Major urinary metabolites of 6-keto-prostaglandin F2α in mice.

Western diets are enriched in omega-6 vs. omega-3 fatty acids, and a shift in this balance toward omega-3 fatty acids may have health benefits. There is limited information about the catabolism of 3-series prostaglandins (PG) formed from eicosapentaenoic acid (EPA), a fish oil omega-3 fatty acid that becomes elevated in tissues following fish oil consumption. Quantification of appropriate urinary 3-series PG metabolites could be used for noninvasive measurement of omega-3 fatty acid tone. Here we describe the preparation of tritium- and deuterium-labeled 6-keto-PGF2α and their use in identifying urinary metabolites in mice using LC-MS/MS. The major 6-keto-PGF2α urinary metabolites included dinor-6-keto-PGF2α (∼10%) and dinor-13,14-dihydro-6,15-diketo-PGF1α (∼10%). These metabolites can arise only from the enzymatic conversion of EPA to the 3-series PGH endoperoxide by cyclooxygenases, then PGI3 by prostacyclin synthase and, finally, nonenzymatic hydrolysis to 6-keto-PGF2α. The 6-keto-PGF derivatives are not formed by free radical mechanisms that generate isoprostanes, and thus, these metabolites provide an unbiased marker for utilization of EPA by cyclooxygenases.

Initial hydrolysis of the bicyclic prostaglandin PGI 2 results in the formation of 6-keto-PGF 1 ␣ , whereas the hydrolysis of PGI 3 yields 6-keto-PGF 2 ␣ . The major urinary metabolites of AA-derived 6-keto-PGF 1 ␣ have been described in both man ( 12,17,18 ) and rats ( 19 ) but not in mice. In contrast, no careful, quantitative analysis of the metabolites of PGI 3 have been performed in any species, although homologs of the major PGI 2 metabolites have been measured in urine of primates following dietary fi sh oil (20)(21)(22). In the present study, we prepared 6-keto-PGF 2 ␣ from EPA bearing tritium and deuterium labels to form two different isotopomers in order to facilitate metabolite isolation from the urine matrix and product identifi cation using mass spectrometry, and we studied the metabolism of 6-keto-PGF 2 ␣ in mice.

Materials
All fatty acid methyl esters with purities of greater than 98% were purchased from NuCheck Prep (Elysian, MN). Anhydrous ethyl ether and tetrahydrofuran (THF) with water contents of less than 50 ppm were purchased from EMD Chemicals (Gibbstown, NJ) and used without further purifi cation. Chemicals used for metabolite extraction and solvents used for HPLC/MS analysis were purchased from Fisher Scientifi c (Pittsburgh, PA) as "HPLC grade" and used without further purifi cation. Stable isotopes were purchased from C/D/N Isotopes Inc. (Pointe-Claire, Quebec, Canada). Radioactive isotopes were purchased from and custom radioactive labeling was performed by Vitrax Radiochemicals (Placentia, CA). Silica gel "Selecto" 32-63 m for preparative column chromatography was purchased from Selecto Scientifi c (Suwanee, GA). Lipid standards used as mass spectrometric and chromatographic references (6-keto-PGF 1 ␣ ; 2,3-dinor-6-keto-PGF 1 ␣ ; and 6,15-diketo-PGF 1 ␣ ) were purchased from Cayman Chemicals (Ann Arbor, MI).

Enzymes and tissues
Human recombinant COX-2 was prepared as described previously ( 23 ) and used as a 400 g/ml solution in 50 mM phosphate buffered saline at pH 7.0. Bovine aorta was purchased from Animal Technologies Inc. (Tyler, TX). Soybean lipoxygenase type 1 was from Sigma-Aldrich Chemical Co. (St. Louis, MO).

Injection of mice with d0/d5 [ 3 H]6-keto-PGF 2 ␣
A solution of d0/d5 [ 3 H]6-keto-PGF 2 ␣ containing 0.2 g of d0-6-keto-PGF 2 ␣ , 0.2 g of d5-6-keto-PGF 2 ␣ , and 2 Ci of [ 3 H]6-keto-PGF 2 ␣ per ml of 0.9% NaCl was prepared. Briefl y, an ethanol solution of the d0/d5 [ 3 H]6-keto-PGF 2 ␣ prepared as described in the supplemental data was added to a sterile 10 ml glass vial and the sample dried under a stream of N 2 gas, then sterile saline was added, and the samples were shaken vigorously. Eight 8-week-old C57BL/6J male mice purchased from Jackson Laboratories were divided into two groups. Each mouse was injected with 0.2 g (1 Ci) of d0/d5 [ 3 H]6-keto-PGF 2 ␣ in 0.5 ml of sterile 0.9% NaCl. Following the injections, the mice were placed, four per cage, in metabolic cages for 24 h urine collections. Mice were allowed free access to water and mouse chow (LabDiet 5K67) during urine collections. The mice were euthanized by isofl urane inhalation. The urine was centrifuged to remove contaminating feces and stored at Ϫ 80°C until analyzed. All animal protocols were conducted in conformity with USPHS policy, and the experimental protocols were approved by the University Committee on Use and Care of Animals at the University of Michigan.

Isolation of 6-keto PGF 2 ␣ metabolites from urine
Metabolites of 6-keto PGF 2 ␣ with a 2,3-dinor structure were chemically isolated from urine using a protocol developed by Falardeau et al. ( 12 ). These dinor metabolites have a unique confi guration of a hydroxyl on carbon 9, a ketone at carbon 6, and a carboxylate at carbon 3 that can be manipulated with pH to form a ketal with subsequent lactone structure that can be more selectively purifi ed from the complex urine matrix.

Chemical derivatization of metabolites
Ketone functionalities of metabolites and standards were derivatized with methoxylamine using the following protocol: From 10 ng to 1 g of sample was dried at the bottom of a 10 ml screwtop glass tube. In a matching tube, 1 ml of 1N NaOH and then approximately 5 mg of methoxylamine hydrochloride were added, and the tubes were quickly joined with a female-female union, liquid side down, and incubated at 67 ° C for 2 h. The sample end of the apparatus was removed and suspended in HPLC solvents for RP LC/MS analysis.

Analysis and identifi cation of metabolites
Reverse-phase HPLC analysis was carried out for lipid analysis using a 2 × 150 mm 5 m particle size C-18 Luna column the crude oil [5] was purifi ed by column chromatography on 20 g of silica gel with 4% ethyl acetate in hexane used as a mobile phase to yield a colorless oil of prepurifi ed [5] (20.3 mg, 28%).
The mixture of methyl esters of monoacetylenic derivatives of EPA was subjected to selective hydrogenation using the Lindlar catalyst as detailed in the supplementary data. This hydrogenation method was used for preparation of tritium-labeled EPA by Vitrax Radiochemicals.

Biosynthesis 6-keto-PGF 2 ␣ [12] and d5-6-keto-PGF 2 ␣ [13]
as a mixture with uniformly tritium-labeled 6-keto-PGF 2 ␣ (at a ratio 2 g/ Ci) Bovine aorta (2-3 g) frozen in liquid nitrogen was crushed into fi ne pieces (0.5-5 mm 3 ) while frozen and fragile using an homogenizer. The pieces were mixed with 0.1 M TrisHCl (pH 7, 6 ml) containing 3 mg of fl urbiprofen per 50 ml of the buffer and transferred into a 10 ml glass tissue grinder in which they were ground at 0-5°C. The resulting mixture was centrifuged at 15,000 g for 15 min, and the supernatant was separated from the precipitate. This supernatant was used immediately after preparation, but it could be kept on ice (0-5°C) for 1-3 h. To prepare the substrate, 50 Ci of [ 3 H]EPA in 50 l of ethanol was placed in a 10 ml glass vial, the ethanol was evaporated with stream of nitrogen, buffer A (2.5 ml) was placed in the vial and thoroughly mixed, then 2.5 ml of deionized water was added, and then d5-EPA and d0-EPA (25 l of each as 6 mM solutions in ethanol) were added and mixed well. To this solution was added an aliquot of recombinant huPGHS-2 (200 l of an 0.8 mg/ml solution in 0.1 M TrisHCl, pH 8), and the sample was stirred vigorously for 120 s An aliquot of the ovine aorta extract (2.5 l) prepared as described above was added to the resulting mixture, the sample was mixed well, and was incubated at room temperature for 10 min. The enzymatic reaction was terminated by the addition of 0.2 M citric acid (2.5 ml). The mixture was stirred for 1 min and then centrifuged at 15,000 g for 5 min. The supernatant was carefully removed, and the target 6-keto-PGF 2 ␣ was purifi ed by RP-HPLC as follows. The supernatant (1,000 l) was injected onto an HPLC system equipped with an C18-RP-HPLC (Shodex RSpack DE-413L, 250 mm × 4.6 mm, 5 mm) column tive metabolites were also observed in the urine eluting at 16.5, 18, and 25 min. The two major metabolites became the target to drive purifi cation strategies because they represented over 20% of the initial 6-keto-PGF 2 ␣ administered to the mice based on the summed radioactivity in these two peaks compared with the total radioactivity eluting in Fig. 1A . Based on the observed retention times of the 6K1 and 6K2 being shorter than the starting 6-keto-PGF 2 ␣ , we reasoned that these were chain-shortened metabolites, and we essentially followed a procedure previously published by Falardeau et al. ( 12 ) that purifi ed the dinor metabolites of 6-keto-PGF 1 ␣ . This strategy made use of the unique chemistry innate to a 2,3-dinor 6-keto eicosanoid that could form a hemi-acetal lactone stable to mild basic conditions (supplementary Fig. I). Recovery of radioactivity at each step of the purifi cation procedure is presented in Table 1 . An aliquot of urine (1 ml) was adjusted to pH 10 and let react for 15 min prior to adjusting to pH 3 in order to cyclize a metabolite as a hemi-acetal lactone. This lactone-containing solution was subjected to reversed-phase solid-phase extraction (SPE), and the radioactivity was eluted with methanol (recovery 53%). This SPE extract was dried, redissolved in dichloromethane, and extracted with pH 8 borate buffer that ionized any free carboxylic (Phenomenex, Torrance, CA). Mobile phase A consisted of water and 10 mM ammonium acetate buffered to pH 4.5 with acetic acid. Mobile phase B was composed of acetonitrile and methanol at a ratio of 65:35. Flow rate was 200 l/min, and the gradient used for analysis started at 10% B, isocratic for 5 min, then to 40% B in 20 min, to 95% B in 5 min, then isocratic at 95% B for 5 min. Aliquots of the isolated metabolites from urine were dried and resuspended in HPLC solvents for analysis. The fl ow was split 3:1 between an in-line radioactivity counter (Flo-One Beta; Radiomatic, Tampa, FL) and mass spectrometer (QTRAP® 5500; AB Sciex, Foster City, CA) to establish correlation between chromatographic radioactivity and mass spectra of the component. Mass spectrometer conditions included the following conditions for all negative ion analyses: ion spray voltage, Ϫ 4,000; curtain gas, 30; nebulizer gas, 30; source temperature, 300°C ; declustering potential, Ϫ 30; and exit potential, Ϫ 25. Q3 scan range was 290-700 Da at a rate of 200 Da/s. The linear trap functionality was used intermittently in complimentary enhanced mass spectra mode, which enhanced mass resolution and sensitivity. In this modality, a scan rate of 1,000 Da/s was used with collision gas set at "low." Product ion spectra were acquired in enhanced product ion mode at a rate of 1,000 Da/sec, fi ll time 1 msec, and collision energy of Ϫ 30.

Overview
The metabolism of 6-keto-PGF 2 ␣ was carried out by peritoneal injections of this eicosanoid into male mice (four per group) with 1 Ci/mouse of tritium-labeled eicosanoid that additionally had a deuterium-labeled isotopomer so that the d0/d5 ratio was approximately 1:1. Urine from each group of mice was collected for 24 h. Therefore, these studies had three different isotopic variants of 6-keto-PGF 2 ␣ that facilitated development of optimal extraction and purifi cation protocols using radioactivity to detect elution from the HPLC and structural characterization of metabolites by mass spectrometry detected as isotope doublets. For these studies, the appearance of an isotopic doublet in the mass spectrum enabled assignment of an eluting component as an eicosanoid metabolite. Because the site of tritium label was not on the identical carbon atoms as the deuterium label, it was possible that a change in the isotope doublet pattern could be observed for a radiolabeled metabolite but unlikely that any reasonably large molecular weight metabolite would have lost all tritium atoms. Approximately 50% of the injected 6-keto-PGF 2 ␣ -associated radioactivity was eliminated in the urine; no obvious behavioral effects were noted following administration of this amount of 6-keto-PGF 2 ␣ .

Metabolites of 6-keto-PGF 2 ␣
Direct analysis of an aliquot of the pooled urine sample collected 24 h after injection of 6-keto-PGF 2 ␣ into the mice was surveyed by reversed-phase LC using a radioactivity monitor to detect elution of metabolites ( Fig. 1A ). Several metabolites were apparent with some components eluting early (before 4 min), indicative of very weakly lipophilic metabolites not well retained on the column and two abundant lipophilic metabolites eluting at 14 and 16 min, respectively (labeled 6K1 and 6K2). Other minor radioac- identical ion chemistry for m/z 339 and 344. Collision activation of m/z 339 from peak 6K1 ( Fig. 2A ) yielded losses corresponding to one water molecule ( m/z 321) , two water molecules ( m/z 303), and two water molecules and CO 2 ( m/z 259). In addition, there was an abundant ion corresponding to loss of 98 amu, which would correspond to the loss of three molecules of water and CO 2 , leading to a highly stabilized structure shown in Fig. 2A . This ion structure was supported by appropriate mass shifts in the observed ions from the deuterated metabolite ( m/z 246, supplementary Fig. IA) and the homologous ion formed from authentic 2,3-dinor-6-keto-PGF 1 ␣ ( m/z 243, supplementary Fig. IB). However, the ions formed following collision activation of m/z 339 from 6K2 ( Fig. 2B ) showed only the losses of one and two molecules of water, but not formation of the highly conjugated product ion discussed above ( m/z 241). The isolation strategy and reduction in molecular weight from the starting 6-keto-PGF 2 ␣ for each metabolite was consistent with a chain-shortened 2,3-dinor metabolite, but differences in HPLC retention time ( Table 2 ) suggested that one metabolite could have the prostanoid side chain intact while the other metabolite could have the carbon 13-14 double bond saturated and the hydroxyl group at C-15 oxidized to a keto moiety, which are known metabolic transformations of 6-keto-PGF 1 ␣ ( 12 ). acid groups present in any unrelated molecules excreted into the urine but was not suffi ciently strong to saponify the hemi-acetal lactone ( Table 1 , yield 38%). The organic layer was dried, then treated with pyridine/borate buffer to saponify the hemi-acetal lactone, and then this aqueous solution was extracted with ethyl acetate to remove neutral, unrelated molecules still present from the urine (yield 20%). The radioactive metabolites remaining in the aqueous layer were extracted into dichloromethane after adjusting to pH 2 using HCl to yield 13% purifi ed recovery based on the starting radioactivity in the urine ( Table 1 ). This extract was analyzed by reversed-phase LC ( Fig. 1B ) to reveal the same abundant metabolites eluting at 14 and 16 min as seen in the crude urine analysis. The total yield for this extraction protocol for these hemi-acetal lactones was approximately 13% of the radioactivity in the pooled urine sample, which corresponded to 6.5% of the injected 6-keto-PGF 2 ␣ . Even though there was signifi cant loss of radioactivity at each step ( Table 1 ), the solvent extraction yielded fairly clean metabolite isolation.
This extracted fraction was then subjected to LC-MS to detect the major negative ions for the components eluting from the HPLC ( Fig. 1C ). At the 14 min elution period, an isotope doublet was observed at m/z 339/344, and at the 16 min retention time, a more complex multiplet was observed, including a doublet at m/z 339/344 ( Fig. 1C , inset). Clearly there were radioactive metabolites as well as other components eluting at the 16 min time period, as evident by peaks at m/z 338 and 340. Nonetheless, these two urine components did reveal isotope doublets, thus strongly supporting the radioactivity data that there were metabolites of 6-keto-PGF 2 ␣ eluting at this time.
The major HPLC peaks ( Fig. 1C ) were then directly analyzed by LC-MS/MS following injection of an aliquot of this urine extract ( Fig. 2 ). Collisional activation of each ion at m/z 339 and 344 revealed product ions that were shifted in most cases by 5 amu (d5 isotope data, supplementary  Mouse urine was collected for 24 h after injecting 1 Ci of d0/d5 [ 3 H]6-keto-PGF 2 ␣ into the peritoneal cavity of each of four mice. The isolation steps are detailed in Experimental Procedures and involved converting 6-keto-PGF 2 ␣ metabolites into a unique hemi-acetal lactone that could be separated from other acidic metabolites in urine by differential extraction. DPM, disintegrations per min; RP-SPE, reversephase solid-phase extraction. groups to the metabolite 6K2 that had a [M-H] Ϫ at m/z 339 ( Fig. 2B ). The ion at m/z 397 (peak B) was collisionally activated ( Fig. 5A ), and as was the case for peak A, the deuterium-labeled analog eluting at 21.6 min had an identical CID mass spectrum that revealed 5 amu shifts in most product ions (data not shown). To have 2-keto groups and a molecular mass of 340 Da, one of the hydroxyl substituents in the starting 6-keto-PGF 2 ␣ had to be oxidized to a ketone and one carbon-carbon double bond had to be reduced. This type of metabolism is quite well known for the double bond ⌬ 13,14 and 15-hydroxyl moieties of many prostanoids ( 11,13,28,29 ). The molecular ion for this dimethoxime derivative yielded product ions corresponding to the losses expected for each of the methoxime derivatives (sequential losses of 30 amu) as well as the loss of acetic acid ( m/z 337.2), which is an interesting ion perhaps unique to these 2,3-dinor derivatives ( Fig. 5A ). Very few other product ions were present, making it diffi cult to unambiguously assign the exact structure for each of these metabolites eluting from the HPLC in this region as geometric isomers of the methoxime derivative. To test whether the dimethoxime derivative of this metabolite would form HPLC-separable geometric isomers, authentic 6,15-diketo-13,14-dehydro-PGF 1 ␣ was derivatized (MW 428 ) and separated by the same reversed-phase HPLC system, then analyzed by mass spectrometry ( Fig. 5B ). Four geometric isomers (syn/ anti at both keto positions) were clearly observed in the HPLC separation ( Fig. 5B , inset) at an expected 4 min longer retention time, because this synthetic eicosanoid had two additional carbon atoms. The CID spectra of four peaks in Fig. 3B-E (supplementary Fig. II) were all similar and consistent with the suggested structural features of metabolite 6K2 being 2,3-dinor-6,15-diketo-13,14-dihydro-PGF 2 ␣ .

DISCUSSION
Intact PGs are present only at low concentrations in urine and are not representative of whole-body PG synthesis. Measurements of circulating PGs in blood are also unreliable because of the rapid synthesis of PGs during sampling. Instead, measurements of urinary metabolites Because the difference between the suspected metabolites was the presence of one versus two keto groups, an aliquot of the urine extract was treated with methoxylamine hydrochloride to form the methoxime (MOX) derivative(s). Additional separation of these components was revealed in the radiochromatogram ( Fig. 3A ). Major radioactive peaks eluted at 16.9, 21.6, 23.0, 23.8, and 24.5 min. Because formation of a methoxime from a keto group can lead to two separable syn and anti isomers for this derivative, some of the components clearly could have been due to these geometric isomers ( Table 2 ). Peak A at 16.9 min ( Fig. 3A ) was found to contain an isotope doublet at m/z 368/373 after LC-MS analysis, and extracting the ion current for m/z 368 revealed a single HPLC peak at 16.9 min ( Fig. 3B ). Furthermore, the ions at this retention time revealed the isotope doublet at an approximately 1:1 ratio ( Fig. 3B , inset, m/z 368/373). The components at 21.6 (peak B), 23.0 (peak C), 23.8 (peak D), and 24.5 min (peak E) all revealed isotope doublets at m/z 397 and 402, and an elution profi le of m/z 397 ( Fig. 3C ) identical to the radiochromatogram ( Fig. 3A ) with appropriate doublets at m/z 402 for each retention time ( Fig. 3C , inset).

Characteristic ions corresponding to the loss of the methoxime derivative ( m/z 338, [M-H-30]
Ϫ ) as well as a very diagnostic ion at m/z 230.1 (structure in the inset of Fig. 4A ). Furthermore, all of these ions were advanced by 5 amu in the CID mass spectrum of the ion at m/z 373 ( Fig. 4B ). Authentic 2,3-dinor-6-keto-PGF 1 ␣ was derivatized to the methoxime and found to have a retention time 2 min later than peak A, consistent with one less double bond ( Table 2 ). Comparison of the CID mass spectrum from compound A to that of authentic derivatized 2,3-dinor-6-keto-PGF 1 ␣ , which lacks the -3 double bond ([M-H] Ϫ , m/z 370) ( Fig. 4C ), revealed similar CID behavior. Collectively, these data indicate that 6K1 ( Figs. 1 and 2 ) was derivatized to peak A and that this derivatized metabolite was formed from 2,3-dinor-6-keto-PGF 2 ␣ .
The obvious fi rst step in developing methods for quantifi cation of 3-series PG metabolites formed from the omega-3 fatty acid EPA is to identify the major urinary PG metabolites. The focus of the studies reported here was to identify the major urinary metabolites of PGI 3 . Initially, we prepared both [ 3 H]EPA and a very highly purifi ed form of d5-EPA, neither of which is commercially available. [ 3 H] EPA was synthesized by reduction with tritium gas of a mixture of EPA-alkynes derived from bromohydrin isomers. About 45% of the [ 3 H]label was incorporated into the ⌬ 17 double bound with more modest but similar levels of label at the other double-bond positions. Wittig-coupling of a 17-triphenyl phosphonium iodide derivative of 5Z,8Z,11Z, 14Z-methyl heptadecatetraenoate with d5-propanal was  PGF 2 , PGD 2 , and thromboxane B 2 , can be formed via nonenzymatic, free-radical oxidation reactions ( 29,38,39 ). Metabolites of isoprostanes are found in urine and in some cases have properties (e.g., HPLC retention times and mass spectroscopic behaviors) of enzymatically derived PGs. This can make it complicated to discriminate between enzymatically and nonenzymatically derived urinary metabolites of PGs other than PGI 2 and PGI 3 . PGI 2 can be formed from AA via either COX-1 or COX-2, but PGI 2 is produced, at least in humans, in three times greater abundance via COX-2 ( 40 ). EPA-derived PGI 3 is even more likely to be formed via COX-2 ( 8,9 ). This is because EPA is both an effective inhibitor and a poor substrate for COX-1 ( 8,23,41,42 ), but it is a modestly good substrate for COX-2 ( 8,23,41 ). Accordingly, measuring the formation of 6-keto-PGF 2 ␣ metabolites may prove to be a quantitative biomarker not only for omega-3 fatty acid tone but also, independently under appropriate dietary conditions, for relative COX-2 activity. Measurements of 6-keto-PGF 2 ␣ metabolites could also be used to assess the effectiveness of COX inhibitors on COX-2 in vivo.
The results of our studies of 6-keto-PGF 2 ␣ catabolism reveal that major urinary metabolites of this eicosanoid can be found and measured and that they retain the 3 double bond that distinguishes the 2-series from the 3-series prostacyclin metabolites. In the rat, where some detailed analytical work has been performed, one PGE 3 metabolite is a tetranor derivative in which four carbons have been cleaved from the carboxyl end via ␤ -oxidation ( 15,16 ); however, formation of another metabolite of PGE 3 involved -oxidation and loss of the ⌬ 17 double bond ( 16 ). An unexpected fi nding of the work is the abundance of two isomeric metabolites that retain the ⌬ 17 double bond and that could serve as measures of PGI 3 biosynthesis in mouse that had 3-polyunsaturated fatty acids in its diet.
We have yet to synthesize and test appropriate standards for measuring either of the 6-keto-PGF 2 ␣ metabolites we identifi ed. Therefore, we do not yet know whether measuring a 6-keto-PGF 2 ␣ metabolite in conjunction with a 2-series PG metabolite will be useful in estimating tissue AA/EPA ratios. The advantage of measuring prostacyclin metabolites is that they are only formed enzymatically (35)(36)(37). This is unlike other PGs. Isoprostane stereoisomers, having fi ve-member ring structures similar to PGE 2 ,