A targeted nanopore sequencing-based test method for the rapid diagnosis of drug-resistant TB - Justin O'Grady
Tuberculosis is a "major cause of morbidity and mortality globally". An estimated 10 million cases and >1.6 million deaths are attributed to tuberculosis (TB) each year. The greatest threat to TB treatment and prevention efforts is drug resistant TB; half a million cases of multi-drug resistant TB (MDR-TB) occur annually, and 6.2% of these cases are "even worse" and are now classified as extremely drug resistant (XDR) TB, meaning that they are incredibly difficult to cure. TB is most prevalent in middle- and Southern-Africa, but most instances of new TB cases with drug resistance occur in Russia, as well as China and India. The incidence of XDR-TB is greatest in Russia and South East Asia.
Justin described how next generation sequencing has great potential for rapidly diagnosing drug resistant TB (DR-TB). It overcomes many of the significant challenges associated with conventional testing, which are slow, as "TB takes a long time to grow" in culture, and can be inaccurate for determining resistance to some drugs. It is also more comprehensive than traditional tests which only test a limited number of targets across the genome. Nonetheless, uptake of sequencing can be hindered by concerns regarding its cost, integration into existing lab workflows, skills required for using the technology, and management and interpretation of the sequencing data.
Displaying a world map of MDR/DR-TB cases tested for susceptibility to second-line drugs, Justin described how wealthy places are performing tests for TB identification, but they aren't doing enough testing for second line resistance in places with MDR-TB or XDR-TB. Justin introduced the Seq&Treat programme, which involves a joint effort by the Foundation for Innovative New Diagnostics (FIND) and Unitaid. The aim of this programme is to evaluate the use of targeted next generation sequencing for diagnosis of DR-TB in low-middle income countries, by generating clinical evidence to support WHO guidance for the use of targeted sequencing for DR-TB identification. This involves evaluating existing methods for integrating targeted sequencing into existing diagnostic workflows. The overall aim would be to try and implement such targeted sequencing globally in the near future.
This programme state requirements for a test identifying DR-TB; these include: >98% specificity and sensitivity for drug resistance detection; 100-5,000 CFU/ml as the limit of detection; detection of hetero-resistance from 1-10% (equal to 10 resistant reads in 100-1,000 reads, and 500 resistant bacteria in 4,5000); minimal cross reactivity with non-tuberculous mycobacteria (NTM); and <5% indeterminate results.
Justin described how his project has been to "come up with a test to fulfill these criteria" using the MinION for targeted sequencing and DR-TB detection. Advantages of the MinION include it being easily deployable globally at low cost; rapid turnaround to results and real-time analysis; flexible sample numbers (depending on the size of the sequencing facility and the expected number of samples received); and low cost-per-sample ($15 USD per sample at 40 samples/flow cell - the extent of multiplexing that "we are aiming to do"). Justin stated that MinION sequencers can be deployed cheaply across the globe.
The ~6-24 hour workflow for nanopore sequencing and turnaround involved 1 hour of DNA extraction, a 17-target PCR (2.5 hours), library preparation (4.5 hours for many samples, using the Ligation Sequencing Kit and PCR barcoding expansion), MinION sequencing (overnight, although "if you have twelve you could do this in a few hours"), and finally, analysis using EPI2ME to investigate the TB resistance profile. The 17 targets examined were associated with both first and second line drug resistance.
Justin noted that the library prep workflow (4.5 hours) is a little bit complex and cumbersome at the moment, and he would like to reduce the number of targets and complexity.
In terms of sequencing metrics, the average read length achieved was ~1 kb, with an average quality score of 9.88. The mean depth of coverage achieved was ~12,871x (range 2,382x - 28,301x). Justin said that he would like the depth of coverage to be more even across the 17 targets, but considering it took quite a while to get some of the targets working "we can live with that"!
By performing serial dilutions, the limit of detection was found to be at the 50x threshold, between~100-10 cells per ml, or about 10x depth of coverage on the 10-cell equivalent. Justin stated that he was "very pleased that the primers are working" right down to such low concentrations of DNA.
In terms of specificity, the data were tested against clinical sputum containing a range of mycobacterial species, and compared to publicly available NTM FastQ datasets. It was found that 15/17 assay targets were consistently specific for MTB members. As depth of coverage for the two exceptions was very low, and because it was only 2 of the 17 targets, Justin said that he wasn't really concerned about this lower specificity.
The EPI2ME TB resistance profile identified reads associated with genes involved in resistance, and if any resistance was detected. Showing an example specimen that they had tested, Justin pointed out how a SNP that conferred pyrazinamide resistance was detected within the sample.
Justin said that they now plan to test their method on ~450 well defined sputum samples provided by FIND, to define clinical performance. These samples contain varying levels of TB bacteria, different resistance phenotypes and genotypes, as well as other bacteria, and "we have got to see if we can find all that". The test will proceed to a global trial if performance is sufficient, with the long term aim of rolling it out globally in future.
To conclude, Justin stated that they have been able to develop a rapid and accurate test to detect DR-TB, and that MinION sequencing is "flexible and cost-effective and very easily deployable".