Multi-tissue transcriptional profiling resolves opposing acute and chronic host-response programs along the bone marrow–to–periphery axis in bacterial pneumonia

Background Whether the host response to severe bacterial pneumonia is shared across pathogens or tailored to the infecting organism is unresolved, yet it bears directly on host-directed therapy and on risk stratification in the critically ill. Methods Matched murine models of acute and chronic intranasal infection with Staphylococcus aureus (Gram-positive) and Klebsiella pneumoniae (Gram-negative) were established, and bulk ribonucleic acid (RNA) sequencing was performed on bone marrow, blood, spleen and lymph node. Genes differentially expressed between the marrow and the peripheral compartments were intersected across pathogens to define core mobilization signatures. These were scored by single-sample gene set enrichment analysis (ssGSEA) in four independent animal datasets, five human infection cohorts and three murine datasets containing infected lung tissue, and were refined into compact panels by four-algorithm machine learning and weighted gene co-expression network analysis. Results The host response resolved into two distinct transcriptional programs organized along the bone marrow to periphery axis. Acute infection engaged a conserved 32-gene mobilization core, anchored by a densely interconnected 10-gene hub and enriched for tumour necrosis factor (TNF), interleukin-17 (IL-17), leukocyte transendothelial migration and efferocytosis signalling, that was shared between Gram-positive and Gram-negative pneumonia and preferentially active in peripheral blood. This acute signature marked an effector-mobilization cellular ecology. It was elevated across bacterial, viral and fungal disease, was conserved from mouse to human, and in five human cohorts increased with sepsis severity, scaled with the Sequential Organ Failure Assessment (SOFA) organ-failure score, and independently stratified 28-day mortality (hazard ratio [HR] 2.33). In published murine datasets it was also induced in infected lung by both pathogen classes and rose with progression to sepsis. Chronic infection instead engaged a smaller, pathogen-restricted 14-gene program with tissue- and time-dependent dynamics that defined a regulatory, exhausted and myeloid-inflammatory state and carried no prognostic information. Orthogonal machine-learning and co-expression analyses nominated compact 10-gene acute and four-gene chronic candidate panels. Conclusion The transcriptional response to bacterial pneumonia is a coordinated bone marrow to periphery process rather than a set of isolated tissue reactions. The conserved acute axis is a severity- and outcome-associated, pathogen-agnostic signature and a rational substrate for broadly applicable host-response biomarkers and host-directed therapy in severe infection.