Abstract
Although the human lung cytochrome P450 2A13 (CYP2A13) and its liver counterpart cytochrome P450 2A6 (CYP2A6) are 94% identical in amino acid sequence, they metabolize a number of substrates with substantially different efficiencies. To determine differences in binding for a diverse set of cytochrome P450 2A ligands, we have measured the spectral binding affinities (KD) for nicotine, phenethyl isothiocyanate (PEITC), coumarin, 2′-methoxyacetophenone (MAP), and 8-methoxypsoralen. The differences in the KD values for CYP2A6 versus CYP2A13 ranged from 74-fold for 2′-methoxyacetophenone to 1.1-fold for coumarin, with CYP2A13 demonstrating the higher affinity. To identify active site amino acids responsible for the differences in binding of MAP, PEITC, and coumarin, 10 CYP2A13 mutant proteins were generated in which individual amino acids from the CYP2A6 active site were substituted into CYP2A13 at the corresponding position. Titrations revealed that substitutions at positions 208, 300, and 301 individually had the largest effects on ligand binding. The collective relevance of these amino acids to differential ligand selectivity was verified by evaluating binding to CYP2A6 mutant enzymes that incorporate several of the CYP2A13 amino acids at these positions. Inclusion of four CYP2A13 amino acids resulted in a CYP2A6 mutant protein (I208S/I300F/G301A/S369G) with binding affinities for MAP and PEITC much more similar to those observed for CYP2A13 than to those for CYP2A6 without altering coumarin binding. The structure-based quantitative structure-activity relationship analysis using COMBINE successfully modeled the observed mutant-ligand trends and emphasized steric roles for active site residues including four substituted amino acids and an adjacent conserved Leu370.
Xenobiotic-metabolizing cytochrome P450 (P450) enzymes act on chemically diverse small molecules, often with overlapping specificities, but with the formation of distinct metabolites and/or metabolic rates. The only active human cytochromes P450 in the 2A subfamily are CYP2A13 and CYP2A6. CYP2A13 is expressed primarily throughout the respiratory system including nasal mucosa, trachea, and lung (Su et al., 2000; Zhu et al., 2006), whereas CYP2A6 is primarily a hepatic enzyme (Yamano et al., 1990; Fernandez-Salguero et al., 1995). Coumarin 7-hydroxylation is a characteristic activity for the P450 2A subfamily. However, reports disagree on whether the catalytic efficiency for coumarin is 10-fold higher for CYP2A6 compared with that for CYP2A13 (He et al., 2004b) or similar (von Weymarn and Murphy, 2003). A number of other substrates are metabolized very differently by these two enzymes, including nicotine and the nicotine-derived compounds cotinine and 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK). CYP2A13 metabolizes nicotine and cotinine with 23- and 15-fold higher catalytic efficiency (kcat/Km) than CYP2A6, respectively (Bao et al., 2005). In addition, the metabolism of NNK by CYP2A13 occurs at a catalytic efficiency 215-fold greater than that of CYP2A6 (He et al., 2004a). Both CYP2A13 and CYP2A6 have been implicated in tobacco-related lung cancers and have genetic polymorphisms reported to decrease cancer risk (Tan et al., 2001; Wang et al., 2003).
CYP2A13 also functions in the activation of several other procarcinogens and the metabolism of toxins. Aflatoxin B1 is efficiently activated by CYP2A13 to a mutagenic epoxide, an activity similar to that of CYP1A2, but is not activated by CYP2A6 (He et al., 2006). Phenacetin, an antipyretic drug used for many years and later found to be a kidney toxin, is O-deethylated efficiently by both CYP1A2 and CYP2A13 (kcat/Km 0.36 for both enzymes) but not by CYP2A6 (Fukami et al., 2007; DeVore et al., 2008).
Diversity in the functions of CYP2A13 and CYP2A6 is not likely to originate with the global structure as the enzymes are 94% identical, differing by only 32 amino acids and the crystallographically determined structures have an overall root mean square deviation for the Cα carbons of only 0.5 Å (Smith et al., 2007). Ten of the 32 amino acid differences are in or near the active site and are likely to be directly related to the functional differences observed in substrate metabolism. We have established previously that differential binding and metabolism of the analgesic phenacetin in human CYP2A enzymes is the result of the disparate amino acids at positions 208, 300, 301, and 369 in concert with the effects of those four substitutions on adjacent conserved residues Phe209 and Leu370 (DeVore et al., 2008). The net steric effects of these differences alter active site morphology to determine CYP2A compatibility with phenacetin binding and O-deethylation. Other studies of human CYP2A enzymes have revealed that residues 117 and 372 are critical for both coumarin (He et al., 2004b) and NNK metabolism and that the residue at position 208 is also critical for NNK metabolism (He et al., 2004a).
The goal of the present work is to determine whether these same amino acids are responsible for differential binding of other ligands by human CYP2A enzymes. In several cases, other P450 proteins are known to adapt to different ligands in a more induced-fit mode of binding, frustrating attempts to predict drug binding and metabolism. If the relatively small CYP2A selectivity is based on a more steric-dominated, lock-and-key type model, then xenobiotic binding to these enzymes may be suitably evaluated using docking methodologies based on the known crystal structures. In this study, we have surveyed a series of five structurally diverse cytochrome P450 2A ligands: nicotine, phenethyl isothiocyanate (PEITC), coumarin, 2′-methoxyacetophenone (MAP), and 8-methoxypsoralen (MOP). Coumarin, PEITC, and MAP were identified as having a range of differential affinities for the human CYP2A enzymes and were thus used to investigate the effects of mutations at 10 active site positions differing between CYP2A6 and CYP2A13. These ligand-binding studies identified a number of substitutions that changed the KD significantly for at least one substrate. Substitutions combining three or four of these CYP2A13 amino acid residues into CYP2A6 indicate that the differences between CYP2A13 and CYP2A6 binding affinities observed for the ligands in this study are largely dependent on the same four residues identified in the phenacetin study: 208, 300, 301, and 369 (DeVore et al., 2008). To evaluate the roles that individual features of the active site have in modulating ligand-binding affinity, we performed structure-based quantitative structure-activity relationship (QSAR) analysis via the COMBINE method, and the results were consistent with a steric binding rationale.
Materials and Methods
Materials. Coumarin, PEITC, MOP, (S)-nicotine, and MAP were purchased from Sigma-Aldrich (St. Louis, MO).
Protein Expression and Purification. The human CYP2A13 and CYP2A6 proteins used in these studies were all generated by deleting the N-terminal transmembrane sequence (Δ2–23), altering several residues at the modified N terminus (from 24-WRQRKSR-30 to 24-AKKTSSK-30), and adding four histidine residues at the C terminus. All mutations were also made in this background. Many studies with several different membrane P450 enzymes have shown that truncated versions metabolize the same substrates as the full-length parent enzyme with the same regio- and stereoselectivity (von Wachenfeldt et al., 1997; Scott et al., 2001; Schoch et al., 2004; Wester et al., 2004). Given substantial differences in the availability and purity of full-length and truncated enzymes and differences in the requirement for lipid, direct comparisons of activity rates are difficult, but similar activities and binding have been reported (Wester et al., 2004; Yano et al., 2004). Use of the truncated versions allows a far greater yield of protein than could be attained with the full-length enzyme (Scott et al., 2001). These modifications allow generation of a highly purified protein that retains its catalytic activity and in quantities that are sufficient for ligand binding and other biophysical studies. Such truncated CYP2A proteins were expressed with yields ranging from 320 to 1100 nmol/l of culture at the microsome stage and were further purified as described previously (Smith et al., 2007). The total yield of purified protein ranged from 17.4 to 155 nmol of P450/l of Escherichia coli culture. Two mutant proteins, CYP2A6 I208S/I300F/G301A and CYP2A6 I208S/I300F/G301A/S369G, had specific contents of ≥3 to 3.5 nmol/mg, but all other proteins had specific contents ranging from 10.4 to 19.3 nmol/mg. These CYP2A proteins were competent in the metabolism of phenacetin as reported previously (DeVore et al., 2008).
Site-Directed Mutagenesis. The mutations were generated using the modified CYP2A gene in an expression vector (pKK2A13dH) or (pKK2A6dH) as a template (Smith et al., 2007) and the QuikChange Site-Directed Mutagenesis method (Stratagene, La Jolla, CA). Synthetic oligonucleotides were designed to yield the desired amino acid substitution but also to contain silent restriction site mutations where possible to facilitate identification of mutated genes (Table 1). Oligonucleotides were synthesized by Genosys (Woodlands, TX).
Ligand-Binding Titrations. Binding titrations with the ligands were conducted with purified P450 protein at 20°C using a UV-visible scanning spectrophotometer (UV-2101; Shimadzu Scientific Instruments, Columbia, MD). Protein was diluted to 1 μM in 100 mM potassium phosphate buffer, pH 7.4. Diluted protein was divided equally between two 1.0-ml quartz cuvettes (1-cm path length), and a baseline was recorded (300–500 nm). Freshly prepared aliquots of ligand dissolved in 100% ethanol were added to the sample cuvette. An equal volume of ethanol was added to the reference cuvette. Difference spectra were collected (300–500 nm) after an equilibration period. During the spectral titrations, the total amount of ethanol added did not exceed 2%. Binding to P450 was monitored as the absorbance difference (ΔA) between the minimum (∼420 nm) and the maximum (∼385 nm). The apparent binding constant (KD) and the maximum spectral change (ΔAmax) were determined from nonlinear least-squares regression fitting to the following equation: where P is total P450 concentration and S is total ligand concentration. Nonlinear regression was accomplished using GraphPad Prism 4 (GraphPad Software Inc., San Diego, CA).
Computational Methods. To elucidate the role that individual residues play in determining the relative affinities of ligands as a function of active site structure, we performed structure-based QSAR analysis via the COMBINE method (Ortiz et al., 1995). Protein structures were generated from crystal structures of human CYP2A13 with indole-bound (Protein Data Bank code 2P85) (Smith et al., 2007) and human CYP2A6 complexed with methoxsalen (Protein Data Bank code 1Z11) (Yano et al., 2005). Mutations of CYP2A13 were performed in silico via the Biopolymer suite in Insight-II (Accelrys, San Diego, CA), and each unique mutant protein was allowed to relax via CHARMM simulations (Brooks et al., 1983), entailing a 100-step molecular mechanics preconditioning run to alleviate clashes induced by the point mutation, a 20-ps molecular dynamics warmup, a 20-ps thermal equilibration, and a 100-ps analysis run. In all cases, the native ligand (i.e., indole and methoxsalen for CYP2A13 and CYP2A6, respectively) and crystallographic waters were retained throughout the simulation. All atomic charge and force field parameters corresponded to standard CHARMM terms (Brooks et al., 1983). For each unique protein, the conformation present at the end of the analysis run was used for ligand docking and subsequent COMBINE analysis.
To derive a bound conformer for coumarin bound to CYP2A6, the known crystal structure of this complex (Protein Data Bank code 1Z10) (Yano et al., 2005) was aligned to the relaxed CYP2A6 structure derived in the previous step. This alignment was effected via the Biopolymer module in SYBYL (version 8.0; Tripos, St. Louis, MO) with extra weight being assigned to alignment of the heme moiety (to focus the alignment within the active site region). For all other ligand-protein complexes, the relevant ligand (coumarin, MAP, or PEITC) was docked to each of the distinct proteins (CYP2A6, CYP2A13, and the L110V, A117V, S208I, A213S, F300I, A301G, M365V, L366I, G369S, and H372R mutational variants of CYP2A13) via FlexX (Rarey et al., 1996). In each case the protein structure was stripped of the native ligand and all waters except the one closest to the Asn297 side chain (because of observations in the CYP2A13 crystal structure that it served to bridge between indole and the Asn297 side chain). Fifty poses were then requested per ligand, using the default charge and conformational search setting. For mutant protein CYP2A13 A117V, docking simulations for the ligands MAP and coumarin did not achieve a reasonable bound structure; thus, the initial guess positions of these two ligands were derived by aligning to the conformations attained for the CYP2A13 protein.
The set of COMBINE descriptors used for QSAR analysis consisted of van der Waals and electrostatics interaction terms between each ligand and each residue (i.e., including all amino acids, plus the heme and the one water) present in each of the 12 proteins. The QSAR model was then trained by partial least-squares fitting as implemented in the Simca P program (Umetrics AB, Umea, Sweden) of a weighted linear combination of the COMBINE electrostatic and van der Waals parameters to experimental pKD (= log[KD]) data. By default, the top scoring ligand pose for each ligand/enzyme pair was selected for analysis; however, in instances in which substantial discrepancies were observed between the experimental pKD value and that computed from leave-one-out cross-validated correlation analysis, lower-scoring poses were tested for improved fidelity. In the four cases in which the top-scoring ligand conformer did not correlate well with the experiment (coumarin and MAP binding to H372R and L366I), none of the lower scoring conformers achieved marked improvement; thus, these cases were discarded as outliers. Overall, the use of van der Waals and electrostatic interaction parameters yielded a single COMBINE QSAR model that was reasonably able to predict the experimental binding affinities for 32 of the 36 possible combinations of the three ligands (MAP, PEITC, or coumarin) with CYP2A6, CYP2A13, and the 10 different CYP2A13 mutants (R2 = 0.87, Q2LOO = 0.60; three components).
Results
Binding of Structurally Diverse Ligands to CYP2A6 and CYP2A13 Proteins. Initially a series of five structurally diverse cytochrome P450 2A ligands were surveyed, and binding affinities were determined for both CYP2A6 and CYP2A13 enzymes. Coumarin was chosen because it is an often-used marker substrate for cytochromes P450 in the 2A subfamily (Fernandez-Salguero et al., 1995; He et al., 2004b; Kim et al., 2005). Nicotine (Bao et al., 2005; Murphy et al., 2005) and MAP (Su et al., 2000; von Weymarn et al., 2005) are also CYP2A substrates, whereas 8-methoxypsoralen (von Weymarn et al., 2005) and PEITC (von Weymarn et al., 2006) are inhibitors. Each of these ligands displayed type I binding spectra with a shift from ∼420 to ∼385 nm (Fig. 1), indicating an increase in the high-spin, five-coordinate state of the iron as water bound to the heme iron in the resting state is displaced upon ligand binding in the active site.
All five of the ligands surveyed bound more tightly to CYP2A13 than to CYP2A6 (Table 2) but with a substantial range in both overall affinities and in the differences in affinity between the two enzymes. The ligand that bound the tightest to both enzymes is 8-methoxypsoralen. In contrast, nicotine had the lowest affinity for both enzymes. MAP, PEITC, and coumarin had intermediate KD values.
In terms of differential affinities, the compounds fall into three groups. First, coumarin binds equally well to both enzymes. Second, nicotine, 8-MOP, and PEITC have moderate selectivity for CYP2A13. The substrate nicotine binds to CYP2A13 4.7 times more tightly than to CYP2A6, whereas the inhibitors PEITC and MOP bind to CYP2A13 13- to 14-fold more tightly than to CYP2A6. The ligand with the largest difference in binding affinity is MAP, which binds 74-fold more tightly to CYP2A13 than to CYP2A6.
Ligand Binding to Mutants of CYP2A13. Of the five ligands surveyed above, MAP, PEITC, and coumarin were chosen to characterize mutant CYP2A proteins in which active site amino acids from the opposite CYP2A enzyme were substituted. These three ligands were chosen because they span the range of difference observed in KD values, from MAP with its 74-fold difference in binding affinity, to PEITC with a moderate 14-fold difference in affinity, to coumarin, which binds equally well to both proteins. The 10 amino acids that were mutated were selected based on their location within the CYP2A6 and CYP2A13 active sites (Fig. 2) and their effects on phenacetin binding and metabolism (DeVore et al., 2008).
MAP binding to the single-site CYP2A13 mutant proteins revealed that four modifications resulted in substantially different KD values compared with the parent CYP2A13 enzyme (Table 3). Substitution of Ile for Phe300 (F300I) resulted in a 15-fold increase in KD, S208I had a 12-fold increase, and A301G had an 8-fold increase in KD. A117V had a more moderate 4.1-fold decrease in affinity for MAP. The remaining six mutations had relatively little effect on MAP binding, although L366I improved binding affinity almost 2-fold.
Binding of PEITC to CYP2A13 is characterized by a KD of 0.43 μM, approximately 14-fold higher affinity than CYP2A6 (Table 3). The single substitution of F300I in the CYP2A13 protein resulted in a 14-fold loss in PEITC affinity, such that its KD is essentially that of CYP2A6. The mutation A301G, located directly adjacent to F300I, demonstrated a 7-fold increased KD. The CYP2A13 mutants S208I, M365V, and G369S had KD values that were increased 3- to 4-fold compared with CYP2A13. The remaining single mutants yielded smaller changes in the PEITC binding affinity. Again, CYP2A13 L366I was the only protein to demonstrate an increase in binding affinity for PEITC, although the change was small.
In contrast to PEITC and MAP, which bind more tightly to CYP2A13 than to CYP2A6, the KD value for coumarin is ∼2.9 μM for both CYP2A13 and CYP2A6. Again, the F300I substitution had the most dramatic effect on KD, with a 5-fold increase. Two other substitutions, G369S and S208I, increased KD values by 2- and 3-fold, respectively. The substitution H372R increased coumarin-binding affinity by nearly 50% to a KD of 1.5 μM, whereas L366I increased coumarin-binding affinity by 5-fold to 0.57 μM.
Ligand Binding to CYP2A6 Multiple Mutants. The above studies revealed amino acids that decreased CYP2A13 affinity for MAP and PEITC (Fig. 3), but it is much easier to disrupt ligand binding than to improve it. To verify that these same amino acids would increase CYP2A6 binding affinity for MAP and PEITC, several of these substitutions were incorporated into CYP2A6 and binding of the same three ligands was evaluated. Because mutations at 208, 300, and 301 had the largest negative effects on MAP and PEITC affinity for CYP2A13, the CYP2A6 triple mutant I208S/I300F/G301A was generated. This mutated protein had affinities for MAP and PEITC that were intermediate between those for CYP2A6 and CYP2A13 (Table 3). The affinity for MAP was increased the most, but coumarin affinity was also decreased by 2-fold. Because the effect of the substitution at position 208 was much greater for MAP and the effect of substitution at position 369 was greater for PEITC, a second CYP2A6 triple mutant that consisted of the I300F/G301A/S369G substitutions was examined. Compared with the previous CYP2A6 mutant, this protein had essentially the same affinity for MAP but greater affinity for PEITC. Finally, combination of these substitutions in the CYP2A6 quadruple mutant I208S/I300F/G301A/S369G yielded additional increases in affinity for MAP without further altering PEITC binding and maintaining the approximate KD for coumarin.
COMBINE Results. The ability to engineer into CYP2A6 the capacity to bind phenacetin (DeVore et al., 2008) and the application of these same mutations to rationally increase CYP2A6 affinity for PEITC and MAP suggest that it may be possible to treat the CYP2A active site as a relatively static active site for docking ligands and predicting ligand-binding affinities. To test this hypothesis, receptor-based QSAR studies were undertaken using the comparative binding energy (COMBINE) approach to identify and quantify individual protein/ligand interactions that contribute to the differing affinities for the three ligands with the different proteins under study (Fig. 3). As described under Materials and Methods, the result was a model in which contributions from van der Waals and electrostatic interactions could be used to reasonably predict the observed binding affinities in most cases. The agreement between the binding affinities observed experimentally and those predicted by the COMBINE model had a strong correlation (R2 = 0.87) (Fig. 4). Thus, this model reproduces most of the observed affinity trends for the three different ligands across the 12 CYP2A parent and mutant enzymes under consideration (Table 4). It successfully predicts the two enzyme-ligand complexes with the lowest affinity (MAP interacting with CYP2A6 and with CYP2A13 F300I, respectively) and only narrowly interchanges the order of the next two (coumarin binding to CYP2A13 F300I and MAP binding to CYP2A13 S208I). It also correctly identifies the enzyme-ligand complex with the greatest affinity (PEITC interacting with CYP2A13 L366I) and yields predictions that differ by no more than half a logarithmic unit for 32 of the 36 ligand-enzyme complexes examined experimentally.
Analysis of the electrostatic features that contribute to the binding affinity trends reveals that by far the most significant are favorable coupling with the iron and porphyrin portions of the heme and the water molecule that putatively facilitates H-bond bridging to the Asn297 side chain amide NH2. Each of the three ligands has a polar end that either binds to the water or to Asn297 directly and a more hydrophobic end that couples with the heme via a loose cation-π interaction. The strongly binding CYP2A13/PEITC conformer places its aromatic group in a position that is much more suitable for favorable interactions with the Fe than does CYP2A6/MAP (Fig. 5A), which probably is a key factor behind the stronger activity of the former. Other electrostatic interactions are fairly modest in importance and are located proximal to the heme or fairly distant from the ligand (>6.0 Å) and thus have only small effects on the overall binding affinity predictions. In addition, almost all of the top 10 electrostatic contributors are conserved between CYP2A6 and CYP2A13 and thus are unlikely to discriminate in terms of ligand binding selectivity.
In terms of van der Waals contributions, the largest contribution to binding is also derived from the active site water molecule, but other important van der Waals contacts include amino acids that differ between the two enzymes or that are conserved but have notable interactions. In particular, the nonconserved Phe300, Ala301, and Ala117 are three of the top six most important protein-ligand interactions. In this case, all but one of these top six protein-ligand interactions confer a small penalty to ligand binding. From this, one would argue that substitutions of CYP2A6 residues that further occlude the volume of the active site (e.g., A117V and F300I) should generally lead to poorer ligand binding. This consistently seems to be the case, with the exception of coumarin binding to A117V. In this situation, docking reveals that coumarin is still small enough in proximity to position 117 as to engage in a favorable lipophilic interaction with the larger valine side chain without crossing over into the steric clash domain. One would conversely expect that mutations that substitute smaller amino acid side chains within the active site in place of larger ones would lead to improved ligand activity. In the case of A301G, however, this is universally not the case. However, the modeling results suggest that whereas the alanine to glycine mutation itself opens up space in the active site closer to the polar end of the ligands, there is a slight shift of the I helix backbone and Phe300 toward the ligand that actually leads to slightly greater constriction in the area of the lipophilic group (Fig. 5B). Comparison of the aligned CYP2A6 and CYP2A13 experimentally determined crystal structures reveals a slight helical kinking that arises in the proximity of residue 301 that occludes the cavity by approximately ∼0.2 Å in the case of CYP2A6.
Finally, conserved amino acids Asn297, Leu370, and Phe480 also have significant steric roles in this model. Asn297, which is considered to be the lone available H-bonding site in the CYP2A active sites, has the second largest contribution for an amino acid. The one favorable protein steric contribution to ligand binding in the COMBINE model is the conserved amino acid Leu370, which is located near ligand hydrophobic groups. In this case, the best interpretation of its status as a favorable contact is that mild steric clashes with ligands are compensated for with favorable entropy, desolvation energy, or some combination of the two. The benefits of desolvation are obvious in that in all complexes the ligand places a hydrophobic group in the vicinity of the hydrophobic side chain of Leu370. Entropy is also a plausible contributor in that Leu370 is easily the most mobile side chain observed in the COMBINE modeling studies. The extent of this variability in this side chain in our modeling studies is evident in Fig. 5C, wherein the Leu370 side chain location and conformation are depicted for a series of relevant proteins. Leu370 is also located between Leu366 and His372. Explaining the effects of mutations of these two residues (CYP2A13 L366I and H372R) with MAP and coumarin constitutes the greatest shortcoming of the COMBINE model.
Discussion
Examination of these five structurally diverse cytochrome P450 2A ligands revealed type I binding and a very large range in the affinities of CYP2A13 versus CYP2A6. At the extremes, coumarin bound equally well to both proteins, whereas MAP bound more than 74 times more tightly to CYP2A13 than to CYP2A6.
Coumarin binds to CYP2A6 by hydrogen bonding with Asn297, adopting an orientation within a largely planar active site that positions C7 for hydroxylation (Yano et al., 2005), consistent with the only coumarin metabolite detected for CYP2A6. Asn297 and the planar active site are conserved in CYP2A13 (Smith et al., 2007), as is the generation of 7-hydroxycoumarin. Consistent with these similarities, the nearly identical KD values for both proteins suggest that coumarin binds similarly in CYP2A13. However, CYP2A13 also forms a significant amount of 3,4-epoxide and smaller amounts of 6-hydroxy- and 8-hydroxycoumarin (von Weymarn and Murphy, 2003). This finding suggests that CYP2A13 binds coumarin via N297 hydrogen bonding, but is less constrained in its approach to the activated oxygen on the heme so that C6 or C8 can also be hydroxylated and also binds coumarin in a second, inverted orientation to expose C3/C4 for epoxidation.
MAP is O-demethylated by both enzymes to produce 2′-hydroxy-acetophenone and formaldehyde, and its binding showed the largest difference in affinity among the ligands examined. Although kinetics are not available for MAP metabolism, Su et al. (2004) reported that at 1 mM substrate concentration, CYP2A13 metabolized MAP at a rate 4-fold faster than that of CYP2A6. Thus, the scale of the differences in MAP affinity was unexpected, based on the single productive orientation in the active site suggested by a single product and the preliminary metabolic rate information available.
The remaining ligands had smaller intermediate differences in affinity. 8-Methoxypsoralen and PEITC are both mechanism-based inactivators (Koenigs et al., 1997; von Weymarn et al., 2005, 2006) and had similar ∼14-fold greater affinities for CYP2A13. Because the reactive metabolites are not known, it is difficult to predict the orientation of these ligands in the active site. (S)-Nicotine demonstrated an ∼5-fold preference for CYP2A13. Both CYP2A enzymes hydroxylate nicotine at the pyrrolidine 5′ position and the methyl, although the latter is generated more readily by CYP2A13 (Murphy et al., 2005). Thus, nicotine probably binds similarly in both enzymes, orienting either the 5′ or methyl carbons for oxidation, although the latter may be slightly favored in the case of CYP2A13.
Site-directed mutagenesis and binding titrations with MAP, PEITC, and coumarin revealed several amino acids responsible for disparate ligand KD values. The three substitutions with the largest effects were F300I, A301G, and S208I. CYP2A13 F300I yielded the largest changes in ligand binding, 5- to 18-fold decreases for all three ligands (Fig. 3), suggesting that this residue plays a central role in ligand binding. This finding is consistent with both the steric role suggested by the COMBINE model and a recent crystal structure of the CYP2A6 I208S/I300F/G301A/S369G quadruple mutant (DeVore et al., 2008). Comparison of the CYP2A structures reveals that I300 and F300 overlap at the α carbon atom but that the phenyl ring is torsioned farther away from the bound ligand than the isoleucine side chain in both CYP2A13 and the CYP2A6 quadruple mutant (Yano et al., 2005; Smith et al., 2007; DeVore et al., 2008). The amino acid at this position is key for phenacetin (DeVore et al., 2008) and indole metabolism (Wu et al., 2005) in CYP2A enzymes. Overall, it is clear that this residue is critical for the higher affinity of CYP2A13 for PEITC and MAP, playing a key role in substrate selectivity.
Substitution of the adjacent CYP2A13 residue, A301, to glycine removes the side chain and causes 7- to 8-fold increases in the KD for MAP and PEITC, respectively, but does not significantly alter the KD for coumarin. This result suggests that coumarin interacts differently with the residue at position 300 than does MAP or PEITC. Because MAP and PEITC bind more tightly to CYP2A13, interactions with this residue also contribute to ligand selectivity.
The S208I substitution has a 14-fold increase in the KD for MAP but much smaller ∼3-fold effects on PEITC and coumarin binding. Because MAP is the ligand with the most contrast in affinity between the two CYP 2A enzymes, interactions with this residue may also be significant in ligand selectivity. The more moderate effects on coumarin binding are consistent with the results of He et al. (2004b), who showed that this mutation had little effect on the kinetics of coumarin 7-hydroxylation by CYP2A13 (He et al., 2004b). This mutation is also reported to decrease the catalytic efficiency of CYP2A13 NNK metabolism to both the keto aldehyde and keto alcohol products, neither of which is formed efficiently by CYP2A6 (He et al., 2004a).
COMBINE evaluation of the influence of electrostatics revealed that the extended heme system is found to exert the dominant influence, favoring ligands that orient negatively charged groups toward the positively charged iron. The key nonheme electrostatic interaction is predicted to be with a water molecule that was isolated as an apparent H-bond donor in the CYP2A13/indole crystal structure (Smith et al., 2007) and may help to bridge the ligand with the conserved Asn297 donor site, but whose position was not rigorously conserved during molecular dynamics relaxation of all mutants. Thus, no significant electrostatic differences that are likely to directly affect ligand binding were identified between CYP2A6 and CYP2A13. In contrast, the COMBINE model identified several differing amino acids at positions 300, 301, and 117 that do have important steric contributions. All three residues have substantial effects not only on the ligand binding reported here but also on phenacetin binding and metabolism. The CYP2A13 A117V increased phenacetin affinity 2-fold and activity by >5-fold, whereas the F300I and A301G mutations showed minimal phenacetin binding or metabolism. Structures of the CYP2A6 I208S/I300F/G301A/S369G quadruple mutant in complex with phenacetin indicated steric differences related to the positioning of the amino acids at 300 and 301 that are also probably relevant to the binding of the ligands examined here.
Analysis of the CYP2A6 and CYP2A13 crystal structures reveals that the difference in the position of Leu370 is a root mean squared deviation of 0.74 Å, as opposed to the global residue average of only 0.51 Å. Substantial further variation was also found when the different CYP2A13 mutants were examined computationally in this study. This flexibility of Leu370 observed in these docking studies is further substantiated by the structure of the CYP2A6 I208S/I300F/G301A/S369G quadruple mutant (DeVore et al., 2008), which demonstrates that the position of Leu370 is altered in response to protein mutations, in this case probably primarily by the adjacent S369G mutation. Leu370 may also be capable of differential adaptation to the presence of different ligands in ways that permit the alleviation of steric clashes. It is possible that a primary cause of the mutagenic influence of positions 366, 369, and 372 may be the effect of mutation on the conformation of Leu370, some of which were modeled poorly by the COMBINE model.
Combination of I300F, G301A, and I208S mutations in the CYP2A6 background was sufficient to substantially increase the affinity for MAP but not for PEITC. Because 1) substitution at 369 had the next largest effects on PEITC binding to CYP2A13, 2) modeling suggested effects of mutation at 369 on the position of Leu370, and 3) this substitution made important contributions to the binding and metabolism of phenacetin, G369S was also incorporated into CYP2A6, resulting in an enzyme whose overall binding affinities for all three ligands were much more similar to those of CYP2A13.
In conclusion, the goal of this study was to gain insight into the differences in structure-binding relationships between the human cytochrome CYP2A6 and CYP2A13 enzymes and their diverse ligands. Binding titrations to CYP2A13 and CYP2A6 mutant proteins in conjunction with QSAR identified key active site amino acids at positions 208, 300, 301, and 369. The development of a COMBINE QSAR model that substantially fit the trends observed in the experimental results suggests that the differing CYP2A residues play primarily steric roles in modulating ligand affinity. Difficulties with predictions for a small subset of mutations adjacent to Leu370 probably occur from demonstrated variations in the positioning of Leu370. Thus, apart from the mobility of Leu370, these results suggest that the CYP2A active site can reasonably be treated as a steric docking problem for prediction of ligand affinities.
Acknowledgments
Thanks are due to a number of students involved in constructing the mutants described in this work (Jenny Morrison, Chad Schroeder, Kyle Bailey, Matthew Axtman, and Melanie Blevins) and to Christopher Wood for assistance with purification.
Footnotes
-
This work was supported by the National Institutes of Health National Institute of General Medical Sciences [Grant GM076343].
-
Parts of this work were previously presented at the following conference: Michno NM, Smith BD, Wood CA, Blevins MA, and Scott EE (2007) Interaction of key active site amino acids distinguishing ligand binding in human lung cytochrome P450 2A13 and liver cytochrome P450 2A6. Experimental Biology 2007; 2007 Apr 28–May 2; Washington, DC. The American Society for Pharmacology and Experimental Therapeutics, Rockville, MD.
-
Article, publication date, and citation information can be found at http://dmd.aspetjournals.org.
-
doi:10.1124/dmd.109.026765.
-
ABBREVIATIONS: P450, cytochrome P450; NNK, 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone; PEITC, phenethyl isothiocyanate; MAP, 2′-methoxyacetophenone; MOP, 8-methoxypsoralen; QSAR, quantitative structure-activity relationship.
- Received January 14, 2009.
- Accepted February 26, 2009.
- The American Society for Pharmacology and Experimental Therapeutics