MSACL 2023 Abstract

Introduction:
There are more than 500,000 bloodstream infections (BSI) and 90,000 deaths each year in North America alone. Virulence factors produced by BSI pathogens are known to have a direct impact on the clinical trajectory of infections. Despite this, there are no diagnostic tests for microbial virulence, nor any way of integrating the risk profiles of microbes into clinical decision-making. To address this shortcoming, our group has embarked on a systematic multi-omics survey of a 16-year cohort of BSI microbes collected from the greater Calgary metropolitan area. We are integrating this data with comprehensive medical records to systematically map the relationships between microbial biochemical traits and the severity of infections. Given that most virulence factors are proteins, our quantitative LC-MS proteomic data is playing a pivotal role in identifying new protein-based targets for virulence diagnostics.

Objectives:
The primary objective of our Precision Infection Management (PIM) initiative is to identify microbial proteins or metabolites that are linked to poor patient outcomes and develop LC-MS methods that can be translated into clinical reference labs to enable virulence testing.

Methods:
We have collected 34,000 isolates from every Calgary-area BSI between 2006–2022. Each isolate was grown under standardized conditions and whole-genome sequencing, metabolomics, and proteomics data were collected. For the latter, isolates were cultured in 96-well format in protein-free media and lysed. Proteins were captured on carboxylated Sera-Mag Speed Beads and trypsinized using a modified SP3 protocol [1]. Peptides were TMT-11plex-labelled, pooled, quantified, and identified using LC-MS/MS (Thermo Fusion Lumos Tribrid MS equipped with FAIMS). Data were processed using an automated data analysis pipeline built on RawTools and MaxQuant (https://proteomics.resistancedb.org/). Quantitative proteomics data were then integrated with comprehensive medical histories from each patient and computational strategies (including GWAS and machine learning) were used to map putative virulence traits.

[1] Hughes C., Nature Protocols 14, 68–85 (2019)

Results:
To date, we have collected 16,000 quantitative proteomics profiles covering every infection with Staphylococcus aureus, Escherichia coli, Klebsiella species (pneumoniae and oxytoca), Enterococcus species (faecalis and faecium) in our cohort of isolates. Using these data, we have shown that proteomics profiles can 1) reproduce genomic phylogenetic classifications, 2) identify individual strains of isolates, and 3) identify isolates carrying known resistance or virulence traits. In addition, our machine learning strategies have identified a large cohort of proteins whose expression levels are closely tied to 30-day survival among patients. Moreover, we have identified specific molecular traits that are most commonly found in specific patient cohorts, including individuals suffering from substance abuse disorders and patients with implanted medical devices.

Conclusion:
Our data show that quantitative proteomics profiles are a robust tool for monitoring microbial populations at the molecular level and can be used to identify specific proteins that are predictive of negative clinical outcomes.