Whole-genome sequencing for prediction of Mycobacterium tuberculosis drug susceptibility and resistance: a retrospective cohort study

Abstract

Background: Diagnosing drug-resistance remains an obstacle to the elimination of tuberculosis. Phenotypic drug-susceptibility testing is slow and expensive, and commercial genotypic assays screen only common resistance-determining mutations. We used whole-genome sequencing to characterise common and rare mutations predicting drug resistance, or consistency with susceptibility, for all first-line and second-line drugs for tuberculosis.

Methods: Between Sept 1, 2010, and Dec 1, 2013, we sequenced a training set of 2099 Mycobacterium tuberculosis genomes. For 23 candidate genes identified from the drug-resistance scientific literature, we algorithmically characterised genetic mutations as not conferring resistance (benign), resistance determinants, or uncharacterised. We then assessed the ability of these characterisations to predict phenotypic drug-susceptibility testing for an independent validation set of 1552 genomes. We sought mutations under similar selection pressure to those characterised as resistance determinants outside candidate genes to account for residual phenotypic resistance.

Findings: We characterised 120 training-set mutations as resistance determining, and 772 as benign. With these mutations, we could predict 89·2% of the validation-set phenotypes with a mean 92·3% sensitivity (95% CI 90·7-93·7) and 98·4% specificity (98·1-98·7). 10·8% of validation-set phenotypes could not be predicted because uncharacterised mutations were present. With an in-silico comparison, characterised resistance determinants had higher sensitivity than the mutations from three line-probe assays (85·1% vs 81·6%). No additional resistance determinants were identified among mutations under selection pressure in non-candidate genes.

Interpretation: A broad catalogue of genetic mutations enable data from whole-genome sequencing to be used clinically to predict drug resistance, drug susceptibility, or to identify drug phenotypes that cannot yet be genetically predicted. This approach could be integrated into routine diagnostic workflows, phasing out phenotypic drug-susceptibility testing while reporting drug resistance early.

Funding: Wellcome Trust, National Institute of Health Research, Medical Research Council, and the European Union.

Figure 1

Candidate genes and mutations The number of potentially predictive mutations in genes relevant to each drug after lineage-defining and synonymous mutations have been set aside and are shown by susceptible and resistant phenotypes for 2099 training-set isolates. Genes from which one or more of the 120 resistance-determining mutations were algorithmically characterised are coloured red.

Figure 2

Resistance determinants in training and validation sets Mutations probed by a line-probe assay are coloured red. Mutations that were only noted once in the training set and not again in the validation set (ie, with no additional information to validate them) are not shown. Of the quinolones and aminoglycosides, only ofloxacin and amikacin have been included as representatives of their class.

Figure 3

Phenotypic and genotypic antibiograms for all 3651 isolates The left-hand panel shows the phenotypes for seven drugs for the 3651 isolates. The right-hand panel shows the genotypic predictions based on the mutations characterised after applying the algorithm to all 3651 isolates. INH=isoniazid. RIF=rifampicin. EMB=ethambutol. PZA=pyrazinamide. SM=streptomycin. OFX=ofloxacin. AK=amikacin.

Figure 4

Training-set-characterised mutations Numbers represent the number of mutations for each characterisation. *Among resistance determinants and benign mutations, 15 and 55 insertions and deletions, and 25 and 371 mutations seen in only one isolate respectively, were not or could not be assessed for homoplasy. †gyrA A384V defines the Indian Ocean lineage (all isolates in the lineage have this single-nucleotide polymorphism) but is also in one European American isolate. rpsA A440T defines Mycobacterium bovis but is also in one Central Asian isolate. Both are thereby homoplasic.

Figure 5

Proposed workflow for transition towards whole-genome sequencing-based drug-susceptibility testing *The 30% CI width suggested is arbitrary, and represents how the precise proportion of isolates with a mutation is probably less relevant than understanding whether this proportion is very high, moderate, or low. However, the precise width could be determined by what is regarded as an acceptable degree of clinical risk, and could also vary by the estimate of proportion resistant. For example, with a targeting width of less than 30%, ten phenotypically resistant isolates of ten isolates with a mutation (100%) has a lower 97·5% CI of 69%, so mutations that are uniformly resistant would need to be phenotyped 11 times before confirmatory phenotyping would stop. For a mutation associated with resistance in 50% of isolates, phenotyping would need to happen 48 times, and for a mutation associated with resistance in either 25% or 75% isolates, 36 times.