Biosynthesis of Hemiketal Eicosanoids by Cross-Over of the 5-Lipoxygenase and Cyclooxygenase-2 Pathways

Biosynthesis of hemiketal eicosanoids by cross-over of the 5-lipoxygenase and cyclooxygenase-2 pathways

Markus Griessera,1,2, Takashi Suzukia,1, Noemi Tejeraa, Stacey Montb, William E. Boeglina, Ambra Pozzib, and Claus Schneidera,3

aDivision of Clinical Pharmacology, Department of Pharmacology, and Vanderbilt Institute of Chemical Biology, and bDivision of Nephrology, Department of Medicine, Vanderbilt University Medical School, Nashville, TN 37232

Edited by Jerrold Meinwald, Cornell University, Ithaca, NY, and approved March 14, 2011 (received for review December 27, 2010)

The prostaglandin and leukotriene families of lipid mediators are in the diendoperoxide by incorporation of a second peroxide, formed via two distinct biosynthetic pathways that are initiated connecting carbons 8 and 12. Thus, during transformation of 5S- by the oxygenation of arachidonic acid by either cyclooxygen- HETE, COX-2 catalyzes stereospecific insertion of three mole- ase-2 (COX-2) or 5-lipoxygenase (5-LOX), respectively. The 5-LOX cules of oxygen into the fatty acid substrate. Notably, the “unna- product 5S-hydroxyeicosatetraenoic acid, however, can also serve tural” or nonenzymatically formed enantiomer 5R-HETE is only as an efficient substrate for COX-2, forming a bicyclic diendoperox- a poor substrate for COX-2, and the COX-1 isoform does not ide with structural similarities to the arachidonic acid-derived react with 5S-HETE or 5R-HETE (5). JAm prostaglandin endoperoxide PGH2 [Schneider C, et al. (2006) The transformation of PGH2 to PGE2, PGD2, PGF2α, throm- Chem Soc – 128:720 721]. Here we identify two cyclic hemiketal boxane A2, and prostacyclin is catalyzed by a diverse family of (HK) eicosanoids, HKD2 and HKE2, as the major nonenzymatic prostaglandin synthase enzymes acting to rearrange the unstable rearrangement products of the diendoperoxide using liquid chro- 9,11-peroxide moiety (7, 8). Spontaneous nonenzymatic rearran- matography–mass spectrometry analyses as well as UV and NMR gement of the endoperoxide ring also gives rise to PGE2 and PGD2 spectroscopy. HKD2 and HKE2 are furoketals formed by sponta- and, in addition, can lead to levuglandins and cleavage of the neous cyclization of their respective 8,9-dioxo-5S,11R,12S,15S- carbon chain into malondialdehyde and 12S-hydroxy-5Z,8E, tetrahydroxy- or 11,12-dioxo-5S,8S,9S,15S-tetrahydroxy-eicosadi- 10E-heptadecatrienoic acid (9–11). Here, we show that rearrange- BIOCHEMISTRY 6E,13E-enoic acid precursors, resulting from opening of the ment of both peroxides of the diendoperoxide results in the for- 9S,11R- and 8S,12S-peroxide rings of the diendoperoxide. Further- more, the diendoperoxide is an efficient substrate for the hemato- mation of two hemiketal (HK) eicosanoids, identified as HKE2 poietic type of prostaglandin D synthase resulting in formation of and HKD2, with structural similarities to PGE2 and PGD2. The hemiketals are present in activated human leukocytes and stimu- HKD2, equivalent to the enzymatic transformation of PGH2 to late tubulogenesis of murine pulmonary endothelial cells. PGD2. HKD2 and HKE2 were formed in human blood leukocytes activated with bacterial lipopolysaccharide and calcium ionophore Results A23187, and biosynthesis was blocked by inhibitors of 5-LOX or COX-2. HKD and HKE stimulated migration and tubulogenesis Nonenzymatic transformation of the diendoperoxide. RP-HPLC 2 2 5 ½1 14 5S of microvascular endothelial cells, implicating a proangiogenic role analysis of short-time incubation (< min) of - C -HETE 1 2 of the hemiketals in inflammatory sites that involve expression of with recombinant COX-2 gave the diendoperoxide as the 5-LOX and COX-2. Identification of the highly oxygenated hemike- main product together with 5,11-diHETE and 5,15-diHETE as tal eicosanoids provides evidence for a previously unrecognized by-products (Fig. 1A) (5, 12). When the same incubation was 14 biosynthetic cross-over of the 5-LOX and COX-2 pathways. extracted 30 min after the addition of ½1- C5S-HETE, the diendoperoxide was no longer present, and instead two more polar 3 4 B 3 4 eukotrienes and prostaglandins are lipid mediators derived peaks and were detected (Fig. 1 ). Products and had iden- Lfrom oxidative modification of arachidonic acid. Both families tical UV spectra with a λmax of 236 nm that were unusually broad of eicosanoids exert inflammatory and immunomodulatory func- and extended beyond 260 nm (Fig. 1C). The UV spectra showed tions in disease, as well as homeostatic functions in normal phy- similarity to a conjugated keto-ene moiety rather than the typical siological processes (1). Atherosclerosis, asthma, and many types conjugated diene present in 5S-HETE. Using liquid chromato- of cancer are prototypical inflammatory diseases that are charac- graphy–electrospray ionization–mass spectrometry (LC-ESI-MS) terized by concomitant formation of both leukotrienes and pros- analysis in the negative ion mode a molecular ion ½M-H− of m∕z taglandins (2). The formation of leukotrienes and prostaglandins 399 for 2, 3, and 4 was detected, equivalent to a molecular weight diverges at the point of the initial oxygenation of the common of 400 amu (Fig. 1D). These findings implicated 3 and 4 as stable arachidonic acid substrate by either 5-lipoxygenase (5-LOX) or rearrangement products of the diendoperoxide 2. Products 3 and cyclooxygenase-2 (COX-2), respectively. From there, biosynthesis 4 were also formed when the diendoperoxide was first isolated by proceeds along separate and distinct pathways, each utilizing specific enzymes that catalyze complex reactions using highly unstable substrates (3, 4). Author contributions: A.P. and C.S. designed research; M.G., T.S., N.T., S.M., W.E.B., A.P., and C.S. performed research; M.G., A.P., and C.S. analyzed data; and A.P. and C.S. wrote The possibility of a biosynthetic convergence of the 5-LOX the paper. and COX-2 pathways was implicated when the 5-LOX product The authors declare no conflict of interest. 5S-hydroxyeicosatetraenoic acid (5S-HETE) was identified as an This article is a PNAS Direct Submission. efficient and specific substrate for oxygenation by recombinant 1 COX-2 (5). The COX-2 oxygenation product of 5S-HETE is a M.G. and T.S. contributed equally to this work. 2 bicyclic diendoperoxide with structural similarities to the prosta- Present address: Consumer Safety, Crop Protection, BASF SE, Speyerer Strasse 2, D-67117 Limburgerhof, Germany. glandin (PG) endoperoxide PGH2, including its chemical in- 3To whom correspondence should be addressed at: Department of Pharmacology, stability and potential for further transformation. Both PGH2 Vanderbilt University Medical School, 23rd Avenue South at Pierce, Nashville, TN and the diendoperoxide possess a 9S,11R-endoperoxide and a 37232-6602. E-mail: [email protected]. 15S -hydroxyl (6). The typical cyclopentyl ring present in PGH2 This article contains supporting information online at www.pnas.org/lookup/suppl/ and all prostaglandins is extended to a seven-membered ring doi:10.1073/pnas.1019473108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1019473108 PNAS Early Edition ∣ 1of6 Downloaded by guest on October 1, 2021 o 4 A 5 min, 37 C HO-CH3 10b 18 100 A 16 diendoperoxide 2 C 2 19

cpm) 3 3, 4 4

3 80 6,7 (λ 236 nm) max 14 10a 2 60 13 10b* 20 5,15-diHETE 8 9 15 5 5,11-diHETE 40 5-HETE 13* 10a* 3 1 (λ 236 nm) 5S-HETE 1 20 max 17 Intensity (10 0 0 Rel. absorbance (%) 01020 30 200 220 240 260 280 300 6.06.57.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 ppm Retention time (min) Wavelength (nm) 100 5-HETE 3.94E5 D m/z 319 50 19,20 3 B 30 min, 37oC 1.0 3 0 5,4 2,3 100

cpm) 5,15-diHETE 5,11- 1.76E5 3 2 m/z 335 4 diendoperoxide 2 50 diHETE 9,10b 10a*,10b* 16,17 2.0 5,15-diHETE 0 3,4 1 100 5,11-diHETE 3 4 diendoperoxide 2 4.90E4 9*, m/z 399 3.0

Relative abundance (%) 50

Intensity (10 10a* 10a, 0 0 10b 01020 30 0 2.0 4.0 6.0 8.0 10.0 8,9 4.0 Retention time (min) Retention time (min) 7,8 5.0 Fig. 1. Nonenzymatic transformation of the 5S-HETE-derived diendoperoxide. ½1-14C-5S-HETE was reacted with COX-2 for 5 min (A)or30min 14*, 13* 6.0 (B) followed by extraction and RP-HPLC analysis with radiodetection. (C) Nor- 6,7 malized UV spectra of 5S-HETE and products 3 and 4.(D) LC-ESI-MS analysis 4 (Hemiketal D2) (negative ion mode) of an incubation of 5S-HETE with recombinant COX-2. 14,13 7.0 The extracted ion chromatograms corresponding to 5S-HETE (m∕z 319), 6.06.57.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 ppm diHETEs (m∕z 335), and the diendoperoxide (m∕z 399) are shown.

B O C 201 OH HPLC and then incubated in PBS buffer indicating that COX-2 HO 9 HO 8 CO2H was not involved in the transformation of 2 to 3 and 4. 213 5 CO2H 5 O OH O OH 12 15 11 15 Structural Identification of 3 and 4 as Hemiketal Eicosanoids. Products HO HO 3 4 μ OH O and were synthesized in a large-scale incubation of 600 g 3,HemiketalE2 4,HemiketalD2 5S-HETE with COX-2 and isolated using RP-HPLC and further 100 151 MS2 of HKE2 MS2 of HKD2 345 purified using straight-phase HPLC. The covalent structures of 3 301 - CO 1 2 and 4 were determined using H and 2D homo- and heteronuc- 80 - CO2 A – 381 lear NMR analyses (Fig. 2 , Tables S1 and S2, and Figs. S1 S4). 60 -2 H2O The NMR spectra of both 3 and 4 each showed mixtures of -H O 2 345 -H2O 40 183 381 two diastereomers with partially overlapping signals in a ratio of -H O 201 195 213 2 -H2O 4∶1 to ≈1∶1 depending on the NMR solvent used (d4-methanol, 20 399 399 d3-acetonitrile, or d3-chloroform). The structural identification of 0 the major isomer of 4 will be discussed as representative. 150 200 250 300 350 400 450 150 200 250 300 350 400 450 The 1H NMR spectrum of 4 showed four signals with chemical m/z m/z trans shifts and coupling constants characteristic of double bonds, Fig. 2. Identification of products 3 and 4.(A) NMR analysis of product 4 trans trans A 1 representing the 6,7- and 13,14- double bonds (Fig. 2 ). (hemiketal D2). The H NMR spectrum is shown with the H,H-COSY spectrum The signal for H13 (6.95 ppm) showed a splitting pattern indica- below. The product is a mixture of two isomers in ≈5∶2 ratio. Signals from the tive of a proton with only one direct neighbor, the other proton of minor isomer with a distinct chemical shift are marked with an asterisk. In the the double bond, H14 at 7.01 ppm. Thus, C12 does not carry a H,H-COSY spectrum only the prominent cross-peaks are visible. Weaker proton, implicating a keto group at this position. This was con- signals like the couplings of H5 to H6 and H14 to H15 were discernible at firmed by the carbon chemical shift of 196.0 ppm for C12 deter- lesser signal to noise. (B and C) LC-ESI-MS2 spectra of 3 and 4. mined in the heteronuclear multiple bond correlation experiment tetrahydroxyeicosa-6E,13E-dienoic acid, and cyclization of the through the cross-peak with H13 (Figs. S1 and S2). H,H-COSY 12S 3 cross-peaks of three of the double bond protons (H6, H7, and -hydroxyl and 9-keto gives product that was designated as H14) to signals between 4 and 5 ppm allowed identification of hemiketal E2 (HKE2)(Figs. S3 and S4 and Table S2). The H5 (4.10 ppm), H15 (4.25 ppm), and H8 (4.64 ppm) as geminal absolute configuration of the chiral centers of carbons 5, 11, hydroxy protons. H8 was coupled to another geminal hydroxy 12, and 15 of HKE2, and 5, 8, 9, and 15 of HKD2, respectively, proton, H9 at 4.40 ppm. The two protons of the H10 methylene was assumed to be unchanged from the diendoperoxide precur- 5S 8S 9S 11R 12S 15S group were detected as dd signals at 2.40 ppm and 2.25 ppm, sor, i.e., , , , , , and (5, 6). The characteristic J keto-ene UV chromophore ruled out the possibility of a six-mem- respectively, with a coupling constant of H10a;b of 14.5 Hz. The large coupling constant between the methylenic protons H10a bered (pyran) ring, e.g., C8 through C12 in HKD2, or a furoketal and H10b together with lack of a proton at C11 provided evi- ring of, e.g., carbons 9 through 12 in HKD2. The two isomers dence of a furoketal ring structure in 4 encompassing carbons observed in the NMR analyses of 3 and 4 are the diastereomers 8 through 11 with the hemiketal moiety located at C11. A keto stemming from the anomeric hemiketal carbons C9 or C11, group in conjugation with a double bond and oxygen functional- respectively. The equilibrium between the diastereomers was ities on either side was consistent with the UV spectrum of 4 with dependent on the solvent. The NMR coupling data and chemical λ C max at 236 nm (Fig. 1 ). shifts were not sufficient to unequivocally assign the absolute Product 4 was identified as the furoketal (hemiketal D2, configuration of the anomeric carbons in the diastereomers of HKD2) formed by spontaneous cyclization of the 8S-hydroxyl and HKE2 and HKD2. 11-keto groups of 11,12-dioxo-5S,8S,9S,15S-tetrahydroxyeicosadi- 6E,13E-enoic acid, one of two immediate products of opening Isotopic Labeling Studies. The formation of two keto groups in the of the 9,11- and 8,12-peroxide bridges of the diendoperoxide 2. hemiketals was confirmed by mass spectrometric analyses of the 18 The other immediate product is 8,9-dioxo-5S,11R,12S,15S- transformation of unlabeled 5S-HETE, ½1- O2-5S-HETE, and

2of6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1019473108 Griesser et al. Downloaded by guest on October 1, 2021 [5,6,8,9,11,12,14,15-d8]-5S-HETE. Incubation of a ≈1∶1 mixture ABH-PGDS L-PGDS of unlabeled 5S-HETE and d8-5S-HETE with recombinant 100 4.9E5 100 6.0E4

COX-2 and analysis by LC-ESI-MS showed that the diendoper- HKD2 HKE2 diendoperoxide oxide 2 retained all eight deuterium atoms, whereas HKE2 and 50 50 HKD HKE 2 HKD2 had each lost two of the labels (Fig. S5). Loss of a deu- 2 diendoperoxide terium is indicative of a change in carbon hybridization from sp3 to sp2, as it occurs, for example, when a carbon carrying a hydro- Rel. intensity (%) 0 0 xyl or peroxyl is oxidized to a keto group. The site at the molecule CDnon-enzymatic RAW264.7 where loss of the deuterium label had occurred was determined 100 1.1E5 100 3.7E4 by comparison of the MS2 spectra of HKE2 and HKD2 formed HKD2 18 HKE HKD2 from unlabeled 5S-HETE, from ½1- O2-5S-HETE, and 2 50 50 from d8-5S-HETE. HKE2 diendoperoxide diendoperoxide The MS2 spectrum of HKE2 showed major cleavage ions at

m∕z 213, 195 (loss of H2O from m∕z 213), and 151 (loss of CO2 Rel. intensity (%) 0 0 from m∕z 195) (Fig. 2B). All three fragments were derived from 123456 123456 Time (min) Time (min) the carboxylate end of HKE2 because the first two ions were m∕z ½1 18 5S shifted by 4 amu to 217 and 199 when - O2 - -HETE Fig. 3. Transformation of the diendoperoxide by hematopoietic prostaglan- m∕z was the substrate, whereas 151, formed by loss of CO2 or din D synthase (H-PGDS) to HKD2. LC-ESI-MS analyses of reactions of 5S-HETE 18 C O2, respectively, was unchanged (Fig. S5). When d8-5S-HETE with recombinant COX-2 in the presence of (A) recombinant human H-PGDS, was used as substrate m∕z 213 was shifted by only 2 amu to m∕z (B) recombinant murine L-PGDS, (C) in the absence of PGDS. (D) Transforma- 5S 215 indicating that the loss of the two deuterium atoms in HKE2 tion of exogenous -HETE by LPS-activated RAW264.7 cells. The extracted had occurred at the side of the molecule that contained the ion chromatograms for m∕z 399 are shown. carboxylate, implicating the oxidation of the 8- and 9-sp3 carbons to keto groups. of the COX-2 inhibitor NS-398 (Fig. 4) (18), or when the 5-LOX Comparison of the MS2 spectra of HKD2 derived from unla- inhibitor AA-861 (19) or the five lipoxygenase-activating protein 18 beled 5S-HETE versus ½1- O2-5S-HETE showed that the major inhibitor MK-886 were present (20) (Fig. S7). The levels of hemi- cleavage ions m∕z 201 and m∕z 187 in the MS2 spectrum of ketals were ≈1–5% of the amount of 5-HETE and LTB4 (each C 6 HKD2 also contained the carboxylate group (Fig. 2 and Fig. S5). ≈1 ng∕10 cells) in the stimulated cells based on signal intensi- BIOCHEMISTRY The fragment at m∕z 201 was shifted by 4 mass units to m∕z 205 ties. Hemiketals were not detected in nonstimulated leukocytes. when d8-5S-HETE was the substrate (Fig. S5), indicating that the

loss of the two deuterium atoms in HKD2 had occurred at the HKE2 and HKD2 Stimulate Endothelial Cell Tubulogenesis. Both COX- methyl end of the molecule, in this case from carbons 11 and 12. 2 and 5-LOX mediated pathways have been shown to regulate endothelial cell function and angiogenesis (21). We therefore de- The Diendoperoxide Is a Substrate for Hematopoietic Prostaglandin D termined whether HKD2 and HKE2 could affect endothelial cell Synthase (H-PGDS). HKE2 and HKD2 appeared to be rearrangement products similar to the transformation of PGH2 to PGE2 and PGD2. We therefore tested whether H-PGDS or lipocalin A 100 2.1E6 6.4E5 HKE2 PGDS (L-PGDS) was able to catalyze the transformation of the HKD2 7.1E5 diendoperoxide into HKD2. LC-MS analysis of incubations of 50 5-HETE H-PGDS and L-PGDS with the isolated diendoperoxide showed Rel. int. (%) efficient formation of HKD2 by H-PGDS, whereas L-PGDS did 0 not react with the diendoperoxide (Fig. 3 A–C). In the H-PGDS B 100 reaction, the concentration of the cofactor glutathione was re- HKE 1.1E4 1.9E4 2 HKD 5.9E3 duced from 1 mM (13) to 10 μM in order to prevent otherwise 2 50 5-HETE abundant formation of glutathione-HK adducts (Fig. S6). LPS-

activated RAW264.7 cells that express H-PGDS (14) and show Rel. int. (%) 0 more efficient formation of PGD2 than PGE2 (15), transformed exogenous 5S-HETE mainly to HKD2 (Fig. 3D). Side-by-side C 100 2.9E4 2.1E7 μ HKE2 incubations of 10 M PGH2 or the diendoperoxide generated HKD2 1.0E4 in situ using limiting amounts of recombinant H-PGDS showed 50 5-HETE that both peroxides were about equally effective substrates. The Rel. int. (%) other recombinant PG synthases tested, prostacyclin synthase, 0 thromboxane synthase, and mPGES-1, did not react with the D 100 diendoperoxide in vitro using conditions under which they readily HKE2 4.0E2 6.0E3 reacted with PGH2. HKD2 3.2E2 50 5-HETE Human Leukocytes Form HKE and HKD . We hypothesized that

2 2 Rel. int. (%) 0 human blood leukocytes would be able to biosynthesize the 0 12345678910 hemiketals in response to pertinent stimulation. 5-LOX activity Time (min) in neutrophils and eosinophils can be stimulated using calcium ionophore A23187 (16), and expression of COX-2 in monocytes Fig. 4. Isolated human leukocytes form HKE2 and HKD2. Human leukocytes can be induced using LPS (17). Human peripheral blood leuko- were stimulated with LPS for 6 h and then with A23187 for 15 min before cytes were isolated and treated with 10 μg∕mL LPS for 6 h fol- extraction and LC-ESI-MS-SRM analysis. (A) Standards of HKE2 (red trace) and HKD2 (blue) generated by reaction of recombinant human COX-2 with lowed by treatment with 5 μM A23187 for 15 min and analyzed 5S-HETE (black). (B) HKE2 and HKD2 formed by the leukocytes following using LC-ESI-MS operated in the selected reaction-monitoring stimulation with LPS and A23187. (C) Addition of exogenous 5S-HETE to- (SRM) mode. Formation of HKE2 and HKD2 was detected with- gether with A23187 to the leukocytes increased formation of HKE2 and out addition of exogenous substrate, enhanced about 2-fold when HKD2.(D) Formation of hemiketals was attenuated by preincubation with 5 μM 5S-HETE was added, and reduced to <3% in the presence the COX-2 inhibitor NS-398.

Griesser et al. PNAS Early Edition ∣ 3of6 Downloaded by guest on October 1, 2021 function. Both HKD2 and HKE2 stimulated—in a dose-depen- common structural features of PGD2 (with a 9-hydroxyl and 11- dent manner—the formation of endothelial capillary-like struc- keto) and HKD2 with 8- and 9-hydroxyls and 11- and 12-keto tures, as determined by the ability of these compounds to groups in the open chain form of the molecule. The 11-keto, how- promote endothelial cells to sprout, branch, and form ring-like ever, is masked as the anomeric hemiketal carbon in the cyclized structures (Fig. 5 A and B). When used at 1 μM hemiketals also molecule. stimulated endothelial cell migration (Fig. 5C). Neither HKD2 The major product formed by nonenzymatic rearrangement of nor HKE2 (between 10 nM to 1 μM) induced changes in intra- the diendoperoxide was identified as a structurally related furo- cellular calcium levels of the endothelial cells nor stimulated their ketal, designated “hemiketal E2” (HKE2) in analogy to PGE2. proliferation. Formation of HKE2 and HKD2 upon nonenzymatic rearrangement of the diendoperoxide follows the precedent of sponta- Discussion neous formation of PGE2 and PGD2 from the prostaglandin 5S The reaction of COX-2 with the 5-LOX product -HETE to endoperoxide PGH2 (9). Unexpectedly, of the four possible per- form a diendoperoxide that is structurally equivalent to the mutations for two keto and two hydroxyl groups by opening of prostaglandin endoperoxide suggests a previously unrecognized both peroxide rings of the diendoperoxide only two are manifest biosynthetic connection of the two enzymes. The specific trans- such that the keto groups and the hydroxyl groups are next to formation of the diendoperoxide by H-PGDS lays out a biosyn- each other. A possible driving force for the vicinal arrangement thetic pathway from arachidonic acid via 5-LOX, COX-2, and is the stabilization gained by conjugation of the two keto groups “ H-PGDS to the furoketal eicosanoid that we named hemiketal with the double bond in a 1,2-dioxo-3-ene moiety. The open chain ” “ ” D2 (HKD2) (Fig. 6). The letter D was chosen to indicate the products undergo spontaneous ring closure between hydroxyl and crucial role of H-PGDS in biosynthesis, and to acknowledge the keto groups to form the furoketals HKE2 or HKD2 (Fig. 6). H-PGDS and L-PGDS are structurally unrelated enzymes that catalyze the transformation of PGH2 into PGD2 (7). The two HKD2 enzymes differ markedly in their kinetic behavior with PGH2 A 125 nM 500 nM as substrate: H-PGDS is a low-affinity, high-turnover enzyme, whereas L-PGDS is a high-affinity, low-turnover enzyme (13, 22). The high affinity of PGH2 in the L-PGDS active site is an indica- control tion of tight binding, and this might be one of the reasons why the enzyme did not catalyze the transformation of the larger diendoperoxide. H-PGDS, on the other hand, has only low binding affinity for PGH2 and efficiently catalyzed ring opening of the diendoperoxide. Both enzymes utilize nucleophilic attack of a thiolate anion at the 11-oxygen of PGH2 as their catalytic mechanism. L-PGDS uses an active site cysteine residue (22), whereas in H-PGDS the thiolate is provided by the cofactor glutathione. High concentrations of glutathione (above 100 μM) in the in vitro reactions resulted in noticeable adduct formation 125 nM 500 nM with the hemiketals. Adduct formation was likely due to the pro- HKE 2 pensity of the oxo-ene moiety to undergo Schiff base and Michael B reactions. Adduction with cellular nucleophiles like glutathione, control N-acetylcysteine, and reactive amino acids like lysine and histi- dine (23) could also be a limiting factor for the detection of the 125 * hemiketals in tissue samples. HKD2 500 ** Human leukocytes biosynthesized both hemiketals as a result of stimulation of 5-LOX and COX-2 activities ex vivo. Inhibition 125 * of either enzyme markedly diminished formation of the hemike- HKE2 500 ** tals indicating the crucial role of 5-LOX and COX-2. Few cell [nM] types show concomitant expression of 5-LOX and COX-2 impli- 02468101214161820 cating that formation of hemiketals in the leukocyte preparation C No. of branches/microscopic field occurred through transcellular biosynthesis (24). Thus, biosynth- control esis of hemiketals could be physiologically relevant in sites of inflamed tissue infiltrated by neutrophils (as a source of 5-HETE) serum * and activated macrophages (providing COX-2). This implies a HKD 2 1.0 * possible role for hemiketals in the regulation of discrete steps

HKE2 1.0 * during induction or resolution of inflammation (25). One such [µM] possibility is the contribution of hemiketals to tissue restoration 012345 through the stimulation of endothelial cell migration and tubu- fold change versus control logenesis. The same effect, however, could also lead to enhanced

Fig. 5. HKD2 and HKE2 induce formation of capillary-like structures and mi- tumor angiogenesis and an aggravation of cancer growth, when gration of endothelial cells. (A) Microvascular endothelial cells were plated hemiketals are formed upon neutrophil invasion into COX-2 onto Matrigel in the presence of 125 or 500 nM of HKD2 or HKE2. Represen- expressing tumor tissue or microenvironment. tative images of capillary-like structures taken 6 h after plating are shown. 5S-HETE is considered a by-product of 5-LOX catalysis, which (B) Quantification of capillary network formation. Cells incubated with HKD2 is set up to further transform the initial product, 5S-HPETE, to or HKE2 form tubule-like structures more efficiently than vehicle-treated cells 5S the leukotriene (LT) epoxide, LTA4 (3). Nevertheless, -HETE (control). Values are the mean SD. calculated for 20 images per treatment. is formed in considerable amounts in cells that express 5-LOX Differences between vehicle versus HKD2 or HKE2 treated cells (*), or between 125 versus 500 nM treatment (**) were significant (p < 0.05). (26), and it has a propensity for esterification into specific (C) Migration of cells toward serum-free medium with or without HKD2 or membrane phospholipids depending on the cell type (27, 28). 5S HKE2 or complete medium. Values are fold changes versus serum-free med- -HETE does not appear to carry out a biological function ium from four experiments; (*) p < 0.05 versus control. by itself, but it has a well-documented role as precursor of the

4of6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1019473108 Griesser et al. Downloaded by guest on October 1, 2021 O Prostaglandin CO H biosynthesis 2 11 arachidonic acid COX-1, PGES 9 HO PGE COX-2 O OH 2 CO2H CO2H (nonenzymatically) O HO 2 O 11 2 OH PGDS CO H PGH2 2 11 O2 5-LOX O OH PGD2

OH 5-LOX/COX-2 O O 5 HO cross-over pathway 9 5 O 9 CO2H OH O OH OH 12 12 OH HO HO 15 COX-2 9 OH OH 5 O O (nonenzy- open chain Hemiketal E2 (3) O O matically) intermediates 3 O2 11

OH cyclization) OH 5S-HETE (1) OH H-P (spontaneous GDS HO 8 HO 8 5 CO2H di-endoper- OH OH O O oxide (2) 11 15 11 HO O O OH Hemiketal D2 (4)

Fig. 6. Biosynthesis of hemiketals and prostaglandins. The 5-LOX metabolite 5S-HETE is converted to the diendoperoxide by COX-2 using three molecules of oxygen. The diendoperoxide is a substrate for H-PGDS opening up both endoperoxide groups into keto/hydroxy moieties to give an open chain intermediate. Hemiketal D2 results form spontaneous attack of the 8-hydroxyl at the 11-keto group. Nonenzymatic rearrangement of the diendoperoxide gives a mixture of HKE2 and HKD2. Oxygenation of arachidonic acid by COX-1 or COX-2 yields the prostaglandin endoperoxide PGH2 that rearranges to PGE2 and PGD2.

eosinophil chemoattractant 5-oxo-ETE (29). Our findings sug- HKD2 the flow rate was reduced to 0.4 mL∕ min. The products were further gest an additional role for 5S-HETE as a substrate for the for- purified using an Agilent Zorbax RX-SIL 5 μm column (4.6 × 250 mm) eluted with hexane/isopropanol/acetic acid 80/20/0.1 (by volume) at 1 mL∕ min. 1H mation of hemiketal eicosanoids in the biosynthetic convergence BIOCHEMISTRY of the 5-LOX and COX-2 pathways, acting in tandem in the and 2D NMR spectra were recorded at 285 K using a Bruker AV-II 600-MHz formation of biologically active lipid autacoids. spectrometer equipped with a cryoprobe. Chemical shifts are reported in parts per million relative to the signal of residual nondeuterated solvent Materials and Methods (δ 7.25 for CDCl3, δ 3.30 for CD3OD, and δ 1.93 for CD3CN). Materials. Recombinant human COX-2 was expressed and purified as described (30). 5S-HETE was prepared by chemical synthesis (6, 31). Human Leukocyte Isolation and Incubation. The study was approved by the 18 ½1- O2-5S-HETE was prepared by repeated cycles of methylation and hydro- Vanderbilt Institutional Review Board, protocol #091243. Venous blood 18 14 lysis in H2 O(>97%) (32). ½1- C-5S-HETE and [5,6,8,9,11,12,14,15-d8]-5S- (45 mL) from healthy volunteers was collected into citric acid mixed with HETE were prepared by reacting labeled arachidonic acid with recombinant 10 mL of 6% dextran and allowed to settle for 1 h. Leukocytes were collected human 5-LOX expressed in Sf9 insect cells. Recombinant human H-PGDS and by centrifugation, and remaining red cells were removed by hypotonic lysis. murine L-PGDS were purchased from Cayman Chemical. In addition, H-PGDS Cells resuspended in PBSþ (1.5 × 107 cells) were treated with LPS (10 μg∕mL) was cloned from human colon cancer cDNA (National Center for Biotechnol- for 6 h at 37 °C followed by A23187 (5 μM) for 15 min. 5S-HETE (15 μM), NS- ogy Information accession number NM_014485). The cDNA encoding an 398 (10 μM), MK-886 (10 μM), or AA-861 (10 μM) was added to some of the N-terminal 6xHis-tag was subcloned into the pET3 expression vector, incubations 30 min prior to the addition of A23187. Samples were acidified expressed in Escherichia coli, and purified using Ni-nitrilotriacetate affinity and extracted using Waters HLB cartridges. After evaporation, samples were chromatography. reconstituted in 10 μL of acetonitrile and 40 μL of LC-MS column solvent A. Incubations with Labeled Substrates. Incubation of a 1∶1 mixture of 18 LC-MS Analyses. A ThermoFinnigan Quantum Access triple quadrupole instru- ½1- O2-5S-HETE and unlabeled 5S-HETE (20 μM final) was conducted in ment with an electrospray interface was operated in the negative ion mode. 500 μL Tris-HCl buffer pH 8 containing 500 μM phenol, 1 μM hematin, and 1 μM COX-2 for 10 min at 37 °C. The same reaction conditions were used Settings for sheath and auxiliary gas pressures, temperature, and interface for the incubation of a 1∶1 mixture of d8-5S-HETE and d0-5S-HETE. voltage were optimized using direct infusion of a solution of PGD2. A Waters Symmetry Shield C18 3.5-μm column (2.1 × 150 mm) was eluted with a ∕ 5∕95 Incubations with Recombinant PGDS. The reaction with H-PGDS was con- linear gradient of acetonitrile water, 10 mM NH4OAc ( ; solvent A) to ducted in 500 μL of 100 mM Tris-HCl buffer pH 8 containing 10 μM glu- acetonitrile∕water, 10 mM NH4OAc (95∕5, by volume; solvent B) at tathione (GSH), 10 mM MgCl2,0.5μM H-PGDS, and 10 μM diendoperoxide 0.2 mL∕ min within 10 min. The following transitions were monitored in for 10 min at room temperature. The reaction with L-PGDS was conducted the negative ion SRM mode: m∕z 399 → 151 for HKE2 and m∕z 399 → 183 in the same buffer containing 10 μM GSH, 500 μg BSA, and 0.5 μM L-PGDS. for HKD2 from 0–8 min; m∕z 319 → 115 for 5-HETE from 8–10 min. The signal Control reactions with PGH2 formed in situ were conducted using 10 μM intensities obtained for each ion chromatogram are given in exponential for- μ arachidonic acid and 1 M COX-2, and the same amount of H-PGDS, GSH, mat with the maximum intensity set as 100%. Data acquisition and spectral and MgCl2 or L-PGDS, GSH, and BSA as described for the reaction with analysis was performed using ThermoFinnigan Xcalibur software. the diendoperoxide. For the extraction of glutathione adducts, 30 mg Waters hydrophilic-lipophilic balanced (HLB) cartridges were conditioned with Tubulogenesis Assay. For the Matrigel-based tubulogenesis assay, capillary- 50 mM NH4OAc (pH 3.4), and products were eluted with 95% ethanol. like structure formation was analyzed as described (33). Briefly, 96-well plates were coated with 50 μL of Matrigel and incubated for 30 min at 37 °C. Mouse Preparation and Isolation of the Hemiketals. Twenty 2-mL reactions containing pulmonary endothelial cells [isolated and cultured as described (34)] were 1 μM COX-2 in 100 mM phosphate buffer pH 8, 500 μM phenol, and 1 μM seeded over Matrigel at 1.5 × 104 cells/well in serum-free medium with or hematin were conducted in parallel. The reactions were started by addition of 30 μg 5S-HETE to each vial, allowed to proceed for 30 min, acidified (pH 4), without HKD2 or HKE2. The formation of capillary-like structures was re- and extracted using 30 mg Waters HLB cartridges. The products were ana- corded hourly for the next 10 h. Shown in Fig. 5 are representative images lyzed by RP-HPLC using a Waters Symmetry C18 5-μm column (4.6 × 250 mm) taken 6 h after plating. Cellular nodes were defined as junctions linking at eluted with a gradient of acetonitrile∕water∕acetic acid from 20∕80∕0.01 to least three cells, and counted from digital images. Two independent experi- 70∕30∕0.01 (by volume) within 20 min at 1 mL∕ min. Chromatography was ments were performed in duplicate with a total of 20 images analyzed per monitored using an Agilent 1200 diode array detector. For isolation of treatment.

Griesser et al. PNAS Early Edition ∣ 5of6 Downloaded by guest on October 1, 2021 Migration Assay. Cell migration was assayed in transwell plates. Lower wells ACKNOWLEDGMENTS. We thank Drs. Thomas M. Harris, Markus Voehler, and were coated with matrigel (10 μg∕mL) at 4 °C and then incubated 1 h at 37 °C Ned Porter for careful review of the NMR data and helpful discussion. This with bovine serum albumin (1% in PBS) to inhibit nonspecific cell migration. work was supported by Award R01GM076592 from the National Institute of Serum-free medium with or without HKD2 or HKE2 was added to the lower General Medical Sciences and Award BC063074 from the Department of De- wells and endothelial cells (5 × 104 cells in 100 μL of serum-free medium) to fense Breast Cancer Research Program (to C.S.), and Award 2P01DK065123 the upper wells. Migration toward complete endothelial cell medium was from the National Institute of Diabetes and Digestive and Kidney Diseases used as positive control. After 6 h at 37 °C, cells on the top of the filter were and a Merit Review from the Department of Veterans Affairs (to A.P.). removed by wiping, and the filters were fixed (4% formaldehyde in PBS). The content is solely the responsibility of the authors and does not necessarily Migrating cells were stained with hematoxylin and eosin and counted. represent the official views of the National Institutes of Health. We acknowl- The number of migrated cells is expressed as fold change relative to control edge National Institutes of Health Grant S10 RR019022 for funding of the in four independent experiments. Bruker 600-MHz AV-II NMR spectrometer and accessories.

1. Funk CD (2001) Prostaglandins and leukotrienes: Advances in eicosanoid biology. 19. Yoshimoto T, et al. (1982) 2,3,5-Trimethyl-6-(12-hydroxy-5,10-dodecadiynyl)-1,4- Science 294:1871–1875. benzoquinone (AA861), a selective inhibitor of the 5-lipoxygenase reaction and the 2. Shimizu T (2009) Lipid mediators in health and disease: Enzymes and receptors as biosynthesis of slow-reacting substance of anaphylaxis. Biochim Biophys Acta therapeutic targets for the regulation of immunity and inflammation. Annu Rev 713:470–473. Pharmacol Toxicol 49:123–150. 20. Dixon RA, et al. (1990) Requirement of a 5-lipoxygenase-activating protein for leuko- 3. Murphy RC, Gijon MA (2007) Biosynthesis and metabolism of leukotrienes. Biochem J triene synthesis. Nature 343:282–284. – 405:379 395. 21. Gonzalez-Periz A, Claria J (2007) New approaches to the modulation of the cycloox- 4. Rouzer CA, Marnett LJ (2009) Cyclooxygenases: Structural and functional insights. ygenase-2 and 5-lipoxygenase pathways. Curr Top Med Chem 7:297–309. J Lipid Res 50(Suppl):S29–34. 22. Kumasaka T, et al. (2009) Structural basis of the catalytic mechanism operating in 5. Schneider C, Boeglin WE, Yin H, Stec DF, Voehler M (2006) Convergent oxygenation open-closed conformers of lipocalin type prostaglandin D synthase. J Biol Chem of arachidonic acid by 5-lipoxygenase and cyclooxygenase-2. J Am Chem Soc 284:22344–22352. 128:720–721. 23. Sayre LM, Lin D, Yuan Q, Zhu X, Tang X (2006) Protein adducts generated from 6. Griesser M, Boeglin WE, Suzuki T, Schneider C (2009) Convergence of the 5-LOX and products of lipid oxidation: Focus on HNE and ONE. Drug Metab Rev 38:651–675. COX-2 pathways. Heme-catalyzed cleavage of the 5S-HETE-derived di-endoperoxide 24. Folco G, Murphy RC (2006) Eicosanoid transcellular biosynthesis: From cell-cell inter- into aldehyde fragments. J Lipid Res 50:2455–2462. – 7. Urade Y, Eguchi N (2002) Lipocalin-type and hematopoietic prostaglandin D synthases actions to in vivo tissue responses. Pharmacol Rev 58:375 388. as a novel example of functional convergence. Prostag Oth Lipid M 68–69:375–382. 25. Serhan CN (2007) Resolution phase of inflammation: Novel endogenous anti- 8. Samuelsson B, Morgenstern R, Jakobsson PJ (2007) Membrane prostaglandin E inflammatory and proresolving lipid mediators and pathways. Annu Rev Immunol synthase-1: A novel therapeutic target. Pharmacol Rev 59:207–224. 25:101–137. 9. Hamberg M, Samuelsson B (1967) Oxygenation of unsaturated fatty acids by the 26. Powell WS, Rokach J (2005) Biochemistry, biology and chemistry of the 5-lipoxygenase vesicular gland of sheep. J Biol Chem 242:5344–5354. product 5-oxo-ETE. Prog Lipid Res 44:154–183. 10. Hamberg M, Samuelsson B (1974) Prostaglandin endoperoxides. Novel transforma- 27. Stenson WF, Parker CW (1979) Metabolism of arachidonic acid in ionophore-stimu- tions of arachidonic acid in human platelets. Proc Natl Acad Sci USA 71:3400–3404. lated neutrophils. Esterification of a hydroxylated metabolite into phospholipids. 11. Brame CJ, Salomon RG, Morrow JD, Roberts LJ, 2nd (1999) Identification of extremely J Clin Invest 64:1457–1465. reactive gamma-ketoaldehydes (isolevuglandins) as products of the isoprostane 28. Richards CF, Johnson AR, Campbell WB (1986) Specific incorporation of 5-hydroxy- pathway and characterization of their lysyl protein adducts. J Biol Chem 6,8,11,14-eicosatetraenoic acid into phosphatidylcholine in human endothelial cells. 274:13139–13146. Biochim Biophys Acta 875:569–581. 12. Mulugeta S, et al. (2010) Identification and absolute configuration of dihydroxy- 29. Powell WS, Gravelle F, Gravel S (1992) Metabolism of 5(S)-hydroxy-6,8,11,14-eicosate- arachidonic acids formed by oxygenation of 5S-HETE by native and aspirin-acetylated traenoic acid and other 5(S)-hydroxyeicosanoids by a specific dehydrogenase in human – COX-2. J Lipid Res 51:575 585. polymorphonuclear leukocytes. J Biol Chem 267:19233–19241. 13. Urade Y, Fujimoto N, Ujihara M, Hayaishi O (1987) Biochemical and immunological 30. Schneider C, Boeglin WE, Brash AR (2004) Identification of two cyclooxygenase active characterization of rat spleen prostaglandin D synthetase. J Biol Chem 262:3820–3825. site residues, leucine-384 and glycine-526, that control carbon ring cyclization in pros- 14. Joo M, et al. (2009) Lipopolysaccharide-dependent interaction between PU.1 and c-Jun taglandin biosynthesis. J Biol Chem 279:4404–4414. determines production of lipocalin-type prostaglandin D synthase and prostaglandin 31. Corey EJ, Hashimoto S (1981) A practical process for large-scale synthesis of D2 in macrophages. Am J Physiol Lung Cell Mol Physiol 296:L771–779. (S)-5-hydroxy-6-trans-8,11,14,cis-eicosatetraenoic acid (5-HETE). Tetrahedron Lett 15. Rouzer CA, et al. (2005) RAW264.7 cells lack prostaglandin-dependent autoregulation 22:299–302. of tumor necrosis factor-alpha secretion. J Lipid Res 46:1027–1037. 16. Borgeat P, Samuelsson B (1979) Arachidonic acid metabolism in polymorphonuclear 32. Murphy RC, Clay KL (1990) Preparation of labeled molecules by exchange with oxygen- – leukocytes: Effects of ionophore A23187. Proc Natl Acad Sci USA 76:2148–2152. 18 water. Methods Enzymol 193:338 348. 17. Patrignani P, et al. (1994) Biochemical and pharmacological characterization of 33. Pozzi A, et al. (2010) The anti-tumorigenic properties of peroxisomal proliferator- the cyclooxygenase activity of human blood prostaglandin endoperoxide synthases. activated receptor alpha are arachidonic acid epoxygenase-mediated. J Biol Chem J Pharmacol Exp Ther 271:1705–1712. 285:12840–12850. 18. Barnett J, et al. (1994) Purification, characterization and selective inhibition of human 34. Pozzi A, et al. (2000) Elevated matrix metalloprotease and angiostatin levels in integ- prostaglandin G/H synthase 1 and 2 expressed in the baculovirus system. Biochim rin alpha 1 knockout mice cause reduced tumor vascularization. Proc Natl Acad Sci USA Biophys Acta 1209:130–139. 97:2202–2207.

6of6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1019473108 Griesser et al. Downloaded by guest on October 1, 2021