Interferon-gamma driven elevation of CXCL9: a new sepsis endotype independently associated with mortality

Abstract

Background: Endotype classification becomes the cornerstone of understanding sepsis pathogenesis. Macrophage activation-like syndrome (MALS) and immunoparalysis are the best recognized major endotypes, so far. Interferon-gamma (IFNγ) action on tissue macrophages stimulates the release of the cytotoxic chemokine CXCL9. It was investigated if this mechanism may be an independent sepsis endotype.

Methods: In this cohort study, 14 patient cohorts from Greece, Germany and Italy were studied. The cohorts were 2:1 randomly split into discovery and validation sets. Sepsis was defined by the Sepsis-3 definitions and blood was sampled the first 24 h from meeting the Sepsis-3 definitions. Concentrations of IFNγ, CXCL9, IP-10 (IFNγ induced protein-10), soluble CD163 and ferritin were measured. The endotype of IFNγ-driven sepsis (IDS) was defined in the discovery set as the combination of a) blood IFNγ above a specified cut-off associated with the minimal risk for immunoparalysis (defined as ≥8000 HLA-DR receptors on CD45/CD14-monoytes); and b) increase of CXCL9. Results were compared to the validation set.

Findings: 5503 patients were studied; 3670 in the discovery set and 1833 in the validation set. IDS was defined as IFNγ more than 3 pg/ml and CXCL9 more than 2200 pg/ml. The frequency of IDS in the discovery set was 19.9% (732 patients; 95% confidence intervals-CIs 18.7-21.3%) and in the validation set 20.0% (366 patients; 95% CIs 18.2-21.9%). Soluble CD163, a marker of macrophage activation, was greater in IDS and IDS had features distinct from MALS. The mortality in IDS patients was 43.0% (315 patients; 95% CIs 39.5-46.6%) in the discovery set and 40.4% in the validation set (148 patients; 95% CIs 35.5-45.5%) (p = 0.44 compared to patients of the discovery set). IDS was an independent risk factor for death in the presence of other endotypes, severity scores and organ dysfunctions of the multivariate model [hazard ratio 1.71 (95% CIs 1.45-2.01) in the discovery set and 1.70 (95% CIs 1.34-2.16) in the validation set]. Decreases of IFNγ and CXCL9 blood levels within the first 72 h were associated with better outcome.

Interpretation: IDS is a new sepsis endotype independently associated with unfavorable outcome.

Funding: Hellenic Institute for the Study of Sepsis; Horizon 2020 project ImmunoSep; Swedish Orphan BioVitrum AB (publ) and German Federal Ministry of Education and Research.

Keywords: CXCL9; Interferon-gamma; Macrophages; Outcome; Sepsis.

Fig. 1

Development of the IFNγ-driven sepsis (IDS) endotype in the discovery set. A) Bar graphs of 28-day mortality in patients with IFNγ more than 3 pg/ml and IFNγ 3 pg/ml or less. The p-value of comparison by the Fisher's exact test is provided. B) Distributions of the absolute count of HLA-DR receptors on CD45/CD14-monocytes in patients with IFNγ more than 3 pg/ml and IFNγ 3 pg/ml or less. Lines overlying the distributions represent medians and interquartile ranges. The p-value of comparison by the Mann–Whitney U test is provided. C) Distributions of sCD163 in patients with IFNγ more than 3 pg/ml and IFNγ 3 pg/ml or less. Lines overlying the distributions represent medians and interquartile ranges. The p-value of comparison by the Mann–Whitney U test is provided. D) Distributions of CXCL9 in survivors and non-survivors. This comparison involves only the patients with IFNγ levels more than 3 pg/ml. Lines overlying the distributions represent medians and interquartile ranges. The p-value of comparison by the Mann–Whitney U test is provided. Abbreviations: CI, confidence interval; IDS, IFNγ-driven sepsis; IFN, interferon; n, number of patients.

Fig. 2

Distribution of endotypes in the discovery and the validation sets of patients. Patients are classified into four endotypes, MALS, IDS, Low IFNγ and High IFNγ without CXCL9 increase. Patients with low IFNγ (i.e. blood IFNg ≤3 pg/ml) most likely represent patients with sepsis-induced immunoparalysis. The overlaps between MALS and IDS in each set are provided. In the discovery set, 126 patients with MALS have IFNγ ≤3 pg/ml. In the validation set, 65 patients with MALS have IFNγ ≤3 pg/ml. Abbreviations: IDS, IFNγ-driven sepsis; IFN, interferon; MALS, macrophage activation-like syndrome; n, number of patients; w/o, without.

Fig. 3

The impact of the endotype of IFNγ-driven sepsis on mortality the first 28 days. Kaplan–Meier survival curves in the discovery (A) and in the validation (B) sets according to sepsis endotype are shown. The values of the indicated log-rank comparisons and of the respective p-values after Bonferroni corrections for multiple comparisons are provided. In the analysis, patients overlapping in classification with both IDS and MALS are analyzed as IDS. Abbreviations: IDS, IFNγ-driven sepsis; IFN, interferon; MALS, macrophage activation-like syndrome; n, number of patients; w/o, without.

Fig. 4

Mortality risk by the presence of the IFNγ-driven sepsis endotype in patient sub-groups. The HRs of the impact of IDS on 28-day mortality coming from Cox regression analyses within each indicated subgroup are provided. The p-values of significances are shown. Abbreviations: CAP, community-acquired pneumonia; CI, confidence interval; HAP, hospital-acquired pneumonia; HR, hazard ratio; ICU, Intensive Care Unit; VAP, ventilator-associated pneumonia.

Fig. 5

Changes of IFNγ and CXCL9 over the first 72 h and 28-day outcome. Measurements of blood IFNγ and of CXCL9 were repeated for 84 patients after 72 h. A) Paired changes of IFNγ between the two time points, separately for 28-day survivors and 28-day non-survivors. The p-values of paired comparisons by the Wilcoxon's signed rank test are provided. B) Paired changes of CXCL9 between the two time points, separately for 28-day survivors and 28-day non-survivors. The p-values of paired comparisons by the Wilcoxon's signed rank test are provided. C) 28-day mortality for patients with less than 50% decrease of blood IFNγ and for patients with ≥50% decrease of blood IFNγ within the first 72 h. The 50% cut-off is defined as the Youden index following ROC curve analysis of the relative changes of IFNγ to predict 28-day mortality (area under the curve 0.69; 95% confidence intervals 0.55 to 0.82; p = 0.016). D) 28-day mortality for patients with less than 30% decrease of blood CXCL9 and for patients with ≥30% decrease of blood CXCL9 within the first 72 h. The 30% cut-off is defined as the Youden index following ROC curve analysis of the relative changes of CXCL9 to predict 28-day mortality (area under the curve 0.65; 95% confidence intervals 0.525 to 0.82; p = 0.046). Abbreviations: CI: confidence interval; IFN; interferon; OR: odds ratio; n: number of patients; ROC, receiver operator characteristics.