The cytokines HGF and CXCL13 predict the severity and the mortality in COVID-19 patients

Abstract

The objective of the present study was to identify biological signatures of severe coronavirus disease 2019 (COVID-19) predictive of admission in the intensive care unit (ICU). Over 170 immunological markers were investigated in a 'discovery' cohort (n = 98 patients) of the Lausanne University Hospital (LUH-1). Here we report that 13 out of 49 cytokines were significantly associated with ICU admission in the three cohorts (P < 0.05 to P < 0.001), while cellular immunological markers lacked power in discriminating between ICU and non-ICU patients. The cytokine results were confirmed in two 'validation' cohorts, i.e. the French COVID-19 Study (FCS; n = 62) and a second LUH-2 cohort (n = 47). The combination of hepatocyte growth factor (HGF) and C-X-C motif chemokine ligand 13 (CXCL13) was the best predictor of ICU admission (positive and negative predictive values ranging from 81.8% to 93.1% and 85.2% to 94.4% in the 3 cohorts) and occurrence of death during patient follow-up (8.8 fold higher likelihood of death when both cytokines were increased). Of note, HGF is a pleiotropic cytokine with anti-inflammatory properties playing a fundamental role in lung tissue repair, and CXCL13, a pro-inflammatory chemokine associated with pulmonary fibrosis and regulating the maturation of B cell response. Up-regulation of HGF reflects the most powerful counter-regulatory mechanism of the host immune response to antagonize the pro-inflammatory cytokines including CXCL13 and to prevent lung fibrosis in COVID-19 patients.

Fig. 1. Distribution of CD4 T cell lineage and phosphoprotein signaling profiles in non-ICU and ICU COVID-19 patients.

a Frequencies of Th1 (CXCR3+T-bet+), Th2 (CCR4+Gata-3+), Th17 (CCR6+RoR-γt+) and Treg (CD25+CD127-FoxP3+) CD4 T cell sub-populations in healthy subjects (N = 146), non-ICU (N = 50) and ICU (N = 25) patients. b Mean signal intensity of ex vivo phospho-STAT1 (pSTAT1), pSTAT3, pSTAT5, p38, pMAPKAP2, pNFkB, pCREB, pS6 and pERK1/2 in healthy subjects (N = 39), non-ICU (N = 33) and ICU (N = 29) patients. Blue plots correspond to healthy subjects (H.S), red plots correspond to non-ICU patients and green plots correspond to ICU patients. Black stars indicate statistical significance between ICU or non-ICU patients and healthy subjects. Statistical significance (P values) was obtained using two-sided Kruskal–Wallis test, using a Bonferroni correction. *P < 0.05; **P < 0.01; ***P < 0.001. Exact P values are available in Source Data file.

Fig. 2. Serum cytokine, soluble cytokine receptor, chemokine, and growth factor profiles in non-ICU and ICU COVID-19 patients.

a Heat-map representing the mean serum cytokine levels detected in healthy subjects (N = 450), non-ICU (N = 55) and ICU (N = 43) patients. Blue-to-yellow color code represents low-to-high average cytokine levels. Cytokine level similarities are represented by a dendrogram constructed by hiearachical clustering. b Levels of cytokines (IL-1β, IL-6, IL-10, and IL-15), cytokine receptor (IL-1RA), chemokines (CCL2, CCL4, CCL11, CXCL9, CXCL10, and CXCL13) and growth factors (NGF-β, EGF, HGH, LIF, PIGF-1, and VEGF-A) in healthy subjects (N = 450), non-ICU (N = 55) and ICU (N = 43) patients. Blue plots correspond to healthy subjects (HS), red plots corresponds to non-ICU patients and green plots correspond to ICU patients. Dotted line represents the upper normal values. Black stars indicate statistical significance between ICU or non-ICU patients and healthy subjects. Statistical significance (P values) was obtained using two-sided Kruskal–Wallis test, using a Bonferroni correction. *P < 0.05; **P < 0.01; ***P < 0.001. Exact P values are available in Source Data file.