Abstract
BACKGROUND: Sepsis, a life-threatening inflammatory syndrome with incompletely understood genetic mechanisms, necessitates identification of pathogenic proteins to advance precision medicine. This study integrated Mendelian randomization (MR) and multi-omics analyses to systematically investigate causal proteins and their regulatory mechanisms in sepsis. METHODS: We employed a multi-step framework combining two-sample MR, expression validation, immune infiltration analysis, single-cell RNA sequencing and mediation analysis. Exposure variables included cis-protein quantitative trait loci (cis-pQTLs) from two plasma proteome datasets (4,907 proteins from Decode and 2,640 proteins from UKB-PPP), with sepsis (ieu-b-4980) and 28-day sepsis (ieu-b-5086) serving as the outcome. Intersection analysis revealed the candidate proteins common between the two cis-pQTL datasets, followed by multi-omics validation using microarray expression profiling, receiver operating characteristic (ROC) analysis, and methylation quantitative trait loci (mQTLs). We also performed mediation analyses to assess whether immune cells potentially mediate causal pathways linking proteins to sepsis. Drug prediction was performed using DGIdb and molecular docking was subsequently used to identify the candidate drugs. RESULTS: MR analysis identified 97 (Decode) and 130 (UKB-PPP) sepsis-associated proteins, with 15 overlapping candidates. Among these, SFRP1, IL1RL1, and INHBB were validated as risk factors, while GSTP1 exhibited protective effects. They exhibited directionally concordant expression differences across an independent microarray dataset GSE95233. Furthermore, the associations of IL1RL1, INHBB, and GSTP1 with sepsis were replicated in the 28-day mortality dataset. ROC analysis demonstrated the superior diagnostic performance of GSTP1 for sepsis. MR further revealed that methylation at the GSTP1 locus cg25135322 increased the risk of sepsis (OR = 1.0844, 95% CI = 1.0339-1.1374). Immune infiltration analysis showed that GSTP1 expression was positively correlated with resting NK cells and CD8 + T cells, but negatively correlated with neutrophils. Single-cell RNA sequencing revealed the elevated GSTP1 expression in T cells and B cells, but reduced expression in monocytes of patients with sepsis. Mediation analysis suggested that CD39 + CD8br Treg cells might partially mediate the protective effects of GSTP1, accounting for 9.8% (Decode) and 11.47% (UKB-PPP) of the total effect, respectively. Four candidate drugs were identified, with Exatecan mesylate and Misonidazole showing optimal binding with GSTP1. CONCLUSIONS: This study identified three proteins with putative causal associations with sepsis risk, providing valuable clues for the development of biomarkers and therapeutic targets for sepsis.