In-silico immune cell deconvolution of the airway proteomes of infants with pneumonia reveals a link between reduced airway eosinophils and an increased risk of mortality

Rationale Pneumonia is a leading cause of mortality in infants and young children. The mechanisms that lead to mortality in these children are poorly understood. Studies of the cellular immunology of the infant airway have traditionally been hindered by the limited sample volumes available from the young, frail children who are admitted to hospital with pneumonia. This is further compounded by the relatively low frequencies of certain immune cell phenotypes that are thought to be critical to the clinical outcome of pneumonia. To address this, we developed a novel in-silico deconvolution method for inferring the frequencies of immune cell phenotypes in the airway of children with different survival outcomes using proteomic data. Methods Using high-resolution mass spectrometry, we identified > 1,000 proteins expressed in the airways of children who were admitted to hospital with clinical pneumonia. 61 of these children were discharged from hospital and survived for more than 365 days after discharge, while 19 died during admission. We used machine learning by random forest to derive protein features that could be used to deconvolve immune cell phenotypes in paediatric airway samples. We applied these phenotype-specific signatures to identify airway-resident immune cell phenotypes that were differentially enriched by survival status and validated the findings using a large retrospective pneumonia cohort. Main Results We identified immune-cell phenotype classification features for 33 immune cell types. Eosinophil-associated features were significantly elevated in airway samples obtained from pneumonia survivors and were downregulated in children who subsequently died. To confirm these results, we analyzed clinical parameters from >10,000 children who had been admitted with pneumonia in the previous 10 years. The results of this retrospective analysis mirrored airway deconvolution data and showed that survivors had significantly elevated eosinophils at admission compared to fatal pneumonia. Conclusions Using a proteomics bioinformatics approach, we identify airway eosinophils as a critical factor for pneumonia survival in infants and young children.