Consequences of rpoB mutations missed by the GenoType MTBDRplus assay in a programmatic setting in South Africa

Background Rifampicin resistance missed by commercial rapid molecular assays but detected by phenotypic assays may lead to discordant susceptibility results and affect patient management. Objective This study was conducted to evaluate the causes of rifampicin resistance missed by the GenoType MTBDRplus and its impact on the programmatic management of tuberculosis in KwaZulu-Natal, South Africa. Methods We analysed routine tuberculosis programme data from January 2014 to December 2014 on isolates showing rifampicin susceptibility on the GenoType MTBDRplus assay but resistance on the phenotypic agar proportion method. Whole-genome sequencing was performed on a subset of these isolates. Results Out of 505 patients with isoniazid mono-resistant tuberculosis on the MTBDRplus, 145 (28.7%) isolates showed both isoniazid and rifampicin resistance on the phenotypic assay. The mean time from MTBDRplus results to initiation of drug-resistant tuberculosis therapy was 93.7 days. 65.7% of the patients had received previous tuberculosis treatment. The most common mutations detected in the 36 sequenced isolates were I491F (16; 44.4%) and L452P (12; 33.3%). Among the 36 isolates, resistance to other anti-tuberculosis drugs was 69.4% for pyrazinamide, 83.3% for ethambutol, 69.4% for streptomycin, and 50% for ethionamide. Conclusion Missed rifampicin resistance was mostly due to the I491F mutation located outside the MTBDRplus detection area and the L452P mutation, which was not included in the initial version 2 of the MTBDRplus. This led to substantial delays in the initiation of appropriate therapy. The previous tuberculosis treatment history and the high level of resistance to other anti-tuberculosis drugs suggest an accumulation of resistance.

http://www.ajlmonline.org Open Access Due to the limitations of the Xpert and the MTBDRplus assays, phenotypic methods remain the gold standard for tuberculosis DST. Both assays demonstrate variable performance in detecting heteroresistance and do not detect rpoB gene mutations outside the RRDR. 6 Until recently, mutations outside the RRDR were believed to only account for less than 5% of overall rifampicin resistance. 7,8 However, in a national drug resistance survey conducted in Eswatini between 2009 and 2010, 30% of multidrug-resistant tuberculosis (MDR-TB: resistant to rifampicin and isoniazid) isolates carried the I491F rpoB gene mutation located outside the RRDR. 9 This caused concerns, especially in neighbouring countries like South Africa, because while this mutation is rare globally, it might be more common in certain geographical settings. A subsequent study conducted in the northern provinces of South Africa showed that 15% of isoniazid mono-resistant strains carried the I491 mutation, meaning they were MDR-TB strains. 10 The same study also revealed that strains carrying this mutation may be driving outbreaks of MDR-TB in Eswatini and South Africa. 10 Rapid molecular tests that only detect mutations in the RRDR may fail to detect rifampicin resistance in patients with tuberculosis caused by strains carrying mutations outside the RRDR, and this may lead to inappropriate management, resulting in resistance selection, accumulation of resistance, treatment failure and increased transmission. In settings where molecular and phenotypic rifampicin DST are performed concurrently, discordant results often occur, especially with liquid culture-based assays. 11,12 Given the fact that rifampicin is the key determinant of the choice of a treatment regimen, the hesitancy caused by discordant results may also affect the decision to start appropriate treatment in the affected patients. Often, an attempt is made to confirm a discordant result by either repeating the test or using another confirmatory assay (if available), thus causing further delay in initiating appropriate therapy.
Because the KwaZulu-Natal province accounts for almost 30% of South Africa's drug-resistant tuberculosis (DR-TB) cases, and since Eswatini forms part of its northern border, we conducted this study in KwaZulu-Natal, South Africa, to determine why phenotypically resistant isolates were reported as rifampicin susceptible on the MTBDRplus. 13 Considering the dearth of information on the clinical management of patients with rifampicin-discordant tuberculosis results globally, we also report on the programmatic management of these patients in our setting.

Ethical considerations
Ethics approval was obtained from the University of KwaZulu-Natal Biomedical Research Ethics Council (BE267/18). Individual patient consent was not required as only routine programmatic data was accessed; however, permission was obtained from the provincial Department of Health. For anonymity, patients' names were only used for data collection and were not used during analysis.

Study design and setting
The study was conducted in the KwaZulu-Natal province in South Africa. The province has the second-highest population in the country with more than 11 million people. There are 11 districts in the province and one MDR-TB treatment facility per district.
In health facilities in the KwaZulu-Natal province, the initial diagnosis of tuberculosis and rifampicin resistance is routinely done using the Xpert ( -type probes  and mutation probes for the commonly occurring mutations  (S450L, H455Y, H455D, and D435V for rifampicin). The labelled polymerase chain reaction products from an amplified target are hybridised with specific probes immobilised on a strip (reverse hybridisation). Resistance is reported when there is a lack of binding to the wild-type probe with or without binding to a mutation probe. 14 Isolates that were resistant to either rifampicin or isoniazid on the MTBDRplus assay were further tested for resistance to critical concentrations of isoniazid (0.2 µg/mL), rifampicin (1 µg/mL), ofloxacin (2 µg/mL), streptomycin (2 µg/mL), and kanamycin (5 µg/mL) using the 1% agar proportion method on Middlebrook 7H10 agar (Becton Dickinson, Sparks, Maryland, United States). 15 The simultaneous performance of molecular and phenotypic rifampicin DST allowed the detection of discordance between these two tests.

Laboratory analysis
Routine clinical isolates from specimens received at the Inkosi Albert Luthuli Central Hospital laboratory of the KwaZulu-Natal province between January 2014 and December 2014 were used for this study. Isolates were selected if they showed rifampicin susceptibility on the MTBDRplus but were rifampicin resistant on the 1% agar proportion method on Middlebrook 7H10 agar at a critical rifampicin concentration of 1 µg/mL. The isolates from 2014 were chosen because simultaneous molecular and phenotypic rifampicin DST was performed during this time but was subsequently stopped. The selected isolates were then stored at -70 °C and later used for this study. Of the isolates that had discordant rifampicin results, 36 were randomly selected for further evaluation using whole-genome sequencing.

Clinical data
Patients with discordant rifampicin susceptibility results were identified from the laboratory. Further laboratory results (phenotypic DST, HIV status, CD4 count, and viral load results) were obtained from the laboratory information system. Treatment data was obtained from the electronic drug-resistant tuberculosis treatment register of the KwaZulu-Natal provincial Department of Health. Treatment outcomes were defined according to the WHO definitions. 17

Data analysis
The data were captured into an Excel file (Microsoft Corp., Redmond, Washington, United States) and cleaned and coded before being imported into STATA version 13 (StataCorp, College Station, Texas, United States) for statistical analysis. Patient names and ages were used to remove duplicate entries. Descriptive analysis was conducted on data for all patients with rifampicin-discordant tuberculosis results, as well as those selected for wholegenome sequencing. Categorical variables such as sex, HIV status, previous tuberculosis treatment, as well as the Xpert, phenotypic DST and MTBDRplus results, were presented as proportions and percentages. Continuous variables such as age, CD4 count, and the time taken to treatment initiation were presented as means with standard deviation. A bivariate analysis was conducted using the two-sample t-test to compare the mean time taken to treatment initiation between the Xpert-susceptible and Xpert-resistant results, and a p-value of < 0.05 was considered indicative of statistical significance.

Results
In 2014, out of 12 279 M. tuberculosis complex cases detected using the MTBDRplus assay, 505 (4.1%) were isoniazid monoresistant. From the 505 isoniazid mono-resistant cases, 145 (28.7%) were MDR-TB based on the phenotypic 1% agar proportion method (i.e., had discordant rifampicin DST results). The median age of the patients with discordant rifampicin DST results was 33.8 years, and 52.4% were male (Table 1).
The most common isoniazid resistance-conferring mutation was the S315T mutation, (29 isolates; 80.6%). This was followed by the inhA promoter region mutation T-8A (

Discussion
In this study, almost 29% of the isoniazid mono-resistant tuberculosis cases detected using the MTBDRplus assay were MDR-TB cases. This led to significant delays in the initiation of DR-TB treatment. The main cause of rifampicin resistance missed by the MTBDRplus assay was the presence of mutations outside the RRDR (mainly I491F), as well as the L452P rpoB mutation. Mutations outside the RRDR are not detected by the currently used WHO-endorsed rapid molecular assays, while the L452P mutations were missed by the previous version of the MTBDRplus assay. Importantly, isolates carrying these mutations were also resistant to other first-line anti-tuberculosis drugs whose resistance is not routinely tested in tuberculosis patients globally, and the isolates also clustered into distinct groups with unique mutation profiles.
The rpoB L452P mutation was left out of the earlier version (version 2, released in 2011) of the MTBDRplus assay as it was thought to be clinically insignificant. 18 This was later corrected in an updated version of the assay launched in 2014. 18,19 At the time of this study, the older version 2 was still in use, hence the discordant rifampicin results between the MTBDRplus assay and the phenotypic assay in isolates harbouring this mutation. The MTBDRplus assay may also miss heteroresistance. One Belgian study from 2019 found that the limit of detection of rifampicin heteroresistance was 5% -10%. 6 This may explain the other RRDR mutations missed by the MTBDRplus assay in this study. Notably, among isolates that had an Xpert result and had the L452P mutation as detected by whole-genome sequencing, the Xpert assay detected rifampicin resistance.
Mutations outside the RRDR were found in just over half (19/   Belgium by Torrea et al., the agar proportion method detected rifampicin resistance in 75% of isolates with I491F that was missed by the Mycobacteria Growth Indicator Tube DST. 12 It is therefore likely that the occurrence of these mutations is more frequent than what we found in this study. While the overall prevalence of I491F mutation among tuberculosis patients is reportedly low, in patients with isoniazid resistance, the prevalence is high. 9,10,20 The WHO defines universal access to DST as performing rapid DST for at least rifampicin in all patients with bacteriologically confirmed tuberculosis plus additional DST for at least fluoroquinolones and second-line injectable agents in patients with rifampicin resistance. 1 The use of Xpert as an entry point to tuberculosis care without investigating isoniazid resistance would prove disastrous for patients infected with M. tuberculosis strains that have mutations outside the area of detection and are resistant to all other first-line drugs. Recent studies conducted between 2015 and 2017 have shown that isoniazid resistance generally develops before rifampicin resistance. 21,22 Notwithstanding the importance of testing for rifampicin resistance, the neglect of isoniazid testing leads to inappropriate therapy, treatment failure and accumulation of resistance in patients with initial isoniazid resistance. 23 We therefore propose an algorithm to optimise DR-TB detection ( Figure 2). We submit that the initial DST should include both isoniazid and rifampicin. Importantly, if resistance is found to any of these two drugs, it should trigger further DST of other first-line and second-line drugs that will be used for treatment. Moreover, an attempt should be made to look for the I491F mutation in isolates from patients with isoniazid mono-resistant tuberculosis as this mutation may be missed by both phenotypic and genotypic DST methods that are routinely used for the detection of rifampicin resistance.
The largest global cluster of extensively drug-resistant tuberculosis that was ever reported was from Tugela Ferry in KwaZulu-Natal in 2005 and it was caused by a strain named F15/LAM4/KZN. 24   indications for performing M. tuberculosis culture, phenotypic DST was probably performed for these patients because they had already failed tuberculosis therapy. The presence of resistance to streptomycin suggests that these patients may have failed a few rounds of tuberculosis therapy because streptomycin was previously used as part of a standard re-treatment regimen in patients who had failed first-line therapy. In South Africa, this regimen was stopped after the rollout of Xpert, which allowed universal testing of all tuberculosis patients. The rollout of Xpert was completed towards the end of 2013.
Isolates in this study belonged predominantly to lineage 4, which is known to predominate among DR-TB cases in the KwaZulu-Natal province. 30

Limitations
This study reports old data on M. tuberculosis isolates from 2014. However, we examined this period because this was when both phenotypic and genotypic rifampicin DST were performed simultaneously in our setting. Moreover, the tuberculosis diagnostic algorithm has not changed since then, although phenotypic rifampicin DST was subsequently stopped. Another limitation of this study was our use of phenotypic DST to select isolates with possible mutations outside the RRDR instead of using molecular screening. This may have underestimated the prevalence of isolates with these mutations as some of them remain susceptible on the phenotypic assay. Due to limited resources, we sequenced only a subset of the isolates with discordant results.
The data from the susceptible tuberculosis treatment register was not available to compare with that on the DR-TB register to determine if patients not listed on the DR-TB register were treated with first-line tuberculosis therapy. Finally, the study was performed in one province of South Africa so the findings may not apply to other regions. Nonetheless, this province has the highest prevalence of DR-TB cases in the country and similar findings have been reported in the northern provinces.

Conclusion
The presence of highly drug-resistant M. tuberculosis strains with mutations missed by the routine rapid molecular assays highlights the need for the revision of the WHO definition of universal access to DST so that tuberculosis diagnostic algorithms include testing for both isoniazid and rifampicin in all patients with bacteriologically confirmed tuberculosis. The recent endorsement of the Xpert MTB/XDR by the WHO for detection of isoniazid, fluoroquinolone and second-line injectable agent resistance in Xpert (Ultra)-confirmed tuberculosis cases provides an opportunity to close the gap in isoniazid testing. 1 The I491F mutation remains the most commonly detected mutation outside the RRDR and its frequent occurrence in isoniazid-resistant cases calls for its inclusion in assays that detect rifampicin resistance. This codon is not too far away from the RRDR, so current assays can be upgraded to include it to avoid the use of inappropriate therapy, prevent the accumulation of resistance, and reduce community spread.