Background
Comprehensive and in-depth research on the immunophenotype of septic patients remains limited, and effective biomarkers for the diagnosis and treatment of sepsis are urgently needed in clinical practice.
Methods
Blood samples from 31 septic patients in the Intensive Care Unit (ICU), 25 non-septic ICU patients, and 18 healthy controls were analyzed using flow cytometry for deep immunophenotyping. Metagenomic sequencing was performed in 41 fecal samples, including 13 septic patients, 10 non-septic ICU patients, and 18 healthy controls. Immunophenotype shifts were evaluated using differential expression sliding window analysis, and random forest models were developed for sepsis diagnosis or prognosis prediction. Findings: Septic patients exhibited decreased proportions of natural killer (NK) cells and plasmacytoid dendritic cells (pDCs) in CD45+ leukocytes compared with non-septic ICU patients and healthy controls. These changes statistically mediated the association of Bacteroides salyersiae with sepsis, suggesting a potential underlying mechanism. A combined diagnostic model incorporating B.salyersia, NK cells in CD45+ leukocytes, and C-reactive protein (CRP) demonstrated high accuracy in distinguishing sepsis from non-sepsis (area under the receiver operating characteristic curve, AUC = 0.950, 95% CI: 0.811-1.000). Immunophenotyping and disease severity analysis identified an Acute Physiology and Chronic Health Evaluation (APACHE) II score threshold of 21, effectively distinguishing mild (n = 19) from severe (n = 12) sepsis. A prognostic model based on the proportion of total lymphocytes, Helper T (Th) 17 cells, CD4+ effector memory T (TEM) cells, and Th1 cells in CD45+ leukocytes achieved robust outcome prediction (AUC = 0.906, 95% CI: 0.732-1.000), with further accuracy improvement when combined with clinical scores (AUC = 0.938, 95% CI: 0.796-1.000). Interpretation: NK cell subsets within innate immunity exhibit significant diagnostic value for sepsis, particularly when combined with B. salyersiae and CRP. In addition, T cell phenotypes within adaptive immunity are correlated with sepsis severity and may serve as reliable prognostic markers. Funding: This project was supported by the National Key R&D Program of China (2023YFC2307600, 2021YFA1301000), Shanghai Municipal Science and Technology Major Project (2023SHZDZX02, 2017SHZDZX01), Shanghai Municipal Technology Standards Project (23DZ2202600).