Chem. Res. Toxicol. 1998, 11, 703-707
703
A Simple Alternate Substrate Test Can Help Determine the Aqueous or Bilayer Location of Binding Sites for Hydrophobic Ligands/Substrates on Membrane Proteins Ralf Morgenstern* Institute of Environmental Medicine, Division of Biochemical Toxicology, Karolinska Institutet, Box 210, S-17177 Stockholm, Sweden Received January 19, 1998
It is of general interest to determine the location of binding sites in membrane protein receptors and enzymes. A simple method is described that relies on the use of alternate substrates with defined hydrophobicity and decipherable trends in relative kcat/Km values. The rates at first-order conditions (low substrate concentrations) for at least two substrates are determined at a defined lipid/detergent volume fraction. Ratios of the rates (which at firstorder conditions are linearly related to substrate concentration) are compared to the predicted partitioning of the substrates. Since the relative rates depend also on the kcat/Km ratios, the contribution of this parameter and its limits is discussed. When alternate substrates that fulfill reasonable criteria regarding kcat/Km values are used, the location of binding sites can be tentatively predicted. Examples are given describing evaluation of the location of the xenobiotic binding site in detoxication enzymes (microsomal glutathione transferase and cytochrome P450). Furthermore, it is argued that unspecific binding sites for hydrophobic molecules, such as those that are important in many detoxication enzymes, can only benefit from hydrophobic interactions facing the aqueous compartment of the cell. The membrane location of many drug-metabolizing enzymes thus implies that an aqueous active site located close to the membrane is advantageous, an advantage that could be realized if a concentration gradient of hydrophobic molecules extends into the aqueous phase at lipid interfaces.
Introduction Many proteins interact with hydrophobic substances that partition to varying degrees into lipid bilayers. When the functional analysis of such proteins involves the addition of lipid or detergent, as is most often obligatory in the case of membrane proteins, the question arises whether the binding site faces the aqueous or lipid phase (or the interface). A simple variation of the detergent/lipid concentration and evaluation of binding or kinetic parameters does not answer this question since concentrations in the aqueous and lipid compartments vary in parallel (Figure 1). Given the lack of structural information on membrane proteins, simple methods that can determine the location of the active/binding site are needed for a correct evaluation of binding and kinetic parameters. Sophisticated multiphasic modeling has provided a framework within which the question of facedness and properties of acceptor sites can be evaluated (1). The method described here is to be regarded as a simpler diagnostic approach implicit in the earlier work. The diagnostic procedure, since it is aimed primarily at determining the location of binding sites, involves the measurement of enzymic rates under first-order conditions and is thus greatly simplified. Furthermore, the diagnostic approach is directly applicable to binding data and, therefore, of general utility.
Experimental Procedures Materials. 4-Chloro-3-nitrobenzamide (CNBAM)1 and 4-chloro-3-nitrobenzanilide (CNBAN) were from Alfred Bader Library
of Rare Chemicals, Division of Aldrich Chemical Co. (Milwaukee, WI). Glutathione and Triton X-100 were purchased from Sigma Chemical Co. (St. Louis, MO). All other chemicals were standard commercial products of highest purity. Methods. Enzyme Purification. Microsomal glutathione transferase was prepared from male Sprague-Dawley rat liver (180-200 g), as previously described (2). Enzyme Assay. Measurement of enzyme activity with CNBAM and CNBAN was performed as described (3). In short, 100 µL consisted of 5 mM glutathione, 0.1 and 0.05 mM CNBAM/CNBAN (dissolved in ethanol yielding a final concentration of 2.5%), 0.1 M potassium phosphate (pH 6.5), and 0.1% or 1.1% Triton X-100. Enzyme was added so that the final concentration was below 4 µM. Protein Determination. Protein concentration was measured by the method of Peterson with bovine serum albumin as the standard (4). Calculations. It is assumed that a partition coefficient governs the distribution of the hydrophobic substrate between two compartments (aqueous and detergent in the present case). These compartments are in rapid equilibrium so that enzymic rates are not limited. These are reasonable assumptions as discussed in ref 5 provided that the substrate/detergent ratio is below 0.1. Interfacial sites are not treated explicitly in this work, and for this the reader is referred to other analytical approaches (1). The relative concentration of substrate in the aqueous and detergent phases is given by:
Cw/Ct ) 1/(1 - R + PR)
(1)
1 Abbreviations: CNBAM, 4-chloro-3-nitrobenzamide; CNBAN, 4-chloro-3-nitrobenzanilide.
S0893-228x(98)00013-7 CCC: $15.00 © 1998 American Chemical Society Published on Web 05/15/1998
704 Chem. Res. Toxicol., Vol. 11, No. 6, 1998
Morgenstern
Figure 1. Calculated relative concentrations in the aqueous and hydrophobic phases at varying volume fraction detergent (R) and selected partition coefficients (P).
Cd/Ct ) 1/((1 - R)/P + R)
(2)
where Ct ) total concentration (substrate or ligand), Cw ) concentration in the aqueous phase, Cd ) concentration in the detergent phase, P ) partition coefficient of the substrate (octanol/water), and R ) volume fraction of detergent. Calculations and figures were made employing Microsoft Excel. Incremental binding energies were calculated by the equation: ∆∆G ) -RT ln(kcat/Km)(S/R)/(kcat/Km)(S/H) as described in ref 6.
Results and Discussion It is obvious from Figure 1 that variation of the detergent amount in an assay causes the same relative substrate concentration changes in the aqueous and hydrophobic (detergent) compartments, regardless of substrate hydrophobicity. Thus it is not possible to use these changes as tools to determine true kinetic parameters or binding constants unless assumptions are made about the location of the protein site being studied. Measurement of enzyme initial rate at first-order conditions is directly proportional to substrate concentration [S] as given by: v ) kcat/Km[E][S]. If two substrates display equal kcat/Km values, the ratio of enzymic rates at first-order conditions will exactly reflect the relative substrate concentrations. If these substrates display different hydrophobicities, one can exactly predict the ratio of their rates at any detergent concentration. Furthermore these ratios are completely different if the substrate approaches from the aqueous or hydrophobic phase. For example, two hypothetical substrates A and B with equal kcat/Km values are added at equal concentration (,Km) to an assay medium containing 0.001 volume fraction (R) detergent. A and B display different partition coefficients (Poctanol/water) of 1000 and 10. In the case of an aqueous active site, the ratios of the activities A/B
would be 0.5, whereas the corresponding ratio for a site exposed to the hydrophobic phase would be 50 (exactly reflecting their partition behavior, see Figure 1). Clearly if one knew the true kcat/Km values of different substrates, it would be very simple to determine the true position of an active site and vice versa. Since the position of the active site can only be unambiguously determined from the protein structure, of which information is scarce for most membrane proteins, alternative approaches must be taken. One such alternative approach described here involves studying the rate behavior of several substrate analogues at first-order conditions with the aim to find substrates that display different partition coefficients but similar kcat/Km values. Microsomal glutathione transferase conjugates numerous substrates to glutathione thereby serving in detoxication (7). All substrates are hydrophobic molecules that contain reactive electrophilic centers. Taking advantage of the broad substrate specificity, precise manipulations can be made to substrates producing analogues with varying hydrophobicities and reactivities. For instance, microsomal glutathione transferase catalyzes the conjugation of reactive activated nitroarenes such as CNBAM (Scheme 1a) and the isoreactive but more hydrophobic CNBAN (Scheme 1b) (3). These compounds form part of a series of incrementally reactive analogues modified in the (1-) position whose log kcat/Km values are linearly dependent on Hammett σ- values (3). Thus reactivity is a prominent feature in the first-order reaction of these compounds with the enzyme and not minor structural changes at the (1-) position para to the reacting 4-carbon and the chlorine leaving group. Therefore CNBAM/CNBAN were investigated as potential tools to determine the location of the hydrophobic binding site in microsomal glutathione transferase.
Binding Site Orientation
Chem. Res. Toxicol., Vol. 11, No. 6, 1998 705
Figure 2. Calculated relative concentrations in the aqueous and hydrophobic phases at varying volume fraction detergent (R) of CNBAM and CNBAN. Table 1. Activity of Microsomal Glutathione Transferase toward CNBAM and CNBAN at Different Detergent Concentrations (Assay Conditions Given under Methods) specific activity (v) (µmol/min mg) detergent concentration substrate
Triton X-100 (0.1%)
Triton X-100 (1.1%)
v1.1%/v0.1% obsd (expta)
CNBAM (50 µM) CNBAM (100 µM) CNBAN (50 µM) CNBAN (100 µM) vCNBAN/vCNBAM (mean value)b
0.13 ( 0.02 (n ) 5) 0.27 ( 0.05 (n ) 5) 0.037 ( 0.003 (n ) 4) 0.074 ( 0.004 (n ) 4) 0.28
0.089 ( 0.018 (n ) 5) 0.18 ( 0.02 (n ) 5) 0.0085 ( 0.0006 (n ) 4) 0.018 ( 0.002 (n ) 4) 0.098
0.67 (0.86) 0.66 (0.86) 0.23 (0.15) 0.24 (0.15)
a The expected ratio was calculated by eqs 1 and 2 (see also Figure 2). b Values at 50/100 µM were 0.28/0.27 and 0.096/0.1 at 0.1% and 1.1% detergent, respectively.
The approximate log P values for these compounds were calculated by the program ACDLogP, and P was inserted into eqs 1 and 2 giving the estimated relative concentrations in aqueous and detergent compartments shown in Figure 2.2 The corresponding predicted firstorder rate ratios are 0.46 and 0.08 (vmore hydrophobic/vless hydrophobic) assuming an aqueous approach and 32 and 5.8 if the substrate approaches from the hydrophobic phase at 0.001 and 0.011 volume fraction of detergent, respectively. The measured ratios of CNBAN/CNBAM activity were 0.28 and 0.098 (Table 1). These values are in good agreement to those expected for an aqueous-faced active site in microsomal glutathione transferase. However there is no experimental verification of the assumption of equal kcat/Km values, and neither is one possible. One must therefore also examine the hypothetical ratio of kcat/ Km values that is required should the substrate binding site reside in the hydrophobic phase. In this case kcat/ Km has to be 70-fold lower for the more hydrophobic substrate. Clearly, if anything, the opposite is expected for substrates that can contribute an incremental hydro2 Calculation of log P values by the commercial program ACDLogP was performed on-line at URL: http://www.acdlabs.com/products/logp/ logp_predict.htm on Aug 20, 1997.
Scheme 1
phobic interaction to catalysis. The only situation where the false assignment as an aqueous-faced active site (that in truth was lipid-exposed) could occur, given the present results, is if a lowering of the kcat/Km by 70-fold is occurring as a result of restricted access to the active site
706 Chem. Res. Toxicol., Vol. 11, No. 6, 1998
Figure 3. Estimated incremental binding energies of a series of 4-hydroxyalk-2-enals to microsomal glutathione transferase when the total (9), calculated aqueous (2), or detergent (b) substrate concentration was used. The dashed line shows the theoretical curve obtained if all potential hydrophobic binding energy is realized (2.9 kJ/mol of methylene, based on the octanol/ water partitioning in ref 6).
due to steric hindrance. This situation is however unlikely in this particular case, given that many bulky substrates are accommodated by the enzyme (3) and that the additional benzene ring is para to the reacting carbon. This example illustrates that substrate analogues of different hydrophobicity but displaying similar kcat/Km values can be used to tentatively predict the location of an active site. It is suggested that this favorable situation can be found by exploring any series of substrates (to reach a point where the potential binding energy from additional hydrophobic groups cannot translate into increases in kcat/Km but will only change the partition behavior). The analysis becomes less clear-cut if hypothetical ratios of 32 and 5.8 (vmore hydrophobic/vless hydrophobic at 0.001 and 0.011 volume fraction detergent, respectively) would have been obtained indicating a hydrophobic binding site. These same ratios are expected upon aqueous approach if kcat/Km is increased by 70-fold for the more hydrophobic compound, realizing a potential binding energy of 10.5 kJ/mol [which would be in the range expected for a benzene ring binding to a site with hydrophobicity similar to n-octanol (12.2 kJ/mol; 6)]. The favorable situation encountered with CNBAN/ CNBAM does not always present itself directly but may or may not be arrived at by exploring a series of substrate analogues. The latter unfavorable situation is illustrated, again employing microsomal glutathione transferase as an example, in Figure 3 where incremental binding energies of a series of 4-hydroxyalk-2-enals were determined from the corresponding kcat/Km values comparing the 7-, 9-, 11-, and 14-carbon compounds (from ref 8). The kcat/Km values were calculated using the total added substrate concentration, making no assumptions on substrate partitioning or location of the substrate binding site, and yielded the original curve given (filled squares in Figure 3). The figure also shows the recalculated incremental binding energies when the predicted substrate concentrations in the detergent (0.5%) and aqueous phases are used to obtain the corresponding kcat/ Km values. Clearly, an aqueous approach results in all
Morgenstern
potential binding energy being translated into catalytic efficiency. Assuming a hydrophobic binding site facing the detergent phase, no incremental binding is apparent upon increase of the carbon chain length. However, discrimination between an aqueous faced binding site that can interact effectively with hydrophobic substrates or a binding site facing the lipid interior that does not cannot be made since a limiting condition is not reached in this case. It is interesting to note that the hypothetical kcat/Km value for the 4-hydroxytetradec-2-enal is 1.3 × 106 M-1 s-1, approaching those of very efficient enzymes (assuming aqueous location), whereas the kcat/Km values obtained assuming detergent-phase approach yield modest catalytic efficiencies of 20-90 M-1 s-1 for all four compounds. Thus, either an aqueous active site that has evolved high catalytic efficiency in response to a substrate spectrum or a lipid-facing active site that has not (the lipid-facing site largely taking advantage of the higher local substrate concentration) was, in theory, favored during evolution. It is difficult to envision a totally unspecific hydrophobic binding site facing the membrane interior as there appears to be no possibility to utilize unspecific hydrophobic interactions in a lipid environment. Rather, very precisely architectured binding sites (as in aqueous binding sites for hydrophilic solutes) are expected. The possibility remains that lipid-exposed hydrophobic active sites simply utilize the increased local concentration of hydrophobic substrates in the membrane phase. In this particular case, however, the CNBAN/CNBAM data for microsomal glutathione transferase contradicts this notion since a substantial decrease in the kcat/Km for the more hydrophobic substrate, should it approach through lipid, is predicted (vide supra). The magnitude of the incremental binding energies to microsomal glutathione transferase (8) indicates that the binding site has a hydrophobicity similar to that of n-octanol (6). Higher binding energies (up to 5-fold) are deduced when specific binding pockets that rely on hydrophobic interactions are analyzed (such as the one in chymotrypsin, in ref 6). Thus the hydrophobic binding site in microsomal glutathione transferase, being less specific, appears to utilize hydrophobic interactions in a general way. Linear hydrocarbons can interact favorably, at least up to a 14-carbon length, whereas the extra benzene ring in CNBAN does not appear to interact at all. This observation does fit with the known specificity of the enzyme toward fatty acid and phospholipid hydroperoxides that are formed endogenously during oxidative stress (8). The experiments with hydroxyalkenals, although favoring an aqueous binding site, are inconclusive since, apparently, the favorable situation where only partition behavior is affected was not reached. Another example from the literature, now analyzing the binding of homologous ligands to cytochrome P450, serves to illustrate the more favorable situation. In Figure 4, a graph is resketched showing the dependence of the negative logarithm of binding constants for 4-alkyl-substituted pyridines on increasing carbon chain length (5, 9). The apparent binding constants were calculated making no assumption on the location of the binding site. The bellshaped curve obtained is compatible with an aqueousfaced hydrophobic binding site considering the behavior of the relative concentration changes of ligand in the aqueous and lipid phases (compare Figure 1). Analogues
Binding Site Orientation
Chem. Res. Toxicol., Vol. 11, No. 6, 1998 707
zyme in whole cells (11). Furthermore, although cytochrome P450 displayed a bell-shaped curve in its affinity for a series of ligands, the disposition of the same ligands in the intact perfused organ displayed a monotonic increase (12). Of course, the notion of a high local concentration gradient extending from membranes has to be experimentally verified. In summary, it is expected that membrane-bound detoxication enzymes, cytochrome P450, epoxide hydrolase, microsomal glutathione transferases, and UDPglucuronysyl transferases all possess aqueous-faced binding sites to accommodate hydrophobic substrates. Provided that substrate analogues can be obtained that fulfill the criteria depicted above, these sites can be tentatively verified by the present approach.
Figure 4. Negative logarithm of the apparent dissociation constant (Kd) of cytochrome P450 for a homologous series of 4-alkyl-substituted pyridines. Redrawn from ref 5.
Acknowledgment. This research was supported by the Swedish Cancer Society, the Carl Trygger Foundation, and funds from Karolinska Institutet.
References that are less hydrophobic are present in the aqueous phase at close to the added concentration allowing incremental hydrophobic interactions to translate into tighter binding that is visible in the apparent binding constant (ascending limb in Figure 4). At a certain carbon chain length, dependent on the volume fraction of the hydrophobic compartment, the concentration decrease in the aqueous phase accompanying the increased ligand hydrophobicity exactly balances the incremental binding (a plateau is reached). When further methylene groups are added to the ligand, which do not interact with the binding site, the decreasing concentration in the water phase translates into weaker apparent binding (the descending limb in Figure 4). The apparent weaker binding is analogous to the decreased rates for the more hydrophobic substrate in the CNBAN/CNBAM experiment (Table 1). Actually, the descending limb of Figure 4 has strong diagnostic value in favor of an aqueous approach since, were the approach from the lipid phase, it can only result from a strongly decreasing trend in binding affinities. In conclusion, bearing in mind the caveat that steric hindrance is a possibility, one can draw the tentative conclusion that both microsomal glutathione transferase and cytochrome P450 contain hydrophobic binding sites that are aqueous-faced (i.e., exposed to the cytosol). Why then, are many detoxication enzymes, especially those metabolizing the most hydrophobic substrates, membrane proteins? Perhaps a local concentration gradient exists, upheld by the great concentration difference between the lipid and soluble compartments and the presence of cytosolic binding proteins [which are important in ensuring an efficient diffusion of hydrophobic and amphipatic molecules in liver cells (10)]. Indeed, this idea is supported by experimental evidence. The conjugation of a very hydrophobic compound by soluble and microsomal glutathione transferases, although occurring at similar rates in isolated subcellular fractions, was preferentially carried out by the membrane-bound en-
(1) Heirwegh, K. P., Vermeir, M., Zaman, and Z. (1994) Multiphasic modelling of ligand/acceptor interactions. The hydrophobicitydependent binding of relatively small amphiphilic substances to acceptor proteins and the nature and facedness of acceptor sites. J. Biochem. Biophys. Methods 29, 23-47. (2) Morgenstern, R., and DePierre, J. W. (1983) Microsomal glutathione transferase, Purification in unactivated form and further characterization of the activation process, substrate specificity and amino acid composition. Eur. J. Biochem. 134, 591-597. (3) Morgenstern, R., Lundqvist, G., Hancock, V., and DePierre, J. W. (1988) Studies on the activity and activation of rat liver microsomal glutathione transferase, in particular with a substrate analogue series. J. Biol. Chem. 263, 6671-6675. (4) Peterson, G. L. (1977) A simplification of the protein assay method of Lowry et al. which is more generally applicable. Anal. Biochem. 83, 346-356. (5) Heirwegh, K. P., Meuwissen, J. A., Vermeir, M., and De Smedt, H. (1988) Liposomes as carriers of poorly water-soluble substrates: linear modelling of membrane systems with catalytic or binding sites of different facedness. Significance of experimental membrane partition coefficients and of kinetic and equilibrium parameters. Biochem. J. 254, 101-108. (6) Fersht, A. (1985) Enzyme Structure and Mechanism, W. H. Freeman & Co., New York. (7) Andersson, C., Mosialou, E., Weinander, R., and Morgenstern, R. (1994) Enzymology of microsomal glutathione S-transferase. In Conjugation-dependent carcinogenicity and toxicity of foreign compounds (Anders, M. W., and Dekant, W., Eds.) Vol. 27, pp 19-35, Academic Press, San Diego. (8) Mosialou, E., Piemonte, F., Andersson, C., Vos, R., Van Bladeren, P. J., and Morgenstern, R. (1995) Microsomal glutathione transferase-lipid-derived substrates and lipid dependence. Arch. Biochem. Biophys. 320, 210-216. (9) Born, J. L., and Hadley, W. M. (1980) Ligand interactions of 4-pyridyl ethers with cytochrome P-450. Proc. West. Pharmacol. Soc. 23, 259-262. (10) Tipping, E., and Ketterer, B. (1981) The influence of soluble binding proteins on lipophile transport and metabolism in hepatocytes. Biochem. J. 195, 441-452. (11) Hargus, S. J., Fitzsimmons, M. E., Aniya, Y., and Anders, M. W. (1991) Stereochemistry of the Microsomal Glutathione S-Transferase Catalyzed Addition of Glutathione to Chlorotrifluoroethene. Biochemistry 30, 717-721. (12) Yih, T. D., and van Rossum, J. M. (1977) Ks Values of some homologous series of barbiturates and the relationship with the lipophilicity and metabolic clearance. Biochem. Pharmacol. 26, 2117-2120.
TX980013E
