Surviving septic patients endotyped with a functional assay demonstrate active immune responses

Abstract

Introduction: Sepsis is a complex clinical syndrome characterized by a heterogenous host immune response. Historically, static protein and transcriptomic metrics have been employed to describe the underlying biology. Here, we tested the hypothesis that ex vivo functional TNF expression as well as an immunologic endotype based on both IFNγ and TNF expression could be used to model clinical outcomes in sepsis patients.

Methods: This prospective, observational study of patient samples collected from the SPIES consortium included patients at five health systems enrolled over 17 months, with 46 healthy control patients, 68 ICU patients without sepsis, and 107 ICU patients with sepsis. Whole blood was collected on day 1, 4, and 7 of ICU admission. Outcomes included in-hospital and 180-day mortality and non-favorable discharge disposition defined by skilled nursing facility, long-term acute care facility, or hospice. Whole blood ELISpot assays were conducted to quantify TNF expression [stimulated by lipopolysaccharide (LPS)] and IFNγ expression (stimulated by anti-CD3/CD28 mAb), which were then used for assignment to one of four subgroups including an 'immunocompetent', 'immunosuppressed endotype', and two 'mixed' endotypes.

Results: Whole blood TNF spot-forming units were significantly increased in septic and CINS patients on days 4 and 7 compared to healthy subjects. In contrast, TNF expression per cell on days 1, 4, and 7 was significantly lower in both septic and critically ill non-septic (CINS) patients compared to healthy subjects. Early increases in total TNF expression were associated with favorable discharge disposition and lower in-hospital mortality. 'Immunocompetent' endotype patients on day 1 had a higher proportion of favorable to non-favorable discharges compared to the 'immunosuppressed' endotype. Similarly, 'immunocompetent' endotype patients on day 4 had a higher in-hospital survival compared to the 'immunosuppressed' endotype patients. Finally, among septic patients, decreased total TNF and IFNγ expression were associated with 180-day mortality.

Conclusions: Increased ex vivo whole blood TNF expression is associated with improved clinical outcomes. Further, the early 'immunocompetent' endotype is associated with favorable discharge and improved in-hospital and 180-day survival. The ability to functionally stratify septic patients based on blood cell function ex vivo may allow for identification of future immune modulating therapies.

Keywords: IL-6; critical illness; late mortality; prediction modeling; procalcitonin.

Figure 1

LPS-stimulated TNF expression as determined by ELISpot spot forming units (SFU) and spot size (SS) in Sepsis, and CINS patients at timepoints 1, 2, and 3 following ICU admission, and in healthy control subjects. Values represent mean and individual subject responses for TNF SFU (A) and SS (B) at T1, T2, and T3. T1: healthy cohort n=45, CINS n=67, septic n=103; T2: healthy cohort n=45, CINS n=55, septic n=90; T3: healthy cohort n=45, CINS n=45, septic n=77. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, as determined by Kruskal-Wallis ANOVA and post hoc analyses using Dunn’s test. SFU, spot-forming units.

Figure 2

LPS-stimulated total TNF expression at timepoint 2 is associated with in-hospital mortality. (A). Comparison of in-hospital mortality of the septic patient cohort. (B) Area under the Receiver Operator Curve (AUROC) for total LPS-stimulated TNF expression in differentiating in-hospital mortality. **P < 0.01, as determined by unpaired Mann-Whitney test. AUROC=0.7572, p=0.0059.

Figure 3

Total LPS-stimulated TNF expression at timepoint 1 is associated with discharge. (A). Comparison of favorable and non-favorable discharge patient populations in the septic patient cohort. (B) Area under the Receiver Operator Curve (AUROC) for total LPS-stimulated TNF expression in differentiating favorable vs. unfavorable discharge. *P < 0.05, as determined by unpaired Mann-Whitney test. AUROC curve=0.6218, p=0.0345.

Figure 4

Area under the Receiver Operator Curve (AUROC) using both TNF⁺ and IFNγ⁺ on day 4 following ICU admission in differentiating 180-day mortality. A multiple Cox regression to predict survival time using both total TNF and IFNγ was conducted to construct a combined metric. When predicting 180-day mortality using the constructed metric, AUROC curve was 0.8133, 95% confidence interval was 0.7173-0.9094. p=0.0001.