Emergence of Known and Novel Healthcare-Associated Pathogens in the Gut Microbiome after Antibiotic Treatment
Mentor: Dr. Zhihua Li, The Food and Drug Administration (CDER/OMPT/CDER/OTS/OCP/DARS)
Date/Time: May 13, 2019 at 1:30pm
Location: Room 1300, Harris Building
Oral antibiotics can have a devastating impact on the gut microbiome. Changes in normal gut flora can lead to the proliferation of antibiotic resistance and opportunistic pathogens. Many of these pathogens are the cause of healthcare-associated infections (HAIs). While HAIs are very common, the emergence of these pathogens is not well understood. In an ongoing research project in the Division of Applied Regulatory Science at the FDA, we took a metagenomics approach to characterize the changing taxa in the guts of mice who were inoculated with a uropathogenic strain of E. Coli and treated with antibiotics commonly used for urinary tract infections (ciprofloxacin, fosfomycin, or ampicillin). Previously in this project, we focused on informatic tools that aligned our short shotgun reads against marker databases for different taxa and antibiotic resistance genes. Marker databases contain short and unique sequences that serve as indicators for a certain taxa or gene. While this approach is fast and not that computationally intensive, low abundance and divergent microbes are often not detected. This can leave many reads uncharacterized. Our previous experiment was not able to assign many reads to taxa at the species level, presumably because these reads came from microbes that are too divergent from known microorganisms already characterized in any databases. Recently, Kowarsky et al. demonstrated that assembling short reads into longer continuous and semi-continuous regions (contigs and scaffolds) can help to characterize reads with better definition and even assign previously “unmappable” reads to a taxa or genomic feature (Kowarsky, PNAS 2017). While they were able to use this method to uncover numerous uncharacterized and highly divergent microbes from human plasma, no bioinformatics pipeline has been established to apply this method to the gut microbiome, which is more divergent and complex than the cell-free DNA-derived circulating microbiome.
In this project, we developed and fine-tuned a bioinformatics pipeline to apply this method to characterize previously unmappable reads from the gut microbiome following antibiotic treatment. We retroactively assembled the reads in samples from the aforementioned mouse study into scaffolds using the metagenomic alignment tool, MetaSpades. Using the metagenomics gene prediction tool, MetaGene, we found “high complexity” scaffolds, or scaffolds with high gene density. Next, scaffolds were then classified as known, divergent, or novel using cutoffs based on their percent nucleotide and protein identity to existing sequences in the Refseq and NCBI non-redundant databases. Scaffolds showing relatively low nucleotide identity across the entire sequence and low protein identity within their predicted genes will be classified as novel. Scaffolds with very high nucleotide identity are considered known and divergent scaffolds will fall between the novel and known groups. From the protein alignments, we created a python script that can assign taxonomy based on the lowest common ancestor for all gene hits on a scaffold. We then looked at the most commonly assigned species within each antibiotic cohort and found that each antibiotic lead to the enrichment of different opportunistic pathogens. In this study, we were the first to show the proliferation of certain species as a direct result of antibiotic treatment. This work highlighted the importance of studying the emergence of HAI pathogens within a controlled laboratory setting.