Lipid features were first annotated by comparison with authentic standards and by MS/MS spectral matching against the LipidBlast spectral library. Matching was performed using precursor mass tolerances of 3.0 ppm or 3.0 mDa and fragment mass tolerances of 10.0 ppm or 10.0 mDa. Fragments with intensities below 1% of the precursor ion were discarded as potential noise. A minimum spectral match score threshold of 400 (on a scale of 0–1000) and a minimum match peak coverage of 25% were required for candidate annotations.
The dataset was further processed using the Buzatto Research Group’s LipidQuest routine (https://github.com/Buzattoresearch/BRG-LipidQuest). Features not annotated by MS/MS were matched to the LIPID MAPS database using accurate mass matching with a tolerance of 3 mDa. Orthogonal constraints were applied to all annotated lipids to reduce false-positive assignments that may arise when relying solely on accurate mass or spectral matching. These included retention-time filtering relative to annotated features of the same lipid class, Kendrick mass-defect filtering within lipid classes, empirically determined adduct preferences for each lipid class, and biochemical plausibility based on the biological matrix under study. We note, however, that tentative lipid annotations derived from mass or MS/MS matching remain putative and do not constitute definitive lipid identification.
Chromatographic separations were performed on a Thermo Vanquish UHPLC system using a Waters ACQUITY Premier CSH C18 column with VanGuard FIT pre-column (2.1 × 100 mm, 1.7 µm particle size). The column temperature was maintained at 45 °C and the autosampler temperature at 6 °C. The following gradient was applied: 0 min, 10% MPB, 0.350 mL/min; 0.7 min, 10% MPB, 0.350 mL/min; 3.5 min, 38% MPB, 0.350 mL/min; 9.5 min, 48% MPB, 0.350 mL/min; 10.6 min, 68% MPB, 0.350 mL/min; 14.0 min, 88% MPB, 0.320 mL/min; 14.7 min, 99% MPB, 0.300 mL/min; 15.4 min, 10% MPB, 0.270 mL/min; 16.3 min, 10% MPB, 0.350 mL/min (re-equilibrium); 18.5 min, 10% MPB, 0.350 mL/min. Each sample extract was injected once in positive-ion mode (5 µL, equivalent to 2.3 × 108 bacterial cells) and once in negative-ion mode (10 µL, equivalent to 4.6 × 108 cells) to maximize lipid class coverage. At least 3 blank injections preceded each sequence, and samples were bracketed by pooled QC injections to monitor instrument stability throughout the analytical sequence.
"],"publication":["Nutrient environment drives strain-specific remodelling of Escherichia coli lipidomes."],"submitter_name":["Adriana Zardini Buzatto"],"submitter_affiliation":["University of Calgary"],"organism_part":["cell pellet"],"technology_type":["mass spectrometry assay"],"disease":[""],"extraction_protocol":["Lipids were extracted immediately after pellet collection or quenching using a modified Folch method (2:1 dichloromethane/methanol, followed by addition of water to 8:4:3 dichloromethane/methanol/aqueous). Samples were mixed with 155 µL cold methanol, 320 µL cold dichloromethane, and 120 µL chilled water, with 20 s of vortexing between each solvent addition. The resulting biphasic extracts were equilibrated for 10 min at 4 °C and centrifuged at 13,000 × g for 10 min at 4 °C. An aliquot of the organic layer (200 µL) was evaporated to dryness using a SpeedVac for 45 min at 4ºC. The remaining organic extracts were combined to generate a pooled quality control (QC) sample. The QC pool was divided into 200 µL aliquots and evaporated to dryness. Dried extracts were briefly purged with nitrogen and stored at -80ºC protected from light for up to 7 days.
The dried extracts were resuspended in 4:6 mobile phase B (MPB: 10 mM ammonium formate in 96:3:1 isopropanol/acetonitrile/water) / mobile phase A (MPA: 10 mM ammonium formate in 2:2:1 methanol/acetonitrile/water) (v/v) to maintain lipid solubility while remaining compatible with the starting LC conditions. Each extract was first vortexed with 15.4 µL mobile phase B to ensure lipid solubilization, and diluted with 23.5 µL mobile phase A to reach 4.2 (quenching experiment) to 4.6 × 107 cells/µL (medium and strain comparison). Resuspended samples were centrifuged at 13,000 × g for 5 min at 4 °C. Supernatants were transferred to 2 mL amber injection vials fitted with 300 µL polypropylene inserts and sealed with PTFE/silicone caps. Samples were allowed to equilibrate at 4 °C for 4 h before analysis to ensure complete solvent equilibration, and were injected within 32 h of resuspension.
"],"organism":["Escherichia coli"],"full_dataset_link":["https://www.ebi.ac.uk/metabolights/MTBLS14565"],"author":["Adriana Zardini Buzatto. University of Calgary. 2500 University Drive NW, University of Calgary, Calgary, AB, Canada. adriana.zardinibuzat@ucalgary.ca."],"data_transformation_protocol":["Chromatographic data were processed using an in-house workflow comprising feature detection, alignment, filtering, lipid annotation, handling of missing values, and normalization. Processing was restricted to a retention time window of 0.75–16.2 min, excluding the initial void volume (0-0.75 min) and the column re-equilibration phase (16.2-20.0 min). After alignment, MS1 features and MS/MS precursor ions below 2.0 × E5 intensity (empirically defined as 10× above the chromatogram-wide baseline noise) or local signal-to-noise ratio (S/N) < 6 were excluded. Peaks detected in fewer than 85% of samples within at least one experimental group were removed. Further details are provided as Supplementary Materials.
Lipid features were first annotated by comparison with authentic standards and by MS/MS spectral matching against the LipidBlast spectral library. Matching was performed using precursor mass tolerances of 3.0 ppm or 3.0 mDa and fragment mass tolerances of 10.0 ppm or 10.0 mDa. Fragments with intensities below 1% of the precursor ion were discarded as potential noise. A minimum spectral match score threshold of 400 (on a scale of 0–1000) and a minimum match peak coverage of 25% were required for candidate annotations.
The dataset was further processed using the Buzatto Research Group’s LipidQuest routine (https://github.com/Buzattoresearch/BRG-LipidQuest). Missing values were imputed with small, biologically reasonable values derived from the dataset, preserving signal structure. Peak intensities were then normalized by internal standards (deuterated lipid standards from different classes; Avanti Splash Lipidomix Mass Spec Standard, Avanti Research). Internal standards were assigned to lipids based on class match or structural similarity (i.e., similar physicochemical properties and retention behaviour). The normalized intensity ratio was calculated as the feature intensity (peak height) divided by the internal standard intensity, and then multiplied by the internal standard concentration in the sample. All features within the same lipid class and ionization polarity used the same internal standard as the denominator to maintain comparability across samples. Finally, features exhibiting QC relative standard deviation (RSD) ≥30% after normalization were removed due to high variability. No further normalization strategies were applied to maintain the data structure, interpretability, and comparability.
Positive and negative ionization features were merged into a single feature list through row-wise concatenation. For annotated features, a single “best” polarity per lipid feature was selected when both polarities annotate the same underlying compound within a retention time tolerance of 12 s and a neutral mass tolerance of 5 ppm or 5 mDa, based on stronger annotation evidence (highest MS/MS score, lowest m/z error) and reproducibility (lowest RSD for QCs). The final annotated results are available as supplemental information. Non-annotated features were not used for statistical analysis and interpretation.
"],"study_factor":["Experiment","Strain"],"submitter_email":["adriana.zardinibuzat@ucalgary.ca"],"sample_collection_protocol":["We investigated the lipidomes of Escherichia coli strains ATCC 25922 (serotype O6, biotype 1), E2348/69 (serotype O127:H6) and EHEC 86/24 (serotype O157:H7). Stocks were generated by streaking strains from cryostocks onto tryptic soy agar plates and incubating for 24h at 37 °C. Individual colonies were transferred to liquid tryptic soy medium and cryopreserved as frozen glycerol stocks. A fresh stock was used for each experiment to minimize contamination and genetic drift. Cryostocks were not thawed before streaking and culturing to prevent loss of viability. Cells were cultured in 4 mL liquid medium at 37°C with shaking at 120 rpm. 15 mL Falcon tubes were loosely capped to allow gas exchange. Cultures were incubated under standardized conditions used across all experiments. Cell density was estimated by optical density at 600 nm (OD600).
Culture volumes corresponding to equivalent biomass (3 × E9 cells, based on an approximate conversion factor of 8 × E8 cells per OD600 unit) were collected for lipid extraction. The aliquots were centrifuged at 3,000 × g for 4 min to pellet the cells. The medium was removed, and the pellet was washed three times with 0.5 mL cold phosphate-buffered saline (PBS). We recognize that cold shock may lead to leakage of metabolites; however, we opted for cold PBS wash cycles to prevent interference from extracellular compounds. Nevertheless, this work focused on intracellular measurements of low-polarity compounds that are mostly insoluble in cold PBS.
"],"omics_type":["Metabolomics"],"study_design":["ultra-performance liquid chromatography-mass spectrometry","cell pellet","untargeted analysis","Thermo Scientific Vanquish Flex UHPLC System","Thermo Q Exactive HF Quadrupole-Orbitrap","Escherichia coli","Lipidomics","medium","data-dependent acquisition","quenching","experimental sample"],"curator_keywords":["ultra-performance liquid chromatography-mass spectrometry","cell pellet","untargeted analysis","Thermo Scientific Vanquish Flex UHPLC System","Thermo Q Exactive HF Quadrupole-Orbitrap","Escherichia coli","Lipidomics","medium","data-dependent acquisition","quenching","experimental sample"],"mass_spectrometry_protocol":["Mass spectrometry analyses were performed using a Thermo Q Exactive HF Quadrupole-Orbitrap mass spectrometer equipped with a heated electrospray ionization (HESI) source. Full MS scans were acquired at a resolving power of 240,000 over an m/z range of 140–2000. Data-dependent MS/MS spectra were acquired in dd-MS2 mode with a resolving power of 45,000.
Full MS scans were acquired with an AGC target of 3 × E6 and a maximum injection time of 200 ms. Data-dependent MS/MS spectra were acquired with AGC target of 1 × E5, maximum injection time of 80 ms, and topN of 4. Intensity thresholds were set to 1.3 × E6 for positive ionization and 8.8 × E5 for negative ionization. The fixed first mass was set to 140 m/z, the apex trigger to 2–5 s, and the isolation window to 1.5 m/z. Fragmentation was performed using stepped normalized collision energy (NCE) of 15, 25, and 35. Dynamic exclusion was applied for 5.0 s. The HESI source was operated at 4,000 V in positive-ion mode and 3,500 V in negative-ion mode. The capillary temperature was set to 220°C, the probe heater temperature to 180°C, and the S-lens RF level to 80. Sheath gas, auxiliary gas, and sweep gas were set to 25, 10, and 1 arbitrary units, respectively.
"],"additional_accession":[]},"is_claimable":false,"name":"Nutrient environment drives strain-specific remodelling of Escherichia coli lipidomes","description":"Background/Objectives: Escherichia coli causes more than 2.8 million infections annually and is widely used as a model organism. Despite its importance, strain-level and handling differences in E. coli lipidomes remain poorly defined. Lipid composition can reveal membrane organization, metabolic state, and strain-specific features, but measured lipidomes depend strongly on experimental design. This study aimed to characterize the lipidomes of reference and pathogenic E. coli strains and to determine how sample handling and growth conditions influence their interpretation. Methods: Three E. coli strains (ATCC 25922, O127:H6, and O157:H7) were grown in minimal (M9) and rich (tryptic soy) media to late stationary phase. Lipids were extracted from bacterial pellets using direct extraction or a methanol-based quenching strategy and analyzed by reversed-phase LC–MS/MS. Statistical analysis included uni- and multivariate testing, along with evaluation of coordinated changes across lipid classes. Results: Methanol-based quenching altered lipid composition, increasing lysophospholipids and phosphatidylalcohols and shifting quinone balance, without improving reproducibility. Direct extraction minimized variability and preserved lipid composition. The growth medium defined the lipidome structure, with tryptic soy broth increasing phospholipid abundance while minimal medium constrained composition and increased remodelling-associated species. Pathogenic strains exhibited distinct lipidomic profiles, including increased cardiolipin and phosphatidic acid, elevated LPE/PE ratios, reduced diacylglycerol levels, and depletion of prenol lipids. Conclusions: The observed E. coli lipidome is defined by experimental design, with sample handling and growth medium constraining the measurable lipid space within which strain-specific differences emerge. These findings establish a context for the design and interpretation of lipidomics experiments across bacterial strains.","dates":{"publication":"2026-09-23","submission":"2026-05-22"},"accession":"MTBLS14565","cross_references":{}}