Human NT2 teratocarcinoma cells differentiate into neuron-like NT2-N cells when treated with retinoic acid. GABA evoked concentration-dependent whole-cell currents in NT2-N cells with an EC50 of 21.8 μm and a Hill slope of 1.2. GABAA receptor (GABAR) currents reversed atE Cl− and did not display voltage-dependent rectification. GABAR single channels opened in bursts to a 23 pS main conductance level and a 19 pS subconductance level, with infrequent openings to a 27 pS conductance level. Kinetic properties of the main conductance level were similar to other native and recombinant GABAR channels. Diazepam and zolpidem enhanced GABAR currents with moderate affinity, whereas methyl-6,7-dimethoxy-4-ethyl-β-carboline-3-carboxylate inhibited GABAR currents. Loreclezole enhanced GABAR currents with high affinity, but furosemide antagonized GABAR currents with low affinity. The neurosteroids alphaxalone and pregnenolone sulfate appropriately modulated GABAR currents. Zinc blocked GABAR currents with low affinity, but lanthanum did not significantly alter NT2-N GABAR currents. Reverse transcription PCR (RT-PCR) performed on RNA from NT2-N cells clearly detected transcripts encoding human α2, α3, α5, β3, γ3, and π subtypes. The combined pharmacological and RT-PCR results are most consistent with a single or predominant GABAR isoform composed of an α2 and/or α3 subtype combined with the β3 and γ3 subtypes. The data do not rule out receptors containing combinations of α2 and/or α3 subtypes with the α5 subtype or receptors with both β1 and β3 subtypes. The presence or absence or the π subunit in functionally expressed receptors could not be determined.
GABA is the major inhibitory neurotransmitter in the vertebrate brain. Fast IPSPs are mediated by GABAA receptors (GABARs), which contain binding sites for many modulators including benzodiazepines, barbiturates, general anesthetics, penicillin, picrotoxin, bicuculline, and zinc. GABARs consist of five subunits that form a chloride ion channel. Four different subunit families (α, β, γ, and δ) have been studied extensively (Macdonald and Olsen, 1994), and two new subunit families, π (Hedblom et al., 1997) and ε (Davies et al., 1997), have been identified recently. In addition, in mammals, including humans, six α (α1–α6), three β (β1–β3), and three γ (γ1–γ3) subtypes have been described. Pharmacological studies of recombinant receptors have shown that individual subunits and their subtypes confer different sensitivities to GABAR modulators such as benzodiazepines (Pritchett et al., 1989; Wieland et al., 1992) and zinc (Draguhn et al., 1990). The subunit subtypes are differentially expressed throughout development and in different CNS regions (Wisden et al., 1992), reducing the total number of possible isoforms that can be formed in different brain regions and in individual cells.
Studies of GABARs in human CNS neurons have been difficult because of heterogeneity of cell types and unknown GABAR subtype expression. For example, in isolated human central neurons, the heterogeneity of cell types has led to a wide range of responses to GABAR modulators, presumably because of differing subunit composition (Gibbs et al., 1996). It has not been possible to isolate a homogeneous population of postmitotic neurons from human brain and maintain them in culture for detailed pharmacological and biochemical studies. Electrophysiological studies of neuronal GABARs, therefore, have relied on nonhuman neurons in primary cell cultures or acutely isolated neuron preparations. However, the transcriptional, translational, and post-translational processing, assembly, transport, and subcellular localization performed by a heterogeneous population of rodent neurons may not adequately reflect the behavior of human neurons.
The NT2-N human neuronal cell line offers a useful bridge between studies of defined recombinant receptors expressed in non-neuronal cell lines and studies of native nonhuman receptors. Prolonged treatment of the human NT2 teratocarcinoma cell line with retinoic acid causes these pluripotential cells to become terminally differentiated into neurons termed NT2-N cells (Pleasure et al., 1992). After differentiation, NT2-N cells extend both dendritic and axonal processes and express neuron-specific cell surface, cytoskeletal, and secretory markers and functional neurotransmitter receptors (Pleasure et al., 1992; Younkin et al., 1993; Munir et al., 1995; Beczkowska et al., 1997). Levels of protein kinases A and C have also been shown to increase in NT2 cells during differentiation (Abraham et al., 1991), which may be important in regulating the function of these receptors and other cellular processes. In addition, recent reports have shown the existence of synaptic connections between cells in culture (Hartley et al., 1997). In this study we demonstrated that NT2-N cells expressed functional human GABARs with relatively homogeneous properties, suggesting that NT2-N cells may provide a useful model system for investigation of human GABARs assembled and expressed in a homogeneous population of human neurons.
MATERIALS AND METHODS
Cell culture. NT2 cells were grown and maintained in DMEM high glucose (HG) with 10% fetal bovine serum, and penicillin and streptomycin were added as previously described (Andrews, 1984). Cells were plated at 2 × 106 cells/75 cm2 flask and differentiated by treatment with 1 μm retinoic acid (RA) for 4 weeks. After RA treatment, cells were washed with versene and then treated with trypsin to dislodge the cells. Cells were resuspended after being triturated and replated at a 1:10 dilution with DMEM HG and maintained in 5% CO2. The following day the media was removed and saved as conditioned media to feed replates II and III. Cells were again treated with trypsin and then spun for 5 min at 1000 rpm. The pellet was resuspended in 1 ml of media containing mitotic inhibitors (in μm: 10 5-fluoro-2′deoxyuridine plus 10 uridine and 1 cytosine arabinoside) and replated (replate II). The same treatment was performed again after 1 week in culture (replate III) to obtain ∼100% pure neurons. These cells were replated onto 35 mm Corning (Corning, NY) dishes for electrophysiological recording. For these experiments we limited selection of cells to those that were isolated and had neuronal morphology. NT2-N cells tend to form clusters as they differentiate in culture, and cell–cell interactions could possibly effect the differentiation process and resulting cell morphology and might affect neuronal properties. In addition, cell cultures were at best only 99% pure neurons. The contaminating non-neuronal cells were morphologically distinct and could be eliminated by visual selection for electrophysiological experiments.
Solutions and drug application. Cells were removed from the 5% CO2 incubator, and the feeding medium was replaced with recording medium (in mm: 142 NaCl, 1 CaCl2, 6 MgCl2, 8.09 KCl, 10 glucose, and 10 HEPES, 315–325 mOsm, pH 7.4). Patch-clamp electrodes of 5–10 MΩ were filled with pipette solution (in mm: 153.33 KCl, 1 MgCl2, 10 HEPES, 5 EGTA, and 2 ATP, 300–310 mOsm, pH 7.3). This combination of solutions results in nearly equivalent intracellular and extracellular chloride ion concentrations, hence an E Cl− of ∼0 mV, and produced an inward current when cells were voltage clamped at negative potentials.
Compounds were applied to cells using a modified U-tube “multipuffer” application system (Greenfield et al., 1996). The U-tube system enabled us to position a micropipette with a 40–50 μm tip next to the cell for the duration of the recording and apply multiple concentrations of individual drugs to each cell. Stock solutions of GABA, diazepam, lanthanum, pentobarbital, phenobarbital, picrotoxin, 3-5-pregnen-3β-ol-20-one sulfate sodium salt (pregnenolone sulfate), zinc, and zolpidem were made by dissolving each in sterile water. Stock solutions of (3α)-hydroxy-(5α)-pregnan-11,20-dione (alphaxalone), bicuculline, methyl-6,7-dimethoxy-4-ethyl-β-carboline-3-carboxylate (DMCM), furosemide, and loreclezole were dissolved in dimethylsulfoxide (DMSO) and diluted with sterile water (final DMSO concentration was <0.1% v/v). Individual drugs were diluted in recording medium to their final concentration. Alphaxalone and pregnenolone sulfate were purchased from Research Biochemicals (Natick, MA). Loreclezole was obtained from Janssen Biochimica (Berse, Belgium). Bicuculline was purchased from Calbiochem (La Jolla, CA). All other compounds were from Sigma (St. Louis, MO).
Electrophysiology. Whole-cell voltage-clamp and single-channel recordings using the patch-clamp technique were obtained as described previously (Hamill et al., 1981). Patch-clamp electrodes were pulled from Labcraft microhematocrit capillary tubes (Curtin Matheson Scientific, Inc., Houston, TX) or WPI (Sarasota, FL) borosilicate electrode glass (1B150F-3) using a Flaming-Brown P-87 micropipette puller (Sutter Instrument Co., San Rafael, CA). Borosilicate pipettes were fire-polished on a Narashige MF-3 microforge and subsequently coated with polystyrene Q-dope (GC Electronics, Rockford, IL) to reduce electrode tip capacitance.
Whole-cell recordings were performed using an Axopatch 1-B amplifier (Axon Instruments, Foster City, CA). Signals were digitized on-line at 200 Hz using a Labmaster TL-1 analog-to-digital converter (ADC), recorded, and subsequently analyzed off-line using Axotape 2.0 software (Axon Instruments). Single-channel recordings were obtained from “outside-out” patches formed using standard techniques (Hamill et al., 1981) with an Axopatch 200A amplifier. Single-channel currents were low-pass-filtered at 2 kHz using an eight pole Bessel filter and subsequently digitized on line at 20 kHz using Axotape 2.0 software. For some later patch recordings, digitization was performed using a Digidata 1200A ADC and Axoscope software (Axon Instruments).
Data analysis. The magnitude of the enhancement or inhibition of GABAR current by a drug was measured by dividing the peak amplitude of GABAR currents elicited in the presence of a given concentration of the drug (with GABA) by the peak amplitude of control current elicited by GABA alone and multiplying the fraction by 100 to express it as percent of control. Thus the control response was 100%. Peak GABAR currents at various drug concentrations were fitted to a sigmoidal function using a four-parameter logistic equation (sigmoidal concentration–response) with a variable slope. The equation used to fit the concentration–response relationship was: where I was the GABAR current at a given GABA concentration, and I max was the maximal GABAR current. Maximal current and concentration–response curves were obtained after pooling data from all cells tested for GABA and for all drugs. Concentration–response curves were also obtained from individual cells. The curve-fitting algorithm minimized the sum of the squares of the actual distance of points from the curve. Convergence was reached when two consecutive iterations changed the sum of the squares by <0.01%. The curve fit was performed on an IBM PC compatible personal computer using Prism 1.0 or 2.0 (Graph Pad, San Diego, CA). Data are presented as mean ± SEM.
To quantify whole-cell current rectification, peak amplitudes of responses to GABA (at EC50 concentrations) were measured at holding potentials of −50 and +50 mV. These responses exhibited no visible desensitization. An amplitude ratio (+50:−50 mV) was calculated, and rectification was determined with respect to a linear ratio of 1.0 using the predicted E Cl of 0 mV. An amplitude ratio >1.0 indicated outward rectification. To quantify voltage-dependent differences in desensitization, the amount of desensitization produced by a high concentration of GABA (100 μm) was first determined at holding potentials of −50 and +50 mV. Desensitization was expressed as (1 − amplitude at end of GABA application)/(peak GABA amplitude). A desensitization ratio (+50:−50 mV) was then calculated using the method used to characterize rectification. A desensitization ratio <1.0 was indicative of less desensitization at depolarized membrane potentials.
Single-channel current analysis. Analysis of single-channel currents off-line consisted of amplitude and interval detection with Fetchan 6.0 (Axon Instruments) using a 50% threshold detection system, followed by pStat 6.0 (Axon Instruments) for amplitude histograms, and subsequent kinetic investigation using Interval 5 (Dr. Barry S. Palotta, University of North Carolina, Chapel Hill, NC) and software developed in our laboratory (Macdonald et al., 1989). To reduce errors resulting from multichannel patches, recordings were included in the kinetic analysis only if overlaps of simultaneous openings were very brief and occurred in <1% of openings. Open and shut duration histograms were constructed as described by Sigworth and Sine (1987)and fit by the maximum likelihood method. The number of exponential functions required to fit the binned data was increased until additional components did not significantly improve the fit, as determined by the log-likelihood ratio test (Horn 1987; McManus et al., 1988). Intervals with durations <1.5 times the system dead time were displayed in the interval histograms but not included in the fit. Bursts were defined as clusters of openings separated by closures longer than the two shortest closed components, which were thus considered intraburst closures. A critical gap for each patch was calculated from the closed interval distribution to equalize the proportion of misclassified events (Colquhoun and Sakmann, 1985). Prolonged (30 sec to several minutes) drug applications were used to ensure stationarity of channel activity.
Reverse transcription PCR. Total RNA was isolated from NT2-N neurons using the Ultraspec method by Biotecx (Houston, TX). One microgram of total RNA was treated with DNase in a total volume of 10 μl composed of 1 μl of 10× DNase I buffer, 1 μl of DNase I (1 U/μl), and 7 μl of DEPC water. This mixture was incubated for 15 min at 25°C. Addition of 1 μl of 25 mm EDTA was followed by a 10 min incubation at 70°C to heat inactivate the DNase I. The reaction mixture was then chilled immediately on ice. One microliter of a random primer (0.6 μg/μl) was then added to the mixture that was then incubated at 25°C for 15 min. Reverse transcription used the product of the above reaction mixed with the following: 1 μl of 5× first-strand buffer, 2 μl of 0.1m DTT, 1 μl of 10 mm dNTP mix, and 1 μl of RNase inhibitor (10 U/μl). The mixture was prewarmed for 2 min at 42°C before addition of 1 μl of Superscript II (10 U/μl) and then was incubated for 45 min at 42°C. Using the same procedure as described above, with the exception that 1 μl of DEPC water was substituted for Superscript II, acted as a negative control. The reverse transcription product was heat-inactivated for 15 min at 70°C before PCR. The PCR reaction was performed for each subunit in 100 μl of the following mixture: 1 μl of a 20 μm primer mixture (3′ primer and 5′ primer), 2 μl of the reverse transcription product, 10 μl of 10× buffer, 16 μl of 25 mmMgCl2, 1 μl of 25 mm dNTP mixture, 0.5 μl of Amplitaq (Perkin-Elmer, Norwalk, CT; 5 U/μl), and 72.5 μl of DEPC-treated water. The templates for the PCR reaction using cDNA for each subunit subtype were used as positive controls. PCR was performed as follows: a 2 min period at 94°C to denature the mixture, 35 cycles at 94°C for 1 min, 55°C for 2 min, 72°C for 1 min, and a 7 min extension period at 72°C. PCR products (5 μl for positive controls, 15 μl for both test and negative controls) were mixed with dye and run on a 1.5% agarose gel at 120 V for 40 min and then stained with ethidium bromide and photographed. Reaction reagents were purchased from Amplitaq; random primer, Superscript II, RNase inhibitor, and DNase I were purchased from Life Technologies (Gaithersburg, MD); and dNTP mix and DNA molecular weight marker VI were purchased from Boehringer Mannheim (Indianapolis, IN).
NT2-N cell GABAR current
NT2-N cells were voltage-clamped in the whole-cell configuration. Neurons can first be isolated 5 weeks after retinoic acid treatment has begun (Pleasure et al., 1992). At this time multipolar cells were the predominant cell type found in the cultures, and they were readily identifiable. Immediately on rupturing the membrane for intracellular access, the resting membrane potential was measured for each NT2-N cell by shifting to current-clamp mode with current set at zero. Resting membrane potentials ranged from −49 to −33 mV with an average of −40.9 ± 0.4 mV (n = 118). Cells were then voltage-clamped at −75 mV, and 300 μm GABA was applied for 6–10 sec to determine whether functional GABARs were expressed on the membrane surface. Whole-cell currents resembled those seen in recordings of native GABARs in mouse cortical neurons (Kume et al., 1996) and recombinant receptors (Angelotti et al., 1993) with a rapid concentration-dependent rising phase, followed by a decrease in current (despite continued application of GABA), consistent with desensitization. The apparent desensitization occurred at GABA concentrations >10 μm and was typically monoexponential in time course, with faster rates at increased GABA concentrations. Typical current traces from an NT2-N cell are shown in Figure1 A. Peak amplitudes of NT2-N cell responses (n = 12) to a range of GABA concentrations (300 nm to 300 μm) were pooled and fitted to a sigmoidal logistic function (EC50, 21.8 μm; Hill slope, 1.2) (Fig. 1 B). Individual EC50 values ranged from 7 to 80 μmwith an average of 31.3 ± 8.3 μm. NT2-N cells had an average maximal current of 472.4 ± 70.5 pA at 300 μm GABA (n = 12).
Voltage dependence of GABAR currents
The degree and direction of GABAR current rectification depended on the GABAR isoform (Burgard et al., 1996). Whole-cell current–voltage (I–V) relations were generated in NT2-N cells by measuring peak currents evoked by 30 μmGABA at holding potentials from −100 to +75 mV at 25 mV increments. In all cells studied (n = 6), there was no evidence of inward or outward current rectification. A plot of the average peak current at each holding potential for the four cells in which the complete I–V relations were generated resulted in a linearI–V plot (p < 0.001) (Fig.2). An amplitude ratio of 0.90 ± 0.90 was calculated for six cells by dividing the peak current evoked at +50 mV by the peak current evoked at −50 mV. This again indicates that there was no rectification in either direction, because the amplitude ratio was not significantly different from unity (see Materials and Methods).
Voltage-dependent desensitization to high concentrations of GABA was also dependent on the GABAR isoform expressed (Burgard et al., 1996). In the continued presence (15 sec) of 100 μm and higher GABA concentrations, there was a prominent exponential decrease in amplitude from peak until the agonist was removed. The rate of apparent desensitization increased with higher GABA concentration. With our application system, rapid desensitization (time constants <50 msec) could not be recorded, but slower desensitization (time constants >100 msec) could be characterized. To determine whether this apparent desensitization was voltage-dependent, cells were held at +50 and −50 mV for sequential 15 sec applications of 100 μm GABA. The current amplitude at the end of the application was divided by the peak (initial) current for each holding potential. The desensitization ratio for the four cells tested was 1.01 ± 0.12 (see Materials and Methods), indicating no voltage dependence of desensitization during applications of 100 μm GABA at these holding potentials.
Single-channel currents were recorded in the outside-out configuration from a total of 116 GABA applications onto 19 excised patches in six separate experiments. However, most of these patches were not used for kinetic analysis because of the large number of overlapping or multiple openings, suggesting a high density of GABAR channels in the NT2-N cell membrane. Five patches were obtained in four of these experiments in which single openings predominated during applications of 1, 3, and 10 μm GABA. Data from these patches were used for most of the kinetic and amplitude analysis (see Materials and Methods).
GABAR single-channel currents opened singly and in bursts of multiple openings (Fig. 3 A). At least two conductance levels were observed at most holding potentials: a main conductance level of 23 pS in which the majority of openings were classified and a subconductance level of ∼19 pS (Fig. 3 B). Rare openings to a conductance level of ∼27 pS were observed more commonly at negative than positive holding potentials. A current–voltage relationship was obtained by holding the membrane potential at voltages from −80 to +80 mV in 20 mV steps during sequential applications of 3 μm GABA. Openings were detected using a 50% threshold based on the amplitude of the main (23 pS) conductance level and thus included both the subconductance (19 pS) and larger conductance level (27 pS) openings. Event amplitude histograms were best fit using three Gaussian functions; the statistical means of these functions for a single patch are displayed in the current–voltage relationship shown in Figure 3 B. Linear regressions of mean amplitudes at each holding potential showed reversal of single-channel currents at ∼0 mV for all three conductance levels, consistent with the reversal potential for chloride with these solutions. None of the three conductance levels demonstrated either inward or outward rectification at the holding potentials examined. An example of the amplitude histogram for openings in a different patch held at +50 mV (Fig. 3 C) and −50 mV (Fig.3 D) demonstrates symmetrical conductance levels (i.e., no inward or outward current rectification), although the relative proportion of main to subconductance level openings was somewhat larger at more negative potentials. The relative proportion of 27 pS openings was small (5–10%) at both positive and negative holding potentials in most patches.
The frequency of single-channel openings increased with increasing GABA concentration, from 10.4 openings/sec during application of 1 μm GABA to 28.8 openings/sec with 10 μmGABA (Table 1). Mean open time also increased slightly from 1.4 to 1.6 msec, and the percentage of time open increased from 1.8 to 5.2%, consistent with both longer burst openings and increased opening frequency at increased GABA concentration. The mean closed time diminished as GABA concentration increased. Open-duration histograms from selected patches with minimal overlapping or multiple openings were fitted best with three exponential functions, suggesting three open time constants with durations of 0.2, 1.5, and 4.5 msec (Fig.4 A, Table2). These time constants did not change significantly with GABA concentration, but the percentage of openings classified in each exponential distribution varied with GABA concentration. With 1 μm GABA, most of the openings were classified in the shortest (O1) distribution (57%) with smaller proportions of openings in the two longer opening classes (23% O2 openings and 12% O3 openings). As GABA concentration increased, the percentage of O1 openings decreased, and the percentage of longer openings increased (Fig. 4 A), as predicted by models in which the binding of more than one GABA molecule drives the channel into longer open states (Macdonald et al., 1989). However, with 10 μm GABA, a concentration at which almost all openings would be expected to be in the O2 and O3 classes, there were still a large number of O1 openings. This may have resulted from “trapping” of liganded channels in a desensitized distal closed state, from which recovery was slow enough that brief applications of GABA (lasting 30–60 sec) were biased toward sampling channels in the O1 and O2 states.
Closed-duration histograms were fitted best with five exponential functions (Fig. 4 B). As GABA concentration increased, the durations of the three longer closed states (C3, C4, and C5) decreased, whereas the durations of the two short states (C1 and C2) did not change (Fig. 4 B). This was consistent with previous observations in our laboratory (Twyman et al., 1990; Angelotti and Macdonald, 1993) and the interpretation that the C1 and C2 states represent intraburst closures. Histograms of burst durations pooled for each GABA concentration (based on a critical gap determined for each patch) were fitted best with three exponential distributions whose time constants did not change significantly between 1 and 3 μm GABA (1.27, 3.07, and 24.4 msec for 1 μmGABA; 1.96, 2.32, and 24.7 msec for 3 μm GABA) (data not shown), but the proportions shifted from shorter to longer burst durations (64.5, 22.5, and 12.8% for shortest to longest durations for 1 μm GABA vs 31.0, 28.2, and 40.6% for 3 μm GABA). There was an increase in mean burst duration from 3.9 at 1 μm GABA to 10.8 msec at 3 μmGABA, with a concomitant increase in mean openings per burst from 3.6 to 4.9 (Table 1). At 10 μm GABA, however, mean burst duration and the number of openings per burst decreased to 6.6 and 3.1 msec, respectively, again probably because of entry into desensitized states from which recovery may be slow.
Pharmacology of GABAR currents
Individual NT2-N neurons were tested for their responsiveness to a variety of GABAR modulators. In each of the cells tested with all of the following drugs (n = 3), the positive allosteric modulators diazepam (300 nm), loreclezole (3 μm), and pentobarbital (100 μm) enhanced currents evoked by 10 μm GABA (Fig.5 A), whereas the negative allosteric modulators zinc (100 μm), bicuculline (100 μm), and picrotoxin (30 μm) inhibited currents evoked by 30 μm GABA (Fig. 5 B). GABAR currents in NT2-N cells were further studied to characterize the pharmacology of NT2-N GABARs in greater detail with a variety of compounds that modulate GABAR function (Table3).
Both pentobarbital (Fig.6 A) and phenobarbital (Fig. 6 B) enhanced whole-cell GABAR currents evoked by 10 μm GABA. Pentobarbital maximally enhanced GABA-evoked currents by 434 ± 60% at 200 μm(n = 6) with an apparent EC50 of 41.3 μm and Hill slope of 1.47 (Fig. 6 C). Phenobarbital maximally enhanced GABAR current to 331 ± 37% of control at 3 mm but was ∼10-fold less potent with an apparent EC50 of 412 μm and a Hill slope of 2.21 (n = 5) (Fig. 6 C).
Diazepam (300 nm) enhanced whole-cell GABAR currents evoked by 10 μm GABA (Fig.7 A). Diazepam maximally enhanced GABAR current to 340 ± 42% (n = 5) of control at 100 nm diazepam with an apparent EC50 of 74.2 nm and a Hill slope of 1.08 (Fig. 7 D). Concentration–response curves were obtained in five cells by coapplication of diazepam with 10 μmGABA.
The imidazopyridine zolpidem is another GABAR benzodiazepine site agonist sensitive to the α subunit subtype. Zolpidem (10 μm) enhanced GABAR currents evoked by 10 μmGABA in all but one of the NT2-N cells studied (eight of nine) (Fig.7 B). Zolpidem enhanced GABAR currents with moderate affinity when coapplied with 10 μm GABA (EC50, 527 nm; Hill slope, 0.99) (Fig. 7 D).
DMCM is a β-carboline that acts as an inverse agonist at the benzodiazepine site. DMCM (3 μm) reduced whole-cell GABAR currents evoked by 30 μm GABA (Fig. 7 C). DMCM produced a maximal reduction of 44.6 ± 5.2% of GABAR current (n = 9) with an IC50 of 44.8 nmand Hill slope of −1.12 (Fig. 7 D). Eight of the nine cells tested showed inhibition of the current across the range of DMCM concentrations tested. However, in one cell GABAR currents were enhanced at concentrations up to 30 nm and inhibited at higher concentrations.
The novel anticonvulsant drug loreclezole interacts with a site on the β2 and β3 subtypes to enhance GABAR currents (Wafford et al., 1994). Loreclezole enhanced GABAR currents evoked by 10 μm GABA in every cell tested (n = 5) (Fig.8 A). Peak enhancement occurred at 10 μm loreclezole (240 ± 46%), but the degree of enhancement decreased at 30 μm loreclezole (148 ± 23%) for all five cells tested (Fig. 8), as previously reported for recombinant α5β3γ2 GABARs (Donnelly and Macdonald, 1996). To determine the EC50 for loreclezole, only the responses between 100 nm and 10 μm were used to generate the concentration–response relationship. Under these conditions, the loreclezole EC50 was 1.22 μmwith a Hill slope of 3.26 (n = 5) (Fig.8 B).
Neurosteroids have been shown to differentially modulate GABAR currents. Alphaxalone (5α-pregnan-3α-ol-11,20-dione) and pregnenolone sulfate (5-pregnan-3β-ol-20-one sulfate sodium salt) are neurosteroids that enhance and inhibit GABAR currents, respectively. Alphaxalone (10 μm) enhanced currents elicited by GABA (10 μm) by 431 ± 31% (n = 4). Increasing concentrations of alphaxalone (30 nm to 30 μm) resulted in greater enhancement of the peak current (Fig. 9 A). A concentration–response curve of the enhancement of peak GABAR current relative to control was fit with a four-parameter logistic equation yielding an apparent EC50 of 414 nm and a Hill slope of 0.94 (Fig. 9 B). Pregnenolone sulfate (10 pm to 10 μm) decreased GABAR currents (Fig.9 C) in a concentration-dependent manner starting at 1 nm with a maximal inhibition of 67.5 ± 1.9% at 10 μm (Fig. 9 D). An accurate IC50 for pregnenolone sulfate could not be calculated, because 10 μm was near the limit of solubility. However, fitting the data with a logistic equation where I max was set at 0% yielded an equation with an apparent IC50 of 4.9 μm and a Hill slope of −0.50.
Bicuculline is a competitive GABAR antagonist and has no effect on GABAB receptors, GABAC receptors, or the phylogenetically and structurally similar glycine receptor (Macdonald and Olsen, 1994). Bicuculline inhibits current at any functional GABAR isoform, irrespective of subunit composition. Currents evoked by 30 μm GABA in NT2-N cells were inhibited by bicuculline (10 nm to 30 μm) (EC50, 1.1 μm; Hill slope, −0.72) (Fig. 10 A,C). Currents elicited by 30 μm GABA were completely blocked by coapplication of 100 μm bicuculline.
Furosemide, an anthranilic acid derivative, inhibits recombinant GABAR currents with IC50s in the micromolar range, but only when either an α4 or α6 subunit is present (Wafford et al., 1996). Furosemide (10 μm to 3 mm) inhibited NT2-N GABAR currents evoked by 30 μm GABA only at concentrations >100 μm (Fig. 10 B). The greatest inhibition observed was 78.2 ± 1.2% at 3 mm furosemide; concentrations >3 mm could not be achieved because of limits of solubility; hence, true maximal inhibition could not be assessed. However, fitting the data with a logistic equation where I max was set at 0% produced an apparent IC50 of 1.4 mm and a Hill slope of −1.3 (Fig. 10 C).
Picrotoxin, a noncompetitive GABAR antagonist, also inhibited NT2-N GABAR currents. Picrotoxin (30 μm) blocked 71.5% of the peak current evoked by 30 μm GABA (n = 8) (data not shown). A complete concentration–response curve was not generated because of the long-lasting effects of moderate concentrations of picrotoxin (Table3).
The divalent cation zinc (Zn2+) is a noncompetitive antagonist of GABAR currents. Zn2+(300 nm) reduced whole-cell GABAR currents evoked by 30 μm GABA (Fig.11 A). Inhibitory concentration–response curves were generated by increasing the concentration of Zn2+ (100 nm to 1 mm) coapplied with a concentration of GABA near the EC50 for NT2-N cells (30 μm) (Fig.11 C). The IC50 for Zn2+ was 14.7 μm with a Hill slope of −0.43 (n = 5). GABAR currents were almost completely blocked by 1 mmZn2+ (94.8 ± 1.7%).
The trivalent cation lanthanum (La3+), when coapplied with a GABA concentration near the EC50, inhibits α6-containing GABARs and enhances α1-containing GABARs (Saxena et al., 1997). La3+ (100 nm to 1 mm) was coapplied with either 10 μm (data not shown) or 30 μm GABA to NT2-N cells (Fig. 11 B). La3+ had little effect on peak GABAR currents elicited by either GABA concentration. There were no statistically significant changes at lower La3+ concentrations. A small (22.5 ± 4.8%) but significant inhibition of GABAR currents evoked by 30 μm GABA was seen at 1 mmLa3+ (n = 11) (Fig.11 C).
Distribution of maximal effects on individual neurons
The pharmacological data for the allosteric modulators that were tested are summarized in Table 3. The maximal effect was statistically compared with control GABA currents for each cell using pairedt tests. Statistical significance was reached for each compound (*p < 0.05; **p < 0.01; and ***p < 0.001). Normalized maximal effects of all compounds tested on individual cells were evenly distributed around the mean and not separated into distinct groups.
PCR analysis of GABAR subunit mRNAs
Total RNA was isolated from cultures of NT2-N cells at the same 5 week time point that electrophysiological studies were performed. Reverse transcription PCR (RT-PCR) was used to determine the presence or absence of GABAR subunits using primers specific for human α1–6, β1–3, γ1–3, π, and ε subunit subtypes (Table4). In repeated experiments, major bands were found for α2, α3, α5, β3, γ3, and π subtypes (Fig.12), whereas bands for α1, β2, γ1, and γ2 subtypes were not found. Faint bands for the remaining subtypes (α4, α6, and β1) were inconsistently found. Primers designed to amplify the ε subunit did not amplify products of the predicted molecular size. Multiple primers were designed for the γ3 subtype. The molecular masses of the PCR products were slightly larger than the cDNA-positive control when the primers amplified the entire N-terminal domain to the first intracellular loop [amino acids (aa) 18–267]. However, when the primers were designed to amplify only part of the N-terminal domain (aa 110) to midway through the large intracellular loop between M3 and M4 (aa 413), the PCR products were the same molecular mass as the cDNA control (data not shown). These same primers were unable to amplify either the γ1 or γ2 subtype, suggesting that the γ3 subtype expressed in NT2-N cells may represent a longer splice variant than previously described.
NT2-N cells express primarily α2, α3, α5, β3, γ3, and π GABARs subtype mRNAs
High mRNA levels for the α2, α3, α5, β3, γ3, and π subtypes were found using RT-PCR performed on total RNA isolated from NT2-N cells. Minimal mRNA levels for α4, β1, and γ1 subtypes were found in some, but not all, of the experiments, and no transcripts encoding human α1, α6, β2, or γ2 subtypes were detected. The pharmacological data suggested that a limited number of subunit subtypes were endogenously expressed in functional GABARs in NT2-N cells, which was confirmed by the use of RT-PCR. These subtype mRNAs were consistent with most of the pharmacological properties of GABAR currents recorded from NT2-N cells. Because RNA was isolated from an entire 10 cm dish of cells with no preference for cell type, we cannot rule out the possibility that there were subpopulations of NT2-N cells that expressed different isoforms based on either the stage of differentiation or differences in cell fate. It is unlikely that retinoic acid treatment alone would cause multiple fates of clonal NT2 stem cells. However, it is possible that multiple cell fates could occur during the differentiation process caused by secondary factors such as cell–cell or cell–substrate interactions that may alter the major GABAR isoforms being expressed. The possible existence of subpopulations of cells may also explain the low levels of α4, α6, and β1 mRNA and the inconsistency of their presence across RNA preparations. Although it is unlikely that NT2-N cells differentiate into a specific neuronal population found in the developing vertebrate brain, it is interesting to note that the subunits found expressed at high levels in NT2-N cells at this time point are located on chromosomes 15 (α5, β3, and γ3) and X (α3). With the exception of the α2 subtype mRNA, those barely detected were located on chromosome 4 (α2, α4, β1, and γ1), and those not expressed at all were located on chromosome 5 (α1, α6, β2, and γ2) (Rabow et al., 1995). The subunits expressed in NT2-N cells have been shown to be primarily expressed early in the development of the rat brain (Laurie et al., 1992), whereas those located on chromosome 5 are the predominate subunits found in adult rat brains (Wisden et al., 1992). NT2-N cells resemble immature CNS neurons, as determined by markers such as fetal τ, the incomplete phosphorylation state of high molecular weight neurofilament protein, and no MAP2b expression (Pleasure et al., 1992). Further studies are under way to determine whether all NT2-N cells contain all subtype mRNAs or whether there are multiple populations of cells with identifiable morphologies that contain different combinations of messages and whether the NT2 cell line undergoes a developmental switch in subunit expression similar to that found in rat brain.
NT2-N cells express GABARs likely composed primarily of α2, α3, α5, β3, and γ3 subtypes
Differentiated NT2-N neurons expressed GABARs with a distinct set of pharmacological and biophysical properties and expressed a limited number of GABAR subtype mRNAs. High levels of α2, α3, α5, β3, γ3, and π subunit subtype mRNAs were consistently amplified using RT-PCR performed on total RNA isolated from NT2-N cells. These data suggest that NT2-N cells express a limited number of GABAR isoforms with relatively uniform biophysical and pharmacological properties. NT2-N cells constitute the only known human-derived clonal cell line that express functional GABARs; therefore, they provide a unique preparation for investigation of human GABAR expression, assembly, structure, and regulation.
Studies using nonhuman recombinant GABAR subunits have shown that the affinity and efficacy for many GABAR modulators, as well as certain biophysical properties, depend on the expressed isoform. Although no definitive conclusions about the subunit subtype composition of GABARs can be made by determination of pharmacology and kinetic properties, several pharmacological and biophysical properties of NT2-N GABARs limit the possible expressed subtypes. Possible subtypes can be further limited by comparison of the pharmacological, biophysical, and RT-PCR data.
All NT2-N cells responded to GABA with a mean GABA EC50 of 21.8 μm, and individual EC50 values and maximal peak amplitudes were evenly distributed around the mean. Recombinant GABARs expressing rat or bovine α, β, and γ subunits had EC50 values ranging from 1–50 μm, whereas GABARs expressing αβε, αβδ, or αβ subunits alone had lower GABA EC50 values, ranging from 0.5 to 5 μm (Angelotti et al., 1993; Saxena and Macdonald, 1994) (T. R. Neelands and R. L. Macdonald, unpublished results). Thus, NT2-N cell GABARs had a GABA EC50 that was consistent with an isoform containing α, β, and γ subunits.
NT2-N GABARs likely contain a γ3 subtype
Enhancement of GABAR currents by benzodiazepine site ligands requires the presence of a γ subunit in the receptor (Knoflach et al., 1991; Luddens et al., 1994; Hadingham et al., 1995; Benke et al., 1996). Because NT2-N cells expressed high levels of mRNA for γ3, but not γ1 or γ2 subtypes, and GABAR currents were enhanced by diazepam, zolpidem, and reduced by DMCM, it is likely that NT2-N GABARs contain a γ3 subtype.
NT2-N GABARs likely contain α2 and/or α3 subtypes
Recombinant GABARs with high affinity for benzodiazepines contain the α1 subtype, and low-affinity receptors contain either α4 or α6 subtypes when coexpressed with β and γ subtypes. Receptors with moderate affinity for diazepam can be further divided based on moderate affinity (α2- or α3-containing receptors) or insensitivity (α5-containing receptors) to zolpidem. In addition, DMCM, an inverse agonist at the benzodiazepine site, inhibited GABARs containing α1, α2, α3, and α5 subtypes but enhanced GABARs containing α4 and α6 subtypes. Furosemide is a high-affinity antagonist of recombinant rat GABARs that contain either α4 or 6 subtypes (Wafford et al., 1996) when coexpressed with γ and β2 or β3, but not β1, subtypes (Korpi et al., 1995). The cation lanthanum enhances α1-containing receptors but inhibits α6-containing GABARs (Saxena et al., 1997) when coexpressed with a β and γ or δ subunit. The α subtype alters the magnitude and affinity of zinc block. GABARs containing α2 and α3 subtypes were blocked by zinc to a greater extent but with lower affinity than GABARs containing the α1 subtype (White and Gurley, 1995).
GABAR currents evoked from NT2-N cells were potentiated by both diazepam and zolpidem with only moderate affinity, suggesting the presence of an α2 and/or α3 subtype. DMCM inhibited GABAR currents with a moderate affinity; furosemide blocked GABAR current with low affinity; lanthanum had no significant effect on the currents; and zinc had an apparent affinity of 14.7 μm, almost completely blocking (95%) the current at 1 mm, suggesting that the major isoform expressed in NT2-N GABARs did not contain α1, α4, and α6 subtypes. The zinc affinity was similar to that reported for dentate granule cell GABARs (28 μm; Kapur and Macdonald, 1996) and for α5β3γ2L receptors (22 μm;Burgard et al., 1996) but was higher than described for αβ combinations (∼560 nm) (Draguhn et al., 1990). A combination of the subunits identified by RT-PCR (α2, α3, and/or α5, β3, and γ3) would be consistent with the effect of zinc as well as the potentiation by diazepam.
These data suggest that NT2-N cells might express α2 and/or α3 subtypes and likely do not express α1, α4, or α6 subtypes. The RT-PCR study demonstrated the presence of α5 in addition to α2 and α3 subtype mRNAs. Although α5 is expressed, the enhancement of NT2-N currents by zolpidem argues against incorporation of α5 into the major isoform expressed by NT2-N cells. Taken together, these data are consistent with the presence of α2 and/or α3 subtypes coassembled with β and γ subtypes.
NT2-N GABARs likely contain the β3 subtype
Loreclezole enhancement of recombinant rat GABARs acts at a specific modulatory site on the β2 and β3, but not β1, subtypes (Wingrove et al., 1994). In addition, when β1 and β3 were coexpressed, loreclezole sensitivity was similar to that of the low-affinity β1-containing isoforms (Donnelly and Macdonald, 1996). High levels of message for β3 in NT2-N cells with little or no message for β1 or β2 would predict high-affinity enhancement of GABA-evoked currents by loreclezole. GABA-evoked currents in NT2-N cells were enhanced by loreclezole with a calculated affinity consistent with those previously reported for recombinant rat GABARs. These data, along with the RT-PCR findings, support the incorporation of only the β3 subtype into the major NT2-N GABAR isoform.
NT2-N GABAR biophysical properties were consistent with an α, β, and γ subunit subtype containing isoform
Single GABAR channels pulled from NT2-N cells opened in bursts to three different conductance levels: 19, 23, and 27 pS. These conductance levels were similar to those observed in native GABAR isolated in membrane patches from cultured mouse spinal neurons (Bormann et al., 1987) and cultured chick neuron patches (Weiss et al., 1988). Single-channel studies of recombinant rat GABARs have shown similar conductance levels of 29 and 21 pS for GABAR isoforms composed of α1, β1, and γ2S subunits but 15 and 10 pS for GABARs composed of α1 and β1 (Angelotti and Macdonald, 1993; Moss et al., 1990).
Kinetic analysis of the main conductance level openings suggested the presence of three open states and five closed states, similar to those determined in our laboratory for native GABAR channels in outside-out patches obtained from mouse spinal cord neurons in primary cell culture (Macdonald et al., 1989). A limited burst analysis showed concentration-dependent burst distributions similar to previously reported native and recombinant receptors. In most respects, however, the kinetic behavior of NT2-N single-channel GABAR openings was consistent with that reported in earlier studies.
Although studies of recombinant rat or bovine GABAR single-channel currents have not demonstrated a particular conductance level or kinetic “fingerprint” for specific isoforms, isoforms containing only an α and β subunit generally resulted in smaller single-channel conductance levels compared with γ-containing receptors and only two rather than three open states. To date no other detailed single-channel analysis has been done on human recombinant receptors or native isoforms from human tissue. Therefore, because this is the first single-channel description of human GABARs, only limited comparisons can be made between this study and previous work. However, the single-channel currents of NT2-N cells are remarkably similar in the pattern and level of their openings to other native GABA receptors and are consistent with the RT-PCR and pharmacological results, suggesting that the native NT2-N GABAR is composed of an α, β, and γ subunit.
We thank Nadia Esmaeil and Yunning Yang for their technical support in differentiating and maintaining the NT2-N cells and Dr. Ewen Kirkness for contribution of π cDNA, which was used as positive controls in the RT-PCR experiments.
Correspondence should be addressed to Dr. Robert L. Macdonald, Neuroscience Laboratory Building, 1103 East Huron Street, Ann Arbor, Michigan, 48104-1687.