NEW YORK (GenomeWeb) – A retrospective study published online today in Lancet Infectious Diseases suggests mutation patterns found by sequencing many Mycobacterium tuberculosis isolates can provide a framework for predicting drug resistance or susceptibility.
An international team sequenced the genomes of almost 2,100 M. tuberculosis isolates collected in the UK, Sierra Leone, and South Africa between the fall of 2010 and late 2013, uncovering a set of mutations that coincided with drug resistance phenotypes reported in the past.
The team used these mutations to make drug response predictions in 1,552 more M. tuberculosis isolates from Germany, South Africa, and Uzbekistan, accurately ascertaining drug susceptibility or resistance more than 89 percent of the time with 92.3 percent sensitivity and 98.4 percent specificity.
"We have established a kind of dictionary for mutations in the genomes of TB pathogens," co-corresponding author Stefan Niemann, a researcher affiliated with the Research Center Borstel and the German Center for Infection Research, said in a statement. "If changes to the genetic code are found in a patient isolate, then certain medications are no longer effective and should therefore not be used for treatment."
Niemann and his colleagues used Illumina platforms to perform paired-end sequencing on 2,099 isolates from a M. tuberculosis training set collected in three countries where parasites from diverse M. tuberculosis clades are present. More than 18 percent of isolates were resistant to at least one tuberculosis drug, while 91 isolates — 4.3 percent — were multidrug resistant and 0.2 percent were extensively drug resistant.
The team then analyzed these genome sequences, paying particular attention to benign, resistance-related, or uncharacterized alterations in the coding and promoter sequences for 23 genes implicated in tuberculosis drug resistance in the past.
From the nearly 9,000 drug susceptibility-related features and 701 different resistance phenotypes in the pathogens detected in the parasites, the researchers began teasing apart specific mutations that appeared to be most informative.
After tossing out variations not related to susceptibility or resistance — such as SNPs associated with population structure — they were left with 120 key resistance-related mutations falling in or around 14 genes.
The team uncovered resistance-related alterations from this 120-mutation set in nearly 94 percent of the 701 resistant isolates sequenced initially.
In a validation set comprised of 1,552 isolates from Germany, Uzbekistan, and South Africa, meanwhile, the researchers found that their mutation-based classification scheme could distinguish between susceptible and resistant strains roughly 89 percent of the time.
Less than 8 percent of phenotypic features in that isolate collection were classified as susceptible when they were actually resistance related, the study's author explained, while 1.6 percent of susceptible features were incorrectly deemed resistance-related.
Across the full set of isolates in the training and validation set, the team uncovered another 112 mutations with apparent ties to drug resistance and hundreds of new susceptibility-related variations, which appeared to boost the sensitivity and specificity of the mutation-based classification method to around 95 percent and 98 percent, respectively.
Those involved in the study noted that whole-genome sequencing is being tested by Public Health England to help diagnose and treat mycobacterial infections. As such, they explained, "[p]arallel phenotypic drug susceptibility testing will lend support to the status of some mutations, and characterize further ones."
"The cosmopolitan nature of tuberculosis in the UK will enhance our understanding of molecular determinants of resistance," they concluded, "as will the global accumulation of data from whole-genome sequencing."