Positional determination of the carbon–carbon double bonds in unsaturated fatty acids mediated by solvent plasmatization using LC–MS

Fatty acids (FAs) are the central components of life: they constitute biological membranes in the form of lipid, act as signaling molecules, and are used as energy sources. FAs are classified according to their chain lengths and the number and position of carbon–carbon double bond, and their physiological character is largely defined by these structural properties. Determination of the precise structural properties is crucial for characterizing FAs, but pinpointing the exact position of carbon–carbon double bond in FA molecules is challenging. Herein, a new analytical method is reported for determining the double bond position of mono- and poly-unsaturated FAs using liquid chromatography-mass spectrometry (LC–MS) coupled with solvent plasmatization. With the aid of plasma on ESI capillary, epoxidation or peroxidation of carbon–carbon double bond in FAs is facilitated. Subsequently, molecular fragmentation occurs at or beside the epoxidized or peroxidized double bond via collision-induced dissociation (CID), and the position of the double bond is elucidated. In this method, FAs are separated by LC, modified by plasma, fragmented via CID, and detected using a time-of-flight mass spectrometer in a seamless manner such that the FA composition in a mixture can be determined. Our method enables thorough characterization of FA species by distinguishing multiple isomers, and therefore can uncover the true diversity of FAs for their application in food, health, and medical sciences.

After phase separation at 200 × g for 5 min, the upper organic phase containing FAs was collected and placed in a glass sample vial with a polytetrafluoroethylene (PTFE)-lined cap (GL Sciences, Tokyo, Japan, #1030-51023 and #1030-45260). The organic phase was evaporated under a stream of nitrogen gas and the recovered FAs were reconstituted in 100 µL of acetone and subjected to LC-MS analysis.
Ethics. This study was approved by the Ethical Committee of the Graduate School of Medicine, Gifu University (permission number: 29-286). For the use of a clinical sample obtained from an infant patient, written informed consent was obtained from the patient's parents. All experiments in this manuscript were conducted according to the guidelines and regulations provided by Gifu University.  (Fig. S1B) was also generated on solvent plasmatizaton as a side product. This side reaction always occurred irrespective of given electric voltages so that was unable to eliminated. However, we found that the peroxidized form can also be used for determining the double bond positions as shown in the following section. When methanol and isopropanol were used for the solvent, same reaction happened to form epoxide and peroxide, though use of isopropanol unstabilized plasma formation at the tip of ESI capillary. In this study, we chose to use acetonitrile because it has better elution capability on LC for FA analysis.

Results and discussion
The same protocol was attempted using cis-vaccenic acid (C18:1 ω-7, cis-11) and petroselinic acid (C18:1 ω-12, cis-6), which have the same chain length and number of double bonds, but different double bond positions compared to those in oleic acid. Following epoxidation and CID, similar results were obtained while producing the diagnostic fragments for the determination of the double bond position (Fig. 2B-D). Thus, solvent plasmatization as corona discharge was found to be useful for the epoxidation of FAs and this method is referred to as plasma-ESI-MS.
We next applied plasma-ESI-MS with LC analysis. The entire scheme of the analysis, LC-plasma-ESI-MS, is depicted in Fig. 1B; and the diagnostic fragment ions for determining the position of carbon-carbon double bond in MUFAs and PUFAs found in the above and following experiments are shown in Fig. 1C,D.

Determination of the double bond position of MUFAs by LC-MS.
We analyzed octadecenoic acids (C18:1 FAs) with LC-plasma-ESI-MS. In our previous study 27 , the proximity of the double bond to the omega carbon was found to be the major determinant of retention time of unsaturated FAs in reverse-phase LC. Cisvaccenic acid (ω-7) elutes from the LC column first, oleic acid (ω-9) elutes second, and petroselinic acid (ω-12) Scientific RepoRtS | (2020) 10:12988 | https://doi.org/10.1038/s41598-020-69833-y www.nature.com/scientificreports/ elutes last 27 . Here, this was also confirmed by determining the exact position of the double bond. C18:1 FAs were injected independently and simultaneously as a mixed sample. In the independent injection, each C18:1 species yielded a sharp peak at a distinct retention time (Fig. S2A). The epoxidized form of each C18:1 species was detected, and by CID, the diagnostic fragment ions were produced (Figs. S2B,C; fragment a and fragment b in Fig. 1C). In the mixture injection, peaks were imperfectly separated ( Fig. S3A-C). The epoxidized form was detected for all peaks. Upon CID, the fragment ions of the epoxidized cis-vaccenic acid were observed at the earliest retention time at the first half of the first peak ( Fig. S3A), the fragments of epoxidized oleic acid were observed at the second half of the first peak (Fig. S3B), and the fragments of epoxidized petroselinic acid were observed in the second peak (Fig. S3C). Epoxidized cis-vaccenic acid showed fragments at m/z 199.13 and 183.14 with a peak position at 17.82 min (Fig. S3A). The fragments for the epoxidized oleic acid, m/z 171.10 and 155.11, showed peaks at 18.24 and 18.14 min, respectively (Fig. S3B). The fragments for the epoxidized petroselinic acid, m/z 129.06 and 113.06, showed a peak at 18.66 min (Fig. S3C). These time differences confirmed the elution order of C18:1 FA species: cis-vaccenic acid, oleic acid, and then petroselinic acid. The above data confirmed our previous observations that showed the order of elution timing of unsaturated FAs correlated with the proximity of the double bond to the omega carbon. As shown in the previous section, FAs can be peroxidized by the solvent plasmatization. We found that the fragmentation spectra of the peroxidized FAs are also informative for determining the position of the double bonds. On solvent plasmatiozation, peroxidized oleic acid (m/z 313.24) was found simultaneously with the epoxidized oleic acid as a side reaction, though the intensity is lower than the epoxidized form (Fig. S1B). Upon CID, peroxidized oleic acid produced the same fragment ions as the epoxidized form (Fig. S4A,B), such as the alpha fragments at m/z 171.10 and 155.11. Additional fragment ions were also observed at m/z 143.11 and 127.11 corresponding to the alpha fragments with and without terminal oxygen (fragment c and fragment d in Fig. 1C). Minor fragment ions such as peroxidized alpha fragment at m/z 201.11, and the fragment at m/z 141. 13 corresponding to the omega carbon-side fragment (omega fragment) were also detected ( Fig. S4A,B). Similar results were obtained for the analysis of peroxidized cis-vaccenic acid and petroselinic acid ( Fig. S4C-F). These additional fragments yielded positional information of the double bond, and similar fragments were informative for determining the position of double bonds in PUFAs, which will be discussed in later sections.  Figure S5B) were analyzed and compared with corresponding cis counterparts (palmitoleic acid and cis-vaccenic acid, respectively). The trans FAs yielded the same fragmentation patterns with their cis counterparts, including diagnostic fragment ions, when their epoxide or peroxide forms were fragmented ( Fig. S5C and data not shown). There were no cis FA or trans FA specific fragments detected, therefore the cis/trans conformation could not be determined by their fragmentation patterns. However, the trans FAs could be distinguished from cis FAs on mass chromatogram; the trans FAs had longer retention times compared to the cis-FA counterparts (Fig. S5A,B). Therefore, this property can be used to distinguish trans FAs in analytical samples when comparable cis FAs are available.

Determination of the position of double bonds in PUFAs.
We next examined if the plasma-mediated modification of PUFAs could yield any informative fragment spectra for determining the position of carboncarbon double bonds. Two types of octadecatrienoic acids (C18:3), α-linolenic acid (C18:3 ω-3, cis-9, 12, 15) and γ-linolenic acid (C18:3 ω-6, cis- 6,9,12) were first examined by LC-plasma-ESI-MS. The form of epoxide was detected with predicted m/z value (m/z 293.21). Additionally, we found two types of peroxide with mass difference of two units (m/z 309.21 and 311.22) (Figs. S6A and S7A). The first one with m/z 309.21 was considered as regular peroxide as found in MUFA above, and the second one with m/z of 311.22 was predicted as cyclic peroxide (Figs. S6B and S7B), which the similar form was reported previously 28 .
We compared the fragmentation patterns of each epoxidized/peroxidized form. When epoxidized form was subjected to CID, highly prominent fragment spectra at m/z 71.05 in α-linolenic acid and m/z 113.10 in γ-linolenic acid (Fig. 3A,B, left panels) were yielded. These spectra corresponded to the omega fragments epoxidized at the proximal double bond to the omega carbon. These types of fragmentation spectra always appeared in extremely high intensities in PUFAs and can be used for determining the double bond that is closest to the omega carbon for their categorization into ω-n families. The two types of peroxide were examined separately. When regular peroxide of α-linolenic acid and γ-linolenic acid were fragmented, ions at m/z 87.05 and 129.09, respectively, were detected in high intensities, and were peroxidized omega fragments at the proximal double  www.nature.com/scientificreports/ bond to the omega carbon (Fig. 3A,B, right panels). The alpha fragments at m/z 223.17 and 181.12, which are the counterpart fragments of the above omega fragments, were also detected in high intensities (Fig. 3A,B, right panels). From the omega and alpha fragments, the positions of the first and second double bonds can be determined. Similar alpha fragments were observed at the second and third double bonds (m/z 183.14 and 141.09 for α-and γ-linolenic acid, respectively; Fig. 3A,B). The fragmentation yielding this type of alpha fragments only effectively occurs when the neighboring double bonds are separated by a single methylene group (hereafter referred to as "methylene-interrupted double bonds"; Fig. 1D) and not by multiple methylene groups (hereafter referred to as "isolated double bond"; Fig. 1D). In addition, there were multiple minor fragment ions that could support the positional information of double bonds as shown in Figs. S6D and S7D. When the cyclic peroxides of α-linolenic acid and γ-linolenic acid were fragmented, they yielded the same diagnostic fragment ions with regular peroxide but higher intensities (Figs. S6E, S7E). The same minor fragment ions with regular peroxide were also found with a few exceptions. The details of the fragmentation patterns of each peroxide were shown in Figs. S6 and S7. From these diagnostic fragments, the positions of all the three double bonds can be determined. Because both types of PUFA peroxide generate the diagnostic fragment ions, they were simultaneously fragmented to yield higher signals of diagnostic fragment ions for quantitative purposes, while independently fragmented for qualitative purposes in this study. Two types of eicosatrienoic acid (C20:3), sciadonic acid (C20:3 ω-6 cis-5, 11, 14) and dihomo-γ-linolenic acid (C20:3 ω-6 cis- 8,11,14), were examined. Dihomo-γ-linolenic acid contains three methylene-interrupted double bonds, whereas sciadonic acid contains two methylene-interrupted double bonds and one isolated double bond (Fig. S8C). From the fragmentation pattern of the peroxidized form, dihomo-γ-linolenic acid showed a similar fragmentation pattern as that of the C18:3 fatty acids described above (Fig. S8A,C; the fragment spectra of cyclic peroxide are shown). In contrast, sciadonic acid yielded the same fragmentation pattern only from its methyleneinterrupted double bonds (cis-11 and cis-14 double bonds) (Fig. S8B,C). The isolated double bond at cis-5 position behaved in a similar manner as a MUFA double bond, producing common alpha fragments (Figs. S4 and S8B,C), enabling the determination of the position of the isolated double bond. The same fragmentation pattern was observed with the isolated double bond of cis-4 DTA, which contains three methylene-interrupted double bonds and an isolated double bond (see below). Therefore, the double bond positions in PUFAs can be determined by the fragment ions specific to the methylene-interrupted and isolated double bonds.
We also evaluated the intraday (n = 5) and interday (n = 5) reproducibilities of diagnostic fragment ions using standard solutions of cis-vaccenic acid and γ-linolenic acid at two different concentrations (10 μM and 100 μM). The intraday precision, expressed as relative standard deviation (RSD), was less than 10% for most of the fragments ( Table 2). The interday precision of RSD without internal standard (IS; 13 C18:1 cis-9) was high; however, it improved upon normalization using IS (Table 2).
Relative quantification was evaluated with mixtures of FA isomers of MUFA (oleic acid and cis-vaccenic acid), and PUFA (α-linolenic acid γ-linolenic acid), with several different ratios of the two isomers ( Table 3). The relative quantity of the two isomers showed good linearity in the range from 0.1 to 2.0. The above data indicate that the relative abundance of FA isomers can be quantified using this method.
Ionization and fragmentation mechanisms. The analytical mechanisms of this method were also studied. To elucidate the origin of oxygen in the epoxide and peroxide groups of the FAs formed by the plasma-ESI, we used 18  O and acetonitrile, and was examined by direct infusion assay (additives like ammonium acetate and ammonia solution were not supplemented to eliminate any contaminant of regular water). Epoxide formed by plasma-ESI had the same m/z values as those obtained using regular water (Fig. 5A), indicating that the oxygen in epoxide group did not came from the solvent, but probably from the surrounding air as in the case of a previous report 15 . In contrast, cyclic peroxide of cis-4 DTA had higher m/z value with two units (2 Da). The fragment ions containing peroxide group also had 2 Da increase (Fig. 5B). These results suggest that one of the two oxygen atoms in the cyclic peroxide group comes from the solvent and the other oxygen atom comes from the surrounding air.
We next used deuterium oxide (D 2 O) as a solvent to reveal the origin of the added hydrogen in the peroxide group and in some fragment ions (i.e. diagnostic fragment ions c, d, e, and f; see Fig. 1C,D) by the same direct infusion assay. We found peroxidized cis-4 DTA and cis-vaccenic acid had increased m/z values, indicating that Scientific RepoRtS | (2020) 10:12988 | https://doi.org/10.1038/s41598-020-69833-y www.nature.com/scientificreports/ the solvent was the origin of the hydrogen in the peroxide group ( Fig. 5 and Fig. S10, respectively). The diagnostic alpha fragment ions (diagnostic fragment ions c, d, and f) also had increased m/z values by 1 Da, again indicating that the hydrogen added to the fragments was derived from the solvent. On the other hand, the omega fragment (diagnostic fragment e) did not have increased m/z value. Taken together, these data indicate that the hydrogens are included in the diagnostic fragment ions in the plasma-ESI method.
Analysis of unsaturated fatty acids from human fibroblasts. The developed method was subsequently applied to biological samples. FAs were extracted from human fibroblasts (NB1RGB) via acid hydrolysis and separated using a reverse-phase gradient LC (see "Materials and methods"). The quadrupole was set to introduce epoxidized MUFA species (epoxi-MUFAs), including C16:1, C18:1, C20:1, C22:1, C24:1, and C26:1, www.nature.com/scientificreports/ in separate time windows. The introduced epoxi-MUFAs were subsequently fragmented by CID. The entire ions ranging from m/z 100 to 1,000 were also scanned and recorded simultaneously in another channel to detect the non-epoxidized FAs. As reported previously 27 , multiple peaks can be observed for some of the MUFAs in their mass chromatograms (Fig. 6A; Figs. S12A, S13A), which may represent FA isomers with different double bond positions. This observation was confirmed by characterizing the isomers in each peak. In the mass chromatogram of C16:1, two peaks were observed with incomplete base separation. Upon CID fragmentation, three C16:1 Table 1. Analytical performances of MUFA and PUFA species. RSD residual standard deviation. a Diagnostic fragment ions with least intensities are underlined. b LOD and LOQ were calculated as 3.3 × RSD/slope and 10 × RSD/slope, respectively, based on the calibration function.   Table 2. Intraday and Interday precisions associated with this method. a Fragment ions of the internal standard, 13 C-labeled oleic acid, was used for normalization.    (Fig. 6) were observed. In the first peak, ω-7 isomer yielding fragment ions at m/z 155.11 and 171.10 was detected. At the beginning area of the second peak, ω-9 isomer was observed yielding fragments at m/z 143.07 and 127.08, while ω-10 isomer yielding fragments at m/z 113.06 and 129.06 was found throughout most of the second peak. The retention order of the isomers followed previous observations where the FAs with more proximal double bonds to the omega carbon eluted faster than those with more distal double bonds. From the C18:1 assay, four isomers were detected ( Figure S11), with the most abundant being oleic acid (C18:1 ω-9), which yielded a fragment ion pair at m/z 171.10 and 155.11. The other C18:1 isomers were ω-7 (producing a fragment ion pair at m/z 199.13 and 183.14), ω-10 (m/z 157.09 and 141.09), and ω-12 (m/z 129.05 and 113.06). ω-7, ω-9, and ω-10 isomers were observed in the first major peak, while ω-12 isomer was detected in the minor second peak (Fig. S11). Comparing the retention times of the isomers, the elution order from the earliest to the latest was ω-7, ω-9, ω-10, and ω-12 isomers (Fig. S11). Even though four isomers of C18:1 were detected, the high abundance of prospective oleic acid (ω-9 isomer) embedded the other isomers within a single large peak, highlighting the importance of isomer segregation. Similarly, multiple isomers were observed in the assays of C20:1 to C26:1. The isomers with their diagnostic fragment ions are shown in Figs. S12-S15. Though the isomeric variety of longer FAs are not well known, the isomers of C16:1 and C18:1 found in our study were consistent with previous reports 15,16,19,20,26,29 . The double bond position of MUFAs from biological samples, thus, were confirmed by this method. www.nature.com/scientificreports/ Finally, PUFA isomers from a clinical sample were analyzed. FAs were extracted from fibroblasts of a human patient with Zellweger syndrome (ZS) and from control fibroblasts (F-12 and NB1RGB, respectively) 12 . In ZS, the biosynthesis of peroxisomes is impaired (Fig. 7A) and the FA composition fluctuates due to metabolic failures in peroxisomes. In our previous study, multiple peaks were obtained for C20:3 FA upon LC separation, where each peak was predicted to be a distinct C20:3 isomer, although the exact double bond position remained undetermined 27 . By analyzing the FAs from the fibroblast samples, multiple peaks for C20:3 were confirmed again www.nature.com/scientificreports/ (Fig. 7B). Peroxidation of C20:3 followed by CID fragmentation indicated that the first peak included a C20:3 isomer with the double bonds at cis-8, 11, and 14 positions in ZS and control fibroblasts (Fig. 7B-D). Moreover, an additional isomer with double bonds at cis-7, 10, and 13 positions was found in both samples but reduced in ZS fibroblasts (Fig. 7C,D, Fig. S16). The second peak corresponded to only a single isomer with double bonds at cis-5, 8, and 11 positions found in both samples but again reduced in ZS fibroblasts (Fig. 7C,D, Fig. S16). These results confirmed our previous prediction about multiple peaks of C20:3 in control and ZS fibroblasts 27 . These data indicate that the PUFA isomers in biological samples can be distinguished by the double bond position. This is the first example that determined the double bond positions of PUFA isomers showing differential abundances in ZS samples, highlighting the usefulness of the method reported herein for medical applications.

conclusions
Determination of the positions and number of double bonds, as well as the chain length, is crucial for understanding the precise physiological functions of each FA species. Many methods have been established for the positional determination of double bonds by modifying the target FAs at the double bond positions or the terminal carboxyl group to yield diagnostic fragment ions [8][9][10][11][13][14][15][16][17][18][19][20][21][22][23][24][25][26] . CID-mediated fragmentation on ESI-MS or EI-mediated fragmentation on GC-MS of unsaturated FAs cannot always produce informative product ions to determine the double bond positions, especially in case of PUFAs with many double bonds. Moreover, the derivatization of FAs would affect their chromatographic retention time, compromising the interpretation of chromatographic results. Herein, a method of post-column epoxidation and peroxidation of unsaturated FAs using corona discharge plasmatization of chromatographic solvent is reported. This method is advantageous for samples containing mixed FAs, such as biological samples. The post-column epoxidation/peroxidation does not compromise the LC, so the chain length and number of double bonds in each FA in the mixture can be easily determined from the LC results 27 . Similar approach was reported recently, where FA epoxidation was facilitated by low voltage on nano-ESI sprayer 30 , although there are significant technical differences such as the existence of hydrochloric acid and the polarity of ESI capillary. The drawback of our method, presently, is its low sensitivity, mainly attributable to the poor epoxidation/ peroxidation efficiency. Moreover, we could not separately induce epoxidation or peroxidation on requirement therefore cannot eliminate unwanted derivatives from the reaction. The reported LOQ of FAs in regular LC-MS or GC-MS analysis is around nanomolar order so that our method is currently only suitable for FAs with high abundance. However, our method can be used as a supporting method with regular LC-MS/GC-MS analyses to show the isomeric heterogeneity of detected FA species. Moreover, our method only requires conventional reagents and equipment for LC-MS therefore easy to try with.
For the improvement, several approaches are conceivable. Because we used a pre-equipped setup of conventional ESI, the development of a specialized setup may improve the sensitivity for detecting FA species with low abundance. Alternatively, addition of an epoxidizing/peroxidizing agent to the solvent would improve the sensitivity of this method. In addition, post-column modification of target molecules with solvent plasmatization can be applied to other types of molecules for structural analysis. In conclusion, the method reported herein enabled a thorough characterization of FA species, i.e. chain length, number of double bonds, and position of double bonds, and can aid for many applications.