Leukotriene A 4 hydrolase: Insights to the molecular evolution by homology modeling and mutational analysis of enzyme from Saccharomyces cerevisiae

Fredrik Tholander‡, Filippa Kull‡, Eva Ohlson‡, Jawed Shafqat‡, Marjolein M.G.M. Thunnissen§ and Jesper Z. Haeggström‡* ‡ Department of Medical Biochemistry and Biophysics, Divisions of Chemistry 1 & 2, Karolinska Institutet, S-171 77 Stockholm, Sweden. § Department of Molecular Biophysics, Kemicentrum, Lund University, S-221 00 Lund, Sweden. Running title: Molecular evolution of LTA4 hydrolase Adress correspondence to: Jesper Z. Haeggström Department of Medical Biochemistry and Biophysics, Division of Chemistry 2, Karolinska Institutet, S-171 77 Stockholm, Sweden Telephone: +46-8-5248 7612, Fax: +46-8-736 0439, E-mail: jesper.haeggstrom@mbb.ki.se


Mammalian leukotriene A 4 (LTA 4 ) hydrolase
is a bifunctional zinc metalloenzyme possessing an Arg/Ala aminopeptidase and an epoxide hydrolase activity, which converts LTA 4 into the chemoattractant LTB 4 . We have previously cloned an LTA 4 hydrolase from S. cerevisiae with a primitive epoxide hydrolase activity and a Leu aminopeptidase activity, which is stimulated by LTA 4 . Here we used a modeled structure of S. cerevisiae LTA 4 hydrolase, mutational analysis, and binding studies to show that Glu-316 and Arg-627 are critical for catalysis, allowing us to a propose a mechanism for the epoxide hydrolase activity. Guided by the structure, we engineered S. cerevisiae LTA 4 hydrolase to attain catalytic properties resembling those of human LTA 4 hydrolase. Thus, six consecutive point mutations gradually introduced a novel Arg aminopeptidase activity and caused the specific Ala and Pro aminopeptidase activities to increase 24 and 63 times, respectively. In contrast to the wild type enzyme, the hexuple mutant was inhibited by LTA 4 for all tested substrates and to the same extent as for the human enzyme. In addition, these mutations improved binding of LTA 4 and increased the relative formation of LTB 4 , whereas the turnover of this substrate was only weakly affected. Our results suggest that during evolution, the active site of an ancestral eukaryotic zinc aminopeptidase has been reshaped to accommodate lipid substrates, while using already existing catalytic residues for a novel, gradually evolving, epoxide hydrolase activity. Moreover, the unique ability to catalyze LTB 4 synthesis appears to be the result of multiple and subtle structural rearrangements at the catalytic center, rather than a limited set of specific amino acid substitutions.
LTA4H is homologous to other aminopeptidases in a variety of species, ranging from mammals to bacteria, in particular those belonging to the M1 family (4). Members of this family share a common Zn binding signature, HEXXH, in which the His residues are the primary Zn binding ligands and Glu the general base catalyst required for peptidolysis. In LTA4H, as well as most other M1 aminopeptidases, the third Zn binding ligand is a Glu located 18 residues downstream from the HEXXH motif, thus defining a HEXXH-(X) 18 -E motif. However, the evolutionary relationship between the M1 aminopeptidases and LTA4H is uncertain. Thus, aminopeptidase 1 from Caenorhabditis elegans is 45% identical (64% similar) at the amino acid level to mammalian LTA4H and exhibits an Arg aminopeptidase activity (5). Despite this high level of sequence identity, the C. elegans enzyme fails to hydrolyze LTA 4 into LTB 4 and no other functional link to LTA4H has been reported. In fact, enzymes carrying a distinct LTA 4 hydrolase activity have only been detected among vertebrates, including birds, frogs and fish (6)(7)(8).
Recently, the crystal structure of humLTA4H was solved (11). The protein is folded into three domains with a deep cavity in between, which harbors the active center. The architecture of the active site was established and a model for binding of LTA 4 was proposed, involving a narrow, L-shaped, hydrophobic side pocket. The structure clearly indicates that the conservation of active-site residues are not confined to residues of the HEXXH-(X) 18 -E motif but extends to residues lining the entire LTA 4 binding pocket. Notably, the active site of LTA4H is composed of residues from all three domains.
Considering the sequence identity between human and yeast LTA4H as well as functional similarities and differences, we chose scLTA4H as a suitable target to study the molecular mechanisms underlying the evolution of mammalian LTA4H:s. To this end, a homology model of the scLTA4H enzyme, based on the crystal structure of humLTA4H, has been constructed and candidate catalytic residues have been subjected to mutational analysis. Using rational protein engineering, scLTA4H was converted to an enzyme with functional properties more similar to those of mammalian LTA4H. Together, our data provide novel insights to both the catalytic mechanisms and the molecular evolution of LTA4H.  (12). From this multiple alignment a pair-wise alignment between scLTA4H and humLTA4H was derived (Fig.  S1, Supplementary Material on line). The structure of humLTA4H (PDB ID 1HS6) was used as the template. The model was constructed using the WHATIF program, which was also used for initial geometry refinement (13). Manual adjustment of the model including rebuilding of loops with insertions and deletions was done with XtalView (14). Finally, the model was energy minimized using the YAMBER force field implementation of the Yasara software (15). After energy minimization the rmsd between the scLTA4H model and the template was 0.8 Å for 567 carbon-α atoms. Concerning model quality the final model fulfilled the requirements of the WHATCHECK program, where applicable (16).
Expression and purification − Expression and purification of recombinant proteins were performed as previously described (10). Briefly, scLTA4H was expressed as a (His) 6tagged fusion protein and purified by affinity chromatography on a Ni-NTA column followed by chromatofocusing and anion exchange chromatography. Protein concentrations were determined according to the method of Bradford, using the Bio-Rad protein-assay reagent and bovine serum albumin as standard. To determine the purity of the protein samples, SDS-polyacrylamide gel electrophoresis (PAGE) was performed on a Phast system (Amersham Pharmacia Biotech) using 10-15% gradient gels. Bands of protein were visualized by staining with Coomassie Brilliant Blue.
Enzyme kinetic assays − For determination of apparent kinetic constants (k cat and K m ) for the aminopeptidase activity at least three sets of initial velocity measurements were performed. Aliquots (50 μl) of equilibrated solutions containing 1 μg protein, 500 mM KCl, 250 mM Tris-HCl, pH 7.5, were added to the wells of a microtiter plate. The reactions were initiated by the addition of 200 μl of aminoacyl-p-NA of various concentrations (12, 10, 5, 2.5, 1.25, 0.625, 0.3125, and 0.1525 mM) yielding a final reaction volume of 250 μl containing 1 μg protein, 100 mM KCl, 50 mM Tris-HCl, pH 7.5, and 9.6 -0.125 mM Ala-p-NA. The formation of p-NA was subsequently monitored at ambient temperature for 15 min in a MCC/340 multiscan spectrophotometer as the increase in absorbance at 405 nm. Spontaneous hydrolysis of the various aminoacyl-p-NA was corrected for by subtracting the absorbance of incubations without enzyme. A molar response factor of 8065 M -1 for p-NA was derived from a set of standard incubations of known concentration.
For determination of apparent kinetic constants (k cat and K m ) for the epoxide hydrolase activity at least three sets of initial velocity measurements were performed. Reactions were carried out in 100 μl aliquots of Tris-HCl, pH 7.5, containing 25 μg of protein. The reactions were initiated by the addition of various amounts of LTA 4 (giving final concentrations ranging from 2 μM -150 μM) and quenched after 30 s by the addition of 200 μl of methanol containing 1.5 μM prostaglandin (PG)B 2 as internal standard. The added volume of LTA 4 never exceeded 2 μl. The samples were subsequently diluted with 1 ml water and acidified to pH ≈ 3.5 with 10 µl of acetic acid (10 %) and metabolites extracted on solid phase Chromabond C 18 columns. Samples were loaded onto the columns, washed with 1 ml 25% methanol, and eluted with 250 μl 100% methanol. The samples were diluted with 250 μl water prior to HPLC analysis. Metabolites of LTA 4 were separated by isocratic reverse-phase HPLC on a Waters Nova-Pak C 18 column eluted with a mixture of methanol/acetonitrile/water/acetic acid (30:31:39:0.01 by vol.) at a flow rate of 1.2 ml/min. The UV detector was set at 270 nm and metabolites were quantified based on peak area measurements and the known extinction coefficients for the internal standard PGB 2 (30 000 M -1 x cm -1 ), LTB 4 (50 000 M -1 x cm -1 ), 5S,6S-DHETE (40 000 M -1 x cm -1 ) and Δ 6trans-Δ 8 -cis-LTB 4 (50 000 M -1 x cm -1 ).
For determination of the specific aminopeptidase and epoxide hydrolase activities the corresponding standard reactions were carried out with 1 mM aminoacyl-p-NA and 40 μM LTA 4 , respectively.
Determination of the effects of LTA 4 on the peptidase activity of LTA4H − As a test for a stimulatory or inhibitory effect, LTA 4 was added at a single concentration (50 µM). To assess dose-response relationships, the specific peptidase activity was assayed in enzyme preparations pre-treated with various amounts of LTA 4 (typically 0.5 -150 μM) for 20 to 30 min, prior to the activity measurements.

SPR Biosensor Analysis of bestatin binding −
To assess the substrate affinity of the inactive mutants [E316Q] and [E316A]scLTA4H an interaction assay using an SPR biosensor, Biacore 3000 (Biacore, Uppsala, Sweden) was performed using the peptide analogue bestatin as the analyte. As a control experiment the same binding affinity of wild-type enzyme was determined. All binding experiments were performed at ambient temperature in a 10 mM Tris pH 7.4 running buffer supplemented with 0.005 % SP20.
Standard amine coupling (BIAapplication handbook;Biacore) was used for covalent protein immobilization onto a CM5 sensor chip. The proteins were kept at a concentration of 20 μg/ml in 10 mM sodium acetate buffer pH 5.5 and injected onto the EDC/NHS (N-ethyl-N'-(3-dimethyl aminopropyl)-carbodiimide hydrochloride /Nhydroxysuccinimide) activated sensor surface at a flow-rate of 5 μl/min. To ensure that similar amounts of protein were immobilized in the three lanes the change in refractive index was monitored and immobilization stopped when the relative response reached approximately 11000 response units. To measure the real-time binding various amounts of bestatin (62.5 125, 250 μM) in running buffer were injected, at a flow-rate of 20 μl/min for 2 minutes, onto the three surfaces containing immobilized protein. The SPR response caused by analyte binding and subsequent self-dissociation was monitored. To correct for background responses several injections containing buffer alone, were performed. For each concentration of bestatin the experiment was run in triplicate. To regenerate the chip surface bound bestatin was removed by washing with 20 μl of 1M ethanolamine.
Competition assay for analysis of substrate binding − To further assess the substrate affinity of the inactive mutant E316A 125μl of a solution containing 0.48 μM wild-type enzyme, 120 μM mutant enzyme, 200 mM KCl and 100 mM Tris pH 7.5 was added to a well of a micro-titre plate. To a reference well the same amount of a mixture without mutant enzyme, but otherwise identical, was added. To initiate the reactions 125 μl of 120 μM Leu-p-NA was added. The resulting timecourse of p-NA formation was subsequently monitored for 3 h. Obtained progress-curves were analyzed by the software package Dynafit (17) allowing the obtained data to be fitted to the following model: E + S <−> ES, ES −> E + P, M + S <−> MP, were M denotes inactive mutant and MS denotes the mutantsubstrate complex. Thus, the model states that both M and E bind the substrate but only E is capable of converting S into P. To evaluate the data the obtained progress-curves were compared to a simulated curve assuming a substrate affinity of the mutant equal to that of wild-type enzyme.
Model fitting and statistical analysis − Appropriate kinetic models were fitted against the experimental data from the enzyme kinetic experiments and the dose-response measurements, using non-linear regression.
The models and the derived parameters were evaluated with the statistical tools provided by the Excel macro SolverStat as described by Comuzzi et al. (18). Reported standard errors are the ones given by the program. Activity measurements were analyzed using ANOVA and pair-wise comparisons performed according to the Tukey-Kramer procedure. All reported differences are significant at the 5% level or less, unless otherwise stated.

RESULTS
Homology modeling − A model of scLTA4H was constructed using humLTA4H as a structural template. Based on a multiple alignment generated with ClustalW all residues from 52 to 671 of the yeast enzyme were modeled using the WHATIF software. Inspection of the final model and the underlying alignment between human and scLTA4H, which served as the base for the modeling, indicates that all insertions and deletions occur in loop regions of humLTA4H (Figs. S1 and S2, Supplementary Material on line). This indicates a reliable alignment and is a prerequisite for homology modeling of good quality. The fact that the structural framework of the active site of humLTA4H is dictated by the fold and of residues of the catalytic domain, which exhibits higher sequence identity than the rest of the protein, further supports the predictive value of our model. Thus, the model indicates that scLTA4H possesses an active site similar to that of humLTA4H, which contains the established catalytic residues (presented as yeast residue/human residue) His-340/His-295, Glu-341/Glu-296, His-344/His-299, Glu-363/Glu-318, and Tyr-429/Tyr-383 as well as a hydrophobic channel possibly binding LTA 4 . (Fig. 1). In addition, three putative catalytic residues Glu-316/Glu-271, Asp-421/Asp-375, and Arg-627/Arg-563 were found at the active site and functionally analyzed (see below).
Four residues directly lining the binding pocket of the human enzyme differ in yeast, namely: Glu-186/Gln-136, Gln-412/Val-367, Thr-418/Ile-372 and Phe-424/Tyr-378. The overall shape of the hydrophobic pocket appears slightly wider and shorter as compared to humLTA4H, in part due to the exchange of Val-367 -> Gln-412 (in yeast) at the bottom of the pocket (cf. Fig. 7). Furthermore, the model structure indicates that Asp-422 (Val-376 in humLTA4H) forms a novel interaction with polar residues of the C-terminal domain of the protein, which potentially could affect the shape of the active site. Also Asn-417, a loop residue close to the active site, which is absent in all mammalian LTA4H could have a potential effect on the chemistry of the active site. Glu-186, Gln-412, Asn-417, Thr-418, Asp-422, and Phe-424 were all subjected to mutational replacements, individually or in groups, in an effort to engineer scLTA4H into en enzyme with catalytic properties more similar to humLTA4H (see below).
Mutagenesis, expression and purification − In addition to wild-type scLTA4H, 17 mutants of scLTA4H were successfully constructed, expressed and purified to apparent homogeneity.
The final yield was approximately 0.5-5 mg of protein per liter cell culture and the purity was at least 90% as judged from SDS-PAGE. The analyzed mutants are presented in Table S1 (Supplementary Material on line) along with a short description of the rational behind the selection of mutations.
Effects of mutations on specific peptidase activities − The specific peptidase activity and effects of LTA 4 on wild type and all 17 mutated proteins were analyzed using different derivatives of aminoacyl-p-NA, i.e., Leu-, Ala-, Arg-, Lys-, Met-, Gly-and Pro-p-NA (Fig. 2). Mutations probing for catalytic residues identified Glu-316, conserved throughout the M1 family of aminopeptidases, and Arg-627, conserved in mammalian LTA4H:s, as critical for the peptidase activity of scLTA4H (Fig. 2). Thus, all enzyme variants of scLTA4H mutated at Glu-316 were virtually devoid of peptidase activity and mutants at position Arg-627 exhibited drastically reduced peptidase activities, even under stringent conditions with higher concentrations of enzyme and substrate (data not shown). The peptidase activity of neither of these mutants could be restored by LTA 4 treatment. However, for E316D, a very small but significant residual activity could be detected. Thus, to restore the normal activity the enzyme concentration was increased approximately 300 times as compared to wildtype enzyme.
Mutations aimed at converting scLTA4H into an enzyme with properties similar to humLTA4H had variable effects on the peptidase activity. Thus, F424Y shows a strong increase in its Leu-and Met-p-NA hydrolyzing activity, whereas the combined mutants 4M (E186Q/Q412V/T418I/F424Y), 5M (E186Q/Q412V/T418I/D422V/F424Y) and 6M (E186Q/Q412V/N417del/T418I/D422V/F424 Y) just like humLTA4H, also hydrolyze Ala-p-NA efficiently. Of note, with increasing number of mutagenetic changes, the combined mutants gradually developed a strong Arg-p-NA hydrolyzing capacity, an enzyme activity that is prominent in humLTA4H but virtually absent in wild type scLTA4H. Also in terms of k cat /K m the 6M mutant displayed highly increased values for all its activities and the preferred substrates were Arg-, Leu-and Alap-NA followed by Pro-, Lys-and Met-p-NA, just as for humLTA4H (Fig. S3, Supplementary Material on-line). For Ala-, Leu-and Arg-p-NA hydrolysis the values even exceeded those of humLTA4H.
The single mutation E186A reduced the Ala aminopeptidase activity but strongly increased the ability of scLTA4H to hydrolyze Pro-p-NA. For the remaining mutants E186Q, Q412V, N417del, N417del/D422V, D421N and 3M (data not shown), no major changes of the catalytic properties were observed. The peptidase activities against Gly-p-NA as a substrate were undetectable for all tested mutants as well as for wild-type enzyme, and for Lys-p-NA the observed activities parallel those of Arg-p-NA (data not shown).
Effects of mutations on the stimulatory effects of LTA 4 on the peptidase activities of scLTA4H − Depending on the substrate, mutants were either stimulated, inhibited or unaffected by treatment with LTA 4 (Fig. 2). Thus, with few exceptions, the Met-p-NA, Arg-p-NA or Ala-p-NA hydrolyzing activities were inactivated when treated with LTA 4 , whereas the Leu aminopeptidase activities were stimulated. Considering specific mutants, pre-treatment of 4M or 5M with LTA 4 inhibited the peptidase activities for all substrates, except Leu-p-NA. However, for 5M the stimulatory effect of LTA 4 on the specific Leu aminopeptidase activity was very low, corresponding to about 25%, as compared to approx. 600% in wild type scLTA4H. A similar reduction of the LTA 4 stimulatory effect was also observed for the D422V mutant. Interestingly, for 6M, the mutant most closely resembling humLTA4H, the peptidase activities against all tested substrates were inhibited upon LTA 4 treatment. In addition, the degree of inhibition for each substrate paralleled well the inhibitory profile of humLTA4H (Fig. 3).
Considering apparent kinetic constants for mutants exhibiting LTA 4 activation or inhibition of the peptidase activity, the LTA 4 treatment mainly affects the apparent k cat , whereas the K m values are only affected to a limited extent (Fig. S4 Determination of apparent K i for LTA 4 inhibition of the peptidase activity − The inhibitory effect of LTA 4 on the Ala and Arg aminopeptidase activities of 4M and 5M, as well as on the Ala, Leu and Arg aminopeptidase activities of 6M and humLTA4H were assayed and values of K i calculated ( Table 1). The data indicate that the mode of inhibition is substrate dependent and gradually changes towards competitive as the mutants become more similar to humLTA4H. Furthermore, for the most extensive mutant, 6M, the observed degree of inhibition is in level with the inhibition of humLTA4H.
Effects of mutations on the epoxide hydrolase activities − The specific epoxide hydrolase activities of all 17 mutants, as well as wildtype enzyme, were assayed. Enzyme variants of scLTA4H mutated at Glu-316 or Arg-627 exhibited very low or non-detectable levels of 5S,6S-DHETE production and were classified as inactive (data not shown). For mutants displaying activity, the typical changes were a moderate to significant decrease in both 5S,6S-DHETE as well as LTB 4 production (Fig. 4). An exception is the single mutant D421N, which actually displayed an increased epoxide hydrolase activity, due to increased formation of 5S, 6S-DHETE under saturating conditions (see further below). Notably, all combined mutants displayed reduced rates of 5S,6S-DHETE production. One mutant, viz. 6M, exhibited a decrease in 5S, 6S-DHETE formation with an intact, or even slightly increased, ability to produce LTB 4 . Similar activities and product profiles were obtained under stringent conditions with higher enzyme and substrate concentrations (data not shown).
Determination of apparent turnover numbers (k cat ) for the epoxide hydrolase activity − Mutants exhibiting a significantly changed specific epoxide hydrolase activity were analyzed under conditions with substrate saturation. For the three enzymatic epoxide hydrolase activities of scLTA4H only the activity generating 5S,6S-DHETE was amenable to kinetic experiments; the other two activities were too low. For assayed mutants, significant changes were observed for Q412V, D421N and D422V which all exhibited between 3-and 4-fold increases in their reaction rates at substrate saturation (data not shown). For 4M, F424Y and N417del slight decreases were observed (data not shown), similarly to what was observed for their specific activities (Fig. 4).

Effects of mutations on the relative LTB 4
formation − If the specific epoxide hydrolase activities are assessed as the sum of 5S,6S-DHETE and LTB 4 production, almost all significant changes were reduced activities (Fig. 4). However, if one considers the relative amounts of the two products, significant differences between mutants are observed. The most pronounced effect was seen for the hexuple mutant, 6M, which displayed an LTB 4 to 5S,6S-DHETE ratio of 0.4. For N417del/D422V and 5M, the corresponding values are 0.1 and 0.2 and for wild-type scLTA4H the ratio is 0.08 (Fig. 4).
Effects of mutations of Glu-316 and Arg-627 on substrate affinity − To assess the role of Glu-316 in substrate tethering of peptide αamino groups, an interaction assay using an SPR sensor chip was performed with wildtype enzyme along with the two inactive mutants E316A and E316Q. The general amino-peptidase inhibitor bestatin, possessing a free amino group corresponding to the αamino group of peptides, was used to mimic this specific enzyme-substrate interaction. The time-course data for the binding responses did not allow proper determination of dissociation and association rate constants. However, in qualitative terms the data show that binding of bestatin is much weaker, or almost absent, in E316Q and E316A compared to wild-type enzyme (Fig. 5).
To directly assess the change in aminopeptidase substrate affinity upon mutation of Glu-316, a progress curve analysis of the formation of p-NA was performed. In this experiment the inactive mutant E316A was present in high concentration in the reaction mixture together with wild-type enzyme and substrate. Thus, the inactive mutant competes for the substrate leading to effects on the observed progress of product formation, which is related to the substrate affinity and turnover of the mutant. Comparison of the experimentally observed progress curve with a simulated progress curve corresponding to a preserved substrate affinity of the mutant, demonstrates that mutation of Glu-316 leads to drastically reduced substrate binding (Fig. 5).
The role of Arg-627 as a carboxylate recognition site in substrate binding was evaluated by determining the effects of the mutations R627A and R627K on the K i of bestatin and a hydroxamic acid inhibitor, each containing carboxylate moieties resembling the carboxy groups of peptide substrates and LTA 4 , respectively. Aliquots of mutated enzymes (1 µg in 250 µl 50 mM Tris-HCl, pH 7.5 containing 100 mM KCl) were incubated with 1 mM Ala-p-NA at increasing concentrations of bestatin and hydroxamic acid, typically 1, 2, 5, 10, 20 μM. For bestatin the IC 50 increased from 3.6±0.3 µM of wildtype enzyme to 24±3 and 79±5 µM of R627K and R627A, respectively. For the hydroxamic acid inhibitor the IC 50 changed from 1.5±0.1 µM of wild-type enzyme to 0.17±0.02 and 0.9±0.1 µM for R627A and R627K, respectively. We also modeled the tripeptide substrate Arg-Ser-Arg into the active site of scLTA4H and found that its C-terminal carboxyl group is well positioned for interaction with Arg-627 ( Fig  S6,  Supplementary Material on line). Similarily, the carboxy moiety of LTA 4 also binds to this residue (Fig.1).

DISCUSSION
Mammalian LTA4H is a bifunctional enzyme with an aminopeptidase and epoxide hydrolase activity, which generates the classical chemoattractant LTB 4 . Interestingly, the two active sites share several catalytic residues and yet provide unique catalytic properties and substrate specificities. The bifunctional LTA4H from S. cerevisiae is the first example of an isoenzyme from an invertebrate organism and was chosen for studies of the molecular evolution of the enzyme's two activities.
S. cerevisiae LTA4H possesses a more spacious LTA 4  Glu-316 is essential for anchoring of the Nterminus of peptide substrates − Glu-316 of scLTA4H belongs to a GXMEN motif, which is conserved among M1 aminopeptidases and believed to function as an N-terminal recognition site for peptide substrates. Mutation of Glu-316 to either a Gln or an Ala residue rendered the enzyme virtually unable to turnover any peptide substrate. We assessed the binding capacity of the inactive mutants for the substrate mimetic bestatin by the SPR biosensor technique (Biacore), and for E316A we also performed a progress curve analysis for the hydrolysis of substrate (Fig. 5). Each method demonstrates that Glu-316 is indeed crucial for substrate binding. Hence, we propose that the role of Glu-316 is to anchor peptide substrates to the active site via binding to their free α-amino group (Fig. S6, Supplementary Material on-line).
Interestingly, catalysis appears to dependent on the precise positioning of the substrate. Thus, the mutant E316D, which has a preserved negative charge possesses some, however drastically reduced, aminopeptidase activity (data not shown). Even though the main role of this residue appears to be substrate anchoring an indirect role in catalysis cannot be excluded. On the other hand, the generally strong effects on catalysis caused by mutations of this residue probably reflects its close proximity to other functional residues, e.g. residues of the Zn binding motif. Thus, mutation of this residue probably leads to perturbation of substrate positioning (with respect to other functional residues) resulting in a significant loss of enzyme function. These conclusions are also supported by structural and mutational data for humLTA4H as well as other mammalian members of the M1 family of aminopeptidases (19)(20)(21)(22).
Glu-316 has a second essential role in LTA 4 hydrolysis in which it acts as a general base and acid catalyst specifically coupled to 5S,6S-DHETE production − All three mutants of Glu-316 were also unable to produce significant amounts of 5S,6S-DHETE from LTA 4 indicating that this residue is essential for the epoxide hydrolase activity. This agrees well with data for humLTA4H, which indicate that Glu-271 is critical for LTB 4 formation by assisting in the initial opening of the epoxide ring of LTA 4 (19).
Since scLTA4H mainly yields 5S,6S-DHETE as product whereas humLTA4H produces LTB 4 , the reaction mechanisms of the two enzymes appear to be different. A possible role for Glu-316 in the epoxide hydrolase reaction, which also fits our structural model of scLTA4H, would be that of general base and acid catalyst. The binding mode of LTA 4 , as outlined in Fig. 6, is supported by the modeled structure of an LTA 4 /scLTA4H complex (Fig. 1), previous structure-activity relationships of hydroxamic acid inhibitors (23) as well as the structure of humLTA4H (11,24). Furthermore, since the stereochemistry of the product, i.e., the vicinal diol 5S,6S-DHETE, has been retained from the substrate, the allylic epoxide LTA 4 (5S-trans-5,6-oxido-7,9-trans-11,14-cis-eicosatetraenoic acid), it could be inferred that the reaction proceeds according to an S N 1 mechanism.
Mutation of Asp-421 in scLTA4H did not inhibit the enzyme's epoxide hydrolase activity, as judged by the formation of 5S,6S-DHETE, LTB 4 , and Δ 6 -trans-Δ 8 -cis-LTB 4 . Instead this mutation increases the formation of 5S,6S-DHETE (Fig. 4). This was somewhat unexpected, at least with respect to the LTB 4 isomers, since the human counterpart, Asp-375, seems to serve as a general base catalyst in LTB 4 formation (25). However, in the model structure of scLTA4H, Asp-421 is too distant from the catalytic zinc to be able to catalyze the introduction of water at C-6 of LTA 4 , which explains why it is not responsible for the formation of 5S,6S-DHETE.

Arg-627 is essential for both LTA 4 hydrolysis as well as peptidolysis as a carboxylate
recognition site − Arg-627 is a residue that is conserved among LTA4H isoenzymes but not in, e.g., aminopeptidase B where a Lys is found instead. Recent structural, mutational and binding data for humLTA4H have identified Arg-563 as a carboxylate recognition site common to both the epoxide hydrolase and peptidase activities (26). Mutation of Arg-627 in scLTA4H to either an Ala or Lys residue drastically reduced the ability of the enzyme to hydrolyze both peptides and LTA 4 . To assess the contribution of Arg-627 to substrate binding, we used inhibition assays with substrate mimics, as well as determinations of apparent kinetic constants. Thus, the IC 50 value of bestatin, a peptide substrate mimic, against the Leu aminopeptidase activity increased 7-fold in the mutant R627K. Furthermore, removal of the positive charge by mutation of Arg-627 into an Ala residue, increased the IC 50 of bestatin 22fold and the K m values for Leu-p-NA and Alap-NA approximately 4-fold (data not shown). Although these data support a role for Arg-627 in peptide substrate binding, it can be argued that neither bestatin nor the nitroanilide substrates carry a carboxyl group that can be a reliable mimic of the C-terminus of a peptide substrate. In fact, the distance between Arg-627 and Glu-316 indicates that scLTA4H is a tripeptidase which was also supported in preliminary competition assays indicating that the tripeptide RSR is an excellent substrate for scLTA4H with a k cat of approx. 7 s -1 (data not shown). Thus, we modeled RSR into the active site of scLTA4H and found that Arg-627 can well bind the carboxylate of a tripeptide (Fig.  S6, Supplementary Material on-line). Hence, together our data support the conclusion that Arg-627 functions as a carboxylate recognition site for peptide substrates.
Due to the chemical instability and reactivity of LTA 4 , it is not possible to assess binding of this molecule to an inactive mutant and thus the role of Arg-627 for carboxylate binding. Instead we used a hydroxamic acid inhibitor, originally developed as a stable mimic of LTA 4 , and studied the effects of mutations on inhibitor potency. In contrast to bestatin, the potency of the hydroxamic acid inhibitor exhibits a limited, if any, dependence on an intact Arg-627. In fact, the inhibitor was even more potent against the mutant as compared to wild type scLTA4H. Presumably, this reflects the flexibility of the fatty acid-like hydroxamic acid inhibitor (in contrast to bestatin) which allows it to find alternative, equally strong, binding conformations in the mutated enzyme. Furthermore, modeling of LTA 4 into the active site of scLTA4H clearly indicates interactions between the C1 carboxyl group of the substrate and the positively charged guanido group of Arg-627 (Fig. 1). Together, these results suggest that Arg-627 acts as a carboxylate recognition site for LTA 4 . However, rather than providing binding strength per se, this residue appears to assure an accurate substrate positioning compatible with catalysis (Figs. 6 and 7), as previously demonstrated for humLTA4H (26).

S. cerevisiae LTA4H can be engineered to attain a peptide substrate specificity and sensitivity to LTA 4 -inhibition paralleling
humLTA4H − With the exceptions of E316A and F424Y, most single point mutations did not generate major effects on the aminopeptidase activity of scLTA4H. The E186A mutation selectively increased the Prop-NA hydrolyzing activity (Fig. 2), which may be explained by the fact that an Ala residue at this position allows a more bulky N-terminal group, such as Pro, to fit into the S1 subsite. In contrast, the F424Y mutation led to strongly increased activities for all substrates normally accepted by wild type scLTA4H, i.e. Leu-, Met-and Ala-p-NA. Since this mutation mainly affects the K m of the enzyme this is probably caused by increased substrate affinity upon introduction of the Tyr hydroxyl group (10).
The combined mutants 4M, 5M and particularly 6M efficiently hydrolyze Arg-p-NA and exhibits increased rates of Ala-p-NA hydrolysis, both in terms of specific activities and k cat /K m values (Figs. 2, 3 and Fig. S3, Supplementary Material on-line). The 6M mutant also exhibits increased rates of Pro-, Met-and Lys-p-NA hydrolysis (Fig. 3). For humLTA4H, Ala-and Arg-p-NA are turned over most efficiently, followed by Pro-, Leu-, Met-and Lys-p-NA (Fig. 3). In contrast, wild type scLTA4H is a Leu aminopeptidase, which hydrolyzes Ala-p-NA at significantly lower rates and Arg-and Pro-p-NA are barely accepted as substrates. Hence, mutants most closely mimicking the human enzyme, have attained aminopeptidase activities that are shared with humLTA4H. With the exception of Leu-p-NA, which is hydrolyzed at significantly higher rates in the hexuple yeast mutant than in humLTA4H, the substrate specificities are also very similar for these two enzymes (Fig. 3). Furthermore, the Ala and novel Arg aminopeptidase activities of mutants 4M and 5M were inhibited by treatment with LTA 4 and for 6M all aminopeptidase activities were inhibited by LTA 4 with a relative inhibitory profile similar to that observed for humLTA4H (Fig. 3). The gradual introduction of mutations in 4M, 5M and 6M also increase the enzyme's sensitivity for LTA 4 inhibition, as judged by decreased K i values, to levels comparable to those of humLTA4H (Table 1). Moreover, these mutations change the mode of inhibition to a competitive type, a feature typical of humLTA4H. Hence, not only the peptide substrate specificity of humLTA4H but also its sensitivity to inhibition by LTA 4 can be engineered into scLTA4H. LTA 4 appears to bind in one productive and one allosteric conformation to scLTA4H − Several lines of evidence indicate that LTA 4 binds in two conformations to scLTA4H, one productive that turns over LTA 4 , and one unproductive leading to stimulation of the peptidase activity (Fig. 7). The two activities of scLTA4H both require the catalytic zinc, Glu-316 and Arg-627, suggesting that the substrates of the two activities are mutually exclusive. Yet, LTA 4 can both inhibit and stimulate the aminopeptidase activity of scLTA4H in a substrate-dependent manner. Thus, LTA 4 treatment inhibits the Met peptidase activity of wild type and mutated enzymes as well as the Arg and Ala peptidase activity of certain mutants, whereas the Leu peptidase activity is generally stimulated (Fig.  2). Since the outcome of LTA 4 treatment for a given enzyme depends on the chemistry of the peptide substrate it is reasonable to assume that LTA 4 , in its activating binding mode, is bound close to where the substrates differ, viz. the side-chain of the N-terminal residue. In humLTA4H this part of the active site is very narrow forcing LTA 4 to bind in an extended conformation. This in turn, suggests that the LTA 4 binding pocket of scLTA4H has to be considerably more spacious to allow LTA 4 to bind differently. Furthermore, LTA 4 acts as a competitive inhibitor of the Ala-p-NA cleaving activity of 4M, 5M and 6M ( Table 1), suggesting that this substrate, which mostly occupies the zinc binding site and wide portion of the active site, is only affected by LTA 4 when bound in its elongated binding mode (Fig. 7). In contrast, LTA 4 gives a mixed type of inhibition against the substrate Arg-p-NA, presumably because its side-chain penetrates deeper into the hydrophobic S1 binding pocket, allowing interactions with LTA 4 bound in both conformations.
Mutations of scLTA4H affect the rate of LTA 4 hydrolysis and product profile − When engineering scLTA4H into an enzyme with properties resembling those of humLTA4H, several different effects on the epoxide hydrolase activity could in principle occur. These include (i) increased turnover of LTA 4 , (ii) reduced formation of 5S,6S-DHETE, and, in particular, (iii) increased formation of LTB 4 . The structure of humLTA4H shows that the putative LTA 4 binding cavity is formed at an interface between the three enzyme domains such that residues of all domains will contribute to the active site architecture and the resulting substrate binding, alignment and turnover.
Given these circumstances, exchanges of major protein segments, which may disrupt the quaternary structure, seemed like a less promising strategy for improving the epoxide hydrolase activity of the yeast enzyme. In fact, we exchanged a 108-residue segment of scLTA4H for the corresponding part of humLTA4H, which resulted in a protein that failed to express and fold properly (data not shown). Instead, we chose a strategy with a limited number of point mutations, selected from the modeled structure, possibly mimicking the molecular evolution of the enzyme.
Mutations assumed to cause significant changes to the active site of scLTA4H, i.e. D422V, D422V/N417del and Q412V, do indeed increase the rate of LTA 4 turnover (data not shown). Thus, these mutations increase the 5S,6S-DHETE production approximately 3-fold as compared to wild-type enzyme while the LTB 4 production remains the same. Since these effects were only observed under conditions with substrate saturation they probably reflect changes in both substrate affinity and turn over. Also the 3-fold increase observed for D421N (data not shown) suggests similar changes.
The combined mutants 3M, 4M and 5M exhibited reduced abilities to produce 5S,6S-DHETE as well as LTB 4 . The hexuple mutant (6M), however, not only displayed a reduced ability to produce 5S,6S-DHETE but also retained (or slightly increased) its ability to convert LTA 4 into LTB 4 . Thus, in relative terms this mutant exhibits increased LTB 4 over 5S,6S-DHETE production (Fig. 4), with a LTB 4 :5S,6S-DHETE ratio of 0.4 compared to 0.08 for wild-type scLTA4H. In addition, substrate binding appeared to be significantly improved since LTA 4 acted as a competitive inhibitor for the aminopeptidase activity of the hexuple mutant, across three tested substrates, with a potency profile similar to what is observed for humLTA4H (Table 1).
Recent data show that the catalytic machinery of humLTA4H is very sensitive to changes in the binding and precise alignment of LTA 4 along the active site (26). Even a minimal change, i.e. an Arg to Lys substitution, in the carboxylate recognition site (i.e. Arg-563 in humLTA4H corresponding to Arg-627 in scLTA4H) renders the enzyme inactive even though LTA 4 binding appears intact. Keeping this in mind it is perhaps not surprising that we were unable to engineer a mutant of scLTA4H with a robust LTB 4 producing activity by a limited number of point mutations. Nonetheless, the catalytic and kinetic properties of the hexuple mutant indicate that structural changes have been introduced which indeed mimic steps in the evolution of the epoxide hydrolase activity thus leading to an LTA 4 -binding cavity resembling the human counterpart.
Structural modifications of LTA4H during evolution − LTA4H is an unusual combination of an M1 zinc aminopeptidase and an epoxide hydrolase with a very strict substrate specificity, for which Glu-271, Asp-375, and Arg-563 are essential. Inspection of multiple sequence alignments, e.g. see the MEROPS peptidase database, merops.sanger.ac.uk (27), reveals that several M1 aminopeptidases of invertebrate organisms have the former two of these three residues conserved, i.e. those corresponding to Glu-316 and Asp-421 of scLTA4H. The carboxylate recognition site (Arg-627 of scLTA4H) is conserved within mammalian LTA4H and also occurs in some related M1 aminopeptidases (26). Thus, one would expect, based on sequence similarity and our biochemical data, that the ability to hydrolyze LTA 4 into 5S,6S-DHETE and possibly into LTB 4 , would exist among other members of this broad enzyme family. However, except for scLTA4H such activities have never been demonstrated.
Given the evolutionary conservation of the peptide cleaving activity it seems likely that LTA4H has been evolved from an ancestral aminopeptidase. Since only six point mutations were sufficient to convert scLTA4H into an enzyme with a peptidase activity resembling that of humLTA4H, it appears likely that scLTA4H represents a close ancestor among the lower eukaryotes. Indeed, according to a phylogenetic tree derived from an alignment of aminopeptidases of the M1 family, see the MEROPS peptidase database, merops.sanger.ac.uk (27), scLTA4H clusters together with mammalian LTA4H. Throughout the M1 family of aminopeptidases the catalytic domain is the most conserved domain. This suggests, and is indeed supported by our data, that changes of the catalytic domain mainly affects the peptidase activity, but also, to some extent, creates the structural basis required for LTA 4 binding. Our data shows that the introduction of a limited number of new functionalities to an aminopeptidase active site is sufficient to improve LTA 4 binding but additional changes are required to create the full machinery for efficient conversion of LTA 4 into LTB 4. How then has this specific ability been introduced? Inasmuch as the C-terminal domain of LTA4H appears to be most unique to LTA4Hs one may speculate that its intrinsic structural features and interactions with the N-terminal and catalytic domains, may be key to the process of fine-tuning the chemistry of the active site. Thus, in the course of evolution the binding pocket has been gradually reshaped in small steps to optimally fit LTA 4 thereby favoring LTB 4 production over other structural isomers. In this process, the substrate has been aligned with residues already serving in the peptidase activity, e.g. Glu-316 and Arg-627, thereby allowing them to perform novel epoxide hydrolase catalysis while at the same time maintaining the original peptidase activity.

Figure Legends
(B), The progress of of p-NA formation was monitored for 2 hours in the presence of wild-type enzyme alone (blue trace) as well as in the presence of wild-type and inactive mutant (E316A) enzyme (magenta trace). A simulated curve assuming preserved substrate affinity, equal to that of wild-type enzyme, of the inactive mutant is shown as a black trace. The addition of mutant enzyme (at a 250-fold higher concentration with respect to wild-type enzyme) to the reaction mixture has a very weak effect on the observed progress of product formation due to a drastically reduced substrate affinity of the mutant enzyme. Note that the axes are broken.
FIG. 6. -Model for the epoxide hydrolase reaction mechanism. The scheme shows a putative S N 1 reaction mechanism for the conversion of LTA 4 into the major product 5S,6S-DHETE by scLTA4H. Arg-627 holds the carboxy moiety of LTA 4 whereas the epoxide oxygen binds to the Zn 2+ ion, which functions as a Lewis acid, thus promoting activation and opening of the oxirane ring. In a second step, Glu-316 acts as a general base and polarizes a water molecule to facilitate its attack at C6 of the carbocation intermediate. In the final step, Glu-316 instead acts as an acid and delivers a proton to the oxyanion at C5, thus forming the second hydroxyl group of the vicinal diol.  4 . '3*' indicate residues exchanged in the 3M mutant and '4*', '5*'and '6*'indicate the additional residues exchanged in the mutants 4M, 5M and 6M, respectively. Panels B and C show the putative pocket of the yeast enzyme with LTA 4 in two binding modes. Residues around the catalytic Zn are highly conserved and also the basic residues binding the carboxy group of LTA 4 (as well as peptides). Residues lining the deeper part of the LTA 4 binding pocket differs between the two enzymes. As discussed in the text, the binding pocket of the yeast enzyme (solid line) appears to be more spacious than the human enzyme (dotted line). The double arrow at Asp-422 of scLTA4H denotes a putative interaction with residues of the C terminal domain that could affect the shape of the deeper part of the pocket. The additional space in scLTA4H could allow LTA 4 to bind in two possible conformations; one elongated (panel B) and one curled-up in the deeper part of the pocket (panel C), see text for further details. The latter binding mode would be the one compatible with activation of the peptidase activity. A peptide substrate would occupy the upper and middle part of the binding pocket. Depending on the side-chain of the peptide substrate it penetrates to a variable extent into the deeper part of the pocket thus interacting differently with LTA 4 , as indicated by the shaded areas in dark and light gray representing the binding of Ala-and Arg-p-NA, respectively. 4 .Aliquots of mutated enzymes (1 µg in 250 µl 50 mM Tris-HCl, pH 7.5 containing 100 mM KCl) were incubated with Ala-, Leu-or Arg-p-NA (9.6 -0.125 mM) at varying concentrations of LTA 4 , typically 1, 2, 5, 10, 20 μM. Values of K i were calculated as described in the methods section and are expressed as mean values ± SE. Note that the experiment is not applicable for the Leu-p-NA hydrolyzing activities of 4M and 5M since these activities are stimulated upon LTA 4 -treatment.        Figure S2.

_________________________________________________________________________
Stereo diagram of the modeled regions of LTA4H (before energy minimization) superimposed on its template LTA4H. The picture depicts the C-trace of LTA4H and modeled regions of LTA4H.
The model is coloured according to the same scheme as in S1.  Figure S5. LTA 4 -induced stimulation of the peptidase activity of wild type and mutated scLTA4H. The specific peptidase activity of enzymes susceptible to LTA 4 stimulation was assayed with different peptide substrates and increasing amounts of LTA 4 , until saturation was achieved. The maximum increase of the specific Leu-, Ala-and Pro-p -NA aminopeptidase activities are presented. The bars denote mean value ± SE.