Feb 27, 2026
Mass spectrometry (MS)-based metabolomics analyses for fecal samples
- Gaute Hovde Bø1,
- Veronika Pettersen1
- 1UiT The Arctic University of Norway
Protocol Citation: Gaute Hovde Bø, Veronika Pettersen 2026. Mass spectrometry (MS)-based metabolomics analyses for fecal samples. protocols.io https://dx.doi.org/10.17504/protocols.io.rm7vzeqp4vx1/v1
License: This is an open access protocol distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Protocol status: Working
We use this protocol and it's working
Created: February 27, 2026
Last Modified: February 27, 2026
Protocol Integer ID: 244150
Keywords: stool metabolomics, LC–MS/MS, targeted metabolomics, untargeted metabolomics, short-chain fatty acids (SCFAs), organic acids, Orbitrap, metabolite profiling, microbial metabolites, fecal metabolome, metabolomics analyses for fecal sample, fecal metabolomic profiling, untargeted metabolomic profiling, resolution profiling of global metabolomic change, targeted microbial metabolite, microbial metabolite, global metabolomic change, based metabolomics workflow, metabolomics workflow, infant gut by bifidobacteria, mass spectrometry, based probiotic, metabolite signal, fecal sample, metabolic, triple quadrupole mass spectrometerin, resolution orbitrap mass spectrometry with data, triple quadrupole mass spectrometerin multiple reaction, resolution orbitrap mass spectrometry, chain fatty acid, infant gut, metaboanalyst, fatty acid
Funders Acknowledgements:
Tromsø Research Foundation
Grant ID: (TFS18_CANS_AS-HVF
Disclaimer
DISCLAIMER – FOR INFORMATIONAL PURPOSES ONLY; USE AT YOUR OWN RISK
This protocol is provided for research use only. Users are responsible for ensuring compliance with applicable institutional, ethical, and safety regulations. The authors assume no liability for misuse of the procedures described.”
Abstract
This protocol describes the mass spectrometry–based metabolomics workflow used in the study “Metabolic reprogramming of the infant gut by bifidobacteria-based probiotics drives exclusion of antibiotic-resistant pathobionts” by Bargheet et al. (2026). Fecal metabolomic profiling was performed using complementary targeted and untargeted LC–MS/MS approaches. Samples collected in OMNImet•GUT tubes were processed by centrifugation and spin filtration, and metabolite signals were normalized to estimated dry weight. Targeted quantification of short-chain fatty acids and organic acids was performed after derivatization with 3-nitrophenylhydrazine using UPLC coupled to a triple quadrupole mass spectrometerin multiple reaction monitoring mode. Quantification was based on isotopically labeled internal standards and external calibration curves.
Untargeted metabolomic profiling was performed using hydrophilic interaction chromatography coupled to high-resolution Orbitrap mass spectrometry with data-dependent acquisition. Quality control procedures included pooled QC samples, internal standards, blanks, and drift monitoring throughout acquisition. Raw data were processed using MZmine for feature detection and alignment, followed by molecular annotation with SIRIUS and molecular networking in GNPS. Statistical analyses and normalization were performed in MetaboAnalyst. Together, this workflow enabled quantitative assessment of targeted microbial metabolites and high-resolution profiling of global metabolomic changes in fecal samples.
Materials
OMNImet•GUT, 10 kDa cutoff spin filter, MilliQ water, 3-nitrophenylhydrazine (3NPH), N-(3-dimethylaminopropyl)-N'-ethylcarbodiimide (EDC), pyridine, 2,6-di-tert-butyl-4-methylphenol (BHT), formic acid, acetonitrile (ACN), analytical standards for SCFAs and OAs, Sigma-Aldrich reagents, Cayman Chemicals reagents, Acetyl-l-carnitine-D3, Hexadecanoyl(palmitoyl)-l-carnitine-D3, l-Leucine-5,5,5-D3, l-Tryptophan-(indole-D5), l-Methionine-(methyl-D3), Stearic acid-18,18,18-D3, Chenodeoxycholic-2,2,3,4,4,6,6,7,8-D9 acid, 18:0-D35 Lyso PC, Dopamine-1,1,2,2-D4 hydrochloride, quercitin, amitriptylin, histidine, arginine, labetalol, doxepin, proline, tryptophan.
Troubleshooting
Before start
Prior to sample injection, a system blank consisting of clean sample solvent (ACN: MilliQ, 90:10, v/v) was injected to ensure that the instrument and mobile phase were free from contamination. Next, several injections of a system control sample consisting of 0.1 mg/L quercitin, amitriptylin, histidine, arginine, labetalol, doxepin, proline, and tryptophan (Merck Life Science, Darmstadt, Germany) were injected to monitor and ensure instrument performance, e.g., retention time, mass accuracy, peak shape, and peak intensity. When all the above quality checks were acceptable, a processed and diluted clean solvent from OMNImet•GUT | ME-200 tubes was injected twice as an extraction blank. Then, four injections of the pooled QC were performed to generate deep-scan ID samples before unknown samples were run in full-scan mode. The pooled QC was injected every 10 samples to monitor drift along the run.
Mass spectrometry (MS)-based metabolomics analyses
For both targeted and untargeted analyses, aliquots of fecal samples collected in OMNImet•GUT were vortexed for 1 minute and centrifuged at 6,500 × g for 10 min at 4 °C. The supernatants were spin-filtered using a 10 kDa cutoff spin filter at 20,800 × g for 10 min at 0 °C. An aliquot of non-processed fecal sample (≈ 100 µL) was dried to determine the wet-to-dry weight ratio. This ratio was used to estimate the theoretical dry weight of the samples analyzed. The metabolite levels were further adjusted for dilution and extraction volume. Final peak areas were normalized by dividing the final internal standard corrected signal area by the estimated dry weight, yielding peak areas per gram of dried feces.
UPLC-MS/MS analysis for quantification of short-chain fatty acids and organic acids
For targeted profiling of short-chain fatty acids (SCFAs) and organic acids (OAs), samples were prepared according to a previously validated method with minor modifications (https://doi.org/10.1016/j.aca.2014.11.015). In brief, cleaned supernatants were derivatized in a 2:1:1 volume ratio with 200 mM 3-nitrophenylhydrazine (3NPH) and a 120 mM N-(3-dimethylaminopropyl)-N'-ethylcarbodiimide (EDC) solution (6% pyridine) prepared in a 50:50 (v/v) acetonitrile (ACN): MilliQ water mixture. The derivatization reaction was allowed to proceed for 30 minutes at 40 °C. The reaction mixture was first diluted 1:2 with 2,6-di-tert-butyl-4-methylphenol (BHT) and subsequently with 2% formic acid, both prepared in ACN:MilliQ (10:90, v/v), to quench the reaction. Lastly, the quenched mixture was diluted at 1:6.25 ratio with ACN: MilliQ (10:90, v/v), yielding a total dilution after derivatization of 1:25. Before MS analysis, the extract was mixed with an equal volume of isotopically labelled internal standards (ISTD_T).
Preparation of external and internal standards and quality controls
Analytical standards for 15 SCFAs and 8 OAs were purchased from Sigma-Aldrich (Darmstadt, Germany): acetate, propionate, butyrate, isobutyrate, valerate, isovalerate, pivalate, 2-methylbutyrate, methyl isobutyrate, hexanoate, 3,3-dimethylbutyrate, 2-ethylbutyrate, 2-methylvalerate, 3-methylvalerate, 4-methylvalerate, lactate, succinate, pyruvate, oxaloacetate, α-ketoglutarate, fumarate, malate, and citrate. Analytical grade 3NPH, EDC, pyridine, formic acid, and BHT were also purchased from Sigma-Aldrich. For ISTD_T preparation, 13C-3NPH was purchased from Cayman Chemicals.
Preparation of external standard (ESTD) curves and ISTD_T for all targeted metabolites was performed as previously described. In brief, 20 mM ESTD mixes of 1) acetate, propionate, and butyrate, 2) the remaining SCFA, and 3) all OAs, were freshly prepared in 50% ACN:MilliQ (v/v). The standard mixes were diluted to 5 mM for acetate, propionate, and butyrate, and one mM for the rest. This mixture was further diluted to form an 8-point calibration curve, with a two-fold dilution between the two highest concentrations, followed by a three-fold dilution between the other concentrations (3.43-5000 µM for acetate, propionate, and butyrate, and 0.69-1,000 µM for the rest). Because lactate levels in infant feces were much higher than those of any other metabolite, an additional standard mixture for lactate was prepared at 30 mM and diluted to a 6-point curve spanning 937.5-30,000 µM, with a 2-fold dilution between points. The two sets of standards were then derivatized and processed under the same conditions and at the same concentrations as described for the fecal samples above. A fecal sample from an adult donor was processed in the same manner as the standards/samples to serve as an internal quality control (QC) sample. It was injected every 10 samples throughout the analytical run. ISTD_T mixes were prepared by diluting acetate to 4 mM, propionate and butyrate to 2 mM, and the remaining metabolites to 1 mM. Fifty µL of each ISTD_T was mixed with 1 mg of 13C-3NPH, 25 µL EDC-6% pyridine, and 25 µL 50% ACN: MilliQ (v/v). The mixture was derivatized under the same conditions as above and, after adding BHT and formic acid, diluted to 100 mL with 10% CAN.
Targeted UPLC-MS/MS instrumentation and data processing
Targeted metabolomic analysis was conducted using Waters XEVO TQ-XS Triple Quadrupole mass spectrometer coupled to Waters Acquity UPLC. Chromatographic separation was achieved by gradient elution on Acquity Premier HSS T3 (1.8 µm, 2.1x100 mm). Mobile phase A consisted of MilliQ water with 0.1% formic acid (v/v), and mobile phase B of ACN with 0.1% formic acid (v/v). The gradient was as follows: 15% B for 2 min, 15-40% B in 10 min, 40-100% B in 1 min, 100% B for 0.1 min, and held at 15% B for 2 min. The flow rate was 0.5 mL/min, the column temperature 50 °C, and the autosampler was maintained at 6 °C. The mass spectrometer was operated in negative electrospray ionization mode using multiple reaction monitoring. Desolvation temperature was 550 °C, desolvation gas flow 1000 L/h, cone gas flow 150 L/h, nebulizer gas flow 7 bar, capillary voltage 0.6 kV, and ion energy 1 and 2 were set to 1 and 2, respectively. Peak integration and data processing were performed using the TargetLynx application in MassLynx.
Analyte responses were calculated as a ratio of the endogenous peak area to the corresponding ISTD_T peak area. Quantification was based on a weighted (1/x2) linear regression of the ESTD curve, using at least six calibration points that covered the biological concentration range, with an R2 value close to 0.99. Two metabolite pairs, isovalerate/methylisobutyrate and pivalate/2-methylbutyrate, did not separate and were thus quantified together.
Untargeted LC-MS/MS profiling
Supernatants from fecal samples were prepared as for the targeted analysis. Five µL of each sample supernatant was pooled to generate a QC sample. Afterwards, all supernatants including the QC were diluted 1:10 with ACN: MilliQ water (90:10, v/v) containing internal standard mix (ISTD_U) of Acetyl-l-carnitine-D3, Hexadecanoyl(palmitoyl)-l-carnitine-D3, l-Leucine-5,5,5-D3, l-Tryptophan-(indole-D5), l-Methionine-(methyl-D3), Stearic acid-18,18,18-D3, Chenodeoxycholic-2,2,3,4,4,6,6,7,8-D9 acid, 18:0-D35 Lyso PC, Dopamine-1,1,2,2-D4 hydrochloride (Merck Life Science, Darmstadt, Germany). Before analysis, the dilution factor, solvent type, sample extraction method, and sample stability in the autosampler, freezer, and at room temperature were evaluated with respect to efficient sample handling, peak shape and intensity, and reproducibility, and were found to be well within the instrument's dynamic range. The final samples for analysis were randomized (RAND function in Excel, Microsoft Office), transferred to 96-well injection plates (Waters, SKU: 186002481), with the QC sample included every 10 injections, and stored in the autosampler (6 °C) throughout acquisition.
LC-MS/MS instrumentation and data acquisition
Untargeted metabolomic profiling was based on a published hydrophilic interaction liquid chromatography (HILIC) method (https://doi.org/10.1016/j.chroma.2024.465230). In summary, the instrument was a Thermo Vanquish UHPLC system coupled with a Thermo Scientific Orbitrap ID-X Tribrid Mass Spectrometer. The data acquisition was performed using Thermo AcquireX Deep Scan module in data-dependent acquisition (DDA). Chromatographic separation was achieved by injecting 3 µL of each sample through a BEH Amide (100x2.1 mm, 1.7 Å) column at a flow rate of 0.45 mL/min. The column temperature was maintained at 50 °C, and the autosampler at 6 °C. The separation was carried out using gradient elution with mobile phase A consisting of H2O with 10 mM NH4Ac at pH 9, and mobile phase B of ACN: H2O (9:1, v/v) with 10 mM NH4Ac at pH 9. Mobile phase B was maintained at 100% for 6 min to equilibrate the system and then kept at 100% for an additional 2.5 min after injection. Then 100-60% from 2.5-9 min and kept at 60% for 0.2 min.
The ion source was operated in heated negative electrospray ionization (2500 V). Gas settings were 50 arbitrary units (arb) for sheath gas, 10 arb for auxiliary gas, and one arb for sweep gas, with an ion transfer tube temperature of 325 °C and vaporizer temperature of 350 °C. Full MS scans were acquired using the Orbitrap at a resolution of 120,000 over a scan range of 70-800 m/z. The automatic gain control (AGC) target was set to custom with a normalized AGC of 25 and a maximum injection time of 50 milliseconds (ms). For MS/MS, up to 8 dependent scans were triggered per MS1 event using higher-energy collision dissociation (HCD, with stepped collision energies of 20%, 35%, and 50%) and collision-induced dissociation (CID, with a fixed energy of 30%). MS2 scans were also acquired in the Orbitrap at a resolution of 30,000 with a quadrupole isolation window of 1.2 m/z and a maximum injection time of 150 ms. Dynamic exclusion was enabled (excluding after one acquisition, for 5 seconds) with isotope exclusion to prevent repeated fragmentation of isotopologues.
Untargeted LC-MS/MS quality control assessment
Prior to sample injection, a system blank consisting of clean sample solvent (ACN: MilliQ, 90:10, v/v) was injected to ensure that the instrument and mobile phase were free from contamination. Next, several injections of a system control sample consisting of 0.1 mg/L quercitin, amitriptylin, histidine, arginine, labetalol, doxepin, proline, and tryptophan (Merck Life Science, Darmstadt, Germany) were injected to monitor and ensure instrument performance, e.g., retention time, mass accuracy, peak shape, and peak intensity. When all the above quality checks were acceptable, a processed and diluted clean solvent from OMNImet•GUT | ME-200 tubes was injected twice as an extraction blank. Then, four injections of the pooled QC were performed to generate deep-scan ID samples before unknown samples were run in full-scan mode. The pooled QC was injected every 10 samples to monitor drift along the run.
Additionally, the ISTD_U spike in every sample was used to monitor drift and injection quality. After the injection sequence, the ISTD_U m/z values were uploaded into the software Skyline (v 24.1.0.414, MacCross Lab, University of Washington, USA), where retention time drift and signal intensity were monitored for each ISTD_U in each sample. Next, extracted ion chromatograms were manually checked and compared in the software Freestyle (Thermo Fisher Scientific, San Jose, USA). Specific masses for tryptophan, proline, and isoleucine were manually verified in each QC to confirm peak shape, retention time, and intensity. The blanks were also checked to verify the absence of the specific masses. Finally, to evaluate stability, QCs were visualized in a PCA plot to detect potential global drift in metabolomic profiles over time.
Raw MS/MS data processing
The raw data files were centroided and processed using MZmine 4.7.8 (https://doi.org/10.1038/s41587-023-01690-2) in two processing batches. The processing workflows were as follows: in the first batch, mass detection in MS1 was performed for all samples using a noise level of 1.0e4 based on precursor intensity. Then, mass detection in MS2 was performed without noise-level filtering across four pooled MS2 samples, which were representative of the MS2 data. Subsequent steps included chromatogram builder, smoothing (Savitzky-Golay), local minimum feature resolver, 13C isotope filter, and isotopic peaks finder with default settings. Processing batch two was performed by first running the join aligner on the merged ID samples, then aligning to all the full-scan samples with a retention time tolerance of 0.2 min. The extraction blank was used in the module feature list subtraction, with a fold change threshold of 300%, to retain relevant features. The following modules were applied: group MS2 scans with features, feature list rows filter, correlation grouping, and ion identity networking. The processed data were extracted for downstream analysis in GNPS (https://doi.org/10.1038/s41592-020-0933-6) and SIRIUS (https://doi.org/10.1038/s41592-019-0344-8). To determine the need for drift correction, the quantification table was uploaded to MetaboAnalyst 6.0, and the QCs were monitored using principal component analysis (PCA).
SIRIUS-based annotation
For feature annotation, the processed data from MZmine were uploaded into SIRIUS 6.2.2 using default settings. Spectral matching, molecular formula prediction, CSI: FingerID, and structure database search with PubChem as a fallback was enabled. Annotations were manually curated and evaluated based on the compound's confidence score. A score above 0.6 was considered good if the substructure matched with databases and had a consistent top structure hit. These features were assigned a level 2 annotation. Features with lower confidence scores, which showed reliable substructure matches, were assigned level 3 annotation and were annotated at the ClassyFire^^83 most specific class level.
Feature-based molecular networking and statistics
For feature-based molecular networking, the processed data from MZmine were uploaded and run in GNPS. The processed network was then uploaded to the software Cytoscape 3.10.3 for network visualization. Edges were labeled with their delta mz and further colored by cosine (blue) or MS1 annotation (red). Furthermore, the node sizes were mapped to the sum of precursor intensity. The normalized quantification table was uploaded to MetaboAnalyst, filtered by standard deviation (40%), normalized by median, log-transformed, and Pareto scaled. To assess significant features, a fold change cutoff of 2, with a p-value threshold of 0.05, and a false discovery rate (FDR) were used, along with unequal group variance, to generate a list of significantly different features between the two groups. This list was further uploaded to Cytoscape. The nodes were colored either for significantly higher (red) or lower (blue) abundance to visualize global trends in the network.