Stereoselective oxidation of regioisomeric octadecenoic acids by fatty acid dioxygenases.

Seven Z-octadecenoic acids having the double bond located in positions 6Z to 13Z were photooxidized. The resulting hydroperoxy-E-octadecenoic acids [HpOME(E)] were resolved by chiral phase-HPLC-MS, and the absolute configurations of the enantiomers were determined by gas chromatographic analysis of diastereoisomeric derivatives. The MS/MS/MS spectra showed characteristic fragments, which were influenced by the distance between the hydroperoxide and carboxyl groups. These fatty acids were then investigated as substrates of cyclooxygenase-1 (COX-1), manganese lipoxygenase (MnLOX), and the (8R)-dioxygenase (8R-DOX) activities of two linoleate diol synthases (LDS) and 10R-DOX. COX-1 and MnLOX abstracted hydrogen at C-11 of (12Z)-18:1 and C-12 of (13Z)-18:1. (11Z)-18:1 was subject to hydrogen abstraction at C-10 by MnLOX and at both allylic positions by COX-1. Both allylic hydrogens of (8Z)-18:1 were also abstracted by 8R-DOX activities of LDS and 10R-DOX, but only the allylic hydrogens close to the carboxyl groups of (11Z)-18:1 and (12Z)-18:1. 8R-DOX also oxidized monoenoic C14-C20 fatty acids with double bonds at the (9Z) position, suggesting that the length of the omega end has little influence on positioning for oxygenation. We conclude that COX-1 and MnLOX can readily abstract allylic hydrogens of octadecenoic fatty acids from C-10 to C-12 and 8R-DOX from C-7 and C-12.

A. fumigatus , all with oleate 8 R -DOX activities, and ovine COX-1 and MnLOX. To facilitate the analysis, we silenced the hydroperoxide isomerase (P450) activities of LDS by point mutation (5,8-LDS Cys1006Ser) or truncation of the carboxyl end (7, ( 27 ). The heme-dependent dioxygenases catalyze antarafacial hydrogen abstraction and oxygenation, whereas MnLOX catalyzes suprafacial abstraction and oxygenation ( 28 ). Our third goal was to determine how the chain length affected the oxidation of monoenoic fatty acids by the 8 R -DOX activity of LDS.
The effl uents from the columns were combined with isopropyl alcohol/water (3/2) from a second HPLC pump ( 33 ), and then introduced by electrospray into a linear ion trap mass spectrometer (LTQ, ThermoFisher). We mixed the column eluates with isopropanol/water (60/40) in a ratio of ‫ف‬ 2:1. This choice was based on the observed signal intensities of HpOME with mixing ratios of 11:1, 5:1, 2:1, 1.5:1, and 1:1. The last three ratios yielded COX transforms certain polyunsaturated C 20 fatty acids sequentially to PGG and PGH compounds, and arachidonic acid is the most important physiological substrate ( 16 ). This enzyme also oxygenates (8 Z ,11 Z ,14 Z )-20:3 to prostaglandins and Mead acid [(5 Z ,8 Z ,11 Z )-20:3], linoleic acid, and oleic acid to hydroperoxides (17)(18)(19). The 3D structure of COX-1 has been determined with polyunsaturated C 20 and C 18 fatty acids in the active site ( 20 ). Arachidonic and linoleic acids bind with the bisallylic methylene, from which hydrogen abstraction takes place, close to the catalytic Tyr-385, and the omega ends align with three phenylalanine residues.
Regioisomeric octadecenoic acids can be used to systematically determine the importance of the position of the double bond for hydrogen abstraction by fatty acid dioxygenases. These studies would be facilitated by methods for steric analysis of the products, either the hydroperoxides or the corresponding alcohols. Unfortunately, there is little information on chromatographic separation of enantiomers of HOME and HpOME.
Chiral phase-HPLC (CP-HPLC) is now routinely used for analysis of polyunsaturated hydroxy and hydroperoxy fatty acids, e.g., HETE and HpETE ( 24 ). Versatile matrices contain chiral selectors linked to cellulose or amylose, which are coated on silica, e.g., 3,5-dimethylphenyl carbamate (Chiralcel OD, Chiralpak AD, Reprosil Chiral-AM) and benzoate (Chiralcel OB). With different alcoholic modifi ers, these matrices can separate a large number of hydroxy and hydroperoxy metabolites ( 24 ), but they cannot be eluted with high pressure and do not tolerate certain solvents. A more stable alternative is a chiral selector of "Pirkle type" on silica (Reprosil Chiral NR 2 ), which can separate enantiomers of hydroperoxyeicosatetraenoic and hydroperoxyoctadecadienoic acids, as well as enantiomers of many drugs ( 25,26 ).
The fi rst goal of the present study was to develop a method for CP-HPLC separation of enantiomers of HpOME and to study their fragmentation during LC-MS 3 analysis. Racemic HpOME were obtained by photooxidation and by autoxidation. Seven octadecenoic acids were investigated [double bonds at positions (6 Z )-(13 Z ), except at (10 Z )], and the hydroperoxides were analyzed by CP-HPLC on Reprosil Chiral NR. Our second goal was to determine the position specifi city of fatty acid dioxygenases for hydrogen abstraction of regioisomeric octadecenoic acids. We chose the 8 R -DOX activities of LDS (7,8-LDS of G. graminis and 5,8-LDS of A. fumigatus ) and 10 R -DOX of The fatty acid hydroperoxides were reduced in some experiments to alcohols by treatment with triphenylphosphine or NaBH 4 .

CP-HPLC separation of enantiomers of HpOME
All pairs of investigated enantiomers of HpOME were resolved on CP-HPLC (Reprosil Chiral NR). The relative retention times of HpOME with trans double bonds are summarized in Table 1 . The elution order of hydroperoxides was S before R . The separations of four enantiomers from photooxidation of (12 Z ) -18:1 and eight enantiomers from autoxidation of (9 Z ) -18:1 are shown in Figs. 1A and 2 , respectively. For comparison, the separation of products formed from (12 Z )-and (13 Z )-18:1 by MnLOX are shown in Fig. 1B, C . This column also seemed to separate enantiomers of HpOME with cis double bond confi guration ( Fig. 2 and Table 1 ), and the elution order of 8 S -and 8 R -HpOME(9 Z ) was deduced with aid of the 8 R -DOX activity of LDS ( Fig. 2 , bottom chromatogram).
The chiral selection of HpOME was likely based on the absolute confi guration of the 1-hydroperoxy-2-propene elements (illustrated in Fig. 1D ) and not on the positions of these elements between C-6 and C-14 ( Table 1 ).

MS fragmentation of HpOME and KOME
The MS/MS spectra ( m/z 313 → full scan) confi rmed that HpOME were dehydrated to KOME as judged from the main ion at m/z 295 analogous to dehydration of other hydroperoxy fatty acids ( 25,37 ). Important ions in the MS 3 spectra ( m/z 313 → 295 → full scan) of HpOME are summarized in Tables 2 and 3 .
The MS 3 spectra of HpOME with the hydroperoxide group located at the omega end of the trans or cis double bonds showed a fragmentation pattern, which was consistently changed from 7-HpOME to 14-HpOME ( Table 2 ). signals of almost equal intensities, whereas the signals intensities of the ratios 5:1 and 11:1 were reduced by 30 and 80%.
The transfer capillary was heated to 315°C, the ion isolation width was set at 1.5 for anions of HOME ( m/z 297 → full scan) and 5 for anions of HpOME ( m/z 313 → full scan) and 1.5 at the fi nal selection of MS 3 analysis of HpOME ( m/z 313 → 295 → full scan). The collision energy was set at 1.7 V, and the ion tube lens at Ϫ 110 to Ϫ 130 V. We recorded fi ve microscans and used the Gaussian algorithm for peak smoothing (Xcalibur Software). Prostaglandin F 1 ␣ was infused for tuning.

Chemical oxidation and steric analysis
Photooxidation of 18:1 was performed with methylene blue in methanol as described ( 25 ). Autoxidation of (9 Z ) -18:1 was modifi ed from the method of Punta et al ( 34 ). 0.2 M oleic acid [with 5-10% ␣ -tocopherol (w/w) in some experiments] reacted under oxygen with 40 µM N-hydroxyphthalimide and 20 µM ceric ammonium nitrate in CH 3 CN at 37°C for a few days. The conversion was followed by TLC and LC-MS; additional ceric ammonium nitrate was added in some experiments. HpOME was reduced to alcohols with NaBH 4 , and methylated with CH 2 N 2 . A mixture of 7-HOME(8 E ) and 8-HOME(6 E ) was oxidized to 7-KOME and 8-KOME with Dess-Martin periodinane ( 35 ), and then separated and analyzed by NP-HPLC-MS. HpOME was reduced with NaBH 4 and methylated. Steric analysis of HOME methyl ester was performed by derivatization into ( Ϫ )-menthoxycarbonyl derivatives followed by gas chromatographic analysis of diastereomeric short chain fragments prepared by oxidative ozonolysis ( 36 ).   a HpOME( E ) were obtained by photooxidation. The retention volumes of the last eluting isomers were 8-9.5 ml (with retention times of 16-19 min) on the analytical Reprosil Chiral NR column. The absolute confi gurations of HpOME( E ) were determined by chemical methods after separation of enantiomers on a preparative Reprosil Chiral NR-R column.

Enzymatic oxidation
b HpOME( Z ) were obtained by oxidation with the 8 R -DOX of LDS and MnLOX.

Oxidation of octadecenoic acids by LDS, 10 R -DOX, and ovine COX-1
The preferred octadecenoic acids of 8 R -DOX and COX-1 were, as expected, (9 Z ) -18:1 and (12 Z ) -18:1, respectively, which were oxidized by hydrogen abstraction and insertion of molecular oxygen at C-8 and C-11 (and in a few percentages at C-10 and C-13, respectively). We fi rst evaluated the effect of shifting the double bond toward the carboxyl group. 8 R -DOX and COX-1 oxidized (8 Z )and (11 Z ) -18:1 by hydrogen abstraction and oxygenation at C-7 and C-10, respectively, but signifi cant oxygenation also occurred at C-8 of (8 Z )-18:1 and C-11 of (11 Z )-18:1. This was deduced from analysis of the corresponding HOME, as HpOME were largely reduced to alcohols by the peroxidase activity of microsomal and purifi ed COX-1. This suggests that these enzymes abstracted both allylic hydrogens. The oxidation of (8 Z ) -18:1 by 8 R -DOX is shown The MS 3 spectrum of 13-HpOME(11 E ) is illustrated in Fig.  3A . A plausible fragmentation mechanism, based on ketoenol tautomerism and ␣ -fragmentation, is summarized in Fig. 3B .
The MS 3 spectra of HpOME with their hydroperoxide groups located at the carboxylic side of the trans or cis double bond, e.g., from 6-HpOME(7 E ) to 13-HpOME(14 E ), could not be explained by a common fragmentation mechanism. Inspection of Table 3 suggests that hydroperoxides positioned from C-11 to C-13 yielded a similar fragmentation. The MS 3 spectrum of 12-HpOME(13 E ) is illustrative ( Fig. 4A ), and a plausible mechanism for its fragmentation is suggested ( Fig. 4B ). Hydoperoxides positioned at C-6 to C-9 yielded a series of unique and rather intense signal pairs (e.g., m/z 211 and 193, 197 and 179, 183 and 165, and 169 and 151, respectively; Table 3 ).  2. Separation by CP-HPLC of stereoisomers of HpOME formed by autoxidation of (9 Z ) -18:1 in the presence of ␣tocopherol. 8 R -HpOME(9 Z ) was added to identify this enantiomer. The chromatograms show selected ion monitoring of characteristic fragment ions (see Tables 2 and 3 ). Separation by CP-HPLC of stereoisomers of 12-HpOME(13 E ) and 13-HpOME(11 E ), which were formed by photooxidation of (12 Z ) -18:1. B: Oxidation of (12 Z ) -18:1 by MnLOX. The top chromatogram shows separation of 13 S -and 13 R -HpOME(11 E ), which were obtained by photooxidation. The middle and bottom chromatograms show analysis of 11 R -HpOME(12 Z ) and 13 R -HpOME(11 E ) formed by MnLOX. C: Oxidation of (13 Z ) -18:1 by MnLOX. The top chromatogram shows separation of 14 S -and 14 R -HpOME (12 E ), which were obtained by photooxidation. The middle and bottom chromatograms show analysis of 12 R -HpOME(13 Z ) and 14 R -HpOME(12 E ), which were formed by MnLOX. D: Comparison of HpOME with 3 S -hydroperoxy-1-propene (top) and 1 S -hydroperoxy-2-propene elements (bottom, reverse orientation). It seems likely that the chiral NR selector interacts in a similar way with the two elements. Panels A-C show selected ion chromatograms as indicated by insets of the selected ions. demonstrated that (9 Z )-16:1, (9 Z )-20:1, and (11 Z )-20:1 were also oxygenated ( 27 ). These results, and the effect of the position of the double bond on hydrogen abstraction, are summarized in Fig. 7 .

Separation of stereoisomers of HOME
An advantage of CP-HPLC analysis of hydroperoxides on Reprosil Chiral NR was resolution of all studied HpOME enantiomers. In contrast, enantiomers of HOME were not generally resolved on this column.
The resolution of enantiomers of HOME required assessment of different CP-HPLC columns and alcoholic modifi ers. We evaluated Reprosil Chiral AM, eluted with 5% ethanol or 5% methanol in hexane, and Chiralcel OB-H, eluted with 5% isopropanol in hexane ( 38 ). A systematic investigation on resolution of HOME by CP-HPLC was beyond the scope of this investigation, but the observations described below are illustrative.
The Reprosil Chiral AM column with methanol as modifi er readily baseline separated R and S stereoisomers of 10-HOME(11 Z ) (see Fig. 6B ), 11-HOME(12 Z ), and 12-HOME(13 Z ). The elution order was as expected, R before S ( 24 ). This was deduced from reference compounds based on the suprafacial oxygenation by MnLOX (to R enantiomers) and the antarafacial oxygenation by COX-1 (to S enantiomers). Under these conditions, 11 R -HOME(12 Z ) eluted before the stereoisomers of 12-HOME(13 E ), which were well separated, whereas 11 S -HOME(12 Z ) coeluted with the fi rst eluting isomer of 12-HOME(13 E ). The stereoisomers of 13-HOME(11 E ) were not resolved. Ethanol as alcoholic modifi er did not improve the situation, as enantiomers of HOME( E ), which were obtained by photooxidation of (11 Z ) -, (12 Z ) -, and (13 Z ) -18:1, were not resolved.
Chiralcel OB-H resolved the isomers of 11-HOME(12 E ) and 12-HOME(13 E ). Separation of HOME with 1-hydroxy-2-propene elements seemed less demanding than separation of HOME with 3-hydroxy-1-propene elements. The Chiralcel OB-H column thus resolved the isomers of 13-HOME(11 E ) but not the isomers of 12-HOME(10 E ).

Oxidation of monoenoic C 14 , C 16 , and C 20 fatty acids by 8 R -DOX
The 8 R -DOX of 7,8-LDS_1-764 rapidly oxidized (9 Z ) -14:1 to its 8-hydroperoxy metabolite. Previous work has  (5) ND, not detected; -, below the range of the instrument. a These ions (X series) were postulated as intermediates, which likely formed strong fragments ions after loss of water but only weak fragments after loss of CO 2 .
b These ions (Z series) contained the omega end (see text for details). c This ion occurs in two columns and could be formed by different fragmentation mechanisms.  ND, not detected. a The mass spectra of the corresponding hydroperoxides with a cis double bond were virtually identical ( 19 ). b These ions (X series) were postulated as intermediates, which formed strong fragment ions after loss of water or CO 2 in some of the spectra.
c These ions (Y series) were formed by loss of the omega end. d This ion occurs in two columns and could be formed by two mechanisms. The base peaks of the MS 3 spectra were usually as listed in the table, except m/z 251 (313-18-44) for 9-HpOME and m/z 277 (313-44) for 6-HpOME. e These ions (Z series) contained the omega end and gave rise to intense signals after loss of water in some spectra. Fig. 3. MS 3 spectrum of 13-HpOME(11 E ) and an interpretation of the fragmentation mechanism. A: MS 3 spectrum ( m/z 313 → 295 → full scan). B: A hypothetical fragmentation mechanism: 13-HpOME is dehydrated to 13-KOME, and the latter might fragment as indicated. The fragment ions are found in the X and Z series of ions in Table 2 .

DISCUSSION
We demonstrated that enantiomers of HpOME can be conveniently separated by CP-HPLC on the Reprosil Chiral NR matrix and that the elution order was consistently " S " before " R ". The chiral selector has not been disclosed, but it is described as an aromatic chiral phase withdonor and -acceptor groups of Pirkle type. This column can be operated in both reverse and normal phase modes, but separation of enantiomers of oxygenated fatty acids has so far been reported only in normal phase mode. Reprosil Chiral NR also resolves the enantiomers of Fig. 4. MS 3 spectrum of 12-HpOME(13 E ) and an interpretation of the fragmentation mechanism. A: MS 3 spectrum ( m/z 313 → 295 → full scan). B: A hypothetical fragmentation mechanisms by which 12-HpOME is dehydrated to 12-KOME and the latter fragments as indicated (see Fig. 3B ). The fragment ions are found in the X and Y series of ions in Table 3 .  Table 3 and its series of Z ions).
The absolute confi guration of enantiomers of HpOME obtained following chromatography on Reprosil Chiral NR was determined by chemical methods and by comparison with enzymatic products described above. The chemical method was based on ( Ϫ )-menthoxycarbonyl derivatives of the corresponding HOME methyl esters, followed by ozonolysis of the double bond, methylation, and separation of the diastereoisomers by gas chromatography ( 36 ). This procedure showed that the S stereoisomers of 14 regioisomeric HpOME with the double bond in trans confi guration eluted before the R stereoisomers on Reprosil Chiral NR. As far as it is known, this elution order also applies to HpOME( Z ) ( Table 1 ).
Decisive for chiral separation is interaction of the analyte to the chiral selector at a minimum of three positions with at least one position dependent on stereochemistry ( 39 ). The two oxygens of the hydroperoxide group and the double bond seem to interact with the chiral selector in this way. HpOME with a 1 S -hydroperoxy-2-propene element eluted before the enantiomer having the 1 R -hydroperoxy-2-propene  element, and the position of this hydroperoxy-propene unit along the carbon chain from C-6 to C-14 did not change the elution order ( Fig. 1D and Table 1 ). Interestingly, chiral separation was lost when the hydroperoxide group of HpOME was replaced by an alcohol. This seems to be a unique property of Reprosil Chiral NR compared with chiral selectors based on derivatized amylose ( 24 ). Enantiomers of 10-HOME(8 Z ) thus interacted with the chiral NR selector without separation, but the stereoisomers of 10-hydroxy-8 E ,12 Z -octadecadienoic acid, with an addition double bond at 12 Z , were resolved ( 35 ), presumably by three-point interaction of the hydroxyl group and the two double bonds with the chiral selector. It follows that enantiomers of polyunsaturated fatty acid hydroperoxides with 1-hydroperoxy-2 E ,4 Z -pentene elements and additional double bonds (e.g., HpETE) may not necessarily elute in the same " S " and " R" order as HpOME, 9-HpODE, and 13-HpODE (see Ref. 25 ).
COX-1 and the oleate 8 R -DOX activities of 7,8-LDS, 5,8-LDS, and 10 R -DOX initiate oxidation of octadecenoic acids by hydrogen abstraction, which likely is catalyzed by a tyrosyl radical ( 16,40,41 ). The selective oxidation of three octadecenoic acids by ovine COX-1 proved that C-11 of (12 Z ) -18:1, C-12 of (13 Z ) -18:1, and both C-10 and C-13 of (11 Z ) -18:1 were positioned suffi ciently close to Tyr-385 for hydrogen abstraction. This was not unexpected, as the 3D structure of COX-1 with linoleic acid in the active site shows that C-11 to C-13 are within relatively short distances from Tyr-385 ( 20 ). A double bond at the 11 Z , 12 Z , or 13 Z position seems to be required for correct positioning, as (9 Z )-18:1 was not oxidized by COX-1.
The 8 R -DOX domain of 7,8-LDS, 5,8-LDS, and 10 R -DOX readily oxidized 18:1 with double bonds at position from (8 Z ) to (11 Z ). Signifi cant oxidation by LDS also occurred at the (12 Z ) position. We do not know whether the active site of 8 R -DOX mimics the COX-1 "tail fi rst" orientation of fatty acids autoxidation. Enantiomers of HpOME can also be separated by CP-HPLC, and HpOME of biological origin can readily be analyzed for stereoisomers by this method.
in the active site. A critical factor for 8 R -DOX catalysis is the presence of a saturated carbon chain of 6-10 carbons between the carboxyl group and the fi rst double bond ( 13,27,42 ), but the total chain lengths of monounsaturated fatty acids from 14:1 and 20:1 or additional double bonds at the end do not affect catalysis.
It is interesting to compare 8 R -DOX with lipoxygenases. Newcomer and coworkers propose from crystal studies of arachidonate 8 R -LOX of Plexaura homomalla that all lipoxygenases may bind their substrates in a U-shaped channel ( 43 ). Each lipoxygenase may have one end of the Ushaped channel closed with the carboxylate of the substrate exposed to solvents at the entrance of the open channel with the carbon chain in the U-shaped channel. MnLOX likely binds fatty acids with the carboxylates at the entrance, and the omega ends could be buried in a substrate channel ( 44 ). Analogous to COX-1, MnLOX oxidized (11 Z ) -18:1, (12 Z ) -18:1, and (13 Z ) -18:1 to hydroperoxides at the allylic carbon of the carboxyl ends, albeit with R confi guration ( 19 ). 8 R -DOX may also bind their substrates in this way, but we cannot exclude that the carboxylate could be buried in the interior analogous to ␣ -DOX ( 45 ).
LC-MS is a convenient and powerful tool for identifi cation of oxylipins. The fragmentation of a large number of hydroxy and dihydroxy fatty acids during MS/MS analysis has now been characterized ( 37 ). Informative fragments are often obtained by ␣ -cleavage at the oxidized carbons. Hydroperoxy fatty acids are dehydrated to keto fatty acids during MS/MS analysis ( 37,46 ). We recorded the MS 3 spectra of HpOME and found that keto-enol tautomerism and ␣ -fragmentation could explain many fragments, as illustrated by the MS 3 spectra of 12-HpOME(13 E ) and 13-HpOME(11 E ) ( Figs. 3 and 4 ).
The MS 3 fragmentation of 6-HpOME, 7-HpOME, 8-HpOME, and possibly 9-HpOME differed from the other HpOME with 1-hydroperoxy-2-propene elements. In all four spectra, one of the most intense ions seemed to be formed by rearrangement with loss of a short chain of 5 to 8 carbons and oxygen transfer to the end ( Table 3 ). We confi rmed that 7-KOME(8 E ) and 7-HpOME(8 E ) yielded identical MS 2 and MS 3 spectra, respectively. The fragmentation of HpOME with 1-hydroperoxy-2-propene elements thus varies with the position. For comparison, 7,10-DiHOME(8 E ) fragments in a different way than 5,8-DiHOME, 7,8-DiHOME, and 8,11-DiHOME ( 47 ). If needed, GC-MS analysis can usually be used to identify regioisomeric HOME and other oxylipins without ambiguity ( 48 ). The MS spectra with electrospray ionization can be less informative than electron impact MS spectra. Comparison of the LC-MS spectra of two different compounds in Figs. 3A and 5A illustrates this point.
We conclude that 8 R -DOX and COX-1 oxygenate octadecenoic acids to HpOME with stereoselectivity, provided the double bond is located at positions (8 Z ) to (12 Z ) for 8 R -DOX and at positions (11 Z ) to (13 Z ) for COX-1. Octadecenoic acids can thus be used to determine the positional effect of a double bond on enzymatic catalysis. Racemic HpOME standards are readily available by photo-or