Defining the arachidonic acid binding site of human 15-lipoxygenase. Molecular modeling and mutagenesis.

Mammalian lipoxygenases have been implicated in the pathogenesis of several inflammatory disorders and are, therefore, important targets for drug discovery. Both plant and mammalian lipoxygenases catalyze the dioxygenation of polyunsaturated fatty acids, which contain one or more 1,4-cis,cis-pentadiene units to yield hydroperoxide products. At the time this study was initiated, soybean lipoxygenase-1 was the only lipoxygenase for which an atomic resolution structure had been determined. No structure of lipoxygenase with substrate or inhibitor bound is currently available. A model of arachidonic acid docked into the proposed substrate binding site in the soybean structure is presented here. Analysis of this model suggested two residues, an aromatic residue and a positively charged residue, could be critical for substrate binding. Validation of this model is provided by site-directed mutagenesis of human 15-lipoxygenase, despite the low amino acid sequence identity between the soybean and mammalian enzymes. Both a positively charged amino acid residue (Arg402) and an aromatic amino acid residue (Phe414) are identified as critical for the binding of fatty acid substrates in human 15-lipoxygenase. Thus, binding determinants shown to be characteristic of non-enzymatic fatty acid-binding proteins are now implicated in the substrate binding pocket of lipoxygenases.

Mammalian lipoxygenases have been implicated in the pathogenesis of various inflammatory conditions such as arthritis, psoriasis, and bronchial asthma (1,2). There is also evidence that 15-lipoxygenase may be involved in atherosclerosis. 15-Lipoxygenase is induced in human atherosclerosis lesions (3) and it is capable of oxidizing low density lipoprotein to its atherogenic form (4,5). All lipoxygenases catalyze the stereo-specific oxygenation of fatty acids containing 1,4-cis,cispentadiene. Currently, three distinct mammalian lipoxygenase enzymes have been characterized, 5-, 12-, and 15-lipoxygenases, named for their positional specificity on arachidonic acid. Molecular cloning and sequencing of mammalian lipoxygenases from a variety of organisms have revealed a high degree of sequence identity. Comparison of plant and mammalian lipoxygenases, however, reveals only 25% amino acid identity between the two families (6).
At the time this study was initiated, the only lipoxygenase for which an atomic resolution structure had been determined was soybean lipoxygenase-1 (7,8). A solvent-accessible cavity within the catalytic domain is proposed to be the substrate binding site (7). Since the structure was solved in the absence of any bound substrate or inhibitor, it is difficult to assign functional relevance to particular structural features. We have chosen site-directed mutagenesis to test hypotheses concerning which amino acids may interact with the substrate. Previous work, using amino acid sequence alignment and site-directed mutagenesis, identified amino acid residues at positions 417 and 418 in the human platelet 12-lipoxygenase and human reticulocyte 15-lipoxygenase (corresponding to sThr 556 and sPhe 557 of soybean lipoxygenase-1) 1 as important features of the binding site for the methyl end of the fatty acid substrate (9 -11). Examination of the soybean structure indicates that the homologous residues, sThr 556 and sPhe 557 , line the internal cavity, providing support that this cavity may be the arachidonic acid binding site in soybean lipoxygenase-1 (7). We present a model for the binding of arachidonic acid to lipoxygenase based on the high resolution structure of the soybean enzyme (12) and have used this model to identify key amino acid residues in human 15-lipoxygenase that could play a role in substrate binding. Iron is an essential cofactor of the enzyme (13). EPR data have indicated that the iron shuttles from its ferric to its ferrous form during catalysis (14). Thus, the region of the enzyme in proximity to the iron center can be considered the catalytic core of the enzyme. Within a 15-Å sphere of the catalytic core of the enzyme, as revealed by this structure, the sequence identity between the human and soybean enzyme is only 28%. We have, however, assumed that the catalytic core of the plant and mammalian lipoxygenases will be structurally similar, given the similar activities of the enzymes (15). Mutagenesis results validate the assumed structural similarities between the plant and mammalian lipoxygenases and provided support for the proposed arachidonic acid model.

MATERIALS AND METHODS
Molecular Modeling-The program Insight II (Biosym) was used to examine the soybean lipoxygenase-1 structure (12) and to calculate the solvent-accessible surface of the internal cavity (probe size 1.4 Å). Arachidonic acid was manually docked into the proposed substrate binding site, and the quality of the binding modes were visually assessed to determine which amino acid residues could be critical for the proposed substrate binding mode. For these modeling experiments, the protein structure was not modified or adjusted.
Mutagenesis, Protein Expression, and Purification-Site-directed mutagenesis of human 15-lipoxygenase (6) was performed using the Muta-Gene phagemid in vitro mutagenesis kit (Bio-Rad). Oligonucleotides containing the appropriate base changes were synthesized on an * This work was supported by National Institutes of Health Grant HL48591 (to E. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Beckman Oligo 1000 DNA synthesizer. Uracil-laden plasmid pSS15LO, used as a mutagenesis template, was prepared as described (16). This plasmid contains a lac promotor to express the 15-lipoxygenase and a bacteriophage f1 origin of replication to generate single-strand DNA. To replace Phe 414 with isoleucine or tryptophan, oligonucleotides with sequence 5Ј-GCTCATTATCTGATCGATAATTCCCATGTC-3Ј and 5Ј-CCAGTGCTCATTATCTGGTCCCAAATTCCCATGTCAGAGAC-3Ј were used respectively (base changes are underlined). The mismatch contained in the first oligonucleotide, T, converted the Phe 414 codon (TTC) to a isoleucine codon (ATC). Another mismatch, A, added a ClaI restriction site to facilitate screening. In the second oligonucleotide, mismatches CC converted the Phe 414 codon to a tryptophan codon (TGG) and introduced an additional AvaII restriction site. To generate Arg 402 3 Leu and Arg 402 3 Lys, primers 5Ј-CCTGGCCAGGACGTT-GATTTCCAGGGTG-3Ј and 5Ј-CCTGGCCTTGACGTTGATTTCCAG-GGTG-3Ј were used. Mismatches converted codons CGG (Arg 402 ) to CTG (Leu) or AAG (Lys). Mismatch G removed a MseI restriction site. After the mutagenesis reaction, transformants were screened with restriction enzymes. The complete coding region of the mutant cDNAs were sequenced to ensure that there were no second-site mutations.
The Bac-to-Bac™ baculovirus expression system (Life Technologies, Inc.) was used for the expression of the wild-type human 15-lipoxygenase as well as the mutant proteins described above. A fragment of cDNA containing the complete coding region of human 15-lipoxygenase was cloned into the donor vector pFastBac1 between NotI and SalI sites, to generate the recombinant donor plasmid for wild-type human 15-lipoxygenase. The recombinant donor plasmids for mutant proteins were obtained by replacing the fragment of wild-type cDNA between the two PpuMI sites with the same fragments of each mutated cDNA. This fragment contains the coding region of human 15-lipoxygenase from codon 117 through the stop codon. The recombinant plasmids were transformed into DH10Bac-competent cells containing the bacmid with a mini-attTn7 target site and the helper plasmid. High molecular weight mini-prep DNAs prepared from selected Escherichia coli clones for each engineered protein were used to transfect the Sf9 insect cells and generate the recombinant baculoviruses.
Sf9 cell cultures (1 liter) at a cell density of 2 ϫ 10 6 cells/ml were infected with each of the recombinant baculoviral stocks (multiplicity of infection of 0.9 plaque-forming units/cell). After 4 days, the cell pellets were collected by centrifugation. All proteins were purified to near homogeneity by high pressure liquid chromatography (17). The iron content of purified enzymes was determined by a colorimetric method using the chromogenic ligand Ferrozine (18).
Lipoxygenase Assay and Kinetics-Lipoxygenase activity was assayed spectrophotometrically (Beckman DU650 spectrometer) by monitoring the increase in absorbance at 234 nm. The standard assay contained 1 ml of 10 mM potassium phosphate buffer (pH 7.0) with 0.2% sodium cholate, 2 M 13-hydroperoxylinoleic acid (13-HPODE), 2 and 30 M linoleate (Biomol) as substrate. The substrate solutions were prepared as described (17). The values of enzyme specific activity were normalized for the iron content of each enzyme.
To estimate the enzyme kinetic parameters of engineered enzymes, assay solutions in the same buffer as above with various substrate concentrations ranging from 5 to 25 uM were used. The reactions were started by addition of 1 g of purified enzyme and performed in quartz cuvettes at room temperature. The linear part of the reaction curves were used to calculate reaction rates using a molar extinction coefficient of 23,000 M cm Ϫ1 for conjugated dienes at 234 nm (19). Values for apparent K m and V max were estimated using the Kinetassist program and normalized with iron content of the enzymes.
HPLC Analysis-For each reaction, 5 g of purified enzyme was added to a substrate solution containing 100 g of the substrate in 1 ml of 10 mM potassium phosphate buffer (pH ϭ 7.0) and incubated at 37°C for 15 min. The reaction mixtures were then reduced with trimethyl phosphite and extracted for HPLC analysis as described previously (20). HPLC analyses of the reaction products formed from different substrates were carried out on a Waters 600 HPLC system equipped with a Waters 996 photo-diode array detector. RP-HPLC was performed on a Beckman C-18 column. The column was developed isocratically with methanol/water/acetic acid (80:20:0.1) at a flow rate of 1.5 ml/min, and the column effluent was monitored continuously at wavelengths between 200 and 300 nm. Chiral phase HPLC of the hydroxy fatty acids was carried out on a Chiralcel OD column (Daicel Chemical Industries) using a solvent system of n-hexane/2-propanol/HAc (100:1.5:0.1, v/v/v) and a flow rate of 1.0 ml/min. Experiments with various amounts of enzyme were performed for the methyl ester substrates. The results showed that the decrease of reactivity toward methyl ester substrates was not caused by the limited solubility of these substrates (data not shown).

RESULTS
Defining the Substrate Binding Cavity-The molecular modeling studies reported here utilize the high resolution (1.4 Å) soybean lipoxygenase-1 structure determined by W. Minor and co-workers (12). We generated the solvent-accessible surface for this structure and confirmed the existence of an internal pocket, similar to the pocket identified previously (Fig. 1A) (7,12,21). We have focused our attention on the inner cavity as the catalytically relevant binding site for arachidonic acid. This inner cavity contains a broad pocket, which is about 20 Å in length and 6 Å in diameter (Fig. 1B). The catalytic iron is positioned just outside of the solvent-accessible cavity, approximately in the middle, with the coordinating ligands, sHis 499 , sHis 504 , and sIle 839 , defining part of the cavity. sHis 690 and sAsn 694 , the other two protein ligands to the iron, do not contact the surface of this cavity. Across from the iron, a narrower pocket, about 3 Å in diameter, extends off the larger pocket at nearly a 90°angle. It is approximately 5 Å from this extension to a narrow neck defined by amino acids sThr 556 and sPhe 557 . Extending further, the pocket widens into a larger cavity. The end of this larger pocket may open to the outer cavity as described by Amzel and co-workers (7), if the side chains of sArg 707 and sVal 354 adopt slightly different positions. The inner cavity is defined by the 30 amino acid residues shown in Table I.
Docking Arachidonic Acid-Docking arachidonic acid into a potential binding site within an enzyme structure is complicated by its conformational flexibility. It has been estimated that arachidonic acid can assume more than 10 7 low energy conformations (22). Our initial attempts to dock arachidonic acid into the inner cavity of soybean lipoxygenase-1 indicated that this substrate could be placed in the cavity in several different conformations, as well as in several orientations. In an effort to clarify which possible binding mode might be more realistic, assumptions concerning the interaction between the enzyme and arachidonic acid had to be made. First, we assumed that the binding site for the methyl end of arachidonic acid would be defined by sThr 556 and sPhe 557 , since the homologous amino acid residues in human 15-lipoxygenase, hIle 417 and hMet 418 , have been shown to effect positional specificity (9). This would allow arachidonic acid to adopt a conformation that positions the LS-hydrogens (as defined in Ref. 23) of C-10 and C-13 near the catalytic iron.
In developing a possible docking mode for arachidonic acid, we postulated that there may be a positively charged residue in the vicinity of the binding site to interact with the carboxylate group of arachidonic acid. Three positively charged amino acid residues, sLys 260 , sArg 707 , and sHis 494 , are possible candidates for this interaction based on the soybean lipoxygenase-1 structure. Another possible interaction between the enzyme and the substrate could be that of an aromatic residue with the cisdouble bonds of the substrate. In considering various docking orientations of arachidonic acid, the existence of possible -electron interactions was noted.
Arachidonic acid was docked into the inner cavity of soybean lipoxygenase-1 as shown in Fig. 1B. Our working model docked the methyl end of arachidonic acid near the narrow neck of the inner cavity defined by sThr 556 and sPhe 557 . The carboxylate group of arachidonic acid was docked near sLys 260 , instead of sArg 707 or sHis 494 , as this orientation of the substrate allowed the reactive 1,4-cis,cis-pentadiene moiety to be very close to the catalytic iron. This docking mode, therefore, proposes that sLys 260 interacts with the carboxylate group of the substrate.
Identification of Amino Acid Residues for Mutagenesis-To extrapolate from our model of arachidonic acid binding to the soybean enzyme, we identified conserved mammalian amino acid residues that could play similar roles in the substrate binding site (Table I and Fig. 1B). We identified sLys 260 as a potential charged interaction with the substrate. Across the FIG. 1. A, ribbon diagram of soybean lipoxygenase-1 structure (12). The NH 2terminal ␤-sheet domain (amino acid residues 1-148) is shown in pink, and the large COOH-terminal domain is colored light blue. The yellow surface is the proposed substrate binding cavity. The orange sphere is the catalytic iron with its coordinating amino acid residues, His 499 , His 504 , His 690 , Asn 694 , and Ile 839 , shown in green. B, stereo diagram of the proposed substrate binding site in soybean lipoxygenase-1 with arachidonic acid, docked manually. Only one of the possible conformations of arachidonic acid is shown. Amino acid residues involved in the ligation of iron and in the mutagenesis experiments are shown. The catalytic iron (orange sphere) is behind the proposed binding site, with its coordinating amino acid residues shown in yellow. Amino acid residues previously suggested to be near the substrate binding site, Thr 556 , Phe 557 , and Ser 747 , are shown in light blue (9,25). The amino acid residues shown in red (electron interaction) and pink (charge interaction) are proposed to be important for fatty acid binding in either soybean lipoxygenase-1 (Lys 260 and Trp 500 ) or human 15-lipoxygenase (Leu 541 and Ile 553 ). The labels in parentheses are the corresponding amino acid residues in human 15-lipoxygenase.

TABLE I
Amino acid residues surrounding the proposed arachidonic acid binding site Sequence comparison of soybean lipoxygenase-1 and human 15-and 12-lipoxygenases. Amino acid residues in the solid boxes, bold for the soybean lipoxygenase-1 and bold italic for human 15-lipoxygenase, are suggested by modeling or mutagenesis in this report to be important for substrate binding. Amino acid residues in the dashed boxes were previously suggested to be near the substrate binding site (9,25). All sequences were accessed from GenBank and compared using the BestFit program in the GCG sequence analysis software package, version 7.3 on a VAX computer.
proposed arachidonic acid binding site from sLys 260 is sLeu 541 . The residue in all mammalian enzymes that corresponds to sLeu 541 is hArg 402 , based on sequence alignments (Figs. 1B and 2). We hypothesized that hArg 402 could be important for interacting with the carboxylate of the acid substrate. With the substrate modeled in this orientation, sTrp 500 could form a -electron interaction with arachidonic acid. The interaction postulated for sTrp 500 and the substrate may be served by hPhe 414 in the mammalian enzymes (Figs. 1B and 2). Another amino acid that was identified from the model was sLeu 546 , which may define the entrance to the narrow pocket occupied by the methyl end of the substrate. The homologous residue in mammalian lipoxygenase is hLeu 407 . It was of interest to investigate the role of this residue in substrate binding, since it is conserved in all of the lipoxygenases. To determine if hArg 402 , hPhe 414 , and hLeu 407 are critical in substrate binding, each was individually replaced by site-directed mutagenesis of human 15-lipoxygenase.
Mutagenesis-In human 15-lipoxygenase, hArg 402 may provide a positive charge to interact with the carboxylate group of the substrate and was, therefore, replaced with leucine and lysine. We postulated that hPhe 414 could be involved in an aromatic interaction with the substrate, and it was replaced with isoleucine and tryptophan. hLeu 407 may be one of the amino acid residues that define the entrance of the narrower pocket of the proposed substrate binding site for the methyl end of the substrate and, therefore, was changed to alanine, isoleucine, and methionine. Preliminary experiments to determine the effects of substituting residues hArg 402 , hPhe 414 , and hLeu 407 were performed using the crude bacteria extracts. All cultures were assessed for protein expression by SDS-polyacrylamide gel electrophoresis, followed by immunoblot detection. To the limits of this technique, the levels of expression were not significantly altered (data not shown). Bacteria expressing the mutated human lipoxygenases were incubated with arachidonic acid. The fatty acid products were extracted and analyzed by RP-HPLC. The activity of both Arg 402 3 Leu and Phe 414 3 Ile were markedly lower than that of the wildtype enzyme in the crude bacteria extracts (data not shown). Introducing a positively charged residue (Arg 402 3 Lys) or an aromatic residue (Phe 414 3 Trp) back into the enzyme led to an apparent increase in enzyme activity. The substitutions of hLeu 407 , however, showed no significant effect on either the total enzyme activity or the positional specificity. Based on these preliminary results, mutant proteins relevant to hArg 402 and hPhe 414 were expressed in the baculovirus/insect cell system and purified to near homogeneity (17). The iron content of the purified mutated protein samples were determined to be similar to that of the wild-type protein.
Steady-state Kinetics-The steady-state kinetic parameters of the purified mutant enzymes were compared with the wildtype enzyme. Substrate concentrations ranging from 5 to 25 M were used. The reactions were started by addition of 1 g of purified enzyme and performed in a quartz cuvette at room temperature. As shown in Table II, all of the mutations altered the values of apparent K m for fatty acid substrates, while V max of the mutant enzymes were similar to that of the wild-type enzyme.
Characterization of hArg 402 Amino Acid Substitutions-Based on the model, we predicted that replacement of hArg 402 with leucine would reduce the binding affinity for fatty acid substrates. The activity of Arg 402 3 Leu is about 5% of the wild-type activity against both arachidonic acid and linoleic acid (Table III). Kinetic characterization of Arg 402 3 Leu under steady-state conditions showed that V max is unaffected, while K m is increased 7-10-fold depending on the substrate (Table   II), consistent with the hypothesis that Arg 402 is energetically involved in binding the substrate by a charge interaction. Chiral phase HPLC analysis of the enzyme reaction products (15-HETE, 12-HETE, and 13-HODE) indicated that they are predominantly in the S-configuration.
Human 15-lipoxygenase has dual positional specificity; oxygenation of arachidonic acid produces two chiral products, 15(S)-hydroxyeicosatetraenoic acid (HETE) and 12(S)-HETE, in a ratio of 10:1. It has previously been suggested that the positional specificity of 15-lipoxygenase is affected by the alignment of a doubly allylic methylene carbon with the catalytic iron of the enzyme (24). To determine the effect of an arginine to leucine substitution at position 402 on the positional specificity, the enzyme reaction products with arachidonic acid as substrate were analyzed by RP-HPLC (Table III). Wild-type human 15-lipoxygenase produced 15-HETE and 12-HETE in a ratio of 10:1 (15-HETE:12-HETE), whereas this ratio was 4:1 for Arg 402 3 Leu. This change in positional specificity implies that hArg 402 is important in positioning the substrate accurately in the binding site relative to the catalytic iron. When Arg 402 is replaced with leucine, the substrate apparently can shift in the pocket positioning the C-10 of arachidonic acid closer to the catalytic iron, leading to a relatively higher yield   15-lipoxygenase in substrate binding The enzymatic activities for the native and engineered proteins were assayed spectrophotometrically by monitoring conjugated formation as measured by an increase in absorbance at 234 nm and normalized for the iron content of each enzyme. The standard assay contained 1 ml of 10 mM potassium phosphate buffer (pH 7.0) with 0.2% sodium cholate, 2 M 13-HPODE, and 30 M fatty acid as substrate. The product ratios were determined using an HPLC assay. For each reaction, 5 g of purified enzyme was added to a substrate solution containing 100 g of the substrate in 1 ml of 10 mM potassium phosphate buffer (pH ϭ 7.0) and incubated at 37°C for 15 min. RP-HPLC was performed on a Beckman C-18 column; the column was developed isocratically with methanol/water/acetic acid (80:20:0.1) at a flow rate of 1.5 ml/min (values Ϯ S.E.; n ϭ 3).
To provide further evidence for the importance of a positive charge in the substrate binding pocket, we constructed the charge-preserving mutation, Arg 402 3 Lys. This amino acid replacement resulted in approximately 2-fold increase in enzyme specific activity when compared to Arg 402 3 Leu, although the activity was 6-and 11-fold lower than that of the wild-type enzyme on arachidonic acid and linoleic acid, respectively (Table III). Importantly, the Arg 402 3 Lys mutation shows the same positional specificity as the wild-type enzyme, which is reflected in the 15-HETE:12-HETE ratio (Table III). pH titration of enzyme specific activity revealed that both wild-type human 15-lipoxygenase and Arg 402 3 Lys mutant have pH optima at 7, while no pH effect on enzyme activity was observed for the Arg 402 3 Leu mutant (data not shown).
The hypothesis that hArg 402 defines a carboxylic acid binding site predicts that the Arg 402 3 Leu variant would no longer have a preference for acid substrate. We tested this prediction by measuring the activity of the Arg 402 3 Leu enzyme on methyl ester substrate. As shown in Table IV, the activity of wild-type enzyme toward a methyl ester substrate is approximately 10% of that toward the corresponding acid substrates, arachidonic acid and linoleic acid. The Arg 402 3 Leu enzyme has comparable activities for both the acid and methyl ester substrates. Importantly, the Arg 402 3 Leu mutant enzyme is as active as the wild-type enzyme against the methyl ester substrates. These results provide further evidence that the positive charge of Arg 402 is important for the binding of fatty acid substrates. Furthermore, it shows that replacing the arginine with leucine did not significantly perturb the overall structure of the enzyme beyond the site of substitution.
Characterization of hPhe 414 Amino Acid Substitutions-In docking arachidonic acid in the soybean lipoxygenase-1 structure, we hypothesized that an aromatic residue could potentially provide an important -electron interaction with a double bond of the substrate. Extrapolating from the model of arachidonic acid docked in soybean lipoxygenase-1, hPhe 414 is the only aromatic amino acid residue within the predicted substrate binding site of human 15-lipoxygenase that could provide such an interaction. To test this hypothesis, hPhe 414 was replaced with both isoleucine and tryptophan. The Phe 414 3 Ile mutation markedly decreased enzyme activity, while the Phe 414 3 Trp mutant behaved essentially like the wild-type enzyme (Table III).
Analysis of the enzyme reaction products reveals that Phe 414 3 Ile has an increase in 12-lipoxygenase activity relative to 15-lipoxygenase activity (Table III). The loss of the aromatic side chain allows the reactive 1,4-cis,cis-pentadiene moiety of the substrate to shift within the pocket thus producing the altered substrate specificity. The replacement of hPhe 414 with tryptophan resulted in a large enhancement in the enzyme's 15-lipoxygenase activity relative to its 12-lipoxygenase activity (15-HETE:12-HETE ϭ 25:1). Chiral phase HPLC analysis showed there was no change in product configuration (data not shown). Comparison of the kinetic parameters for the enzymes where hPhe 414 has been replaced with isoleucine or tryptophan showed only small changes (Table II).
Using our model of arachidonic acid binding, we could not predict whether the ⌬ 8 -, ⌬ 11 -, or ⌬ 14 -double bond of arachidonic acid would interact with hPhe 414 . To identify which, if any, of these three double bonds interacts with hPhe 414 , we examined the activity of the enzymes against two other unsaturated fatty acid substrates, which differ in their double bond character. As mentioned above, Phe 414 3 Ile is less active against arachidonic acid, and the preference for 15-lipoxygenation is lost (15-HETE:12-HETE ϭ 1.6:1) (Table V). Phe 414 3 Trp is more active than the wild-type enzyme, with very strong preference for 15-lipoxygenase activity (15-HETE:12-HETE ϭ 25:1). Fatty acid 11,14-eicosadienoic acid contains only the ⌬ 11 -and ⌬ 14double bonds. This substrate is, therefore, only available for 15-lipoxygenase activity, producing the product 15-HEDE. Replacement of hPhe 414 , with the non-aromatic residue isoleucine again reduced the enzyme activity on this substrate, whereas replacement with tryptophan had little effect on the activity. These results suggest that ⌬ 5 -and ⌬ 8 -double bond of arachidonic acid is not interacting with the aromatic side chain of hPhe 414 .
Next, we investigated the activity of these enzymes against the substrate 5,8,11-eicosatrienoic acid, which does not have ⌬ 14 -double bond and should only produce 12-HETE. Again Phe 414 3 Ile showed 2-fold lower enzyme activity against this substrate, whereas Phe 414 3 Trp exhibited virtually unaltered activity compared with the wild-type enzyme. It is, therefore, most likely that the aromatic side chain of hPhe 414 interacts with the ⌬ 11 -double bond of arachidonic acid. Unfortunately, a substrate lacking the ⌬ 11 -double bond (i.e. 5,8,14-eicosatrienoic acid) is not accessible to either 12-or 15-lipoxygenation and, therefore, cannot be tested.

DISCUSSION
Two internal cavities in proximity to the catalytic iron were apparent in the three-dimensional structures of soybean lipoxygenase-1 (7,21). One of these cavities was initially proposed by Amzel and co-workers to be the arachidonic acid binding site. A similar internal pocket is present in the high resolution soybean lipoxygenase-1 structure used in this study (12). The solvent-accessible surface used in describing this cavity in soybean lipoxygenase-1 was generated in the presence of the polar hydrogen atoms only. If this is the substrate binding site, we can assume that some of the amino acid residues lining the pocket would be repositioned and that the pocket would be larger when occupied by substrate.
Experimental data that support the identification of this inner cavity as the arachidonic acid binding site in soybean lipoxygenase-1 are provided by the results reported here and from previous experiments with 15-lipoxygenase and 12-lipoxygenases (9 -11). Wild-type human 15-lipoxygenase has dual positional specificity, producing both 15-oxygenation and 12oxygenation products in a ratio of approximately 10:1, from arachidonic acid (17). Earlier experiments showed that the substitution of both hIle 417 and hMet 418 in human 15-lipoxygenase with valine altered the positional specificity of the enzyme to favor 12-lipoxygenation. It was, therefore, suggested that these residues could define a portion of the substrate binding pocket of the enzyme (9,16). The homologues of the human 15-lipoxygenase residues, hIle 417 and hMet 418 , are sThr 556 and sPhe 557 in the soybean enzyme (Table I). These residues define a narrow neck of the inner cavity (Fig. 1B). In addition, we have previously shown that hMet 590 of human 15-lipoxygenase is specifically oxygenated by the enzyme product, 13-HPODE, and hence this residue is also likely to be close to the arachidonic acid binding site (25). The corresponding soybean residue of hMet 590 is sSer 747 . Examination of the soybean structure indicates that all of these residues, sThr 556 , sPhe 557 , and sSer 747 , line the inner cavity, providing support that this cavity is the arachidonic acid binding site in soybean lipoxygenase-1.
A central feature of our current working model of arachidonic acid binding is a positively charged residue at one end of the cavity (hArg 402 or sLys 260 ), that interacts with the carboxylate group of the substrate (Fig. 2). Prigge and co-workers also suggested the possibility that sHis 494 could interact with the carboxylate group of the substrate; however, they were able to rule this out using the results of previously published sitedirected mutagenesis experiments (21). We propose that hArg 402 may interact with arachidonic acid. In agreement with our proposal, enzymatic activity and steady-state kinetics data suggested that replacement of hArg 402 dramatically affected substrate binding. These results highlight the importance of a positively charged residue in this region of the binding site. Interestingly, this type of interaction has been widely observed in other proteins that bind fatty acids, such as fatty acid transport proteins (26). The change in positional specificity implies that hArg 402 is also important in positioning the substrate accurately in the binding site relative to the catalytic iron. When hArg 402 is replaced with leucine, there is no longer a charge interaction; thus, the carboxyl terminus of the substrate is free, enabling the substrate to shift in the pocket. This shift positions the C-10 of arachidonic acid closer to the catalytic iron, leading to a relatively higher yield of 12-HETE. The study with methyl ester substrates provides further evidence that the positive charge of hArg 402 is important for the binding of fatty acid substrates.
Characterization of Arg 402 3 Lys provides further support for the hypothesis that a positively charged residue is critical for optimal 15-lipoxygenase activity. The fact that Arg 402 3 Lys has reduced activity compared with wild-type enzyme may be attributed to the various structural features that distinguish arginine and lysine. The subtle difference in the electrostatics of arginine and lysine could also contribute to the decrease in enzyme activity. Although both side chains carry the same formal charge, the distribution of charge in an arginine To identify which double bond of arachidonic acid interacts with the side chain of Phe 414 , the enzyme specific activity and positional specificity were determined using fatty acid substrate containing cis-double bonds at different positions. The enzymatic activity for the native and engineered proteins were assayed spectrophotometrically by monitoring conjugated diene formation as measured by an increase in absorbance at 234 nm and normalized with iron content of each enzyme. The standard assay contained 1 ml of 10 mM potassium phosphate buffer (pH 7.0) with 0.2% sodium cholate, 2 mM 13-HPODE, and 30 M fatty acid as substrate. The product ratios were determined using HPLC assay. For each reaction, 5 g of purified enzyme was added to a substrate solution containing 100 g of the substrate in 1 ml of 10 mM potassium phosphate buffer (pH ϭ 7.0) and incubated at 37°C for 15 min. RP-HPLC was performed on a Beckman C-18 column; the column was developed isocratically with methanol/water/ acetic acid (80:20:0.1) at a flow rate of 1.5 ml/min. The position of hydrogen abstraction and oxygenation are indicated for arachidonic acid. 1, hydrogen abstraction; 2, oxygenation.
FIG. 2. Schematic diagram of the proposed substrate binding determinants of human 15-lipoxygenase. This schematic approximates the shape of the solvent-accessible cavity defined in soybean lipoxygenase-1 (Fig. 1B). Previous mutagenesis experiments suggested that the methyl end of the substrate is near Ile 417 and Met 418 (9). This report provides evidence for the interactions indicated for Arg 402 and Phe 414 with the substrate. The dashed arrows represent the LS-hydrogens. The catalytic iron atom is behind the pocket. side chain is more delocalized, which may provide a more favorable charge interaction. Taken together with earlier work, our current model places binding determinants at both the methyl and carboxyl termini of the substrate. The absence of a charge interaction would explain the lower activity of the wildtype enzyme on esterified fatty acid.
The studies on Phe 414 3 Ile and Phe 414 3 Trp mutant proteins suggest that an aromatic residue near the catalytic iron is also important for enzyme activity and the positioning of the substrate in the correct catalytic register (Fig. 2). In soybean enzyme sTrp 500 , which is analogous to hPhe 414 , was modeled to interact with the docked substrate. It is interesting to note that soybean lipoxygenase-1 produces only the 15-lipoxygenase product. Introduction of tryptophan into the substrate binding site of human enzyme creates a stronger preference for 15-lipoxygenation and higher enzyme specific activity, perhaps because of the particular -electron characteristics of this amino acid residue. Interaction of a tryptophan residue with a fatty acid substrate is present in the structure of cellular retinol-binding protein in complex with all-trans-retinol (27). In this case, the aromatic residue is close to the isoprene tail of retinol. The direct interaction between tryptophan and retinol has been confirmed by site-directed mutagenesis and 19 F nuclear magnetic resonance spectroscopy with protein containing 19 F-labeled tryptophan (28,29).
It is most likely that the aromatic side chain of hPhe 414 interacts with the ⌬ 11 -double bond of arachidonic acid. A recent EPR study of the purple form of soybean lipoxygenase-1 suggested that a reaction intermediate contains an allyl radical delocalized over C-9 through C-11 of linoleic acid (30). A -electron interaction between the substrate and an aromatic amino acid side chain of the enzyme would facilitate the formation of this radical by delocalizing the -electrons. The ⌬ 9 -double bond of linoleic acid and the ⌬ 11 -double bond of arachidonic acid are both 6 carbons away from the methyl terminus. Our finding that the ⌬ 11 -double bond of arachidonic acid may participate in ainteraction with the enzyme provides support for the enzyme mechanism proposed by Nelson et al. (30).
Although the catalytic mechanism of lipoxygenases has yet to be fully elucidated, it has been suggested that the ratelimiting step of the reaction is the abstraction of hydrogen, with the help of the oxidized iron center, to form a bis-allelic methylene (31,32). Molecular oxygen is then inserted stereo-specifically 2 carbons away from the site of hydrogen abstraction, resulting in a hydroperoxy-fatty acid containing a conjugated diene. Studies with stereo-specifically labeled substrates have revealed a large primary kinetic isotope effect and loss of label only when the label is in the pro-S conformation (31,(33)(34)(35). We have docked arachidonic acid into the proposed substrate binding site in an orientation so that both the 10-Ls-and 13-Ls-hydrogens to be positioned close to the catalytic iron and points them directly toward the iron, thereby facilitating hydrogen abstraction.
In summary, we have developed a model for substrate binding in the lipoxygenases and used this model to identify two more amino acid residues in human 15-lipoxygenase that are important substrate binding determinants, hArg 402 and hPhe 414 (Fig. 2). The model for arachidonic acid binding uses the soybean lipoxygenase-1 structure, whereas the proposed substrate-enzyme interactions were tested and proven using the human enzyme. The structural similarity between these two distantly related enzymes is, therefore, confirmed by these experiments. These results also demonstrate the importance of both a positively charged amino acid residue and an aromatic amino acid residue for fatty acid binding (Fig. 2). Thus, what has been shown previously to be characteristic of non-enzymatic fatty acid-binding proteins is also shown to be an important binding determinant for an enzyme that metabolizes arachidonic acid.