Computational mass spectrometry accelerates C=C position-resolved untargeted lipidomics using oxygen attachment dissociation

doi:10.21203/rs.3.rs-727852/v1

Download PDF

Article

Computational mass spectrometry accelerates C=C position-resolved untargeted lipidomics using oxygen attachment dissociation

https://doi.org/10.21203/rs.3.rs-727852/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 19 Dec, 2022

Read the published version in Communications Chemistry →

Version 1

posted

You are reading this latest preprint version

Mass spectrometry-based untargeted lipidomics has revealed the lipidome atlas of living organisms at the molecular species level. Despite the double bond (C=C) position being a crucial factor for enzyme preference, cellular membrane milieu, and biological activity, the C=C defined structures have not yet been characterized. Here, we present a novel approach for C=C position-resolved untargeted lipidomics using a combination of oxygen attachment dissociation and computational mass spectrometry to increase the rate of annotation. We validated the accuracy of our platform as per the authentic standards of 21 lipid subclasses and the biogenic standards of 51 molecules containing polyunsaturated fatty acids (PUFAs) from the cultured cells fed with various fatty acid-enriched media. By analyzing human and mice-derived biological samples, we characterized 675 unique lipid structures with the C=C position-resolved level encompassing 22 lipid subclasses defined by LIPID MAPS. Our platform also illuminated the unique profiles of tissue-specific lipids containing n-3 and/or n-6 very long-chain PUFAs (carbon M 28 and double bonds a 4) in the eye, testis, and brain of the mouse.

Biotechnology and Bioengineering

Computational Biology

mass spectrometry

lipidomics

liquid chromatography (LC)

Lipids are essential compounds for living organisms that play various physiological functions, including the formation of components of the cell membrane, signaling mediators, storage of energy, and the formation of the epithelial barrier¹. These functions are driven by its structural diversity, where the specific composition of its membrane environment is formed by the tissues, cells, organelles, and the microdomain scale^1–3. The structural diversity of lipids is estimated to be over 45000⁴ as curated in LIPID MAPS, yielding from different combinations of the molecular backbone, polar head groups, acyl chains, double-bond positions, sn-positions, and stereo-centers. Disruption of the lipid homeostasis is related to various diseases, such as atherosclerotic lesions⁵, cancer⁶, neurodegeneration⁷, and infertility⁸. To understand alterations in lipid homeostasis, identify specific biomarkers, and elucidate the complex pathology underlying diseases, untargeted lipidomics is a powerful technology in this multi-omics era⁹. Liquid chromatography (LC) coupled to tandem mass spectrometry (MS/MS) with low-energy collision-induced dissociation (CID) is the gold standard technique for characterizing lipid structures¹⁰. In addition, advances in computational mass spectrometry (CompMS) facilitating in silico MS/MS library generation, rule-based mass spectrum elucidation, and predictions of collision cross section and retention time have greatly contributed to expanding the annotation rates and reducing the rates of false positive identification^11–13. While the LC-CID-MS/MS approach provides information for >1000 lipids for a single specimen, the diversity of C=C positional isomers remain obscure as the fragment ions involved in C=C positions are rarely detected by low-energy CID, they significantly affect the enzyme selectivity¹⁴, lipid mediator functions^15,16, tissue-specific membrane composition, and physical properties¹⁷. Therefore, understanding the diversity of C=C positional isomers is important to explore the mechanism of lipid homeostasis in each tissue as well as its alteration in various diseases.

Two types of MS methods can be used to resolve the double bond position of lipids. One utilizes chemical derivatization, where the derivatized C=C moiety is preferentially cleaved in low-energy CID. Paterno-Buchi (PB) photochemical reaction^18–22 and mCPBA epoxidation for lipid double-bond identification (MELDI)^23,24 are popular techniques that have advantages in a higher signal-to-noise ratio without the need for special MS instruments. Another uses a different principle of mass fragmentation, which includes ultraviolet photodissociation (UVPD)^25–28, ozone-induced dissociation (OzID)^29–33, and electron impact excitation of ions from organics (EIEIO)^34–37. These methods provide the fragment ions associated with C=C positions from the native molecular structure and are adaptable to conventional LC systems. The state-of-the-art techniques have accelerated the characterization of C=C positional isomers of glycerophospholipids, glycerolipids, sphingolipids, sterol lipids, and fatty acids. Furthermore, imaging MS coupled with these methodologies revealed that the C=C positional isomers are spatially distributed in the gray and white matter of the mouse brain^38–41 as well as in cancerous and healthy tissues^24,39,42.

Several informatics tools have been developed to facilitate C=C position-resolved lipidomics^18,43; however, resolving the C=C positions of various lipid subclasses in an unbiased manner, that is, C=C resolved-untargeted lipidomics, is still not accomplished owing to the lack of CompMS techniques. Moreover, characterization of double bond positions for two or three unsaturated acyl chains is elusive because of the highly complicated MS/MS spectra. In fact, we estimated that the MS/MS spectrum of oxygen attachment dissociation (OAD)⁴⁴ used in this study contains hundreds of fragment ions on average, which are over three-fold higher than those in CID spectra, which is also true in other fragmentation techniques^{19,24,29,31,36}. Thus, advances in CompMS are an emerging need for lipidology to comprehensively characterize the C=C positional isomers of lipids with fewer false positive annotations.

In this study, we developed a mass spectrometry radical-induced dissociation decipherer (MS-RIDD; https://github.com/Metabolomics-CompMS/MS-RIDD.git) to facilitate C=C position-resolved untargeted lipidomics using oxygen attachment dissociation (OAD)⁴⁴. OAD-based mass fragmentation is known as radical-induced dissociation, where atomic oxygen (O) and hydroxyl radicals (OH) effectively cause C=C-specific fragmentation. OAD-MS/MS method has also demonstrated its practicality for the annotation of C=C positions in a previous study⁴⁵. OAD-MS/MS and OzID have a similar feature in which both fragmentations occur due to the electrophilicity of oxygen atoms; however, compared to an ozone generator, OAD-MS/MS can utilize water as a safe radical source⁴⁶. Importantly, our approach integrates the LC-CID-MS/MS and LC-OAD-MS/MS data. The global profiling of lipids at the molecular species level was performed using LC-CID-MS/MS and the C=C positions were characterized using LC-OAD-MS/MS. We formulated the OAD fragmentation rules by measuring the authentic standards of 21 lipid subclasses and implemented them in the MS-RIDD software program. Furthermore, we validated our platform by 222 experimental MS/MS spectra of 51 biogenic lipid standards obtained from cultured cells with the supplement of various free fatty acid-containing media, in which the double bond positions of lipids containing polyunsaturated fatty acids (PUFAs) are defined. Finally, we applied our untargeted lipidomics platform to human plasma and mouse tissues (liver, brain, eye, skin, testis, and feces), and were able to successfully characterize a total of 675 lipid species from 22 lipid subclasses having a C=C position-defined level.

Structural analysis of C=C positional isomers by OAD-MS/MS

To elucidate the fragmentation rules of OAD, we used 31 authentic lipid standards of 21 lipid subclasses (Supplementary Table 1 and Supplementary Fig. 1). As the terminology, n-description and delta-description are used for acyl chain and sphingoid base moieties, respectively. For instance, the acyl chains of arachidonic acid and sphingosine are 20:4(n-6,9,12,15) and 18:1(Δ4):2O, respectively. In principle, the OAD mass fragmentation provides two major fragment ions for a C=C moiety: neutral losses of 97.1381 Da (cleavage of n-8_n-9 bond) and 139.1487 Da ( n-10_n-11 bond) were observed in lipids containing C18:1(n-9) acyl chains (Fig. 1a and Supplementary Fig. 1). In fact, the “fragment pair” of these two fragment ions was defined as the essential criterion to select the structure candidates. Moreover, we characterized fragment ion structures associated with a single double bond, as exemplified by phosphatidylcholine (PC) 18:0/18:1(n-9) (Fig. 1b): atomic oxygen or hydroxyl radical attaches to double bonds and cleaves at three locations, namely at n-8_n-9, n-9_n-10, and n-10_n-11 in the case of 18:1(n-9). The details of the fragmentation schemes are summarized in Supplementary Fig. 2 and Supplementary Table 2. Moreover, the fragment ions were assigned using the same fragmentation scheme derived from six double bonds of docosahexaenoic acid (DHA) with a mass accuracy of 10 ppm (Fig. 1c). Thus, fragmentation patterns can be formulated to efficiently elucidate any position of C=C, even in polyunsaturated acyl chains.

The determination of C=C locations in plasmalogen and sphingoid bases was also accomplished. The vinyl ether specific fragment ions were confirmed as m/z 521.3148 and 522.3122 derived from the cleavage of the n-17_n-18 bond in the P-18:0 acyl chain (Fig. 1d and Supplementary Fig. 1). Moreover, the fragment ions in the sphingobase were described by the example of ceramide (Cer) 18:2(Δ4,8) and 2O/24:1(n-9) (Fig. 1e). Due to the polar interference of the neighboring oxygen atom, fragmentation patterns of C=C in plasmalogen and Δ4 in sphingoid base³¹ are different when compared to those of the conventional acyl chains, while these dissociation patterns are also within the scope of the systematized fragmentation rules (Supplementary Fig. 2).

The results for N-acyl phosphoethanolamine, PE-N (FA 20:4(n-6,9,12,15)) 18:1(n-9)/18:1(n-9), which is an example of a lipid standard with different unsaturated acyl chains, showed that product ions indicating C=C locations of both oleic- and arachidonic-acyl chains were fully assigned (Fig. 1f). Our observations also indicated the importance of high-resolution mass spectrometry (~30000 full width at half maximum, FWHM) for distinguishing m/z 933.6816 from n-7_n-8 cleavage in C20:4 and m/z 933.6452 from n-8_n-9 cleavage in C18:1 (Δm/z = 0.0364; 21,334 FWHM). Hence, C=C positional isomers can be distinguished theoretically by the time-of-flight mass resolution, even though the molecule contains multiple PUFA chains such as PC 20:4(n-6,9,12,15) C22:6(n-3,6,9,12,15,18) and PC 20:4(n-3,6,9,12) C20:4 (n-6,9,12,15). Importantly, the systematized dissociation rules enable us to predict the MS/MS spectrum of lipids, resulting in the comprehensive creation of in silico lipid reference spectra, empowering the unexpected discovery of C=C positional isomers. In addition, polar head-specific product ions were detected in the cases of PC (m/z 184.0734; C₅H₁₅NO₄P⁺), PG (m/z 695.5052 Da; NL of C₃H₈O₆P), and PE (m/z 589.5543; NL of C₂H₈NO₄P) were detected in OAD-MS/MS, although the fragment that defines the chain length and double bond equivalent is rarely observed (Fig. 1b-d). Therefore, we concluded that the integrated analysis of LC-CID-MS/MS for molecular species level annotation (e.g., PC 18:1_20:4) and LC-OAD-MS/MS for C=C position assignment is required to accomplish untargeted lipidomics addressing precise C=C structural isomers.

Strategy for C=C resolved lipidomics by integrating the CID- and OAD spectral data

We developed a MS-RIDD to facilitate the integrated strategy using LC-CID-MS/MS and LC-OAD-MS/MS for C=C resolved untargeted lipidomics (Fig. 2a). Both CID- and OAD tandem mass spectral data were obtained by data-dependent acquisition under the same LC conditions. First, a biological sample was analyzed by positive- and negative ion modes of LC-CID-MS/MS, and the MS data were analyzed using the MS-DIAL software program¹³ that provides the global profile of lipids at the molecular species level (e.g., PC 18:1_22:6). Next, the same sample was analyzed using the positive ion mode of LC-OAD-MS/MS. The core algorithm of MS-RIDD uses the annotated lipid name (PC 18:1_22:6) and the tandem mass spectrum of OAD-MS/MS, and it assigns the C=C position information such as PC 18:1(n-9) 22:6 (n-3,6,9,12,15,18). For the batch process, the MS-RIDD program can directly import the MS-DIAL result and OAD-MS data where the mass tolerance of 0.01 Da and the retention time (RT) tolerance of 0.15 min were utilized to automatically integrate these MS data in this study.

Here, we describe the core algorithm of the MS-RIDD (Fig. 2b). First, MS-RIDD generates theoretically possible structural candidates based on the lipid molecular species information. In the case of PC 18:1_22:6, a total of 14 candidates for C18:1 acyl chains (n-3 to n-16) and 1,716 candidates for C22:6 acyl chains ranging from (n-3, 5, 7, 9, 11, 13) to (n-10, 12, 14, 16, 18, 20) were generated. Next, the experimental OAD-MS/MS spectrum was evaluated for each of the structure candidates using the in silico fragmentation rule. The fragmentation rule generates reference product ions associated with the C=C positions for C18:1 and C22:6, respectively (see Supplementary Table 2). A candidate was excluded if the essential fragment pairs (Fig. 1a) were not detected where a mass accuracy of 15 ppm and relative intensity tolerance of 0.01% were used in this study. The relative intensity tolerance was set to 0.005% in the case of Δ4 in the sphingoid base because of the low ion abundance. When several candidates remain after the diagnostic ion filtering, the MS/MS reverse dot-product similarity by square-root-transformed value of intensity⁴⁷ is used to prioritize the candidates; the heuristic intensity of the reference spectrum is summarized in Supplementary Table 2. For lipids containing di- and triacyl chains, the double bond position in the acyl chain having the highest double bond equivalent in the molecule was initially evaluated: C22:6 was first evaluated, and C18:1 was then characterized in PC 18:1_22:6.

MS/MS spectra of biogenic lipid standards to validate the MS-RIDD software

To validate the MS-RIDD program, we prepared biogenic samples of HEK293 cells cultured in various PUFAs-enriched media (Supplementary Table 3). Free fatty acids are actively incorporated into membrane phospholipids in cultured cells⁴⁸. Therefore, the reference MS/MS spectrum of C=C position-defined lipids can be obtained by comparing the LC-MS data between PUFA-supplemented and non-supplemented cultured cells. For instance, based on the extracted ion chromatograms (EIC) of m/z 778.54, reference MS/MS spectra of at least three PC molecules having different acyl chains, 14:0_22:6, 16:1_20:5, and 18:3_18:3, were obtained (Fig. 3), where the cells were cultivated with no supplementation as a control (Fig. 3a), docosahexaenoic acid {DHA; C22:6(n-3,6,9,12,15,18)} (Fig. 3b), eicosapentaenoic acid {EPA; C20:5(n-3,6,9,12,15)} (Fig. 3c), α-linoleic acid {α-LA; C18:3(n-3,6,9)} (Fig. 3d), γ-linoleic acid {γ-LA; C18:3(n-6,9,12)} (Fig. 3e), and both α-LA and γ-LA (Fig. 3f). The ion abundance of each EIC was normalized to the cell number (Supplementary Table 4).

As judged from the differential abundance of peak intensity under DHA-supplemented conditions, we annotated the chromatographic peak at R.T. 9.15 min condition as PC 14:0_22:6(n-3,6,9,12,15,18) via the CID-MS/MS and OAD-MS/MS spectra (Fig. 3b). Similarly, the peak from the EPA-supplemented condition was characterized as PC 16:1(n-7) 20:5(n-3,6,9,12,15) by elucidating the CID-MS/MS and OAD-MS/MS spectra (Fig. 3c). In the α-LA-supplemented condition, the peak was determined to be PC 18:3(n-3,6,9) 18:3(n-3,6,9), and the evidence for n-6 C18:3(n-6,9,12) was not fully observed in the OAD-MS/MS spectrum (Fig. 3d). On the other hand, the tandem mass spectrum in the γ-LA-supplemented sample was annotated as PC 18:3(n-6,9,12) 18:3(n-6,9,12), and the diagnostic ions for n-3 18:3(n-3,6,9) were not detected (Fig. 3e). Moreover, the OAD-MS/MS spectrum of the samples cultured with both α-LA and γ-LA contained the diagnostic ions of n-3 18:3(n-3,6,9) and n-6 18:3(n-6,9,12), resulting in the annotation of PC 18:3(n-3,6,9)_18:3(n-6,9,12) (Fig. 3f). This result clearly demonstrates that our methodology can differentiate the C=C positions of isobaric C=C isomers; the details are summarized in Supplementary Fig. 3.

Consequently, these results indicate that the integrated strategy of CID- and OAD-MS/MS in combination with the MS-DIAL and MS-RIDD programs is widely applicable to various C=C positional isomers, which accelerates the elucidation of the structural diversity of lipids. More importantly, we obtained the CID- and OAD-MS/MS spectra of 51 lipid molecules whose acyl chains and C=C positions were defined, which can be utilized as the true positive MS/MS set for the validation of our untargeted lipidomics platform.

Validation of the MS-RIDD program using the true positive spectral records

As mentioned above, we successfully obtained a series of reference OAD-MS/MS spectra of various C=C isomers from the lipid extract of cultured cells supplemented with different PUFAs. The lipid structure was annotated with the double bond positional information (1) if the peak height is 1.5 times increased in the supplemented sample compared to the control sample and (2) if the essential diagnostic ions to define the C=C position were confirmed in the OAD-MS/MS spectrum with the mass tolerance of 0.01 Da and the relative intensity cutoff of 0.01 % (Supplementary Table 5). Finally, a total of 222 OAD-MS/MS spectra from 51 lipid molecules were identified by the C=C defined molecular species level, and they were used as the true positive records to evaluate the performance of the MS-RIDD program (Supplementary Data 1). Two parameters, that is, mass tolerance and relative intensity threshold, should be optimized for MS-RIDD, while the mass tolerance was fixed at 15 ppm in this study. For the optimization of the relative intensity threshold, the value ranged from 0–1 % (Fig. 4a). The accuracy of MS-RIDD was maximized at 97.29% with a threshold of 0.01 %; in detail, the accuracy was 100% and 95.55% for the spectra with a single unsaturated moiety and multiple unsaturated moieties, respectively (Fig. 4b). Complex lipids such as PC 20:4(n-6,9,12,15) 22:6(n-3,6,9,12,15,18) and triacylglycerol (TG) 18:1(n-7)_18:1(n-9)_18:3(n-3,6,9) were also characterized correctly by MS-RIDD (Fig. 4c-d).

On the other hand, the mis-annotation rate of 2.71% was mostly due to the ambiguous OAD-MS/MS spectra, where several candidates such as 20:4(n-3,6,9,12) 20:4(n-6,9,12,15) and 20:4(n-3,9,12,15)_20:4(n-6,9,12,15) were assigned as the top candidates equally. For ambiguous cases, the graphical user interface of MS-RIDD allows one to visualize all candidates that exceed the thresholds for annotation and to modify the annotation result if needed, which is important because biologists have prior knowledge about the biosynthetic pathway to resolve the double bond position of lipid molecules. Importantly, the automated workflow by the MS-DIAL and MS-RIDD programs drastically reduces the time and effort required for data analysis and provides an objective result with ~97% annotation precision.

Application of C=C position-resolved untargeted lipidomics to biological samples

We applied our untargeted lipidomics platform to biological samples, including six mouse tissues (brain, eye, skin, liver, testis, and feces) and NIST SRM 1950 human plasma. In total, we characterized 675 unique structures of 22 lipid subclasses among seven biological samples by the confidence of the C=C position-defined molecular species level (Supplementary Fig. 4). Lipids were semi-quantified according to lipidomics standards initiative (LSI) level 2 or 3 using EquiSPLASH internal standards mixture according to a previous report¹³. On average, >100 unique C=C resolved structures were annotated in each tissue (Supplementary Table 6), and the detailed distribution of C=C isomers to lipid subclasses was determined.

Hierarchical clustering analysis (HCA) using the binary matrix (1=detected and 0=not detected) of lipid molecules was employed to reveal the common and specific lipid molecules in each tissue. The common-lipid segment included lipid molecules containing major fatty acids in mammalian cells: 16:1(n-7), 18:1(n-9), 18:2(n-6,9), 20:4(n-6,9,12,15), and 22:6(n-3,6,9,12,15,18) of acyl chains (Supplementary Fig. 5a-b). In addition, the common segment of phospholipids contained 18:1(n-7), 20:2(n-6,9), 20:3(n-6,9,12), 20:4(n-3,6,9,12), and 22:5(n-3,6,9,12,15) of the acyl chain (Supplementary Fig. 5a), and that of other lipid classes included 18:3(n-3,6,9 and n-6,9,12), and 20:5(n-3,6,9,12,15) acyl chains (Supplementary Fig. 5b). As sphingolipid moieties, 24:1(n-9) of the N-acyl chain, and 17:1(Δ4);2O, 18:1(Δ4);2O and 18:2(Δ4,14);2O of SPB were also included in the common-lipid segment (Supplementary Fig. 5b). Several lipids containing 18:1(n-6, n-10, and n-11) and 18:2(n-7,9), which have rarely been reported in lipidology, were also located in the common lipid segment (Supplementary Fig. 5b). The CID- and OAD-MS/MS spectra of all characteristic C=C isomers are summarized in detail (Supplementary Fig. 6).

In the feces sample, the total number of resolved C=C isomers was lower than that of other tissues because the acyl composition was rich in saturated moieties (Supplementary Table 6). Cyclopropane-containing acyl chains are frequently observed in microbiota lipidome⁴⁹, which could not be distinguished from those containing double bonds by our current methodology. However, the unique sphingoid base SPB 18:1(Δ9);2O in Cer-NS, SPB 18:1(Δ8);3O in ceramide alpha-hydroxy fatty acid-phytosphingosine (Cer-AP) and SPB 19:3(Δ4,8,11);3O in SM, were found in the fecal matter (Supplementary Fig. 5b). We previously reported the presence of sphingolipids containing SPB 19:3;2O in feces at the molecular species level¹³ while in this study, we could further annotate the moiety as SPB 19:3(Δ4,8,11);2O. The polyunsaturated sphingoid base of SPB 19:3;2O has been previously identified as 2-amino-9-methyl-4,8,10-octadecatriene-1,3-diol in starfish spermatozoa⁵⁰ and squid nerve⁵¹ by nuclear magnetic resonance and positive-ion fast atom bombardment tandem mass spectrometry, respectively. Thus, although there is a possibility that we might have annotated the novel SPB structure, the detected SPB could be SPB 19:3(Δ4,8,10, 9-Methyl);2O. The skin tissue contains a rich amount of ceramides⁵², including ceramide esterified omega-hydroxy fatty acid-sphingosine (Cer-EOS), which is responsible for skin barrier function⁵³. Our results successfully identified that the O-acyl chain of Cer-EOS is linoleic acid, 18:2(n-6,9) (Supplementary Fig. 5b), which has been confirmed in a previous study⁵⁴. In addition, we revealed that SPB 17:1;2O(Δ4) containing Cer-NS and hexosylceramide non-hydroxy fatty acid-sphingosine (HexCer-NS) are unique and highly abundant ceramides in mouse skin. In human plasma samples, our approach clarified the unique C=C positional isomers in the ether-link acyl chain, in addition to the plasmalogen (Δ2) structure, such as O-16:1(n-6) and O-18:1(n-6) in ether PC and ester LPC lipid subclasses, respectively (Supplementary Fig. 5a).

We clarified the detailed profile of the C=C positional isomers of PUFAs (C18~26), sphingoid bases (d17:1, d18:1, d18:2), and very long chain-polyunsaturated fatty acids (VLC-PUFAs) (C28–42) in the complex lipids (Fig. 5). The lipid molecules were semi-quantified using the internal standards of EquiSPLASH (LSI level 2 semi-quantification). In addition to the clear distribution of n-3 and n-6 fatty acids, our results also revealed the unique distribution of n-9 PUFAs in the liver eye, human plasma, and testis: n-9 PUFAs were found in C20:3, 22:4, 24:3, and 24:4. Moreover, phospholipids containing C22:5 were characterized as n-3 in the brain, eye, liver, and human plasma, as n-6 in the testis, and both in the skin. The n-6 feature in the testis is consistent with previous studies^55,56, which reported that osbond acid (C22:5n-6) is abundant in the glycerolipids and glycerophospholipids in the testes of mice.

Characterization of C=C locations in sphingoid bases is challenging because the diagnostic ions are often less abundant compared to that of O- and N-acyl chains, which is due to the effect of neighboring hydroxyl functional group³¹. The MS-RIDD program enables a comprehensive understanding of the diversity of C=C positional isomers in sphingoid bases (Fig. 5b). Our results indicated that the Δ4 sphingobase 18:1(Δ4),2O is the major component of sphingolipids, which are biosynthesized by dihydroceramide Δ4-desaturase during the ceramide synthesis pathway converting ceramide non-hydroxy fatty acid-dihydrosphingosine (Cer-NDS) into Cer-NS⁵². The C=C positions in the sphingadienine moiety have recently been reported as 18:2(Δ4,14);2O in several studies^20,57,58. Our platform characterized 18:2(Δ4,14);2O in various tissues containing mouse brain, eye, liver, and human plasma. Interestingly, we also identified 18:1(Δ14);2O which is the sphingosine isomer biosynthesized by Δ14-desaturase from dihydrosphingosine or Cer-NDS⁵⁸. While our results suggest that the C=C location of 18:1;2O is dominantly Δ4 in the seven biological samples, 4-hydroxy-8sphingenine 18:1(Δ8);3O and 8-sphinganine 18:1(Δ8);2O, which are known as major sphingoid bases in plant^59,60, were also annotated in feces and liver, respectively. Moreover, we annotated the unique sphingosine isomer of 18:1(Δ9);2O in Cer-NS, the metabolic pathway of Cer-NS remains to be determined.

We characterized the unique PC lipid molecules containing n-3 and n-6 VLC-PUFAs (C36–42) in the brain tissue. In the eye tissue, in addition to the unique distribution of n-3 fatty acids, such as 22:5 (n-3,6,9,12,15), 24:5(n-3,6,9,12,15), and 22:6 (n-3,6,9,12,15,18) into the phosphatidylserine (PS) lipid subclass, our results also indicated that the n-3 acyl chain of VLC-PUFAs (C30–36) is only present in the PC subclass. Furthermore, we found the specific distribution of n-3, n-6, and n-9 VLC-PUFAs (C28–32) into Cer-NS, PC, sphingomyelin (SM), and TG (Fig. 5c). Our previous study characterized the diversity of O- and N-acyl VLC-PUFAs in the testis, while the C=C positions remain to be elucidated¹³. Direct evidence of C=C positions in VLC-PUFAs is crucial because these physicochemical properties may affect the formation of tissue-specific lipid milieus surrounding rhodopsin conformation⁶¹ and spermatogenesis⁸, as well as elovanoids, a series of VLC-PUFA-derived oxylipins with protective functions toward photoreceptors⁶² and neurons⁶³. In this study, all VLC-PUFAs containing six double bonds were annotated as n-3, while those containing either five or four double bonds contained two isomers, n-3 and n-6, and n-6 and n-9, respectively.

We developed a novel C=C position-resolved untargeted lipidomics platform and applied it to biological samples. The advantage of OAD-based mass fragmentation when compared to other related techniques^19,24,37 is that it produces product ions, which are easy to interpret and associated with double bond positions in O-, P- N- acyl chains from intact precursor ions ranging from various lipid subclasses without any derivatization procedure. However, the lipid MS/MS spectra are hardly elucidated because of the complexity and lower intensity of product ions than that in CID-MS/MS. In this study, we elucidated the fragmentation rules of OAD-MS/MS and developed the MS-RIDD software program, which rapidly and precisely resolves complex OAD-MS/MS spectra obtained from crude samples. Various methods have been devised to resolve different C=C positional isomers, but their data validations are limited owing to the limited availability of lipid standards. In this study, we demonstrated the biogenic synthesis of various lipids with C=C positional isomers. With these, we first overcome this limitation by means of a well-validated biochemical approach: preparing lipid extracts from cultured cells fed with PUFA-enriched media. The analysis of biogenic lipid extracts enabled us to create a true positive dataset of OAD-MS/MS on a large scale. Using the true positive dataset, we estimated the accuracy of MS-RIDD to determine acyl chain C=C positions as 97.29%. Our novel workflow illuminated 675 C=C positional isomers in mammalian tissues, including tissue-specific lipid structures that were previously unknown; n-3, n-6, and n-9 of (VLC-) PUFAs and other isomers of sphingoid bases annotated in this study would provide new insights to update the conventional understanding of C=C positional isomers present in mammalian lipidomes.

For example, in addition to the major PUFAs, namely α-/γ-LA, ARA, EPA, DPA, and DHA, we annotated unique PUFA isomers such as n-3, n-6, and n-9 of C20:3, n-3 of C20:4, and n-6 of C22:4 in the complex lipid structures. Although the Human Metabolome Database (HMDB; https://hmdb.ca/) recorded the concentration of n-3 fatty acids of C20:4 in human serum^64,65 as >1000 times lower than that of arachidonic acid (n-6 of C20:4), our platform showed that >10% of PC containing C20:4 was annotated as n-3 species in human plasma. In addition, our platform could characterize the unique distribution of the n-9 series of C22:4, C24:3, and C24:4 in complex lipids, some of which have been reported in previous reports^66,67. Moreover, we elucidated the unique distribution of VLC-PUFAs in the mouse brain, eye, and testis (Supplementary Fig. 5 and Fig. 5c), and PC molecules containing n-3 VLC-PUFA were present in the eye, while those containing n-6 VLC-PUFA were present in the testis. We also characterized the unique distribution of the acyl chain length of VLC-PUFAs in each tissue: C28–32 in the testis, C30–36 in the eye, and C36–42 in the brain. These results imply that the definite lipid bilayer milieu is involved in tissue-specific cellular functions yet to be characterized.

Further development is needed to increase the annotation coverage of our novel lipidomic platform. For instance, the product ion sensitivity of OAD-MS/MS derived from phosphatidylinositol (PI) and phosphatidylglycerol (PG) lipid classes is not sufficient to be elucidated in biological sample data, although the essential fragment ions to identify C=C positions in these lipid classes were detected by analyzing the authentic standards (Fig. 1c and Supplementary Fig. 1). In addition to the instrumentation issue, we should further develop the MS-RIDD program. In fact, the neutral losses of the PI- and PG polar heads are prioritized in the fragmentation of OAD than the sequential losses of acyl chains associated with double bond positions, which do not follow the fragmentation rules that we formulated. Furthermore, complex lipids with four acyl chains, such as cardiolipin (CL), are also out of the scope of the current program because of the computational cost considering the double bond positions in the four acyl chains. In principle, the OAD-based mass fragmentation can be executed in negative ion mode, as the concept has been demonstrated by hydrogen attachment dissociation⁶⁸ whose principle of mass fragmentation is the same as that of OAD. Several important lipid subclasses, such as PI, PG, phosphatidic acid (PA), CL, and free fatty acids are well ionized in the negative ion mode. Therefore, the integrated methodology using OAD-MS/MS spectra from both positive and negative ion modes will increase the coverage and confidence in lipid annotations.

Our methodology will be more powerful when combined with other state-of-the-art mass spectrometry technologies, for example, ion mobility to separate co-eluted C=C positional isomers, and imaging MS for spatially resolved lipidomics. The structural annotation range in our platform depends on the coverage of the in silico MS/MS library coupled with the predicted retention time established by LC-CID-MS/MS data curation; thus, we will continue to expand the library incorporating new CompMS technology such as feature-based molecular networking⁶⁹. Future efforts should also focus on improving the sensitivity and specificity of our methodology, inventing a more sophisticated deconvolution method to distinguish C=C isomers in crude OAD-MS/MS, constructing a deeper analysis workflow to determine the positions of branched chains and functional groups. Untargeted lipidomics addressing precise C=C structural isomers will open up a new avenue for the discovery of unique lipid biomarkers and metabolic pathways, leading to a deeper understanding of the biology of various C=C isomers in lipids.

Chemicals

All authentic lipid standards, CAR 18:1(9Z), LPC 18:1(9Z), LPE 18:1(9Z), LPG 18:1(9Z), LPI 18:1(9Z), LPS 18:1(9Z), PA 18:1(9Z)/18:1(9Z), PC 14:1(9Z)/14:1(9Z), PC 16:1(9E)/16:1(9E), PC 18:1(9Z)/16:0, PC 18:0/18:1(9Z), PC 18:1(9Z)/18:1(9Z), PC 18:3(9Z,12Z,15Z)/18:3(9Z,12Z,15Z), PC 16:0/20:4(5Z,8Z,11Z,14Z), PC O-16:0/18:1(9Z), PC P-18:0/18:1(9Z), PE 18:1(9Z)/18:1(9Z), PE O-16:0/18:1(9Z), PE P-18:0/18:1(9Z), PE-N(FA 20:4(5Z,8Z,11Z,14Z)) 18:1(9Z)/18:1(9Z), PG 18:1(9Z)/18:1(9Z), PG 22:6(7Z,10Z,13Z,16Z,19Z)/22:6(7Z,10Z,13Z,16Z,19Z), PI 18:0/20:4(5Z,8Z,11Z,14Z), PS 18:1(9Z)/18:1(9Z), HBMP 18:1(9Z)/18:1(9Z)/18:1(9Z), Cer-NS 18:1(4E);2O/18:1(9Z), Cer-NS 18:2(4E,8Z);2O/24:1(15Z), Cer-NS 18:2(4E,8Z);2O/24:0, DG 18:1(9Z)/18:1(9Z), TG 18:1(9Z)/18:1(9Z)/18:1(9Z), and equiSPLASH, which were used as internal standards, were purchased from Avanti Polar Lipids, Inc (Alabama, US). Alpha-linoleic acid (alpha-LA), gamma-linoleic acid (gamma-LA), eicosatetraenoic acid (ETA), arachidonic acid (ARA), docosapentaenoic acid (DPA), osbond acid (OsA), eicosapentaenoic acid (EPA), and docosahexaenoic acid (DHA) were purchased from Cayman Chemical (Michigan, US) and used as supplements for cell cultures. Details are described in Supplementary Tables 1 and 3. The National Institute of Standards and Technology (NIST SRM 1950) human plasma samples were purchased from NIST (Maryland, US). Acetonitrile (ACN), methanol (MeOH), and isopropanol (IPA) of LC-MS grade were purchased from Wako (Tokyo, Japan). Chloroform (CHCl₃) was purchased from Sigma–Aldrich (Tokyo, Japan). Ammonium acetate and ethylenediaminetetraacetic acid (EDTA) were purchased from Wako and Dojindo (Tokyo, Japan), respectively. Milli-Q water was purchased from Millipore (Massachusetts, US).

Sample preparation

The brain, eye, liver, skin from the ear, testis, and plasma of 8 weeks male mice (C57BL/6J background; CLEA Japan, Tokyo, Japan) were harvested according to the ethical protocol approved by the RIKEN Center for Integrative Medical Sciences (2019-015(2)). Tissues were frozen immediately after dissection and stored at −80 °C until lipid extraction.

HEK293 cells were cultured in Dulbecco's modified Eagle medium (DMEM) (high glucose) with L-glutamine (Wako) containing 10% fetal bovine serum (FBS) and penicillin-streptomycin-L-glutamine (PSG), and incubated in 5% carbon dioxide (CO₂) at 37 °C. HEK293 cells (1.0×10⁷cells) were plated on a 10 cm dish and incubated with 5 or 10 µM of fatty acid dissolved in DMEM (high-glucose) with L-glutamine (Wako) containing no FBS and PSG for 1 h at 37 °C. After 1 h of cultivation, the medium was aspirated and DMEM (FBS⁺PSG⁺) and trypsin 0.25% with EDTA were added for cell collection, and the cells were centrifuged at 1200 rpm for 5 min at 4 °C. The supernatant was aspirated and phosphate-buffered saline (PBS) was added for cell counting. After repeating the same centrifugation process, the pellet cells were stored at −80 °C until lipid extraction.

Lipid extraction

Lipid extraction for human plasma was performed according to a previous report¹³. Briefly, an aliquot of 40 µL of NIST SRM 1950 plasma sample was added to 200 µL of ice-cold CHCl₃ and vortexed for 10 s. After 1 h of incubation on ice, 400 µL of ice-cold MeOH containing 10 µL of EquiSPLASH, 20 µM palmitic acid-d3, and 20 µM stearic acid-d3 were added and vortexed for 10 s. After 2 h incubation on ice, the solvent tube was centrifuged at 2000 × g for 10 min at 4 °C, and 400 µL of supernatant was transferred to LC/MS vials (Agilent Technologies).

Lipid extraction for cultured cells and mouse tissues was performed using the same mixed solvent protocol (MeOH:CHCl₃:H₂O, 2:1:0.2, v/v/v) as described above. The details of the solvent volumes and internal standards used in this study are described in Supplementary Table 7. First, MeOH was added to dried cells in a tube, followed by sonication and vortexing for 10 s. After 120 min of incubation on ice, CHCl₃ was added and the mixture was vortexed for 10 s. After 60 min of incubation, Milli-Q water was added, vortexed for 10 s, and the tube was left to stand for 10 min. The cells were then centrifuged at 2000 × g for 10 min at 20°C, and 400 µL of supernatant was transferred to LC/MS vials. To prepare the mixture samples, 80 µL of sample liquid from the pair, alpha-LA and gamma-LA, ETA, ARA, DPA, and OsA, respectively, were mixed into one vial.

The tissues were homogenized using a multi-beads shocker (YASUI KIKAI, Japan) with a metal cone (YASUI KIKAI, Japan) at 1500 × g for 15s, and MeOH was added to the homogenate according to the tissue weight (Supplementary Table 8). After the solvent was homogenized again under the same conditions, an appropriate amount of MeOH (30 mg tissue weight per 200 µL) was transferred to a 2-mL glass tube. After the solvent was removed up to 95% of the total volume of each sample, an appropriate amount of internal standards and CHCl₃ (Supplementary Table 8) were added to the solvent and vortexed for 10 s. After the solvent was incubated for 1 h on ice, 20 µL of Milli-Q water was added, and the mixture was vortexed for 10 s. After 10 min incubation on ice, the solvent was centrifuged at 2000 × g for 10 min at 20°C, and the supernatant was transferred to an LC-MS vial.

LC-CID-MS/MS and LC-OAD-MS/MS

The LC system consisted of a Shimadzu UPLC system (Kyoto, Japan). Lipids were separated on an Acquity UPLC Peptide BEH C18 column (50 × 2.1 mm; 1.7 µm; Waters, Milford, MA, USA). The column was maintained at 45 °C at a flow rate of 0.3 mL/min. The mobile phases consisted of (A) 1:1:3 (v/v/v) ACN:MeOH: water with ammonium acetate (5 mM) and 10 nM EDTA, and (B) 100% IPA with ammonium acetate (5 mM) and 10 nM EDTA. A sample volume of 1−3 µL, which depended on biological samples, was applied to the injection (Supplementary Tables 6 and 7). The separation was conducted under the following gradient: 0 min 0% (B); 1 min 0% (B); 5 min 40% (B); 7.5 min 64% (B); 12 min 64% (B); 12.5 min 82.5% (B); 19 min 85% (B); 20 min 95% (B); 20.1 min 0% (B); and 25 min 0% (B). The temperature of the sample was maintained at 4 °C.

Mass spectrometric detection of lipids was performed using a quadrupole/time-of-flight mass spectrometer LCMS-9030 (Shimadzu, Kyoto, Japan). All analyses were performed in data-dependent acquisition (DDA) mode with a mass range of 70–1250 m/z. The parameters were MS1 accumulation time, 250 ms; CID-MS/MS accumulation time, 100 ms; OAD-MS/MS accumulation time, 500 ms; CID-MS/MS collision energy, 25 eV with an energy spread of 7 eV; OAD-MS/MS collision energy, 6 eV with an energy spread of 0 eV; nebulizer gas flow, 2 L/min; interface temperature, 300 °C(+)/300 °C(-); and interface voltage, +4.0(+)/–3.0 kV(-). The other DDA parameters were dependent on the product ion scan number, 10; intensity threshold for CID-MS/MS, 1000 a.u.; intensity threshold for OAD-MS/MS, 10,000 a.u.; and exclusion time of precursor ion, 0 s. Mass calibration was automatically performed using a calibration delivery system (CDS) with a sub-interface. The experimental setup for OAD-MS/MS is described in detail elsewhere⁴⁶. Briefly, O/OH was generated via the microwave discharge of water vapor. The generated O/OH was introduced into the collision cell (q2) of LCMS-9030. The quadrupole rods in q2 were heated to 150 °C to prevent surface oxidation of the electrodes due to O/OH.

Data Analysis

Raw MS Data Files (.lcd) were converted into mzML using the LabSolutions software. The mzML data were preprocessed using the MS-DIAL software program (http://prime.psc.riken.jp/) for peak picking, MS/MS assignment, annotation in CID-MS/MS, and peak alignment. The following parameters were set: (data collection) RT begin, 0 min; RT end, 18 min; mass range begin, 0 Da; mass range end, 2000 Da; MS1 tolerance, 0.01 Da; MS2 tolerance, 0.025 Da; (peak detection) minimum peak height, 300 amplitude; mass slice width, 0.1 Da; smoothing method, linear weighted moving average; smoothing level, 3 scan; minimum peak width, 5 scan; exclusion list, none; (Deconvolution parameters) sigma window value, 0.5; MS/MS abundance cut off, 0 amplitude; (identification) retention time tolerance, 2 min; MS1 accurate mass tolerance 0.01 Da; MS2 mass tolerance. 0.05 Da; identification score cut-off, 70%; (text file and post identification) RT tolerance, 0.3 min; mass tolerance, 0.01 Da; identification score cut-off, 70%; (alignment) RT tolerance, 0.05 min; MS1 tolerance, 0.015 Da. The mzML data of OAD-MS/MS were imported into MS-DIAL, where the MS-DIAL alignment results (.txt) were utilized as a text library to assign molecular species information to the OAD-MS/MS spectra of the same sample. The OAD-MS/MS spectra with molecular species information were exported in text format (.txt) from the MS-DIAL program, and the result was imported to the MS-RIDD software program, where the mass accuracy was set to 15 ppm and relative intensity thresholds for OAD diagnostic ions were set to 0.01, as%, and 0.005 % for Δ4 the sphingoid base, respectively. In this study, 14 fragment ions associated with a single double bond and related dehydrated ions were utilized as reference ions (Supplementary Table 2). The detailed annotation algorithm is summarized in Supplementary Fig. 7. All annotation results from MS-RIDD were confirmed and collected by human verification.

The source code of MS-RIDD was written in Python (version 3.7.3) using Pandas (version 0.24.2) for data curation, Matplotlib (version 3.0.3) for data visualization, Numpy (version 1.16.2) for statistical analysis, and openpyxl (version 2.6.1), and python-pptx (version 0.6.18) for the results. Hierarchical clustering analysis was performed using Scipy (version 1.5.3). The data visualization for CID-/OAD-MS/MS spectra, extracted ion chromatograms, pairwise plots of retention time and m/z, heatmaps, and bar charts were carried out using Matplotlib (version 3.0.3). The compound structure was drawn using MarvinSketch (version 20.20) from ChemAxon (https://www.chemaxon.com). The fragmentation pathway of OAD (Fig. 1a and Supplementary Fig. 2) was drawn using ChemDraw (version 19.1.1.21; https://perkinelmerinformatics.com/).

Nomenclature

Lipid classifications, structural descriptions, and shorthand notations used in this study followed the definitions of LIPID MAPS. Details are provided on our CompMS website (http://prime.psc.riken.jp/compms/msdial/lipidnomenclature.html). Briefly, the acyl chain moiety is characterized by carbon and double-bond numbers, for instance 20:4, representing 20 carbons and four double bonds in the acyl chain. The ether and vinyl ether linkages were described by O-18:0 and P-18:0, respectively. The N-acyl linkage was described by N-18:0, except for the ceramide species. Underlining (‘_’) is used to describe acyl chain compositions if sn1, sn2 and sn3 positional isomers are uncharacterized, whereas the virgule (‘/’) character is used if the acyl chain position is determined in a specific position, such as the N-acyl chain, ether chain, or acyl sugar. The description of C=C positions followed two manners, delta-description and n-description. For instance, the arachidonic acyl chain is described as 20:4(Δ5, Δ8, Δ11, Δ14) in the former manner, and 20:4(n-6, 9, 12, 15) in the latter. Stereo-isomer information (‘E/Z’) was assigned as 20:4(5Z, 8Z, 11Z, 14Z) and 20:4(n-6Z, 9Z, 12Z, 15Z), respectively. To facilitate the recognition of the C=C positions of acyl chains and sphingoid bases, n-description was applied to the description of acyl chains and delta-description was applied to the description of sphingoid bases.

Data Availability

MS data are available at the DropMet section of the RIKEN PRIMe (http://prime.psc.riken.jp/) via the index DM0040.

Code availability

The MS-RIDD software and representative test datasets are freely available at https://github.com/Metabolomics-CompMS/MS-RIDD.git).

Additional information

Supplementary information

The online version contains supplementary material available at https:~.

Acknowledgements

This work was supported by the JSPS Grant-in-Aid for Scientific Research on Innovative Areas “Lipo- Quality” (15H05897 and 15H05898 to M.A.), JSPS KAKENHI (18H02432, 18K19155 to Hiroshi T., 20H00495 to M.A.), RIKEN Junior Research Associate Program (H.U.), and the National Cancer Center Research and Development Fund (2020-A-9 to Hiroshi T.).

Author contributions

H.T. and M.A. designed the study. H.U. developed the MS-RIDD software program and prepared the biological samples and authentic lipid standards. Hidenori performed LC-CID-MS/MS and LC-OAD-MS/MS analyses. H.U. analyzed the MS data, created a true positive dataset, and evaluated the MS-RIDD program. H.U., H.T., and M.A. wrote the manuscript. All authors have thoroughly discussed this project and helped improve the manuscript.

Competing interest

Hidenori T., who performed LC-MS/MS analysis, was an appreciation specialist in Shimadzu Japan.

1. Harayama, T. & Riezman, H. Understanding the diversity of membrane lipid composition. Nat. Rev. Mol. Cell Biol. 19, 281–296 (2018).

2. Van Meer, G. & De Kroon, A. I. P. M. Lipid map of the mammalian cell. J. Cell Sci. 124, 5–8 (2011).

3. Antonny, B., Vanni, S., Shindou, H. & Ferreira, T. From zero to six double bonds: Phospholipid unsaturation and organelle function. Trends Cell Biol. 25, 427–436 (2015).

4. Liebisch, G. et al. Update on LIPID MAPS classification, nomenclature, and shorthand notation for MS-derived lipid structures. J. Lipid Res. 61, 1539–1555 (2020).

5. Que, X. et al. Oxidized phospholipids are proinflammatory and proatherogenic in hypercholesterolaemic mice. Nature 558, 301–306 (2018).

6. Vriens, K. et al. Evidence for an alternative fatty acid desaturation pathway increasing cancer plasticity. Nature 566, 403–406 (2019).

7. Mori, A. et al. Parkinson’s disease-associated iPLA2-VIA/ PLA2G6 regulates neuronal functions and α-synuclein stability through membrane remodeling. Proceedings of the National Academy of Sciences 116, 20689–20699 (2019).

8. Rabionet, M. et al. Male meiotic cytokinesis requires ceramide synthase 3-dependent sphingolipids with unique membrane anchors. Hum. Mol. Genet. 24, 4792–4808 (2015).

9. Wenk, M. R. The emerging field of lipidomics. Nat. Rev. Drug Discov. 4, 594–610 (2005).

10. Nguyen, D. H., Nguyen, C. H. & Mamitsuka, H. Recent advances and prospects of computational methods for metabolite identification: a review with emphasis on machine learning approaches. Brief. Bioinform. 00, 1–16 (2018).

11. Tsugawa, H. et al. Comprehensive identification of sphingolipid species by in silico retention time and tandem mass spectral library. J. Cheminform. 9, (2017).

12. Koelmel, J. P. et al. LipidMatch: An automated workflow for rule-based lipid identification using untargeted high-resolution tandem mass spectrometry data. BMC Bioinformatics 18, (2017).

13. Tsugawa, H. et al. A lipidome atlas in MS-DIAL 4. Nat. Biotechnol. 38, 1159–1163 (2020).

14. Qian, H. et al. Structural basis for catalysis and substrate specificity of human ACAT1. Nature 581, 333–338 (2020).

15. Ishihara, T., Yoshida, M. & Arita, M. Omega-3 fatty acid-derived mediators that control inflammation and tissue homeostasis. Int. Immunol. 31, 559–567 (2019).

16. Harayama, T. & Shimizu, T. Roles of polyunsaturated fatty acids, from mediators to membranes. J. Lipid Res. 61, jlr.R120000800 (2020).

17. Martinez-Seara, H. et al. Effect of double bond position on lipid bilayer properties: Insight through atomistic simulations. J. Phys. Chem. B 111, 11162–11168 (2007).

18. Zhang, W. et al. Online photochemical derivatization enables comprehensive mass spectrometric analysis of unsaturated phospholipid isomers. Nat. Commun. 10, 1–9 (2019).

19. Cao, W. et al. Large-scale lipid analysis with C=C location and sn-position isomer resolving power. Nat. Commun. 11, 375 (2020).

20. Zhao, X., Wu, G., Zhang, W., Dong, M. & Xia, Y. Resolving Modifications on Sphingoid Base and N-Acyl Chain of Sphingomyelin Lipids in Complex Lipid Extracts. Anal. Chem. 92, 14775–14782 (2020).

21. Xie, X., Zhao, J., Lin, M., Zhang, J.-L. & Xia, Y. Profiling of Cholesteryl Esters by Coupling Charge-Tagging Paternò–Büchi Reaction and Liquid Chromatography–Mass Spectrometry. Anal. Chem. 92, 8487–8496 (2020).

22. Xie, X. & Xia, Y. Analysis of Conjugated Fatty Acid Isomers by the Paternò-Büchi Reaction and Trapped Ion Mobility Mass Spectrometry. Anal. Chem. 91, 7173–7180 (2019).

23. Feng, Y., Chen, B., Yu, Q. & Li, L. Identification of Double Bond Position Isomers in Unsaturated Lipids by m-CPBA Epoxidation and Mass Spectrometry Fragmentation. Anal. Chem. 91, 1791–1795 (2019).

24. Kuo, T.-H. et al. Deep Lipidomics and Molecular Imaging of Unsaturated Lipid Isomers: A Universal Strategy Initiated by mCPBA Epoxidation. Anal. Chem. 91, 11905–11915 (2019).

25. Williams, P. E., Klein, D. R., Greer, S. M. & Brodbelt, J. S. Pinpointing Double Bond and sn-Positions in Glycerophospholipids via Hybrid 193 nm Ultraviolet Photodissociation (UVPD) Mass Spectrometry. J. Am. Chem. Soc. 139, 15681–15690 (2017).

26. Ryan, E., Nguyen, C. Q. N., Shiea, C. & Reid, G. E. Detailed Structural Characterization of Sphingolipids via 193 nm Ultraviolet Photodissociation and Ultra High Resolution Tandem Mass Spectrometry. J. Am. Soc. Mass Spectrom. 28, 1406–1419 (2017).

27. West, H. & Reid, G. E. Hybrid 213 nm photodissociation of cationized Sterol lipid ions yield [M]+. Radical products for improved structural characterization using multistage tandem mass spectrometry. Anal. Chim. Acta 1141, 100–109 (2021).

28. Fang, M., Rustam, Y., Palmieri, M., Sieber, O. M. & Reid, G. E. Evaluation of ultraviolet photodissociation tandem mass spectrometry for the structural assignment of unsaturated fatty acid double bond positional isomers. Anal. Bioanal. Chem. 412, 2339–2351 (2020).

29. Thomas, M. C. et al. Ozone-induced dissociation: Elucidation of double bond position within mass-selected lipid ions. Anal. Chem. 80, 303–311 (2008).

30. Marshall, D. L. et al. Mapping Unsaturation in Human Plasma Lipids by Data-Independent Ozone-Induced Dissociation. J. Am. Soc. Mass Spectrom. 30, 1621–1630 (2019).

31. Barrientos, R. C., Vu, N. & Zhang, Q. Structural Analysis of Unsaturated Glycosphingolipids Using Shotgun Ozone-Induced Dissociation Mass Spectrometry. J. Am. Soc. Mass Spectrom. 28, 2330–2343 (2017).

32. Marshall, D. L. et al. Sequential Collision- and Ozone-Induced Dissociation Enables Assignment of Relative Acyl Chain Position in Triacylglycerols. Anal. Chem. 88, 2685–2692 (2016).

33. Pham, H. T., Maccarone, A. T., Campbell, J. L., Mitchell, T. W. & Blanksby, S. J. Ozone-induced dissociation of conjugated lipids reveals significant reaction rate enhancements and characteristic odd-electron product ions. J. Am. Soc. Mass Spectrom. 24, 286–296 (2013).

34. Campbell, J. L. & Baba, T. Near-complete structural characterization of phosphatidylcholines using electron impact excitation of ions from organics. Anal. Chem. 87, 5837–5845 (2015).

35. Baba, T., Campbell, J. L., Blanc, J. C. Y. L. & Baker, P. R. S. In-depth sphingomyelin characterization using electron impact excitation of ions from organics and mass spectrometry. J. Lipid Res. 57, 858–867 (2016).

36. Baba, T., Campbell, J. L., Le Blanc, J. C. Y. & Baker, P. R. S. Structural identification of triacylglycerol isomers using electron impact excitation of ions from organics (EIEIO). J. Lipid Res. 57, 2015–2027 (2016).

37. Baba, T., Larry Campbell, J., Yves Le Blanc, J. C., Baker, P. R. S. & Ikeda, K. Quantitative structural multiclass lipidomics using differential mobility: electron impact excitation of ions from organics (EIEIO) mass spectrometry. J. Lipid Res. 59, 910–919 (2018).

38. Bednařík, A., Bölsker, S., Soltwisch, J. & Dreisewerd, K. An On-Tissue Paternò-Büchi Reaction for Localization of Carbon-Carbon Double Bonds in Phospholipids and Glycolipids by Matrix-Assisted Laser-Desorption-Ionization Mass-Spectrometry Imaging. Angew. Chem. Int. Ed Engl. 57, 12092–12096 (2018).

39. Klein, D. R. et al. Desorption Electrospray Ionization Coupled with Ultraviolet Photodissociation for Characterization of Phospholipid Isomers in Tissue Sections. Anal. Chem. 90, 10100–10104 (2018).

40. Bednařík, A. et al. Ozonization of Tissue Sections for MALDI MS Imaging of Carbon-Carbon Double Bond Positional Isomers of Phospholipids. Anal. Chem. 92, 6245–6250 (2020).

41. Wäldchen, F., Spengler, B. & Heiles, S. Reactive Matrix-Assisted Laser Desorption/Ionization Mass Spectrometry Imaging Using an Intrinsically Photoreactive Paternò-Büchi Matrix for Double-Bond Localization in Isomeric Phospholipids. J. Am. Chem. Soc. 141, 11816–11820 (2019).

42. Paine, M. R. L. et al. Mass Spectrometry Imaging with Isomeric Resolution Enabled by Ozone-Induced Dissociation. Angew. Chem. Int. Ed Engl. 57, 10530–10534 (2018).

43. Korf, A., Jeck, V., Schmid, R., Helmer, P. O. & Hayen, H. Lipid Species Annotation at Double Bond Position Level with Custom Databases by Extension of the MZmine 2 Open-Source Software Package. Anal. Chem. 91, 5098–5105 (2019).

44. Takahashi, H. et al. Structural Analysis of Phospholipid Using Hydrogen Abstraction Dissociation and Oxygen Attachment Dissociation in Tandem Mass Spectrometry. Anal. Chem. 90, 7230–7238 (2018).

45. Kozakai, R. et al. Acyltransferase that catalyses the condensation of polyketide and peptide moieties of goadvionin hybrid lipopeptides. Nat. Chem. 12, 869–877 (2020).

46. Takahashi, H. et al. Identifying double bond positions in phospholipids using liquid chromatography-triple quadrupole tandem mass spectrometry based on oxygen attachment dissociation. Mass Spectrom. 8, 1–6 (2020).

47. Yen, C.-Y., Houel, S., Ahn, N. G. & Old, W. M. Spectrum-to-spectrum searching using a proteome-wide spectral library. Mol. Cell. Proteomics 10, M111.007666 (2011).

48. Aoyagi, R., Ikeda, K., Isobe, Y. & Arita, M. Comprehensive analyses of oxidized phospholipids using a measured MS/MS spectra library. J. Lipid Res. 58, 2229–2237 (2017).

49. Blevins, M. S. et al. Unsaturation Elements and Other Modifications of Phospholipids in Bacteria: New Insight from Ultraviolet Photodissociation Mass Spectrometry. Anal. Chem. 92, 9146–9155 (2020).

50. Irie, A., Kubo, H. & Hoshi, M. Glucosylceramide having a novel tri-unsaturated long-chain base from the spermatozoa of the starfish, Asterias amurensis. J. Biochem. 107, 578–586 (1990).

51. Ohashi, Y. et al. Squid nerve sphingomyelin containing an unusual sphingoid base. J. Lipid Res. 41, 1118–1124 (2000).

52. Cha, H. J. et al. Intercellular and intracellular functions of ceramides and their metabolites in skin (Review). Int. J. Mol. Med. 38, 16–22 (2016).

53. Hirabayashi, T. et al. PNPLA1 has a crucial role in skin barrier function by directing acylceramide biosynthesis. Nat. Commun. 8, 14609 (2017).

54. Ohno, Y., Kamiyama, N., Nakamichi, S. & Kihara, A. PNPLA1 is a transacylase essential for the generation of the skin barrier lipid ω-O-acylceramide. Nat. Commun. 8, 14610 (2017).

55. Oresti, G. M., García-López, J., Aveldanõ, M. I. & Del Mazo, J. Cell-type-specific regulation of genes involved in testicular lipid metabolism: Fatty acid-binding proteins, diacylglycerol acyltransferases, and perilipin 2. Reproduction 146, 471–480 (2013).

56. Casado, M. E. et al. Hormone-sensitive lipase deficiency disturbs the fatty acid composition of mouse testis. Prostaglandins Leukot. Essent. Fatty Acids 88, 227–233 (2013).

57. Jojima, K., Edagawa, M., Sawai, M., Ohno, Y. & Kihara, A. Biosynthesis of the anti-lipid-microdomain sphingoid base 4,14-sphingadiene by the ceramide desaturase FADS3. FASEB J. 34, 3318–3335 (2020).

58. Karsai, G. et al. FADS3 is a Δ14Z sphingoid base desaturase that contributes to gender differences in the human plasma sphingolipidome. J. Biol. Chem. 295, 1889–1897 (2020).

59. Mashima, R., Okuyama, T. & Ohira, M. Biosynthesis of long chain base in sphingolipids in animals, plants and fungi. Future Science OA 6, (2019).

60. Adem, A. A. et al. Structural characterization of plant glucosylceramides and the corresponding ceramides by UHPLC-LTQ-Orbitrap mass spectrometry. J. Pharm. Biomed. Anal. 192, 113677 (2021).

61. Aveldano, M. I. Phospholipid species containing long and very long polyenoic fatty acids remain with rhodopsin after hexane extraction of photoreceptor membranes. Biochemistry 27, 1229–1239 (1988).

62. Do, K. V. et al. Elovanoids counteract oligomeric β-amyloid-induced gene expression and protect photoreceptors. Proc. Natl. Acad. Sci. U. S. A. 116, 24317–24325 (2019).

63. Bhattacharjee, S. et al. Elovanoids are a novel class of homeostatic lipid mediators that protect neural cell integrity upon injury. Sci Adv 3, e1700735 (2017).

64. Psychogios, N. et al. The human serum metabolome. PLoS One 6, e16957 (2011).

65. Wishart, D. S. et al. HMDB 4.0: The human metabolome database for 2018. Nucleic Acids Res. 46, D608–D617 (2018).

66. Aveldaño, M. I., Rotstein, N. P. & Vermouth, N. T. Occurrence of long and very long polyenoic fatty acids of the n-9 series in rat spermatozoa. Lipids 27, 676–680 (1992).

67. Oresti, G. M., Luquez, J. M., Furland, N. E. & Aveldaño, M. I. Uneven distribution of ceramides, sphingomyelins and glycerophospholipids between heads and tails of rat spermatozoa. Lipids 46, 1081–1090 (2011).

68. Asakawa, D., Takahashi, H., Sekiya, S., Iwamoto, S. & Tanaka, K. Sequencing of Sulfopeptides Using Negative-Ion Tandem Mass Spectrometry with Hydrogen Attachment/Abstraction Dissociation. Anal. Chem. 91, 10549–10556 (2019).

69. Nothias, L.-F. et al. Feature-based molecular networking in the GNPS analysis environment. Nat. Methods 17, 905–908 (2020).

Yes there is potential Competing Interest. Hidenori Takahashi, who performed LC-MS/MS analysis, was an appreciation specialist in Shimadzu Japan.

UchinoetalOADCompMSSupplementaryInformationv6.docx
Supplementary Information
SupplementaryFigure01.pdf
Supplementary Figure 1
SupplementaryFigure02.pdf
Supplementary Figure 2
SupplementaryFigure03.pdf
Supplementary Figure 3
SupplementaryFigure04.pdf
Supplementary Figure 4
SupplementaryFigure05.pdf
Supplementary Figure 5
SupplementaryFigure06.pdf
Supplementary Figure 6
SupplementaryFigure07.pdf
Supplementary Figure 7
SupplementaryTable0108.xlsx
Supplementary Tables
SupplementaryData1.zip
Supplementary Data 1
SupplementaryData2.zip
Supplementary Data 2

Download PDF

Journal Publication

published 19 Dec, 2022

Read the published version in Communications Chemistry →

Version 1

posted

You are reading this latest preprint version

Computational mass spectrometry accelerates C=C position-resolved untargeted lipidomics using oxygen attachment dissociation

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Results

Discussion

Methods

Declarations

References

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1