Current approaches to classifying sepsis are limited because they rely too heavily on clinical features and host-only immune responses, while largely ignoring the role of the infecting microbes. Sepsis is a heterogeneous syndrome, and grouping patients based on clinical signs alone assumes a single underlying mechanism, which is often incorrect. This has hindered progress in understanding pathophysiology and in developing effective targeted therapies.
In a recent editorial, the authors propose that integrating microbial and host molecular data provides a more accurate and biologically meaningful approach to classifying sepsis.
Molecular endotyping has already improved risk stratification in sepsis. Approaches such as the quantitative sepsis response signature (SRSq) and hyper- and hypo-inflammatory phenotypes have demonstrated that host gene expression patterns can predict outcomes. However, these frameworks focus almost exclusively on the host response and do not account for differences in the type or burden of infecting organisms. This omission limits their ability to capture the critical interactions between host and pathogen that influence outcomes and may define treatable subgroups.
A study by Spottiswoode and colleagues extends host-only profiling by integrating metagenomics, transcriptomics, and proteomics in blood samples from 321 patients enrolled in the multicentre EARLI cohort. Patients were divided into detailed categories, including culture-positive sepsis (further separated into bloodstream versus peripheral site infection), culture-negative sepsis, possible sepsis, and critically ill patients without sepsis. This complex classification highlights the difficulty of using syndromic definitions to capture true biological differences in sepsis.
Host-based risk measures remained important. Higher APACHE-III scores, higher SRSq values, classification as SRS1, and the hyperinflammatory phenotype were all associated with increased mortality in patients with sepsis and in the combined group of sepsis plus possible sepsis.
The multi-omic approach revealed additional microbial and host signatures linked to mortality. Whole-blood metagenomic sequencing showed that greater total microbial mass in the bloodstream was associated with increased risk of death, and this was highest in patients with culture-positive sepsis. A second microbial metric, bacterial relative dominance, reflecting overgrowth of a limited number of species, was also linked to mortality. Detection of Staphylococcus aureus and human cytomegalovirus in blood by sequencing correlated with worse outcomes. These findings indicate that the burden and composition of microbes in the blood carry important prognostic information.
Host transcriptomic profiling showed a distinct immune pattern associated with outcome. Mortality was linked to increased neutrophil degranulation and antimicrobial defence programmes, as well as elevated levels of the chemoattractant IL-8. In contrast, survival was associated with stronger T-cell activation and signalling. Together, these findings suggest a seesaw between neutrophil-driven inflammation and adaptive immune function, with excessive innate immune activation and impaired T-cell responses indicating a poor prognosis in critical illness.
Many of the same microbial and host response signatures that predicted death in sepsis were also associated with mortality in patients who were classified as not having sepsis. The authors suggest two explanations. One is microbial translocation from the gut, which could drive inflammatory responses even in the absence of clinically recognised infection. Another is misclassification, as sepsis remains difficult to diagnose at the bedside; around 80% of the non-sepsis control group still received antibiotics, indicating clinical suspicion of infection. These observations imply that host–microbe interactions may influence outcomes across critical illness, not only in formally diagnosed sepsis.
The integrated host–microbe model developed by Spottiswoode and colleagues achieved an area under the curve of 0.79, outperforming APACHE-III and matching models based on host signatures alone. Although external transcriptomic comparisons with the COMET and GAinS cohorts showed encouraging similarities, much more validation is needed before this approach could be adopted as a clinical test.
The value of multi-omic profiling lies in identifying small subsets of patients with distinct, potentially treatable traits, such as dominant neutrophil-mediated injury or impaired T-cell or interferon responses. For this purpose, exploratory and hypothesis-generating approaches are appropriate, with subsequent validation through mechanistic and interventional studies rather than through broad, one-size-fits-all treatments.
In conclusion, the authors argue that integrating host transcriptomics, proteomics and microbial metagenomics from blood provides a powerful way to capture the complex biological interactions that drive outcomes in sepsis and critical illness. This strategy shows that both microbial burden and host immune state contribute to mortality, and that these interactions extend beyond classical sepsis to critically ill patients more broadly. Although further validation is required, this multi-omic framework represents a promising step towards precision medicine in sepsis, with the potential to improve prognostication and to guide future targeted therapies.
Source: AJRCCM
Image Credit: iStock
