Rapid assessment of bacterial virulence for personalized treatment: using bacterial genome wide and mass spectrum wide association studies of invasive and non-invasive bacteria to identify biomarkers linked with sepsis

Background: The course of infections is heterogenous and determined by host/pathogen interactions and time to treatment. In today’s clinical practice, patient-centric factors assess the outcome and thereby influence empiric antibiotics. For decades, antimicrobial resistance (AMR) profiling was the key for subsequent treatment adaptations. Why is the pathogen’s virulence largely ignored? Bacterial virulence factors are as important as AMR and contribute to clinical phenotypes and pathogen invasiveness. Virulence is variably expressed, similar to AMR, and highly diverse across and within bacterial species. In clinical practise, knowledge about virulence is surprisingly scars and not readily available. Few exceptions show the impact on clinical and epidemiological management, e.g., toxins of Clostridioides difficile or enterohemorrhagic Escherichia coli (EHEC). There is clearly a missed opportunity for virulence assessment in current diagnostic practice, which may result in ground breaking changes of clinical management and antibiotic stewardship. But how can we unfold this potential?Objectives: We will develop a diagnostic approach for early bacterial virulence risk assessment using (i) fundamental knowledge on virulence in clinically relevant scenarios (WP1-4), (ii) experimentally validate identified bacterial biomarkers (WP5), monitor markers during infection (WP6), and (iii) translate findings into clinical diagnostic practice (WP7). Our driving hypothesis is that, invasiveness can be determined early.Methods: First, we screen for potential virulence factors linked to invasiveness of primary sterile sites. We will adapt our machine learning (ML) algorithm for AMR (Weis C et al. Nature Med. 2022) and predict invasiveness from an unprecedented large, MALDI-TOF spectral datasets covering diverse species from an established collaborative laboratory network across 7 countries (WP1). Second, we confirm virulence factors in relevant Gram-negative and -positive bacterial species (ESKAPE): performing bacterial genome wide (bGWAS, WP2) and mass spectrum wide association studies (MSWAS, WP3) from in house and public-available data of invasive and non-invasive ESKAPE isolates. Next, we will adjust the model for patient factors e.g., age, immunosuppression, surgery, etc. (WP4). We will genetically modify wildtype bacteria, using promising candidates from WP1-4, and make isogenic isolates with and without virulence factors (Gibson assembly) for validation of ML algorithms (WP5). Next, we will monitor the expression profiles of biomarkers during clinically relevant infections and we will expand routine MALDI-TOF MS profiling with lipid profiling (negative ion mode and super-DHB matrix) and use Fourier Transform Infrared spectroscopy (WP6). Finally, we will conduct a retrospective quasi study to determine the impact on clinical management (WP7).Impact: The proposal may result in using bacterial virulence for early personalized evaluation and prediction of outcomes. Thereby, (i) monitoring of patients may be adapted e.g., with additional follow-ups or earlier triaging; and (ii) treatments may be tailored. We generate (iii) a substantial gain in knowledge about virulence in diagnostics with potential for research and the diagnostic industry; (iv) development of bioinformatic tools focusing on virulence, which can be translated to other microorganisms; a list of (v) clinically validated diagnostic biomarkers; (vi) potential “drug-able" targets for anti-virulence treatments and vaccines; (vii) highly characterized biobanks and datasets, which will be made available for the scientific community (FAIR data/sample sharing concepts) as a reference toolbox on virulence, and (viii) a strong translational use case with model character for virulence related research, which could complement other phenotypes, such as AMR.