Introduction

Tuberculosis (TB) is a disease of global public health importance declared the deadliest infectious disease alongside HIV/AIDS [1–3]. Although new cases have been reducing every year [3, 4], TB has persisted in human populations and it remains among the top 10 causes of death worldwide [3, 4]. In 2014 there were 9.6 million TB cases reported and 1.5 million deaths; more than 95% of the deaths occurred in developing countries [3, 5].

TB control programs in many countries rely on rapid identification of cases for effective treatment of the disease. Importantly, the drug-resistant forms of TB require accurate diagnosis of the type of resistance to guide therapy and interrupt transmission of resistant organisms in communities [6]. However, the conventional culture-based diagnostics for TB are slow, labor intensive and expensive, and they are often not available in the developing countries. The demand for rapid TB diagnostics led to the development and introduction of commercial molecular tests into practice. Two commonly used tests in Uganda are the “Xpert MTB/RIF” (GeneXpert, Cepheid Inc.), which diagnoses TB and detects rifampicin resistance [7–11], and the GenoType MTBDRplus (line probe assay [LPA] from Hain Life Sciences Nehren, Germany), which detects Mycobacterium tuberculosis and its resistance to both rifampicin and isoniazid [12–14]. Rifampicin and isoniazid are the most important anti-tuberculosis drugs, and resistance to both drugs is commonly referred to as “multidrug resistance” (MDR-TB) [1]. Importantly, although the commercial molecular tests rapidly and accurately diagnose TB [15, 16], they may not detect all the resistance especially in strains with novel or unknown resistance mechanisms. The molecular tests are also ineffective at detecting heteroresistance (“the occurrence of populations of both drug-susceptible and drug-resistant isolates within the same clinical sample” [17]) due to mixed-strain TB infections. This has led to reports of varying efficiencies in the performance of rapid molecular tests in detecting drug resistance, depending on the drug in question [6, 18], or location of the strains under study. Without doubt, differences in geographical distribution of TB strains (i.e. M. tuberculosis lineages and/or sub-lineages/genotypes) across continents [6, 18–21] significantly affect the performance of rapid molecular tests [22–24].

Molecular tests for TB target spontaneous point mutations in specific genes and/or loci in the M. tuberculosis chromosome that are associated with drug resistance [25], and only a limited number of mutations accounts for majority of the resistance to anti-TB drugs [6, 18]. Rifampicin-resistance, which is the best understood of all anti-TB drugs, is associated with mutations in an 81 base pair region of the rpoB gene commonly referred to as the rifampicin-resistance determining region (rpoB/RRDR). Mutations within the rpoB/RRDR confer resistance to rifampicin in majority of the rifampicin-resistant isolates [6], and three key mutations i.e. Ser531Leu, His526Tyr, and Asp516Val are reported to be predominant worldwide [6, 26, 27]. For isoniazid, mutations mainly in katG and inhA gene promoter, and infrequently ahpC, oxyR, kasA, furA and ndh genes, confer resistance to the drug [6, 18]. Approx. 64% of the phenotypic resistance to isoniazid globally is attributed to the katG/Ser315Thr mutation [18].

Regional variation in the frequencies of rifampicin and isoniazid resistance-conferring mutations has been reported and this could limit the sensitivity of molecular tests in detecting resistance [6, 18]. As such, it is becoming increasingly clear that the performance of rapid molecular tests in TB cannot be extrapolated from one setting to another, but needs to be validated in each geographic setting [23, 24], through systematic surveys to ascertain the mutation profiles and frequencies in geographic regions where the tests are being deployed [6, 18, 20]. It is through such investigations that the performance of new genotypic tests for TB in particular settings can be assessed. DNA sequencing is presumed to be the most appropriate approach for elucidating the patterns and frequencies of resistance conferring mutations in M. tuberculosis, and hence, evaluating the performance of new genotypic tests for TB [6].

Here, we describe the frequencies, patterns and distribution of rifampicin/isoniazid-resistance conferring mutations in MDR-TB isolates from Uganda based on sequence analysis of the rpoB/RRDR, katG and inhA gene promoter. We demonstrate that sequence analysis of those fragments is useful in detecting rifampicin- and isoniazid-resistance in Uganda however, a significant number of MDR-TB isolates lacks known resistance-conferring mutations implying that rapid molecular assays will miss detecting them as resistant to the drug in question. Also, we describe key associations between the circulating M. tuberculosis genotypes and occurrence of rifampicin/isoniazid resistance conferring mutations, providing new insight as to why the predominant TB strain in Uganda is weakly associated with MDR-TB.

Results

Frequency of rifampicin- and isoniazid-resistance conferring mutations

rpoB and rifampicin-resistance.

Sequence analysis of the rpoB/RRDR revealed mutations in six codons namely 588, 531, 526, 516, 513, and 511. The most frequent mutation was Ser531Leu (40%, 18/45), while the second most frequent mutations affected codon 526 (27%, 12/45) with various non-synonymous mutations i.e. His526Asp (6 isolates), His526Tyr (3 isolates), His526Arg (1 isolate), His526Gly (1 isolate), and His526Leu (1 isolate). The third most frequent mutation was Asp516Tyr (9%, 4/45) while the infrequent mutations were Leu511Pro (4%, 2/45), Glu588Gly (2%, 1/45), and Glu513Lys (2%, 1/45), Table 1 and S1 Table.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Frequency and patterns of resistance conferring mutations in M. tuberculosis isolates that were resistant to rifampicin (n = 45) and isoniazid (n = 50).

https://doi.org/10.1371/journal.pone.0198091.t001

Overall, mutations affecting three codons 531, 526, and 516, which have been reported to be responsible for up to 90% of rifampicin-resistance in most settings, accounted for only 76% of the rifampicin-resistant isolates investigated. One isolate was a double mutant as it possessed both Asp516Val and Glu588Gly mutations, S1 Table. Interestingly, 18% (8/45) of the rifampicin-resistant isolates lacked mutations in the rpoB/RRDR, Table 1 and S1 Table. These isolates were subjected to the Genotype MTBDRplus assay as an extra quality control measure but still, we did not detect resistance conferring mutations in the rpoB/RRDR of those isolates.

katG/inhA gene promoter and isoniazid resistance.

Sequence analysis of katG gene revealed Ser315Thr mutation as the most frequent at 76% (38/50) in isoniazid-resistant isolates (Table 1), and two of these isolates possessed both Ser315Thr and Asp406Ala (S1 Table). Furthermore, six mutations were identified in the inhA gene promoter namely C(-15)T (8%, 4/50), T(-8)C (4%, 2/50), A(-255)C (2%, 1/50), G(-116)A (2%, 1/50), A(-7)C (2%, 1/50), and C(-305)T (2%, 1/50). Overall, mutations in katG and the inhA gene promoter that are frequently associated with isoniazid-resistance in most settings [i.e. Ser315Thr and C(-15)T & T(-8)C, respectively] accounted for 88% of isoniazid-resistance, with katG/Ser315Thr alone accounting for 76% of the resistance. Nevertheless, 20% (10/50) of the isoniazid-resistant isolates lacked mutations in katG and inhA gene promoter, Table 1. These isolates were also re-tested with the Genotype MTBDRplus assay as an extra quality control measure and we confirmed absence of known isoniazid-resistance conferring mutations in those isolates. All the rifampicin- and isoniazid-resistant isolates that lacked resistance conferring mutations were MDR however, none of them simultaneously lacked mutations to both drugs, S1 Table.

No resistance conferring mutations in drug susceptible isolates.

All the 40 isolates that were susceptible to rifampicin and isoniazid did not possess resistance conferring mutations in similar genes/loci analysed, implying that drug susceptible TB isolates in Uganda do not harbour mutations associated with resistance to rifampicin and isoniazid. Furthermore, the association between phenotypic rifampicin/isoniazid-resistance and Sanger DNA sequencing data was evaluated to determine the sensitivity of detecting rifampicin/isoniazid-resistance based on identification of high confidence resistance-conferring mutations. For rifampicin-resistance, the sensitivity of sequence analysis of rpoB/RRDR for the three mutations Ser531Leu, His526Tyr and Asp516Tyr was 81% (95%CI: 65.88%-91.40%). For isoniazid-resistance, the sensitivity of sequence analysis of both katG and inhA gene promoter for mutations Ser315Thr & C(-15)T and T(-8)C respectively, was 77% (95%CI: 63.16–87.47%).

Frequency of resistance conferring mutations in the circulating strains.

Of the 18 rifampicin-resistant isolates that had the most frequent high confidence mutation rpoB/Ser531Leu, six (33%, 6/18) were lineage 3/sub-lineage Delhi/CAS and five (28%, 5/18) lineage 4/sub-lineage Uganda; the remaining seven isolates with the mutation rpoB/Ser531Leu were lineage 4/sub-lineage LAM (22%, 4/18) and lineage 2/sub-lineage Beijing (17%, 3/18), Table 2. Additionally, isolates that carried the two other commonly occurring high confidence for rifampicin-resistance (i.e. His526Tyr, Asp516Tyr) were lineage 3/sub-lineage Delhi/CASI, and lineage 4/sub-lineages TUR, URAL, and Haarlem. Though, three lineage 4/sub-lineage Uganda II isolates possessed His526Asp while a double mutant with Asp516Val and Glu588Gly was a Uganda II genotype isolate, Table 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Frequency of rifampicin/isoniazid-resistance conferring mutations in the circulating M. tuberculosis genotypes/sub-lineages in Uganda (2009–2011).

https://doi.org/10.1371/journal.pone.0198091.t002

Furthermore, of the 38 isoniazid-resistant isolates with the prevalent katG/Ser315Thr mutation, 12 (32%) were lineage 4/sub-lineage Uganda, of which seven were Uganda II (18%) and five Uganda I (13%). The remaining katG/Ser315Thr mutations were distributed as follows: 16% (6/38) each in lineage 4/sub-lineage Haarlem and lineage 2/sub-lineage Beijing; 13% (5/38) in lineage 3/sub-lineage Delhi/CAS, and 3% (1/38) each in lineage 4 sub-lineages S, Ghana, URAL, and NEW-1. Additionally, two isolates with katG/Asp406Ala mutation were lineage 4/sub-lineage NEW-1 and lineage 2/sub-lineage Beijing strains, S1 Table. Overall, katG mutations were most frequent in sub-lineage Uganda strains at 30% (12/50), followed by sub-lineages Beijing (12%, 6/50), Haarlem (12%, 6/50), LAM (10%, 5/50), and Delhi/CAS (10%, 5/50). Additionally, three of the four isoniazid-resistant isolates with the inhA gene promoter mutation C(-15)T were sub-lineage Uganda; likewise, one of the two isolates with the mutation T(-8)C was sub-lineage Uganda while the other was sub-lineage Beijing, S1 Table. The rare inhA gene promoter mutations A(-7)C, G(-116)A, & A(-255)C occurred in sub-lineages Beijing and LAM, respectively, Table 2.

Overall, the high confidence mutations for isoniazid-resistance (i.e. Ser315Thr, C(-15)T, T(-8)C) were more frequent than the high confidence mutations for rifampicin-resistance (i.e. Ser531Leu, His526Tyr, Asp516Tyr) in isolates with a genetic background of “M. tuberculosis Uganda” (P = 0.0298), S1 Table, perhaps explaining its weak association with MDR-TB. The other statistically significant associations between occurrence of resistance conferring mutations and the circulating M. tuberculosis genotypes in Uganda were (a) rpoB/Ser531Leu mutation being more prevalent in lineages 2 and 3 combined compared to lineage 4 strains (P = 0.0251), and (b) isoniazid-resistance conferring mutations Ser315Thr, C(-15)T, and C(-8)T being more prevalent in lineage 4 strains compared to lineage 2 strains (P = 0.0418), S1 Table. All the other noticed differences in association between occurrence of resistance-conferring mutations vis-à-vis circulating TB lineages/sub-lineages were not statistically significant, S1 Table.

Discussion

In this study, we have demonstrated that sequence analysis of rpoB/RRDR, katG and inhA gene promoter is useful for the detection of rifampicin- and isoniazid-resistant M. tuberculosis isolates in Uganda. Indeed, all the susceptible isolates investigated lacked resistance conferring mutations implying that drug susceptible M. tuberculosis isolates in Uganda do not harbour mutations associated with resistance to rifampicin and isoniazid, although susceptible strains with these mutations do occur in other settings [6]. Importantly, we have shown that certain M. tuberculosis clinical isolates in Uganda are phenotypically resistant to rifampicin or isoniazid, but lack high confidence resistance-conferring mutations; these isolates do occur in Uganda at significantly high rates implying that rapid molecular tests like the Xpert MTB/RIF and Genotype MTBDRplus will not detect them as resistant. Also, we have shown that a comparatively low frequency of high confidence mutations for rifampicin-resistance in M. tuberculosis sub-lineage Uganda strains that are predominant in Uganda explains the previously reported weak association between MDR-TB and M. tuberculosis Uganda genotype [28, 29].

Among the 45 rifampicin-resistant isolates investigated, Ser531Leu was the most frequent mutation (40%). This mutation was previously reported at 70% prevalence by an earlier study in Uganda [20] that investigated ≤10 rifampicin-resistant isolates, and at 58% prevalence by another study in Uganda that investigated Xpert/MTB RIF data [7]. Although our findings agree with several other studies that reported predominance of Ser531Leu in rifampicin resistant M. tuberculosis isolates (e.g. 67.2% in Georgia [30] and Mexico [31], 58% Pakistan [32], 71% Germany [33], 59% Italy [34], 56% Greece [35], and 43% Japan [36], rates for Ser531Leu in Uganda are somewhat low, and confirm a previous prevalence of 41% (21/51) for this mutation in a study that used whole genome sequencing to investigate transmission of MDR-TB at Mulago National Referral Hospital in Kampala Uganda [37]. Overall, mutations affecting the three codons 531, 526 and 516 combined, which are frequently associated with rifampicin-resistance in most settings, accounted for 76% of the rifampicin resistance in this study. The second most frequent mutation His526Asp, accounted for 13% of the rifampicin resistant isolates. This mutation was previously reported in Uganda [13, 20] though at a lower frequency. Indeed, codon 526 in rpoB/RRDR of rifampicin resistant isolates in this study possessed several non-synonymous mutations i.e. His526Asp (6 isolates), His526Tyr (3 isolates), His526Arg (1 isolate), His526Gly (1 isolate), and His526Leu (1 isolate). Our findings for codon 526 are in agreement with the findings of Clarke et al (2013), who also reported diverse mutations at codon 526 i.e. His526Asp (3 isolates), His526Arg (1 isolate), His526Tyr (2 isolates), and His526Asn (1 isolate) [37] in MDR-TB isolates at Mulago Hospital. His526Tyr is a well-documented high confidence mutation for rifampicin-resistance in several settings, alongside Ser531Leu and Asp516Val [6, 25]. Overall, there is variation in frequencies of mutations at codon 526 (especially His526Tyr vs. His526Asp) in rifampicin resistant M. tuberculosis isolates from various settings, and we have shown that His526Asp is comparatively more frequent in isolates from Uganda. Note, some of the identified mutations at codon 526 e.g. His526Leu had not been reported before in isolates from Uganda.

Furthermore, only four (8.9%, 4/45) rifampicin-resistant isolates possessed Asp516Tyr, the third most frequent high confidence mutation for rifampicin-resistance; again, our findings are comparable with Clarke et al (2013) who reported five (9.8%, 5/51) MDR-TB isolates with mutations at codon 516 at Mulago Hospital [37]. The Asp516Tyr mutation was previously described by other investigators in Uganda [13, 20], also at low frequencies. Three isolates with Asp516Tyr were lineage 4/sub-lineage Haarlem strains, in accordance with previous reports from Spain [38]. Importantly, an isolate with the Asp516Val mutation also had Glu588Gly (double mutant); Glu588Gly has not been reported before in Uganda but was documented elsewhere in a survey that compared INNO-LiPA and MTBDRplus tests for identification of rpoB point mutations [39]. Another lineage 4/sub-lineage Haarlem isolate possessed His526Gly, a unique mutation given that it is generated by substitution of two bases at nucleotide positions 1333 and 1334. This mutation has not been reported in many settings including Uganda though it was previously associated with rifampicin-resistance [40]. Furthermore, two isolates had Leu511Pro, a mutation that was previously reported to occur in rifampicin-resistant M. tuberculosis from Uganda at 8% prevalence [7]. Elsewhere, Leu511Pro was designated a high confidence mutation in rifampicin resistant M. tuberculosis [41], and it has been reported among rifampicin resistant M. tuberculosis isolates in the UK [42].

For isoniazid, only 38 (76%) of the resistant isolates possessed the katG/Ser315Thr mutation, a finding comparable with reports from many settings worldwide [6, 18] including Mulago National Referral Hospital in Kampala i.e. 58.8% (30/51) frequency for katG/Ser315 [37]. katG/Ser315Thr was found frequent in isoniazid resistant isolates in Belarus (100%), China (55%), Honduras (60%), Iran (40%), Romania (80%) [20], Nepal (62.2%) [43] and Poland (72%) [44]. As such, katG/Ser315Thr is a reliable target for rapid detection of isoniazid resistant M. tuberculosis irrespective of geographic location. However, significant variation by geographic region in its distribution has been reported [18]; for instance, of the 808 isoniazid resistant isolates from Japan and Korea, only 43% had katG/Ser315Thr [18]. In contrast, this mutation was observed in 94% of the 751 isoniazid resistant isolates from Eastern Europe. Importantly, sub-Saharan Africa was poorly represented in this analysis and a need for systematic surveys of regional patterns in mutations in the sub-continent was highlighted [18].

Furthermore, an isoniazid resistant isolate in this study had a frame shift mutation resulting from insertion of adenine at position 845 base pairs, which could have inactivated katG. Thus, isoniazid-resistance in this sole isolate with a frame shift mutation was probably due to deficiency of catalase peroxidase enzyme that activates isoniazid [6, 45, 46]. M. tuberculosis clinical isolates with deletion of the katG gene, which is required for catalase-peroxidase activation of isoniazid, were reported previously for two MDR-TB isolates in Uganda [37]. In addition, two isolates possessed katG/Asp406Ala, a mutation that substitutes a polar amino acid Aspartate with an aliphatic one, Alanine, which may lead to expression of a modified catalase peroxidase leading to insufficient isoniazid breakdown. The mutation Asp406Ala in katG has not been described in previous studies of isoniazid-resistant M. tuberculosis isolates in Uganda.

Within the inhA gene promoter, only 4 (8%) isoniazid-resistant isolates possessed the C(-15)T mutation. This finding is comparable with results of a study that reported a prevalence of 5% for this mutation in MDR-TB isolates from Uganda [20]. The C(-15)T mutation is associated with low level resistance to isoniazid [47] and it occurs at lower frequencies compared to katG/Ser315Thr [6, 18]. Further, two isolates possessed T(-8)C, a mutation that is also associated with low level resistance to isoniazid. This mutation was reported previously in MDR-TB isolates from Uganda, Romania, Honduras, Iran, and Belarus, also at low frequencies [20]. Other inhA gene promoter mutations [A(-255)C, G(-116)A, A(-7)C, C(-8)A] occurred at much lower frequencies compared to C(-15)T and T(-8)C and this is consistent with findings from other investigators [18, 40], implying that isoniazid-resistance in Uganda is mostly attributable to the katG/Ser315Thr mutation. Indeed, majority (32, 78%) of the isoniazid-resistant isolates that lacked mutations in inhA gene promoter harboured katG/Ser315Thr, and they could still be detected by the rapid molecular assays as they target mutations both in katG and inhA promoter [47].

An unusual finding in this study was the relatively high prevalence of rifampicin- and isoniazid- resistant M. tuberculosis isolates that lacked resistance conferring mutations in rpoB/RRDR, katG and inhA promoter. However, similar findings were reported from other TB endemic countries [48, 49]. In fact, MDR-TB isolates that are phenotypically resistant to rifampicin (three isolates) and isoniazid (five isolates) but lacking resistance conferring mutations have already been reported in isolates from Uganda [37]. Yet, we have reported higher prevalence of those isolates in Uganda, implying they could prove to be problematic in the management of drug resistant TB patients. Further studies are required to fully characterize the molecular epidemiology of these isolates in Uganda, especially studies employing high throughput sequencing approaches to uncover all the potential resistance conferring mutations. Since the commercial rapid molecular assays for TB target only high confidence mutations to swiftly detect resistance to anti-TB drugs [31, 32, 42, 50], absence of targets is troublesome as it makes it nearly impossible to employ rapid molecular assays to detect certain resistant strains in clinical samples. On the other hand, strains that lacked mutations could harbour resistance conferring mutations in other genes e.g. ahpC, oxyR, kasA, furA and ndh that are associated with isoniazid-resistance [18, 51], or the use of efflux pumps for rifampicin-resistance [20, 25] but we did not investigate these mechanisms. Furthermore, although the resistant isolates that lacked known-resistance conferring mutations were validated as having no mutations, hetero-resistance due to mixed-strain resistant-plus-susceptible TB strains might have confounded the results, as both Sanger-DNA sequencing and rapid molecular assays (i.e. GeneXpert and LPAs) do not detect mixed-strain tuberculosis infections. In theory, mixed-strain tuberculosis infections can negatively impact the performance of rapid molecular tests for TB, especially in samples with hetero-resistant strains. Nevertheless, drug resistant strains lacking known resistance conferring mutations to rifampicin and isoniazid do occur [6], and our detection of these isolates is in agreement with several other global studies that described them [6, 18, 52–54].

Conclusions

Sequence analysis of rpoB/RRDR, katG and inhA gene promoter is useful in the detection of rifampicin/isoniazid-resistant M. tuberculosis isolates in Uganda, and has revealed that drug susceptible isolates in Uganda do not harbour mutations associated with resistance to rifampicin and isoniazid. The most prevalent high confidence mutation for rifampicin-resistance in Uganda is rpoB/Ser531Leu, although its frequency is low compared to estimates from other countries. Accordingly, a comparatively low frequency of high confidence mutations for rifampicin-resistance in M. tuberculosis sub-lineage Uganda strains explains their weak association with MDR-TB. Furthermore, katG/Ser315Thr is the most frequent high confidence mutation for isoniazid-resistance, and its frequency is comparable to estimates from other countries. Overall, a significant number (≤20%) of rifampicin/isoniazid-resistant isolates in Uganda lack high confidence resistance conferring mutations, implying that the commercial rapid molecular assays (e.g. Xpert MTB/RIF, Genotype MTBDRplus) will not detect them as resistant, potentially complicating the control of MDR-TB.

Materials and methods

Supporting information

Acknowledgments

This work was supported in part by the “Basic HIV and HIV/AIDs-Related Diseases Science Training Program for Uganda (BHSTP)” funded by Fogarty International Center (grant# ID43TW009780). The work was also supported through a DELTAS Africa Initiative [grant# 107743/Z/15/Z]. The DELTAS Africa Initiative is an independent funding scheme of the African Academy of Sciences (AAS), Alliance for Accelerating Excellence in Science in Africa (AESA), and supported by the New Partnership for Africa's Development Planning and Coordinating Agency (NEPAD Agency) with funding from the Wellcome Trust (Grant no. 107743) and the UK Government. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

We thank staff at the National Tuberculosis Reference Laboratory/SRL Uganda for technical assistance. For administrative support, we thank Justine Nakibuuka (Uganda-Case Western Research Collaboration), Ivan Mwebaza, Joan Basemera, Carolyne Atugonza, Peter Katumba, Moses Okee, & Geraldine Nalwadda (Makerere University College of Health Sciences); Joshua Mandre & Allen Mukhwana (MUII-plus, Entebbe).