Researchers have discovered that bacteria that are resistant to antibiotics tend to be shorter and rounder compared to their antibiotic-sensitive counterparts. These physical differences appear to be linked to changes in the activity of genes involved in energy metabolism and resistance to antibiotics. A machine learning technique has successfully differentiated between antibiotic-resistant and antibiotic-sensitive bacteria using microscope images without any drug treatment, indicating that bioinformatics could play a role in identifying antibiotic resistance in patient samples.
Upon its discovery, penicillin was celebrated as “the silver bullet” because of its unique ability to eliminate disease-causing bacteria without causing harm to humans. Since then, numerous other antibiotics have been developed targeting various kinds of bacteria. However, frequent use of these antibiotics increases the risk of creating antibiotic-resistant strains.
A recent study published in Frontiers in Microbiology by researchers from Osaka University has uncovered distinctive differences in the shape of bacteria that have developed resistance to antibiotics.
Antibiotic resistance has become a significant global health issue, leading to fewer options for effectively treating bacterial infections. Rapid identification of antibiotic-resistant bacteria is critical to ensuring patients receive the appropriate treatment. However, the standard method for identifying resistant strains involves several days of growing the bacteria in a laboratory and testing their reactions to drugs.
“There is some evidence suggesting that antibiotic resistance manifests in different ways; for instance, the morphology of Gram-negative rod-shaped bacteria alters when treated with antibiotics,” states Miki Ikebe, the lead author of the study. “We wanted to find out if these morphological characteristics could potentially be used to identify antibiotic resistance without actually exposing the bacteria to antibiotics.”
To investigate this, the researchers exposed Escherichia coli to controlled amounts of various antibiotics, leading them to develop resistance. They then removed the antibiotics and utilized machine learning to analyze the shapes, sizes, and other characteristics of the bacteria through microscope images.
“The findings were very clear,” explains Kunihiko Nishino, the senior author. “The antibiotic-resistant strains were either bulkier or shorter than their parental strains, particularly those that were resistant to quinolone and β-lactams.”
Furthermore, the researchers delved into the genetic characteristics of the antibiotic-resistant bacteria to ascertain any links between their shape and resistance. Their findings indicated that genes associated with energy metabolism and antibiotic resistance indeed correlated with the observed changes in bacterial shape.
“Our results demonstrate that drug-resistant bacteria can be detected from microscope images, without using antibiotics, through machine learning,” states Ikebe.
Given that bacteria resistant to quinolones, β-lactams, and chloramphenicol showed similar shapes and sizes, it suggests that a common genetic mechanism could underlie antibiotic resistance across these strains. In the future, machine learning tools may be employed to quickly analyze samples from patients, aiding in the selection of the most appropriate drug for treating their infections.
