Skip to main content

Defining critical illness using immunological endotypes in patients with and without sepsis: a cohort study



Sepsis is a heterogenous syndrome with limited therapeutic options. Identifying immunological endotypes through gene expression patterns in septic patients may lead to targeted interventions. We investigated whether patients admitted to a surgical intensive care unit (ICU) with sepsis and with high risk of mortality express similar endotypes to non-septic, but still critically ill patients using two multiplex transcriptomic metrics obtained both on admission to a surgical ICU and at set intervals.


We analyzed transcriptomic data from 522 patients in two single-site, prospective, observational cohorts admitted to surgical ICUs over a 5-year period ending in July 2020. Using an FDA-cleared analytical platform (nCounter FLEX®, NanoString, Inc.), we assessed a previously validated 29-messenger RNA transcriptomic classifier for likelihood of 30-day mortality (IMX-SEV-3) and a 33-messenger RNA transcriptomic endotype classifier. Clinical outcomes included all-cause mortality, development of chronic critical illness, and secondary infections. Univariate and multivariate analyses were performed to assess for true effect and confounding.


Sepsis was associated with a significantly higher predicted and actual hospital mortality. At enrollment, the predominant endotype for both septic and non-septic patients was adaptive, though with significantly different distributions. Inflammopathic and coagulopathic septic patients, as well as inflammopathic non-septic patients, showed significantly higher frequencies of secondary infections compared to those with adaptive endotypes (p < 0.01). Endotypes changed during ICU hospitalization in 57.5% of patients. Patients who remained adaptive had overall better prognosis, while those who remained inflammopathic or coagulopathic had worse overall outcomes. For severity metrics, patients admitted with sepsis and a high predicted likelihood of mortality showed an inflammopathic (49.6%) endotype and had higher rates of cumulative adverse outcomes (67.4%). Patients at low mortality risk, whether septic or non-septic, almost uniformly presented with an adaptive endotype (100% and 93.4%, respectively).


Critically ill surgical patients express different and evolving immunological endotypes depending upon both their sepsis status and severity of their clinical course. Future studies will elucidate whether endotyping critically ill, septic patients can identify individuals for targeted therapeutic interventions to improve patient management and outcomes.


Sepsis remains one of the most common causes of mortality and morbidity in critically ill patients, affecting as many as 50 million individuals annually with case mortality rates as high as 40% [1]. Earlier recognition and near-universal implementation of sepsis protocols have improved in-hospital clinical outcomes; however, targeted therapies remain elusive [2,3,4].

Sepsis is defined as a dysregulated host immune response to infection resulting in life-threatening organ dysfunction [5, 6]. However, inherent to this definition is a wide range of insults and trajectories of physiologic decline. This disease heterogeneity likely explains the lack of efficacy in previous randomized controlled trials employing immune modulating therapeutics [7,8,9,10,11]. To address this heterogeneity, efforts have been made to classify patients based on constellations of observable characteristics and commonly available laboratory values, also called phenotypes [12,13,14,15]. However, phenotypes based on these clinical variables may not accurately discriminate differences in the underlying disease mechanisms, also called endotypes. Thus, efforts at phenotyping have not led to substantial changes in patient care or outcomes [16, 17].

Multiplex metabolomics, proteomics, and transcriptomics offer the potential to reveal a spectrum of sepsis endotypes, both illuminating common underlying mechanisms for immunological dyscrasia and providing potential therapeutic targets. Semantically, we choose the term “endotype” to highlight subphenotypes with distinct functional or pathobiological mechanisms amenable to targeted interventions and to contrast against clinically observable phenotypes. While the present classification schema has not been proven to be linked to treatment effect, it does align with previous research by our group and others [7, 18,19,20]. Previous research has identified 2–5 endotypes in diagnosed sepsis, though they vary with regards to domains, data sources, classification algorithms, statistical methodology, duration of observations, and stated goals [18, 21,22,23,24,25,26]. Cumulatively, however, these studies have sparked interest in re-defining aspects of critical illness in terms of underlying physiologic perturbations rather than phenotypic syndromes [9, 27].

In this study, we apply 29- and 33-gene transcriptomic signatures to simultaneously classify severity and endotype, respectively, within both septic and non-septic critically ill patients [25, 28,29,30]. These transcriptomic signatures were originally validated in non-surgical patients with bacterial or viral sepsis, and were classified into adaptative, inflammopathic, and coagulopathic endotypes based on gene ontology analysis [25]. We investigate whether patients admitted to a surgical ICU with sepsis and with high risk of mortality would express similar endotypes to non-septic, but still critically ill patients at-risk of developing sepsis. We hypothesize that patients admitted to a surgical ICU with sepsis and with high risk of mortality would express similar endotypes to non-septic, but still critically ill patients. In addition, we compare differences in endotype on admission between patients with predicted high severity by the transcriptomic metric versus the ground truth of those who clinically developed adverse outcomes. We also examine how these endotypes evolve over time in critically ill patients, tracking those who either rapidly recover versus those who experience adverse outcomes–defined as all-cause (in-hospital, 30-, 90-day) mortality, development or absence of chronic critical illness (CCI), secondary infections, and poor discharge disposition.

Materials and methods

Study designs

This post hoc study performed transcriptomic analyses on samples from two single-site, prospective, observational cohorts that enrolled a total of 522 patients admitted to non-cardiac, surgical ICUs and were classified as either (1) critically ill patients with a diagnosis of sepsis (septic) or (2) non-septic critically ill patients, at high risk of subsequently developing sepsis (at risk or non-septic; Fig. 1) [28, 29]. As a post hoc analysis, it was not powered for any specific outcome. Data and additional samples were obtained from the University of Florida CTSA Biorepository, a resource available to the scientific community [31]. In the first study (INF-05) [29], the parent cohort included 363 patients admitted to a surgical ICU between January 2015 and January 2020 with a diagnosis of sepsis (NCT02276417). Sepsis cohort inclusion criteria were: (1) age greater than or equal to 18 years, (2) clinical diagnosis of sepsis as defined by 2001 consensus guidelines, and (3) entrance into the electronic health record (EHR)-based sepsis clinical management protocol. Although prospectively enrolled using 2001 sepsis consensus criteria, these patients were retrospectively re-adjudicated and reclassified using Sepsis-3 consensus definitions [5, 32]. Detailed descriptions of the inclusion and exclusion criteria are contained in Additional file 1: Supplemental Materials: Methods.

Fig. 1
figure 1

Study inclusion criteria. Study population was derived from two single-site, prospective, observational studies that enrolled a total of 522 patients admitted to a non-cardiac, surgical ICUs. *All data points are at time of enrollment. Therefore, the 11 crossover patients were included in the non-septic group since they were not septic at the time of enrollment

The second prospective diagnostic and prognostic study (INF-06) was conducted between July 2020 and July 2021 [28] and enrolled two cohorts of critically ill patients at the time of surgical ICU admission (NCT04414189). Comparisons between INF-05 and INF-06 are shown in Additional file 1: Table S1. One cohort included patients with a suspected diagnosis of sepsis admitted to the ICU for protocolized sepsis management, as in the aforementioned study. Sepsis was defined according to Sepsis-3 criteria. The second cohort included critically ill patients without sepsis (e.g., severely injured trauma patients, post-operative patients, patients admitted directly to ICU from emergency department, see Additional file 1: Table S2). Inclusion and exclusion criteria, study design, and cohort flow are contained in Fig. 1 with a more detailed flow diagram shown in Additional file 1: Fig. S1.

In both studies, all enrolled subjects underwent post hoc adjudication by physician-investigators within one week of cohort enrollment to confirm sepsis diagnosis, severity, and source. Hospital-acquired secondary infections were adjudicated by physician-investigators during primary data/chart review utilizing current United States Centers for Disease Control definitions and guidelines [10].

Individual clinical outcome variables included all-cause (in-hospital, 30-, 90-day) mortality, development or absence of chronic critical illness (CCI), secondary infections, and poor discharge disposition. Inpatient clinical trajectory was defined as “early death,” “rapid recovery,” or “CCI.” CCI was defined as an ICU length of stay greater than or equal to 14 days with evidence of persistent organ dysfunction (SOFA score ≥ 2) [33, 34]. Hospitalized patients who died after an ICU length of stay greater than 14 days from the index hospitalization were also classified as CCI. Rapid recovery patients were those discharged from the ICU within 14 days following resolution of organ dysfunction. Poor disposition was defined as discharge to a skilled nursing facility, long-term acute care facility, or hospice. Our study was performed in accordance with the STROBE guidelines.

Sample collections

Blood samples were collected in PAXgene™ blood RNA tubes within 12–24 h of ICU admission and were stored at  − 80 °C for subsequent analysis. In the second study, additional blood samples were also collected on ICU days 4 and 7 and weekly thereafter during ICU stay (Additional file 1: Methods). RNA was extracted with the RNeasy® Plus Micro Kit (QIAGEN, Germantown, MD). The IMX-SEV-3 severity and the 33-mRNA endotyping classifiers were quantitated simultaneously from 200 ng of RNA input using the 510(k)-cleared NanoString nCounter FLEX™ profiler (NanoString, Seattle, WA) according to a validated standard operating protocol in a Clinical Laboratory Improvement Amendments-certified diagnostic laboratory (UF Health Medical Laboratories at Rocky Point, Gainesville, FL) by licensed laboratory technicians.

Severity and endotype classification

The probability of an adverse clinical outcome (in-hospital, 30-, and 90-day mortality, development of CCI and discharge disposition) was estimated by a 29 host-messenger RNA (mRNA) test (IMX-SEV-3, Inflammatix Inc., Sunnyvale, CA) that uses a machine learning algorithm to report results as both a continuous variable and stratified “risk bands” to meet clinically actionable performance thresholds: “low,” “moderate,” and “high” likelihood of 30-day mortality (see Additional file 1: Fig. S2) [35]. Severity classification was performed using supervised multi-layer perceptron (MLP) models as previously described [30].

Classification into three endotypes was computed from the whole blood expression of 33 host immune mRNAs using a previously published classifier [25, 28, 36]. These endotypes—adaptive, inflammopathic, and coagulopathic—were derived from the difference of geometric means of gene expression for each of three modules. The inflammopathic module comprises the expression of ARG1, LCN2, LTF, OLFM4, and HLA-DMB; the coagulopathic module comprises KCNMB4, CRISP2, HTRA1, PPL, RHBDF2, ZCCHC4, YKT6, DDX6, SENP5, RAPGEF1, DTX2, and RELB, and the adaptive module comprises YKT6, PDE4B, TWISTNB, BTN2A2, ZBTB33, PSMB9, CAMK4, TMEM19, SLC12A7, TP53BP1, PLEKHO1, SLC25A22, FRS2, GADD45A, CD24, S100A12, and STX1A expression. An overall endotype assignment for each subject was calculated using a 3-class logistic regression model which takes as input the three modules and generates a probability of endotype assignment {for each subject, the total probability [p(Inflammopathic) + p(Adaptive) + p(Coagulopathic)] sums to 1}. Each sample is assigned an endotype according to the highest probability. Numerical values are presented in the Additional file 1: Table S2 [25].

Total leukocyte and absolute lymphocyte counts (ALCs) were determined at the University of Florida Health Clinical and Diagnostic Laboratories. Plasma IL-6 levels were determined using the Luminex MagPix® platform (Austin, TX).

Statistical analysis

Descriptive data are presented as frequencies and percentages or means and standard deviations (SD). The Fisher exact test or Pearson's Chi-squared test and t-test were used for comparison of categorical and continuous variables, respectively. All significance tests were two sided, with a raw p ≤ 0.05 considered statistically significant. Significance levels 0.05 > p > 0.01 are reported precisely; 0.01 > p > 0.001 are reported as ‘p < 0.01,’ and lower values are all reported as ‘p < 0.001’. Univariable and multivariable logistic regressions were performed, controlling for age, sex, WBC, IL-6, SOFA, endotype, Charlson Comorbidity Index, and septic status. Analyses were performed using the R Project statistical package, version 4.2.0 (R Project for Statistical Computing).

Study approval

Ethics approvals were obtained from the University of Florida Institutional Review Board (IRB#201400611 and IRB#201702261). Informed consent was obtained from each subject or their surrogate decision-maker. Self-reported or proxy-reported race and ethnicity category data were collected as per National Institutes of Health reporting guidelines and requirements.


Septic and non-septic cohorts

The overall analytic cohort consisted of 522 critically ill patients from the two consecutive, prospective observational studies (Fig. 1). Prediction of sepsis severity and endotype analyses were conducted on 377 septic and 145 non-septic patients within 24 h of ICU admission (Table 1). A subset of septic (N = 51) and all non-septic (N = 145) patients had repeat blood sampling at designated intervals over their ICU stay. Three hundred and twenty-six (86%) septic patients were drawn from the initial cohort (INF-05), and all non-septic patients were drawn from the second cohort (INF-06) [28, 29]. Demographics of included patients are shown in Table 1, while Table 2 shows outcomes, endotypes, and severity predictions of the two critically ill cohorts.

Table 1 Patient demographics at enrollment
Table 2 Clinical outcomes, endotypes, and severity predictions

As expected, critically ill patients admitted to the ICU with sepsis had significantly higher SOFA and Charlson Comorbidity scores compared to the non-septic cohort, indicating more severe organ dysfunction and greater number of comorbidities. As shown in Table 2, poorer outcomes were observed among the septic cohort, including a higher incidence of secondary infection (30.2 vs. 8.3%, p < 0.001), development of CCI (32.4 vs. 6.9%, p < 0.001), poor discharge disposition (40.4 vs. 16.0%, p < 0.001), in-hospital mortality (7.4 vs. 2.1%, p = 0.02), 30-day (10.2 vs. 4.1%, p = 0.03), and 90-day (16.8 vs. 5.5%, p < 0.01) mortality.

Endotype distributions and outcomes

Endotype distributions were significantly different between septic and non-septic groups (Table 2). In both septic and non-septic cohorts, the adaptive endotype was most frequent, although it was more common in non-septic patients (40.1% vs. 51%). The inflammopathic endotype was second most common in septic patients and third in non-septic patients (34.2 vs. 15.9%) (Table 2). However, septic patients (n = 377) had different clinical outcomes depending upon their endotype at admission (Table 3). Inflammopathic and coagulopathic septic patients had a significantly higher frequency of secondary infections (37% each) compared to septic patients with an adaptive endotype (20%, p < 0.01). Similar increases in the frequency of secondary infections were seen in the inflammopathic non-septic patients (26%) versus patients with coagulopathic (4%) or adaptive (5%) endotypes (p < 0.01). Thirty-day mortality, CCI, and adverse discharge disposition did not reach statistical significance.

Table 3 Endotypes and outcomes on ICU admission

To examine whether endotype at baseline is associated with different patient outcomes, a multivariable logistic regression was conducted by including endotypes and other clinically relevant factors into the model. Of interest, patients with inflammopathic (OR 2.4, 95% CI 1.4–4.1, p < 0.001) and coagulopathic endotypes (OR 1.9, 95% CI 1.1–3.1, p = 0.014) had higher odds of having an adverse outcome compared to those with the adaptive endotype (Additional file 1: Table S3 and Fig. S3).

Endotype transitions

Figure 2 illustrates endotype distributions and transitions over time until death or hospital discharge. Measurements for both the septic (n = 52, Group 1) and non-septic patients (n = 145, Group 2) were obtained only from the second cohort (INF-06). 61 patients had at least one missing value, with 20% of data missing secondary to declined blood draw and 7% due to inadequate samples, labeling errors, or staff unavailability. Endotypes changed in 57.5% of patients during their hospitalization; of the remaining, 19% remained adaptative, 4% inflammopathic, and 3% coagulopathic.

Fig. 2
figure 2

Alluvial Plots of Immunological Endotypes as they Change Over Time in Septic Patients (Group 1) and Non-septic Patients (Group 2). Measurements for both the septic (n = 52, Group 1) and non-septic patients (n = 145, Group 2) were obtained only from the second cohort (INF-06). 74% of patients changed endotypes during their hospitalization, 19% remained adaptative, 3.5% inflammopathic, and 3.5% coagulopathic. “Switched” is defined as those that transitioned into sepsis

We assessed pooled outcome data between septic and non-septic patients obtained after the last collected endotype measurement (Table 4). Based on similar clinical behavior and worse overall clinical outcomes, we also pooled inflammopathic and coagulopathic endotypes. In most cases, the final endotype assessment was drawn on day 7 or 10 of ICU admission. There were significant differences in prognosis among the classes depending on endotype trajectory; patients who remained adaptive (N = 60) had the best nominal outcomes across all endpoints measured, though these were not individually tested. There were non-significant differences between those who transitioned either to or from adaptive endotype.

Table 4 Change in endotypes over time in ICU and subsequent outcomes

Endotypes and predicted severity

To better control for disease severity when comparing endotypes, we employed the severity transcriptomic metric (IMX-SEV-3) and found that endotypes were imbalanced across severity metrics (Fig. 3). Patients predicted to be low severity (n = 54), independent of their ICU admission cause, were near universally adaptive (98%): only one patient expressed a coagulopathic endotype while the patients with moderate severity prediction by IMX-SEV-3 continued to favor adaptive versus inflammopathic and coagulopathic endotypes (septic cohort: 45 vs. 26 vs. 29%, respectively; non-septic cohort: 48 vs. 15 vs. 38%, respectively). In contrast, those patients with high severity prediction based on IMX-SEV-3 were inflammopathic or coagulopathic in the septic (72% vs. 28%) and inflammopathic in the non-septic (100% vs. 0%) cohorts. We noted that inflammopathic (n = 69) and coagulopathic patients (n = 25) with a high risk of predicted mortality by IMX-SEV-3 appeared clinically similar, with nonsignificant differences in SOFA score, secondary infection, CCI, adverse outcomes, or mortality. The only noted difference was that inflammopathic patients demonstrated significantly higher plasma IL-6 concentrations than their coagulopathic counterparts (1870 vs. 642 pg/ml, p < 0.01; Additional file 1: Table S4).

Fig. 3
figure 3

Immunological endotypes of sepsis and at-risk sepsis patients based on their risk of mortality using the IMX-SEV severity prediction model. Sepsis (N = 377) and non-septic (N = 154) patients were stratified based on their mortality prediction model and assigned to either adaptive, inflammopathic, or coagulopathic. Importantly, patients stratified into the low risk of mortality using the IMX severity index were uniformly adaptive, while patients assigned to high risk of mortality were near uniformly inflammopathic or coagulopathic, regardless of whether they were septic or at risk of sepsis


Key findings

This post hoc analysis of a septic and non-septic cohorts of critically ill surgical patients showed similar endotype profiles regardless of Sepsis-3 criteria, with an inflammopathic endotype on admission corresponding to high severity and worse outcomes by composite measure. Endotypes transitions occurred frequently during hospital admission. We found no statistically significant differences in mortality.


Recent advances in sepsis endotyping research benefit from standard comparisons between studies, as advocated by De Merle et al. [7]. By observing endotypes in other septic and non-septic patients, we join efforts to redefine sepsis nosology as a heterogenous condition that shares characteristics across the spectrum of critical illness [9].

Our study shares commonalities and differences with others. The concept of expression-based sepsis subclasses dates to 2009 [37], though the last five years have witnessed an explosion of interest in this field: the MARS consortium investigated and validated patient endotypes in sepsis and identified four groups (MARS1-4) from the expression of 140 genes from 787 septic patients [24]; research by Davenport et al. [38] identified two distinct response signatures reflecting degree of immunosuppression in sepsis patients; and a study by Baghela et al. [18] validated five distinct gene expression profiles across several hospital systems, clustering patients into neutrophilic-suppressive, inflammatory, innate host defense, interferon, and adaptive. These authors assessed mortality using values drawn at a single time point within 24 h of admission and examined the biological plausibility of the identified genes known to cytokine signaling, cell proliferation, and lymphocyte and metabolic pathways, among others. Selected genes varied between studies, possibly due to differences in cohorts or in the classification techniques used to derive the groups [12, 25]. While these expression profiles carried prognostic significance, these groups did not examine changes in gene expression over the hospital course of illness, nor was there follow-up beyond 14 days.

Several studies, however, have analyzed gene expression profiles at different points during admission and following discharge. In patients expected to require at least 72 h of mechanical ventilation, a follow-up study of the PREVAIL trial assessed samples at days 1, 3, 6, 14, 21, and 28 to differentiate septic and non-septic patients using a novel scoring mechanism [23]. While they demonstrated changes in gene expression profiles through patient admission, they did not analyze outcomes. Similarly, Burnham et al. [39] showed that 46% of patients with community-acquired pneumonia and fecal peritonitis had changes to their gene expression profile on days 1, 3, and 5 of admission. Moreover, while patients who transitioned to the more critically ill group had nominally higher mortality rates, this is was not specifically analyzed. However, Cano-Gamez et al. [40], from the same institution, did demonstrate that patients with the largest decrease in genomic sepsis response had the lowest mortality rate. In a pediatric population, Wong et al. [41] showed that 42% of patients transitioned endotypes, and that those who remained in the more severe class had increased odds of mortality with administration of steroids. Finally, Kwok et al. [19] also examined expression patterns in convalescent samples 6 months after the septic event, finding persistent granulocytic dysfunction.

Current work

We were able to both identify and track changes in gene expression profile and severity scores over the course of acute critical illness in an adult, critically ill, surgical patient population. While admission endotype appeared to be the strongest predictor of outcomes, the extensive crossover noted between days 2 and 7 suggests that it may be valuable to continue assessing gene expression profiles, rather than focus on a single timepoint. We also noted that very few patients remained inflammatory or coagulopathic throughout their stay, suggesting a transitory maladaptation. This permits monitoring for resolution of immunologic dyscrasia, severity of condition, as well as possible responses to therapy.

Our study recapitulates observations about the 33-mRNA endotypes shown in previous investigations [22, 25, 42]. In contrast with previous investigations, however, we found that inflammopathic and coagulopathic patients had more similarities than differences, perhaps representing a single endotype. When pooled together, we showed that patients who presented with inflammopathic or coagulopathic endotype had increased incidence of adverse outcomes and secondary infections, though differences in in-hospital, 30-day, and 90-day mortality did not reach statistical significance.

In addition to analyzing septic patients, we also included a non-septic, critically ill cohort. While inflammopathic patterns had higher rates of secondary infections regardless of sepsis status, there were no overall changes in mortality or poor discharge disposition. Interestingly, both inflammopathic and coagulopathic subjects in the high-severity risk category had similar outcomes. These results may contribute to the understanding of sepsis as a part of a spectrum of critical illness rather than a separate entity.

Finally, this study applied the endotyping signature in a surgical cohort, while prior evaluations have mostly been in medical, bacterial sepsis, or COVID-19 patients [9, 18, 24, 37, 39, 41]. A recent report suggested the potential for endotypes to underpin different forms of critical illness [9]: a possibility that an ‘inflammopathic’ COVID-19 patient may be similar to an ‘inflammopathic’ surgical sepsis patient in molecular pathophysiology, further contributing to the idea of sepsis as a critical illness subtype.


We note several limitations to our study. First, this study was performed at a single institution with a predominately Caucasian patient population and may lack generalizability. However, both the IMX-SEV-3 severity and the endotyping classifier have been validated multiple times in external hospitals with similar results [25, 36, 42]. Second, our non-septic cohort was broadly defined and with lower overall APACHE II scores. Age, gender, and Charlson comorbidity index were similar between the cohorts. However, when controlling for high-severity risk, we noted similar demographic and patient characteristics between the cohorts. Third, the majority (86%) of septic patients were derived from the initial cohort. These patients generally had higher SOFA scores and rates of CCI, with similar discharge disposition, complications, and mortality to the septic patients recruited in the second cohort. Fourth, as the first cohort was recruited from 2015 until 2020, there is the possibility of data drift, though standard of care for septic patients did not change during that period for our institution. Fifth, the multiple time series population contained only 196 patients, limiting our ability to draw conclusions based on trends and outcomes; and, as common to the literature in transcriptomics, this study is limited as a post-hoc analysis of an existing dataset and may not be powered for a specific outcome, though we have shown significance in several areas. Another important caveat is that with three endotypes, two cohorts, and multiple outcomes measures, we present numerous hypotheses in this manuscript, and we chose not to apply a multiple-hypothesis correction for ease of readership. Larger prospective studies are needed. Sixth, our findings regarding outcomes in final endotype measurements may not be representative of their endotype closer to the outcome measure, as day 10 measurements may have less impact on 30- and 90-day mortality. Finally, this paper did not seek to investigate the biological underpinnings of the mRNAs used in the two classifiers and their relation to pathophysiology; this has been done elsewhere [21, 25].

Future directions

Results from this study and others could assist in paving the way for personalization of sepsis treatment. By monitoring heterogenous, pathophysiologic responses to therapy, clinicians and researchers may be able to “divide and conquer” the sepsis syndrome and perhaps redefine sepsis along a spectrum of critical illness rather than as a separate entity. Current work into both immunosuppressant and immunostimulant therapies would benefit from targeting specific endotypes. The results of this study may be incorporated into randomized controlled trials or advanced causal analysis techniques employing observational data. From a prognostic standpoint, the conduct of similar endotyping on patients following discharge could also inform our clinical outreach efforts in diverting resources to those with greater follow-up needs.


Critically ill surgical patients with and without sepsis express different immunological endotypes. These endotypes are dynamic across a patient’s admission and are associated with distinct outcomes, and transitions between them may inform patient prognosis and care. Having identified differences among the patient groups using an endotyping classifier, future prospective studies are needed evaluate differences in therapeutic response between the classes.

Availability of data and materials

The complete raw datasets generated and/or analyzed during the current study are maintained and are available at the UF Clinical and Translational Science Institute Biorepository ( Requests for access to the data are made to the Biorepository directly who will provide a complete deidentified dataset containing both the clinical and transcriptomic data upon request (27).



Absolute lymphocyte count


Acute physiology and chronic health evaluation


Chronic critical illness


Intensive care unit


Standard deviation


Sequential organ failure assessment


White blood cell




  1. Rudd KE, Johnson SC, Agesa KM, Shackelford KA, Tsoi D, Kievlan DR, Colombara DV, Ikuta KS, Kissoon N, Finfer S, et al. Global, regional, and national sepsis incidence and mortality, 1990–2017: analysis for the global burden of disease study. Lancet. 2020;395(10219):200–11.

    Article  PubMed  PubMed Central  Google Scholar 

  2. Cecconi M, Evans L, Levy M, Rhodes A. Sepsis and septic shock. Lancet. 2018;392(10141):75–87.

    Article  PubMed  Google Scholar 

  3. Herrán-Monge R, Muriel-Bombín A, García-García MM, Merino-García PA, Cítores-González R, Fernández-Ratero JA, Albalá N, Carriedo D, Moradillo-González S, Álvarez-Martínez B, et al. mortality reduction and long-term compliance with surviving sepsis campaign: a nationwide multicenter study. Shock. 2016;45(6):598–606.

    Article  PubMed  Google Scholar 

  4. van Zanten AR, Brinkman S, Arbous MS, Abu-Hanna A, Levy MM, de Keizer NF. Guideline bundles adherence and mortality in severe sepsis and septic shock. Crit Care Med. 2014;42(8):1890–8.

    Article  PubMed  Google Scholar 

  5. Seymour CW, Liu VX, Iwashyna TJ, Brunkhorst FM, Rea TD, Scherag A, Rubenfeld G, Kahn JM, Shankar-Hari M, Singer M, et al. Assessment of clinical criteria for sepsis: for the third international consensus definitions for sepsis and septic shock (sepsis-3). JAMA. 2016;315(8):762–74.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  6. Shankar-Hari M, Phillips GS, Levy ML, Seymour CW, Liu VX, Deutschman CS, Angus DC, Rubenfeld GD, Singer M. Developing a new definition and assessing new clinical criteria for septic shock: for the third international consensus definitions for sepsis and septic shock (sepsis-3). JAMA. 2016;315(8):775–87.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. DeMerle KM, Angus DC, Baillie JK, Brant E, Calfee CS, Carcillo J, Chang CH, Dickson R, Evans I, Gordon AC, et al. Sepsis subclasses: a framework for development and interpretation. Crit Care Med. 2021;49(5):748–59.

    Article  PubMed  PubMed Central  Google Scholar 

  8. Hotchkiss RS, Moldawer LL, Opal SM, Reinhart K, Turnbull IR, Vincent J-L. Sepsis and septic shock. Nat Rev Dis Primers. 2016;2(1):16045.

    Article  PubMed  PubMed Central  Google Scholar 

  9. Maslove DM, Tang B, Shankar-Hari M, Lawler PR, Angus DC, Baillie JK, Baron RM, Bauer M, Buchman TG, Calfee CS, et al. Redefining critical illness. Nat Med. 2022;28(6):1141–8.

    Article  CAS  PubMed  Google Scholar 

  10. Vincent JL. The clinical challenge of sepsis identification and monitoring. PLoS Med. 2016;13(5):e1002022.

    Article  PubMed  PubMed Central  Google Scholar 

  11. Vincent JL, van der Poll T, Marshall JC. The end of “one size fits all” sepsis therapies: toward an individualized approach. Biomedicines. 2022;10:9.

    Article  Google Scholar 

  12. Baek MS, Kim JH, Kwon YS. Cluster analysis integrating age and body temperature for mortality in patients with sepsis: a multicenter retrospective study. Sci Rep. 2022;12(1):1090.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  13. Xu Z, Mao C, Su C, Zhang H, Siempos I, Torres LK, Pan D, Luo Y, Schenck EJ, Wang F. Sepsis subphenotyping based on organ dysfunction trajectory. Crit Care. 2022;26(1):197.

    Article  PubMed  PubMed Central  Google Scholar 

  14. Yehya N, Fitzgerald JC, Hayes K, Zhang D, Bush J, Koterba N, Chen F, Tuluc F, Teachey DT, Balamuth F, et al. Temperature trajectory sub-phenotypes and the immuno-inflammatory response in pediatric sepsis. Shock. 2022;57(5):645–51.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  15. Zhao H, Kennedy JN, Wang S, Brant EB, Bernard GR, DeMerle K, Chang CH, Angus DC, Seymour CW. Revising host phenotypes of sepsis using microbiology. Front Med. 2021;8:775511.

    Article  Google Scholar 

  16. Seymour CW, Kennedy JN, Wang S, Chang CH, Elliott CF, Xu Z, Berry S, Clermont G, Cooper G, Gomez H, et al. Derivation, validation, and potential treatment implications of novel clinical phenotypes for sepsis. JAMA. 2019;321(20):2003–17.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  17. Sweeney TE, Perumal TM, Henao R, Nichols M, Howrylak JA, Choi AM, Bermejo-Martin JF, Almansa R, Tamayo E, Davenport EE, et al. A community approach to mortality prediction in sepsis via gene expression analysis. Nat Commun. 2018;9(1):694.

    Article  PubMed  PubMed Central  Google Scholar 

  18. Baghela A, Pena OM, Lee AH, Baquir B, Falsafi R, An A, Farmer SW, Hurlburt A, Mondragon-Cardona A, Rivera JD, 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 

  19. Kwok AJ, Allcock A, Ferreira RC, Cano-Gamez E, Smee M, Burnham KL, Zurke YX, McKechnie S, Mentzer AJ, Monaco C, et al. Neutrophils and emergency granulopoiesis drive immune suppression and an extreme response endotype during sepsis. Nat Immunol. 2023;24(5):767–79.

    Article  CAS  PubMed  Google Scholar 

  20. Leligdowicz A, Matthay MA. Heterogeneity in sepsis: new biological evidence with clinical applications. Crit Care. 2019;23(1):80.

    Article  PubMed  PubMed Central  Google Scholar 

  21. He YD, Wohlford EM, Uhle F, Buturovic L, Liesenfeld O, Sweeney TE. The optimization and biological significance of a 29-host-immune-mRNA panel for the diagnosis of acute infections and sepsis. J Pers Med. 2021;11:8.

    Article  Google Scholar 

  22. Iglesias J, Vassallo AV, Liesenfeld O, Levine JS, Patel VV, Sullivan JB, Cavanaugh JB, Elbaga Y, Sweeney TE. 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;11:1.

    Article  Google Scholar 

  23. Maslove DM, Shapira T, Tyryshkin K, Veldhoen RA, Marshall JC, Muscedere J. Validation of diagnostic gene sets to identify critically ill patients with sepsis. J Crit Care. 2019;49:92–8.

    Article  CAS  PubMed  Google Scholar 

  24. Scicluna BP, van Vught LA, Zwinderman AH, Wiewel MA, Davenport EE, Burnham KL, Nürnberg P, Schultz MJ, Horn J, Cremer OL, et al. Classification of patients with sepsis according to blood genomic endotype: a prospective cohort study. Lancet Respir Med. 2017;5(10):816–26.

    Article  PubMed  Google Scholar 

  25. Sweeney TE, Azad TD, Donato M, Haynes WA, Perumal TM, Henao R, Bermejo-Martin JF, Almansa R, Tamayo E, Howrylak JA, et al. Unsupervised analysis of transcriptomics in bacterial sepsis across multiple datasets reveals three robust clusters. Crit Care Med. 2018;46(6):915–25.

    Article  PubMed  PubMed Central  Google Scholar 

  26. Bodinier M, Monneret G, Casimir M, Fleurie A, Conti F, Venet F, Cazalis MA, Cerrato E, Peronnet E, Rimmelé T, et al. Identification of a sub-group of critically ill patients with high risk of intensive care unit-acquired infections and poor clinical course using a transcriptomic score. Crit Care. 2023;27(1):158.

    Article  PubMed  PubMed Central  Google Scholar 

  27. Heijnen NFL, Hagens LA, Smit MR, Cremer OL, Ong DSY, van der Poll T, van Vught LA, Scicluna BP, Schnabel RM, van der Horst ICC, 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(12):1503–11.

    Article  PubMed  Google Scholar 

  28. Brakenridge SC, Chen UI, Loftus T, Ungaro R, Dirain M, Kerr A, Zhong L, Bacher R, Starostik P, Ghita G, et al. Evaluation of a multivalent transcriptomic metric for diagnosing surgical sepsis and estimating mortality among critically Ill patients. JAMA Netw Open. 2022;5(7):e2221520.

    Article  PubMed  PubMed Central  Google Scholar 

  29. Brakenridge SC, Starostik P, Ghita G, Midic U, Darden D, Fenner B, Wacker J, Efron PA, Liesenfeld O, Sweeney TE, et al. A transcriptomic severity metric that predicts clinical outcomes in critically Ill surgical sepsis patients. Crit Care Explor. 2021;3(10):e0554.

    Article  PubMed  PubMed Central  Google Scholar 

  30. Mayhew MB, Buturovic L, Luethy R, Midic U, Moore AR, Roque JA, Shaller BD, Asuni T, Rawling D, Remmel M, et al. A generalizable 29-mRNA neural-network classifier for acute bacterial and viral infections. Nat Commun. 2020;11(1):1177.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  31. UF Sepsis and Critical Illness Research Center (SCIRC) Biospecimen and Clinical Data Archive []

  32. Levy MM, Fink MP, Marshall JC, Abraham E, Angus D, Cook D, Cohen J, Opal SM, Vincent JL, Ramsay G. 2001 SCCM/ESICM/ACCP/ATS/SIS international sepsis definitions conference. Crit Care Med. 2003;31(4):1250–6.

    Article  PubMed  Google Scholar 

  33. Stortz JA, Cox MC, Hawkins RB, Ghita GL, Brumback BA, Mohr AM, Moldawer LL, Efron PA, Brakenridge SC, Moore FA. Phenotypic heterogeneity by site of infection in surgical sepsis: a prospective longitudinal study. Crit Care. 2020;24(1):203.

    Article  PubMed  PubMed Central  Google Scholar 

  34. Stortz JA, Mira JC, Raymond SL, Loftus TJ, Ozrazgat-Baslanti T, Wang Z, Ghita GL, Leeuwenburgh C, Segal MS, Bihorac A, et al. Benchmarking clinical outcomes and the immunocatabolic phenotype of chronic critical illness after sepsis in surgical intensive care unit patients. J Trauma Acute Care Surg. 2018;84(2):342–9.

    Article  PubMed  PubMed Central  Google Scholar 

  35. Ducharme J, Self WH, Osborn TM, Ledeboer NA, Romanowsky J, Sweeney TE, Liesenfeld O, Rothman RE. A Multi-mRNA host-response molecular blood test for the diagnosis and prognosis of acute infections and sepsis: proceedings from a clinical advisory panel. J Pers Med. 2020;10:4.

    Article  Google Scholar 

  36. Galtung N, Diehl-Wiesenecker E, Lehmann D, Markmann N, Bergström WH, Wacker J, Liesenfeld O, Mayhew M, Buturovic L, Luethy R, et al. Prospective validation of a transcriptomic severity classifier among patients with suspected acute infection and sepsis in the emergency department. Eur J Emerg Med. 2022;29(5):357–65.

    Article  PubMed  PubMed Central  Google Scholar 

  37. Wong HR, Wheeler DS, Tegtmeyer K, Poynter SE, Kaplan JM, Chima RS, Stalets E, Basu RK, Doughty LA. Toward a clinically feasible gene expression-based subclassification strategy for septic shock: proof of concept. Crit Care Med. 2010;38(10):1955–61.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  38. Davenport EE, Burnham KL, Radhakrishnan J, Humburg P, Hutton P, Mills TC, Rautanen A, Gordon AC, Garrard C, Hill AV, et al. Genomic landscape of the individual host response and outcomes in sepsis: a prospective cohort study. Lancet Respir Med. 2016;4(4):259–71.

    Article  PubMed  PubMed Central  Google Scholar 

  39. Burnham KL, Davenport EE, Radhakrishnan J, Humburg P, Gordon AC, Hutton P, Svoren-Jabalera E, Garrard C, Hill AVS, Hinds CJ, et al. Shared and distinct aspects of the sepsis transcriptomic response to fecal peritonitis and pneumonia. Am J Respir Crit Care Med. 2017;196(3):328–39.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  40. Cano-Gamez E, Burnham KL, Goh C, Allcock A, Malick ZH, Overend L, Kwok A, Smith DA, Peters-Sengers H, Antcliffe D, et al. An immune dysfunction score for stratification of patients with acute infection based on whole-blood gene expression. Sci Transl Med. 2022;14(669):eabq44339.

    Article  Google Scholar 

  41. Wong HR, Cvijanovich NZ, Anas N, Allen GL, Thomas NJ, Bigham MT, Weiss SL, Fitzgerald JC, Checchia PA, Meyer K, et al. Endotype transitions during the acute phase of pediatric septic shock reflect changing risk and treatment response. Crit Care Med. 2018;46(3):e242–9.

    Article  PubMed  PubMed Central  Google Scholar 

  42. Sweeney TE, Liesenfeld O, Wacker J, He YD, Rawling D, Remmel M, Coyle S, Midic U, Kotsaki A, Kanavou A, et al. Validation of inflammopathic, adaptive, and coagulopathic sepsis endotypes in coronavirus disease 2019. Crit Care Med. 2021;49(2):e170–8.

    Article  CAS  PubMed  Google Scholar 

Download references


The authors wish to acknowledge the staff of the Sepsis and Critical Illness Research Center and the UF Health Medical Laboratory at Rocky Point whose professional conduct of the two studies was essential to the success of the program. In addition, the authors wish to thank the patients and their families for their willingness to participate in this institutionally approved research with the hope that these findings could lead to improved outcomes in sepsis. We thank Purvesh Khatri, Ph.D. for his external review of the manuscript prior to publication. We appreciate expert statistical and scientific support from Drs. James Wacker and Florian Uhle.


Supported in part by Grants RM1 GM139690-02, R35 GM140806-02 and P50 GM111152, awarded by the National Institute of General Medical Sciences (NIGMS), U.S.P.H.S. JB is a recipient of a T32 GM008721-24 training Grant in burns, trauma, and sepsis, awarded by the NIGMS. This work is also supported in part by a research subcontract 75A50119C00044 between Inflammatix Inc and the University of Florida, with funding awarded from the Department of Health and Human Services, Office of the Assistant Secretary for Preparedness and Response, BARDA, DRIVe.

Author information

Authors and Affiliations



OL, TS, SB, and LLM conceived and designed the overall study design. JAB, TJL, and LLM provided the original draft of the manuscript, and OL, TS, SB, FU, PAE, UIC edited and revised the final draft. PS developed and validated the ‘laboratory derived test’ used for IMX-SEV and endotyping analyses. UIC, JW, and JAB developed the statistical approaches used to analyze the data and conducted the statistical analyses. All of the authors have reviewed and approved the final draft. JAB and UIC share first author position for equivalent contribution to manuscript drafting and data analysis, with JAB appearing first given his time in drafting the manuscript.

Corresponding author

Correspondence to Lyle L. Moldawer.

Ethics declarations

Ethical approval and consent to participate

Ethics approval was obtained from the University of Florida Institutional Review Board (#201702261 and #201400611). Informed consent was obtained from each subject or their surrogate decision-maker. Self-reported or proxy-reported race and ethnicity category data were collected as per National Institutes of Health reporting guidelines and requirements.

Consent for publication

Not applicable.

Competing interests

Dr. Balch reported receiving Grants from the National Institute of General Medical Sciences during the conduct of the study. Drs. Chen, Liesenfield, and Sweeney reported being an employee and stock option holder of Inflammatix Inc during the conduct of the study, and outside the submitted work. Dr. Starostik reported funding from the U.S. Army Medical Research Acquisition. Drs. Loftus and Efron reported receiving Grants from the National Institute of General Medical Sciences during the conduct of the study. Dr. Brakenridge reported receiving joint academic-industry Department of Health and Human Services, Biomedical Advanced Research and Development Authority (BARDA) funding from Inflammatix Inc during the conduct of the study. Dr. Sweeney reported receiving Grants from BARDA Division of Research, Innovation, and Ventures (DRIVe); being an employee and shareholder during the conduct of the study and outside the submitted work; and being the licensed inventor on several patents pending and issued that cover the IMX-SEV metric. Dr. Moldawer reported a subcontract on BARDA Grant to Inflammatix Inc during the conduct of the study and receiving grants from the National Institute of General Medical Sciences. No other disclosures were reported.

Additional information

Publisher's Note

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

Supplementary Information

Additional file 1. 

Supplemental Materials: Methods.

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 The Creative Commons Public Domain Dedication waiver ( 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

Balch, J.A., Chen, UI., Liesenfeld, O. et al. Defining critical illness using immunological endotypes in patients with and without sepsis: a cohort study. Crit Care 27, 292 (2023).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: