Photomotor Responses in Zebrafish and Electrophysiology Reveal Varying Interactions of Anesthetics Targeting Distinct Sites on γ-Aminobutyric Acid Type A Receptors

Zebrafish (Danio rerio, Tubingen AB strain) were used with approval from the Massachusetts General Hospital Institutional Animal Care and Use Committee (protocol No. 2014N000031) and in accordance with Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines. Adult zebrafish were mated to produce embryos and larvae as needed. Embryos and larvae were maintained in Petri dishes (140 mm diameter, fewer than 100 per dish) filled with E3 medium (in mM: 5.0 NaCl, 0.17 KCl, 0.33 CaCl₂, 0.33 MgSO₄, 2 HEPES, pH 7.4) in a 28.5°C incubator under a 14/10 h light/dark cycle. The light intensity inside the incubator ranged from 3,000 to 4,000 lux (measured with a calibrated light meter from Thermo Fisher, USA). Experiments were performed on larvae at 7 days after fertilization. After use in experiments or at 8 days after fertilization, larvae were euthanized.

Xenopus laevis oocytes were used in electrophysiology experiments. Female frogs were housed in a veterinarian-supervised facility and used with approval from the Massachusetts General Hospital Institutional Animal Care and Use Committee (Boston, Massachusetts; protocol No. 2005N000051). Oocytes were harvested via mini-laparotomy from frogs anesthetized by immersion in 0.2% Tricaine, in accordance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health (Bethesda, Maryland).

Propofol (2,6 di-isopropyl phenol), GABA, and salts were purchased from Sigma-Aldrich (USA). R-etomidate (a gift from Douglas E. Raines, MD, Massachusetts General Hospital) was synthesized by Bachem America (USA). R-mTFD-MPAB was synthesized by KareBay Biochem Inc. (USA). Alfaxalone was purchased from Steraloids Inc. (USA). Anesthetic stock solutions in dimethylsulfoxide were stored at –20ºC and diluted to a final concentration in experimental buffers on the day of use. cDNAs for human GABA_A receptor wild-type α1, β3, and γ2L subunits and mutants (α1S270I and γ2LS280W) were cloned in pCDNA3.1 vectors.^15,17  Capped messenger RNAs were synthesized on linearized cDNA templates using mMessage mMachine kits (Sigma-Aldrich) and stored in nuclease-free water at –80°C.

Binarized zebrafish larvae motor responses to photic stimuli were measured in quadruplicate for each animal. Development of the inverted photomotor response assay used approximately 2,000 zebrafish larvae. We next established hypnotic drug IC50s by varying anesthetic concentrations (n = 752 to 1,680 larvae per drug, for four anesthetics). Drug pair experiments were then performed with one drug fixed at a fraction of its IC50 and the other varied. Each such drug pair study was used to calculate a pairwise interactive parameter, α (n = 240 to 480 larvae per α determination). Six or seven such α values under different fixed drug conditions were combined for isobolographic analysis of each drug pair (n = 1,440 to 1,920 larvae per isobologram). Six anesthetic pairs were studied.

Drug interactions in α1β3γ2L GABA_A receptors were studied using voltage clamp electrophysiology in Xenopus oocytes. We compared anesthetic modulation of receptors activated by GABA versus another anesthetic as primary agonist, based on fitted log(d) values from allosteric coagonist shift analyses. The ratio of log(d) values indicated whether anesthetic pairs interacted in accord with coagonist models. For these experiments, estimated open probability was assessed at 9 to 11 primary agonist concentrations in five oocytes per concentration–response curve. Concentration-dependent activation by four primary agonists (GABA, etomidate, propofol, and R-mTFD-MPAB) and 11 studies of anesthetic modulation (4 with GABA, 3 with etomidate, 3 with propofol, and 1 with R-mTFD-MPAB) were assessed (n = 75 oocytes in total). Another set of experiments assessed anesthetic additivity in receptors activated by low GABA, using an isobolographic approach (three drug pairs; n = 6 oocytes per anesthetic pair).

Single zebrafish larvae (7 days after fertilization, sex indeterminate) were randomly selected and placed into wells of standard 96-well plates containing 150 µl E3 buffer. Anesthetic stocks were diluted in E3 buffer to four times final concentrations, and 50 µl of four times solution was transferred to each well using a multi-pipetter, bringing the final well volumes to 200 µl. Drug transfers took less than 5 min per plate. All control and final drug solutions contained 0.1% dimethylsulfoxide. Each 96-well plate was placed in a Zebrabox (Viewpoint Life Sciences, Canada) and maintained at 28°C in a thermostatically controlled water bath. The duration of acclimation to the Zebrabox environment was varied during assay development and set at 30 min for experiments. During experiments, Zebralab v5.15 software (Viewpoint Life Sciences‚ USA) coordinated stimuli while recording and integrating the motor activity of individual larvae during programed epochs, using infrared video data as previously described.¹⁸ 

To assess the hypnotic effects of anesthetics singly and in pairs for isobolographic analyses, we used an inverted photomotor response assay that provided a baseline response probability near 100%. Larvae in groups of eight were placed in E3 buffer containing no drug or a specified drug solution. Larvae were acclimated in a Zebrabox illuminated with white light at 4,000 lux for 30 min while spontaneous activity stabilized. During inverted photomotor response trials, baseline activity was measured for 10 s followed by a 1 s exposure to darkness (less than 15 lux) followed by renewed exposure to bright white light. Trials were repeated 4 times (up to 10 times during assay development) with 3 min recovery periods between trials. Activity was integrated over 0.5-s epochs during baseline periods and 0.1-s epochs during each dark flash. For analysis, activity values were normalized to epoch duration. To establish binary outcomes, the means and SDs for normalized activity during prestimulus basal periods (80 total epochs) were calculated for individual larvae. Binary responses for each trial were scored as positive (1) if normalized activity in any of the ten 0.1-s epochs during a dark flash exceeded the upper 99% CI (mean + 2.8 × SD, using a Bonferroni adjustment for four comparisons) for normalized basal activity. Otherwise, the photomotor response was scored as negative (0). Cumulative response probability (P_Resp) for each larva was calculated by averaging its four binary trial results. For statistical analyses, results from all larvae in each exposure group were pooled.

P_Resp results for all larvae within exposure groups (mean ± 95% CI; n > 24 per condition; 240 to 1,680 larvae per concentration–response study) were plotted against log[drug]. The data were fitted to three-variable logistic functions (eq. 1) using nonlinear least-squares in GraphPad Prism version 8 (GraphPad Software, USA).

Logistic fits calculated mean and standard errors values for maximum response probability at 0 drug (P_max), log half-maximal inhibitory concentration (log[IC₅₀]), and Hill slope (nH). The software also calculated the log drug concentration at which P_Resp = 0.5 (log[IC50]), and 95% CI for IC₅₀ and IC50 and other parameters. Note that IC₅₀ and IC50 (drug concentration resulting in 50% response probability) are identical when P_max = 1.0 but diverge when P_max < 1.0.

We adopted the approach used by Kent et al.¹⁴  IC50 values for photomotor response inhibition by single drugs were determined multiple times. An initial set of single-drug IC50s was established to guide subsequent experiments on drug pairs. In drug pair experiments, one drug was held at a fixed fraction (0.25, 0.5, or 0.75) of its IC50 while the other drug was varied. Each pair was studied with both drugs fixed at three or more values and the other varied, resulting in six or more separate concentration–response studies per combination. The IC50 for the varied drug (at P_Resp = 0.5) was calculated as described (n = 240 to 480 larvae for each concentration–response; n = 1,440 to 1,920 per combination). Single-drug experiments with the varied drug were performed at least once on the same day as drug pair experiments, using larvae from the same batches. These single-drug data were then combined with earlier results, and an updated overall single-drug IC50 was calculated (final n = 752 to 1,680 larvae per drug). After all pairs were studied, fixed drug concentrations were recalculated as fractions of final IC50s.

Two-dimensional isobolograms display plotted pairs of drug concentrations, each normalized to its single-drug IC50, that produce equivalent effects, namely P_Resp = 0.5. The sum of normalized drug concentrations is an interaction index, α (eq. 2). If α is 1.0, the drug interaction is additive, and if α is less than 1.0, the interaction is supra-additive (synergistic).¹⁹ 

Harvested Xenopus oocytes were treated with collagenase and defolliculated as previously described,¹⁵  then microinjected with 0.5 to 1.0 ng of mRNA mixtures in ratio 1α:1β:5γ. Before use in electrophysiologic experiments, injected oocytes were maintained for up to 48 h at 18°C in ND-96 (in mM: 96 NaCl, 2 KCl, 1.8 CaCl₂, 1 MgCl₂, 5 HEPES, pH 7.4) supplemented with 0.05 mg/ml gentamicin and 0.01 mg/ml ciprofloxacin.

Two-microelectrode voltage clamp electrophysiology was performed at room temperature (21 to 23°C) as previously described.¹⁵  Oocytes in a 0.2 ml flow chamber were impaled with glass microelectrodes filled with 3 M KCl (tip resistance 0.5 to 2 MΩ) and clamped at –50 mV (model OC-725C; Warner Instruments‚ USA). Gravity-driven solutions in ND-96 were delivered at 4 ml/min from glass syringes via polytetrafluoroethylene tubing and a micro-manifold. LabScribe 2 software and an RA834 interface (both from iWorx, USA) coordinated timing of solutions through valves and digitized currents at 200 Hz. Current data were digitally filtered with a 10-Hz low-pass Bessel function and analyzed offline using Clampfit8 software (Molecular Devices, USA).

Voltage clamp electrophysiology in Xenopus oocytes expressing α1β3γ2L GABA_A receptors was used to compare anesthetic modulation of GABA-activated receptors and receptors activated with another anesthetic as primary agonist (etomidate, propofol, or R-mTFD-MPAB). Modulation of primary agonist responses was tested in the presence of 3.2 µM etomidate, 5 µM propofol, 2.5 µM alfaxalone, or 8 µM R-mTFD-MPAB (n > 5 oocytes per condition), which similarly modulate GABA-activated GABA_A receptors.²⁰  Voltage clamped oocytes expressing GABA_A receptors were exposed to solutions of variable primary agonist concentrations alternating with maximal agonist controls (3 mM GABA, 1 mM etomidate, 300 µM propofol, or 150 µM R-mTFD-MPAB) with at least 5 min of ND-96 wash between recordings. After baseline correction, peak currents were normalized to the average of preceding and following control responses.

For modulation experiments, oocytes were pre-exposed to the modulating drug for 10 s before activation with primary agonist plus modulator. Peak currents were normalized to the average of preceding and following responses to maximal agonist plus modulator.

To explore whether the two R-mTFD-MPAB binding sites (α+/β– and γ+/β- transmembrane interfaces) differentially contributed to etomidate–R-mTFD-MPAB interactions, we studied etomidate-dependent activation in the absence and presence of 8 µM R-mTFD-MPAB in oocytes expressing α1S270Iβ3γ2L and α1β3γ2LS280W GABA_A receptors, using the same approach used for wild-type receptors.

Estimated open probability P_open was calculated as previously described.¹⁵  This calculation explicitly adds spontaneous basal channel activity and re-normalizes experimental peak currents to account for maximal agonist efficacy, based on anesthetic-enhanced maximal GABA-induced currents that are assumed to represent 100% activated channels (eq. 3).

Spontaneous channel activation was assessed in wild-type and mutant GABA_A receptors by measuring outward currents from oocyte-expressed receptors exposed to 2 mM picrotoxin, which inhibits channel activity. Picrotoxin-sensitive currents were normalized to maximal inward currents elicited with 3 mM GABA (⁠$IPTXIGABAmax$⁠; n = 5 oocytesper receptor type). Maximal agonist efficacies in wild-type receptors were assessed by activating oocyte currents with maximally activating concentrations and normalizing to currents elicited with 3 mM GABA plus a modulator (⁠$IGABA+ModmaxIGABAmax$⁠; n = 5 oocytes per condition). Alfaxalone- or etomidate-enhanced GABA responses were used to assess GABA and etomidate maximal efficacy in α1S270Iβ3γ2L and α1β3γ2LS280W mutant receptors.

P_open data (mean ± 95% CI; n ≥ 5 per condition) were plotted against log([Agonist]) and parametrically analyzed by fitting to four-variable logistic equations (eq. 4) using nonlinear least-squares (GraphPad Prism). Fitted parameters reported from these fits include log(EC₅₀), which was used to calculate 95% CI for EC₅₀, nH, P_min (non-zero in some mutant receptors), and P_max. Pairwise parameter comparisons between fits to different data sets were performed using F tests with P = 95% as a significance threshold.

To compare allosteric gating energy shifts caused by modulating anesthetics with different primary agonists, we utilized Monod-Wyman-Changeux log(d) shift analysis as described previously.¹⁵  This approach simultaneously fits agonist-dependent P_open results in the absence and presence of a specified modulator concentration to a simplified two-state (resting and active) coagonist model (eq. 5).¹⁵ 

L₀ is the equilibrium constant for resting:active receptors in the absence of ligands. K_Ag is the dissociation constant for primary agonist binding to resting receptors, and c is the ratio of dissociation constants in active versus resting receptors, reflecting agonist efficacy. Log(d) is a fitted model parameter that is proportional to the allosteric gating free energy contributed by the modulating drug at its specified concentration. Modulator-induced EC₅₀ shifts and changes in both P_min and P_max are incorporated into the single log(d) parameter, making it suitable for quantitative comparison of modulator effects on receptors activated with different primary agonists or in different receptor types.^15,21  For analyses of results in wild-type, we compared log(d, GABA) for a modulator when GABA was the primary agonist to log(d, AN) when another anesthetic was the primary agonist. Fits to allosteric shift models designated the modulator as a binary factor (0 if absent, 1 if present) and were performed using nonlinear least-squares in Origin 6.1 (OriginLab, USA). The models were constrained to two primary agonist (GABA or anesthetic) sites and a single modulatory (coagonist) site. L₀ was constrained to a value of 25,000 for wild-type receptors²¹  and for mutant receptors was calculated based on spontaneous P_open estimates (L₀ = [1 – P₀]/P₀).

All the anesthetics we studied act as coagonists with GABA in α1β3γ2L GABA_A receptors, resulting in supra-additive activation. Two-state coagonist models predict that a modulator at a selected concentration should result in the same log(d) value with any primary agonist. Thus a ratio of log(d, AN):log(d, GABA) near 1.0 is indicative of strict energy additivity and activation synergy between the coagonist modulator and the anesthetic primary agonist. In calculating log(d) ratios, standard errors were propagated as described by Bevington and Robinson.²²  To estimate log(d) values for additive (competitive) drug interactions, we used the primary agonist logistic model for etomidate and calculated the effect of adding 3.2 µM etomidate as a modulator to the varied etomidate agonist concentrations. The etomidate agonist model combined with its “concentration-enhanced” model (20 points each) was then fitted with equation 5. The resulting log(d) value (–0.008 ± 0.010) was not significantly different from zero. Thus, a log(d, AN) near zero is indicative of a strictly additive interaction between the modulator and the primary agonist anesthetic, consistent with competition for shared agonist sites.

We also electrophysiologically assessed whether etomidate combined with R-mTFD-MPAB, propofol, or alfaxalone produced supra-additive modulation of low GABA responses in Xenopus oocytes expressing α1β3γ2L receptors. These experiments used approximately equieffective drug concentrations that were close to the hypnotic IC50s observed in zebrafish larvae. We first compared the effects of 10 µM GABA (approximately 8% maximal activation) alone and combined with each drug individually at its zebrafish IC50 in independent sets of three oocytes. We then adjusted anesthetic concentrations to produce responses to 10 µM GABA plus anesthetic near 40% of maximal GABA activation (3 mM) in additional sets of oocytes. This process identified 0.50 µM etomidate, 1.70 µM propofol, 1.03 µM alfaxalone, and 1.85 µM R-mTFD-MPAB as approximately equally modulatory, based on one-way ANOVA and Tukey multiple comparisons (n = 5 each).

We next tested whether drug pairs were supra-additive in separate sets of six oocytes for each combination. To illustrate, in one set of oocytes, we measured currents elicited with 10 µM GABA plus 0.5 µM etomidate, 10 µM GABA plus 1.03 µM alfaxalone, and 10 µM GABA plus 0.25 µM etomidate plus 0.51 µM alfaxalone (a 1:1 mixture of the first two solutions). The other drug combinations were similarly tested in separate groups of six oocytes. Anesthetics were preapplied to oocytes for 10 s before activation using combinations of GABA plus anesthetics, and normalizing responses to 3 mM GABA were recorded both before and after each GABA plus anesthetic response. To control for the order of drug applications, each oocyte in a set of six was studied using a different permutation of the three drug-containing solutions.

We used data from electrophysiologic concentration–responses in the presence and absence of modulators to estimate enhancement ratios at agonist concentrations eliciting 1% of maximal response. Primary agonist concentration responses were examined to identify concentrations that produced activation with P_open in the range 0.5 to 1%. P_open data from studies with modulator plus the same primary agonist concentration were then identified. Each data set was used to calculate mean ± SD, and the enhancement ratio was calculated as P_open with modulator:P_open without modulator. SDs for ratios were propagated according to Bevington and Robinson.²² 

Hypnotic concentration–response studies in zebrafish larvae were performed in multiple groups of randomly selected larvae (sex indeterminate) from at least three separate breeding clutches, and included 9 to 11 concentrations (n = 24 to 48 larvae per experimental condition; n = 240 to 480 per study). No a priori statistical power calculations were conducted. The number of zebrafish larvae studied was based on our previous experience with this approach. Photomotor response outcomes in individual zebrafish larvae were binarized based on quantitative and automated motion analysis, eliminating observer bias. Investigator blinding was not used in performance of experiments or analyses. Each animal was tested four times. No corrections were made for larvae that may have died during experiments. For isobolographic analyses of each drug pair, six or seven conditions from separate sets of trials were tested, with one drug fixed and the other varied. Controls with no drugs were also included in every 96-well plate to confirm high photomotor response probability. Between 1,440 and 1,920 zebrafish larvae (5,760 to 7,680 photomotor response trials) were studied to generate multiple independent α values for each of the six drug pairs, from which averages and variances were calculated. Each set of α values was compared to the additivity model (α = 1.0) using a single-sample Student’s t test. Supra-additivity was inferred for a drug pair when the upper 95% CI for mean α was less than 1.0. This also corresponded with the majority of CI for individual α values in a drug pair set less than 1.0.

In electrophysiologic experiments, the number of oocytes studied for each condition (n ≥ 5) was based on our extensive previous experience with this method. Oocytes were randomly assigned to groups and tested for adequate and stable receptor expression before experiments. Results were normalized to frequently measured within-oocyte controls to account for variations in functional receptor expression levels. Nonlinear least-squares fits to normalized electrophysiologic results (eqs. 4 and 5) were performed in GraphPad Prism v8 or MicroCal Origin 6.1 (MicroCal‚ USA). Statistical comparisons based on ANOVA, F tests, and t tests were performed in GraphPad Prism. The threshold for statistical significance was P < 0.05.

To determine whether anesthetic pairs at low concentrations produced supra-additive effects on GABA_A receptors, we used a variation of isobolography. In sets of six oocytes, we compared within-oocyte averaged normalized peak responses recorded with GABA plus single anesthetics to the normalized response elicited by the 1:1 mixture of GABA plus anesthetics (i.e., GABA plus two anesthetics, each at half concentration), using paired Student’s t tests (GraphPad Prism 8). Supra-additivity was inferred when the combination of two drugs at half concentrations was significantly (P < 0.05) larger than the average responses to individual drugs alone at approximately equally modulating concentrations.

Correlations between zebrafish α values and metrics of receptor modulation ([log(d) and enhancement ratios] were calculated as Pearson r values in GraphPad Prism 8.

Binarized photomotor responses in our previous studies of hypnotic drugs in zebrafish larvae were based on movement of dark-adapted animals after brief exposures to bright white light.¹⁶  Using that approach, control response probabilities were under 90% and diminished with repeated trials. However, consistent control response rates near 100% are desirable for studying pharmacodynamic interactions between sedative–hypnotic drugs using isobolographic analyses. It is known that zebrafish larvae are more active in lighted than dark environments and reliably respond to sudden darkness with vigorous movements.^23,24  We therefore tested response probabilities using an inverted photomotor response assay, with larvae maintained in a lighted environment and exposed to brief periods of darkness (dark flashes). Initial trials showed that inverted photomotor response probabilities were consistently near 95% for zebrafish adapted to white light intensities ranging from 100 to 5,000 lux. We chose to use 4,000 lux white background light to approximate the conditions used in cultivating zebrafish embryos and larvae. In comparison to traditional photomotor responses, inverted photomotor responses resulted in consistently higher initial probabilities of movement and a significantly smaller drop in cumulative response probability with repeated trials at 3-min intervals (fig. 2A). Varying the period of environmental adaptation before testing motor responses to photic stimuli revealed unchanged dark flash response probability with longer adaptation periods in light but diminishing light flash response probability with longer dark adaptation periods (fig. 2B), as previously reported.¹⁶  This suggested that exposure to a dark environment alone caused zebrafish larvae to become sedated and less responsive to photic stimuli, possibly through onset of natural sleep. Consistent with this idea, we found that anesthetic IC50s for sedation (reduced spontaneous activity) in a lighted environment (table 1) were much higher than when measured in dark-adapted larvae¹⁸  and also higher than IC50s for hypnosis (reduced responses to photic stimuli) using either approach (fig. 2C; table 1).

We assessed the interactions of etomidate, alfaxalone, R-mTFD-MPAB, and propofol alone and in paired combinations in 7 days postfertilization zebrafish larvae using inverted photomotor response hypnosis assays. Multiple single-drug concentration–responses, including on days when two drugs were studied in combination, were performed to define IC50 values (table 1). Multiple drug pair interaction studies, with one drug fixed at a fraction of its IC50 and the other drug varied, were performed. IC50s fitted to the varied drug were used to calculate α values (eq. 2).

Figure 3 illustrates experimental data for etomidate as the varied drug alone (red circles) or combined with 0.43 µM R-mTFD-MPAB (~0.24 × IC50; green squares) or 0.3 µM alfaxalone (~0.29 × IC50; blue triangles). For larvae studied with etomidate alone in these trials, the control (no drug) response probability was 94% (95% CI = 92 to 97%]. The fitted etomidate IC₅₀ = 0.36 µM (95% CI = 0.33 to 0.38] results in a 47% (half-maximal) response probability, while the IC50 eliciting 50% response probability (dotted line) is 0.35 µM (95% CI = 0.32 to 0.37]. Unsurprisingly, response rates at 0 etomidate are slightly lower in the presence of R-mTFD-MPAB or alfaxalone. The IC50 for etomidate combined with 0.43 µM R-mTFD-MPAB is 0.23 µM (95% CI = 0.21 to 0.26] and combined with 0.3 µM alfaxalone is 0.15 µM (95% CI = 0.13 to 0.17]. The calculated α value from these data for R-mTFD-MPAB plus etomidate is 0.93 and for alfaxalone plus etomidate is 0.74. Six or more such drug pairs with different drugs held constant were studied to construct isobolograms (fig. 4) and calculate mean ± SD α values for each set of experiments for a drug pair (table 2).

Consistent with previous studies in tadpoles¹⁴  and rodents,¹³  combining the neurosteroid alfaxalone with etomidate, R-mTFD-MPAB, or propofol resulted in synergistic hypnotic interactions with mean α values significantly less than 1.0 (table 2; fig. 4). Combinations of R-mTFD-MPAB plus propofol also resulted in α values indicative of weak synergy (mean ± SD = 0.89 ± 0.110; P = 0.033). However, contrasting with results in tadpoles,¹⁴  our results in zebrafish larvae for combinations of etomidate plus propofol (mean α = 1.0 ± 0.07; P = 0.88) and etomidate plus R-mTFD-MPAB (mean α = 0.96 ± 0.11; P = 0.44) were inconsistent with synergy, and instead consistent with additive interactions.

To test whether our zebrafish results reflect molecular effects of hypnotic combinations in synaptic α1β3γ2L GABA_A receptors, we used two-microelectrode voltage clamp electrophysiology to compare anesthetic modulation of currents activated by GABA (an orthosteric agonist) versus another anesthetic (allosteric agonist). Two-state coagonist models predict that anesthetic modulation of GABA_A receptor activation will be independent of the agonist applied as long as agonist and modulator sites are structurally distinct.^14,25,26  Thus, log(d, AN):log(d, GABA) ratios near 1.0 are predicted for anesthetic pairs acting through distinct sites, and ratios near 0 are consistent with competition for shared sites. Etomidate at 3.2 µM, propofol at 5 µM, R-mTFD-MPAB at 8 µM, and alfaxalone at 2.5 µM were previously shown to produce similar modulation of GABA-dependent receptor activation.^15,20 Figure 5A shows GABA-dependent receptor activation (estimated P_open) in the absence (black circles) and presence of anesthetics. Spontaneous activation of α1β3γ2L receptors was undetectable, indicating basal P_open < 0.0005. GABA EC₅₀ and maximal efficacy values, derived from nonlinear least-squares fits to logistic functions (eq. 4 in Materials and Methods), are reported in table 3. The shifts in GABA responses produced by R-mTFD-MPAB (green squares) or etomidate (red diamonds) appeared larger than those produced by propofol (hexagons) or alfaxalone (blue triangles). At the concentrations used, R-mTFD-MPAB and etomidate resulted in lower GABA EC₅₀ values than alfaxalone or propofol (P < 0.01 by F tests), while all four drugs similarly increased maximal GABA responses. Figure 5B shows the same data subjected to allosteric shift analysis (eq. 5 in Materials and Methods), which calculates log(d) values proportional to the gating energy attributable to modulating drugs. The fitted parameters from these analyses are summarized in table 4. Of note, the fitted values for K and c, reflecting GABA binding affinity and efficacy, are similar in all four shift models. The quality of fits, reflected in R² values (tables 3 and 4), are slightly better for logistic fits than for Monod-Wyman-Changeux models, likely because of the variable Hill slope.

Of particular interest was whether interactions between etomidate and other anesthetics in synaptic GABA_A receptors corresponded to the varying interactions observed in zebrafish experiments (fig. 4) or to predictions of two-state coagonism. Figure 5C shows etomidate-dependent receptor activation in the absence (red diamonds) and presence of alfaxalone (blue triangles), propofol (open hexagons), or R-mTFD-MPAB (green squares) at the same concentrations coapplied with GABA. Etomidate EC₅₀ values and maximal efficacy values from logistic fits are reported in table 3. In contrast to results in GABA-activated receptors, alfaxalone produced larger leftward shifts in etomidate responses than R-mTFD-MPAB or propofol. The etomidate EC₅₀ in the presence of alfaxalone was lower than in the presence of R-mTFD-MPAB (9.6 vs. 19 µM; P = 0.036 by F test) or propofol (9.6 vs., 83 µM; P < 0.0001 by F test), and the maximum efficacy of etomidate plus alfaxalone was higher than that of etomidate plus R-mTFD-MPAB or etomidate plus propofol (0.75 vs. 0.48 or 0.27; both P < 0.0001 by F test). The allosteric shift model fits for etomidate-activated receptor currents are shown in figure 5D, and the fitted parameters are reported in table 4. Log(d, AN):log(d, GABA) derived from shift analysis (table 4) indicated that alfaxalone interacts with etomidate with 87 ± 5.6% of the energy observed with alfaxalone plus GABA. In stark contrast, the interaction between propofol and etomidate resulted in a log(d) value of –0.11 ± 0.028, which was only 8.6 ± 2.1% of the log(d) calculated for propofol modulation of GABA responses. We also assessed the effect of 3.2 µM etomidate on propofol-activated receptor currents and observed a similarly low log(d) value (Supplemental Digital Content fig. S1, http://links.lww.com/ALN/C914; table 4). The R-mTFD-MPAB log(d) for etomidate responses was 38 ± 2.4% of the value for GABA responses, intermediate between alfaxalone and propofol.

We also studied the effects of R-mTFD-MPAB and alfaxalone on propofol-activated GABA_A receptors and the effects of alfaxalone on R-mTFD-MPAB–activated receptors. Results are displayed in Supplemental Digital Content figure S1 (http://links.lww.com/ALN/C914). Logistic fit parameters and results of allosteric shift calculations are reported in tables 3 and 4, respectively. The effects of R-mTFD-MPAB and alfaxalone on propofol-activated currents were similar, producing a six- to sevenfold reduction in propofol EC₅₀, but without increasing maximal responses at high propofol. These results were poorly fit with Monod-Wyman-Changeux allosteric shift models, reflected in low R² values (table 4). R-mTFD-MPAB was a weak agonist of α1β3γ2L GABA_A receptors, and at up to 30 µM, R-mTFD-MPAB–activated currents were enhanced up to 20-fold by alfaxalone. At higher R-mTFD-MPAB concentrations in the presence of alfaxalone, inhibition in the form of surge currents was observed. The alfaxalone log(d) for R-mTFD-MPAB–activated receptors was 55 ± 7.0% of the value for GABA-activated receptors.

For all six drug pairs tested, correlation between zebrafish α values and log(d) ratios resulted in a Pearson r = –0.46 (P = 0.35).

Figure 5E displays representative current traces recorded from oocytes used in experiments shown in figure 5,C and D. These revealed additional details about the interactions of etomidate with R-mTFD-MPAB, propofol, and alfaxalone in α1β3γ2L receptors. Traces elicited with high etomidate (100 µM) alone showed channel activation and deactivation only. With the addition of 2.5 µM alfaxalone, peak currents were amplified about fourfold, and very small surge currents were evident during drug washout. In contrast, addition of 8 µM R-mTFD-MPAB to 100 µM etomidate increased peak current about twofold, and large surge currents were evident after drug exposure ended. Addition of 5 µM propofol to 100 µM etomidate increased peak currents by only approximately 15% without inducing surge currents.

The receptor-activating anesthetic concentrations used in coagonist shift experiments were mostly higher than those present in zebrafish hypnosis experiments. We assessed anesthetic interactions at lower concentrations in two ways. Concentration eliciting 1% of maximal response enhancement ratios (table 3) were calculated from subsets of data generated for allosteric shift analysis, specifically modulation of anesthetic concentrations activating approximately 1% of maximal GABA currents. These enhancement ratios were largest for etomidate plus alfaxalone, and lower but similar for R-mTFD-MPAB plus alfaxalone, propofol plus R-mTFD-MPAB, and propofol plus alfaxalone. The two drug pairs that appeared additive in zebrafish hypnosis assays (etomidate plus R-mTFD-MPAB and etomidate plus propofol) produced the two lowest enhancement ratios.

We also assessed interactions between etomidate plus R-mTFD-MPAB, etomidate plus alfaxalone, and etomidate plus propofol in α1β3γ2L receptors activated with low GABA (10 µM) using concentrations close to zebrafish hypnosis IC50 values. We found that 10 µM GABA combined with 0.5 µM etomidate, 1.03 µM alfaxalone, 1.85 µM R-mTFD-MPAB, or 1.70 µM propofol, all elicited peak currents that were approximately 40% of maximal GABA (3 mM) responses. To test anesthetic interactions, we compared the average of two currents elicited with GABA plus single anesthetic solutions to the current elicited with GABA plus both anesthetics at half concentrations (n = 6 oocytes per drug pair). Figure 6A shows results for alfaxalone and etomidate. The normalized currents elicited with GABA plus half each alfaxalone plus etomidate (mean ± SD = 0.45 ± 0.079) were consistently higher than the within-oocyte averages of normalized currents elicited with GABA plus alfaxalone and GABA plus etomidate (0.35 ± 0.062; P = 0.023 by paired Student’s t test), indicating a supra-additive interaction. Based on the within-oocyte ratios of currents with half each alfaxalone plus etomidate to the averages of single-anesthetic responses (1.28 ± 0.25), the degree of synergy was not large. Figure 6B shows results for R-mTFD-MPAB and etomidate in another group of oocytes. Within-oocyte normalized currents elicited with GABA plus half each R-mTFD-MPAB plus etomidate (0.43 ± 0.043) were mostly higher than averaged results elicited with GABA plus R-mTFD-MPAB and GABA plus etomidate (mean ± SD = 0.38 ± 0.038; ratio = 1.15 ± 0.17; n = 6), but the results did not reach statistical significance (P = 0.064 by paired Student’s t test). Figure 6C shows results for propofol and etomidate. The normalized currents elicited with GABA plus half each propofol plus etomidate (0.27 ± 0.035) were very similar to average within-oocyte currents elicited with GABA plus propofol and GABA plus etomidate (mean ± SD = 0.26 ± 0.050; ratio = 1.0 ± 0.14; n = 6; P = 0.86 by paired Student’s t test). These results are consistent with an additive drug interaction. In summary, our studies of anesthetic interactions in GABA-activated α1β3γ2L receptors echoed findings in zebrafish larvae at comparable drug concentrations. Combining alfaxalone and etomidate produced modest but significant supra-additive effects, while R-mTFD-MPAB plus etomidate or propofol plus etomidate interacted additively.

The appearance of surge currents after coapplication of etomidate plus R-mTFD-MPAB (fig. 5E) indicated that both activation and inhibition were induced in α1β3γ2L GABA_A receptors. We considered whether the two distinct R-mTFD-MPAB subsites might differentially contribute to these opposing effects, as observed for diazepam by McGrath et al.²⁷  We therefore tested the effects of 8 µM R-mTFD-MPAB on etomidate-activated GABA_A receptors containing a mutation in either the α+/β– or γ+/β– site. Results are shown in figure 7.

Both α1S270Iβ3γ2L and α1β3γS280W receptors displayed spontaneous activation of approximately 2%, consistent with previous results.¹⁵  Based on logistic fits, etomidate as an agonist in α1S270Iβ3γ2L receptors (fig. 7, open red circles) is characterized by EC₅₀ = 19 µM and high efficacy (P_max ≈ 1.13), while in α1β3γS280W receptors (solid red circles), etomidate activates with lower potency (EC₅₀ = 200 µM) and lower efficacy (P_max ≈ 0.40). In α1S270Iβ3γ2L receptors, addition of 8 µM R-mTFD-MPAB (open green squares) did not significantly alter peak currents elicited by etomidate at any tested concentration, and etomidate EC₅₀ was not significantly changed (P = 0.45 by F test). In α1β3γ2LS280W receptors, addition of 8 µM R-mTFD-MPAB (solid green squares) moderately reduced currents elicited by etomidate (P_max ≈ 0.28; P = 0.15 by F test) without altering apparent potency (EC₅₀ = 160 µM; P = 0.73 by F test). Shift analysis was also applied to quantify the allosteric effects of 8 µM R-mTFD-MPAB in the two mutant GABA_A receptors. In α1S270Iβ3γ2L receptors, 8 µM R-mTFD-MPAB resulted in log(d) = –0.36 ± 0.71, indicating no significant overall gating effect. In α1β3γ2LS280W receptors, 8 µM R-mTFD-MPAB resulted in log(d) = 0.20 ± 0.038, indicating a small but significant inhibitory effect.

We also tested whether R-mTFD-MPAB directly activated the mutant GABA_A receptors (data not shown). At concentrations up to 300 µM, R-mTFD-MPAB activated less than 1% of maximal GABA currents in α1β3γ2S280W receptors. R-mTFD-MPAB at concentrations above 10 µM activated up to 18% of α1S270Iβ3γ2L receptors.

We tested the hypothesis, based on two-state allosteric coagonist models, that pairs of four anesthetics (etomidate, propofol, R-mTFD-MPAB, and alfaxalone) acting through distinct sets of GABA_A receptor sites would interact synergistically in both zebrafish larvae hypnosis assays and electrophysiologic measures of α1β3γ2L receptor activation.^11,14 

Hypnotic drug interactions in zebrafish larvae were not consistently synergistic. All three anesthetic pairs containing alfaxalone displayed synergy in zebrafish, agreeing with previous studies of similar neurosteroids in tadpoles and rodents.^13,14  Propofol plus R-mTFD-MPAB also displayed synergy in zebrafish larvae. However, etomidate plus R-mTFD-MPAB and etomidate plus propofol pairs were additive (not synergistic) in zebrafish, contrasting with tadpole results.¹⁴  When observed, anesthetic synergy in zebrafish was also generally weaker than that reported in Xenopus tadpole righting reflex tests, reflected in higher α values (table 2).¹⁴  This difference probably reflects lower drug IC50 values for zebrafish inverted photomotor responses versus tadpole righting reflexes along with differences in species and neurologic development. These factors may also have contributed to the contrasting results in zebrafish and tadpoles for etomidate plus R-mTFD-MPAB and etomidate plus propofol.

Based on allosteric shift analysis of α1β3γ2L receptor activation assessed electrophysiologically (table 4), etomidate plus alfaxalone interacted in close accord with predictions of allosteric coagonism [log(d) ratio near 1.0], but the other five drug pairs did not. Notably, etomidate plus propofol interacted with log(d) near 0, consistent with competition for shared sites. Thus, most of our log(d) results appear to conflict with previous electrophysiologic studies reporting drug interactions fully in accord with allosteric coagonism. However, most of the drug interactions that we studied in α1β3γ2L receptors have not previously been reported. Additionally, electrophysiologic experiments by others used α1β2γ2L receptors formed from concatenated subunit assemblies and low concentrations of agonists and coagonists.^11,12  For all six drug pairs in our study, no significant correlation was found between log(d, AN):log(d, GABA) ratios and zebrafish α values. Enhancement ratios at agonist concentrations eliciting 1% maximal response (table 3), based on results at relatively low anesthetic concentrations, correlated better than log(d) ratios with zebrafish α values (Spearman r = –0.66; P = 0.15).

We also examined the three drug pairs that included etomidate in receptors activated with low GABA and low anesthetic concentrations near zebrafish hypnotic IC50s. Under these conditions, alfaxalone plus etomidate interacted supra-additively, while R-mTFD-MPAB plus etomidate and etomidate plus propofol were additive (fig. 6). Thus, these three low concentration drug pair interactions in synaptic receptors correlate well with zebrafish photomotor response results.

Oocyte current recordings (fig. 5E) provided evidence that different hypnotic pair interactions in GABA_A receptors induce different functional state mixtures. Currents elicited with etomidate alone or combined with alfaxalone or propofol showed primarily activation and deactivation phases, consistent with two dominant functional states: resting and active. In contrast, during etomidate plus R-mTFD-MPAB washout, we observed large surge currents indicating reactivation from additional nonconductive states. Channel blockade by anesthetics is a possible explanation, but neither etomidate nor R-mTFD-MPAB alone inhibits receptors at the relevant concentrations. Another explanation is receptor desensitization. GABA_A receptors can reopen during recovery from desensitization, prolonging deactivation and generating surge currents.²⁸  We speculate that R-mTFD-MPAB enhanced both activation and desensitization by high etomidate concentrations.⁶  Nonetheless, under nondesensitizing conditions, low concentrations of R-mTFD-MPAB and etomidate also cooperated more weakly than alfaxalone and etomidate (fig. 6). We also explored the possibility that the α+/β– and γ+/β– binding sites might mediate opposing R-mTFD-MPAB effects, finding that both α+ (S270I) and γ+ (S280W) mutations essentially eliminated R-mTFD-MPAB modulation (fig. 7) of etomidate-activated receptors. The log(d) shifts for etomidate plus R-mTFD-MPAB in mutant receptors were similar to those for GABA plus R-mTFD-MPAB,¹⁵  suggesting that R-mTFD-MPAB binding at both β- sites cooperatively enhances activation without selectively affecting inhibition.

Overall, our results indicate that anesthetic interactions in GABA_A receptors were not consistently explained by two-state allosteric coagonism. Qualitatively, these models correctly predicted synergistic interactions in both zebrafish and synaptic GABA_A receptors for four hypnotic drug pairs at low concentrations: alfaxalone plus etomidate, alfaxalone plus propofol, alfaxalone plus R-mTFD-MPAB, and propofol plus R-mTFD-MPAB. However, for higher drug concentrations, this mechanism quantitatively approximated only the interaction of alfaxalone plus etomidate. Indeed, our findings suggest that allosteric coagonist models best describe the interactions of agonists and coagonists that selectively bind in β+/α– subunit interfaces: GABA, etomidate, and alfaxalone. The interactions of etomidate plus R-mTFD-MPAB and etomidate plus propofol in both zebrafish and GABA_A receptors clearly disagreed with coagonism. Receptor structure–function¹⁵  and cryoelectron microscopy⁵  studies provide alternative frameworks for interpreting these drug pair interactions.

The weak interaction of etomidate plus R-mTFD-MPAB is explained by cryoelectron microscopy images of GABA_A receptors with etomidate occupying β+/α– sites, showing collapse of α+/β– and γ+/β– cavities where R-mTFD-MPAB binds.⁵  The cryoelectron microscopy structures are thought to represent desensitized–inactive receptors but likely maintain open–active state features, because activation and desensitization involve rearrangement of different ion channel regions.^17,29  Allosteric shift analyses of mutant function also indicate that etomidate binding may weaken R-mTFD-MPAB interactions with GABA_A receptors. Mutations in etomidate sites weaken R-mTFD-MPAB modulation, and mutations in R-mTFD-MPAB sites weaken etomidate modulation.¹⁵  This negative cross-talk echoes our current findings that cooperativity is much weaker between these drugs than between either drug and GABA.

Etomidate and propofol interacted additively in both zebrafish and α1β3γ2L receptors, consistent with competition for the same coagonist sites. Indeed, evidence from photolabeling, mutant analyses, substituted cysteine modification and protection, and cryoelectron microscopy all indicate that propofol and etomidate bind at overlapping sites in transmembrane β+/α– inter-subunit pockets.^5,20,30–33  However, most studies also locate propofol binding in β– sites overlapping those for R-mTFD-MPAB. Assuming that propofol binds at the two R-mTFD-MPAB (β–) and the two etomidate (β+) sites and that all four sites contribute equal gating energy, one might expect a log(d) ratio near 0.5, reflecting coagonism between etomidate and propofol at the β– sites. In contrast, cryoelectron microscopy shows propofol binding only in β+/α– pockets.⁵  Also, mutations in β+/α– sites nearly obliterate modulation by both etomidate and propofol, while α+/β– and γ+/β– mutations produce smaller effects.¹⁵  Thus, the β+/α– sites may mediate nearly all of the allosteric gating effects of propofol, and our initial assertion that propofol and etomidate act via different sets of sites needs reconsideration.

Given that propofol and etomidate both act through β+/α– sites in GABA_A receptors, one might also expect that the combination of propofol plus R-mTFD-MPAB would interact similarly to etomidate plus R-mTFD-MPAB (i.e., weakly). However, the 1% response enhancement ratio for propofol plus R-mTFD-MPAB was larger than that for etomidate plus R-mTFD-MPAB, and similar to that for propofol plus alfaxalone (Supplemental Digital Content fig. S1, http://links.lww.com/ALN/C914, table 3). Both propofol plus R-mTFD-MPAB and propofol plus alfaxalone pairs also displayed synergy in zebrafish. Cryoelectron microscopy suggests why the propofol plus R-mTFD-MPAB interaction is more cooperative than the etomidate plus R-mTFD-MPAB interaction. In contrast to etomidate, propofol bound in the outer β+/α– sites is associated with accessible α+/β– and γ+/β– pockets where R-mTFD-MPAB binds,⁵  suggesting that propofol and R-mTFD-MPAB can simultaneously and cooperatively occupy their distinct binding sites.

This study has several limitations that influence comparison of results in zebrafish versus tadpoles and GABA_A receptors. The nervous systems of larval zebrafish are not fully developed, and generalizability of our findings needs testing in older animals and in other species. Our previous photomotor response studies in larval zebrafish showed strong correlations with tadpole results.¹⁶  However, our α values for drug interactions based on inverted photomotor response results differ quantitatively from results in Xenopus tadpoles.¹⁴  For example, etomidate and R-mTFD-MPAB were weakly cooperative in GABA_A receptors and additive in zebrafish, but synergized in tadpoles. There are multiple advantages of zebrafish over tadpoles for hypnotic drug assays,¹⁶  while one disadvantage is the slightly lower control response probability. Inverted photomotor response assays also resulted in different hypnotic IC50 values than our earlier photomotor response assays in zebrafish larvae,¹⁸  while greatly increasing IC50s for spontaneous activity. Thus, zebrafish inverted photomotor response results may not correlate as well with tadpole results.

We found that allosteric shift models are of limited relevance for analyzing anesthetic interactions at concentrations affecting zebrafish, because direct GABA_A receptor agonism requires high anesthetic concentrations. Studies in GABA-activated receptors and 1% response enhancement ratios using lower anesthetic concentrations were alternatives that also better correlated with our zebrafish results.

Finally, our analysis is framed by the assumption that most of the behavioral effects of the studied drugs are mediated by GABA_A receptors with pharmacology similar to α1β3γ2L. However, nonsynaptic GABA_A receptors and other molecular targets may also contribute to the actions of anesthetics we studied.³⁴ 

In conclusion, our findings in both zebrafish larvae and α1β3γ2L GABA_A receptors correlate modestly, while challenging the general validity of simple two-state allosteric coagonism when more than one anesthetic is present. Our results indicate that cross-talk between allosteric coagonist sites on GABA_A receptors varies, probably influencing drug interactions in animals. They also support the functional and pharmacologic relevance of distinct cryoelectron microscopy structures of GABA_A receptors bound to different anesthetics, despite the nonphysiologic conditions used in these studies.

Douglas E. Raines, M.D. (Department of Anesthesia Critical Care & Pain Medicine, Massachusetts General Hospital, Boston, Massachusetts) provided R-etomidate used in these experiments.

This research was supported by funding from the National Institutes of Health (Bethesda, Maryland) to Dr. Forman (R01GM089745, R01GM128989, and R35GM141951).

The authors declare no competing interests.

Supplemental Figure, http://links.lww.com/ALN/C914