LPS-induced lipid alterations in microglia revealed by MALDI mass spectrometry-based cell fingerprinting in neuroinflammation studies

Pathological microglia activation can promote neuroinflammation in many neurodegenerative diseases, and it has therefore emerged as a potential therapeutic target. Increasing evidence suggests alterations in lipid metabolism as modulators and indicators in microglia activation and its effector functions. Yet, how lipid dynamics in activated microglia is affected by inflammatory stimuli demands additional investigation to allow development of more effective therapies. Here, we report an extensive matrix-assisted laser desorption/ionization (MALDI) mass spectrometry (MS) whole cell fingerprinting workflow to investigate inflammation-associated lipid patterns in SIM-A9 microglial cells. By combining a platform of three synergistic MALDI MS technologies we could detect substantial differences in lipid profiles of lipopolysaccharide (LPS)- stimulated and unstimulated microglia-like cells leading to the identification of 21 potential inflammation-associated lipid markers. LPS-induced lipids in SIM-A9 microglial cells include phosphatidylcholines, lysophosphatidylcholines (LysoPC), sphingolipids, diacylglycerols and triacylglycerols. Moreover, MALDI MS-based cell lipid fingerprinting of LPS-stimulated SIM-A9 microglial cells pre-treated with the non-selective histone deacetylase inhibitor suberoylanilide hydroxamic acid revealed specific modulation of LPS-induced-glycerolipids and LysoPC(18:0) with a significant reduction of microglial inflammation response. Our study introduces MALDI MS as a complementary technology for fast and label-free investigation of stimulus-dependent changes in lipid patterns and their modulation by pharmaceutical agents.


Scientific Reports
| (2022) 12:2908 | https://doi.org/10.1038/s41598-022-06894-1 www.nature.com/scientificreports/ phospholipase activity (e.g. phospholipase A 2 ; PLA 2 ) with concomitantly increased synthesis of lipid mediators associated with microglia inflammatory response 18,19 has also been reported. These data indicate a role of lipid metabolism in microglia activation and response, yet, our understanding of how activated microglia orchestrates lipid metabolism to mount a protective or detrimental response remains to be clarified. Advances in the lipidomics field have allowed the investigation of physiological and pathological functions of lipids in vitro and in vivo 20 . As an emerging technology for the study of lipid metabolism in neurodegenerative states 8 , lipidomics approaches offer the possibility to track pathophysiological relevant molecules in complex cell and tissue extracts. However, lipidomics studies usually rely on the combination of liquid chromatography to mass spectrometry (MS) which requires additional extraction steps, labeling and long analysis time. Recently, we introduced fast, automatable, robust and label-free matrix-assisted laser desorption/ionization (MALDI) MS cell fingerprinting workflows 21,22 for cellular assays capable of characterizing drug on-and off-target responses with high-throughput screening capability and minimal sample preparation.
In this study, we extended this approach to microglia research and combined it with MALDI Magnetic Resonance MS (MRMS) and Trapped Ion-Mobility (TIMS) MS to simultaneously map LPS-induced lipid pattern alterations in microglia-like cells and to identify potential lipid markers of microglial activation. Our data indicate that increased levels of sphingomyelin [SM 34:1 + K] + , the ceramides (Cer) [Cer d42:1 + H-H 2 O] + and [Hex-Cer d34:1 + Na] + , the protonated isoforms and alkali adducts of C14-, C16-and C18 lysophosphatidylcholines (LysoPC) as well as the protonated isoforms and alkali adducts of C30-, C32-and C34 phosphatidylcholines (PC), three diacylglycerols (DG) and two triacylglycerols (TG) are present in LPS-activated SIM-A9 microglia cell line.
In addition, to test applicability of our workflow, when cells were pre-treated with suberoylanilide hydroxamic acid (SAHA), a known non-selective histone deacetylase (HDAC) inhibitor with reported anti-inflammatory activity 23,24 , we observed inhibition of LPS-induced IL-6 and TNF-α release as well as a specific modulation of glycerolipids (DG and TG) and LysoPC (18:0). All together, we present a proof-of-concept study for the application of whole cell MALDI MS lipid fingerprinting to the discovery of inflammation-associated lipid biomarkers in microglia and to evaluate their modulation by anti-inflammatory agents. Further understanding of the role of the microglia inflammation-associated lipid markers may provide insight into the interplay between microglia activation and lipid metabolism that could support new therapeutic options in neuroinflammatory diseases.

MALDI-TOF MS-based cell fingerprinting detects changes in LPS-stimulated SIM-A9 microglia lipid patterns.
Whole cell MALDI MS has been successfully applied to monitor biomolecular changes in mammalian cells allowing the profiling of overall content of lipids/metabolites in a label-free manner providing a snapshot of the physiological and pathological status of a biological system 21,[25][26][27][28][29] . However, it has not yet been used for microglia studies. Therefore, in this study, we optimized our previously stablished MALDI-TOF MSbased cell fingerprinting workflow 21,30 combining it to a platform of three synergistic MALDI MS technologies to interrogate lipid patterns in activated microglia and extract potential metabolic markers of inflammation.
The workflow is simple, robust and reproducible starting with treatment and freezing of microglia cells that are then automatically spotted on MALDI target plates and measured with a MALDI-TOF mass spectrometer (Fig. 1). For the extraction of highly significant differential mass-to-charge ratio (m/z) features, processing of mass spectra includes a step of multiple-testing correction of final peak lists 31 . Significantly altered m/z features are then investigated by high-resolution MALDI Magnetic Resonance MS (MRMS), Trapped Ion-Mobility (TIMS) MS and tandem MS (MS/MS) for identification. The use of automation for sample preparation, MALDI MS measurements and data analysis in this workflow offered us highly defined experimental parameters to overcome both biological and experimental variances. The average Pearson's correlation coefficient of r = 0.92 ± 0.05 and r = 0.89 ± 0.06 for the replicates in this study (vehicle (VEH)-treated cells) in positive and negative ion modes, respectively, indicate a high degree of repeatability of the MS fingerprints (Suppl . Table ST1 and ST2). Finally, accurate mass and fragmentation MS/MS data used in combination with collisional cross sections (CCS) values offered increased confidence in target molecules identification.
To investigate lipid patterns associated with microglia activation, we used mouse SIM-A9 microglial cells stimulated with LPS as a model system 32 . LPS is a chemical stimulant that strongly activates microglial cells triggering rapid inflammatory responses through the TLR4 pathway promoting the release of several proinflammatory cytokines such as IL-6 and TNF-α 5 . Initially, to control the capacity of SIM-A9 microglial cells to mimic an effective pro-inflammatory LPS-response, we evaluated IL-6 and TNF-α release in the supernatants of LPS-stimulated cells by ELISA. Following the treatment with 2.5, 10 or 100 ng/ml of LPS for 18 h, we observed a strong (P < 0.0001) concentration-dependent increase in IL-6 and TNF-α release compared to VEH-control treatment corresponding to an effective activation of SIM-A9 microglial cells ( Fig. 2a and b).
We next performed MALDI-TOF MS-based cell fingerprinting to track LPS-induced changes in SIM-A9 microglial cells lipid patterns. A total of 298 and 93 m/z features in positive and negative ion mode, respectively, were detected and volcano plots highlighted m/z features with significantly different ion intensities (TIC normalized; signal-to-noise ratio (SNR) > 3) between the VEH-treated and each LPS-treated group (Fig. 3a-c; Suppl. Fig. S1). LPS-responsive m/z features with adjusted P values ≤ 0.01 were only detected in positive ion mode. In total, we found 51 significantly altered m/z features. 38, 45 and 39 such m/z features were extracted for LPS concentrations of 2.5, 10 and 100 ng/ml, respectively (Suppl . Table ST3). Of these features, 29 (56.8%) were shared between all LPS-treated groups with increased intensities compared to unstimulated microglial cells (Fig. 3d). Among the features with greatest alterations, m/z 772.55 showed the highest change for LPS 2.5 ng/ ml (Padj < 0.001; Fig. 3e and Table 1) and m/z 524.38 for LPS concentrations 10 and 100 ng/ml (Padj < 0.0001; Fig. 3f and Suppl. Table ST3). Since cells were serum starved for 6 h prior to LPS stimulation and remained without serum for the next 18 h as well as due to the use of a vehicle control these differences are not likely an  www.nature.com/scientificreports/ effect of serum starvation or due to serum contents (e.g. exogenous metabolites). Instead, they could be rendered as prominent candidate inflammation-associated markers. Viability of SIM-A9 cells upon LPS treatment was reduced to 74% and 67% following LPS treatment with 10 and LPS 100 ng/ml, respectively. Therefore, we considered SIM-A9 microglial cells treated with LPS 2.5 ng/ml (viability 85%) for subsequent analysis and experiments (Suppl. Fig. S2). Overall, our data indicate that LPS stimulation induce significant changes in SIM-A9 microglial cell lipid patterns that can be rapid detected, and in a label-free manner, by MALDI-TOF MS cell lipid fingerprinting allowing the extraction of putative markers of inflammation. www.nature.com/scientificreports/ Identification of SIM-A9 microglia inflammation lipid-associated markers. Next, to characterize the prominent candidate markers molecules, we first remeasured the m/z signals using an Ultra-high resolution SolariX FTICR MS. In total, the identity of 21 out of the 29 potential inflammation-associated markers could be assigned based on stringent accurate mass determination (∆m/z < 1 ppm) ( Table 1). The remaining unassigned m/z features were either isotopic peaks or unavailable in the database (Suppl . Table ST3). 12 out of the 21 m/z species were later confirmed by high resolution MS/MS experiments with fragmentation patterns (Fig. 4e and Suppl. Figs. S3-S7, S9 and S11-S16) corresponding to structure-specific characteristic ions compared to the literature 33   www.nature.com/scientificreports/ deviation < 1%) with those predicted and previously reported in CCS databases and the literature where available (Suppl . Table ST4) [34][35][36][37][38] .
Representative MALDI-TOF MS, high-resolution MS spectrum and timsTOFfleX spectra with the tandem MS analysis are provided for m/z 772.55 in Fig. 4. The mean MALDI-TOF MS spectra for VEH and LPS-treated microglial cells shows the upregulation of the ion at m/z 772.55 with a neighboring unresolved peak (Fig. 4a). High-resolution FTICR MS allowed to distinguish both m/z values and assign the mass 772.5253 as [PC 32:0 + K] + with high accuracy (< 1 ppm error) (Fig. 4b). Additionally, the extracted ion mobility with 1/K 0 = 1.407 ± 0.01 (CCS = 286.8 Å 2 ) revealed the occurrence of one feature under for m/z 772.525 (Figs. 4c and d). Collision-induced dissociation (CID) of the precursor ion m/z 772.525 in both FTICR-MS and timsTOFflex-MS (Fig. 4e) (Table 1). Together, these lipids represent a potential panel of discriminant metabolic markers in activated microglia that should be considered in further validation studies, to increase mechanistic understanding of therapeutics and to be monitored in early drug development.
SAHA inhibit cytokines secretion and specifically modulate LPS-stimulated SIM-A9 microglia glycerolipids response and LysoPC(18:0). Once the automated MALDI MS cell-based fingerprinting setup was established and inflammation-associated lipid markers were identified, we next sought to test the general applicability of our workflow by investigating the therapeutic potential of the broad spectrum HDAC inhibitor SAHA for the modulation of LPS-dependent lipid inflammation responses in microglial cells (Fig. 5). Emerging data have shown the beneficial effects of HDAC inhibitors in inflammation-related diseases [39][40][41][42][43] and recent research supports the interplay between lipid metabolism and lysine acetylation during inflammation responses 20,44,45 with acetyl coenzyme A (acetyl-CoA), a key intermediary metabolite in lipid metabolism, being the main substrate for lysine acetylation by histone acetyltransferases (HATs) 45,46 . On the contrary, HDAC-containing protein complexes catalyze the removal of N-acetyl groups from lysine residues 47 which activity and expression is increased during microglial activation and particularly important for TLR-activated inflammatory responses 23,48 . Therefore, we hypothesized that anti-inflammatory action of SAHA might also involve changes in lipid metabolism.
We found that 1 h pre-treatment with SAHA (1 µM) completely abrogated the LPS-induced IL-6 release and significantly reduced LPS-induced TNF-α levels in SIM-A9 microglial cells compared to VEH-treated cells (Figs. 5a and b) Fig. 5h). Cell viability was confirmed by MTT assay (Suppl. Fig. S18). Altogether, these results demonstrated that SAHA significantly lowered the LPS-induced inflammatory cytokine response in SIM-A9 microglial cells which was accompanied by specific modulation of some inflammation-associated lipid markers indicating glycerolipids metabolism as possible off-target pathway for the SAHA anti-inflammatory effect.

Discussion
In this work, we present a multiplatform MALDI MS-based cell lipid fingerprinting approach to explore inflammation-associated lipid markers in LPS-stimulated microglia-like cells. To the best of our knowledge, no study has been reported on the use of whole cell MALDI MS to interrogate lipid patterns in activated microglia. Together, the current findings reveal that microglia inflammation-evoked activation of TLR4 leads to significant changes in the cell's lipid pattern with alterations in major lipid pathways such as the sphingolipid (SM and Cer species), glycerophospholipid (LysoPC and PC species) and glycerolipid (DG and TG species) metabolism pathways corroborating previous microglia lipidomics studies 12,17 .
LPS treatment is the most employed activation strategy in microglia studies with well understood signaling processes inducing transcription and expression of proinflammatory and neurotoxic genes 5,48 . Commonly used methods to monitor microglia activation are the determination of changes in levels of activation markers by ELISA, fluorescence assays, western blots and quantitative RT-PCR 49 . However, microglia activation also leads to metabolic reprogramming and changes in lipid homeostasis 10,12 . Therefore, our MALDI MS-based cell lipid fingerprinting strategy represents a further potential tool to investigate microglia activation and detect inflammation-associated metabolic alterations. A key feature of our untargeted workflow is that we initially perform an unbiased exploration of m/z features from LPS-stimulated cells which allow the rapid detection and extraction of a panel of highly significant differential lipid markers. The combination with high-resolution MS and tandem MS supports the identification of such markers, which can then be further validated and used to probe multiple targets of small-molecule inhibitors of microglial cell activation. www.nature.com/scientificreports/ Exacerbated brain inflammatory responses are present in many neurodegenerative diseases 3 . With the current knowledge on lipids as important mediators of immune responses 20,50 , differentially expressed lipids in activated microglia can be potentially used as markers for neuroinflammation. It has been demonstrated the involvement of LysoPCs, PCs, Ceramides and Sphingomyelins in microglial inflammatory responses [51][52][53] . Particularly in AD, characteristic lipids, such as LysoPC(16:0), LysoPC(18:0), LysoPC(18:1) and Ceramides, were found associated with beta-amyloid plaques 54 . In like manner, these LysoPCs were also found up-regulated in a PD mouse model brain 55 . Accordingly, we found increased levels of LysoPC(14:0), LysoPC(16:0) and LysoPC(18:0) as well as Ceramides in activated microglia-like cells. As a response to an inflammatory stimulus, pro-inflammatory cytokines can activate PLA 2 and Phosphatidylcholine-specific phospholipase C -mediated processes as well as Sphingomyelinases to hydrolyze membrane PCs and SMs generating LysoPCs and Cer, respectively 20,56 . Increased www.nature.com/scientificreports/ PLA 2 activity in microglial cells upon LPS-treatment, which may be related to higher levels of LysoPCs, have also been demonstrated 18,57 . Remarkably, LysoPCs and Ceramides are well known cell lipid mediators that act as second messengers having both pro-survival and death roles triggering alterations in mitochondrial function, oxidative stress and gene transcription with a central role in inflammation signaling 11,14,51,53,58,59 . Our whole-cell MALDI MS approach was also capable of detecting LPS-induced alterations in glycerolipids (DG and TG species). Accordingly, accumulation of DG was demonstrated in AD and PD brains [60][61][62] , in activated murine macrophages 63 and in LPS-stimulated BV2 cells 12,17 . DGs are minor components of the lipid bilayer that can also act as glycerophospholipid-derived lipid mediators in response to immune activation 64 . As a second messenger, it regulates activation of many protein kinases (PK), such as PKC and DG kinases (DGK), directly modulating nuclear signal transduction 64,65 . Ceramide signal may also be converted to one that is mediated by DG through the action of sphingomyelin synthases that breaks PC species releasing DG while adding a PC headgroup to ceramide to form SM 56 . Besides, DG in the lipid biosynthesis serve as precursors for the major glycerophospholipids, like PC, as well as for TG synthesis 66 . We also found increased levels of two TGs [TG(48:0) and TG(50:0)] in LPS-activated murine SIM-A9 microglial cells. This finding is consistent with recent reports that demonstrated higher levels of TG(48:0) and TG(50:0) and accumulation of TG-enriched lipid droplets in LPS-treated microglial cells 12,17 . TG is the main component of lipid droplets involved in energy storage and in the sequestration of FAs to reduce their availability for the activation of several signaling pathways (e.g. TLR and NF-κB inflammatory pathways, activation of protein phosphatase 2A (PP2A) by ceramides or of PKC by DG) protecting cells from injury as well as in storing lipid precursors to regulate signaling pathways 50,67 . Moreover, the metabolic reprogramming of activated microglia involves changes in mitochondrial function and in the tricarboxylic acid (TCA) cycle that can result in increased production of TG within lipid droplets 16,17,68 . In this way, our observations further corroborate an involvement of glycerolipids in TLR4-dependent inflammation response in microglia-like cells and also support the use of our MALDI MS-based cell lipid fingerprinting approach to detect alterations in activated microglia lipid metabolism.
Here, we also showed that modulation of acetylation by blocking HDACs activity with SAHA profoundly impaired LPS-induced pro-inflammatory cytokines in SIM-A9 microglial cells and partially rescued LPS-induced glycerolipid levels, as detected by our MALDI-TOF MS-based cell lipid fingerprinting. Lysine acetylation regulates the activity of intermediate metabolism enzymes of the FA pathways and of the TCA cycle 69 with emerging evidences on inflammation-associated metabolic diseases and aging supporting a multilayered lipid-epigenetic interplay [70][71][72] . Our findings are consistent with reports showing a decrease of LPS-induced cytokine expression after SAHA treatment in microglial cells, mixed glial cultures and macrophages 23,42,43,73 as well as with studies of lipid accumulation inhibition by SAHA in hepatic cells 74 and in cultured adipocytes 75 and by MS-275 (class I-specific HDAC inhibitor) in human macrophages 39 . A reduction in DG levels can sign a shift to a beneficial metabolic state since its generation is associated to inflammatory response as previously reported 61,64 . Additionally, the reduction in TGs may indicate an attempt of the cells to restore energy metabolism and cell membrane synthesis along with regulation of inflammatory signaling pathways through release of FAs by TG lipolysis from LPS-induced lipid droplets 9,67,68 , an aspect that requires additional investigation. Another noteworthy finding from the current study was the specific effect of SAHA on LPC(18:0) indicating involvement of this particular LysoPC with the anti-inflammatory effects observed which may indicate regulation of PLA 2 activity 18,57 and/or of LPC acyltransferase activity, an enzyme that converts LysoPCs back to PC in the presence of acetyl-CoA 76 . Nevertheless, follow-up studies are needed to verify these observations and validate these potential markers.
Overall, we present here a comprehensive MALDI MS-based cell lipid fingerprinting workflow for microglia research by showing the modulation of inflammatory lipid-associated markers in LPS-stimulated microglia-like cells. Our workflow also displays promise application for the investigation of lysine acetylation involvement in neuroinflammation through the use of HDAC tool inhibitors and bring to focus the cross-talk between lipid pathways, inflammation and lysine acetylation in microglial cells. Furthermore, our combined label-free automated MALDI MS-based cell fingerprinting approach is robust and reproducible with little sample preparation which can be scaled-up to enable high-throughput and support drug screening in neuroinflammation as well as contribute to the understanding of the role of lipids in microglia.

Materials and methods
Chemicals. All  www.nature.com/scientificreports/ SIM-A9 cells were seeded 0.2 × 10 6 per mL in 96-well plates and incubated for 24 h. Medium was changed to serum-free medium for 6 h before treatments. For LPS treatment without SAHA, cells were treated with PBS (vehicle; VEH) or stimulated with LPS diluted in PBS (2.5, 10 or 100 ng/ml) for 18 h. For LPS treatment in the presence of SAHA, cells were treated for 1 h with compound or DMSO prior to LPS-stimulation for additional 18 h (LPS 2.5 ng/ml). SAHA was dissolved in DMSO as stock solution and the final concentration of DMSO in the medium was kept under 0.1%. The inhibitor concentration was chosen based on known IC 50 and previous reports 24 . Culture supernatants were stored at − 80 °C for ELISAs. Cells were washed once with cold PBS and the 96-well plate was snap-frozen in liquid nitrogen and stored at -80 °C until use. Cytokine levels in culture medium collected from LPS-stimulated cells and/or cells treated with inhibitor with or without LPS were measured by ELISA according to the manufacturer's instructions. Absorbance was read at 450 nm on a spectrophotometer (Multiscan Spectrum, Thermo Fisher Scientific, Germany), and IL-6/TNF-α concentrations calculated based on calibration curves. Automated MALDI-TOF mass spectrometry. Automated preparation of cells on MALDI MS target plates (MTP) was done with a CyBio ® FeliX pipetting platform (Analytik Jena AG, Germany). 96-well plates with frozen cells were taken out of the − 80 °C freezer, and cells were immediately resuspended at ~ 2,000 cells per µl in ddH 2 O/ACN (70:30). 2 µl were applied onto ground steel MTP (Bruker Daltonics, Bremen, Germany). Four technical replicates of each plate well were prepared during transfer to a 384-well MTP. After air drying of sample spots, 20 mg/ml DHB-matrix in ACN/ddH 2 O (50:50) supplemented with 2.5% TFA was sprayed using an HTX M3 Sprayer (HTX Technologies Carrboro, USA). The spray protocol included a 50 °C spray nozzle temperature with 60 µl/min matrix flow rate. Four layers of matrix were sprayed with a spray-head velocity of 1000 mm/min and the distance between sprayed lines was 2 mm 21,22 .
MTPs were measured on a rapifleX MALDI-TOF MS (Bruker Daltonics, Bremen, Germany). The FlexControl 4.0 AutoXecute tool (Bruker Daltonics) was used to perform automated data acquisition. Measurements were done in reflector-positive and -negative ion modes in a mass range of m/z 300 − 1800. The acquisition mode was set to random walk, and 6000 laser shots were accumulated in 60 shot steps per spot using 10 kHz laser frequency. External calibration was performed using the TBLE calibration mixture spotted in close proximity to the sample spots. All spots from the same assay were measured on the same day. Samples originating from different passage numbers of cultured cells that were analyzed on different days are referred to as biological replicates. Additionally, full scan MS and MS/MS spectra were recorded on the timsTOFfleX mass spectrometer to assist with molecular identification 78,79 . Spectra were acquired in both qTOF and TIMS-on modes of operation in positive ion mode within the m/z range 100 -1000. The qTOF was calibrated using "quadratic enhanced"-fit and the TIMS dimension was calibrated linearly using selected ions from the ESI LC/MS tuning mix (m/z, CCS: 118.0862, 120.8; 322.0481, 152.8; 622.0289, 201.6 and 922.0097, 241.8) in positive mode. MALDI parameters were optimized to maximize intensity and resolution. The ion mobility was scanned from 0.6 to 1.70 Vs/cm 2 . Fragmentations were performed by CID with a collision energy ranging from 40 to 55 eV.

MALDI-FTICR and MALDI-timsTOF mass spectrometry. MALDI-Magnetic
Processing of mass spectra and data analysis. Initial manual recalibration of mass spectra was done in flexAnalysis 4.0 software (Bruker Daltonics, Bremen, Germany). Quadratic calibration was performed internally using the lysophosphatidylcholine LysoPC .0 software (Bruker Daltonics, Bremen, Germany) was used for spectra grouping, mean spectra visualization and peaks calculation with the following parameters: TIC normalization; Resolution of 10,000; TopHat baseline subtraction with 10% Minimal Baseline Width; Savitzky − Golay Spectra smoothing with 0.1 m/z width and 1 cycle; Null spectra and Noise spectra exclusion enabled. Peaks were picked on total average spectra with SNR thresholds of three (See supplementary methods for processing of negative ion mode mass spectra).
All data represent the mean ± s.d. unless indicated otherwise. Unpaired T-test was used to compare two groups and One-way ANOVA with Tukey's post hoc test for multiple comparisons. P-values ≤ 0.001, ≤ 0.01 and ≤ 0.05 were used to indicate highly statistically significant, very statistically significant and statistically significant, respectively. www.nature.com/scientificreports/ For FTICR and timsTOF MS data visualization and evaluation, the Bruker Compass DataAnalysis 5.3 (Bruker Daltonics, Bremen, Germany) was used. Lipid identifications were determined based on mass accuracy using the LIPIDMAPS lipidomics gateway [lipidmaps.org; 81 ] and on the fragmentation patterns observed in MS/MS spectra, when available 33

Data availability
There is currently no public data repository for MALDI-MS fingerprinting data. However, key raw data used in this paper are available at figshare: https:// doi. org/ 10. 6084/ m9. figsh are. 17205 908. Additional data are available from the corresponding author upon reasonable request.