Crystallographic Studies with Xenon and Nitrous Oxide Provide Evidence for Protein-dependent Processes in the Mechanisms of General Anesthesia

THE mechanisms by which general anesthetics, including the remarkably safe medicinal gases xenon and nitrous oxide, produce their pharmacological action are still poorly known. However, over the past 15 yr, major progress has been made with regard to the molecular mechanisms by which xenon and nitrous oxide act to produce their anesthetic and neuroprotective effects.^1–11  In contrast with most inhalational clinical anesthetics thought to act by potentiating the type A γ-aminobutyric acid receptor, xenon and nitrous oxide have been shown to act mainly by antagonizing the N-methyl-d-aspartate glutamatergic receptor^2,9,12–14  and by modulating other neuronal targets of physiological and clinical interest such as the nicotinic acetylcholine receptor, the TREK-1 two-pore–domain K⁺ channel, and enzymes.^14–18 

Crystallographic studies under gas pressure have allowed a better understanding of the mechanisms by which the chemically and metabolically inert gases act, particularly by demonstrating in protein models of possible neuronal targets that such gases bind to proteins through weak but specific interactions^19–26  within hydrophobic cavities and/or surface pockets. In that way, recent investigations have further shown that xenon and nitrous oxide bind to and compete for the same hydrophobic sites.^18,19 

In the present study, we performed crystallography studies in a series of four proteins—urate oxidase, lysozyme, myoglobin, and neuroglobin—to better understand the basic mechanisms that determine xenon and nitrous oxide binding to proteins. We confirmed that xenon and nitrous oxide bind in most cases to the same binding sites and further demonstrated mainly that the binding of xenon and nitrous oxide to proteins does not depend on hydrophobicity alone but on complex processes in which hydrophobicity and volume interact in different ways.

Urate oxidase (EC = 1.7.3.3; uricase) is a homotetrameric enzyme of 135 kDa that catalyzes the oxidation of uric acid. Purified recombinant urate oxidase from Aspergillus flavus, expressed in Saccharomyces cerevisiae, was provided by Sanofi-Aventis (Montpellier, France). Urate oxidase crystals were grown by the batch or hanging drop techniques at room temperature using a 10–15 mg/ml solution of urate oxidase with an excess of its inhibitor 8-azaxanthine (Sigma-Aldrich, Lyon, France) in 50 mM Tris/HCl pH 8.5 in the presence of 5–8% polyethylene glycol 8000. This led to crystals in orthorhombic space group I222 with one monomer per asymmetric unit.²⁷ 

Hen egg white lysozyme (E.C. 3.2.1.17) is a monomeric enzyme of 15 kDa that hydrolyzes specific linkages in peptidoglycans. Purified lysozyme was purchased from Hampton Research (Aliso Viejo, CA) and dissolved in sodium acetate buffer 0.02 M pH 4.6. Lysozyme crystals were grown by the batch technique using a 15–30 mg/ml solution of lysozyme in 0.1 M sodium acetate buffer, pH 4.6, in the presence of 0.6 to 1.1 M NaCl. This led to crystals in the tetragonal space group P4₃2₁2 with one enzyme per asymmetric unit.²⁸ 

Sperm whale myoglobin is a monomeric globin of 19 kDa involved in oxygen storage by reversibly binding oxygen molecules that have been transported by hemoglobin. Myoglobin crystals were grown by the batch technique by mixing 50 mg/ml myoglobin in 50 mm potassium phosphate buffer, pH 7, and saturated ammonium sulfate. This led to crystals in the monoclinic space group P2₁ with one globin per asymmetric unit.²⁹ 

Murine neuroglobin is a hexacoordinate globin of 18 kDa expressed in the brain of vertebrates, which is involved in neuroprotection during hypoxia and ischemia.^30–32  Neuroglobin production, purification, and crystallization have been described previously.³³  Crystals grew in a 1:1 mixture of neuroglobin (10 mg/ml) and reservoir solution (1.6 M ammonium sulfate, 0.1 M morpholino-ethanesulfonic acid, pH 6.5) in the hanging-drop method. This led to neuroglobin crystals in the trigonal space group R32 with one globin per asymmetric unit.

Data collections were recorded gas-less and under gas pressure of xenon or nitrous oxide of 10, 20, and 30 bar. For each data collection, a crystal mounted in a quartz capillary fitted to a specially designed cell was pressurized and maintained under gas pressure as described earlier.^34,35  Diffraction data were collected at room temperature at the BM14, BM16, and BM30A beamlines at the European Synchrotron Radiation Facility in Grenoble, France. Detectors used were a MAR CCD detector for BM14, an ADSC Q210r CCD detector for BM16, and an ADSC Q315r CCD detector for BM30A. Data were indexed and integrated by DENZO and scaled independently and reduced using SCALEPACK, both programs from the HKL package³⁶  or indexed and integrated by MOSFLM³⁷  and scaled by SCALA. Intensities were converted in structure factor amplitudes using TRUNCATE, and structure refinements were carried out using REFMAC,³⁸  all from the CCP4 package.³⁹  Protein Data Bank entries 2IBA for urate oxidase,¹⁹  1C10 for lysozyme,²³  2JHO for myoglobin,⁴⁰  and 3GKT for neuroglobin²¹  in which heteroatoms and alternate side-chain positions were removed were used as starting model for rigid body refinement. The graphics program COOT⁴¹  was used to visualize |2F_obs− F_calc| and |F_obs− F_calc| electron density maps and for manual rebuilding. All data collections were made at wavelength in the range 0.95–1 Å far from the xenon K-edge (0.3587 Å) or LI-edge (2.2738 Å) restraining the use of xenon anomalous signal to accurately determine occupancies in xenon complexes. However, with a residual of 3 to 4 electrons, the Δf″ signal was enough to assess the presence of xenon from peaks in anomalous difference maps.18 Determination of gas occupancies have been deduced, as in Reference 18, from the height of the peaks corresponding to the gas in the omit maps. The resulting gas B-factors were constant for a given binding site whatever the applied gas pressure. Hydrophobicity of the gas binding site has been estimated by calculating the percentage of carbon atoms in the total number of atoms (carbon, oxygen, nitrogen, and sulfur, ignoring hydrogen) that line the binding site. Cavity and pocket volumes were calculated using the program CASTp⁴²  with a probe radius of 1.4 Å. Summary of the x-ray data collections and refinements statistics for each sample is reported in Supplemental Digital Content 1, http://links.lww.com/ALN/B90. Atomic coordinates and structure factors have been deposited in the Protein Data Bank (4NWH lysozyme under 30 bar of xenon, 4NWE lysozyme under 30 bar of nitrous oxide, 4NXA myoglobin under 30 bar of xenon, 4NXC myoglobin under 30 bar of nitrous oxide, 4O4T for neuroglobin under 30 bar of xenon, and 4O4Z for neuroglobin under 30 bar of nitrous oxide).

Raw data of binding site occupancies by xenon and nitrous oxide were modeled versus pressure using the Origin8 software (OriginLab, Northampton, MA) according to the following logistic equation:

where y₁ is the lower (or initial) experimental value, y₂ the higher (or final) experimental value, x₀ the x-axis value that corresponds to the y-axis value y₀ = y₁ + (y₂ − y₁)/2, and p the power of the sigmoidal fit (fixed to 3). Then, the dose–response occupancies of xenon and nitrous oxide as defined by the predicted values of occupancy obtained from equation (1) were compared by performing a Poisson analysis for distribution.

The relationships between the predicted values of occupancy of xenon and nitrous oxide were modeled versus the gas binding sites’ hydrophobicity and volume, as well as their product and ratio, using equation (1). The level of statistical significance for the Poisson analysis and the r values obtained from equation (1) was set at P less than 0.05.

Xenon and nitrous oxide bound to hydrophobic sites by either intramolecular cavities or solvent-accessible pockets. As an example, figure 1 illustrates the binding of xenon and nitrous oxide to lysozyme (fig. 1A), myoglobin (fig. 1B), and neuroglobin (fig. 1C). As reported previously,^34,43  the presence of xenon or nitrous oxide in the binding sites did not change unit cell parameters and only had little effect on the proteins structure (root mean square differences between native structures and structures with gas were about 0.2 Å for all protein atoms). As already reported,^18,19  the only gas-induced structural effect was a dose (pressure)-dependent expansion of the volume of the gas binding site that did not result, however, in an increase of the number of xenon and nitrous oxide atoms within the gas binding sites.

As reported previously,^18,19  we found that xenon and nitrous oxide bound to urate oxidase within an internal cavity, called UOX-I, located close to the active site of urate oxidase. UOX-I has 90% of lining atoms that are carbons. In the gas-less structure of urate oxidase, UOX-I is empty and has a volume of 116 ų. Xenon and nitrous oxide bound to UOX-I in a pressure-dependent manner of 10 to 30 bar (fig. 2A). Statistical analysis showed a significant difference between xenon and nitrous oxide occupancy of UOX-I (Z = 4.34, P < 0.001), indicating that xenon has a higher affinity than nitrous oxide for UOX-I.

Xenon and nitrous oxide also bound weakly to a small extension of a preexisting pocket, called UOX-II, previously suggested to be not involved in the enzymatic activity of urate oxidase¹⁸  and further shown to be a negligible factor in a mechanistic model of anesthesia based on xenon and nitrous oxide in vivo pharmacology and crystallography data.¹⁹  Therefore, this gas binding site was not taken into account in our analysis.

As described previously,^23,44,45  we found that xenon bound to lysozyme within a small buried internal cavity, called LYSO-I (fig. 1A). LYSO-I is a “full” hydrophobic cavity with 100% of lining atoms being carbons. In the gas-less structure of lysozyme, LYSO-I is empty and has a volume of 32 ų. Xenon bound to LYSO-I in a pressure-dependent manner of 10 to 30 bar; nitrous oxide did not bind to LYSO-I (fig. 2B). This led to a significant difference between xenon and nitrous oxide occupancy (Z = 6.56, P < 0.001).

Also, xenon and nitrous oxide bound to a second pocket called LYSO-II located at a crystallographic interface, a condition indicating that this site is an experimental artifact. Therefore, this gas binding site was not taken into account in our analysis.

In agreement with previous observation,^46–48  we found that xenon bound to myoglobin within four binding sites, called MB-I to MB-IV (fig. 1B). Here, we further showed that xenon also bound to an additional site, called MB-V, and that nitrous oxide bound to the same sites as xenon. MB-III and MB-V are with no doubt internal cavities. Likewise, if the heme is taken as being an intrinsic part of the globin, MB-I, MB-II, and MB-IV can be considered as internal cavities buried between the heme and the globin.

MB-I is located on the proximal side of the heme and has 83% of lining atoms that are carbons. In the gas-less structure of myoglobin, MB-I is empty and has a volume of 66 ų. Xenon and nitrous oxide bound to MB-I in a pressure-dependent manner of 10 to 30 bar (fig. 2C). No significant difference was found between xenon and nitrous oxide occupancy of MB-I (Z = 0.00, P < 1), indicating that both gases have a similar affinity for MB-I.

MB-II is located close to the heme and possesses 73% of lining atoms that are carbons. In the gas-less structure of myoglobin, MB-II is empty and has a volume of 42 ų. Xenon and nitrous oxide bound to MB-II in a pressure-dependent manner of 10 to 30 bar (fig. 2D). Statistical analysis resulted in a significant difference between nitrous oxide and xenon occupancy of MB-II (Z = −7.93, P < 0.001), indicating that nitrous oxide has a higher affinity than xenon for MB-II.

MB-III has 59% of lining atoms that are carbons. In the gas-less structure of myoglobin, MB-III is filled with a water molecule and has a volume of 74 ų. Xenon and nitrous oxide bound to MB-III in a pressure-dependent manner of 10 to 30 bar (fig. 2E). Statistical analysis showed a significant difference between nitrous oxide and xenon occupancy of MB-III (Z = −6.59, P < 0.001), indicating that nitrous oxide has a higher affinity than xenon for MB-III.

MB-IV is located on the distal side of the heme and has 83% of lining atoms that are carbons. In the gas-less structure of myoglobin, MB-IV is empty and has a volume of 40 ų. Both xenon and nitrous oxide bound to MB-IV in a pressure-dependent manner of 10 to 30 bar (fig. 2F). Statistical analysis revealed no significant difference between xenon and nitrous oxide occupancy of MB-IV (Z = 0.51, P < 0.7), a result indicating that xenon and nitrous oxide have a similar affinity for MB-IV.

MB-V has 93% of lining atoms that are carbons. In the gas-less structure of myoglobin, MB-V is empty and has a volume of 35 ų. Xenon bound to MB-V in a pressure-dependent manner of 10 to 30 bar (fig. 2G); nitrous oxide did not bind to MB-V. This led to a significant difference between xenon and nitrous oxide occupancy (Z = 4.47, P < 0.001).

A large internal cavity is located behind the heme and is involved in the heme sliding mechanism in neuroglobin.^33,49  We found that xenon bound to neuroglobin within four binding sites called NGB-I to NGB-IV as reported previously²¹  and to three additional sites called NGB-V to NGB-VII (fig. 1C). Nitrous oxide bound to the same sites as xenon, except NGB-V and NGB-VII.

NGB-I, NGB-V, NGB-VI, and NGB-VII are located in crystallographic interfaces, indicating that these sites are experimental artifacts. Therefore, these sites were not taken into account in our analysis.

NGB-II is a solvent-accessible pocket with 72% of lining atoms being carbons. In the gas-less structure, NGB-II is empty with a volume of 51 ų. Xenon and nitrous oxide bound to NGB-II in a pressure-dependent manner of 10 to 30 bar (fig. 2H). Statistical analysis showed a significant difference between nitrous oxide and xenon occupancy of NGB-II (Z = −3.04, P < 0.005), indicating that nitrous oxide has a higher affinity for NGB-II than xenon.

NGB-III is located at the back of the large internal cavity, close to a tunnel connecting the heme cavity to the exterior of the protein, and thought to be a secondary access for oxygen.^21,50  NGB-III has 86% of lining atoms being carbons. In the gas-less structure, NGB-III is empty and has a volume of 62 ų. Xenon and nitrous oxide bound to NGB-III in a pressure-dependent manner of 10 to 30 bar (fig. 2I). Statistical analysis showed a significant difference between xenon and nitrous oxide occupancy of NGB-III (Z = 7.73, P < 0.001), indicating that xenon has a higher affinity for NGB-III than nitrous oxide.

NGB-IV is also located in the large internal cavity, between the heme and NGB-III, with 89% of lining atoms being carbons. In the gas-less structure, NGB-IV is empty with a volume of 55 ų. Xenon and nitrous oxide bound to NGB-IV in a pressure-dependent manner of 10 to 30 bar (fig. 2J). Statistical analysis showed a significant difference between xenon and nitrous oxide occupancy of NGB-IV (Z = 3.42, P = 0.001), indicating that xenon has a higher affinity for NGB-IV than nitrous oxide.

Taken together, these data confirm and extend previous studies that have first shown in two intramolecular binding cavities that xenon and nitrous oxide bind to and compete for the same binding sites.^18,19 

As shown in figure 3, it is obvious that (1) xenon bound with a higher occupancy than nitrous oxide in the binding sites whose lining atoms include more than 83% of carbon atoms; (2) nitrous oxide bound with a higher occupancy than xenon in the binding sites whose lining atoms include less than 83% of carbon atoms; (3) xenon and nitrous oxide bound with a similar affinity to the binding sites whose lining atoms include 83% of carbon atoms.

Whatever the pressure (10, 20, or 30 bar), occupancy of the gas binding sites by xenon showed a significant logistic relationship with the gas binding site’s volume (10 bar: r = 0.7798, P < 0.01, df = 8; 20 bar: r = 0.784, P < 0.01, df = 8; 30 bar: r = 0.778, P < 0.01, df = 8) but not with the gas binding site’s hydrophobicity (r = 0.000, P = 1) (data not shown). Likewise, occupancy of the gas binding sites by nitrous oxide showed a significant logistic relationship with the gas binding site’s volume (10 bar: r = 0.727, P < 0.02, df = 8; 20 bar: r = 0.778, P < 0.01, df = 8; 30 bar: r = 0.792, P < 0.01, df = 8) but not with the gas binding site’s hydrophobicity (0.591 < r < 0.621, P < 0.1) (data not shown). In addition, because both hydrophobicity and volume play a critical role in gas binding to proteins, we further examined possible interactions between these factors. Occupancy by xenon of the gas binding sites showed a higher significant logistic relationship with the product of hydrophobicity by volume (10 bar: r = 0.910, P < 0.001, df = 8, fig. 4A; 20 bar: r = 0.909, P < 0.001, df = 8, fig. 4B; 30 bar: r = 0.897, P < 0.001, df = 8, fig. 4C) compared to the ratio of hydrophobicity to volume (10 bar: r = 0.612, P < 0.1, df = 8; 20 bar: r = 0.647, P < 0.05, df = 8; 30 bar: r = 0,620, P < 0.1, df = 8; data not shown). By contrast, occupancy of the gas binding sites by nitrous oxide resulted in a significant logistic relationship with the ratio of hydrophobicity to volume (10 bar: r = 0.861, P < 0.005, df = 8, fig. 5A; 20 bar: r = 0.885, P < 0.001, df = 8, fig. 5B; 30 bar: r = 0.898, P < 0.001, df = 8, fig. 5C), but not with the product of hydrophobicity by volume (0.357 < r < 0.497, 0.2 < P < 0.9; data not shown).

Alternatively, we further found that the ratio of xenon to nitrous oxide occupancy of the gas binding sites decreased from 1.47 at 10 bar to 1.21 at 30 bar, and further showed whatever the pressure a significant logistic relationship with the gas binding sites’ hydrophobicity (10 bar: r = 0.960, P < 0.001, df = 5, fig. 6A; 20 bar: r = 0.984, P < 0.001, df = 5, fig. 6B; 30 bar: r = 0.862, P < 0.005, df = 6, fig. 6C).

This research was designed to better understand the mechanisms of gas binding to proteins, particularly the role of the binding sites’ hydrophobicity and volume. By performing crystallography studies with xenon and nitrous oxide, we confirmed that these gases bind in most cases to the same binding sites and further showed that the binding of these gases to proteins does not depend on hydrophobicity alone but on complex processes in which hydrophobicity and volume interact in different ways.

All experiments were performed at 10, 20, and 30 bar of gas pressure, which are pressures at least 10-fold that required to obtain in vivo narcotic/anesthetic effects²  and in vitro inhibitory effects of enzyme activity.^15,17,18,51  The relevance of such pressures to the pressure (concentration) used in clinical anesthesia could therefore be questioned. It is a well-known condition, due to slow gas penetration in crystalline systems, that the gas pressure to be used in crystallography studies must be at least 10 times higher the gas physiological concentration to allow approximating gas binding saturation in a reasonable time period.⁵²  In addition, the possibility that these pressures could have biased our findings should also be discussed since these are pressures under which the phenomenon of pressure reversal of anesthesia may occur. Crystallography studies with helium, an inert gas that virtually possesses no narcotic action and can therefore be considered to mediate the effect of pressure per se,⁵³  have found no effect on the structure of urate oxidase at pressures up to 45 bar.¹⁸  In addition, studies with urate oxidase have further demonstrated that hydrostatic pressures higher than 1,000 bar are required to induce structural changes mainly characterized by compression of the gas binding sites.⁵⁴  Interestingly, such pressures are also approximately 10-fold higher than the physiological pressure at which seizures occurred in mammals⁵⁵  as the ultimate manifestation before death of the high pressure neurological syndrome, a set of excitatory symptoms whose main characteristics oppose those of inert gas narcosis.⁵⁶  Alternatively, one could also question as to whether soluble proteins are good models of membrane proteins, the actual targets of inert gases and volatile anesthetics. For instance, apoferritin has been used as a model for the type A γ-aminobutyric acid receptor.^57,58  Also, there is evidence that membrane proteins such as the Gloeobacter violaceus and Erwinia chrysanthemi pentameric ligand-gated ion channels possess anesthetic binding sites similar to those found in globular proteins that are hydrophobic pockets or cavities of various sizes. In the E. chrysanthemi channel, xenon bound to a hydrophobic cavity within the pore,⁵⁹  and bromoform has multiple hydrophobic binding sites.⁶⁰  In the G. violaceus channel, general anesthetics bound to a large intramolecular hydrophobic pocket that is accessible from the lipid bilayer.⁶¹  Interestingly, these ligand-gated ion channels are homolog to the nicotinic acetylcholine receptor, whose electrophysiological activity is reduced by xenon and nitrous oxide.¹⁴  Taken together, these data indicated that the pressures and models used herein are relevant to the study of the mechanisms of clinical anesthesia.

We found that hydrophobicity of the gas binding sites has a screening effect on xenon and nitrous oxide binding, with a threshold value of 83% beyond which and below which xenon and nitrous oxide binds preferentially compared to each other. Xenon that is more polarizable than nitrous oxide tends to prefer very hydrophobic environment while nitrous oxide, which has a dipolar moment, tends to prefer less hydrophobic environment (table 1). In addition, we found that volume, but not hydrophobicity, was further significantly correlated to xenon and nitrous oxide occupancy of the gas binding sites. This suggests that, despite the threshold effect of hydrophobicity in gas binding, volume but not hydrophobicity plays a major role in the degree of gas binding sites’ occupancy. Also, we found that occupancy of the gas binding sites by xenon and nitrous oxide was significantly better correlated to the product of hydrophobicity by volume and to the ratio of hydrophobicity to volume, respectively. This indicates that hydrophobicity and volume are binding parameters that complement and oppose each other’s effects to determine, respectively, xenon and nitrous oxide occupancy of the gas binding sites.

Alternatively, in line with the well-known Meyer–Overton rule of a high correlation between anesthetic potency and solubility in lipids, we found that the ratio of occupancy of xenon to nitrous oxide was significantly correlated to hydrophobicity of the gas binding sites. In addition, we further found that the mean ratio of occupancy of xenon to nitrous oxide decreased with pressure from 10 to 30 bar, a result consistent with previous studies in rodents that also showed a reduction of the ratio of the in vivo effects of these gases from low gas pressures (concentrations) producing narcosis to higher gas pressures producing anesthesia.^2,19,62–66  Together these data indicate with little doubt that xenon and nitrous oxide would act weakly at multiple proteins rather than strongly at specific targets to induce their narcotic and anesthetic effects. This does not preclude, however, that among the multiple target proteins bound by these gases, some of them could play a preponderant role in some particular effect of these gases, for example, the neuroprotective effects of xenon and nitrous oxide that are thought to result largely from their antagonistic properties at the N-methyl-d-aspartate glutamatergic receptor whose overactivation is well known to be involved in neuronal death.^{2–4,6,7,9,11,13,67–69}

Whether gas binding to proteins within internal cavities, thought to be crucial for conformational flexibility and domain motion,^70,71  or solvent-accessible pockets actually disturbs protein and cell functions is a major question. We recently showed that the ability of xenon and nitrous oxide to bind to urate oxidase disturbs the enzymatic activity of this globular protein.^18,72  Likewise, we further showed that xenon inhibited the catalytic activity of elastase by binding to its active site⁵¹  and demonstrated that xenon and nitrous oxide disrupt in a concentration-dependent manner both the catalytic and thrombolytic activities of tissue-type plasminogen activator,^15,17  a serine protease whose human recombinant form is the only approved therapy for acute ischemic stroke. This agrees with studies that have shown that anesthetic binding to proteins leads to an increase in protein stability^73–75  and to a reduction of kinetics in channel receptors⁷⁶  and provides evidence that gas binding to proteins may actually disturb protein and cell functions.

The Meyer–Overton rule has been often used to infer mechanistic models or hypotheses of general narcosis and anesthesia. Particularly, probably due to the use of olive oil or benzene as models to assess solubility in lipids of general anesthetics, the binding sites of general anesthetics and inert gases are often thought to be highly hydrophobic sites. Also, due to the limited number of exceptions to the Meyer–Overton rule, hydrophobicity is often considered as the preponderant binding parameter involved in the processes of gas binding to proteins. By contrast, here we show that (1) xenon and nitrous oxide bound preferentially to hydrophobic sites of high and mild hydrophobicity, respectively; (2) hydrophobicity, in contrast with volume, did not play a significant role in the degree of occupancy of the gas-binding sites by xenon or nitrous oxide; and (3) hydrophobicity is a good predictor of the ratio of the gas binding site occupancy of xenon to nitrous oxide. Taken together, these data demonstrate that xenon and nitrous oxide obey different binding mechanisms, a finding that in turn argues against all unitary hypotheses of narcosis and anesthesia and further indicates that the Meyer–Overton rule, which is often overinterpreted particularly in the field of undersea physiology and medicine, should only be used in its restricted sense of a high correlation between anesthetic potency and solubility in lipids of general anesthetics.

Also, it is noteworthy from this study that both xenon and nitrous oxide bound to binding sites that were either empty or filled with a water molecule (thereby taking the place of the water molecule). Although all (except one) of these binding sites filled with a water molecule were experimental artefacts located at crystallographic interfaces that were therefore not taken into account for data analysis, the fact that xenon and nitrous oxide bound to these binding sites should be considered with interest in future studies since recent data in membrane proteins have reported anesthetic binding in binding sites containing water.^59–61,77  Particularly, it would be of interest to investigate whether our findings in regard to the respective role of hydrophobicity and volume in gas binding to proteins would also apply to binding sites filled with a water molecule.

In conclusion, the present crystallography study provides evidence that the mechanisms of gas binding to proteins and therefore of general anesthesia may no longer be regarded as the result from the sole ability of the general anesthetics to bind to hydrophobic sites according to their solubility in lipids. Also and importantly, as suggested previously for halogenated compounds^78,79  and recently in mutation studies for xenon,⁸⁰  our findings strongly support that the mechanisms of gas binding to proteins should be considered as the result of an interaction between a drug ligand and a receptor like in classical pharmacology, that is, as the ability of a drug ligand (here, a gas) to bind to a receptor (a hydrophobic pocket or cavity) and the reciprocal ability of this receptor to bind the drug ligand in a fully reversible manner. With no doubt, such complex mechanisms may explain a number of alterations and exceptions to the Meyer–Overton rule of a high correlation between anesthetic potency and solubility in lipids of general anesthetics, such as the virtual lack of narcotic effect of helium and the so-called nonimmobilizers, which are large inhalational anesthetic-like alkanes.^53,81 

The authors gratefully thank Bertrand Castro, Ph.D., D.Sc., and Mohamed El Hajji, Ph.D. (Sanofi-Aventis, Montpellier, France), for their generous gifts of urate oxidase and Hassan Belhari, Ph.D., D.Sc., from the European Molecular Biology Laboratory (Grenoble, France), Gavin Fox, Ph.D., from the European Synchrotron Radiation Facility (Grenoble, France), Jean-Luc Ferrer, Ph.D., D.Sc., and Michel Pirrochi, Ph.D., from the Institut de Biologie Structurale (Grenoble, France), and the staff members of the BM14, BM16, and BM30A beamlines at the European Synchrotron Radiation Facility for their technical help and assistance.

This research was supported by the Centre National de la Recherche Scientifique (Paris, France), the University of Caen Basse Normandie (Caen, France), and the University Paris-Descartes (Paris, France).

Guillaume Marassio was supported by a grant from the Ministère de l’Education et de la Recherche (Paris, France). The other authors declare no competing interests.