Antibiotic‑resistant bacteria are outpacing our ability to fight them, turning once‑routine infections into life‑threatening challenges. Yet the tools we use to evaluate drugs haven’t kept up. Most tests still judge antibiotics by how well they slow bacterial growth in bulk bacterial populations, not by whether they actually kill the pathogens inside the body. It’s a mismatch that leaves clinicians guessing and too often leads to treatments that fail when patients need them most.
A research team led by Lucas Boeck, MD, from the department of biomedicine at the University of Basel and University Hospital Basel has come up with a new method to predict treatment success. The paper is titled “Large-scale testing of antimicrobial lethality at single-cell resolution predicts mycobacterial infection outcomes,” and was published in Nature Microbiology.
Bacteria can additionally evade antibiotics if in a dormant state, allowing the bacteria to survive treatment and cause another infection later. Choosing drugs that not only stop the infection but also completely kill the bacteria is important, especially in cases like tuberculosis, which can require many months of treatment. This method, called Antimicrobial Single-Cell Testing (ASCT), is “a large-scale live-cell imaging approach that quantifies bacterial killing in real-time at single-cell resolution,” the authors wrote.
“In ASCT, bacteria are dispensed into [1,536] multiwell plates, immobilized in agar pads containing the viability dye propidium iodide (PI), and exposed to drugs,” explained the authors, continuing, “High-content live-cell imaging captures brightfield and fluorescence images of over 10,000 fields every two to four hours for up to seven days, generating up to one million images per experiment. Time-lapse images of every field are sequentially analyzed using sparse and low-rank decomposition to correct for background fluorescence, supervised random forest classifiers for bacterial segmentation and viability classification, drift correction, and single-cell tracking based on object position and homology. To analyze time–kill kinetics at the population level, single-cell data are pooled, and overall killing is quantified using the area under the time–kill curve.”
The team tracked over 140 million mycobacteria and analyzed approximately 20,000 time–kill curves, allowing them to identify key determinants of antibiotic killing, as well as its clinical relevance. “We use it to film each individual bacterium over several days and observe whether and how quickly a drug actually kills it,” explained Boeck.
Using ASCT, the researchers wanted to analyze its potential to identify more effective tuberculosis treatments. By testing 65 drug regimens in two strains of Mycobacterium tuberculosis, in nutrient-rich and starvation conditions, the team showed that different drug regimens were more effective given these respective conditions. “The better bacteria tolerate an antibiotic, the lower the chances of therapeutic success are for the patients,” added Boeck.
The team found that in M. tuberculosis, drug-specific killing dynamics in starved bacteria better predicted regimen efficacy in mice and humans, as compared to growth inhibition or killing of growing cells. They also showed that antibiotic killing is a “genetically encoded bacterial trait,” after applying this approach to 405 bacterial strains in Mycobacterium abcessus.
While homogenous and heterogeneous bacterial growth can also be quantified by ASCT, the authors reported that small well size and overlapping bacteria are constraints.
The team hopes for future uses to be applied in the clinic and industry. “Our test method allows us to tailor antibiotic therapies specifically to the bacterial strains in individual patients,” said Boeck, adding, “Last but not least, the data can help researchers to better understand the survival strategies of pathogens and thus lay the foundation for new, more effective therapeutic approaches.”
