Multidimensional high-throughput screening approaches to antimicrobial drug discovery

The rapid rise of antimicrobial resistance (AMR) challenges human health on a global scale. Accumulating resistance mechanisms in an ever-growing complement of clinical pathogens reduce the effectiveness of a limited supply of antimicrobial drugs, increasing patient morbidity and mortality. Concomitantly, the rate of antimicrobial drug discovery has experienced a decline in recent years, further restricting our capacity to manage infectious diseases and exacerbating the AMR crisis. New strategies for drug discovery are required to address these challenges. Exploiting the phenomenon of collateral sensitivity (CS), whereby increased resistance to one drug results in increased sensitivity to another, represents one such strategy. CS drug combination therapies or drug cycling regimens have been shown to successfully curtail the emergence of AMR. Herein this work, we report the design and application of a high-throughput screening platform to identify CS interactions in drug-resistant Escherichia coli. Screening of an 80-member commercial antimicrobial library reproduced several known CS interactions, while screening of a 6,195-member microbial natural product (NP) extract library revealed extensive CS interactions in nature. In pursuit of CS-active natural products, we identified the borrelidin family of macrolides as exhibiting potent activity against cephalosporin-resistant mutants but not the wildtype strain. Co-dosing borrelidin A with ceftazidime suppressed the emergence of ceftazidime resistance below breakpoints over a two-week period and reduced overall cephalosporin resistance by up to 40-fold, compared to controls. In a follow-up study, we reformulate screening analysis to create unique drug resistance profiles for each antimicrobial compound and NP extract. Using hierarchal clustering algorithms and software published earlier by our group, these drug resistance profiles were integrated with their cognate mass spectrometry-based metabolomics data. The resulting similarity networks allowed for the accurate prediction of bioactive species within each extract based on phenotypic profile and by high-resolution mass; we demonstrate the successful dereplication of multiple NPs using this approach. Lastly, we provide a detailed, step-by-step protocol for an optimized, high-throughput antimicrobial susceptibility screen consisting of 20 clinically relevant pathogens. Efficient, flexible, and adherent to international testing standards, this protocol has contributed to numerous publications regarding drug discovery, structure-activity relationship studies, and NP synthesis.