An Inhibitory Monoclonal Antibody Binds in Close Proximity to a Determinant for Substrate Binding in Cytochrome P45OIIC5”

We used the expression of chimeric proteins and point mutants to identify amino acids of the hepatic progesterone 21-hydroxylase P450IIC5 which are part of an epitope recognized by an inhibitory monoclonal antibody and which affect substrate binding. Three amino acids of P450IIC5 at positions 113, 115, and 118 were introduced into P450IIC4, which is 95% identical to P450IIC5. The resultant chimeric protein acquired binding of the monoclonal antibody 1F11, which is highly specific and inhibitory for P450IIC5. Point mutants in P450IIC4 showed that two of the three changes, T115S and N118K, contribute to the epitope recognized by this antibody. The T115S mutant bound the antibody weakly (Kd greater than 30 nM) whereas the N118K mutant bound the antibody as tightly as P450IIC5 (Kd less than or equal to 0.7 nM). Thus, residues 115 and 118 are located on the surface of these enzymes, and the Lys/Asn difference at amino acid 118 is largely responsible for the high degree of discrimination which this antibody exhibits between P450IIC5 and P450IIC4. The valine to alanine mutation at position 113 conferred to P450IIC4 a lower apparent Km for progesterone 21-hydroxylation. Because antibody binding was not affected by this mutation, it is tempting to speculate that this residue is buried in the protein where it exerts its effect on the catalytic activity by interaction with the substrate or alters the positions of residues of the active site. The close proximity of the epitope at positions 115 and 118 to Ala113 suggests that the inhibitory monoclonal antibody interferes with substrate binding.

The mammalian enzymes of the cytochrome P450 (P450)' superfamily metabolize a multitude of substrates and only few details are known about how this ability is encoded in the primary and tertiary structures of these proteins. This is largely due to the lack of direct three-dimensional structural information for the mammalian P450 enzymes. We have used previously the expression of chimeric P450s to define segments in the primary structures of P450IIC4 and P450IIC5 which determine substrate binding (1). Epitopes of antibodies which are directed against functional proteins are expected to lie on the surface of the molecule and therefore have the potential to yield information about the three-dimensional organization of the protein. Thus, the identification of epitopes provides information regarding surface residues of the protein. Antipeptide antibodies that react with functional proteins and that are directed toward specific segments of the protein provide similar information, and they have been useful in the elucidation of the topology of P450 enzymes (2, 3). In this report, we describe the mapping of amino acids of the epitope recognized by a monoclonal antibody, IF11, to obtain structural and functional information about the two progesterone 21-hydroxylases P4501IC5 and P450IIC4. When compared with P4501IC4, P450IIC5 catalyzes progesterone 21hydroxylation with a >lO-fold lower apparent K,. The lFl1 antibody is highly specific for and inhibitory of P450IIC5 (41, but it does not react with P450IIC4. We have expressed chimeric enzymes derived from these two P450s and point mutants of P450IIC4 to identify not only residues on the surface of these proteins which contribute to the epitope recognized by the 1F11 monoclonal antibody, but also to identify a specific residue which modulates the apparent K , for progesterone 21-hydroxylation.

MATERIALS AND METHODS
The monoclonal antibodies IF11 and 2F5 were developed against P450 1 (P450IIC5) which had been purified from rabbit liver as reported earlier (4). Chimeric cDNAs were constructed in pBluescript or pCMV plasmids. Single-stranded DNA was prepared from pBluescript by use of helper phage R408 (Stratagene, La Jolla, CA) or M13K07 (Pharmacia LKB Biotechnology Inc.). Site-directed mutagenesis was performed with the phosphorothionate method ( 5 ) , using reagents supplied by Amersham Corp.
The abbreviations and conventions used in this article are: The generic term "P450" is used to indicate a cytochrome P-450. Individual forms of P450 are designated according to the uniform system of nomenclature described by Nebert et al. (26) with the following exceptions: The common name, P450cam, is used for P450CI. Two forms of mouse P450IIA (14) which had not been assigned designations in the last published listing (26) are designated as P45OI5,, and P450,,h to distinguish the steroid 15a-hydroxylase and the coumarin hydroxylase, respectively. Mutations are designated by indicating the one-letter abbreviation for the residue that was replaced, its position in the sequence, and the one-letter designation of the new residue in the indicated order.

6215
COSl cells were grown in T75 or T225 flasks in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (GIBCO), modified Eagle's medium-nonessential amino acids (GIBCO), and 50 units/ml each of penicillin and streptomycin (GIBCO). The cells were transfected by the DEAE-dextran method (6). Microsomal fractions were prepared from cells 48-72 h after transfection by differential centrifugation as described (1). The 21hydroxylation of progesterone was determined in vivo a t 48 h after transfection by supplementation of the culture medium for 1 h with 2 p~ ["C]progesterone, followed by extraction and quantification of the 21-hydroxyprogesterone as described earlier (1).
A filter-binding assay was used to determine the binding of the l F l l monoclonal antibody to the expressed proteins. Ten pg of microsomal protein from transfected cells were transferred onto a nylon filter using a dot blot apparatus (Schleicher & Schuell). Free binding sites were blocked with a blocking solution consisting of 2% nonfat dry milk, 0.05% Tween 20,0.01% sodium azide, 0.15 M NaCl, and 0.01 M sodium phosphate, pH 7.4. The filter was then cut into strips which were incubated separately in 15-ml centrifuge tubes with 2 ml of monoclonal antibody which had been diluted to the appropriate concentrations in blocking solution. After four washes with blocking solution, the strips were incubated with '2511-labeled sheep-antimouse IgG (Amersham) and washed extensively in blocking solution. Bound label was visualized on the filter by autoradiography after it was reassembled. Individual dots were then cut from the strips, and the radioactivity bound to each was determined using a y-counter. These values were fit with either a four-parameter logistic equation or the following equation, where B is the amount of radioactivity bound to the sample (counts/ rnin), €Irnax is the estimate for maximum binding (counts/min), [Ab] is the total molar concentration of antibody, K d is the estimated dissociation constant, and C is an estimate for a constant amount of nonspecific binding (counts/min). Both methods yield indistinguishable estimates of Kd by nonlinear regression using Sigma Plot 4.0 (Jandel Scientific, Corte Madera, CA).
Immunoblotting was performed by electrophoretically transferring 20 pg of microsomal protein separated on a 10% polyacrylamide gel containing 1% sodium dodecyl sulfate (7) onto nitrocellulose (8). Free binding sites were blocked with blocking solution, and the proteins were detected with 2F5, a monoclonal antibody which reacts with both, P450IIC4 and P450IIC5 (9). This antibody recognizes an epitope that is spatially distinct from that which binds the l F l l monoclonal antibody and, in contrast to the latter, i t detects P450IIC5 on immunoblots. Monoclonal antibody 2F5 was detected using autoradiography with either "'1-labeled goat-anti-mouse IgG or with a sheepanti-mouse IgG conjugated to horseradish peroxidase and a luminescent substrate (ECL, Amersham).
Immunoinhibition was determined by incubating COSl microsomes for 15 min at 22 "C in a volume of 90 pl containing 10 p~ progesterone with or without 2 pmol of monoclonal antibody 1F11. The reaction was started by addition of 10 pl of a NADPH-regenerating system (lo), and the sample was incubated for 60 min at 37 "C.
The substrate dependence of progesterone 21-hydroxylation was determined as described earlier (1). Briefly, 25 pg of microsomal protein were incubated in a final volume of 100 pl with varying amounts of [14C]progesterone and a NADPH-regenerating system for 30 min. After extraction with chloroform and thin layer chromatography, progesterone and 21-hydroxyprogesterone were quantified by scintillation counting. The values for apparent K,,, and VmaX were estimated by nonlinear fitting of the Michaelis-Menten equation to the experimentally determined rates.

RESULTS
Differential Binding of the Monoclonal Antibody l F l l to P45OZZC4, P45OZZC5, and Chimera G-We have shown previously that P450IIC5 and P450IIC4 metabolize progesterone to deoxycorticosterone by 21-hydroxylation when the proteins are expressed from their corresponding cDNAs in COSl cells (1). The >lO-fold higher K, of P450IIC4 for progesterone can be lowered to the apparent K, of P450IIC5 in a chimeric enzyme which carries three amino acids of P450IIC5 at positions 113, 115, and 118 in P450IIC4 (1). This chimera is referred to as chimera G.
To compare the binding of the l F l l monoclonal antibody to chimera G with that to P450IIC5 and P450IIC4, we used a filter-binding assay to determine the reactivity of the lFll monoclonal antibody with microsomal fractions prepared from COSl cells, expressing each protein following transfection with an expression plasmid harboring cDNAs encoding the individual proteins. Fig. 1 shows that the l F l l monoclonal antibody binds tightly to chimera G indicating that the three amino acid differences, which determine the apparent K, toward progesterone, also transfer onto P450IIC4 a crucial determinant of the epitope recognized by the l F l l monoclonal antibody. This binding affinity is similar to that of the l F l l monoclonal antibody with P450IIC5 ( Fig. 2), whereas no binding to IIC4 is detectable.
Monoclonal Antibody 1 F l 1 Is a Potent Inhibitor of Chimera G-The efficient binding of the l F l l monoclonal antibody to hybrid G suggested that this antibody also would inhibit . 10 pg of microsomal protein were applied onto a nylon filter with a dot blot apparatus. The filter was then cut into vertical strips and each strip was incubated with monoclonal antibody lFll at the indicated concentrations. T h e l F l l antibody was then detected with a '2511-labeled sheep anti-mouse IgG, the strips were reassembled in the original order and exposed to x-ray film. B, Western blot of the samples in A with the monoclonal antibody 2F5. Ala, Ser, and Lys refer to the V113A, T115S, and N118K mutants of P450IIC4, respectively. 20 pg of microsomal protein were denatured and separated on a 10% polyacrylamide gel containing sodium dodecyl sulfate. The P450s were detected with the monoclonal antibody 2F5, a sheep-anti-mouse antibody conjugated to horseradish peroxidase, and subsequent luminescence detection by exposure to x-ray film for 30 s. Monoclonal antibody 2F5 reacts with both P450IIC4 and P450IIC5. efficiently the progesterone 21-hydroxylase activity of this enzyme. To test this, we incubated 15 pg of microsomes prepared from COSl cells which expressed chimera G with 10 p~ progesterone in the presence or absence of either 2 pmol (20 nM) of the 1Fll monoclonal antibody or of a monoclonal antibody specific for P450IIC3 (11). The antibody directed against P45011C3 did not inhibit progesterone 21-hydroxylation; however, the lFl1 monoclonal antibody inhibited 93% of this activity. Thus, changing three amino acids of P450IIC4 to those of P450IIC5 in chimera G makes this chimeric enzyme indistinguishable from P450IIC5 for IF11 binding and inhibition.
Relative Contributions of Ala"', Ser"', and Lys118 to the Epitope of Monoclonal Antibody 1F11-To determine the relative contributions of the individual amino acid changes in chimera G to this epitope, we constructed the point mutants by site-directed mutagenesis of the P450IIC4 cDNA. These mutants were expressed in COSl cells, and microsomal fractions prepared from these cells were assayed for their capacity to bind the lFl1 monoclonal antibody. Fig. 1 shows that the V113A substitution did not confer binding of the l F l l monoclonal antibody to P450IIC4, that the T115S mutant showed weak binding, and that the N118K change had a strong effect leading to extensive binding at low concentrations of antibody.
The level of expressed protein was examined for each mutant by immunoblotting and immunodetection with 2F5, a monoclonal antibody which reacts with both P450IIC5 and P450IIC4 (9). According to the immunoblot (Fig. 1B) for the samples employed in Fig. lA, the V113A mutant and P450IIC4 are expressed at about the same concentration as the N118K mutant which reacts strongly with the IF11 monoclonal antibody. Therefore, the absence of immunoreactivity of the l F l l monoclonal antibody with the V113A mutant or with P450IIC4 is not due to the absence of expressed protein.
From the data in Fig. 1, the Kd of antibody binding was estimated to be <1 nM for P450IIC5, chimera G, the N118K mutant and >30 nM for the T115S mutant, respectively. Because the concentration of antigen bound to the filter is roughly equivalent to the estimated K d for P450IIC5, chimera G, and the N118K mutant, the true K d could not be estimated but is clearly less than 1 nM. Kd values for P450IIC4 or P450IIC4-V113A could not be obtained because binding was not detected. Fig. 2 shows the results together with the estimated binding curves determined by a nonlinear least squares fit. The Kd values for chimera G and the mutant P450IIC4 N118K are indistinguishable from the values for P450IIC5. Taken together this indicates that the Lys/Asn difference between P450IIC5 and P450IIC4 at position 118 is the only major determinant leading to the capacity of the l F l l monoclonal antibody to distinguish between the two enzymes. Moreover, these results indicate that the determinant of the K, for the substrate progesterone described earlier (1) and the epitope recognized by the inhibitory monoclonal antibody, 1F11, are in close proximity. The Effect of Point Mutations in P45OIIC4 on Progesterone 21-Hydroxylase Activity-The three amino acid changes V113A, T115S, and N118K in chimera G had been used previously to localize a region in P450IIC4 and P450IIC5 which is responsible for the difference in apparent K, between these enzymes (1). We have shown previously that the N118K mutant did not alter the apparent K, of P450IIC4 (1). The two additional point mutants were expressed from their cDNAs in COSl cells and after 48 h, the culture medium was supplemented with 2 pM ['4C]progesterone. At this concentration of substrate, large differences of 21-hydroxylation between P450IIC5 and P450IIC4 are seen due to their different K, values. Fig. 3 compares the enzymatic activities and protein levels of cells transfected without DNA (Mock) to those transfected with either expression plasmids encoding P450IIC4, P450IIC5, two independent clones of P4501IC4-V113A, two independent clones of P450IIC4-T115S, or the P450IIC4-Nl18K point-mutants.
The V113A mutants showed elevated 21-hydroxylase activity at this substrate concentration when compared with P4501IC4, although they are expressed at lower levels as judged by their reactivity toward the monoclonal antibody 2F5 (Fig. 3B). The T115S and N118K mutations have no detectable effect on the catalytic activity of the P450IIC4.
Kinetic Analysis of P450IIC4-V113A--When compared with the K, for progesterone 21-hydroxylation of P450IIC4, >25 pM, the V113A mutant showed a markedly lower apparent K , of 7.5 WM. This indicates that the increased catalytic activity of this chimera is mainly due to an increase in the apparent affinity toward the substrate progesterone (Fig. 4). However, the chimera G had shown an apparent K, of 2.2 WM which is very close to the K, of 1.7 p~ for P450IIC5. It is likely, therefore, that the T115S and/or the N118K differences together with the V113A difference contribute to the lower apparent K,,, of chimera G toward progesterone as compared with P4501IC4, although individually they do not appear to affect the activity of P4501IC4. Only a rough comparison of V, , , values can be made due to limitations in the precise estimation of the concentration of the enzyme by immunoblotting. When expressed relative to the amount of microsomal protein, the values obtained for the V113A mutant are similar to those obtained for P450IIC5 using the same transfection protocol.

DISCUSSION
The monoclonal antibody l F l l is highly specific for P45011C5 and does not react measurably with P450IIC4. The After 1 h at 37 "C the products were extracted and analyzed by TLC and autoradiography (27). P denotes the mobility of progesterone. DOC indicates the mobility of the metabolite 11deoxycorticosterone (21-hydroxyprogesterone). The two lanes for both the V113A and T115S mutants represent results obtained with two independent clones selected in the mutagenesis experiment and confirmed by sequence analysis. B, immunoblot of microsomal fractions obtained from the cells in A. 20 pg of microsomal protein was separated on a 10% polyacrylamide-sodium dodecyl sulfate gel, blotted to nitrocellulose, and reacted sequentially with monoclonal antibody 2F5 and a "'I-labeled sheep anti-mouse IgG. An autoradiogram of the blot is shown. exchange in P450IIC4 of the three amino acids at positions 113, 115, and 118 of P450IIC5 to form chimera G was sufficient, however, to confer immunoreactivity with the l F l l monoclonal antibody to this enzyme, and the antibody was found to inhibit the microsomal enzyme. This indicates that at least part of this segment is exposed at the surface of this functional protein. A similar finding was made for P450IIB1 in a previous study in which its topology was investigated with antipeptide-antibodies (2). This study showed that antibodies directed against two peptides corresponding to amino acids 108-116 and 122-131, respectively, of P450IIB1 bound to the native enzyme in an enzyme-linked immunosorbent assay. Moreover, this antibody correctly localized the enzyme in rat liver microsomes as judged by immunoelectron microscopy and enzyme-linked immunosorbent assay (2). These two peptides align to residues 107-115 and 121-130, respectively, of P450IIC5. Thus the segment of P450IIC5 which we show in this study to contribute to the epitope recognized by the l F l l monoclonal antibody lies between these two segments of P450IIB1. It is therefore likely that residues in this region are on the surface of both P450IIB1 and P450IIC5.
Because chimera G also exhibits the apparent K,,, of P450IIC5 for progesterone 21-hydroxylation rather than that of P450IIC4, we asked whether the changes necessary to confer immunoreactivity are identical or independent from the changes which modulate the apparent K,,,. Two individual mutations in P450IIC4, N118K and T115S, were able to transfer immunoreactivity with the l F l l monoclonal antibody onto P450IIC4. This indicates, that Ser1Is and Lys'" of P450IIC5 are part of the epitope recognized by the l F l l monoclonal antibody and that their side chains are on the surface of the protein. The T115S change had a clearly detectable but weak effect when compared to the N118K mutation which alone conferred most of the immunoreactivity to P450IIC4 yielding a K d close to that of P450IIC5. This indicates that for monoclonal antibody 1F11, LYS"~ is the major determinant which makes P450IIC5 distinguishable from P450IIC4. The T115S and the N118K changes had no detectable effect individually on the catalytic activity of the mutants. The Only differences in the sequence are shown in the second row of each pair. The number of residues which have been shown to contribute to the catalytic differences between each pair of enzymes and the total number of sequence differences between the pairs is given in the right column. The vertical line shows the alignment position of the residue which modifies the catalytic activity of all six proteins. Segments of P450IIB1 corresponding to the two peptides used to elicit inhibitory antibodies that recognize the native enzyme (13) are underlined. Results for P45015u (1IA3-15~~) and P450coh (IIA3coh) were taken from Lindberg and Negishi (14) and for P450IIB1 and P450IIB2-1,2 from Aoyama et al. (15). mation of the protein in the vicinity of the active site. It seems likely that Ala113 in P450IIC5 is not a strong determinant of the epitope of the l F l l monoclonal antibody in that it alone does not effect measurable binding with the IF11 monoclonal antibody when this change is made to P4501IC4. We suspect that the alanine side chain is localized toward the interior of the protein and influences the interaction of the enzyme with bound substrate, and thus it does not participate in binding the antibody.
The mechanism by which l F l l inhibits the function of P450IIC5 is unknown. It is interesting to note that the two antipeptide antibodies directed against residues 108-116 and 122-131 of P450IIB1 had been shown previously to be potent inhibitors of 7-ethoxycoumarin 0-deethylation catalyzed by P450IIB1 when compared to antibodies directed toward 13 other peptides corresponding to distinct segments of P4501IB1 (13). The l F l l monoclonal antibody is also inhibitory of P4501IC5 and chimera G. The close vicinity of the epitope of monoclonal antibody l F l l to Ala113 which determines the apparent K,,, for the 21-hydroxylation of progesterone, suggests that this antibody interferes with substrate access or binding.
The valine to alanine replacement at position 113 of P450IIC4 conferred a marked improvement in the K,,, for progesterone 21-hydroxylation. Fig. 5 demonstrates that this exact position of the primary structure corresponds to a Val-Ala difference at position 117 between the mouse P45OCoh and P45OIs,, (14) and to the Ile-Phe difference at position 114 between rat P450IIB1 and a variant P450IIB2 (15) when these sequences are aligned (16). In both instances it has been shown, that these residues play a critical role in determining differences in the catalytic activity of the respective pairs of enzymes. The V117A change in P45OCoh decreased the coumarin hydroxylase activity three fold (14). A I114F change, in combination with the L58F mutation in P450IIB1, changed the ratio between 16a and 16p metabolites of testosterone by about 50-fold (15). Variation at this position of alignment (18) also underlies an allelic polymorphism affecting coumarin hydroxylation in the mouse (19)(20)(21). This suggests that this residue in general may be a critical residue for substrate recognition and that P450IIC5 shares the three-dimensional organization of P450IIB1, P450IIB2-1, P450,,,,,and P45OI5,,. This functionally important residue lies in a region in which class IIC P450s show considerable sequence variability. Alignment with P450cam shows that this hypervariable region, residues 90-120 of class IIC P450s, maps to a surface loop in the bacterial protein which contains Tyrg6 (16,22), a residue which determines binding of the substrate camphor in P450cam by positioning the substrate by hydrogen bonding (23,24). This surface loop is one of five flexible loops which are thought to control the access of the substrate to P450cam (25). Thus, the V113A mutation which affects the K,,, for progesterone 21-hydroxylation could affect either the binding of progesterone directly or the access of progesterone to its binding site. This suggests that despite their low identity in amino acid sequence, P450 enzymes share important topological features. The substrate-contact loop of the P450cam structure which harbors Tyrg6 (23) could therefore be a general determinant of substrate specificity in P450 enzymes. Because this loop aligns to the hypervariable region of the eukaryotic P450s which we have characterized here, it is tempting to speculate that the loop structure of this region allows genetic variation to occur which alters substrate selectivity but does not disrupt the overall topology of the P450s thereby contributing to the functional diversity of closely related P450 enzymes.
This study has defined structural features of P450s IIC5 and JIC4 which are likely to be conserved in all members of class 11, the largest and most diverse of the P450 gene family. Residue 113 markedly influences catalytic activity and other residues in close proximity are located at the surface where they act as epitopes for inhibitory antibodies.