Integrated Host/Microbe Metagenomics Enables Accurate Lower Respiratory Tract Infection Diagnosis in Critically Ill Children

Lower respiratory tract infections (LRTI) are a large source of morbidity and mortality in pediatric patients. Metagenomic next-generation sequencing is an important emerging diagnostic tool that can identify both specific pathogens and host gene expression in response to these infections. However, metagenomic next-generation sequencing hasn’t been applied at scale or to critically ill pediatric populations. This study aimed to use RNA sequencing of tracheal aspirate to contribute to a classifier that integrates host, pathogen, and microbiome features to improve diagnosis and treatment of causal pathogens.

Those enrolled included 261 patients aged 31 days to 18 years with acute respiratory failure requiring mechanical ventilation from 8 hospitals in the United States between February 2015 and December 2017. Patients with tracheostomy and those with “do not resuscitate” orders in place were excluded.

The study consisted of a secondary analysis of a prospective cohort study. Microbiological diagnostic tests were obtained on the study population tracheal aspirate samples, and the results were uploaded to a research database. The patient samples were divided into groups with “definite evidence” of infection (n = 117) and “no evidence” of infection (n = 50) based on clinical, imaging, and laboratory data. The tracheal aspirates then underwent next-generation RNA sequencing. Sequencing reads were aligned to an index of human protein encoding RNA genes, and the most frequently encountered genes were included for analysis to develop a gene expression classifier for LRTI. A second classifier was then developed to integrate the host LRTI probability, abundance of respiratory viruses, and the lung microbiome of pathogenic bacteria and fungi.

The integrated classifier could classify 93% of patients in the “definite evidence” group as LRTI+ and 88% of patients in the “no evidence” group as LRTI-, achieving a median area under the curve of 0.986 (range, 0.953–1.000). In 94 patients with an uncertain diagnosis, the integrated classifier indicated 52% of these cases to be LRTI and nominated likely pathogens in 98% of cases. Among the patients with no evidence of infection, 40% still had potentially pathogenic microbes identified by next-gen sequencing.

Profiling host, pathogen, and microbiome gene expression in response to infection is a validated tool to aid in LRTI diagnosis in children.

Current diagnostic tests for LRTI do not account for the relationship among pathogen, lung microbiome, and host immune response. The classifier produced by this study has important utility in infection classification in immunocompromised hosts as well as in the practice of antimicrobial stewardship. The frequent challenge in treating critically ill children who are suffering from invasive infections often involves a delicate balance surrounding whether the patient’s severe symptoms are due to the underlying organism or the host’s inflammatory response. Metagenomics has the potential to yield valuable clinical insight in such challenging situations.