Introduction

Top Abstract Introduction Materials & Methods Results Discussion References

Studies of cGMP in mammalian brain begin with the early reports that activation of glutamate receptors in cerebellum causes a dramatic increase in the cGMP content of this tissue (Mao et al., 1974; Rubin and Ferrendelli, 1977; Garthwaite and Balázs, 1978). Subsequently, the signaling pathway mediating cGMP production in neurons was shown to involve an increase in intracellular Ca⁺⁺, leading to a Ca⁺⁺-calmodulin-dependent activation of NOS (Bredt and Snyder, 1989; Mayer et al., 1992) and NO stimulation of soluble guanylate cyclase typically located in cells in close proximity (Garthwaite, 1991). In the cerebellum, NOS appears to be localized to granule cells (Bredt et al., 1991) and the NO-sensitive soluble guanylate cyclase is localized to granule cells (Southam et al., 1992; Southam and Garthwaite, 1993) and to Purkinje neurons (Zwiller et al., 1981; Ariano et al., 1982; Matsuoka et al., 1992), which are thought to be largely responsible for the dramatic kainate-mediated increase in cerebellar cGMP (Rubin and Ferrendelli, 1977).

The consequences of glutamate receptor-linked elevation of intracellular cGMP in Purkinje and granule neurons are poorly understood, but it is likely that cGMP activates G-kinase in the Purkinje neurons (DeCamilli et al., 1984), which contain high levels of a 23-kd cytosolic protein called G-substrate (Aswad and Greengard, 1981). cGMP is also the intracellular ligand for cyclic nucleotide-gated nonselective cation channels, thus depolarizing a number of cell types (Yau and Baylor, 1989; Zagotta and Siegelbaum, 1996). An important role is proposed for cGMP in LTD in cerebellum (Ito and Karachot, 1992; Daniel et al., 1993; Hartell, 1994; Lev-Ram et al., 1997). In single Purkinje neurons, for example, LTD has been linked to an increase of cGMP, in synergy with an increase of intracellular Ca⁺⁺ and of NO production (Lev-Ram et al., 1997). The mechanism by which an elevation of intracellular cGMP contributes to synaptic depression is as yet unknown.

A less widely known area of investigation is the interaction between the extracellular domain of glutamate receptor channels and guanosine nucleotides. Interestingly, most guanosine compounds appear to interact with the extracellular domains of glutamate receptors. Early studies of goldfish brain kainate binding proteins suggested that a G protein-coupled mechanism may be involved in guanosine compound-mediated inhibition of [³H]kainate binding (Willard et al., 1991; Willard and Oswald, 1992; Ziegra et al., 1992a; 1992b). Subsequent work provided evidence that GTPS-mediated inhibition of [³H]kainate binding in goldfish brain membranes was at least partially explained by competitive interactions at the agonist binding site on either the kainate binding proteins or ionotropic glutamate receptors (Barnes et al., 1993). Gorodinsky et al. (1993) demonstrated a low-affinity (IC₅₀ = 1 mM) competitive inhibition by GTP for the low-affinity [³H]AMPA binding site in rat cerebral cortex and a noncompetitive inhibition of high-affinity [³H]AMPA binding to rat cortex non-NMDA receptors. Likewise, a competitive mechanism was suggested by studies of [³H]kainate binding to detergent-solubilized recombinant kainate binding proteins where GTP and GDP analogs were equipotent competitors; in this study, however, cGMP was minimally effective (Paas et al., 1996a). By contrast, in a crude preparation of rabbit cerebellar membranes (P₂ fraction) high-affinity [³H]kainate binding was inhibited competitively by all the guanosine compounds tested, including cGMP, GMP, GDP and GTP, with comparable low affinity (IC₅₀ = 1.5 mM) (Poulopoulou, 1994). The differences between the binding data from partially purified receptors and a crude synaptosomal preparation, coupled with reports that cGMP does inhibit glutamate responses in electrophysiological studies when applied at the extracellular surface (Linden et al., 1995), suggested that cGMP may preferentially interact with intact non-NMDA receptors in cell membranes, in contrast to the other guanosine nucleotides that interact with sites on soluble receptors (Paas et al., 1996a).

We have examined the nature of the extracellular cGMP inhibition of non-NMDA responses of cultured neurons to slow bath application of kainate, which activates primarily AMPA receptor channels because kainate receptor channels are rapidly and profoundly desensitized by continuous exposure to kainate (Lerma et al., 1993). Patch-clamp recordings in whole-cell mode for cerebellar granule cells and in outside-out patches excised from Purkinje neurons were used to determine whether the effects of cGMP may depend on differences in the expression of non-NMDA receptor channel subunits by different cells. Cerebellar granule cells were chosen because there is evidence that their responses to different non-NMDA agonists are not homogeneous from cell to cell (Wyllie et al., 1993; Poulopoulou, 1994).

	Materials and Methods

Top Abstract Introduction Materials & Methods Results Discussion References

Primary cultures. Purkinje cell cultures were prepared as micro-explants from newborn mouse pups (12-36 h after birth) by the method of Moonen et al. (1982). Pups were anesthetized and killed with CO₂ according to an IACUC-approved protocol. Brains were rapidly dissected and placed in ice-cold Puck's saline with sucrose for ~10 min until the cerebella were dissected. Vermal, floccular and parafloccular areas were removed and discarded. The neocortical lobes of cerebellum were minced with irridectomy scissors and lightly dissociated by trituration (3-4 passes) through a large-bore Pasteur pipet in a MEM (Gibco BRL, Grand Island, NY) solution containing 10% fetal bovine serum and 10% heat-inactivated horse serum (Hyclone, Logan, UT). Small pieces of tissue and suspended cells were collected and plated on collagen-coated (Sigma Chemical Co., St. Louis, MO) 35-mm dishes (Falcon, Lincoln Park, NJ). When the explants were well attached (12-24 h), 1 ml of medium containing FUDR, Sigma was added. Cultures were grown in MEM/10-10 with FUDR for 2 days before changing to MEM/10% horse serum medium. Subsequently, culture solutions were partially changed at 3 to 4-day intervals. Purkinje neurons were recognized by their size, morphology and synaptic activity; recordings were performed between 10 and 20 days in vitro.

Recording conditions. Experiments were performed at room temperature (22°-25°C). Agonist responses were recorded in conventional whole-cell and outside-out patch modes as described by Hamill et al. (1981) with a List EPC-7 amplifier (Medical Systems) using borosilicate glass pipets (2-8 M; WPI TW150). Glass recording pipets were coated with Sylgard (Dow Corning, Midland, MI) and their tips lightly fire-polished. Patch pipets contained (in mM): 140 or 145 CsCl, 10 K-EGTA, 1 CaCl₂, 10 K-HEPES (pH 7.2). The majority of recordings were made with 4 mM ATP-Mg added to the pipet solution. Extracellular ("bath") solutions contained (in mM): 150 NaCl, 2.8 KCl, 1.0 CaCl₂, 10 Na-HEPES (pH 7.2) with 300 nM TTX (Sigma).

Pharmacological agents. Kainate (Sigma Chemicals and Research Biochemicals Inc.; RBI, Natick, MA), AMPA (RS--amino-3-hydroxy-5-methylisoxazole-4-propionic acid; Cambridge Research Biochemicals, and RBI, Natick, MA), cGMP and 8 bromo-cGMP (Sigma) were prepared as concentrated stock solutions, divided into aliquots and frozen. Frozen stock solutions were used within 20 days of preparation.

Data acquisition and analysis. Amplified data (List EPC7) were initially filtered at 4 kHz through an 8-pole low-pass Bessel filter (902LPF, Frequency Devices, Haverhill, MA), converted to digital format (Medical Systems, PCM-1B) and recorded on video tapes (Sony Beta SL HF450 using Maxell Gold, Fuji Beridox and BASF-HG tapes). Data were reconverted (PCM-1B) to analog signals and refiltered (Butterworth) at the desired cutoff frequency. Data analysis was done on personal computers (AST Premium) using CAP software (R. C. Electronics, Goleta, CA) and additional software written in the laboratory. Whole-cell and large-amplitude responses in excised patches were subjected to noise analysis.

Power spectral density analysis. Power spectral density analysis as described by Wright et al. (1991) was performed on control and agonist recordings from cells and excised patches that gave large responses. Currents were filtered (3 db point at 1.0 or 1.5 kHz) using an 8-pole Butterworth filter to minimize frequency-dependent changes in power density. Data were sampled at the Nyquist frequency (2 or 3 kHz) in blocks of 4096 points. Agonist-evoked current noise and control noise (instrument plus background membrane patch noise) were fast-Fourier-transformed into component frequencies. Power spectra were obtained by averaging between 40 and 70 blocks of data. Equal numbers of control and agonist blocks were used to generate any given pair of power spectra; agonist noise spectra are shown with control noise subtracted. Spectra were fitted (Simplex algorithm) by one Lorentzian function or the sum of two Lorentzian functions of the form

(1)

where S_(f) is the spectral density frequency f in Hz, S₀ is the zero frequency asymptote and f_c is the frequency at which the spectral density is 0.5 S₀. The proportion of the two components in two Lorentzian fit spectra is shown on the figures as S₀₁ and S₀₂, and the time constants ( values) underlying the noise were calculated as follows:

(2)

Variance analysis. The variance method was used to estimate channel conductance () in patches exhibiting large-amplitude responses, because individual channel events could not be detected, and in whole-cell recordings. Variance (²) of the agonist noise response was determined in pseudostationary conditions as the drug concentration increased from control to its equilibrium concentration. Data segments (4096 points) were chosen in control, during the onset and growth of the response and after the steady-state drug concentration was reached. A line was fitted to each data segment to determine the mean current level (I) in it, and the ² of the current noise about the I mean was calculated and plotted against I (²/I plots). Under the conditions of our experiments, where the final agonist concentration is less than the agonist EC₅₀, the data were fitted by a straight line using linear regression. The slope of the line was taken to be the elementary current of the active channels. The  of the channels was estimated by the formula

(3)

where i_e is the elementary current, V_R is the reversal potential of the response and V_H is the membrane holding potential during the response. The value i_e was determined as the slope of the ratio of the variance in the agonist-evoked current noise (² in pA²) to the mean transmembrane current (I in pA) as the agonist concentration is increasing from control to the full agonist response. Typically, if the final concentration is less than the EC₅₀ for the agonist; the pseudostationary variance analysis as performed here provides a condition where less than 50% of channels are likely to be active, and the slope of the ²/I plots was fitted by linear regression. In performing the power spectral density and variance analyses, we took care to use kainate concentrations 3- to 4-fold higher than needed to obtain a threshold agonist response, but less than the EC₅₀. Where noise analysis data are compared from different cells or under different conditions, mean values +/ S.E.M. are given.

	Results

Top Abstract Introduction Materials & Methods Results Discussion References

Whole-cell and outside-out patches from cerebellar granule cells and patches from some Purkinje neurons responded to application of 25 to 300 µM kainate with large-amplitude inward currents at a holding potential of 60 mV. All the neurons and patches examined responded to kainate. During the 20 to 75-min recordings, the cells and patches were subjected to multiple solution changes to ensure reproducibility of the drug effects. Cerebellar granule cells were grouped according to the characteristics of the agonist-evoked noise. Figure 1 shows two different types of responses to kainate observed in these cells. Variance analysis on these responses gave an average estimated conductance of 5.51 +/ 0.83 pS (n = 8) for the group of cells with high-noise responses referred to as type I cells. Meanwhile, the average conductance for the cells with low-noise responses (type II cells) was 0.88 +/ 0.55 pS (n = 18). None of the cells in this sample had estimated conductances between 2 and 4 pS.

View larger version (18K): [in this window] [in a new window]   Fig. 1.   Whole-cell currents from mouse cerebellar neurons from primary cultures in response to kainate application were initially classified into two groups according to the estimated aggregate conductance of the current responses. A) Application of 11 to 100 µM kainate to some cerebellar neurons evoked inward current responses associated with robust increases in the noise. Variance analysis on the current noise of one such response to 100 µM kainate gave an estimated ie of 0.37 pA in this cell for an estimated average  of ~6.2 pS. B) In other cells, 33 to 300 µM kainate routinely evoked inward currents that were accompanied by a modest increase in the noise. Variance analysis of the current noise in the cell depicted here gave an estimated  = 0.83 pS. In both examples, the holding potential was 60 mV and the reversal potential was 0 mV.

Concentration-effect data were collected from eight cerebellar granule cells for kainate between 8 µM and 1 mM; these data also indicated that mouse cerebellar granule cells are not uniformly sensitive to kainate. The high-noise responses in kainate (n = 3) were more sensitive to this agonist than the cells exhibiting low-noise responses (n = 5). Typical type I cells showed a small response to 11.1 µM kainate, whereas 33.3 µM kainate was required to evoke small-amplitude responses routinely in type II cells. Approximate EC₅₀ values were between 150 and 200 µM kainate for the type I cells and between 300 and 400 µM kainate for the type II cells. The type II cells exhibited larger average current responses in 100 to 300 µM kainate than the type I cells, a result consistent with the low-noise cells containing low-conductance non-NMDA channels, but many more active ones than the cells with high-noise responses to kainate.

The effects of cGMP/kainate coapplication were reversible and highly reproducible in the type I (high-noise) cells and patches. cGMP inhibited the kainate responses in all type I cells and in patches excised from Purkinje neurons (cells/patches = 18, trials = 43). Preliminary experiments had indicated that both cGMP and the membrane-permeable analog 8 Br-cGMP inhibited the kainate responses in type I patches at the same concentrations, and their time courses for inhibition and washout were similar. Subsequently, the membrane-impermeable form cGMP was used for recordings to control for the possibility that the patches/cells might contain some cyclic nucleotide-gated ion channels.

The cGMP-mediated inhibition of these kainate currents was rapidly reversed by subsequent perfusion with 25 µM kainate alone (fig. 2). A typical example of a large-current response to 25 µM kainate from a granule neuron is shown in figure 2A. After the response to the first application of 25 µM kainate had reached its steady-state amplitude, kainate and cGMP were coapplied. As illustrated, kainate currents were inhibited by cGMP in a dose-dependent fashion (fig. 2, B and C). Coapplication of 200 µM cGMP with 25 µM kainate resulted in a 60% reduction of the kainate current.

View larger version (21K): [in this window] [in a new window]   Fig. 2.   High-noise inward currents are inhibited by cGMP in a dose-dependent manner. A) The response of a type I cerebellar granule neuron to 25 µM kainate is inhibited ~60% by coapplication of 200 µM cGMP. B) Responses of an outside-out patch from a Purkinje neuron are shown. 25 µM kainate evoked large increases in the current and the noise (VH = 60 mV), a result consistent with the patch containing many kainate-sensitive receptor channels. Slow bath coapplication of 200 µM cGMP with 25 µM kainate in this patch decreased the kainate current by 70% (solid arrow). The current was further reduced to 92% (open arrow) by subsequent application of 600 µM cGMP, which indicates concentration-dependent inhibition of the kainate current by cGMP. C) Dose-inhibition data for cGMP in the kainate current responses from mouse cerebellar cells/patches. Currents activated by 25 µM kainate in the absence and presence of increasing concentrations of cGMP (5-1000 µM) were recorded from mouse cerebellar neurons and patches that contained many kainate-activated channels. Patches from cells as small as the granule neurons used in this study were generally nucleated (Sather et al., 1990). Standard-error bars are shown. The curve was drawn between the data points and the IC50 estimated from the graph.

Recordings from four cells and 12 outside-out patches that gave high-noise responses were used for estimating an IC₅₀ for cGMP-mediated inhibition in the following set of conditions: 25 µM kainate was applied alone, and then it was coapplied with three or four different concentrations of cGMP between 5 and 1000 µM. The steady-state control kainate current amplitude was measured, and the percent inhibition for each dose of cGMP was calculated. The dose-inhibition curve in figure 2C was constructed by plotting the percent of kainate current remaining in each concentration of cGMP against the log of the nucleotide concentration for each patch/cell. The IC₅₀ under these conditions was 150 (+/25) µM cGMP. Increasing the kainate concentration to 100 µM increased the agonist response amplitude in the absence and presence of cGMP. Cells with high-variance noise responses to 100 µM kainate were inhibited by 49.5% (n = 4 cells; 7 trials) in 500 µM cGMP. Approximately half of the granule cells with low-noise responses were comparably affected (fig. 3, A and B).

View larger version (23K): [in this window] [in a new window]   Fig. 3.   cGMP-mediated inhibition appears to be competitive. A) Whole-cell current responses of a granule cell to the application of 33 µM kainate (panel A1) and 100 µM kainate (panel A2) show the dose-dependent increase in the agonist response. B) Coapplication of 33 µM kainate and 500 µM cGMP (panel B1) and 100 µM kainate plus 500 µM cGMP (panel B2) responses on the same cell. Recordings in panels A and B were at 60 mV. Open triangles indicate changing to control solution. C) A current-voltage (I-V) relationship for a type I response from an outside-out patch is shown. The I-V relationships shown were obtained by subtracting the I-V obtained at control solution from the ones obtained in the presence of 25 µM kainate, with and without cGMP coapplication. The I-V relationship of the kainate responses in the absence of cGMP rectifies similar to macroscopic currents recorded in whole cells with kainate (Ascher and Nowak, 1988). In the presence of 500 µM cGMP, the I-V relationship is smaller at all voltages away from the reversal potential. D) The kainate current inhibited by 500 µM cGMP is plotted as a function of membrane potential. The relationship is linear, which indicates that the nucleotide inhibition of the kainate current is independent of voltage.

The mechanism of cGMP-mediated inhibition of kainate currents was investigated in several experiments: increasing the kainate concentration in the coapplied cGMP/kainate mixture to overcome the inhibition of cGMP (fig. 3, A and B), examining the voltage sensitivity of the cGMP-mediated inhibition (fig. 3C) and analyzing the effects of the cGMP on the noise variance and power spectra (fig. 4).

View larger version (31K): [in this window] [in a new window]   Fig. 4.   Noise analysis of type I kainate responses in the absence and presence of cGMP. Variance (panel A1) and power spectral density (panel A2) analyses of the kainate-activated noise recorded at 60 mV (raw data shown in fig. 2B) are presented for comparison with results of similar analyses conducted in the presence of 200 µM cGMP (panel B). The estimated elementary current during the application of 25 µM kainate alone was 0.45 pA (panel A1), and it was 0.44 pA during coapplication of 200 µM cGMP and 25 µM kainate (panel B1). Spectra of the currents from the same outside-out patch in response to 25 µM kainate in the absence of cGMP (panel A2) were fitted by the sum of two Lorentzian curves with slow and fast time constants of 3.5 and 0.6 ms, respectively. In the presence of 200 µM cGMP, the spectra (panel B2) of the remaining kainate current had a profile similar to the one in kainate alone. The spectrum for the 25 µM kainate (panel A2) was obtained by subtracting the power density in the control noise from the power density in the agonist-evoked current noise; control noise was also subtracted from the spectrum for the kainate-activated noise remaining during cGMP/kainate coapplication (panel B2). Control and agonist noise data were filtered at 1500 Hz (Butterworth) and sampled at 3000 Hz. The recording was made at 60 mV.

Figure 3 shows current traces from a whole-cell recording where application of 33 µM kainate evoked a type II noise and inward current. In the presence of 500 µM cGMP, this current response was nearly abolished. In the subsequent coapplication of a mixture of 100 µM kainate with 500 µM cGMP, the kainate response was larger than in 33 µM kainate alone, which indicates that increasing the kainate concentration overcomes cGMP-mediated inhibition. This observation, which is consistent with cGMP being a competitive antagonist at the kainate binding site, was obtained in the four cells and two excised patches tested.

The possible voltage dependence of the cGMP inhibition of kainate steady-state currents was examined in the presence and absence of the nucleotide, over holding potentials ranging from 100 to +40 mV, with measurements made at 20-mV increments in whole cells and outside-out patches that had high-amplitude kainate responses. The current-voltage (I-V) relationships of the cells and patches with 5 to 8-pS kainate responses were obtained by subtracting the I-V obtained in the control solution from that obtained in the presence of 25 to 100 µM kainate, plus or minus 500 µM cGMP. One example of such a pair of I-V curves from such an outside-out patch is presented in figure 3C. Macroscopic kainate currents exhibited a small outward rectification. The fraction of kainate current inhibited by cGMP was a linear function of holding potential (fig. 3D), which indicates that the observed inhibition of the kainate-induced currents by the nucleotide was independent of voltage. The reversal potential of the kainate currents, plus or minus the nucleotide, was near 0 mV in the symmetrical cation solutions used in the recordings. Thus cGMP did not appear to block the channels, nor did it alter their selectivity in type I cells/patches.

The noise characteristics of large inward currents evoked by 25 µM kainate in the absence and presence of 200 µM cGMP were examined in type I cells and patches. An example is presented in figure 4. Variance analysis was performed to determine whether cGMP affected the conductance underlying the type I kainate responses. Kainate ² in the absence (fig. 4A₁) and presence of cGMP (fig. 4B₁) indicated that cGMP did not affect the i_e of the remaining kainate-activated currents. This observation indicates that the conductances of the non-NMDA channels underlying the noise were unchanged by cGMP, a result consistent with the actions of a competitive antagonist. Likewise, power spectral density analysis of the agonist noise in the absence (fig. 4A₂) and presence of cGMP (fig. 4B₂) yielded similar results, which indicates that the cGMP reduced the kainate response amplitude without modifying the kinetics of the active receptor channels. In both cases, the power spectra were fitted by the sum of two Lorentzians, which suggests complicated kinetics of the channels and/or the activation of more than one channel population with different kinetics. Power spectra parameters in cGMP were virtually identical to those with kainate alone. This behavior, which is consistent with competitive antagonist actions, was observed in recordings from all high-variance noise responses, and for some cGMP-sensitive low-noise kainate responses (e.g., fig. 3A and B; power spectra not shown).

Effects of cGMP on low-noise kainate responses. Type II cells have estimated conductances between 0.30 and 2.0 pS (0.88 +/ 0.55 pS; n = 18). They apparently are not a single functional class, however, because their power spectra, obtained under similar experimental conditions, differed considerably from cell to cell (see below). Moreover, granule cells exhibiting low-noise kainate responses are found to differ considerably with respect to their sensitivity to bath-applied AMPA (Poulopoulou and Nowak, in preparation). The effects of cGMP on some low-noise kainate-responsive cells were investigated systematically to test for cell-to-cell differences.

View larger version (18K): [in this window] [in a new window]   Fig. 5.   Examples of low-noise granule cells and the variability of their sensitivity to cGMP coapplication. A) The response to 100 µM kainate in this cell was decreased 32% by coapplication of 500 µM cGMP. Variance analysis gave an estimated conductance of 0.63 pS (mean of 2 measurements) for kainate and of 0.51 pS for the coapplication of cGMP. B) In this cell, the response to 100 µM kainate was decreased by an average of 35% by coapplication of 500 µM cGMP (two trials). The estimated conductance was 1.57 pS for kainate and was 0.34 pS during coapplication of cGMP. C) The response to 100 µM kainate was inhibited by 11% (average of two trials). cGMP decreased the estimated conductance from 0.4 to 0.25 pS. All measurements were made at 60 mV, and the cells were recorded in the same session so that they were exposed to identical solutions.

View larger version (34K): [in this window] [in a new window]   Fig. 6.   Kinetics of cGMP-inhibited non-NMDA receptor channel noise in type II cells were modified by the presence of the antagonist. A) The spectrum for 100 µM kainate alone (panel A1) was fitted by the sum of two Lorentzian functions as described in "Materials and Methods." In the presence of 500 µM cGMP (panel A2), the shape of the spectrum was altered. It was now fitted with a single Lorentzian function with  = 1.2 ms, which suggests that this is the principal time constant of the cGMP-insensitive component of the kainate-activated noise. B) The power density spectrum of another cell in 100 µM kainate was also fitted by the sum of two Lorentzians. In this example, the power spectrum in cGMP/kainate resembled the one from the cell in panel A and again was fitted by a single time constant. This particular pair of spectra were generated from the responses exemplified by the raw data in figure 5B. We performed power density analysis of the underlying noise for kainate responses and of cGMP-insensitive and cGMP-sensitive fractions of kainate noise. All spectra shown have a control noise file of equal length subtracted from the response spectra. Control and agonist noise data were filtered at 1500 Hz (Butterworth) and sampled at 3000 Hz. The recordings were made at 60 mV.

View larger version (35K): [in this window] [in a new window]   Fig. 7.   Additional examples of power spectra from low-noise kainate responses illustrate the kinetic diversity and varying degrees of cGMP sensitivity of the so-called type II cells. A) The noise in this cell was analyzed to determine the nature of the kinetics of cGMP-sensitive currents. In panel A1, the spectrum shows that the noise in 100 µM kainate is fitted by the sum of two Lorentzians, and there is significantly more low-frequency noise in this spectrum than in many others shown in this report. In this cell, cGMP partially inhibited the kainate response, and the shape of the spectrum of the remaining noise (panel A2) was different from the one obtained in kainate alone (panel A1). Subtracting the cGMP/kainate spectrum (panel A2) from the kainate spectrum (panel A1) revealed the kinetics of the non-NMDA channels inhibited by cGMP in this cell (panel A3). Both low and high frequencies were inhibited. The relatively cGMP-insensitive portion contained mainly the intermediate frequencies ( = 3 ms). B) In this spectrum, from the patch in figure 5C, the kainate noise was fitted by the sum of two Lorentzians (1 = 95.4 ms, S01 = 94.1%; 2 = 2.12 ms, S02 = 5.9%). The coapplication of 500 µM cGMP did not change the amplitude of the kainate response more than 12.5%, nor did it change the shape of the spectrum remarkably (data not shown). However, the spectrum was no longer fit using our software, because it would not converge to the sum of two theoretical Lorentzian functions using the Simplex algorithm. Noise data were filtered at 1500 Hz (Butterworth) and sampled at 3000 Hz. The control files subtracted from the response files were of identical length. Recordings were made in cells voltage-clamped to 60 mV.

	Discussion

Top Abstract Introduction Materials & Methods Results Discussion References

Diverse expression of kainate-activated receptor channels in cerebellar granule cells. Noise analysis of the kainate-activated responses from cultured mouse cerebellar granule cells points out the diversity of these small neurons with respect to their expression of functional non-NMDA receptor channels. Variance analysis suggested that there were at least two different groups of cells, those with high-noise kainate responses (5.51 +/ 0.83 pS) and a second group with low-noise responses (0.88 +/ 0.55 pS), referred to respectively as type I and type II cells. Surprisingly, none of the 26 cells analyzed had estimated conductances between 2 and 4 pS, which suggests that there may be two distinct populations of cells. Preliminary concentration-effect studies also suggest that the non-NMDA receptors in type I cells may be more sensitive to kainate than those in type II cells. However, the differences in the effects of cGMP on low-noise kainate responses indicate that the hypothesis of there being two distinct populations of granule cells is too simple.

Extracellular cGMP inhibits kainate responses. Extracellular application cGMP to cerebellar granule cells demonstrated that this nucleotide inhibits kainate responses in a concentration-dependent fashion in the majority of neurons tested. In the case of the high-noise kainate responses, and also in the outside-out patches from Purkinje neurons, cGMP appears to be a weak competitive antagonist. The classical approach of comparing complete concentration-effect curves in the absence of the inhibitor, and observing a rightward shift in the curve in the presence of the cGMP, was not pursued here because of the technical problems associated with gathering such data from cells 4 to 6 µm in diameter. The conclusion that cGMP acts as a competitive inhibitor of the high-noise kainate responses is based on several observations: 1) The inhibition was voltage-independent. 2) The conductance of kainate-activated channels, estimated by variance analysis, is unchanged. 3) The kinetics of the agonist noise in the power spectra are unaffected by cGMP. 4) cGMP-mediated inhibition was overcome by increasing the kainate concentration.

Does extracellular cGMP play a physiological role in cerebellum? The direct inhibition of non-NMDA receptor-mediated responses by cGMP requires relatively high concentrations (100-1000 µM), which raises the question of what, if any, physiological role extracellular cGMP may play. The most recent estimates from in vivo microdialysis studies in rat cerebellum indicate the basal extracellular cGMP to be ~200 nM (Luo et al., 1994). Two groups have reported large increases in extracellular cGMP levels in vivo, in rat hippocampus stimulated with NO donor agents (Vallebuona and Raiteri, 1994) and in rat cerebellum after harmaline stimulation of climbing fiber input (Luo et al., 1994). The cerebellum study is of particular interest, because although the increase in extracellular cGMP measured by microdialysis was more modest (about a 5-fold increase above the basal level) compared with the hippocampal study, the cGMP release was inhibited by local TTX and nifedipine infusion, and it was mimicked by direct application of NMDA and non-NMDA agonists in the presence of TTX. Luo et al. (1994) also indicated that the increase in extracellular cGMP was rapid and transient, which suggests that cGMP may be released from Purkinje neurons. They did not speculate on the nature of the release mechanism but implied that it is more rapid than a purely diffusional process.

What is the site of the competitive cGMP interaction? Ionotropic glutamate receptors and kainate binding proteins are thought to have a large extracellular N-terminal "lobe," three transmembrane spanning portions and a cytoplasmic C-terminal (Hollmann et al., 1994; Wo and Oswald, 1994, 1995). The hypothesized sites of ligand-receptor interaction include portions of the N-terminal lobe and the extracellular loop between transmembrane spanning regions 3 and 4, which suggests that the agonist may span a pocket between the two lobes during binding (Stern-Bach et al., 1994; Mano et al., 1996).

Pharmacology of the cGMP binding site. There appears to be at least one common determinant of guanosine nucleotide binding across the kainate binding proteins and non-NMDA receptor subunits. In addition to observing comparable IC₅₀ values for the guanosine nucleotides (GTP, GDP, GMP and cGMP) in a crude synaptosome preparation, Poulopoulou (1994) found that the structurally related molecule theophylline also weakly inhibited [³H]kainate binding to rabbit cerebellar membranes. Preliminary electrophysiology experiments indicated that GTP and theophylline also inhibited kainate responses recorded in cerebellar cells, a result consistent with the pharmacological profile obtained in binding assays (Poulopoulou, 1994). Although the guanosine compounds and the methyl-xanthine compete for [³H]kainate binding in rabbit cerebellar membranes, none of the adenosine purine analogs affected the radioligand binding, which suggests that the double-bonded oxygen on the pyrimidine ring of the purine base imparts the selectivity of the guanosine nucleotides for interaction with the non-NMDA glutamate receptor subunit binding site (Poulopoulou, 1994). Paas et al. (1996a) compared the affinity of adenosine nucleotides and inosine to the guanosine nucleotides and concluded that the double-bonded oxygen on the guanosine base is important for the observed pharmacological selectivity for guanine nucleotides in kainate binding proteins. Paas et al. (1996a) also observed that GTP and GDP had higher affinity than cGMP for the triton X-100-treated receptors and postulated that the presence of at least one free phosphate moiety increased the affinity of the guanosine nucleotides over that of cGMP and guanosine. However, in the rabbit brain membranes, GTP and cGMP were of equally low affinity, a phenomenon that reflects either some GTPase activity in preparation or a difference between non-NMDA receptors and kainate binding proteins.