Enhancing pathogen detection and characterization of respiratory microbiome signatures of infection through next-generation sequencing

Abstract

Next-generation sequencing-based (NGS) methods for pathogen detection, which can capture the entire microbial composition in a clinical sample, are used strategically in parallel to culture-based methods. While NGS-based methods cannot confirm organism viability, culture-based methods can but are limited by the number of culture conditions routinely utilized in clinical microbiology laboratories. The relatively wide scope provided by sequencing compensates for the limited scope of culture-based methods but creates obstacles in distinguishing pathogens from an expected body-site-specific microbiome or benign transient colonizers. This study compares NGS-based methods for pathogen detection in respiratory samples to current diagnostics at Texas Children's Hospital in Houston, Texas. Current diagnostics include various culture-based assays, targeted PCR assays, and reference laboratory-generated sequencing. We hypothesize that employing sample metadata (source, type, etc.) and patient data (clinical history, chronic conditions, etc.) during data analysis will help distinguish pathogens from expected flora in respiratory samples. To develop these methods, 27 respiratory samples with suffcient volume for both routine diagnostic testing and this study were collected. Nucleic acid was extracted using the Qiagen DNeasy PowerSoil Pro Kit after collection, and targeted short-read sequencing (16S, 18S/ITS) was performed. These data were then analyzed using a custom bioinformatic pipeline built around Qiime2-2022. In brief, sequences are denoised and quality filtered using DADA2, an algorithm that uses a statistical model for correcting sequencing errors, after which similar amplicon sequence variants (ASVs) are clustered. Next, a feature table of ASVs that are 100% identical is generated, and taxonomy assigned using a naive Bayes taxonomy classifier against the Silva 13.2 database. Finally, unassigned and low abundance ASVs are removed. When species-level identification couldn't be determined through targeted short-read sequencing, untargeted long-read sequencing was used. In one case, we found that a bronchoalveolar lavage sample from the lung of a patient with a history of recurrent pneumonia contained a singular pathogenic bacterium, H. influenzae. This finding agreed with results obtained through reference laboratory-generated sequencing. In another case, we identified >14 unique bacteria at varying concentrations in a wash sample collected from the lung of an asymptomatic patient during a routine bronchoscopy. These findings, however, were determined to be expected respiratory flora. Our NGS-based methods’ enhanced pathogen detection compared to current diagnostics and discriminates between respiratory flora and active infections. Setting thresholds for pathogen detection, evaluating clinical relevance in the context of body site and presentation, and deploying additional molecular techniques when necessary are crucial steps to increase the accuracy of pathogen detection in the clinic. This study was supported by the Texas Children's Hospital Department of Pathology. This is the full abstract presented at the American Physiology Summit 2024 meeting and is only available in HTML format. There are no additional versions or additional content available for this abstract. Physiology was not involved in the peer review process.