Development of a Chiral Supercritical Fluid Chromatography–Tandem Mass Spectrometry and Reversed-Phase Liquid Chromatography–Tandem Mass Spectrometry Platform for the Quantitative Metabolic Profiling of Octadecanoid Oxylipins

Octadecanoids are broadly defined as oxylipins (i.e., lipid mediators) derived from 18-carbon fatty acids. In contrast to the well-studied eicosanoids, there is a lack of analytical methods for octadecanoids, hampering further investigations in the field. We developed an integrated workflow combining chiral separation by supercritical fluid chromatography (SFC) and reversed-phase liquid chromatography (LC) coupled to tandem mass spectrometry detection for quantification of a broad panel of octadecanoids. The platform includes 70 custom-synthesized analytical and internal standards to extend the coverage of the octadecanoid synthetic pathways. A total of 103 octadecanoids could be separated by chiral SFC and complex enantioseparations could be performed in <13 min, while the achiral LC method separated 67 octadecanoids in 13.5 min. The LC method provided a robust complementary approach with greater sensitivity relative to the SFC method. Both methods were validated in solvent and surrogate matrix in terms of linearity, lower limits of quantification (LLOQ), recovery, accuracy, precision, and matrix effects. Instrumental linearity was good for both methods (R2 > 0.995) and LLOQ ranged from 0.03 to 6.00 ng/mL for SFC and 0.01 to 1.25 ng/mL for LC. The average accuracy in the solvent and surrogate matrix ranged from 89 to 109% in SFC and from 106 to 220% in LC, whereas coefficients of variation (CV) were <14% (at medium and high concentrations) and 26% (at low concentrations). Validation in the surrogate matrix showed negligible matrix effects (<16% for all analytes), and average recoveries ranged from 71 to 83%. The combined methods provide a platform to investigate the biological activity of octadecanoids and expand our understanding of these little-studied compounds.


I) Analytical workflow
The developed analytical workflow includes chiral and achiral characterization of octadecanoids, which are first extracted using solid phase extraction (SPE), then analyzed by chiral supercritical fluid chromatography coupled to tandem mass spectrometry (SFC-MS/MS) and, finally, by reversed phase liquid chromatography coupled to MS/MS (LC-MS/MS). To monitor column and instrument performance, as well as retention time stability, a system suitability test (SST) solution containing all of the target octadecanoids (including all available regioisomers and diastereoisomers) at concentrations ranging 1.0-10.0 ng/mL was prepared, aliquoted in single-use vials, and stored at -80℃. Single vials were thawed prior to analysis of samples and repeatedly injected until a stable signal for all analytes was achieved. Figure S1. Scheme of the analytical workflow Figure S1. Scheme of the analytical workflow. Workflow overview from sample preparation to SFC-and LC-MS/MS analyses, including the primary parameters employed during the various workup steps. Residual sample volumes enable the further analysis of eicosanoids (e.g., prostaglandins, isoprostanes, and leukotrienes) by employing our previously published method 1 .  b Peak numbers refer to the elution order in the two analytical methods. c Compounds were purchased from either Cayman Chemical (Ann Arbor, MI, USA) or Larodan AB (Solna, Sweden). Non-commercial compounds are described as "in-house" and were custom synthesized as described in Section III Description and characterization of in-house octadecanoid syntheses. Alternatively, a citation is provided for those compounds previously synthesized. d These compounds were in-house synthesized and the details of their synthesis have been submitted for publication. e These compounds are formed from the 20-carbon PUFA DGLA and are subsequently not octadecanoids. f Following discussions generated by the first version of this manuscript during early stages of the publication process, Cayman Chemicals started the production of these compounds and now include them in its catalogue. * Order of elution not confirmed with enantiopure standard due to unavailability but inferred by comparison with similar compounds under the same conditions. # Eluted as three non-resolved peaks, integrated as a single peak. S12 Figure S2. Octadecanoid formation by auto-oxidative processes Figure S2. Octadecanoid formation by auto-oxidative processes. Major pathways of ROS-and lipid autoxidation-derived octadecanoid formation from linoleic acid (LA) and α-linolenic acid (ALA). See Table S1 for a list of complete octadecanoid nomenclature. S13

III) Description and characterization of in-house octadecanoid syntheses
Analytical and chromatographical methods. Gas chromatography-mass spectrometry (GC-MS) was performed using an Agilent mass selective detector model 5977E connected to an Agilent model 7820A gas chromatograph. A capillary column of 5% phenylmethylsiloxane (12 m, 0.33 µm film thickness) with helium as the carrier gas was used. The temperature was raised from 80℃ to 320℃ at a rate of 10℃/min. Reversed-phase HPLC (RP-HPLC) was performed with a column of 250 x 10 mm Nucleosil 100-7 C18 whereas a column of 250 x 10 mm Nucleosil 50-7 was used for straight-phase HPLC (SP-HPLC). In both cases the flow rate was 4 mL/min.  10(R,S)-Hydroxy-12(Z),15(Z)-octadecadienoic acid. The title compound was prepared by two successive acetylene couplings as shown in Scheme A. Trimethylsilylacetylene (28.6 mmol) in 85 mL of THF was treated under argon at -78℃ with n-butyllithium (30 mmol). After stirring at -78℃ for 1 h a solution of methyl 10,11-epoxyundecanoate (23.8 mmol) in 16 mL of THF was added followed by BF3-Et2O (34.8 mmol). After stirring at -78℃ for 1.5 h, saturated ammonium chloride was added and the product extracted with diethyl ether. The product (7.65 g of a pale-yellow oil) was dissolved in 300 mL of methanol-THF (1:1, v/v) and stirred with K2CO3 (17 mmol) at 22℃ for 15 h. Extraction with diethyl ether followed by silica gel column chromatography (diethyl ether -hexane (3:7, v/v)) afforded the deprotected alkyne as a pale-yellow viscous oil (4.8 g, 19.8 mmol, yield from epoxyundecanoate, 83%).
By using 9(R,S)-and 13(R,S)-HpODEs 6 as starting materials instead of lipoxygenase-derived hydroperoxides it was possible to generate TriHOMEs in racemic form using the same methodology. Another way of preparing racemic TriHOMEs was by autoxidation of linoleic acid followed by separation of products by TLC and SP-HPLC 4,5 .

Dihydroxyoctadecenoates (DiHOMEs) and dihydroxyoctadecadienoates (DiHODEs).
Vicinal diol derivatives of linoleic and -linolenic acids were prepared by perchloric acid-catalyzed hydrolysis of the above EpOMEs and EpODEs or purchased from Larodan Co. S16

Threo-12,13-Dihydroxy-9(Z)-octadecenoic acid enriched with the 12(R),13(R) enantiomer.
Earlier work in our laboratory has shown that acetolysis of (+)-vernolic acid (12(S),13(R)epoxy-9(Z)-octadecenoic acid) takes place with a slight preference for the homoallylic position, thus producing an excess of the 12(R)-acetoxy-13(R)-hydroxy derivative relative to the 12(S)-hydroxy-13(S)-acetoxy compound. In addition, a slight separation of these two regioisomers on TLC allowed further enrichment of the first-mentioned compound. Accordingly, (+)-vernolic acid methyl ester (25 mg) and glacial acetic acid (1 mL) were kept at 70 o C for 4 h. The product was subjected to preparative TLC using a solvent system of ethyl acetate-hexane (25:75, v/v). The upper half of the band visualized by spraying with 2´,7´dichlorofluorescein was recovered and the product consisting of methyl 12-acetoxy-13hydroxy-and 12-hydroxy-13-acetoxyoctadecenoates in proportion 3:1 according to analysis by GC-MS was saponified by treatment with 0.2 M NaOH in methanol-water 7:3 at room temperature for 18 h. The product was crystallized from diethyl ether-hexane at -20 o C to provide a sample ready for analysis by chiral-phase HPLC.
Threo-9,10-Dihydroxy-12(Z)-octadecenoic acid enriched with the 9(R),10(R) enantiomer. The title compound was prepared from (+)-coronaric acid methyl ester (methyl 9(R),10(S)-epoxy-12(Z)-octadecenoate, 56 mg) using the protocol given above. However, in this case the lower part of the TLC band, in which the 9-hydroxy-10-acetoxy isomer was enriched, was used. Method costs. The expense of purchasing analytical standards is a concern in establishing a new method. For the octadecanoid platform, the broadly estimated cost of purchasing the 47 currently commercially available standards is 11,500€ (including labeled IS). If the custom synthesized standards are costed at the same price as their commercially available structural analogs (e.g., the price of trans-12,13-EpOME, commercially available, was used to estimate the price of trans-9,10-EpOME, not yet commercially available), the estimated cost for all standards is 19,500€ (including labeled IS). Based upon these costs, a minimum of 8500 calibration curves can be prepared (with the labeled IS as the limiting factor, which would need to be re-purchased every 10,000 samples) at a cost ~2.45€ per curve. While many standards are not yet commercially available, they are being actively produced by commercial companies and availability should increase in the near future. In the interim, all co-authors have agreed to make analytical quantities available for research purposes upon request. S17 Synthesis scheme A. Synthesis of 10-hydroxy-12(Z),15(Z)-octadecadienoic acid. a, nbutyllithium, BF3-Et2O; b, K2CO3, MeOH, THF; c, 1-bromo-2-pentyne, CuI, Cs2CO3, NaI; d, Lindlar catalyst, quinoline, H2; e) NaOH, aq. MeOH. "TMS", trimethylsilyl.

Conditioning procedure for Waters AMY-1 chiral column
The following procedure was applied to newly purchased AMY-1 chiral columns in order to tune the selectivity to harmonize the method performance and to obtain reproducible separations between new columns ( Figure S3A) and the column previously in use ( Figure  S3D). The new column was first equilibrated as suggested by the manufacturer: 30 min with 100% CO2 at a flow rate of 1.5 mL/min, followed by 30 min in CO2:MeOH 1:1 at a flow rate of 1.5 mL/min, and subsequently equilibrated in 100% MeOH for 15 min at 0.5 mL/min and 15 min at 1 mL/min (Step 0). The column was then equilibrated at the method initial conditions for 30 minutes and a test sequence consisting of 3 solvent and 3 SST injections was performed and evaluated to benchmark the column performance ( Figure S3A). The following step consisted in flowing through the column 5000 mL of CO2:MeOH, CH3COONH4 5 mM 1:1 at a flow rate of 1.0 mL/min, setting the ABPR at 1500 psi (Step 1). This step required a total of 72 hours and was followed by evaluation, performed in the same way as described above ( Figure S3B). The final step provided the fine-tuning resolution of more polar species and consisted of flowing 70 mL of ACN:IPA 1:1, HCOOH 0.2% v/v over 70 minutes at a flow rate of 1.0 mL/min (no CO2, ABPR turned off; Step 2). The results were evaluated by the same test sequence ( Figure S3C) and can be compared with a reference chromatogram acquired with the previous column (Reference panel), reported in Figure S3D.
To evaluate the reproducibility and stability of the procedure, three AMY-1 columns from different production lots were conditioned as described above. The procedure yielded the same results in terms of selectivity alteration on all the tested columns (data not shown). Figure S3. Effects of conditioning on the AMY-1 stationary phase Figure S3. Effects of conditioning on the AMY-1 stationary phase. Evolution of the selectivity of new AMY-1 columns during the different steps of the initial conditioning procedure. The column could be used after Step 2 (purple trace). Panel D shows a reference chromatogram obtained on a different AMY-1 column, subjected to the same conditioning procedure. The procedure was tested on 3 AMY-1 columns from different production lots providing the same results. Refer to Table S1 for peak numbering and identity. Peaks labeled with * in the TriHOME panels refer to different positional isomers that share a transition with the illustrated TriHOMEs and appear in the same MRM channel. The optimal result for each class of compounds is highlighted in bold, and the second best is underlined All tests were performed without changing other parameters: ABPR 2000 psi, column T=35.0℃, gradient as reported in Table S2 The best result for each solvent composition is reported in the The effect of different acid additives in the mobile phase and of the variation of other parameters such as the ABPR pressure, column temperature, as well as flow rate and composition of the make-up solvent were evaluated but not reported in the present work.

S27
b Not quantified, but only screened qualitatively. Standards of 9,10,11-TriHOME and 11,12,13 were synthesized, it was not possible to assign stereochemistry. In addition, full chromatographic resolution was not achieved for all isomers, which are therefore reported as peak numbers.
* Order of elution and identity not confirmed with an enantiopure standard but inferred by comparison with similar compounds under the same conditions. N.A.=not applicable.   Table S5).     Table S8).

Results and Discussion: Breakthrough determination of polar species
The scarce solubility of water in supercritical CO2 limits its use in SFC mobile phases to additive level (up to 5% of water in alcohol can be used) 14 . Additionally, according to manufacturer instructions, AMY-1 columns cannot tolerate water even at trace levels since it would cause permanent alterations to their selectivity. The impossibility to use water in both mobile phase and sample solvent imposes the necessity for a quantitative removal of polar species during SPE, to avoid precipitation in the initial steps of the SFC chromatographic gradient under low MeOH conditions. SPE performance were evaluated regarding phosphates and proteins, major components of surrogate matrix and extraction solution, which precipitation can be the source of pressure instability and column clogging. The presence of residual proteins and phosphates in SPE eluates was evaluated for the original SPE method 1 and by including 1-4 additional 3 mL water wash steps ( Figure S8). While a strong phosphate signal could be detected in the eluate from the original SPE method, a significant decrease could be obtained already with the first additional wash (98.4% phosphates signal decrease), with further decreases until the third wash (Table S9). The final SPE method was designed to include 3 X 3 mL water wash steps after the initial 3 mL MeOH : H2O 9:1 wash.
Quantification of proteins via BCA assay showed that 0.4-1.2 µg of protein were present in SPE eluates, irrespective of the number of water wash steps (Table S9). The use of the SPE method including 9 mL of additional water wash resulted in the injection of about 15 ng of protein for each sample (SFC method; 35 ng for the LC method, given the higher injection volume).  Figure S8. HILIC evaluation of phosphates in eluates with increasing wash volumes Figure S8. HILIC evaluation of phosphates in eluates with increasing wash volumes. Overlaid chromatogram obtained with the HILIC method described in Table S8 showing the variation of the phosphates peak in different SPE washing conditions. The brown track refers to the phosphates breakthrough by using the original SPE method. Additional wash steps are magnified in the inset.

Experimental: parameters evaluated during method validation
Sensitivity and linearity. Solvent-matched calibration curves were prepared in three replicates to determine the linear range for the target octadecanoids and the LLOQ. The curves were built by plotting the ratio of the analyte area on the related IS (response) against the theoretical concentration, with a weighting factor of 1/X and a linear fit. The LLOQ for each analyte was determined as the lowest concentration presenting a peak with a signal-to-noise ratio (S/N) 5, and relative error (RE) on the back-calculated concentration <30%. For higher levels, the accepted threshold for the RE was lowered to 15%. Matrix-matched calibration curves were prepared by spiking calibration levels and IS solution in surrogate matrix extracts. These curves were used to determine linearity in matrix, to calculate matrix effects, and in the quantification of accuracy and precision in matrix.

Accuracy and precision.
Inter-and intra-day accuracy and precision for the whole procedure were determined by spiking quality control (QC) in surrogate matrix at three concentration levels (Low, 3 times LLOQ; Medium, middle point of the linear range; High, 75% of calibration level 9). Six replicates of each QC were extracted each day over a course of three consecutive days to evaluate inter-day precision and accuracy (n=18), while the extracts of the third day were analyzed in triplicate (n=18) to evaluate intra-day precision and accuracy. Accuracy was defined as the closeness between the theoretical concentration of each QC and the average concentration obtained by the repeated measurements, expressed as percentage difference. Precision was defined as the consistency of repeated measurements and was expressed as percentage relative standard deviation (%RSD).
Instrumental accuracy and precision were evaluated in solvent, by preparing at the same concentration levels in pure solvent and directly quantifying them on solvent-matched calibration curves over the course of three consecutive days, in the same number of replicates as described for the matrix validation.
Recovery. The recovery was calculated by spiking QC in surrogate matrix before extraction and comparing the obtained concentration with the concentration measured in the same QC, spiked after extraction (in both cases the IS solution was spiked before SPE). The recovery was calculated at the same three concentration levels used for the determination of accuracy and precision (n=6 per level) and was reported as percentage of the concentration measured in the pre-spiked QC over the concentration measured in the post-spiked QC. IS recovery was calculated both in plasma and in surrogate matrix by comparing the peak area obtained by spiking the IS solution in the biological matrix before and after SPE extraction (n=6).
Matrix effect. Matrix effect was calculated by comparing the slope of the matrix-matched calibration curves (n=3) with the slope of solvent-matched calibration curves (n=3) and was expressed as % RE between the two slopes. IS matrix effect was calculated both in plasma and in surrogate matrix by comparing the peak area of IS spiked in matrix extracts (n=6) with the peak area in solvent (n=6) and expressed as %RE between the average of the two measurements.
Analyte stability. Autosampler stability of the analytes was determined over the course of 96 hours at 8℃, to mimic typical conditions experienced during sample analysis. Stability was evaluated by the analysis of four SST solution aliquots, spiked with the IS, stored in the autosampler, and quantified at four time points (0, 24, 48, and 96 hours) on freshly prepared calibration curves. The feasibility of leaving the samples for the investigated time range was evaluated by assessing the stability of the measured concentrations. Long-term storage stability at -80℃ was not evaluated, because this has been previously reported 15 .

Results and discussion: evaluation of instrumental accuracy for deviating species
While most of the investigated octadecanoids showed instrumental accuracy within the acceptable threshold for method validation, 9-and 13-HODE, 11(E)-10-KOME, 11(E)-10-HOME, and 9-OH-trans-12,13-EpOME evidenced extremely high deviation from the theoretical concentration value at both low and medium concentration (300-800%, Table  S11). While the last three species are, to the best of our knowledge, included in a broad profiling method for the first time, HODEs are amongst the most studied octadecanoids, and are routinely included in oxylipin analytical platforms 1, [16][17][18] . HODEs data acquired with our general eicosanoid and oxylipin profiling method 1 over the course of one year were used to evaluate the background levels and accuracy of quantification at low concentrations. Background levels ranging 0.28-1.12 ng/mL appeared randomly 50% of the time (data not shown). These levels are comparable with the alterations observed during the method validation and exert a strong effect on the accuracy at low and medium concentrations.
Varying background levels may depend on a multitude of factors, including the ubiquitous presence of the analytes in solvents, glassware, and other consumables. It is therefore necessary to perform a careful evaluation of solvent and extraction blanks on each analysis day to avoid reporting of artifacts. Method validation for HODEs has been reported in the literature; however, the investigated linear ranges are generally shifted towards higher concentrations [16][17][18] . HODEs are relatively high-concentration metabolites, and their reported levels in plasma range 1.5-25 ng/mL 19 . At these levels, comparable to the high-concentration QC used in the LC method validation, HODEs were quantified with an accuracy of 110-113%. The average accuracy levels at low and medium concentrations were affected by the poor performance of the aforementioned compounds. It was therefore concluded that the current LC method was fit for purpose for quantification of the HODEs at the levels expected to be observed in most biological matrices; however, caution should be exerted if levels of these compounds are reported at lower concentrations.
The SFC platform was not significantly affected by background levels of any of the octadecanoids, which is most likely due to the higher detection limit of the technique and the different linear range. 43% 45%