A non-targeted metabolomics analysis of SARS-CoV-2 and influenza infection in children

Significance of the topic

Acute respiratory infections (ARIs) in children present diagnostic challenges because diverse viral etiologies can produce similar clinical signs. Rapid, specific biomarkers that differentiate infections such as SARS-CoV-2 and influenza could improve triage, treatment decisions and prognosis. Non-targeted metabolomics of blood offers a systems-level readout of host metabolic response and the potential to identify discriminating lipid and metabolite signatures that are not evident from routine clinical tests.

Objectives and study overview

This study applied a non-targeted LC-MS/MS metabolomics workflow to whole-blood extracts from pediatric patients with laboratory-confirmed SARS-CoV-2 or Influenza A and from matched healthy controls. The goals were to: (1) phenotype systemic metabolic changes associated with each infection, (2) identify statistically significant features and putative compound identifications, and (3) demonstrate an integrated, automated data-processing pipeline for QTOF data to accelerate discovery of candidate biomarkers.

Materials and methods

Clinical cohort and sample handling:

• Whole blood from children aged 1 month to 16 years: SARS-CoV-2 (n=21), Influenza A (n=28), controls (n=10).
• Age- and gender-matched; subjects with immunosuppression or significant comorbidities were excluded.
• Metabolite extraction: ice-cold methanol; pooled QC prepared by mixing aliquots of each extract.

Analytical strategy and data processing:

• Non-targeted metabolomics using high-resolution reversed-phase LC-MS/MS (data-independent acquisition, DIA) on a Q-TOF instrument.
• Automated end-to-end data processing in Insight Profiler: integrated feature detection, alignment, statistical testing, and compound annotation with MS/MS library matching.
• Statistical workflow: PCA for overview, ANOVA with FDR correction to select significant features (pFDR < 0.05), followed by Tukey’s HSD post hoc pairwise comparisons.
• Compound annotation via an in-house library plus MassBank, LipidBlast and HMDB for MS/MS matching; identifications reported as putative.

Used instrumentation

• Mass spectrometer: LCMS-9050 Q-TOF (Shimadzu).
• LC column and conditions: C18 BEH, 2.1 x 100 mm, 1.7 µm, 50 °C, 0.4 mL/min, binary gradient (water + 0.1% formic acid / acetonitrile + 0.1% formic acid), sample cycle ~35 min.
• MS acquisition: positive ion mode survey scan m/z 60–1000 (100 ms); 27 DIA MS/MS windows covering m/z 40–1000 (33 ms per scan) with 35 Da precursor isolation windows; collision energy spread 5–55 V; external mass calibration; total cycle time <1 s.

Main results and discussion

Global findings:

• Insight Profiler detected and annotated 75 metabolite and lipid features that differed significantly between groups by ANOVA (FDR-adjusted p < 0.05).
• Post hoc testing indicated 73 features significantly altered in SARS-CoV-2 versus controls, 35 in Influenza A versus controls, and 9 features distinguishing SARS-CoV-2 from Influenza A.

Key biochemical signatures:

• Phosphatidylcholines (PCs), including several ether-linked PCs (e.g., PC O-16:1_18:2, PC O-16:0_18:2) and diacyl PCs (PC 18:2_18:2, PC 18:2_20:4), were among the most significantly changed features.
• These PCs were generally reduced in both SARS-CoV-2 and Influenza A cases compared with controls. Two ether PCs (PC O-16:1_18:2 and PC O-16:0_18:2) showed significantly lower abundances in SARS-CoV-2 relative to Influenza A.

Interpretation:

• Depletion of specific phosphatidylcholines suggests infection-associated perturbation of membrane lipid metabolism and host lipid remodeling during viral ARI; ether phospholipids may reflect differences in inflammatory or oxidative processes between viral infections.
• The overlap in metabolic changes between SARS-CoV-2 and Influenza A is consistent with shared host responses to respiratory viruses, while the subset of features that differ between the two infections points to potential discriminatory biomarkers.

Benefits and practical applications

• Demonstrates feasibility of an integrated automated pipeline (Insight Profiler) for high-throughput non-targeted QTOF metabolomics, reducing manual processing and enabling reproducible batch analysis.
• Identifies candidate lipid biomarkers (notably PCs) that could be further validated and potentially developed into adjunctive assays to improve ARI diagnosis or stratify disease phenotypes in pediatric populations.
• Pooled QC strategy and DIA MS/MS acquisition provide robust spectral coverage to support putative identifications and cross-sample comparisons.

Limitations

• Moderate cohort sizes (especially controls) limit statistical power and generalizability; findings are preliminary and require independent validation in larger, diverse cohorts.
• Annotations are putative based on MS/MS library matching; orthogonal confirmation (authentic standards, targeted assays) is needed for definitive identification and quantification.
• The software and workflow are research-use-only and not validated for clinical diagnostic decision-making.

Future trends and applications

• Validation and translation: targeted LC-MS/MS panels based on the top candidate PCs and other discriminating metabolites could be developed for rapid clinical assays after multi-cohort validation.
• Integration with multi-omics: combining metabolomics with proteomics, lipidomics and host transcriptomics could improve mechanistic understanding and biomarker specificity for different ARIs.
• Refinement of DIA and data-processing: advances in spectral deconvolution, larger curated MS/MS libraries and machine-learning-based annotation will enhance confidence in non-targeted identifications and accelerate discovery.
• Point-of-care potential: if validated, simplified lipid or metabolite signatures could inform rapid triage tools to guide antiviral use, isolation decisions and resource allocation in pediatric care settings.

Conclusion

This work shows that non-targeted LC-MS/MS metabolomics of pediatric whole-blood extracts can detect reproducible metabolic perturbations associated with SARS-CoV-2 and Influenza A infections. A predominance of decreased phosphatidylcholines characterized infected children relative to controls, with specific ether-linked PCs lower in SARS-CoV-2 than in Influenza A. The study also illustrates the utility of an integrated software pipeline to streamline feature finding, statistical testing and putative compound annotation, enabling rapid hypothesis generation for biomarker development. However, findings remain exploratory and require targeted validation before clinical application.

References

1. Study authors and affiliations as provided: Emily G. Armitage et al., Shimadzu Corporation; Aristotle University School of Medicine; Biomic AUTh.
2. Software and libraries referenced: Insight Profiler (Shimadzu), MassBank, LipidBlast, HMDB (used for MS/MS matching).