LC-Orbitrap-HRMS method for analysis of traces of triacylglycerols featuring furan fatty acids

Furan fatty acids (FuFAs) are valuable antioxidants that are highly relevant for the protection of polyunsaturated fatty acids (PUFAs) in biological systems and food. Despite their low contributions to the total fatty acids, their widespread occurrence has been documented in food and biological samples. Like other fatty acids, FuFAs are also stored esterified, e.g., in triacylglycerols. However, FuFA-containing triacylglycerols had not been detected in lipidomics analyses. Here, we present a screening method that allows for the identification of traces of FuFA-containing triacylglycerols (TAGs) utilizing LC-Orbitrap-HRMS. Initially developed with the help of purposefully synthesized FuFA-containing TAGs, the screening method was successfully applied to the analysis of two fish oil samples and one mushroom extract sample. Several FuFA-containing TAGs could be identified by direct analysis using the method and database developed in this study. Graphical Abstract Supplementary Information The online version contains supplementary material available at 10.1007/s00216-022-04480-y.


Introduction
Furan fatty acids (FuFAs) are powerful radical scavengers that can effectively prevent the oxidation of polyunsaturated fatty acids (PUFAs) and other susceptible molecules [1][2][3]. Arguably, FuFAs thus belong to the most valuable fatty acids in living organisms and the human diet. For instance, some beneficial health effects of fish consumption currently linked to PUFAs were proposed to be caused by FuFAs [1]. Structurally, the most relevant FuFAs feature a furan moiety with an odd-numbered carboxyalkyl chain of mainly nine or eleven carbon atoms attached in α-position and an alkyl chain of three or five carbon atoms in α′-position (Fig. 1). In addition, either one or two methyl groups are located in β-and β′-position [1,4,5]. Instead of the complex chemical names (e.g., 9-(3-methyl-5-pentylfuran-2-yl)-nonanoic acid, Fig. 1), FuFAs can be presented by number-letter-number short forms which give direct structural information (for this example, 9M5) [4]. Namely, the first and the second number denote the number of carbon atoms of the carboxyalkyl and the alkyl chain, respectively, while the central letter indicates the methylation degree on the furan moiety ("M" = monomethyl-substituted in β-position, "D" = dimethyl-substituted in β,β′-positions) [4] (Table 1). For better readability in TAGs, numbers in these short forms (e.g., 9D5) will be presented in superscript and subscript style (e.g., 9 D 5 ) [6].
Similar to other antioxidants such as tocopherols, FuFA levels in food are considerably low (typically < 1% of the lipids) [7,8]. Like other fatty acids, FuFAs are almost exclusively found esterified in different lipid classes. However, the structural similarity with conventional fatty acids makes the analysis of FuFAs more challenging [9] than, e.g., of tocopherols which can be easily separated from the bulk of fatty acids. Typical analysis protocols for FuFA determinations include lipid extraction followed by saponification and/or transmethylation followed by GC/MS or LC/MS determination [5,7,[9][10][11]. Almost exclusively, an enrichment step by silver ion chromatography has to be carried out prior to the quantification [12]. On the one hand, all information on the presence of FuFAs in particular lipid classes is lost after these sample processing steps. On the other hand, existing methods of FuFA enrichment cannot be applied to intact lipids such as triacylglycerols (TAGs) which are characterized by a huge plethora of structural variants [13,14].
In lipidomics, TAGs are usually analyzed by high-performance liquid chromatography in combination with electrospray tandem mass spectrometry (LC-ESI-MS 2 ) [15][16][17]. Regardless of immense progress in LC/MS instrumentation, contemporary analysis is still hampered by the wide concentration range of individual TAGs. As a consequence, very low abundant fatty acids such as FuFAs are usually overlooked although they could be highly bioactive. Namely, the direct measurement of FuFA-containing TAGs has not been reported, yet. However, the inclusion of FuFA-containing TAGs in lipidomic studies seems to be an important matter given the high relevance of FuFAs for the prevention of lipid oxidation.
Here, we provide a method for the direct analysis of FuFA-containing TAGs in food samples by LC-Orbitrap-HRMS. Initial measurements of purposefully synthesized FuFA-containing TAGs [6] enabled us to select FuFA-specific m/z values. While the direct identification of FuFAcontaining TAGs in first-dimension mass spectra (MS 1 ) via the exact mass was found to be equivocal, the selection of diagnostic m/z values in second-dimension mass spectra (MS 2 ) enabled their detection in two relevant sample matrices and created a database for the direct determination of FuFA-containing TAGs in food and biological samples.

Lipid extraction
About 400 mg freeze-dried king oyster mushroom powder was extracted twice with 4 mL cyclohexane/ethyl acetate (46:54, w/w) via ultrasonication (5 min) followed by centrifugation (8 min). The whole procedure was repeated in the same way with 4 mL iso-propanol/n-hexane (1:4, v/v). The combined supernatant of the four extractions was evaporated to dryness in a pre-weighed tube by means of a gentle stream of nitrogen. The weighed residue (~ 8.8 mg) was dissolved in 1 mL ethanol. After membrane filtration, the sample was measured by LC/MS. For the fish oil II, a gilthead liver was freeze-dried and the oil was extracted by accelerated solvent extraction (ASE) with a Dionex ASE 350 (Thermo Scientific, Waltham, MA, USA) instrument using the instrumental parameters of Weichbrodt et al. [19]. The used solvent system was an azeotropic mixture of cyclohexane/ethyl acetate (46/54, w/w). The extract was reduced and made up to a total volume of 4 mL. An aliquot (~ 10 mg) was taken from this, the solvent removed in a heating block maintained at 40 °C with a gentle stream of nitrogen and dissolved in 1 mL ethanol.

High-performance liquid chromatography with mass spectrometry (LC-Orbitrap-HRMS)
Samples were measured on a HPLC 1290 (Agilent, Waldbronn, Germany) instrument interfaced with a Q Exactive Plus highresolution mass spectrometer (Thermo Scientific, Waltham, MA, US). Separations were performed with a 2.1-mm-long, 150-mm-i.d., and 1.7-µm-particle size ACQUITY UPLC CSH C18 column (Waters, Milford, MA, USA). The column temperature was set to 60 °C and 5 µL of the sample was injected. Eluent A (acetonitrile/water (6:4, v/v) with 5 mM ammonium formiate) and eluent B (iso-propanol/acetonitrile (9:1, v/v) with 5 mM ammonium formiate) additionally featured 0.2% formic acid. The flow rate of 0.325 mL/min was selected because it performed best during testing of 0.320-0.340 mL/min in steps of 0.005 mL/min. A gradient program from 55 to 40% A within 7 min, then 40-18% A within 21 min, was used. Finally, A decreased to 1% during 1 min, which was held for 5.90 min. After 34.90 min, the run was finished. The pump needed a further 0.10 min to get the start conditions, and these were held for 5 min. The mass spectra were acquired using electrospray ionization (ESI) in positive mode. The scan range of MS 1 covered m/z 100-1200 with a resolution (FMHW) of 70,000. MS 2 covered m/z 50-2000.

Identification of triacylglycerols containing FuFAs
For identification of TAGs, characteristic fragment ions are available [20][21][22][23]. The molecular ion [M] + and the high abundant [M-RCOO] + are of great importance. [M-RCOO] + results from the release of an entire acyl group from the molecule [24,25]. TAGs with three identical fatty acids  [17]. But often, it is not clear in which order the fatty acids are arranged on the glycerol backbone. It is assumed that isomers co-elute due to the same mass. For the sake of simplicity, a scheme was drawn up in which fatty acids were listed in a specific order. Namely, FuFAs were always assigned to the sn-3 position, while conventional fatty acids were arranged in ascending order of molecular weight. Note that this must not necessarily agree with the real order in the corresponding TAG.  Table 2) in a given TAG. Further characteristic fragment ions show a positive mass difference compared to the acyl ion (namely, + 74 u and + 128 u). These and other less prominent peaks are indicative of TAGs [24].
Beyond that, the characteristic and abundant GC/MS fragment ions of FuFA-ME (e.g., m/z 165 (base peak)) are also present in the LC/MS spectra of FuFA-containing TAGs (FuFA-TAGs) [26].

Results and discussion
Structural and mass spectrometric features of FuFA-containing TAGs FuFA-containing TAGs shared the HPLC elution range with TAGs featuring (only) conventional fatty acids. FuFAs with nine or eleven carbons in the α-connected carboxyalkyl chain, one or two methyl groups in β,β′-positions of the furan moiety, and three or five carbons in the alkyl chain in α′-position (Table 1) give rise to 2 3 structural variants including two pairs of isomers, i.e., 11D3 and 9D5 as well as 11M3 and 9M5. These most prominent FuFAs feature 17 to 22 carbon atoms (  4 ] + ions, the latter due to the presence of ammonium acetate in the mobile phase (Fig. 2a) [27]. The molecular formulae of 11D5 and 23:2 isomers (O vs. CH 4 ) differ by 0.0364 Da in their exact masses with the one of FuFAs being lower compared to those of conventional fatty acids ( Table 3, Table S1).
In Despite these distinct mass differences between nominally isobaric molecular ions, extraction of the exact masses  Fig. S1). Clicking through the peak in MS 1 , the [M + H] + ion (here: m/z 1019.7698) varied in abundance and was accompanied by about ten other, typically more abundant, peaks with other masses (Fig. S1b). Hence, the assignment of FuFA-containing TAGs by LC-Orbitrap-HRMS necessitated the involvement of MS 2 data.
However, screening for FuFA-containing [M-RCOO] + fragment ions in MS 2 spectra did not yet lead to an unequivocal assignment of a TAG, since a fragment ion such as [L 9 M 5 ] + , for example, is formed from both LL 9 M 5 and OL 9 M 5 . Therefore, filtering for FuFA-containing TAGs in MS 2 should include [M + H] + (e.g., m/z 907.7391 in case of LL 9 M 5 ) (Fig. 2b).
The mass difference between theoretical and measured masses of [M + H] + in MS 2 was mostly < 2 ppm (  (Fig. 1), and also the ion formed by McLafferty rearrangement of FuFAs (here: m/z 109.0653 for 9M5). The simultaneous occurrence of these three types  (7D5, 9D3), m/z 291.2324 (9M5, 11M3) (Fig. 1), m/z 305.2481 (9D5, 11D3), m/z 319.2637 (11M5, 13M3), and m/z 333.2794 (11D5, 13D3). Since each exact mass of the [FCO] + fragment ion was formed by two FuFA isomers (Fig. 1), the list additionally covers four less relevant FuFAs with a shorter (7M5, 7D5) or longer (11M3, 13D3) carboxyalkyl chain. Additional, very rare alternatives of [FCO] + fragment ions of FuFAs with 7 or 13 C-atoms in the carboxyalkyl chain are compiled in the supporting information (Table S2). Still, distinguishing these pairs of positional isomers required the implementation of additional fragment ions (see below).  10,13,16,19,DPA; in TAGs: Dp), and DHA (Dh). Due to the low abundance of FuFAs, it was reasonable to assume that, if present, (detectable) TAGs will mainly feature only one FuFA. Therefore, emphasis was put on TAGs consisting of one FuFA and two conventional fatty acids (Table S3). For the sake of completeness, examples of TAGs with two FuFAs are listed in the supporting information (Table S4).  (Table S3) were software-filtered (Thermo Xcalibur) in the presence of the [FCO] + ion, the furan core ion, and the ion formed by McLafferty rearrangement (Figs. 3 and 4) allowing a maximum deviation of 4 ppm from the exact masses. Despite these precisely fitting selections, FuFA-containing TAGs could not be detected in case of an insufficient chromatographic separation of the sample. Apparently, overlaying highly abundant peaks from non-FuFA-TAGs inhibited the filter process. While this problem could be overcome by choosing a sufficiently good chromatographic separation (conditions are shown in the supporting information, Table S5), this problem is likely to exist in also other LC-HRMS studies where the selectivity is appropriate but filtering still fails.

Method of LC-Orbitrap-HRMS detection of FuFA-containing
Once this chromatographic problem was solved, the presence of a FuFA-containing TAG via [M + H] + ion could be verified by means of the [FCO] + ion in MS 2 spectra and narrowed down to two positional isomers (e.g., #7 and #8, Table 5). In a second step, the concrete structure of the FuFA was derived from the m/z value of the furan core ions (e.g., #7, m/z 305.2481). Finally, the full structure of the FuFA-containing TAG could be determined by means of the combined presence of (i) [M-RCOO] + fragment ions, (ii) ion formed by McLafferty rearrangement, and (iii) [M + H] + /[M + NH 4 ] + (Fig. 3). E.g., the presence of 9M5 in  (Fig. 2b).
Since only primary fatty acids are removed from TAGs, either one [M-RCOO] + ion (in the case of the same fatty acids in sn-1 and sn-3 position) or two [M-RCOO] + ions can result from this fragmentation. Specifically, LL 9 M 5 will generate [LL] + and [L 9 M 5 ] + while L 9 M 5 L will (almost) exclusively generate [L 9 M 5 ] + . Since LL 9 M 5 and L 9 M 5 L will likely co-elute in HPLC, the presence of only LL 9 M 5 or both positional isomers is difficult to determine if [LL] + and [L 9 M 5 ] + are detected in MS 2 . For the reason of simplicity, TAGs were named according to a specific scheme, mentioned at the beginning.
Application of the method to FuFA-containing TAGs in king oyster mushroom (Pleurotus eryngii) LC-Orbitrap-MS 2based filtering of a king oyster mushroom (Pleurotus eryngii) sample according to Table S3 and the follow-up steps described above enabled the identification of 18 FuFA-containing TAGs (Table 6). In agreement with a predominance of 9D5 >> 9M5 > 7D5, these FuFAs were present in 13, 3, and 2 TAGs, respectively. Similarly, the conventional fatty acids were represented by the most relevant of mushrooms, namely linoleic acid (18:2n-6, L) >> oleic acid (18:1n-9, O) > palmitic acid (16:0, P) [28]. Specifically, the most prominent peak originated from LL 9 D 5 followed by LO 9 D 5 . In the MS 2 spectrum, presence of LL 9 Table 6). Presence of other FuFA-containing TAGs in the sample was verified the same way.
The abundance ratio between the most abundant (LL 9 D 5 , intensity (I) = 66,800) and the least abundant (OO 9 M 5 , I = 121) FuFA-containing TAG was spread over more than two orders of magnitude (Table 6). Still, OO 9 M 5 could be assigned with the LC-Orbitrap-HRMS method. However, structural assignments were more difficult for FuFA-containing TAGs due to their low abundance in the sample. Co-elutions (see the first two TAGs in Table 6) still enabled the assignment of the characteristic [FCO] + fragment ion but not of the corresponding [M-RCOO] + fragment ions. Hence, only the FuFA but not the conventional fatty acids could be identified in the TAG (see the first example, Table 6).

FuFA-containing TAGs in two fish oil samples
The fish oil I was dominated by the conventional fatty acids EPA and DHA and FuFAs 11D3, 11D5, and 9M5 [18]. Combinations of these fatty acids were found in nine of the thirteen FuFA-containing TAGs that could be detected in the sample (Table S6). Four TAGs contained 9M5 with the most abundant one (EpEp 9 M 5 ) having an intensity of I = 2820 (Table S6). Representatives containing 11D3 (n = 4) were less abundant (maximum I = 626 for EpEp 11 D 3 ) while the Fig. 4 Filter of the MS 2 mass spectra in LC-Orbitrap-HRMS measurements of four TAGs obtained from the synthesis of LLL and 9M5 [6]. Trigger points of the precursor ions are shown   (Table S6).

Method information
The reproducibility of the method was tested by three injections of fish oil II (gilthead liver oil) and comparison of the relative abundances of four TAGs and five FuFA-containing TAGs (Table S8). The most abundant non-FuFA and FuFAcontaining TAG of the first run was used as the reference (100%). Based on these data, conventional (and more abundant) TAGs showed good reproducibility (10-15% relative STDEV) with the exception of OOO due to its low relative abundance in run 3 (Table S8). Compared to that, relative standard deviations of FuFA-containing TAGs were in the range of 12-37% which was deemed acceptable given the fact that abundances were lower.
The limit of detection (LOD) of FuFA-containing TAGs could not be directly determined due to the lack both of reference standards and MSMS responses. Hence, it was aimed to estimate the minimum amount of a FuFA-containing TAG indirectly from previous data in the corresponding samples which were obtained after transmethylation and determination of the FuFAs as fatty acid methyl esters (FAMEs). Specifically, the king oyster mushroom was found to contain 33 mg 9D5/100 g fungi dry weight (determined as FAME) [29]. Based on the prerequisites that (i) all 9D5 in the mushroom was stored in the TAG fraction and that (ii) LC-Orbitrap-MSMS responses of all FuFA-containing TAGs were similar (which was not unlikely due to the formation of the highly abundant ion formed via McLafferty rearrangement), we could calculate the individual contribution of FuFA-containing TAGs from the peak areas and related them to the reported sum value (Table S9). According to that, the minimum amount that could be determined in a FuFA-containing TAG of the mushroom was estimated at 0.1 mg/100 g dry weight (Table S9). Based on a lipid content of ~ 5% lipids (estimated) in the dry weight of the mushroom, this corresponds with the lowest detectable amount of 2 mg/100 g lipids FuFA-containing TAG.
In the same way, we also estimated the content of 9M5, 11D3, and 11D5 in fish oil I with a FuFA amount of 1030 mg/100 g lipids as determined via methyl esters [18]. Accordingly, the lowest measurable contribution originated from EpDp 11 D 5 at 4.6 mg/100 g lipids (Table S10).
Overall, the minimum amount that could be estimated in this way in two samples was in the same range of 1-10 mg FuFA-containing TAGs/100 g lipids. This concentration range may be considered in the planning of further studies on FuFA-containing TAGs. * Fatty acids are listed as follows: first two conventional fatty acids with the one with higher mass listed first, followed by the FuFA in position three of the glycerol backbone. This order must not reflect the correct order of fatty acids ** From top to bottom: on the left: ion formed by McLafferty rearrangement, furan core ion, base peak ([FCO]+)-formally [FuFA-OH]+, printed in bold), and on the right: finally two or three [R-COOH]+ ions *** Signal intensity of the base peak, divided by a factor of 1000. In the case of FuFA-containing TAGs, this is the [FCO] + ion (formally [FuFA-OH] + , formed by α-cleavage) **** XY stands for two conventional fatty acids that could not be clearly assigned. Based on the characteristic fragment ions, the FuFA could be assigned