Untargeted Metabolomics Analysis of Pancreatic Cancer Using LC/QTOF

Untargeted Metabolomics Analysis of Pancreatic Cancer Using LC/QTOF — Expert Summary

Significance of the topic:

Metabolomics reveals small-molecule signatures that reflect disease state, biochemical pathway perturbations and potential biomarkers. Robust untargeted workflows that are accessible to routine laboratories are essential to accelerate discovery in clinical metabolomics, biomarker research, and translational studies. This work demonstrates an end-to-end LC/QTOF pipeline applied to serum from pancreatic ductal adenocarcinoma (PDAC) patients versus healthy controls, illustrating practical considerations for acquisition, quality control and open-source data processing.

Study objectives and overview:

- Develop a reliable UHPLC–QTOF MS method for untargeted metabolomics that supports both data-independent (DIA) and data-dependent (DDA) acquisition while remaining accessible to users with minimal mass‑spectrometry expertise.
- Apply the workflow to serum samples from PDAC patients and matched healthy controls to identify differentially expressed metabolites.
- Use open-source tools (MS-DIAL and MetaboAnalyst) for feature detection, alignment, annotation and statistical interpretation to produce a reproducible analysis pipeline.

Sample preparation and extraction:

- Serum proteins were precipitated/extracted with methanol (sample:solvent 1:3, v/v).
- For positive‑ion untargeted analysis, extracts were diluted 1:10 prior to UHPLC–MS/MS.

Methodology and data acquisition:

- Chromatography: reversed‑phase UHPLC (alternative HILIC options noted) using Nexera X2 and a C18 column (2.1 × 100 mm, 1.7 µm) at 50 °C. Flow 0.4 mL/min. Mobile phases: A = water + 0.1% formic acid; B = acetonitrile + 0.1% formic acid. Typical gradient progressed from 2% B to 90–100% B with a 35‑minute total run including re-equilibration. Injection volume 0.5 µL; autosampler at 4 °C.
- Mass spectrometry: Shimadzu LCMS‑9030 QTOF in positive electrospray ionisation. Interface 300 °C, 1.5 kV; heat block 400 °C; desolvation (DL) 250 °C. Gas flows: nebulizing 3 L/min, heating 10 L/min, drying 15 L/min. CID (collision) gas pressure ~230 kPa.
- Acquisition strategy: TOF‑MS scan m/z 60–1000 plus DIA‑MS/MS (27 windows covering m/z 40–1000, ~20 Da isolation windows, collision energy spread 5–55 V). Combined MS+MS/MS cycle ~1 s to support high‑throughput feature detection and MS/MS‑level annotation across the chromatogram.

Pooled QCs and quality control strategy:

- Pooled QC samples were prepared by combining equal aliquots of each study sample (mQACC recommendations followed).
- Randomised injection order; system conditioning followed by periodic pooled QC injections (e.g., after every 6 samples) to monitor retention time, mass accuracy and signal variance.
- Feature filtering criteria: present in ≥50% of pooled QC injections with RSD <20%, and present in ≥80% of samples in at least one biological group (PDAC or control) prior to statistical testing.

Data processing and annotation:

- MS-DIAL read LabSolutions raw files directly (no conversion required) to perform peak detection, retention‑time alignment and DIA deconvolution. The lipidomics workflow and additional .msp start‑up libraries (including MassBank repositories) were used for annotation.
- MS‑DIAL produces deconvoluted MS/MS spectra from DIA windows and matches these against reference spectra to provide candidate annotations and match scores. Navigation features (peak spot viewer, peak tables) facilitated manual review and verification of annotations.
- All reported metabolite annotations were assigned MSI level 2 (putative annotation via MS/MS match to external libraries).

Statistical analysis and marker selection:

- The aligned feature matrix (positive mode) was imported into MetaboAnalyst for statistical processing. Missing values were imputed using the default replacement (1/5 of the minimum positive value per feature).
- Univariate filtering: volcano‑plot selection using fold change >1.5 and p‑value <0.1 identified 229 features as significant candidates.
- Manual curation in MS‑DIAL removed unannotated features, probable exogenous compounds (e.g., drugs), retention‑time discordant annotations for reversed‑phase separation, and redundant adducts, yielding 45 annotated features considered potential PDAC markers.

Main results and discussion:

- The workflow revealed widespread lipid perturbations and changes in several small metabolites in PDAC serum versus controls. Annotated classes included lysophosphatidylcholines (e.g., LPC 18:2, LPC 18:3, LPC 20:5), multiple phosphatidylcholines (PCs, including ether forms PC O‑ and plasmalogens PE P‑), sphingomyelins (SM species), acylcarnitines (e.g., CAR 18:1, acetylcarnitine), amino acid–related metabolites (L‑glutamic acid), and hippuric acid.
- Directionality: a mix of increases and decreases in PDAC samples was observed across lipid classes and small molecules, indicating complex metabolic dysregulation rather than a single pathway perturbation. Representative box plots demonstrated reproducible group separation for several lipids annotated by MS‑DIAL.
- Annotation confidence: all reported candidates are MSI level 2; further targeted validation (authentic standards, orthogonal MS/MS, retention index confirmation) will be required to elevate identifications to level 1 for clinical or mechanistic interpretation.

Practical benefits and applicability of the method:

- The described LC/QTOF + DIA workflow provides an accessible untargeted metabolomics solution suitable for routine research labs, with minimal need for advanced MS expertise thanks to automated DIA deconvolution and user‑friendly interfaces in MS‑DIAL and MetaboAnalyst.
- Raw data are compatible with multiple open‑source tools (MS‑DIAL, MetaboAnalyst, MZmine, GNPS, XCMS), enabling flexible downstream processing and community‑standard analyses.
- Periodic pooled QCs and objective filtering criteria increase data robustness and reduce false discoveries in comparative studies.

Used instrumentation:

• UHPLC: Shimadzu Nexera X2
• LC column: C18, 2.1 × 100 mm, 1.7 µm, column thermostat 50 °C
• Mass spectrometer: Shimadzu LCMS‑9030 Quadrupole Time‑of‑Flight (QTOF)
• Software: MS‑DIAL for feature detection/annotation and MetaboAnalyst for statistical analysis (data compatible with MZmine, GNPS, XCMS)

Limitations and considerations:

- Only positive ion mode data were reported; negative mode analysis may reveal additional, complementary markers and should be included in follow‑up studies.
- Annotations remain at putative (MSI level 2) confidence; confirmatory targeted assays are necessary for biomarker validation.
- Statistical thresholds used (fold change >1.5, p < 0.1) are permissive; adjustment for multiple testing and larger cohort sizes will improve specificity.

Future trends and potential applications:

• Expand acquisition to include negative‑ion mode, orthogonal chromatographic separation (HILIC) and ion mobility to increase metabolite coverage and improve isomer separation.
• Integrate larger, multi‑center cohorts with standardized QC and normalization pipelines to enhance biomarker reproducibility and clinical translatability.
• Combine untargeted metabolomics with targeted follow‑up assays, isotope‑labelled standards and pathway‑level interpretation for mechanistic insight.
• Adopt machine‑learning and multi‑omics integration (proteomics, genomics) to refine predictive signatures for PDAC detection and staging.
• Leverage open‑data repositories and spectral libraries to improve automated annotation confidence and share validated spectral evidence.

Conclusion:

The authors present a practical, reproducible LC/QTOF untargeted metabolomics workflow that uses DIA‑MS/MS acquisition and open‑source software (MS‑DIAL, MetaboAnalyst) to identify candidate serum markers associated with PDAC. The pipeline produced 45 putative metabolite/lipid markers after QC filtering and manual curation. The approach is well suited for routine discovery studies but requires orthogonal validation and expansion (negative mode, larger cohorts) before clinical application.

References:

1. Kirwan JA, Gika H, Beger RD, et al. Quality assurance and quality control reporting in untargeted metabolic phenotyping: mQACC recommendations for analytical quality management. Metabolomics. 2022;18(9).
2. Spicer R, Salek R, Steinbeck C. A decade after the metabolomics standards initiative it's time for a revision. Scientific Data. 2017;4:170138.