Asymmetric Cross-inhibition between GABAA and Glycine Receptors in Rat Spinal Dorsal Horn Neurons*

Presynaptic nerve terminals of inhibitory synapses in the dorsal horn of the spinal cord and brain stem can release both GABA and glycine, leading to coactivation of postsynaptic GABAA and glycine receptors. In the present study we have analyzed functional interactions between GABAA and glycine receptors in acutely dissociated neurons from rat sacral dorsal commissural nucleus. Although the application of GABA and glycine activates pharmacologically distinct receptors, the current induced by a simultaneous application of these two transmitters was less than the sum of currents induced by applying two transmitters separately. Sequential application of glycine and GABA revealed that the GABA-evoked current is more affected by glycine than glycineevoked responses by GABA. Activation of glycine receptors decreased the amplitude and accelerated the rate of desensitization of GABA-induced currents. This asymmetric cross-inhibition is reversible, dependent on the agonist concentration applied, but independent of both membrane potential and intracellular calcium concentration or changes in the chloride equilibrium potential. During sequential applications, the asymmetric cross-inhibition was prevented by selective GABAA or glycine receptor antagonists, suggesting that occupation of binding sites did not suffice to induce glycine and GABAA receptors functional interaction, and receptor channel activation is required. Furthermore, inhibition of phosphatase 2B, but not phosphatase 1 or 2A, prevented GABAA receptor inhibition by glycine receptor activation, whereas inhibition of phosphorylation pathways rendered cross-talk irreversible. Taken together, our results demonstrated that there is an asymmetric cross-inhibition between glycine and GABAA receptors and that a selective modulation of the state of phosphorylation of GABAA receptor and/or mediator proteins underlies the asymmetry in the cross-inhibition.

Presynaptic nerve terminals of inhibitory synapses in the dorsal horn of the spinal cord and brain stem can release both GABA and glycine, leading to coactivation of postsynaptic GABA A and glycine receptors. In the present study we have analyzed functional interactions between GABA A and glycine receptors in acutely dissociated neurons from rat sacral dorsal commissural nucleus. Although the application of GABA and glycine activates pharmacologically distinct receptors, the current induced by a simultaneous application of these two transmitters was less than the sum of currents induced by applying two transmitters separately. Sequential application of glycine and GABA revealed that the GABAevoked current is more affected by glycine than glycineevoked responses by GABA. Activation of glycine receptors decreased the amplitude and accelerated the rate of desensitization of GABA-induced currents. This asymmetric cross-inhibition is reversible, dependent on the agonist concentration applied, but independent of both membrane potential and intracellular calcium concentration or changes in the chloride equilibrium potential. During sequential applications, the asymmetric cross-inhibition was prevented by selective GABA A or glycine receptor antagonists, suggesting that occupation of binding sites did not suffice to induce glycine and GABA A receptors functional interaction, and receptor channel activation is required. Furthermore, inhibition of phosphatase 2B, but not phosphatase 1 or 2A, prevented GABA A receptor inhibition by glycine receptor activation, whereas inhibition of phosphorylation pathways rendered cross-talk irreversible. Taken together, our results demonstrated that there is an asymmetric cross-inhibition between glycine and GABA A receptors and that a selective modulation of the state of phosphorylation of GABA A receptor and/or mediator proteins underlies the asymmetry in the cross-inhibition.
Although neurotransmission involves specific activation of receptor channels by distinct neurotransmitters, different classes of receptor can be colocalized at the same postsynaptic site and may be activated by the corelease of more than one type of neurotransmitter from the same presynaptic nerve terminal. Recently it has been demonstrated that simultaneous activation of different postsynaptic receptors by the coapplication of their specific neurotransmitter induces cross-modulation of their activation properties. This cross-talk phenomenon has been proposed to represent a fast adaptive process in controlling signal transmission (1). Negative cross-talk was demonstrated between ATP P2X and nicotinic acetylcholine receptors (2)(3)(4)(5), between dopamine and adenosine receptors (6), between ␥-aminobutyric acid type A (GABA A ) 1 and dopamine (1) or P2X receptors (7), as well as between dopamine and N-methyl-D-aspartic acid receptors (8). In general, the crosstalk between these receptors is characterized by a partial occlusion of transmitter-evoked currents, i.e. the sum of the amplitudes of responses evoked by each agonist is larger than the amplitude of responses evoked by coapplication of the two neurotransmitters. The molecular mechanism involved in the interaction between P2X and nicotinic acetylcholine receptors is unknown. Negative cross-talk between dopamine and adenosine receptors, between GABA A and dopamine receptors, and between dopamine and N-methyl-D-aspartic acid receptors involves direct intramembranous protein-protein interaction (1,6,8), whereas cross-talk between P2X receptor and GABA A receptors (GABA A R) is modulated by chloride efflux and intracellular Ca 2ϩ (7).
In the spinal cord, brain stem (9 -12), and cerebellum (13), GABA and glycine can be coreleased by the same synaptic terminal, whereas GABA A R and glycine receptors (GlyR) can be coaggregated at the same postsynaptic site (13,14). The ionotropic GABA A and GlyR both gate chloride-permeable channels but have distinct structures, resulting from oligomerization of specific subunits (15,16). There is evidence suggesting that GABA A R and GlyR can interact negatively in mammalian (17,18) or lamprey (19) spinal cord neurons, in olfactory bulb cells (20), and in the hippocampus (21), i.e. the effects of GABA and glycine are less than additive when they are coapplied. The mechanisms responsible for this cross-inhibition between GABA and glycine remain poorly understood. One possibility is that it reflects the presence, at least in the olfac-tory bulb, of a receptor subpopulation that can bind either GABA or glycine (20). However the identity of these postulated receptors remains unclear.
In the present study, we explored the mechanisms of the functional interaction between GlyR and GABA A R by analyzing cross-inhibition between glycine-and GABA-induced Cl Ϫ currents (I Gly and I GABA ) in acutely dissociated rat sacral dorsal commissural nucleus neurons, using whole cell, patch clamp recording. Our results showed that cross-inhibition is asymmetric between GABA A R and GlyR and that glycine-induced inhibition of GABA responses depends largely on protein dephosphorylation processes.

EXPERIMENTAL PROCEDURES
Cell Preparation-Rat sacral dorsal commissural nucleus neurons were acutely dissociated according to the method of Wu et al. (11). In brief, pentobarbital-sodium-anesthetized (45-50 mg kg Ϫ1 , intraperitoneally) Wistar rats (2 weeks old) were decapitated. A segment about 10 -15 mm long of lumbosacral (L 5 -S 3 ) spinal cord was quickly dissected out and immersed in the standard external solution at freezing temperatures. After removing attached dorsal rootlets and the pia matter on the lateral aspects of the cord, the spinal segment was fixed with cyanoacrylic glue to a 15 ϫ 15-mm 2 agar block to support the spinal cord tissue. The tissue block was then placed in the cutting chamber of a vibratome tissue slicer (LEICA VT1000S, Leica Instruments Ltd., Wetzlar, Germany). A cold standard external solution (about 4°C) bubbled with O 2 was subsequently placed in the chamber to immerse the tissue block. The spinal segment was sectioned to yield several transverse slices of thickness 400 m. Slices were preincubated in oxygenated incubation solution for 30 min at room temperature (22-25°C) and then treated enzymatically in oxygenated incubation solution containing Pronase (1 mg/5 ml) for 20 min at 31°C. This treatment was followed by exposure to thermolysin (1 mg/5 ml) for 15 min. After the enzyme treatment, slices were kept in enzyme-free incubation solution for 1 h. Then a portion of dorsal horn region was micropunched out and trans-ferred into a culture dish filled with the standard external solution. Neurons were mechanically dissociated with fire-polished Pasteur pipettes under visual guidance under a phase contrast microscope (IX70, Olympus Optical Co., Ltd., Tokyo, Japan). Within 20 min, isolated neurons had attached to the bottom of the culture dish and were ready for electrical recording. The care and use of animals in these experiments followed guidelines and protocols approved by our institutional Animal Care and Use Committee.
Electrophysiology-Whole cell, voltage clamp recordings were made at room temperature (22-25°C). Culture dishes (Corning 430165, Corning, Inc., Corning, NY) were used as recording chambers and were perfused at 0.5-2.0 ml/min with the standard external solution. Patch pipettes were pulled from glass capillaries with an outer diameter of 1.5 mm (Narishige, Tokyo, Japan) on a two-stage puller (PP-830, Narishige), and had a resistance of 4 -6 megohms. Membrane potentials were corrected for the liquid junction potential, which had a measured value of 3-4 mV. The series resistance (R s ), estimated from visual cancellation of the capacity transient, was 5-15 megohms. In most experiments, 80 -90% series resistance compensation was applied. Artifacts caused by inadequate voltage clamp and space clamp were minimized by selecting for experiment neurons bearing no or short processes. Transmitter-evoked currents recorded from the cell soma normally did not exceed 5 nA. Unless otherwise noted, CsCl solution was used in the recording pipette. After taking these precautions, the largest recorded currents did not differ in time course from the smaller ones. To ensure cell dialysis, data for measurements were obtained at Solutions-The composition of incubation solution was (in mM) 124 NaCl, 24 NaHCO 3, 5 KCl, 1.2 KH 2 PO 4 , 2.4 CaCl 2 , 1.3 MgSO 4 , and 10 glucose, and it was aerated with 95% O 2 and 5% CO 2 to a final pH of 7.4. The standard external solution contained (in mM) 150 NaCl, 5 KCl, 1 MgCl 2 , 2 CaCl 2 , 10 HEPES, and 10 glucose. The pH was adjusted to 7.4 with Tris. The osmolarity of all bath solutions was adjusted to 310 -320 mosm/liter with sucrose. The ionic composition of the internal solution medium was (in mM) 120 CsCl, 30 NaCl, 0.5 CaCl 2 2H 2 O, 1 MgCl 2 6H 2 O, 5 EGTA, 2 MgATP, and 10 HEPES with the pH adjusted to 7.2. Stocks of MgATP stored at Ϫ20°C were dissolved in the intracellular solution shortly before use to a final concentration of 2 mM. Unless otherwise noted, the membrane potential was held at Ϫ50 mV in the voltage clamp studies.
Drugs and Application System-Cyclosporin A (CSPN), staurosporine, and Li 4 ATP␥S were obtained from Biomol. Okadaic acid (OA) and GF 109203X were from Tocris (Bristol, UK), and H89, KN93, and genistein were from Sigma. All other drugs were from Sigma. Phosphatase inhibitors and protein kinase inhibitors were added to the pipette solution. Agonists or antagonists of GlyR and GABA A R were diluted with extracellular solution to a final concentration and applied via the "Y-tube" method as described previously (22). The tip of the drug tube was positioned between 50 and 100 m away from the patched neurons. This system allows a complete exchange of external solution surrounding a neuron within 20 ms. Throughout the experiment the bath was perfused continuously with the standard external solution.
Data Acquisition and Analysis-Signals were filtered at 1 kHz, data were collected with an Axopatch 200B patch clamp amplifier (Axon Instruments, Foster City, CA), which was connected to a Pentium III computer equipped with Digidata 1320A, Clampex, and Clampfit software (Axon Instruments). Data were analyzed with the pCLAMP software (Axon Instruments) and ORIGIN for Windows (Microcal Software, Northampton, MA). Results are presented as the mean Ϯ S.E. (n ϭ number of cells), with statistical significance assessed by Student's t test for two groups' comparison or one-way analysis of variance test for multiple comparisons. A p value of Ͻ0.05 or 0.01 was considered statistically significant. To evaluate the strychnine and bicuculline concentrations at half-maximal inhibition of I Gly and I GABA (IC 50 ), the mirror image of the Michaelis-Menten equation was fitted to the data by the least squares method. Current desensitization was fitted by exponential functions, beginning shortly after the peak of response, using Clampfit 8.0 software. Curve fitting was per-formed using a simplex algorithm least squares exponential fitting routine with single or double exponential equations of the form where I f and I s are the amplitudes of the fast and slow components, respectively; I p was the amplitude of the nondesensitized current; and f and s are respective time constants.

GABA-and Glycine-induced
Responses-Cross-talk between GABA A R and GlyR was analyzed on acutely dissociated sacral dorsal commissural nucleus neurons from P14 rats. At this age, neurons express the mature form of GABA A R and GlyR (15,23). Whole cell recording at the holding potential (V h ) of Ϫ50 mV showed that application of GABA or glycine evoked inward chloride currents in all cells tested. The reversal potentials of GABA-and glycine-induced currents (E GABA and E Gly ) were Ϫ2.5 Ϯ 1.1 (S.E., n ϭ 6) and Ϫ1.3 Ϯ 0.8 mV (n ϭ 5), respectively, and were not significantly different (p Ͼ 0.1; unpaired t test). They were close to the theoretical Cl Ϫ equilibrium potential (E Cl ) calculated from the Nernst equation (Ϫ1.3 mV), based on the external and internal Cl Ϫ concentration used in the recording (see "Experimental Procedures").
The saturating concentration and EC 50 for GABA or glycine were determined by analyzing concentration-response curves for responses to a 10-s application of agonists ( Fig. 1). I Gly and I GABA were normalized to the peak amplitude of the currents evoked by the application of glycine and GABA at 1 mM, respectively. The normalized data were fitted using a single isotherm function of the form, where I max is the maximum current amplitude, I/I max is the normalized current amplitude, EC 50 is the agonist concentration producing 50% of the I max , and H is the Hill coefficient. This fit produced an EC 50 of 8.1 and 34 M for GABA and glycine, respectively. The Hill coefficients for GABA and glycine were 1.18 and 1.27, respectively.

FIG. 3. Effect of bicuculline and strychnine on I GABA and I Gly .
A and B, 30 M bicuculline (Bic) was an effective antagonist of 1 mM I GABA but had little effect on 1 mM I Gly . To the contrary, 1 M strychnine (Str) had little effect on 1 mM I GABA but inhibited 1 mM I Gly . C and D, concentration-response relationships for inhibition of 1 mM I GABA and 1 mM I Gly by bicuculline (C) and strychnine (D), respectively. The antagonists were perfused 30 s before simultaneous application of the agonists and the antagonists. The amplitudes of I GABA and I Gly were measured at the peak and expressed as values relative to the control response induced by 1 mM GABA or 1 mM Gly alone, respectively. Each point represents the mean of 6 -10 neurons.
Cross-inhibition between I GABA and I Gly -The above experiments suggest that the saturating concentration for GABA or glycine for inducing neuronal responses was Ն1 mM. This suggests that responses evoked by the application of 1 mM GABA or 1 mM glycine should be additive if they are induced by the activation of independent receptors. We therefore compared peak amplitudes of currents induced by the application of either GABA or glycine (at 1 mM) and by the simultaneous application of both agonists. As shown in Fig. 2A, coapplication of GABA and glycine evoked currents with an amplitude that was significantly lower than the expected sum (I EX ) of currents evoked separately by GABA and glycine (Fig. 2B), suggesting a cross-inhibition between I GABA and I Gly (Fig. 2B; n ϭ 36; p Ͻ 0.01; t test). A quantitative analysis of this cross-inhibition and its dependence on concentrations of both agonists is shown in Fig. 2C. At 10 M GABA and glycine, no significant crossinhibition occurred (91.5 Ϯ 5.3% of the I EX ; n ϭ 9; p Ͼ 0.05). Significant cross-inhibition was observed at 30 M GABA and glycine (55.1 Ϯ 6.4% of the I EX ; n ϭ 11; p Ͻ 0.01), and the effect appears to saturate beyond 30 M concentrations of the two agonists.
Cross-inhibition Was Not Caused by Nonspecific Receptor Activation by GABA and Glycine-It has been suggested that at high concentrations GABA may activate GlyR (24, 25), resulting in an apparent cross-talk between GABA A R and GlyR (20). We therefore used selective antagonists to determine the specificity of GABA and glycine in their receptor activation (Fig. 3). At 30 M bicuculline, a selective GABA A R antagonist, the I GABA induced by 1 mM GABA was totally abolished, whereas the I Gly induced by 1 mM glycine was unaffected (Fig.  3, A and B). Conversely, 1 M strychnine, a potent GlyR antagonist, completely abolished I Gly without affecting I GABA . Doseresponse curves for bicuculline and strychnine inhibition were obtained from currents evoked by glycine or GABA at 1 mM. The bicuculline inhibited I GABA in a concentration-dependent manner with an IC 50 of 0.68 M (Fig. 3C). Strychnine also inhibited I GABA , but only at concentrations higher than 1 M (Fig. 3D), with IC 50 of 2.83 M (Fig. 3D). The inhibitory effect of bicuculline on I Gly was also evident only at concentrations higher than 30 M (Fig. 3C), a concentration that completely suppressed I GABA . Strychnine was effective in suppressing I Gly at concentrations Ն 0.01 M (Fig. 3D). In contrast, the IC 50 values for the inhibition of I Gly were 0.04 and 74.2 M for strychnine and bicuculline, respectively. Taken together, these results suggest that at a concentration of 1 mM, GABA and glycine are unlikely to cross-activate GlyR or GABA A R, respectively, significantly.
Cross-inhibition between GABA A R and GlyR Was Asymmetric-A different mode of drug application was used to explore further the interactions between I Gly and I GABA and to examine relative contributions of GABA A R and GlyR to the cross-inhibition. As shown in Fig. 4A, GABA or glycine responses obtained in the presence of the other agonist were reduced in amplitude, with GABA-evoked responses more strongly affected by glycine than the reverse condition. The cross-inhibi-

FIG. 4. Asymmetric cross-interaction between I GABA and I Gly .
A and B, sample recordings demonstrating application of 1 mM GABA and 1 mM Gly with two different drug application modes. C, pooled percentage inhibition of I GABA or I Gly after Gly (IЈ GABA in B) or GABA (IЈ Gly in B) prepulse. Ordinates represent IЈ GABA /I GABA or IЈ Gly /I Gly ϫ 100%. The dashed line indicates the control I GABA or I Gly in the absence of Gly or GABA prepulse. Note the much stronger reduction of I GABA by Gly than contrariwise. In this and subsequent figures, unless otherwise noted, sequential application means that one agonist was applied immediately after (200 -400 ms) another agonist application. *, p Ͻ 0.05; **, p Ͻ 0.01.

FIG. 5. Direction of current flow did not influence cross-inhibition.
A, left, inhibition of I GABA by Gly prepulse was similar at Ϫ50 mV and at ϩ50 mV (E Cl ϭ 0 mV). A, right, the inhibition of I Gly by GABA prepulse was similar at Ϫ50 mV and at ϩ50 mV. The traces in A were from the same cells. B, representative current trace obtained from the ramp voltage command. The experimental protocol is shown above the current trace. Three voltage ramps ranging from Ϫ80 to ϩ80 mV were applied. Gly and GABA were applied to the cell and covered the last two ramps. Traces obtained from the first ramp measured background or leakage currents. C, current-voltage (I-V) curves for I GABA and I Gly obtained from traces shown in B.
tion was also observed during sequential application of 1 mM GABA and 1 mM glycine or vice versa. As shown in Fig. 4B, when glycine was applied before GABA, the amplitude of I GABA became 37.4 Ϯ 2.5% (n ϭ 12; p Ͻ 0.01) of the control values observed in the absence of glycine. Similarly, I Gly was reduced to 77.9 Ϯ 4.0% (n ϭ 15; p Ͻ 0.05) of the control value after a prepulse of 1 mM GABA. Thus the suppression of I GABA by GlyR activation was significantly larger than inhibition of I Gly by GABA A R activation (p Ͻ 0.01). Fig. 4C summarizes the percent reduction in the mean amplitude of I GABA and I Gly when different prepulse concentrations of glycine and GABA were used, respectively. Glycine became effective in cross-inhibition at concentrations Ն 30 M, whereas GABA was effective at 100 M. This asymmetric cross-inhibition between GABA A R and GlyR did not depend on chloride fluxes or changes in E Cl during receptor activation. For E Cl close to 0 mV, cross-inhibition was observed at both V h ϭ Ϫ50 and ϩ50 mV (Fig. 5A). Preactivation of GlyR decreased I GABA to 37.4 Ϯ 2.5% of the control at V h ϭ Ϫ50 mV (n ϭ 12) and to 38.5 Ϯ 4.1% of control at V h ϭ ϩ50 mV (n ϭ 12). Similarly, preactivation of GABA A R decreased I Gly to 77.9 Ϯ 4.0% of the control at V h ϭ Ϫ50 mV (n ϭ 10) and to 79.5 Ϯ 4.5% of the control at V h ϭ ϩ50 mV (n ϭ 10). The differences between percent changes at two V h values were all not significant (p ϭ 0.1, paired t test). Reversal potentials of response evoked by successive applications of glycine and GABA at 1 mM were estimated by voltage ramp application during the steady-state phase of the evoked currents (Fig. 5B). As shown in Fig. 5C, successive glycine-and GABA-evoked responses had similar reversal potentials. During the glycine prepulse E Cl was Ϫ4.15 Ϯ 0.4 mV, whereas it wasϪ3.98 Ϯ 0.58 mV during successive GABA application (n ϭ 8).
Receptor Channel Activation Is Required for Cross-inhibition-The data shown in Fig. 4 suggest that cross-inhibition depends on the prepulse agonist concentration. However, it remained unclear whether specific receptor channel opening and thus changes in receptor conformation triggered by the prepulse application are required for asymmetric cross-inhibition. To address this issue, we have used specific competitive antagonists for GABA A R and GlyR, bicuculline and strychnine, respectively. Sequential applications of glycine-GABA at 1 mM (Fig. 6A), or GABA-glycine at 1 mM (Fig. 6B) were performed in the presence of 1 M strychnine and 30 M bicuculline, respectively, at a concentration that selectively inhibits GlyR or GABA A R completely (Fig. 3, C and D). As shown in Fig. 6, A and B, during sequential applications of transmitters, application of the antagonist known to compete with transmitter binding of the preactivated receptor prevented cross-inhibition. Fig.  6C summarizes the results from all experiments (n ϭ 9) by comparing the amplitudes of control currents (in the absence of prepulses and antagonists) and currents observed after prepulse application, in the absence or presence of either strychnine (for I GABA ) or bicuculline (I Gly ). We found that in the presence of bicuculline, I Gly observed after the GABA prepulse had the same amplitude as the control I Gly (2.151 Ϯ 0.243 versus 2.213 Ϯ 0.259 with or without bicuculline, p Ͼ 0.1). Similarly, in the presence of strychnine, I GABA observed after the glycine prepulse had the same amplitude as the control I GABA (1.615 Ϯ 0.226 versus 1.638 Ϯ 0.204 with or without strychnine, p Ͼ 0.1). These data indicate that cross-inhibition between GlyR and GABA A R cannot be explained solely by ligand binding and suggest that activation of receptor channel is required.
Kinetics of Cross-inhibition-Inhibition of I GABA by preactivation of GlyR was accompanied by significant changes in the time course of the desensitizing I GABA component. This was not the case for I Gly when GABA A R was preactivated. To examine potential changes in the desensitization of I Gly or I GABA during cross-inhibition, time constants of the desensitizing phase of the corresponding current were measured. The desensitization time course of I Gly evoked by 10-s applications of 1 mM glycine with or without GABA prepulses could be fitted by the sum of two exponential curves with time constants f and s . As shown in Fig. 7, preactivation of GlyR accelerated the desensitization of GABA A R (Fig. 7, A and B). In contrast, preactivation of GABA A R did not change GlyR desensitization kinetics significantly (Fig. 7C).
We have also examined whether cross-inhibition between GlyR and GABA A R was reversible and whether it requires simultaneous activation of the two amino acid receptors. As shown in Fig. 8A, we found that increasing the time interval between sequential applications of glycine and GABA at 1 mM (or vice versa) resulted in a progressively reduced cross-inhibition. The effect of time interval between successive applications of agonist on cross-inhibition was analyzed on 12 and 14 cells for I GABA and I Gly inhibition, respectively. Half-inhibition of I GABA was obtained at interpulse intervals close to 41 s (Fig.  8B 1 ). Cross-inhibition cannot be evoked for I GABA for interpulse intervals Ն55 s. The time interval for I Gly inhibition evoked by GABA prepulses at which no cross-talk occurred was shortened (Ն22.5). Half-inhibition of I Gly occurred at time intervals close to 14 s (Fig. 8B 2 ).
Inhibition of I GABA by GlyR Activation Depends on Phosphatase 2B-It is known that phosphatase 2B (calcineurin) activity can inhibit GABA A R activation (26 -28) and modulate GABA A R desensitization kinetics (29,30). We therefore asked whether phosphatase activity might be involved in the cross-inhibition between GABA A R and GlyR. We first examined the effect of the phosphatase 2B inhibitor CSPN on the asymmetric cross-inhibition elicited by sequential application of 1 mM glycine and GABA at 1-s intervals. When loaded into neurons via the patch electrode, 500 nM CSPN prevented I GABA inhibition by preactivation of GlyR. In contrast, CSPN had no significant effect on the inhibition of I Gly evoked by GABA A R activation (Fig. 9, A1 and B). When glycine and GABA were coapplied, 500 nM CSPN blocked cross-talk by 84.6 Ϯ 5.2% (n ϭ 9).
Loading the neuron with 1 M okadaic acid at a concentration that specifically blocks phosphatase 1 and 2A (31) did not prevent I GABA inhibition by preapplication of glycine or I Gly inhibition by preapplication of GABA. In contrast, loading the neuron with a concentration of OA (5 M) known to inhibit phosphatase 2B (30) prevented I GABA inhibition by GlyR pre-activation ( Fig. 9, A2, A3, and B). OA became nonspecific at concentration higher than 1 M and mimics the effect of CSPN. Similar to that found for CSPN, 5 M OA had no significant effect on I Gly inhibition evoked by GABA prepulses. Because phosphatase 2B activation may depend on intracellular Ca 2ϩ concentration, we have examined the effect of buffering intracellular Ca 2ϩ with BAPTA. When the neurons were loaded with 15 mM BAPTA via the recording pipette, I GABA was decreased to 39.7 Ϯ 5.3% of its control value, whereas I Gly was reduced to 81.2 Ϯ 4.5% of the control following the glycine or GABA prepulse, respectively (Fig. 9, A4 and B). These values did not differ significantly from the cross-inhibition obtained in control conditions (p Ͼ 0.1; unpaired t test). Thus the results indicate that changes in intracellular Ca 2ϩ were not directly involved in the cross-inhibition. Inhibition of I GABA , but not of I Gly , depends on phosphatase 2B activity, whereas phosphatase 1 and 2A did not appear to be involved.
The level of substrate phosphorylation depends on the action of kinases and phosphatases in the neuronal cytoplasm. Prevention of cross-inhibition by phosphatase 2B inhibition may result from a shift in balance in favor of phosphorylation of the substrate by protein kinases. We first tested this hypothesis by loading the neuron with ATP␥S, which facilitates protein phosphorylation by donating a thiophosphate group in a kinasemediated reaction that resists hydrolysis by phosphatases (32). As shown in Fig. 10 time after the onset of whole cell configuration and completely abolished the inhibition within 15 min. The effect is unlikely to result from the loading of Li ϩ because loading of LiCl (up to 20 mM) did not affect the cross-inhibition (data not shown; n ϭ 4). In contrast to the effect on I GABA cross-inhibition, I Gly inhibition evoked by GABA prepulse application was not affected by loading of ATP␥S (Fig. 10).
To confirm further the role of protein phosphorylation, we analyzed the effect of 5 M staurosporine, a nonselective protein kinase inhibitor, on the cross-inhibition between GABA A R and GlyR (Fig. 11). If changes in GABA A R activation properties resulted from changes in the balance between phosphorylation and dephosphorylation processes, protein kinase inhibition should render cross-inhibition irreversible. The protocol used to determine time interval dependence of cross-inhibition in the presence of staurosporine was identical to that described in Fig. 8. In control conditions, recovery from cross-inhibition occurred in less than 50 s (Fig. 8). However, staurosporine evoked a rundown of I GABA when applied alone. This was not the case for I Gly . A 50% rundown of I GABA was found after 40 -50 min. To overcome this problem, the effects of staurosporine on cross-inhibition were tested when I GABA amplitude had declined to 50%. Change in GABA response amplitude evoked by glycine preapplication in the presence of staurosporine was compared with change in GABA response amplitude with time in the presence of staurosporine but without preapplication of glycine. As shown in Fig. 11B, I GABA was only slightly decreased 100 s after attaining 50% rundown. Staurosporine cannot block the inhibition of I GABA or I Gly evoked by preactivation of GlyR or GABA A R, respectively. However, in these conditions we observed no recovery from I GABA inhibition up to 2 min after evoking cross-inhibition. In contrast, staurosporine had no effect on the recovery of I Gly (Fig. 11B). These results indicate that kinase activity is needed for GABA A R but not for GlyR to recover from cross-inhibition. Interestingly, specific antagonists of protein kinase A (H89, 10 M, n ϭ 6 cells), protein kinase C (GF 109203X, 3 M, n ϭ 6 cells), tyrosine kinase (genistein, 5 M, n ϭ 10 cells) or calmodulin-dependent protein kinase II (KN93, 10 M, n ϭ 10 cells) had no effect on cross-inhibition and its recovery. However, applying a mixture of all four kinase inhibitors mimicked the staurosporine effect on the recovery from cross-inhibition. DISCUSSION In this study, we have demonstrated that activation of GABA A R and GlyR in acutely dissociated rat spinal dorsal horn neurons can induce specific asymmetric cross-inhibition of currents induced by GABA and glycine. Our results reveal a novel form of interaction between signal events mediated by two distinct anionic channels and show that specific involvement of intracellular phosphorylation pathways triggered by GlyR activation underlies the asymmetry in the cross-inhibition between the GABA A R and GlyR.
Cross-inhibition Depends on Intracellular GABA A R and GlyR Signaling-Although negative cross-talk between GABA A R and GlyR has been suspected previously (20, 21, 33), the underlying mechanisms remain controversial. Trombley et al. (20) proposed that the interactions between I GABA and I Gly in the olfactory bulb reflect direct actions of GABA on GlyR or glycine on GABA A R. This is not the case in rat spinal dorsal horn neurons. In this spinal cord preparation, we found a small minority of neurons lacking either glycine or GABA responses. The currents evoked by GABA and glycine were not affected by prepulse of glycine and GABA, respectively (data not shown). Similar findings were reported for acutely isolated hippocampal neurons (21). In the present study, we have shown that the specific competitive antagonist bicuculline, when applied at a concentration that blocked 100% of I GABA , did not alter I Gly . Similarly, I GABA was not modified by strychnine when applied at a concentration that totally suppressed I Gly . Furthermore, the inhibition of I GABA by the application of bicuculline prevented the inhibition of I Gly by preapplication of GABA. At a prepulse concentration of 1 mM, there is no direct effect of GABA and glycine on GlyR and GABA A R, respectively. Likewise, I GABA was not depressed by prepulse application of glycine in the presence of strychnine. These results indicate that activation of GABA A and glycine receptor channels is required for the cross-inhibition.
What downstream signaling events associated with the receptor channel activation are involved in the cross-inhibition? It is unlikely that alterations in the Cl Ϫ gradient resulting from channel activation mediate the cross-inhibition, as proposed previously by Grassi (33). We found that the inhibition remained when I Gly and I GABA were changed from inward to outward, indicating that the movement of Cl Ϫ is not critical for the inhibition. Furthermore, measurements of agonist-induced Cl Ϫ currents with the application of a voltage ramp also showed no change in the reversal potential for the Cl Ϫ current (34). These data indicate that cross-inhibition is a receptormediated event unrelated to the Cl Ϫ flux across the membrane.
Asymmetry in Cross-inhibition-When GABA and glycine were coapplied, the Cl Ϫ current was smaller than the sum of the two individual currents evoked by the application of each agonist. The occlusion level depended on the concentration of the two agonists, and it was nearly maximal when the concentration was Ն30 M. Thus, the intensity of the cross-inhibition between GABA A R and GlyR might be determined by the number of activated receptors, as proposed previously for the interactions between P2X and GABA A R (7).
To determine the respective contribution of the two agonists to the cross-inhibition, we examined the responses evoked by sequential application of these two compounds. Preapplication of glycine more strongly inhibited I GABA than preapplication of GABA did to I Gly . The level of current inhibition depended on the concentration of the prepulse agonist, in a manner consistent with that found when GABA and glycine were coapplied. Furthermore, the effect of prepulse glycine on I GABA is accompanied by a decrease in the desensitization time constants of the current response. This was not the case for the effect of GABA prepulse on I Gly . Because the effect of desensitization was observed during successive applications of the two agonists, the changes in GABA A R desensitization kinetics by GlyR activation remain even when GlyR is fully deactivated. This effect on the nonactivated GABA A R appears to persist for more than 45 s after the end of the application of glycine. This effect cannot be accounted for simply by cross-desensitization between these two receptors (35) because only GABA A R desensitization properties were modified. These results are also in sharp contrast to the previous reports on the inhibitory crosstalk between ionotropic receptors. No changes in receptor desensitization properties were observed during cross-inhibition between GABA A R and P2X (7) or between P2X and nicotinic acetylcholine receptors (2).
Cross-inhibition of GABA A R by Glycine Depends on Dephosphorylation-Our results indicate that inhibition of I GABA by preapplied glycine involves phosphatase 2B activity, whereas its recovery requires protein kinase activity. Our findings suggest the following model of cross-inhibition: Glycine preapplication results in dephosphorylation by phosphatase 2B and inhibition of I GABA , whereas recovery from inhibition depends on rephosphorylation of GABA A R or of associated proteins. Activation of phosphatase 2B requires the activation of GlyR, but rephosphorylation of GABA A R or of an associated protein does not require the activation of either GlyR or GABA A R. Our observations also indicate that in the absence of glycine prepulse, the balance between basal phosphorylation/dephosphorylation results in stable kinetic properties of the GABA A R during repetitive applications of GABA.
There are three potential mechanisms by which GlyR activation may evoke dephosphorylation by phosphatase 2B. First, activated GlyR may directly activate phosphatase 2B. To our knowledge, there is no existing evidence in support of this FIG. 11. Inhibition of I GABA by GlyR activation became irreversible after the inhibition of protein kinases. A and B, illustrations of the experimental protocols. A, in the presence of staurosporine (a nonspecific protein kinase), the I GABA decreased with time. To overcome this problem paired application of glycine and GABA was performed when I GABA reached 50% of the control response amplitude (measured at the beginning of the recording). The response amplitudes were then compared with currents recorded on different cells without paired application at the same latency after the beginning on the recordings (empty squares in C). In the presence of staurosporine preapplication of glycine can still inhibit I GABA , but in this case inhibitory cross-talk became irreversible for interpulse intervals up to 2 min. B, in the presence of staurosporine the preapplication of GABA can still inhibit I GABA , but in this case inhibitory cross-talk was still reversible. C, normalized amplitude of I GABA and I Gly after glycine or GABA prepulse recorded at various time intervals (n ϭ 6 -12). Empty squares are normalized control amplitude values of I GABA without glycine prepulse. possibility. Alternatively, phosphatase 2B is activated indirectly by elevation of the intracellular Ca 2ϩ concentration (27). Although activation of immature GlyR could elevate the intracellular Ca 2ϩ concentration of neurons (36), it is unlikely that glycine application evoked a Ca 2ϩ increase in the present neuronal preparation obtained from mature animal in the voltage clamp mode. Moreover, loading the neuron with BAPTA, a high affinity calcium chelator, did not modify the cross-inhibition between the two receptors. Thus the activation of phosphatase 2B by glycine prepulse does not seem to be mediated by a calcium-dependent process. Finally, GlyR activation may directly change GABA A R conformation through receptor-receptor interaction as proposed previously for some ionotropic receptors as for P2X2 and 5-HT3 receptors (37). This could result in the exposure of phosphorylation sites of GABA A R favorable to phosphatase 2B and/or unfavorable to protein kinase binding.
A large fraction of phosphatase 2B is known to be associated with the plasma membrane (38), a condition favorable for its rapid modulation of membrane receptors. Effectively, phosphatase 2B can modulate GABA A R desensitization kinetics (29) even during outside-out recordings (30). The slow recovery from cross-inhibition may reflect a slower process of rephosphorylation by cytoplasmic protein kinases. However, such a mechanism would require that GlyR is closely associated with GABA A R. Indeed, GABA A R and GlyR are colocalized at inhibitory synapses in the spinal cord and brain stem (10,39), and it has been demonstrated recently that GABA and glycine can be coreleased from inhibitory synapses in rat sacral dorsal commissural nucleus (11).
Cross-inhibition of GlyR Function by GABA-GABA preapplication did not change GlyR desensitization kinetics, its effect on current amplitude was less pronounced compared with the effect of glycine on I GABA , and the recovery from inhibition was faster for I Gly than I GABA . The effects of GlyR subunit phosphorylation on receptor function remain controversial, and they largely depend on subunit combination (15). Our results indicate that inhibition of GlyR activity by GABA is unlikely to result from receptor subunit dephosphorylation or phosphorylation processes because inhibition of phosphatases or protein kinases did not alter the GABA effect on I Gly . The most likely mechanism for the inhibition of I Gly by GABA A R activation is through direct coupling between these two receptors. However, we cannot exclude the possibility that other unknown cytoplasmic signaling molecules may mediate the functional interaction between these two receptors.
Physiological Significance of the Cross-inhibition-GABA and glycine can be colocalized in the same presynaptic terminals (40) and can be coreleased in several brain regions (9 -13), and their associated receptors are colocalized in the postsynaptic density (39), thus providing the physiological conditions for the cross-talk between GABA A R and GlyR. A recent study demonstrating that GABA and glycine interact at the synaptic level is in favor of such a speculation (11). The cross-inhibition of these two major inhibitory neurotransmitter systems could have important functional implications because it may represent an alternative means of limiting excessive inhibition at the synapses where GABA and glycine act as cotransmitters. In particular, GlyR activation and the subsequent phosphatase 2B-dependent GABA A R dephosphorylation would influence the shape and amplitude of GABAergic postsynaptic currents by speeding up GABA A R desensitization kinetics. Indeed, GABA A R desensitization is fast enough to control the time course of GABA-evoked synaptic events (41). Interestingly, a recent work has demonstrated that activity-dependent physical and functional interaction between phosphatase 2B and GABA A R is necessary and sufficient for inducing long term depression at hippocampal CA1 inhibitory synapses (42). Although corelease of GABA and glycine has not yet been detected in the hippocampus, a functional cross-inhibition has been shown for hippocampal GABA A R and GlyR (21). Whether or not the phosphatase 2B-dependent cross-inhibition between GABA A R and GlyR observed in the present study could regulate the synaptic transmission and plasticity remains to be established. Finally, because both GABA A R and GlyR are targets for many clinical therapeutics, it would be of interest to know whether and how drugs acting at one receptor may shift the synaptic influence toward the other transmitter.