366 - Transcriptomic Profiling Identifies Shared Immunopathogenic Pathways in Children with Bronchiolitis and Community-Acquired Alveolar Pneumonia that Correlate with Clinical Findings

Publication Number: 2355.366

Zhaohui Xu, St. Jude Children's Research Hospital, Memphis, TN, United States; Sara E. Mertz, Nationwide Children's Hospital, Columbus, OH, United States; Hannah Kim, University of Cincinnati College of Medicine, Columbus, OH, United States; Amy L.. Leber, Association for Molecular Pathology and South Central Association for Clinical Microbioloyg, Columbus, OH, United States; Guy Hazan, Ben Gurion University of The Negev, Beer Sheva, HaDarom, Israel; Asuncion Mejias, St. Jude Children's Research Hospital, Memphis, TN, United States; Ron Dagan, Ben-Gurion University, Beer-Sheva, israel, Beer Sheva, Tel Aviv, Israel; Octavio Ramilo, St. Jude Children's Research Hospital, Memphis, TN, United States

Background: Bronchiolitis (BRO) and Community Acquired Alveolar Pneumonia (CAAP) are common lower respiratory infections (LRI), and respiratory viruses including respiratory syncytial virus (RSV), influenza, parainfluenza and human metapneumovirus are well-established etiologies. While both are lung infections, they have distinct clinical characteristics

Objective: Identify transcriptomic profiles of BRO and CAAP to investigate the immunopathogenesis of LRI and the association with clinical findings

Design/Methods: Children < 2yr hospitalized with LRI and healthy controls (HC) were prospectively enrolled in Israel. CAAP was radiographically confirmed per the WHO protocol and BRO was defined as acute respiratory infection with bronchiolitis (dyspnea, wheezing and negative chest radiography). Biofire Filmarray and RT-PCR were used to detect virus and bacteria from nasal swabs (NP). Blood samples collected for transcriptome analysis through RNA sequencing and data analysis performed with R

Results: Overall 75% of patients (n=71) had ≥ 1 virus, RSV was the most frequently detected virus (88%) in viral LRI cases. Comparing RSV-associated LRIs to HC (n=17), more differentially expressed genes (DEGs) were found in CAAP (n=23) than BRO (n=23). However, a direct comparison between these 2 groups revealed only 38 DEGs. An unsupervised machine learning approach identified 4 LRI clusters with distinct gene expression patterns (Fig. 1). Both BRO and CAAP cases were distributed in all 4 clusters, suggesting a continuum of transcriptomic profiles rather than clear separation between 2 groups. We validated these findings by analyzed non-RSV-associated LRIs (BRO=13, CAAP=12) and observed similar transcriptomic profiles. Combining all LRIs (BRO=36, CAAP=35) into these 4 predefined clusters (Fig. 2) revealed that interferon (IFN) responses were present in all clusters. Cluster 4, which included older patients, had elevated inflammatory profiles, pronounced suppression of adaptive immune responses, higher neutrophil percentages, increased CRP concentrations, and prolonged fever duration (Fig. 3). In contrast, cluster 1 exhibited overexpression of neutrophil activation genes and moderate IFN gene expression, but, unexpectedly, reduced expression of other inflammation-related genes. This cluster also showed a lower nasopharyngeal detection rate of S pneumoniae

Conclusion(s): Transcriptomic analysis of children with LRIs revealed a surprising degree of overlap between BRO and CAAP. Unsupervised clustering highlighted shared immunopathogenic pathways in both conditions, and these molecular signatures correlated closely with clinical findings

Figure 1. Transcriptomic profiling of RSV-asscoiated LRI samples.
Heatmaps show the differential gene expression analysis of BRO with HC (A), CAAP with HC (B) and BRO with CAAP (C). Venn diagram (D) shows the DEGs of these 3 pairwise comparisons. Adjusted p value < 0.05, fold change > 1.25. Red: over-expression, blue: under-expression.

Figure 2. Unsupervised clustering of LRI samples.
K-means clustering was applied to group RSV-associated LRI samples into four distinct clusters. Non-RSV-associated LRI samples were subsequently assigned to these clusters based on nearest distance. Heatmap (A) shows 4 clusters of all LRI samples with annotation of clinical defination and detected viruses on top (RSV: respiratory syncytial virus, Flu: influenza, PIV: parainfluenza and hMPV: human metapneumovirus). Quantitative gene set enrichment analysis was performed on LRI samples within each cluster, compared to HC, using a significance threshold of padj < 0.05. Fold changes of significant gene sets are presented (B). Red: over-expression, blue: under-expression.

Figure 3. Clinical and laboratory findings of 4 LRI groups.
Four groups of LRI samples, identified through transcriptomic profiling, were analyzed with their corresponding clinical and laboratory data. Detection rates of Streptococcus pneumoniae (A), age in months (B), total duration of fever (C), C-reactive protein (CRP) concentrations (D), neutrophil percentages (E), and lymphocyte percentages (F) in peripheral blood are presented. Fisher's exact test or Mann-Whitney test, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.