Identification of subunits contributing to synaptic and extrasynaptic NMDA receptors in Golgi cells of the rat cerebellum

Abstract

1. To investigate the properties of N-methyl-D-aspartate receptors (NMDARs) in cerebellar Golgi cells, patch-clamp recordings were made in cerebellar slices from postnatal day 14 (P14) rats. To verify cell identity, cells were filled with Neurobiotin and examined using confocal microscopy.

2. The NR2B subunit-selective NMDAR antagonist ifenprodil (10 μM) reduced whole-cell NMDA-evoked currents by ≈80 %. The NMDA-evoked currents were unaffected by the Zn²⁺ chelator N,N,N′,N′-tetrakis-(2-pyridylmethyl)-ethylenediamine (TPEN; 1 μM) suggesting the absence of NMDARs containing NR2A subunits.

3. Outside-out patches from Golgi cells exhibited a population of ‘high-conductance’ 50 pS NMDAR openings. These were inhibited by ifenprodil, with an IC₅₀ of 19 nM.

4. Patches from these cells also contained ‘low-conductance’ NMDAR channels, with features characteristic of NR2D subunit-containing receptors. These exhibited a main conductance of 39 pS, with a sub-conductance level of 19 pS, with clear asymmetry of transitions between the two levels. As expected of NR2D-containing receptors, these events were not affected by ifenprodil.

5. The NMDAR-mediated component of EPSCs, evoked by parallel fibre stimulation or occurring spontaneously, was not affected by 1 μM TPEN. However, it was reduced (by ≈60 %) in the presence of 10 μM ifenprodil, to leave a residual NMDAR-mediated current that exhibited fast decay kinetics. This is, therefore, unlikely to have arisen from receptors composed of NR1/NR2D subunits.

6. We conclude that in cerebellar Golgi cells, the high- and low-conductance NMDAR channels arise from NR2B- and NR2D-containing receptors, respectively. We found no evidence for NR2A-containing receptors in these cells. While NR2B-containing receptors are present in both the synaptic and extrasynaptic membrane, our results indicate that NR1/NR2D receptors do not contribute to the EPSC and appear to be restricted to the extrasynaptic membrane.

At many central excitatory synapses, the transmitter L-glutamate activates a mixture of NMDARs and non-NMDARs. The NMDARs are unique among mammalian synaptic receptors in displaying a voltage-dependent block by Mg²⁺ ions, and requiring the binding of both the transmitter and the co-agonist, glycine, for their activation (reviewed by Johnson & Ascher, 1996). The functional properties of NMDARs are influenced by a wide variety of endogenous modulators including Zn²⁺ ions, protons and polyamines (see Dingledine et al. 1999). Several functionally distinct subtypes of NMDARs are present in central neurons and these differ in their responses to glutamate, glycine and modulators (reviewed by Feldmeyer & Cull-Candy, 1996). Thus, the type of NMDAR present at a particular synapse has clear functional implications for transmission.

Molecular techniques have identified three families of NMDAR subunits. The NR1 family exists in a variety of splice variants, and is widespread throughout the central nervous system (CNS). The NR2 family comprises four members, NR2A, -2B, -2C and -2D, which are differentially distributed. More recently, the NR3 subunit has been identified, which has a relatively restricted distribution (see McBain & Mayer, 1994; Dingledine et al. 1999). Our knowledge of the relationship between NMDAR properties and their subunit composition has been greatly advanced by studies of native receptors, whose subunit complement has been inferred from in situ hybridization, immunocytochemistry or single cell PCR (Farrant et al. 1994; Momiyama et al. 1996; Paoletti et al. 1997; Plant et al. 1997; Stocca & Vicini, 1998) and by experiments on recombinant receptors (see for example, Stern et al. 1992; Monyer et al. 1994; Vicini et al. 1998; Wyllie et al. 1998).

Two functionally distinct classes of native NMDAR channels have been identified in the CNS: ‘high-conductance’ NMDAR channels, with conductance levels of 40 and 50 pS, and ‘low-conductance’ channels with levels of ∼18 and ∼38 pS (see Momiyama et al. 1996, and references therein). Studies of both recombinant and native receptors have ascribed the high-conductance events to receptors formed from NR1/NR2A or NR1/NR2B subunits (Stern et al. 1992; Farrant et al. 1994; Brimecombe et al. 1997). The low-conductance class of receptor channel is associated with receptor assemblies containing NR1/NR2C (Stern et al. 1992; Farrant et al. 1994; Takahashi et al. 1996) or NR1/NR2D subunits (Momiyama et al. 1996; Wyllie et al. 1996). We have been particularly interested in the low-conductance NR1/NR2D channels since they exhibit a number of unusual properties, including a low sensitivity to voltage-dependent block by extracellular Mg²⁺ (see Monyer et al. 1994; Momiyama et al. 1996) and unusually slow deactivation rate, following a brief pulse of glutamate (Monyer et al. 1994; Vicini et al. 1998; Wyllie et al. 1998). Thus, if present at the synapse, such receptors might be expected to produce currents, and Ca²⁺ entry, lasting several seconds.

In situ hybridization data suggest that the NR2D-containing NMDARs may be widespread in the cerebellum, being the only NR2 subunit mRNA detected in young Purkinje cells, stellate cells and Golgi cells of the rat (Watanabe et al. 1994; Akazawa et al. 1994). Previous work on immature Purkinje cells has demonstrated that while these cells express a homogeneous population of receptors with properties characteristic of NR1/NR2D assemblies, the receptors appear not to be activated synaptically (Momiyama et al. 1996).

Golgi cells are GABAergic interneurons, classically thought to contribute to the filtering of mossy fibre sensory input during its relay via granule cells to Purkinje cells (see Gabianni et al. 1994 for references). The importance of Golgi cells to cerebellar function is underlined by the fact that their selective ablation causes acute disruption of motor co-ordination (Watanabe et al. 1998). Recently, it has been suggested that the feedback excitation of Golgi cells via granule cell axons (parallel fibres) acts to synchronise activity of both cell types (Maex & DeSchutter, 1998; Vos et al. 1999). Although parallel fibre input to Golgi cells activates both AMPA- and NMDA-type glutamate receptors (Dieudonné, 1998), the subunit composition of the latter and their importance to parallel fibre efficacy remain unclear.

In the present study we have examined the biophysical and pharmacological properties of synaptic and extrasynaptic NMDARs in cerebellar Golgi cells. Contrary to expectations based on in situ hybridization data, our results indicate that Golgi cells express at least two types of NMDAR channel, only one of which appears to be expressed at the synapse.

METHODS

Slice preparation

Cerebellar slices were prepared from Sprague-Dawley rats (P14) as previously described (Farrant et al. 1994). Following decapitation, the brain was rapidly removed and placed in cold (2–4°C) ‘slicing solution’ (composition, mM: NaCl 125; KCl 2.5; CaCl₂ 1; MgCl₂ 4; NaHCO₃ 26; NaH₂PO₄ 1.25; glucose 25; pH 7.4 when bubbled with 95 % O₂ and 5 % CO₂). Slices (200–250 μm) were cut from the dissected cerebellar vermis in a sagittal or coronal orientation using a moving blade microtome (DTK-1000; Dosaka EM Co. Ltd, Kyoto, Japan). Slices were incubated in slicing solution at either room temperature (22–25°C) or 34°C for 1 h, and subsequently maintained at room temperature for up to 8 h. For patch-clamp experiments, slices were then transferred to a chamber on an Axioskop-FS microscope (Zeiss, Welwyn Garden City, UK).

Solutions

The standard external solution used during experiments was the same as the slicing solution, except that Mg²⁺ was omitted. In experiments where synaptic currents were recorded, external Ca²⁺ was increased from 1 to 2 mM. The internal solution, used for recording from outside-out patches, contained (mM): CsCl 140; NaCl 4; CaCl₂ 0.5; Hepes 10; EGTA 5; Mg-ATP 2 (adjusted to pH 7.3 with CsOH). This solution was also used for whole-cell recording from cerebellar granule cells. Whole-cell recordings from Golgi cells were usually made with an internal solution containing (mM): CsF 95; CsCl 25; Hepes 10; EGTA 10; NaCl 2; Mg-ATP 2; QX-314 10; tetraethylammonium chloride (TEA-Cl) 5; 4-aminopyridine (4-AP) 5.

In outside-out patch experiments and some whole-cell experiments, the following drugs were added to the standard Mg²⁺-free external solution: 10 μM bicuculline methobromide (Research Biochemicals International, Natick, MA, USA), 0.5–1 μM strychnine hydrochloride (Sigma, Poole, UK), 5 μM 6-cyano-7-dinitroquinoxaline-2,3-dione (CNQX; Tocris Cookson, Bristol, UK). In some experiments ifenprodil (Sigma), or N,N,N′,N′-tetrakis-(2-pyridylmethyl)-ethylenediamine (TPEN; Sigma) was applied in the external solution. For activation of NMDARs, 10–50 μM NMDA (Tocris) was applied with 10 μM glycine (BDH) in the presence of 300 nM tetrodotoxin (TTX, Sigma). Experiments investigating spontaneous and evoked EPSCs were made in the presence of 10 μM bicuculline and 0.5 μM strychnine to block GABAergic and glycinergic spontaneous inhibitory postsynaptic currents (sIPSCs) (Dieudonné, 1995).

Neurobiotin-filling procedure

N-(2-Aminoethyl)biotinamide hydrochloride (Neurobiotin; Vector Laboratories Inc., Burlingame, CA, USA) was added to the internal solution at a concentration of 5 mg ml⁻¹. This solution was used for whole-cell recording from Golgi cells. The slice was then washed and stored overnight in a cold solution of 0.1 M phosphate buffered saline (PBS; Sigma) containing 3 % paraformaldehyde. Slices were then washed in 0.1 M PBS and incubated for 1 h in 4 μl ml⁻¹ Triton X-100 (4 %) before incubation for a further hour in fluorescein-streptavidin (Vector Laboratories; 30 μg ml⁻¹ in 0.1 M PBS). The slices were mounted in VectaShield mounting medium (Vector Laboratories) and viewed under a confocal microscope (LEICA TCS SP; Leica Microsystems UK Ltd, Milton Keynes, UK).

Recording procedures

All recordings were made using an Axopatch 200A amplifier (Axon Instruments, Foster City, CA, USA). Cerebellar cell types were identified visually with Nomarski differential interference contrast (DIC) optics (× 40 water-immersion objective, total magnification × 320–1000). Patch-pipettes were made from thick-walled borosilicate glass capillary tubing (GC-150F; Clark Electromedical, Pangbourne, UK) on a two-stage puller. Pipettes were coated with Sylgard (Dow Corning 184) and fire polished. During whole-cell recording from Golgi cells, series resistance compensation (60–70 %, typically 7 μs lag time) was employed. In some experiments, drug application was performed by pressure ejection from theta-glass pipettes (tip diameter ∼5–8 μm; TGC150–10; Clark Electromedical) positioned at least 40 μm from the recorded cell. Afferent inputs (parallel fibres) were stimulated via a glass pipette that contained external solution or 1 M NaCl, using a 10–30 V pulse of 20–40 μs duration delivered at 0.2 Hz (Digitimer D4030 and Neurolog DS2 isolator; Digitimer Ltd, Welwyn Garden City, UK).

Data analysis

Current records were stored, for subsequent computer analysis, on digital audio tape (DTR-1204; BioLogic, Claix, France; DC to 20 kHz) with the amplifier filter (4-pole Bessel type) set at 10 kHz. Single-channel currents were replayed from tape, filtered at 2 kHz and digitised at 10 kHz (pCLAMP 6.1, Axotape; Axon Instruments). To determine the slope conductance and reversal potential of single-channel currents, recordings were made at potentials between −80 and +20 mV and histograms constructed (pCLAMP 6.1, Fetchan). The mean single-channel current, determined from Gaussian fits to these amplitude distributions, was plotted against voltage and the data fitted by linear regression. Channel openings to levels other than the main conductance were identified by eye and fitted with cursors.

To examine transitions between the various conductance states in more detail, single-channel currents were also analysed using the method of time-course fitting (Colquhoun & Sigworth, 1995; EKDIST; http://www.ucl.ac.uk/Pharmacology/dc.html). Currents were replayed from tape, filtered at 2 kHz and digitised at 10 kHz (CED 1401+ interface; Cambridge Electronic Design, Cambridge, UK). Individual openings were fitted with the step response function of the recording system; only those openings that were longer than two filter rise times (i.e. reaching 98.8 % of their full amplitude) were included in the distributions. The mean amplitude levels of single-channel currents were determined from fits of Gaussian distributions to the cursor fitted amplitudes. To examine the frequency of direct transitions (with durations longer than 2 filter rise times), conductance ranges were visually determined from the corresponding amplitude histogram. In asymmetry plots, each occurrence of a transition between two states (represented as the i th conductance level followed by the (i+ 1)th conductance level, where i is an integer) was shown as a single dot when displaying the frequency of the various transition types. The effect of drugs on single-channel activity was determined by integrating 100 ms epochs of channel activity to yield an estimate of mean charge transfer over a 30 s period (Momiyama et al. 1996).

Excitatory postsynaptic currents (EPSCs) were filtered at 2 kHz and digitised at 10 kHz. EPSCs were identified by eye from the digitised records and were analysed using ‘N’ v. 1.0 software (written by Stephen Traynelis; Emory University, Atlanta, GA, USA). Average waveforms were constructed by aligning EPSCs on their initial rising phase; current decays were fitted using either ‘N’ or Origin 4.10 (Microcal, Northampton MA, USA). The decay of both spontaneous and evoked synaptic currents could be described by the sum of two exponential functions (representing the non-NMDAR- and NMDAR-mediated components). The effect of drugs on the amplitudes of the fast and slow components of the EPSCs was examined by determining the average of three points at the peak of the EPSC, and the average of 10 points, 20 ms after the peak (i.e. when the fast-component had largely decayed).

Ifenprodil inhibition curves were fitted with a form of the Hill equation:

where I_max is the maximal inhibition of the response, [A] is the concentration of ifenprodil, IC₅₀ is the concentration of ifenprodil required to reduce the response to 50 % of control, and n is the Hill slope. All values are expressed as means ±s.e.m. and differences between groups were tested using either Student's paired or Student's unpaired two-tailed t test and considered significant at the P < 0.05 level.

RESULTS

Identification of cerebellar Golgi cells

Golgi cells were identified using a combination of morphological and electrophysiological criteria. Initially they were distinguished from the more numerous granule cells present in the internal granular cell layer by their large soma, which was readily apparent under DIC optics. A more thorough morphological identification was based on confocal reconstructions of a subset of cells filled with Neurobiotin during electrophysiological examination (see Methods). Golgi cells had a large soma (diameter 26.0 ± 2.6 μm; n = 11), with at least two sagitally orientated ascending dendrites, which arborised within the molecular layer (see Fig. 1). Other dendrites originated near the base of the soma and remained within the internal granule cell layer. All Golgi cell dendrites lacked spines, although these could be readily resolved in Purkinje cells in the same slice that had been filled with Neurobiotin (data not shown). Axonal projections of Golgi cells were often severed within the slice but when present arborised exclusively within the internal granule cell layer. This axonal plexus is indicated by the arrows in Fig. 1. Our observations are in good general agreement with previous descriptions of cerebellar Golgi cells (Hamori & Szentagothai, 1966; Palay & Chan-Palay, 1974; Dieudonné, 1998).

Figure 1. Morphology of a P14 Golgi cell.

The figure shows a confocal reconstruction of a Golgi cell in a sagittal slice. The cell was filled with Neurobiotin and processed using the streptavadin-fluorescein technique. The position of the cell is typical of the Golgi cells recorded in this study. Note the ascending dendrites, which enter the molecular layer, and the fine calibre axonal plexus (demarcated by arrows) within the granule cell layer.

Various functional characteristics were also used to identify Golgi cells. In the cell-attached configuration the cells exhibited regular spiking (mean frequency 6.4 ± 0.48 Hz, n = 52). This contrasted with the cerebellar granule cells, which were electrically silent in cell-attached recordings. In the whole-cell configuration Golgi cells had a capacitance of 17.4 ± 1.3 pF (n = 30), which was invariably larger than that of the granule cells (1.92 ± 0.1 pF; n = 31). Golgi cells exhibited spontaneous EPSCs and IPSCs (sEPSCs and sIPSCs). As expected, the sIPSCs were abolished in the presence of bicuculline and strychnine (Dieudonné, 1995, 1998).

Pharmacological properties of whole-cell NMDA-responses in Golgi cells

Pressure application of 10 μM NMDA (together with 10 μM glycine), from a pipette positioned close to the cell soma, elicited a mean inward current of −120 ± 18.3 pA (V_m = −60 mV; n = 14 cells). To investigate the subunit composition of the receptors mediating this response, we examined the effect of various drugs which show selectivity for particular NMDAR subunits. Ifenprodil acts as an atypical non-competitive NMDAR antagonist with an apparent affinity for NR1/NR2B receptors that is ∼400-fold higher than for NR1/NR2A receptors (Williams, 1993). Furthermore it has little effect on recombinant NR1/NR2C or NR1/NR2D receptors (Williams, 1995; Whittemore et al. 1997). However, its effects on native receptors containing these subunits has not been tested. As shown in Fig. 2, 10 μM ifenprodil caused a 78.9 ± 2.7 % (n = 8) inhibition of the inward current. This degree of block is slightly less than the maximal block produced by ifenprodil on recombinant NR1–1a/NR2B receptors (∼90 %; Mott et al. 1998; Masuko et al. 1999), suggesting that the residual current might reflect the presence of both unblocked NR1/NR2B receptors and ifenprodil-insensitive receptors.

Figure 2. Pharmacology of whole-cell NMDA-evoked currents in Golgi cells.

A and B show whole-cell recordings (−60 mV) from two different Golgi cells. Application of 10 μM NMDA and 10 μM glycine (filled bars) produces an inward current, which in A is blocked by co-application of 10 μM ifenprodil (open bar). In B, co-application of 1 μM TPEN (open bar) produces little change in the inward current. C, shows a similar experiment performed on a cerebellar granule cell in the internal granule cell layer at P6; in this case TPEN causes a substantial potentiation of the NMDA-evoked current. D, summary of results with TPEN showing a significant potentiation of currents in granule cells (P < 0.05) but no effect in Golgi cells (P > 0.05).

To determine the possible contribution of NR2A-containing receptors to NMDA-evoked responses in Golgi cells, we examined the effect of the Zn²⁺ chelator TPEN. This has previously been shown to enhance NMDA-evoked responses from recombinant NR1/NR2A receptors, whilst having little effect on NR1/NR2B receptors (Paoletti et al. 1997). This effect is due to the chelation of contaminating Zn²⁺, which tonically suppresses the response of NMDARs containing the NR2A subunit (Paoletti et al. 1997). As shown in Fig. 2B, the magnitude of the response to NMDA was unaffected by the co-application of 1 μM TPEN (−4.9 ± 3.6 %, n = 6; P = 0.88). To confirm the action of TPEN under our recording conditions, we examined its effect on NMDA-evoked currents in internal granule cells at P6 (a stage at which these cells contain mRNA for NR2A and NR2B subunits; Akazawa et al. 1994; Watanabe et al. 1994; Monyer et al. 1994). As expected, there was a marked potentiation of the NMDA response in these cells (Fig. 2C and D; 38.0 ± 0.1 %, n = 15; P = 0.02).

These data suggest that at P14 a high proportion of the functional NMDARs expressed by Golgi cells contain NR2B subunits, with little if any expression of NR2A-containing NMDARs. As whole-cell responses reflect the activation of both synaptic and extrasynaptic NMDARs (see, for example, Tovar & Westbrook, 1999), we next examined the properties of extrasynaptic NMDARs in isolation, by recording single-channel currents in outside-out patches from the cell soma.

Extrasynaptic NMDA receptor channels in Golgi cells

Application of 10 μM NMDA (in the presence of 10 μM glycine, strychnine, bicuculline and CNQX) gave rise to single-channel activity in all outside-out patches examined (n = 50 patches; Misra et al. 1998). As illustrated in Fig. 3A, we observed openings to three different conductance levels. The slope conductance of the most frequently visited state (level 3, Fig. 3A) was approximately 50 pS (49.7 ± 0.9 pS; n = 18 patches) with a mean reversal potential of −1.9 ± 1.0 mV (n = 7; Fig. 3B). These values were obtained from all-point amplitude histograms examined over a range of membrane potentials (−80 to +20 mV). The 50 pS openings were present in all patches and were associated with a ∼40 pS sub-conductance state with clear transitions between levels; 50/40 pS events of this type are characteristic of ‘high-conductance’ native NR1/NR2A and NR1/NR2B NMDARs (see Cull-Candy et al. 1995).

Figure 3. Single-channel properties of Golgi cell NMDARs.

A, single-channel records from an outside-out patch recording at −70 mV from a P14 Golgi cell, in response to 10 μM NMDA and 10 μM glycine. The three conductance states (1, 2 and 3) and the closed state (C) are indicated by dashed lines. The filled circles indicate transitions between the main-conductance and sub-conductance of the high-conductance openings and the open circles indicate the corresponding transitions for the low-conductance openings. B, current-voltage relationship obtained from all-point histograms for the high-conductance channels shown in A. Linear regression analysis in this cell provided an estimated slope-conductance of 55 pS and a reversal potential of −1.5 mV. C, amplitude distribution of NMDA-evoked single-channel events from a patch held at −70 mV. Amplitudes were obtained from time-course fitting analysis and the histogram was fitted with the sum of three Gaussians, giving conductance estimates of 18.7 pS (7.2 %), 37.3 pS (39.3 %) and 49.9 pS (53.5 %). D, amplitude stability plot for a single outside-out patch held at −70 mV. Three bands of events are detectable corresponding to different conductance states shown in C.

In addition to these events, ‘low-conductance’ channel openings were observed in ∼50 % of patches. Although too infrequent to be identified in the all-point amplitude histograms, these could be measured by visually placed amplitude cursors. Taking the mean reversal potential of the high-conductance events, we estimated these low-conductance channels had a mean chord conductance of 39.4 ± 1.4 pS (level 2 in Fig. 3A). Furthermore, there were frequent direct transitions between this level and a lower 19.4 ± 0.9 pS (n = 9) conductance state (level 1 in Fig. 3A). Transitions occurred between levels 1 and 2 (19 and 39 pS), and between levels 2 and 3 (40 and 50 pS), but not between levels 1 and 3 (19 and 50 pS). This is consistent with the idea that ‘high-’ and ‘low-conductance’ events arise from separate channels. For clarity we will usually refer to low-conductance events as ‘39/19 pS’ openings, and the high-conductance events as ‘50/40 pS’ openings. We did not observe any patches that contained only low-conductance openings.

Certain low-conductance NMDARs exhibit asymmetry in their sequence of transitions between the main- and sub-conductance states (Cull-Candy & Usowicz, 1987). This behaviour is known to be highly characteristic of native (Momiyama et al. 1996; Cull-Candy et al. 1998) and recombinant (Wyllie et al. 1996) NR2D-containing NMDARs. To obtain further information about transitions between the main- and sub-conductance states of NMDAR channels in Golgi cells, single-channel currents were analysed in more detail, using time-course fitting. An amplitude histogram showing the three different conductance states identified in this way is illustrated in Fig. 3C. The conductance levels were 18.8 ± 0.8, 38.3 ± 1.5 and 49.2 ± 2.4 pS (n = 3 patches). As shown in Fig. 3D, the amplitudes of these openings remained stable during individual recordings. The lower two levels match closely previous conductance measurements (Momiyama et al. 1996) from native NR2D-containing receptors.

Figure 4A shows examples of transitions between the 50 and 40 pS states of the high-conductance channel. We found no evidence for asymmetry of opening in three cells where such transitions were examined in detail. Thus, transitions originating in the 40 pS level and progressing to the 50 pS level were as prevalent (51.8 ± 1.4 %) as those from the 50 to 40 pS state (48.3 ± 1.4 %). On the other hand, the low-conductance openings displayed clear asymmetry of transitions. In all patches, transitions from the 39 to 19 pS state (level 2 → level 1 in Fig. 4B) were more prevalent (63.7 ± 4.1 %) than transitions from the 19 to 39 pS level (36.3 ± 4.1 %). The ‘asymmetry plot’ in Fig. 4C illustrates this characteristic imbalance of transitions in a single patch. Here, as in all patches, the main-conductance of the low-conductance channels and the sub-conductance of the high-conductance channels cannot be separated. As both are approximately 40 pS they were grouped together as level 2. The number of points scattered in the level 2 → 1 quadrant clearly exceeds that in the level 1 → 2 quadrant, while transitions between level 2 and 3 (i.e. 2 → 3 and 3 → 2) exhibited no apparent differences in number. The majority of ‘transitions’ occurred between the closed state and main open states (C ⇌ 3 or C ⇌ 2). The absence of clear transitions between the 19 pS and the 50 pS state (1 ⇌ 3) is consistent with the idea that the single-channel events arise from at least two distinct receptor populations: a high-conductance (NR2A- or NR2B-containing) channel; and a low-conductance channel, which has properties typical of NR2D-containing receptors. This situation resembles that previously described for certain other cells types (Momiyama et al. 1996; Cull-Candy et al. 1998).

Figure 4. The asymmetry of low-conductance NMDAR openings.

A, high-conductance openings from an outside-out patch at −70 mV. A small section of the recording is shown on an expanded time scale (lower trace). The dashed lines indicate different conductance states (2 and 3) and the closed state (C; numbering as in Fig. 3A). Clear transitions between the main (3) and sub-conductance (2) of the high-conductance openings are visible. Transitions from level 3 to 2 (50 → 40 pS) were as prevalent as those from 2 to 3. B, low-conductance openings from a different outside-out patch at −70 mV. Clear transitions between the main (2) and sub-conductance (1) of the low-conductance openings are clearly visible. Transitions from level 2 to 1 were more prevalent than those from level 1 to 2 (19 → 39 pS). C, scatter plot illustrating the frequency of occurrence of transitions between two consecutive conductance states (excluding the closed state). Dots above the diagonal continuous line represent transitions in the opening direction: 2 → 3 and 1 → 2. Dots below the line represent transitions in the closing direction: 3 → 2 and 2 → 1. There is scatter associated with each of the conductance levels. The asymmetry associated with the low-conductance channels is evident: 2 → 1 transitions (steps from 39 to 19 pS) greatly exceed 1 → 2 transitions (19 → 39 pS). This contrasts with the symmetry observed in transitions associated with the high-conductance channels.

Ifenprodil sensitivity of extrasynaptic NMDARs in Golgi cells

To characterise further the subunit composition of the extrasynaptic NMDARs we examined the effect of ifenprodil on single-channel activity. As shown in Fig. 5A, high-conductance channel openings became much briefer in the presence of 10 μM ifenprodil. The reduction in charge transfer through NMDA-activated channels was determined for a range of ifenprodil concentrations, yielding an IC₅₀ of 19 nM and a maximal inhibition of 71 % (Fig. 5B). On the other hand, low-conductance openings remained detectable in the presence of ifenprodil, as illustrated in the three-dimensional asymmetry plot in Fig. 5C. This is consistent with the idea that the low-conductance events arose from ifenprodil-insensitive receptors. However, it was not possible to quantify the contribution that these receptors made to the residual current seen in Fig. 5B. We therefore examined the effect of ifenprodil on NMDARs in two other well-characterised cell types – migrating granule cells and Purkinje cells.

Figure 5. Ifenprodil sensitivity of high-conductance NMDAR openings.

A, single-channel records from an outside-out patch from a P14 Golgi cell at −70 mV. Channel activity was evoked in response to 10 μM NMDA and 10 μM glycine. In the presence of 10 μM ifenprodil (lower trace), single-channel openings became briefer. B, concentration-inhibition relationship for the action of ifenprodil on single-channel charge transfer of the type shown in A (4–6 patches). The relationship was fitted with the Hill equation giving an IC₅₀ of 19 nM and a maximal inhibition of 71 %. C, three-dimensional plots of transitions between conductance states in a single outside-out patch in the presence and absence of 10 μM ifenprodil (compare with Fig. 4C). In the presence of ifenprodil the high-conductance peaks are lost (background), leaving only the asymmetric transitions of the low-conductance channels (foreground).

Ifenprodil blocks high-conductance NMDARs in migrating granule cells

Migrating granule cells possess mRNA for the NR1 and NR2B subunits (Akazawa et al. 1994). Furthermore, they exhibit high-conductance NMDAR channel openings (Farrant et al. 1994). As there is no evidence for the presence of any other NR2 subunit in migrating granule cells at this stage, it is likely that these cells express a ‘pure’ population of NR1/NR2B receptors. Recordings were made from outside-out patches from migrating granule cells in cerebellar slices from P10 rats. Application of 10 μM NMDA (and 10 μM glycine) resulted in 50/40 pS channel activity, as shown in the control record in Fig. 6A. This activity was significantly reduced, but not completely blocked, in the presence of 10 μM ifenprodil (lower trace, Fig. 6A).

Figure 6. Ifenprodil sensitivity of native NR2B- and NR2D-containing receptors.

A, single-channel record from a P10 migrating granule cell in response to 10 μM NMDA and 10 μM glycine. All channel openings were to the high-conductance (50 pS) level (note occasional presence of doubles). The lower trace corresponds to recordings from the same cell in the presence of 10 μM ifenprodil. B, concentration-inhibition relationship for the action of ifenprodil. Charge transfer measurements were made in 5 patches and the resulting data were fitted with the Hill equation yielding an IC₅₀ of 26 nM and a maximal inhibition of 84 %. C, outside-out patch recordings from a P6 Purkinje cell in response to 10 μM NMDA and 10 μM glycine (−60 mV). All openings are of the low-conductance type. Transitions between the conductance states are evident. Ifenprodil (lower trace) had no effect on either the frequency or conductance of openings. D, pooled data from 5 Purkinje cells showing the lack of effect of ifenprodil.

From the inhibition curve illustrated in Fig. 6B, the NMDA response was reduced by ifenprodil in a concentration-dependent manner, giving an IC₅₀ of 26 nM and a maximal inhibition of 85 %. Thus, as expected, the degree of maximal inhibition produced by ifenprodil differed between granule and Golgi cells (85 vs. 71 %). This would suggest that at least some of the residual response in Golgi cell soma could be arising from ifenprodil-insensitive receptors.

NR2D-containing receptors in Purkinje cells are not blocked by ifenprodil

We examined the effect of ifenprodil on native NR1/NR2D receptors in Purkinje cells. These cells possess mRNA for the NR2D subunit during their first two postnatal weeks (along with NR1; Akazawa et al. 1994), and express functional low-conductance NMDARs as illustrated in Fig. 6C (see also Momiyama et al. 1996). From a cursory examination, NR2D-containing receptor channels appeared to be insensitive to 10 μM ifenprodil (lower trace, Fig. 6C), as previously described for recombinant NR2D receptors (Williams, 1995). This was confirmed by analysis of the charge transfer through NMDAR channels, which revealed no significant reduction in the presence of ifenprodil (Fig. 6D; −14.2 ± 10 %, n = 5;P = 0.25).

Taken together, these results suggest that Golgi cells express a mixed population of NMDARs in their somatic (extrasynaptic) membrane. The properties of these channels are consistent with the presence of NR1/NR2B and NR1/NR2D subunit assemblies. To investigate the subunit composition of the synaptic NMDARs, we next examined spontaneous and evoked EPSCs in these cells.

NMDAR-mediated component of EPSCs in Golgi cells

sEPSCs were observed at frequencies ranging from 0.1 to 0.68 Hz (mean 0.2 ± 0.04 Hz, n = 9 cells). Figure 7 shows sEPSCs recorded at −30 mV to minimise the effect of residual Mg²⁺ on the NMDAR-mediated component. At this potential, the mean peak amplitude of sEPSCs was −35.5 ± 6.4 pA (n = 9; range −3 to −161 pA). Their 10–90 % rise time was 0.8 ± 0.1 ms, and the decay phase could be described by the sum of two exponentials (Fig. 7B) with average time constants of 1.6 ± 0.2 ms (τ_fast) and 64.2 ± 8.3 ms (τ_slow) (n = 9 cells). The slow component contributed 18 ± 3 % to the peak amplitude. This was mediated by NMDARs (Dieudonné, 1998; Misra et al. 1999a), since 10 μM 2-amino-5-phosphonopentanoic acid (AP5) inhibited the current (measured 20 ms after the initial peak; see Fig. 7B inset) by 91.4 ± 3.3 %, (n = 7; P < 0.001). The fast component could be completely blocked by 5 μM CNQX (data not shown).

Figure 7. Properties of spontaneous EPSCs in Golgi cells.

A, whole-cell recording of sEPSCs recorded in a P14 Golgi cell held at −30 mV, in the presence of bicuculline and strychnine. Each downward deflection represents an individual sEPSC. Note the variable amplitude of sEPSCs. B, average waveform obtained by aligning 122 individual events on their rising phase. The decay of the waveform was best fitted by the sum of two exponentials: τ₁ of 2.2 ms (82.7 %) and τ₂ of 107.3 ms. The inset shows the initial portion of the average waveform indicating the points at which the amplitude of the AMPA and NMDA currents were determined (dashed line represents single exponential with τ of 2.2 ms). C, amplitude histogram for sEPSCs recorded from a single cell. In this example, the peak amplitudes ranged from −3 to −69 pA. The right hand panel shows a plot of the 10–90 % rise time against peak amplitude for sEPSCs from the same cell. All rise times are below 1 ms, with a mean of 0.5 ± 0.2 ms.

It is likely that most sEPSCs arose from parallel fibre inputs, which form the majority of excitatory synaptic contacts onto Golgi cells (see Dieudonné, 1998). However, we could not exclude the possibility that some may arise from mossy fibre or climbing fibre inputs. To distinguish between different Golgi cell afferents we used selective extracellular stimulation. As illustrated in Fig. 8A, stimulation in the internal granular layer resulted in a complex response which may reflect direct activation of climbing and/or mossy fibre inputs, or indirect activation of granule cells. Such responses were not amenable to analysis. Thus, to allow us to examine a population of monosynaptic NMDAR-mediated EPSCs, we stimulated parallel fibre inputs within the molecular layer of coronal slices. Examples of individual events obtained in this manner are shown in Fig. 8B (inset). Figure 8B shows the distribution of evoked EPSC amplitudes from a single cell.

Figure 8. Evoked EPSCs in Golgi cells.

A, individual evoked EPSC from a Golgi cell held at −30 mV, obtained by stimulating in the internal granule layer (activating putative climbing and mossy fibres). Note the complex nature of the response. B, amplitude histogram for parallel fibre-evoked events from a different cell. The black histogram represents superimposition of noise data. The noise histogram and a subset of events in the amplitude histogram overlap, indicating the proportion of events contributing to failures. Inset shows examples of parallel fibre-evoked EPSCs from a single cell at −30 mV. The top trace depicts a failure and the lower two traces show examples of evoked EPSCs of different amplitudes. It was possible therefore to identify failures both visually and from analysis of amplitudes. C, normalised average parallel fibre-evoked EPSC from a single cell. Note similarity in deactivation kinetics to sEPSCs (Fig. 7B). Inset shows scatter plot of rise times against amplitude from the same cell. There was no significant correlation. D, parallel fibre-evoked response to trains of stimuli (5 shocks at 50 Hz), during control conditions and in the presence of 10 μM AP5.

As shown in Fig. 8C, the deactivation kinetics of the parallel fibre-evoked response was similar to that of sEPSCs (see also Table 1). Furthermore, the slow component of the EPSC was significantly reduced in the presence of 10 μM AP5 (Fig. 9A and Table 1), and the fast component was blocked by CNQX (data not shown). We also examined synaptic currents evoked with short trains of stimuli (5 shocks at 50 Hz; n = 5 cells). This protocol evoked an NMDAR-mediated EPSC that decayed with a time constant similar to that produced by a single stimulus (Fig. 8D).

Table 1.

Summary of NMDAR-mediated EPSCs

	Spontaneous	Evoked
Control
τslow(ms)	64.2 ± 8.3	126.7 ± 25.9
Effect of ifenprodil
Percentage inhibition†	57.2 ± 8.0*	63.7 ± 4.4*
Residual τslow(ms)	57.9 ± 8.1	88.1 ± 38.9
Effect of AP5
Percentage inhibition†	91.4 ± 3.3*	91.5 ± 3.5*
Effect of TPEN
Percentage potentiation†	13.5 ± 10.7	0.8 ± 16.2

^†Percentage inhibition/potentiation of peak NMDA response at 20 ms.

^*Significant effect (P < 0.05).

Figure 9. The NMDAR-mediated component of EPSCs.

A, normalized average parallel fibre-evoked EPSCs recorded from a single cell held at −30 mV. The averages are from 97, 73 and 89 individual evoked EPSCs for control, ifenprodil and AP5, respectively. The inset shows the NMDAR-mediated component of the evoked EPSCs in isolation following subtraction of the AP5 average waveform. B, summarised results for evoked EPSCs (upper panel, n = 5) and spontaneous EPSCs (lower panel, n = 7) showing the conductance of the NMDAR-mediated component (measured 20 ms after the peak of the EPSC). * P < 0.05.

For both sEPSCs and parallel fibre EPSCs, the NMDAR-mediated component was substantially reduced by 10 μM ifenprodil (−57.2 ± 8.0 %, n = 7 and −63.7 ± 4.4 %, n = 5, respectively; Fig. 9). Synaptic currents evoked by short trains of stimuli were also reduced, by a similar extent, by ifenprodil (n = 3; data not shown). In contrast, the Zn²⁺ chelator TPEN (1 μM) had no apparent effect on sEPSCs or parallel fibre EPSCs (n = 5 in each case; Fig. 10 and Table 1), consistent with an absence of synaptic NR1/NR2A receptors.

Figure 10. TPEN has no effect on spontaneous and evoked EPSCs.

A, normalised average waveforms from parallel fibre-evoked EPSCs in the presence and absence of 1 μM TPEN. B, traces show normalised average waveforms from sEPSCs in the presence and absence of TPEN and were constructed from 217 and 66 events, respectively. The insets to the two average waveforms in B indicates stability plots for sEPSCs in the presence and absence of TPEN. C, graphs of mean effect of TPEN on the NMDAR-mediated component (lower panel) and the non-NMDAR-mediated component (upper panel) of the evoked EPSCs (filled bars) and spontaneous EPSCs (open bars). Average results from 5 cells are indicated and there was no significant difference in conductance for either the NMDAR-mediated or non-NMDAR-mediated component calculated from evoked or spontaneous EPSCs.

The decay kinetics of the NMDAR-mediated component was analysed after subtraction of the initial non-NMDAR-mediated component. For both spontaneous and evoked events the magnitude of this slow component was reduced in ifenprodil (Fig. 9A, inset), but its decay time was unaffected (Table 1). It has previously been demonstrated that recombinant NR1/NR2D receptors exhibit deactivation times (following rapid agonist removal) that are in the time scale of seconds (Monyer et al. 1994; Vicini et al. 1998; Wyllie et al. 1998). Furthermore, recent experiments have shown that native NR2D-containing receptors in Purkinje cells also exhibit unusually slow deactivation times that are an order of magnitude slower than the Golgi cell ‘residual’ NMDAR-mediated EPSC (Misra et al. 1999b). Thus, the relatively rapid decay of the NMDAR-mediated component of EPSCs seen in ifenprodil suggests that NR1/NR2D receptors do not contribute to the synaptic current in Golgi cells.

No change in the time course, or frequency, of sEPSCs was observed in the presence of TPEN (see Fig. 10). Moreover, the failure rate for evoked EPSCs was not changed by AP5, ifenprodil or TPEN, suggesting that the pharmacological agents used in this study did not have marked presynaptic effects.

DISCUSSION

Our experiments have provided two main findings. First, patches from cerebellar Golgi cells exhibit mixed populations of high- and low-conductance NMDAR channels. The properties of the low-conductance openings are consistent with the presence of NR2D-containing NMDARs, as suggested from the in situ hybridization data (Akazawa et al. 1994). On the other hand, the high-conductance channels have features typical of NR2B-containing receptors, the presence of which was unexpected in Golgi cells. Second, the NMDAR-mediated component of the EPSC at the parallel fibre to Golgi cell synapse is carried by NR2B-containing receptors. We found no evidence that other types of NMDARs were activated during transmission in these cells. Our results would therefore suggest that NR2D-containing receptors are restricted to extrasynaptic sites, while NR2B-containing receptors are present in both synaptic and extrasynaptic membrane. We will consider these various points below.

Comparison with in situ hybridization data

In situ hybridization experiments have demonstrated that Golgi cells contain mRNA for the NR2D subunit, along with isoforms of NR1 (Akazawa et al. 1994). Our results have demonstrated the presence of functional channels with properties characteristic of both native NR2D-containing (see Momiyama et al. 1996) and NR2B-containing receptors. While the low-conductance channel openings were readily detected, they were less prevalent than high-conductance openings. Furthermore, they were not expressed in all patches. Thus, despite the lack of earlier evidence for NR2B-containing receptors in these cells, they appear to constitute the majority of functional NMDAR channels. This apparent contradiction may result from the difficulties associated with accurately interpreting in situ hybridization data in cells that occur at low density or are irregularly distributed.

Pharmacological identification of NR2B-containing NMDARs

To assess the contribution of NR2B-containing receptors to macroscopic NMDA-evoked currents and synaptic currents in Golgi cells we have made use of the fact that ifenprodil selectively blocks recombinant NMDARs of the NR1/NR2B subtype (see Williams, 1993; Priestley et al. 1995; Stocca & Vicini, 1998), while NR2A-containing receptors are potentiated with the Zn²⁺ chelator TPEN (Paoletti et al. 1997). At the concentration used here (10 μM) ifenprodil has little effect on recombinant assemblies composed of NR1/NR2A, NR1/NR2C or NR1/NR2D subunits (Williams, 1993, 1995; Whittemore et al. 1997). However, its action has been less thoroughly examined on native channels. Since our experiments suggested that, along with the NR2B-containing receptors, NR2D-containing receptors were also present in Golgi cells, it was necessary to verify that ifenprodil had little effect on the NMDARs present in Purkinje cells, which express a pure population of NR1/NR2D receptors (Momiyama et al. 1996). We also verified that, under the conditions of our experiments, TPEN potentiates responses in granule cells that express NR2A-containing receptors. Hence, the pharmacological properties of the NMDA response in Golgi cells are consistent with it being carried largely by NR2B-containing receptors.

Several studies have reported that ifenprodil produces a maximal inhibition of ∼90 %, with an IC₅₀ in the order of a few hundred nanomolar, when tested against NMDA- or glutamate-evoked responses from recombinant NR1-1a/NR2B receptors (Williams, 1993; Whittemore et al. 1997; Mott et al. 1998; Masuko et al. 1999). The IC₅₀ values that we have obtained from charge transfer measurements in Golgi cells and granule cells (19 and 26 nM, respectively) were significantly lower than previous estimates from recombinant receptors. The reason for this difference is unclear, although it may be that native NMDARs show a greater sensitivity to block by ifenprodil. Consistent with this idea, Kirson & Yaari (1996) have obtained IC₅₀ values in the nanomolar range (10 to 29 nM), for the action of ifenprodil on NMDAR-mediated EPSCs in mouse hippocampal neurones.

Multiple NR2 subunit-types within single cells

We have distinguished between individual NR2D- and NR2C-containing receptor channels on the basis of the characteristic pattern of transitions between conductance states of the former (Momiyama et al. 1996; Wyllie et al. 1996). These channels tend to open via the 19 pS level and close via the 39 pS level. It is worth noting that the degree of asymmetry that we obtained for low-conductance channels in Golgi cells matched closely that previously reported for NR2D channels in Purkinje cells (Momiyama et al. 1996). This would suggest that any contribution from NR2C receptors is extremely small.

While the combined approach of single-channel analysis and examination of pharmacological properties has proved useful in identifying NMDAR types, there are some necessary points of caution. There seems little doubt, from in situ hybridization data and patch-clamp studies, that many central neurones express at least two types of NMDAR. Furthermore, there is good evidence to suggest that some native receptors can contain more than one type of NR2 subunit (Wafford et al. 1993; Sheng et al. 1994; Chazot et al. 1994). Although recent experiments on recombinant trimeric receptors (NR1/NR2A/NR2D) have shown that these can exhibit distinct single-channel properties (Cheffings & Colquhoun, 1999), the proportion of such receptors is thought to be low compared with assemblies containing one type of NR2 (see Chazot & Stephenson, 1997). Our results would suggest that if such assemblies are present in Golgi cells they would be composed of NR1/NR2B/NR2D subunits. The fact that the present results, and previous studies on cells expressing mixed populations of channel conductances (Farrant et al. 1994; Momiyama et al. 1996) have found no evidence for channels with ‘intermediate’ properties suggests that these occur at low density or are functionally indistinguishable from one of the dimeric assemblies. Thus, the contribution of such trimeric receptor channels to the functional population in neurones remains uncertain.

NR2B-containing receptors are present at parallel fibre synapses

NR2B subunit-selective drugs have been shown previously to block synaptic NMDARs in developing hippocampal neurons (Kirson & Yaari, 1996), cortical neurons (Stocca & Vicini, 1998) and medial septal neurons (Plant et al. 1997). However, the extent of this block can vary. In hippocampal neurons the degree of ifenprodil block changes during development, consistent with a changing contribution of NR2B-containing NMDARs (Kirson & Yaari, 1996). Also, in cortical neurons the degree of block varies between synaptic and extrasynaptic receptors, possibly suggesting a differential distribution of NR2B subunits. In a recent study on rat cultured hippocampal neurons (Tovar & Westbrook, 1999) ifenprodil sensitivity was used to assess the subunit composition of synaptic and extrasynaptic NMDARs. A high ifenprodil sensitivity was observed only for extrasynaptic NMDARs, and for those NMDARs present at ‘immature’ synapses. Lower ifenprodil sensitivity was observed at ‘mature’ synapses. The authors have interpreted this in terms of the developmental appearance of NMDARs with a distinct heterotrimeric subunit composition.

In the present study, the NMDAR-mediated component of spontaneous and evoked EPSCs in Golgi cells was substantially reduced by ifenprodil (>60 %). Although we cannot rule out the presence of trimeric receptors with novel pharmacological properties, the most parsimonious interpretation of this result is that NR2B-containing receptors predominate at this synapse. In which case, the residual current observed in ifenprodil is likely to reflect incomplete block. The fact that the Zn²⁺ chelator TPEN did not affect EPSCs indicates that NR1/NR2A receptors are absent. Moreover, given the distinctive kinetic properties of recombinant (Monyer et al. 1994; Vicini et al. 1998; Wyllie et al. 1998) and native (Misra et al. 1999b) NR1/NR2D receptors, it is unlikely that these are responsible for the rapidly decaying residual synaptic NMDAR-mediated component observed in ifenprodil.

It remains to be seen whether NR2D-containing NMDARs make a contribution to the EPSC at any central synapse. In Golgi cells we found no evidence of a contribution from NR2D-containing NMDARs to synaptic currents, even during trains of stimuli which may result in activation of extrasynaptic receptors by glutamate spillover. In Purkinje cells there is no evidence of NR2D-containing receptors at synaptic sites, although it is clear that such receptors are present in the soma (Momiyama et al. 1996). Dorsal horn neurons also express mRNA for the NR2D subunit (Monyer et al. 1994) and exhibit low-conductance NR2D-like channel openings (Momiyama et al. 1996). However the NMDAR-mediated component of EPSCs in these cells is rapid, consistent with a strong influence from other NR2 subunits (Bardoni et al. 1998). Thus, in the cerebellum, as elsewhere in the brain, the precise role of the NR2D subunit remains to be determined.