Identification of functional bacterial-viral pneumotypes associated with airway inflammation and all-cause mortality in critically ill patients

Identification of functional bacterial-viral pneumotypes associated with airway inflammation and all-cause mortality in critically ill patients | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Identification of functional bacterial-viral pneumotypes associated with airway inflammation and all-cause mortality in critically ill patients Hussein Anani, Grégory Destras, Quentin Semanas, Florian Martin, and 12 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9114464/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Alterations in lung virome and bacteriome compositions have been associated with hospital-acquired pneumonia in critically ill patients, but data on lung microbiome functions underlying progression to severe complications remain unclear. Combining viral-metagenomics with metatranscriptomics analyses of 184 endotracheal aspirates (ETAs) collected in 94 intubated critically ill patients, we identified a high-risk functional pneumotype associated with lung inflammation and mortality. This unfavourable pneumotype was characterised by enrichment in virulent Streptococcus- bacteriophages and temperate Klebsiella- bacteriophages, decreased transcriptional activity of commensal taxa such as Streptococcus and Alloprevotella , and increased transcriptional activity of Klebsiella pneumoniae . Machine-learning models defined two-viral- and four-bacterial-factor signatures predicting unfavourable pneumotype (AUC 0.9 and 0.8, respectively). Validation in an independent cohort of 117 ICU patients (239 ETAs) demonstrated the robustness of the association of these signatures with the risk of mortality. Causal inference identified virulent bacteriophages associated with Streptococcus and Alloprevotella and host genes including CCL2 , PLGRKT and PTX3 , as key regulators of severe lung dysbiosis linked to mortality in critically ill patients. Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Critically ill patients are particularly vulnerable to hospital-acquired pneumonia (HAP), which is one of the most common causes of nosocomial infections in the intensive care units (ICU) 1 . Despite advances in guideline-based management and antibiotic optimization, HAP remains associated with a high risk of mortality and treatment failure, with approximately 30% of cases progressing to acute respiratory distress syndrome (ARDS) and mortality rates reaching up to 40% 2,2–5 . Recent data suggest that, in addition to the presence of a pathogen, the disruption of healthy host-respiratory microbiome interactions plays a central role in HAP severity and inter-individual heterogeneity in the response to treatment 6,7 . Several studies have identified an association between lung microbiome dysbiosis and respiratory complications. Notably, the reduction in the healthy lung microbiome core (i.e., the microbial taxa typically found in healthy lung samples) was associated with HAP and acute respiratory failure 8 . This “healthy” bacterial core likely contributes to lung homeostasis 8,9 . Using viral metagenomics, we have recently demonstrated significant shifts in the lung viral community composition of intubated critically ill patients, suggesting a contribution of the lung virome to HAP pathogenesis and treatment outcomes 10 . Historically, pneumotypes have been defined as conserved combinations of microorganisms based on bacterial genus-level profiles from 16S rRNA data and a pneumotype enriched in supraglottic-associated taxa was linked to heightened subclinical lung inflammation in asymptomatic individuals 11 . More recently, a pneumotype enriched in oral-associated taxa at the time of diagnosis was associated with successful pneumonia therapy in patients with HAP and CAP 12 . In the present study, we extend this framework by defining functional pneumotypes, integrating the active bacteriome and the virome to capture the microbiome heterogeneity among critically ill patients before antibiotic exposure and HAP diagnosis. Two functional pneumotypes were identified as distinct meta-clusters through integrated analysis of viral metagenomics and metatranscriptomic data from endotracheal aspirates (ETAs) and associated with clinical outcomes. Results Patients characteristics and multi-omic sequencing approach We analysed ETAs from ICU patients enrolled in the multicentre, placebo-controlled, randomised clinical trial (PREV-HAP study) testing interferon-gamma-1b for the prevention of HAP (inclusion across 11 European centres, April-October 2021). Adults (18-80 years), under invasive mechanical ventilation, with one or more acute organ failure were included within the first 48 hours of ICU hospitalisation, randomised to interferon-γ-1b or placebo (100 microg every 48 hours for 9 days), and followed clinically for 3 months. ETAs were collected at day 0 (immediately before the first treatment injection) and at days 3 and 7. Based on sample availability, we analysed 184 ETAs from 94 randomised patients enrolled. The integrated dataset comprises viral taxa and functions from metatranscriptomics, as well as bacterial taxa, functions, and predicted metabolic pathways. It also includes viral composition and functions from viral metagenomics ( Fig. 1 , Supplementary Tables S1-3 and Extended data Fig. 1-3 ). Clinical characteristics did not differ between patients with and without HAP. The causes of ICU admission were 48% trauma, 37% surgical, and 15% medical. HAP developed in 45 patients (48%) at a median of 4 days after inclusion (interquartile range [IQR] 3–6). Out of the 94 patients, 22 (23%) died by day 90. During ICU stay, 26% received corticosteroids and 49% interferon-γ-1b ( Supplementary Table S4 ). These treatments were found to have no effect on lung microbiome functions ( Extended data Fig. 4-5). Prior to HAP cluster derivation using metatranscriptomic and viral metagenomic data To investigate the heterogeneity in the lung microbiome without antimicrobial treatment exposure, we selected 64 ETA samples collected from 39 patients before HAP diagnosis (defined as Day 0, with earlier samples indexed relative to this reference point; Supplementary Fig. S1). To identify distinct lung meta-clusters (so-called functional pneumotypes), we applied a three-step unsupervised analytical framework. First, we used MEFISTO (Method for the Functional Integration of Spatial and Temporal Omics Data) on viral metagenomics and metatranscriptomics to reduce the dimensionality of the selected features over time relative to HAP onset. Features found in at least 50% of the samples were retained. Second, unsupervised hierarchical clustering was subsequently applied to the first two resulting MEFISTO factors capturing the most variance to define discrete community structures. Model fit metrics indicated that a two-cluster solution (k = 2) best represented the data (Fig. 2a and Supplementary Fig. S2-S3). Third, because patients contributed multiple longitudinal samples before HAP onset, cluster assignment was determined by the cluster of the sample collected closest to HAP onset. Patients maintained consistent cluster assignments across their longitudinal samples, with minor cluster transition over time (Supplementary Fig. S4). This series of analyses demonstrated that two main functional pneumotypes of virome-microbiome status can be observed in ICU patients before HAP onset. To investigate the clinical significance of this result, we compared the mortality rates between the two pneumotypes. No major difference in baseline clinical characteristics was observed between the two pneumotypes (Supplementary Table S3). However, the hazard ratio of all-cause mortality and early successful extubation in pneumotype 1 was 0.2 and 2.33 (95% confidence interval (CI) [0.04, 0.58], log-rank test p = 0.00153, and 95% CI [0.94-5.8], p = 0.05, respectively; Fig. 2b and Supplementary Fig. S5). Association of pneumotype 2 with all-cause mortality remained statistically significant after multivariate analysis taking into account major baseline risk factors of death (RR 9.60, 95%CI 1.14-80.8, Fig. 2c). To investigate differences in the human immune response across pneumotypes, we performed a negative binomial generalized linear model (NB-GLM) identifying 85 differentially expressed (DE) human genes (false discovery rate [FDR] ≤ 0.2; Supplementary Table S5) ( Fig. 2d ). Pneumotype 2 exhibited elevated expression of pathways associated with immune and inflammatory responses, including canonical NF-κB (nuclear factor kappa-B) signaling, interleukin-6 (IL-6) production, and cellular responses to viral or other biotic stimuli. In contrast, pneumotype 1 was associated with negative regulation of cytokine production, lymphocyte differentiation, interleukin (IL)-1b and T cell response ( Fig. 2e ). These findings showed that the investigation of meta-clusters including of viral and bacterial composition and transcriptomic activity enables the identification of two functional pneumotypes associated with robust clinical outcomes in critically ill patients. Transcriptional bacterial signature and score in proinflammatory high-risk pneumotype To determine how microbiome composition drove differences between the pneumotypes at ICU hospitalisation, we compared transcriptionally active bacteria, gene expression, functional pathways and diversity between groups using metatranscriptomic data. Patients with the pneumotype 1 (hereafter referred to as the low-risk pneumotype) exhibited greater β-diversity of transcriptionally active taxa across patients (Wilcoxon test, p = 4.88 × 10⁻¹¹; Fig. 3a , Supplementary Table S6 and Extended data Fig. 6 ) than those with pneumotype 2 (high-risk pneumotype). The relative transcriptional activity of taxa constituting the healthy lung core microbiome, including Veillonella , Streptococcus , Prevotella , Haemophilus , and Fusobacterium 8 , was significantly reduced in the high-risk pneumotype 2 compared to the low-risk group (34% vs. 52%, respectively; Wilcoxon test, p = 0.03; Fig. 3b ). We next identified a panel of 23 differentially transcriptionally active bacterial species between low- and high-risk pneumotypes ( Fig. 3c and Supplementary Table S6 ). Bacteria with higher transcriptional activity in the high-risk pneumotype included Klebsiella pneumoniae, Ureaplasma urealyticum , Haemophilus influenzae , while Alloprevotella_unclassified , Prevotella pallens and Streptococcus constellatus and anginosus were more transcriptionally active in the low risk pneumotype. The overall expression of bacterial protein-coding genes was higher in the low-risk pneumotype (Wilcoxon test, p < 0.00001; Fig. 3d) , which showed a stronger COG (Clusters of Orthologous Groups) functional profile. Eleven differentially expressed bacterial functions were significantly enriched in the low-risk pneumotype, including cell motility, cell division and defense mechanisms (multivariate permutation test, p = 0.025, R² = 0.08; Fig. 3 e ). Similarly, a pathway-level analysis revealed increased transcription of metabolic pathways in the low-risk group, with 28 upregulated pathways versus only three in the high-risk group (Fisher’s exact test, p = 2.379e-06; Supplementary Fig. S 6 ). Finally, to create a simple and interpretable predictive tool for pneumotype classification, we applied a stacked machine learning approach that combined spectral clustering with a fast-and-frugal decision tree (FFT) model. The resulting four-factor decision tree robustly discriminated high-risk from low-risk pneumotypes (accuracy= 80%, sensitivity = 100%, specificity = 70%; Fig. 3 f ). The high-risk pneumotype was defined by lower relative transcriptional activity of Alloprevotella_unclassified (2.21%) and Klebsiella pneumoniae (>2.31%). The model’s predictive accuracy, assessed via tenfold cross-validation, achieved a mean area under the curve (AUC) of 0.8 ( Fig. 3 g ). To assess the clinical utility of such biomarkers in patients regardless of future HAP status, we used the entire cohort, selecting all samples within 0 to 4 days after admission, individuals classified as high risk pneumotype by XGBoost using the four bacterial features, had an increased mortality risk compared to those classified as low risk, as indicated by a hazard ratio of 3.9 (95% CI: 1.1–14, p = 0.025; Fig. 3 h and Extended data Fig. 7 ). Taken together, this series of analyses demonstrated that the pre-HAP high-risk pneumotype was characterized by a reduction in the transcriptional activity of the healthy core microbiome, with several pathobionts showing high transcriptional activity. Bacteriophage patterns and score in pre-HAP high-risk patients We next evaluated the virome composition and function across pneumotypes using viral metagenomic. Bacteriophages were found to be the most abundant viral entities in the two pneumotypes, followed by giant viruses ( Fig. 4a , Supplementary Fig. S7 and Supplementary Table S7 ). Patients classified as high-risk pneumotype exhibited significantly lower Bray-Curtis dissimilarity than those assigned to the low-risk pneumotype (Wilcoxon test, p = 4.73×10⁻¹¹; Fig. 4b , Supplementary Table S7 and Extended data Fig. 6 ). We subsequently identified 66 bacteriophage genera with significant differences in relative abundance between the low- and high-risk pneumotypes ( Fig. 4c and Supplementary Table S7 ). Functional profiling of the virome revealed distinct viral activity patterns across pneumotypes. Notably, the majority (69%) of the bacteriophages that were more abundant in the high-risk pneumotype were predicted to display a virulent lifestyle ( Supplementary Table S7 ). Using bacteriophage metagenomics data, we observed that ten viral functions assigned were significantly abundant in the high-risk pneumotype, the top ones being the transport of chemicals molecules/nutrients, and the viral adhesion and invasion (GLM with multivariate permutation test, p = 0.005, R² = 0.07; Fig. 4d ). In addition, we identified using bacteriophage metatranscriptomics data, that the regulation of gene expression and the host-related functions (stress-associated response) were more transcriptionally expressed in high risk pneumotype (p = 0.001, R² = 0.14; Fig. 4e ). A two-feature decision model using FFT achieved robust discrimination between high- and low-risk pneumotypes (accuracy= 90%, sensitivity = 94%, specificity = 84%; Fig. 4f ). The high-risk signature was primarily driven by elevated relative abundances of Rhizobium bacteriophage, Cuauhnhuacivirus (>0.06%), whose lifestyle is unknown, and Lactobacillus bacteriophage, Lidleenavirus (>0.09%), which is predicted to be virulent. Model performance, evaluated through tenfold cross-validation, yielded a mean AUC of 0.9 ( Fig. 4g ). Using XGBoost with the two bacteriophage features, the entire patient cohort showed a consistent trend of higher mortality among predicted high-risk individuals during the first 0 to 4 days following admission, which was analyzed independently of HAP status (HR = 4.8, 95% CI: 1.7-14, p = 0.0012; Fig. 4 h and Extended data Fig. 7 ). These findings demonstrate that the high-risk pneumotype harbors a virome signature reflecting lung viral convergence, characterized by the expansion and increased activity of virulent (lytic) bacteriophages. Such enrichment suggests an active regulatory role of bacteriophages within the disturbed microbial ecosystem. Validation of the functional pneumotype scores in an independent cohort Finally, we aimed to validate the accuracy of these scores in an independently recruited cohort of patients admitted in the Nantes University Hospital ICU for severe brain injury and who required invasive mechanical ventilation (IBIS cohort). We analyzed 239 ETAs samples collected on days 0, 3 and/or 7 after ICU admission (n=117 patients). The HAP rate was 51%; with a median time of four days (IQR= 3-6 days) after being admitted to the ICU. The characteristics of the study population are described in Supplementary Table S8 . The PREV-HAP and IBIS cohorts enrolled critically-ill patients in different years and different clinical centers, with different inclusion and exclusion criteria but similar sample processing and sequencing. In the entire patient cohort during the first 0 to 3 days following ICU inclusion, patients with high-risk pneumotype classification based on two viral, four bacterial, or six combined FFT features had higher mortality rates than those classified as low-risk. The corresponding HRs were 2 (95% CI 0.96-4), 3.25 (95% CI 1.1-9.5), and 2.4 (95% CI 1.1-4.9), respectively ( Fig. 5a-c and Extended data Fig. 7 ). We found that patients classified using the combined FFT signature overlapped more with the viral signature (Supplementary Fig S8-9) . Consistent with our previous findings, the relative transcriptional abundance of the healthy lung bacterial core metatranscriptome taxa was significantly lower in the pre-HAP samples with a high-risk pneumotype signature (using bacterial FFT) than in those with a low-risk pneumotype signature (22% vs. 38%, respectively; p = 0.02; Fig. 5d ). Additionally, bacteriophages dominated the predicted high-risk (log₁₀ RPKM median = 4.3, IQR = 3.9-4.5) and low risk pneumotype (predicted using viral FFT) virome composition (log₁₀ RPKM median = 3.8, IQR = 3.5-4.4; Fig. 5e ; Supplementary Table S8 ). Finally, we investigated whether host immune responses could further discriminate patients’ pneumotypes in the validation cohort using samples collected before HAP onset. A partial least squares discriminant analysis (PLS-DA) using 450 inflammatory, immune-related genes distinguished between high- and low-risk pneumotype patients ( Supplementary Fig. S10 ). The predicted high-risk group's inflammatory profile was confirmed (Fig. 5f and Supplementary Fig. S11) . Overall, these findings demonstrate that viral and/or bacterial pneumotype classification schemes identify patients at high risk of mortality, whose lung dysbiosis is characterised by depletion of actively transcribed commensal bacteria, enrichment of bacteriophages, and expression of pro-inflammatory cytokines. Causality inference and pathophysiological mechanisms of the high-risk pneumotype To further investigate the causal transkingdom interactions that may contribute to the high-risk pneumotype trajectory, we applied Transkingdom Network Analysis (TkNA) by integrating data from both PREV-HAP and IBIS cohorts to ensure statistical power. The resulting network comprised 62 nodes (including six bacterial species, thirteen bacteriophages, and forty-three human inflammatory genes) and 151 edges ( Fig. 6a ; Supplementary Table S9 and Extended data ). Twenty-seven nodes (three bacterial species, eight bacteriophages, and sixteen inflammatory genes) exhibiting a median importance score of 0.5 (IQR = 0.2-0.6) ( Fig. 6a , Supplementary Fig. S12 and Supplementary Table S9 ), supporting their potential for causal involvement in severe pre-HAP pneumotype development. Among these nodes, we observed negative correlations between several bacteriophages (i.e., Colossusvirus , Fletchervirus , Kisquinquevirus and Magiavirus ) ( Supplementary Fig. S13) and the bacterial genera Streptococcus_unclassified and Alloprevotella_ unclassified , which were correlated with each other but anticorrelated with Klebsiella pneumoniae . Moreover, Streptococcus_ unclassified and Alloprevotella_ unclassified showed strong negative interactions with a cluster of potent inflammatory genes encoding cytokines and receptors involved in immune activation including NOS2 , IFNG , IL2RA , CCL2 , IL9 , IL12B , TNFRSF4 , PTX3 , HRH1 , HRH4 , TIRAP , and CD40 ( Fig. 6a and Supplementary Table S9 ). Two potential bacteria-bacteriophage interactions were identified within the network. In the high-risk pneumotype, we observed a significant reduction in Streptococcus_ unclassified transcriptional activity (p = 0.0095; Fig. 6b ) and an increase in the abundance of virulent Streptococcus bacteriophage ( Brussowvirus ; Wilcoxon test, p = 0.00006; Fig. 6c ), leading to high Streptococcus_ unclassified virus-to-host ratio (VHR) ( Fig. 6d ). In contrast, there was an enrichment of Klebsiella pneumoniae (p = 0.0012; Fig. 6e ) and its associated temperate bacteriophage ( Eowynvirus ; p = 0.0001; Fig. 6f ) with no differences in Klebsiella VHR among the groups (p = 0.18; Fig. 6g ). These results suggest that different relationships between bacteriophage-bacteria leading to high-risk pneumotype may apply, with a potential predator-prey relationship between Streptococcus-Brussowvirus and piggyback-the-winner between Klebsiella-Eowynvirus . Furthermore, GLM modeling of gene expression across groups revealed increased expression of CCL2 (p=0.00019; Fig. 6h ), PLGRKT (p=0.00029; Fig. 6i ) and PTX3 (p=0.01; Fig. 6j ) in high-risk patients. Overall, these results support the hypothesis that bacteriophage-driven predation contributes to bacterial dysbiosis, promoting heightened inflammatory responses in patients who progress toward severe, high-risk HAP. Discussion In this study, we used an unsupervised and multi-omic approach to analyse endotracheal aspirates of critically ill patients. We integrated metatranscriptomic and viral metagenomic profiles in order to investigate lung microbiome heterogeneity. This strategy revealed two pre-HAP functional pneumotypes associated with all-cause mortality before antibiotic treatment. The high-risk pneumotype showed increased mortality, a strong proinflammatory signature, and loss of healthy lung bacterial transcriptional activity. It also displayed elevated transcriptional activity of pathobionts such as K. pneumoniae . Additionally, this pneumotype showed a virome convergence driven by increased absolute abundance of bacteriophages. Finally, we defined two predictive lung signature scores derived from active bacteriome and virome features. These scores were validated in an external prospective cohort, predicting both pneumotypes across all ICU patients. Together, these distinct functional pneumotypes identified within the first days of ICU admission support plausible pathophysiological mechanisms underlying their divergent clinical outcomes. The observation of two pre-HAP patient groups was notable, as it mirrors previous reports of two subphenotypes among 3,889 critically ill patients at HAP diagnosis. These subphenotypes, which are linked to mortality and inflammation, were defined by clinical characteristics and routine biological tests 6 . In the present study, RNA expression profiling revealed that high-risk pneumotype showed lymphocyte and leukocyte proliferation, IL-6 production, and NF-κB activation, sustaining a proinflammatory profile related to poor outcomes 13–16 . These suggest that the lung virome-bacteriome profile of ICU patients reflects the evolution of their clinical status in real time. As these pneumotypes precede HAP onset, early lung inflammation may predispose patients for clinical deterioration and high-mortality HAP. This underscores the role of lung viral-microbial communities in both disease susceptibility and progression 8,10 , highlighting the potential of longitudinal bedside microbiome profiling to guide personalized interventions. High-risk patients exhibited viral convergence and enrichment with virulent bacteriophages as recently documented 10 . In this latest work, we reported significant viral convergence in ETA samples from patients who later developed HAP compared with those who did not 10 . In the present study, we observed an even stronger viral convergence in the high-risk pneumotype viromes than in the low-risk within the upcoming HAP group. Additionally, we identified an enrichment of virulent Caudoviricetes bacteriophages, which are anticorrelated with bacteria belonging to the core healthy respiratory group. Members of this tailed bacteriophage class are described among the key contributors to HAP onset in the lung viromes of critically ill patients 10 . We next derived a bacteriophage-based predictive score using viral metagenomic data. The decision tree incorporated two bacteriophages enriched in the high-risk pneumotype. Cuauhnahuacvirus , a Rhizobium bacteriophage previously associated with prolonged mechanical ventilation in ICU patients 8 , and Lidleunavirus , a virulent bacteriophage increased in the ileal mucosal virome of IBD patients 17 of Lactobacillus that was reported in ICU patients with early successful extubation 8 . The link between virome and lung disease has been described previously. Sputum metagenomes from 99 COPD patients and 36 controls revealed disrupted virus-bacteria ecological dynamics and progressive loss of bacteriophage diversity, particularly Porphyromonas -associated bacteriophages, in patients with frequent exacerbations 18 . Similarly, metatranscriptomic analysis of 278 bronchoalveolar lavage samples from 229 paediatric HCT patients identified viral enrichment as a key driver of fatal lung injury and high-risk in-hospital outcomes 19 . In our previously described pathophysiological model, we hypothesised that Prevotella and Streptococcus were targeted by virulent bacteriophages via predator-prey interactions, resulting in lung microbial dysbiosis and promoting HAP onset 10 . Consistent with this model, we observe a significant increase in the Streptococcus virus-to-host ratio suggesting that virulent Streptococcus bacteriophage ( Brussowvirus ) infected and depleted their bacterial hosts via a classic kill-the-winner (predator–prey) dynamic. This recurrent pattern further supports our hypothesis that bacteriophage-bacteria interactions contribute to dysbiosis associated with worse clinical outcomes. However, this study identifies the presence of K. pneumoniae exhibiting an opposing pattern: when bacterial transcription activity was high, its temperate bacteriophage ( Eowynvirus ) multiplied as a result of bacterial division and adopted the lysogenic lifestyle following the Piggyback-the-Winner hypothesis 20 . This mechanistic scenario provides a plausible explanation for the observed lung microbiome destabilization that has led to enhanced inflammatory responses. In line with these patterns, both DNA and RNA profiles showed elevated bacteriophage activity in the high-risk pneumotype. These results suggest that high-risk lung virome is actively undergoing functional remodelling, which is associated with shifts in bacteriophage abundance and potentially with changes in the genetic composition of individual bacteriophages. Using metatranscriptomics, we defined the transcriptionally active bacteriome in lung fluids and identified a bacterial dysbiosis in the high-risk pneumotype, reflected by reduced activity of core healthy pulmonary taxa. This mirrors the dysbiosis previously identified using 16S profiles in the severe HAP subphenotype 6 and highlights the added resolution of metatranscriptomics in distinguishing active pathogens and identifying microbial drivers of infection 19,21–23 . Besides, metatranscriptomic data has contributed to the identification of the functional lung pneumotypes in critically ill patients before antibiotic exposure. Recent work has demonstrated that, at pneumonia diagnosis and under antibiotic treatment, the lung microbiota segregates into four distinct 16S rRNA gene-based pneumotypes reflecting the disruption of the microbial landscape during pneumonia 12 . Notably, a pneumotype enriched in oral-associated taxa, including Streptococcus , was associated with successful treatment of HAP and with upregulation of IL-1 signalling pathways in alveolar macrophages, partially mirroring our low-risk functional pneumotype. We further derived a four-factor bacterial transcriptomic signature, in which high-risk status was defined by low Alloprevotella expression, a commensal bacteria in ventilated ICU patients 24 also linked to treatment-responsive lung masses 25 and improved COPD outcomes 26 , and by elevated Paracoccus and K. pneumoniae . Paracoccus is enriched in ARDS-associated microbiomes 8 , while K. pneumoniae is a common VAP pathogen 27 associated with higher mortality, disease severity, and respiratory failure in ICU cohorts 28 . Viral- and bacterial-derived scores effectively identified pre-HAP pneumotypes in an external ICU cohort, even within the first days of admission, underscoring their utility for early risk stratification. These results support the concept that stabilizing the lung microbiome through prudent antimicrobial use, probiotic approaches, or targeted bacteriophage-based strategies may help prevent the dysbiosis preceding severe HAP 29 . This study has limitations: The discovery cohort was small, with only 39 pre-antibiotic ETA samples from patients who later developed HAP used for multi-omic pneumotype identification. The predictive model could not be fully validated externally, as pneumotypes are inferred rather than being clinically recorded. Causality cannot be confirmed without in vivo experimentation. Nevertheless, a key strength is the validation of our findings in a prospective cohort. In conclusion, our multi-omic analysis revealed substantial heterogeneity in the transcriptomic activity of the lung virome and bacteriome of ICU patients prior to HAP. The distinct functional pneumotypes associated with inflammation and increased mortality highlight the clinical importance of identifying airway community structure early on. Machine learning-based scores incorporating four bacterial and two bacteriophage markers accurately stratified patients upon admission and identified those at elevated risk. Furthermore, we detected a bacterial-viral dysbiosis signature accompanied by inflammatory gene expression several days prior to HAP onset, suggesting that early microbial disruption may contribute to disease development. Together, these findings refine our understanding of HAP pathophysiology and lay the groundwork for future diagnostic and microbiome-targeted interventions in critical illness. Methods Cohorts Description and ETAs Collection The discovery cohort included 94 patients enrolled in the PREV-HAP clinical trial. This was an investigator-initiated, multicenter, parallel-group, double-blind, randomized study conducted in 11 ICUs in France, Spain, and Greece. The study protocol was approved by the Ouest II Angers Ethics Committee in France in March 2021, and the trial adhered to the Declaration of Helsinki. The trial was registered on ClinicalTrials.gov (NCT04793568) in the same month. Written informed consent was obtained from each patient’s legal surrogate. In accordance with local regulations, patients could be included before surrogate consent was obtained if the next of kin could not be contacted within the allowed enrollment window. Follow-up consent was requested from the patient within 90 days of inclusion, when feasible. Eligible patients were aged 18–85 years, receiving invasive mechanical ventilation, and had at least one acute organ failure at the time of enrollment. Participants were randomized in a 1:1 ratio to receive either interferon gamma-1b (100 µg/0.5 mL vials, IMUKIN®, Clinigen®) or placebo (normal saline) in fixed blocks of six, stratified by hospitalization cause (sepsis vs. other) and country (France, Greece, Spain). Key exclusion criteria included pregnancy or breastfeeding, hypersensitivity to interferon gamma-1b, pre-existing immunosuppression (e.g., recent chemotherapy or radiotherapy, AIDS, leukopenia), severe hepatic insufficiency (Child-Pugh B or C), liver cytolysis (transaminases >5× normal), chronic renal failure (MDRD GFR <10 mL/min/1.73 m²), persistent coma post-resuscitated cardiac arrest, cervical spinal cord injury hospitalization, previous hospital-acquired pneumonia during the current admission, and sustained hyperlactatemia (>5 mmol/L). Patients received five subcutaneous injections of either interferon gamma-1b or matching placebo from day 1 to day 9 (one injection every 48 hours). Follow-up continued for 90 days. HAP and ARDS diagnoses were blindly reviewed by intensivists for compliance with international definitions. HAP was defined as pneumonia occurring ≥48 hours after hospital admission, requiring at least two clinical signs (fever >38°C, leukocytosis >12,000 cells/µL, leukopenia <4,000 cells/µL, or purulent pulmonary secretions), accompanied by a new or worsening infiltrate on chest imaging, and confirmed with semi-quantitative or quantitative respiratory cultures obtained prior to initiating new antimicrobial therapy. All respiratory infection diagnoses were reviewed by an independent adjudication committee following European HAP guidelines. ARDS was defined as severe hypoxemia (PaO₂/FiO₂ < 200 mmHg with PEEP ≥5 cm H₂O) and bilateral opacities on chest imaging within one week of worsening respiratory symptoms. Information on ancestry, race, ethnicity, and socioeconomic status was not available for either cohort due to legal restrictions in France. In the validation cohort IBIS (Immunology and Brain Injury Study), 117 patients were enrolled in two French Surgical Intensive Care Units of one university hospital (Nantes, France). The collection of human samples has been declared to the French Ministry of Health (Programme de recherche “Immunologie”, DC-2017-2987), and was approved by the Comite de Protection des Personnes Ouest IV (7/04/2015 and 08/10/2020) (number clinicaltrials.gov NCT02003196). Written informed consent from a next-of-kin was required for enrolment. Retrospective consent was obtained from patients when possible. Appropriate consent was obtained for the release of information from deceased individuals. Participants received no compensation. Inclusion criteriawere male or female, 18-80 years old, brain injury (Glasgow Coma Scale below or equal to 12 and abnormal brain-CT scan) and receiving invasive mechanical ventilation. Exclusion criteria were cancer with radiotherapy or chemotherapy in the last 90 days, AIDS, leukopenia in the previous five years, immunosuppressive drugs, preexisting immunosuppression and pregnancy. All patients were clinically followed up for 90 days. Intensivists blindly reviewed HAP and ARDS diagnoses for compliance with international definitions 5,30,31 . In this study, a total of 423 endotracheal aspirates (ETAs) were collected from critically ill patients enrolled in two independent cohorts: PREV-HAP (184 ETAs from 94 patients) and IBIS (239 ETAs from 117 patients), based on sample availability. Within the PREV-HAP cohort, 122 ETAs were selected for metatranscriptomic analyses and 133 ETAs for viral metagenomic analyses. In the IBIS cohort, all 239 ETAs were included in both metatranscriptomic and viral metagenomic investigations. Metatranscriptomic sequencing for integrated microbiome and host analysis Metatranscriptomic analysis was conducted using the Revelo kit (Tecan) according to the protocol described by Destras et al. 2024. Briefly, 50 µL of ETA samples or no-template controls (NTCs) were spiked with an RNA internal control (MS2 bacteriophage, 2800 bp) at a concentration corresponding to a Ct value of 36. Nucleic acids were then extracted using the Maxwell platform (Promega) with the DNA Blood Kit. Sequencing was performed on a NovaSeq 6000 system using an SP flow cell for PREV-HAP cohort and S2 cartridge for IBIS, with 2 × 100 bp paired-end reads. Metagenomic sequencing for viral profiling ETAs were preserved diluted (dilution factor of 1:7.4) in a viral transport medium (LMR1925VTM, Labomoderne) and stored at -20°C. Samples from both cohorts were treated according to an adapted version of the protocol from Bal and colleagues and Anani et al. Briefly, the MS2 RNA bacteriophage (MS2, IC1 RNA internal control; bioMérieux R-GENE® ref. 71-110) was spiked in each sample to validate the whole metagenomic process. The samples were then filtered through 0.45 µm filters and incubated with DNase (Life Technologies ®, Carlsbad, USA) for 90 min. Total nucleic acid extraction was performed on the Qiagen EZ1 Advanced XL Extractor using the DSP Virus Kit, Qiagen®. After extraction, both RNA and DNA amplification were performed using a Whole Transcriptome Amplification kit (WTA2, Sigma-Aldrich®, Darmstadt, Germany). Amplified DNA was purified using QIAquick spin columns (Qiagen®, Hilden, Germany). No-template controls (NTC) and MS2 spiked-in NTC were prepared for each sample batch using nuclease-free water that followed the whole process starting from dilution in the transport media. The viral transport medium was also sequenced to confirm the absence of viral contamination. Before sequencing, quality control testing was performed with MS2 qPCR (R-GENE® ref. 71-110). The libraries were prepared using the post-PCR portion of the Illumina (San Diego, USA) COVIDSeq ® Test Kit for PREV-HAP samples and the Nextera XT Kit for IBIS samples. The libraries were then sequenced on a NovaSeq 6000 using SP (PREV-HAP) and S2 (IBIS) flow cells with 2×100 bp reads. Bioinformatic pipeline for virome analysis Raw sequencing reads were dehosted using SraHumanScrubber v2.2.1 to remove human-derived sequences and trimmed with fastp v0.23.4 to eliminate adapters, low-complexity regions, and low-quality bases, retaining only reads longer than 30 bp. Taxonomic classification was performed with Kraken2 v2.1.2 (parameters: minimum-hit-groups = 2, confidence = 0.1) using a custom database incorporating viral taxa from RVDB (v29.0) and Inphared (v2, January 2025) 32 , as well as archaeal, bacterial, fungal, and human references from RefSeq (downloaded on 06/01/2025) 33 . Non-viral reads assigned to Eukaryota, Bacteria, Archaea, Fungi, and their taxonomic descendants were filtered out using Krakentools, retaining only putative viral reads. These viral reads were assembled de novo with metaSPAdes 34 v3.15.5, and the resulting contigs were dereplicated using CD-HIT 35 v4.7 to generate operational taxonomic units (OTUs) at ≥95% nucleotide identity and ≥85% coverage, following MIUViG (Minimum Information about an Uncultivated Virus Genome) standards. OTUs shorter than 500 bp were discarded. The length-filtered OTUs were analyzed with PhaBOX2 36 v2.1.13 to identify viral OTUs (vOTUs), assign taxonomy according to the latest International Committee on Taxonomy of Viruses (ICTV) release, and predict host range and lifestyle for bacteriophages using CHERRY 36 and PhaTYP 36 , respectively. Finally, quality-filtered viral reads were realigned to the vOTU sequences using the Burrows–Wheeler Aligner (BWA) 37 v0.7.17, and count tables were generated with samtools. Viral decontamination was performed following the recently published MS2 internal control based method described by Anani and colleagues 10 . Bioinformatics analysis of metatranscriptomic data Metatranscriptomic data were analyzed using a pipeline that processed each kingdom independently. Prior to downstream analyses, all reads (FASTQ files) were trimmed for quality using fastp v0.23.4 (reference). First, human transcriptomic analyses were performed on non-dehosted, trimmed reads using RASflow 38 v2.0, HISAT2 v2.2.1, and the GRCh38 reference genome with default parameters. Duplicate reads were removed using Samtools, and FeatureCounts v2.0.3 was used to generate a gene count matrix. Only protein-coding genes detected in at least 10% of the samples were retained for further analysis. Second, for bacterial metatranscriptomic analyses, trimmed reads were first dehosted using SraHumanScrubber v2.2.1 with the GRCh38 human reference genome. Bacterial reads were then identified using Kraken2 32 v2.1.2 (minimum-hit-groups = 2, confidence = 0.1) against the previously described custom database. Reads assigned to the bacterial taxon (taxid = 2) were extracted using Krakentools. From these, 16S rRNA reads were removed using SortMeRNA 39 v4.3.6 with the SILVA database (v138.2), and the remaining reads were assembled de novo with rnaSPAdes 34 v3.15.5. Assembled bacterial transcripts were dereplicated using CD-HIT 35 v4.7 to generate operational taxonomic units (OTUs) with ≥95% sequence identity and ≥95% coverage. OTUs shorter than 1 kb were discarded. The remaining length-filtered OTUs were taxonomically assigned using DIAMOND 40 v2.1.6.160 with the lowest common ancestor (LCA) algorithm against the NCBI RefSeq Bacteria database (version 4). Quality control of bacterial OTUs (bOTUs) was performed to assess completeness and contamination. Functional annotation was carried out with Prokka 41 v1.14.6, and bOTUs lacking coding sequences were excluded. Quality-filtered bacterial reads used for de novo assembly were realigned to the curated bOTUs using the Burrows–Wheeler Aligner (BWA) 37 v0.7.17, and count tables were generated with samtools. Finally, potential bacterial contaminants were identified using the decontam 42 R package v1.20.0, applying the prevalence method with a threshold of 0.5 on read count tables normalized to RPKM. Third, for viral metatranscriptomic analyses, we performed preprocessing, quality assessment, and dehosting of raw paired-end reads using fastp v0.23.4 and SraHumanScrubber v2.2.1, applying the same parameters used in the viral metagenomic pipeline. All subsequent steps followed the viral metagenomic workflow parameters, except for the de novo assembly step. For this step, rnaSPAdes was used instead of metaSPAdes. Viral decontamination was carried out as described for viral metagenomic analyses. Functional microbiome and virome analysis For the bacterial component, high-quality reads obtained after ribosomal RNA filtration were analyzed using the HUMAnN 43 v3.9 pipeline from Biobakery to infer bacterial metabolic pathways. For functional annotation, all bacterial transcripts or bOTUs were annotated with Prokka 41 v1.14.6 (using the bacterial kingdom option), and the resulting protein sequences were searched against the latest NCBI database of Clusters of Orthologous Genes (COGs) 44 using BLASTp. For the viral component, derived from viral metagenomic or metatranscriptomic datasets, all vOTUs were annotated with Prodigal v2.6.3. The predicted viral proteins were then compared against latest UniProt SPROT curated viral databases, and Gene Ontology (GO) terms were retrieved from UniProtKB to classify viral proteins into functional categories following Cao and colleagues 17 . Statistical analysis: Unsupervised clustering analysis . Microbiome data characterized by virome metagenomes, virome and bacteriome transcriptomes and their corresponding identified proteins in addition to the bacterial metabolic pathway dataset were analyzed all using MEFISTO 45 (Method for the Functional Integration of Spatial and Temporal Omics data) to reduce dimensionality and identify a core set of factors over time by launching the python/R coupled library MOFA2 46 v1.18.0. This method effectively handles different data structures and distributions while being robust to collinearity. The data were filtered to retain features present in more than 50% of the samples and subsequently subjected to a log₁₀(x + 1) transformation. MEFISTO identified 15 latent factors. The factors 1 and 2 were then subjected to unsupervised hierarchical clustering, and two clusters were determined to be optimal based on Silhouette, Elbow, and Gap statistics. Then, each patient was assigned principal component coordinates derived from the first two components. These coordinates were visualized in a two-dimensional space using the FactoMineR v2.12 R package. Clinical characteristics. We fitted a Cox proportional hazards model to assess the association between clinical variables and ICU mortality. This model included cluster assignment, demographic factors, and key clinical parameters. We verified model performance and proportionality assumptions and visualized relative risks (RRs) with 95% confidence intervals using a forest plot generated with the ggforest function from the survminer v0.5.1 R package. We conducted Kaplan–Meier survival analyses to evaluate in-hospital mortality and the probability of invasive mechanical ventilation across the identified clusters. To account for competing risks, death was considered a censoring event for ICU length of stay and mechanical ventilation duration. The survival v3.8.3 R package was used to generate survival curves and compare them using the log-rank test to assess differences in survival distributions between clusters. Host gene expression. Differentially expressed genes (DEGs) were identified using edgeR v4.6.3 within each cluster based on normalized counts. We visualized the DEGs using volcano plots to illustrate significance and fold-change distribution. Gene set enrichment analysis (GSEA) was then performed on upregulated DEGs of each cluster using the clusterProfiler v4.16.0 and org.Hs.eg.db v3.21.0 R packages. Microbiome comparisons and predictions. Alpha diversity, as measured by the Shannon index and richness, was calculated at the species level for bacteria and at the genus level for viruses. This calculation was performed using the vegan v2.7.1 package in R (v4.5.1). Diversity distributions were visualized using the ggplot2 v3.5.0 package, and longitudinal trends were modeled with loess regression and 95% confidence intervals. The ANCOM-BC 47 v2.10.1 was applied to identify features significantly associated with the identified clusters and to account for sampling bias. Signature outputs were visualized using the SIAMCAT 48 v2.12.0 package. To identify bacterial/viral functional categories associated with each cluster, a generalized linear model (GLM) with a binomial family was applied. Briefly, normalized features (RPKM values) were compared between clusters by fitting individual GLMs for each function, modeling cluster membership as the binary outcome. For each function, log-odds estimates and p-values were extracted, and features with p < 0.05 were considered significant. In addition, a multivariate PERMANOVA based on Bray–Curtis dissimilarities was performed using the adonis2 function of vegan v2.7.1 to assess the overall functional compositional differences between clusters. PLS-DA analyses were performed using the mixOmics R package v6.32.0. Comparisons between the virome and bacteriome compositions in both the discovery and validation cohorts were performed using the UpSetR v1.4.0 R package for visualization of shared and unique features. Machine learning analyses were conducted using the FFTrees 49 v2.1.0, XGBoost 50 v1.7.11.1 and SIAMCAT 48 v2.12.0 R packages to evaluate the predictive power of the selected signatures, and receiver operating characteristic (ROC) curves were used to assess model performance. Transkingdom Network Analysis (TkNA 51 v1.2.1) was applied to perform causal inference using log 10 (relative abundance data) from the viral/bacterial signatures and the log 10 (TMM-normalized counts) from the host genes signature, with the resulting correlation network visualized in Cytoscape 52 v3.10.2. To identify the key nodes in the network, we computed an importance score by integrating three topological features: node degree, betweenness-based bipartite centrality (BiBC), and the probability of randomly selecting each node. Each metric was normalized between 0 and 1, and the probability was inverted so that lower values indicated higher importance. The final importance score was obtained by averaging the normalized degree, BiBC, and probability values. For comparisons of numerical and categorical variables, the Wilcoxon test, two-tailed Mann–Whitney U test, and Fisher’s exact test were applied. Statistical significance was defined as p ≤ 0.05, and symbols were used as follows: ns, not significant (p > 0.05); *, p ≤ 0.05; **, p ≤ 0.01; ***, p ≤ 0.001; ****, p ≤ 0.0001. Declarations Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Data availability The metatranscriptomic and the viral metagenomic raw data have been deposited in the BioProject repository PRJNA1373167 and PRJNA1371704, respectively. Code availability All code used for the analyses in this study is available on GitHub at https://github.com/genepii/HAP2-PREVHAP-HAP_pneumotypes. Acknowledgements We thank all clinicians and staff involved in the collection of endotracheal aspirates from ICU patients. The study was funded by the European Union's Horizon 2020 research and innovation program under grant agreement number 847782 (HAP2 project) and MSD Avenir grant (Phenomenon project). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Author contributions H.A., C.B.B., A.R. and L.J. conceived the study. S.B., G.B., F.P.M., C.P., M. P., E.O. and Q.S. performed the clinical data curation, sample preparations and sequencing. H.A. performed bioinformatic analysis. H.A., G.D., F.P.M., V.G., C.B.B., A.R. and L.J. interpreted results and drafted the manuscript. H.A., G.D., F.P.M., C.P., C.B.B., A.R. and L.J revised the manuscript. All authors reviewed and approved the final version of the manuscript. Competing interests L.C., E.M. and A.R. have a patent on respiratory microbiome composition in ICU patients ( WO2025078559A2 ). The authors declare no competing interests. Correspondence and requests for materials should be addressed to Laurence Josset. References He, Q. et al. The epidemiology and clinical outcomes of ventilator-associated events among 20,769 mechanically ventilated patients at intensive care units: an observational study. Crit. Care Lond. Engl. 25 , 44 (2021). Luyt, C.-E. et al. Pulmonary infections complicating ARDS. Intensive Care Med. 46 , 2168–2183 (2020). Lagarde, C. et al. A multifaceted strategy to optimize pharmacokinetics of antimicrobial therapy in patients with hospital-acquired infections—a monocentre quality improvement project. J. Antimicrob. Chemother. 78 , 1378–1385 (2023). Poulain, C. et al. Clinical evaluation of the BioFire Respiratory Pathogen Panel for the guidance of empirical antimicrobial therapy in critically ill patients with hospital-acquired pneumonia: A multicenter, quality improvement project. Anaesth. Crit. Care Pain Med. 43 , 101353 (2024). Torres, A. et al. International ERS/ESICM/ESCMID/ALAT guidelines for the management of hospital-acquired pneumonia and ventilator-associated pneumonia: Guidelines for the management of hospital-acquired pneumonia (HAP)/ventilator-associated pneumonia (VAP) of the European Respiratory Society (ERS), European Society of Intensive Care Medicine (ESICM), European Society of Clinical Microbiology and Infectious Diseases (ESCMID) and Asociación Latinoamericana del Tórax (ALAT). Eur. Respir. J. 50 , 1700582 (2017). Martin, F. P. et al. Identification and validation of robust hospital-acquired pneumonia subphenotypes associated with all-cause mortality: a multi-cohort derivation and validation. Intensive Care Med. 51 , 692–707 (2025). Deng, J. et al. Macrophage MR1 antigen presentation promotes MAIT cell immunity and lung microbiota modulation. Science eadr6322 (2025) doi:10.1126/science.adr6322. Montassier, E. et al. Robust airway microbiome signatures in acute respiratory failure and hospital-acquired pneumonia. Nat. Med. 29 , 2793–2804 (2023). Li, R., Li, J. & Zhou, X. Lung microbiome: new insights into the pathogenesis of respiratory diseases. Signal Transduct. Target. Ther. 9 , 19 (2024). Anani, H. et al. Lung virome convergence precedes hospital-acquired pneumonia in intubated critically ill patients. Cell Rep. Med. 6 , 102289 (2025). Segal, L. N. et al. Enrichment of lung microbiome with supraglottic taxa is associated with increased pulmonary inflammation. Microbiome 1 , 19 (2013). Sumner, J. T. et al. Transitions in lung microbiota landscape associate with distinct patterns of pneumonia progression. Cell Host Microbe 33 , 2148-2166.e8 (2025). Ishihara, K. & Hirano, T. IL-6 in autoimmune disease and chronic inflammatory proliferative disease. Cytokine Growth Factor Rev. 13 , 357–368 (2002). Rubini, A. Interleukin-6 and lung inflammation: evidence for a causative role in inducing respiratory system resistance increments. Inflamm. Allergy Drug Targets 12 , 315–321 (2013). Stotts, C., Corrales-Medina, V. F. & Rayner, K. J. Pneumonia-Induced Inflammation, Resolution and Cardiovascular Disease: Causes, Consequences and Clinical Opportunities. Circ. Res. 132 , 751–774 (2023). Guo, Q. et al. NF-κB in biology and targeted therapy: new insights and translational implications. Signal Transduct. Target. Ther. 9 , 53 (2024). Cao, Z. et al. The gut ileal mucosal virome is disturbed in patients with Crohn’s disease and exacerbates intestinal inflammation in mice. Nat. Commun. 15 , 1638 (2024). Cook, R. A. et al. Bacteriophage diversity declines with COPD severity in the respiratory microbiome. Cell Rep. 44 , 116413 (2025). Zinter, M. S. et al. Pathobiological signatures of dysbiotic lung injury in pediatric patients undergoing stem cell transplantation. Nat. Med. 30 , 1982–1993 (2024). Ritz, N. L. et al. The gut virome is associated with stress-induced changes in behaviour and immune responses in mice. Nat. Microbiol. 9 , 359–376 (2024). Langelier, C. et al. Integrating host response and unbiased microbe detection for lower respiratory tract infection diagnosis in critically ill adults. Proc. Natl. Acad. Sci. 115 , (2018). Kalantar, K. L. et al. Integrated host-microbe plasma metagenomics for sepsis diagnosis in a prospective cohort of critically ill adults. Nat. Microbiol. 7 , 1805–1816 (2022). Zinter, M. S. et al. The pulmonary metatranscriptome prior to pediatric HCT identifies post-HCT lung injury. Blood 137 , 1679–1689 (2021). Kitsios, G. D. et al. Longitudinal multicompartment characterization of host-microbiota interactions in patients with acute respiratory failure. Nat. Commun. 15 , 4708 (2024). Yang, Y. et al. Characteristics of gut and lung microbiota in patients with lung masses and their relationship with clinical features. BMC Microbiol. 25 , 541 (2025). Opron, K. et al. Loss of Airway Phylogenetic Diversity Is Associated with Clinical and Pathobiological Markers of Disease Development in Chronic Obstructive Pulmonary Disease. Am. J. Respir. Crit. Care Med. 210 , 186–200 (2024). Papazian, L., Klompas, M. & Luyt, C.-E. Ventilator-associated pneumonia in adults: a narrative review. Intensive Care Med. 46 , 888–906 (2020). Grosjean, V. et al. Community-acquired Klebsiella pneumoniae pneumonia in ICU: a multicenter retrospective study. Ann. Intensive Care 14 , 69 (2024). Roquilly, A. et al. Pathophysiological role of respiratory dysbiosis in hospital-acquired pneumonia. Lancet Respir. Med. 7 , 710–720 (2019). ARDS Definition Task Force et al. Acute respiratory distress syndrome: the Berlin Definition. JAMA 307 , 2526–2533 (2012). Leone, M. et al. Hospital-acquired pneumonia in ICU. Anaesth. Crit. Care Pain Med. 37 , 83–98 (2018). Lu, J. et al. Metagenome analysis using the Kraken software suite. Nat. Protoc. 17 , 2815–2839 (2022). Wood, D. E., Lu, J. & Langmead, B. Improved metagenomic analysis with Kraken 2. Genome Biol. 20 , 257 (2019). Nurk, S., Meleshko, D., Korobeynikov, A. & Pevzner, P. A. metaSPAdes: a new versatile metagenomic assembler. Genome Res. 27 , 824–834 (2017). Fu, L., Niu, B., Zhu, Z., Wu, S. & Li, W. CD-HIT: accelerated for clustering the next-generation sequencing data. Bioinforma. Oxf. Engl. 28 , 3150–3152 (2012). Shang, J., Peng, C., Liao, H., Tang, X. & Sun, Y. PhaBOX: a web server for identifying and characterizing phage contigs in metagenomic data. Bioinforma. Adv. 3 , vbad101 (2023). Jung, Y. & Han, D. BWA-MEME: BWA-MEM emulated with a machine learning approach. Bioinforma. Oxf. Engl. 38 , 2404–2413 (2022). Zhang, X. & Jonassen, I. RASflow: an RNA-Seq analysis workflow with Snakemake. BMC Bioinformatics 21 , 110 (2020). Kopylova, E., Noé, L. & Touzet, H. SortMeRNA: fast and accurate filtering of ribosomal RNAs in metatranscriptomic data. Bioinformatics 28 , 3211–3217 (2012). Buchfink, B., Reuter, K. & Drost, H.-G. Sensitive protein alignments at tree-of-life scale using DIAMOND. Nat. Methods 18 , 366–368 (2021). Seemann, T. Prokka: rapid prokaryotic genome annotation. Bioinforma. Oxf. Engl. 30 , 2068–2069 (2014). Davis, N. M., Proctor, D. M., Holmes, S. P., Relman, D. A. & Callahan, B. J. Simple statistical identification and removal of contaminant sequences in marker-gene and metagenomics data. Microbiome 6 , 226 (2018). Beghini, F. et al. Integrating taxonomic, functional, and strain-level profiling of diverse microbial communities with bioBakery 3. eLife 10 , e65088 (2021). Galperin, M. Y. et al. COG database update 2024. Nucleic Acids Res. 53 , D356–D363 (2025). Velten, B. et al. Identifying temporal and spatial patterns of variation from multimodal data using MEFISTO. Nat. Methods 19 , 179–186 (2022). Argelaguet, R. et al. MOFA+: a statistical framework for comprehensive integration of multi-modal single-cell data. Genome Biol. 21 , 111 (2020). Lin, H. & Peddada, S. D. Analysis of compositions of microbiomes with bias correction. Nat. Commun. 11 , 3514 (2020). Wirbel, J. et al. Microbiome meta-analysis and cross-disease comparison enabled by the SIAMCAT machine learning toolbox. Genome Biol. 22 , 93 (2021). Phillips, N. D., Neth, H., K. Woike, J. & Gaissmaier, W. FFTrees: A toolbox to create, visualize, and evaluate fast-and-frugal decision trees. Judgm. Decis. Mak. 344–368 (2017). Chen, T. & Guestrin, C. XGBoost: A Scalable Tree Boosting System. https://doi.org/10.48550/ARXIV.1603.02754 (2016) doi:10.48550/ARXIV.1603.02754. Newman, N. K. et al. Transkingdom Network Analysis (TkNA): a systems framework for inferring causal factors underlying host-microbiota and other multi-omic interactions. Nat. Protoc. 19 , 1750–1778 (2024). Shannon, P. et al. Cytoscape: A Software Environment for Integrated Models of Biomolecular Interaction Networks. Genome Res. 13 , 2498–2504 (2003). Additional Declarations The authors declare no competing interests. Supplementary Files SupplementalfiguresAnanietal.pdf Supplemental figures ExtendeddataAnanietal.pdf Extended data Supplementaltablestitles.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9114464","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":605639979,"identity":"7bb3b177-1260-4314-a83e-a8f02b3263fa","order_by":0,"name":"Hussein Anani","email":"","orcid":"","institution":"Hospices Civils de Lyon","correspondingAuthor":false,"prefix":"","firstName":"Hussein","middleName":"","lastName":"Anani","suffix":""},{"id":605639980,"identity":"e24c5d8f-0b58-4313-8bde-deb09c4ce460","order_by":1,"name":"Grégory Destras","email":"","orcid":"","institution":"Hospices Civils de Lyon","correspondingAuthor":false,"prefix":"","firstName":"Grégory","middleName":"","lastName":"Destras","suffix":""},{"id":605639981,"identity":"ed1ab573-46f5-45a0-8bc6-cb589565217a","order_by":2,"name":"Quentin Semanas","email":"","orcid":"","institution":"Hospices Civils de Lyon","correspondingAuthor":false,"prefix":"","firstName":"Quentin","middleName":"","lastName":"Semanas","suffix":""},{"id":605639982,"identity":"e8349981-e01f-4d64-b724-a9a78c83c4ef","order_by":3,"name":"Florian Martin","email":"","orcid":"","institution":"Nantes Université","correspondingAuthor":false,"prefix":"","firstName":"Florian","middleName":"","lastName":"Martin","suffix":""},{"id":605639983,"identity":"6237b5a1-53d2-4456-83c8-2dae2dc35c05","order_by":4,"name":"Mélanie Petrier","email":"","orcid":"","institution":"Nantes Université","correspondingAuthor":false,"prefix":"","firstName":"Mélanie","middleName":"","lastName":"Petrier","suffix":""},{"id":605639984,"identity":"9ec2aeb9-959f-4aa3-87c6-b7e4d701c40b","order_by":5,"name":"Eva Oddoux","email":"","orcid":"","institution":"Hospices Civils de Lyon","correspondingAuthor":false,"prefix":"","firstName":"Eva","middleName":"","lastName":"Oddoux","suffix":""},{"id":605639985,"identity":"5c73edbf-e558-4497-9677-4a132b577d6c","order_by":6,"name":"Gwendolyne Burfin","email":"","orcid":"","institution":"Hospices Civils de Lyon","correspondingAuthor":false,"prefix":"","firstName":"Gwendolyne","middleName":"","lastName":"Burfin","suffix":""},{"id":605639986,"identity":"ca533df4-10dc-4f1c-a1a2-8f30dab6845d","order_by":7,"name":"Simon Bulteau","email":"","orcid":"","institution":"Nantes Université","correspondingAuthor":false,"prefix":"","firstName":"Simon","middleName":"","lastName":"Bulteau","suffix":""},{"id":605639987,"identity":"2697c2cd-98c9-43e9-802f-cb50626bc82a","order_by":8,"name":"Cécile Poulain","email":"","orcid":"","institution":"Nantes Université","correspondingAuthor":false,"prefix":"","firstName":"Cécile","middleName":"","lastName":"Poulain","suffix":""},{"id":605639988,"identity":"672cf7c3-e779-4110-a8a6-c288d00afb2b","order_by":9,"name":"Ana Motos","email":"","orcid":"","institution":"Nantes Université","correspondingAuthor":false,"prefix":"","firstName":"Ana","middleName":"","lastName":"Motos","suffix":""},{"id":605639989,"identity":"366df249-c3ce-4c50-8394-bdfbfc86c0ea","order_by":10,"name":"Victor Gourain","email":"","orcid":"","institution":"Nantes Université","correspondingAuthor":false,"prefix":"","firstName":"Victor","middleName":"","lastName":"Gourain","suffix":""},{"id":605639990,"identity":"a35d1808-4634-46a1-8b27-be5de3507bc8","order_by":11,"name":"Lise Cremet","email":"","orcid":"","institution":"Nantes Université","correspondingAuthor":false,"prefix":"","firstName":"Lise","middleName":"","lastName":"Cremet","suffix":""},{"id":605639991,"identity":"a431a579-45f7-4f2c-9cc0-d038340ba1a5","order_by":12,"name":"Emmanuel Montassier","email":"","orcid":"","institution":"Nantes Université","correspondingAuthor":false,"prefix":"","firstName":"Emmanuel","middleName":"","lastName":"Montassier","suffix":""},{"id":605639992,"identity":"eb1fc77a-b304-40df-9acf-4538add054ef","order_by":13,"name":"Céline Bressollette-Bodin","email":"","orcid":"","institution":"Nantes Université","correspondingAuthor":false,"prefix":"","firstName":"Céline","middleName":"","lastName":"Bressollette-Bodin","suffix":""},{"id":605639993,"identity":"c9fb53c2-7045-4aef-ae64-1628fde50c4a","order_by":14,"name":"Antoine Roquilly","email":"","orcid":"","institution":"Nantes Université","correspondingAuthor":false,"prefix":"","firstName":"Antoine","middleName":"","lastName":"Roquilly","suffix":""},{"id":605639994,"identity":"38e7cc1a-87d5-478c-82a1-bb8dc8ebc1a1","order_by":15,"name":"Laurence Josset","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABFklEQVRIiWNgGAWjYBACPgTzAJiUY5AAkg9ATOYDWLWwoWkxMAZrSQDLJRDSAgYGiQ0EtbCfMfvwcYedHAPj4cMff1T8Se+f3WPAkNhmk8fAxvsAqxaeHOOZM88kGzMwHEuT5jljkDvjzhmQlrRioHEG2B2WY8zM23YgsYHhjBkzY5tB7gaJHPMfiW2HExvk27A7jP+NMfPftgP1QC3GH3/+M0g3kMgB2QLUwobuUagWCaAtjG0HgD49YyDB22CQQISWZ8WMvW3Jhm1gvxwzNpxxI62AIeFcWmIbDi38/MmbGX622cnzS4BCrEZOnn9G8gaGD2U2if04tCBZdwCJx8iGGWVY7GtA5v0hqH4UjIJRMApGDgAAsV5XBMZpdboAAAAASUVORK5CYII=","orcid":"","institution":"Hospices Civils de Lyon","correspondingAuthor":true,"prefix":"","firstName":"Laurence","middleName":"","lastName":"Josset","suffix":""}],"badges":[],"createdAt":"2026-03-13 11:54:39","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-9114464/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9114464/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104699403,"identity":"5edf9319-b449-4565-be9b-4972d2219d02","added_by":"auto","created_at":"2026-03-16 08:14:28","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":2408599,"visible":true,"origin":"","legend":"\u003cp\u003eFlowchart representation of sample handling and analytical workflow using metatranscriptomics and viral metagenomics approaches.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-9114464/v1/239674a0445def879a1e943e.png"},{"id":104782538,"identity":"0359e906-f55c-4d79-859d-78058b32590a","added_by":"auto","created_at":"2026-03-17 07:57:29","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1845226,"visible":true,"origin":"","legend":"\u003cp\u003eIdentification of high-risk and inflammatory pre-HAP functional pneumotype prior to HAP onset. a, pre-HAP clusters identified using unsupervised clustering of MEFISTO factors. The x- and y-axes present factors 1 and 2. b, Kaplan-Meier curves for survival probability between pre-HAP pneumotypes (log-rank p\u0026lt;0.05). c, Forest plot showing the relative risk for 90-day mortality between clinical variables stratified by pneumotype (Global log-rank p = 0.009). d, Volcano plot showing the differentially gene expression in each pre-HAP pneumotype. e, Gene set enrichment analysis of the upregulated differentially expressed genes in pneumotypes. Only genes with adjusted p\u0026lt;0.05 were selected. Data from 64 endotracheal aspirate sample of the discovery cohort (PREV-HAP) were used in (a-e).\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-9114464/v1/bee16325f1e04e91a3da4f4b.png"},{"id":104699397,"identity":"d4f489d5-6df6-4746-881b-a780dccad6fe","added_by":"auto","created_at":"2026-03-16 08:14:27","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":3487300,"visible":true,"origin":"","legend":"\u003cp\u003eCharacterization of the transcriptionally active bacteria in functional pneumotypes. a, boxplots showing the median of beta diversity (Bray-Curtis dissimilarity) between pneumotypes. Data are presented through box plots, showcasing the 25th to 75th percentiles along with the median, while the whiskers extend to the minimum and maximum observed values. Statistical significance was calculated by the Wilcoxon test (**** for p\u0026lt; 0.0001). b, Relative abundance of the healthy lung core metatranscriptome core (\u003cem\u003eVeillonella\u003c/em\u003e, \u003cem\u003eStreptococcus\u003c/em\u003e, \u003cem\u003ePrevotella\u003c/em\u003e, \u003cem\u003eHaemophilus\u003c/em\u003eand \u003cem\u003eFusobacterium\u003c/em\u003e) in pneumotypes (Wilcoxon test, p=0.03). c, Top features of the differentially transcribed (DT) bacteria at species level in pneumotypes. Statistical significance for each species was assessed using a non-parametric Wilcoxon test. The strength and direction of association were further characterized using multiple effect size measures, including the false discovery rate (FDR)–adjusted \u003cem\u003ep\u003c/em\u003e-value, area under the receiver operating characteristic curve (AUC), and fold change. d, Heatmap showing the expression of bacterial proteins across functional pneumotypes. e, Generalized linear model (GLM) represents the bacterial differentially transcriptional functions in pneumotypes. A function-wise logistic regression model was employed to assess the association between function expression levels and pneumotype membership. For each function, a GLM with a binomial distribution and logit link function was fitted. The model estimated the regression coefficient (log-odds) and corresponding p-value for each function. Functions with p \u0026lt; 0.05 were considered statistically significant, and the direction of association was determined by the sign of the regression coefficient. In addition, a permutational multivariate analysis of variance (PERMANOVA) based on Bray-Curtis dissimilarities was performed (p=0.025 and R2=0.08). f, FFTree-based staging scheme using differentially transcribed bacteria signature to predict the high risk pneumotype. g, ROC of the high risk pneumotype classification applying a tenfold cross-validation scheme with ten repetitions. h, Kaplan-Meier curves for survival probability between predicted pneumotypes for all patients. Data from 42 and 73 endotracheal samples were analysed in a-g and h, respectively.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-9114464/v1/43a1c81532b54e3495478f8e.png"},{"id":104699400,"identity":"184e2dc5-568f-497b-9f18-35fd62851427","added_by":"auto","created_at":"2026-03-16 08:14:28","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":3095940,"visible":true,"origin":"","legend":"\u003cp\u003eCharacterization of the viral signatures in functional pneumotypes. a, Bubble plot showing the composition of the whole virome in functional pneumotypes. b, boxplots showing the median of beta diversity (Bray-Curtis dissimilarity) between pneumotypes. Data are presented through box plots, showcasing the 25th to 75th percentiles along with the median, while the whiskers extend to the minimum and maximum observed values. Statistical significance was calculated by the Wilcoxon test (**** for p\u0026lt; 0.0001). c, Top features of the differentially abundant (DA) bacteriophage at genus level in pneumotypes. Statistical significance for each genus was assessed using a non-parametric Wilcoxon test. The strength and direction of association were further characterized using multiple effect size measures, including the false discovery rate (FDR)–adjusted \u003cem\u003ep\u003c/em\u003e-value, area under the receiver operating characteristic curve (AUC), and fold change. d-e, Generalized linear model (GLM) represents the differentially abundant (d) and differentially expressed (e) functions in pneumotypes. A function-wise logistic regression model was employed to assess the association between function expression levels and pneumotype membership. For each function, a GLM with a binomial distribution and logit link function was fitted. The model estimated the regression coefficient (log-odds) and corresponding p-value for each function. Functions with p \u0026lt; 0.05 were considered statistically significant, and the direction of association was determined by the sign of the regression coefficient. *, ** and *** for p \u0026lt; 0.05, p \u0026lt; 0.01 and p \u0026lt; 0.001, respectively. Besides, Permanova based on Bray-Curtis dissimilarities was computed (p=0.005-R2=0.07 in d and p=0.001-R2=0.14 in e). f, FFTree-based staging scheme using differentially abundant bacteriophage signature to predict the high risk pneumotype. g, ROC of the high risk pneumotype classification applying a tenfold cross-validation scheme with ten repetitions. h, Kaplan-Meier curves for survival probability between predicted pneumotypes for all patients. Data from 64 and 98 endotracheal samples were analysed in a-g and h, respectively.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-9114464/v1/011caefd0a08f88794b7d7fb.png"},{"id":104699401,"identity":"9b9edc09-8d0a-42c5-9bbe-f36342abf20b","added_by":"auto","created_at":"2026-03-16 08:14:28","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1448559,"visible":true,"origin":"","legend":"\u003cp\u003eValidation in an external independent cohort (IBIS). a-c, Kaplan-Meier curves for survival probability between predicted pneumotypes using viral, bacterial and combined FFT features in all patients. d, Relative abundance of the core healthy lung microbiome in the predicted pneumotypes. e, Viral composition of the predicted pneumotypes. The size of the circle corresponds to the absolute abundance (Log\u003csub\u003e10\u003c/sub\u003eRPKM), with larger circles indicating higher abundance levels. f, Gene set enrichment analysis of the upregulated differentially expressed genes in the predicted high risk pneumotype. Data from 81 and 236 endotracheal samples were analysed in d-f and a-c, respectively.\u003c/p\u003e","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-9114464/v1/b2d16f828c6fcc7ba3955a28.png"},{"id":104699398,"identity":"04a36e8d-2ded-4598-9cb1-9c50097bcbb6","added_by":"auto","created_at":"2026-03-16 08:14:27","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":3534414,"visible":true,"origin":"","legend":"\u003cp\u003ePathophysiological mechanism of the inflammatory and high risk pre-HAP functional pneumotype. a, Transkingdom network illustrating significant correlations between differentially transcribed bacteria, bacteriophage abundance, and inflammatory immune gene expression before HAP onset. Edges are colored according to the TkNA correlation type: blue indicates negative correlations and red indicates positive correlations. Edge width reflects the node importance score, and edge transparency indicates the level of statistical significance (p-value). Rectangles represent bacteria, circles represent bacteriophages, and octagons represent immune genes. b-g, Boxplots plots showing the abundance (in log10 RPKM) of \u003cem\u003eStreptococcus_unclassified \u003c/em\u003e(b), \u003cem\u003eBrussowvirus \u003c/em\u003e(c), virus-to-host ratio in \u003cem\u003eStreptococcus_unclassified \u003c/em\u003e(d), \u003cem\u003eKlebsiella pneumoniae \u003c/em\u003e(e), \u003cem\u003eEowynvirus \u003c/em\u003e(f), virus-to-host ratio in \u003cem\u003eKlebsiella \u003c/em\u003e(g) between pneumotypes. h-j, Boxplots plots showing the expression (in log10 TMM-normalised) of \u003cem\u003eCCL2 \u003c/em\u003e(h), \u003cem\u003ePLGRKT \u003c/em\u003e(i) and \u003cem\u003ePTX3 \u003c/em\u003e(j) genes across pneumotypes. Symbols used: Statistical significance was calculated by the Wilcoxon test. ns for not significant (p \u0026gt; 0.05), * for p ≤ 0.05, ** for p ≤ 0.01, *** for p ≤ 0.001, and **** for p ≤ 0.0001. Data from 75 endotracheal samples were analysed in a-j.\u003c/p\u003e","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-9114464/v1/34afa9726b8081481ba9f8fe.png"},{"id":104785003,"identity":"809fb5d4-73c2-4526-9896-54772ce86129","added_by":"auto","created_at":"2026-03-17 08:09:12","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":17846094,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9114464/v1/24e7a7e8-745f-4aa4-96f0-6c228ae62e2c.pdf"},{"id":104699402,"identity":"a708eb57-1f5e-4052-bbe4-0f91b40227f1","added_by":"auto","created_at":"2026-03-16 08:14:28","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":1230229,"visible":true,"origin":"","legend":"\u003cp\u003eSupplemental figures\u003c/p\u003e","description":"","filename":"SupplementalfiguresAnanietal.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9114464/v1/2d2afe8ba31854e7febcd201.pdf"},{"id":104699405,"identity":"a6df76bc-8a01-45a0-8dcb-50abff71d651","added_by":"auto","created_at":"2026-03-16 08:14:28","extension":"pdf","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":2028823,"visible":true,"origin":"","legend":"\u003cp\u003eExtended data\u003c/p\u003e","description":"","filename":"ExtendeddataAnanietal.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9114464/v1/1520de4563c3ec2e004c8b17.pdf"},{"id":104699404,"identity":"2c5a576c-eae7-4623-985e-3c0ade18f91b","added_by":"auto","created_at":"2026-03-16 08:14:28","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":15079,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementaltablestitles.docx","url":"https://assets-eu.researchsquare.com/files/rs-9114464/v1/fbe4d61ceb2e54980ba6380f.docx"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eIdentification of functional bacterial-viral pneumotypes associated with airway inflammation and all-cause mortality in critically ill patients\u003c/p\u003e","fulltext":[{"header":"Introduction","content":"\u003cp\u003eCritically ill patients are particularly vulnerable to hospital-acquired pneumonia (HAP), which is one of the most common causes of nosocomial infections in the intensive care units (ICU)\u003csup\u003e1\u003c/sup\u003e. Despite advances in guideline-based management and antibiotic optimization, HAP remains associated with a high risk of mortality and treatment failure, with approximately 30% of cases progressing to acute respiratory distress syndrome (ARDS) and mortality rates reaching up to 40%\u003csup\u003e2,2\u0026ndash;5\u003c/sup\u003e.\u0026nbsp;Recent data suggest that, in addition to the presence of a pathogen, the disruption of healthy host-respiratory microbiome interactions plays a central role in HAP severity and inter-individual heterogeneity in the response to treatment\u003csup\u003e6,7\u003c/sup\u003e. Several studies have identified an association between lung microbiome dysbiosis and respiratory complications. Notably, the reduction in the healthy lung microbiome core (i.e., the microbial taxa typically found in healthy lung samples) was associated with HAP and acute respiratory failure\u003csup\u003e8\u003c/sup\u003e. This \u0026ldquo;healthy\u0026rdquo; bacterial core likely contributes to lung homeostasis\u003csup\u003e8,9\u003c/sup\u003e. Using viral metagenomics, we have recently demonstrated significant shifts in the lung viral community composition of intubated critically ill patients, suggesting a contribution of the lung virome to HAP pathogenesis and treatment outcomes\u003csup\u003e10\u003c/sup\u003e. Historically, pneumotypes have been defined as conserved combinations of microorganisms based on bacterial genus-level profiles from 16S rRNA data and a pneumotype enriched in supraglottic-associated taxa was linked to heightened subclinical lung inflammation in asymptomatic individuals\u003csup\u003e11\u003c/sup\u003e. More recently, a pneumotype enriched in oral-associated taxa at the time of diagnosis was associated with successful pneumonia therapy in patients with HAP and CAP\u003csup\u003e12\u003c/sup\u003e. In the present study, we extend this framework by defining functional pneumotypes, integrating the active bacteriome and the virome to capture the microbiome heterogeneity among critically ill patients before antibiotic exposure and HAP diagnosis. Two functional pneumotypes were identified as distinct meta-clusters through integrated analysis of viral metagenomics and metatranscriptomic data from endotracheal aspirates (ETAs) and associated with clinical outcomes.\u0026nbsp;\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003ePatients characteristics and multi-omic sequencing approach\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe analysed ETAs from ICU patients enrolled in the multicentre, placebo-controlled, randomised clinical trial (PREV-HAP study) testing interferon-gamma-1b for the prevention of HAP (inclusion across 11 European centres, April-October 2021). Adults (18-80 years), under invasive mechanical ventilation, with one or more acute organ failure were included within the first 48 hours of ICU hospitalisation, randomised to interferon-γ-1b or placebo (100 microg every 48 hours for 9 days), and followed clinically for 3 months. ETAs were collected at day 0 (immediately before the first treatment injection) and at days 3 and 7.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eBased on sample availability, we analysed 184 ETAs from 94 randomised patients enrolled. The integrated dataset comprises viral taxa and functions from metatranscriptomics, as well as bacterial taxa, functions, and predicted metabolic pathways. It also includes viral composition and functions from viral metagenomics (\u003cstrong\u003eFig. 1\u003c/strong\u003e, \u003cstrong\u003eSupplementary Tables S1-3 and Extended data Fig. 1-3\u003c/strong\u003e). Clinical characteristics did not differ between patients with and without HAP. The causes of ICU admission were 48% trauma, 37% surgical, and 15% medical. HAP developed in 45 patients (48%) at a median of 4 days after inclusion (interquartile range [IQR] 3–6). Out of the 94 patients, 22 (23%) died by day 90. During ICU stay, 26% received corticosteroids and 49% interferon-γ-1b (\u003cstrong\u003eSupplementary Table S4\u003c/strong\u003e). These treatments were found to have no effect on lung microbiome functions (\u003cstrong\u003eExtended data Fig. 4-5).\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePrior to HAP cluster derivation using metatranscriptomic and viral metagenomic data\u003c/p\u003e\n\u003cp\u003eTo investigate the heterogeneity in the lung microbiome without antimicrobial treatment exposure, we selected 64 ETA samples collected from 39 patients before HAP diagnosis (defined as Day 0, with earlier samples indexed relative to this reference point;\u0026nbsp;Supplementary Fig. S1). To identify distinct lung meta-clusters (so-called functional pneumotypes), we applied a three-step unsupervised analytical framework. First, we used MEFISTO (Method for the Functional Integration of Spatial and Temporal Omics Data) on viral metagenomics and metatranscriptomics to reduce the dimensionality of the selected features over time relative to HAP onset. Features found in at least 50% of the samples were retained. Second, unsupervised hierarchical clustering was subsequently applied to the first two resulting MEFISTO factors capturing the most variance to define discrete community structures. Model fit metrics indicated that a two-cluster solution (k = 2) best represented the data (Fig. 2a\u0026nbsp;and\u0026nbsp;Supplementary Fig. S2-S3). Third, because patients contributed multiple longitudinal samples before HAP onset, cluster assignment was determined by the cluster of the sample collected closest to HAP onset. Patients maintained consistent cluster assignments across their longitudinal samples, with minor cluster transition over time (Supplementary Fig. S4). This series of analyses demonstrated that two main functional pneumotypes of virome-microbiome status can be observed in ICU patients before HAP onset.\u003c/p\u003e\n\u003cp\u003eTo investigate the clinical significance of this result, we compared the mortality rates between the two pneumotypes. No major difference in baseline clinical characteristics was observed between the two pneumotypes (Supplementary Table S3). However, the hazard ratio of all-cause mortality and early successful extubation in pneumotype 1 was 0.2 and 2.33 (95% confidence interval (CI) [0.04, 0.58], log-rank test p = 0.00153, and 95% CI [0.94-5.8], p = 0.05, respectively;\u0026nbsp;Fig. 2b\u0026nbsp;and\u0026nbsp;Supplementary Fig. S5). Association of pneumotype 2 with all-cause mortality remained statistically significant after multivariate analysis taking into account major baseline risk factors of death (RR 9.60, 95%CI 1.14-80.8,\u0026nbsp;Fig. 2c).\u003c/p\u003e\n\u003cp\u003eTo investigate differences in the human immune response across pneumotypes, we performed a negative binomial generalized linear model (NB-GLM) identifying 85 differentially expressed (DE) human genes (false discovery rate [FDR] ≤ 0.2; Supplementary Table S5) (\u003cstrong\u003eFig. 2d\u003c/strong\u003e). Pneumotype 2 exhibited elevated expression of pathways associated with immune and inflammatory responses, including canonical NF-κB (nuclear factor kappa-B) signaling, interleukin-6 (IL-6) production, and cellular responses to viral or other biotic stimuli. In contrast, pneumotype 1 was associated with negative regulation of cytokine production, lymphocyte differentiation, interleukin (IL)-1b and T cell response (\u003cstrong\u003eFig. 2e\u003c/strong\u003e).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThese findings showed that the investigation of meta-clusters including of viral and bacterial composition and transcriptomic activity enables the identification of two functional pneumotypes associated with robust clinical outcomes in critically ill patients.\u003c/p\u003e\n\u003cp\u003eTranscriptional bacterial signature and score in proinflammatory high-risk pneumotype\u003c/p\u003e\n\u003cp\u003eTo determine how microbiome composition drove differences between the pneumotypes at ICU hospitalisation, we compared transcriptionally active bacteria, gene expression, functional pathways and diversity between groups using metatranscriptomic data. Patients with the pneumotype 1 (hereafter referred to as the low-risk pneumotype) exhibited greater β-diversity of transcriptionally active taxa across patients (Wilcoxon test, p = 4.88 × 10⁻¹¹; \u003cstrong\u003eFig. 3a\u003c/strong\u003e, \u003cstrong\u003eSupplementary Table S6\u0026nbsp;\u003c/strong\u003eand \u003cstrong\u003eExtended data Fig. 6\u003c/strong\u003e) than those with pneumotype 2 (high-risk pneumotype). The relative transcriptional activity of taxa constituting the healthy lung core microbiome, including \u003cem\u003eVeillonella\u003c/em\u003e, \u003cem\u003eStreptococcus\u003c/em\u003e, \u003cem\u003ePrevotella\u003c/em\u003e, \u003cem\u003eHaemophilus\u003c/em\u003e, and \u003cem\u003eFusobacterium\u003c/em\u003e\u003csup\u003e8\u003c/sup\u003e, was significantly reduced in the high-risk pneumotype 2 compared to the low-risk group (34% vs. 52%, respectively; Wilcoxon test, p = 0.03; \u003cstrong\u003eFig. 3b\u003c/strong\u003e). We next identified a panel of 23 differentially transcriptionally active bacterial species between low- and high-risk pneumotypes (\u003cstrong\u003eFig. 3c\u003c/strong\u003e and \u003cstrong\u003eSupplementary Table S6\u003c/strong\u003e). Bacteria with higher transcriptional activity in the high-risk pneumotype included \u003cem\u003eKlebsiella\u003c/em\u003e pneumoniae, \u003cem\u003eUreaplasma urealyticum\u003c/em\u003e, \u003cem\u003eHaemophilus influenzae\u003c/em\u003e, while \u003cem\u003eAlloprevotella_unclassified\u003c/em\u003e, \u003cem\u003ePrevotella pallens\u003c/em\u003e and \u003cem\u003eStreptococcus constellatus\u003c/em\u003e and \u003cem\u003eanginosus\u003c/em\u003e were more transcriptionally active in the low risk pneumotype.\u003c/p\u003e\n\u003cp\u003eThe overall expression of bacterial protein-coding genes was higher in the low-risk pneumotype (Wilcoxon test, p \u0026lt; 0.00001; \u003cstrong\u003eFig. 3d)\u003c/strong\u003e, which showed a stronger COG (Clusters of Orthologous Groups) functional profile. Eleven differentially expressed bacterial functions were significantly enriched in the low-risk pneumotype,\u0026nbsp;including cell motility, cell division and defense mechanisms (multivariate permutation test, p = 0.025, R² = 0.08; \u003cstrong\u003eFig. 3\u003c/strong\u003e\u003cstrong\u003ee\u003c/strong\u003e). Similarly, a pathway-level analysis revealed increased transcription of metabolic pathways in the low-risk group, with 28 upregulated\u0026nbsp;pathways versus only three in the high-risk group (Fisher’s exact test, p =\u0026nbsp;2.379e-06; \u003cstrong\u003eSupplementary Fig. S\u003c/strong\u003e\u003cstrong\u003e6\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eFinally, to create a simple and interpretable predictive tool for pneumotype classification, we applied a stacked machine learning approach that combined spectral clustering with a fast-and-frugal decision tree (FFT) model. The resulting four-factor decision tree robustly discriminated high-risk from low-risk pneumotypes (accuracy= 80%, sensitivity = 100%, specificity = 70%; \u003cstrong\u003eFig. 3\u003c/strong\u003e\u003cstrong\u003ef\u003c/strong\u003e).\u0026nbsp;The high-risk pneumotype was defined by lower relative transcriptional activity of \u003cem\u003eAlloprevotella_unclassified\u003c/em\u003e (\u0026lt;0.23%) and \u003cem\u003ePrevotellaceae_unclassified\u003c/em\u003e (≤0.29%), as well as higher relative expression of \u003cem\u003eParacoccus\u003c/em\u003e\u003cem\u003e_unclassified\u0026nbsp;\u003c/em\u003e(\u0026gt;2.21%) and \u003cem\u003eKlebsiella pneumoniae\u003c/em\u003e (\u0026gt;2.31%). The model’s predictive accuracy, assessed via tenfold cross-validation, achieved a mean area under the curve (AUC) of 0.8 (\u003cstrong\u003eFig. 3\u003c/strong\u003e\u003cstrong\u003eg\u003c/strong\u003e). To assess the clinical utility of such biomarkers in patients regardless of future HAP status, we used the entire cohort, selecting all samples within 0 to 4 days after admission, individuals classified as high risk pneumotype by XGBoost using the four bacterial features, had an increased mortality risk compared to those classified as low risk, as indicated by a hazard ratio of 3.9 (95% CI: 1.1–14, p = 0.025; \u003cstrong\u003eFig. 3\u003c/strong\u003e\u003cstrong\u003eh\u0026nbsp;\u003c/strong\u003eand \u003cstrong\u003eExtended data Fig. 7\u003c/strong\u003e). Taken together, this series of analyses demonstrated that the pre-HAP high-risk pneumotype was characterized by a reduction in the transcriptional activity of the healthy core microbiome, with several pathobionts showing high transcriptional activity.\u003c/p\u003e\n\u003cp\u003eBacteriophage patterns and score in pre-HAP high-risk patients\u003c/p\u003e\n\u003cp\u003eWe next evaluated the virome composition and function across pneumotypes using viral metagenomic. Bacteriophages were found to be the most abundant viral entities in the two pneumotypes, followed by giant viruses (\u003cstrong\u003eFig. 4a\u003c/strong\u003e, \u003cstrong\u003eSupplementary Fig. S7\u003c/strong\u003e and\u003cstrong\u003e\u0026nbsp;Supplementary Table S7\u003c/strong\u003e). Patients classified as high-risk pneumotype exhibited significantly lower Bray-Curtis dissimilarity than those assigned to the low-risk pneumotype (Wilcoxon test, p = 4.73×10⁻¹¹; \u003cstrong\u003eFig. 4b\u003c/strong\u003e, \u003cstrong\u003eSupplementary Table S7\u0026nbsp;\u003c/strong\u003eand \u003cstrong\u003eExtended data Fig. 6\u003c/strong\u003e). We subsequently identified 66 bacteriophage genera with significant differences in relative abundance between the low- and high-risk pneumotypes (\u003cstrong\u003eFig. 4c\u003c/strong\u003e and \u003cstrong\u003eSupplementary Table S7\u003c/strong\u003e). Functional profiling of the virome revealed distinct viral activity patterns across pneumotypes. Notably, the majority (69%) of the bacteriophages that were more abundant in the high-risk pneumotype were predicted to display a virulent lifestyle (\u003cstrong\u003eSupplementary Table S7\u003c/strong\u003e). Using bacteriophage metagenomics data, we observed that ten viral functions assigned were significantly abundant in the high-risk pneumotype, the top ones being the transport of chemicals molecules/nutrients, and the viral adhesion and invasion (GLM with multivariate permutation test, p = 0.005, R² = 0.07; \u003cstrong\u003eFig. 4d\u003c/strong\u003e). In addition, we identified using bacteriophage metatranscriptomics data, that the regulation of gene expression and the host-related functions (stress-associated response) were more transcriptionally expressed in high risk pneumotype (p = 0.001, R² = 0.14; \u003cstrong\u003eFig. 4e\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eA two-feature decision model using FFT achieved robust discrimination between high- and low-risk pneumotypes (accuracy= 90%, sensitivity = 94%, specificity = 84%; \u003cstrong\u003eFig. 4f\u003c/strong\u003e). The high-risk signature was primarily driven by elevated relative abundances of \u003cem\u003eRhizobium\u0026nbsp;\u003c/em\u003ebacteriophage, \u003cem\u003eCuauhnhuacivirus\u0026nbsp;\u003c/em\u003e(\u0026gt;0.06%), whose lifestyle is unknown, and \u003cem\u003eLactobacillus\u0026nbsp;\u003c/em\u003ebacteriophage, \u003cem\u003eLidleenavirus\u0026nbsp;\u003c/em\u003e(\u0026gt;0.09%), which is predicted to be virulent. Model performance, evaluated through tenfold cross-validation, yielded a mean AUC of 0.9 (\u003cstrong\u003eFig. 4g\u003c/strong\u003e). Using XGBoost with the two bacteriophage features, the entire patient cohort showed a\u0026nbsp;consistent trend of higher mortality among predicted high-risk individuals during the first 0 to\u0026nbsp;4\u0026nbsp;days following admission, which was analyzed independently of HAP status (HR = 4.8, 95% CI: 1.7-14, p = 0.0012; \u003cstrong\u003eFig. 4\u003c/strong\u003e\u003cstrong\u003eh\u0026nbsp;\u003c/strong\u003eand \u003cstrong\u003eExtended data Fig. 7\u003c/strong\u003e). These findings demonstrate that the high-risk pneumotype harbors a virome signature reflecting lung viral convergence, characterized by the expansion and increased activity of virulent (lytic) bacteriophages. Such enrichment suggests an active regulatory role of bacteriophages within the disturbed microbial ecosystem.\u003c/p\u003e\n\u003cp\u003eValidation of the functional pneumotype scores in an independent cohort\u003c/p\u003e\n\u003cp\u003eFinally, we aimed to validate the accuracy of these scores in an independently recruited cohort of patients admitted in the Nantes University Hospital ICU for severe brain injury and who required invasive mechanical ventilation (IBIS cohort). We analyzed 239 ETAs samples collected on days 0, 3 and/or 7 after ICU admission (n=117 patients). The HAP rate was 51%; with a median time of four days (IQR= 3-6 days) after being admitted to the ICU. The characteristics of the study population are described in \u003cstrong\u003eSupplementary Table S8\u003c/strong\u003e. The PREV-HAP and IBIS cohorts enrolled critically-ill patients in different years and different clinical centers, with different inclusion and exclusion criteria but similar sample processing and sequencing. In the entire patient cohort during the first 0 to 3 days following ICU inclusion, patients with high-risk pneumotype classification based on two viral, four bacterial, or six combined FFT features had higher mortality rates than those classified as low-risk. The corresponding HRs were 2 (95% CI 0.96-4), 3.25 (95% CI 1.1-9.5), and 2.4 (95% CI 1.1-4.9), respectively (\u003cstrong\u003eFig. 5a-c\u0026nbsp;\u003c/strong\u003eand \u003cstrong\u003eExtended data Fig. 7\u003c/strong\u003e). We found that patients classified using the combined FFT signature overlapped more with the viral signature \u003cstrong\u003e(Supplementary Fig S8-9)\u003c/strong\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eConsistent with our previous findings, the relative transcriptional abundance of the healthy lung bacterial core metatranscriptome taxa was significantly lower in the pre-HAP samples with a high-risk pneumotype signature (using bacterial FFT) than in those with a low-risk pneumotype signature (22% vs. 38%, respectively; p = 0.02; \u003cstrong\u003eFig. 5d\u003c/strong\u003e).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAdditionally, bacteriophages dominated the predicted high-risk (log₁₀ RPKM median = 4.3, IQR = 3.9-4.5) and low risk pneumotype (predicted using viral FFT) virome composition (log₁₀ RPKM median = 3.8, IQR = 3.5-4.4; \u003cstrong\u003eFig. 5e\u003c/strong\u003e; \u003cstrong\u003eSupplementary Table S8\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eFinally, we investigated whether host immune responses could further discriminate patients’ pneumotypes in the validation cohort using samples collected before HAP onset. A partial least squares discriminant analysis (PLS-DA) using 450 inflammatory, immune-related genes distinguished between high- and low-risk pneumotype patients (\u003cstrong\u003eSupplementary Fig. S10\u003c/strong\u003e). The predicted high-risk group's inflammatory profile was confirmed \u003cstrong\u003e(Fig. 5f\u003c/strong\u003e and \u003cstrong\u003eSupplementary Fig. S11)\u003c/strong\u003e. Overall, these findings demonstrate that viral and/or bacterial pneumotype classification schemes identify patients at high risk of mortality, whose lung dysbiosis is characterised by depletion of actively transcribed commensal bacteria, enrichment of bacteriophages, and expression of pro-inflammatory cytokines.\u003c/p\u003e\n\u003cp\u003eCausality inference and pathophysiological mechanisms of the high-risk pneumotype\u003c/p\u003e\n\u003cp\u003eTo further investigate the causal transkingdom interactions that may contribute to the high-risk pneumotype trajectory, we applied Transkingdom Network Analysis (TkNA) by integrating data from both PREV-HAP and IBIS cohorts to ensure statistical power. The resulting network comprised 62 nodes (including six bacterial species, thirteen bacteriophages, and forty-three human inflammatory genes) and 151 edges (\u003cstrong\u003eFig. 6a\u003c/strong\u003e; \u003cstrong\u003eSupplementary Table S9\u0026nbsp;\u003c/strong\u003eand \u003cstrong\u003eExtended data\u003c/strong\u003e). Twenty-seven nodes (three bacterial species, eight bacteriophages, and sixteen inflammatory genes) exhibiting a median importance score of 0.5 (IQR = 0.2-0.6) (\u003cstrong\u003eFig. 6a\u003c/strong\u003e, \u003cstrong\u003eSupplementary Fig. S12\u003c/strong\u003e and \u003cstrong\u003eSupplementary Table S9\u003c/strong\u003e), supporting their potential for causal involvement in severe pre-HAP pneumotype development. Among these nodes, we observed negative correlations between several bacteriophages (i.e., \u003cem\u003eColossusvirus\u003c/em\u003e, \u003cem\u003eFletchervirus\u003c/em\u003e, \u003cem\u003eKisquinquevirus\u0026nbsp;\u003c/em\u003eand \u003cem\u003eMagiavirus\u003c/em\u003e) (\u003cstrong\u003eSupplementary Fig. S13)\u003c/strong\u003e and the bacterial genera \u003cem\u003eStreptococcus_unclassified\u003c/em\u003e and \u003cem\u003eAlloprevotella_ unclassified\u003c/em\u003e, which were correlated with each other but anticorrelated with \u003cem\u003eKlebsiella\u003c/em\u003e \u003cem\u003epneumoniae\u003c/em\u003e. Moreover, \u003cem\u003eStreptococcus_ unclassified\u003c/em\u003e and \u003cem\u003eAlloprevotella_ unclassified\u003c/em\u003e showed strong negative interactions with a cluster of potent inflammatory genes encoding cytokines and receptors involved in immune activation including \u003cem\u003eNOS2\u003c/em\u003e, \u003cem\u003eIFNG\u003c/em\u003e, \u003cem\u003eIL2RA\u003c/em\u003e, \u003cem\u003eCCL2\u003c/em\u003e, \u003cem\u003eIL9\u003c/em\u003e, \u003cem\u003eIL12B\u003c/em\u003e, \u003cem\u003eTNFRSF4\u003c/em\u003e, \u003cem\u003ePTX3\u003c/em\u003e, \u003cem\u003eHRH1\u003c/em\u003e, \u003cem\u003eHRH4\u003c/em\u003e, \u003cem\u003eTIRAP\u003c/em\u003e, and \u003cem\u003eCD40\u0026nbsp;\u003c/em\u003e(\u003cstrong\u003eFig. 6a\u0026nbsp;\u003c/strong\u003eand \u003cstrong\u003eSupplementary Table S9\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eTwo potential bacteria-bacteriophage interactions were identified within the network. In the high-risk pneumotype, we observed a significant reduction in \u003cem\u003eStreptococcus_ unclassified\u003c/em\u003e transcriptional activity (p = 0.0095;\u003cstrong\u003e\u0026nbsp;Fig. 6b\u003c/strong\u003e) and an increase in the abundance of virulent \u003cem\u003eStreptococcus\u003c/em\u003e bacteriophage (\u003cem\u003eBrussowvirus\u003c/em\u003e; Wilcoxon test, p = 0.00006; \u003cstrong\u003eFig. 6c\u003c/strong\u003e), leading to high \u003cem\u003eStreptococcus_ unclassified\u003c/em\u003e virus-to-host ratio (VHR) (\u003cstrong\u003eFig. 6d\u003c/strong\u003e). In contrast, there was an enrichment of \u003cem\u003eKlebsiella\u003c/em\u003e \u003cem\u003epneumoniae\u003c/em\u003e (p = 0.0012; \u003cstrong\u003eFig. 6e\u003c/strong\u003e) and its associated temperate bacteriophage (\u003cem\u003eEowynvirus\u003c/em\u003e; p = 0.0001; \u003cstrong\u003eFig. 6f\u003c/strong\u003e) with no differences in \u003cem\u003eKlebsiella\u003c/em\u003e VHR among the groups (p = 0.18; \u003cstrong\u003eFig. 6g\u003c/strong\u003e). These results suggest that different relationships between bacteriophage-bacteria leading to high-risk pneumotype may apply, with a potential predator-prey relationship between \u003cem\u003eStreptococcus-Brussowvirus\u003c/em\u003e and piggyback-the-winner between \u003cem\u003eKlebsiella-Eowynvirus\u003c/em\u003e.\u003c/p\u003e\n\u003cp\u003eFurthermore, GLM modeling of gene expression across groups revealed increased expression of \u003cem\u003eCCL2\u0026nbsp;\u003c/em\u003e(p=0.00019; \u003cstrong\u003eFig. 6h\u003c/strong\u003e), \u003cem\u003ePLGRKT\u0026nbsp;\u003c/em\u003e(p=0.00029; \u003cstrong\u003eFig. 6i\u003c/strong\u003e) and \u003cem\u003ePTX3\u0026nbsp;\u003c/em\u003e(p=0.01; \u003cstrong\u003eFig. 6j\u003c/strong\u003e) in high-risk patients. Overall, these results support the hypothesis that bacteriophage-driven predation contributes to bacterial dysbiosis, promoting heightened inflammatory responses in patients who progress toward severe, high-risk HAP.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this study, we used an unsupervised and multi-omic approach to analyse endotracheal aspirates of critically ill patients. We integrated metatranscriptomic and viral metagenomic profiles in order to investigate lung microbiome heterogeneity. This strategy revealed two pre-HAP functional pneumotypes associated with all-cause mortality before antibiotic treatment. The high-risk pneumotype showed increased mortality, a strong proinflammatory signature, and loss of healthy lung bacterial transcriptional activity. It also displayed elevated transcriptional activity of pathobionts such as \u003cem\u003eK. pneumoniae\u003c/em\u003e. Additionally, this pneumotype showed a virome convergence driven by increased absolute abundance of bacteriophages. Finally, we defined two predictive lung signature scores derived from active bacteriome and virome features. These scores were validated in an external prospective cohort, predicting both pneumotypes across all ICU patients. Together, these distinct functional pneumotypes identified within the first days of ICU admission support plausible pathophysiological mechanisms underlying their divergent clinical outcomes.\u003c/p\u003e\n\u003cp\u003eThe observation of two pre-HAP patient groups was notable, as it mirrors previous reports of two subphenotypes among 3,889 critically ill patients at HAP diagnosis. These subphenotypes, which are linked to mortality and inflammation, were defined by clinical characteristics and routine biological tests\u003csup\u003e6\u003c/sup\u003e. In the present study, RNA expression profiling revealed that high-risk pneumotype showed lymphocyte and leukocyte proliferation, IL-6 production, and NF-\u0026kappa;B activation, sustaining a proinflammatory profile related to poor outcomes\u003csup\u003e13\u0026ndash;16\u003c/sup\u003e. These suggest that the lung virome-bacteriome profile of ICU patients reflects the evolution of their clinical status in real time. As these pneumotypes precede HAP onset, early lung inflammation may predispose patients for clinical deterioration and high-mortality HAP. This underscores the role of lung viral-microbial communities in both disease susceptibility and progression\u003csup\u003e8,10\u003c/sup\u003e,\u0026nbsp;highlighting the potential of longitudinal bedside microbiome profiling to guide personalized interventions.\u003c/p\u003e\n\u003cp\u003eHigh-risk patients exhibited viral convergence and enrichment with virulent bacteriophages as recently documented\u003csup\u003e10\u003c/sup\u003e. In this latest work, we reported significant viral convergence in ETA samples from patients who later developed HAP compared with those who did not\u003csup\u003e10\u003c/sup\u003e. In the present study, we observed an even stronger viral convergence in the high-risk pneumotype viromes than in the low-risk within the upcoming HAP group. Additionally, we identified an enrichment of virulent \u003cem\u003eCaudoviricetes\u0026nbsp;\u003c/em\u003ebacteriophages, which are anticorrelated with bacteria belonging to the core healthy respiratory group. Members of this tailed bacteriophage class are described among the key contributors to HAP onset in the lung viromes of critically ill patients\u003csup\u003e10\u003c/sup\u003e. We next derived a bacteriophage-based predictive score using viral metagenomic data. The decision tree incorporated two bacteriophages enriched in the high-risk pneumotype. \u003cem\u003eCuauhnahuacvirus\u003c/em\u003e, a \u003cem\u003eRhizobium\u0026nbsp;\u003c/em\u003ebacteriophage previously associated with prolonged mechanical ventilation in ICU patients\u003csup\u003e8\u003c/sup\u003e, and \u003cem\u003eLidleunavirus\u003c/em\u003e, a virulent bacteriophage increased in the ileal mucosal virome of IBD patients\u003csup\u003e17\u003c/sup\u003e of \u003cem\u003eLactobacillus\u0026nbsp;\u003c/em\u003ethat was reported in ICU patients with early successful extubation\u003csup\u003e8\u003c/sup\u003e. The link between virome and lung disease has been described previously. Sputum metagenomes from 99 COPD patients and 36 controls revealed disrupted virus-bacteria ecological dynamics and progressive loss of bacteriophage diversity, particularly \u003cem\u003ePorphyromonas\u003c/em\u003e-associated bacteriophages, in patients with frequent exacerbations\u003csup\u003e18\u003c/sup\u003e. Similarly, metatranscriptomic analysis of 278 bronchoalveolar lavage samples from 229 paediatric HCT patients identified viral enrichment as a key driver of fatal lung injury and high-risk in-hospital outcomes\u003csup\u003e19\u003c/sup\u003e. In our previously described pathophysiological model, we hypothesised that \u003cem\u003ePrevotella\u0026nbsp;\u003c/em\u003eand \u003cem\u003eStreptococcus\u0026nbsp;\u003c/em\u003ewere targeted by virulent bacteriophages via predator-prey interactions, resulting in lung microbial dysbiosis and promoting HAP onset\u003csup\u003e10\u003c/sup\u003e. Consistent with this model, we observe a significant increase in the \u003cem\u003eStreptococcus\u0026nbsp;\u003c/em\u003evirus-to-host ratio suggesting that virulent \u003cem\u003eStreptococcus\u0026nbsp;\u003c/em\u003ebacteriophage (\u003cem\u003eBrussowvirus\u003c/em\u003e) infected and depleted their bacterial hosts via a classic kill-the-winner (predator\u0026ndash;prey) dynamic. This recurrent pattern further supports our hypothesis that bacteriophage-bacteria interactions contribute to dysbiosis associated with worse clinical outcomes. However, this study identifies the presence of \u003cem\u003eK. pneumoniae\u0026nbsp;\u003c/em\u003eexhibiting an opposing pattern: when bacterial transcription activity was high, its temperate bacteriophage (\u003cem\u003eEowynvirus\u003c/em\u003e) multiplied as a result of bacterial division and adopted the lysogenic lifestyle following the Piggyback-the-Winner hypothesis\u003csup\u003e20\u003c/sup\u003e. This mechanistic scenario provides a plausible explanation for the observed lung microbiome destabilization that has led to enhanced inflammatory responses. In line with these patterns, both DNA and RNA profiles showed elevated bacteriophage activity in the high-risk pneumotype. These results suggest that high-risk lung virome is actively undergoing functional remodelling, which is associated with shifts in bacteriophage abundance and potentially with changes in the genetic composition of individual bacteriophages.\u003c/p\u003e\n\u003cp\u003eUsing metatranscriptomics, we defined the transcriptionally active bacteriome in lung fluids and identified a bacterial dysbiosis in the high-risk pneumotype, reflected by reduced activity of core healthy pulmonary taxa. This mirrors the dysbiosis previously identified using 16S profiles in the severe HAP subphenotype\u003csup\u003e6\u003c/sup\u003e and highlights the added resolution of metatranscriptomics in distinguishing active pathogens and identifying microbial drivers of infection\u003csup\u003e19,21\u0026ndash;23\u003c/sup\u003e. Besides, metatranscriptomic data has contributed to the identification of the functional lung pneumotypes in critically ill patients before antibiotic exposure. Recent work has demonstrated that, at pneumonia diagnosis and under antibiotic treatment, the lung microbiota segregates into four distinct 16S rRNA gene-based pneumotypes reflecting the disruption of the microbial landscape during pneumonia\u003csup\u003e12\u003c/sup\u003e. Notably, a pneumotype enriched in oral-associated taxa, including \u003cem\u003eStreptococcus\u003c/em\u003e, was associated with successful treatment of HAP and with upregulation of IL-1 signalling pathways in alveolar macrophages, partially mirroring our low-risk functional pneumotype. We further derived a four-factor bacterial transcriptomic signature, in which high-risk status was defined by low \u003cem\u003eAlloprevotella\u0026nbsp;\u003c/em\u003eexpression, a commensal bacteria in ventilated ICU patients\u003csup\u003e24\u003c/sup\u003e also linked to treatment-responsive lung masses\u003csup\u003e25\u003c/sup\u003e and improved COPD outcomes\u003csup\u003e26\u003c/sup\u003e, and by elevated \u003cem\u003eParacoccus\u0026nbsp;\u003c/em\u003eand \u003cem\u003eK. pneumoniae\u003c/em\u003e. \u003cem\u003eParacoccus\u0026nbsp;\u003c/em\u003eis enriched in ARDS-associated microbiomes\u003csup\u003e8\u003c/sup\u003e, while \u003cem\u003eK. pneumoniae\u003c/em\u003e is a common VAP pathogen\u003csup\u003e27\u003c/sup\u003e associated with higher mortality, disease severity, and respiratory failure in ICU cohorts\u003csup\u003e28\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eViral- and bacterial-derived scores effectively identified pre-HAP pneumotypes in an external ICU cohort, even within the first days of admission, underscoring their utility for early risk stratification. These results support the concept that stabilizing the lung microbiome through prudent antimicrobial use, probiotic approaches, or targeted bacteriophage-based strategies may help prevent the dysbiosis preceding severe HAP\u003csup\u003e29\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eThis study has limitations: The discovery cohort was small, with only 39 pre-antibiotic ETA samples from patients who later developed HAP used for multi-omic pneumotype identification. The predictive model could not be fully validated externally, as pneumotypes are inferred rather than being clinically recorded. Causality cannot be confirmed without \u003cem\u003ein vivo\u003c/em\u003e experimentation. Nevertheless, a key strength is the validation of our findings in a prospective cohort.\u003c/p\u003e\n\u003cp\u003eIn conclusion, our multi-omic analysis revealed substantial heterogeneity in the transcriptomic activity of the lung virome and bacteriome of ICU patients prior to HAP. The distinct functional pneumotypes associated with inflammation and increased mortality highlight the clinical importance of identifying airway community structure early on. Machine learning-based scores incorporating four bacterial and two bacteriophage markers accurately stratified patients upon admission and identified those at elevated risk. Furthermore, we detected a bacterial-viral dysbiosis signature accompanied by inflammatory gene expression several days prior to HAP onset, suggesting that early microbial disruption may contribute to disease development. Together, these findings refine our understanding of HAP pathophysiology and lay the groundwork for future diagnostic and microbiome-targeted interventions in critical illness.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eCohorts Description and ETAs Collection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe discovery cohort included 94 patients enrolled in the PREV-HAP clinical trial. This was an investigator-initiated, multicenter, parallel-group, double-blind, randomized study conducted in 11 ICUs in France, Spain, and Greece. The study protocol was approved by the Ouest II Angers Ethics Committee in France in March 2021, and the trial adhered to the Declaration of Helsinki. The trial was registered on ClinicalTrials.gov (NCT04793568) in the same month. Written informed consent was obtained from each patient’s legal surrogate. In accordance with local regulations, patients could be included before surrogate consent was obtained if the next of kin could not be contacted within the allowed enrollment window. Follow-up consent was requested from the patient within 90 days of inclusion, when feasible. Eligible patients were aged 18–85 years, receiving invasive mechanical ventilation, and had at least one acute organ failure at the time of enrollment. Participants were randomized in a 1:1 ratio to receive either interferon gamma-1b (100 µg/0.5 mL vials, IMUKIN®, Clinigen®) or placebo (normal saline) in fixed blocks of six, stratified by hospitalization cause (sepsis vs. other) and country (France, Greece, Spain). Key exclusion criteria included pregnancy or breastfeeding, hypersensitivity to interferon gamma-1b, pre-existing immunosuppression (e.g., recent chemotherapy or radiotherapy, AIDS, leukopenia), severe hepatic insufficiency (Child-Pugh B or C), liver cytolysis (transaminases \u0026gt;5× normal), chronic renal failure (MDRD GFR \u0026lt;10 mL/min/1.73 m²), persistent coma post-resuscitated cardiac arrest, cervical spinal cord injury hospitalization, previous hospital-acquired pneumonia during the current admission, and sustained hyperlactatemia (\u0026gt;5 mmol/L). Patients received five subcutaneous injections of either interferon gamma-1b or matching placebo from day 1 to day 9 (one injection every 48 hours). Follow-up continued for 90 days. HAP and ARDS diagnoses were blindly reviewed by intensivists for compliance with international definitions. HAP was defined as pneumonia occurring ≥48 hours after hospital admission, requiring at least two clinical signs (fever \u0026gt;38°C, leukocytosis \u0026gt;12,000 cells/µL, leukopenia \u0026lt;4,000 cells/µL, or purulent pulmonary secretions), accompanied by a new or worsening infiltrate on chest imaging, and confirmed with semi-quantitative or quantitative respiratory cultures obtained prior to initiating new antimicrobial therapy. All respiratory infection diagnoses were reviewed by an independent adjudication committee following European HAP guidelines. ARDS was defined as severe hypoxemia (PaO₂/FiO₂ \u0026lt; 200 mmHg with PEEP ≥5 cm H₂O) and bilateral opacities on chest imaging within one week of worsening respiratory symptoms. Information on ancestry, race, ethnicity, and socioeconomic status was not available for either cohort due to legal restrictions in France.\u003c/p\u003e\n\u003cp\u003eIn the validation cohort IBIS (Immunology and Brain Injury Study), 117 patients were enrolled in two French Surgical Intensive Care Units of one university hospital (Nantes, France). The collection of human samples has been declared to the French Ministry of Health (Programme de recherche “Immunologie”, DC-2017-2987), and was approved by the Comite de Protection des Personnes Ouest IV (7/04/2015 and 08/10/2020) (number clinicaltrials.gov NCT02003196). Written informed consent from a next-of-kin was required for enrolment. Retrospective consent was obtained from patients when possible. Appropriate consent was obtained for the release of information from deceased individuals. Participants received no compensation. Inclusion criteriawere male or female, 18-80 years old, brain injury (Glasgow Coma Scale below or equal to 12 and abnormal brain-CT scan) and receiving invasive mechanical ventilation. Exclusion criteria were cancer with radiotherapy or chemotherapy in the last 90 days, AIDS, leukopenia in the previous five years, immunosuppressive drugs, preexisting immunosuppression and pregnancy. All patients were clinically followed up for 90 days. Intensivists blindly reviewed HAP and ARDS diagnoses for compliance with international definitions\u003csup\u003e5,30,31\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eIn this study, a total of 423 endotracheal aspirates (ETAs) were collected from critically ill patients enrolled in two independent cohorts: PREV-HAP (184 ETAs from 94 patients) and IBIS (239 ETAs from 117 patients), based on sample availability. Within the PREV-HAP cohort, 122 ETAs were selected for metatranscriptomic analyses and 133 ETAs for viral metagenomic analyses. In the IBIS cohort, all 239 ETAs were included in both metatranscriptomic and viral metagenomic investigations.\u003c/p\u003e\n\u003cp\u003eMetatranscriptomic sequencing for integrated microbiome and host analysis\u003c/p\u003e\n\u003cp\u003eMetatranscriptomic analysis was conducted using the Revelo kit (Tecan) according to the protocol described by Destras et al. 2024. Briefly, 50 µL of ETA samples or no-template controls (NTCs) were spiked with an RNA internal control (MS2 bacteriophage, 2800 bp) at a concentration corresponding to a Ct value of 36. Nucleic acids were then extracted using the Maxwell platform (Promega) with the DNA Blood Kit. Sequencing was performed on a NovaSeq 6000 system using an SP flow cell for PREV-HAP cohort and S2 cartridge for IBIS, with 2 × 100 bp paired-end reads.\u003c/p\u003e\n\u003cp\u003eMetagenomic sequencing for viral profiling\u003c/p\u003e\n\u003cp\u003eETAs were preserved diluted (dilution factor of 1:7.4) in a viral transport medium (LMR1925VTM, Labomoderne) and stored at -20°C. Samples from both cohorts were treated according to an adapted version of the protocol from Bal and colleagues and Anani et al. Briefly, the MS2 RNA bacteriophage (MS2, IC1 RNA internal control; bioMérieux R-GENE® ref. 71-110) was spiked in each sample to validate the whole metagenomic process. The samples were then filtered through 0.45 µm filters and incubated with DNase (Life Technologies ®, Carlsbad, USA) for 90 min. Total nucleic acid extraction was performed on the Qiagen EZ1 Advanced XL Extractor using the DSP Virus Kit, Qiagen®. After extraction, both RNA and DNA amplification were performed using a Whole Transcriptome Amplification kit (WTA2, Sigma-Aldrich®, Darmstadt, Germany). Amplified DNA was purified using QIAquick spin columns (Qiagen®, Hilden, Germany). No-template controls (NTC) and MS2 spiked-in NTC were prepared for each sample batch using nuclease-free water that followed the whole process starting from dilution in the transport media. The viral transport medium was also sequenced to confirm the absence of viral contamination.\u0026nbsp;Before sequencing, quality control testing was performed with MS2 qPCR (R-GENE® ref. 71-110). The libraries were prepared using the post-PCR portion of the Illumina (San Diego, USA) COVIDSeq ® Test Kit for PREV-HAP samples and the Nextera XT Kit for IBIS samples. The libraries were then sequenced on a NovaSeq 6000 using SP (PREV-HAP) and S2 (IBIS) flow cells with 2×100 bp reads.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eBioinformatic pipeline for virome analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eRaw sequencing reads were dehosted using SraHumanScrubber v2.2.1 to remove human-derived sequences and trimmed with fastp v0.23.4 to eliminate adapters, low-complexity regions, and low-quality bases, retaining only reads longer than 30 bp. Taxonomic classification was performed with Kraken2 v2.1.2 (parameters: minimum-hit-groups = 2, confidence = 0.1) using a custom database incorporating viral taxa from RVDB (v29.0) and Inphared (v2, January 2025)\u003csup\u003e32\u003c/sup\u003e, as well as archaeal, bacterial, fungal, and human references from RefSeq (downloaded on 06/01/2025)\u003csup\u003e33\u003c/sup\u003e. Non-viral reads assigned to Eukaryota, Bacteria, Archaea, Fungi, and their taxonomic descendants were filtered out using Krakentools, retaining only putative viral reads. These viral reads were assembled de novo with metaSPAdes\u003csup\u003e34\u003c/sup\u003e v3.15.5, and the resulting contigs were dereplicated using CD-HIT\u003csup\u003e35\u003c/sup\u003e v4.7 to generate operational taxonomic units (OTUs) at ≥95% nucleotide identity and ≥85% coverage, following MIUViG (Minimum Information about an Uncultivated Virus Genome) standards. OTUs shorter than 500 bp were discarded. The length-filtered OTUs were analyzed with PhaBOX2\u003csup\u003e36\u003c/sup\u003e v2.1.13 to identify viral OTUs (vOTUs), assign taxonomy according to the latest International Committee on Taxonomy of Viruses (ICTV) release, and predict host range and lifestyle for bacteriophages using CHERRY\u003csup\u003e36\u003c/sup\u003e and PhaTYP\u003csup\u003e36\u003c/sup\u003e, respectively. Finally, quality-filtered viral reads were realigned to the vOTU sequences using the Burrows–Wheeler Aligner (BWA)\u003csup\u003e37\u003c/sup\u003e v0.7.17, and count tables were generated with samtools. Viral decontamination was performed following the recently published MS2 internal control based method described by Anani and colleagues\u003csup\u003e10\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eBioinformatics analysis of metatranscriptomic data\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMetatranscriptomic data were analyzed using a pipeline that processed each kingdom independently. Prior to downstream analyses, all reads (FASTQ files) were trimmed for quality using fastp\u0026nbsp;v0.23.4 (reference). First, human transcriptomic analyses were performed on non-dehosted, trimmed reads using RASflow\u003csup\u003e38\u003c/sup\u003e v2.0, HISAT2 v2.2.1, and the GRCh38 reference genome with default parameters. Duplicate reads were removed using Samtools, and FeatureCounts v2.0.3 was used to generate a gene count matrix. Only protein-coding genes detected in at least 10% of the samples were retained for further analysis. Second, for bacterial metatranscriptomic analyses, trimmed reads were first dehosted using SraHumanScrubber v2.2.1 with the GRCh38 human reference genome. Bacterial reads were then identified using Kraken2\u003csup\u003e32\u003c/sup\u003e v2.1.2 (minimum-hit-groups = 2, confidence = 0.1) against the previously described custom database. Reads assigned to the bacterial taxon (taxid = 2) were extracted using Krakentools. From these, 16S rRNA reads were removed using SortMeRNA\u003csup\u003e39\u003c/sup\u003e v4.3.6 with the SILVA database (v138.2), and the remaining reads were assembled de novo with rnaSPAdes\u003csup\u003e34\u003c/sup\u003e v3.15.5. Assembled bacterial transcripts were dereplicated using CD-HIT\u003csup\u003e35\u003c/sup\u003e v4.7 to generate operational taxonomic units (OTUs) with ≥95% sequence identity and ≥95% coverage. OTUs shorter than 1 kb were discarded. The remaining length-filtered OTUs were taxonomically assigned using DIAMOND\u003csup\u003e40\u003c/sup\u003e v2.1.6.160 with the lowest common ancestor (LCA) algorithm against the NCBI RefSeq Bacteria database (version 4). Quality control of bacterial OTUs (bOTUs) was performed to assess completeness and contamination. Functional annotation was carried out with Prokka\u003csup\u003e41\u003c/sup\u003e v1.14.6, and bOTUs lacking coding sequences were excluded. Quality-filtered bacterial reads used for de novo assembly were realigned to the curated bOTUs using the Burrows–Wheeler Aligner (BWA)\u003csup\u003e37\u003c/sup\u003e v0.7.17, and count tables were generated with samtools. Finally, potential bacterial contaminants were identified using the decontam\u003csup\u003e42\u003c/sup\u003e R package v1.20.0, applying the prevalence method with a threshold of 0.5 on read count tables normalized to RPKM.\u0026nbsp;Third, for viral metatranscriptomic analyses, we performed preprocessing, quality assessment, and dehosting of raw paired-end reads using fastp v0.23.4 and SraHumanScrubber v2.2.1, applying the same parameters used in the viral metagenomic pipeline. All subsequent steps followed the viral metagenomic workflow parameters, except for the de novo assembly step. For this step, rnaSPAdes was used instead of metaSPAdes. Viral decontamination was carried out as described for viral metagenomic analyses.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunctional microbiome and virome analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor the bacterial component, high-quality reads obtained after ribosomal RNA filtration were analyzed using the HUMAnN\u003csup\u003e43\u003c/sup\u003e v3.9 pipeline from Biobakery to infer bacterial metabolic pathways. For functional annotation, all bacterial transcripts or bOTUs were annotated with Prokka\u003csup\u003e41\u003c/sup\u003e v1.14.6\u0026nbsp;(using the bacterial kingdom option), and the resulting protein sequences were searched against the latest NCBI database of Clusters of Orthologous Genes (COGs)\u003csup\u003e44\u003c/sup\u003e using BLASTp. For the viral component, derived from viral metagenomic or metatranscriptomic datasets, all vOTUs were annotated with Prodigal v2.6.3. The predicted viral proteins were then compared against latest UniProt SPROT curated viral databases, and Gene Ontology (GO) terms were retrieved from UniProtKB to classify viral proteins into functional categories following Cao and colleagues\u003csup\u003e17\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical analysis:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eUnsupervised clustering analysis\u003c/strong\u003e. Microbiome data characterized by virome metagenomes, virome and bacteriome transcriptomes and their corresponding identified proteins in addition to the bacterial metabolic pathway dataset were analyzed all using MEFISTO\u003csup\u003e45\u003c/sup\u003e (Method for the Functional Integration of Spatial and Temporal Omics data) to reduce dimensionality and identify a core set of factors over time by launching the python/R coupled library MOFA2\u003csup\u003e46\u003c/sup\u003e v1.18.0. This method effectively handles different data structures and distributions while being robust to collinearity. The data were filtered to retain features present in more than 50% of the samples and subsequently subjected to a log₁₀(x + 1) transformation. MEFISTO identified 15 latent factors. The factors 1 and 2 were then subjected to unsupervised hierarchical clustering, and two clusters were determined to be optimal based on Silhouette, Elbow, and Gap statistics. Then, each patient was assigned principal component coordinates derived from the first two components. These coordinates were visualized in a two-dimensional space using the FactoMineR v2.12 R package.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical characteristics.\u0026nbsp;\u003c/strong\u003eWe fitted a Cox proportional hazards model to assess the association between clinical variables and ICU mortality. This model included cluster assignment, demographic factors, and key clinical parameters. We verified model performance and proportionality assumptions and visualized\u0026nbsp;relative risks (RRs) with 95% confidence intervals using a forest plot generated with the ggforest function from the survminer v0.5.1 R package. We conducted Kaplan–Meier survival analyses to evaluate in-hospital mortality and the probability of invasive mechanical ventilation across the identified clusters. To account for competing risks, death was considered a censoring event for ICU length of stay and mechanical ventilation duration. The survival v3.8.3 R package was used to generate survival curves and compare them using the log-rank test to assess differences in survival distributions between clusters.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHost gene expression.\u0026nbsp;\u003c/strong\u003eDifferentially expressed genes (DEGs) were identified using edgeR v4.6.3 within each cluster based on normalized counts. We visualized the DEGs using volcano plots to illustrate significance and fold-change distribution. Gene set enrichment analysis (GSEA) was then performed\u0026nbsp;on upregulated DEGs of each cluster using the clusterProfiler v4.16.0 and org.Hs.eg.db v3.21.0 R packages.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMicrobiome comparisons and predictions.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAlpha diversity, as measured by the Shannon index and richness, was calculated at the species level for bacteria and at the genus level for viruses. This calculation was performed using the vegan v2.7.1 package in R (v4.5.1). Diversity distributions were visualized using the ggplot2 v3.5.0 package, and longitudinal trends were modeled with loess regression and 95% confidence intervals. The ANCOM-BC\u003csup\u003e47\u003c/sup\u003e v2.10.1 was applied to identify features significantly associated with the identified clusters and to account for sampling bias. Signature outputs were visualized using the SIAMCAT\u003csup\u003e48\u003c/sup\u003e v2.12.0 package.\u0026nbsp;To identify bacterial/viral functional categories associated with each cluster, a generalized linear model (GLM) with a binomial family was applied. Briefly, normalized features (RPKM values) were compared between clusters by fitting individual GLMs for each function, modeling cluster membership as the binary outcome. For each function, log-odds estimates and p-values were extracted, and features with p \u0026lt; 0.05 were considered significant. In addition, a multivariate PERMANOVA based on Bray–Curtis dissimilarities was performed using the adonis2 function of vegan v2.7.1 to assess the overall functional compositional differences between clusters. PLS-DA analyses were performed using the mixOmics R package v6.32.0. Comparisons between the virome and bacteriome compositions in both the discovery and validation cohorts were performed using the UpSetR v1.4.0 R package for visualization of shared and unique features. Machine learning analyses were conducted using the FFTrees\u003csup\u003e49\u003c/sup\u003e v2.1.0, XGBoost\u003csup\u003e50\u003c/sup\u003e v1.7.11.1 and SIAMCAT\u003csup\u003e48\u003c/sup\u003e v2.12.0 R packages to evaluate the predictive power of the selected signatures, and receiver operating characteristic (ROC) curves were used to assess model performance.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTranskingdom Network Analysis (TkNA\u003csup\u003e51\u003c/sup\u003e v1.2.1) was applied to perform causal inference using log\u003csub\u003e10\u003c/sub\u003e(relative abundance data) from the viral/bacterial signatures and the log\u003csub\u003e10\u003c/sub\u003e (TMM-normalized counts) from the host genes signature, with the resulting correlation network visualized in Cytoscape\u003csup\u003e52\u003c/sup\u003e v3.10.2. To identify the key nodes in the network, we computed an importance score by integrating three topological features: node degree, betweenness-based bipartite centrality (BiBC), and the probability of randomly selecting each node. Each metric was normalized between 0 and 1, and the probability was inverted so that lower values indicated higher importance. The final importance score was obtained by averaging the normalized degree, BiBC, and probability values. For comparisons of numerical and categorical variables, the Wilcoxon test, two-tailed Mann–Whitney U test, and Fisher’s exact test were applied. Statistical significance was defined as p ≤ 0.05, and symbols were used as follows: ns, not significant (p \u0026gt; 0.05); *, p ≤ 0.05; **, p ≤ 0.01; ***, p ≤ 0.001; ****, p ≤ 0.0001.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eReporting summary\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFurther information on research design is available in the Nature Portfolio Reporting Summary linked to this article.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe metatranscriptomic and the viral metagenomic raw data have been deposited in the BioProject repository PRJNA1373167 and PRJNA1371704, respectively.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll code used for the analyses in this study is available on GitHub at https://github.com/genepii/HAP2-PREVHAP-HAP_pneumotypes.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe thank all clinicians and staff involved in the collection of endotracheal aspirates from ICU patients. The study was funded by the European Union\u0026apos;s Horizon 2020 research and innovation program under grant agreement number 847782 (HAP2 project) and MSD Avenir grant (Phenomenon project). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eH.A., C.B.B., A.R. and L.J. conceived the study. S.B., G.B., F.P.M., C.P., M. P., E.O. and Q.S. performed the clinical data curation, sample preparations and sequencing. H.A. performed bioinformatic analysis. H.A., G.D., F.P.M., V.G., C.B.B., A.R. and L.J. interpreted results and drafted the manuscript. H.A., G.D., F.P.M., C.P., C.B.B., A.R. and L.J revised the manuscript. All authors reviewed and approved the final version of the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eL.C., E.M. and A.R. have a patent on respiratory microbiome composition in ICU patients (\u003cstrong\u003eWO2025078559A2\u003c/strong\u003e). The authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCorrespondence and requests\u0026nbsp;\u003c/strong\u003efor materials should be addressed to Laurence Josset.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eHe, Q. \u003cem\u003eet al.\u003c/em\u003e The epidemiology and clinical outcomes of ventilator-associated events among 20,769 mechanically ventilated patients at intensive care units: an observational study. \u003cem\u003eCrit. Care Lond. Engl.\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 44 (2021).\u003c/li\u003e\n\u003cli\u003eLuyt, C.-E. \u003cem\u003eet al.\u003c/em\u003e Pulmonary infections complicating ARDS. \u003cem\u003eIntensive Care Med.\u003c/em\u003e \u003cstrong\u003e46\u003c/strong\u003e, 2168\u0026ndash;2183 (2020).\u003c/li\u003e\n\u003cli\u003eLagarde, C. \u003cem\u003eet al.\u003c/em\u003e A multifaceted strategy to optimize pharmacokinetics of antimicrobial therapy in patients with hospital-acquired infections\u0026mdash;a monocentre quality improvement project. \u003cem\u003eJ. Antimicrob. Chemother.\u003c/em\u003e \u003cstrong\u003e78\u003c/strong\u003e, 1378\u0026ndash;1385 (2023).\u003c/li\u003e\n\u003cli\u003ePoulain, C. \u003cem\u003eet al.\u003c/em\u003e Clinical evaluation of the BioFire Respiratory Pathogen Panel for the guidance of empirical antimicrobial therapy in critically ill patients with hospital-acquired pneumonia: A multicenter, quality improvement project. \u003cem\u003eAnaesth. Crit. Care Pain Med.\u003c/em\u003e \u003cstrong\u003e43\u003c/strong\u003e, 101353 (2024).\u003c/li\u003e\n\u003cli\u003eTorres, A. \u003cem\u003eet al.\u003c/em\u003e International ERS/ESICM/ESCMID/ALAT guidelines for the management of hospital-acquired pneumonia and ventilator-associated pneumonia: Guidelines for the management of hospital-acquired pneumonia (HAP)/ventilator-associated pneumonia (VAP) of the European Respiratory Society (ERS), European Society of Intensive Care Medicine (ESICM), European Society of Clinical Microbiology and Infectious Diseases (ESCMID) and Asociaci\u0026oacute;n Latinoamericana del T\u0026oacute;rax (ALAT). \u003cem\u003eEur. Respir. J.\u003c/em\u003e \u003cstrong\u003e50\u003c/strong\u003e, 1700582 (2017).\u003c/li\u003e\n\u003cli\u003eMartin, F. P. \u003cem\u003eet al.\u003c/em\u003e Identification and validation of robust hospital-acquired pneumonia subphenotypes associated with all-cause mortality: a multi-cohort derivation and validation. \u003cem\u003eIntensive Care Med.\u003c/em\u003e \u003cstrong\u003e51\u003c/strong\u003e, 692\u0026ndash;707 (2025).\u003c/li\u003e\n\u003cli\u003eDeng, J. \u003cem\u003eet al.\u003c/em\u003e Macrophage MR1 antigen presentation promotes MAIT cell immunity and lung microbiota modulation. \u003cem\u003eScience\u003c/em\u003e eadr6322 (2025) doi:10.1126/science.adr6322.\u003c/li\u003e\n\u003cli\u003eMontassier, E. \u003cem\u003eet al.\u003c/em\u003e Robust airway microbiome signatures in acute respiratory failure and hospital-acquired pneumonia. \u003cem\u003eNat. Med.\u003c/em\u003e \u003cstrong\u003e29\u003c/strong\u003e, 2793\u0026ndash;2804 (2023).\u003c/li\u003e\n\u003cli\u003eLi, R., Li, J. \u0026amp; Zhou, X. Lung microbiome: new insights into the pathogenesis of respiratory diseases. \u003cem\u003eSignal Transduct. Target. Ther.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 19 (2024).\u003c/li\u003e\n\u003cli\u003eAnani, H. \u003cem\u003eet al.\u003c/em\u003e Lung virome convergence precedes hospital-acquired pneumonia in intubated critically ill patients. \u003cem\u003eCell Rep. Med.\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 102289 (2025).\u003c/li\u003e\n\u003cli\u003eSegal, L. N. \u003cem\u003eet al.\u003c/em\u003e Enrichment of lung microbiome with supraglottic taxa is associated with increased pulmonary inflammation. \u003cem\u003eMicrobiome\u003c/em\u003e \u003cstrong\u003e1\u003c/strong\u003e, 19 (2013).\u003c/li\u003e\n\u003cli\u003eSumner, J. T. \u003cem\u003eet al.\u003c/em\u003e Transitions in lung microbiota landscape associate with distinct patterns of pneumonia progression. \u003cem\u003eCell Host Microbe\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 2148-2166.e8 (2025).\u003c/li\u003e\n\u003cli\u003eIshihara, K. \u0026amp; Hirano, T. IL-6 in autoimmune disease and chronic inflammatory proliferative disease. \u003cem\u003eCytokine Growth Factor Rev.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 357\u0026ndash;368 (2002).\u003c/li\u003e\n\u003cli\u003eRubini, A. Interleukin-6 and lung inflammation: evidence for a causative role in inducing respiratory system resistance increments. \u003cem\u003eInflamm. Allergy Drug Targets\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 315\u0026ndash;321 (2013).\u003c/li\u003e\n\u003cli\u003eStotts, C., Corrales-Medina, V. F. \u0026amp; Rayner, K. J. Pneumonia-Induced Inflammation, Resolution and Cardiovascular Disease: Causes, Consequences and Clinical Opportunities. \u003cem\u003eCirc. Res.\u003c/em\u003e \u003cstrong\u003e132\u003c/strong\u003e, 751\u0026ndash;774 (2023).\u003c/li\u003e\n\u003cli\u003eGuo, Q. \u003cem\u003eet al.\u003c/em\u003e NF-\u0026kappa;B in biology and targeted therapy: new insights and translational implications. \u003cem\u003eSignal Transduct. Target. Ther.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 53 (2024).\u003c/li\u003e\n\u003cli\u003eCao, Z. \u003cem\u003eet al.\u003c/em\u003e The gut ileal mucosal virome is disturbed in patients with Crohn\u0026rsquo;s disease and exacerbates intestinal inflammation in mice. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 1638 (2024).\u003c/li\u003e\n\u003cli\u003eCook, R. A. \u003cem\u003eet al.\u003c/em\u003e Bacteriophage diversity declines with COPD severity in the respiratory microbiome. \u003cem\u003eCell Rep.\u003c/em\u003e \u003cstrong\u003e44\u003c/strong\u003e, 116413 (2025).\u003c/li\u003e\n\u003cli\u003eZinter, M. S. \u003cem\u003eet al.\u003c/em\u003e Pathobiological signatures of dysbiotic lung injury in pediatric patients undergoing stem cell transplantation. \u003cem\u003eNat. Med.\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 1982\u0026ndash;1993 (2024).\u003c/li\u003e\n\u003cli\u003eRitz, N. L. \u003cem\u003eet al.\u003c/em\u003e The gut virome is associated with stress-induced changes in behaviour and immune responses in mice. \u003cem\u003eNat. Microbiol.\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 359\u0026ndash;376 (2024).\u003c/li\u003e\n\u003cli\u003eLangelier, C. \u003cem\u003eet al.\u003c/em\u003e Integrating host response and unbiased microbe detection for lower respiratory tract infection diagnosis in critically ill adults. \u003cem\u003eProc. Natl. Acad. Sci.\u003c/em\u003e \u003cstrong\u003e115\u003c/strong\u003e, (2018).\u003c/li\u003e\n\u003cli\u003eKalantar, K. L. \u003cem\u003eet al.\u003c/em\u003e Integrated host-microbe plasma metagenomics for sepsis diagnosis in a prospective cohort of critically ill adults. \u003cem\u003eNat. Microbiol.\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 1805\u0026ndash;1816 (2022).\u003c/li\u003e\n\u003cli\u003eZinter, M. S. \u003cem\u003eet al.\u003c/em\u003e The pulmonary metatranscriptome prior to pediatric HCT identifies post-HCT lung injury. \u003cem\u003eBlood\u003c/em\u003e \u003cstrong\u003e137\u003c/strong\u003e, 1679\u0026ndash;1689 (2021).\u003c/li\u003e\n\u003cli\u003eKitsios, G. D. \u003cem\u003eet al.\u003c/em\u003e Longitudinal multicompartment characterization of host-microbiota interactions in patients with acute respiratory failure. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 4708 (2024).\u003c/li\u003e\n\u003cli\u003eYang, Y. \u003cem\u003eet al.\u003c/em\u003e Characteristics of gut and lung microbiota in patients with lung masses and their relationship with clinical features. \u003cem\u003eBMC Microbiol.\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 541 (2025).\u003c/li\u003e\n\u003cli\u003eOpron, K. \u003cem\u003eet al.\u003c/em\u003e Loss of Airway Phylogenetic Diversity Is Associated with Clinical and Pathobiological Markers of Disease Development in Chronic Obstructive Pulmonary Disease. \u003cem\u003eAm. J. Respir. Crit. Care Med.\u003c/em\u003e \u003cstrong\u003e210\u003c/strong\u003e, 186\u0026ndash;200 (2024).\u003c/li\u003e\n\u003cli\u003ePapazian, L., Klompas, M. \u0026amp; Luyt, C.-E. Ventilator-associated pneumonia in adults: a narrative review. \u003cem\u003eIntensive Care Med.\u003c/em\u003e \u003cstrong\u003e46\u003c/strong\u003e, 888\u0026ndash;906 (2020).\u003c/li\u003e\n\u003cli\u003eGrosjean, V. \u003cem\u003eet al.\u003c/em\u003e Community-acquired Klebsiella pneumoniae pneumonia in ICU: a multicenter retrospective study. \u003cem\u003eAnn. Intensive Care\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 69 (2024).\u003c/li\u003e\n\u003cli\u003eRoquilly, A. \u003cem\u003eet al.\u003c/em\u003e Pathophysiological role of respiratory dysbiosis in hospital-acquired pneumonia. \u003cem\u003eLancet Respir. Med.\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 710\u0026ndash;720 (2019).\u003c/li\u003e\n\u003cli\u003eARDS Definition Task Force \u003cem\u003eet al.\u003c/em\u003e Acute respiratory distress syndrome: the Berlin Definition. \u003cem\u003eJAMA\u003c/em\u003e \u003cstrong\u003e307\u003c/strong\u003e, 2526\u0026ndash;2533 (2012).\u003c/li\u003e\n\u003cli\u003eLeone, M. \u003cem\u003eet al.\u003c/em\u003e Hospital-acquired pneumonia in ICU. \u003cem\u003eAnaesth. Crit. Care Pain Med.\u003c/em\u003e \u003cstrong\u003e37\u003c/strong\u003e, 83\u0026ndash;98 (2018).\u003c/li\u003e\n\u003cli\u003eLu, J. \u003cem\u003eet al.\u003c/em\u003e Metagenome analysis using the Kraken software suite. \u003cem\u003eNat. Protoc.\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 2815\u0026ndash;2839 (2022).\u003c/li\u003e\n\u003cli\u003eWood, D. E., Lu, J. \u0026amp; Langmead, B. Improved metagenomic analysis with Kraken 2. \u003cem\u003eGenome Biol.\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 257 (2019).\u003c/li\u003e\n\u003cli\u003eNurk, S., Meleshko, D., Korobeynikov, A. \u0026amp; Pevzner, P. A. metaSPAdes: a new versatile metagenomic assembler. \u003cem\u003eGenome Res.\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 824\u0026ndash;834 (2017).\u003c/li\u003e\n\u003cli\u003eFu, L., Niu, B., Zhu, Z., Wu, S. \u0026amp; Li, W. CD-HIT: accelerated for clustering the next-generation sequencing data. \u003cem\u003eBioinforma. Oxf. Engl.\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 3150\u0026ndash;3152 (2012).\u003c/li\u003e\n\u003cli\u003eShang, J., Peng, C., Liao, H., Tang, X. \u0026amp; Sun, Y. PhaBOX: a web server for identifying and characterizing phage contigs in metagenomic data. \u003cem\u003eBioinforma. Adv.\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, vbad101 (2023).\u003c/li\u003e\n\u003cli\u003eJung, Y. \u0026amp; Han, D. BWA-MEME: BWA-MEM emulated with a machine learning approach. \u003cem\u003eBioinforma. Oxf. Engl.\u003c/em\u003e \u003cstrong\u003e38\u003c/strong\u003e, 2404\u0026ndash;2413 (2022).\u003c/li\u003e\n\u003cli\u003eZhang, X. \u0026amp; Jonassen, I. RASflow: an RNA-Seq analysis workflow with Snakemake. \u003cem\u003eBMC Bioinformatics\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 110 (2020).\u003c/li\u003e\n\u003cli\u003eKopylova, E., No\u0026eacute;, L. \u0026amp; Touzet, H. SortMeRNA: fast and accurate filtering of ribosomal RNAs in metatranscriptomic data. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 3211\u0026ndash;3217 (2012).\u003c/li\u003e\n\u003cli\u003eBuchfink, B., Reuter, K. \u0026amp; Drost, H.-G. Sensitive protein alignments at tree-of-life scale using DIAMOND. \u003cem\u003eNat. Methods\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 366\u0026ndash;368 (2021).\u003c/li\u003e\n\u003cli\u003eSeemann, T. Prokka: rapid prokaryotic genome annotation. \u003cem\u003eBioinforma. Oxf. Engl.\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 2068\u0026ndash;2069 (2014).\u003c/li\u003e\n\u003cli\u003eDavis, N. M., Proctor, D. M., Holmes, S. P., Relman, D. A. \u0026amp; Callahan, B. J. Simple statistical identification and removal of contaminant sequences in marker-gene and metagenomics data. \u003cem\u003eMicrobiome\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 226 (2018).\u003c/li\u003e\n\u003cli\u003eBeghini, F. \u003cem\u003eet al.\u003c/em\u003e Integrating taxonomic, functional, and strain-level profiling of diverse microbial communities with bioBakery 3. \u003cem\u003eeLife\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, e65088 (2021).\u003c/li\u003e\n\u003cli\u003eGalperin, M. Y. \u003cem\u003eet al.\u003c/em\u003e COG database update 2024. \u003cem\u003eNucleic Acids Res.\u003c/em\u003e \u003cstrong\u003e53\u003c/strong\u003e, D356\u0026ndash;D363 (2025).\u003c/li\u003e\n\u003cli\u003eVelten, B. \u003cem\u003eet al.\u003c/em\u003e Identifying temporal and spatial patterns of variation from multimodal data using MEFISTO. \u003cem\u003eNat. Methods\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 179\u0026ndash;186 (2022).\u003c/li\u003e\n\u003cli\u003eArgelaguet, R. \u003cem\u003eet al.\u003c/em\u003e MOFA+: a statistical framework for comprehensive integration of multi-modal single-cell data. \u003cem\u003eGenome Biol.\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 111 (2020).\u003c/li\u003e\n\u003cli\u003eLin, H. \u0026amp; Peddada, S. D. Analysis of compositions of microbiomes with bias correction. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 3514 (2020).\u003c/li\u003e\n\u003cli\u003eWirbel, J. \u003cem\u003eet al.\u003c/em\u003e Microbiome meta-analysis and cross-disease comparison enabled by the SIAMCAT machine learning toolbox. \u003cem\u003eGenome Biol.\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 93 (2021).\u003c/li\u003e\n\u003cli\u003ePhillips, N. D., Neth, H., K. Woike, J. \u0026amp; Gaissmaier, W. FFTrees: A toolbox to create, visualize, and evaluate fast-and-frugal decision trees. \u003cem\u003eJudgm. Decis. Mak.\u003c/em\u003e 344\u0026ndash;368 (2017).\u003c/li\u003e\n\u003cli\u003eChen, T. \u0026amp; Guestrin, C. XGBoost: A Scalable Tree Boosting System. https://doi.org/10.48550/ARXIV.1603.02754 (2016) doi:10.48550/ARXIV.1603.02754.\u003c/li\u003e\n\u003cli\u003eNewman, N. K. \u003cem\u003eet al.\u003c/em\u003e Transkingdom Network Analysis (TkNA): a systems framework for inferring causal factors underlying host-microbiota and other multi-omic interactions. \u003cem\u003eNat. Protoc.\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 1750\u0026ndash;1778 (2024).\u003c/li\u003e\n\u003cli\u003eShannon, P. \u003cem\u003eet al.\u003c/em\u003e Cytoscape: A Software Environment for Integrated Models of Biomolecular Interaction Networks. \u003cem\u003eGenome Res.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 2498\u0026ndash;2504 (2003).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":true,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-9114464/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9114464/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAlterations in lung virome and bacteriome compositions have been associated with hospital-acquired pneumonia in critically ill patients, but data on lung microbiome functions underlying progression to severe complications remain unclear. Combining viral-metagenomics with metatranscriptomics analyses of 184 endotracheal aspirates (ETAs) collected in 94 intubated critically ill patients, we identified a high-risk functional pneumotype associated with lung inflammation and mortality. This unfavourable pneumotype was characterised by enrichment in virulent \u003cem\u003eStreptococcus-\u003c/em\u003ebacteriophages and temperate \u003cem\u003eKlebsiella-\u003c/em\u003ebacteriophages, decreased transcriptional activity of commensal taxa such as \u003cem\u003eStreptococcus \u003c/em\u003eand \u003cem\u003eAlloprevotella\u003c/em\u003e, and increased transcriptional activity of \u003cem\u003eKlebsiella pneumoniae\u003c/em\u003e. Machine-learning models defined two-viral- and four-bacterial-factor signatures predicting unfavourable pneumotype (AUC 0.9 and 0.8, respectively). Validation in an independent cohort of 117 ICU patients (239 ETAs) demonstrated the robustness of the association of these signatures with the risk of mortality. Causal inference identified virulent bacteriophages associated with \u003cem\u003eStreptococcus \u003c/em\u003eand \u003cem\u003eAlloprevotella \u003c/em\u003eand host genes including \u003cem\u003eCCL2\u003c/em\u003e, \u003cem\u003ePLGRKT \u003c/em\u003eand \u003cem\u003ePTX3\u003c/em\u003e,\u003cem\u003e \u003c/em\u003eas key regulators of severe lung dysbiosis linked to mortality in critically ill patients.\u003c/p\u003e","manuscriptTitle":"Identification of functional bacterial-viral pneumotypes associated with airway inflammation and all-cause mortality in critically ill patients","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-16 08:14:22","doi":"10.21203/rs.3.rs-9114464/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"9548d6fc-12ec-4532-af6f-3f8af9574bdb","owner":[],"postedDate":"March 16th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-03-16T08:14:23+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-16 08:14:22","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9114464","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9114464","identity":"rs-9114464","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}