One necessary step to fend off a dystopian future of medical care without antibiotics is the development of new antibiotics. 

Another is improved deployment of existing ones, a feat which will take, among other things, better antibiotic susceptibility testing (AST). 

“I’m astounded that we can get men to the moon, and we are using practices [dating] almost back to the age of Robert Koch to identify bacteria,” Deborah Hung told BioWorld. “The standard practice takes amazingly long.” 

Hung is a core member of the Broad Institute of Harvard and MIT as well as a practicing infectious disease clinician. In the Nov. 25, 2019, online issue of Nature Medicine, she and her colleagues reported on an RNA-based method that was able to determine antibiotic susceptibility of bacteria within four hours. 

Current standard AST methods take 36 hours or longer to achieve the same feat.  

Those 36 hours are not just a problem in the clinic, where getting the right antibiotic early on can send a patient home sooner, while getting the wrong one can send them to intensive care. They are also yet another hurdle to the development of new antibiotics.   

“It’s extremely difficult to run clinical trials on new antibiotics,” Hung said. In order to test whether the experimental drug is working without confounders, patients ideally need to be enrolled in a trial immediately, and at the very least “before they’ve gotten more than 24 hours of a different antibiotic.” 

That timeline is often challenging to meet while doing justice to the patient’s needs. “Our hope and expectation is that this would make the path to enrollment… much more feasible,” Hung said. 

Clinical isolate sequencing to look for resistance genes is one possible solution, but it has its own drawbacks, at least as a standalone method.  

For one, Hung said, “genotypic causes of antibiotic resistance are predicated on the idea that we have a complete database” of resistance mutations, which is simply not the case. 

Even known resistance mutations can mislead. Bacteria can carry resistance genes that they don’t express due to other alterations in their genomes, making them a sheep in wolf’s clothing. 

For their method, which they have named GoPhAST-R for “genotypic and phenotypic AST through RNA detection,” the researchers took advantage of the fact that bacteria react to antibiotics much more rapidly than is discernible from their growth rate, which is the readout of current AST methods.  

When bacteria are exposed to antibiotics they are susceptible to, “bugs will feel the stress of that antibiotic and change their transcription profile within five to 10 minutes,” Hung said. “Resistant bacteria don’t have that response, because they don’t feel that stress.” 

The team used transcription-level changes rather than growth in culture as their phenotypic readout, combined with sequencing of key resistance genes as a genotypic adjunct to improve accuracy. 

Bioinformatics methods enabled them to identify 10 key phenotypic transcripts, and to translate the changes they observed into the final yes or no classification of whether a bacterium was susceptible or resistant to a given antibiotic. 

Testing more than 100 bug/drug combinations, the team found that their approach was as good as or better than standard methods for predicting meropenem, ciprofloxacin, levofloxacin or gentamicin susceptibility of susceptible and resistant strains of Klebsiella pneumoniae, Escherichia coli, Acinetobacter baumannii, Staphylococcus aureus, and Pseudomonas aeruginosa.  

The work reported by Hung and her colleagues has been licensed to molecular diagnostics and analysis company Nanostring Technologies Inc., in “the first important step,” Hung said, to bringing the technology into clinical use. The company is developing the technology into a next-generation platform. 

Hung and her team, too, are working on improving the method. Some resistance mechanisms do not kick in until late, and “because we are determining antibiotic susceptibility very early, if there are strains that only develop resistance late, we won’t pick that up,” Hung said, though the resistance gene sequencing that is part of the method catches some such late resistance genes. 

Likewise, polymicrobial infections “could get complicated,” she said. “We are trying to develop methods to work around that.” 

The team is also trying to get rid of the culturing step altogether. Currently, the clinical isolates still need to be cultured in order to test them with Go-PhAST-R.   

But to identify and test bacteria “directly from blood without culture,” Hung said, “that’s kind of the holy grail.”