Whole Genome Sequencing Use in Predicting TB Resistance

Whole genome sequencing (WGS) of Mycobacterium tuberculosis isolates has great potential for identifying complete drug susceptibility profiles, which could enable precise and individualized treatment early in the course of tuberculosis (TB) infection. Traditional phenotypic drug-susceptibility testing takes many weeks and requires expensive lab resources, which are often inadequate in settings with the highest TB burden.

The rollout of Xpert MTB/RIF — a nucleic acid amplification test for simultaneous diagnosis of TB and antibiotic sensitivity testing — has improved the early detection of resistance to first-line TB agents. The cartridge-based test identifies the most clinically relevant mutations in the RNA polymerase beta (rpoB) gene causing resistance to rifampin. However, it is not designed to either identify or exclude alternative resistance mechanisms. The consequent problem of suboptimal diagnosis of TB resistance has the potential to lead to resistance amplification, which could possibly be exacerbated by the use of assays exerting selective pressure on micro-organisms.

WGS screens all known genetic loci, both those associated with resistance and other loci, providing an opportunity to characterize new resistance-conferring mutations. Successful prediction of phenotypic sensitivity depends on accurate and comprehensive knowledge of resistance conferring these mutations.

A study from the CRyPTIC Consortium¹ published in the New England Journal of Medicine with an accompanying editorial² assessed the sensitivity and specificity of WGS for identifying susceptibility to first-line anti-tuberculosis drugs.

Genotypic predictions were based on mutations in 9 genes or their promoter regions associated with drug resistance to the standard first-line agents: isoniazid (ahpC, inhA, fabG1, and katG), rifampin (rpoB), ethambutol (embA, embB, and embC), or pyrazinamide (pncA).

All major TB lineages were analyzed from 10,209 isolates from 16 countries representing 6 continents. Predictions were made only if there were no other mutations present, with unknown association with drug resistance leading to between 89.8% (ethambutol) and 95.4% (rifampin) phenotypes being predicted for each first-line agent. Prediction of resistance was concordant with phenotype in 97.1%, 97.5%, 94.6%, and 91.3% (sensitivity) and 99.0%, 98.8%, 93.6%, and 96.8% (specificity) for susceptibility to isoniazid, rifampin, ethambutol, and pyrazinamide, respectively.

Incomplete prediction profiles of susceptibility were reported in 22.0% of patients due to unknown or failed predictions for at least one drug¹. Of those profiles that had complete genotypic predictions and were determined to be susceptible to all 4 drugs, 97.9% were predicted correctly. Simulation of the effect of increasing prevalence of resistance on negative predictive value was conducted by random sub-sampling of the isolates. The negative predictive value remained above 95% for 97.5% of profiles below a 47.0% prevalence of resistance to any drug, exceeding that seen in the countries with the highest multidrug-resistant-TB burden.

Infectious Disease Advisor talked to Dr Timothy Walker, department of microbiology, John Radcliffe Hospital, Oxford, United Kingdom (CRyTIC Consortium writing group).

Infectious Disease Advisor: Does the CRyPTIC Consortium plan to investigate the predictive value of WGS for testing drug susceptibility directly on sputum³ rather than via mycobacterial growth indicator tube samples?

Timothy Walker, MD:  This is something we are pursuing in parallel with the CRyPTIC project through a separate funding stream. If successful, I imagine the CRyPTIC consortium could well play a role in further studies into such an application.

Infectious Disease Advisor: Do we have sufficient information concerning resistance mutations to second-line agents to extend genotypic predictions to multidrug-resistant TB strains?

Dr Walker:  The sensitivity remains a little lower than for first-line drugs; for example, it is approximately 85% for quinolone class of antibiotics. CRyPTIC aims to increase the knowledge base for these drugs, among other objectives. I imagine we’ll be doing much better in a year’s time. 

Infectious Disease Advisor: What are the logistical barriers to the use of this technology in low resource settings?

Dr Walker:  That’s a big question. Briefly, this technology involves availability of laboratory facilities required for DNA extraction, training sufficient staff, overcoming customs issues, developing IT infrastructure, and ensuring user expertise in clinicians receiving the results. These barriers are all surmountable, and we are giving them plenty of thought. But it’s not a straightforward process.

Conclusion

Prediction of pan-susceptibility to all 4 first-line anti-tuberculosis drugs using whole genome sequencing has been shown to meet World Health Organization (WHO) standards for sensitivity and specificity. Based on this finding, public health authorities in England, The Netherlands and the United States (New York) have now stopped phenotypic drug susceptibility testing for isolates predicted by sequencing to be pan-susceptible to first-line drugs.²

References

1. The CRyPTIC Consortium and the 100, 000 Genomes Project. Prediction of susceptibility to first-line tuberculosis drugs by DNA sequencing [published online October 11, 2018]. N Engl J Med. doi:10.1056/NEJMOA1800474
2. Cox, H., & Mizrahi, V. The coming of age of drug-susceptibility testing for tuberculosis [published online October 11, 2018]. N Engl J Med. doi:10.1056/NEJMe1811861
3. Doyle RM, Burgess C, Williams R., et al. Direct whole-genome sequencing of sputum accurately identifies drug-resistant Mycobacterium tuberculosis faster than MGIT culture sequencing [published online July 26, 2018]. J Clin Micro. doi:org/10.1128/JCM.00666-18

This article originally appeared on Infectious Disease Advisor