Dioxygenase Activity of Epidermal Lipoxygenase-3 Unveiled

Epidermal lipoxygenase-3 (eLOX3) exhibits hydroperoxide isomerase activity implicated in epidermal barrier formation, but its potential dioxygenase activity has remained elusive. We identified herein a synthetic fatty acid, 9E,11Z,14Z-20:3ω6, that was oxygenated by eLOX3 specifically to the 9S-hydroperoxide. Reaction showed a pronounced lag phase, which suggested that eLOX3 is deficient in its activation step. Indeed, we found that high concentrations of hydroperoxide activator (e.g. 65 μm) overcame a prolonged lag phase (>1 h) and unveiled a dioxygenase activity with arachidonic acid; the main products were the 5-, 9-, and 7-hydroperoxyeicosatetraenoic acids (HPETEs). These were R/S mixtures (ranging from ∼50:50 to 73:27), and as the bis-allylic 7-HPETE can be formed only inside the enzyme active site, the results indicate there is oxygen availability along either face of the reacting fatty acid radical. That the active site oxygen supply is limited is implied from the need for continuous re-activation, as carbon radical leakage leaves the enzyme in the unactivated ferrous state. An Ala-to-Gly mutation, known to affect the positioning of O2 in the active site of other lipoxygenase enzymes, led to more readily activated reaction and a significant increase in the 9R- over the 5-HPETE. Activation and cycling of the ferric enzyme are thus promoted using the 9E,11Z,14Z-20:3ω6 substrate, by continuous hydroperoxide activation, or by the Ala-to-Gly mutation. We suggest that eLOX3 represents one end of a spectrum among lipoxygenases where activation is inefficient, favoring hydroperoxide isomerase cycling, with the opposite end represented by readily activated enzymes in which dioxygenase activity is prominent.

Epidermal lipoxygenase-3 (eLOX3) 2 is one of the two lipoxygenases (LOX) involved in skin barrier formation (1,2). Inactivating mutations in either the gene encoding human eLOX3 or the gene encoding 12R-LOX lead to impaired skin barrier function that manifests itself in the disease autosomal recessive congenital ichthyosis (3,4). The scaly skin phenotype of this disease is a compensatory response to the fundamental defect in the barrier function (5). Targeted gene knockouts or ethylnitrourea mutagenesis in mice confirm the role of these two LOX enzymes, and in both cases the mice die soon after birth due to severe transepidermal water loss (6 -8).
In addition to its involvement in skin barrier formation, eLOX3 also attracts attention because of its unusual enzymology. It is classified as a LOX based on its amino acid sequence (e.g. 58% identity to 12R-LOX) yet its name eLOX3 lacks the usual designation such as 5-, 12-, or 15-LOX. The reason is the enzyme appears to be devoid of typical dioxygenase activity with polyunsaturated fatty acid substrates such as arachidonic acid (9,10). Instead, it displays a novel hydroperoxide isomerase activity that efficiently converts fatty acid hydroperoxides, the usual LOX products, to epoxyalcohols and ketones (10). This hydroperoxide isomerase activity, acting on a hydroperoxide product of 12R-LOX, has been proposed to be the physiological function of eLOX3 (1).
The non-heme iron in the active site of LOX enzymes exists in two oxidation states, ferrous and ferric. Native LOX are usually in the ferrous state, yet only the ferric enzyme is catalytically active as a dioxygenase (11). Therefore, to enter the catalytic cycle the ferrous enzyme has to be oxidized to the ferric form first, usually by 1 eq of their own hydroperoxide product. This process, commonly known as the activation step, accounts for the initial lag phase in the progress curve of LOX reactions (12,13). Because the reaction between typical LOX and fatty acid hydroperoxides is only a single turnover event, no significant products would accumulate unless an iron-reducing agent such as nordihydroguaiaretic acid (NDGA) is given to regenerate the ferrous enzyme (14) (Fig. 1A).
Unlike typical LOX, eLOX3 does not require iron-reducing agents for its hydroperoxide isomerase activity. That this reaction is catalysis, rather than a single turnover, implies that the ferrous state of eLOX3 is largely preserved. However, NDGA does speed up the reaction (10), suggesting that the enzyme can be oxidized by fatty acid hydroperoxides to some extent. We reasoned that the activation step of the LOX activity in eLOX3, if it does occur, would be very inefficient, requiring many molecules of fatty acid hydroperoxide to oxidize one molecule of eLOX3 (Fig. 1B). In other words, higher than usual concentrations of fatty acid hydroperoxide might be required to activate eLOX3. This inefficiency in the activation step of eLOX3 could potentially account for its apparent lack of dioxygenase activity, which must be performed by the activated ferric enzyme.
An additional explanation for the lack of dioxygenase activity in eLOX3 might be that the active site iron has a low redox potential. This redox requirement is imposed by the first chemical step of the dioxygenase cycle, which involves homolytic cleavage of a bis-allylic C-H bond in the fatty acid substrate by the active site Fe(III)-OH complex (15). The moderately high bond dissociation energy of this C-H bond (e.g. ϳ73 kcal/mol for C 7 -H, C 10 -H, or C 13 -H in arachidonic acid (16)) dictates that the redox potential of the active site iron should also be moderately high for the bond cleavage to occur (17). Indeed, a moderately high redox potential of ϳ600 mV has been demonstrated experimentally in the prototypical LOX, soybean LOX-1 (18). Unfortunately, a direct test of this hypothesis, by measuring the redox potential of eLOX3 for comparison with soybean LOX-1, is not yet possible due to difficulties in preparing multiple milligrams of purified eLOX3 for EPR spectroscopy.
In this study, we set out to test indirectly the redox potential hypothesis by incubating eLOX3 with a panel of synthetic fatty acids with extended conjugation. Because of the extended conjugation, these fatty acids are easier to oxidize than typical polyunsaturated fatty acids such as arachidonic acid (16) and thus might be accepted by eLOX3 as substrate. The study on this panel of synthetic fatty acids revealed some unusual characteristics of eLOX3 that prompted us to consider the other possibility that eLOX3 does not react with natural fatty acids because it is deficient in the activation step. To further test this possibility, we reexamined the reaction eLOX3 with arachidonic acid using high concentrations of hydroperoxide activator. We also studied the effect of an Alato-Gly mutation in the active site that is known to affect oxygenation specificity (19,20), and we investigated the effect of substrate orientation on product specificity using the large ester arachidonoyl-lysophosphatidic acid (AA/lyso-PA). "Discussion" ties the findings together and proposes a model with mechanistic relevance to the understanding of LOX enzymes in general.
Chemical Synthesis of Conjugated Fatty Acids-9E,11Z,14Z-20:36 was synthesized by Dr. Jin K. Cha, Wayne State University, as described previously (22). 8Z,11Z,13E-20: 37 was a kind gift from Dr. K. C. Nicolaou at the Scripps Research Institute. 11Z,14Z,16E-20:34 and 12E,14Z,17Z-20: 33 were synthesized by alkaline isomerization (23) of 11Z,14Z,17Z-20:33 (NuChek Prep Inc., Elysian, MN) and 9Z,12Z,14E-18:34 by alkaline isomerization of ␣-linolenic acid (NuChek Prep Inc., Elysian, MN)). Briefly, 10 mg of the precursor fatty acid was dissolved in 1 ml of 1 N KOH in ethylene glycol, flushed with argon, and placed in an oven at 144°C. The extent of the reaction was monitored every 10 min by UV spectrophotometry assay of an aliquot, and the reaction was stopped when UV absorbance at 235 nm ceased to increase (total reaction time was about 20 -30 min). The reaction mixture was then acidified by 1 N HCl to pH 3 and extracted by methylene chloride. The extract was analyzed by silver ion chromatography, and the two desired products were identified based on their UV (identical to that of 9E,11Z,14Z-20:36) and 1 H and 1 H, 1 H-COSY NMR spectra.
Expression and Purification of Human eLOX3-The cDNA of human eLOX3 with an N-terminal His tag was subcloned into the pCW vector, and the protein was expressed and purified according to a previously published protocol (24). Flatbottomed 500-ml flasks, which were found to give a higher yield of eLOX3 than baffled flasks, were used for the 50-ml cultures.
Reaction Incubations-The room temperature reaction with 9E,11Z,14Z-20:36 was monitored by UV spectrophotometry. By contrast, 0°C reactions and the room temperature reaction with arachidonic acid were not monitored by UV spectroscopy. For the latter, incubation was typically conducted in a 5-ml reactivial. First, the fatty acid substrate and the auxiliary components to be investigated (e.g. hydroperoxide activators, 4-hydroxy-TEMPO) were added to an empty vial and then taken to dryness under a stream of N 2 . Next, 2 ml of 100 mM sodium phosphate buffer, pH 7.5, was added, and the vial was capped and subjected to vortex. Next, the enzyme was added to start the reaction. Finally, the reaction was quenched by addition of methanol (usually 1 ml). For time course experiments with arachidonic acid, at specified time points, an aliquot (usually 150 -200 l) of the incubation mixture was taken out and immediately added to a separate vial containing 833 ng of 9E,11Z,14Z-20:36, the internal standard, in 0.5 ml of methanol. The 0°C reaction with AA/ lyso-PA was performed under the same conditions as the 0°C reaction with arachidonic acid. The lysophospholipid was chosen for this study because of its better solubility in aqueous solutions than those of diacyl phospholipids (25).
Extraction of Reaction Products-The incubation mixture was acidified to pH 5 using 1 N HCl, treated with more than 1 M eq of SnCl 2 on ice for 10 min, and then extracted by an Oasis HLB cartridge (Waters Corp.).
For reaction with arachidonic acid, RP-HPLC analysis of HETE products was performed using a Waters Symmetry C18 column (0.46 ϫ 25 cm), a solvent system of methanol/water/ acetic acid (80:20:0.01), and a flow rate of 1 ml/min. For time course experiments, RP-HPLC analysis used the same column, methanol/water/acetic acid (95:5:0.01), and 1 ml/min. The remaining arachidonic acid substrate and accumulated 9and 5-HPETE products in the time course experiments were quantified by comparison of their corresponding peak area with that of the internal standard, 9E,11Z,14Z-20:36. To avoid rearrangement of bis-allylic HETEs, which may occur under even mildly acidic conditions (26), HPLC analyses with no acid-containing solvent systems were also performed on reaction extract pretreated with ethereal diazomethane, which converts free acids to methyl esters. 7-HETE and 11-HETE methyl esters were separated using a Beckman silica column (0.46 ϫ 25 cm), an isocratic solvent system of hexane/isopropyl alcohol (100:2, by volume), and a flow rate of 1 ml/min. The relative abundance of each HETE product was estimated based on the 205 nm absorbance. (This is justified by a previous study showing that the radioactive trace of [1-14 C]HETEs matches well with the corresponding UV trace at 205 nm (26).) Chiral analysis used a Daicel ChiralPak AD column (0.46 ϫ 25 cm), isocratic solvent of hexane/methanol (100: 1.5), and a flow rate of 1 ml/min. The stereo-configuration of methyl 7-HETE enantiomers on chiral HPLC was deduced based on the finding that bis-allylic HETEs would undergo mainly "suprafacial" rearrangement upon mild acid treatment (26). Briefly, the two enantiomers (1-2 g) were purified by chiral HPLC, repeatedly (4 -5 cycles) treated with 1 ml of methanol/water/acetic acid (75:25:1) followed by evaporation under N 2 , and then subjected to chiral HPLC analysis. For example, the second peak of 7-HETE methyl ester on chiral HPLC was found to rearrange mainly to 5S-HETE and 9R-HETE methyl esters and was thus deduced to be 7R, because 7R-hydroxyl sits on the same face of the double bond system in the fully extended configuration as does the 5S-or 9R-hydroxyl.
For reaction with AA/lyso-PA, the products were transmethylated after extraction using a previously published protocol (27), followed by RP-HPLC analysis using a Waters Symmetry C18 column (0.46 ϫ 25 cm), an isocratic solvent system of methanol/water (80/20), and a flow rate of 1 ml/min.
Derivatization and GC-MS Analysis-Catalytic hydrogenations were performed in 100 l of ethanol using about 1 mg of palladium on alumina and bubbling with hydrogen for 2 min at room temperature. The hydrogenated products were recovered by the addition of water and extraction with ethyl acetate. Trimethylsilyl (TMS) ether derivatives were prepared by treatment with bis(trimethylsilyl)-trifluoroacetamide (10 l) and pyridine (2 l) at room temperature for 2 h. Subsequently, the reagents were evaporated under a stream of nitrogen, and the samples were dissolved in hexane for GC-MS.
Analysis of the methyl ester trimethylsilyl ether derivatives of the products was carried out in the positive ion electron impact mode (70 eV) using a Thermo Finnigan Trace DSQ ion trap GC-MS with the Xcalibur data system. Samples were injected at 150°C, and after 1 min the temperature was programmed to 300°C at 20°C/min. The spectra shown under Results were averaged from about 10 spectra collected during elution of the GC peaks.

RESULTS
Dioxygenase Activity of eLOX3 with 9E,11Z,14Z-20:36-To examine the potential oxygenase activity of eLOX3, we tested eLOX3 with a series of synthetic fatty acids with extended conjugation (their structures are shown in supplemental Fig. S1). Among these fatty acids, only 9E,11Z,14Z-20:36 turned out to be a good substrate. The initial reactions of eLOX3 (0.1 M) with 9E,11Z,14Z-20:36 (18 M) were conducted at room temperature and monitored by scanning UV spectrophotometry. The substrate contains a conjugated diene chromophore and absorbs strongly at 235 nm ( Fig. 2A, dotted line). Upon addition of enzyme, the UV absorbance of the substrate gradually decreased, whereas a new chromophore typical of conjugated trienes came to prominence ( max ϳ270 nm; Fig. 2A). Monitoring of the rate of reaction at 235 nm revealed a pronounced lag phase of ϳ5 min during which time no significant reaction is apparent (Fig. 2B); addition of exogenous hydroperoxide activator (1 M 13S-HPODE) almost eliminated the lag phase. Kinetic measurements of initial rates over a range of substrate concentrations gave estimates of K m and k cat as 50 M and 1.6 s Ϫ1 , respectively (Fig. 2C).
Products of the 9E,11Z,14Z-20:36 Reaction at Room Temperature-SP and RP-HPLC analyses revealed a complex product pattern, including at least four products with max at 270 nm, four at 235 nm, and four at 320 nm (data not shown). We reasoned that the substrate was first oxygenated to conjugated triene hydroperoxides ( max 270 nm), which were then further converted by the hydroperoxide isomerase activity of eLOX3 to epoxyalcohols and ketones ( max 235 and 320 nm; for postulated structures, see supplemental Fig. S2). The fact that the epoxyalcohol products also absorb at 235 nm explains why a substantial UV absorbance at 235 nm remained upon completion of the reaction ( Fig. 2A).
Trapping of the Primary Products-To focus on the primary LOX reaction, we identified the following three key factors that almost eliminated secondary product formation: (i) incubation at 0°C preferentially reduces hydroperoxide isomerase activity compared with dioxygenase activity (28 -31); (ii) an excess of 15S-HPETE included in the incubation competes with the conjugated triene hydroperoxide products for the hydroperoxide isomerase activity, and the excess 15S-HPETE helps keep the enzyme in the ferric state (see below); and (iii), addition of 4-hydroxy-TEMPO (1.5 mM), a water-soluble antioxidant, prevents nonenzymatic lipid peroxidation without causing significant enzyme inhibition (32). At the end of the incubation, SnCl 2 was added to convert hydroperoxides to the more stable alcohols, and the products were extracted and analyzed by SP-HPLC with a diode-array detector (Fig. 2, D Dioxygenase Activity of eLOX3 DECEMBER 17, 2010 • VOLUME 285 • NUMBER 51

JOURNAL OF BIOLOGICAL CHEMISTRY 39869
and E). By comparison with authentic standards obtained in a previous study (22) and by GC-MS analysis, the major peak 3 (76%) was identified as 9-hydroxy-10E,12E,14Z-eicosatrienoic acid, and the two minor products 1 and 2 as positional and geometric isomers of 3. (Products 1 and 2 both have a 15hydroxyl but differ in double bond configuration; data not shown). Chiral phase HPLC resolved product 3 into the R and S isomers with a ratio of 6:94 (Fig. 2F). Thus, 9E,11Z,14Z-20: 36 was oxygenated by eLOX3 in a regio-and stereospecific fashion to its 9S-hydroperoxy derivative, 9S-hydroperoxy-10E,12E,14Z-eicosatrienoic acid. The 9S-oxygenation occurs four carbons removed from the site of hydrogen abstraction at C-13. None of the other conjugated fatty acids tested were significantly oxygenated, indicating substrate selectivity of eLOX3.
Dioxygenase Activity of eLOX3 with Arachidonic Acid-The pronounced lag phase seen with 9E,11Z,14Z-20:36 led us to question whether there might be an even longer lag phase with arachidonic acid, which might be overcome using fatty acid hydroperoxide activator. As shown in Fig. 3A, in the absence of fatty acid hydroperoxides, eLOX3 (0.4 M) showed only weak and slow activity with arachidonic acid (42 M) as monitored by RP-HPLC assay of the remaining substrate. In the presence of high concentrations of fatty acid hydroperoxides, however, the rate of conversion of arachidonic acid was greatly accelerated, more so in the presence of 13S-HPODE than in the presence of 12R-HPETE. Also, the effect of fatty acid hydroperoxides is dose-dependent, as 65 M 13S-HPODE proved to be more effective than 27 M 13S-HPODE. In fact, the reaction initiated with 27 M 13S-HPODE was found to stall when 13S-HPODE was fully consumed and to resume only after a second dose of 13S-HPODE (58 M) was given. Therefore, the reaction of eLOX3 with arachidonic acid coincided with and was likely driven by its reaction with fatty acid hydroperoxides.
Dioxygenase Activity of A451G eLOX3 with Arachidonic Acid-The stereospecificity of oxygenation in LOX enzymes is partly controlled by an active site residue conserved as Ala in S-LOX and Gly in R-LOX (19). eLOX3 has Ala in this position. Recently, Cristea and Oliw (33) reported that mutation of the natural Gly to Ala in manganese lipoxygenase strongly augments hydroperoxide isomerase activity. Accordingly, it was of interest to examine whether changing Ala-451 to Gly in eLOX3 would produce the opposite effect, i.e. a switch from hydroperoxide isomerase to dioxygenase activity. Therefore, we mutated Ala-451 to Gly and incubated the mutant enzyme (0.5 M) with arachidonic acid (42 M). Unlike wildtype eLOX3, A451G eLOX3 reacted well with arachidonic acid even in the absence of fatty acid hydroperoxides, yet only after an unusually long lag phase of ϳ20 min (Fig. 3B). This lag phase could be eliminated by addition of 11 M 13S-HPODE at the beginning of the reaction (Fig. 3B).
Effect of Temperature on Accumulation of HPETE Products-In the room temperature reactions mentioned above, the expected HPETE products occurred only as transient intermedi-  ates with either the wild-type or A451G eLOX3 (Fig. 4, A and C); the primary hydroperoxide products are further converted to epoxyalcohols and ketones by the hydroperoxide isomerase activity of eLOX3 (data not shown). However, similar to the observations with 9E,11Z,14Z-20:36, when reaction with arachidonic acid was performed at 0°C and in the presence of excess 13S-HPODE, the HPETE products continued to accumulate during the entire course of the reaction (Fig. 4, B and  D), and the epoxyalcohol and ketone secondary products were formed only in negligible amounts.
Identification of the HPETE Products-After the reaction, the HPETE products from the 0°C incubation were reduced with SnCl 2 , extracted using a C 18 cartridge, and subjected to RP-and SP-HPLC analyses. As shown in Fig. 5A, the products consisted of predominantly 7-, 9-, and 5-H(P)ETEs (25,29, and 22%, respectively) that stemmed from C-7 hydrogen abstraction, followed by 15-and 11-H(P)ETEs (8 and 9% respectively) from C-13 hydrogen abstraction. 12-and 8-H(P)ETEs were formed only in minor amounts (less than 3% each). The identity of the unusual bis-allylic 7-H(P)ETE product was unequivocally established by GC-MS analysis of its methyl ester TMS ether derivative, before and after catalytic hydrogenation (supplemental Fig. S3 and Fig. 5C).
Consistent with this lack of positional specificity, further analysis by chiral HPLC indicated no remarkable stereospecificity in any of the positional isomers (Fig. 6, A, C, and E). Among the three major products, 7-H(P)ETE showed the highest degree of stereospecificity, being 73% 7R. The 5-H(P)ETE product ranked the second, being 62% 5R, and the 9-H(P)ETE product was essentially racemic.
The active site Ala-Gly substitution in other LOX enzymes can lead to relatively dramatic changes in product profile (19,22,34) or mainly to changes in stereochemistry (35). The most noticeable effect of the A451G mutation in eLOX3 was a switch toward more 9-H(P)ETE, increased from 28.9 Ϯ 1.4% (mean Ϯ S.E., n ϭ 4) to 38.5 Ϯ 1.0%, seemingly at the expense of 5-H(P)ETE, decreased from 21.7 Ϯ 0.8 to 11.6 Ϯ 1.0% (cf. Fig. 5, A and B). These differences were statistically significant (p Ͻ 0.01). Chiral HPLC analysis further revealed that the 9-H(P)ETE in the A451G FIGURE 5. Identification of products from reactions of wild-type eLOX3 and A451G eLOX3 with arachidonic acid. The incubation conditions were the same as in Fig. 4, B and D. A and B, RP-HPLC analysis of products from wild-type and A451G eLOX3, respectively. C, GC-MS analysis of TMS ether methyl ester derivative of hydrogenated 7-HETE product. Before hydrogenation and derivatization, the 7-HETE product had been further purified by SP-HPLC. eLOX3 reaction was 67% 9R (Fig. 6F), an improvement in stereospecificity compared with the racemic 9-H(P)ETE product of wild-type eLOX3 (Fig. 6E).
Dioxygenase Activity of eLOX3 with AA/Lyso-PA-Experiments using AA/lyso-PA as substrate, a molecule with a bulky ester on the arachidonate carboxyl, were used to gain insight into the possible "tail first" or "head first" substrate binding orientation (36,37). Reaction occurred at less than 0.1% of the rate with free arachidonic acid, and products arising mainly from C-13 hydrogen abstraction were formed (supplemental Fig. S4). These results strongly suggest that the tail first orientation of substrate binding is mainly associated with H-13 abstraction in the reaction of wild-type or A451G eLOX3 with AA/lyso-PA. It can be further inferred that the head first orientation is associated with H-7 abstraction and is the preferred orientation in the reaction of eLOX3 with free arachidonic acid.

Discovery of Dioxygenase Activity in eLOX3
This study demonstrates that eLOX3 in its active form is capable of oxygenating polyunsaturated fatty acids to the fatty acid hydroperoxides. As summarized in Fig. 7, eLOX3 oxygenates 9E,11Z,14Z-20:36 in a regio-and stereospecific fashion to give the 9S-hydroperoxide. Although the reaction with arachidonic acid is much slower and exhibits considerably less specificity (Fig. 7), the enzymatic nature of this reaction is confirmed by the formation of 7-HPETE, which cannot be formed nonenzymatically under the conditions used in this study. Such bis-allylic hydroperoxides are formed from the corresponding peroxyl radical only in the presence of a highly efficient hydrogen atom donor (38,39), which in this case is not present in the incubation buffer and thus must be part of the enzyme machinery, most likely the Fe(II)-H 2 O complex. eLOX3 joins manganese lipoxygenase (40) and the recently characterized mini-LOX from cyanobacterium Cyanothece sp. (41) in the ability to make a bis-allylic hydroperoxide.

Unusual Aspects of eLOX3 and Reasons for Its Elusive Dioxygenase Activity
Our study also reveals aspects of eLOX3 that may explain why its dioxygenase activity has remained undetected for years. (i) With fatty acid substrate alone, the reaction shows an unusually long lag phase. This ranged from 4 to 5 min with 9E,11Z,14Z-20:36 (Fig. 2B) to over 1 h with arachidonic acid (Fig. 3A). (ii) Compared with typical LOX reactions, high concentrations of fatty acid hydroperoxide are required to abolish FIGURE 6. Chiral HPLC analysis of individual HETE products from reactions of wild-type eLOX3 and A451G eLOX3 with arachidonic acid. A, wild-type eLOX3, 5-HETE methyl ester. B, A451G eLOX3, 5-HETE methyl ester. C, wild-type eLOX3, 7-HETE methyl ester. D, A451G eLOX3, 7-HETE methyl ester. E, wild-type eLOX3, 9-HETE methyl ester. F, A451G eLOX3, 9-HETE methyl ester. the lag phase. (iii) As known with other LOX enzymes (42)(43)(44), not every fatty acid hydroperoxide is equally effective in the enzyme activation. The best substrate for the hydroperoxide isomerase activity of eLOX3, 12R-HPETE, turns out to be only marginally effective. This is not a coincidence, as 12R-HPETE allows for efficient hydroperoxide isomerase cycling of eLOX3. (iv) At room temperature (as opposed to at 0°C), the primary oxygenation products occur only as transient intermediates, i.e. they will be further converted to epoxyalcohols and ketones by the hydroperoxide isomerase activity. Therefore, the reaction gives no signal with the usual spectrophotometric (UV 235 nm) or HPLC-UV assays.

Our Interpretations of the Unusual Aspects
The evidence points to a deficiency in the activation of eLOX3. Whereas 1 eq of fatty acid hydroperoxide activator is sufficient in the activation step of typical LOX, many equivalents are required in the activation step of eLOX3, hence the pronounced lag phase and the requirement for high concentrations of a fatty acid hydroperoxide activator.
The deficiency in the activation step of eLOX3 is somewhat relieved by the A451G mutation. A451G eLOX3 appears to stand at the midpoint between wild-type eLOX3 and typical LOX. On the one hand, A451G eLOX3 still shows a pronounced lag phase in its reactions with arachidonic acid in the absence of fatty acid hydroperoxides. On the other hand, the lag phase of A451G eLOX3 is much shorter than that of wildtype eLOX3 and can be eventually overcome by only trace amounts of fatty acid hydroperoxide present in the substrate preparation.

Explaining the Difference between eLOX3 and Typical LOX
When the hydroperoxide isomerase activity of eLOX3 was originally characterized, our group had speculated that the enzyme fails to oxygenate natural fatty acids because the active site iron has a lower redox potential than the iron in typical LOX (10). As already outlined in the Introduction, hydrogen abstraction on typical fatty acid substrates, which is the first step of LOX catalysis, calls for a moderately high redox potential, and conceivably the redox potential could be exceptionally low in eLOX3 due to some unfavorable protein environment of the enzyme active site. However, this is not immediately evident on inspection of the protein sequence, as all the typical conserved features of LOX enzymes, including the conserved amino acid ligands to the iron, are retained in eLOX3. Most importantly, this issue is now addressed as we have demonstrated in this study that arachidonic acid can be oxidized by eLOX3 (with H-7 preferentially abstracted by the iron complex), thus definitively ruling out this low redox potential possibility.
To account for the atypical dioxygenase activity of eLOX3, we invoke an alternative explanation, that oxygen access is limited in the enzyme active site. Because molecular oxygen is one of the two substrates in LOX catalysis, limited oxygen access to the reacting fatty acid radical intermediate would undoubtedly impede catalysis. In the extreme scenario, when molecular oxygen is not available at all either because oxygen access is completely blocked or because the atmosphere is rendered anaerobic, even though the initial C-H bond cleavage step may proceed normally, the fatty acid radical thus produced would eventually escape from the enzyme active site in the absence of reacting molecular oxygen, leaving the enzyme in the inactive ferrous state (Fig. 8). Such a single turnover event would be undetectable in typical enzyme assays. Under such circumstances, redox cycling can occur only if the ferrous enzyme is re-oxidized by hydroperoxide back to the ferric form (Fig. 8, dotted line), as seen classically in the anaerobic reaction of soybean LOX with linoleic acid and 13S-HPODE (45,46). However, under the usual aerobic conditions, such an extreme scenario hardly applies to typical LOX, indicating that molecular oxygen is readily accessible to the reacting fatty acid radical in their active site. For example, it is estimated that free radical leakage occurs only once per hundred turnovers in the reaction of soybean LOX-1 with linoleic acid (13). We suggest that limited oxygen access is the actual case in eLOX3, based on the following interpretations of the following evidence.
Carbon Radicals Escape from the Active Site during Dioxygenase Cycling-In the transformation of arachidonic acid at room temperature, the hydroperoxide isomerase activity is persistent throughout the reaction, and HPETE products fail to accumulate (cf. Fig. 4, A and C), indicating that the ferrous enzyme is constantly produced. Also, the reaction of eLOX3 with arachidonic acid is clearly driven by 13S-HPODE (Fig.  3A, the curve with two doses of 13S-HPODE), just as the anaerobic conversion of linoleic acid by soybean LOX-1 is driven by 13S-HPODE. These observations can be explained if a significant portion of fatty acid carbon radical intermediates escapes from the active site, which could be the result of lack of reacting oxygen (cf. Fig. 8). Although leakage of peroxyl radical instead of carbon radical is also conceptually plausible, we consider this rather unlikely, because the subsequent trapping of peroxyl radical by the enzyme is remarkably efficient as indicated by formation of the bis-allylic product, 7-HPETE.
Lack of Regio-and Stereo-specificity within the Active Site-The lack of oxygenation specificity with arachidonic acid is not entirely due to radical leakage and subsequent nonenzymatic oxygenation in solution. This is indicated by the appearance of 7-HPETE, which must be generated entirely within the enzyme active site. The R/S ratio of 73:27 in 7-HPETE must reflect a combination of the differences in oxygen supply to the two faces of C-7 and in trapping effi- Dioxygenase Activity of eLOX3 DECEMBER 17, 2010 • VOLUME 285 • NUMBER 51 ciency of the respective 7R-and 7S-peroxyl radicals. Indeed, it is likely that oxygen is added nonselectively to every reacting site on the delocalized pentadienyl radical (e.g. C-5, C-7, and C-9 when H-7 is abstracted), suggesting that oxygen supply is similarly low in every reacting site.
A451G Mutation Augments Dioxygenase Activity-A451G eLOX3 shows "improved" specificity in its reaction with arachidonic acid compared with wild-type eLOX3, in that both the proportion of 9-HPETE among all isomers and its R/S ratio increase considerably. We deduce that the Ala-to-Gly mutation opens up space and introduces an oxygen "pocket" that is aligned with the "9R" site on arachidonic acid.
eLOX3 Readily Oxygenates 9E,11Z,14Z-20:36-This synthetic fatty acid with extended conjugation is a much better substrate than arachidonic acid and is oxygenated regio-and stereospecifically to the 13S-hydroperoxide. This is consistent with the availability of an oxygen pocket in this unusual 9S position, which is four carbons removed from the site of hydrogen abstraction at C-13. Obviously, such a pocket or channel would play no role in the oxygenation of typical pentadienyl fatty acids that require oxygen to be delivered to sites that are two carbons removed from the site of hydrogen abstraction.

Connection between Oxygen Access and Enzyme Activation
The natural question at this point is how our proposed difference of eLOX3 from typical LOX, i.e. limited oxygen access, would lead to a deficiency in the enzyme activation step. A kinetic study of rabbit reticulocyte 15-LOX-1 suggests that molecular oxygen is in fact directly involved in the LOX activation step (47). We have evidence from comparison of soybean LOX-1 and eLOX3 that agrees with this proposition and further suggests that molecular oxygen is actually a driving force in the activation step (52). Therefore, it is not surprising that the activation step would be impaired when oxygen access in the active site becomes limited.

Concluding Remarks
In this study we have unveiled dioxygenase activity of eLOX3 with fatty acids and investigated both the usual and unusual aspects. The usual aspects suggest that the basic machinery for LOX catalysis is fully functional in eLOX3, whereas the unusual aspects of eLOX3 point to the possibility that eLOX3 has limited oxygen access in the active site. It may be useful to consider a spectrum of activities among LOX enzymes that are partly dependent on the availability of molecular oxygen within the respective enzyme active sites. At one end of the spectrum are the conventional LOX, exemplified by soybean LOX-1, in which dioxygenase activity is dominant. Toward the other end of the spectrum, with limited access to molecular oxygen and radical leakage from the active sites, are atypical LOX enzymes such as eLOX3, plant type-2 lipoxygenases (48,49), moss and maize LOX with hydroperoxide isomerase, and lyase activities (50,51).
As noted at the beginning of the Introduction, eLOX3 performs a vital function in epidermal barrier formation. Where do our new findings stand in relation to this physiological role, and what is their importance? When the hydroperoxide isomerase activity of eLOX3 was discovered originally, it was presumed that this atypical isomerase activity, acting on a hydroperoxide product of its partner 12R-LOX, represents the function of eLOX3 in vivo. As the evidence stands at present, we continue to favor this interpretation, although whether this dioxygenase activity in eLOX3 is physiologically relevant should be considered in future studies. In our view, the latent dioxygenase activity of eLOX3 is a reflection of mechanistic issues biasing eLOX3 toward hydroperoxide isomerase activity.