Skip to main content
Download PDF

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

Critical Care volume 28, Article number: 186 (2024) Cite this article

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.

Graphical abstract

Background

Critical illness syndromes including sepsis, acute respiratory distress syndrome (ARDS), and acute kidney injury (AKI) affect pediatric and adult patients admitted to intensive care units (ICUs) across the world 1,1,2,3,4,5, and are associated with high mortality. Moreover, survivors of critical illness syndromes remain at risk of long-term health consequences including chronic debilitation 6, technology dependence 7, and late death 8,1,2,3,4,5,6,7,8,9,10. Despite this burden of disease, care for such patients remains largely limited to antibiotics and intensive organ support. One-size-fits-all approaches to modulate the host response among critically ill patients have been met with repeated failure in clinical trials 11. This lack of efficacy of drugs has been attributed to administration of therapies in unselected populations and due to unaccounted variation among patients 12, 13. Thus, identification of the right drug in the right patient at the right time remains a major challenge.

Era of precision critical care medicine

Precision medicine approaches seek to address these challenges by identifying subsets of critically ill patients based on shared features including clinical, laboratory, biomarker, or ‘omic’ data 13. Broadly, these include (1) prognostic enrichment tools that seek to identify groups of patients based on risk of outcomes such as mortality, and (2) predictive enrichment tools that seek to identify groups of patients based on shared biological pathways, which may be amenable to therapeutic intervention 14. While numerous promising subclassification schemes have emerged within the previous two decades, each with prognostic, predictive implications, or both, few have translated into the clinical realm. Moreover, we lack a comprehensive understanding of underlying disease mechanisms, a challenge further hampered by the limited translation of current disease models used to study critical illness pathobiology. Thus, much progress needs to be made in identifying treatable traits underlying critical illness subclasses. 13

In this state-of-the-art review, we summarize subclassification schemes that have demonstrated reproducibility in identifying biologically distinct subgroups of patients. While the relevance of critical illness subclasses have been previously considered elsewhere 15, 16, we seek to highlight key molecular similarities and differences across the subclassification schemes. Finally, we highlight current knowledge gaps, translational approaches, and disease models that may help augment our mechanistic understanding of subclass-specific biology and accelerate progress toward drug repurposing as well as the discovery of de novo targeted therapies.

Evolving understanding of the host response in critical illness

The classical paradigm of dysfunctional host responses among critically ill patients imply a sequential activation of the host innate and adaptive immune response resulting in a systemic inflammatory response syndrome (SIRS) followed by a compensatory anti-inflammatory response (CARS) syndrome 17, 18. However, it is increasingly understood that both the innate and adaptive arms of the host immune response may contribute to pro- and anti-inflammatory responses relatively early in the course of illness with significant crosstalk between them. An evolving paradigm of dysfunctional host response among those critically ill suggests that patients manifest a spectrum of maladaptive responses at illness onset that are subject to change over the course of disease under the influence of myriad pathogen-host-environment related factors. This concept is illustrated in Fig. 1.

Fig. 1
figure 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

Full size image

Recognition of pathogen- and damage- associated molecular patterns (PAMPs and DAMPs) initiate and propagate the host response in critical illness. Although not the primary objective of this review, we provide an overview of the perturbations in intra-cellular pathways linking inflammatory signaling, metabolic state, and cellular fate central to the dysregulated host response among critically ill patients in Fig. 2. Further, the shifts in immune states and cellular landscape among critically ill patients that contribute to pro- and anti-inflammatory responses are summarized in Fig. 3. Our objective here is to provide the reader the necessary context to understand the biological basis of critical illness subclasses discussed herein.

Fig. 2
figure 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

Full size image
Fig. 3
figure 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

Full size image

Subclassification schemes in critical illness

It is worth noting that the subclassification schemes that have emerged sample various facets of the host response–the proverbial ‘biological pie’– in differing permutations and combinations and at varying depth. Given that gene-expression and clinico-biomarker models using peripheral blood have been most frequently utilized to identify critical illness subclasses and likely provide the most comprehensive understanding of biological mechanisms thus far, we have chosen to focus our attention on these approaches. In Table 1, we have compiled biological similarities and differences across numerous subclassification schemes among critically ill patients. Key studies have been summarized in the following paragraphs.

Table 1 Summary of critical illness subclassification schemes with gene-expression data used to comprehensively characterize biological basis of subclasses
Full size table

Gene-expression based endotypes

Wong and colleagues first used whole blood RNA microarrays to identify the gene-expression signatures associated with mortality among children with septic shock 19, 20. Subsequently, among the most differentially expressed genes (DEGs), 100 subclass defining genes broadly reflective of the adaptive arm of the host immune response, T-cell function, and glucocorticoid signaling were used to identify patient subclasses through unsupervised clustering analyses 21, 22. Subsequently, two subclasses or endotypes A and B were validated, with patients belonging to endotype A being characterized by repression of the adaptive immune and glucocorticoid signaling, relative to patients with endotype B. 23 Of note, endotype A was associated with a nearly threefold higher odds of mortality 21. Further, patients with endotype A who received adjunctive corticosteroids had a fourfold higher rate of ICU mortality compared to endotype B who received corticosteroids 23. In follow up studies, Grunwell et al. identified that 11,630 gene probes were differentially regulated between patient endotypes A and B 24. Targeted analyses of nuclear factor erythroid-related factor 2 (Nrf2), a transcription factor that regulates expression of anti-oxidant genes, revealed greater downregulation among patients with endotype A. More recently among children with septic shock in secondary analyses of the Coagulation and Fibrinolysis in Pediatric Insulin Titration (CAF-PINT) study, Yang et al. identified two septic shock subclasses using bulk mRNA sequencing 25. Although the mortality among patients included in this cohort was low, Subclass 1 had higher burden or organ dysfunction compared to Subclass 2 and characterized by upregulation of the innate immune and downregulation of the adaptive immune pathways, with evidence of systemic inflammation and endothelial injury based on plasma protein biomarkers. Deconvolution analyses of cellular composition suggested a decrease of CD4 T- and B-lymphocytes, and a lower diversity of T-cell receptors (TCR). 25

Several efforts to identify gene-expression endotypes have been undertaken among adults. Davenport et al. conducted peripheral blood leukocyte expression profiles among patients with community acquired pneumonia with sepsis recruited through the U.K. Genomic Advances in Sepsis (GAinS) study 26. They identified two distinct sepsis response signatures (SRS), where SRS1 was associated with an immune-suppressed phenotype with downregulation of pathogen recognition receptors (PRRs), decreased expression of human leukocyte antigen (HLA) class II on antigen presenting cells (APCs), and T-cell exhaustion, relative to those with SRS2 endotype. Patients with SRS1 had a two to threefold higher odds of mortality, relative to those with SRS2 membership. In secondary analyses of the vasopressin, norepinephrine, and steroids (VANISH) trial, Antcliffe et al. identified that patients with the immunocompetent SRS2 endotype, had an ~ eightfold higher mortality with receipt of randomized hydrocortisone relative to patients designated as SRS1. 27 The authors speculated that among patients with SRS2, receipt of corticosteroids may have contributed to greater downregulation of HLA-DR expression contributing to a detrimental response to corticosteroids in this subgroup. The authors took the illuminating step of integrating whole genome genotyping with gene-expression data allowing them to identify expression quantitative trait loci (eQTL) associated with endotypes 26. Interleukin 18 receptor (IL18RAP) and chemokine receptor 1 (CCR1) were upregulated, while CCR3 was downregulated among patients with SRS1. Moreover, hypoxia inducible transcription factors HIF1α (HIF1A) and HIF2α (EPAS1), mammalian target of rapamycin (mTOR) pathway a key determinant of cell metabolism and fate (Fig. 2), were among the most upregulated genes with evidence of cis-eQTL among patients with SRS1.

Sweeney et al. pooled pediatric and adult datasets across 14 discovery cohorts comprising 700 patients and utilized meta-clustering approach to identify 3 sepsis endotypes, which they termed Inflammopathic, Coagulopathic, and Adaptive. 28 The Inflammopathic subclass was characterized by pathogen recognition receptor signaling, pro-inflammatory cytokine signaling and complement activation and had ~ 30% mortality. The Coagulopathic subclass was characterized by pathways involved in platelet degranulation, glycosaminoglycan-, heparin-, and fibrinogen-binding, and broadly reflective of coagulation cascades and had ~ 25% mortality. Finally, the Adaptive subclass was characterized by activation of pathways related to antigen presentation activity, T-cell receptor binding, activation of lymphocytes, and had a mortality of ~ 8%. This group deployed a 33-gene classifier to assign endotypes which has demonstrated reproducibility in when applied in secondary analyses of observational cohort of patients with surgical sepsis 29, COVID19 patients 30, and heterogeneous treatment effects in a randomized trial assessing Outcomes of Metabolic Resuscitation using Ascorbic acid, Thiamine, and Glucocorticoids in the Early Treatment of Sepsis (ORANGES) trial 31, and most recently interleukin-1 receptor antagonist in the SAVE-MORE trial among patients with COVID19. 32

Scicluna et al. identified 4 endotypes among adults with sepsis enrolled in the prospective observational cohort by the molecular diagnosis and risk stratification (MARS) consortium based in the Netherlands 33. The Mars1 endotype was associated with higher hazard of mortality relative to the other 3 endotypes, and characterized by repression of both innate and adaptive genes including PRR signaling, nuclear factor κB (NFκB) signaling, APC signaling, T-helper cell differentiation and TCR signaling (Figs. 2 and 3). Clinical outcomes were similar among Mars 2–4 endotypes. This group developed a 140-gene classifier which demonstrated reproducibility. Of note, Mars3 endotype showed a significant association with the immunocompetent SRS2 and Adaptive endotypes, described previously. 26, 34

Transcriptomic studies among patients withs ARDS focused on identifying subclasses have been limited 35. Yehya et al. identified 3 subclasses named the Children’s Hospital of Philadelphia ARDS Transcriptomic Subtypes (CATS) among 96 pediatric patients using whole blood mRNA. CATS1 subclass was characterized by persistent hypoxemia and ~ 30% mortality. Nearly half of the patients assigned to CATS2 had an immunocompromised status but rapidly resolved hypoxemia and ~ 25% mortality. Patients among CATS3 had the lowest mortality ~ 8%. CATS1 was characterized by T helper cell differentiation and programmed cell death (PD-1) signaling. CATS2 showed enrichment for complement activation and CATS3 had repression of TCR signaling. Pathway analyses suggested activation of upstream regulators of inflammatory cytokines among CATS1 and repression among CATS3 patients. Receipt of and response to corticosteroids were not reported in this cohort.

Latent class or clustering based phenotypes

In seminal work, Calfee et al. leveraged latent class analyses using clinical, laboratory, and biomarker data to identify two subphenotypes of ARDS among patients recruited through the ARDS network trials 36. Patients with higher probability of membership among phenotype 2, identified as hyperinflammatory based on higher plasma concentrations of proinflammatory cytokines and markers of endothelial activation including IL-6, IL-8, soluble tumor necrosis factor receptor-1 (sTNFr1) and plasminogen activating inhibitor-1 (PAI-I) and lower concentrations of protein C, had a higher prevalence of sepsis and worse clinical outcomes relative to those without this phenotype. Dahmer et al. showed reproducibility and prognostic utility of this approach among children with ARDS in secondary analyses of the Randomized Evaluation of Sedation Titration for Respiratory Failure (RESTORE) trial 37; similar findings have been made in other observational cohorts 38. Of note, latent profile phenotypes have demonstrated interaction with several interventions 36, 39 and simvastatin therapy in the HARP-2 trial, on clinical outcomes 40. Using a different set of clinical variables and fewer biomarker inputs, Sinha et al. identified two phenotypes among adults with COVID19 with treatment with corticosteroids being associated with improved survival among those with hyperinflammatory phenotype and higher mortality among those with a hypoinflammatory phenotype. 41

Studies conducted by investigators in the MARS consortium have reproduced the findings of Calfee and colleagues. Bos et al. used cluster analyses to identify a reactive and uninflamed phenotype, with worse outcomes observed among those with a reactive phenotype 42. Assessment of leukocyte-expression profiles revealed genes were broadly representative of innate immune signaling and neutrophil activation 43. Pathway analyses revealed that that oxidative phosphorylation pathways, cholesterol biosynthesis, and antioxidant signaling through Nrf2 were upregulated among patients with a reactive phenotype (Fig. 2). In contrast mitogen activated protein kinase (MAPK) pathways, T-cell signaling, and apoptosis were upregulated among patients with an uninflamed phenotype. Heijnen et al. identified that these phenotypes maintained their biological distinctiveness among mechanically ventilated patients irrespective of whether they met ARDS criteria, and showed similarity with outputs of latent class analyses. 44

Sinha and colleagues recently published on molecular phenotypes among adults with sepsis 45. In two prospective observational cohorts, the authors established the prognostic utility of latent class phenotypes among adults with sepsis with high concordance with phenotypes among patients with ARDS. When re-analyzing data from the PROWESS-SHOCK study 46, the authors noted a differential response to recombinant human activated protein C between phenotypes, with patients with a greater probability of a hyperinflammatory phenotype demonstrating a beneficial effect with receipt of therapy. Of note, there was no evidence of differential responses among phenotypes based on vasopressor choice when re-examining results of the Vasopressin in Septic Shock Trial (VASST); these results are concordant with findings with similar studies. 47, 48

Recent studies by Neyton et al. among adults with sepsis revealed overexpression of genes representative of the innate arm among those with a hyperinflammatory phenotype and those of the adaptive arm among patients with an hypoinflammatory phenotype 49. Patients in the VANISH study were subsequently assigned to the phenotypes based on a gene-expression classifier and the authors identified that patients with hypoinflammatory phenotype overlapped with the immune competent SRS2 endotype and demonstrated higher mortality with receipt of hydrocortisone compared to placebo. Such an effect was not noted among patients with the hyperinflammatory phenotype. Sinha et al. conducted transcriptomic analyses among a subset of patients assigned to ARDS phenotypes enrolled in the re-evaluation of the systemic early neuromuscular blockade (ROSE) trial yielding similar results 50. Of note, temporal shifts in gene-expression identified among subclasses influenced by patient survival, rather than phenotypic assignment, highlighting the dynamic nature of critical illness and molecular complexities involved.

Finally, Bhatraju et al. used latent class analyses using biomarkers of inflammation and endothelial activation to identify two subphenotypes of AKI- AKI-SP1 and AKI-SP2 in independent discovery and validation cohorts 51. The AKI-SP2 subgroup characterized by elevated Angiopoietin-2/Angiopoietin-1 ratio had a ~ 2.5-fold higher risk of 28-day mortality and increased risk of renal non-recovery, relative to patients with AKI-SP1. In secondary analyses of the VASST trial, patients with AKI-SP1 had a differential response to vasopressors with lower 28- and 90-day mortality among those receiving vasopressin compared to norepinephrine. However, this was not the case with the high-risk AKI-SP2 subphenotype. Studies by Wiersema et al. in secondary analyses of the Finnish AKI (FINNAKI) study yielded similar subclasses with prognostic utility 52. It is worth noting that the biomarkers used for phenotyping patients are not specific to AKI and generalizable across critical illness syndromes.

Corroboration in pre-clinical models

Few studies have attempted to develop animal models capable of recapitulating biological heterogeneity noted among critically ill patients. In cecal ligation puncture (CLP) model of sepsis, Wong et al. identified that murine analogues of pediatric sepsis mortality risk biomarkers (PERSEVERE) could be utilized to risk-stratify experimental mice 53. In an analogous approach, Seymour et al. have used biotelemetry-enhanced CLP coupled with latent class analyses to identify two phenotypes among experimental mice, with one of the classes demonstrated shorter time to deterioration and greater concentrations of biomarkers of systemic inflammation 54. Interestingly, in a subset of mice that were randomized to receive immediate vs. delayed fluid resuscitation and antibiotics; only the sicker subset demonstrated improvements with therapy. To model complexity, investigator groups have utilized large animal models. Millar et al. have developed an ovine model of ARDS using a two-hit approach 55. Importantly, animals received intensive care. A priori clustering analysis was used to identify two phenotypes, with one of the phenotypes characterized by higher plasma concentrations of IL-6, IL-8, and IL-10 and manifesting similar features as ARDS patients with a hyperinflammatory phenotype. Targeted study of gene-expression revealed that this phenotype was characterized by overexpression of neutrophil genes also implicated in the human host response 56. In subsequent studies the group identified that randomization to corticosteroids was associated with a benefit only among the ovine hyperinflammatory group 57. A few important limitations worth considering including the fact that experimental approaches described may merely account for variance in the procedures to induce critical illness in animals. More importantly, models either use animals with a homogenous genetic background or those with knockout of select genes. While the latter are useful to understand the mechanism of individual genes, they do not reflect the tremendous genetic diversity among patients that vitally contribute to patient-, organ-, and cell-specific heterogeneity, thus inherently limiting their translational potential. Lastly, to the best our knowledge, no in vitro models have thus far been developed to recapitulate genetic heterogeneity nor the organotypic responses observed among critically ill patients –an area of interest emphasized by the U.S National Advisory General Medical Sciences Council (NAGMSC) working group on sepsis. 58

Current translational gaps

The identification of subclasses with prognostic and predictive relevance is a commendable feat. Yet, we remain in the relative infancy of our understanding of subclass-specific disease mechanisms. As an analogy, the endotypes and phenotypes described thus far may serve as lenses of a compound microscope through which we are able to begin to grasp a more granular understanding of the mechanisms underlying critically ill patients. However, there is an urgent need for implementation of advanced scientific tools and disease models to open new vistas for the discovery of biological drivers underlying subclasses. Further, it remains possible that there exist relevant yet undiscovered biological pathways that are not restricted by the current framework for critical illness subclasses. Accordingly, development of both top-down and bottom-up approaches, which consider current subclasses, are necessary to better understand causal biological pathways underlying critical illness.

Towards consensus endophenotypes

In Fig. 4a, activation (green) and inactivation (red) of intra-cellular signaling pathways identified through gene expression studies are shown according to critical illness syndrome and the subclassification scheme used; evidently some approaches are more congruent than others. In a recent study van Amstel et al. compared class outputs of several critical illness subtyping schemes within the same set of patients and identified relatively low to moderate overlap between clinical, biomarker, and transcriptomic data-based approaches 59. Furthermore, several studies have identified that integrated subclassification approaches, for example those that combine biomarker and transcriptomic approaches, may be more informative than using a single approach alone 59,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61. Moreover, it is established that nearly half of the patients switch between gene-expression endotypes assigned on day 1 by day 3 among adults 62 and children 63. In contrast, limited data among adult ARDS phenotypes suggest that ~ 95% of patients maintained original class assignment. It remains possible that given the greater molecular depth afforded by gene-expression, in comparison with clinico-biomarker approaches, that subclass switching is more readily apparent when using the former approach.

Fig. 4
figure 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

Full size image

Given these challenges, achieving consensus approaches to identify clinically relevant, biological informative, and temporally characterized subclasses agnostic of underlying syndromic assignments and host developmental age is crucial, as illustrated in Fig. 4b. Towards this end, efforts towards identifying consensus endotypes are underway through the SUBtyping in SePsis And Critical illnEss (SUBSPACE) consortium – a partnership between academia and industry to identify consensus gene-expression endotypes among critically ill adults and children. Similar outputs may be expected from the recently formed ARDS, Pneumonia, Sepsis (APS) consortium—funded by the U.S National Institute of Health. Forward thinking adaptive study designs as with the Precision medicine Adaptive Network platform Trial in Hypoxemic acutE respiratory failuRe (PANTHER) 64 and the treatable traits in acute critical illness (TRAITS) study 65 are likely to further inform the biology of critical illness subclasses. Finally, temporally sampled biospecimens from both prospective cohort and randomized clinical trials before and after intervention hold tremendous potential to inform efficacy of biological response to therapies among critical illness subclasses 66. Several challenges are worth considering. (1) Clinical phenotyping alone in the absence of biologically informative inputs may limit reproducibility of subclasses. (2) Inclusion of appropriate non-critically ill controls is essential, especially to gain insights into critically ill patients who have relatively repression of protein or genes in comparison with the sickest subset of patients. (3) Sampling the host response at multiple-time points is necessary to characterize longitudinal disease trajectories 67 among patients with emphasis on understanding the molecular drivers that predispose patients toward these trajectories. (4) Given the statistical power necessary, investment in federated databases that allow for pooled analyses of a large sample size of patients with rich clinical annotations will be needed. (5) Finally, prioritization of subtyping methods, especially in the absence of obvious overlaps between class outputs, with preferential weighting of approaches that yield reproducible subclasses with demonstrated heterogeneity in treatment effect to interventions and therapies will be crucial.

Multi-compartment sampling of the host response

Progress and challenges in profiling the dysregulated immune response in critical illness has been excellently summarized elsewhere 68. A major limitation has been that most studies among critically ill patients have focused on the peripheral blood compartment in isolation and not considered the compartmentalized effects of the host response among patients 69. Heijnen et al. profiled a limited set of inflammatory cytokines and lung microbiota using mini-BAL specimens among 26 patients sampled among patients with reactive and uninflamed phenotypes identified based on blood biomarkers 44. Although limited by sample size, the authors did not find significant differences between phenotypes leading them to conclude that the blood biomarker-based designation of phenotypes may not inform the alveolar compartment. Given the potential for both concordant and discordant effects of therapies on systemic and local pro- and anti-inflammatory responses, concerted efforts are necessary to sample multiple compartments among patients. To this end, some groups have had success with use of minimally invasive sampling to sample the tissue-level host response. 70, 71

Multi-omic profiling to unravel subclass-specific biology

As opposed to Mendelian diseases that arise from very rare genetic variants with a high penetrance, critical illnesses are inherently complex with several relatively common genetic variants with low penetrance acting in concert to influence patient outcomes. This polygenic nature of critical illness 72 has meant that while innumerable studies have characterized the roles of individual genes and related variants among critically ill patients, we are yet to identify approaches to synthesize this complexity and tailor interventions based on knowledge of patient genotype or haplotype. Thus, systematic efforts to identify the influence of the host genome on subclass-specific biology is imperative. Further, given that sampling different omic layers provides orthogonal evidence, integrated multi-omic analyses are likely to provide greater confidence in subclass-specific biological pathways identified. Lastly, few small-scale studies have evaluated epigenome wide shifts among critically ill patients 73, 74. Given the increasing appreciation of the role of epigenetic regulation in the coordination of orchestrated transcriptomic responses 75, simultaneous study of epigenome and transcriptome both at a patient- and cellular-levels may allow for identification of gene-regulatory networks and drivers of subclass-specific biology.

Dawn of single-cell omics in critical illness pathobiology

Advances in single cell RNA sequencing (scRNAseq) approaches have increasingly allowed for a greater understanding of molecular changes among cellular subsets among critically ill patients. Reyes et al. published a comprehensive atlas of the immune landscape among critically ill adults and identified a unique CD14 + monocyte state specific to septic patients 76. Notably, Kwok et al. recently published a whole-blood multi-omic atlas and identified that developing neutrophils and emergency granulopoiesis were key drivers of an extreme endotype among critically ill adults with sepsis 77. Lastly, numerous well designed multi-omics studies 78 including those by the COVID-19 Multi-Omic Blood Atlas (COMBAT) consortium 79 have shed light on organ- and cell-specific molecular perturbations among critically ill patients. Integration of such advanced sequencing technologies with conventional approaches to endophenotype patients holds potential to inform subclass-specific biology. It remains yet unknown whether such knowledge can be harnessed to tailor cell-subset specific precision therapies.

Understanding host–pathogen and microbiome interactions

Advances in scientific methods have enhanced the ability to identify pathogens which contribute to critical illness but often not captured by conventional culture techniques. Using 16s ribosomal RNA sequencing Dickson et al. identified that in humans and experimental sepsis, gut-specific bacteria were more common and abundant in bronchoalveolar lavage fluid and correlated with degree of systemic inflammation, relative to healthy controls 80. Further alveolar inflammation was correlated with perturbations in the lung microbiome. Subsequent studies have identified that among critically ill patients enrolled through the Biomarker analyses in Septic ICU patients (BASIC) - a subset of the MARS consortium, patients with increased lung bacterial burden had fewer ventilator free days and associated with development of ARDS. Of note, the presence of gut-associated bacteria including Lachanospiraceae and Enterobacteriaceae in the lung were predictive of the worst outcomes. 81

Kalantar et al. deployed metagenomic next generation sequencing (mNGS) to distinguish patients with and without critical illness due to infectious etiologies and enhance sepsis diagnosis 82. More recently, patients with a hyperinflammatory phenotype were observed to have a greater abundance of bacterial reads particularly Enterobacteriaceae species identified through mNGS relative to those with a hypoinflammatory phenotype, substantiating findings that patients with a hyperinflammatory phenotype were more likely to be bacteremic based on conventional results of culture 45. In summary, the interaction between host genetics, pathogen and antibiotic exposure, and alterations in host microbiome at a systems-level remains relatively understudied. Understanding their contribution to disease progression among critical illness subclasses may hold potential to inform targeted interventions.

Urgent need for human relevant disease models

Current in vitro models based largely on culture of monolayers of cells are reductionist and limited in their ability to reveal organ-specific and compartmentalized responses in critical illness. Recent advances in the field of regenerative medicine hold potential to begin to address these challenges. Human induced pluripotent stem cells (hiPSCs), reprogrammed from somatic cells including PBMCs, are replenishable sources of cells which can be differentiated into any cell type 83. Human iPSCs have been used to develop sophisticated in vitro 3-D organoid models including multi-compartment models 84 and used for therapeutic drug screening and monitoring. Importantly, hiPSCs have been shown to recapitulate functional, phenotypic, and transcriptomic responses of primary—both circulating and tissue-resident cells 85,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87. Several studies have used to study cellular responses to pathogenic agents, 88,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90 including a recent study where the investigators treated vascularized organoids with plasma from COVID19 patients to recapitulate in vivo responses 91. Further, novel gene-editing tools including CRISPR-cas9 hold potential to develop in vitro models by efficiently knocking down or out one or more genes simultaneously and accelerate mechanistic studies of candidate targets identified through reverse translational approaches. Lastly, patient-specific iPSCs have been used for precision disease modeling in a variety of chronic conditions with monogenic or polygenic inheritance patterns 92,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95.Although untested, it is possible that patient-specific iPSCs derived from critically ill patients by capturing human genetic diversity may facilitate the development of avatars for precision medicine to study critical illness subclass-specific biology.

Multi-modal approaches to inform patient care

The wealth of biological information generated among critically ill patients is likely to inundate human capacity to meaningfully process and interpret the data, thus necessitating the deployment of supervised and unsupervised machine learning approaches to distinguish signal from noise. Moreover, as illustrated in Fig. 5, future intensive care practitioners will be required to process a wealth of multi-modal data including electronic and biological data, the latter generated through rapid molecular assays. As such the role of artificial intelligence (AI) for data integration, synthesis, and decision-support cannot be overemphasized 96. Thus, concurrent advances are necessary to bridge the divide between systems-biology based understanding of patient-level pathobiology and clinical informatics to inform patient care at the bedside.

Fig. 5
figure 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

Full size image

Conclusions

Advances in precision medicine approaches of the previous two decades have led to the identification of numerous subclassification schemes among critically ill patients. Given the disparate outputs of current subclassification schemes, progress toward identifying consensus endophenotypes is necessary. Furthermore, we remain at the very beginning of the path to identifying treatable traits underlying patient subclasses. Deployment of advanced ‘omic’ technologies and analytic tools are likely to yield subclass-specific candidate biological pathways for hypotheses testing. Moreover, development of robust human relevant disease models are necessary to disentangle mechanistic basis of molecular drivers of critical illness subclasses. Embracing such multi-dimensional approaches may help catalyze the translation of subclass-specific insights into targeted and efficacious interventions and help deliver on the promise of precision critical care medicine to improve patient outcomes.

Availability of data and materials

Not applicable.

References

  1. Vincent J-L, et al. Assessment of the worldwide burden of critical illness: the intensive care over nations (ICON) audit. Lancet Respir Med. 2014;2:380–6.

    Article  PubMed  Google Scholar 

  2. Weiss SL, et al. Global epidemiology of pediatric severe sepsis: the sepsis prevalence, outcomes, and therapies study. Am J Respir Crit Care Med. 2015;191:1147–57.

    Article  PubMed  PubMed Central  Google Scholar 

  3. Bellani G, et al. Epidemiology, patterns of care, and mortality for patients with acute respiratory distress syndrome in intensive care units in 50 countries. JAMA. 2016;315:788–800.

    Article  CAS  PubMed  Google Scholar 

  4. Kaddourah A, Basu RK, Bagshaw SM, Goldstein SL. Epidemiology of acute kidney injury in critically ill children and young adults. N Engl J Med. 2017;376(1):11–20. https://doi.org/10.1056/NEJMoa1611391.

    Article  PubMed  Google Scholar 

  5. Rudd KE, et al. Global, regional, and national sepsis incidence and mortality, 1990–2017: analysis for the Global Burden of Disease Study. The Lancet. 2020;395:200–11.

    Article  Google Scholar 

  6. Patel BK, et al. Effect of early mobilisation on long-term cognitive impairment in critical illness in the USA: a randomised controlled trial. Lancet Respir Med. 2023;11:563–72.

    Article  PubMed  Google Scholar 

  7. Carlton EF, Donnelly JP, Hensley MK, Cornell TT, Prescott HC. New medical device acquisition during pediatric severe sepsis hospitalizations. Crit Care Med. 2020;48:725–31.

    Article  PubMed  PubMed Central  Google Scholar 

  8. Prescott HC, Osterholzer JJ, Langa KM, Angus DC, Iwashyna TJ. Late mortality after sepsis: propensity matched cohort study. BMJ. 2016;353:i2375.

    Article  PubMed  PubMed Central  Google Scholar 

  9. Hartman ME, et al. Readmission and late mortality after critical illness in childhood. Pediatr Crit Care Med. 2017;18:e112–21.

    Article  PubMed  PubMed Central  Google Scholar 

  10. Zimmerman JJ, et al. Critical illness factors associated with long-term mortality and health related quality of life morbidity following community-acquired pediatric septic shock. Crit Care Med. 2020;48:319–28.

    Article  PubMed  PubMed Central  Google Scholar 

  11. Marshall JC. Why have clinical trials in sepsis failed? Trends Mol Med. 2014;20:195–203.

    Article  PubMed  Google Scholar 

  12. Shah FA, et al. A research agenda for precision medicine in sepsis and acute respiratory distress syndrome: an official american thoracic society research statement. Am J Respir Crit Care Med. 2021;204:891–901.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  13. Maslove DM, et al. Redefining critical illness. Nat Med. 2022;28:1141–8.

    Article  CAS  PubMed  Google Scholar 

  14. Stanski NL, Wong HR. Prognostic and predictive enrichment in sepsis. Nat Rev Nephrol. 2020;16:20–31.

    Article  PubMed  Google Scholar 

  15. Reddy K, et al. Subphenotypes in critical care: translation into clinical practice. Lancet Respir Med. 2020;8:631–43.

    Article  PubMed  Google Scholar 

  16. Sinha P, Meyer NJ, Calfee CS. Biological phenotyping in sepsis and acute respiratory distress syndrome. Annu Rev Med. 2023;74:457–71.

    Article  CAS  PubMed  Google Scholar 

  17. Hotchkiss RS, Monneret G, Payen D. Sepsis-induced immunosuppression: from cellular dysfunctions to immunotherapy. Nat Rev Immunol. 2013;13:862–74.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  18. Delano MJ, Ward PA. Sepsis-induced immune dysfunction: can immune therapies reduce mortality? J Clin Invest. 2016;126:23–31.

    Article  PubMed  PubMed Central  Google Scholar 

  19. Wong HR, et al. Genome-level expression profiles in pediatric septic shock indicate a role for altered zinc homeostasis in poor outcome. Physiol Genomics. 2007;30:146–55.

    Article  CAS  PubMed  Google Scholar 

  20. Cvijanovich N, et al. Validating the genomic signature of pediatric septic shock. Physiol Genomics. 2008;34:127–34.

    Article  CAS  PubMed  Google Scholar 

  21. Wong HR, et al. Identification of pediatric septic shock subclasses based on genome-wide expression profiling. BMC Med. 2009;7:34.

    Article  PubMed  PubMed Central  Google Scholar 

  22. Wong HR, et al. Validation of a gene expression-based subclassification strategy for pediatric septic shock. Crit Care Med. 2011;39:2511–7.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  23. Wong HR, et al. Developing a clinically feasible personalized medicine approach to pediatric septic shock. Am J Respir Crit Care Med. 2015;191:309–15.

    Article  PubMed  PubMed Central  Google Scholar 

  24. Grunwell JR, et al. Differential expression of the Nrf2-linked genes in pediatric septic shock. Crit Care. 2015;19:327.

    Article  PubMed  PubMed Central  Google Scholar 

  25. Yang JO, et al. Whole blood transcriptomics identifies subclasses of pediatric septic shock. Crit Care. 2023;27:486.

    Article  PubMed  PubMed Central  Google Scholar 

  26. Davenport EE, et al. Genomic landscape of the individual host response and outcomes in sepsis: a prospective cohort study. Lancet Respir Med. 2016;4:259–71.

    Article  PubMed  PubMed Central  Google Scholar 

  27. Antcliffe DB, et al. Transcriptomic signatures in sepsis and a differential response to steroids. From the VANISH randomized trial. Am J Respir Crit Care Med. 2019;199:980–6.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  28. Sweeney TE, et al. Unsupervised analysis of transcriptomics in bacterial sepsis across multiple datasets reveals three robust clusters. Crit Care Med. 2018;46:915–25.

    Article  PubMed  PubMed Central  Google Scholar 

  29. Balch JA, et al. Defining critical illness using immunological endotypes in patients with and without sepsis: a cohort study. Crit Care. 2023;27:292.

    Article  PubMed  PubMed Central  Google Scholar 

  30. Sweeney TE, et al. Validation of inflammopathic, adaptive, and coagulopathic sepsis endotypes in coronavirus disease 2019. Crit Care Med. 2021;49:e170–8.

    Article  CAS  PubMed  Google Scholar 

  31. Iglesias J, et al. A 33-mRNA classifier is able to produce inflammopathic, adaptive, and coagulopathic endotypes with prognostic significance: the outcomes of metabolic resuscitation using ascorbic acid, thiamine, and glucocorticoids in the early treatment of sepsis (ORANGES) trial. J Pers Med. 2020;11:9.

    Article  PubMed  PubMed Central  Google Scholar 

  32. Kyriazopoulou E, et al. Transitions of blood immune endotypes and improved outcome by anakinra in COVID-19 pneumonia: an analysis of the SAVE-MORE randomized controlled trial. Crit Care. 2024;28:73.

    Article  PubMed  PubMed Central  Google Scholar 

  33. Scicluna BP, et al. Classification of patients with sepsis according to blood genomic endotype: a prospective cohort study. Lancet Respir Med. 2017;5:816–26.

    Article  PubMed  Google Scholar 

  34. Shankar-Hari M, et al. Reframing sepsis immunobiology for translation: towards informative subtyping and targeted immunomodulatory therapies. Lancet Respi Med. 2024;12(4):323–36.

    Article  CAS  Google Scholar 

  35. Yehya N, et al. Peripheral blood transcriptomic sub-phenotypes of pediatric acute respiratory distress syndrome. Crit Care. 2020;24:681.

    Article  PubMed  PubMed Central  Google Scholar 

  36. Calfee CS, et al. Subphenotypes in acute respiratory distress syndrome: latent class analysis of data from two randomised controlled trials. Lancet Respir Med. 2014;2:611–20.

    Article  PubMed  PubMed Central  Google Scholar 

  37. Dahmer MK, et al. Identification of phenotypes in paediatric patients with acute respiratory distress syndrome: a latent class analysis. Lancet Respir Med. 2022;10:289–97.

    Article  PubMed  Google Scholar 

  38. Yehya N, et al. Identification of molecular subphenotypes in two cohorts of paediatric ARDS. Thorax. 2024;79:128–34.

    Article  PubMed  Google Scholar 

  39. Famous KR, et al. Acute respiratory distress syndrome subphenotypes respond differently to randomized fluid management strategy. Am J Respir Crit Care Med. 2017;195:331–8.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  40. Calfee CS, et al. ARDS subphenotypes and differential response to simvastatin: secondary analysis of a randomized controlled trial. Lancet Respir Med. 2018;6:691–8.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  41. Sinha P, et al. Latent class analysis reveals COVID-19-related acute respiratory distress syndrome subgroups with differential responses to corticosteroids. Am J Respir Crit Care Med. 2021;204:1274–85.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  42. Bos LD, et al. Identification and validation of distinct biological phenotypes in patients with acute respiratory distress syndrome by cluster analysis. Thorax. 2017;72:876–83.

    Article  CAS  PubMed  Google Scholar 

  43. Bos LDJ, et al. Understanding heterogeneity in biologic phenotypes of acute respiratory distress syndrome by leukocyte expression profiles. Am J Respir Crit Care Med. 2019;200:42–50.

    Article  CAS  PubMed  Google Scholar 

  44. Heijnen NFL, et al. Biological subphenotypes of acute respiratory distress syndrome show prognostic enrichment in mechanically ventilated patients without acute respiratory distress syndrome. Am J Respir Crit Care Med. 2021;203:1503–11.

    Article  PubMed  Google Scholar 

  45. Sinha P, et al. Identifying molecular phenotypes in sepsis: an analysis of two prospective observational cohorts and secondary analysis of two randomised controlled trials. Lancet Respi Med. 2023;11(11):965–74.

    Article  CAS  Google Scholar 

  46. Ranieri VM, et al. Drotrecogin alfa (Activated) in adults with septic shock. N Engl J Med. 2012;366:2055–64.

    Article  CAS  PubMed  Google Scholar 

  47. Shankar-Hari M, et al. Defining Phenotypes and Treatment Effect Heterogeneity to Inform Acute Respiratory Distress Syndrome and Sepsis Trials: Secondary Analyses of Three RCTs. Southampton (UK): NIHR Journals Library; 2021.

    Google Scholar 

  48. Antcliffe DB, et al. Patient stratification using plasma cytokines and their regulators in sepsis: relationship to outcomes, treatment effect and leucocyte transcriptomic subphenotypes. Thorax. 2024. https://doi.org/10.1136/thorax-2023-220538.

    Article  PubMed  Google Scholar 

  49. Neyton LPA, et al. Host and microbe blood metagenomics reveals key pathways characterizing critical illness phenotypes. Am J Respir Crit Care Med. 2024. https://doi.org/10.1164/rccm.202308-1328OC.

    Article  PubMed  Google Scholar 

  50. Sinha P, et al. Molecular phenotypes of ARDS in the ROSE trial have differential outcomes and gene expression patterns that differ at baseline and longitudinally over time. Am J Respir Crit Care Med. 2024. https://doi.org/10.1164/rccm.202308-1490OC.

    Article  PubMed  Google Scholar 

  51. Bhatraju PK, et al. Identification of acute kidney injury subphenotypes with differing molecular signatures and responses to vasopressin therapy. Am J Respir Crit Care Med. 2019;199:863–72.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  52. Wiersema R, et al. Two subphenotypes of septic acute kidney injury are associated with different 90-day mortality and renal recovery. Crit Care. 2020;24:150.

    Article  PubMed  PubMed Central  Google Scholar 

  53. Wong HR, et al. Prospective clinical testing and experimental validation of the Pediatric Sepsis Biomarker Risk Model. Sci Transl Med. 2019. https://doi.org/10.1126/scitranslmed.aax9000.

    Article  PubMed  PubMed Central  Google Scholar 

  54. Seymour CW, et al. Murine sepsis phenotypes and differential treatment effects in a randomized trial of prompt antibiotics and fluids. Crit Care. 2019;23:384.

    Article  PubMed  PubMed Central  Google Scholar 

  55. Millar JE, et al. Characterizing preclinical sub-phenotypic models of acute respiratory distress syndrome: an experimental ovine study. Physiol Rep. 2021;9:e15048.

    Article  PubMed  PubMed Central  Google Scholar 

  56. Wildi K, et al. Validation of messenger ribonucleic acid markers differentiating among human acute respiratory distress syndrome subgroups in an ovine model of acute respiratory distress syndrome phenotypes. Front Med. 2022. https://doi.org/10.3389/fmed.2022.961336.

    Article  Google Scholar 

  57. Wildi K, et al. Application of anti-inflammatory treatment in two different ovine acute respiratory distress syndrome injury models: a preclinical randomized intervention study. Sci Rep. 2023;13:17986.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  58. NAGMSC Working Group on Sepsis.

  59. van Amstel RBE, et al. Uncovering heterogeneity in sepsis: a comparative analysis of subphenotypes. Intensive Care Med. 2023;49:1360–9.

    Article  PubMed  PubMed Central  Google Scholar 

  60. Wong HR, et al. Combining prognostic and predictive enrichment strategies to identify children with septic shock responsive to corticosteroids. Crit Care Med. 2016;44:e1000-1003.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  61. Cano-Gamez E, et al. An immune dysfunction score for stratification of patients with acute infection based on whole-blood gene expression. Sci Trans Med. 2022;14:eabq4433.

    Article  CAS  Google Scholar 

  62. Burnham KL, et al. Shared and distinct aspects of the sepsis transcriptomic response to fecal peritonitis and pneumonia. Am J Respir Crit Care Med. 2017;196:328–39.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  63. Wong HR, et al. Endotype transitions during the acute phase of pediatric septic shock reflect changing risk and treatment response. Crit Care Med. 2018;46:e242–9.

    Article  PubMed  PubMed Central  Google Scholar 

  64. PANTHER. https://panthertrial.org.

  65. TRAITS - Time Critical Precision Medicine programme. The University of Edinburgh https://www.ed.ac.uk/inflammation-research/clinical-trials/traits-ci-trial (2023).

  66. Karakike E, et al. Effect of intravenous clarithromycin in patients with sepsis, respiratory and multiple organ dysfunction syndrome: a randomized clinical trial. Crit Care. 2022;26:183.

    Article  PubMed  PubMed Central  Google Scholar 

  67. Maslove DM, Wong HR. Gene expression profiling in sepsis: timing, tissue, and translational considerations. Trends Mol Med. 2014;20:204–13.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  68. Cajander S, et al. Profiling the dysregulated immune response in sepsis: overcoming challenges to achieve the goal of precision medicine. Lancet Respi Med. 2023;12(4):305–22.

    Article  Google Scholar 

  69. Morris AC, Rynne J, Shankar-Hari M. Compartmentalisation of immune responses in critical illness: Does it matter? Intensive Care Med. 2022;48:1617.

    Article  Google Scholar 

  70. Williams JG, et al. Multi-omic characterization of pediatric ARDS via nasal brushings. Respir Res. 2022;23:181.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  71. McNeil JB, et al. Novel method for noninvasive sampling of the distal airspace in acute respiratory distress syndrome. Am J Respir Crit Care Med. 2018;197:1027–35.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  72. Engoren M, et al. Genetic variants associated with sepsis. PLoS ONE. 2022;17:e0265052.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  73. Binnie A, et al. Epigenetic profiling in severe sepsis: a pilot study of DNA methylation profiles in critical illness. Crit Care Med. 2020;48:142–50.

    Article  CAS  PubMed  Google Scholar 

  74. Chen T, et al. The independent prognostic value of global epigenetic alterations: an analysis of single-cell ATAC-seq of circulating leukocytes from trauma patients followed by validation in whole blood leukocyte transcriptomes across three etiologies of critical illness. EBioMedicine. 2022;76:103860.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  75. Cross D, Drury R, Hill J, Pollard AJ. Epigenetics in sepsis: understanding its role in endothelial dysfunction, immunosuppression, and potential therapeutics. Front Immunol. 2019;10:1363.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  76. Reyes M, et al. An immune cell signature of bacterial sepsis. Nat Med. 2020;26:333–40.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  77. Kwok AJ, et al. Neutrophils and emergency granulopoiesis drive immune suppression and an extreme response endotype during sepsis. Nat Immunol. 2023;24:767–79.

    Article  CAS  PubMed  Google Scholar 

  78. Yoshida M, et al. Local and systemic responses to SARS-CoV-2 infection in children and adults. Nature. 2022;602:321–7.

    Article  CAS  PubMed  Google Scholar 

  79. Ahern DJ, et al. A blood atlas of COVID-19 defines hallmarks of disease severity and specificity. Cell. 2022;185:916-938.e58.

    Article  Google Scholar 

  80. Dickson RP, et al. Enrichment of the lung microbiome with gut bacteria in sepsis and the acute respiratory distress syndrome. Nat Microbiol. 2016;1:16113.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  81. Dickson, R. P. et al. Lung microbiota predict clinical outcomes in critically Ill patients. Am J Respir Crit Care Med 201, 555–563.

  82. Kalantar KL, et al. Integrated host-microbe plasma metagenomics for sepsis diagnosis in a prospective cohort of critically ill adults. Nat Microbiol. 2022;7:1805–16.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  83. Karagiannis P, et al. Induced pluripotent stem cells and their use in human models of disease and development. Physiol Rev. 2019;99:79–114.

    Article  CAS  PubMed  Google Scholar 

  84. Sharma A, Sances S, Workman MJ, Svendsen CN. Multi-lineage human iPSC-derived platforms for disease modeling and drug discovery. Cell Stem Cell. 2020;26:309–29.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  85. Douvaras P, et al. Directed differentiation of human pluripotent stem cells to microglia. Stem Cell Reports. 2017;8:1516–24.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  86. Pavlovic BJ, Blake LE, Roux J, Chavarria C, Gilad Y. A comparative assessment of human and chimpanzee iPSC-derived cardiomyocytes with primary heart tissues. Sci Rep. 2018;8:15312.

    Article  PubMed  PubMed Central  Google Scholar 

  87. Monkley S, et al. Optimised generation of iPSC-derived macrophages and dendritic cells that are functionally and transcriptionally similar to their primary counterparts. PLoS ONE. 2020;15:e0243807.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  88. Martins Gomes SF, et al. Induced pluripotent stem cell-derived brain endothelial cells as a cellular model to study neisseria meningitidis infection. Front Microbiol. 2019. https://doi.org/10.3389/fmicb.2019.01181.

    Article  PubMed  PubMed Central  Google Scholar 

  89. Perez-Bermejo JA, et al. SARS-CoV-2 infection of human iPSC–derived cardiac cells reflects cytopathic features in hearts of patients with COVID-19. Sci Trans Med. 2021;13:eabf872.

    Article  Google Scholar 

  90. Wang R, et al. Human airway lineages derived from pluripotent stem cells reveal the epithelial responses to SARS-CoV-2 infection. Am J Physiol Lung Cell Mol Physiol. 2022;322:L462–78.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  91. Kawakami E, et al. Complement factor D targeting protects endotheliopathy in organoid and monkey models of COVID-19. Cell Stem Cell. 2023;30:1315-1330.e10.

    Article  CAS  PubMed  Google Scholar 

  92. Hamazaki T, El Rouby N, Fredette NC, Santostefano KE, Terada N. Concise review: induced pluripotent stem cell research in the era of precision medicine. Stem Cells. 2017;35:545–50.

    Article  PubMed  Google Scholar 

  93. Musunuru K, et al. Induced pluripotent stem cells for cardiovascular disease modeling and precision medicine: a scientific statement from the american heart association. Circ Genomic Precis Med. 2018;11:e000043.

    Google Scholar 

  94. Ortiz-Vitali JL, Darabi R. iPSCs as a platform for disease modeling, drug screening, and personalized therapy in muscular dystrophies. Cells. 2019;8:E20.

    Article  Google Scholar 

  95. Rivetti di Val Cervo P, Besusso D, Conforti P, Cattaneo E. hiPSCs for predictive modelling of neurodegenerative diseases: dreaming the possible. Nat Rev Neurol. 2021;17(6):381–92. https://doi.org/10.1038/s41582-021-00465-0.

    Article  PubMed  PubMed Central  Google Scholar 

  96. Acosta JN, Falcone GJ, Rajpurkar P, Topol EJ. Multimodal biomedical AI. Nat Med. 2022;28:1773–84.

    Article  CAS  PubMed  Google Scholar 

  97. Maslove DM, Tang BM, McLean AS. Identification of sepsis subtypes in critically ill adults using gene expression profiling. Crit Care. 2012;16:R183.

    Article  PubMed  PubMed Central  Google Scholar 

  98. Tang BMP, McLean AS, Dawes IW, Huang SJ, Lin RCY. The use of gene-expression profiling to identify candidate genes in human sepsis. Am J Respir Crit Care Med. 2007;176:676–84.

    Article  CAS  PubMed  Google Scholar 

  99. Tang BMP, et al. Gene-expression profiling of gram-positive and gram-negative sepsis in critically ill patients. Crit Care Med. 2008;36:1125–8.

    Article  CAS  PubMed  Google Scholar 

  100. Yao L, et al. Gene expression scoring of immune activity levels for precision use of hydrocortisone in vasodilatory shock. Shock. 2022;57:384–91.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  101. Baghela A, et al. Predicting sepsis severity at first clinical presentation: the role of endotypes and mechanistic signatures. EBioMedicine. 2022;75:103776.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  102. Mathew D, et al. Deep immune profiling of COVID-19 patients reveals distinct immunotypes with therapeutic implications. Science. 2020;369:eabc8511.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  103. López-Martínez C, et al. Transcriptomic clustering of critically ill COVID-19 patients. Eur Respi J. 2023;61:2200592.

    Article  Google Scholar 

  104. Neyton LPA, et al. Molecular patterns in acute pancreatitis reflect generalizable endotypes of the host response to systemic injury in humans. Ann Surg. 2022;275:e453–62.

    Article  PubMed  Google Scholar 

Download references

Acknowledgements

Figures were created with www.BioRender.com

Funding

M.R.A was supported by a Procter K-to-R Scholar award through the Cincinnati Children’s Research Foundation and NIH awards R21GM151703 and R21GM150093.

Author information

Authors and Affiliations

  1. Division of Immunobiology, Graduate Program, College of Medicine, University of Cincinnati, Cincinnati, OH, 45267, USA

    Joseph Stevens

  2. Division of Critical Care Medicine, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH, 45229, USA

    Oğuzhan Tezel, Matthew Hapstack & Mihir R. Atreya

  3. Department of Pediatrics, University of Cincinnati College of Medicine, Cincinnati, OH, 45627, USA

    Valentina Bonnefil & Mihir R. Atreya

Authors
  1. Joseph Stevens
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. OÄŸuzhan Tezel
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Valentina Bonnefil
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Matthew Hapstack
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Mihir R. Atreya
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

J.S and M.R.A conceptualized the article. J.S and M.R.A wrote the first draft. J.S, O.T, V.B, M.H, and M.R.A contributed key portions, provided critical input, and approved the final version of the manuscript.

Corresponding author

Correspondence to Mihir R. Atreya.

Ethics declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

Cincinnati Children’s Hospital Medical Center (CCHMC) and the estate of the late Dr. Hector R. Wong hold patents for the pediatric sepsis biomarker risk model (PERSEVERE) for risk-stratification of pediatric sepsis patients and gene-expression based adaptive endotypes. M.R.A and CCHMC hold provisional patents for provisional patents (1) for a unified biomarker model – PERSEVERENCE that incorporates PERSEVERE and endothelial dysfunction markers to predict risk of multiple organ dysfunctions in sepsis, (2) a provisional patent for a PERSEVERENCE SA-AKI model that incorporates endothelial biomarkers predictive of persistent sepsis associated acute kidney injury for enrichment in clinical trials, and (3) a provisional patent for gene-expression-based multiple organ dysfunction syndrome (MODS) subclass identification, reflective of the host innate immune response.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Stevens, J., Tezel, O., Bonnefil, V. et al. Biological basis of critical illness subclasses: from the bedside to the bench and back again. Crit Care 28, 186 (2024). https://doi.org/10.1186/s13054-024-04959-3

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s13054-024-04959-3

Keywords

Critical Care

ISSN: 1364-8535