Implementation and Application of Genomic Association Methods to Clostridium Difficile Toxicity and Clinical Infection Outcomes

Abstract

Clostridium difficile is a major cause of healthcare-associated infections in the United States. A C. difficile infection can lead to a range of outcomes including diarrhea, intensive care unit admission, abdominal surgery, or death. Pathogenesis is mediated by the release of toxin from C. difficile cells growing in the intestines. Some patients are more vulnerable to infection, including those with previous antibiotic exposure and advanced age. Host factors can affect the likelihood of infection but also the severity of infection. Additionally, infection severity can be influenced by the genome of the infecting strain(s). Host-pathogen interactions are extremely complex and very little is known about the interplay between host factors and C. difficile genomic variation with respect to infection likelihood and outcomes. With the recent deluge of whole genome sequencing data, the contribution of bacterial genomic variation to infections can be more comprehensively evaluated than ever before. The work described in this dissertation used two different approaches to test for associations between C. difficile genomic variation and clinically relevant phenotypes.

In the first approach we implemented and applied a novel convergence-based bacterial genome-wide association study (bGWAS) algorithm for quantitative traits. We introduce the algorithm using a set of data generated in silico to realistically model bacterial genome variation and phenotypes under various evolutionary regimes. When the algorithm was applied to C. difficile genomic variants and toxin activity our bGWAS identified known toxin regulatory genes associated with toxin activity, supporting the value of our approach. Besides identifying key cis-regulatory variants in the toxin-producing locus, we observed several associations that connect toxin activity to a complex network of trans-regulatory genes. Many highly associated variants occur in flagellar genes and indicate coregulation of toxicity and motility. We propose new variants associated with toxin activity for future functional validation. This study focused on a complex phenotype, toxin activity, within a highly controlled in vitro system.

We next investigated the impact of bacterial genetic variation on human infections. The increased complexity of this human-pathogen interaction justified a different association approach to better understand the independent contribution of bacterial genomic variation to infection. In a set of clinically derived isolates, we tested for the association between variants in trehalose metabolism operons and infection severity while incorporating and controlling for infection severity-modulating patient characteristics. Trehalose utilization variants were recently proposed to modulate C. difficile infections in a mouse model. Interestingly, we observed that this in vivo result did not translate to our clinical cohort as we found no evidence of an association between any of the trehalose utilization variants and patient infection outcomes. Taken together, these results demonstrate the utility of applying multiple approaches for identifying genomic variants associated with clinical outcomes that account for either bacterial population structure or host factors.