NEW YORK (GenomeWeb) – Scientists from Japan's Riken have used a combination of phenotypic analysis and gene expression profiling to quantitatively predict antibiotic resistance in laboratory-evolved strains of Escherichia coli.
The study, published yesterday in Nature Communications, provided novel insights into the mechanism of action of some well-known antibiotic resistance-related genes and implicated some less-well-characterized genes in the development of resistance. In addition, the research has paved the way for the potential prevention of antibiotic resistance through the inhibition of phenotypic changes, the study's authors noted.
The threat of multi-drug-resistant bacteria to public health is well-documented, but the extent of this problem may still be understated, as bacteria are increasingly developing resistance even to antibiotics they have never encountered before, the Riken team noted in a statement coinciding with the publication of their paper.
In addition, the relationship between a mutation and drug resistance is not always a simple one-to-one correspondence, with multiple mutations often needed to confer high levels of resistance to a particular drug. Meantime, single mutations can cause phenotypic changes that affect the resistance and susceptibility to various drugs simultaneously, the researchers wrote.
In order to provide deeper insight into these conundrums, the Riken researchers employed laboratory evolution of bacteria, in which bacterial cells are exposed to fixed drug concentrations, resulting in partial or complete inhibition of cell growth but allowing for the selection and propagation of antibiotic-resistant strains.
Specifically, they laboratory-evolved E. coli using 11 antibiotics covering a wide range of action mechanisms such as disruption of cell wall synthesis, protein synthesis, folic acid biosynthesis, and DNA replication, establishing minimum inhibitory concentration (MIC) values of each antibiotic as an indicator of whether bacteria had established resistance. For all resistant strains, they confirmed resistance for at least 30 generations in the absence of the drug.
Then, in order to better understand the relationships between various antibiotic-resistance profiles, they measured MICs for 25 different antibiotics in each of the bacterial strains with laboratory-evolved resistance. This work revealed that the antibiotic-resistant strains generally exhibited significant changes in the MICs of multiple drugs as compared to control strains, and also demonstrated well-known phenomena such as cross-resistance to drugs with the same mechanism of action, and something called hyper-susceptibility, in which resistant strains to a drug become more susceptible than the parent strain.
The results of this initial analysis suggested that the phenotypic changes that occurred in resistant strains were not always restricted to specific factors, such as modification of the drug target protein structure, but instead were caused by changes in several intra-cellular properties. As such, the researchers hypothesized that these phenotypic changes were represented by changes in gene expression profiles — information they then tried to extract from transcriptome data obtained via microarray analysis.
Specifically, they used custom-designed Agilent 8 X 60K arrays for E. coli W3110, in which 12 probes were prepared for each gene analyzed. In addition, they constructed a simple mathematical model to predict resistances using the obtained gene expression profiles and showed that they could use the linear model to not only discriminate resistant from non-resistant strains but also quantitatively predict the resistances of non-resistant strains by only using the cross-resistance and hyper-susceptibility data.
Their analyses identified several genes that were frequently selected by the genetic algorithm trials, including several well-known resistance-related genes such as acrB and ompF. However, these trials also revealed several less-characterized genes. For instance, the expression of oppA — which encodes for a peptide-binding protein that is an essential component of the oligopeptide transporter — was generally upregulated with quinolones and beta-lactams. The researchers noted that some previous studies implied that the deletion of oppA contributes to aminoglycoside resistance.
In addition, several less-characterized genes that have never been reported as related to resistance, such as yhfL and yijD, were also suggested to relate to the resistance acquisitions — targets for future studies, the Riken team noted.
Finally, the researchers used high-throughput sequencing on a Roche FLX+ and Illumina HiSeq instrument to analyze genomic DNA samples from the antibiotic-resistant strains in an attempt to identify "fixed" mutations. Using Sanger sequencing as a confirmatory procedure, they uncovered fewer than 20 such fixed mutations, suggesting that these mutations contribute to the resistance acquisition of a given class of antibiotics.
"Our experimental data indicated that, although different mutations were fixed in resistant strains to the same drug, the expression changes among these strains were similar, suggesting that different mutations can cause a similar antibiotic resistance through common expression changes," the researchers concluded in their paper. "We suspect that investigating common resistance acquisition mechanisms, such as [horizontal gene transfer], will show similar expression changes for the resistance. Thus, it is important to compare precise phenotype–genotype comparisons when investigating common resistance acquisition mechanisms."
Further elaborating on this, in a statement, Riken researcher and study co-author Chikara Furusawa said that "by making it possible to quantitatively determine what genes contribute to the development of antibiotic resistance, this research could lead to new methods for blocking acquisition of resistance and to the development of new antibiotic compounds."