Bioactive products formed in humans from fish oils.

Resolvins, maresins, and protectins can be formed from fish oils. These specialized pro-resolving mediators (SPMs) have been implicated in the resolution of inflammation. Synthetic versions of such SPMs exert anti-inflammatory effects in vitro and when administered to animal models. However, their importance as endogenous products formed in sufficient amounts to exert anti-inflammatory actions in vivo remains speculative. We biased our ability to detect SPMs formed in healthy volunteers by supplementing fish oil in doses shown previously to influence blood pressure and platelet aggregation under placebo-controlled conditions. Additionally, we sought to determine the relative formation of SPMs during an acute inflammatory response and its resolution, evoked in healthy volunteers by bacterial lipopolysaccharide (LPS). Bioactive lipids, enzymatic epoxyeicosatrienoic acids (EETs), and free radical-catalyzed prostanoids [isoprostanes (iPs)] formed from arachidonic acid and the fish oils, served as comparators. Despite the clear shift from ω-6 to ω-3 EETs and iPs, we failed to detect a consistent signal, in most cases, of SPM formation in urine or plasma in response to fish oil, and in all cases in response to LPS on a background of fish oil. Our results question the relevance of these SPMs to the putative anti-inflammatory effects of fish oils in humans.

work (hematology, biochemistry, and urinalysis) at the time of screening and on completion of the study. Subjects were counseled not to make major changes in their diet, and to refrain from consuming any additional fi sh foods to minimize exposure to contaminant heavy metals, including mercury, apparent in fi sh products. Compliance with this regimen was assessed daily by real-time text messaging or emailing the intake of Lovaza TM capsules, weekly by capsule count in the Clinical and Translational Research Center, by the change of lipid ratios in red blood cell (RBC) membranes ( Fig. 1C ), and by diversion from AA-derived iPs and epoxides toward EPA/DHA-derived species ( Fig. 1D-H ).
Plasma samples were also available from a subset of healthy volunteers (n = 6), studied to evaluate the systemic infl ammatory response to experimental endotoxemia in subjects treated with lower doses of marine lipids delivered as Lovaza fi sh oil; please refer to ( 38 ) for further details (clinicaltrials.gov registration number: NCT01048502). In brief, this group of subjects was randomized to receive two capsules Lovaza TM twice a day (bid) for 8 weeks in a double-blinded fashion, followed by an intravenous bolus of US standard reference endotoxin (LPS; lot number CCRE-LOT-1 +2; Clinical Center, Pharmacy Department at the National Institutes of Health, Bethesda, MD) dosed at 0.6 ng/kg body weight ( Fig. 2A ).
A pilot study in 21 healthy participants established baseline values of urinary 5-epi -8,12-iso-iPF 3 ␣ -VI, 8,12-iso-iPF 2 ␣ -VI, and iPF 2 ␣ -III. All studies were approved by the Institutional Review Board and informed consent was obtained from the participants prior to study initiation. Blood (plasma). For the study High Doses of Lovaza Fish Oils , venous blood was drawn into Na-heparin vacutainer tubes after an overnight fast on six occasions: two times as baseline (12.2 ± 6.0 days and 3.0 ± 0.0 days before the start of fi sh oil supplementation), two times during fi sh oil supplementation (12.0 ± 0.0 days and 24.2 ± 2.3 days after the start of fi sh oil supplementation), and two times after discontinuation of fi sh oil supplementation (7.9 ± 0.3 days and 13.3 ± 0.8 days after the stop of taking fi sh oil). Plasma was separated by immediate centrifugation (3,000 rpm, 15 min, 4°C) and stored at Ϫ 80°C until extraction.

Biospecimens
For the study Evoked Endotoxemia , venous blood was drawn into citrate vacutainer tubes after an overnight fast on seven occasions: one time as baseline (before the start of fi sh oil supplementation), one time after 50.0 ± 6.8 days of fi sh oil supplementation (which coincided as baseline prior to the injection of LPS), and fi ve times after the intravenous injection of LPS (time course of 2, 6, 12, and 24 h post-LPS administration, as well as a followup any time between 48 and 72 h post-LPS). Plasma was separated by immediate centrifugation (3,000 rpm, 15 min, 4°C) and stored at Ϫ 80°C until extraction.

Blood (RBC membranes). For the study High Doses of Lovaza Fish
Oils , venous blood was drawn into EDTA vacutainer tubes after an overnight fast on six occasions: two times as baseline (12.2 ± 6.0 days and 3.0 ± 0.0 days before the start of fi sh oil supplementation), two times during fi sh oil supplementation (12.0 ± 0.0 days and 24.2 ± 2.3 days after the start of fi sh oil supplementation), and two times after discontinuation of fi sh oil supplementation (7.9 ± 0.3 days and 13.3 ± 0.8 days after the stop of taking fi sh oil). The tubes were immediately centrifuged at 3,000 rpm for 15 min (4°C) to separate the RBCs from plasma and the buffy coat. RBCs were transferred to a clean tube and washed twice with ice-cold isotonic saline, followed each time by centrifugation for 10 min (3,000 rpm, 4°C). Washed RBCs were stored in saline (33:67 v/v) of exogenous SPMs in these models, however, are substantial; 0.6 g resolvin (Rv)D1 per mouse, for example, was used to attenuate the lipopolysaccharide (LPS)-induced infl ammatory response in lung ( 18 ). These quantities are in marked contrast to the low picograms per milliliter concentrations reported in humans ( 19 ). Thus, the importance of SPMs as endogenous products formed in suffi cient amounts to exert an anti-infl ammatory action in vivo remains speculative. A particular limitation has been the use of assays in biological systems related to capacity, such as serum, rather than actual biosynthesis in a fi eld where a marked discrepancy between the two approaches has long been recognized ( 20 ). Here, the fact that detection of SPMs is confounded by cells activated ex vivo in the test tube often remains unrecognized. An additional limitation has been of a technical nature. There are few data attesting to their formation in humans based on rigorous mass spectrometric methodology. Most clinical investigations reporting the formation of SPMs rely on immunoassays ( 21 ) or MS without synthetic internal standards structurally similar to the SPMs under investigation (22)(23)(24)(25)(26)(27)(29)(30)(31)(32), often in ex vivo stimulated systems ( 17,(33)(34)(35)(36).
Here we biased our ability to detect SPMs formed in healthy volunteers by administering to them high doses of fi sh oils, such as have been shown to infl uence blood pressure and platelet aggregation under placebo-controlled conditions ( 11,12 ). As a comparator, we also analyzed enzymatic [epoxyeicosatrienoic acids (EETs)] and free radical [isoprostanes (iPs)] catalyzed prostanoids formed from AA and the fi sh oils. Additionally, we sought to determine the relative formation of SPMs and these bioactive lipids during an acute infl ammatory response and its resolution, evoked in healthy volunteers by administration of LPS. Despite a clear shift from -6 to -3 EETs and iPs, we failed to detect a consistent signal in most cases of SPM formation in urine or plasma in response to fi sh oil supplementation and in all cases in response to LPS on a background of fi sh oil supplementation.

Human studies
Samples were available from a human study (clinicaltrials.gov registration number: NCT00682318) assessing the interaction of high doses of marine lipids delivered as Lovaza fi sh oil with ethanol on lipid peroxidation. Healthy volunteers (n = 12) received, in an open single-arm design, seven capsules Lovaza TM three times a day (tid) for 24.2 ± 2.3 days ( Fig. 1A ). This dose delivered a total of 17.6 g/day -3 PUFA, consisting of 55.1% EPA (9.7 g/day) and 44.9% DHA (7.9 g/day). Prior to study enrollment, informed consent, approval by the Institutional Review Board, and authorization by the Food and Drug Administration (IND#79,750) had been obtained . Subjects were nonsmokers, not pregnant, and abstained from the use of high-dose vitamins, NSAIDs, and illicit drugs as assessed by cotinine (Craig Medical, Vista, CA) and pregnancy tests, history, platelet aggregometry ( 37 ), and a urine drug screen (RDI, Poteau, OK), respectively, for at least 2 weeks before enrollment and throughout the study. Health status and safety was assessed by routine medical history, physical exam, and laboratory AcN followed by 0.25 ml water. The cartridge was washed with 1 ml water, dried by application of vacuum, and eluted with 1 ml 50% AcN/ethyl acetate. The eluate was dried under a stream of N 2 and dissolved in 15 l AcN. For analysis, 135 l water was added and 75 l was injected into the tandem mass spectrometer .

URINE. Deuterium-labeled internal standards dissolved in 50 l
AcN were added to 2 ml of urine, gently mixed, and allowed to equilibrate for 15 min. The internal standard consisted of 5 ng each of d 4 -RvE1, d 2 -RvD1, d 2 -PD1, and d 2 -maresin. The sample was then acidifi ed with 20 l formic acid and applied to a StrataX SPE cartridge which was washed and eluted as described above. iPs were quantitated as described previously ( 47 ). In brief, 5 ng of each internal standard was added to 1 ml of urine. The SPE cartridge was conditioned with 1 ml of AcN and equilibrated with 0.25 ml of water. The sample was applied to the cartridge, which was then washed with 1 ml of 5% AcN in water and dried with vacuum for 15 min. The analyte and internal standards were eluted from the cartridge using 1 ml of 5% AcN in ethyl acetate. The eluate was collected and dried under a gentle stream of nitrogen. The resulting residue was reconstituted in 200 l of 20% AcN in water and fi ltered by centrifugation using 0.2 m Nylon Microspin fi lters (Spin-X HPLC Filter; Fisher Scientifi c). One hundred microliters were injected. Metabolite levels were corrected for urinary creatinine.
URINARY CREATININE. To 20 l urine, 1 ml AcN containing 10 g d 3 -creatinine was added. Ten microliters of this solution was injected into the tandem mass spectrometer. Mobile phase A consisted of AcN and mobile phase B consisted of 5 mM ammonium formate at pH 4.0. A Waters XBridge BEH HILIC (2.1 mm × 50 mm × 2.5 um) was used as the LC column. The LC was programmed at 3.5% mobile phase B at 350 l/min fl owrate.
LIPIDS FROM RBC MEMBRANES. Lipids were extracted using a modifi ed Folch method ( 48 ). RBCs (50 l), mixed in 1:3 ratio with saline, were used per sample, and further diluted with an additional 950 l of saline after thawing from the freezer. The RBCs were added to an internal standard (dried using nitrogen gas) containing labeled fatty acids (d 8 -AA, d 5 -EPA, and d 5 -DHA). The RBCs were homogenized in chloroform and methanol (2:1 v/v) containing 0.01% butylated hydroxytoluene as an antioxidant . The samples were then base hydrolyzed (KOH 7.5% in water) for 30 min. The pH was adjusted with formic acid to ‫ف‬ 2.5. The samples were purifi ed by SPE.
at Ϫ 80°C until extraction. For the study Evoked Endotoxemia , RBCs were obtained from venous blood as described in ( 38 ).
Urine. For the study High Doses of Lovaza Fish Oils , 12 h overnight urine samples were collected on eight occasions: two times as baseline (12.2 ± 6.0 days and 3.0 ± 0.0 days before the start of fi sh oil supplementation), four times during fi sh oil supplementation (5.0 ± 0.0 days, 12.0 ± 0.0 days, 19.1 ± 0.3 days, and 24.2 ± 2.3 days after the start of fi sh oil supplementation), and two times after discontinuation of fi sh oil supplementation (7.9 ± 0.3 days and 13.3 ± 0.8 days after the stop of taking fi sh oil supplementation). Urine aliquots with quality control samples were stored until extraction at Ϫ 80°C. . We verifi ed that the deuterated internal standards for EETs, DHA, and EPA were not contaminated by authentic unlabeled lipids (supplementary Fig. 6a). Please note that protectin D1 (PD1) and neuroprotectin D1 are structurally equivalent, while protectin DX (PDX) is an isomer ( 41 ). The d 3 -creatinine was purchased from CDN Isotopes (Quebec, Canada). Synthetic d 4 -8,12-iso -iPF 2 ␣ -VI and d 4 -5-epi -8,12-iso -iPF 3 ␣ -VI were synthesized by Joshua Rokach, PhD, Florida Institute of Technology, Melbourne, FL, as previously described, and used as internal standards (42)(43)(44)(45)(46). Burdick and Jackson solvents were purchased from Honeywell. All water was freshly generated Millipore, 18.2 M ⍀ .  . Authentic PD1 appears in plasma in concentrations close to the level of detection; limitations are, however, that wide peaks compromise clear identifi cation and that concentrations are not dose dependent. C: Mean ± SD plasma concentrations of plasma PD1 before, during, and after supplementation of healthy volunteers (n = 12) with high doses of fi sh oil; the red dotted line represents the approximated limit of detection for PD1 (left). The relationship of PD1 (ordinate) and plasma 19(20)-EpDPE (abscissa) shows a weak association between the two species derived from DHA; the area shaded in gray depicts the confi dence curve fi t (right). D: Ratios of EPA/AA and DHA/AA in RBC membranes before, during, and after supplementation with high doses of fi sh oil. E: Urinary 8,12-iso -iPF 3 ␣ -VI, derived from EPA and DHA, as percent change from baseline (red dotted line). F: Urinary 8,12-iso -iPF 2 ␣ -VI, derived from AA, as percent change from baseline (red dotted line). Baseline was averaged from n = 3 measurements during the 2 weeks prior to the start of supplementation with fi sh oil. G: Plasma 17(18)-EpETE (left), derived from EPA, and plasma 19(20)-EpDPE (right), derived from DHA, as percent change from baseline (red dotted line). H: Several EETs in plasma, derived from AA, as percent change from baseline (red dotted line). Boxplots indicate median, 25% and 75% quartiles and whiskers drawn to the furthest point within 1.5× interquartile range from the box. Note that the unit range on the ordinates varies considerably.   (20)-EpDPE (abscissa) shows a weak association between the two species derived from DHA; area shaded in gray depicts the confi dence curve fi t (top right). Formation of maresin, also derived from DHA, is not associated with 19(20)-EpDPE (bottom right). E-K: The time course of TNF-␣ (E), CRP (F), MCP-1 (G), IL-6 (H), IL-10 (I), SAA (J), and IL1-RA (K) in plasma is shown as percent change from baseline, i.e., pre-fi sh oil measurements. Note that time point zero (0) denotes the condition pre-LPS which coincides with day 50 ± 6.8 of supplementing the healthy volunteers with two capsules Lovaza TM bid. L: Urinary 8,12-iso -iPF 3 ␣ -VI, derived from EPA and DHA, as percent change from baseline (pre-fi sh oil measurements, red dotted line) (left); urinary 8,12-iso -iPF 2 ␣ -VI, derived from AA, as percent change from baseline (pre-fi sh oil measurements, red dotted line) (right). Note that time point zero (0) denotes the condition pre-LPS which coincides with day 50 ± 6.8 of supplementing the healthy volunteers with two capsules Lovaza TM bid. M: Plasma concentrations in picograms per milliliter of EETs formed from AA (left), EpETEs derived from EPA (center), and EpDPEs formed from DHA (right) before and after 8 weeks of supplementing with two capsules Lovaza TM bid. Boxplots indicate median, 25% and 75% quantiles and whiskers drawn to the furthest point within 1.5× interquartile range from the box. Note that the unit range on the ordinates varies considerably.

MS
LC/MS/MS analysis. All lipid analyses were performed on a Xevo TQ-S tandem mass spectrometer interfaced with an Acquity ultra-performance liquid chromatograph (Waters, Milford, MA), except for the analysis of urinary iPs, which was run on a TSQ Quantum Ultra instrument (Thermo; please see instrument parameters below). The mass spectrometer was operated in the negative ion ESI multiple reaction monitoring (MRM) mode using argon as the collision gas. The ESI probe capillary potential was 2.8 kV. The source was maintained at 150°C; the desolvation temperature was 350°C. All samples were chromatographed on a BEH C18 UPLC column (150 mm × 2.1 mm × 1.7 particle size; Waters) with a fl ow rate of 350 l/min. Mobile phase A consisted of Millipore water containing 0.05% acetic acid, adjusted to pH 5.7 with ammonium hydroxide. Mobile phase B was AcN/ methanol (95:5). Solvent AcN Optima® (Fisher Scientifi c) and methanol (Burdick and Jackson) were HPLC grade and met ACS specifi cations .

LC/MS/MS analysis of SPMs in urine.
The UPLC gradient was held at 10% B for 1 min, followed by a linear gradient to 22% at 10 min, 38% at 13 min, and 48% at 20 min. The column was then washed for 2 min with 100% B before returning to initial conditions, which were held for a 5 min equilibration period.

LC/MS/MS analysis of iPs in urine.
The HPLC included an Accela solvent delivery system (Thermo) and a Hypersil GOLD C18 ( 2 ), 200 mm × 2.1 mm, 1.9 m particle size column (Thermo). The mobile phase consisted of water (solvent A) and acetonitrile: methanol (95:5, solvent B), both with 0.005% acetic acid adjusted to pH 5.7 with ammonium hydroxide. The fl ow rate was 350 l/min. The mobile phase gradient began at 5% B, increased linearly to 18.5% B at 45 min, then to 38% B at 65 min. A TSQ Quantum Ultra instrument (Thermo) equipped with a heated electrospray source and a triple quadrupole analyzer were used in these studies. The ESI source used nitrogen for both sheath and auxiliary gas, set to 70 and 0 arbitrary units, respectively. The mass spectrometer was operated in the negative ion mode with a capillary temperature of 350°C and a spray voltage of 0 kV. The source offset was 6 V. The analyzer was operated in the MRM mode for the analysis of urinary iPs. The transitions monitored were: m/z 353 > 115 for the endogenous 8,12-iso -iPF 2 ␣ -VI and m/z 357 > 115 for the corresponding tetradeuterated internal standard. The CE was 24 eV. For the endogenous 5-epi -8,12-iso -iPF 3 ␣ -VI, the transitions monitored were m/z 351 > 115 and m/z 355 > 115 for the corresponding tetradeuterated internal standard. The CE was 23 eV.
LC/MS/MS analysis of SPMs/epoxides in plasma. The UPLC was held at 10% B for 1 min, and then programmed linearly to 22% at 10 min, 40% at 14.5 min, and 53% at 27.5 min. The column was then washed with 100% B for 2 min and allowed to equilibrate at initial conditions for 5 min prior to the next injection.

LC/MS/MS analysis of creatinine in urine.
A Thermo Ultima LC/MS/MS interfaced to an Accela UPLC (Thermo) was used in the positive ion chemical ionization MRM mode using argon as the collision gas. The MS interface capillary was held at 350°C. The chemical ionization probe was set at 350°C. The discharge current was 4.0 A. The column was an XBridge BEH HILIC, 50 mm × 2.1 mm × 2.5 particle size (Waters). Mobile phase A was AcN; mobile phase B was 100 mM ammonium formate, adjusted to pH 4.0 with acetic acid. The UPLC program was isocratic at 7.5% B. Transitions monitored were m/z 114 → 86 (creatinine) and m/z 117 → 89 (d 3 -creatinine) at CE 12 eV .

Statistical analyses
Descriptive statistics were executed in JMP® Pro Version 10. Means ± SDs are depicted unless otherwise stated. Averaged baseline measurements served as the reference point to calculate percent changes from baseline. Tests of signifi cance were not conducted due to the exploratory nature of this study.

Human subjects and safety
Twelve subjects concluded the study High Doses of Lovaza Fish Oils (supplementary Table 1); one subject discontinued due to personal reasons (CONSORT diagram in supplementary Fig. 1a). Three subjects deviated from the supplementation regimen of seven capsules Lovaza TM tid; that is, subject identifi cation number (ID#) 31 took six capsules instead of seven on four occasions; morning dose on days 5 and 10 of supplementation, evening dose on day 8, and midday dose on day 12 during the 24 day dosing period due to diffi culties in swallowing the capsules. Subject ID# 34 missed the morning dose on day 14. Due to oily diarrhea, the period of fi sh oil supplementation was shortened by 11 days and reduced to four capsules Lovaza TM tid (5,580 mg/day EPA and 4,500 mg/day DHA) for the last 5 days by subject ID# 34. Similar adverse events in subject ID# 37.1 accounted for the reduction to four capsules of Lovaza TM tid after 5 days of protocol-conform fi sh oil supplementation. The safety laboratory assessments at the screening and exit visits were both within normal range for the complete blood count, blood chemistry, lipid, and coagulation panels (supplementary Table 2). Spontaneous blood pressure was, on average (across all clinical visits), 120 ± 13 mmHg over 71 ± 10 mmHg (supplementary Fig. 1b); spontaneous heart rate was 71 ± 12 beats/min (supplementary Fig. 1c); oral temperature was 97.6 ± 0.9°F. Serious adverse events did not occur. The adverse event profi les are listed in supplementary Table 6 for the High Doses of Lovaza Fish Oils study and in supplementary Table 7 for the Evoked Endotoxemia study .

Lipidomics
SPMs. RvD1 and RvE1 were not detected in plasma samples from both clinical studies, as evident in the chromatograms of Figs. 1B, 2B . For PD1, peaks in plasma were seen which met the LC and MS/MS criteria for authentic material (chromatograms of Figs. 1B, 2B ). Baseline PD1 concentrations in plasma amounted to 1.0 ± 0.8 pg/ml and 1.2 ± 0.8 pg/ml, which rose to 4.1 ± 2.8 pg/ml and 2.7 ± 1.4 pg/ml after 12 and 24 days, respectively, of supplementation with high doses of fi sh oil; PD1 dropped to 1.7 ± 1.3 pg/ml and 1.4 ± 0.9 pg/ml at 8 and 13 days after LC/MS/MS analysis of lipids from RBC membranes. The analyzer was operated in the MRM mode for the analysis of the fatty acids EPA, DHA, and AA. The transitions monitored were m/z 301.

Limit of detection calculations
RvD1, RvE1, and maresin were not observed by UPLC/MS/ MS in the study involving administration of high doses of Lovaza fi sh oils . To determine the actual limit of detection of the samples in question, we identifi ed and integrated one or more peaks, with a signal-to-noise ratio of three or greater, that eluted near the expected retention time for the compound in question. When the areas of these peaks are divided by the areas of the deuterated standards, we obtain a value in picograms (supplementary Fig. 3a). Because no peak was observed at the correct retention time, the amount is therefore less than that of the selected peaks. For the RvD1 samples shown in Fig. 1B , the integrated peaks ranged from 3.4 to 4.3 pg. When the volume of the sample is taken into account, this calculates to a range of 3.9 to 4.2 pg/ml of plasma. For the RvE1 samples in Fig. 1B , the limit of detection ranges from 7.8 to 10.4 pg/ml of plasma. For maresin, the limit of detection ranges from 3.4 to 3.6 pg/ml of plasma. A limit of detection of 0.2-0.3 pg/ml of plasma was approximated for PD1 using the signals depicted in Fig. 1 . Noise was calculated as root mean square from selected time windows in the chromatograms. The limit of detection was then extrapolated to discriminate a signal from noise at a ratio of 3:1.
In the study of lower dose Lovaza fi sh oils in which volunteers were administered LPS, RvD1 and RvD2 were not observed. Calculations analogous to those above show the limit of detection for the RvD1 samples shown in Fig. 2B to range from 3.4 to 6.1 pg/ml of plasma. For the RvE1, the range is from 6.8 to 10.0 pg/ml of plasma. The limit of detection for PD1, based on the signals of Fig. 2 and determined as described above, ranged from 0.1 to 0.3 pg/ml. Notably for one sample, the limit of detection was one order of magnitude lower at 0.02 pg/ml.

Recovery
The analytes with homologous internal standards were RvE1, PD1, maresin, 5 (6) The recovery of each compound for which a homologous internal standard was available was calculated from the average area of the UPLC/MS/MS internal standard peak from each sample divided by the average area of two internal standards that were not subjected to SPE. In detail, the recoveries amounted to 76% for RvE1, 77% for PD1, 71% for maresin, 69% for 5(6)-EET, 81% for 8(9)-EET, 84% for 11(12)-EET, and 83% for 14(15)-EET. The recovery was not calculable for the compounds for which a homologous internal standard was unavailable.

Standard curve
The relative UPLC/MS/MS response of compounds, for which d 11 -14(15)-EET was used as a heterologous internal standard, was obtained by generating a six-point standard curve, 0, 1, 5, 10, 50, and 100 ng d 11 -14(15)-EET in triplicate, in which the amount of d 11 -14(15)-EET was constant and the amount of the target analyte varied (supplementary Fig. 5). The equations derived from the linear regression from these standard curves were used to adjust the response of the analytes in the actual samples.

Ratios of -3 and -6 PUFA in RBCs.
The ratios of -3 to -6 PUFA were assessed in erythrocytes to monitor the compliance with the high dose supplementation regimen. The relationship of EPA/AA rose by one order of magnitude from pretreatment 0.012 ± 0.004 and 0.011 ± 0.003 to 0.096 ± 0.031 and 0.113 ± 0.033 by days 12 and 24 of supplementation, falling to 0.073 ± 0.021 and 0.065 ± 0.017 at 1 and 2 weeks post-intervention. The ratio of DHA/AA in erythrocytes increased from 0.108 ± 0.025 and 0.103 ± 0.022 to 0.136 ± 0.024 and 0.158 ± 0.02 on days 12 and 24 on fi sh oil to 0.169 ± 0.021 for both 1 and 2 weeks post-fi sh oil ( Fig. 1D ).

Markers of infl ammation
Administration of LPS elevated a panel of cytokines and acute phase proteins ( Fig. 2E-K ). The maximum increase in plasma concentrations of TNF-␣ occurred 2 h postadministration of LPS, from initially 0.8 ± 0.3 pg/ml prior to LPS to 12.1 ± 4.3 pg/ml. CRP rose from 0.6 ± 0.5 pg/ml to 10.5 ± 7.1 pg/ml at 24 h after LPS; MCP-1 from 142.0 ± 22.9 pg/ml to 1,919.6 ± 703.4 pg/ml at 4 h after LPS; IL-6 discontinuation of fi sh oil ( Fig. 1C , left). PD1 concentrations in plasma after 4 g fi sh oil per day over 8 weeks of 3.1 ± 1.2 pg/ml were not different from baseline PD1 of 3.0 ± 1.5 pg/ml. Acute LPS had no discernable impact on PD1 concentrations in plasma, either in the infl ammatory or resolution phases, as evident from 2.9 ± 1.0 pg/ml at 2 h post-LPS, followed by 2.5 ± 1.8 pg/ml, 2.8 ± 2.8 pg/ml, 2.7 ± 1.5 pg/ml, and 2.8 ± 0.4 pg/ml at 6, 12, 24, and 48-72 h post-LPS, respectively ( Fig. 2D , top left). PDX, the isomer of PD1, was present in plasma samples from the evoked endotoxemia study in low picogram per milliliter concentrations, but without apparent modulation by either fi sh oil or LPS (supplementary Table 3, supplementary Fig.  3b). While maresin was consistently below the limit of detection before and after supplementation with high doses of fi sh oil, plasma samples before and after the administration of LPS showed mean concentrations between 4 and 6 pg/ml. In detail, 4.2 ± 1.4 pg/ml maresin as baseline before the initiation of fi sh oil was comparable to 4.3 ± 1.8 pg/ml after 8 weeks of supplementation. At 2 h post-LPS, maresin amounted to 6.3 ± 2.8 pg/ml, followed by 4.2 ± 3.5 pg/ml, 5.0 ± 1.6 pg/ml, 4.5 ± 3.0 pg/ml, and 5.8 ± 3.5 pg/ml at 6, 12, 24, and 48-72 h post-LPS administration, respectively ( Fig. 2D , bottom). Rvd1, RvE1, maresin, and PD1 were not detected in urine from subjects supplemented with high doses of fi sh oil (chromatograms not shown).  Table 4). The -6 PUFAderived epoxides in plasma, EETs, were not infl uenced by fi sh oil administration ( Fig. 1 , supplementary Table 4).

DISCUSSION
Derivatives of fi sh oils have been implicated in a variety of biological activities ( 49 ), including the resolution of infl ammation ( 50 ). Although the therapeutic effi cacy of dietary supplementation with fi sh oils remains controversial ( 6,7 ), observational studies have correlated diets rich in fi sh with desirable clinical outcomes ( 1,51 ). Whether this relates to a substitution of dietary saturated fatty acids, a direct effect of fi sh oils, or confounding lifestyle factors associated with a diet rich in fi sh is unknown. When fi sh oils, such as EPA and DHA, substitute for dietary AA, there is a shift in formation from dienoic prostanoids (e.g., PGE 2 ) to trienoic compounds (e.g., PGE 3 ), which may differ in biological properties when applied in vitro ( 52 ). In recent years, there has been considerable interest in the transcellular formation of bioactive lipids, such as might result from substrate exchange between platelets, neutrophils, and endothelial cells ( 50 ). In particular, transcellular products formed from fi sh oils (resolvins, maresins, and protectins) have exhibited anti-infl ammatory properties when administered in vitro or to animal models ( 53,54 ). While such observations highlight the therapeutic potential of synthetic SPMs, their relevance as endogenous mediators of infl ammation is more controversial. Interpretation of the literature is constrained by two limitations: i ) the use of nonspecifi c immunoassays; and ii ) the extrapolation of assays of cellular capacity to form SPMs, often with manipulation ex vivo, to infer actual rates of biosynthesis. Both limitations confounded estimates of prostanoid formation in the past; thus the capacity to form prostaglandins greatly exceeds actual biosynthetic rates ( 55 ), and immunoassays often grossly overestimated amounts of these compounds detected in biological fl uids by more precise physicochemical methods ( 56 ).
Here we sought evidence for formation of bioactive lipids at baseline in healthy volunteers receiving either a high dose of marine lipids delivered as Lovaza fi sh oil (21 g/day), known to modulate blood pressure and platelet function ( 11,12 ), or a dosing regimen (4 g/day) more commonly used in clinical trials designed to seek therapeutic effi cacy ( 57 ). In the latter case, we further characterized the lipidomic response to LPS, both during the evoked infl ammation and during the period of its resolution.
In both studies, supplementation with fi sh oils shifted the bioactive lipid profi le in vivo. This was apparent in a supplementation-related increase in the ratio of EPA and DHA to AA in erythrocyte membranes and in the endogenous biosynthesis of -3-derived products of the epoxygenase enzymes and free radical-catalyzed formation of iPs. Dose-dependent changes in these compounds were observed, exemplifi ed by the respective 3-and 10-fold increase from baseline for the EPA-derived epoxide, 18(19)-EpETE, at the two doses of fi sh oils. By contrast, SPMs were either undetectable or, when identifi ed in trivial amounts, appeared largely unrelated to the periods or doses of fi sh oil supplementation. The infl ammatory response to LPS administration was refl ected by evoked cytokines; SPMs were not altered in response to LPS during either the infl ammatory (2-4 h post-administration) or resolution phases (samples measured out to 48-72 h) in these healthy volunteers.
In the case of RvD1 and RvE1 in plasma, we utilized deuterated authentic derivatives as internal standards for their detection by MS and established lower limits of detection in plasma ranging from 3.4 to 4.3 pg/ml for RvD1 and from 7.8 to 10.4 pg/ml for RvE1. By contrast, PD1 levels were detectable, and doubled from ‫ف‬ 2 to 4 pg/ml in plasma during supplementation with 21 g/day Lovaza and fell after the period of supplementation ( Fig. 1C ). PD1 was also detectable on the lower dose of fi sh oils, but did not alter during the infl ammatory or resolution phases post-LPS ( Fig. 2D ), suggesting that nonenzymatic formation of PD1 has to be considered (supplementary Table 8, supplementary Fig. 6b). Low concentrations of maresins were also detected, but apparently unaltered by either dietary supplementation or administration of LPS. For our analyses, we have no evidence that signals of the authentic endogenous SPMs might be suppressed by the respective deuterated internal standards (supplementary Table 9, supplementary Fig. 6c). SPMs were not detectable in urine samples from subjects with high dose supplementation of fi sh oils, however, some maresin and RvD1 could be present in its glucuronidated form (supplementary Tables  10-13, supplementary Fig. 6e, f).
Few studies have utilized MS to assess SPM formation in response to fish oil supplementation in humans ( 19 ) and none have assessed their formation during resolution of evoked infl ammation in vivo. Average plasma levels of RvE1 of 0.521 nM (182.6 pg/ml) and of RvD1 of 0.0454 nM (17.1 pg/ml) were reported in a study of healthy volunteers on an ad libitum diet ( 58 ). By contrast, a randomized and placebo-controlled supplementation of 16 volunteers with 3 g/day fi sh oil over 10 weeks failed to detect RvD1 and RvE1 in plasma ( 59 ). However, labeled authentic standards were not applied in either study.
Although there is a paucity of data relating synthetic SPM administration with attained systemic exposure (60)(61)(62)(63)(64) and functional effects in mouse models ( 28 ), administration of 100 ng per mouse RvD1 resulted in plasma levels of immunoreactive RvD1 peaking at 850 pg/ml 3 h later and descending to 200 pg/ml over the ensuing 24 h, during which leukocyte infi ltration in zymosan A-induced acute peritonitis was reduced. While no such data are available using physicochemical estimates of RvD biosynthesis, assumption of the analytical validity of the approach in this study infers an anti-infl ammatory effect at concentrations greatly in excess of the limits of detection of our analysis.
In conclusion, we have failed to detect RvD1 or RvE1 in the plasma of volunteers administered fi sh oils or in response to an acute infl ammatory stimulus after administration of LPS. While maresins and PD1 were detected at low concentrations, they were unaltered during evoked infl ammation or its resolution. This was in marked contrast to the formation of epoxygenase and free radical-catalyzed products of fi sh oils. For now, the results of these studies question the relevance of endogenous SPMs to the putative anti-infl ammatory effects of fi sh oils in humans.

Note added in proof
The author Xuanwen Li was inadvertently left out of the author list of the accepted version of this article. All other authors and the Journal's accepting Associate Editor approved the addition after the article was in proof stage. Dr. Li will appear as an author in all forms of the article except in the originally accepted Paper in Press .