# Clinical and mechanistic relevance of high-dimensionality analysis of the paediatric sepsis immunome

## TL;DR

This study uses high-dimensionality analysis to identify immune signatures in children with sepsis that predict poor outcomes.

## Contribution

The study introduces a novel model using specific immune subsets to diagnose sepsis and predict outcomes in children.

## Key findings

- Sepsis in children is associated with a loss of coordinated immune communication and regulatory interactions.
- Four immune subsets were identified that predict sepsis with high accuracy in both discovery and validation cohorts.
- The model based on these subsets has strong diagnostic potential with AUC values above 0.9.

## Abstract

By employing a high-dimensionality approach, this study aims to identify mechanistically relevant cellular immune signatures that predict poor outcomes.

This prospective study recruited 39 children with sepsis admitted to the intensive care unit and 19 healthy age-matched children. Peripheral blood mononuclear cells were studied with mass cytometry. Unique cell subsets were identified in the paediatric sepsis immunome and depicted with t-distributed stochastic neighbour embedding (tSNE) plots. Network analysis was performed to quantify interactions between immune subsets. Enriched immune subsets were included in a model for distinguishing sepsis and validated by flow cytometry in an independent cohort.

The median (interquartile range) age and paediatric sequential organ failure assessment (pSOFA) score in this cohort was 5.6(2.0, 11.3) years and 6.6 (IQR: 2.5, 10.1), respectively. High-dimensionality analyses of the immunome in sepsis revealed a loss of coordinated communication between immune subsets, particularly a loss of regulatory/inhibitory interaction between cell types, fewer interactions between cell subsets, and fewer negatively correlated edges than controls. Four independent immune subsets (CD45RA−CX3CR1+CTLA4+CD4+ T cells, CD45RA−17A+CD4+ T cells CD15+CD14+ monocytes, and Ki67+ B cells) were increased in sepsis and provide a predictive model for diagnosis with area under the receiver operating characteristic curve, AUC 0.90 (95% confidence interval, CI 0.82–0.98) in the discovery cohort and AUC 0.94 (95% CI 0.83–1.00) in the validation cohort.

The sepsis immunome is deranged with loss of regulatory/inhibitory interactions. Four immune subsets increased in sepsis could be used in a model for diagnosis and prediction of poor outcomes.