Measuring signatures of host resistance, disease tolerance, and damage in human sepsis: a prospective cohort study.

Measuring signatures of host resistance, disease tolerance, and damage in human sepsis: a prospective cohort study - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Search Log in Primary site navigation Search Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice Intensive Care Med . 2026 Apr 7;52(4):662–672. doi: 10.1007/s00134-026-08404-7 Measuring signatures of host resistance, disease tolerance, and damage in human sepsis: a prospective cohort study Arnab Chowdhury 1, 2, ✉ , Rachel E Powell 2, 4 , Jason N Kennedy 1, 2 , Kelly L Urbanek 1, 2 , Derek C Angus 1, 2 , Chung-Chou H Chang 3 , Lu Tang 3 , Sebastian Weis 5, 6, 7 , Michael Bauer 5 , Manu Shankar-Hari 8, 9 , Christopher W Seymour 1, 2 Author information Article notes Copyright and License information 1 Department of Critical Care Medicine, University of Pittsburgh School of Medicine, Pittsburgh, USA 2 The CRISMA Center, Department of Critical Care Medicine, University of Pittsburgh School of Medicine, Pittsburgh, USA 3 Department of Biostatistics, University of Pittsburgh, Pittsburgh, USA 4 Division of Pulmonary, Allergy, and Critical Care Medicine, University of Pittsburgh Medical Center, Pittsburgh, USA 5 Department of Anesthesiology and Intensive Care Medicine, Jena University Hospital, Friedrich-Schiller University, Jena, Germany 6 Institute of Infectious Diseases and Infection Control, Jena University Hospital, Friedrich-Schiller University, Jena, Germany 7 Leibniz Institute for Natural Product Research and Infection Biology, Hans-Knöll Institute-HKI, Jena, Germany 8 Centre for Inflammation Research, University of Edinburgh, Edinburgh, UK 9 Centre for Critical Illness Research, King’s College, London, UK ✉ Corresponding author. Received 2026 Jan 9; Accepted 2026 Mar 19; Issue date 2026. © The Author(s) 2026 Open Access This article is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License, which permits any non-commercial 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-nc/4.0/ . PMC Copyright notice PMCID: PMC13221389  PMID: 41944864 Abstract Purpose To understand how protein biomarkers in blood and urine that are aligned with host resistance to infection, disease tolerance, and damage are associated with clinical outcomes and sepsis subtypes in community-onset sepsis. Methods Adults meeting Sepsis-3 criteria were prospectively enrolled within 6 h of emergency department arrival and assigned clinical subtypes (α, β, γ, δ), using the Sepsis ENdotyping in Emergency CAre (SENECA) approach. Using structured expert ranking with consensus adjudication, 16 plasma and urinary biomarkers obtained from remnant biospecimens were grouped into three mechanistic axes contributing to sepsis pathophysiology: host resistance to infection, disease tolerance, and damage to the host. Biomarkers for each concept were analyzed with principal component analysis as a signature, and multivariable logistic regression tested associations of each signature with 90-day mortality and sepsis subtype membership. Results Among 444 adults, the mean age was 60 years [SD: 16], the mean SOFA score was 4.3 [SD: 2.3], and 90-day mortality was 17%. After adjustment for age, sex, and race, greater damage to the host was associated with increased 90-day mortality (adjusted odds ratio (aOR) = 1.70; 95% CI 1.38–2.11; p < 0.001), while greater host resistance (aOR = 0.83; 95% CI 0.54–1.10; p = 0.4) and greater disease tolerance (aOR = 0.83; 95% CI 0.68–1.01; p = 0.06) were not. Differences across sepsis subtypes were most pronounced for disease tolerance and damage signatures, where the δ‑type patients exhibited higher damage and lower disease tolerance and the α‑type patients had lower damage and higher disease tolerance. Conclusion Biomarker signatures aligned with host resistance to infection, disease tolerance, and damage to the host, informed by expert consensus, were associated with clinical outcomes and sepsis subtype membership. Graphic abstract Supplementary Information The online version contains supplementary material available at 10.1007/s00134-026-08404-7. Keywords: Sepsis, Host resistance, Disease tolerance, Damage, Biomarkers Take-home message We tested whether biomarkers in the circulation and urine, grouped into signatures of host resistance, disease tolerance, and damage, are associated with 90-day mortality and subtypes in sepsis. The clinical subtype with the greatest mortality showed low disease tolerance and high damage, a hypothesis-generating finding that highlights the opportunity to develop disease tolerance-targeted strategies to complement traditional host resistance-focused sepsis care. Open in a new tab Introduction Sepsis is a common and deadly syndrome, and current treatments focus on early antimicrobial administration, source control, and organ support. A key barrier to the development of targeted treatments is the diverse and heterogeneous underlying biologic mechanisms of sepsis, which are not easily discerned clinically. A recent step toward a greater understanding of sepsis heterogeneity was characterization of multiple sepsis subtypes using machine learning methods. Though widely validated, few subtypes are predictive of treatment response or implemented at the bedside [ 1 ]. Recent work suggests the prevailing conceptual model for sepsis may limit the broader mechanistic understanding of sepsis and how new treatments could target the individual [ 2 , 3 ]. Current models focus primarily on the pathogen and host immune response, without fully capturing additional aspects about the role of the host, including host resistance to infection, disease tolerance, and the magnitude of tissue damage [ 4 ]. Specifically, host resistance mechanisms are innate and adaptive immune responses that aim to decrease pathogen load and reduce pathogen-associated damage. This process is distinct from antimicrobial resistance. Disease tolerance mechanisms are tissue-protective stress responses and metabolic adaptations that preserve function without directly reducing pathogen burden and limit the parenchymal dysfunction and subsequent direct damage caused by the pathogen, their by-products, and indirect damage from host resistance. Damage is the injury to tissue caused by the pathogen or the host’s immune response that ensues when stress and damage responses fail and disease tolerance mechanisms are overwhelmed. These have been described in detail in prior publications [ 3 – 5 ]. Consistent with this framework, experimental sepsis models show that metabolic and endocrine programs can determine survival and organ injury without affecting pathogen burden, with examples including iron and glucose metabolic adaptations [ 6 ] and fasting and glucose utilization effects [ 7 ]. Although well-described in preclinical models and immunity, limited work maps this framework to sepsis patients and their outcomes. To address this knowledge gap, we proposed measurable signatures of host resistance, disease tolerance, and damage to the host in a prospective cohort of sepsis patients and examined their associations with clinical sepsis subtypes and outcomes. Methods The study was approved by the University of Pittsburgh Human Research Protection Office. The data were obtained under a waiver of informed consent. Reporting followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. This study had four steps: (1) structured expert consensus process with members of the International Sepsis Forum (ISF) to categorize candidate plasma and urinary biomarkers into mechanistic domains aligned with host resistance, disease tolerance, and damage, (2) prospective cohort study where candidate biomarkers for mechanistic concepts were measured from patients with community-onset sepsis receiving emergency care, (3) application of Sepsis ENdotyping in Emergency CAre (SENECA) sepsis subtype labels using clinical data, and (4) determination of the association between host resistance, disease tolerance, and damage signatures with both sepsis subtypes and outcome. Mechanistic concept and biomarkers We preselected 16 candidate biomarkers with established links to host-response pathways relevant to host resistance, disease tolerance and damage, and feasibility of measurement in remnant plasma and urine: interleukin-6 [IL-6], interleukin-8 [IL-8], interleukin-10 [IL-10], angiopoietin-1 [Ang-1], angiopoietin-2 [Ang-2], syndecan-1 [Syn-1], E-selectin, intercellular adhesion molecule 1 [ICAM-1], antithrombin III [ATIII], plasminogen activator inhibitor 1 [PAI-1], high mobility group box 1 [HMGB1], heme oxygenase-1 [HO-1], NephroCheck® tissue inhibitor of metalloproteinase-2 and insulin-like growth factor binding protein-7 [TIMP-2]·[IGFBP7], lactic acid, bicarbonate, and uric acid. Candidate biomarkers were grouped into mechanistic concepts of host resistance, disease tolerance, and damage using a structured expert consensus process. Ten ISF experts independently assigned each biomarker a percentage attribution (0–100%) to each of the three concepts, constrained to sum to 100%, allowing cross-domain attribution. A consensus committee of senior ISF members reconciled aggregated attributions alongside established biological evidence to determine a concept assignment. In two instances, the consensus committee selected literature-supported primary assignments: HMGB1 was assigned to damage as a prototypical damage-associated molecular pattern [ 8 ], and [TIMP-2]·[IGFBP7] was assigned to disease tolerance given its linkage to cell-cycle arrest stress responses that may represent early tissue-protective mechanisms [ 9 ]. Prospective cohort study We enrolled patients with community-onset sepsis at a tertiary care center in southwestern Pennsylvania from October 2017 to June 2019. This cohort was initially enrolled in the VESPRE study [ 10 ]. Eligible patients were adults (≥ 18 years) who met Sepsis-3 [ 11 ] criteria within 6 h of emergency department (ED) presentation and had sufficient remnant biospecimen volume. Patients were identified using an EHR alert for suspected infection (antibiotics plus blood cultures) and sequential organ failure assessment (SOFA) ≥ 2, followed by online adjudication to confirm Sepsis-3. Required remnant volumes were ≥ 450 µL EDTA plasma and ≥ 125 µL urine. Data sources Clinical data We used electronic health record (EHR; CERNER Co., Kansas City, MO) data from encounters in the prospective cohort. Clinical data included patient demographics, vital signs, laboratory values, organ support such as mechanical ventilation and vasopressor use, length of stay, and mortality up to 90 days after admission. Patient-reported race was derived from the University of Pittsburgh Medical Center (UPMC) registration system data. Due to small sample sizes, race was collapsed into two categories (White, non-White). Comorbid conditions were identified using hospitalization diagnosis International Classification of Diseases, Ninth Revision (ICD-9) codes. The most abnormal vital sign and laboratory values within six hours of sepsis onset were used for assessment of baseline illness severity. Biospecimen data Remnant plasma and urine specimens were obtained from the UPMC clinical laboratory within 48 h of acquisition after clinical testing was complete. Samples were transported to the Center for Research, Investigation, and Systems Modeling of Acute Illness (CRISMA) research laboratory and processed the day of acquisition, aliquoted, and stored at − 80 °C in a secure biorepository. Biomarkers were quantified using commercially available kits. Statistical analyses Statistical analyses included the following steps: (1) calculating, for each biomarker, the mean expert-attributed proportion assigned to host resistance, disease tolerance, and damage; (2) mapping biomarkers of host resistance, disease tolerance, and damage using principal component analysis into a signature; (3) comparing baseline characteristics and outcomes by signature; (4) modeling associations of signatures with sepsis subtypes and clinical outcomes. For biomarkers grouped into mechanistic concepts, we conducted a one-dimensional principal component analysis (PCA) to map the measurements into a single quantitative signature. The directionality of biomarkers Ang-1, ATIII, and bicarbonate was inverted such that higher values reflected more abnormal values. Additionally, we compared baseline characteristics and outcomes across high vs low strata of each mechanistic signature, defined by a median split of the PCA-derived signature scores. We used EHR-derived presentation data that included vital signs, clinical laboratory data and comorbidities to assign patients to one of four SENECA clinical subtypes (α, β, γ, δ). Subtype membership was defined by the minimum Euclidean distance to the published centroids [ 12 ]. We did not model subtype classification uncertainty described in a recent work [ 13 ]. We compared baseline characteristics and outcomes across SENECA sepsis subtypes. Categorical data were presented as no. (%), normally distributed continuous data as mean (standard deviation [SD]), and non-normally distributed continuous data as median (interquartile range [IQR]). Prior to modeling, we assessed variable distributions, missingness, and correlation. Log transformation was used for non-symmetrically distributed data. Missing data were imputed using random forest imputation with predictive mean matching. To determine the risk-adjusted association between the PCA-derived mechanistic signatures and 90-day mortality, we used multivariable logistic regression with robust standard errors. Model covariates were chosen a priori based on factors known to be associated with sepsis-specific outcomes, including age, race, and sex. We illustrated 90-day survival using a Kaplan–Meier analysis and log-rank tests utilizing high vs low strata of each mechanistic signature, defined by a median split of the PCA-derived signature scores. To understand the potential contributions of each mechanistic signature to sepsis subtype membership, we used random forest models and measures of feature importance to predict membership in each sepsis subtype. Sensitivity analyses We performed five sensitivity analyses. First, we repeated primary analyses after eliminating biomarkers with high degrees of missingness (missingness > 20%). Second, we repeated models using 2-dimensional PCA. Third, we repeated the primary analyses with biomarkers grouped into mechanistic concepts using the proportion of expert agreement rather than discrete categorization. Fourth, we stratified patients by median PCA-derived host resistance and disease tolerance signatures (high/high, high/low, low/high, low/low) and compared 90-day survival using Kaplan–Meier curves with log-rank tests and estimated adjusted associations with 90-day mortality via multivariable logistic regression with group as the exposure (reference: low host resistance + low disease tolerance), adjusting for age, sex, and race. Finally, to ensure these mechanistic signatures were not proxies for illness severity, we compared median-split host resistance, disease tolerance, and damage scores with admission SOFA categories (≤ 2, 3–4, ≥ 5) and visualized overlap using alluvial plots (eFigure 4). Analyses were performed in Stata version 17.0 (StataCorp) and R Statistical Software version 4.2.2. Results were significant at a two-sided p value of ≤ 0.05. Results Patients Of 1,027 screened patients who met Sepsis-3 criteria within 6 h of presentation to the ED, 444 (43%) were enrolled (eFigure 1). Patients were predominantly white (85%), male (57%), and had a mean age of 60 (SD: 16) years (Table 1 ). The mean Elixhauser comorbidity index was 4.0 (SD: 2.3), and the mean SOFA score was 4.3 (SD: 2.3). Overall, 110 patients (25%) required mechanical ventilation, 83 patients (19%) required vasopressors, and 75 patients (17%) died at 90 days. Across sepsis subtypes, 135 (30%) patients belonged to the α-type, 152 (34%) patients to the β-type, 94 (21%) patients to the γ-type, and 63 (14%) patients to the δ-type (eTable 2). The δ-type patients had the highest SOFA scores and worst outcomes, including the greatest in-hospital mortality (eTable 2). Table 1. Baseline characteristics Patient characteristics a ( n = 444) Enrollment period 2017–2019 Age, mean (sd), years 60 (16) Sex, no. (%) Male 252 (57%) Female 192 (43%) Race, no. b (%) White 375 (85%) Non-White 69 (15%) ICU admission during encounter, no. (%) 213 (48%) Mechanical ventilation during encounter, no. c (%) 110 (25%) Vasopressor use during encounter, no. d (%) 83 (19%) SOFA score, mean (SD) 4.3 (2.3) Elixhauser, mean (SD) 4.0 (2.3) In-hospital mortality, no. (%) 40 (9%) 90-Day mortality, no. (%) 75 (17%) Open in a new tab Values represent mean (SD) or number (%) SOFA Sequential Organ Failure Assessment a All variables reflect data from the first 6 h of presentation unless otherwise specified b Race was derived from structured electronic health record documentation c Mechanical ventilation refers to invasive positive-pressure ventilation d Vasopressor use includes norepinephrine, epinephrine, vasopressin, or equivalent agents Mapping sepsis biomarkers into mechanistic concepts The mean expert percentage attributions of each biomarker to host resistance, disease tolerance, and damage, along with the primary domain assignment used for analysis, are shown in eFigure 2. After review by the consensus committee, we proposed four biomarkers that could indicate host resistance (IL-6, IL-8, E-selectin, ICAM-1), five biomarkers that could indicate disease tolerance (IL-10, HO-1, Ang-1, [TIMP-2]·[IGFBP7], ATIII), and seven biomarkers that could indicate damage (Ang-2, Syn-1, HMGB1, PAI-1, uric acid, lactic acid, bicarbonate). We calculated one-dimensional principal components for each of the three mechanistic concepts (eFigure 3). The variance explained by the principal component signature for host resistance, disease tolerance, and damage was 46.8%, 28.4%, and 31.4%, respectively. Patient characteristics Across dichotomized groups of the host resistance signatures, the distributions of age, race, sex, comorbidities, and SOFA score were similar (eTable 3). Outcomes such as ICU admission (range 46–51%), mechanical ventilation (range 23–27%), and in-hospital mortality (9%) were similar. Vasopressor use was more common in the high host resistance group (24% in the high group vs. 13% in the low group), while 90-day mortality was lower (14% in the high group vs. 20% in the low group, eTable 3, Fig. 1 A). Among disease tolerance groups, the in-hospital mortality was lower for patients with high disease tolerance (6%) compared to patients with low disease tolerance (12%, eTable 4, Fig. 1 B). For damage groups, the SOFA score was greater (4.7 [SD 2.6] for high vs. 3.9 [SD 1.9] for low), vasopressor use was more common (27% for high vs. 11% for low), and in-hospital mortality was greater in those with high damage (14% for high vs. 4% for low, eTable 5, Fig. 1 C). Fig. 1. Open in a new tab Kaplan–Meier survival curves stratified by high vs. low (median split) for each mechanistic concept, A host resistance, B disease tolerance, C damage Risk-adjusted association between mechanistic signatures and 90-day mortality In adjusted models using PCA-derived mechanistic signatures, greater host resistance and disease tolerance were not significantly associated with lower 90-day mortality (host resistance: aOR 0.83; 95% CI 0.54–1.10; p = 0.40; Table 2 A, disease tolerance: aOR 0.83; 95% CI 0.68–1.01; p = 0.06; Table 2 B). Increased damage to the host was independently associated with greater odds of 90-day mortality (aOR 1.70; 95% CI 1.38–2.11; p < 0.001; Table 2 C). Results were similar when adjusting for initial clinical sepsis subtype in an exploratory analysis (eTable 6). Table 2. Risk-adjusted association between mechanistic concepts and 90-day mortality Open in a new tab Adjusted odds ratios and p -values derived from multivariable logistic regression, controlling for age, sex, and race Contribution of mechanistic signatures to sepsis subtype membership Among the four clinical sepsis subtypes, we illustrated how each concept contributed to membership using random forest feature importance (Fig. 2 ), with principal component scores indicating directionality. For the α-type, low damage (component mean − 0.62 [SD: 0.6]) and high disease tolerance (component mean 0.23 [SD: 1.25]) contributed more than host resistance (component mean -0.02 [SD: 1.1]). In contrast, the δ-type was primarily associated with high damage (component mean 1.92 [SD: 2.81]). Low host resistance contributed most to membership in the β-type (component mean − 0.28 [SD: 0.48]). Fig. 2. Open in a new tab A Distribution of host resistance, disease tolerance, and damage concepts across clinical sepsis subtypes, illustrated as a chord diagram (left). B The heat map (top right) shows standardized Z ‑scores for the mechanistic concepts (host resistance, disease tolerance, and damage) across the four SENECA sepsis subtypes (α, β, γ, δ), C Bar plots (bottom right) display the mean principal component (PC1) scores for host resistance ( R ), disease tolerance ( T ), and damage ( D ) across the α, β, γ, and δ-types Sensitivity analyses Model results were robust to sensitivity analysis. First, after excluding [TIMP-2]–[IGFBP7] and uric acid for high missingness, there were no changes in risk-adjusted associations between damage (aOR 1.75, 95% CI 1.40–2.20, p < 0.001) and host resistance (aOR 0.83, 95% CI 0.54–1.10, p = 0.4) with mortality (eTable 7). Second, when using two-dimensional PCA, damage remained significantly associated with 90-day mortality (aOR = 1.56, p < 0.001, eTable 8). Third, when biomarkers were concept mapped by the proportion of expert agreement, the results were similar (eTable 9). Fourth, the exploratory four-group analysis of combined host resistance and disease tolerance supported the primary results: patients with both high host resistance and disease tolerance had the lowest mortality, whereas those with both low host resistance and low tolerance had the highest mortality (Fig. 3 , eTable 10, eTable 11). Finally, high and low strata of host resistance, disease tolerance, and damage were represented across all SOFA categories, suggesting these mechanistic signatures were not simply proxies for illness severity (eFigure 4). Fig. 3. Open in a new tab Kaplan–Meier 90-day survival curves for the four combinations of high/low host resistance ( R ) and high/low disease tolerance ( T ). The groups shown are High R + High T , High R + Low T , Low R + High T , and Low R + Low T Discussion In a prospective sepsis cohort, we operationalized circulating biomarker signatures of host resistance, disease tolerance, and damage and evaluated associations with outcomes and subtype membership. A greater damage signature was independently associated with increased 90-day mortality. δ-type membership was characterized by high damage and low disease tolerance signatures. Most sepsis conceptual models focus on pathogen-mediated injury or the predominant immune state (e.g ., hyperinflammatory, hypoinflammatory, immune-paralysis, innate vs. adaptive immunity). They may not fully capture why similar patients with similar pathogen loads experience distinct outcomes. A complementary approach is the framework described by Medzhitov et al., which includes host resistance to infection, disease tolerance, and damage. Herein, host survival during infection depends on three interrelated processes: (i) the ability to directly eliminate pathogens (host resistance) through innate and adaptive immune responses, (ii) the capacity to maintain organ function of the host without reducing pathogen load (disease tolerance) which is distinct from immune mediated pathogen clearance, and (iii) and the extent of tissue injury caused by both the pathogen and the host’s immune response (damage) [ 3 , 4 ]. This framework has been instrumental in furthering our understanding of diseases like malaria. In murine models of malaria, biomarkers that reflect disease tolerance such as heme oxygenase-1 (HO-1) and its heme-catabolizing activity, confer protection without reducing parasite burden [ 14 , 15 ]. In murine polymicrobial sepsis, detoxification of free heme and induction of HO-1 similarly improve survival at unchanged bacterial load [ 16 ]. To translate concepts to human data and extend prior work, we used a structured expert consensus process and expert adjudication to assign protein biomarkers to these domains for study in sepsis patients and biospecimens. In this work, after integrating damage-aligned biomarkers into a PCA-derived signature, higher damage was associated with greater 90-day mortality. The biomarkers selected as plausible indicators of host damage included Ang-2, Syn-1, HMGB1, and PAI-1. Notably, Ang-2 is associated with endothelial dysfunction [ 17 ], Syn-1 with vascular dysfunction [ 18 ], HMGB1 with endothelial barrier disruption [ 8 ] and PAI-1 with impaired coagulation and fibrinolysis [ 19 ]. Host resistance and disease tolerance signatures were similarly associated with outcome but did not reach statistical significance in adjusted models. Among the plausible constituent markers of host resistance, IL-6 is a cytokine that activates the acute phase response [ 20 ], IL-8 promotes neutrophil recruitment and E-selectin and ICAM-1 similarly promote endothelial–leukocyte adhesion [ 21 , 22 ]. Together, these reflect mechanisms that aim at accelerating pathogen clearance. Lastly, among the plausible constituent biomarkers of disease tolerance, IL-10 inhibits the production of pro-inflammatory cytokines [ 23 ], HO-1 catabolizes heme and is linked to improved outcomes in sepsis models [ 16 ], Ang-1 supports endothelial stability [ 24 ], ATIII contributes to endogenous anticoagulation and reduces microvascular thrombosis [ 25 ] and TIMP-2 and IGFBP7 are implicated in cell-cycle arrest which can be protective in the early phases of cellular stress [ 9 ]. Importantly, it is possible that signatures of host resistance and disease tolerance are underpowered in current models, require additional biomarkers, or reflect overlapping and compartmentalized mechanisms not fully captured in circulation. Notably, patients with concurrently low host resistance and low disease tolerance signatures had the highest mortality, consistent with a phenotype of reduced protective reserve and greater risk of severe disease trajectories. The recently proposed sepsis subtypes have been reproduced in multiple international cohorts of more than 60,000 patients [ 26 ]. Yet, the mechanisms that distinguish subtypes are less understood. We now observe that classification by mechanistic concepts—host resistance, disease tolerance, and damage—may unpack key drivers of sepsis subtype differentiation. For example, δ-type patients have the greatest mortality and also a high damage signature. This is consistent with prior studies of damage biomarkers including circulating danger-associated molecular patterns (DAMPs), markers of cell death, and endothelial alarmins such as HMGB1, cell-free hemoglobin, and angiopoietin-2 [ 27 , 28 ]. The δ-type patients exhibited low disease tolerance. This is hypothesis generating, but consistent with lower levels of hemopexin (a key heme scavenging protein) among patients with the worst sepsis outcomes [ 16 ]. Limitations There are multiple limitations of the study. First, as with all observational studies and biomarkers measured at a single time point, these results are subject to unmeasured confounding. Similarly, SENECA sepsis subtype assignments based on a single time point may be subject to misclassification, as highlighted in a recent work [ 13 ]. Second, in a single-center prospective cohort, generalizability has not yet been tested. Third, the classification of biomarkers into host resistance, disease tolerance, and damage was based on expert consensus which introduces subjectivity. Additionally, biomarker attribution may overlap across domains, depend on context, and may oversimplify complex sepsis biology. Fourth, disease tolerance mechanisms are proposed to be independent of pathogen burden, and the burden of pathogens was not measurable. Additionally, we did not have adjudicated, analyzable data on infection source, nor measures of antibiotic appropriateness, which may confound associations. Fifth, biomarker concentrations were measured in the urine and plasma, not at the level of the organ. For example, HO-1 exerts a cytoprotective effect when induced within parenchymal tissues and circulating HO-1 may be an imprecise surrogate for organ-level expression. Even though HO-1 plasma levels may reflect expression of HMOX1 (which plays a role in antimicrobial defense) [ 29 ], they may also reflect cytosolic enzyme release due to cell damage. Sixth, the results are hypothesis generating, requiring refinement of biomarker signatures, integration with multi-omic pathways [ 30 ], and validation in external cohorts. Conclusion In a prospective cohort study, the mechanistic concepts of host resistance, disease tolerance, and damage to the host contributed to clinical sepsis subtypes and were associated with 90-day mortality. Supplementary Information Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 2621 KB) (2.6MB, docx) Supplementary file2 (DOCX 35 KB) (35.1KB, docx) Author contributions Christopher W. Seymour, Rachel E. Powell, and Arnab Chowdhury conceived the study. Jason N. Kennedy and Arnab Chowdhury performed the statistical analyses. Arnab Chowdhury, Rachel E. Powell, and Christopher W. Seymour drafted the manuscript. Rachel E. Powell, Jason N. Kennedy, Kelly L. Urbanek, Derek C. Angus, Chung-Chou H. Chang, Lu Tang, Sebastian Weis, Michael Bauer, Manu Shankar-Hari, and Christopher W. Seymour contributed to study design, interpretation of the data, and critical revision of the manuscript for important intellectual content. All authors read and approved the final version of the manuscript and agree to be accountable for all aspects of the work. Funding Dr. Seymour was supported in part by grants R35GM119519 and R21GM144851 from the National Institute of General Medical Sciences (NIGMS), National Institutes of Health (NIH). Dr. Chowdhury was supported in part by training grant T32HL007820 from the National Heart, Lung, and Blood Institute (NHLBI), NIH. No additional external funding was received for this work. Role of the funder/sponsor : The funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication. Data availability The datasets generated and/or analyzed during the current study are not publicly available owing to institutional privacy and data-use restrictions, but are available from the corresponding author on reasonable request and with approval from the University of Pittsburgh. Code availability The R scripts used for data processing and statistical analysis are available from the corresponding author on reasonable request. Declarations Conflicts of interest The authors declare that they have no competing interests related to the submitted work. Ethical approval and consent to participate This study followed the STROBE reporting guideline for cohort studies. The protocol was reviewed and approved by the University of Pittsburgh Human Research Protection Office. Because the study used clinical data and remnant biospecimens collected as part of routine care and posed no more than minimal risk, the Institutional Review Board granted a waiver of informed consent with authorization under the Health Insurance Portability and Accountability Act (HIPAA). Consent for publication This manuscript does not contain any individual person’s identifiable data. Consent for publication is therefore not applicable. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References 1. DeMerle KM, Angus DC, Baillie JK et al (2021) Sepsis subclasses: a framework for development and interpretation. Crit Care Med 49:748–759. 10.1097/CCM.0000000000004842 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 2. Maslove DM, Tang B, Shankar-Hari M et al (2022) Redefining critical illness. Nat Med 28:1141–1148. 10.1038/s41591-022-01843-x [ DOI ] [ PubMed ] [ Google Scholar ] 3. Powell RE, Soares MP, Weis S (2023) What’s new in intensive care: disease tolerance. Intensive Care Med 49:1235–1237. 10.1007/s00134-023-07130-8 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 4. Medzhitov R, Schneider DS, Soares MP (2012) Disease tolerance as a defense strategy. Science 335:936–941. 10.1126/science.1214935 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 5. Soares MP, Gozzelino R, Weis S (2014) Tissue damage control in disease tolerance. Trends Immunol 35:483–494. 10.1016/j.it.2014.08.001 [ DOI ] [ PubMed ] [ Google Scholar ] 6. Weis S, Carlos AR, Moita MR et al (2017) Metabolic adaptation establishes disease tolerance to sepsis. Cell 169(1263–1275):e1214. 10.1016/j.cell.2017.05.031 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 7. Wang A, Huen SC, Luan HH et al (2016) Opposing effects of fasting metabolism on tissue tolerance in bacterial and viral inflammation. Cell 166(1512–1525):e1512. 10.1016/j.cell.2016.07.026 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 8. Denning NL, Aziz M, Gurien SD et al (2019) DAMPs and NETs in Sepsis. Front Immunol 10:2536. 10.3389/fimmu.2019.02536 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 9. Kellum JA, Chawla LS (2016) Cell-cycle arrest and acute kidney injury: the light and the dark sides. Nephrol Dial Transplant 31:16–22. 10.1093/ndt/gfv130 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 10. DeMerle KM, Kennedy JN, Peck Palmer OM et al (2021) Feasibility of embedding a scalable, virtually enabled biorepository in the electronic health record for precision medicine. JAMA Netw Open 4:e2037739. 10.1001/jamanetworkopen.2020.37739 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 11. Seymour CW, Liu VX, Iwashyna TJ et al (2016) Assessment of clinical criteria for sepsis: for the third international consensus definitions for sepsis and septic shock (sepsis-3). JAMA 315:762–774. 10.1001/jama.2016.0288 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 12. Seymour CW, Kennedy JN, Wang S et al (2019) Derivation, validation, and potential treatment implications of novel clinical phenotypes for sepsis. JAMA 321:2003–2017. 10.1001/jama.2019.5791 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 13. Kennedy JN, Iyer S, Nauka PC et al (2026) Fuzzy classification of sepsis subtypes and implications for trajectory and treatment. EBioMedicine 124:106125. 10.1016/j.ebiom.2026.106125 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 14. Pamplona A, Ferreira A, Balla J et al (2007) Heme oxygenase-1 and carbon monoxide suppress the pathogenesis of experimental cerebral malaria. Nat Med 13:703–710. 10.1038/nm1586 [ DOI ] [ PubMed ] [ Google Scholar ] 15. Seixas E, Gozzelino R, Chora A et al (2009) Heme oxygenase-1 affords protection against noncerebral forms of severe malaria. Proc Natl Acad Sci U S A 106:15837–15842. 10.1073/pnas.0903419106 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 16. Larsen R, Gozzelino R, Jeney V et al (2010) A central role for free heme in the pathogenesis of severe sepsis. Sci Transl Med 2:51ra71. 10.1126/scitranslmed.3001118 [ DOI ] [ PubMed ] [ Google Scholar ] 17. Hakanpaa L, Sipila T, Leppanen VM et al (2015) Endothelial destabilization by angiopoietin-2 via integrin beta1 activation. Nat Commun 6:5962. 10.1038/ncomms6962 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 18. Colbert JF, Schmidt EP (2016) Endothelial and microcirculatory function and dysfunction in sepsis. Clin Chest Med 37:263–275. 10.1016/j.ccm.2016.01.009 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 19. Madoiwa S, Nunomiya S, Ono T et al (2006) Plasminogen activator inhibitor 1 promotes a poor prognosis in sepsis-induced disseminated intravascular coagulation. Int J Hematol 84:398–405. 10.1532/IJH97.05190 [ DOI ] [ PubMed ] [ Google Scholar ] 20. Kishimoto T (2010) IL-6: from its discovery to clinical applications. Int Immunol 22:347–352. 10.1093/intimm/dxq030 [ DOI ] [ PubMed ] [ Google Scholar ] 21. Cambier S, Gouwy M, Proost P (2023) The chemokines CXCL8 and CXCL12: molecular and functional properties, role in disease and efforts towards pharmacological intervention. Cell Mol Immunol 20:217–251. 10.1038/s41423-023-00974-6 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 22. Ley K, Laudanna C, Cybulsky MI et al (2007) Getting to the site of inflammation: the leukocyte adhesion cascade updated. Nat Rev Immunol 7:678–689. 10.1038/nri2156 [ DOI ] [ PubMed ] [ Google Scholar ] 23. Latifi SQ, O’Riordan MA, Levine AD (2002) Interleukin-10 controls the onset of irreversible septic shock. Infect Immun 70:4441–4446. 10.1128/IAI.70.8.4441-4446.2002 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 24. Witzenbichler B, Westermann D, Knueppel S et al (2005) Protective role of angiopoietin-1 in endotoxic shock. Circulation 111:97–105. 10.1161/01.CIR.0000151287.08202.8E [ DOI ] [ PubMed ] [ Google Scholar ] 25. Levi M, Schouten M, van der Poll T (2008) Sepsis, coagulation, and antithrombin: old lessons and new insights. Semin Thromb Hemost 34:742–746. 10.1055/s-0029-1145256 [ DOI ] [ PubMed ] [ Google Scholar ] 26. van Amstel RBE, Kennedy JN, Scicluna BP et al (2023) Uncovering heterogeneity in sepsis: a comparative analysis of subphenotypes. Intensive Care Med 49:1360–1369. 10.1007/s00134-023-07239-w [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 27. Janz DR, Bastarache JA, Peterson JF et al (2013) Association between cell-free hemoglobin, acetaminophen, and mortality in patients with sepsis: an observational study. Crit Care Med 41:784–790. 10.1097/CCM.0b013e3182741a54 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 28. Kumpers P, Hafer C, David S et al (2010) Angiopoietin-2 in patients requiring renal replacement therapy in the ICU: relation to acute kidney injury, multiple organ dysfunction syndrome and outcome. Intensive Care Med 36:462–470. 10.1007/s00134-009-1726-7 [ DOI ] [ PubMed ] [ Google Scholar ] 29. Sponholz C, Huse K, Kramer M et al (2012) Gene polymorphisms in the heme degradation pathway and outcome of severe human sepsis. Shock 38:459–465. 10.1097/SHK.0b013e31826ae951 [ DOI ] [ PubMed ] [ Google Scholar ] 30. Cohn O, Yankovitz G, Peshes-Yaloz N et al (2022) Distinct gene programs underpinning disease tolerance and resistance in influenza virus infection. Cell Syst 13(1002–1015):e1009. 10.1016/j.cels.2022.11.004 [ DOI ] [ PubMed ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials Supplementary file1 (DOCX 2621 KB) (2.6MB, docx) Supplementary file2 (DOCX 35 KB) (35.1KB, docx) Data Availability Statement The datasets generated and/or analyzed during the current study are not publicly available owing to institutional privacy and data-use restrictions, but are available from the corresponding author on reasonable request and with approval from the University of Pittsburgh. The R scripts used for data processing and statistical analysis are available from the corresponding author on reasonable request. Articles from Intensive Care Medicine are provided here courtesy of Springer ACTIONS View on publisher site PDF (3.0 MB) Cite Collections Permalink RESOURCES Similar articles Cited by other articles Links to NCBI Databases Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top

Related documents

• Host Dysregulation-Driven Sepsis Prediction Model#853379
• Immune Biomarker Profiling Across Clinical States and Development of a SOFA-Integrated Model for 28-Day Mortality Prediction in Sepsis: A Prospective Cohort Study.#662672
• Reframing sepsis through resistance, disease tolerance and damage: a clinically meaningful conceptual shift?#284386
• Global burden, projections, and causal factors of maternal sepsis and other maternal infections: A comprehensive epidemiological and mendelian randomization study.#255344
• The Causal Effects of Inflammation and Coagulation-Related Biomarkers on Sepsis-Related Complications: A Mendelian Randomization Study.#182419