8R-Lipoxygenase-catalyzed synthesis of a prominent cis-epoxyalcohol from dihomo-γ-linolenic acid: a distinctive transformation compared with S-lipoxygenases.

Conversion of fatty acid hydroperoxides to epoxyalcohols is a well known secondary reaction of lipoxygenases, described for S-specific lipoxygenases forming epoxyalcohols with a trans-epoxide configuration. Here we report on R-specific lipoxygenase synthesis of a cis-epoxyalcohol. Although arachidonic and dihomo-γ-linolenic acids are metabolized by extracts of the Caribbean coral Plexaura homomalla via 8R-lipoxygenase and allene oxide synthase activities, 20:3ω6 forms an additional prominent product, identified using UV, GC-MS, and NMR in comparison to synthetic standards as 8R,9S-cis-epoxy-10S-erythro-hydroxy-eicosa-11Z,14Z-dienoic acid. Both oxygens of (18)O-labeled 8R-hydroperoxide are retained in the product, indicating a hydroperoxide isomerase activity. Recombinant allene oxide synthase formed only allene epoxide from 8R-hydroperoxy-20:3ω6, whereas two different 8R-lipoxygenases selectively produced the epoxyalcohol.A biosynthetic scheme is proposed in which a partial rotation of the reacting intermediate is required to give the observed erythro epoxyalcohol product. This characteristic and the synthesis of cis-epoxy epoxyalcohol may be a feature of R-specific lipoxygenases.


Expression and purifi cation of 8R-lipoxygenase
cDNA of the 8 R -LOX domain of the P. homomalla peroxidaselipoxygenase fusion protein ( 19 ) was subcloned into the pET3a vector (with an N-terminal His4 tag), and the protein was expressed in Escherichia coli BL21 (DE3) cells and purifi ed by nickel affi nity chromatography according to a previously published protocol ( 20 ). For clarity, this 8 R -lipoxygenase is referred to herein as the recombinant 8 R -LOX.
The second P. homomalla 8 R -lipoxygenase tested here was the soluble enzyme purifi ed in 1996 ( 21 ); aliquots from the original purifi cation were stored at Ϫ 70°C and these retained suffi cient activity for use 15 years later. This enzyme is referred to here as the soluble 8 R -LOX.

Incubation with enzymes
Side-by-side incubations were performed at room temperature in 1 ml of 50 mM Tris pH 8.0 containing 500 mM NaCl, 2 mM CaCl 2 and 0.01% Emulphogene detergent using [ 14

GC-MS analysis of 18 O 2 incorporation in product from coral
Incubation of 20:3 6 (100 M) with an extract of P. homomalla acetone powder (3 mg powder/ml pH 8 buffer) was conducted under an atmosphere of 18 O 2 . The products were purifi ed initially by RP-HPLC (MeOH/H 2 O/HAc, 80/20/0.01 by volume), and then further purifi ed by SP-HPLC (Hex/IPA/HAc, 100/5/0.1 by volume for the epoxyalcohol). Aliquots of HETrE (prepared by TPP reduction) and epoxyalcohol from the 18 O 2 incubation, together with unlabeled samples, were hydrogenated (H 2 , palladium on carbon in ethanol for 2 min) and after addition of water and extraction with ethyl acetate, they were converted to the pentafl uorobenzyl (PFB) ester TMS ether derivative. The 18 O content was determined by GC-MS analysis in the negative ion/chemical ionization mode using a Nermag R10-10B instrument with a 5 m SPB-1 capillary column programmed from 150° to 300° at 20°/min. The samples were subjected to rapid repetitive scanning over a 10 a.m.u. mass range (0.2 s per scan) covering the prominent M-181 ion (loss of PFB, resulting in the RCOO ion of product); approximately 30 scans were collected during elution of the GC peak, and these were averaged for calculation of the relative ion abundances. For analysis of hydrogenated Although the naturally occurring prostaglandin products in P. homomalla are all 2-series derived from arachidonic acid, we included a study of the metabolic fate of 20:3 6 because it was originally reported as a substrate for the enzymatic activity in the coral ( 18 ) and because study of 20:3 6 metabolism in P. homomalla is not complicated by the presence of large amounts of endogenous products. With the availability of cloned recombinant enzymes from P. homomalla , we recently returned to the issue of the origin of this extra product from 20:3 6. The novel product we characterize herein is formed specifi cally by 8 R -lipoxygenase metabolism, and its unusual stereochemistry may represent a feature of the secondary reactions of R -as opposed to S -lipoxygenases.

Incubation with coral extracts
Frozen P. homomalla was cut into small pieces with scissors and placed in 10 vols of 50 mM Tris, pH 8, containing 1 M NaCl on ice and homogenized using a Polytron blender (Brinkmann) in 10-s bursts. The homogenate was allowed to settle under gravity for up to 30 min; aliquots of the supernatant were diluted 10-fold into fresh buffer for incubations with fatty acid substrates (100 M), typically for 5 min at room temperature. Products were extracted by the addition of 1 M KH 2 PO 4 plus suffi cient 1 N HCl to give pH 4, followed by extraction with 2 vols of ethyl acetate. The organic phase was collected, washed with water to remove traces of acid, and taken to dryness under nitrogen. The extracts were redissolved in a small volume of MeOH before HPLC analysis.
Acetone powders of P. homomalla were prepared as described ( 5 ) and stored at Ϫ 70°C until use. Typically, a 3 mg/ml suspension/solution in 50 mM Tris (pH 8) containing 1 M NaCl was prepared for incubations with substrates (5 min at room temperature). For recovery of 8-hydroperoxides from these incubations, the 3 mg/ml suspension was diluted 10-fold, and the incubation time was extended to 20 min; a few milligrams of 8 R -HPETE or 8 R -HPETrE could be prepared and purifi ed from 0.5 l of the dilute acetone powder incubations. Products were extracted as described above. If required, before HPLC, hydroperoxides were reduced using a molar excess of triphenylphosphine in MeOH (5 min at room temperature).

HPLC analyses
Typically, aliquots of the extracts were analyzed initially by RP-HPLC using an ODS Ultrasphere 5 column (Beckman) (25 × 0.46 cm) or Waters Symmetry column (25 × 0.46 cm) using a solvent of MeOH/H 2 O/HAc (80/20/0.01 or 75/25/0.01 by volume) at a fl ow rate of 1 ml/min with on-line UV detection (1100 series diode array detector; Agilent, Santa Clara, CA) and radioactive monitoring (Radiomatic Flo-One). Larger amounts (0.5-1 mg of total fatty acids) were injected for collection of products, or a semi-preparative column (Ultrasphere ODS, 25 × 1 cm; Beckmann) was used for larger quantities. Further analysis and purification was achieved by SP-HPLC using a 5-silica column (Alltech) or a Beckmann Ultrasphere 5 silica column using a

Determination of the C-10 hydroxyl confi guration
To establish the relative stereochemistry of the epoxide to the C-10 hydroxyl, two saturated analogs of the natural product were prepared by total chemical synthesis as outlined in the supplementary data and in supplementary Scheme I.
These synthetic standards, 8 R ,9 S -cis -epoxy-10-hydroxy-eicosanoates with the 9,10 erythro and threo relative confi gurations, were fi rst analyzed by GC-MS (EI mode) in comparison to the hydrogenated natural product as the methyl ester TMS ether derivatives. The threo 8,9-cis -epoxy-10-hydroxy-eicosanoate standard eluted before the erythro diastereomer (5 m SPB-1 capillary column, 150° to 300° at 20°/min) each as well resolved peaks with retention times of 4 min 52 s and 5 min 1 s, respectively. Their mass spectra had a noticeably different pattern of ion fragments, especially at the lower m/z values (see supplementary  , because it had been used three times previously for 18 O syntheses). The [ 18 O 2 ]8 R -HPETrE labeled in the hydroperoxy group was purifi ed by SP-HPLC and reacted with recombinant 8 R -LOX under a normal atmosphere to produce the corresponding epoxyalcohol. The 18 O contents of the 8 R -HPETrE and its corresponding epoxyalcohol product (which share the same molecular weight, 338 for the unlabeled species) were measured by negative ion electrospray LC-MS using a Ther-moFinnigan TSQ Quantum instrument by rapid repetitive scanning over the mass range encompassing the M-H anions ( m/z 330-350, 5 scans/sec). A total of 20-30 scans over the HPLC peaks were averaged to obtain the partial mass spectra of labeled and unlabeled epoxyalcohol and 8 R -HPETrE.

Metabolism in extracts of P. homomalla
As originally reported ( 5 ), when arachidonic acid (20:4 6) is incubated with extracts of P. homomalla , the fatty acid is rapidly metabolized by 8 R -LOX, and the resulting 8 R -HPETE is further transformed by allene oxide synthase, leading to the appearance of ␣ -ketol and cyclopentenone end products ( Fig. 1 , lower panel). Metabolism of dihomo-␥ -linolenic acid (20:3 6) is similar, except for the appearance of a prominent, more polar product that is absent (or present in insignifi cant amounts) in the arachidonic acid incubations ( Fig. 1 ).

Identifi cation of the novel 20:3 6 product
The structure was established based on UV, NMR, and GC-MS data. The purifi ed polar product displayed only end absorbance in the UV data, indicating no conjugated double bonds. A quantity of ‫ف‬ 100 g was prepared, and the proton NMR and COSY spectra were recorded in CDCl 3 . These results (see supplementary Table I ( 22 )], with ␣ -hydroxyl at C-10 and two cis double bonds at 11,12 and 14,15. So far this established the covalent structure as 8,9-cis -epoxy-10-hydroxy-eicosa-11 Z , 14 Z -dienoic acid, an epoxyalcohol of the hepoxilin B-type that is distinctive in being a cis -epoxide ( 23 ). Because the precursor of the epoxyalcohol is 8 R -HPETrE (an assumption proved formally using purifi ed enzymes, vide infra), these results are compatible with essentially complete retention of the hydroperoxy oxygens from the precursor 8 R -HPETrE.

Lack of product using allene oxide synthase
There are several precedents for the transformation of fatty acid hydroperoxides to epoxyalcohols catalyzed by allene oxide synthase (AOS) and related enzymes (24)(25)(26), and it seemed possible that this might account for the formation of the 20:3 6-derived epoxyalcohol. However, experiments with the expressed AOS domain of the P. homomalla AOS-LOX fusion protein ( 19 ) produced only allene oxide as product [detected as the major ␣ -ketol hydrolysis product and cyclopentenone ( 5 )] from 8 R -HPETE or 8 R -HPETrE (data not shown).

Formation of epoxyalcohol by 8 R -LOX enzymes
By contrast, use of the recombinant LOX domain of the AOS-LOX fusion protein gave positive results. When suffi cient enzyme was used to quickly transform (<1 min) all the fatty acid to the corresponding 8 R -hydroperoxide, further reaction generated secondary products. When observed by repetitive scanning in the UV, the rapid appearance of the derivatives (data not shown). The erythro standard had an indistinguishable mass spectrum and retention time to the hydrogenated epoxyalcohol product of P. homomalla . Their structural identity was confi rmed by comparison of the NMR spectra of the saturated natural product with the synthetic standards ( Fig. 2 ). These data confi rmed the erythro relative confi guration at 9,10 in the natural product. Because P. homomalla exhibits only 8 R -LOX activity, the cis epoxide moiety can be assigned as the 8 R ,9 S enantiomer. Thus, the complete structure of the novel product from 20:3 6 is established as 8 R ,9 S -cis -epoxy-10 S -hydroxy-eicosa-11 Z ,14 Z -dienoic acid. Metabolism in the coral extracts is summarized in Scheme 1 . We also tested the soluble 76-kDa 8 R -LOX from P. homomalla , which was available in limited quantities from the original purifi cation ( 21 ). It reacted very similarly to the recombinant 8 R -LOX from the AOS-LOX fusion protein. The substrates 20:4 6 and 20:3 6 were comparable for oxygenation to the corresponding 8 R -hydroperoxide; however, 8 R -HPETrE was converted to further products at over twice the rate of 8 R -HPETE. When reactions with identical amounts of enzyme were analyzed and stopped at the same time (with half of the 20:3 6 hydroperoxide consumed), subsequent RP-HPLC analysis confi rmed the more extensive metabolism of 8 R -HPETrE and the appearance of a single prominent, more polar peak detected at 205 nm, with no comparable prominent product from 8 R -HPETE ( Fig. 3B ). This polar product from 20:3 6 was identifi ed as the same epoxyalcohol identifi ed earlier by its identical UV profi le and cochromatography on both RP-HPLC and SP-HPLC with the epoxyalcohol formed by the recombinant 8 R -LOX.

Retention of hydroperoxy oxygens in the epoxyalcohol
When 8 R -HPETrE containing an ‫ف‬ 1:2 mixture of 2 16 O and 2 18 O in the hydroperoxide group was reacted with the recombinant 8 R -LOX, the 18 O contents of the substrate and epoxyalcohol product were almost indistinguishable ( Fig. 4 ). Close inspection indicated 98% retention of both hydroperoxy oxygens in the epoxyalcohol, pointing to a conjugated diene at 237 nm was followed by the gradual decrease in intensity at this wavelength, with the appearance of a new chromophore characteristic of a conjugated triene(s) centered on ‫ف‬ 270 nm and a weaker broad absorbance in the area of 300-350 nm. The main product of the 20:3 6 reaction absorbs relatively weakly, at 205 nm, and is not detected by UV scanning (see below). In side-by-side incubations monitored in the UV, it was apparent that the 20:3 6-derived 8 R -HPETrE disappeared more quickly than the corresponding arachidonic acid-derived 8 R -HPETE. These side-by-side reactions were also conducted using 14 Clabeled fatty acid substrate, and, after extraction of these samples using C18 cartridges, RP-HPLC analysis showed distinctly different profi les of products ( Fig. 3A ). The results confi rmed the more extensive metabolism of the 20:3 6-derived 8R-HPETrE (less remaining compared with 8 R -HPETE) and, more signifi cantly, the prominent appearance of a polar product unique to 20:3 6 metabolism. This distinctive peak at ‫ف‬ 10 min is the most abundant secondary product from 20:3 6, detected at 205 nm in the UV. In larger-scale incubations, this polar product from 20:3 6 was prepared in suffi cient amounts for structural analysis by 1 H-NMR (see supplemental Tables I and II). On the basis of these data, the 8 R -LOX product was shown to be identical to the coral epoxyalcohol 8 R ,9 S -cis -epoxy-10 Shydroxy-eicosa-11 Z ,14 Z -dienoic acid. ing for the erythro confi guration of the epoxyalcohol product (discussed in the following subsection), and therefore it is imperative that the structural assignment is secure. For trans -epoxy epoxyalcohols, there are empirical rules that reliably allow assignment of the erythro or threo confi guration. These rules relate to their relative polarity on TLC, relative retention time on GC, and both the relative chemical shifts and coupling constants on NMR ( 22,32,33 ). However, for cis -epoxy products there are fewer closely analogous examples in the literature (e.g., all fatty acidrelated epoxyalcohols with data available are trans epoxides), and the differences for erythro and threo on NMR are small or nonexistent ( 34 ). Our assignment is founded on the well precedented threo product in Sharpless' hydroxyldirected epoxidation of Z-allylic alcohols with Ti(OiPr) 4 (see supplementary Scheme I, epoxidation of 10 R -3) (34)(35)(36)(37). This allowed assignment of the two epoxide diastereomers ( erythro and threo ) obtained via Sharpless asymmetric epoxidation (see supplementary Scheme I). Indeed, the latter assignment shows good agreement with precedent using closely related model compounds ( 34,35 ). For example, the asymmetric epoxidation (using L-(+)diisopropyl tartrate) of 3-hydroxy-4 Z -undecenol yields an unreactive 3 R enantiomer with 2:3 ratio of erythro : threo products ( 35 ); our results concur exactly with this precedent and others ( 34,36,37 ).

Proposed catalytic cycle
The reaction is catalyzed and controlled by the active site iron, which must fi rst cleave the hydroperoxide and subsequently catalyze an oxygen rebound and hydroxylate the intermediate epoxyallylic radical while both hydroperoxy oxygens are retained in the epoxyalcohol product ( Fig. 5 ). This is easy to conceptualize for the reactions of S -confi guration fatty acid hydroperoxides because all steps occur on the same face of the reacting molecule, allowing formation of a trans epoxide and threo alcohol ( Fig. 5 , box). Our results with the R -confi guration hydroperoxide indicate not only formation of a cis -epoxide, which itself presents no conceptual problem, but also the erythro confi guration of the alcohol. Assuming the iron is in control, this necessitates either a 9,10 bond rotation before hydroxylation or fl ipping over of the reacting epoxyallylic radical intermediate ( Fig. 5 , right and left options). Perhaps the 8 R -hydroperoxide sits partly turned away from square so that the epoxyallylic intermediate, when formed, further rotates to expose the opposite face of the intermediate for hydroxylation. We note too that the formation of cis -epoxides may be a characteristic of 8 R -LOX because the activity in P. homomalla extracts was shown to convert 5 S -HPETE to cis -epoxy LTA 4 , not to the well known trans -epoxy leukotriene A 4 ( 38 ). Although the mechanisms of epoxyalcohol and LTA 4 synthesis differ, the reactions being initiated by the ferrous and ferric enzymes, respectively, the substrate conformation that predisposes to cis -epoxide formation is dictated by binding in the active site and thus could be dictated in similar fashion by an enzyme that favors R versus S oxygenation. mechanism involving close control of the transformation by the 8 R -LOX enzyme.

Hydroperoxide isomerase activity
The typical dioxygenase activity of lipoxygenase enzymes involves activation of the resting ferrous enzyme to the ferric form, then cycling of the ferric enzyme as it catalyzes reaction with polyunsaturated fatty acid and O 2 ( 27 ). By contrast, the epoxyalcohol biosynthesis we characterize here fi ts the criteria for a LOX enzyme acting as a hydroperoxide isomerase ( 28,29 ). In this case, the reaction cycle is initiated by the ferrous enzyme. Several lines of evidence suggest that a lack of access of molecular oxygen within the active site promotes hydroperoxide isomerase activity ( 30 ). If present, molecular oxygen reacts readily with radical intermediates, thus intercepting and blocking hydroperoxide isomerase cycling. Furthermore, molecular oxygen promotes enzyme activation to the ferric form, also inhibiting isomerase activity ( 29,31 ). Therefore, one can deduce that the 8 R -HPETrE is an acceptable substrate for interaction with the ferrous iron and that O 2 is excluded from intercepting the radical intermediates. With the arachidonic acid-derived 8 R -hydroperoxide, the overall rate of reaction is comparatively sluggish, and very little epoxyalcohol product is formed. The main products are dihydroperoxides or leukotriene A-related diols, both of which are products of the ferric enzyme. This suggests that the selective reaction with the 20:3 8 R -hydroperoxide is facilitated by exclusion of O 2 within a critical part of the active site and that this does not occur with binding of the arachidonate analog.

Assignment of the 10S (erythro, anti) confi guration
In postulating a mechanism for the hydroperoxide cycling with 8 R -HPETrE, there is some diffi culty in account- lipoxygenases can diffuse out of the active site or be subject to interception by molecular oxygen, an event that promotes lipoxygenase activation to the ferric form ( 30 ). Accordingly, one might expect there is more time in the 8 R -LOX reaction for the rotation required to form the observed erythro epoxyalcohol product ( Fig. 5 ).

Wrap-up of a historical issue
The striking and unexpected difference between 20:4 6 and 20:3 6 metabolism in P. homomalla was detected in the original investigations of prostaglandin biosynthesis by Corey and Ensley, and the prominent extra product from 20:3 6 was partially characterized ( 17 ). For example, it was shown to exhibit only weak end absorbance in the UV, to not react with sodium borohydride, to contain two double bonds and an alcohol and a possible epoxy functionality, and to have a molecular formula as the methyl ester of C 21 H 36 O 4 , all a perfect match for the epoxyalcohol we identify. Furthermore, the reported mass spectrum of the hydrogenated product as the methyl ester TMS derivative [listed in tabular form in the thesis ( 17 )] contains all the major ions and similar ion abundances as reported in our Results section. There is little doubt that this product and our epoxyalcohol are the same compound. The existence of 8 R -LOX metabolism in P. homomalla was not uncovered until the mid-1980s, a decade after these early biosynthetic studies ( 8 ), and it was only around the years 1995-2000 that the origin of the coral prostaglandins via cyclooxygenase was fi rmly established ( 6,7,(47)(48)(49).  (8,9-cis -epoxy, 9,10 erythro ) and the complete retention of the hydroperoxy oxygens. Assuming that the active site iron cleaves the hydroperoxide and momentarily binds the distal hydroperoxy oxygen, the epoxyallylic radical intermediate must either rotate at the 9,10 bond (left) or fl ip over (right) to produce the epoxyalcohol product. In the box: reaction of S -confi guration fatty acid hydroperoxide forms a trans -epoxy threo -hydroxy epoxyalcohol.

Other biosyntheses of cis-epoxyalcohols
Although heretofore only trans -epoxyalcohols have been reported from lipoxygenase catalysis (e.g., 23,28,[39][40][41], other enzymes can make the cis -epoxides. The majority of these are mechanistically quite distinct, however, because the epoxide is formed via oxygen transfer. The epoxyalcohol synthase activities in the fi sh parasitic fungus Saprolegnia parasitica ( 42 ) and in potato leaves and beetroot ( 43,44 ) catalyze oxygen transfer from the hydroperoxy fatty acid to the adjacent conjugated diene; the original hydroperoxide moiety is reduced to an alcohol, while the transferred oxygen produces trans -or cis -epoxidation of the trans and cis double bonds, respectively. In the case of plant peroxygenases, epoxidation may occur via intermolecular or intramolecular oxygen transfer from a fatty acid hydroperoxide to a cis double bond ( 45,46 ). More similar to our reaction, but forming the threo product, is the conversion of 13 S -hydroperoxylinoleic acid to the 11 S -threo -hydroxy-12 R ,13 S -cis -epoxide by a cytochrome P450 in the amphioxus Branchiostoma fl oridae ( 26 ). Notably, the oxygen rebound step in P450 catalysis is very fast ( ‫ف‬ 10 Ϫ 9 s), tending to favor suprafacial hydroxylation of the intermediate, forming the threo epoxyalcohol. By comparison, the equivalent intermediate in the hydroperoxide isomerase activity of