Introduction

Top Abstract Introduction Experimental Procedures Results Discussion References

The molecular pharmacology of the inhaled anesthetics remains poorly understood, despite widespread clinical use. The literature is replete with examples of how these small volatile compounds alter the activity of a remarkably wide variety of proteins and in a pharmacologically narrow concentration range of about 10-fold. Thus, identification of important molecular targets has been frustrating and contentious. The paradigm under which many investigators currently work is that few important targets underlie anesthetic action, whereas others are beginning to suspect that many diverse targets are involved. An analysis of volatile anesthetic binding targets may help to differentiate between few and many target paradigms.

Unfortunately, the inhaled anesthetics are unique pharmacologic ligands in that their clinical IC₅₀ values are in the low millimolar range, implying weak binding interactions and rapid binding kinetics. This makes conventional radioligand binding approaches impractical, which in turn has limited progress in developing a model for how these drugs produce their clinically important effects (Eckenhoff and Johansson, 1997).

Photoaffinity labeling converts weak, transient, binding interactions to stable covalent bonds to allow detection of binding targets in complex mixtures of proteins and lipids. We have previously shown that halothane, a clinical inhaled anesthetic, is a reasonable photoaffinity probe capable of providing residue-level resolution of halothane binding location in a protein matrix (Eckenhoff and Shuman, 1993; Eckenhoff, 1996a). Using this approach in rat brain slices, we have also found widespread incorporation of label (Eckenhoff and Eckenhoff, 1998). Furthermore, substantial inhibition of photolabeling by unlabeled halothane strongly suggests that incorporation was to protein, a notion reinforced somewhat by the preferential distribution of label to regions of high synaptic density (Eckenhoff and Eckenhoff, 2000). Although this may represent binding to a few widely distributed proteins, we suspect that the necessary binding motif is sufficiently degenerate that the widespread labeling indicates binding to many proteins. In this study, we test this prediction using quantitative halothane photoaffinity labeling in suspensions of rat cerebellar homogenates.

	Experimental Procedures

Top Abstract Introduction Experimental Procedures Results Discussion References

Photoaffinity Labeling

Animal protocols were approved as required by the Institutional Animal Care and Use Committee and rats were treated in accordance with American Pain Society/National Institutes of Health guidelines. A total of four adult male Sprague-Dawley rats were used in this study. Each animal was deeply anesthetized with halothane, the chest opened, and the brain perfused through the left ventricle of the heart with ice-cold saline. The exsanguinated cerebellum was quickly removed and placed in 2 ml of ice-cold argon-equilibrated incubation solution (120 mM NaCl, 4.75 mM KCl, 1.18 mM MgSO₄, 26 mM NaHCO_3, 1.2 mM KH₂PO₄, 1.77 mM CaCl₂, 5.5 mM glucose, 58.5 mM sucrose, pH 7.5), homogenized with a glass homogenizer, and diluted to 4 ml with the same incubation solution. Quartz cuvettes for photolabeling contained 100 µl of cerebellar homogenate (for a final protein concentration of about 2 mg/ml); 100 µl of 50 mCi/mmol [¹⁴C]halothane (final concentration = 0.2 mM) in argon-equilibrated phosphate-buffered saline (PBS), pH 7.0; and amounts of one of the following unlabeled compounds to achieve concentrations: halothane (0.0, 0.1, 0.5, 1.0, 3.0, 6.0 mM), isoflurane (0.0, 0.2, 0.6, 1.3, 3.8, 7.5 mM), chloroform (0.0, 0.5, 2.0, 4.3, 12.5, 25.0 mM), cyclobutanes F3 (15 mM) and F6 (250 µM), plus incubation solution to bring the final volume to 2 ml. The last addition filled the cuvette to prevent loss of halothane into an air space. The cuvettes were exposed to 60 s of 254-nm UV light (cuvette pathlength ~1 mm) with constant stirring from enclosed microstir bars and at room temperature. After UV exposure, the solution was centrifuged at 14,000g for 15 min, and the supernatant (containing primarily soluble protein) was immediately frozen at 80°C. The pellet was washed once with incubation solution containing 1 mM EDTA and frozen at 80°C.

Gel Electrophoresis

The membrane protein aliquots were thawed and washed again by centrifugation in PBS containing EDTA, and the resulting pellets were resuspended in 100 µl of sample buffer. The soluble protein aliquots were thawed and centrifuged again at 14,000g for 30 min, lyophilized, and suspended in sample buffer (significant contamination of these samples with membranes was ruled out in pilot studies). The membrane and soluble proteins were loaded onto 16 × 20 cm, 1-mm-thick, 12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) (80-100 µg of protein in each lane). Invitrogen (Gaithersburg, MD) and Bio-Rad (Hercules, CA) low molecular weight standards and mass standards were run with each set of samples. Coomassie Blue-stained gels were vacuum dried onto filter paper and placed on X-ray film at 4^oC for 12 days. In addition, samples of membrane and soluble proteins that had not been exposed to UV were also run on SDS-PAGE to determine whether photolysis or proteolysis fragment bands were present in the exposed gels.

Quantitation

All gels and autoradiograms were scanned and quantified using Bio-Rad Quantity One 4.1 software. Twelve protein bands spanning the molecular weight range were randomly selected from the gels for quantitation of both membrane and soluble proteins, and these same bands were used consistently throughout the study. No attempt at identification was made.

Control Lanes. The control lanes were those in which no competing anesthetic was added, thus one control lane was available on each gel; approximately the same concentration of [¹⁴C]halothane (0.2 mM) was used for each set of samples. The peak reflective densities (RD) from the gels and the peak optical densities (OD) from the autoradiograms were measured for each of the selected protein bands. A rolling ball background subtraction was used for each control lane analysis on each gel. The data from all control lanes (n = 30) were grouped together to determine the mean OD/RD, using GraphPad Prism 3.0 software (San Diego, CA). We calculated stoichiometry as the ratio of picomoles of halothane/picomoles of protein for both membrane and soluble proteins. The picomoles of halothane incorporation was determined from standard curves using a protein of known specific activity (¹⁴C-labeled bovine serum albumin). Mass standards were determined from standard curves of RD versus mass for three proteins of different molecular weights (bovine serum albumin, carbonic anhydrase, and myoglobin). These standards were included on each gel.

Competitor Lanes. Using volume analyses instead of peak, the RD from the gels and OD from the autoradiograms were measured for the same selected protein bands, in both the control lanes and the remaining five lanes, in which labeling had been conducted in the presence of increasing concentrations of the competing anesthetics (halothane, isoflurane, and chloroform). The OD/RD (label incorporation/protein mass) was calculated for each band in each lane, using GraphPad Prism 3.0 software. These values were normalized by dividing them by the control lane OD/RD volumes. Nonlinear regression was used to fit sigmoid curves of negative slope for each anesthetic for each band; control lanes were taken as 0.01 mM halothane. Any negative bottom values were set to zero.

Molecular Weights

The molecular weights were determined for each of the membrane and soluble protein bands analyzed. Molecular weight standards were run with each gel and from these a standard curve was constructed to determine the approximate molecular weights of the bands chosen for quantitation. Approximately 10 values for each band for both the soluble and membrane protein were averaged.

Materials

Halothane was obtained from Halocarbon Laboratories (River Edge, NJ) and isoflurane from Abbott Labs (Chicago, IL). F3 and F6 cyclobutanes were obtained from PCR Chemical, Inc. (Gainesville, FL). The [¹⁴C]halothane was synthesized by PerkinElmer Life Sciences (Boston, MA). Quartz cuvettes were purchased from Markson LabSales, Inc. (Wayne, NJ), and the 254-nm UV low pressure Hg(Ar) pencil calibration lamp was obtained from Thermo Oriel (Stratford, CT). Prism software used was version 3.0 from GraphPad, and Quantity One software version 4.1 was from Bio-Rad. Hyperfilm-MP was obtained from Amersham Biosciences (Arlington Heights, IL). Gibco low molecular weight standard was obtained from Invitrogen, and Bio-Rad low molecular weight standard was purchased from Bio-Rad. Chemicals for incubation solution, PBS, sample buffer, Coomassie Blue, SDS-PAGE, bovine serum albumin, carbonic anhydrase, myoglobin, chloroform, and EDTA were all obtained from either Sigma (St. Louis, MO) or Fisher (Pittsburgh, PA).

	Results

Top Abstract Introduction Experimental Procedures Results Discussion References

Many cerebellar membrane and soluble proteins incorporate label in the presence of [¹⁴C]halothane and 254-nm UV light (Fig. 1). Of the 12 membrane protein bands examined, calculated stoichiometry of label incorporation varied by over 2 orders of magnitude at this concentration (0.2 mM) of halothane (Table 1). Soluble protein stoichiometry varied less, by only 1 order of magnitude, but still reaching a similar maximum (~3:1-4:1) as observed for the membrane proteins. There was a significant correlation between labeling stoichiometry and molecular weight in both soluble and membrane protein (Fig. 2, C and D) and also between label incorporation and protein mass in the soluble protein (Fig. 2B). No apparent correlation between protein mass (RD) and labeling (OD) existed for the membrane protein (Fig. 2A). We verified that the more strongly labeled protein bands are not simply photolysis or proteolysis fragments, as gels run with unexposed cerebellar proteins were essentially the same. Also, we verified that our soluble protein samples were not significantly contaminated by small membrane fragments by comparing with samples prepared with more extensive centrifugation.

View larger version (137K): [in this window] [in a new window]   Fig. 1.   An example of a Coomassie Blue-stained membrane protein SDS-PAGE (A) and autoradiogram (B) and a similar soluble protein SDS-PAGE (C) and autoradiogram (D). In these examples, the concentration of unlabeled halothane was increased from 0.0 to 6.0 mM, whereas the concentration of [14C]halothane remained constant (0.2 mM). Other competitors, not shown in this figure, were chloroform and isoflurane. The 12 protein bands selected for quantitation are shown on the left side of the figure.

View larger version (15K): [in this window] [in a new window]   Fig. 2.   Plotted are the reflective density and optical density from the control lane for each membrane (A) and soluble (B) band. Also plotted are the stoichiometry and molecular weight for each membrane (C) and soluble (D) band. Each point represents a mean of 30 individual bands. Shown is the regression line and the R2 value for each plot.

There was significant inhibition of label incorporation in most bands by the two anesthetics, indicating competitive interactions within this concentration range, a reflection of saturable halothane binding. Table 1 shows the parameters for sigmoidal fits to the competitor data. IC₅₀ values varied 10-fold, from 0.2 to 2.0 mM, with all of the soluble bands being below 1 mM. Hill slopes averaged about 1 overall, with greater variation in the membrane bands. Table 2 shows the actual maximal displacement achieved with the highest concentration of competitor. Because isoflurane or F3 inhibited label incorporation by only an average of 30 to 50% at maximal concentrations, reliable calculation of competition parameters could not be determined and were not included in Table 1. Maximal (saturating) concentrations of the nonimmobilizer, F6, produced the least inhibition of halothane labeling in either membrane or soluble bands. To graphically illustrate differences between inhibitors, all 12 bands were grouped for each anesthetic within either membrane or soluble protein (Fig. 3). Both halothane and chloroform inhibit [¹⁴C]halothane incorporation in a competitive fashion, but halothane was more effective. The difference between the halothane and chloroform IC₅₀ values for membrane protein averaged about 1.3 mM, whereas that for soluble proteins averaged about 0.3 mM. It is interesting to note that the difference between the clinical IC₅₀ values for halothane and chloroform is intermediate, about 1.0. Figure 4 illustrates differences between membrane and soluble proteins for each anesthetic inhibitor. Inhibition of label incorporation by the three anesthetics was most effective in soluble protein, and the difference was most apparent for chloroform.

View larger version (21K): [in this window] [in a new window]   Fig. 3.   The ability of the different compounds to inhibit halothane photolabeling in the 12 membrane (A) and soluble (B) protein bands. Points represent the mean of all bands (120 individual points from three experiments) for each inhibitor. S.E.M. values are included but are smaller than the size of the symbols. The curve represents a least-squares fit to sigmoid curves of variable slope. Halothane was the most effective inhibitor, and isoflurane was the least effective in both membrane and soluble protein. Chloroform was intermediate in both cases but closer to halothane in the soluble protein bands.

View larger version (17K): [in this window] [in a new window]   Fig. 4.   Similar to Fig. 3, except grouped according to inhibitor, so that differences in membrane and soluble protein bands are more apparent. Shown only are the regression lines for clarity. Halothane inhibits photolabeling similarly in both membrane and soluble proteins, although chloroform inhibits photolabeling of soluble more effectively than membrane. Isoflurane causes less inhibition than either halothane or chloroform but like chloroform is more effective in the soluble protein bands.

	Discussion

Top Abstract Introduction Experimental Procedures Results Discussion References

Multiple Binding Targets. These results confirm our prediction that halothane binds to multiple proteins in the mammalian brain and not to a few widely distributed proteins. This is consistent with the relative lack of provocative molecular features on these ligands. The small surface area and weak electrostatic features (small permanent dipole and a few polarizable halogens) imply low stringency in binding sites; thus multiple and low affinity targets are predicted. Taking this a step further, low site stringency suggests that stoichiometry for individual targets could be relatively high, consistent with these and other results (Eckenhoff, 1996a,b; Bhattacharya et al., 2000; Xu et al., 2000). Furthermore, there should be a relationship between stoichiometry and protein size, which was also found in this study (Fig. 2, C and D). These many binding sites for halothane and other inhaled anesthetics are likely to be internal protein cavities, because a cavity can involve more of the ligand surface area in van der Waal interactions than a surface pocket or patch and therefore have higher binding energy. It is relevant to note, then, that a considerable number (~15%) of proteins with high resolution structures available have internal cavities of sufficient size to accommodate halothane (volume ~130 Å) (Fig. 5). The fact that more than 15% of brain proteins appear to be labeled in this study likely relates to the fact that the structures deposited in the Protein Data Bank (PDB) are smaller, more easily crystallized, and therefore more likely to have smaller cavities than the neuronal proteins examined here. But cavity volume alone does not explain why isoflurane, an anesthetic with only slightly larger volume than halothane, was so much less effective at inhibiting halothane labeling than halothane or chloroform. This suggests that some structural or electrostatic features in the cavities underlie this modest selectivity.

View larger version (17K): [in this window] [in a new window]   Fig. 5.   We used GRASP (Nicholls et al., 1991) to determine the largest internal cavity size in the PDBselect (Hobohm et al., 1992; Hobohm and Sander, 1994) data set (high stringency). This graph represents the relative frequency of cavity size, ordered in bins of 20 Å3, in this 550-protein data set. The approximate volume of halothane is shown, demonstrating that 13% of PDBselect entries have cavities large enough to accommodate halothane.

Identification of Proteins. Testing of our hypothesis did not require identification of the selected proteins, other than determining their molecular weights for analytical purposes. Nevertheless, identification of selected bands may yield interesting and novel targets for future study. Unfortunately, bands in one-dimensional gels may include more than one protein or even unrelated proteolytic fragments. This possibility renders our stoichiometry calculations tentative if only a fraction of proteins are labeled in a given band. Furthermore, it is probable that some currently attractive anesthetic targets, such as the -aminobutyric acid type A receptors, were entirely missed in our analysis, primarily because of low molar representation. However, our question was the number and distribution of binding targets and not whether we could identify binding in a preselected subset. Even if considered more relevant or plausible based on in vitro functional studies, it is unlikely that such targets will show enhanced binding over those selected for quantitation in this study, since we calculated that many had stoichiometry values of greater than 1 at 0.2 mM halothane concentration. It is the distribution of binding character across the conformational ensemble that controls the functional influence of a ligand; thus a different set of methods and reagents will be required to dissect this in selected protein systems.

Membrane Versus Soluble Protein. We did not expect that soluble proteins would have a higher apparent affinity for halothane than membrane proteins, because many investigators currently believe that most important effects of anesthetics are mediated by direct interactions with membrane protein (Franks and Lieb, 1994). However, given that the relationship between binding affinity and the effect at the cellular, or certainly organism level, is not a priori predictable, lower affinity does not reduce the potential importance of membrane protein actions. In addition, the lack of correlation between mass (RD) and stoichiometry (OD) in the membrane-bound proteins (Fig. 2A) suggests that these targets are more selective than soluble proteins. A potential reason why membrane proteins bind halothane with lower affinity relates to the lipid milieu, which may form a portion of the anesthetic binding site. This may result in lipid competition for binding on both an equilibrium and a photochemical basis. Furthermore, since solute/protein affinity is ultimately defined as the energetic difference between solvation of a solute in a protein site versus that in the solvent, the fact that inhaled anesthetics are more soluble in lipid than in water dictates that affinity for a protein site immersed in lipid will be lower than that of an otherwise identical site immersed in water. Of course, the final possibility is that the character of protein sites immersed in lipid is simply less favorable for anesthetic binding than those of soluble proteins. This may relate to the unusually hydrophobic character of transmembrane segments and the well known requirement for a degree of polarity in anesthetics (Eckenhoff and Johansson, 1997). The fact that F6 inhibits halothane labeling more in the membrane proteins is consistent with this view.

Binding Selectivity. Chloroform, a haloalkane much like halothane, inhibits halothane labeling by about 80% in soluble proteins and about 50% in membrane proteins. This indicates that the binding sites for halothane and chloroform are the same, or at least overlapping, which is consistent with the fact that these two anesthetics have quite similar clinical characteristics (Cousins and Seaton, 1995). Interestingly, the separation of IC₅₀ values for inhibition of halothane labeling between halothane and chloroform was much smaller in the soluble proteins than in membrane proteins (Fig. 3), but both were in the direction of the clinical potency difference (chloroform is less potent than halothane). In contrast, isoflurane, a haloether, inhibits label incorporation significantly less than the two haloalkanes and in a direction not consistent with clinical potency differences (isoflurane is more potent than chloroform). Therefore, in agreement with previous work (Eckenhoff and Eckenhoff, 1998; Greenblatt and Meng, 2001), these results strongly suggest that binding sites (not necessarily binding targets) for these two different classes of anesthetic are different.

The Cyclobutanes. F3 and F6 were included because they have been touted as a test of relevance for molecular targets (Kendig et al., 1994; North and Cafiso, 1997). Thus, F3, an anesthetic, should not bind to and change the function of the same molecular target that F6, a nonanesthetic, does, if that target is to be considered important functionally. The enormous water solubility difference between F3 and F6 opens the possibility of energetic explanations for lack of an F6 effect (as opposed to steric). Nevertheless, this "criterion" appears to be satisfied in this study since F6 had a considerably smaller effect on halothane labeling than F3 or any of the other anesthetic compounds.

Photolabeling Methodology. Short UV (254 nm) light is required for photolysis and has the capacity to produce cross-linking and alterations in side chain character (Johansson and Eckenhoff, 2001), which may limit the ability to cleanly separate some proteins, and was occasionally manifest in our gels as SDS-resistant aggregation. Also, the halothane photolysis product, a carbon-centered radical, has a half-life (low microseconds) that could permit diffusion to sites of preferential labeling. However, we have found excellent agreement of photolabeling residue-level localization with crystallographic (Bhattacharya et al., 2000; Tang et al., 2000) and spectroscopic (Johansson et al., 1995; Johansson et al., 2000) data within a protein matrix. Furthermore, halothane labeling reliably distinguishes specific from nonspecific binding (Eckenhoff and Tanner, 1998), so we are confident that the technique has the resolution and specificity to select out preferential binding targets from a complex mixture. Finally, photolabeling is a nonequilibrium binding method, and we have used it here to estimate some equilibrium binding parameters for the different protein bands. Again, the estimates of affinity derived from these studies agree with those obtained by equilibrium methods in purified protein models like serum albumin (Johansson et al., 1995; Eckenhoff, 1996a), sometimes being lower by a factor of 2 to 3. Thus, our photolabeling methodology should be reliably reporting both stoichiometry and affinity values for these samples of brain protein.

Multiple Functional Targets? It should be here emphasized that multiple binding targets do not necessarily indicate multiple functional targets. But it is important to note that binding alone will alter protein energetics, which implies that activity might be altered in many proteins. For example, any functional protein has an ensemble of conformations in dynamic equilibrium, and the relative population of each is dictated by the free energy differences between them. Changes in activity are brought about by stabilizing and thereby populating one conformation at the expense of others. Any binding event that favors a given conformer must influence the conformational equilibrium and therefore the activity or function. Since the distribution of internal cavities or other binding sites for small anesthetic molecules is unlikely to be precisely preserved across the entire conformational ensemble, anesthetic binding may favor a given conformer and therefore alter activity. It is not clear how much change in the conformational equilibrium is required for a biologically significant change in function for many proteins (Eckenhoff and Johansson, 1999), but it is suspected that in some receptors, like the G protein-coupled receptors, it is quite small (Ross, 1996). It is possible that many of the site-directed mutants of ion channels and receptors alter protein activity by modulation of protein stability, as opposed to any specific steric changes to a ligand binding site. In summary, the clear presence of multiple binding targets probably reflects multiple sites of action, albeit with widely varying contributions to the final in vivo effect. A multiple target model for anesthetic action is consistent with genetic approaches to date (Campbell and Nash, 1994; Morgan and Sedensky, 1994), the lack of chemical antagonists, the lack of inherent or acquired tolerance, and the remarkably conserved potency across the entire animal kingdom.