Colocalization of GABA and glycine in synaptic terminals of the superficial dorsal horn raises the question of their relative contribution to inhibition of different classes of neurons in this area. To address this issue, miniature IPSCs (mIPSCs) mediated via GABAA receptors (GABAARs) and glycine receptors (GlyRs) were recorded from identified laminae I-II neurons in adult rat spinal cord slices. GABAAR-mediated mIPSCs had similar amplitude and rise times, but significantly slower decay kinetics than GlyR-mediated mIPSCs. Lamina I neurons appeared to receive almost exclusively GlyR-mediated mIPSCs, even after application of hypertonic solutions. Yet, all neurons responded to exogenous applications of both GABA and glycine, indicating that they expressed both GABAARs and GlyRs. Given that virtually all glycinergic interneurons also contain GABA, the possibility was examined that GABAARs may be located extrasynaptically in lamina I neurons. A slow GABAAR-mediated component was revealed in large, but not minimally evoked monosynaptic IPSCs. Administration of the benzodiazepine flunitrazepam unmasked a GABAAR component to most mIPSCs , suggesting that both transmitters were released from the same vesicle. The isolated GABAAR component of these mIPSCs had rising kinetics 10 times slower than that of the GlyR component (or of GABAAR mIPSCs in lamina II). The slow GABAAR components were prolonged by GABA uptake blockers.
It is concluded that, whereas GABA and glycine are likely released from the same vesicle of transmitter in lamina I, GABAARs appear to be located extrasynaptically. Thus, glycine mediates most of the tonic inhibition at these synapses. This differential distribution of GABAARs and GlyRs confers distinct functional properties to inhibition mediated by these two transmitters in lamina I.
The superficial laminae I and II of the dorsal horn play a pivotal role in the integration and relay of pain-related information (Perl, 1984; Willis, 1985; Light, 1992; Craig, 1996), and thus elucidating the nature of inhibitory control in this area is crucial for our understanding of nociceptive processing. Both GABA and glycine function as inhibitory neurotransmitters in the mammalian spinal cord (for review, see Todd and Spike, 1993), and blocking either of these control mechanisms produces a hypersensitivity characteristic of neuropathic pain syndromes (Yaksh, 1989; Sivilotti and Woolf, 1994; Sherman and Loomis, 1996; Sorkin and Puig, 1996). Previous studies report the coexistence of GABA and glycine as well as their respective receptors at many synapses in the superficial dorsal horn of the rat spinal cord (van den Pol and Gorcs, 1988; Bohlhalter et al., 1994, 1996; Todd et al., 1995, 1996), and it appears that nearly all glycine-immunoreactive cells in this area are also GABA-immunoreactive (although only half of GABAergic cells contain glycine) (Todd and Sullivan, 1990; Mitchell et al., 1993). Recent evidence indicates the possible co-storage of GABA and glycine within the same vesicles at some synapses (Burger et al., 1991; Christensen and Fonnum, 1991; Chaudhry et al., 1998; Jonas et al., 1998). These findings raise the question of whether both transmitters are contained within the same synaptic vesicles, co-released, and act together at the same synaptic junction in the superficial dorsal horn, and therefore whether they play distinct roles in inhibition of laminae I and II neurons.
Previous studies focused on GABAA and glycine receptor-mediated inhibition in lamina II of the spinal dorsal horn (Yoshimura and Nishi, 1995) or trigeminal medulla (Grudt and Williams, 1994). They did not directly address the question of co-release from the same vesicles, nor whether inhibition mediated by GABA and glycine differed among identified classes of neurons, in part because they did not identify lamina I neurons. Thus, data are lacking on this layer that represents one of the main spinal nociceptive output pathways (Willis, 1989; Light, 1992; Craig, 1994).
Recordings from lamina I in slices have remained limited by the difficulty to maintain and delineate this thin layer in conventional preparations. To overcome this, we used a parasagittal slice of adult rat spinal cords that respects the natural rostrocaudal orientation of marginal layer neurons (Light et al., 1979; Woolf and Fitzgerald, 1983;Lima and Coimbra, 1986; De Koninck et al., 1992), thus allowing visual identification of lamina I (Chéry and De Koninck, 1997; De Koninck and Chéry, 1998) while still providing visual access to deeper layers. Using this approach, we performed whole-cell patch-clamp recordings in identified spinal laminae I-II neurons to study spontaneously occurring miniature (action potential-independent) IPSCs (mIPSCs) that are thought to reflect transmitter release from single vesicles.
An important initial finding was that lamina I neurons were almost exclusively bombarded by glycine receptor (GlyR)-mediated mIPSCs even though all cells appeared to express both GABAAreceptors (GABAARs) and GlyRs, suggesting that perhaps GABAARs were not located at synaptic junctions in this layer. To test this hypothesis more directly, we manipulated release, receptor sensitivity, and uptake system using electrical stimulation, benzodiazepines, and GABA uptake inhibitors. The results are consistent with the interpretation that GABAARs are likely located at extrasynaptic sites in lamina I.
MATERIALS AND METHODS
Slice preparation. Adult male Sprague Dawley rats (30- to 60-d-old) were anesthetized with Na+ pentobarbital (30 mg/kg) and perfused intracardially for 15–20 sec with ice-cold oxygenated (95% O2 and 5% CO2) sucrose-substituted ACSF (S-ACSF) containing (in mm): 252 sucrose, 2.5 KCl, 2 CaCl2, 2 MgCl2, 10 glucose, 26 NaHCO3, 1.25 NaH2PO4, and 5 kynurenic acid, pH 7.35; 340–350 mOsm. The rats were then rapidly decapitated, and the spinal cord was immediately removed by hydraulic extrusion and immersed in ice-cold S-ACSF for 1–2 min. In some cases, a laminectomy was performed before the perfusion for subsequent surgical extraction of the cord rather than by hydraulic extrusion. Lumbar and cervical segments (2-cm-long) were isolated and glued, lateral side down, on a brass platform with cyanoacrylate cement, in a chamber filled with oxygenated ice-cold S-ACSF, and 400-μm-thick parasagittal sections were cut. The slices were then incubated in S-ACSF at room temperature (23–28°C) for 30 min and subsequently transferred to a storage chamber filled with oxygenated normal ACSF (126 mm NaCl instead of sucrose, 300–310 mOsm) at room temperature. After a minimum incubation of 1 hr, the slices were transferred to a recording chamber under a Zeiss Axioscope equipped with infrared differential interference contrast (IR-DIC) and water immersion-objectives for visualization of neurons in thick live tissue. The slices were perfused at ∼2 ml/min with oxygenated ACSF containing the glutamate receptor antagonists 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX, 10 μm; Tocris Cookson, Bristol, UK), andd-2-amino-5-phosphonovaleric acid (d-AP-5; 40 μm; Tocris Cookson). All recordings were performed at room temperature.
Labeling and reconstruction of neurons. All neurons were labeled during the recordings by including Lucifer yellow (dipotassium salt, 0.5–1%; Sigma, St. Louis, MO) in the recording pipette. Simple diffusion of the dye from the pipette into the cell during the course of the recording was sufficient to obtain complete labeling. Immediately after the end of the recording, the slice was placed between wet filter paper to prevent wrinkling and fixed by immersion in 4% paraformaldehyde in 0.1 m phosphate buffer. Subsequently, the slices were cryoprotected by infiltration with 30% sucrose in 0.1 m phosphate buffer, overnight at 4°C, and processed later for confocal light microscopy. The slices were examined under a Zeiss LSM 410 inverted laser scanning microscope equipped with argon/krypton and helium/neon lasers. From 40–60 serial optical sections (1 μm apart) of the Lucifer yellow-labeled cell were obtained. With this approach, it was possible to reconstruct the entire dendritic tree and perform three-dimensional rotations in different planes for full morphological identification of each neuron.
Drug application. Bicuculline methiodide (10 μm; Research Biochemicals, Natick, MA), SR-95531 (3 μm; Research Biochemicals) and strychnine hydrochloride (100 nm to 1 μm; Research Biochemicals) were added to the ACSF from frozen, aliquoted stock solutions. For recording of action potential-independent mIPSCs, 1 μm tetrodotoxin (TTX; Research Biochemicals) was added to the bathing solution. GABA and glycine were prepared similarly and dissolved in ACSF containing 10 mm HEPES in replacement for the bicarbonate buffer, to reach a concentration of 1 mm. These amino acids were applied locally onto lamina I-II dorsal horn neurons by pressure ejection through glass micropipettes. Two pipettes, connected to a two-channel Picospritzer, were positioned close to the cells and contained GABA and glycine, respectively, except in control experiments in which one of the pipettes contained only vehicle solution. In some cases, GABA and glycine were bath-applied. In experiments aimed at forcing additional vesicle release, pressure-application of hypertonic ACSF onto the recorded neurons was used (sucrose was added to HEPES-buffered ACSF to obtain an osmolarity of 590–610 mOsm). In some experiments, to potentiate possible subliminal GABAAR-mediated events, the benzodiazepine flunitrazepam (Hoffman-La Roche, Basel, Switzerland) was bath-applied (1 μm). The GABA uptake blockers tiagabine (25–50 μm; Abbott Labs, Irving, TX) and 1-(2-(((Diphenylmethylene)imino)oxy)ethyl)-1,2,5,6-tetrahydro-3-pyridine-caboxylic acid hydrochloride (NO-711; 10–30 μm; Research Biochemicals) were also used to study accumulation of synaptically released GABA.
Whole-cell recording and data acquisition. For whole-cell voltage-clamp recordings, patch pipettes were pulled from borosilicate glass capillaries (with an inner filament, WPI) using a two-stage vertical puller (PP-83; Narishige, Tokyo, Japan). The pipettes were filled with an intracellular solution composed of (in mm): 140 CsCl, 10 HEPES, 2 MgCl2, and 0.5% Lucifer yellow (Sigma). The pH was adjusted to 7.2 with CsOH, and the osmolarity ranged from 260–280 mOsm (pipette resistance, 3 MΩ). In >80% of the recordings, we added 2 mm ATP, 0.4 mm GTP, 11 mm BAPTA (all from Sigma), and 1 mm CaCl2 to the intracellular solution. Recordings were obtained by lowering the patch electrode onto the surface of visually identified neurons in lamina I or II. Neurons with a healthy appearance presented a smooth surface, and the cell body and parts of the dendrites could be clearly seen. These neurons also had fusiform or oval cell bodies (usually 10–20 μm in length). Neurons with round, swollen cell bodies were avoided because recording from them revealed low resting membrane potential and poor membrane integrity. While monitoring current responses to 5 mV pulses, a brief suction was applied to form >1GΩ seals. An Axopatch 200B amplifier (Axon Instruments, Foster City, CA) with >80% series resistance compensation was used for the recording. The access resistance was monitored throughout each experiment. Only recordings with access resistance between 7 and 20 MΩ (average 14 ± 1 MΩ; mean ± SEM) were considered acceptable for analysis of IPSCs, and only recording with stable access throughout the entire administration of antagonists were used for classification of GABAAR- and GlyR-mediated IPSCs. Traces were low-pass filtered at 10 kHz and stored on a videotape, using a digital data recorder (VR-10B, Instrutech Corp.). Off-line, the recordings were low-pass filtered at 2–3 kHz and sampled at 10–20 kHz, on an Intel Pentium-based computer using the Strathclyde Electrophysiology software (CDR; by J. Dempster, Department of Physiology and Pharmacology, University of Strathclyde, Glasgow, UK).
Eliciting monosynaptic IPSCs and detection of spontaneous IPSCs. Monosynaptic GABAAR- and GlyR-mediated IPSCs were evoked by electrical stimulation via either bipolar tungsten electrodes for large intensity stimuli, or a patch micropipette for focal stimulations. Square-wave constant pulses (200–300 μsec duration) were applied at a frequency of ≤0.3 Hz. The electrode was placed within 20–50 μm of the neuron cell body (for proximal stimuli) or within 20–50 μm of a distal branch of a dendrite at 150–300 μm away from the soma (for distal stimulation) along the bipolar axis of the dendritic tree of lamina I neurons. Individual spontaneously occurring IPSCs (sIPSCs and mIPSCs ) were detected off-line using a software trigger as previously described (De Koninck and Mody, 1994). For each experiment, the detected events were examined; any noise that spuriously met the trigger specifications was rejected. For analysis of the decay phase of sIPSCs or mIPSCs, the events were selected on the basis of the following criteria: (1) traces containing multiple events were discarded and (2) only events that had stable baselines before the rise and after the end of the decay were kept for analysis. Rise times were determined between 10 and 90% of the peak amplitude of the IPSCs. For averaging of IPSCs, the events were software aligned by their initial rising phase. All software for analysis was designed by Y. De Koninck.
Statistical analysis and curve fitting. Peak amplitudes, rise times, and decay time constants were calculated for each of several hundreds of sIPSCs or mIPSCs per cell, using an automated algorithm (De Koninck and Mody, 1994, 1996). Averages of several hundred mIPSC were also superimposed for comparison. Decay time constants were fitted using a least square method based on a simplex algorithm as previously described (De Koninck and Mody, 1994). The goodness of fit was evaluated on the basis of fitting subsets of points drawn from the whole set of data points as well as from evaluation of the reduced χ2: χv 2 = χ2/v, where the factor ν = N − n is the number of degrees of freedom left after fitting N data points to then parameters. The necessity to introduce additional exponential components to the fits was judged first on the basis of visual inspection of the fitted curves superimposed onto the data. When the merit of additional components was not obvious, an Ftest was used to assess how the additional component improved the value of the reduced χ2:F χ = Δχ2/χv 2(df1 = 1 and df2 = ν). The critical value for the merit of additional components was set at a low level (p < 0.01) to favor parsimony of the fitted function. When focusing on comparisons of the late component of mIPSCs, fits were started at a fixed interval after the peak of the event to allow for nonequivocal monoexponential fits that provide an easier and fairer reference when dealing with nonaveraged, individual traces (De Koninck and Mody, 1994; Williams et al., 1998). This also avoided contamination of the values of decay time constants with variable weighting factors. Membrane time constants were estimated in voltage-clamp mode from analysis of current transients following 5 mV, 200-msec-long voltage pulses (Rall, 1969; Jackson, 1992; Spruston and Johnston, 1992).
Student t tests, were used to analyze the differences between parameters of the GABAAR- and GlyR-mediated IPSCs. ANOVA was used to determine the differences in rise times of proximally versus distally evoked IPSCs, and post hoc tests were obtained with Bonferroni or Tukey corrections. To evaluate the relationship between different parameters, we used the Pearson correlation matrix, and the significance of the r value was determined with an ANOVA followed by at t test using a Bonferroni correction. χ2 tests for contingency tables were performed to determine the correlation between the laminar location of the cells and the presence of GABAAR- and GlyR-mediated sIPSCs. The critical value for statistical significance was set at p < 0.05. All the data are expressed as mean ± SEM, unless otherwise indicated.
Recordings from identified laminae I and II neurons
With our slice preparation, laminae I and II were easily identified. Lamina I neurons had a distinct rostrocaudal orientation (Light et al., 1979; Woolf and Fitzgerald, 1983; Lima and Coimbra, 1986; De Koninck et al., 1992; Ma et al., 1996; Zhang et al., 1996;Zhang and Craig, 1997) and appeared densely packed (Fig.1). Most of the cells in lamina I had a bipolar morphology in the parasagittal plane, with rostrocaudal dendritic arborization (Fig. 1), but a few neurons with dendrites extending dorsally and/or ventrally could also be clearly identified. The cell bodies in lamina I were usually small (5–20 μm in length, in the longitudinal axis). Lamina II neurons, in contrast, had larger cell bodies (20–50 μm in length). Consistent with previous reports, the thickness of lamina I ranged from 15 to 50 μm (Ribeiro-da-Silva, 1995; Todd et al., 1998), being closer to 20 μm in the medial slices, whereas the more lateral sections had thicker lamina I because of the curving of the superficial layers at the lateral border of the dorsal horn (Fig. 1). The recorded lamina I neurons were at an average distance of 10 ± 2 μm from the dorsal border between the gray and white matter, indicating that these neurons clearly belonged to the marginal layer (Fig. 1). In contrast, the recorded lamina II neurons were located at an average distance of 67 ± 9 μm from the dorsal border of the gray matter to ensure a clear distinction between the sample of cells belonging to these two laminae.
Differential distribution of GABAAR- and GlyR-mediated mIPSCs in laminae I-II neurons
Spontaneous mIPSCs were observed in 322 of 387 recorded laminae I-II neurons of the dorsal horn, in the presence of 10 μmCNQX, 40 μm APV and 1 μm TTX, with a recording time varying from 2 to 90 min. To determine whether these miniature events were mediated by activation of GABAARs or GlyRs, we used the selective GABAAR antagonists bicuculline (10 μm) or SR-95531 (3 μm) and the GlyR antagonist strychnine (100 nm; Fig.2). For a comparison of the distribution of GABAAR- and GlyR-mediated mIPSCs, only data from 51 neurons (31 in lamina I and 20 in lamina II) were retained, because (1) the recordings were of sufficient duration, (2) the access resistance was stable throughout the testing of antagonists, and (3) it was possible to test the effects of both bicuculline (or SR-95531) and strychnine. Almost all lamina I neurons (30 of 31) displayed mIPSCs that could be entirely, selectively, and reversibly abolished by 100 nm strychnine, indicating that they were mediated by glycine (Table 1). In contrast, 11 of the 20 neurons in lamina II received exclusively SR-95531- or bicuculline-sensitive mIPSCs, suggesting that they were mediated via activation of GABAARs (Fig.2 B); the remaining neurons in lamina II displayed exclusively strychnine-sensitive mIPSCs, indicating their mediation by glycine (Fig. 2 A, Table 1). None of the cells tested in laminae I and II displayed both GlyR- and GABAAR-mediated mIPSCs together. In cells with GlyR-mediated mIPSCs (that were selectively blocked by 100 nm strychnine; Fig. 2 A) administration of 10 μm bicuculline or 3 μm SR-95531 did not decrease the frequency of events (Fig. 2 A) or affect their amplitude, rise times, and decay kinetics (n = 34; p > 0.2; paired comparisons; Fig. 2 C). To test for the possibility that, in these cases, bicuculline might have decreased the frequency of the events by an undetectable amount (i.e., a very small proportion of the events might have been GABAergic), in some experiments we added 100 nm strychnine first, and in all cases it completely blocked all of the mIPSCs (n= 5). Conversely, in cells with GABAergic mIPSCs, the synaptic events were selectively abolished by 10 μm bicuculline (Fig. 2 B), although up to 1 μm strychnine failed to affect these synaptic events or decrease their frequency (n = 12;p > 0.2; Fig. 2 B). These findings indicate that mIPSCs have distinct distributions in laminae I-II neurons.
Additional experiments were performed in the absence of TTX to record spontaneous IPSCs (sIPSCs; n = 9). As for mIPSCs, virtually all of these sIPSCs were also completely blocked by strychnine in lamina I. Thus, results with sIPSCs were pooled with those from mIPSCs (Table 1).
Using our recording technique, no apparent run down of either type of mIPSCs was observed, even with recordings lasting >2 hr as we previously reported in other tissues in absence of ATP or additional calcium buffers in the recording pipettes (De Koninck and Mody, 1994,1996; Otis et al., 1994). Nevertheless, because of the possibility that some GABAAR-mediated currents may be more sensitive to the lack of ATP and calcium buffering (Chen et al., 1990; however, see De Koninck and Mody, 1996), after the initial recordings, we always added 2 mm ATP, 0.4 mm GTP, 11 mm BAPTA, and 1 mm CaCl2to the intracellular solution for all subsequent recordings. We found no difference in the recordings obtained with the two different intracellular solutions; thus all the data were pooled.
Distinct kinetic properties of GABAAR- and GlyR-mediated mIPSCs
Miniature IPSCs mediated by GlyRs in lamina I had amplitude and kinetic properties that were not different from those in lamina II and therefore were pooled for comparison with properties of GABAAR-mediated mIPSCs recorded in lamina II.
A number of parameters were similar for both GlyR- and GABAAR-mediated mIPSCs. Their frequency was variable with an average of 1.0 ± 0.2 Hz (range, 0.1–5.0 Hz;n = 19) for GlyR mIPSCs, and an average frequency of 0.6 ± 0.2 Hz (range, 0.1–1.8 Hz; n = 17) for GABAAR mIPSCs. The difference in frequency was not significant between these two populations of mIPSCs (p > 0.05; Table2). In all cases, the mIPSCs amplitude distribution was skewed. The mean amplitude of GlyR-mediated events was 87.5 ± 6.1 pA (n = 30). The GABAAR-mediated mIPSCs had a mean amplitude of 72.9 ± 10.3 pA (n = 11), not significantly different from that of GlyR-mediated mIPSCs (Fig.3, Table 2). Similarly, the average 10–90% rise time for GABAAR-mediated mIPSCs was comparable to that of GlyR-mediated events (455 ± 64 μsec vs 403 ± 36 μsec, respectively; p > 0.5).
The decay time course of GABAAR-mediated mIPSCs was however significantly slower than that of GlyR-mediated mIPSCs (Fig. 3, Table 2). For quantitative comparison, the decay phase of individual mIPSCs were fit by a monoexponential function (De Koninck and Mody, 1994, 1996; Williams et al., 1998). Figures 3 Aband 3 Bb illustrate the difference in kinetics of the two populations of mIPSCs. The mean decay time constant (τD) of GlyR-mediated mIPSCs was 5.8 ± 0.3 msec versus 10.5 ± 0.6 msec for GABAAR-mediated mIPSCs (p< 0.001). The average τD at +40 mV were 19.4 ± 1.0 msec for the GlyR-mediated mIPSCs against 25.9 ± 2.4 msec for GABAAR-mediated mIPSCs , respectively (p < 0.05). Figure 3 illustrates representative distributions of the rise time, decay time constant, and amplitude of GlyR- and GABAAR-mediated mIPSCs, respectively. The average access resistances during the recordings of GABAAR- and GlyR-mediated inhibitory events were equivalent (14.0 ± 1.8 MΩ vs 13.9 ± 1.1 MΩ, respectively). No correlation was found between the access resistance and the rise times for each type of mIPSC (r = 0.236;p > 0.2). In addition, the input resistance of the neurons displaying GABAAR-mediated mIPSCs was not different from that of cells showing GlyR-mediated mIPSCs (544 ± 240 MΩ vs 422 ± 69 MΩ; respectively, p > 0.5). Finally, we found no correlation between the rise time and decay values of miniature GABAAR- and GlyR-mediated events (r = 0.335; p > 0.4), indicating that the differences in decay time constants could not be accounted for by differences in electrotonic distance from the soma for these two types of mIPSCs (i.e., GABAAR- mediated mIPSCs are unlikely to occur at synapses further away from the soma than GlyR-mediated mIPSCs).
Co-occurrence of both GABAAR- and GlyR-mediated mIPSCs in deeper neurons
Because miniature synaptic currents are thought to reflect release from single vesicles of transmitters, they can serve as a useful tool to test whether GABA and glycine are contained within the same vesicle. For example, if individual mIPSCs included both a GABAAR- and GlyR-mediated component, it could indicate co-packaging of GABA and glycine. In the neurons tested in laminae I and II, the mIPSCs were completely blocked by either strychnine or bicuculline/SR-95531. This cannot rule out the possibility of co-packaging because it may be a consequence of selective expression of one of the two respective receptors in these cells. Because there is clearer evidence that the same cell in lamina III expressed GABAAR and GlyRs (many dendrites in lamina I-II do not belong to laminae I-II cells) (Todd et al., 1996), we recorded mIPSCs in deeper (lamina III) dorsal horn neurons (n = 5; Fig. 4). In these cells, two populations of mIPSCs were identified on the basis of their decay kinetics (distribution of decay time constants best fitted by two Gaussians; Fig. 4 B). After administration of strychnine or bicuculline, only one of the two Gaussian populations of decay time constant remained; for example, in Figure 4, the faster mIPSCs were blocked by application of strychnine. No mIPSC with the combined fast and slow kinetics were found, indicating that separate populations of GABAAR- and GlyR-mediated mIPSCs were found within the same cell. This indicated that, for these cells, either GABA and glycine were likely released from separate vesicles of transmitter or that GABAARs and GlyRs were clustered at separate synaptic junctions.
Responses to exogenous applications of GABA and glycine
To determine whether the fact that GABA- and glycine-mediated mIPSCs were never observed jointly within the same cell was caused by a differential expression of their respective postsynaptic receptors among laminae I-II neurons, GABA and glycine were locally applied by pressure ejection to 27 cells. Interestingly, all neurons tested responded to both GABA and glycine (1 mm) regardless of the type of mIPSCs they displayed. Figure 5illustrates examples of responses induced by application of GABA and glycine in cells showing either exclusively GlyR- or exclusively GABAAR mIPSCs. GABA- and glycine-induced currents could be blocked by bicuculline/SR-95531 or strychnine, respectively (Fig.5 B–D), indicating that the responses to these agonists involved receptors similar to those mediating the mIPSCs.
Hypertonic solution-induced release of inhibitory neurotransmitters
To test whether the lack of detection of either GABAAR- or GlyR-mediated mIPSCs reflected a low-frequency of release of one of the two transmitter, in some experiments we applied hypertonic ACSF (590–610 mOsm, adjusted with sucrose) onto laminae I-II neurons (n = 14). The hypertonic stress appears to provoke exocytosis of the releasable pool of vesicles and thus substitutes for the calcium-induced release following action potential invasion (Rosenmund and Stevens, 1996). Yet, as the frequency of the synaptic events was increased but remained nonsynchronized, it was still possible to continue recording spontaneous action potential-independent IPSCs and thus compare the results with those from the control conditions. A 25-fold increase in frequency of mIPSCs was observed on average (range, 17- to 33-fold increase). In all cases, the mIPSCs were blocked either exclusively by bicuculline (or SR-95531) or by strychnine (Fig.6).
Large evoked IPSCs in lamina I
The observation that lamina I neurons displayed only GlyR-mediated mIPSCs, yet appeared to express both GlyRs and GABAARs, together with the finding that virtually all glycine-immunoreactive interneurons in the superficial dorsal horn also express GABA (Todd and Sullivan, 1990), suggests that GABAARs may not be located at synaptic junctions in lamina I neurons. By recruiting more synaptic input synchronously, accumulation of transmitter may lead to sufficient spillover from synapses to activate distant receptors (Isaacson et al., 1993). To test this possibility, we recorded monosynaptic evoked IPSCs in lamina I by placing a bipolar tungsten electrode within 20–50 μm of the recorded neurons. Large intensity stimulation (200–500 μA; 200–300 μsec duration) were used at a frequency of ≤0.3 Hz. The evoked IPSCs (eIPSCs) were only partially attenuated by 1–2 μmstrychnine; the remaining component was blocked by 20 μmbicuculline (n = 5; Fig.7 A). Thus, in lamina I, there was a GABAAR-mediated component of large eIPSCs. The rise time of the GABAAR-mediated component of these eIPSCs was significantly slower than that of the GlyR-mediated component (6.2 ± 0.5 msec vs 2.8 ± 0.6 msec, respectively;p < 0.005, Fig. 7 A). The decay time course of the GlyR-mediated component (17.1 ± 3.2 msec), was also significantly faster than that of the GABAAR component (80.4 ± 30.3 msec; p < 0.05).
Minimal versus large evoked IPSCs in lamina I
To investigate whether the differences in rise times of the evoked GlyR- and GABAAR-mediated IPSCs could reflect a difference in the location of their respective receptors, we used minimal stimuli (producing all-or-none IPSCs), using patch micropipettes placed either close to the soma (within 20–50 μm) or close to a dendrite at a distance from the cell body (150–300 μm). Because most of the neurons in lamina I have a bipolar morphology in the parasagittal plane (Light et al., 1979; Woolf and Fitzgerald, 1983;Lima and Coimbra, 1986; Ma et al., 1996), the positioning of the stimulating electrode could be achieved to reach specifically one portion of the dendrite of the recorded neuron (Fig.8, diagram). In all cases tested, minimally evoked IPSCs (meIPSCs) obtained from stimulations within 20 μm from the cell body were completely blocked by 1 μm strychnine (n = 12; Figs.7 B, 8). After complete block by strychnine and after a 5- to 10-fold increase in the stimulus intensity, a slower IPSC could be evoked that was blocked by 10–20 μmbicuculline (Figs. 7 B, 8). The rise time of the GABAAR-mediated IPSCs was 2.2 ± 1.2 msec, significantly slower that that of GlyR-mediated meIPSCs (0.9 ± 0.5 msec; Fig. 7 B, inset).
The GABAAR-evoked IPSCs may have had slower rise times because they systematically originated at a more distant site from the soma than the GlyR-evoked IPSCs. For this to be the case, the GABAAR evoked IPSCs resulting from stimulation proximal to the soma would have to have a significantly longer latency than GlyR-evoked IPSCs (i.e., proximal stimuli producing GABAAR IPSCs would have to activate fibers that travel a longer distance to reach distal portions of the postsynaptic cell). We thus compared the latencies of the different components with each type of stimuli.
The latency of both GlyR and GABAAR-mediated components was comparable (1.2 ± 0.2 vs 1.7 ± 0.4 msec;p > 0.1). Similarly, the latency of GlyR-mediated meIPSCs elicited by stimulation close to the soma versus that in the vicinity of a distal portion of a dendrite were not significantly different (1.2 ± 0.2 vs 1.7 ± 0.6 msec, respectively). Moreover, stimulation at a >100 μm distance from the soma but also away from the vicinity of a dendrite always failed to produce IPSCs. Evoked GABAAR-mediated IPSCs (using larger stimulus intensities) at these same proximal and distal sites of stimulation had also comparable latencies (1.7 ± 0.4 vs 2.1 ± 0.9 msec, respectively). Finally, a comparison of the rise time of GlyR-mediated meIPSCs originating at a site proximal or distal to the soma revealed little prolongation (0.7 ± 0.6 vs 0.9 ± 0.5 msec, respectively; Fig. 8 A), consistent with the short membrane time constant of these cells (15.2 ± 6.8 msec; range, 2.5–45.3 msec). Proximal and distal GABAAR-mediated IPSCs also had comparable rise times (2.2 ± 1.2 vs 2.3 ± 1.1 msec, respectively; Fig.8 B). The decay time constant of the GlyR and GABAAR components evoked by these focal stimulations were 10.8 ± 1.7 msec versus 42.7 ± 6.0 msec, respectively.
Thus, it appeared that both GABAAR- and GlyR-mediated IPSCs could originate at similar electrotonic distance from the soma. Eliciting GABAAR-mediated IPSCs, however, always required stronger stimulus intensities that produced synaptic events with significantly slower rising and decaying kinetics than their GlyR-mediated counterpart. Taken together, these results indicated that the slower rising phase of evoked GABAAR-mediated IPSCs appeared not to be caused by a more distant site of origin but rather a slower, likely diffuse or distant activation of GABAAR after sufficient accumulation of released GABA.
Effect of flunitrazepam on miniature IPSCs in lamina I
The results from mIPSCs and stimulus-evoked IPSCs suggested that either different threshold for glycine and GABA release or subthreshold activation of GABAARs by single vesicles of GABA. To distinguish between these two possibilities, we aimed at raising the sensitivity of GABAARs with the benzodiazepine flunitrazepam, to potentiate possible subliminal responses to GABA released by single vesicles of transmitter. In the presence of 1 μm flunitrazepam, the rise time of mIPSCs was significantly prolonged to 2.4 ± 0.3 msec, and the decay time constant increased to 27 ± 4.2 msec (vs 0.6 ± 0.1 and 6.0 ± 0.3 msec in control; n = 7; Fig.9). Close examination of the distribution of rise times and decay time constants (Fig. 9 B–D) indicates that the large majority of mIPSCs had their kinetics prolonged by flunitrazepam. Addition of 100 nmstrychnine to the bath solution containing flunitrazepam revealed mIPSCs with very slow rise time (4.1 ± 0.9 msec); 10 times more prolonged than that of GlyR mIPSCs in lamina I neurons and of GABAAR mIPSCs in neurons of deeper laminae. Similarly, the decay kinetics (52.8 ± 8.9 msec) were ∼10× slower than that of the GlyR mIPSCs in lamina I neurons and 5× slower that of GABAAR mIPSCs in lamina II. Addition of 10 μm bicuculline completely abolished all of these remaining slow mIPSCs, indicating that they were GABAAR-mediated. The frequency of mIPSCs in flunitrazepam was slightly increased (20–30%) over that in control conditions, most likely because, at some junctions, no synaptic glycine receptors were present. Importantly however, the frequency of the very slow GABAAR mIPSCs that persisted in the presence of strychnine and flunitrazepam was always >75% of that of the number of events in control conditions and more than twice the increase in frequency observed in the presence of flunitrazepam alone (Fig.9 B,C). There was also <11% overlap in areas between the distributions of decay time constants in control versus flunitrazepam (Fig. 9 D). This therefore indicated that the majority of individual mIPSCs had a dual GlyR and GABAAR component. Given the fact that mIPSCs are likely reflecting responses to single vesicles of transmitter, these results indicate co-storage of GABA and glycine in the same synaptic vesicles.
Effects of GABA uptake inhibitors on IPSCs in lamina I
Because GABA uptake inhibitors do not affect the amplitude or the time course of mIPSCs, it is thought that clearance of GABA from the synaptic cleft is not a limiting factor determining the decay time course of mIPSCs (Thompson and Gahwiler, 1992; Isaacson et al., 1993). On the other hand, the uptake blockers prolong large evoked GABAAR-mediated IPSCs. If the GABA released by single vesicles at lamina I synapse is insufficient to activate distant GABAARs, blocking uptake should not significantly affect mIPSCs. On the other hand, with “spillover” of GABA from several neighboring synapses activated synchronously, blocking uptake may favor temporal summation of extracellular GABA originating from adjacent sources (Isaacson et al., 1993). We first tested the effects of tiagabine (25–50 μm) and NO-711 (10–30 μm) of normal mIPSCs in lamina I (n = 7). Neither uptake blocker had a detectable effect on these mIPSCs (Fig.10). After tiagabine or NO-711 administration, eIPSCs had significantly reduced peak amplitude, and their decay time courses were prolonged (n = 5; mean increase in decay of the late component of 44 ± 3 msec; p < 0.01; Fig. 10).
Uptake may be important in limiting the extent of extrasynaptic receptor activation (Isaacson et al., 1993; Asztely et al., 1997). Yet, failure to detect an effect of the uptake inhibitor on miniature events may simply reflect that the amount of “spillover” during a miniature event is subliminal for activation of these receptors. At glutamate synapses for example, spillover of transmitter after release of a single vesicle appears to be relevant only for NMDA receptors because they have a much higher affinity for the transmitter than AMPA receptors. It may be expected, therefore, that raising the affinity of extrasynaptic GABAA receptors with a benzodiazepine could reveal a significant effect of the uptake blocker on mIPSCs. To test this, we recorded pure GABAAR mIPSCs in lamina I revealed in the presence of 1 μmflunitrazepam and 100 nm strychnine to test for the effect of the uptake inhibitors. Tiagabine and NO-711 administration decreased the frequency of mIPSCs and significantly prolonged their decay time course (from 62.3 ± 8.7 to 106.7 ± 10.2 msec;p < 0.05; n = 5; Fig. 10). In all cases, subsequent addition of 10 μm bicuculline or 3 μm SR-95531 abolished all of these mIPSCs.
Our results indicate that GABA and glycine mediate separate miniature IPSCs with distinct kinetics in superficial dorsal horn neurons of the rat spinal cord, and these cells are usually bombarded by one type of mIPSC only. Even in cells with both types of mIPSCs present, these synaptic events represent separate populations. The differences in kinetics of the GlyR- versus GABAAR-mediated mIPSCs are consistent with previous findings in the dorsal horn (Takahashi and Momiyama, 1991;Takahashi et al., 1992; Baba et al., 1994; Yoshimura and Nishi, 1993,1995) and in medullary neurons (Lewis and Faber, 1996; Grudt and Henderson, 1998), in contrast with retinal ganglion cells in which GABAAR-mediated IPSCs have a faster decay than GlyR-mediated ones (Protti et al., 1997).
Immunocytochemical studies by Todd and Sullivan (1990) demonstrated that the proportion of glycine-immunoreactive neurons in laminae I, II, and III were 9, 14, and 30%, respectively. The proportions of GABA-immunoreactive neurons in laminae I, II, and III were higher: 28, 31, and 46%, respectively. They also reported that virtually all of the glycine-immunoreactive cells in this area were also GABA-immunoreactive, but many GABA-immunoreactive cells do not show immunoreactivity to glycine. Thus, while recording from laminae I-II, one would expect that the neurons displaying GlyR-mediated mIPSCs would also show GABAAR-mediated mIPSCs. On the contrary, we found namely that lamina I neurons were bombarded exclusively by GlyR-mediated mIPSCs in normal conditions. Thus, glycine appears to be solely responsible for tonic inhibition of second order neurons in this layer.
Co-release of glycine and GABA from the same vesicle
Unmasking of a very slow rising and decaying GABAA component of mIPSCs in lamina I with flunitrazepam provided an ideal tool to address the issue of co-storage of GABA and glycine in the same vesicles. Importantly, the whole distribution of rise times and decay time constants was shifted in the presence of the benzodiazepine with little overlap with the control distribution, indicating that the majority of the events were affected. Accordingly, the increase in frequency of events observed in the presence of the benzodiazepine was much lower than the number of GABAAR-mediated mIPSCs remaining in the presence of strychnine. Thus, it can be concluded that the majority of mIPSCs had a dual GlyR- and GABAAR-mediated component. Because mIPSCs appear to reflect postsynaptic responses to the release of single vesicles of transmitter, the data strongly indicated the co-storage of glycine and GABA in the same vesicles. This is consistent with similar recent evidence at the immature motoneuron synapse (Jonas et al., 1998) as well as with the observation that both transmitters can be carried by the same vesicular transporter (Christensen and Fonnum, 1991; Burger et al., 1991) present at both glycinergic and GABAergic synapses (Chaudhry et al., 1998; Dumoulin et al., 1999). Whereas the evidence presented byJonas et al. (1998) relied on template fits of the decay phase of mIPSCs, our evidence has the added advantage that the rising kinetics of the GABAA component was also dramatically slower than that of the GlyR component and that the events were sufficiently altered after the benzodiazepine administration that it was possible to quantify the degree of overlap between the distributions of events for each kinetic parameter, providing an even more compelling argument for the co-storage of GABA and glycine.
Extrajunctional GABAAR activation in lamina I
Several pieces of evidence converge to indicate that, in contrast to GlyRs, GABAARs are likely located extrasynaptically in lamina I: (1) the absence of GABAAR-mediated spontaneous and miniature IPSCs while the cells received GlyR-mediated mIPSC and all responded to exogenous application of GABA in a bicuculline-dependent matter; (2) activation of several inhibitory synapses synchronously (large evoked IPSCs) revealed a GABAA component; (3) the evoked GABAAR component had a similar latency, yet slower rising kinetics than the GlyR component; (4) benzodiazepines can unmask a GABAA component to individual mIPSCs , and this component has extremely slow rising kinetics (10× slower than GlyR mIPSCs in lamina I or GABAAR mIPSCs in deeper laminae); (5) specific blockers of GABA uptake did not affect normal mIPSCs in lamina I but significantly prolonged the decay kinetics of evoked IPSCs that involve synchronous activation of neighboring synapses as well as of GABAAR mIPSCs unmasked by flunitrazepam (in the presence of strychnine).
The slow kinetics of GABAAR-mediated IPSCs in lamina I could suggest that they specifically originate at distant sites from the soma [therefore subject to greater space clamp attenuation (Spruston et al., 1994)]. However, the short time constant of lamina I neurons and the lack of difference in the kinetics of proximally versus distally evoked IPSCs argues against this possibility. More importantly, the dual-component mIPSCs in flunitrazepam clearly show that the slow GABAAcomponents originated from the same release sites as the fast-rising GlyR components.
Extrasynaptic distribution of GABAA receptor subunits have often been described in the spinal cord and other brain regions (Somogyi et al., 1989; Soltesz et al., 1990; Bohlhalter et al., 1994; Nusser et al., 1998). In the dorsal horn, β2/β3 subunit immunoreactivity, possibly the most widely expressed subunits in the spinal cord, appeared often extrasynaptically (Alvarez et al., 1996). It may be however difficult to draw definitive conclusions from studies on the subcellular distribution of GABAARs with pre-embedding approaches because of the limited access of some antigenic sites (Nusser et al., 1995). Nevertheless, even with pre-embedding approaches, antibodies directed against an intracellular loop of the receptor may have better access to postsynaptic densities (active zones) (Todd et al., 1996). Interestingly, whereas Todd et al. (1996) found punctate staining for the β3subunit in deeper laminae, the labeling often extended beyond active sites in laminae I-II. In fact, it was often difficult to localize precisely the β3 staining in these laminae (A. J. Todd, personal communication). For this reason, they refrained from further quantifying β3 versus gephyrin immunoreactivity in lamina I. Thus, although immunocytochemical evidence may be consistent with the possibility of a prominent extrasynaptic distribution of GABAARs in lamina I, no direct data are available.
Evidence with benzodiazepines rules out the possibility of a specifically higher threshold for activation of interneurons releasing GABA. The very slow rise time of GABAAR-mediated mIPSCs revealed by flunitrazepam suggests that either very little GABA is contained in glycine-containing synaptic vesicles or that the GABAARs are located at a distance from the site of release. No evidence is available, however, to indicate that a very low level of GABA in synaptic vesicles may occur. In fact, all available evidence suggest the opposite (Mody et al., 1994). Furthermore, this possibility would appear highly unlikely in the present case, because the vesicular transporter has a greater affinity for GABA than glycine (Burger et al., 1991; Christensen and Fonnum, 1991) and because substantial levels of immunoreactivity for GABA is detected in these cells (Todd and Sullivan, 1990) and their terminals (Todd et al., 1996). Finally, whereas results from rapid applications of GABA to excised membrane patches would tend to link slow rise times with low agonist concentration (Maconochie et al., 1994), it is interesting to note that with very brief transients of transmitter [as seems to be the case in the cleft (Destexhe and Sejnowski, 1995;Clements, 1996)], even very low concentrations of GABA generate currents with fast rise times (Galarreta and Hestrin, 1997). Low concentration of GABA is also not expected to generate longer decay time courses (Maconochie et al., 1994; Galarreta and Hestrin, 1997) that could not account for the much slower decay of the GABAAR components in lamina I versus those recorded in lamina II [see also discussion in (Rossi and Hamann, 1998)]. Thus, the most plausible explanation for our results is that GABAARs are located at a distance from the vesicle release site, suggesting a prominent perisynaptic distribution of functional GABAARs at these synapses (Fig.11).
Functional significance of the difference in GABAAR- and GlyR-mediated inhibition in lamina I
Symptoms very similar to those observed in neuropathic pain models involving peripheral nerve constriction can also be obtained with intrathecal administration of subconvulsant doses of strychnine or bicuculline (Yaksh, 1989; Sivilotti and Woolf, 1994). Importantly, strychnine-induced hypersensitivity is selective to non-noxious input (allodynia) and is morphine-insensitive (Sherman and Loomis, 1994,1995, 1996; Sorkin and Puig, 1996). Moreover, the sensitization produced does not only affect spinal segmental nociceptive reflexes, but also ascending nociceptive pathways (Sherman et al., 1997a,b). Thus GABAA and glycine receptor-mediated inhibition appear to play an important role in the regulation of excitability in specific nociceptive sensory pathways in the dorsal horn, and blockade of this control appears to unveil subliminal innocuous input to neurons in these pathways.
The present finding of the differential distribution of GABAA and glycine receptor-mediated mIPSCs among cell types and especially the finding that only glycine is responsible for the tonic inhibition of lamina I neurons suggests separate roles for these two inhibitory systems. The selective tonic inhibition exerted by glycine on lamina I neurons may be contrasted to evoked inhibition mediated by GABA acting on GABAAreceptors. Blockade of either type of control may differentially affect nociceptive integration. For example, the difference in kinetics of the GABAergic versus glycinergic IPSCs and the type of activity that recruit them may match the nature of excitatory input they control. Interestingly, metabotropic receptor antagonists attenuate bicuculline-induced allodynia but are not effective against strychnine-induced allodynia (Onaka et al., 1996). Metabotropic glutamate receptors appear to be mainly located at extrasynaptic sites (Baude et al., 1993) and may therefore be activated selectively after release of a sufficient concentration of glutamate that can spill over the synapses (Rusakov and Kullmann, 1998). This property may match that guiding activation of GABAARs in lamina I (also because of their perisynaptic distribution). Thus, subcellular segregation of GABAARs and GlyRs confers distinct functional roles to inhibition mediated by these two transmitters in lamina I.
This work was supported by National Institute for Neurological Disorders and Stroke Grant NS 34022 and by Canadian Medical Research Council (MRC) Grant MT 12170 to Y.D.K. Y.D.K. is a Scholar of the Canadian MRC. N.C. was the recipient of a Faculty of Medicine Graduate Award and an Eileen Peters McGill Major Fellowship. We thank Drs. R. C̆apek, M. W. Salter, and A. J. Todd for helpful comments on this manuscript and Ms. A. Constantin for expert technical assistance. We thank Hoffman-La Roche for the generous donation of flunitrazepam and Abbott Labs for tiagabine.
Correspondence should be addressed to Dr. Yves De Koninck, Department of Pharmacology and Therapeutics, McGill University, 3655 Drummond, #1317, Montreal, Quebec, H3G 1Y6 Canada.