Robert L. Macdonald MD, PhD
Professor and Chairman, Department of Neurology, Vanderbilt University, Nashville, Tennessee
Benzodiazepines (BDZs) are used as anticonvulsant drugs primarily to treat status epilepticus and to terminate serial seizures, but they are also used clinically for antianxiety, as muscle relaxants, and for their hypnotic activity. The primary target of BDZs is a receptor for the neutral amino acid γ-aminobutyric acid (GABA). GABA, the major inhibitory neurotransmitter in the central nervous system, is released from GABAergic neurons and binds to both GABAA receptors (GABARs) and GABAB receptors. The GABAR is a macromolecular protein that forms a chloride ion-selective channel and contains binding sites for GABA, anticonvulsant BDZs and barbiturates, anesthetic steroids, and the convulsant β-carbolines and picrotoxin (26,47,49) (Figure 16.1; Table 16.1). The GABAB receptor does not form an ion channel but is coupled by guanosine triphosphate binding proteins to calcium or potassium ion channels. BDZs enhance GABAR-mediated inhibition but do not alter GABAB receptor-mediated inhibition.
BENZODIAZEPINES ENHANCE GABAERGIC INHIBITION BY BINDING TO BENZODIAZEPINE RECEPTORS
An effect of BDZs on GABAergic inhibition was initially suggested by finding that diazepam enhanced presynaptic inhibition in the cat spinal cord (70), and it was further supported by the finding that diazepam had a GABA-mimetic action on cerebellar cyclic guanosine monophosphate content (21,22). The first direct demonstration that BDZs enhance postsynaptic GABA responses was made using vertebrate spinal cord neurons in cell culture (20,44). Based on the direct effect of diazepam and chlordiazepoxide on GABA responses, it was suggested that BDZs interact directly with a postsynaptic GABA receptor to enhance GABA receptor current by an allosteric mechanism.
The basis for the interaction of BDZs with GABARs was established when a BDZ binding site on the GABAR (the BDZ receptor) was discovered and was characterized (54,81) (Figure 16.1). The BDZ receptor represented the site for BDZs to enhance GABAR current allosterically. The demonstration of high-affinity BDZ binding to mammalian brain fractions and the reciprocal interaction of GABA and BDZ binding were consistent with the physiologic demonstration that BDZs were acting by direct modulation of GABAR function (31,44,81,84).
However, the nature of the BDZ receptor was shown to be more complex by the discovery of compounds that bound to the BDZ receptor but had unexpected actions (10,11). The ethyl ester of β-carboline-3-carboxylate was isolated from human urine and rat brain and was shown to displace [3H]diazepam potently from BDZ receptors. However, this and related β-carbolines were shown to reduce GABAR current and to be convulsant and anxiogenic despite their binding to the BDZ receptor. Thus, these compounds were described as inverse agonists. In addition, an imidazobenzodiazepine, flumazenil, was shown to bind to BDZ receptors but to have little intrinsic activity and thus appeared to be a BDZ receptor antagonist. Furthermore, compounds with partial agonist or partial antagonist activities were described (11). Thus, the BDZ receptor can effectively increase (BDZ agonist action) or decrease (BDZ inverse agonist action) postsynaptic GABAergic inhibition, depending on the nature of the BDZ receptor ligand (9).
In addition to binding BDZ receptor ligands with positive and negative efficacy, the BDZ receptor was shown to exist in multiple forms. Triazolopyridazines such as CL 218,872 were shown to displace [3H]flunitrazepam more potently from cerebellar than from hippocampal membranes, a finding suggesting the existence of BZ1 (CL 218,872-sensitive) and BZ2 (CL 218,872-insensitive) receptors (39). Both BZ1 and BZ2 receptors were coupled to the GABA binding site, thus suggesting GABAR heterogeneity.
FIGURE 16.1. Schematic “donut” model of the GABAA receptor-chloride ion channel complex. Each of the sections represents a distinct functional binding domain for the drug class indicated, as well as the chloride channel itself in the middle. No implications about the protein subunit structure are made in this model. (From Olsen RW, Sapp DM, Bureau MH, et al. Allosteric actions of CNS depressants including anesthetics on subtypes of the inhibitory GABAA receptor-chloride channel complex. In: Rubin E, Miller KW, Roth SH, eds. Molecular and cellular mechanisms of alcohol and anesthetics. Ann NY Acad Sci 1991;625:145-154, with permission.)
MOLECULAR BIOLOGY OF GABA RECEPTORS
Our understanding of the molecular biology of GABARs evolved rapidly in the 1990s. The molecular structure of the GABAR was determined by photoaffinity labeling the GABAR with the BDZ [3H]flunitrazepam (54,75). Initially, a single 51-kd polypeptide was specifically labeled in crude brain homogenates. Subsequently, additional 53-, 55-, and 59-kd BDZ binding polypeptides were obtained and were shown to vary among brain regions, during development, in one-dimensional peptide mapping of proteolytic fragments, and in binding specificity (73), consistent with the presence of a family of gene products. This conclusion was subsequently confirmed by molecular cloning.
The GABAR was then purified (76), and it was demonstrated that the purified receptor bound BDZs and muscimol and was composed of two subunits, an α subunit and a β subunit (78). Photoaffinity labeling demonstrated that α subunits were labeled with BDZs and β subunits were labeled with GABA agonists (18,37,38). Immunolabeling, immunoprecipitation, and Western blot analysis of GABA and BDZ binding sites demonstrated conservation of GABARs within different brain regions and across species (32,52). Subsequently, microheterogeneity of both subunits was observed by protein staining, photoaffinity labeling, and immunoblotting (15,16,30,57). What originally appeared to be two polypeptide bands actually contained more than a dozen polypeptides of similar size.
TABLE 16.1. γ-AMINOBUTYRIC ACID A RECEPTOR PHARMACOLOGY
Purification of bovine GABAR α and β subunits allowed Barnard and colleagues to clone the complementary DNA (cDNA) sequences (71). Several important features were initially identified in these two subunits but were later found in all subsequent GABAR subunits (Figure 16.2). First, both subunit proteins were homologous to each other, having approximately 35% amino acid identity, and they had conserved hydropathy profiles, a finding suggesting a conserved transmembrane topology and common evolution. This general structure was similar to the four-transmembrane domain model of the nicotinic cholinergic receptor (nAchR) (58), a finding suggesting that these receptors (and later glycine and serotonin receptors) formed a supergene family of ligand-gated ion channels. Second, there existed a β-structural loop formed from the disulfide linkage of two cysteine residues that were 14 amino acids apart in the N-terminal segment of the proteins. Third, numerous sites for N- and O-linked glycosylation were identified in the N-terminal segment of the proteins. Finally, there was a high concentration of Thr and Ser residues in the second transmembrane region (TM2) that, again based on analogy to nAchRs, was believed to line the ion channel pore. However, a major difference between the two subunits was the proposed intracellular loop located between TM3 and TM4, the region most dissimilar between the two subunits. Moreover, a consensus phosphorylation sequence for cyclic adenosine monophosphate-dependent protein kinase was located within this loop of the β, but not the α, subunit.
FIGURE 16.2. Generic GABAA receptor protein subunit sequence and topologic structure. The N-terminal half of the polypeptide is suggested to be extracellular, with probable sites for asparagine glycosylation (polymeric black circles at positions 10 and 110) and the conserved cystine bridge at positions 138 to 152. Four putative membrane-spanning domains are shown as α-helical cylinders within the cell membrane (stippled), with the C-terminal at the extracellular end of the fourth membrane-spanning region. A large putative intracellular loop is present between the third and fourth membrane-spanning regions. (Modified from Olsen RW, Tobin AJ. Molecular biology of GABAA receptors.FASEB J 1990;4:1469-1480, with permission.)
After the identification of the first two GABAR subunit cDNAs, it was discovered that mammals had multiple families of subunits (α,β,γ,δ,ε,π,θ, and ρ) and multiple subunit subtypes including α (1, 2, 3, 4, 5, 6), β (1, 2, 3), γ (1, 2, 3), and ρ (1—3) (4,23,24,34,36,40,42,43,60,72), which all share similar structural features and basic functional properties (Table 16.2). Within a family, the amino acid similarity ranges from 70% to 80%, and between families, it ranges from 30% to 40%. Further GABAR subunit diversity has been shown to result from alternative splicing of subunit messenger RNA (mRNA) transcripts. Two forms of the γ2 subtype, a short form (γ2S, the original clone) and a long form (γ2L), were cloned (81). The γ2L variant differed from the γ2S variant by differential splicing and insertion of eight amino acids into the intracellular loop between TM3 and TM4. Interestingly, this insert contained a consensus phosphorylation sequence for protein kinase C.
TABLE 16.2. γ-AMINOBUTYRIC ACID A RECEPTOR SUBUNIT BIOCHEMISTRY
MOLECULAR STRUCTURE OF GABA RECEPTORS
The current understanding of the molecular structure of the GABAR-ion channel complex is that it is heteropentameric glycoprotein (approximately 275 kd), composed of multiple combinations of polypeptide subunits. The subunits form a quasisymmetric structure around the ion channel, with each subunit contributing to the wall of the channel (50). The model is based heavily on analogy with the nAchR.
The existence of 19 or so subunit subtypes leads to several questions about the molecular structure of GABARs. It is not known how many oligomeric GABAR isoforms exist in nature, and the subunit composition of each GABAR isoform is unclear. It is likely that different combinations of subunit subtypes form different GABAR isoforms in different neuronal populations. Thus, it was not surprising to discover differential regional expression in the central nervous system and spinal cord of various subunit subtype mRNAs (41). Localization of the GABAR gene products, including mRNA by in situ hybridization and polypeptides by immunocytochemistry, is giving a picture of where each is found and their possible overlap or coexistence in the same cells. The distribution of mRNAs in the central nervous system determined by in situ hybridization is very different for each subunit subtype. Some subtype mRNAs are only expressed in specific cell types; for example α6 mRNA is demonstrated only in cerebellar granule cells (18). Other subtypes, such as for the β2 subtype, have a more ubiquitous
distribution, whereas various α,β, and γ subtypes and the one δ subtype show very different regional as well as developmental distributions (51,59,65,72). Thus, differential expression and assembly of various GABAR subtypes could produce a multitude of receptor isoforms. However, the specific subunit composition, stoichiometry, and number of different GABAR isoforms are unknown.
BENZODIAZEPINE REGULATION OF RECOMBINANT GABA RECEPTORS
After cloning of GABAR subunits, GABAR expression in Xenopus oocytes or mammalian cell lines was performed to determine whether functional GABAR channel assembly occurred and to determine the pharmacologic properties of expressed GABARs (5,6,62,63,64,76,82). Coexpression of α1, β1, and γ2 subtypes in Xenopus oocytes produced GABARs with reproducible BDZ pharmacology. GABAR responses were potentiated by diazepam and were inhibited by inverse agonist β-carbolines, and these effects were blocked by the BDZ receptor antagonist flumazenil. Complete BDZ receptor pharmacology was also demonstrated by binding assays to human embryonic kidney (HEK) 293 cell membranes after transient coexpression of the three GABAR subunits. An appropriate rank order of potency of several BDZ receptor ligands was demonstrated.
The molecular basis for BI and BZII binding sites was determined using transient expression of αxβ12S (x = 1, 2, or 3) subunit combinations in HEK 293 cells (63). The combination of α1β1γ2S GABAR subtypes produced BZI binding sites, and either α2β1γ2S or α3β1γ2S GABAR subtype combination produced BZII binding sites. These receptor isoforms were differentiated based on binding of the BZI selective compounds, zolpidem and CL 218,872. BDZ receptor pharmacology was not altered by substituting any other β subtype. Seeburg and colleagues further studied the effect of expressing α4, α5, and α6 subtypes with β1 and γ2S subtypes. Expression of the α5 subtype with β1 and γ2S subtypes also created BZII BDZ binding sites, although they had even lower affinity for zolpidem than either α2 or α3 subtype-containing receptors (60). Expression of the α6 subtype with β1 and γ2S subtypes produced a receptor isoform that did not bind the prototypical BDZs, diazepam and flunitrazepam, or β-carbolines, but it did bind the inverse agonist imidazobenzodiazepine Ro 15-4513 and flumazenil (74). Binding of GABA agonists was not impaired. A similar BDZ receptor profile was discovered for the α4 subtype expressed with β1 and γ2S subtypes (82). Thus, BDZ pharmacology of recombinant GABARs appears to depend on the α subtype. The original BZI and BZII classification was altered to include the increased BDZ receptor heterogeneity and contains BZI (α1), BZIIA (α2 and α3), BZIIB (α5), and BZIII (α4 and α6) receptors (27).
FUNCTIONAL DOMAINS FOR BENZODIAZEPINE ACTION
A single amino acid was identified as the covalent attachment site for photoaffinity labeling with [3H] flunitrazepam, H102 in bovine α1 (H101 in rat, also present in α2, α3, α5) (28,79,80). Replacement of the H in the α1 subtype with R as in the α6 subtype reduced BDZ binding, whereas replacement of R100 in the α6 subtype with H as in the α1 subtype resulted in BDZ binding (87). A mouse knock-in for point mutation H101R in the α1 subtype lost sensitivity to the sedative-hypnotic action of BDZs, but not the anxolytic or motor-impairing effects, which must be mediated by α2, α3, α5 subtypes (67). Comparison of other amino acids with histidine at this position revealed that α1 H101 is critical for BDZ binding and efficacy (29). Additional studies revealed that the α4 subtype, also producing GABARs with low affinity for BDZ agonists, has R, not H, in the corresponding position 101. Mutagenesis also implicated P161 and I211 in the BDZ binding site on α6 (88). In addition, the variable selectivity of α subunits for BDZs allowed determination of a residue involved in this specificity: residue α1G200 = α3E225 was shown to be responsible for differences in α1 versus α2/3 subtypes in affinity for CL 218,872 and other ligands (61), a finding suggesting that these residues may participate in the BDZ binding pocket.
The γ subunit is required for benzodiazepine binding by GABARs, and γ2T142 was shown to affect BDZ binding (53). Other residues in the γ2 subtype (F77, T55, M57) have been implicated in BDZ binding (12,13). Chimeras between γ2 and α1 subtypes showed that γ2 K41-W82 and γ2 R114-D161 are needed for BDZ binding in HEK cells, but the chimera of N-terminal 1-161γ2 with the rest of the α1 subtype sequence (expressed with the β2 subtype) gave poor BDZ enhancement of GABAR currents in oocytes and poor GABA enhancement of BDZ binding in HEK cells (7). Residues found in β subunits to affect GABA binding (1) were mutagenized in the α1 subtype (α1Y159, Y161, T206, Y209) and were found to affect BDZ binding (2,14,89). The latter are situated near the already mentioned G200 residue (69). This correspondence of domains for BDZ binding pockets with those for GABA binding in the same pentamer was noted (12,77,79) as a possible evolutionary modification of an agonist site into one for an exogenous allosteric modulator. A sequence homology surrounding the residues in the α1 subtype implicated in binding muscimol and flunitrazepam (79) suggested the structural connection, as well as a functional correlate: in whatever manner GABA binding on β/α subunits is conformationally
coupled to channel opening, BDZ binding on α/γ subunits may also promote that physical change.
REGULATION OF GABA RECEPTOR CHANNELS BY BENZODIAZEPINE RECEPTOR LIGANDS
Development of the single-channel recording technique has permitted direct study of native GABAR channels on neurons and recombinant GABARs expressed in mammalian cells (8,33,48). When GABA is applied to outside-out patches obtained from mouse spinal cord neurons in cell culture, the GABAR channel opens and closes rapidly, so relatively square current pulses are recorded (Figure 16.3). The GABAR channel opens to multiple conductance levels. A 27- to 30-ps conductance level is the predominant or main conductance level (Figure 16.3C, double asterisks) and conductance levels of 17 to 19 ps and 11 to 12 ps occur less frequently (Figure 16.3C, single asterisk). Although the GABAR channel opens to multiple conductance levels, current through the main conductance level is responsible for >95% of the current through the channel.
The equilibrium single-channel gating properties of the main conductance level of the GABAR in murine spinal cord neurons in culture have been characterized (48,68,85). The GABAR opens in bursts of openings interrupted by brief closures (Figure 16.3). The GABAR has been shown to open into three different open states with mean durations of 0.5, 2.5, and 7.3 milliseconds. Increasing GABA concentration produces GABAR single-channel currents that increase in duration. The studies suggest that the receptor has two binding sites for GABA molecules, and the singly bound receptor opens primarily to the brief 0.5-millisecond state, whereas the doubly bound receptor opens primarily to the longer 2.5- and 7.3-millisecond states. The mean open channel duration increases with GABA concentration because more doubly bound receptor is formed. The GABAR has been shown also to enter into multiple closed states including two brief closed states with mean durations of 0.2 and 2.8 milliseconds and at least three closed states with longer mean durations.
FIGURE 16.3. Single-channel currents are shown at increasing time resolution (a-c) for GABA (2 µmol/L). GABA-evoked bursting single-channel inward (downgoing) currents with at least two current amplitudes when outside-out patches are voltage clamped at -75 mV. The larger 2.04-pA (27-picosecond) channel (double asterisks) occurred more frequently compared with a smaller 1.48-pA (20-picosecond) channel (single asterisk). The portion outlined for each tracing is presented expanded in time in the tracing beneath it. Attenuated brief channel openings seen at lower temporal resolution may be seen at true amplitudes at higher temporal resolution. Time calibration for each trace is shown on the right below the trace. Current calibration applies throughout. (Modified from Twyman RE, Macdonald RL. Neurosteroid regulation of GABAA receptor single channel kinetic properties. J Physiol (Lond) 1992;456:215-245, with permission.)
BDZs enhance GABAR current (20,44,45) by binding to the BDZ receptor on GABARs (55,56,83,86). The β-carboline, methyl 6,7-dimethoxy-4-ethyl-β-carboline-3-carboxylate (DMCM), reduces GABA-mediated inhibition also by interacting with the BDZ receptor. To enhance GABAR current, a drug may increase channel conductance, increase channel open and burst frequencies, or increase channel open and burst durations. Conversely, to reduce GABAR current, a drug may decrease channel conductance, decrease channel open and burst frequencies, or decrease channel open and burst durations. By determining the drug-induced alterations produced in the open, closed, and burst properties of GABAR single-channel currents, the mechanisms of action of the BDZs and β-carbolines on GABAR channels were determined. At clinically relevant concentrations (<100 nmol/L), single-channel recordings demonstrated diazepam increased receptor opening frequency without altering mean open time or conductance (66,86) (Figure 16.4). These results contrast with the
increase in burst duration with little effect on burst frequency seen in the presence of phenobarbital (86). For diazepam, these results could be explained by an increased affinity of the GABAR at one, but not both, of the GABA binding sites. Another explanation is that BDZs could reduce the rate of entry into a desensitized state without altering the gating of the bound GABAR channel.
FIGURE 16.4. A: GABA opened chloride channels resulting in single and bursting inward currents. B,C: GABA (2 µm) with diazepam (50 nm) resulted in increased opening frequency. D: GABA with DMCM (50 nm) resulted in decreased opening frequency. Data were obtained from different excised outside-out patches. Time and current calibration bars are applicable to all traces shown. (From Rogers CJ, Twyman RE, Macdonald RL. Benzodiazepine and β-carboline regulation of single GABAA receptor channels of mouse spinal neurones in culture. J Physiol 1994;475:69-82, with permission.)
Reduction of GABAR currents by an inverse agonist for the BDZ receptor is produced by a mechanism opposite to the action of BDZ receptor agonists. The inverse agonist β-carboline, DMCM, did not alter GABAR conductance or average open and burst durations (66), but it did reduce open and burst frequencies. These results suggest that modulation of GABAR single-channel kinetics by DMCM could be explained by a reduction of the affinity of GABA binding at the first, but not second, GABA binding site. Again, an alternative interpretation is that β-carbolines increase the rate of entry into a desensitized state without altering the gating of the bound GABAR.
OTHER ACTIONS OF BENZODIAZEPINES
Although the primary action of anticonvulsant BDZs is to enhance GABAergic inhibition by binding to the BDZ receptor on GABARs, other actions of BDZs have been described. BDZs have been demonstrated to modify sodium channel function in a manner similar to that of the anticonvulsants phenytoin, carbamazepine, and sodium valproate (46). Diazepam has been shown to block trains of action potential in rat diaphragm muscle fibers and to decrease peak inward sodium current of frog myelinated nerve fibers. This effect on the sodium channel produces a voltage-dependent block of high-frequency repetitive firing of cultured mammalian neurons (50). The effect of diazepam to limit repetitive firing occurred at diazepam concentrations achieved in the treatment of status epilepticus and was not blocked by the BDZ receptor antagonist flumazenil. Thus, the effects of BDZs on sodium channels are not mediated through the BDZ receptor, but instead represent direct effects of BDZs on sodium channels.
BDZs have also been demonstrated to reduce voltage-dependent calcium currents (69). Diazepam and the convulsant Ro-54864 reduced the duration of voltage-dependent calcium currents in mouse neurons in cell culture. However, the effect was produced at supertherapeutic concentrations of diazepam not likely achieved in ambulatory patients. Similarly, diazepam reduced the calcium conductance in identified leech neurons (35) and guinea pig myenteric neurons (19) at supertherapeutic concentrations. Thus, although BDZs decrease voltage-dependent calcium currents in neurons, the effect is unlikely to be clinically relevant because of the high concentrations of BDZs that are required.
There have been additional reports of BDZ interactions with other channels and neurotransmitter receptors. BDZs were reported to enhance calcium-mediated potassium conductance (17), and they reduced the effect of excitatory amino acids (3,25). However, these observations have not been substantiated, and it is generally accepted that BDZs do not have significant interactions with excitatory amino acid receptors or potassium channels.
It is firmly established that BDZs have primary action as anticonvulsants by interacting with GABARs at the BDZ binding site and allosterically modifying GABAR current to enhance inhibition. However, the great advances made in understanding the molecular biology of the GABAR demonstrate that there are multiple GABAR isoforms with differential sensitivity to BDZ receptor ligands. Thus, BDZ receptor ligands targeted to BZI, BZII, or BZIII receptors have different clinical actions. Studies are currently under way to determine specific GABAR isoforms in the central nervous system. Once this is accomplished, then specific anticonvulsant drugs will be developed that will target those specific GABAR isoforms. It is certainly possible that in the future there will be BDZ receptor ligands with selective anxiolytic or anticonvulsant actions without undesired side effects.
Although BDZs have their primary action to enhance GABAR current, a significant interaction of BDZs with sodium channels has been reported. This interaction occurs only at concentrations of BDZs that are achieved in the treatment of status epilepticus. The effects of BDZs at these high concentrations are similar to those of phenytoin, carbamazepine, and sodium valproate, findings suggesting that BDZs may recruit a second mechanism of action when these drugs are given in high doses to patients with status epilepticus.