IDENTIFICATION OF RESIDUES 99, 220, AND 221 OF HUMAN CYTOCHROME P450 2C19 AS KEY DETERMINANTS OF OMEPRAZOLE HYDROXYLASE ACTIVITY

Human P450 2C19 is selective for 4′-hydroxylation of S-mephenytoin and 5-hydroxylation of omeprazole, while the structurally homologous P450 2C9 has low activity toward these substrates. To identify the critical amino acids that determine the specificity of human P450 2C19, we constructed chimeras of P450 2C9 replacing various proposed substrate binding sites (SRS) with those of P450 2C19 and then replaced individual residues of P450 2C9 by site-directed mutagenesis. The 339 NH-terminal amino acid residues (SRS-1-SRS-4) and amino acids 160-383 (SRS-2-SRS-5) of P450 2C19 conferred omeprazole 5-hydroxylase activity to P450 2C9. In contrast, the COOH terminus of P450 2C19 (residues 340-490 including SRS-5 and SRS-6), residues 228-339 (SRS-3 and SRS-4) and residues 292-383 (part of SRS-4 and SRS-5) conferred only modest increases in activity. A single mutation Ile His increased omeprazole 5-hydroxylase to 51% of that of P450 2C19. A chimera spanning residues 160-227 of P450 2C19 also exhibited omeprazole 5-hydroxylase activity which was dramatically enhanced by the mutation Ile His. A combination of two mutations, Ile His and Ser Pro, converted P450 2C9 to an enzyme with a turnover number for omeprazole 5-hydroxylation, which resembled that of P450 2C19. Mutation of Pro Thr enhanced this activity. Residue 99 is within SRS-1, but amino acids 220 and 221 are in the F-G loop and outside any known SRS. Mutation of these three amino acids did not confer significant S-mephenytoin 4′-hydroxylase activity to P450 2C9, although chimeras containing SRS-1-SRS-4 and SRS-2-SRS-5 of P450 2C19 exhibited activity toward this substrate. Our results thus indicate that amino acids 99, 220, and 221 are key residues that determine the specificity of P450 2C19 for omeprazole.

The P450 1 cytochromes represent a ubiquitous superfamily of monooxygenases, which metabolize a vast array of endogenous and exogenous substrates (1)(2)(3). Multiple P450 enzymes appear to have arisen from a single ancestral gene by duplication and diverged by mutation and gene conversion to produce families of structurally related enzymes with overlapping but often distinct substrate specificities. The regio-and stereoselectivity of specific enzymes for particular substrates appears to be encoded in certain defined regions of the primary sequence. Considerable progress has been made in recent years in elucidating the structural determinants of substrate specificity (4). In some cases, substrate specificity between highly related members of the same subfamily has been shown to be defined by a few critical residues or even a single amino acid (5).
P450 2C19 is a member of the human CYP2C subfamily, which includes four structurally related enzymes (6). P450 2C9 and 2C19 are the most highly conserved of these forms, showing 91% structural identity, but have very distinctive substrate specificities. For example P450 2C19, which is polymorphic in man, is the principal enzyme responsible for the stereoselective 4Ј-hydroxylation of S-mephenytoin (7) and is highly selective for the 5-hydroxylation of the popular anti-ulcer drug omeprazole (46). The structurally related P450 2C9 exhibits little activity toward either of these substrates, but exhibits a high turnover number for hydroxylation of tolbutamide and phenytoin and specifically 7-hydroxylates S-warfarin (6). The present studies were designed to elucidate the key amino acids that determine the marked specificity of P450 2C19 for omeprazole and S-mephenytoin.
Since none of the mammalian P450 enzymes have been crystallized, the experimentally derived three-dimensional crystal structures of the bacterial enzymes, P450cam, P450BM-3, and P450terp, have served as models for predicting their structure and function. The current paradigm for such predictions has focused on alignment-based sequence similarities, ␣and ␤-secondary structure, hydropathy indices, computerized modeling based on the crystal structures of the bacterial P450 enzymes and protein engineering using chimeric constructs and sitedirected mutagenesis (8 -12). Although the amino acid sequences of the soluble bacterial P450 differ considerably from those of the membrane-bound mammalian P450 enzymes, these methods have proven resourceful in identifying regions that determine the substrate specificities of the mammalian proteins.
Gotoh (13) predicted six potential substrate recognition sites (SRS) in the mammalian family 2 P450 enzymes based on an alignment with bacterial P450cam whose substrate-binding residues have been identified by x-ray crystallography and analysis of mutations that altered substrate specificity in experimental studies of the P450 2 subfamily. For example, sitedirected mutagenesis studies have shown that the substrate and regiospecificity of the 2C subfamily of cytochrome P450 can be altered by substitution of critical amino acid residues at positions 112 to 115, 301, 359, 364 (14 -19). Replacement of residues 117, 209, 365, and 481 in the 2A subfamily (5, 20 -23) as well as residues 114, 206, 302, 363, 367, and 478 (24 -28) in the 2B subfamily and residue 380 (29) in 2D1 have also been documented to alter substrate or regiospecificity.
To identify key amino acid residues responsible for the marked specificity of human P450 2C19 for two key substrates, omeprazole and S-mephenytoin, we constructed a total of 24 chimeras and mutant enzymes from the structurally related P450 2C9. Constructs were expressed in a yeast cDNA expression system. Initial chimeras replaced regions of P450 2C9 with similar segments of P450 2C19, using restriction enzymes sites located outside the six putative substrate binding domains proposed by Gotoh (13). Subsequently, we used primary sequence alignment and comparative analysis of amino acid variations within the appropriate segments of the human P450 2C proteins to identify residues that might play a role in the substrate specificity of P450 2C19. Site-directed mutagenesis was then used to identify key residues responsible for the substrate specificity of P450 2C19.

EXPERIMENTAL PROCEDURES
Chemicals and Reagents-Restriction and modification enzymes were purchased from New England Biolabs and sequenase sequencing kits and deoxyadenosine 5Ј-[ 35 S]thiotriphosphate from Amersham Life Sciences. 14 C-Labeled omeprazole (53 mCi/mmol), a gift from Glaxo Research and Development (Hertfordshire, United Kingdom (UK)), was unstable and therefore periodically purified to ϳ98% homogeneity by HPLC and stored at Ϫ70°C under absolute ethanol. 14 C-Labeled Smephenytoin (20.7 mCi/mmol) was prepared as described previously (7). Purified P450 reductase was obtained from Human Biologics Inc., Phoenix, AZ.
Plasmids-The cloning of P450 2C9 and P450 2C19 into the yeast vector pAAH5 has been described previously (30). To facilitate restriction enzyme manipulations, the modified cDNAs were recloned in both orientations in the HindIII site of pUC19. P450 2C9/2C19 hybrid clones were constructed by ligation of homologous fragments of the cDNA inserts. All plasmids were propagated in Escherichia coli DH5␣ (Life Technologies Inc.) and recombinant DNA identified by sequencing. The modified cDNAs were excised, inserted into pAAH5 vector, and recombinants in the correct orientation selected by restriction mapping and sequencing of the 5Ј-cloning junctions.
Site-directed Mutagenesis-Amino acid changes were introduced in P450 2C9 and the 2C9/2C19 hybrids using synthetic oligonucleotides (Table I) containing the desired point mutations and a second primer, which abolished a unique restriction site in the pAAH5 plasmid. The procedure was essentially as described by Deng and Nickoloff (31) with minor modifications. The concentration of the plasmid selection oligonucleotide was reduced 5-fold in the annealing reaction. After extension with unmodified T7 DNA polymerase, the product was purified by Sephadex G50 spin column chromatography. BMH71-18, a mismatch repair-deficient bacteria strain, was transformed. Plasmid DNA isolated from the pooled colonies was digested to linearize non-mutated clones prior to retransformation in DH5␣ cells. Individual colonies were sequenced to confirm incorporation of the mutagenic primers.
Enzymatic Assays-S-Mephenytoin 4Ј-hydroxylation was assayed as described previously (7) except that cytochrome b 5 was not added to the reactions. To assay omeprazole 5-hydroxylase, recombinant yeast microsomes were preincubated at 37°C for 5 min with purified P450 reductase (100 units/10 pmol of P450) and dilaurophosphatidylcholine (0.3 g/pmol of P450) and then placed on ice. 14 C-Labeled omeprazole (400 M, 22 mCi/mmol) was added to 50 mM HEPES buffer (pH 7.4) containing 0.1 mM EDTA and 1.5 mM MgCl 2 . The reaction was preincubated at 37°C with shaking for 5 min, initiated by addition of NADPH (2 mM), and terminated with equal volume of methanol after 15 min. After centrifugation, products were analyzed by reverse phase HPLC using an isocratic solvent phase consisting of 45 parts methanol and 55 parts water. Detection and quantitative analysis of radioactive peaks was accomplished through an on-line radiochemical detector.
Modeling and Structural Alignments-Omeprazole was docked into two structures using the molecular modeling package SYBYL at the Molecular Modeling Laboratory, University of North Carolina, Chapel Hill, NC. The first structure was a model for P450 2B1 described by Szklar et al. (19), which is based on the x-ray structures of P450cam, P450terp, and P450BM-3. The second structure was that of P450cam. Omeprazole was docked into the binding pockets of both structures by requiring C19 of omeprazole to be 3-4 Å from the heme iron and rotating the substrate to minimize Van der Waals contacts with the protein. P450 2C9 and P450 2C19 were aligned with P450cam using GAP from the Wisconsin GCG package with standard defaults.

RESULTS AND DISCUSSION
Chimeras-Twenty-four chimeric 2C9/2C19 and mutant P450 2C9 proteins were analyzed for their ability to hydroxylate omeprazole and S-mephenytoin. The structures of the chimeras and their relationships to the SRS proposed by Gotoh (13) are shown in Fig. 1A. The 2C9/2C19 339 -490 chimera exhibited only a slight increase in omeprazole 5Ј-hydroxylase activity over the wild-type P450 2C9, suggesting that the carboxyl terminus of P450 2C19 has relatively minor influence on its capacity to metabolize omeprazole (Fig. 2). Although this region encompasses SRS-5 and SRS-6, it contains some of the most highly conserved P450 domains including the heme-binding region. P450 cytochromes 2C9 and 2C19 differ by only 11 amino acids in this 152-residue domain, with a preponderance of conservative replacements. It is, therefore, not surprising that this region conferred only a small increase in omeprazole hydroxylase activity. In contrast, the chimeric 2C19 1-339 /2C9 protein exhibited the largest (13-fold) increase in omeprazole 5-hydroxylase activity. The 2C19 segment of this chimera spanned the putative SRS-1-SRS-4 with 31 nonidentical residues, including 16 nonsynonymous changes. A nearly comparable increase was observed with the 2C9/2C19 160 -383 chimera spanning SRS-2-SRS-5 of 2C19. The 2C19 region of these two chimeras overlap from residues 160 -339, suggesting that this region may be requisite to substrate specificity. The 2C19 1-339 / 2C9 chimera contained 12 amino acid changes at the NH 2terminal end, which were not present in the 2C9/2C19 160 -383 chimera, presumably accounting for the slightly higher activity of the former protein. Two other chimeric proteins containing only parts of the 160 -383 fragment of 2C19, i.e. 2C9/2C19 228 -339 (containing SRS-3 and SRS-4) and 2C9/ 2C19 292-383 (containing SRS-5 and part of SRS-4), exhibited a smaller increase (3-7-fold) in omeprazole hydroxylase activity than the 2C9/2C19 160 -383 chimera, suggesting that the region spanning residues 160 -227 of P450 2C19 could be involved in substrate specificity.
Mutants-We initially generated five single amino acid mutants I99H (m1), Y243D (m3), M257I (m5), L362I (m4), and S451F (m2) and combinations of these mutations (m34, m45, m134, m145, m1245, and m1345) in P450 2C9 (Fig. 1B). Each amino acid was mutated to the corresponding amino acid in P450 2C19. The amino acids were chosen as residues that might be involved in substrate recognition based on alignment with P450cam and mouse 2A4/5 and selection of amino acids in  P450 2C19 that were unique compared to other members of the P450 2C subfamily. Ile 99 3 His was chosen because it is the only difference between P450 cytochromes 2C9 and 2C19 in SRS-1, and His 99 is unique to P450 2C19. Ile 257 is a unique amino acid in P450 2C19. Tyr 243 3 Asp is a unique nonsynonymous amino acid change near SRS-3. Ile 362 corresponds to contact residue Val 296 in P450cam and is the only unique amino acid in SRS-5. Ser 451 3 Phe is a unique nonsynonymous change located between SRS-5 and SRS-6. Of the single amino acid mutants, only I99H (Ile 99 3 His) in SRS-1 increased omeprazole 5Ј-hydroxylase activity markedly (Fig. 3). The other mutations had essentially no effect. Multiple mutants containing the Ile 99 3 His mutation had comparable activity to that of I99H. In contrast, multiple mutants m34 and m45, which did not contain the Ile 99 3 His mutation, exhibited no increase in omeprazole hydroxylase activity (data not shown). Residue 99 maps to the interhelical region of B and BЈ helices, three residues from the NH 2 terminus of BЈ, which comprises part of the substrate-contact loop in the structure of P450cam (36 -38). Many mutations that alter the substrate specificity of members of the 2A, 2B, and 2C subfamilies of cytochrome P450 map to the BЈ-C interhelical region near the carboxyl terminus of SRS-1 (5, 39 -43). Three-dimensional analysis of P450cam, P450terp, and P450BM-3 suggests that this substrate-contact loop, the largest in P450, extends from proximal end of the ␤1-5 sheet through the distal end of the BЈ-C coil (8). His 99 of P450 2C19 corresponds to Pro 87 of P450cam in the alignment of Gotoh (13), which is one of three amino acids in SRS-1 that contact camphor. Moreover, it represents the only amino acid change between P450 2C9 and P450 2C19 in SRS-1 and is the only residue in SRS-1 that is unique to P450 2C19. This change involves the replacement of a nonpolar, hydrophobic isoleucine with a polar and possibly charged histidine residue. This polar histidine has the capacity for hydrogen bonding. It is possible that the residue is involved in substrate binding or alternatively changes the configuration of the active site. Comparison of P450 2C19 with a three-dimensional model of P450 2B1 (19) predicts that this amino acid would be approximately 15 Å from the center of the substrate binding pocket. Alternatively, when omeprazole is docked into binding pocket in the three-dimensional model of P450cam, we find that Pro 87 (which aligns with His 99 in 2C19) is near the docked substrate (approximately 6 Å) as shown in Fig. 4. Moreover, the adjacent Phe 88 of P450cam is even closer to the substrate (ϳ2 Å). Thus, the effect of the Ile 99 3 His mutation is consistent with this model.
Analysis of Residues 160 -227-Analysis of the chimeric data indicated that amino acids 160 -227 may also play a crucial role in the specificity of P450 2C19 for omeprazole. We therefore constructed a 2C9/2C19 160 -227 chimera and an identical construct containing an Ile 99 3 His mutation. The 2C9/ 2C19 160 -227 chimera exhibited a 7-fold increase in omeprazole 5-hydroxylase activity compared to P450 2C9 (Fig. 5). More- over, when combined with the an Ile 99 3 His mutation, this chimera exhibited a dramatic 38-fold increase in activity. The activity of this mutated chimera was even greater than that of native P450 2C19.
We next sequentially modified amino acids in this region to those of the corresponding residues in P450 2C19 as shown in Fig. 1B. Six amino acids of P450 2C19 (three conservative and three nonconservative changes) differ from those of P450 2C9 between amino acids 160 and 227, with five of these mapping within 20 tandem amino acid residues of the carboxyl end of SRS-2. Mutations Lys 206 3 Arg (K206R), Leu 208 3 Val (L208V), Pro 221 3 Thr (P221T), and Ser 210 3 Thr in combination with Ile 99 3 His did not increase omeprazole 5Ј-hydroxylase activity over that of I99H alone (Fig. 5), indicating that these residues are not the critical amino acids in this region. However, mutation of the polar serine at position 220 to a nonpolar hydrophobic proline residue (Ser 220 3 Pro) increased activity to ϳ90% of that of P450 2C19. When a second nonconservative mutation was made in the tandem residue at position 221 (P221T) in conjunction with S220P and I99H, the omeprazole hydroxylase activity of the triple mutant (I99H/S220P/ P221T) was equivalent to that of P450 2C19. Sequence alignment indicates that residues 220 and 221 reside midway between the COOH-terminal end of the F-helix and the NH 2terminal end of G-helix in the interhelical loop, outside of any known substrate binding sites. This region corresponds to the F-G helical turn. This region is extremely variable in length in the P450 superfamily, and bacterial P450cam contains no amino acids corresponding to Pro 220 or Thr 221 of P450 2C19. The three-dimensional model of P450 2B1 would predict that these amino acids are far from the substrate-binding pocket; however, it has been suggested that this region may form a flexible lid over the substrate-binding pocket and thus form part of the substrate access channel (8,11). Proline residues give loops greater flexibility. Therefore, the introduction of Pro 220 may refold the F-G loop and change accessibility to the active site or bring these residues closer to the substrate pocket. Interestingly, Uno and Imai (44) reported that residues 210 -262, which contain part of the same domain, cooperate with amino acids 90 -125 to confer substrate specificity to laurate (-1)-hydroxylase P450 2C2. Several other investigators have also identified important key residues in or near this domain, including residue 209 of mouse 2A5 as well as 206 and 209 of rat 2B1 (5,20,22,26,45). The present results clearly show that residues 220 and 221, which are in the F-G loop of human P450 2C19 and outside the SRS proposed by Gotoh (13), are critical determinants of omeprazole hydroxylase activity.
S-Mephenytoin 4Ј-Hydroxylase Activity-Somewhat surprisingly, S-mephenytoin hydroxylase activity did not parallel omeprazole hydroxylase activity in the various chimeras and mutants. As shown in Fig. 6, only two chimeric proteins, 2C19 1-339 /2C9 and 2C9/2C19 160 -383 , exhibited significant increases in S-mephenytoin hydroxylase activity over P450 2C9. Activities of these proteins were ϳ30% of that of P450 2C19. Furthermore, S-mephenytoin hydroxylase activity was not detected in any of the mutant proteins (data not shown). The differences in activity of the His 99 mutants toward S-mephenytoin and omeprazole could be due to the fact that although S-mephenytoin is a smaller molecule (16 heavy atoms) than omeprazole (24 heavy atoms), it is more branched. The side chain of histidine can potentially interact with both omeprazole and mephenytoin since it has the capacity for hydrogen bonding, whereas this cannot happen with isoleucine. When Smephenytoin is docked with the 4-position of the phenyl ring replacing the 5-position of omeprazole (Fig. 4), the remaining rings of both molecules would be placed in an orientation that could allow strong hydrogen bonding with His 99 . However, the branched structure of S-mephenytoin may require additional three-dimensional accommodation in the active site. These results suggest that specificity of P450 2C19 for S-mephenytoin may require a more complex enzyme configuration than that for omeprazole, possibly involving multiple substrate recognition domains acting in concert.
Conclusions-Chimeric proteins and amino acid mutations have been widely used to determine critical residues responsible for substrate specificity of closely related P450 proteins. We have utilized this approach to identify amino acids that are critical determinants of omeprazole 5-hydroxylase activity of human P450 2C19. We first identified regions important in substrate specificity by analyzing the catalytic activity of chimeras and then sequentially replaced individual residues in P450 2C9 by site-directed mutagenesis. We have identified three key amino acids at positions 99, 220, and 221, which are critical determinants of omeprazole 5-hydroxylase activity. These residues map to SRS-1 and the F-G loop.