Sepsis is a life-threatening consequence of the dysregulated host response to infection and is characterized by profound physiologic derangements. In dogs, sepsis is associated with mortality rates of 20-68%, with increased mortality in those with organ dysfunction. Early therapeutic intervention is crucial because delayed definitive care increases mortality, and correct diagnosis is essential to avoid unnecessary antimicrobial drug administration.

Omics technologies are uniquely able to generate large-scale unbiased data from interacting biological systems, making them attractive ways to study complex disease states. Metabolomics comprehensively evaluates the small molecule compounds that are the end-products of cell metabolism and is a means to probe and quantify metabolic disturbances in sepsis. We previously observed numerous metabolic pathway alterations in dogs with sepsis compared to healthy controls and identified potential diagnostic and prognostic markers. However, we do not know if these metabolic disturbances are due to sepsis or are common to other critical illnesses. This has implications for use of metabolic profiling for sepsis diagnosis. This is the knowledge gap we intend to address. We hypothesize that dogs with bacterial sepsis have distinct metabolomic profiles compared to dogs with non-septic critical illness and that the relative abundance of previously identified biomarkers is discriminating.

To test these hypotheses, we will leverage an existing, novel stored sample set (n=55) derived from a previous CVM resident research project. Plasma samples from 20 dogs with bacterial sepsis with the highest illness severity scores will be selected and matched by illness severity with 20 dogs with non-septic critical illness. These 40 samples will be submitted to the Cornell Biotechnology Resource Center for untargeted metabolomics. After data curation, plasma metabolomes will be analyzed with Compound Discoverer 3.3 to identify and annotate discovered compounds. Subsequently, orthogonal partial least-squares discriminant analysis (OPLSDA) will be used to construct score plots and determine how distinct the metabolomes of the two populations are. Metaboanalyst 6.0 will be used to perform differential analyses and generate volcano plots, to construct heatmaps with hierarchical clustering analysis and visualize altered metabolic pathways. Univariate and multivariate receiver operator characteristic (ROC) curve analysis will be used for biomarker evaluation.

Sample analyses for metabolomics are expected to take 3 months, 2 months are allotted to initial metabolomic data analyses with a further 3 months allowed for follow-up compound and pathway assessments. The remaining 4 months are intended to be used for manuscript preparation to complete the project within 1 year.