We provide important insights into the pathophysiology of septic shock and the heterogeneity of biological perturbations within patients with septic shock. We compared the expression profiles of children with septic shock to mechanically ventilated controls and found upregulated expression of innate immune and neutrophil pathways and downregulation of adaptive immune pathways in children with septic shock. On further analysis, the expression profiles of patients with septic shock clustered around two subclasses with differential upregulation of innate immunity and downregulation of adaptive immunity. While both subclasses were clinically identified as “septic shock,” Subclass 1 was characterized by upregulation of innate immunity and biomarker profiles consistent with a hyperimmune response along with concomitant downregulation of both B and T cells and lesser T cell receptor diversity. This same subclass was associated with worse clinical outcomes, underscoring the importance of addressing sepsis heterogeneity. Subclass 2, while having the same clinical diagnosis, appeared biologically much more similar to our critically ill controls.

Our study suggests that patients with severe initial perturbations of the innate and adaptive immune systems, such as those in Subclass 1, are likely to have worse outcomes including higher PELOD scores, higher maximum inotrope usage, and longer intensive care unit and hospital stays, despite the lack of significant differences in illness severity or the pediatric risk of mortality (PRISM) score on initial presentation, indicating that conventional clinical scores alone are often are not sufficient to capture the pathophysiological complexity or outcomes of septic shock. An interesting clinical difference in our cohort is that Subclass 1 patients tended to be older and had modestly higher average blood glucose levels at baseline. [23]. IL-6 is known to contribute to hyperglycemia through insulin resistance, and multiple retrospective analyses have reported that hyperglycemia is associated with adverse outcomes [24].

Not only did Subclass 1 have upregulated genes related to neutrophil immunity compared to Subclass 2, but these patients also had significantly higher levels of proinflammatory cytokines (PAI1, IL6, IL8, ANG2, TREM1, and TNFR1), and a greater degree of organ dysfunction. Early unchecked innate immune-driven inflammation has been associated with a more profound degree of organ injury. Concomitantly, sepsis has also been shown to cause adaptive immunosuppression with a marked loss of T and B cells [25]. Subclass 1 had significant downregulation of genes related to T and B cell activity, and a lower percentage of CD4 T cells and B cells compared to Subclass 2. Severe T cell dysfunction, leading to decreased T cell cytotoxicity and T cell apoptosis in the setting of a high antigen load and elevated cytokines, is known to occur in septic shock [25].It is a source of debate whether it is the innate immune-driven hyperinflammation, adaptive immunosuppression, or a contribution of both is the driver of morbidity and mortality in sepsis. A recent study supported the latter, as they demonstrated that during sepsis, the proliferation of a large population of immature neutrophils inhibited the proliferation and activation of CD4 + T cells, and a subset of patients with higher frequencies of the immature neutrophils had poorer outcomes [26]. A potential consideration here is that the different sepsis subclasses could represent different chronological stages of the same underlying pathobiology. We attempted to mitigate this confounder by excluding patients who initiated vasoactive support > 72 h prior to enrollment.

In the adult population, at least five independent research groups have identified sepsis subgroups describing an adaptive immunity suppression phenotype with corresponding higher mortality such as Scicluna et al.’s MARS1 endotype [27], Davenport et al.’s SRS1 subgroup [28], and Sweeney et al.’s Inflammopathic subtype [29,30,31].

In the pediatric population, Hector Wong et. al. were the first to identify two pediatric sepsis endotypes, A and B. Endotype A was characterized by upregulation of innate immunity pathways and repression of pathways related to the adaptive immune system and glucocorticoid receptor signaling. However, the gene expression pattern that differentiates adult SRS groups was not enriched in the pediatric endotypes [32, 33]. This suggests that there may be differences between adults and children, highlighting the importance of studying sepsis specifically in the pediatric population.

Our findings and subclassification are consistent with previously published subclasses in the pediatric sepsis population. Our Subclass 1 was analogous to the Wong et al. pediatric Endotype A, which was identified in a separate pediatric cohort. Similar to Endotype A in Dr Wong’s cohort, Subclass 1 was also characterized by repression of the adaptive immune system related to T cell activation and had worse clinical outcomes than Subclass 2. We compared our gene list with the previously validated 100 gene signature that differentiated Endotype A and B, and found 12 of the 100 genes present in our list (GNAI3, PLCG1, CD3E, CD247, NCR3, ARPC5, ZAP70, FYN, SEMA6B, TLR8, CAMK2D, TLR8) [34]. In 2021, Muszynski et al. also identified a subclass of septic shock children with immunoparalysis with worse clinical outcomes [35]. They found that this subclass’s transcriptomic profiles demonstrated upregulated pathways in leukocyte extravasation, and downregulation in adaptive immunity pathways [35]. When comparing our Subclass 1 DEGs to Musznski’s immunoparalysis subclass, shared upregulated DEGs included COL17A1, LAMA2, FFAR3, RAP1GAP, XCR1, MMP27, and MMP8, and shared downregulated DEGs included KLRC1 and IL2RB. Thus, while Wong’s Endotype A, Muszynski’s immunoparalysis subclass, and our Subclass 1 are similar, challenges remain in identifying an appropriate panel of candidate genes out of the often large lists of DEGs that can be generalized across the pediatric population. More research in different pediatric cohorts is needed to develop a consensus subclassification system.

At the molecular level, T cell diversity plays a key role in the ability of the adaptive immune system to effectively mount a response to invading pathogens. Each TCR is made up of alpha (TCRA) and beta chains (TCRB) or delta (TCRD) and gamma (TCRG) chains. TCR diversity is generated through V(D)J recombination in the early stages of T cell maturation in the thymus and is critical to effectively recognizing antigen peptides. In the adult sepsis population, studies have shown that septic patients present with a marked decrease in TCR diversity after the onset of shock, which is associated with mortality and the development of nosocomial infections [36, 37]. However, there is a paucity of literature looking at TCR diversity in pediatric sepsis, and one study demonstrated that the TCR repertoires in adults and children are discrepant and thus difficult to directly compare [38]. Our study demonstrated that Subclass 1 patients demonstrated reduced TCR diversity and clonality, including decreased diversity of TCRA, TCRB, and TCRD, with associated worse clinical outcomes at the onset of septic shock.

In the adult population, there is evidence that in sepsis, a decrease in circulating B cells is associated with a poor prognosis [39]. Additionally, adult studies suggest that the B cell depletion is selective, and IL-10-producing B cells may actually increase and exacerbate immunosuppression [39]. In this pediatric study, Subclass 1 had decreased levels of B cells, but significantly higher levels of IL-10 with worse outcomes, similar to adult findings. The mechanism behind the depletion of B cells in sepsis is not well understood. One theory is that profound sepsis can impair bone marrow production leading to decreases in B cell numbers. Alternatively, some studies suggest sepsis signals can possibly trigger B cell apoptosis [39]. More studies are needed to investigate this phenomenon.

Our study is novel, as for the first time it not only provides a comprehensive evaluation of dysregulated pathways based on genome-wide differential gene expression but also enhances it with cell deconvolution and T and B cell receptor diversity estimations in the context of plasma biomarkers and clinical characteristics in septic shock to better characterize and sub-phenotype the biological perturbations in septic shock that are biologically plausible and clinically relevant in the pediatric population. A potential future direction of this subclassification study would be to create a classifier model that could eventually become useful in the clinical setting to classify pediatric septic shock patients and inform clinical decisions.

First, given that this was a retrospective analysis of a prospective pediatric clinical trial, a limitation was that a suitable case and control population were identified post-retrospectively. We assumed accurate and timely documentation of sepsis and initiation of inotropes for cases, and clinician-documented reasons for PICU admission and initiation of mechanical ventilation and negative blood cultures to create a control group. However even in the unlikely scenario that an underlying infection was missed in our controls, our cases required inotropes, indicating that on the clinical spectrum, sepsis in our cases was certainly much more severe. The use of mechanically ventilated controls ensured that our results were specific to septic shock and not just a result of generalized critical illness. Another limitation was that we could not differentiate between the different sources of sepsis each patient may have had. Finally, our sample sizes were relatively small. A larger sample size would have increased statistical power to detect clinically meaningful differences within each subclass; therefore, mortality was not a reported outcome in this study given that only three of the patients died within our cohort. Therefore, while Subclass 1 had worse outcomes, we cannot state that they had higher mortality.