Characterization and biological effects of di-hydroxylated compounds deriving from the lipoxygenation of ALA

to ALA relevant because Abstract We have recently described a di-hydroxylated compound called protectin DX (PDX) which derives from docosahexaenoic acid (DHA) by double lipoxygenation. PDX exhibits anti-aggregatory and anti-inﬂ ammatory properties, that are also exhibited by similar molecules, called poxytrins, which possess the same E , Z , E conjugated triene geometry, and are synthesized from other polyunsaturated fatty acids with 22 or 20 carbons. Here we present new biological activities of di-hydroxylated metabolites deriving from (cid:2) -linolenic acid (18:3n-3) treated by soybean 15-lipoxygenase (sLOX). We show that 18:3n-3 is converted by sLOX into mainly 13(S)-OH-18:3 after reduction of the hydroperoxide product. But surprisingly, and in contrast to DHA which is metabolized into only one di-hydroxylated compound, 18:3n-3 leads to four di-hydroxylated fatty acid isomers. We here the complete characterization of these compounds using high ﬁ eld NMR and GC-MS techniques, and some of their biological activities. These compounds are:


NMR of 9,16-di-HOTEs
The NMR experiments were performed on a Bruker-Biospin NMR spectrometer operating at 23.4 Tesla (1,000 MHz of Proton Larmor frequency) and equipped with a Cryoprobe. All the spectra were acquired in a CDCl 3 solution, working at 298 K and referencing the chemical shifts to tetramethylsilane as an internal standard. Standard sequences were used for 1D, DQF COSY (26)(27)(28), and J-resolved ( 29-31 ) spectra. The 1 H pulse lengths of 9.0 s, 1 s, and 5 s were used as recycle delay for J-resolved and COSY spectra . A weak selective decoupling of 270 Hz of power during the acquisition time (900 ms) was used in 1D selective homo-decoupling experiments to simplify some coupling patterns. 2D DQF COSY was acquired with 8 scans and 2,048 × 1,024 real points (F2 × F1) with acquisition times of 114 ms and 57 ms for the direct and indirect dimensions. J-resolved 2D was acquired with 16 scans and 8,192 × 256 real points, acquiring 458 ms in the direct dimension and 1,280 ms in the indirect J-coupling dimension. COSY spectra were processed with a 4,096 × 4,096 point matrix using square cosine and square sine window function for the direct and indirect dimensions. J-resolved 2D spectra were processed with an 8,192 × 2,048 point matrix with square cosine window function for both the dimensions.

Preparation of cell suspensions
Platelet suspensions. Venous blood from healthy volunteers who had not taken any medication for at least one week was collected at the local blood bank (Etablissement Français du Sang). Blood was drawn onto anticoagulant and centrifuged at 200 g for 15 min at 20°C to obtain the platelet-rich plasma upper phase, and platelets were isolated as described previously ( 34 ). Briefl y, platelet-rich plasma was acidifi ed to pH 6.4 with 0.15 M citric acid and immediately centrifuged at 900 g for 10 min, and platelet pellets were suspended into Tyrode-HEPES buffer (140 mM NaCl, 2.7 mM KCl, 1.0 mM MgCl 2 , 12 mM NaHCO 3 , 0.41 mM NaH 2 PO 4 , 5 mM HEPES, and 5.5 mM glucose, pH 7.35).
Leukocyte suspensions. Leukocytes were isolated from blood obtained from Etablissement Français du Sang, as previously described ( 35 ). Dextran (2% fi nal concentration) was added to the lower phase obtained after the fi rst blood centrifugation at 200 g . Erythrocytes were then sedimented under gravity for 45 min at 15°C. The erythrocyte-depleted supernatant containing leukocytes this fatty acid is taken in large amounts (1-2 g/day in human adults) as a nutrient and hardly accumulates in plasma, suggesting an active metabolism. Moreover, its conversion into EPA and DHA is relatively limited. A conversion of ALA into its long-chain product DHA was found to be around 1% in men ( 20 ) and 2% in young women ( 21 ). Formation of 9,16-dihydroxyoctadecatrienoic acid (9,16-di-HOTE) isomers from ALA by sLOX-2 have already been reported fi rst in 1984 by Feiters et al. ( 22 ) without establishing the double bond confi guration or the stereochemistry of the asymmetric carbons. Further studies presented by Sok and Kim ( 23,24 ) reported four isomers, but again their characterization was not fully established, especially that of the all trans isomers. In 1991 Grechkin et al. ( 25 ) showed that potato tuber lipoxygenase (LOX) generated mainly 9(S)-hydroperoxy-octadeca-10 E ,12 Z ,15 Z -trienoic acid [9(S)-HpOTE] which is different from the sLOX which produces mainly 13(S)-hydroperoxy-9 Z ,11 E, 15 Zoctadecatrienoic acid [13(S)-HpOTE]. Also, two compounds after reduction by NaBH 4 were characterized as 9,16-di-HOTE (10 E ,12 Z ,14 E ) except for the stereochemistry of the carbon 16 which remains undetermined and trans isomers were not detected. Moreover, the biological properties of the 9,16-di-HOTE isomers have not been investigated yet. Because ALA is the most abundant n-3 PUFA in human diet, it is relevant to investigate more thoroughly its oxygenated metabolism, especially by 15-LOX that is expressed in blood leukocytes and most endothelial cells, and compare the biological activity of the metabolites to that of PDX, the main 15-LOX product of DHA.

Purifi cation of mono-and di-hydroxylated fatty acids
Mono-and di-hydroxylated fatty acids were analyzed by RP-HPLC on a Waters XBridge C18 column (4.6 × 250 mm, 3.5 m) to pH 3 with acetic acid. The oxygenated metabolites were then extracted as previously described ( 18 ).

GC-MS measurement of purifi ed COX-1 and COX-2 products
Commercial COX-1 (40 units) and COX-2 (40 units) were preincubated for 15 min in a 0.1 M Tris-HCl buffer pH 8.1 with or without 1 M of different di-hydroxylated metabolites and incubated for 10 min with 10 M arachidonic acid used as a substrate. Albumin (45 g/l) was added at the end of incubation to facilitate the formation of PGD 2 and PGE 2 from PGH 2 ( 37 ). After acidifi cation to pH 3 with acetic acid, PGE 2 and PGD 2 were extracted by 10 vol of diethyl ether stabilized by 10% ethanol. Deuterated PGD 2 and PGE 2 used as internal standards were added for the quantifi cation. Prostaglandins were derivatized into methoxime, pentafl uorobenzyl esters, and TMS and further analyzed by negative-ion chemical ionization GC-MS. PGD 2 and PGE 2 were measured using the SIM mode. Selected ions corresponding to [M-181] Ϫ (loss of the pentafl uorobenzyl group): m/z 524 for both derivatized PGD 2 and PGE 2 and m/z 528 for their corresponding deuterated internal standards ( 38,39 ) were measured.

Statistical analysis
All results are expressed as means ± standard error of the mean (SEM) or means ± standard deviation (SD). P < 0.05 was set as the level of signifi cance as determined by two-tailed Student's t -tests. One-way ANOVA analysis was used to assess differences between groups.
was centrifuged at 400 g for 10 min. The resulting leukocyte-rich pellets were resuspended into saline buffer, and layered over a Ficoll-Paque Plus (density 1.077). After centrifugation, the fraction above the erythrocyte layer was removed and centrifuged, and resuspended in water for 20 s to lyse the contaminant erythrocytes. Osmotic pressure was immediately restored with NaCl, followed by centrifugation, which provided a white pellet consisting of whole leukocytes that was resuspended into Tyrode-HEPES buffer. The morphological examination and trypan blue exclusion tests were performed to determine the cell count and purity, and evaluate the viability of neutrophils.

Measurement of platelet aggregation and incubation with human leukocytes
Blood platelet aggregation was determined by the turbidimetric method of Born ( 36 ) using a dual aggregometer (ChronoLog, Havertown, PA). To measure changes in the light transmission rate, 400 l of platelet suspension was heated at 37°C for 1 min with stirring, incubated in the presence or absence of di-hydroxylated compounds (added in 1 l of ethanol) for one more minute, and then with collagen as a platelet aggregatory agent for 4 min at 37°C. Concentrations of collagen were adjusted to obtain nearly 75% aggregations for control platelets.
Isolated human leukocytes, suspended in a Tyrode-HEPES buffer containing 2 mM Ca 2+ were preincubated in the presence or absence of 1 M of different 9,16-di-HOTE isomers for 3 min at 37°C, and triggered with 1 M ionophore A23187 plus 10 M arachidonic acid for 10 min. The reaction was stopped by acidifying the media

Localization by GC-MS of hydroxyl groups on the fatty chain
Mono-and di-hydroxylated compounds from ALA mentioned above were hydrogenated and derivatized as methyl esters and TMS ethers as described in Materials and Methods and analyzed by GC-MS using the EI mode.
The EI mass spectrum of the hydrogenated monohydroxylated ALA derivative ( Fig. 2A ) shows characteristic ions at: m/z 371 (M-15) loss of CH 3 , m/z 315 (M-71) which corresponds to the loss of CH 2 -(CH 2 ) 3 -CH 3 , and m/z 173 [TMS-O-CH-(CH 2 ) 4 -CH 3 ]. Such a fragmentation pattern clearly indicates the presence of the hydroxyl group on carbon 13 of the fatty chain. So ALA was mainly converted into 13(S)-HpOTE , further reduced by NaBH 4 into hydroxylated product for analytical purposes.
All four hydrogenated di-hydroxylated compounds from ALA exhibited the same mass spectra ( Fig. 2B ). Their fragmentation pattern shows the following char-

Characterization of metabolites obtained from ALA incubated with sLOX
From ALA incubated with sLOX as described in Materials and Methods, hydroxylated derivatives analyzed by RP-HPLC provided one main compound detected at 50 min with a max of 235 nm ( Fig. 1A ). Its retention time and its ultraviolet (UV) spectrum were similar to that of commercial 13(S)-HOTE (result not shown).
Regarding di-hydroxylated products ( Fig. 1B ), and in contrast to DHA which is only converted into one major dihydroxylated metabolite called PDX, four di-hydroxylated isomers were detected from ALA. They all exhibited characteristic UV spectra with a maximum absorption ( max ) at 270 nm with two shoulder peaks at 260 nm and 280 nm, indicating the presence of a conjugated triene. Moreover, the UV spectra of the fi rst two isomers eluted at 32.1 and 32.6 min are superimposable to that of commercial 12-epiall trans-leukotriene B 4 (LTB 4 ) with an E , E , E conjugated triene geometry (result not shown), whereas the UV spectra of the two other isomers eluted at 33.2 and 33.5 min are superimposable to that of PDX, suggesting an E , Z , E geometry of these conjugated trienes ( Fig. 1B ). The spectra of these four isomers differ from those with a Z , E , E confi guration such as LTB 4 and 12-epi-LTB 4 (result not shown). from those with ALA as a substrate ( Fig. 4 ). As a matter of fact, another major E , Z , E product (retention time 33.5 min, see also Fig. 1B ) is likely 9(R),16(S)-di-HOTE. In addition, the same compound appeared when the racemic 9(±)-HOTE was incubated with sLOX ( Fig. 3B , retention time 33.7 min).

NMR determination of the conjugated triene double bound geometry
A NMR investigation was performed to determine the E / Z confi guration in the triene moiety. A change in the E / Z confi guration signifi cantly modifi es the 1  Determination of carbon stereochemistry of di-hydroxylated metabolites synthesized from ALA Commercial 9(S)-HOTE and racemic 9(±)-HOTE, prepared as described in Materials and Methods, were incubated separately with sLOX as described with ALA, and further reduction by NaBH 4 . 9(S)-HOTE was metabolized into a main compound eluted at 33.4 min ( Fig. 3A ). This peak was attributed to 9(S),16(S)-di-HOTE because sLOX produces almost exclusively the S enantiomer via a double dioxygenation mechanism ( 40 ), although soybean LOX generates almost exclusively 13(S)-HpOTE from ALA ( 40 ) as it was already shown ( Fig. 1A ). On the other hand, the 9(R),16(S)-di-HOTE peak, eluting at 33.7 min, was also formed ( Fig. 3B ) when racemic 9(±)-HOTE was incubated with sLOX, compared with 9(S)-HOTE. The two diastereoisomers 9(S),16(S)-di-HOTE and 9(R),16(S)-di-HOTE eluted at 33.4 min and 33.7 min ( Fig. 3B ), respectively, exhibited the same intensity in agreement with the equivalent amounts of 9(S)-and 9(R)-HOTE present in the racemic mixture. The latter chromatogram ( Fig. 3B ) is similar to that presented in Fig. 1B , suggesting that the two peaks (retention time 33.2 and 33.5 min, Fig. 1B ) observed with ALA as a substrate correspond to the two diastereoisomers 9(S),16(S)di-HOTE and 9(R),16(S)-di-HOTE, respectively.
This assignment was also confi rmed by the fact that chromatograms of di-hydroxylated isomers from 9(S)-HpOTE and 9(S)-HOTE were almost identical, but different shift. Nevertheless, because the determination of the confi guration in the E , Z , E molecules is demonstrated above, we can safely assign by exclusion the confi guration of the E , E , E molecules on the basis of the marked differences among the spectra of E , E , E and E , Z , E .

Metabolism of ALA by human recombinant 15-LOX-2
This experiment was done to check whether the human recombinant 15-LOX type 2 is also able to synthesize such metabolites. For this purpose ALA, 9(S)-HOTE, and 9(±)-HOTE were separately incubated with this enzyme. We found that human recombinant 15-LOX-2 is able to convert ALA into 13-HOTE as previously described with the sLOX (result not shown). However, ALA was less markedly metabolized into di-hydroxylated fatty acids, the main ones being all trans 9,16-di-HOTE isomers. Interestingly enough, the mono-hydroxylated derivatives 9(S)-HOTE and 9(R)-HOTE were well converted into 9(S),16(S)-E , Z , Edi-HOTE and 9(R),16(S)-E , Z , E -di-HOTE, respectively (result not shown) by the human recombinant LOX.

Inhibitory effect of 9,16-di-HOTEs on COX-1 and antiaggregatory effects
The inhibition of sheep COX-1 by different 9,16-di-HOTE isomers was tested at 1 M by measuring both PGD 2 and PGE 2 as described in Materials and Methods. Results reported in Table 1 show that 1 M of 9(S),16(S)-dihydroxy-10 E ,12 Z ,14 E -octadecatrienoic and 9(R),16(S)-dihydroxy-10 E ,12 Z ,14 E -octadecatrienoic acids decreased signifi cantly the synthesis of both PGD 2 and PGE 2 by around 28% and 38%, respectively. Such results are in agreement with previous data showing that poxytrins, characterized by a E , Z , E (33.5 min, Fig. 1B ), which only differ for the inversion of the stereochemical confi guration at the chiral center of position 9, should show very similar spectra with minor differences in chemical shift and J-couplings, especially in the triene region. On this basis, the inspection of the 1D NMR spectra, and in particular of the triene resonance region (4-7 ppm), allows us to rapidly distinguish among the ensemble of molecules having the same triene confi guration [e.g., the E , E , E compounds (32.1 min and 32.6 min, Fig. 1B ) from the E , Z , E compounds (33.2 min and 33.5 min, Fig.1B )], as it can be seen in Figs. 5 and 6 . The resonance assignment has been determined by using DQF COSY, while a large part of the J-couplings were measured through the analysis of 2D J-resolved spectra (data not shown). In such a way, it was possible to determine that the  Such results are in agreement with data relating to platelet aggregation induced by collagen. Platelet aggregation was signifi cantly decreased by 64% and 65% in presence of 1 M 9(S),16(S)-dihydroxy-10 E ,12 Z ,14 E -octadecatrienoic acid and 9(R),16(S)-dihydroxy-10 E ,12 Z ,14 E -octadecatrienoic acid, respectively ( Fig. 7 ). In addition, the aggregation started to be reversed from 2 min following the agonist addition. As previously shown ( 19 ), all trans isomers had no effect on platelet aggregation.

Anti-infl ammatory effect of 9,16-di-HOTEs
After showing that 9,16-dihydroxy-10 E ,12 Z ,14 E -octadecatrienoic acid isomers possess anti-aggregatory properties as those previously described for poxytrins ( 19 ), it was important to know whether such di-hydroxylated compounds issued from ALA may exhibit anti-infl ammatory properties. We then investigated the effects of 9,16-di-HOTEs on COX-2 as well as the 5-LOX pathway.
Effect of 9,16-di-HOTEs on the metabolism of arachidonic acid by the human polymorphonuclear leukocyte 5-LOX. The inhibition of the human 5-LOX pathway by 1 M of 9,16-di-HOTE isomers was assessed by measuring the following metabolites: 5(S)-HETE, LTB 4 and its two all trans isomers. As reported in Table 2 , only the 9(R),16(S)-dihydroxy-10 E ,12 Z ,14 E -octadecatrienoic acid isomer signifi cantly decreased the formation of leukotrienes and 5(S)-HETE by around 10% and 20%, respectively. All other isomers did not inhibit such a production (results not shown). Such data suggest that the inhibition concerns only 5-LOX but not leukotriene A 4 hydroxylase. Taken together, these data suggest that the 9(R),16(S)-dihydroxy-10 E ,12 Z ,14 E -octadecatrienoic acid isomer may have a potential anti-infl ammatory effect.

DISCUSSION
ALA is an essential fatty acid present in vegetable oils, and as such available in human diet. It is the precursor of EPA and DHA, but its conversion into EPA and DHA is rather limiting with around 10% to the former and a further 10% to the latter, which means only 1% conversion of ALA to DHA ( 47,48 ). ALA is also rather well ␤ -oxidized ( 49,50 ), that could partially explain its low accumulation in plasma and tissues. In addition, we describe here a pathway which could explain some benefi cial effects of ALA in the vascular system at large by reducing both the Purifi ed COX-1 and COX-2 were incubated with or without 1 M of different 9,16-di-HOTE isomers and with 10 M arachidonic acid. PGD 2 and PGE 2 were measured by GC-MS. Results expressed in ng of PGD 2 and PGE 2 represent the mean ± SEM of four determinations. Only the inhibitory effects of E,Z,E isomers are shown. a P < 0.05. b P < 0.01 versus control.  Isolated human leukocytes were incubated in the presence or absence of 1 M of different 9,16-di-HOTE isomers at 37°C and triggered with 1 M of ionophore A23187 and 10 M of arachidonic acid. Oxygenated metabolites were then extracted and quantifi ed by RP-HPLC. Results expressed in nmol of total LTB 4 s (LTB 4 + LTB 4 isomers) and 5-HETE represent the mean ± SEM of fi ve determinations. Only the inhibitory effect of 9(R),16(S)-di-HOTE is shown. a P < 0.05, versus control. production of pro-infl ammatory eicosanoids PGD 2 /PGE 2 and LTB 4 (51)(52)(53). We recently showed that a new di-hydroxylated metabolite from DHA called PDX, which is an isomer of the anti-infl ammatory PD1, exhibits anti-aggregatory properties, both by inhibiting collagen-induced prostanoid production, including thromboxane A 2 , and thromboxane A 2 -induced aggregation ( 19 ). In the present paper we report the full characterization of the lipoxygenation of ALA, and the anti-infl ammatory and anti-aggregatory effects of two LOX end-products. The structure of the four LOX metabolites that we herein describe have already been partially reported ( 23,24 ), but the proposed stereochemistry structures of the 9(S),16(R)-di-HOTE do not fi t with our experimental data. On the other hand, their biological properties have not been investigated.
Our data show that ALA is mainly converted into its 13oxygenated derivative 13-HOTE by 15/ 6-LOX. However, formation of a small amount of the 9-oxygenated product 9-HOTE could also be observed. In contrast to DHA, which is converted into only one main di-hydroxylated compound (PDX) via a double lipoxygenation, four di-hydroxylated compounds were observed from ALA following incubation with 15-LOX, characterized as 9,16-di-HOTEs. The stereochemistry of these di-hydroxylated compounds was assessed by incubating 9(S)-HOTE and racemic 9(±)-HOTE separately with sLOX (see the above results). The UV spectra of the two main di-hydroxylated products are superimposed to that of PDX, which possesses a E , Z , E conjugated triene geometry. The diastereoisomers are likely produced via a dioxygenation mechanism as previously described for PDX ( 18 ) and these two compounds are likely to have an E , Z , E conjugated triene geometry. Such a geometry was further confi rmed by high fi eld NMR which allowed us to determine precisely the 1 H-1 H J-couplings among the protons bound to the sp 2 carbons in C=C double bonds. The geometry of the conjugated trienes of two additional but less abundant (around half the amount of the E , Z , E -9(S),16(S)di-HOTE and 9(R),16(S)-di-HOTE) di-hydroxylated products was supposed to be E , E , E because their UV spectra were superimposed to that of all trans LTB 4 , which has such an E , E , E conjugated triene geometry. This all trans geometry was also confi rmed by NMR to determine the 1 H-1 H J-couplings among the protons bound to the sp 2 carbons in C=C double bonds. Based on data from the literature, we may suggest that the stereochemistry of these two minor products ( Fig. 1B ) is 9(R),16(S)-and 9(S),16(S)-all trans-di-HOTE, in order of increasing retention time, respectively. Such a confi guration is also in agreement with the 5(S),12(R)-and 5(S),12(S)-all trans LTB 4 and the 5(S),12(R)-and 5(S),12(S)all trans LTB 5 diastereoisomers already described (43)(44)(45). We conclude that the four di-hydroxylated compounds issued from ALA treated by sLOX are 9(R),16(S)-dihydroxy-10 E ,12 E , 14 E -octadecatrienoic acid, 9(S),16(S)-dihydroxy-10 E ,12 E ,14 Eoctadecatrienoic acid, 9(S),16(S)-dihydroxy-10 E ,12 Z ,14 Eoctadecatrienoic acid, and 9(R),16(S)-dihydroxy-10 E ,12 Z , 14 E -octadecatrienoic acid, in order of increasing retention time, respectively. Such data suggest that there are two different mechanisms explaining the formation of these 9,16-di-HOTE isomers. The E , Z , E compounds 9(S),16(S)-di-HOTE and 9(R),16(S)-di-HOTE would result from a dioxygenation mechanism as previously shown for PDX ( 18 ), whereas the all trans 9,16-di-HOTE isomers might derive from unknown intermediates.
We also investigated the metabolism of ALA incubated with human recombinant 15-LOX-2. Interestingly, 9(S)-HOTE or 9(R)-HOTE were both well converted into the corresponding 9,16-di-HOTEs whereas ALA was less efficiently metabolized into those di-hydroxylated fatty acids than by sLOX. Such a result might be biologically relevant because 9(S)-HOTE can be produced through cyclooxygenases, and 9(R)-HOTE can be produced via aspirinated COX-2, both in an aborted cyclooxygenation process ( 54 ).
The present study also shows some biological effects of the E , Z , E di-hydroxylated 9,16-di-HOTE isomers. We observed that 9(S),16(S)-dihydroxy-10 E ,12 Z ,14 E -octadecatrienoic and 9(R),16(S)-dihydroxy-10 E ,12 Z ,14 E -octadecatrienoic acids decreased signifi cantly prostaglandin synthesized by recombinant COX-1. Moreover, these E , Z , E isomers inhibit platelet aggregation triggered by collagen both by decreasing the extent of aggregation and making it partially reversible ( Fig. 7 ). However, these E , Z , E di-hydroxylated ALA metabolites appear to be less potent inhibitors than PDX under the same conditions, although the anti-aggregatory effects observed are in agreement with our previous data showing that poxytrins, characterized by an E , Z , E conjugated triene, are inhibitors of COX-1 ( 19 ).
COX-2 is an inducible enzyme becoming abundant at sites of infl ammation, and is upregulated in various carcinomas, having a central role in tumorigenesis ( 55 ). The inhibition of COX-2 is a major mode of action to reduce infl ammation ( 56 ). In our study, we found that 9(S),16(S)dihydroxy-10 E ,12 Z ,14 E -octadecatrienoic and 9(R),16(S)dihydroxy-10 E ,12 Z ,14 E -octadecatrienoic acids exhibit an inhibitory effect on the human recombinant COX-2.