The clinical course of infection due to respiratory viruses such as Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV2), the causative agent of Coronavirus Disease 2019 (COVID-19) is thought to be influenced by the community of organisms that colonizes the upper respiratory tract, the oropharyngeal microbiome. In this study, we examined the oropharyngeal microbiome of suspected COVID-19 patients presenting to the Emergency Department and an inpatient COVID-19 unit with symptoms of acute COVID-19. Of 115 enrolled patients, 74 were confirmed COVID-19+ and 50 had symptom duration of 14 days or less; 38 acute COVID-19+ patients (76%) went on to require respiratory support. Although no microbiome features were found to be significantly different between COVID-19+ and COVID-19-patients, when we conducted random forest classification modeling (RFC) to predict the need of respiratory support for the COVID-19+ patients our analysis identified a subset of organisms and metabolic pathways whose relative abundance, when combined with clinical factors (such as age and Body Mass Index), was highly predictive of the need for respiratory support (F1 score 0.857). Microbiome Multivariable Association with Linear Models (MaAsLin2) analysis was then applied to the features identified as predicative of the need for respiratory support by the RFC. This analysis revealed reduced abundance of Prevotella salivae and metabolic pathways associated with lipopolysaccharide and mycolic acid biosynthesis to be the strongest predictors of patients requiring respiratory support. These findings suggest that composition of the oropharyngeal microbiome in COVID-19 may play a role in determining who will suffer from severe disease manifestations.
Importance: The microbial community that colonizes the upper airway, the oropharyngeal microbiome, has the potential to affect how patients respond to respiratory viruses such as SARS-CoV2, the causative agent of COVID-19. In this study, we investigated the oropharyngeal microbiome of COVID-19 patients using high throughput DNA sequencing performed on oral swabs. We combined patient characteristics available at intake such as medical comorbidities and age, with measured abundance of bacterial species and metabolic pathways and then trained a machine learning model to determine what features are predicative of patients needing respiratory support in the form of supplemental oxygen or mechanical ventilation. We found that decreased abundance of some bacterial species and increased abundance of pathways associated bacterial products biosynthesis was highly predictive of needing respiratory support. This suggests that the oropharyngeal microbiome affects disease course in COVID-19 and could be targeted for diagnostic purposes to determine who may need oxygen, or therapeutic purposes such as probiotics to prevent severe COVID-19 disease manifestations.