Abstract
Regional distinctions in GABA type A (GABAA) miniature IPSC responses are thought to be determined by postsynaptic receptor composition. The kinetics of receptor activation and deactivation were studied using rapid exchange (100 μs) of GABA at excised patches containing recombinant (α1β1γ2 or α2β1γ2) and native (cortical) GABAA receptors.
Receptors activated by brief (< 1 ms) pulses of GABA demonstrated a characteristic current response, hereby referred to as the ‘receptor system response’. System response properties included agonist concentration-dependent peak amplitudes and concentration-independent maximal rates of activation and deactivation. Receptor subtypes were characterized functionally and phenotyped using the system response characteristics.
System responses obtained for α1β1γ2 receptors exhibited a single phenotype while α2β1γ2 receptors exhibited either a predominant slow deactivation (type I) or a relatively infrequent faster (type II) phenotype. Receptor system responses of α2β1γ2 receptors reached peak currents twice as fast as those of α1β1γ2 receptors (0.5 versus 1.0 ms) but decayed 2 or 6 times more slowly (τlong of ∼190 and 62 ms for type I and II α2β1γ2, and ∼34 ms for α1β1γ2 receptors).
Receptor system responses from cultured fetal mouse cortical neurons could be statistically separated and classified into five major types with little intragroup variability, primarily based on variations in the current deactivation phases.
Receptors subjected to pharmacological modulation exhibited alterations in system response properties consistent with known mechanisms of action, such that distinctions between binding and gating modulations were possible.
Brief agonist exposure places limits on receptor activation and deactivation response kinetics. Consequently, receptor system responses may be used to characterize and functionally phenotype an excised patch receptor population. Furthermore, since synaptic exposure to transmitter is postulated to be similarly brief, IPSC kinetics may reflect a functional fingerprint of synaptic receptors.
Results from studies using in situ hybridization, immunocytochemistry and electrophysiological techniques have indicated regional variability in receptor affinity and pharmacology, and have provided evidence that a diversity of ligand-gated ion channels exists in vivo and in vitro. The extent of this diversity is unknown, especially for the amino acid receptors, which are major sources of both excitatory and inhibitory control in the central nervous system (see Roberts, 1986, for review). Given the possibilities of different combinations of subunit isoforms, subunit isoform splice variants and post-transcriptional modifications, there are a vast number of potential receptor subtypes (see Whiting et al. 1995, for review). Heterologous expression of cloned receptors has provided evidence that subunit composition can play an important role in the functional characteristics of the resultant receptor ion channel subtype (Levitan et al. 1988; Pritchett et al. 1989; Verdoorn et al. 1990; Wafford et al. 1993). Thus, there exists the potential for an overwhelming complexity of functional receptor subtypes that may be difficult to separate and study individually.
Whole-cell and single channel electrophysiological methods have been used to identify and classify receptor subtypes based upon functional ion channel properties. They have provided considerable information at least about agonist affinity, pharmacological sensitivity, ion channel conductance and single channel kinetics. However, receptor channels are generally studied at or near equilibrium agonist conditions that do not reflect the brief, transient agonist exposure that occurs at the synapse. Additionally, these approaches are time intensive and thus multiple agonist or drug concentrations frequently cannot be completed in the same cell or patch recording. Thus, efforts to identify and study functional characteristics of receptors isolated in a particular recording have been limited.
We used rapid ligand exchange techniques to characterize directly intrinsic kinetic properties of receptors in excised patches. Rapid exposure of an excised multichannel patch to a brief pulse of agonist can evoke a transient current response defined here as the ‘receptor system response’. Computer modelling indicated that similar to the engineering concept of the impulse system response, stimulation of a ligand-gated receptor with a brief pulse of agonist (‘impulse’ input) can produce a current response (output response) that defines the system in terms of its frequency properties. The duration of ligand exposure defining the ‘impulse’ is dependent upon receptor properties (see below). The limit can be estimated as a duration sufficient for ligand binding of some receptors but more brief than the time required for maximal channel activation (approximately 80 % of the current rise time; see below). Under these conditions, the kinetics of the receptor system response are uniquely determined by the maximal activation and deactivation properties of the receptor population that are undistorted by slow ligand application or by slow exit from receptor desensitization. Characteristics of the receptor system response allow determination of two important intrinsic gating properties of ligand-gated receptors: maximal ion channel activation rate (approximately equal to maximal opening rate) and receptor deactivation rate. We applied this approach to study structure-function relationships of the GABA type A (GABAA) receptor.
In this study, computer-modelling techniques were used to evaluate the properties of transient and equilibrium receptor activation. The results indicate that for at least some ligand-gated ion channels, maximal activation and deactivation kinetic rates can be determined by simulating synaptic conditions of agonist exposure to produce a receptor system response. It was also found that this information is sufficient for functional receptor subtype identification or characterization. Data are presented showing the consistency in receptor subtype between transfected cells and between some cultured neurons. This suggested that the receptor system response may provide a means for functional phenotyping of native receptor subtypes. Finally, these results are compared with known miniature IPSC properties to elucidate synaptic determinants of these characteristic responses to transient agonist exposure.
METHODS
Computer modelling
Computer-simulated macroscopic currents were generated and fitted using available software (SCoP, Simulation Resources). Kinetic schemes were based on previous transient and steady-state kinetic models (Twyman et al. 1990; Maconochie et al. 1994; Jones & Westbrook, 1995; Lavoie & Twyman, 1996; Lavoie et al. 1997). Models were fitted to experimental data using internal optimization functions which calculated the least square error following variations in specific parameters.
Cell line transfection
Transformed human embryonic kidney cells (HEK 293 cell line, American Type Culture Collection CRL 1573) were maintained at 5 % CO2 in minimal essential medium (MEM) with added glutamine and glucose, and 10 % fetal calf serum. Standard Ca2PO4 precipitation techniques (Chen & Okayama, 1987) were used to transfect HEK 293 cells with plasmid (pCIS2; gift of D. B. Pritchett; Pritchett et al. 1989) containing either human α1β1 or α2β1 (on a single plasmid) and γ2short cDNAs (2 : 1 ratio, 3 μg). Recordings were obtained at least 48 h after transfection.
Cell culture
Cortical neurons were obtained by anaesthetizing timed pregnant mice with CO2. The mice were killed by cervical dislocation after removing the 15–18 day old fetuses. The dorsal portions of the fetal skulls were removed and the exposed cortex harvested. Cortical neurons were dissociated from fetal mouse cortex and maintained at 5 % CO2 in MEM (with added glutamine and glucose) and 10 % fetal bovine serum for 24 h, at which time the fetal bovine serum was replaced with 10 % horse serum. After 4–6 days, cultures were treated with cytosine arabinoside (ARA-C, 10 nM, Sigma) for 24 h to inhibit growth of non-neuronal cells. Neuronal cultures were used between 2 and 5 weeks in culture.
Recording solutions
Prior to recording, culture media was exchanged for external salt solution which consisted of the following (mM): 142 NaCl, 1.5 KCl, 1 CaCl2, 1 MgCl2, 10 glucose, 30 sucrose, and 10 Na-Hepes at pH 7.4, adjusted to 320 mosmol l−1 with sucrose. Micropipette recording electrodes (borosilicate glass, 3–5 MΩ) were filled with internal salt solution consisting of the following (mM): 140 CsCl, 4 MgCl2, 10 Na-Hepes and 5 EGTA, at pH 7.4, adjusted to 290 mosmol l−1 with water. This combination of external and internal salt solutions resulted in a chloride equilibrium potential (ECl) of about 0 mV. All recordings were performed at room temperature (20–23°C).
GABA and drug solutions
A 1 M GABA (Sigma) stock solution in distilled water was prepared prior to experiments and frozen in 0.5 ml aliquots. Additionally, 10 mM stocks of sodium pentobarbital (Sigma, in distilled water) and diazepam (Sigma, in DMSO) were prepared, frozen in 0.5 ml aliquots and protected from light. The stock solution of GABA was diluted in external solution to a final concentration of 0.01–10 mM on the day of the experiment. Solutions of pentobarbital (50 μM) and diazepam (25 nM) were prepared in external solution less than 4 h prior to use.
Rapid application techniques
The essential features of rapid ligand exchange have been described previously (Lavoie & Twyman, 1996; Lavoie et al. 1997) and similar techniques have been used by others (see Jonas, 1995, for review). Briefly, a double lumen glass tube (thin septum theta tube pulled to a tip diameter of 50–100 μm; R and D Scientific Glass Co.) was used to provide adjacent streams of known solution concentrations with a small (∼4–6 μm) interface between solutions. Solution exchange was achieved via activation of a piezoelectric transducer (Burleigh Instruments) which translated the solution interface ∼40 μm laterally to expose the electrode tip to experimental solution. Exchange time was measured at the open electrode tip by switching between solutions of differing osmolalities (∼100 μs exchange shown for a step and pulse application, Fig. 3A and C). Exchange time at an excised patch containing open glycinergic chloride channels was measured between solutions of different chloride concentrations (data not shown). Solution exchange was reproducible and complete within 150 μs, allowing transient ligand applications to an excised patch of < 1 ms duration. These results are similar to those for previous reports where solution exchange at an excised patch was slower than that for an open electrode by approximately a factor of two (Colquhoun et al. 1992; Maconochie et al. 1994). Open electrode tip potentials were measured at the conclusion of each patch experiment to confirm the adequacy of the exchange rates. Excised outside-out multichannel patches (10–200 channels) were placed in the stream of control solution and serially (1–1.5 s intervals, 10 total) exposed to brief pulses (800 μs to 1 ms) or prolonged steps (> 50 ms) of experimental solution.
Figure 3. Activation/deactivation kinetics of recombinant GABAA receptors following pulse and step applications of GABA.
A, current responses of α1β1γ2 recombinant receptors in different patches to step (50 ms to 2 s) applications of 10 μM, 100 μM, 1 mM, or 10 mM GABA were normalized to compare rise times to peak (peak = original peak amplitude for a single application, 20–200 pA; n = 40 averaged traces). Under these conditions receptor activation rates were dependent on GABA concentration. Peak current response to 10 μM GABA was extremely slow and off scale. Noise in averaged traces was predominantly due to stochastic variability between individual responses. Note the tip potential recording (top) indicating the period of solution exchange. B, comparison of normalized current responses from two different α1β1γ2 patches during a pulse (peak = 40 pA, n = 40 averaged responses) and step (peak = 20 pA, n = 40 averaged responses) application to a saturating concentration of GABA was consistent with modelling predictions that activation rate (inset) in the receptor system response was characterized by the channel maximal activation rate. C, maximal activation rates of system responses following pulse applications of 100 μM (peak = 20 pA, n = 10 averaged responses) or 10 mM (peak = 40 pA, n = 40 averaged responses) GABA indicated that activation (inset) and deactivation rates were independent of agonist concentration. The tip potential recording indicates the period of solution exchange (top).
Current recording
Currents were recorded using an Axopatch 200A amplifier (Axon Instruments) with an interposed 8-pole Bessel filter (low pass 2–10 kHz, Frequency Devices). Filter bandwidth was chosen to maximize the resolution of current onset. Data were digitized at 10 times the filter bandwidth (20–100 kHz; pCLAMP 6.0 software, Axon Instruments) and simultaneously stored on a digital audio tape (75ES digital audio tape deck modified to 0–20 kHz, 14-bit resolution, 44 kHz sampling frequency, SONY) and a strip chart recorder (Gould Inc.). Long duration current responses could be re-digitized at a later date to resolve fully long duration deactivation currents.
Current analysis
Ensemble averages of sequential current responses of the same patch aligned to the stimulus pulse were obtained using locally written software (R. E. T.). Rise time was calculated as the time elapsed between 10 and 90 % of the peak current amplitude and is reported as the mean ±s.e.m. Comparisons between means were made using Student's t test. Least squares multiexponential curve fitting of the current decay phase was performed using programs described previously (Twyman et al. 1990). Error ranges for the estimates were calculated using maximum likelihood ranges (m = 2) which corresponded to about a 95 % confidence interval. The number of significant exponential components was determined by fitting with increasing numbers of exponentials until the maximum likelihood estimate could no longer be greatly improved by addition of more exponential components (log likelihood ratio, LLR > 4; McManus & Magleby, 1989; Twyman et al. 1990).
Clustering analysis
Methods for statistical comparison between two individual waveforms (i.e. two individual patch responses) with different noise levels have not been previously established. Consequently, we adapted pattern recognition techniques typically used for classification of EEG or ECG spike waves to compare, sort and group our individual receptor system responses (averaged from at least 10 responses in a single patch). After peak responses were normalized and initiation of current onsets aligned in time (current amplitude > 2 standard deviations from baseline noise), two different techniques were employed which utilized the pattern recognition engineering concept of distance between groups. The first, CBD (city block distance or, more simply, residual), was calculated as the sum of the absolute difference between current amplitudes at each data point for two different response traces. The second, SED (squared Euclidean distance), was calculated as the sum of the squared difference between response amplitudes at each data point. SED was more useful for evaluation of traces which included substantial baseline samples within the length of the recording (no open channel current). Both distance measurements were calculated over the entire length of the receptor system response (20 ms prior to current onset, 800 ms total sample length). Although both activation and deactivation distances were measured by these techniques, the relatively greater area of deactivation determined that distance ratios primarily distinguished between responses with different deactivation rates. Furthermore, both SED and CBD easily separated responses with different kinetics, particularly near the peak. Differences in activation could be determined by application of these methods to current onsets at a higher time resolution, but onset kinetics are more easily distinguished by standard 10–90 % rise times or single exponential fits of the current rising phase. CBD/ and SED/ ratios were used to group responses (described below).
In image analysis applications, 95 % confidence is attained with CBD distances greater than 5 times or SED distances greater than 7 times the squared error () of the average noise present in the two responses (Dinning & Sanderson, 1981; Wheeler & Heetderks, 1982; Bankman et al. 1993). These CBD and SED ratios are often determined empirically and are greatly dependent on the signal to noise ratio and the type of noise present. The current responses reported here had extremely large signal to noise ratios compared with those frequently encountered in image analysis.
The squared error due to noise in individual (single patch) current responses was determined after removing the slower components of the current response by subtracting a piece-wise linear spline fit. The noise mean squared error was determined for each trace and multiplied by the number of points used in the spline fit to obtain the squared error for each response. The average of the squared error of the two responses, , was used for calculating distance ratios. Signal to noise ratios were sufficiently similar within a patch that averaging did not degrade these ratios. Accurate calculation of the noise component could be achieved by beginning the piece-wise fit just from the peak of the current response (piece-wise time points in ms: 7.68, 10.24, 12.80, 15.36, 20.48, 25.60, 30.72, 35.84, 40.96, 46.08, 51.20, 56.32, 61.44, 81.92 and 102.40, and every 20.48 ms until the end of the recording). This procedure was used to estimate the noise contribution during the entire length of the sample because noise variability in the recording was greater immediately following agonist application due to opening and closing of channels. Resolution of discrete single channel current levels in patches containing few channels considerably degraded the accurate estimation of receptor system response kinetics. Thus, patch responses containing at least 10 channels are included in the analysis. Autocorrelation analysis of noise present in the current responses revealed no intrinsic correlation, and that noise amplitude was of Gaussian distribution. Consequently, addition of white noise was not necessary to determine CBD and SED distance ratios (Bankman et al. 1993).
The sensitivity of these distance methods was tested using simulated macroscopic current data (SCoP, Simulation Resources). The simple models included two sequential binding sites, multiple closed states and a single open state (Fig. 1). Incorporated kinetic transition rates were based on previous models (Macdonald et al. 1989; Twyman et al. 1990; Jones & Westbrook, 1995; Lavoie & Twyman, 1996; Lavoie et al. 1997). Simulated responses were tested with the addition of Gaussian-distributed random noise and various alterations in kinetic transition rates. It was empirically determined that current responses from two different models with only a 25 % difference in a single kinetic rate constant could be distinguished when both CBD/ and SED/ ratios were greater than 3 and 28, respectively (detailed in Results).
Figure 1. Activation/deactivation characteristics of simulated responses to pulse and step agonist conditions.
A simple two binding step kinetic model with positive co-operativity is illustrated, where C represents the closed states of the channel, O represents a single open state and D represents a desensitized state. Kinetic transitions are also shown. Parameters are as follows for the simulations shown below the model: agonist concentration, [A]= 10 μM to 10 mM; association constants, k+1 = 1.5e6 M−1 s−1, k+2 = 1e7 M−1 s−1; dissociation constants, k−1, k−2 = 1200 s−1; gating open transition, β= 2000 s−1; gating closed transition, α= 260 s−1; desensitization, δ= 330 s−1; resensitization, ρ= 80 s−1. Sample points in current responses were reduced by an order of 8 in the graphical representation. A, simulated current response following 1 ms pulse applications of agonist. Peak current amplitudes were agonist concentration dependent (top), but normalization of responses to peak current (bottom) indicated that activation (10–90 % rise time, inset) and deactivation kinetics (10–90 % rise time) of the response were independent of agonist concentration. Note, the 10 μM GABA response was too small to resolve adequate peaks for normalization. B, current responses following 50 ms step applications of agonist allowed receptors sufficient time to reach equilibrium. Under these conditions, peak currents (top) and times to reach peak (normalized, bottom) were agonist concentration dependent. Normalization of responses to peak currents indicated that the deactivation kinetics were independent of agonist concentration (bottom). C, in the model with decreased resensitization rate (ρ= 1 s−1), serial pulse applications (1 ms duration, 10 Hz frequency) showed decremental peak amplitudes (continuous trace, top) due to accumulation of receptors in the desensitized state. Peak amplitudes measured from simulations with a faster resensitization (ρ= 80 s−1) are plotted as points below the continuous trace. Normalization of individual responses during either desensitization condition (bottom) indicated that desensitization does not affect receptor system response kinetics. D, comparisons between deactivation rates (normalized to onset of current decay) following pulse (continuous line) and step applications (dotted line) with a saturating concentration of agonist indicated that exit rates from desensitization following 50 ms agonist applications altered deactivation kinetics, but that differences were minimal and may not be resolvable under experimental conditions. E, normalized system responses are overlaid to illustrate the effect of altering desensitization rate in this simple kinetic model (ρ= 80 s−1). Time constants for deactivation were best fitted with a single time constant of 11 ms when δ= 0 and up to 100 s−1, and two time constants when δ= 500 s−1 (τshort = 7.75 ms, 0.66 amplitude, and τlong = 22.1 ms, 0.36 amplitude) and δ= 1000 s−1 (τshort = 6.55 ms, 0.63 amplitude, and τlong = 27.9 ms, 0.39 amplitude).
The above approach allowed the unbiased analytical determination that two waveforms (normalized to peak) were different. Current responses with indistinguishable kinetics were then grouped analytically according to relative distances, determined by CBD and SED ratios (above) using a clustering technique (originally written for analysis of action potential spikes by E. Maynard & R. Normann (Bioengineering Department, University of Utah), in Matlab, Mathworks). In this clustering program, distances between each response were calculated and the minimum distance ratio between two individual averaged responses was found. These two current responses were then averaged and distance ratios were calculated for the resultant average against the remaining traces. Responses were grouped until the distance ratio to other traces exceeded the threshold values for CBD and SED to noise ratios (3 and 28, respectively). Small differences in the time required to reach peak for responses within a group could cause a relatively minor ‘flattening’ of the average group response peak (Fig. 10), where the average group rise time is still determined by the 10–90 % measurement. Group membership for each method was compared for agreement by visual inspection. Curve fitting of decay time constants from peak amplitude to the end of the 800 ms sample was also used for comparison.
Figure 10. Different activation and deactivation phases for five major classes of neuronal receptor system response.
A, superposition of the normalized five major classes of cortical neuron receptor system responses at higher time resolution indicated that they differ primarily in deactivation kinetics. Receptor system responses of types III and IV were similar at this time scale, but were distinctly different between 150 and 250 ms as evident at low resolutions (Fig. 2). B, at expanded time scales, current onsets differ, for example a difference between group III and IV is evident. Pulse durations varied slightly between patches, but were verified to range between 0.6 and 1.0 ms using open electrode junction potential recording. C, normalized type I α2β1γ2 receptor response (peak = 750 pA, n = 10 averaged responses) and one response from a mouse cortical neuron (peak = 150 pA, n = 10 averaged responses) which did not belong to the major classes were analytically indistinguishable and are shown superimposed. Inset indicates the similarity of activation rates between a native (continuous line) and recombinant α2β1γ2 (dotted line) receptor system response.
RESULTS
Modelling predictions
GABA probably uses two binding sites to activate fully the GABAA receptor (Sakmann et al. 1983). Incorporating two positively co-operative agonist binding sites into a simple desensitizing kinetic model produced simulated macroscopic current responses to 800 μs pulse and 50 ms step applications which varied with agonist concentration (Fig. 1A and B, top). For brief pulse applications, peak current amplitudes were markedly dependent on agonist concentration. Although at higher time resolution the shape of the simulated current response onsets was dependent upon agonist concentration (i.e. sigmoidal and exponential for low and high concentrations, respectively), maximal activation rates measured as the 10–90 % rise time were independent of agonist concentration (Fig. 1A, bottom). Thus, maximal activation kinetic rates could be estimated from 10–90 % rise times at any ligand concentration sufficient to evoke a resolvable response.
Simulated responses to pulse applications of variable duration (0.2–500 ms) were evaluated in the simple model in Fig. 1 (data not shown). Loss of resolution of rise times at low agonist concentrations was not appreciable with pulse durations approaching 3 times (α+β)−1 s (or ∼3β−1 if β≫α with maximal opening rate =α+β; data not shown). Generally, loss of resolution occurred with agonist application durations longer than 2.2β−1 that approximated the 10–90 % rise time for the maximally activated receptor. The measured 10–90 % rise time was prolonged in direct proportion to pulse durations > 3β−1 at non-saturating GABA concentrations. Although modelled using relatively slowly binding receptors such as GABAA, these results indicate that pulse durations of ∼3β−1 should not cause significant distortion of activation or deactivation kinetics even at higher β transition rates (i.e. for glutamate, acetylcholine or glycine receptors).
The concentration independence of activation rates for simulations using the brief pulse protocol accurately predicted experimental results (see below). Rates for current activation were concentration dependent when receptors were activated by agonist application durations that approached equilibrium or steady-state conditions (Maconochie et al. 1994; Jones & Westbrook, 1995; Lavoie & Twyman, 1996). When application durations were relatively long, peak responses and rates of current onset to peak/steady-state levels were concentration dependent up to a maximal activation rate at saturating concentrations (Fig. 1B). Comparison between pulse and step applications at a saturating agonist concentration indicated that the maximal activation rate of the receptor population using either protocol was the same (data not shown).
Desensitization and run-down of ligand-gated receptors result in reduction of current over time with prolonged or repeated agonist exposure. Following a model incorporating a desensitized state with moderate entry (δ= 330 s−1) and minimal exit (ρ= 1 s−1) rates from doubly liganded receptors (Fig. 1; based on Jones & Westbrook, 1995), decremental peak currents following repetitive brief pulse agonist applications were predicted (Fig. 1C, top trace). Using exit rates from the desensitized state that more closely described the α1β1γ2 receptor data (ρ= 80 s−1; see below and Fig. 6), no decrement in peak amplitude was predicted (Fig. 1C, peak current amplitude denoted by filled circles). Normalization of the peak currents and superposition of the responses indicated that absorption of receptors into desensitized states had little impact on the kinetics of the pulse response (Fig. 1C, bottom trace). However, consistent with earlier reports (Jones & Westbrook, 1995), aligning current decay for brief pulse and prolonged step applications indicated that desensitization during prolonged agonist applications could alter the deactivation time course (Fig. 1D) even at these moderate rates. Therefore, the effects of desensitization on pulse response kinetics were simulated. The results demonstrated minimal impact on deactivation kinetics for desensitization rates of < 500 s−1 (δ≪β; Fig. 1E), while the impact of resensitization rate (ρ < 100 s−1) was even less (data not shown). Thus, modelling results indicate that the pulse response provides an accurate estimate of maximal activation and deactivation kinetics under these conditions. Notably, Jones & Westbrook (1995) determined native hippocampal receptor kinetics could be described using two desensitized states, the fastest of which had an entry rate > 500 s−1, which corresponds to a time constant of 1–2 ms. This fast desensitization would permit more receptors to enter desensitized states during the 1 ms pulse, and subsequent exit rates from these states could provide an additional or prolong an existing time constant of deactivation (Fig. 1E). Similarly fast desensitization was not demonstrated for recombinant receptors.
Figure 6. Simulated responses from a simple kinetic model closely approximate experimental applications of GABA to recombinant α1β1γ2 receptors.
A simple positively co-operative kinetic model (Fig. 1; δ= 330 s−1, ρ= 80 s−1) was fitted to concentration-dependent responses of recombinant α1β1γ2 receptors to pulse and step GABA applications. Current responses were normalized for comparison. A, concentration-dependent responses to 50 ms steps of GABA (100 μM to 10 mM), overlaid with responses generated from the simple kinetic model described above. Note, the noise variance of the 100 μM response was smaller than the width of the fitted curve. B, overlaying isolated decay currents from the three responses in A demonstrates deactivation rates of recombinant α1β1γ2 receptors were not dependent on GABA concentration. C, the α1β1γ2 receptor system response overlaid with the system response (thick line) generated by the simple kinetic model. The same kinetic rates were used to generate both the pulse and step simulated responses. D, decay currents from pulse (10 mM, B) and step (100 μM, A) GABA applications aligned at the onset of deactivation and normalized for direct comparison. Deactivation kinetics of recombinant α1β1γ2 receptors were not visually different for the brief pulse and 50 ms step protocols.
Maximal activation and deactivation times were thus revealed by brief pulse applications using moderate or high concentrations of agonist. The kinetics of the current response resulted from synchronization of bound receptors that eventually opened. Receptors that did not bind agonist, and receptors that bound agonist then unbound without opening, did not contribute to channel currents. Those channels that opened were synchronized to a resolution within the pulse duration, provided that receptor activation and not ligand availability was the rate-limiting factor. If the unbinding rate (k−2) is considerably longer than the pulse duration, channel deactivation occurs without agonist rebinding since agonist has been washed away. The effects of desensitization on the deactivation time course are minimized if receptors have limited rates of entry into and subsequent exit from desensitized states. Thus, the current response from a brief agonist application can provide a fingerprint of the maximal activation and deactivation rates of the receptor. This characteristic response is hereafter referred to as the receptor system response.
For the simple model in Fig. 1, the maximal rate of macroscopic current onset was determined by the microscopic transition rate from the fully bound to the open state (β) and the closing rate of the channel (α). However, since β≫α, direct measurements of maximal macroscopic current onset were close approximations of maximal microscopic opening rate (β; Maconochie et al. 1994). Decay kinetics of the receptor system response were a function of the channel bursting properties determined by the gating transitions (α and β) and the rate of ligand unbinding for the second agonist molecule (k-2). Perturbations of these transition rates altered the onset and/or decay kinetics of the receptor system response, while alterations in other binding transitions (k+1, k+2 and k-1) affected only the amplitude of the peak response, but not the activation or deactivation kinetics (not shown). Deactivation phases were not altered by pulse duration in this simple, moderately desensitizing model (Fig. 1C, lower trace).
Differences in system response kinetic properties are predicted by alterations in transition rates in the model. Thus, system response differences between ligand-gated ion channels with different properties for open or burst durations, opening rates or ligand unbinding rates may be readily detectable. The sensitivity for analytical detection of differences in receptor system responses was evaluated by varying individual transition rates in the simple kinetic model in Fig. 1. The distance measurements CBD and SED were used to compare and quantitatively evaluate whether individual system responses were similar. Gaussian-distributed noise was added to simulated receptor system responses in which individual transition rates were altered by 5, 10, 25, 50 and 100 % above, and 5, 10, 25 and 50 % below the control values used in the model of Fig. 1. The ability to resolve differences between responses was dependent on the noise present, and thus a magnitude of noise equivalent to the root mean square noise typically found in experimental data was added to the responses. CBD and SED distance ratios were used as a quantitative measure to detect differences in traces (see Methods). Ratios greater than 3 CBD and 28 SED to the average noise squared error () of the two receptor system responses were empirically determined to be sufficient cut-off criteria for detection of a greater than 25 % change in transition rates α, β and k−2. For example, these measurements could distinguish between receptor system responses resulting from a closing rate (α) difference of at least 25 % (Fig. 2A and B). Thus, two receptor populations with 4 and 5 ms mean open durations could be differentiated by these criteria. Furthermore, two neuronal system responses that were visually distinct (Fig. 2C) had CBD and SED ratios that were consistent with the distinction criteria (CBD/ ratio = 14.0, SED/ ratio = 187.0). This method was less sensitive for detecting differences in k+1, k+2, k−1, δ and ρ.
Figure 2. Distance measurements analytically distinguish between receptor system responses with different kinetics.
Simulated responses to the simple dual agonist kinetic model in Fig. 1 were generated with either a 10 % decrease in the rate of channel closing, α (CBD/ ratio = 2.1, SED/ ratio = 6.3) (A), or a 25 % decrease in α (CBD/ ratio = 8.3, SED/ ratio = 56.7) (B). Random Gaussian-distributed noise was added to each simulation. C, receptor system responses of neuronal type III and type IV normalized to peak and superimposed. Analytically (CBD/ ratio = 14.0, SED/ ratio = 187.0), these two responses are different.
We compared the characteristics of the computer-simulated current responses with GABA currents recorded from recombinant receptors and native receptors.
Recombinant GABAA receptors
GABAA receptors composed of α1β1γ2 or α2β1γ2 subunits (cDNA transfection ratio 2 : 2 : 1) have distinguishable maximal activation and deactivation rates (Lavoie et al. 1997), and were therefore good candidates for experimental evaluation of receptor system responses. Activation rates of α1β1γ2 (Fig. 3A) and α2β1γ2 receptors (not shown, Lavoie et al. 1997) were concentration dependent and reached a plateau at saturating concentrations of GABA (> 1 mM). This result is similar to the concentration-dependent current onsets reported for cerebellar neurons following more prolonged GABA exposure (> 10 ms; Maconochie et al. 1994). Maximal current onset (measured as the time elapsed between 10 and 90 % of peak current and reflecting maximal activation rate) was 1.0 ms for α1β1γ2 receptors and 0.5 ms for α2β1γ2 receptors (see also Lavoie & Twyman, 1996; Lavoie et al. 1997). The maximal activation rate was not different between responses obtained using a step application with a saturating GABA concentration and the receptor system response obtained with a pulse application (shown for α1β1γ2, Fig. 3B). Although the number of activated receptors differed, the kinetics of the receptor system response were independent of agonist concentration (100 μM to 10 mM; shown for α1β1γ2, Fig. 3C). System responses for α1β1γ2- and α2β1γ2-containing receptors differed in both activation and deactivation rates, with α1β1γ2 receptors having a slower onset and faster decay (Fig 4 and Fig 5, see below).
Figure 4. Recombinant receptor system response kinetics are modulated by the α subunit isoform.
Normalized sample receptor system responses of patches from different cells containing α1β1γ2 (Aa) or α2β1γ2 (Ba and Ca) receptor populations were characteristic. For α1β1γ2 receptors, similarities of receptor system responses between cells and across transfection lots indicated a single functional population (Aa, peak = 20–200 pA, n > 10 averaged responses). For α2β1γ2 receptors, two consistent types of receptor system response were found (Ba, peak = 40–800 pA, n > 10 averaged responses; Ca, peak = 100–800 pA, n > 10 averaged responses). Ab, grouping by CBD and SED distance ratios indicated that receptor system responses can be classified as a single class in α1β1γ2 receptors, shown as the mean with s.e.m. error bars indicated at various individual points (n = 9 patches). Bb and Cb, for α2β1γ2 receptors, distance ratios classified two distinct subtypes, shown as type I (n = 17 patches) and II (n = 5 patches). Greater variability for α1β1γ2 receptor system responses resulted from greater numbers of responses with an overall lower signal to noise ratio (fewer active channels).
Figure 5. Receptor system response kinetics of α1β1γ2 and α2β1γ2 receptor populations are different.
Receptor system response group means for α1β1γ2 (mean peak = 40 pA, n = 40 averaged responses), α2β1γ2 type I (mean peak = 750 pA, n = 10 averaged responses) and α2β1γ2 type II (mean peak = 500 pA, n = 10 averaged responses) receptors are primarily distinguished by different deactivation kinetics. Responses were normalized and are also shown at higher time resolution for clarity (inset). Type I and II α2β1γ2 receptor system responses activate at the same rate, but α1β1γ2 receptors activate slower and deactivate faster.
Remarkably, system responses for α1β1γ2 receptors were characteristic and visually similar across cells in the same dish and across transfection lots (9 patches, Fig. 4Aa), although the number of activated α1β1γ2 receptors per patch was often relatively low, causing increased variability. Likewise, receptor system responses from α2β1γ2 cells were predominantly characterized by a single receptor system response (17/22 patches; Fig. 4Ba). However, some α2β1γ2 responses were similar to each other, but characterized by faster decay time constants (5/22 patches; Fig. 4Ca). CBD and SED distance measurements were used to test for similarities in the response profiles. Responses from α1β1γ2 receptors had CBD/ (2.15 ± 0.02, mean ±s.e.m.; n = 9 patches) and SED/ (14.3 ± 0.3) ratios that indicated that they were not different from each other. Similarly, the CBD/ (2.74 ± 0.05, n = 17) and SED/ (20.2 ± 1.5) ratios between the majority (17 of 22) of α2β1γ2 responses indicated that they were similar and indistinguishable. However, five α2β1γ2 responses were visually distinct for this group, but were similar to each other based upon CBD/ (2.47 ± 0.06) and SED/ (21.4 ± 1.5) ratios. Similar results were obtained when tested independently by allowing computer-processed cluster analysis on the entire mixed group of α1β1γ2 and α2β1γ2 receptor system responses. Clustering based upon minimization of the distance within an average set grouped all of the α1β1γ2 responses together (Fig. 4Ab) and separated two types (I and II) of α2β1γ2 receptor system responses (Fig. 4Bb and Cb).
The type I α2β1γ2 response was the most common α2β1γ2 response identified. The type I α2β1γ2 response was significantly different from the type II response with CBD/ and SED/ ratios of 5.75 ± 0.23 and 103.8 ± 13.6, respectively, between the two groups (Fig. 5). Deactivation kinetics were faster for type II α2β1γ2 receptors than for type I, but still somewhat slower than that found for α1β1γ2 receptors (Fig. 5). Interestingly, maximal activation rates for type I and II α2β1γ2 receptors were indistinguishable, but time constants for deactivation were different (Table 1). Response type did not correspond with peak amplitude (< 100–800 pA each), suggesting that receptor density was not responsible for these kinetic differences.
Table 1.
GABAA receptor system response properties
Deactivation | ||||
---|---|---|---|---|
Receptor | n | Activation 10–90% rise time(ms) | τshort(ms; Amp) | τlong(ms; Amp) |
α1β1γ2 | 9 | 1.0(± 0.06) | 9.4(+0.1, −0.1); 0.66 | 33.6(+0.3, −0.3); 0.34 |
α2β1γ2 type I | 17 | 0.5(± 0.06) | 35.5(+0.6, −0.5); 0.70 | 190(+0.7, −0.8); 0.30 |
α2β1γ2 type II | 5 | 0.5(± 0.06) | 16.8(+1.0, −0.5); 0.25 | 61.9(+1.0, −1.0); 0.75 |
Neuronal type I | 6 | 1.7(± 0.25) | 7.3(+0.4, −0.4); 0.56 | 125(+59, −57); 0.44 |
Neuronal type II | 6 | 1.7(± 0.25) | 14.5(+3.2, −0.7); 0.39 | 172(+13, −15); 0.61 |
Neuronal type III | 9 | 1.6 (± 0.19) | 13.7(+1.1, −1.1); 0.20 | 135(+14, −13); 0.80 |
Neuronal type IV | 7 | 1.8(± 0.18) | 28.4(+4.7, −1.3); 0.35 | 208(+17, −16); 0.65 |
Neuronal type V | 4 | 1.3(± 0.06) | 14.7(+2.6, −0.7); 0.55 | 166(+19, −22); 0.45 |
Current rise and decay parameters for receptor system responses evoked from recombinant and native receptors are listed. Three types of recombinant receptor system responses were identified and five major types of neuronal receptor system responses were classified. The number of patches analysed in each group is represented in column n. Rise time was measured as the mean time elapsed between 10 and 90 % of peak current (± S.E.M.). Current decay was fitted best to two exponentials and is reported as the time constant (± maximum log likelihood range) and the relative fractional amplitudes (Amp) of two time constants (τshort and τlong).
The simple kinetic model predicted experimental responses for α1β1γ2 receptors to 50 ms step applications of various GABA concentrations (Fig. 6). Although rates of activation and desensitization were concentration dependent (Figs 3A and 6A), deactivation rate was not dependent on agonist concentration (Fig. 6B). The same kinetic parameters closely predicted experimental α1β1γ2 system response data (Fig. 6C), and aligning the current decay phases from pulse and step applications indicated that deactivation rate also was independent of application duration or desensitization (Fig. 6D).
Peak currents of the α1β1γ2 and α2β1γ2 receptor responses had variable amplitudes which decreased little with serial brief pulse applications of 1 mM GABA, indicating the proportion of receptors remaining in desensitized states was small using this protocol (shown for α1β1γ2, Fig. 7A). Superposition of ensemble averages of earlier and later responses from serial brief applications revealed that the receptor system responses did not vary over time (Fig. 7B). Comparison of the first and last applications within a set also indicated consistent system response kinetics (data not shown). Step application of 1 mM GABA, however, produced a decay in evoked current during prolonged agonist exposure, perhaps indicating that some channels can enter a moderate duration desensitized state or an equilibrium closed state (Fig. 7C). For this 1 mM GABA application, desensitization was best fitted with two exponentials (τshort = 12.4 ms, 0.5 relative amplitude; τlong = 294 ms), similar to desensitization time constants previously reported for α1β2γ2 recombinant receptors (∼30 and ∼300 ms; Gingrich & Burkat, 1998). The fast desensitization time constant was similar to the single channel mean burst duration (17 ms; Lavoie et al. 1997). Fast desensitization was probably responsible for the majority of current decay seen in short step applications of GABA (Fig. 6A), since 50 ms is insufficient time to accumulate a detectable percentage of receptors in a slow desensitized state using a time constant of ∼300 ms (modelled data not shown). Increasing application frequency in a paired pulse protocol (Fig. 7D) demonstrated no decrement in peak amplitude. Notably, using brief interpulse intervals (8 ms) the subsequent peak was 25 % greater than the initial peak, even at a GABA concentration that saturated activation under equilibrium conditions (10 mM). Normalization of subsequent responses to the first initial peak indicated similar kinetics for all responses (Fig. 7E). These results suggest that unlike native receptors, α1β1γ2 recombinant receptors probably do not enter fast, prolonged desensitized states (δ≡ 1000 s−1, τdesensitization≡ 1 ms; ρ < 25 s−1, τresensitization > 40 ms; Jones & Westbrook, 1995), and that receptor efficacy is submaximal under these dynamic conditions.
Figure 7. Desensitization does not influence recombinant receptor system response kinetics.
A, serial pulse activation of α1β1γ2 receptors (< 1 ms duration, 1 Hz frequency, 10 pulses, 2 separate trials in the same patch) indicated that desensitization did not greatly decrease peak current amplitude in recombinant receptors. B, comparisons between ensemble averages of the first and second set of 10 receptor system responses indicated that desensitization did not influence system response kinetics for recombinant receptor populations. C, responses to prolonged agonist application are shown for α1β1γ2 recombinant receptors indicating that more prolonged applications can result in desensitization of recombinant receptors. Current decay was best fitted with two time constants (τshort = 12.4 ms, 0.5 relative amplitude; τlong = 294 ms). D, current responses to paired 1 ms pulse applications of 10 mM GABA at 8, 20, 30, 50 and 100 ms interpulse intervals. All traces were normalized to the peak of the first response (900–1300 pA). Subsequent responses with > 20 ms interpulse intervals were within 5 % of the initial peak value; however, the response to the 8 ms interpulse interval was 25 % greater. E, overlaying normalized subsequent responses from the paired pulse experiment in D to the peak of the first response indicated similar kinetics for all responses.
Pharmacological alteration of system responses
Benzodiazepines, such as diazepam, and barbiturates, such as phenobarbital, modulate GABAA receptor channel function through different mechanisms. Single channel recording under steady-state conditions has revealed that diazepam increases channel burst frequency while barbiturates increase average channel open duration and burst durations (Twyman et al. 1989). Diazepam increases GABA affinity to its receptor, and recently it has been demonstrated that diazepam probably alters GABA binding at the first GABA binding site (Lavoie & Twyman, 1996). Modelling a diazepam-like allosteric increase in affinity at the first binding site revealed that while peak currents were increased severalfold (Fig. 8A), normalized receptor system responses from modulated receptors were indistinguishable from control responses (Fig. 8B). In contrast, modelling a pentobarbital-like allosteric increase in average open duration under dynamic (< 1 ms pulse) conditions revealed that the deactivation phase was primarily altered in the system response (Fig. 8B) with little effect on peak current (Fig. 8A). Thus, pharmacological manipulation of binding affinity or alterations in gating mechanisms could be distinguished using receptor system responses.
Figure 8. Pharmacological modulation of binding or gating kinetics of the receptor system response.
An adaptation of the simple model (Fig. 1) is shown where a benzodiazepine (diazepam; dz) increases GABA association at the first binding site (increased k+1; Lavoie & Twyman, 1996), or a barbiturate (pentobarbital; pb) prolongs open duration (modelled as a decrease in α). A, simulated current responses showed enhancement of current evoked in the presence of either agent. Increased association (k+1 = 108 s−1) resulted in increased peak response to a non-saturating agonist concentration (1 mM), while decreased closing rate (α= 100 s−1) resulted in prolongation of decay currents with minimal change in peak response. B, normalization of simulated responses to peak current indicated that alterations in the first binding transition had no effect on receptor system response kinetics, whereas changes in a channel gating transition significantly affected system response deactivation. High resolution comparisons of current onset are shown for clarity (inset). C, normalized recombinant receptor system responses to 1 mM GABA in the presence or absence of 25 nM diazepam shows no alteration of system response kinetics although peak GABA receptor currents were enhanced by diazepam in the same patch (not shown). Pre-application of diazepam did not alter the results. Normalized system responses evoked in the presence of 50 μM pentobarbital demonstrated that the deactivation phase had slower decay time constants. Pre-application of pentobarbital was necessary in this patch.
Predictions for pharmacological alterations of the receptor system response were studied using diazepam and pentobarbital on α1β1γ2 receptors. Diazepam enhanced the peak current of the receptor system response (not shown, but see Lavoie & Twyman, 1996) when compared with control receptor system responses in the same patch. However, when the two responses were normalized and superimposed, they were not different (Fig. 8C). In contrast, pentobarbital did not enhance the peak currents under these non-equilibrium conditions. When the control and pentobarbital-modulated responses were normalized and superimposed, there was a visually apparent prolongation of the deactivation current (Fig. 8C).
Neuronal receptors
Cultured cortical neurons contain an unknown population of GABAA receptor subtypes. Receptor system responses were obtained from multichannel patches of 37 neurons. Visually apparent differences were readily noted from different cells in the set, but responses from some cells were remarkably similar to each other (not shown). Small CBD and SED to noise ratios between individual (single patch) responses in the set indicated that some neurons had receptor system responses that were remarkably similar. Cluster analysis of the set of native receptor responses revealed five major groups containing six, six, nine, seven and four responses (types I-V, respectively) from different neurons (Fig. 9). Four minor groups composed of one to two responses each were also identified (not shown).
Figure 9. Receptor system responses reveal functional subtypes of GABAA receptors obtained from mouse cortical neurons.
Receptor system responses, evoked with < 1 ms applications of GABA to excised patches from cultured fetal mouse cortical neurons (37 patches; peak = 20–200 pA, n > 10 averaged responses), were analytically grouped using CBD and SED distance ratios into five major classes I-V (normalized and shown as means ±s.e.m. over various points; see Results). Corresponding current onset for group averaged responses is shown in the insets.
Onset differed slightly between the five major groups (Fig. 10A), with rise times that varied over a range of 1.0–1.8 ms, while deactivation phases were unique for the five major cortical neuron groups (Fig. 10B). Deactivation phases were consistently composed of two exponential time constants that varied greatly between the five major groups (Table 1). Although early deactivation phases of types III and IV were similar, later phases were visually distinct (Fig. 2C). It is interesting to note that although both fast and slow decay time constants were similar for types II and V, they were visually and analytically distinguishable by the relative amplitude contributions of each time constant.
The clustering of certain native receptor system responses suggests that, at least functionally, neurons from which these patches were obtained contain similar receptors. Notably, it is not known whether the receptor populations in the patches are homogeneous. Heterogeneous populations would probably result in receptor system responses containing different kinetics unless the differences in parameter rates were sufficiently small to be unresolvable. It would be expected that if two independent populations were present in the patch, a weighted sum of two receptor system responses would produce the heterogeneous population response. Initial attempts at fitting weighted sums of two individual receptor system responses to receptor system responses that did not fall into the major groups were inconclusive.
Using visual comparison and distance ratios, none of the native GABAA receptor system responses were similar to the α1β1γ2 receptor system response. One mouse cortical neuron receptor system response comprising a minor group was visually similar and analytically indistinguishable from the type I α2β1γ2 response (Fig. 10C). While this result suggests functional overlap between native and recombinant receptors, it does not prove that the two receptors expressed the same subunit isoforms.
Peak currents from native neuronal receptors often decreased with serial brief pulse applications of 1 mM GABA, indicating more predominant run-down or desensitization (Fig. 11A) than the expressed recombinant receptors (Fig. 6A), and consistent with other neuronal GABA receptors (Jones & Westbrook, 1995). Superposition of normalized responses from serial brief applications revealed that the receptor system responses did not vary over time (Fig. 11B). However, prolonged agonist exposure (2 s, 1 mM GABA) produced a decay in evoked current, indicating that some channels may have entered a desensitized state(s) (Fig. 11C).
Figure 11. Desensitization did not influence native receptor system response kinetics.
A, serial pulse activation of native mouse cortical neurons (< 1 ms duration, 1 Hz frequency, 10 pulses, 2 separate trials in the same patch) demonstrated a progressive decrease in peak current suggesting absorption of some native receptors into desensitized states. B, comparisons between ensemble averages of the first and second set of 10 receptor system responses indicated that desensitization did not influence system response kinetics for these native receptor populations. C, a single response to a prolonged 100 μM GABA application demonstrates faster desensitization kinetics in this excised patch receptor population than that seen in recombinant α1β1γ2 receptors (Fig. 7C). Current decay was best fitted with two time constants (τshort = 66.7 ms, 0.3 relative amplitude; τlong = 833 ms).
DISCUSSION
The receptor system response results obtained in this study using pulse applications of agonist have several implications in regard to the expression of GABAA receptor subtypes and to ligand-gated receptor function under synaptic-like conditions. First, the receptor system response provides close estimates of maximal activation and deactivation kinetics for the receptor population independent of ligand concentration. Second, the kinetics for maximal activation and deactivation are dependent on GABA subunit composition and receptor subtype. Third, receptor system responses are reproducible within a defined receptor population, suggesting a mechanism for recombinant receptor preferred assembly. Additionally, cultured mouse cortical neuronal GABAA receptors have some distinctively different kinetic properties, but receptor system responses from some neurons are analytically similar suggesting that some neurons express homogeneous populations of GABAA receptor functional subtypes (see below). Fourth, the receptor system response is a novel approach to identify receptor subtypes functionally, and potentially make inferences about receptor subunit composition.
Activation
Previous studies of the GABAA receptor have demonstrated regional differences in agonist binding affinity, pharmacological specificity and IPSC kinetics, and have suggested a regional distribution of receptor subtypes (Vicini et al. 1986; Barker & Harrison, 1988; Otis & Mody, 1992; Pearce, 1993; Zhang et al. 1993; Frerking et al. 1995). Functional properties of the GABAA receptor are also dependent upon the receptor subunits and isoforms expressed (Levitan et al. 1988; Pritchett et al. 1989; Verdoorn et al. 1990; Wafford et al. 1993). Brief, transient exposure to high agonist concentrations has been proposed to occur at fast GABA synapses (Maconochie et al. 1994). Thus, the kinetic properties of miniature IPSCs probably follow those for the receptor system response (i.e. maximal rates of current onset and decay).
The rates of current onset for the receptor system response represent upper physiological limits of receptor activation. At the conclusion of the agonist pulse, receptors that eventually open are synchronized in a bound but closed state, such that transformation of the channel into an open conformation (β) is the primary determinant of current onset (where β≫α). The rates of these gating transitions are intrinsic properties of the subunit proteins which form the receptor-channel complex. For example, receptors containing α2β1γ2 subunits open twice as fast as α1β1γ2-containing receptors. Rise times can be fitted to the following equation (Legendre, 1998):
![]() |
where h is the Hill coefficient. Since the α1β1γ2 and α2β1γ2 receptors have similar values for the EC50 of activation and for closing rate (α; Lavoie et al. 1997), differences in rise time can be explained by a relative doubling of gating open rate (β in Fig. 1; Lavoie et al. 1997). Alternatively, using whole-cell EC50 calculations for α1β1γ2 and α2β1γ2 (i.e. EC50 of ∼130 and 10 μM, Hill number = 1.3 and 2.1, respectively; N. L. Harrison, personal communication) yielded similar results (∼2.5-fold differences in β).
Although characteristic activation rates can be used to distinguish between receptor subtypes, the limited range of onset (10–90 % rise times) observed in recombinant and native receptors limits the practical use of maximal receptor activation as a tool for identifying receptor subtypes from a large pool.
Deactivation
Rapid removal of unbound agonist from the excised patch membrane probably ensures that current decay kinetics are both independent of ligand rebinding and primarily determined by the combined rates of channel closing (α) and the relative probabilities of channel reopening (β/(β+k-2)) or unbinding of the ligand (k-2/(β+k−2)) for the model in Fig. 1. Since the differences in rates of α and β between α1β1γ2 and α2β1γ2 receptors are insufficient to produce 6-fold differences in deactivation (Lavoie et al. 1997), the prolongation in receptor system response deactivation is probably due to slower GABA unbinding rates (k-2) for α2β1γ2 receptors. This is supported by results indicating that α2-containing receptors have a lower whole-cell EC50 for peak GABA responses (Levitan et al. 1988). Similarly, the prolonged deactivation phases for some of the native receptor groups probably resulted from receptors that either had higher apparent affinities for GABA, perhaps resulting from the different α subunit isoforms, or had a greater tendency to produce bursts of openings. It is currently unknown whether β and/or γ subunit isoform expression have similar influences on receptor system response kinetics.
In contrast to previous reports for native GABAA receptors, desensitization did not result in resolvable slowing of deactivation rates for recombinant GABAA receptors following prolonged agonist exposure (Jones & Westbrook, 1995). This previous study in hippocampal neurons demonstrated that prolongation of deactivation was probably the result of slow exit from two desensitized states. The biexponential decay of our prolonged agonist applications suggests that our recombinant receptors may also have two desensitized states; however, the data reported here are consistent with more moderate desensitization and faster resensitization rates than native receptors. Since no decrease in peak was seen following sequential pulse applications with an 8 ms interval, it is likely that the resensitization rate was sufficiently fast to prevent distortion of deactivation time constants. A similarly fast recovery from desensitization was recently reported for α1β2γ2 recombinant receptors (Tia et al. 1996). The simple kinetic model with a single desensitization rate (δ= 330 s−1; Fig. 1) was able to predict closely the short step application data (Fig. 6A) despite its lack of a second, slower desensitized state. This was possible because entry into the second desensitized state is slow compared with entry into the first desensitized state; thus, a small fraction of the receptors enter this slower state during short step (< 50 ms) applications. It would be necessary to invoke a more complex model with at least two desensitized and two open states in order to predict accurately all of the data represented here, particularly the long desensitizing decay currents from 2 s agonist applications. However, the simple kinetic model is sufficient to understand the basic properties of activation and deactivation during brief agonist exposure as in the receptor system response. As such, the system response can be used to phenotype receptor kinetics despite differences in desensitization between recombinant and native receptors.
Structure-function relationships of recombinant and native receptors
The data demonstrate that the kinetic characteristics of receptor system responses were dependent on recombinant GABAA receptor structure, and suggest a similar functional dependence may exist for native GABAA receptors. Subunit composition or structural modification plays a great role in altering gating kinetics and receptor system response deactivation. α2β1γ2 receptors open faster and have a prolonged deactivation phase compared with α1β1γ2 receptors. Furthermore, recent reports indicate that (1) α1β2γ2 receptors preferentially combine two α, two β and one γ subunit protein (Chang et al. 1996), and (2) subunit order, as well as number, can influence function at least in cyclic nucleotide-gated channels (Liu et al. 1996). Thus, the single characteristic receptor system response for α1β1γ2 receptors may indicate either that stoichiometry and order are conserved across a receptor population, or that order does not significantly influence function for these receptor subtypes.
The two characteristic receptor system response types for α2β1γ2 receptors may indicate that subunits combine in at least two different orders and produce two distinctly different kinetic responses. Alternatively, multiple system responses could have resulted from a heterogeneous population of receptor subtypes, differentiated perhaps by open or burst durations. Although these differences are characteristic of GABA receptors composed of different subunits, the prospect of heterogeneous receptor expression (i.e. α2β1 versusα2β1γ2) in the patches examined was considered unlikely for several reasons. First, heterogeneous receptor populations would be expected to have produced a series of graded responses from patch to patch that constituted a weighted average of the represented receptors. Instead, two distinct α2β1γ2 response types were consistently and uniquely identified suggesting that uniform populations existed in individual patches. Second, although whole-cell currents could be recorded with acute transfection of α2β1, excised patches contained insufficient channel densities for adequate system response measurements (authors’ unpublished observations). Also, single channel recordings of α2β1 receptors revealed lower main conductance and shorter burst durations compared with patches of cells transfected with α2β1γ2. This suggests that in the presence of γ2, α2β1γ2 combines more readily to form receptors than α2β1 alone. Third, curve fitting of the type I and II α2β1γ2 deactivation phases revealed non-overlapping time constants. Taken together and with the sensitivity of the analytical clustering technique, it appears likely that patches of α2β1γ2 receptors contained functionally distinct but uniform populations. It is not known whether these two characteristic responses resulted from α2β1γ2 receptors with different stoichiometry or subunit configuration, or perhaps from a single type of receptor that exhibits two functional modes. These functional modes could perhaps be regulated by post-translational modification processes such as phosphorylation.
Native receptor populations from cultured fetal mouse cortical neurons exhibited five main functional classes, each primarily distinguished by deactivation phases. It is likely that these responses represent at least five different functional receptor subtypes, but conclusions about the subunit composition of these native receptors are presently not possible. However, it is remarkable that different neurons from different cultures can have similar receptor system responses. If these individual neurons were expressing a heterogeneous population of functionally different receptor subtypes, then excised patches would be expected to contain varying proportions of each functional receptor subtype. Likewise, individual (single patch) receptor system responses would be expected to vary between neurons according to the cumulative function of the sampled population. Instead, the similarity of some receptor system responses suggests that homogeneous populations of functional receptor subtypes were expressed in some different neurons. This finding, taken together with the results obtained from recombinant receptors here and by others (Angelotti & Macdonald, 1993; Chang et al. 1996), strongly suggests that neurons express specific subtypes of receptors and that some subunits appear to combine and assemble in a preferential fashion.
The number and stoichiometry of native GABAA receptor subtypes remain unknown. Certain subunit combinations are likely to occur and the predominant GABAA receptor subtype in the adult cortex probably contains α1β2γ2 subunits (see McKernan & Whiting, 1996, for review). Receptor system responses from cells expressing these combinations and comparison with responses from native GABA receptors may provide supporting functional evidence for these receptor subtypes. None of the neuronal receptor system responses were similar to the α1β1γ2 or the type II α2β1γ2 receptors, indicating that these are rare or perhaps non-existent conformations in these cultured fetal cortical neurons. Only one neuron had a receptor system response that was analytically indistinguishable from that of the type I α2β1γ2 receptor. It remains to be seen whether this receptor system response uniquely identifies this type of receptor.
Implications for synaptic function
If excised patches of GABA receptors obtained from neuronal soma reflect the same receptors found in the synapse (Fritschy et al. 1992; Caruncho et al. 1993; Craig et al. 1994; Nusser et al. 1995), maximal limits for IPSC onset and decay could be determined by the receptor system response kinetics. Similarly brief agonist conditions between receptor system response and synaptic activation suggest that constraints on the modulation of the receptor system response also apply to the synapse. For example, both modelling and experimental data show that rates of receptor system response activation are practically independent of ligand binding factors such as agonist concentration and binding affinity. Likewise, receptors absorbed into prolonged desensitized states do not distort deactivation rates from impulse agonist applications. Although agonist concentration at the synapse may determine IPSC amplitude (Fatt & Katz, 1951; Frerking et al. 1995), it is not likely to influence activation or deactivation kinetics if the agonist is cleared quickly. Thus, parallels between the receptor system response and postsynaptic currents implicate the system response as a useful tool in determining upper limits of postsynaptic response onset and decay kinetics, as well as estimates of synaptic agonist concentration, and pharmacological modulation at the synapse.
In this study we have demonstrated that brief pulses of agonist elicit a receptor system response that is characterized by maximal receptor activation and deactivation rates independent of agonist concentration or receptor desensitization. Pharmacological modulation of the receptor system response suggested that modulation at the synapse was qualitatively different from modulation under equilibrium conditions, which may be influenced by desensitized states. Furthermore, modulation of IPSC amplitude or kinetics by drugs that increase binding affinity can be used to estimate the ligand concentration and time course at the synapse.
Acknowledgments
This work was supported in part by University of Utah Graduate Fellowship Program to A.M.L.M., the Huntsman Cancer Institute, NIH NINDS grant NS31519 and a Mallinckrodt Scholar award to R.E.T.
References
- Angelotti TP, Macdonald RL. Assembly of GABAA receptor subunits: α1β1 and α1β1γ2σ subunits produce unique ion channels with dissimilar single-channel properties. Journal of Neuroscience. 1993;13:1429–1440. doi: 10.1523/JNEUROSCI.13-04-01429.1993. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Bankman IN, Johnson KO, Schneider W. Optimal detection, classification, and superposition resolution in neural waveform recordings. IEEE Transactions on Biomedical Engineering. 1993;40:836–841. doi: 10.1109/10.238472. [DOI] [PubMed] [Google Scholar]
- Barker JL, Harrison NL. Outward rectification of inhibitory post-synaptic currents in cultured rat hippocampal neurones. The Journal of Physiology. 1988;403:41–55. doi: 10.1113/jphysiol.1988.sp017237. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Caruncho HJ, Puia G, Slobodyansky E, da Silva PP. Freeze-fracture immunocytochemical study of the expression of native and recombinant GABAA receptors. Brain Research. 1993;603:234–242. doi: 10.1016/0006-8993(93)91242-k. [DOI] [PubMed] [Google Scholar]
- Chang Y, Wang R, Barot S, Weiss DS. Stoichiometry of a recombinant GABAA receptor. Journal of Neuroscience. 1996;16:5415–5424. doi: 10.1523/JNEUROSCI.16-17-05415.1996. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Chen C, Okayama H. High-efficiency transformation of mammalian cells by plasmid DNA. Molecular and Cellular Biology. 1987;7:2745–2752. doi: 10.1128/mcb.7.8.2745. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Colquhoun D, Jonas P, Sakmann B. Action of brief pulses of glutamate on AMPA/kainate receptors in patches from different neurones of rat hippocampal slices. The Journal of Physiology. 1992;458:261–287. doi: 10.1113/jphysiol.1992.sp019417. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Craig AM, Blackstone CD, Huganir RL, Banker G. Selective clustering of glutamate and γ-aminobutyric acid receptors opposite terminals releasing the corresponding neurotransmitters. Proceedings of the National Academy of Sciences of the USA. 1994;91:12373–12377. doi: 10.1073/pnas.91.26.12373. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Dinning GJ, Sanderson AC. Real-time classification of multiunit neural signals using reduced feature sets. IEEE Transactions on Biomedical Engineering. 1981;12:804–811. doi: 10.1109/TBME.1981.324679. [DOI] [PubMed] [Google Scholar]
- Fatt P, Katz B. An analysis of the end-plate potential recorded with an intra-cellular electrode. The Journal of Physiology. 1951;115:320–370. doi: 10.1113/jphysiol.1951.sp004675. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Frerking M, Borges S, Wilson M. Variation in GABA mini amplitude is the consequence of variation in transmitter concentration. Neuron. 1995;15:885–895. doi: 10.1016/0896-6273(95)90179-5. 10.1016/0896-6273(95)90179-5. [DOI] [PubMed] [Google Scholar]
- Fritschy JM, Benke D, Mertens S, Oertel WH, Bachi T, Mohler H. Five subtypes of type A γ-aminobutyric acid receptors identified in neurons by double and triple immunofluorescence staining with subunit-specific antibodies. Proceedings of the National Academy of Sciences of the USA. 1992;89:6726–6730. doi: 10.1073/pnas.89.15.6726. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Gingrich KJ, Burkat PM. Zn2+ inhibition of recombinant GABAA receptors: an allosteric, state-dependent mechanism determined by the γ-subunit. The Journal of Physiology. 1998;506:609–625. doi: 10.1111/j.1469-7793.1998.609bv.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Jonas P. Fast application of agonists to isolated membrane patches. In: Sakmann B, Neher E, editors. Single Channel Recording. New York: Plenum Press; 1995. pp. 231–242. [Google Scholar]
- Jones MV, Westbrook GL. Desensitized states prolong GABAA channel responses to brief agonist pulses. Neuron. 1995;15:181–191. doi: 10.1016/0896-6273(95)90075-6. 10.1016/0896-6273(95)90075-6. [DOI] [PubMed] [Google Scholar]
- Lavoie AM, Tingey JJ, Harrison NL, Pritchett DB, Twyman RE. Activation and deactivation rates of recombinant GABAA receptor channels are dependent on a subunit isoform. Biophysical Journal. 1997;73:2518–2526. doi: 10.1016/S0006-3495(97)78280-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Lavoie AM, Twyman RE. Direct evidence for diazepam modulation of GABAA receptor microscopic affinity. Neuropharmacology. 1996;35:1383–1392. doi: 10.1016/s0028-3908(96)00077-9. 10.1016/S0028-3908(96)00077-9. [DOI] [PubMed] [Google Scholar]
- Legendre P. A reluctant gating mode of glycine receptor channels determines the time course of inhibitory miniature synaptic events in zebrafish hindbrain neurons. Journal of Neuroscience. 1998;18:2556–2570. doi: 10.1523/JNEUROSCI.18-08-02856.1998. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Levitan ES, Schofield PR, Burt DR, Rhee LM, Wisden W, Kohler M, Fujits N, Rodriguez HF, Stephenson FA, Darlison MG, Barnard EA, Seeburg PH. Structural and functional basis for GABAA receptor heterogeneity. Nature. 1988;335:76–79. doi: 10.1038/335076a0. 10.1038/335076a0. [DOI] [PubMed] [Google Scholar]
- Liu DT, Tibbs GR, Siegelbaum SA. Subunit stoichiometry of cyclic nucleotide-gated channels and effects of subunit order on channel function. Neuron. 1996;16:983–990. doi: 10.1016/s0896-6273(00)80121-x. 10.1016/S0896-6273(00)80121-X. [DOI] [PubMed] [Google Scholar]
- Macdonald RL, Rogers CJ, Twyman RE. Kinetic properties of the GABAA receptor main conductance state of mouse spinal cord neurones in culture. The Journal of Physiology. 1989;420:479–499. doi: 10.1113/jphysiol.1989.sp017545. [DOI] [PMC free article] [PubMed] [Google Scholar]
- McKernan RM, Whiting PJ. Which GABAA-receptor subtypes really occur in the brain? Trends in Neurosciences. 1996;19:139–143. doi: 10.1016/s0166-2236(96)80023-3. 10.1016/S0166-2236(96)80023-3. [DOI] [PubMed] [Google Scholar]
- McManus OB, Magleby KL. Kinetic time constants independent of previous single-channel activity suggest Markov gating for a large conductance Ca-activated K-channel. Journal of General Physiology. 1989;94:1037–1070. doi: 10.1085/jgp.94.6.1037. 10.1085/jgp.94.6.1037. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Maconochie DJ, Zempel JM, Steinbach JH. How quickly can GABAA receptors open? Neuron. 1994;12:61–71. doi: 10.1016/0896-6273(94)90152-x. 10.1016/0896-6273(94)90152-X. [DOI] [PubMed] [Google Scholar]
- Nusser Z, Roberts JD, Baude A, Richards JG, Somogyi P. Relative densities of synaptic and extrasynaptic GABAA receptors on cerebellar granule cells as determined by a quantitative immunogold method. Journal of Neuroscience. 1995;15:2948–2960. doi: 10.1523/JNEUROSCI.15-04-02948.1995. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Otis TS, Mody I. Modulation of decay kinetics and frequency of GABAA receptor-mediated spontaneous inhibitory post-synaptic currents in hippocampal neurons. Neuroscience. 1992;49:13–32. doi: 10.1016/0306-4522(92)90073-b. 10.1016/0306-4522(92)90073-B. [DOI] [PubMed] [Google Scholar]
- Pearce RA. Physiological evidence for two distinct GABAA responses in rat hippocampus. Neuron. 1993;10:189–200. doi: 10.1016/0896-6273(93)90310-n. 10.1016/0896-6273(93)90310-N. [DOI] [PubMed] [Google Scholar]
- Pritchett DB, Sontheimer H, Shivers B, Ymer S, Kettenmann H, Schofield PR, Seeburg PH. Importance of a novel GABAA receptor subunit for benzodiazepine pharmacology. Nature. 1989;338:582–585. doi: 10.1038/338582a0. 10.1038/338582a0. [DOI] [PubMed] [Google Scholar]
- Roberts E. GABA: the road to neurotransmitter status. In: Olsen RW, Venter JC, editors. Benzodiazepine/GABA Receptors and Chloride Channels: Structural and Functional Properties. New York: Alan R. Liss; 1986. pp. 1–39. [Google Scholar]
- Sakmann B, Hamill OP, Bormann J. Patch-clamp measurements of elementary chloride currents activated by the putative inhibitory transmitters GABA and glycine in mammalian spinal neurons. Journal of Neural Transmission Supplementum. 1983;18:83–95. [PubMed] [Google Scholar]
- Tia S, Wang JF, Kotchabhakdi N, Vicini S. Distinct deactivation and desensitization kinetics of recombinant GABAA receptors. Neuropharmacology. 1996;35:1375–1382. doi: 10.1016/s0028-3908(96)00018-4. [DOI] [PubMed] [Google Scholar]
- Twyman RE, Rogers CJ, Macdonald RL. Differential regulation of γ-aminobutyric acid receptor channels by diazepam and phenobarbital. Annals of Neurology. 1989;25:213–220. doi: 10.1002/ana.410250302. [DOI] [PubMed] [Google Scholar]
- Twyman RE, Rogers CJ, Macdonald RL. Intraburst kinetic properties of the GABAA receptor main conductance state of mouse spinal cord neurones in culture. The Journal of Physiology. 1990;423:193–220. doi: 10.1113/jphysiol.1990.sp018018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Verdoorn TA, Draguhn A, Ymer S, Seeburg PH, Sakmann B. Functional properties of recombinant rat GABAA receptors depend upon subunit composition. Neuron. 1990;4:919–928. doi: 10.1016/0896-6273(90)90145-6. [DOI] [PubMed] [Google Scholar]
- Vicini S, Alho H, Costa E, Mienville J-M, Santi MR, Vaccarino FM. Modulation of γ-aminobutyric acid-mediated inhibitory synaptic currents in dissociated cortical cell cultures. Proceedings of the National Academy of Sciences of the USA. 1986;83:9269–9273. doi: 10.1073/pnas.83.23.9269. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Wafford KA, Whiting PJ, Kemp JA. Differences in affinity and efficacy of benzodiazepine receptor ligands at recombinant γ-aminobutyric acidA receptor subtypes. Molecular Pharmacology. 1993;43:240–244. [PubMed] [Google Scholar]
- Wheeler BC, Heetderks WJ. A comparison of techniques for classification of multiple neural signals. IEEE Transactions on Biomedical Engineering. 1982;29:752–759. doi: 10.1109/TBME.1982.324870. [DOI] [PubMed] [Google Scholar]
- Whiting PJ, McKernan RM, Wafford KA. Structure and pharmacology of vertebrate GABAA receptor subtypes. International Reviews in Neurobiology. 1995;38:95–138. doi: 10.1016/s0074-7742(08)60525-5. [DOI] [PubMed] [Google Scholar]
- Zhang L, Weiner JL, Carlen PL. Potentiation of γ-aminobutyric acid type A receptor-mediated synaptic currents by pentobarbital and diazepam in immature hippocampal CA1 neurons. Journal of Pharmacology and Experimental Therapeutics. 1993;266:1227–1235. [PubMed] [Google Scholar]