Turetsky, 1996). However, these effects depend on the subunit isoforms involved in AMPA receptor formation. Each subunit can exist as flip or flop splice variants (Sommer et al., 1991; Seeburg, 1993; Hollmann and Heinemann, 1994); the sensitivity to cyclothiazide varies between flip andflop (Partin et al., 1994, 1996). For receptors containing GluR1_flip subunits, deactivation is markedly slowed by cyclothiazide (Partin et al., 1996). Furthermore, several studies have suggested that potentiation of AMPARs by cyclothiazide may also depend on the identity of the AMPAR subunits forming the receptor (Yamada and Turetsky, 1996; Dai et al., 2001), with the extent of potentiation being significantly reduced by hetero-oligomerization (Cotton and Partin, 2000). We examined evoked and miniature EPSCs to determine whether the involvement of the GluR2 subunit isoform modified the influence of cyclothiazide on the synaptic current properties. We included spermine in the pipette solution and compared the effects of cyclothiazide on EPSCs mediated by different complements of Ca²⁺-impermeable and Ca²⁺-permeable receptor subtypes at the stellate cell synapse.

Evoked EPSCs were examined at +40 and −40 mV in control conditions and in the presence of 100 μm cyclothiazide (Fig.5A,B). As expected, the decay time course of EPSCs was slowed in the presence of CTZ both at +40 mV (weighted decay time constant, τ, changing from 1.0 ± 0.13 to 6.57 ± 1.35 msec; n = 5; p < 0.02 by paired t test) and at −40 mV (from 0.97 ± 0.16 to 2.58 ± 0.35 msec; n = 5;p < 0.02) (Fig. 5; Table2). Because the receptors composed offlop isoforms are less sensitive to modulation of their desensitization and deactivation kinetics by cyclothiazide thanflip splicing variants (Patneau et al., 1993; Partin et al., 1996; Yamada and Turetsky, 1996), this result is consistent with the idea that EPSCs are mediated, at least in part, by flipisoform-containing AMPARs at these synapses. Furthermore, the change in decay time was more pronounced at positive potentials (τ_CTZ/τ_control = 6.6) than at negative potentials (τ_CTZ/τ_control = 2.66;p < 0.02 by paired t test, comparing τ values at −40 mV in the presence of cyclothiazide with those at +40 mV) (Fig. 5A,B; Table 2). Thus, it is possible that moreflop isoforms were present in Ca²⁺-permeable receptors than in GluR2-containing receptors.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Effect of cyclothiazide on the time course and amplitude of EPSCs and miniature EPSCs at hyperpolarized and depolarized potentials. A, B, The evoked EPSCs were recorded at −40 and +40 mV before (A) and during (B) perfusion of 100 μmcyclothiazide. The bottom traces show EPSCs normalized to the peak current amplitude at −40 and +40 mV. Note the difference in the effect on time course at −40 and +40 mV. C, D, Miniature EPSCs were recorded at −60 and +40 mV in the presence of 1 μm TTX before (C) and during (D) bath application of 100 μm cyclothiazide. Note the increase in amplitude at −60 mV and the absence of change at +40 mV after cyclothiazide treatment. All traces were average EPSCs of 60–90 evoked EPSCs and of 30–120 miniature EPSCs.

Cyclothiazide affected EPSC amplitude differently at positive and negative potentials. At +40 mV, when the current is expected to be mediated by GluR2-containing receptors, the EPSC amplitude was unaffected (23.1 ± 2.3 pA in controls and 22.0 ± 3.2 pA in cyclothiazide). In contrast, at −40 mV, the amplitude was increased by ∼50% (from −82 ± 17 pA to −122 ± 22 pA;n = 5; p < 0.02 by paired ttest) (Fig. 5A,B; Table 2). Because cyclothiazide affects several properties of AMPARs, its effects on EPSCs could be complicated (see Discussion). One possibility would be that receptors mediating EPSCs at negative potentials (mixed receptor population) desensitize more rapidly than those at positive potentials (GluR2-containing receptors); slowing desensitization kinetics by cyclothiazide could cause an increase in EPSC amplitude at hyperpolarized potentials. Work on both recombinant and native receptors has shown that the receptors composed of flop isoforms exhibit rapid desensitization compared with flip isoforms (Mosbacher et al., 1994; Geiger et al., 1995; Lambolez et al., 1996). Therefore, this could result from more flop isoforms present in Ca²⁺-permeable receptors than in GluR2-containing receptors, which is consistent with our observation that cyclothiazide produced a more pronounced change in decay time at positive than at negative potentials.

It is worth noting that the percentage increase in evoked EPSC amplitude at −40 mV is likely to give an underestimate of the effects of cyclothiazide on GluR2-lacking receptors. Thus, we subtracted the synaptic current at +40 mV (which was mediated by GluR2-containing receptors) from the EPSC at −40 mV (mixed receptor population) to obtain a current mediated by GluR2-lacking receptors. In the absence of the GluR2-containing component of the EPSC, the percentage increase of the current amplitude from GluR2-lacking receptors was ∼60% (from −63.5 ± 18.1 to −101.0 ± 14.0 pA; p < 0.02 by paired t test), and the weighted decay time constants increased from 0.73 ± 0.09 to 2.16 ± 0.32 msec (n = 5; p < 0.01 by pairedt test).

To exclude the possibility that the effects we observed could simply be ascribed to an alteration in transmitter release properties (Diamond and Jahr, 1995; Ishikawa and Takahashi, 2001), we examined mEPSCs (in 1 μm TTX) in the presence and absence of cyclothiazide. The frequency of mEPSCs appeared to change from 0.043 ± 0.009 to 0.265 ± 0.109 Hz (n = 4; p = 0.13 by paired t test), indicating some presynaptic action of cyclothiazide. However, as found with the evoked EPSCs, the decay time course of mEPSCs was prolonged at both +40 mV and −60 mV after application of cyclothiazide (Fig. 5C,D). Furthermore, their amplitude was increased by 56%, from −55.2 ± 2.7 to −85.9 ± 5.0 pA at −60 mV (n = 4; p < 0.0001 by paired t test), whereas at +40 mV, mEPSC amplitude remained unaltered (19.5 ± 2.7 and 21.0 ± 4.4 pA). The fact that cyclothiazide has similar effects on evoked EPSC and mEPSC properties in stellate cells suggests that the changes we observed were attributable primarily to its postsynaptic actions.

I–V relationship of EPSCs varies between stellate cells

Does synaptic activity regulate the targeting of GluR2 subunits under physiological conditions? Synaptic activity can modulate AMPAR-mediated transmission by changing the number of AMPARs at the postsynaptic membrane and by regulating the targeting of the AMPAR subtypes to the synapse (O'Brien et al., 1998;Turrigiano et al., 1998; Carrol et al., 1999; Shi et al., 1999). We have shown previously that EPSCs at the parallel fiber input to the stellate cell synapse exhibit the characteristics expected of GluR2-lacking AMPARs and that expression of GluR2-containing receptors at synapses increases after high-frequency synaptic stimulation (Liu and Cull-Candy, 2000). If this activity-dependent targeting of GluR2-containing AMPARs occurs under normal physiological conditions, one would expect a wide range of rectification values of synaptic currents at different synapses. As predicted, we found that synaptic currents show various degrees of rectification and that the degree of rectification correlated with the rate of spontaneous activity in these cells. This result is consistent with the idea that intrinsic synaptic activity could increase the expression of GluR2-containing AMPARs at these synapses. The rate of spontaneous EPSCs is likely to be related to the low spontaneous activity of presynaptic granule cells (D'Angelo et al., 1995). Furthermore, the frequency of spontaneous synaptic currents in slices may be reduced by the removal of some granule cells during the slicing procedure and therefore could be lower than thatin vivo. However, it is also possible that changes in synaptic activity during slice preparation could potentially alter the rectification of EPSCs in cerebellar slices.

The observation that the rectification property of synaptic currents varied from cell to cell is not unique. A wide range of I–Vrelationships of synaptic current was also found at the mossy-fiber input on stratum lucidum interneurons in the hippocampus (Toth and McBain, 1998).

Activity-dependent change in AMPAR subtypes alters Ca²⁺ permeability of receptors

On the basis of the correlation between the rectification properties and Ca²⁺ permeability, we postulated previously that the activity-dependent change in the rectification properties of EPSCs was accompanied by a decrease in the Ca²⁺ permeability of receptors. Because it is difficult to directly assess the Ca²⁺permeability of synaptic AMPARs by examining the reversal potentials of EPSCs in different external Ca²⁺concentrations, we made use of the fact that in stellate cells, neuronal activity also regulates the targeting of somatic AMPAR subtypes.

AMPARs in the soma displayed an outwardly rectifying I–Vrelationship and therefore would be expected to be Ca²⁺ impermeable. After suppression of action potential activity with TTX, AMPARs in the soma displayed an inwardly rectifying I–V relationship. We thus determined the Ca²⁺ permeability of these somatic AMPARs. As predicted, the Ca²⁺ permeability of AMPARs in outside-out patches from TTX-treated cells was higher than in patches from control cells. The Ca²⁺ permeability of AMPARs in patches from control stellate cells (P_Ca/P_Na = 0.25) appears to be slightly lower than the value described previously in nucleated patches from mouse stellate cells (P_Ca/P_Na = 0.46) (Bureau and Mulle, 1998). The Ca²⁺ permeability of AMPARs from TTX-treated cells was comparable with the values estimated for dentate gyrus basket cells and Hilar interneurons (Geiger et al., 1995). Thus the difference in the rectification properties of AMPAR-mediated currents does indeed correlate with a change in Ca²⁺ permeability of the receptors in stellate cells, suggesting that the change in EPSC properties after high-frequency synaptic stimulation is likely to result from a change from Ca²⁺-permeable to Ca²⁺-impermeable synaptic AMPARs.

Does the presence of the GluR2 subunit affect other properties of EPSCs in stellate cells?

The decay time constant of synaptic currents in stellate cells is comparable with the deactivation time constant of somatic AMPARs, indicating that receptor deactivation is primarily responsible for the rapid decay of these EPSCs (Barbour et al., 1994). The decay time course of evoked EPSCs in our experiments appeared to be faster than in the previous report (Barbour et al., 1994), possibly because EPSCs were evoked with minimal stimulation to reduce the asynchronous release of transmitter from presynaptic sites.

Synaptic currents that display a rapid time course have also been described at the mossy fiber–cerebellar granule cell synapse (Silver et al., 1996), at the granule cell–basket cell synapse in the hippocampus (Geiger et al., 1997), and at many auditory synapses (Trussell et al., 1993; Barnes-Davies and Forsythe, 1995; Otis et al., 1996; Gardner et al., 1999). Studies of native and recombinant AMPARs have indicated that their desensitization and deactivation kinetics are critically dependent on their subunit composition (Mosbacher et al., 1994; Geiger et al., 1995). Furthermore, it has been proposed that AMPAR-mediated currents display slow gating properties in neurons expressing high levels of GluR2_flip subunits (Geiger et al., 1995).

Synapses in stellate cells contain a mixed AMPAR population that can undergo a rapid activity-dependent switch in subtypes. This allowed us to test whether the expression of the GluR2 subunit affects the kinetics of synaptic currents at this synapse. We have shown previously that high-frequency synaptic activity induces a change from GluR2-lacking to GluR2-containing receptors, with little effect on the decay time course of EPSCs (Liu and Cull-Candy, 2000). In the present study, we found that the decay time constant of the synaptic current mediated by GluR2-containing receptors is not significantly different from that mediated by a mixed AMPAR population. Previous studies have shown that a fraction of Ca²⁺-permeable AMPARs are tonically blocked by polyamines even at negative potentials (Bowie et al., 1998; Rozov and Burnashev, 1999). Could the lack of difference in the decay time course observed in our experiments reflect the fact that a higher-than-expected proportion of GluR2-containing receptors contributed to the evoked EPSCs at hyperpolarized potentials? We have shown that the amplitude of synaptic currents was significantly reduced at depolarized potentials and that synaptic currents mediated by GluR2-lacking AMPARs (measured as the difference between EPSCs at −40 and +40 mV) constituted ∼77% of total synaptic currents at −40 mV. Therefore, a substantial amount of the synaptic current at negative potentials was mediated by GluR2-lacking receptors. Together, these results provide evidence that the involvement of GluR2-containing AMPARs did not significantly alter the decay time course of the synaptic current in these cells, in contrast to the general expectation that synaptic currents mediated by GluR2-containing receptors exhibit slower kinetic properties than GluR2-lacking receptors (Geiger et al., 1995).

This result is consistent with the idea that both Ca²⁺-permeable and Ca²⁺-impermeable AMPARs display rapid deactivation kinetics in stellate cells. Several studies on recombinant receptors demonstrate that homomeric GluR1, GluR2, and GluR4 receptors display rapid deactivation kinetics that appear to be similar forflip and flop isoforms (Mosbacher et al., 1994;Koike et al., 2000). Unfortunately, the deactivation time constants of GluR3 homomeric and of GluR2/3 heteromeric receptors, the ones likely to be present at stellate cell synapses (Keinanen et al., 1990; Sato et al., 1993), have not yet been described. Our results imply that the primary effect of inclusion of GluR2 is to modify Ca²⁺ permeability, with little effect on EPSC time course in stellate cells.

Does inclusion of the GluR2 subunit influence the effect of cyclothiazide on EPSCs? Cyclothiazide preferentially reduces AMPAR desensitization for flip splice variants (Partin et al., 1994, 1995). Furthermore, Partin et al. (1996) demonstrated that cyclothiazide selectively slows the rate of deactivation of GluR1_flip but not deactivation of GluR1_flop. The results of our experiments showing that cyclothiazide slowed the decay time course of synaptic currents would suggest the presence of the flip isoform of these subunits in at least a proportion of the synaptic AMPARs.

We also found that cyclothiazide induced a large increase in the amplitude of synaptic currents mediated by a mixed receptor population and little change in the current mediated by GluR2-containing receptors. Studies on the potentiation of recombinant AMPARs by cyclothiazide (Yamada and Turetsky, 1996; Cotton and Partin, 2000) show a greater degree of potentiation for GluR3_fliphomomeric receptors than for GluR3_flip/GluR2_fliphetero-oligomeric receptors, in agreement with our observation. Another possible explanation would be that the Ca²⁺-permeable receptors are desensitized more rapidly than GluR2-containing receptors, because GluR3_flop homomeric receptors have faster desensitization kinetics (τ_des = 1.4 msec) (Mosbacher et al., 1994).

In conclusion, our experiments demonstrate that synaptic currents in stellate cells display various degrees of rectification. This finding is in agreement with the idea that spontaneous synaptic activity increases the expression of GluR2 subunits at these synapses. Our results raise the possibility that intrinsic synaptic activity, at least in part, controls the targeting of GluR2 subunits at synapses in stellate cells under normal physiological conditions. Expression of GluR2-containing receptors reduces the Ca²⁺ permeability of AMPARs in outside-out patches but does not significantly alter the kinetics of synaptic currents. Thus, the activity-dependent switch in AMPAR subtypes reduces the Ca²⁺ entry and sensitivity to block by intracellular spermine. However, inclusion of GluR2-containing receptors produces surprisingly little effect on the EPSC kinetics.