The World Health Organization has declared that antimicrobial resistance (AMR) is one of the top global public health threats. One well documented possibility for why an infection would return after proper antibiotic treatment is that the bacteria are becoming inert, and evading antibiotics. Standard antibiotics target active bacteria but are not active on dormant bacterial cells, or ‘persister cells’. Once the danger passes, and the antibiotic course is done, the ‘sleeper’ bacteria can again return to active metabolism and cause infection again, such as with chronic infections. To get technical, these ‘phenotypic variants of wild-type genetically identical bacteria can resume growth again after removing the antibiotic, and give rise to compound-sensitive bacterial cells. ‘ In fact, these persisters may actually promote AMR and facilitate the selection of resistant mutants, so strategies to target persisters is extremely important.
Traditional antibiotics favor active bacterial cells that are metabolizing, by targeting processes that occur in growing cells such as translation, replication, or peptidoglycan synthesis. However, stationary-phase screening methods coupled with deep learning (a type of artificial intelligence) can discover compounds that work against metabolically dormant bacteria. Tales of ‘sleeper’ bacteria are not new- in fact ancient stains have been discovered in this quiescent state on the ocean floor in the South Pacific. Scientists have found aerobic microbial life in marine sediment 101.5 million years old, and when the bacteria were revived and incubated up to 68 days, they could incorporate carbon and nitrogen substrates and start dividing.
Researchers at MIT are making headlines for using AI to discover new classes of antibiotics. In fact, their mission is to discover new antibiotics against seven types of deadly bacteria over seven years. How will they tackle this? They are using the power of AI to figure out how the deep learning models were making their predictions that certain molecules would make good antibiotics. James Collins, the Termeer Professor of Medical Engineering and Science at MIT’s Institute for Medical Engineering and Science (MES) and Dept of Biological Engineering, said their work provides a framework that is ‘time-efficient, resource-efficient; and mechanistically insightful, from a chemical-structure standpoint.’
To figure out how the model was making its predictions, they adapted an algorithm known as the Monte Carlo tree search, and this search algorithm allowed the model to generate not only an estimate of each model’s antimicrobial activity, but also predict which substructure of the molecule accounts for that activity. They are also working on new candidates for methicillin-resistant Staphylococcus aureus (MRSA).
In terms of dormant bacteria, machine learning could help screen compounds that could actually be lethal to these dormant bacteria. Researchers in the Collins lab used AI to speed up the process of finding antibiotic properties in known compounds and found an experimental compound called Semapimod from their high-throughput screening efforts. This compound is an anti-inflammatory used in Crohn’s disease research, but they discovered it could be used for stationary-phase E. coli and Acinetobacter baumannii. It can disrupt these Gram-negative bugs by targeting the lipopolysaccharide component of the outer bacterial membrane.
Other strategies involve looking for antimicrobial agents whose antimicrobial activity is independent of the metabolic status of the bacteria. Since biofilms may harbor dormant subpopulations, anti-persister strategies should include agents to destroy biofilm matrix or inhibit biofilm development such as cell wall hydrolases, polysaccharide depolymerases, and antimicrobial peptides, or combine standard antibiotics with these types of agents to combat persisters.
References
Morono Y et al (2020) Aerobic microbial life persists in oxic marine sediment as old as 101.5 million years. Nat. Commun 11:3626 link.
Ouyang A (2024) When an antibiotic fails: MIT scientists are using AI to target ‘sleeper’ bacteria. MIT News.link.
Persisters: Stojowska-Swedrynska, Kuczynska-Wisnik D and Laskowska E (2023) New strategies to kill metabolically-dormant cells directly bypassing the need for active cellular processes. Antibiotics 12(6):1044 link.