Streptococcus pyogenes (Group A Streptococcus, GAS) can cause severe invasive infections (iGAS) with substantial early and delayed mortality. Host genetic and immune responses are thought to influence disease severity. Peripheral blood transcriptome profiling can characterize host responses in sepsis, and methods such as weighted gene correlation network analysis (WGCNA) help group co-regulated genes into functional modules that can be related to clinical traits and candidate biomarkers.
Most sepsis deaths occur within the first days to weeks, but a meaningful proportion of deaths occur after the first week. Prior work suggests distinct pathophysiological mechanisms may underlie early (<7 days) and delayed (8–90 days) sepsis mortality, and that transcriptome signatures of acute sepsis versus chronic critical illness differ. This study aimed to compare gene expression profiles in adult iGAS cases sampled early and again in the second week of illness, testing whether expression patterns differ between early and delayed deaths and whether profiles might inform prediction of delayed mortality.
The investigators conducted a prospective observational study at Tampere University Hospital and Turku University Hospital, Finland, enrolling adult patients with culture-confirmed iGAS from normally sterile sites between 30 June 2018 and 2 June 2020. Clinical data were obtained from interviews and electronic records with informed consent, and records were pseudonymized.
Whole blood for RNA isolation (PAXgene Blood RNA tube) was collected at two predefined timepoints: timepoint A (on average two days after admission, within two days of recruitment) and timepoint B (five to seven days after timepoint A). The primary clinical comparators used in the analysis were need for intensive care, death, and a composite severe disease outcome.
Forty-five patients with invasive GAS infection were recruited. Eight patients died within 90 days, of whom four died within the first week. After RNA quality control, 34 samples at timepoint A and 31 samples at timepoint B remained for transcriptome analysis. Demographic and clinical confounders considered included age, sex, and Charlson comorbidity index; these factors were compared across outcome groups.
RNA sequencing and preprocessing followed methods described previously by the authors. Expression levels were normalized by spike-in controls. Endogenous protein-coding genes were selected based on variability relative to technical variation estimated from spike-ins (selection threshold p < 0.05 with Benjamini-Hochberg adjustment). Genes with significantly altered expression in at least one timepoint were entered into WGCNA to define modules of co-expressed genes. The WGCNA soft threshold used was 16 and p-values reported for module–trait associations were unadjusted. Gene ontology enrichment for modules was performed with Enrichr and adjusted using Benjamini-Hochberg procedure.
HLA alleles were inferred from RNA-seq using arcasHLA with the IMGT/HLA database version 3.34.0. Associations between observed HLA alleles and clinical traits were evaluated using the Boruta feature-selection algorithm. Statistical comparisons between patients who died after timepoint B and those who survived employed unpaired, nonparametric Mann–Whitney U tests.
Of the 45 enrolled patients, eight died within 90 days; four deaths occurred in the first week. RNA quality control reduced analyzable sample counts to 34 at timepoint A and 31 at timepoint B. The gene expression patterns associated with severe disease and mortality differed substantially between the early (timepoint A) and later (timepoint B) samplings.
Using WGCNA, the investigators identified modules of co-expressed genes whose associations with clinical traits varied by timepoint. Notably, at timepoint B (the later sample, approximately one week after admission), higher expression of genes related to necroptosis and lower expression of HLA genes were associated with death. In contrast, the expression profile linked to severe disease at timepoint A did not mirror the timepoint B signature, indicating temporal evolution in host responses during iGAS.
The authors emphasize that the observed association at timepoint B—necrotic cell-death pathway upregulation combined with reduced antigen-presentation related HLA expression—suggests a different pathophysiology for deaths occurring in the later phase compared with early deaths. However, exact module details, effect sizes, and statistical significance values beyond the described associations are reported in the full article.
The divergent transcriptome signatures between early and later samples support the concept that separate biological processes may contribute to early and delayed mortality in iGAS. The late association of increased necroptosis-related gene expression and decreased HLA expression with death is consistent with mechanisms that could reflect ongoing immune dysregulation, defective antigen presentation, or cell-death–mediated inflammation during the second week of illness.
If validated in larger cohorts, temporal transcriptome profiling could help distinguish patients at risk of delayed deterioration and guide targeted research into mechanistic pathways or therapeutic strategies. The authors caution, however, that their findings are exploratory and not predictive given the small sample size.
Key limitations acknowledged include the small sample size and reduced numbers after RNA quality control, which limit statistical power and generalizability. The study protocol used unadjusted p-values for some WGCNA associations; details and full results are available in the main article. The datasets are not publicly available because of Finnish and EU data-protection regulations, but de-identified data can be requested from the University of Turku subject to approvals and controlled-access arrangements.
In this prospective study of adult iGAS cases, peripheral blood transcriptome profiles associated with severe disease and death changed between the early admission period and approximately one week later. At the later timepoint, higher expression of necroptosis-related genes and lower HLA gene expression were associated with death, suggesting potentially distinct mechanisms for delayed mortality. Due to the small sample size, these observations are preliminary and should not be used as predictive markers without further validation.