Marie-Angélique Cazalis ^1,2 Louis Kreitmann ^1,2 Guillaume Monneret ^1,3 Alexandre Pachot ^1,2 Karen Brengel-Pesce ^1,2 Jean-François Llitjos ^1,2,4*

• ¹ Joint Research Unit HCL-bioMérieux, EA 7426 “Pathophysiology of Injury-Induced Immunosuppression” (Université Claude Bernard Lyon 1 – Hospices Civils de Lyon, bioMérieux), Lyon, Rhône-Alpes, France
• ² Open Innovation and Partnerships (OIP), bioMérieux S.A., Marcy l'etoile, France
• ³ Immunology Laboratory, Edouard Herriot Hospital – Hospices Civils de Lyon, Lyon, Rhône-Alpes, France
• ⁴ Anaesthesia and Critical Care Medicine Department, Hospices Civils de Lyon, Edouard Herriot Hospital, Lyon, Rhône-Alpes, France

Background. Transcriptomics biomarkers have been widely used to predict mortality in patients with sepsis. However, the association between mRNA levels and outcomes shows substantial variability over the course of sepsis, limiting their predictive performance. We aimed to: a) identify and validate an mRNA biomarker signature whose association with all-cause intensive care unit (ICU) mortality is consistent at several timepoints; and to b) evaluate how this mRNA signature could be used in association with lactate levels for predictive and prognostic enrichment in sepsis. Methods. We conducted a gene expression analysis study at two timepoints (day 1 and day 2-3 following ICU admission) using microarray data from adult septic patients to identify candidate biomarkers predictive of all-cause ICU mortality. We validated mRNA biomarkers using reverse transcriptase quantitative polymerase chain reaction (RT-qPCR) on an external validation cohort.The predictive performance of the mRNA biomarkers combination was assessed in association with lactate level to refine ICU mortality prediction. Main Results. Among 180 chips from 100 septic patients, we identified 39 upregulated and 2 downregulated differentially expressed genes (DEGs) in survivors vs. non-survivors, both at day 1 and days 2-3 following ICU admission. We combined CDK1, the hub gene of upregulated DEGs, and CX3CR1 and IL1b to compute expression ratios. The CDK1/CX3CR1 ratio had the best performance to predict all-cause ICU mortality, with an area under the ROC curve (AUROC) of 0.77 (95% confidence interval [CI] 0.88-0.66) at day 1 and of 0.82 (95% CI 0.91-0.72) at days 2-3 after ICU admission. This performance was better than that of each individual mRNA biomarker. In the external validation cohort, the predictive performance of the CDK1/CX3CR1 ratio, measured using RT-qPCR, was similar to that of lactate when measure at day 1, and higher when measured at days 2-3. Combining lactate levels and the CDK1/CX3CR1 ratio, we identify 3 groups of patients with an increasing risk of ICU-mortality, ranging from 9% to 60% with an intermediate-risk group mortality rate of 28%. Conclusion . With stable predictive performance over the first three days following ICU admission, the CDK1/CX3CR1 ratio identifies three groups of septic patients with increasing ICU mortality risk.