The degree to which different volatile anesthetics depress carotid body hypoxic response relates to their ability to activate TASK potassium channels. Most commonly, volatile anesthetic pairs act additively at their molecular targets. We examined whether this applied to carotid body TASK channels.
We studied halothane and isoflurane effects on hypoxia-evoked rise in intracellular calcium (Ca2+i, using the indicator Indo-1) in isolated neonatal rat glomus cells, and TASK single-channel activity (patch clamping) in native glomus cells and HEK293 cell line cells transiently expressing TASK-1.
Halothane (5%) depressed glomus cell Ca2+i hypoxic response (mean ± SD, 94 ± 4% depression; P < 0.001 vs. control). Isoflurane (5%) had a less pronounced effect (53 ± 10% depression; P < 0.001 vs. halothane). A mix of 3% isoflurane/1.5% halothane depressed cell Ca2+i response (51 ± 17% depression) to a lesser degree than 1.5% halothane alone (79 ± 15%; P = 0.001), but similar to 3% isoflurane alone (44 ± 22%; P = 0.224), indicating subadditivity. Halothane and isoflurane increased glomus cell TASK-1/TASK-3 activity, but mixes had a lesser effect than that seen with halothane alone: 4% halothane/4% isoflurane yielded channel open probabilities 127 ± 55% above control, versus 226 ± 12% for 4% halothane alone (P = 0.009). Finally, in HEK293 cell line cells, progressively adding isoflurane (1.5 to 5%) to halothane (2.5%) reduced TASK-1 channel activity from 120 ± 38% above control, to 88 ± 48% (P = 0.034).
In all three experimental models, the effects of isoflurane and halothane combinations were quantitatively consistent with the modeling of weak and strong agonists competing at a common receptor on the TASK channel.
Volatile anesthetic–induced activation of constitutively active TASK potassium channels in the carotid bodies blunts ventilatory response to hypoxia
Whether volatile anesthetics differ in their efficacy of suppressing hypoxic ventilatory response is incompletely understood
While additive effects after coadministration of pairs of volatile anesthetics have been reported, the possibility that the less efficacious of the two agents might antagonize the more efficacious one has not been previously demonstrated
In vitro, both halothane and isoflurane depressed hypoxia-evoked rise in intracellular calcium in rat carotid body cells, and activated TASK potassium channels
At equivalent concentrations, halothane exhibited stronger effects on hypoxic responses and TASK channel activation than isoflurane, but coapplications of these two drugs resulted in lesser effects than halothane alone
These observations suggest that isoflurane and halothane act as competitive agonists on TASK channels where the weaker agonist (isoflurane) antagonizes the effects of the stronger agonist (halothane)
VOLATILE anesthetics are widely used during surgery to maintain unconsciousness, but among undesirable side effects is the depression of breathing, especially postoperatively, contributing to increased morbidity and mortality, particularly in patients with pre-existing cardiorespiratory disease.1 This is exacerbated by simultaneous use of opioids and residual neuromuscular blockade.2 In the United Kingdom alone, there are approximately 3 million general anesthetic procedures annually.3 With an incidence of postoperative hypoxemia as high as 30%,4 even small risk reductions could yield large benefits.
Under normal conditions, hypoxemia evokes hyperventilation. This protective hypoxic ventilatory response is suppressed by volatile anesthetics even at the very low levels that can persist long after surgery.5 Understanding this mechanism could facilitate the development of agents without respiratory side effects and/or other drugs that can selectively antagonize these depressive effects.6
Hypoxic ventilatory response is mediated by the carotid bodies. Under normoxic conditions, the resting membrane potential of their type I (glomus) cells is maintained negative by constitutively active potassium channels (TWIK-related acid-sensitive potassium, TASK, channels; where TWIK is “tandem pore domains in a weak inwardly-rectifying potassium” channel). Hypoxia decreases TASK channel activity causing membrane depolarisation, activating voltage gated calcium (Ca2+) entry, and neurosecretion. These neurotransmitters excite afferent (glossopharyngeal) nerve terminals, which project to the ventilatory control centers in the brainstem and induce hyperventilation.4,7–10
In rat and mouse glomus cells, the predominant channel present is a heterodimer of TASK-1 and TASK-3 (termed TASK 1/3) together with some homodimeric TASK-1 and TASK-3, all of which are inhibited by hypoxia.8–10 In contrast, TASK channels are typically activated by volatile general anesthetics.10–12 These opposing effects on TASK channels could explain the depressive effect of anesthetics on hypoxic response. Furthermore, the observation that TASK channels appear differently sensitive to volatile agents (i.e., activated more strongly by some agents than others)11–13 could underlie established differences between volatile anesthetics on hypoxic response depression at the organismal level.5,14–17 Our initial aim in this study was to test the simple hypothesis that anesthetic agents (we selected halothane and isoflurane) differ in their efficacy of supressing hypoxic response of the cell and TASK channel activity.
A second aim was to test the hypothesis that the less efficacious of the two agents might antagonize the more efficacious in a manner predicted by classical drug-receptor theory. This would be consistent with the current paradigm that anesthetics act on discrete target receptors.18 However, for pairs of volatile inhaled anesthetics, only additive effects have hitherto been observed using immobility as the outcome,19–21 and molecular target studies (albeit not investigating TASK channels) produce similar findings.22,23 We therefore employed mixtures of these agents in different concentrations to test this second hypothesis at three hierarchies of biologic organisation: (a) Ca2+ responses to hypoxia, (b) activation of endogenous TASK 1/3 channels, and (c) activation of cloned TASK-1 channels.
Materials and Methods
Single Glomus Cell Isolation
Carotid bifurcations were surgically removed from isoflurane anesthetized rat pups at day 10 to 15. Animals were then killed by decapitation or exsanguination while still deeply anesthetized, and ethical and regulatory approvals were obtained in accordance with project and personal licenses (United Kingdom Animals [Scientific Procedures] Act, 1986). The bifurcations were placed in ice-cold phosphate-buffered saline, and carotid bodies were subdissected and placed in 37°C F12-medium containing trypsin and collagenase. Enzymatic digestion was followed by trituration, leading to a single cell suspension in fetal bovine serum and insulin supplemented with F12 Ham’s or F12 Ham’s/Dulbecco’s Modified Eagle’s Medium (Fisher Scientific/Gibco, United Kingdom), which was plated out on poly-l-lysine–coated glass coverslips in a culture dish. After adding further culture media (2 h), cells were stained with rhodamine-peanut-agglutinin (10 µg/ml, Vector Labs, United Kingdom; incubated for at least another hour); a bright orange fluorescence when excited at 540 nm aids cell identification. Cells were used from 2 to 8 h after plating.
Transient Expression of TASK-1 in HEK293 Cell Line
Cells from the HEK293 cell line were grown in Dulbecco’s Modified Eagle’s Medium high glucose plus sodium pyruvate (110 mg/l) supplemented with 10% fetal calf serum, 100 units/ml penicillin/streptomycin, and 6 mM L-glutamine (all from Fisher Scientific/Gibco). Rat TASK-1 was inserted in the internal ribosome entry site–enhanced green fluorescent protein plasmid (pIRES-EGFP, Clontech, Takara Bio, France). This was transiently transfected using lipofectamine and PLUS reagent (Life Technologies, United Kingdom): following the manufacturer’s recommendations, 24 to 48 h after transfection, cells were replated onto poly-l-lysine–coated glass coverslips and used for recordings. Transfected cells expressing TASK-1 were identified by their green fluorescence.
Measurements of [Ca2+]i by Spectrofluorimetry
Rat glomus cells were loaded with Indo-1-acetyloxy methyl ester (AM) in culture media at room temperature and incubated for 1 hr (2.5 μM added from a dimethyl sulfoxide stock; Invitrogen, Fisher Scientific/Gibco, United Kingdom). Indo-1 was excited at 340 nm light; [Ca2+]i was calculated from the ratio of Indo-1 fluorescence intensities at 405 and 495 nm, corrected for minimum (zero Ca2+) and maximum (2.5 mM Ca2+) fluorescence ratios measured in each cell.
Anesthetics were introduced into the gas mixtures used to equilibrate Tyrode solutions using calibrated vaporizers with a range of up to 5%. Tyrode solutions contained, in mM, 117 sodium chloride, 4.5 potassium chloride, 1 magnesium chloride, 23 sodium bicarbonate, 11 glucose, and 2.5 calcium chloride, and were equilibrated with 95% air/5% CO2 (control) or 1% O2/5% CO2/94% N2 (hypoxia). At the start of each recording, peanut agglutinin–labeled glomus cells (see Single Glomus Cell Isolation) were identified and exposed to a brief (e.g., 30 s) hypoxic stimulus to confirm a robust [Ca2+]i response. Next the cell was exposed to halothane and/or isoflurane for approximately 1 min in normoxia, followed by a second exposure to hypoxia (30 s) in the presence of halothane and/or isoflurane. This was followed by a recovery period in normoxia without anesthetic of approximately 1 min. At the very end of a recording, a control of approximately 30 s exposure to hypoxia was performed without anesthetic to ensure appropriate recovery in the magnitude of the original hypoxic response. Only cells that demonstrated this recovery were included in our analysis.
For each hypoxic exposure, the magnitude of the Ca2+ transient was calculated as the absolute difference between the mean [Ca2+]i in the period before hypoxic exposure and the mean [Ca2+]i during hypoxia. The hypoxic response in the presence of anesthetic was then expressed as a percentage of the response to hypoxia in the absence of anesthetic.
Cell-attached patch clamp recordings on native glomus cells and HEK293 cells were performed using an Axopatch 200B (Molecular Devices LLC, USA). Single-channel recordings were filtered at 2 to 5 kHz, and current was digitized and recorded at 20 kHz. Voltage clamp control, data acquisition, and analysis were performed using Spike2 (Cambridge Electronic Design, United Kingdom).
Recordings were made with borosilicate pipettes (Harvard Apparatus Ltd, United Kingdom), which were sylgard-coated and fire-polished before use. The pipette filling solution contained, in mM, 140 potassium chloride, 1 magnesium chloride, 1 EGTA, 10 HEPES, 10 tetra-ethyl-ammonium, and 5 4-aminopyridine (included to inhibit other, non-TASK channels). pH was adjusted to 7.4 at 37°C. Cells were placed in a recording chamber and initially superfused with a standard Tyrode solution equilibrated with 5% CO2/95% air in which seal formation was attempted. Upon obtaining a gigohm seal, the superfusate was changed to a Tyrode containing 100 mM potassium chloride (with an equimolar reduction in sodium chloride and no calcium chloride) but otherwise identical to the standard Tyrode. Pipette potential was then set to +80 mV. Under the recording conditions used in this study, the glomus cell background potassium channel conducts inward currents exhibiting short, flickering openings interspersed with clear sections of baseline. The protocol was (all in normoxic solutions and in high-potassium Tyrode), (1) no anesthetic, (2) anesthetic for approximately 30 to 90 s, and then (3) no anesthetic.
Analysis of channel activity was performed on 20-s sections of recording obtained at least 10 s after any solution exchange, to allow full exchange of solutions within the recording chamber. For each condition in each cell/patch, we constructed an all-points histogram to determine the modal single channel current. This modal current was then used to analyze channel open probability (i.e., the product of the number of channels in the patch and the probability of a channel being open) using a 75% threshold crossing method (with 175% representing double openings). For each condition/anesthetic, measurements of channel open probability were also expressed as percentage increase relative to channel open probability during the control period. If a stimulus resulted in unchanged channel activity, the increase in channel activity compared with control would be 0%.
Monitoring of Anesthetic Delivery
For all experiments, the gas phase above the bubbled solutions was constantly monitored by an infrared analyzer (Capnomac Ultima, Datex, United Kingdom). This, in turn, was calibrated against a quadropole mass spectrometer (Airspec MGA 3000, United Kingdom) using calibration gases. Moreover, the relationship between vapor levels/setting on the vaporizer and anesthetic levels in solution was linear.27 Mixtures of agents were delivered by arranging the vaporizers in series (with the isoflurane vaporizer sited first in the pathway of gas flow). We verified using infrared analysis that, with one vaporizer empty, the output of the other was not influenced by turning the dial of the empty vaporizer (i.e., to exclude any potential back-pressure effect). We also confirmed, using mass spectrometry, that when using two vaporizers in series (one with halothane and the other with isoflurane), the anesthetic concentrations in the headspace gas were as dialed on the vaporizer (i.e., there was no significant loss of agent from the gas phase due to arranging the vaporizers in series). Finally, we confirmed using gas chromatography–mass spectrometry that the concentrations of each anesthetic in solution was identical when using a mixture, to the concentration when each agent was bubbled separately (i.e., there was no selective loss of one agent over another within the Tyrode solution). The concentrations of anesthetic in solution sampled in both the bottles containing the Tyrode and in the perfusion chamber were not significantly different (i.e., there was no measurable anesthetic loss from the perfusion system for either agent). Separately we have confirmed that our gas delivery system and perfusion system is mostly gas-tight: nitrogen equilibrated solutions have a PO2 in the perfusion chamber of just approximately 2 to 3 mmHg (i.e., there is minimal ingress of oxygen). A detailed analysis of the performance of this system, especially the gas chromatography–mass spectrometry data in relation to mixtures of agents, has been published elsewhere.27
For sample size, we were guided by the approach of Johnstone et al.,28 who modeled methods for ion channel studies based on a minimum of three separate concentrations (plus the zero point) with a minimum of three repeats (in independent cells from different animals) at each concentration, to construct a suitable dose-response curve. The exact concentration we used at any time was guided by the results of the effects of a previous concentration (where we had no previous knowledge of likely drug effect). For studies of mixtures of agents, we examined at least three concentrations, with studies conducted in at least five cells (each from different animals) at each concentration level, in paired studies (i.e., each agent plus mix in the same cell preparation).29 This exceeded the minimum sample size per data point indicated by the “resource equation.”30 Based on the a priori estimate of the degrees of freedom for the error term in an ANOVA, and the number of groups under study (three in our study; one for each agent, plus the mix), the formula for minimum sample size is (10/3) + 1. On average, 1 to 2 data points are achievable in a single day’s experimentation, for either Ca2+ imaging or single-channel recording, reflecting the workload involved.
Where concentration-response relationships were plotted, lines of best fit were constructed using nonlinear least squares regression, with models fit empirically to a simple ligand binding equation,
where Bmax is the estimated maximum effect of drug, [X] is drug concentration (%), and Kd is the dissociation constant (%). For drug mixtures, especially where one drug appeared to behave as a partial agonist, we predicted the result of drug admixtures according to the method proposed by Howard and Webster,29
where A and B are the respective drugs in their concentrations with their respective maximum effect alone (Bmax), and KA and KB are their respective dissociation constants.
Data are presented as mean ± SD. Anesthetic concentrations are expressed as percentage delivered; these can be converted to millimolar concentrations in aqueous solution using the data of Franks and Lieb,18 where 1% halothane equates to 0.24 mM and 1% isoflurane to 0.28 mM (for presentation of results in terms of millimoles, see Supplemental Digital Content 1, http://links.lww.com/ALN/C467). We conducted statistical analysis using factorial analysis of variance (SPSS for Windows version 25, SPSS Inc., USA), where the “response” was, respectively for each of the experiments described, (a) the [Ca2+]i transient, (b) the normalized channel open probability for native TASK channels in glomus cells, and (c) normalized channel open probability of TASK-1 channels in HEK293 cells. The comparisons of interest were first, between halothane and isoflurane (to assess if these agents differed in effect) and second, between halothane and the mix of agents (to assess if the mix produced an infra-additive response). This was therefore a more stringent approach than comparing the results of mix with the results of a predicted additive effect between the two agents. There were two factors: “agent,” a nominal factor (one level for each anesthetic/mix), and “dose” (a scalar factor, one level of each concentration). The locations of any significant effects were explored using a post hoc Bonferroni test, and P < 0.05 (two-tailed) was considered significant.
Effect of Single Anesthetic Agents on Hypoxia-induced [Ca2+]i Transients in Glomus Cells
Baseline [Ca2+]i was unaffected by either anesthetic at any concentration used in this study ([Ca2+]i was 70 ± 15 nM control, 55 ± 18 nM with halothane [1 to 5%], 56 ± 13 nM with isoflurane [0.5 to 5%]; ANOVA for “agent,” “dose,” and interactive term being P = 0.517, 0.796, and 0.595, respectively). In contrast, the [Ca2+]i response to hypoxia (the “Ca2+ transient”) was depressed by both agents, with halothane virtually abolishing it at concentrations of approximately 3% and higher (fig. 1, A and B). Isoflurane also depressed glomus cell hypoxic response in a concentration-dependent manner, but this was more modest such that approximately 50% of the response remained, even at supraclinical isoflurane concentrations of 5% (fig. 1, A and B). Quantitatively, halothane 5% depressed glomus cell Ca2+i hypoxic response by 94 ± 4% (P < 0.001 vs. control). Isoflurane 5% had a less pronounced effect: 53 ± 10% depression (P < 0.001 vs. halothane).
Statistically, the effect of “dose” was significant, confirming that drug dose had a depressive effect (P < 0.001, F = 9.77, df = 7). The difference between the agents was also significant (P < 0.001, F = 114.2, df = 1 for factor “drug” in the ANOVA model). The interactive term “dose×agent” was not significant (P = 0.200, F = 1.50, df = 5), indicating that the effect of dose did not differ across the agents. Post hoc testing for comparison of specific data points was not necessary for further confirmation of the conclusion that both agents were depressive, with halothane more so than isoflurane.
When these data were plotted as a function of “anesthetic effect” from 0 to 100% (fig. 1B, with 100% being complete abolition of a response), the effects of both anesthetics appeared to show saturability but with differing maximal efficacies. Thus isoflurane appeared to have a lower intrinsic activity than halothane. We applied Equation 1, which assumes a ligand binding to a single receptor, to estimate a Kd for halothane of 0.65 ± 0.07% and Bmax 109 ± 7%; and for isoflurane a Kd of 0.80 ± 0.18% and Bmax of 60 ± 6%.
Effect of Anesthetic Combinations on Hypoxia-induced [Ca2+]i Transients in Glomus Cells
We selected a fixed background level of halothane of 1.5% so we could detect additive or infra-additive effects of adding increasing concentrations of isoflurane (1%, 1.5%, and 3%). Figure 2 shows that the degree of inhibition of the hypoxic [Ca2+]i response caused by 1.5% halothane was progressively diminished—not enhanced—with the addition of increasing amounts of isoflurane to the mix. Thus a mix of 3% isoflurane/1.5% halothane depressed cell Ca2+i response (51 ± 17% depression) to a lesser degree than 1.5% halothane alone (79 ± 15%; P = 0.001), but similar to 3% isoflurane alone (44 ± 22%; P = 0.224), suggesting antagonism by isoflurane toward halothane.
Statistically, comparison between the halothane results and mix results was significant (P = 0.001, F = 14.49, df = 1, for “agent,” ANOVA), as was comparison for effect of progressive concentration of isoflurane (P = 0.002, F = 6.09, df = 3 for “dose,” ANOVA). Post hoc testing for comparison of specific data points was not necessary for further confirmation of the conclusion that isoflurane was infra-additive to halothane in the mix.
We next used the individual estimates for Kd and Bmax (fig. 1) and applied Equation 2, which assumes competition of a partial with a full agonist at a single receptor. Figure 2 (green line) shows the results of this modeling,which predicts that, over the range of anesthetic concentrations tested, adding isoflurane will progressively antagonize the effects of halothane. This model prediction (Equation 2) is in reasonable agreement with the data—albeit to a slightly greater degree than the model (see Discussion).
Electrophysiology: Anesthetic Effect on Native TASK Channel Activity
We next investigated the effects of halothane and isoflurane on the native TASK channels in glomus cells, which mediate the cellular effects of both hypoxia and volatile anesthetics on [Ca2+]i. Figure 3A shows cell attached patch recordings of TASK activity in rat glomus cells, which typically display short flickery openings, increased in frequency by both halothane and isoflurane. Across the concentration ranges examined, halothane and isoflurane again differed in their efficacy, with halothane being the more efficacious (fig. 3B). The response to halothane did not show saturation, whereas for isoflurane it did. Statistically (ANOVA), there was a significant effect of “dose” (P < 0.001, F = 0.452, df = 7; indicating for both agents there was a dose-related effect) and “agent” (P < 0.001, F = 16.43, df = 1; reflecting the difference between agents), but not of the interactive term “dose×agent” (P = 0.331, F = 1.16, df = 4; indicating that dose did not differentially influence the agent effect). Post hoc testing for comparison of specific data points was not necessary for further confirmation of the conclusions that the agents activated the native channel, and that they did so to different degrees.
Modeling to Equation 1 thus resulted in a very high estimated Bmax for halothane of 830 ± 356% and a Kd of 10.8 ± 6.3%. For isoflurane, the respective values were 156 ± 33% and 1.5 ± 1.2%, respectively, for Bmax and Kd.
Electrophysiology: Effect of Anesthetic Combinations on Native TASK Channel Activity
We next examined the effects of combining isoflurane and halothane on TASK channels in glomus cells, applied in equal amounts (1.5, 2.5, and 4% of each). Again there was a concentration-dependent increase in channel activity with the progressive total concentration in mixes of these anesthetic agents, but the response to “halothane plus isoflurane” produced a far lesser increase in channel open probability across the dose range, than that of “halothane alone” (fig. 3B, green symbols).
For the mix condition, post hoc testing revealed (a) that the effects of “isoflurane alone” and of “isoflurane plus halothane” were significantly less than the effects of “halothane alone” (P < 0.001, respectively, with Bonferroni correction for multiple testing), and (b) that the effects of “isoflurane plus halothane” and of “isoflurane alone” were similar (P = 0.602). In quantitative terms, 4% halothane/4% isoflurane yielded channel open probabilities 127 ± 55% above control, vs. 226 ± 12% for 4% halothane alone (P = 0.009). Moreover, we employed Equation 2 using the estimated values of Kd and Bmax for the individual agents (fig. 3). The resulting line for predicted effect, assuming competition by a partial agonist at a single receptor, lay very close to the actual data points for the mix (fig.3, green line).
We analyzed the all-points histograms for the channel activities displayed in figure 3A. The predominant form of TASK channel present in the glomus cell is a heterodimer of TASK-1 and TASK-3, but there is also evidence of smaller numbers/subpopulations of TASK-1 and TASK-3 homodimeric channels.8,9 To complicate single-channel analysis, channel density is high in these cells, so most patches contain multiple channels, and these channels often display more than one conductance state. Consequently, all-points histograms derived from these recordings often show a single broad peak at approximately 2.6 pA under the conditions used in this study (fig. 4A), reflecting the predominant activity of the heterodimer. There was an increase in channel openings by isoflurane and halothane and of the mix at all conductance levels. In one recording (fig. 4B), however, we also captured another type of channel, which was probably a homodimeric TASK-1 channel, judging by its peak channel current of approximately 1.2 pA.8,9 This channel seemed to be little affected by isoflurane alone but was activated by halothane. Moreover, this activation by halothane was antagonized by isoflurane, suggesting competition between different volatile anesthetic agents. We therefore investigated the effects of halothane and isoflurane on homodimeric TASK-1 channels expressed in HEK293 cells.
Electrophysiology: Effect of Single Anesthetic Agents on TASK-1 Channels Expressed in HEK293 Cells
We next investigated the effects of isoflurane and halothane on homodimeric TASK-1 channels heterologously expressed in HEK293 cells. These channels exhibited similar short openings with an average single channel current of 1.2 pA. Halothane clearly increased TASK-1 channel activity, isoflurane less so (fig. 5A). This was confirmed in quantitative analysis over a wide range of concentrations, with the response to halothane saturating at a maximum increase in channel activity (channel open probability) of 122 ± 38% (fig. 5B). Isoflurane increased channel activity much less, saturating at a maximum increase of just 51 ± 27% (fig. 5B). Statistically, the effect of “dose” was significant (P < 0.001, F = 5.22, df = 6, ANOVA), indicating dose-related effects. The comparison between agents was also significant (P < 0.001, F = 27.8, df = 1, ANOVA), indicating agents differed. The interactive term “dose×agent” was not significant (P = 0.425, F = 0.94, df = 3), indicating dose effects did not differ across agents. Post hoc testing for comparison of specific data points was not necessary for further confirmation of the conclusion that halothane and isoflurane both activated the channel yet differed in their efficacies.
The shape of these dose-response curves in transiently expressed single channels was similar to those obtained in whole-cell Ca2+ transients (compare fig. 5 with fig. 1). It was notable that the increase in channel activities of expressed TASK-1 (~50% to ~120% for isoflurane and halothane, respectively) was less than observed in native TASK 1/3 (c.100% to c.220% for the same concentration range; fig. 3). Modeling to Equation 1 yielded estimates as follows: for halothane, Kd 1.15 ± 0.45% and Bmax 160 ± 36%; and for isoflurane, Kd 1.2 ± 0.54% and Bmax 66 ± 20%.
Electrophysiology: Effect of Anesthetic Combinations on TASK-1 Channels Expressed in HEK293 Cells
We next investigated the effects of combining isoflurane with halothane in expressed TASK-1 channels. We applied a slightly different approach to that used in the previous experiment on native TASK channels, in that we selected a fixed concentration of halothane (2.5%) that caused significant but not maximal TASK-1 activation (fig. 5B). Against this background halothane concentration, we tested the effects of adding isoflurane in varying concentrations (1.5%, 2.5%, 4%, and 5%; fig. 6). The effect of isoflurane under these conditions is not as striking as for the glomus cell or native channel (compare with fig. 3B), but the addition of isoflurane with halothane clearly reduced, rather than increased, the level of channel activity compared to that seen with halothane alone. The overall reduction of activity (channel open probability) from 122 ± 38% (with 2.5% halothane only) to 88 ± 48% (with 2.5% halothane plus 5% isoflurane) was statistically significant (P = 0.034, F = 4.91, df = 1, ANOVA by “agent”).
Moreover, we employed Equation 2 using the estimated values of Bmax and Kd for the individual agents at the expressed channel (fig. 5). The resulting line for predicted effect lay very close to the actual data points for the mix (fig. 6, green line), consistent with the model assumption of competition by a partial agonist at this single receptor.
Competitive Behavior between Anesthetics
The major finding of this study is that two volatile anesthetic agents—both agonists—exhibit competition by the weaker antagonizing the effects of the stronger at both whole-cell (Ca2+ signaling) and molecular level (TASK channels). This conclusion was strengthened by a quantitative approach in fitting data to classic binding isotherms to derive apparent Bmax and Kd values, then using these to calculate predicted effects of weak and strong agonist combinations in a competitive receptor-mediated effect model. These measured pharmacologic effects potentially link to the important clinical effect of volatile anesthetics suppressing hypoxic response. However, the clinical relevance of these results will need whole-animal studies.
This result of infra-additivity is notable when viewed in the context of the extensive review of anesthetic interactions by Hendrickx et al.21 They reported that for pairs of volatile inhaled agents, only additive effects are found, using immobility as the outcome.19,20 They also found experimentally that many candidate molecular targets produced similar additive findings (but TASK channels were not examined).21–23 Additive or synergistic anesthetic interactions, rather than subadditive, are also observed when intravenous agents are included in other animal/molecular studies.31–33 As these authors discussed, the ability of studies to distinguish whether two drugs compete for occupancy of a common receptor site34 is dependent on the occupancy of those receptors at the endpoint studied. If the typical behavioral endpoints used to characterize general anesthetics are associated with low site occupancy, then subadditivity may not be detectable.
A considerable body of work indicates anesthetics bind to specific sites on a range of channels, which implies that competition between agents is to be expected (although this has not been directly reported for isoflurane/halothane combinations). Andres-Enguix et al. reported that in truncated TASK-1 channels a single point mutation (lysine 159 to arginine) caused reduction of anesthetic sensitivity to both agents, indicating a common binding site.13 Luethy et al. found that mutations close to a proposed binding site rendered TASK-3 resistant to activation by both trichloroethanol and isoflurane.35 There is also evidence for site-specific anesthetic pharmacology in γ-aminobutyric acid type A (GABAA) receptors, restricted to intravenous agents, supported among other observations by the finding of a novel drug (naphthalene–etomidate) with little or no efficacy with regard to GABAA activation, but which antagonizes the actions of etomidate, pentobarbital, and propofol.32 So on the one hand, our result of subadditivity is consistent with receptor theory; on the other hand, it differs from the general findings of additivity reported in many other channels for anesthetic combinations.
The generally additive interactions between anesthetics has historically been interpreted, together with the correlation of anesthetic potency with lipid solubility (the Overton–Meyer rule), as implying a nonreceptor, physico-chemical mechanism of action (e.g., the lipid solubility theory).36 Shafer et al. modeled additivity and synergy of agent interactions within a framework of anesthetic binding to single (or multiple) receptors, but they did not address the possibility of an anesthetic behaving as a partial agonist at a receptor, in the manner we now describe for isoflurane at the TASK channel.23
Possible Molecular Mechanisms for Competitive Antagonism between Anesthetic Agents
We are unaware of any existing theory, other than classic ligand binding models to receptors, that would be consistent with our data. A nonreceptor theory akin to lipid solubility theory might be developed post hoc, given our new data, but we cannot imagine what that might be. The attempts to reconcile the classical lipid hypothesis with receptor theory should be noted. Several proposed binding sites are hydrophobic (or accessible to drugs only via previous dissolution in lipid).37 This could lead to sufficient disruption of cell membrane “rafts” by anesthetic as to influence channel function, as Petersen et al. noted for the TREK (TWIK-related potassium channel).38 Pavel et al. have recently reported that modulation of raft-related phospholipase D2 can confer anesthetic sensitivity on a previously insensitive potassium channel (TRAAK; TWIK-related arachidonic acid activated potassium channel): a putative receptor could therefore lie within the associated membrane raft site, and not necessarily on the protein moiety of the channel, at least for some channels.39
The most parsimonious interpretation is therefore that our results strongly support general anesthetics acting through specific target sites to exert their effects. Our quantitative modeling is not absolute proof that interactions must reflect competition at a single channel, but rather that our observations are consistent with this proposition.
Brief discussion is needed as to what is meant by a “receptor” in this context. Although a channel might be conceived as a whole (including its related lipid elements39 ) as a single receptor, it may consist therein of more than one binding site, particularly so where the functional channel is made of multiple subunits (two in the case of TASK). The simplest explanation of our data is to assume only one type of binding site that may or may not be present in multiple identical copies per channel. At these putative sites, either halothane or isoflurane can bind but have differing efficacies with respect to increasing channel open probability. Thus, when present together, these agents compete to bind to this single type of site, and the lower-efficacy agent acts as a competitive inhibitor to the higher-efficacy agent. A single-site model is probably best suited to studies conducted in expressed TASK-1 channels (fig. 6). In native heterodimeric TASK 1/3 channels (fig. 3B), there could be two different sites, given two different subunits per channel. With regard to Ca2+ signaling at the cell level (fig. 2), anesthetics may have additional effects on voltage-gated Ca2+ channels and other channels, or on other mechanisms putatively involved in oxygen sensing.40,41 From such considerations we would therefore predict that the best fit of a single-site model (Equation 2) would be seen for data obtained with pure TASK-1, and this is what we observed (compare the proximity of mix data results to the predicted line for fig. 6,vs.figs. 3B or 2).
Potential Limitations of the Study
Our results on the effects of single agents on TASK channels agree with Patel et al.,42 Putzke et al.,12 and Sirois et al.,43 who collectively reported that both halothane and isoflurane activate TASK-1 (in various cell preparations). However, Berg et al. found almost no effect of isoflurane on rat TASK-1 channels in HEK293 cells (although this would also predict possible antagonism between isoflurane and any other stronger activator of this channel).11
A limited the number of data point intervals at delivered concentrations less than 1% limited the precision of our Kd estimates. Conversely, we could not deliver concentrations greater than 5%, which limited the precision of Bmax estimates; the constraint was the upper and lower limits and calibration intervals on the vaporizers. Our experimental setup is designed primarily to study the effects of hypoxia on isolated chemoreceptor cells in vitro, requiring tight control of oxygen, and continuous bubbling of gas mixtures and sustained perfusion of cells achieve the gas equilibria in an “open” system.27 A “closed” delivery system using directly dissolved agents might have permitted a wider concentration range to be assessed, but not allowed rapid switches into and out of hypoxia. That said, our Kd and Bmax estimates for individual agents in Equation 2 resulted in predictions that agreed reasonably well with the data across all experiments.
While we have observed competitive interactions between halothane and isoflurane in both the carotid body TASK 1/3 and the TASK-1 channel, we do not yet know what role the TASK-3 subunit plays. It is entirely feasible that this may have a slightly different anesthetic binding site to that of TASK-1.
We do not have a complete molecular explanation for the interactions we describe, and detailed molecular modeling as used by other groups will be required.13,32–35 A hierarchy of organization also absent from this work is the study of intact animal hypoxic response. Although the literature on animal hypoxic responses and anesthesia has been reviewed elsewhere,44 the effect of halothane/isoflurane combinations in both wide-type and TASK knockout animals would be important. However, our observations detecting subadditive effects of drug combinations are mostly at high (supraclinical) anesthetic concentrations. The inferred competitive interactions may be harder to detect under clinical or subanesthetic conditions due to presumed lower receptor occupancy.21,23
Implications of Our Results
Our discovery was prompted by reports of apparent selectivity between volatile anesthetics in both their depression of hypoxic response5,14–17 and activation of TASK channels.11–13,18 It will be interesting to see if other agents behave like isoflurane, as partial agonists, at TASK channels and so in turn are able to compete with other more efficacious anesthetics.
Exploiting competitive behavior at TASK channels could be important in managing postoperative hypoxia.1,2 Some TASK channel inhibitors offer promising results in cell/channel preparations45,46 as new “respiratory stimulants.” The potential of an antagonist to anesthetic binding could offer an alternative approach: anesthetic-induced depression of hypoxic ventilatory response could be reversed in a targeted manner, with fewer side effects (e.g., as compared with more centrally acting respiratory stimulants). A similar approach to the reversal of general anesthesia has been previously suggested for GABAA receptor antagonists: the notion of reversing general anesthesia “on demand.”31–33 It is proposed by others that two-pore potassium channels are involved in mediating the hypnotic/narcotic effects of anesthesia47,48 ; it would be further interesting if putative antagonists to anesthetic at the TASK channel also restored consciousness.
The authors thank Daniel Lunn, Associate Professor, Department of Statistics, University of Oxford, Oxford, United Kingdom, and Devinder Sivia, Lecturer in Mathematics for the Sciences, St. John’s College, Oxford, United Kingdom, for their statistical advice with this article.
This work was funded by grants from the Medical Research Council, British Journal of Anesthesia, National Institute of Academic Anesthesia, Difficult Airway Society (all located in London, United Kingdom), and a personal Higher Education Funding Council (London, United Kingdom) New Blood Clinical Senior Lectureship award to Dr. Pandit.
The authors declare no competing interests.