[Lipid metabolomics-based biomarker analysis of neonatal sepsis in serum and cerebrospinal fluid]

Neonatal sepsis remains a leading cause of morbidity and mortality among newborns worldwide. Despite advances in neonatal care， early diagnosis of sepsis remains challenging due to the lack of sensitive and specific biomarkers. While serum-based indicators have been widely studied， lipid metabolism in cerebrospinal fluid （CSF） remains relatively underexplored， limiting our understanding of central nervous system involvement （CNS） in the early stages of neonatal sepsis. This study aimed to systematically investigate lipid metabolic alterations in both serum and CSF samples from neonates with confirmed sepsis and to identify potential lipid biomarkers for early diagnosis. Seventeen neonates with blood culture-positive sepsis and seventeen controls with negative blood culture results were enrolled from the Neonatal Intensive Care Unit of Guangdong Women and Children Hospital （Women and Children's Hospital， Southern University of Science and Technology） between February 2020 and August 2023. Paired serum and CSF samples were collected and analyzed using targeted lipidomics based on liquid chromatography-tandem mass spectrometry （LC-MS/MS）. Univariate analyses， including Student's t-tests and Mann-Whitney U tests， were applied to identify statistically significant differences in lipid levels between groups. Multivariate analyses， including principal component analysis （PCA） and orthogonal partial least squares discriminant analysis （OPLS-DA）， were employed to further evaluate group separation and identify discriminatory lipid species. Pathway enrichment analysis was performed using the Kyoto Encyclopedia of Genes and Genomes （KEGG） database， and candidate biomarkers were selected using the Boruta feature selection algorithm and evaluated for diagnostic performance using receiver operating characteristic （ROC） curve analysis. A total of 322 lipid metabolites were identified in serum， with cholesteryl esters （CE）， triacylglycerols （TAG）， and phosphatidylcholines （PC） being the most abundant lipid classes. In the sepsis group， levels of nearly all lipid subclasses were significantly decreased compared to controls （P<0.05）， except for TAG and diacylglycerols （DAG）， which were not significantly altered. In CSF， 300 lipid species were detected， dominated by CE， PC， and phosphatidylethanolamines （PE）. Significantly reduced levels of PE， ceramides （Cer）， and lyso phosphatidylethanolamines （LPE） were observed in septic neonates （P<0.05）. PCA plots demonstrated tight clustering of quality control （QC） samples， indicating high analytical reproducibility and stable instrument performance. In serum， PCA accounted for 66.1% of total variance， showing preliminary group separation that was further confirmed by OPLS-DA （R²Y=0.601， Q²Y=0.271）， which identified 107 significantly downregulated lipid metabolites. Similarly， CSF PCA explained 75.7% of the variance， and OPLS-DA （R²Y=0.579， Q²Y=0.368） revealed 34 significantly downregulated lipid metabolites. Pathway enrichment analysis （FDR-P<0.05， pathway impact>0.10） showed that glycerophospholipid metabolism was the most significantly enriched pathway in both serum and CSF， followed by ether lipid and sphingolipid metabolism in serum. Key shared metabolites included PE（42：9）， PC（38：0）， LPC（22：6）， and LPE（22：6）， while PS（40：6） and PI（40：4） were specific to serum. Notably， thirteen differential lipid species were consistently identified in both serum and CSF， among which LPE（18：2）， ePE（36：4）， and Cer（d18：1/25：0） exhibited significant positive correlations between the two fluids （Pearson r=0.369-0.382， P<0.05）， suggesting potential trans-barrier lipid communication or shared regulatory mechanisms. Boruta-based machine learning analysis identified LPC（28：1）， LPE（18：2） and ePE（36：4） in serum as candidate biomarkers. These exhibited excellent diagnostic performance， with area under the curve （AUC） values of 0.96， 0.94， and 0.94， respectively， sensitivities ranging from 82.4% to 88.2%， and specificities from 94.1% to 100%. In CSF， Cer（d18：1/26：0）， Cer（d18：1/25：0）， and Cer（d18：1/24：1） were identified as high-importance variables. These demonstrated diagnostic AUCs of 0.89， 0.91， and 0.80， with sensitivities between 88.2% and 100% and specificities ranging from 64.7% to 70.6%. In summary， this study provides the first integrated lipidomic profiling of serum and CSF in neonatal sepsis， highlighting a consistent disruption in lipid metabolism， particularly within the glycerophospholipid pathway. Serum lipid biomarkers show promise as non-invasive early screening tools， while CSF lipid alterations offer valuable insights into CNS involvement and potential early neuroinflammatory responses. These findings support the potential of lipid-based biomarkers in improving the precision and timeliness of neonatal sepsis diagnosis. Nevertheless， the relatively small sample size and single-center design may limit the generalizability of the results. Future multicenter studies with larger cohorts are warranted to validate these findings and support clinical translation into neonatal care.