Anesthetic-sensitive ion channel modulation is associated with a molar water solubility cut-off

Background NMDA receptor modulation by hydrocarbons is associated with a molar water solubility cut-off. Low-affinity phenolic modulation of GABAA receptors is also associated with a cut-off, but at much lower molar solubility values. We hypothesized that other anesthetic-sensitive ion channels exhibit distinct cut-off effects associated with hydrocarbon molar water solubility, and that cut-off values are comparatively similar between related receptors than phylogenetically distant ones. Methods Glycine or GABAA receptors or TREK-1, TRESK, Nav1.2, or Nav1.4 channels were expressed separately in frog oocytes. Two electrode voltage clamp techniques were used to study current responses in the presence and absence of hydrocarbon series from eight functional groups with progressively increasing size at saturated aqueous concentrations. Null response (cut-off) was defined by current measurements that were statistically indistinguishable between baseline and hydrocarbon exposure. Results Ion channels exhibited cut-off effects associated with hydrocarbon molar water solubility in the following order of decreasing solubility: Nav1.2 ≈ Nav1.4 ≳ TRESK ≈ TREK-1 > GABAA >> glycine. Previously measured solubility cut-off values for NMDA receptors were intermediate between those for Nav1.4 and TRESK. Conclusions Water solubility cut-off responses were present for all anesthetic-sensitive ion channels; distinct cut-off effects may exist for all cell surface receptors that are sensitive to volatile anesthetics. Suggested is the presence of amphipathic receptor sites normally occupied by water molecules that have dissociation constants inversely related to the cut-off solubility value. Poorly soluble hydrocarbons unable to reach concentrations sufficient to out-compete water for binding site access fail to modulate the receptor.


Background
Protein-ligand interactions are commonly described using a "lock-and-key" model in which the protein and ligand have to fit together to have a chemical effect [1]. Although inhaled anesthetics are presumed to exert their effects through ion channels and cell surface receptor proteins, the diverse nature of both anesthetic ligands and the proteins they modulate challenge assumptions of a conserved complementary structure [2,3]. Current and historical inhaled anesthetics include single atoms, triatomic molecules, various alkanes (with and without halogens), various ethers (with and without halogens), and various alkenes (with and without halogens); experimental inhaled anesthetics demonstrate even greater structural variety. Hence, these drugs act as structurally diverse keys that each can open the same lock. Moreover, a single inhaled anesthetic can allosterically modulate function of a large number of structurally diverse and phylogenetically unrelated ion channels-including many different ligand-gated ion channels, voltage-gated ion channels, and leak channelsand cell surface receptors-including many different channel-linked receptors, enzyme-linked receptors, and G protein-coupled receptors. Hence, these agents also act like a key that can open many different locks.
Inhaled anesthetics typically have high median effective concentrations consistent with low-affinity target interactions [4], and these agents can bind multiple amphipathic sites on a protein of which at least some may be associated with water molecules [5]. Anesthetic modulation of protein function firstly depends upon the presence of an agent in sufficient concentration to desolvate, displace weaklybound water molecules within an allosteric binding site, and only then bind to the allosteric site. We postulate that drugs with insufficient water solubility to competitively displace water from this amphipathic allosteric site would be unable to bind and modulate the protein-a cut-off effect-even at a saturated aqueous phase drug concentration. These low-affinity amphipathic drug-receptor interactions could be associated with different hydrocarbon solubility cut-off values for different proteins since the water dissociation constants within critical allosteric sites could also be different for different proteins. In support, N-methyl-d-aspartate (NMDA) receptor whole-cell currents are unchanged by saturated concentrations of organic compounds once the hydrocarbon ω-end or ring size is increased such that the predicted molar water solubility is less than approximately 1.1 mM; yet these same hydrocarbons still are able to allosterically modulate γ-aminobutyric acid receptor type A (GABA A ) currents [6]. However, GABA A receptors may have their own hydrocarbon cut-off, as GABA A receptor whole-cell currents are unchanged by a series of substituted phenol and benzene rings once the predicted molar water solubility of these compounds is less than 0.1 mM [7].
We hypothesize that cut-off effects associated with molar water solubility is a generalizable feature of inhaled anesthetic-sensitive ion channels and receptors. As a test, we evaluated responses of anesthetic-sensitive channels and receptors to eight different functional groups of organic compounds, differing only by carbon additions to the ω-end of a chain or to a ring, so that we could study compounds of different molecular volumes and carbon atoms but similar molar water solubility values (and vice versa) and therefore distinguish between effects caused by drug size versus solubility. Although inhaled anesthetics bind many different proteins, we studied six channels and receptors reported to contribute to immobilizing effects in vivo: voltage-gated sodium channels type II and type IV (Na v 1.2 and Na v 1.4), [8] TWIK-related spinal cord channel (TRESK) [9], TWIK-related potassium channel type I (TREK-1) [10], GABA A receptors [11], and glycine receptors [12].

Oocyte collection
Adult female Xenopus laevis vivarium-maintained frogs (Xenopus Express, Brooksville, FL) were anesthetized with chilled 0.2% buffered tricaine, surgically ovariectomized, and were either administered morphine analgesia and recovered from anesthesia (first ovariectomy) or euthanized (second ovariectomy) by decapitation and double pithing while anesthetized. The theca externa and mesovarium of the removed ovary were disrupted manually, and oocytes were defolliculated enzymatically by use of 0.2% collagenase. Oocytes were stored in a modified Barth's electrolyte solution until ready for use. This protocol was approved by the Institutional Animal Care and Use Committee at the University of California, Davis.

Receptor expression
Each anesthetic-sensitive receptor type was separately expressed in different oocytes by injecting plasmids containing sequences for receptor subunits which were provided as gifts from the laboratory of R. Adron Harris (University of Texas, Austin). All plasmids were sequenced and their identity confirmed by comparison to National Center for Biotechnology Information databases. Human voltage-gated Type II sodium channels (Na v 1.2) were expressed by intracytoplasmic injection of 5 ng RNA per oocyte containing the SCN2A gene; human voltage gated Type IV sodium channels (Na v 1.4) were similarly expressed using RNA encoding the SCN4A gene. The human TRESK channel (K 2P 18.1) was expressed by intracytoplasmic injection of 5 ng RNA per oocyte containing the KCNK18 gene; the human TREK-1 channel (K 2P 2.1), was similarly expressed using RNA encoding the KCNK2 gene. Heterotrimeric GABA A receptors were expressed by a 1 ng oocyte intranuclear co-injection of three different plasmids containing cytomegalovirus promoters and coding DNA for one of thee subunits-human α 1 (GABRA1), rat β 2 (GABRB2), or rat γ 2s (GABRG2 short)-in a 1:1:10 ratio to ensure incorporation of the γ-subunit; this was confirmed by demonstrating receptor potentiation to 10 μM chlordiazepoxide during co-application with GABA. Glycine receptors were expressed by intracytoplasmic injection of 5 ng RNA per oocyte containing the gene for the human glycine α 1 subunit (GLRA1). Nucleasefree water-injected oocytes served as negative controls for all experiments.

Electrophysiology studies
Oocytes were incubated at 18°C for 1-4 days in fresh and filtered modified Barth's solution composed of 88 mM NaCl, 1 mM KCl, 2.4 mM NaHCO3, 20 mM HEPES, 0.82 mM MgSO 4 , 0.33 mM Ca(NO 3 ) 2 , 0.41 mM CaCl 2 , 5 mM sodium pyruvate, gentamycin, penicillin, streptomycin, and corrected to pH = 7.4. All salts and antibiotics were A.C.S. grade (Fisher Scientific, Pittsburgh, PA). Oocytes were studied in a 250 μL linear-flow chamber perfused with frog Ringer's (FR) solution composed of 115 mM NaCl, 2.5 mM KCl, 1.8 mM CaCl 2 , and 10 mM HEPES prepared in 18.2 MΩ H2O and filtered and adjusted to pH = 7.4. All solutions were delivered to the perfusion chamber via a syringe pump with gastight glass syringes and foil-wrapped Teflon tubing.at a rate of 1.5 ml/min. Oocytes were impaled by two 3 M KCl-filled 0.2-1 MΩ borosilicate glass electrodes (KG-33, King Precision Glass, Claremont, CA) connected to separate headstages (Axon Instruments HS2A, Molecular Devices, San Jose, CA) through which voltage was measured and current was passed by use of a computer-controlled amplifier (Axon Instruments GeneClamp 500B, Molecular Devices, San Jose, CA) [6].
For oocytes expressing Na v channels, the cell transmembrane potential was clamped at -80 mV with a 200 ms step increase to 0 mV applied approximately every minute. The Na v channel response was calculated as the difference between the current required to maintain the holding potential and the largest negative current deflection produced by the voltage step. For oocytes expressing K 2P channels, the transmembrane potential was clamped at −60 mV, and voltage was stepped to + 60 mV each minute. The K 2P channel response was calculated as the difference between holding potential current and voltage step plateau current. For both channel types, baseline responses were measured during oocyte perfusion with FR for a minimum of three times to demonstrate response variability <10%. The perfusate was then switched to the hydrocarbon test drug dissolved at a saturated aqueous phase concentration in FR for 2 min after which the current response to the step voltage was measured. The hydrocarbon drug was then washed out for 5-10 min, and the current response was again measured to demonstrate that it was within 10% of the initial baseline response.
For two-electrode voltage clamp studies in oocytes expressing GABA A receptors or glycine receptors, the transmembrane potential was clamped at −80 mV while being perfused with FR. Approximately every 5 min, the perfusate was switched for 30 s to FR + EC 10 agonist-20 μM 4-aminobutanoic acid for GABA A receptors or 40 μM aminoethanoic acid for glycine receptors-and then the perfusate was switched back to FR for agonist washout. The current response for each ligand-gated ion channel was calculated as the difference between the peak current measured in the presence of agonist and the whole-cell current measured immediately prior to agonist exposure. Baseline responses were measured in triplicate to confirm <10% variability of current responses in the absence of drug exposure. Next, the perfusate was switched to the hydrocarbon test drug dissolved at a saturated aqueous phase concentration for 2 min followed by a 30s exposure the same concentration of hydrocarbon dissolved in FR + agonist. The hydrocarbon drug was then washed out, and oocytes were exposed to FR + agonist 5 and 10 min later to verify return of the current response to within 10% of pre-drug baseline levels.

Drug solution preparation
Using predicted density values published in SciFinder (Chemical Abstracts Service, American Chemical Society, Columbus OH), drug solutions were prepared by anaerobic addition of a hydrocarbon volume to gastight glass syringes containing either FR or FR + agonist to create a concentration equal to its calculated molar water solubility (Table 1). After vigorous agitation, an immiscible hydrocarbon microbubble could be found in the syringe, confirming that a saturated solution had been achieved. However, butane is a gas a room temperature and pressure, and a saturated aqueous phase cannot be achieved because butane vapor pressure exceeds atmospheric pressure. Instead, using FR or FR + agonist containing 1 atm O 2 partial pressure in gastight glass syringes, butane drug solutions were prepared by repeated syringe headspace exchanges to yield solutions with gas partial pressures that were 90% butane and 10% O 2.

Data analysis
Drug responses were calculated as a percent change from the control (baseline) current responses as follows: where I D and I B are the current responses measured during perfusions with drug and without drug (baseline), respectively. Average current responses for each drug and channel were described by mean ± SEM. A positive current change indicated drug-induced positive allosteric modulation of the channel, whereas a negative current change indicated negative allosteric modulation (inhibition) of channel function. A drug-receptor cut-off response was defined as an absolute value change in current <10% from baseline that was statistically indistinguishable from zero using a two-tailed Student t-test. The log 10 of the calculated solubility (log 10 S) for compounds immediately below and above the cut-off for each hydrocarbon functional group were used to determine the receptor cut-off. For each hydrocarbon, there was a "grey area" of indeterminate solubility effect between sequentially increasing hydrocarbon chain lengths. Mean solubility cut-offs were calculated as the average log 10 S for the least soluble compound that modulated receptor function and the most soluble neighboring compound for which no effect was observed. From this result, a 95% confidence interval for log 10 S was calculated for receptor solubility cut-offs.

Results
Sample channel recordings are shown in Fig. 1, and hydrocarbon effects on each anesthetic-sensitive ion channel are summarized in Table 2. With increasing carbon additions to the ω-end or ring, all anesthetic-sensitive channels exhibited a cut-off effect, defined by <10% channel modulation, the smallest effect size that is reliably resolved given the given baseline variability allowed in these studies. All current modulation above this value was statistically different from baseline. The mean hydrocarbon molar water solubility cut-off value and 95% confidence interval range for each ion channel is as follows (Fig. 2 For each ion channel, the calculated molar water solubility was the only physical-chemical property examined that was associated with similar cut-off values between different hydrocarbon functional groups. An example of this is shown in Fig. 3 in which the grey indeterminate bars that separate known regions of positive and absent modulation are clustered around a relatively narrow molar water solubility cut-off range. Other channels, when similarly graphed as a function of hydrocarbon molar water solubility, show qualitatively similar results but with clustering around each channel's own distinct molar water solubility cut-off value. In contrast, GABA A receptor cut-off effects are not associated with the number of molecular carbon atoms or molecular volume of the hydrocarbon (Fig. 4), and these results mirror those of the other anesthetic-sensitive ion channels studied here.

Discussion
As shown previously for NMDA receptors, all of the anesthetic-sensitive ligand-gated, voltage-gated and leak channels examined in this study exhibited cut-off effects for each class of organic compounds, and these cut-offs were associated with the calculated molar water solubility of the hydrocarbon. The cut-offs occurred in a predictable order, with Na v channels K 2P channels, and GABA A receptors cut-offs all clustered within roughly one order magnitude of saturated drug concentrations. In contrast, the glycine receptor cut-off was associated with drug molar water solubility values over two orders of magnitude lower. If previously determined NMDA receptor results are included [6], cut-off responses proceed in order of decreasing hydrocarbon solubility as follows: Na v 1.2 ≈ Na v 1.4 ≳ NMDA ≳ TRESK ≈ TREK-1 > GABA A >> glycine (Fig. 2).
A second explanation for hydrocarbon molar water solubility cut-offs is possible. Desolvation of water molecules around the ligand is necessary for protein binding to occur. However, if the difference in free energy between the solvated protein-drug complex and the separated solvated protein and solvated drug is too great, then drug binding will be energetically unfavorable and no modulation of protein function will occur [19]. Larger hydrophobic drug ligands are surrounded by larger and more rigid water shells for which desolvation may be associated with greater enthalpy. The strength of water complexes around the ligand might increase to the point that drug receptor binding-and thus drug-receptor modulation can only occur if there remains sufficient entropy to overcome increased enthalpy of binding [20].
In either case, a molar water solubility hypothesis is not predicated on any particular molecular size, shape, polarity, functional groups, or atomic arrangement of the drug for low affinity interactions to take place. Indeed, the diversity of conventional anesthetics and other hydrocarbons capable of modulating a single anesthetic-sensitive ion channel, as well as the relatively minor effect differences produced by dug enantiomers [21], suggest that ligand structure itself is not crucial for binding. However, ligand structure does determine the magnitude and type of modulation for the protein it binds. Whether there is inhibition versus potentiation of K 2P channel currents, for example, will depend on the functional group of the hydrocarbon ligand (Table 2). Modulation magnitude also differed between functional groups; inhibition of Na v 1.4 channels was approximately four times greater for alcohols, alkenes, and ethers than for amines or cycloalkanes. The importance of structural elements within receptor binding pockets has also been demonstrated through mutation studies that confer resistance to conventional anesthetic or other hydrocarbon modulation [22][23][24], although it is unknown whether some of these changes may also have altered the water K d or drug desolvation enthalpy that, in turn, could have affected the ability of the drug to bind the allosteric pocket.
The hydrocarbon molar water solubility cut-off value for each channel is expressed as confidence intervals, and certain error is inherent in their measurement. Butane, the smallest n-alkane studied here, is a gas at room temperature and pressure, and therefore cannot be studied at a saturated aqueous phase concentration under normobaric conditions unlike the other liquid and solid hydrocarbons. All receptors were nonetheless modulated by 90% atm of butane, but had current modulation not been observed, a true cut-off could have been inferred at this submaximal concentration. Furthermore, carbon additions to the ω-end of the hydrocarbon chain produce discrete, non-continuous changes in molar water solubility. For each series of functional groups, there is a range of solubility values that lie between the C N modulating hydrocarbon and the C N + 1 cut-off hydrocarbon where the receptor effect is unknown. Most important, however, is the reliance on calculated solubility values for hydrocarbons in pure water at 25°C and pH = 7.0 rather than measured solubility values under study conditions with a 250 mOsm electrolyte solution at 22°C and pH = 7.4. Both accurate measurement and accurate prediction of solubility values are challenging for extremely hydrophobic compounds or for large or complex molecules with multiple functional groups. To limit this problem, only simple aliphatic compounds with the functional group on the first carbon, or central and symmetrical in the case of the dialkyl ethers, were studied. Even so, increasing hydrocarbon chain length is frequently accompanied by greater divergence between calculated and measured water solubility values [25,26].
Whole cell current cut-off responses were measured using hydrocarbons at saturated aqueous concentrations. This was done to ensure that each cut-off was independent of any particular endpoint (e.g., amnesia, unconsciousness, or immobility). Lack of receptor modulation at a saturated hydrocarbon concentration implies absent modulation at a lower pharmacologic concentration, including concentrations relevant to anesthetic endpoints. Anesthetic efficacy is in mammals is unknown for many, but not all, of the hydrocarbons tested. Primary alcohol anesthetic potency increases with increasing carbon chain length from methanol to dodecanol, after which further carbon additions do not produce anesthesia at all [27]. This anesthetic cut-off corresponds to the alcohol molar water solubility cut-off for GABA A receptors (Table 2). However, a general anesthesia cut-off effect has been reported to occur with n-alkanes and dialkyl ethers having around 11-to-15 or more carbon atoms [28] and cycloalkanes having eight or more carbons atoms [29]. These molecules are far longer and have molar water solubility values far lower than occur with the GABA A receptor cut-off. Although observed anesthetic effects might be due to glycine receptor modulation, high affinity effects on one or more other anesthetic-sensitive receptors, or even systemic toxicity, it seems very possible that anesthetic effects could be the result of potent alcohol metabolites produced by oxidation of alkanes, cycloalkanes, and ethers by cytochrome P450 enzymes [30,31]. Since the hydroxyl group confers greater molar water solubility, primary alcohols have receptor cut-offs at longer chain lengths than either alkanes or cycloalkanes or dialkyl ethers. These long-chain alcohol metabolites are also much more potent general anesthetics than their parent compounds [32], so even tiny quantities can have narcotic effects. Identifying simple parallels between in vitro receptor cut-offs and in vivo anesthetic cut-offs thus may be complicated for certain classes of organic compounds.
Channel studies were conducted using a reductionist biological system. However, in vitro electrophysiologic responses conducted at room temperature for relevant anesthetic-sensitive ion channels in oocytes seem to correlate with anesthetic potency in animals [33][34][35]. Likewise, the hydrocarbons studied in Table 1 administered in vivo would be expected to similarly modulate ion channels; in the case of drugs below the solubility cut-off, no in vivo modulation would be expected at all.
Finally, the molar water solubility hypothesis could offer practical applications to the development of new and novel inhaled anesthetic agents. Conventional volatile anesthetics bind promiscuously to a variety of cell proteins, but not all of these receptor interactions are essential to their ability to produce general anesthesia. For example, NMDA receptors contribute to immobilizing actions of inhaled anesthetics able to inhibit their function, but experimental inhaled anesthetics can still be immobilizers without producing NMDA receptor inhibition [36]. Since there is nearly a 10-fold separation in the hydrocarbon molar water solubility cut-off effects between NMDA versus GABA A receptors, a volatile anesthetic might be modified to target calculated aqueous solubility values within this range to confer selectivity against higher cut-off NMDA receptors while preserving activity at lower cut-off GABA A receptors. With sufficient GABA A receptor potentiation and contributions of lower cut-off receptors, such an agent might retain immobilizing potency but lose adverse effects associated with the modulation of higher cut-off receptors. Consequently, molar water solubility could be key to identifying volatile anesthetics with new molecular mechanisms of action and improved pharmacodynamic profiles.

Conclusions
Anesthetic-sensitive ion channels and receptors all appear to exhibit allosteric cut-off effects associated with drug molar water solubility. These results support the Molar Water Solubility Hypothesis mechanism for non-specific protein interactions with drugs of varied sizes and shapes administered at relatively high aqueous phase concentrations, such as occurs with inhaled anesthetic agents. Low-affinity binding at allosteric sites may occur following successful displacement of water by a hydrocarbon molecule (or portion thereof). However, when a hydrocarbon has a molar water solubility below the allosteric cut-off for a channel, as likely determined by the dissociation constant for water at the allosteric site, that hydrocarbon cannot be delivered at a concentration sufficient to outcompete the water for binding site access. In addition to describing how anesthetics of various sizes and shapes may be able to modulate functions of different anesthetic-sensitive ion channels and receptors, the Molar Water Solubility Hypothesis and order of receptor cut-offs offers a unique tool for the development of new inhaled anesthetic agents. By changing substituents on molecules to decrease their molar water solubility, it may be possible to generate novel inhaled anesthetics that are selective against some of the ion channels or receptors normally modulated by conventional volatile anesthetics.
Abbreviations GABA A : γ-aminobutyric acid receptor type A; Na v 1.2: Voltage-gated sodium channels type II; Na v 1.4: Voltage-gated sodium channels type IV; NMDA: N-methyl-d-aspartate; TREK-1: TWIK-related potassium channel type I; TRESK: TWIK-related spinal cord potassium channel Funding This work was funded by the National Institutes of Health National Institutes of General Medical Sciences (Grant GM092821-02).

Availability of data and materials
The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.
Authors' contributions RJB: Conceived the idea, designed the experiments, assisted with some measurements, analyzed the data, and wrote the manuscript. TLP: Performed study measurements and reviewed the manuscript draft. Both authors read and approved the final manuscript.

Ethics approval
This project was conducted under Protocol #19753 which was approved by the Institutional Animal Care and Use Committee of the University of California, Davis.

Consent for publication
Not applicable.

Competing interests
Aspects of this work are included in intellectual property filings and issued patents regarding the identification and use of new inhalation anesthetics and analgesics. Dr. Brosnan is the inventor of record, and patents are owned by the University of California. The University of California has licensed this intellectual property to Expanesthetics, Inc. (Davis, CA), and Dr. Brosnan is Chair of the Expanesthetics Scientific Advisory Board. Neither Dr. Brosnan nor Mr. Pham has received financial compensation from Expanesthetics, Inc., and neither owns equity in Expanesthetics, Inc.

Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.