Biological basis of critical illness subclasses: from the bedside to the bench and back again

Abstract

Critical illness syndromes including sepsis, acute respiratory distress syndrome, and acute kidney injury (AKI) are associated with high in-hospital mortality and long-term adverse health outcomes among survivors. Despite advancements in care, clinical and biological heterogeneity among patients continues to hamper identification of efficacious therapies. Precision medicine offers hope by identifying patient subclasses based on clinical, laboratory, biomarker and 'omic' data and potentially facilitating better alignment of interventions. Within the previous two decades, numerous studies have made strides in identifying gene-expression based endotypes and clinico-biomarker based phenotypes among critically ill patients associated with differential outcomes and responses to treatment. In this state-of-the-art review, we summarize the biological similarities and differences across the various subclassification schemes among critically ill patients. In addition, we highlight current translational gaps, the need for advanced scientific tools, human-relevant disease models, to gain a comprehensive understanding of the molecular mechanisms underlying critical illness subclasses.

Keywords: Critical illness subclass; Endotype; Phenotype; Precision medicine.

Fig. 1

An evolving paradigm of dysfunctional host response in critical illness. Critically ill patients exhibit a spectrum of dysfunctional pro- and anti- inflammatory responses at illness onset, which are subject to change over the longitudinal course of critical illness influenced by host–pathogen-environment related factors. Each column of figures represent time points across the course of critical illness (T0-T90 in days). Each row (E1-4) represents critical illness endotypes. Viewed through an alternate lens, the top two rows together represent a hyperinflammatory phenotype (P1) and the bottom two rows together represent a hypoinflammatory phenotype (P2). The uppermost row represents patients characterized by overexpression of the innate immune response (red) and repression of the adaptive immune response (blue). Given the lack of negative feedback by the adaptive immune system, these patients continue to have sustained and unchecked hyperinflammation. On the other end of the spectrum, are patients with overactivation of anti-inflammatory pathways resulting in severely immunosuppressed state. Further, patients may exhibit temporal subclass switching over the course of critical illness, including those in response to treatments or interventions received

Fig. 2

Intra-cellular signaling links immune response, metabolic state, and cellular fate. Pathogen recognition receptors (PRR) including toll like receptors (TLR), nucleotide oligomerization domain (NOD)-like receptors, and retinoic acid inducible gene-I-like receptors (RIG-I) detect extra- and intra-cellular pathogens. Activation of PRRs results in transcription of key pro-inflammatory pathways including nuclear factor kappa B (NFκB) and mitogen activated protein kinase (MAPK) signaling. This is facilitated by adaptor proteins including myeloid differentiation primary response 88 (MyD88) protein or toll-interleukin receptor domain–containing adaptor protein–inducing interferon-β (TRIF) – the latter being dominant in the host response to viral infections. Under hypoxic conditions, hypoxia inducible factor 1α (HIF1α) signaling triggers metabolic shift from oxidative phosphorylation towards glycolysis and cholesterol biosynthesis through mammalian target of rapamycin (mTOR) signaling. Although an efficient mechanism to maintain cellular function, a side effect of anerobic respiration is the production of reactive oxygen species (ROS). ROS induce a second hit and induce activation of inflammasome through nod like receptor pyrin domain 3 (NLRP3) protein resulting in activation of caspases. The latter serve to propagate the host response by cleaving protein precursors of inflammatory cytokines including interleukin-1 and 18 or serve to activate cell death pathways including apoptosis. Thus, the host immune response is inextricably linked to cellular metabolism and cellular fate. As detailed in the Table and illustrated in Fig. 4, although current subclassification schemes among critically ill patients sample the same set of key biological pathways they yield non-synonymous class outputs

Fig. 3

Summary of shifts in cellular states and landscape among critically ill patients. The left half of the image shows the innate arm while the right half shows the adaptive arm of the immune response. The top half of the figure shows molecular features that drive a pro-inflammatory state. Neutrophils release extracellular traps (NETs) resulting in NETosis that serve to facilitate phagocytosis of pathogens. Monocytes may be driven toward a pro-inflammatory M1 macrophage under the influence of NFκB, HIF1α, signal transducer and activator of transcription 1 (STAT1), and interferon regulatory factor 3 (IRF3). Innate immune antigen presenting cells (APC) engage T-helper (Th) cells through expression of human leukocyte antigen (HLA) DR. B-lymphocytes produce antibodies driving the humoral adaptive response. Th1 differentiation of helper cells results in secretion of pro-inflammatory cytokines including interferon gamma (IFNγ) and tumor necrosis factor alpha (TNFα) which further propagates the innate immune response. Finally, cytotoxic T cells release granzyme and result in cell death. The bottom half depicts features that drive an anti-inflammatory state in patients. There may be a shift towards immature innate cells including developing neutrophils and myeloid derived suppressor cells (MDSCs)– the early and late phases of critical illness, respectively. Monocytes may be polarized to an immunosuppressive M2 phase under the influence of interleukins -4, 10, and 13. There is decrease in HLA-DR expression and activation of co-stimulatory programmed cell death (PD-1L-PD-1) pathway. Differentiation towards a Th2 phenotype results in secretion of IL-4 and 13. Expansion of regulatory T cells result in IL-10, transforming growth factor beta (TGFβ) and vascular endothelial growth factor (VEGF) that are immunosuppressive and thought to promote tissue repair and remodeling. With prolonged illness, B- and T- lymphocytes may exhibit immune exhaustion with metabolic failure ultimately triggering cell death. As detailed in the Table and illustrated in Fig. 4, although current subclassification schemes among critically ill patients largely represent shifts across similar cellular states and landscape they yield non-synonymous class outputs

Fig. 4

a Similarities and differences in activation of key signaling pathways based on gene-expression data among critical illness subclasses according to scheme used. Key biological facets shown include pathogen recognition receptor signaling, pro-inflammatory cytokine signaling, hypoxia induced factor signaling, oxidative phosphorylation, antioxidant signaling, antigen-presenting cell signaling, T-helper lymphocyte activation, B-lymphocyte receptor signaling, and T-lymphocyte apoptosis. The green dots indicate activation while the red dots represent inactivation, as detailed in the referenced articles. The studies are grouped based on age and critical illness syndromes including sepsis, septic shock, and acute respiratory distress syndrome as detailed in the y-axis. b Conceptual overview of (1) current critical illness subclassification schemes sampling the proverbial 'biological pie' in varying slices and depth, (2) the need to move toward consensus critical illness endophenotypes through which the underlying molecular mechanisms can be unraveled, and (3) identification of subclass-specific molecular features or treatable traits that may be amenable to targeted therapeutic intervention

Fig. 5

Overview of bedside-to bench-to bedside approaches necessary to better understand drivers of biological mechanisms underlying critical illness subclasses and inform patient care. Starting from center top illustration in a clockwise direction A. Multi-compartment sampling among critically ill patients including whole blood including single cell suspensions, nasal brushings, humidified moisture exchange (HME) filter, tracheal aspirate, broncho-alveolar lavage, urine, and stool (in light pink), B. Multi-omics profiling including epigenome, genome, transcriptome, metabolome, and metagenome sequencing of biospecimens, including at single-cell resolution where feasible (in light purple). Exploratory multi-omic-data generated from human biospecimens require validation and experimental testing to gain mechanistic insights, necessitating biologically relevant disease models. C. Humanized animal models to capture biological heterogeneity including use of ‘knock in’ of human genes in place of murine analogues and biomarker-based stratification or sub-classification of experimental animals and may be used to recapitulate biology of human critical illness phenotypes in vivo. In addition, large animal models subject to environmental factors such as invasive mechanical ventilation are needed to improve disease modeling and testing efficacy of interventions. D. Human derived organoid models including those healthy donor and patient-specific induced pluripotent stem cells (iPSCs) treated with sera/plasma from critically ill patients may be potentially used to recapitulate biology of patient endotypes in vitro. The use of CRISPR-cas9 gene-editing technology is anticipated to facilitate a more rapid understanding of genes identified in cell- and organ-specific responses in critical illness. Moreover, these human relevant models can be used to understand compartment-specific responses in vitro, facilitate therapeutic drug screening, and drug monitoring. Importantly, concerted efforts are necessary to integrate biological data with other data streams back at the bedside. E. Integration of multi-modal data including vital sign trajectories, physiological and radiological data, point of care ultrasound, although not directly biological informative, are essential to integrate with point of care diagnostic assays that provide biological insights. Built-in artificial intelligence (AI) systems will be essential in the future to synthesize data streams and provide decision-making support to treating clinicians