Phenotype versus genotype discordant rifampicin susceptibility testing in tuberculosis: implications for a diagnostic accuracy

ABSTRACT The World Health Organization recommends culture-based techniques as the gold standard for rifampicin-resistant (RR) tuberculosis. However, as rapid molecular assays are rolled out, discordance in resistance has been observed. The current study compared 516 clinical isolates on Lowenstein–Jenson (LJ) and Mycobacteria Growth Indicator Tube (MGIT) drug susceptibility testings (DSTs) with rpoB sequencing. The discordance rate was higher in MGIT DST (24%) compared to LJ (9.4%). We report discordant results for RpoB mutations 430P (MGIT: 14/20; LJ: 19/20), 435Y (MGIT: 5/21; LJ: 17/21), 445N (MGIT: 11/18; LJ: 15/18), 445L (MGIT: 0/9; LJ: 8/9), 452P (MGIT: 5/20; LJ: 17/20), and 445C (MGIT: 1/6; LJ: 2/6) on LJ and MGIT DSTs, respectively. Mutations N438K and T444A, sensitive on LJ and MGIT DSTs, are novel. Two novel discordant mutations, Q436H+N437del and S428I, were also detected. One novel large deletion (Del) ggaccagaacaacccg/G (Del of DQNNP at positions 435–439) demonstrated concordance on LJ and MGIT DST. Numerous double and triple mutations including some novel combinations, M434V+H445N, L430R+D435Y, H445N+S450W, S428R+L430P, M434L+H445N, D435G+A451V, and M434I+D435G+A451V, have also shown higher discordant results on MGIT DST. The discordance results of RR are higher in MGIT DST compared to LJ. We suggest that the genotypic result should be preferred instead of phenotypic DST in case of discordant mutations. Physicians should be aware of such discordance to prescribe an effective treatment. Untargeted whole-genome sequencing to capture MGIT-sensitive and large deletions may be reconsidered for better treatment outcomes in high-burden countries. IMPORTANCE An accurate diagnosis of drug resistance in clinical isolates is an important step for better treatment outcomes. The current study observed a higher discordance rate of rifampicin resistance on Mycobacteria Growth Indicator Tube (MGIT) drug susceptibility testing (DST) than Lowenstein–Jenson (LJ) DST when compared with the rpoB sequencing. We detected a few novel mutations and their combination in rifampicin resistance isolates that were missed by MGIT DST and may be useful for the better management of tuberculosis (TB) treatment outcomes. Few novel deletions in clinical isolates necessitate the importance of rpoB sequencing in large data sets in geographic-specific locations, especially high-burden countries. We explored the discordance rate on MGIT and LJ, which is important for the clinical management of rifampicin resistance to avoid the mistreatment of drug-resistant TB. Furthermore, MGIT-sensitive isolates may be subjected to molecular methods of diagnosis for further confirmation and treatment options.

to two first-line drugs against M. tuberculosis, constituting a standard therapy, isonia zid (INH) and rifampicin (RIF), is rising.The multidrug-resistant (MDR)-TB (defined as resistance to both INH and RIF) (1,2) has further complicated the TB therapy by controlling about half a million new MDR-TB per year.An estimated 3.4% of new and 18% of previously treated cases were rifampicin-resistant (RR) or MDR-TB globally (3).According to the latest TB report, MDR-TB or RR-TB (MDR/RR-TB) cases were relatively stable from 2015 to 2020 but increased in 2021 (n = 450,000), showing a 3.1% rise from 2020 (n = 437,000) (4).In Pakistan, an estimated 4.2% of new TB cases were MDR/RR-TB in 16% of previously treated patients (5).
MGIT DST is routinely performed on a WHO-defined critical concentration of 1 µg/mL in defined conditions.However, at a critical concentration of 1 µg/mL, it has been shown that strains harboring discordant mutations are detected as rifampicin-susceptible as this critical concentration is above the epidemiological cutoff (ECOFF) values (20).Discordant mutations and the consequent discordance between LJ DST, MGIT DST, and genotypic DST may lead to uncertainty in the appropriate TB management and make it hard to treat the patients.The diagnostic accuracy of phenotypic and genotypic assays varies as the frequency of mutations in different geographic regions differs (21,22).
Recently, the WHO has revised the rifampicin critical concentration from 1 to 0.5 µg/mL (23) on MGIT DST to reduce this issue; however, this study was conducted before the revised rifampicin critical concentration.
Pakistan has been ranked fifth among 30 high-burden countries; data on discord ant are not available for better TB management.Phenotypic DST and MTBDRplus are performed on all cases initially diagnosed as RR by GeneXpert to identify other susceptible drugs that can be included in the regimen and for the confirmation of GeneXpert results.Discordant results have been encountered in our region, causing clinical and diagnostic dilemmas.For the first time, the current study was designed to identify the extent of discordant RR cases on LJ and MGIT DSTs and design strategies for accurately detecting resistance, leading to better treatment options.

Sampling
All clinical samples from January 2014 to October 2016 included all available discordant isolates RR on LJ and susceptible on MGIT DST or sensitive on MGIT and LJ or RR on LJ, and MGIT was isolated in the National Tuberculosis Reference Laboratory Pakistan followed by a random selection of concordant isolates before and after the discordant isolate for this study.The National TB Reference Laboratory receives all RR cases from Punjab, the largest province of Pakistan, Islamabad capital territory, Kashmir, and Gilgit Baltistan for comprehensive phenotypic DST results.

Specimen processing
Clinical specimens were decontaminated using the NALC-NaOH method and inocu lated on Middlebrook 7H9 broth (Bactec MGIT 960 tube) LJ media for Mycobacte rium tuberculosis complex (MTBC) growth.A rapid BD MGIT TBc identification test was performed on positive cultures to confirm MTBC.Liquid culture-based DST was performed on isolated MTBC cultures on BD Bactec MGIT 960 using the SIRE kit (Bactec MGIT 960 SIRE Kit) (24).The standard LJ proportion method for susceptibility testing of M. tuberculosis isolates was also performed as per the standard procedure (25) using rifampicin at 40 µg/mL.A susceptible strain, H37Rv, and two known rifampicin-resistant strains selected from National Proficiency testing round 19 received from the Suprana tional Reference Laboratory, Antwerp, Belgium, were used for quality assessment and accuracy in DST results (26,27).

DNA extraction
DNA for polymerase chain reaction (PCR) was extracted from 516 randomly selected RR isolates including 120 rpoB wild type (sensitive on MGIT and LJ) and 396 RR by heating bacterial colonies at 90°C for 20 minutes in 100 µL of water followed by sonication on an ELMA E30H Sonicator (Elma Schmidbauer GmbH) for 15 minutes.

Polymerase chain reaction
PCR was performed using 25 µL of Qiagen HotStarTaq Master Mix containing 2.5 units of HotStarTaq DNA polymerase 1× PCR buffer (contains 1.5 mM MgCl 2 ) and 200 µM of each dNTP, 2 µL of each forward and reverse 10 µM primers, 8 µL of DNA, and 5 µL of water with a total volume of 50 µL in a GeneAmp PCR system 9700 (Applied Biosystems).The following cycling conditions were used: 15-minute hot start followed by 45 cycles at 94°C for 45 seconds; annealing and extension for 1 minute and 30 seconds at 72°C; and a final amplification at 72°C for 10 minutes.Primers rpoBgeneSA and rpoBgeneRB (5-GGTTCG CCGCGCTGGCGCGAAT-3) and (5-GACCTCCTCGATGACGCCGCTTTCT-3) (28) were used for the amplification of rpoB gene (28).PCR amplicons were analyzed on 2% agarose gel.

RESULTS
Rifampicin resistance results were available on GeneXpert, LJ, and MGIT DST for 1,986 strains during the study period.A total of 1,761 were concordant, and 225 discord ant isolates were analyzed (data have not been shown).The rpoB sequencing results were available for 516 strains, including 120 wild types and 396 (225 discordant and 171 concordance) randomly selected RR isolates (BioProject: PRJNA952823, GenBank submission ID is 2693281).The RR RpoB sequencing, LJ, and MGIT DST results are presented in Table 1; Table S1 based on the M. tuberculosis and Escherichia coli rpoB codon numbering system.We also calculated the 95% confidence interval for each mutation identified in our study to check the sensitivity of LJ and MGIT DSTs (Table 1; Table S1).Mutations were identified in 396 strains in rpoB, in which 42 out of 396 were susceptible in LJ DST, whereas MGIT DST reported 95 isolates as RIF-susceptible.
Ten isolates containing different combinations of mutations (two to three per combination), including two long resistant combinations T427I+D435Y+I491L and S428G+H445N+L452P, showed concordant results on both LJ and MGIT DST.Overall, four mutations, L430R (n = 2), N438K (n = 1), T444A (n = 1), and S431G (n = 1), were sensitive on both types of phenotypic DST in which T444A is novel in the current study.One novel large deletion, ggaccagaacaacccg/G, caused Del of amino acids (aa) DQNNP (435-439) showing concordant results of resistance on LJ and MGIT DST.One single amino acid, Del (Del of N at 519), and insertion (R at 432) were also resistant on both phenotypic DST.A single isolate harboring double mutations (H445>P+K446>Q) was resistant on both phenotypic DST (Table S1).However, to our knowledge, this double mutation in a single isolate is novel in our study.Mutations at Q436H+N437del (Table S1) and S428I were sensitive on MGIT DST at a critical concentration, while both of these were resistant on LJ DST and were classified as disputed mutations on MGIT DST.These mutations are also novel (Q436H+N437del, S428I) and were not previously reported as discordant.Double mutations S428T+D435G and S428R+H445Y, which show concordant results on both types of DST, are also a novel combination in a single RIF-resistant isolate for which the minimum inhibitory concentration (MIC) has not been determined.
MIC of the mutant rpoB was determined on MGIT and LJ (Table 2).Sixteen isolates harboring different discordant mutations exhibited MIC greater than 1 µg/mL; 12 had less than 1 µg/mL (<1 µg/mL) on MGIT, whereas 25 mutant RpoB isolates showed MIC greater than 40 µL/mL and 4 less than 40 µL/mL on LJ.The MIC of some samples harboring mutation His445Asn and Leu452Pro is still under processing on MGIT and LJ.Some of these have been sent to whole-genome sequencing (data are not available) to find some compensatory mutations.Asp435Tyr is a disputed mutation resulting in  a WGS, whole genome sequencing.
discordance results on liquid and solid DSTs.In order to look closely, we performed MIC on LJ and MGIT for four strains and WGS for two strains (BioSample Accession SAMN34126583).One strain with MIC > 1 on MGIT DST and >40 on LJ medium was closely analyzed on WGS and found another mutation Pro358Leu in the rpoB region, which resulted in higher MIC.Similarly for Leu430Pro, a second mutation was found Cys701Trp (BioSample accession SAMN34126582) which resulted in increased MIC > 1 on MGIT and >40 on LJ medium.For a more strongly disputed mutation Leu452Pro, a second mutation was detected at His445Tyr (BioSample accession SAMN34126584) causing increased MIC > 1 on MGIT and >40 on LJ medium.

DISCUSSION
The prevalence of discordant (generally missed by rapid phenotypic DST methods) rpoB mutations is largely limited, and only a few studies addressed this problem and its impact on clinical implications (15,31,32).Population-based molecular studies of RIF mutation screening are required for better understanding and treatment options, particularly in early MDR.
Similarly, the D435>Y sensitivity frequencies were also different on MGIT and LJ.The confidence grading of this mutation is unavailable on MGIT DST, while a moderate confidence grading has been given on solid DST (32).Rigouts et al. detected D435>Y in six resistance strains, completely missed on MGIT DST (9); however, they were resist ant on LJ DST.The possible reason for these variations may be due to compensatory mutations (40) responsible for the fitness cost in different media.Similarly, other studies have reported this mutation as discordant (9,15,(33)(34)(35)(36).
Siu et al. reported that 4% of RIF resistance cases do not harbor mutations in RRDR.In the current study, two mutations (I491>F and V170>F) were detected outside the RRDR of RpoB in which I491>F was detected in five RIF resistance isolates, and only one (20%) was MGIT-sensitive; however, all five isolates were LJ-resistant, which have been classified in minimal confidence grading (35,44).The mutant V170>F showed concord ance results on LJ and MGIT.According to a WHO multi-country surveillance study, V170>F and I491>F have elevated RIF MICs in M. tuberculosis [0.5% (95% CI 0.2%-1.2%)](45-47).However, I491>F (E. coli I572F) have been accounted for more than half of RR in Eswatini, while its frequency in South Africa, a neighboring country, was <1% (48,49).Several studies reported the MIC for mutants I491F and V170>F as 100 and 125 µg/mL, respectively (50)(51)(52).One notable point from Ma et al. 's study is the experimental proof of compensatory effects of non-RRDR mutations, which may have molecular markers' importance in predicting the fitness of clinical M. tuberculosis strains.
M. tuberculosis isolates harboring double and triple mutations have also shown discordant results on MGIT DST only; however, all these strains were resistant on LJ DST (Table S1).A single triple mutation isolate (S428G+H445N+L452P) exhibited concordance results on both phenotypic DSTs.This combination is also novel and has not been reported in earlier studies.Mutation D435G, detected in one RIF-resistant strain, is associated with high-level resistance on LJ but not MGIT (32).However, we reported this mutation as discordance on LJ and MGIT.Double mutations M434V+H445N (n = 5) and L430R+D435Y (n = 2) were all LJ-resistant.Double mutants (M434V+H445N and L430R+D435Y), which showed different frequencies on MGIT and LJ resistance, were reported as MGIT-resistant only without performing LJ DST (32).Miotto et al. (32) reported M434I+H445N mutant as undisputed and MGIT-resistant without confirmation on LJ.However, isolates harboring double mutation (M434I+H445N) are RR on both LJ and MGIT DSTs; even DST was repeated twice for these isolates.Previously, L430R was reported as a low-level resistance mutation to RIF, and S431G is sensitive and has no effect on RIF interaction with RpoB (53).In an earlier study (43), L430>R and L430>P were detected in MDR isolates and other RpoB mutations at positions 431, 435, and 445, except for four isolates where L430P was detected alone.Mutation N438K (n = 1), which was found LJ-and MGIT-sensitive, was not classified (unpublished) in WHO technical report 2021 (44).One novel mutation (T444A), LJ-and MGIT-sensitive, is also missing in WHO report 2021 on the critical concentration of drugs.
The frequencies of double discordance mutations are low, and one could comment that the data are insignificant; however, in either case, MDR-TB misclassification could not be ignored to achieve the Global TB Control Program.Let us consider that such instances are very rare.Still, if a TB outbreak occurs because of strains harboring these resistance mechanisms, it will cause misclassification, leading to mistreatment and further transmission in populations.
Our results show that both LJ and MGIT DSTs are prone to miss rifampicin resistance; however, the discordance rate in MGIT DST is higher than that in LJ.Our study population represents a substantial number of common and less common mutations and their sensitivity on LJ and MGIT DSTs.Rigouts et al. (9) showed that the automated MGIT DST is more prone to demonstrate rifampicin-susceptible results in discordant rpoB mutations than LJ DST.Similarly, a study involving the Supra-National TB Reference Laboratories network of the WHO has shown that more RIF-susceptible results were observed in liquid DST compared to solid DST in samples with discordance rpoB mutations (15).Miotto et al. hypothesized that discordant mutations lower the RIF binding affinity but do not prevent it completely (32).Some other studies pointed out that a loss of fitness in M. tuberculosis strains may slow their growth (58,59), while MGIT is a fast-growth system.The L430>P, D435>Y, H445>C/L/N, and L452>P mutations have already been reported as disputed mutations on MGIT (9,32,43,53), and isolates harboring it will be misclassified on phenotypic results (MGIT).
About 14% of all rpoB mutations are discordant (17) when subjected to conventional phenotype-based drug susceptibility testing.The M. tuberculosis isolates with discordant mutations are thus classified as RIF-susceptible, and the patients are treated as drug-sus ceptible TB (16,60), leading to poor treatment outcomes (16,17,39).Many researchers suggested that the critical concentration for MGIT DST should be lowered as low as 0.0625-0.5 µg/mL (39,44,61).Though earlier studies showed that performing MGIT DST at a critical concentration of 0.25 µg/mL failed to address the problem as some discordant isolates were still missed, some susceptible isolates were classified as resistant (36).
Recently, the WHO has revised rifampicin critical concentration from 1 µg/mL to 0.5 µg/mL on MGIT DST to reduce this issue (44).The misclassification rate of RIF resistance isolates harboring discordant mutations may be higher for MGIT DST than for LJ due to the new MGIT critical concentrations, which need further clarification.There will be a higher overlap between the MIC of discordant resistant and susceptible isolates with MGIT as it has a shorter incubation period (33).
WHO found six RRDR mutations L430>P, D435>Y, H445>L, H445>N, H445>S, and L452>P and some outside the RRDR (non-RRDR).We identified some mutations outside RRDR (170>F, 491>L, and 491>F), which cannot be disregarded as GeneXpert test ing services are decentralized, and RR cases are only systematically referred for DST.Therefore, samples from patients reported as rifampicin-sensitive on GeneXpert but with poor response to treatment should be referred for LJ DST or gene sequencing to avoid misdiagnosis of resistant strains; otherwise, ineffective treatment will result in proliferation and transmission of M. tuberculosis strains in the population.Mutations with significant frequencies outside the RRDR region may be considered in genotypic DST, as evidenced by a previous study in which 30% of RR cases were detected outside the RRDR (62) region.
The variability in MICs with the same mutation (Table 2) may be due to the growth rate (63).The growth rate has been associated with mutations, a useful indicator of fitness costs (64).M. tuberculosis isolates with a known RR mutation have comparable fitness costs irrespective of the genetic makeup.Many factors impact M. tuberculosis fitness costs, including mutations in rpoA and rpoC, which compensate for RIF resist ance behind mutations in RpoB (65).Thus, the in vitro M. tuberculosis growth rate is significantly improved in the presence of rpoB mutations along with compensatory mutations.Moreover, M. tuberculosis isolates harbor mutations in other targets confer ring resistance, which may sustain uncertain fitness costs.Estimating accurate fitness costs is difficult as the diversity of mutations under different environmental conditions may also have diverse impacts on fitness.Therefore, precision in determining the MIC with variable growth rates may not be possible.Moreover, some studies closely look only for a single mutation, so double mutations or mixed population is overlooked.In the case of well-known disputed mutations (430>Pro, 435>Tyr, 441>Leu, 445>Asn, 445>Leu, 452>Pro, and 491>Phe), if found resistant on MGIT DST, a close look using deep sequencing should be applied to explore any other mutations that may cause increased MIC.

Recommendation
It is also important to note that some mutations are geographically specific, which may not be in MGIT detection but have a significant frequency in M. tuberculosis isolates.This has been evidenced by our recent study (66) from the same geography of Khyber Pakhtunkhwa Province, in which we reported the majority of MDR-TB isolates to have a high frequency of rpoC mutations (32/46), a locus recognized to restore the fitness of RIF resistance.Purely genotypic DST with untargeted whole-genome sequence for MGIT susceptibility and LJ resistance, as recently initiated by the United Kingdom (67), may be useful for countries with high-burden TB.Alternatively, rpoB, rpoC, and rpoA sequencing of MGIT-sensitive and LJ-resistant isolates might be useful for reducing the gap in RIF resistance and for GeneXpert comparison.Moreover, geographic-specific mutations may play a role in evading the detection power of current MGIT, as discussed in the text.Identifying the complete range of RIF resistance and compensatory mutations will improve clinical management and surveillance decision-making using long-read sequencing data.While genotypic DST with untargeted WGS will be an important initiative, long-read WGS (lrWGS) (68) is most important to capture the compensatory mutations.In a programmatic setting, we would support the implementation of lrWGS for MGIT and targeted genotypic wild-type isolates to reduce and unveil the discordance and compensatory mutations associated with drug resistance.Such steps may be useful to reduce the TB and drug resistance burden.

Conclusion
Numerous discordant and concordant novel mutations and their combination have been detected on LJ and MGIT DST.Mutations N438K and T444A, sensitive on LJ and MGIT DSTs, are novel.Two novel discordant combinations, Q436H+N437del and S428I, were also detected as LJ-resistant and MGIT-sensitive.One novel large deletion (Del) ggaccagaacaacccg/G (Del of DQNNP at positions 435-439) demonstrated concordance on LJ and MGIT DST.Numerous double and triple mutations, including some novel combinations, have also shown higher discordant results on MGIT DST.The discordant level in MGIT DST is more than that in LJ DST, which may cause poor treatment outcomes.In discordant mutations, phenotypic DST should be replaced and preferred with RIF target genes or untargeted whole sequencing.All clinicians involved in TB treatment management should be well trained to tackle such discordance mutations for proper treatment regimens.The gold standard for rifampicin resistance detection may be reconsidered as a known link between discordant mutations and poor clinical outcomes.Isolates showing discordance on MGIT may undergo sequencing to unveil the hidden resistance mechanism.

TABLE 1
Summary of rifampicin concordance and discordance on LJ and MGIT DSTs and sensitivity, stratified by RpoB mutations in M. tuberculosis isolates from Pakistan a

Codon change Amino acid change M. tuberculosis numbering (E. coli) Rifampicin DST results on LJ Rifampicin DST results on MGIT Total
a R, resistant; S, susceptible; CI, confidence interval.

TABLE 2
RpoB mutations and MICs of RIF