Skip to main content

NGS Can Aid in Detecting Drug Resistance in Tuberculosis

NEW YORK (GenomeWeb) – Two commercially available sequencing platforms are capable of producing drug-resistance profiles for Mycobacterium tuberculosis, according to a new Genome Medicine study.

There has been a rise in recent years in the incidence of multidrug-resistant and extensively drug-resistant tuberculosis, but gauging the resistance profile of M. tuberculosis can be a slow and expensive process.

In the new study, researchers from the London School of Hygiene & Tropical Medicine examined how well two next-generation sequencing approaches and two analytical pipelines could detect drug resistance in a resistant reference strain of tuberculosis as well as in a set of resistant clinical samples. In their assessment, the researchers found that the Illumina MiSeq and Thermo Fisher Scientific Ion PGM sequencing platforms gave comparable results, while the TBProfiler bioinformatics pipeline had a higher concordance with phenotypic susceptibility than did the Mykrobe predictor pipeline.

"Our work suggests that [sequencing platforms] are capable of delivering high-quality data regarding resistance to anti-TB drugs…" senior author Taane Clark from the London School of Hygiene & Tropical Medicine and his colleagues wrote in their paper. "It is expected that drug-resistance profiling using next-generation sequencing will gain accuracy and reliability with the gathering of improved knowledge of the drug-target genes and resistance-causing mutations."

For their analysis, Clark and his colleagues used clinical isolates of M. tuberculosis collected from 10 patients with known drug-resistant TB and the reference TB strain H37Rv. They sequenced the 10 clinical samples in triplicate and the reference sample once on the Illumina MiSeq platform. Four samples were also sequenced six times as technical replicates. At the same time, they sequenced duplicate DNA samples from three clinical isolates on the Ion PGM.

They reported a median 1.2 million reads and a median depth of coverage of fifty-one fold for samples sequenced on the MiSeq, while the Ion PGM samples had a median 990,854 reads and a depth of coverage of fifty-three fold. About a quarter of the M. tuberculosis genome had low coverage in the Ion PGM samples, which the researchers suspected was due to its high numbers of guanine and cytosine bases.

The coverage across the three dozen candidate drug resistance genes was high for the MiSeq samples, the researchers reported, and, for the three samples that were sequenced by both the MiSeq and PGM, the researchers noted greater variability in candidate genes among the Ion PGM samples, again likely due to GC content.

Processing the raw reads with Samtools and GATK initially gave diverse outputs, but the researchers noted that filtering them for coverage and allelic frequency effects led them to converge on similar lists of SNPs that likely cause resistance.

Clark and his colleagues also tested how well two rapid bioinformatics tools, TBProfiler and Mykrobe predictor, could gauge resistance from the raw sequencing reads. Using the MiSeq data, the TBProfiler pipeline exhibited high concordance, 95.3 percent, with results of phenotypic susceptibility testing.

However, there were discordant results where the samples were phenotypically resistant to pyrazinamide or para-aminosalicylic acid, but TBProfiler didn't pick it up. The researchers traced this discordance to novel mutations in known candidate resistance genes, including a double dfrA-thyA deletion in a PAS-resistant sample.

That finding, the researchers said, "highlights the need to look at non-SNP variants." When these novel variants aren’t included, they said the TBProfiler is fully concordant with the phenotypic assessment.

The Mykrobe predictor, meanwhile, showed 73.6 percent concordance with phenotypic testing. Of the 72 resistance calls it made, 19 were incorrect "susceptible" calls, the researchers reported.

"Ultimately, drug resistance profiling using next-generation sequencing offers rapid assessment of resistance-associated mutations, thus accelerating access to effective treatment," Clark and his colleagues wrote in their paper.