The Hidden Epidemic of Isoniazid-Resistant Tuberculosis in South Africa

Abstract

Rationale

Isoniazid-resistant tuberculosis (Hr-TB) is often overlooked in diagnostic algorithms because of reliance on first-line molecular assays testing only for rifampicin resistance.

Objectives

To determine the prevalence, outcomes, and molecular mechanisms associated with rifampin-susceptible, isoniazid-resistant TB (Hr-TB) in the Eastern Cape, South Africa.

Methods

Between April 2016 and October 2017, sputum samples were collected from patients with rifampin-susceptible TB at baseline and at Weeks 7 and 23 of drug-susceptible TB treatment. We performed isoniazid phenotypic and genotypic drug susceptibility testing, including FluoroTypeMTBDR, Sanger sequencing, targeted next-generation sequencing, and whole-genome sequencing.

Results

We analyzed baseline isolates from 766 patients with rifampin-susceptible TB. Of 89 patients (11.7%) who were found to have Hr-TB, 39 (44%) had canonical katG or inhA promoter mutations; 35 (39%) had noncanonical katG mutations (including 5 with underlying large deletions); 4 (5%) had mutations in other candidate genes associated with isoniazid resistance. For 11 (12.4%), no cause of resistance was found.

Conclusions

Among patients with rifampin-susceptible TB who were diagnosed using first-line molecular TB assays, there is a high prevalence of Hr-TB. Phenotypic drug susceptibility testing remains the gold standard. To improve the performance of genetic-based phenotyping tests, all isoniazid resistance–associated regions should be included, and such tests should have the ability to identify underlying mutations.

Drug-resistant tuberculosis (TB) continues to contribute to significant global morbidity and mortality, and rifampin-susceptible, isoniazid-resistant TB (Hr-TB) may be the most common form of drug-resistant TB among new TB cases worldwide (1). In clinical use since the 1950s, isoniazid remains a critical first-line anti-TB drug. Hr-TB is generally associated with worse treatment outcomes, in high-burden settings and in large surveillance or meta-analysis studies alike (2–5), and is considered a gateway to more difficult-to-treat forms of drug-resistant TB (6). However, some studies note similar outcomes between Hr-TB and pansusceptible TB (7), although drug susceptibility monitoring and individualized treatment have been noted to play a role (8–10). The most widely implemented rapid molecular TB assays have focused on rifampin resistance, resulting in underdetection and consequent mismanagement of Hr-TB in global high-burden settings (11).

Limited laboratory capacity contributes to incomplete data in some national surveillance systems (1, 11), and careful studies (7–9) documenting Hr-TB in high-burden settings are rare. In particular, poor understanding of the range of molecular determinants of Hr-TB contributes to low-quality evidence within current treatment guidelines (10). South Africa has one of the highest TB burdens in the world, with an estimated incidence of 513 per 100,000 people (12). The Eastern Cape Province is one of the worst affected regions in South Africa, with 28% of TB cases coinfected with HIV (13). Hr-TB prevalence in the country as a whole has increased between 2002 and 2018 from 2.7% (95% confidence interval [CI] = 2.2–3.2) to 4.9% (95% CI = 4.1–5.8) (14, 15). Despite this trend, local diagnostic algorithms rely on detection of rifampicin resistance by Xpert Ultra to guide further drug resistance investigations.

In practice, Hr-TB is mainly identified in high-burden settings among patients failing to respond to first-line therapy (approximately 2% of TB cases) (16), typically after months of functional rifampin monotherapy in the continuation phase of treatment (17). Encouragingly, newer first-line rapid molecular TB assays such as the cobas assay (Roche) and the BD MAX (Becton Dickinson) (18), as well as the reflex test Xpert MTB/XDR (Cepheid), which detects specific canonical isoniazid resistance conferring mutations, are planned for countrywide implementation.

To determine the prevalence and the genotypic causes of Hr-TB in a high-burden setting, we analyzed isolates from patients with rifampicin-susceptible TB in the Eastern Cape Province of South Africa between 2016 and 2017. We utilized a combination of phenotypic drug susceptibility testing alongside targeted and whole-genome sequencing (WGS) methods to comprehensively explore the genetic basis of resistance.

Methods

Study Setting

A convenience sample of consenting adult (⩾18 yr) patients with TB attending 10 high-burden clinics in the rural Kouga subdistrict of the Eastern Cape Province were enrolled between April 2016 and October 2017. Patients were eligible for inclusion if they had TB-positive, rifampin-susceptible Xpert MTB/RIF assay results or a rifampin-susceptible TB culture at routine diagnosis. Resistance to other drugs was not considered. Study-specific sputum samples were collected before treatment initiation (baseline), at 7 weeks (end of the intensive phase), and at 23 weeks (end of the continuation phase). All patients received a regimen of isoniazid, rifampicin, pyrazinamide, and ethambutol for 2 months, followed by 4 months of isoniazid and rifampicin. Study results were available after patients had completed treatment and, therefore, did not influence treatment regimens. Clinical information, including HIV status, treatment regimen, and treatment outcome, were extracted from the national electronic TB register ETR.Net (19) (for further details, see the data supplement, including Tables E1 and E2). Ethical approval was obtained from the Stellenbosch University Health Research Ethics Committee (N15/07/058) and the Eastern Cape Department of Health (EC_2015RP9_641).

Phenotypic Drug Susceptibility Testing

Sputum samples were decontaminated using the N-acetyl-L-cysteine–sodium hydroxide method (20). Culture and drug susceptibility testing were initially conducted on 7H10 media at 0.2 mg/L and 1.0 mg/L isoniazid (21). When results were uninterpretable, or when discordance between phenotypic and genotypic resistance was observed, phenotypic susceptibility was repeated in BACTEC MGIT960 at 0.1 mg/L and 1.0 mg/L isoniazid (21). MGIT960 results were interpreted as resistant when the growth in the drug-containing tube reached 100 growth units (GU) before the growth control (GC) reached 400 GU; intermediate when the growth reached 100 GU within 7 days after the GC reached 400 GU; and susceptible when no growth was observed within 7 days of the GC reaching 400 GU (22). When intermediate growth was observed, and no causal mutation could be found by genotyping, the drug-containing tube (enriched with the resistant population) was subjected to repeat phenotypic drug susceptibility testing in MGIT 960 as described earlier, followed by targeted deep sequencing (TDS) or WGS.

Genotypic Investigation

Isolates with either a resistant or intermediate phenotypic isoniazid susceptibility result were subjected to genotypic testing using FluoroType MTBDR (Hain Lifescience), according to the manufacturer’s instructions (23), or Sanger sequencing using published primers for inhA promoter and the codon-315 region (“Fragment 2”) of katG (24). Additional Sanger sequencing of the wider katG gene was conducted in overlapping fragments in cases where canonical mutations were not detected in resistant isolates (for primer sequences, see Table E3 in the data supplement). Where no causal mutation could be found by the aforementioned methods, isolates, sampled from the drug-containing tubes, were subjected to TDS to determine the presence of mutations in otherwise undetectable minor subpopulations using methods that were previously described (25, 26). Regions interrogated include the inhA promoter region, as well as the entire katG gene. Analysis of amplicon sequence data was conducted using the Amplicon Sequencing Analysis Pipeline with single-molecule overlapping read technology (25, 26), to facilitate high confidence detection of minor variants. Where a genotypic correlate of isoniazid resistance continued to remain elusive, we performed WGS to investigate the full genome of resistant isolates. Briefly, after genomic DNA extraction, libraries to be sequenced were prepared using the Illumina DNA Prep Kit according to the manufacturer’s instructions. Libraries were sequenced on a NextSeq500 with 150 bp, V2, paired-end chemistry. PhiX (1%) was included in sequencing runs to facilitate run monitoring. Analysis was conducted using a pipeline of freely available software, as described by Black and colleagues (27). The output was examined for variants in genes associated with isoniazid resistance according to the TBProfiler (v.4.2.0) (28), which takes into account the World Health Organization 2022 mutation catalog (29), including ahpC and fabG1 with their promoter regions, inhA, kasA, furA, and katG. Alignment files were visually inspected using Artemis (v.18.2.0) (30, 31) to determine the presence of subpopulations with large deletions of the katG region. WGS data are available in the European Nucleotide Archive (PRJEB67660).

Strain Typing

Spoligotyping of selected isolates was performed according to the method of Kamerbeek and colleagues (32) to detect the possible reinfection or unmasking of subpopulations between baseline and follow-up samples.

Results

During the study period, 1,120 consenting adult patients with TB were enrolled, of which 47 were excluded (22 were not matched to ETR, 4 were rifampin-resistant, 21 had no baseline sample available, and 39 were duplicate entries). In total, 993 patients met the inclusion criteria, of which 766 (77.1%; 95% CI = 74.5–79.8) provided culturable Mycobacterium tuberculosis isolates (Figure 1). In addition, 688 follow-up specimens were collected, of which 124 (18.0%) were culture positive, including 116 Week-7 isolates and eight Week-23 isolates; only one patient provided culture-positive samples at all three time points. Note that we use the term “culture positive” in reference to decontaminated study isolates from which M. tuberculosis was successfully cultured, which may not correlate with the concurrent routine diagnosis.

Figure 1.

Sample inclusion process and retention in the study. Isoniazid susceptibility profiles are based on phenotypic testing. *Follow-up samples were not available from all patients. ^†Includes an rpoB mutant. MGIT = mycobacterial growth indicator tube; RIF = rifampicin; ZN = Ziehl-Neelsen staining.

Phenotypic and Genotypic Drug Susceptibility Testing

Baseline Isolates

Results of the 766 culture-positive baseline isolates available for drug susceptibility testing are presented (see Figure E1). Phenotypic testing on 7H10 media showed isoniazid resistance in 90 (11.7%) isolates, susceptibility in 645 (84.2%) isolates, and an ambiguous result (growth on the drug-containing media but less than the 1:100 control) in 31 (4.0%) isolates. Seventeen isolates were not tested further. Initial results were supplemented by MGIT 960 testing where ambiguous results were observed or where standard molecular tests failed to detect canonical mutations. This resulted in reclassification as isoniazid susceptible (n = 15) or confirming isoniazid resistance (n = 64). Fluorotype MTBDR or Sanger sequencing of the canonical regions confirmed resistance in an additional 25 isolates; thus, the total number of confirmed resistant isolates was 89 (11.6%; 95% CI = 9.4–14.1; Figure 2). Genotypic investigations ascribed isoniazid resistance (Table 1) to canonical katG and/or inhA promoter mutations (n = 39; 43.8%); noncanonical katG mutations (n = 30; 33.7%); underlying large (hetero-)deletions, including katG (n = 5; 5.6%) (see Figure E2); or mutations in other candidate isoniazid resistance–causing genes (n = 4; 4.5%), including efpA, fabG1, and furA (29). We were unable to determine a cause of resistance in 11 (12.4%) isolates.

Figure 2.

General isoniazid (INH) resistance testing workflow and final results. Isolates (n = 766) were subjected to phenotypic INH susceptibility testing using 7H10 media. Isolates with ambiguous results or lacking canonical INH resistance–causing mutations were subjected to further testing using MGIT 960. INH-resistant isolates were investigated for the genotypic cause of resistance using FluoroType MTBDR, Sanger sequencing, targeted deep sequencing, and whole-genome sequencing, terminating the workflow when a plausible cause of resistance was found. (For more details, see Figure E1.)

Table 1.

Putative isoniazid resistance-conferring mutations detected in baseline isolates and method of detection

Mutation	Total Detected	FluoroType MTBDR/Sanger Sequencing	TDS*	WGS†
Canonical mutations (n = 39, 43.8%)
inhA promoter c-15t	11	10	1	—
katG 315AAC	1	1	—	—
katG 315ACC	24	14	10	—
katG 315AAG	1	0	1	—
katG 315AGA	1	0	1	—
katG 315ACA‡	1	1	—	—
Noncanonical mutations (n = 39, 43.8%)
Noncanonical katG	30	5	5	20
efpA R119H§	1	—	—	1
fabG1 R25L§	1	—	—	1
furA 148* > G §	1	—	—	1
furA 166 syn§	1	—	—	1
Heterodeletion including katG	5	—	—	5
No mutation founde (n = 11, 12.4%)ǁ	11	N/A	10	1
Total	89	—	—	—

Definition of abbreviations: N/A = not applicable; TDS = targeted deep sequencing; WGS = whole-genome sequencing.

^*TDS was only conducted if no mutation was found on FluoroType MTBDR or Sanger sequencing.

^† WGS was only conducted if no mutation was found by FluoroType MTBDR, Sanger sequencing, or TDS.

^‡ WGS performed on the enriched culture also detected an rpoB S450L mutation.

^§ Genomic regions that were not interrogated by FluoroType MTBDR, Sanger sequencing, or TDS.

^ǁ Wild-type by TDS or WGS, or no results obtained.

Evidence of heteroresistance was demonstrated in baseline isoniazid-resistant samples by phenotypic methods (n = 41, intermediate MGIT results) or genotypic methods (n = 39, mutations only evident by TDS or WGS of enriched culture).

Follow-up Isolates

Among patients with Hr-TB at baseline, 15/54 (27.7%) and 1/18 (5.6%) study isolates were culture positive at Weeks 7 and 23, respectively; among patients with isoniazid-susceptible TB at baseline, 101/421 (24.0%) and 7/195 (3.6%) study isolates were culture positive at Weeks 7 and 23. Phenotypic testing of follow-up isolates in the isoniazid-susceptible group identified new isoniazid resistance in 9/108 (8.3%) isolates (7 at Week 7 and 2 at Week 23) from unique patients (Figure 1). Causal mutations were identified in only six follow-up isolates (Table 2).

Table 2.

Isoniazid resistance-associated mutations identified in six follow-up samples

Baseline			7-wk Follow-up			23-wk Follow-up
Isoniazid Phenotype	Genotype	Detection Method	Isoniazid Phenotype	Genotype	Detection Method	Isoniazid Phenotype	Genotype	Detection Method
S	ND	ND	R	rpoB D516V, katG S315T, and inhA prom t-17 g	FluoroType MTBDR	No sample	ND	ND
R	Noncanonical katG	WGS	R	Noncanonical katG	WGS	No sample	ND	ND
R	WT	WGS	R	rpoB D516V, katG S315T, and inhA prom t-8c	FluoroType MTBDR	No sample	ND	ND
R	inhA promoter c-15t	FluoroType MTBDR	R	inhA prom c-15t	FluoroType MTBDR	No sample	ND	ND
S	ND	ND	R	Noncanonical katG	WGS	Contaminated	ND	ND
S	ND	ND	No culture	ND	ND	R	Heterodeletion of katG	WGS

Definition of abbreviations: ND = not determined; R = resistant; S = susceptible; WT = wild-type; WGS = whole-genome sequencing.

Three follow-up isolates gained rifampin resistance. One 23-week follow-up isolate had a 6-bp deletion (codons 517–518) and was not associated with isoniazid resistance (rifampin monoresistance). In the remaining two 7-week follow-up isolates (one baseline resistant and one baseline susceptible), the mutation patterns (a combination of inhA promoter c-17t or t-8a, katG S315T, and rpoB D516V mutation) were indicative of multidrug-resistant (MDR) atypical Beijing or Latin American MDR strains, which are endemic to the region (33–35). Spoligotyping confirmed an S-family strain (pansusceptible) at baseline and a Beijing family strain at follow-up in one case, whereas the other had a Beijing strain at baseline and at follow-up.

In 9 patients, isoniazid resistance detected in the baseline isolates reverted to isoniazid susceptibility in the follow-up isolates. In only one of these patients, different spoligotypes were present at baseline (MANU2) and follow-up (Beijing). Six of these baseline isolates had an intermediate isoniazid susceptibility phenotype suggestive of heteroresistance.

Strain Typing

Spoligotyping data were available for isolates at baseline (n = 746), 7 weeks (n = 121), and 23 weeks (n = 8). Evidence of mixed infections were apparent in 33 isolates at baseline, six isolates at 7 weeks, and one at 23 weeks, in a total of 38 patients (5.1%). Spoligotype signatures differed across sampling time points for 24/129 (18.6%) patients.

Discussion

Our prospective observational study in a high TB-burden province of South Africa demonstrated a high (11.6%) prevalence of Hr-TB. Unexpectedly, the majority of putative resistance mutations (71.9%) were not detected by standard molecular tests, because of a combination of mutations outside of the canonical regions and canonical mutations present below the detection limit of standard tests.

Isoniazid remains a key bactericidal component of first-line TB treatment regimens. Commercial TB diagnostic assays have traditionally focused on rifampin resistance, with Hr-TB consequently much less understood, although estimated to have the highest global prevalence of any M. tuberculosis resistance pattern (1). However, observational studies and meta-analyses suggest that Hr-TB is associated with an increased risk of poor treatment outcome and/or additional acquired drug resistance (2, 5, 7–9), and World Health Organization guidelines call for regimens that replace isoniazid with levofloxacin (36, 37). In many settings, Hr-TB is only detected when poor treatment response prompts additional drug susceptibility testing (38–40). We demonstrate an Hr-TB prevalence of 11.6%, almost twice that reported in the South African national drug resistance survey of 2012–2014 for the same region (6.4%) (14), thus clearly highlighting a growing burden of drug resistance.

To evaluate the utility of a targeted molecular diagnostic approach for rapid identification of Hr-TB at diagnosis, we analyzed the hotspot regions of the M. tuberculosis genome (katG 315 and inhA promoter region). Conventionally, mutations in these regions confer isoniazid resistance among approximately 90% of isoniazid-resistant clinical isolates (41). Surprisingly, such canonical resistance conferring mutations were detected in fewer than half of phenotypically isoniazid-resistant isolates. Further, in nine (10.1% of confirmed isoniazid-resistant) baseline isolates, katG or inhA promoter mutations were found only by using next-generation sequencing, suggesting that the mutant population was present below the detection limit of FluoroType MTBDR and Sanger sequencing. In five isolates, the cause of resistance was determined to be a large deletion including the entire katG gene, present in a subpopulation (i.e., heterodeletion; see Figure E2) that cannot easily be detected by methods relying on polymerase chain reaction amplification of a targeted region.

Culture-positive follow-up study specimens were only available from 123/766 (16%) patients, with only one patient submitting culture-positive samples at both follow-up time points, suggesting culture conversion during treatment, as expected. However, 8 patients remained culture positive at 23 weeks, suggesting slow response to antibiotic therapy in these few patients.

Acquisition of rifampin resistance was limited to three patients, one of which was not associated with isoniazid resistance. With regard to the other two patients, one was progressing from Hr-TB (with katG S315T) to MDR and the other, from pansusceptible to MDR; the latter is likely a result of reinfection or unmasking of underlying populations, given patterns of resistance that are associated with endemic MDR strains (34, 35). The combination of inhA promoter c-17t, katG S315T, and rpoB D516V is highly specific of an atypical Beijing strain endemic to the region, suggesting reinfection rather than acquisition of resistance, which was confirmed by spoligotyping. Similarly, the combination of the inhA promoter t-8a, katG S315T, and rpoB D516V is characteristic of the LAM family of strains (42). The presence of Beijing strain types at both baseline (katG mutation present) and follow-up (three mutations present) suggests a mixed infection that was imperfectly delineated in our analysis.

In this study, we also observed the counterintuitive loss of isoniazid resistance during treatment in nine follow-up isolates. A review of baseline MGIT results showed that six isolates had an intermediate isoniazid susceptibility, likely reflecting heteroresistance. One explanation is that the Hr-TB strain was effectively eliminated by treatment, leading to an isoniazid-susceptible phenotype. This implies that the standard first-line anti-TB drugs was sufficient in these patients to reduce the proportion of isoniazid-resistant bacilli to below levels that are detectable with phenotypic testing. However, the observed decrease or increase in isoniazid susceptibility may also be explained by the sampling of different populations of mycobacteria from different lesions within the lung (43).

Together, these results suggest the presence of heteroresistance in several patients, as shown by ambiguous 7H10 or intermediate MGIT results, changing spoligotypes across serial isolates, heterodeletions, and detection of minority variants by TDS. Heteroresistance may be the result of unmasking existing isoniazid resistance, early acquisition of isoniazid resistance, or reinfection with an already isoniazid-resistant strain. All three mechanisms have previously been described for rifampin-resistant and MDR TB in studies that focused on treatment failure (44–46). The presence of mixed infections are further concerning, as it has been documented to increase the risk of poor outcomes (47). These findings demonstrate that, at least, molecular tests need to be expanded to include additional targets. However, persistent lack of knowledge of all resistance markers, and the difficulty of detecting heteroresistance by most molecular methods, suggest that culture-based drug susceptibility testing should be included in diagnostic algorithms to maintain optimal detection of Hr-TB.

Although our study contributes important epidemiological information relating to the understudied rural Eastern Cape, we note some limitations. First, patients were recruited on the basis of the convenience of clinic staff, constraining the number of samples available for analysis and resulting in possible selection biases. Second, many clinical data fields were insufficient for analyses, especially those involving treatment outcomes, including the association of isoniazid preventative therapy with outcomes and comparisons of outcomes in INH-susceptible and INH-resistant groups. Third, specimens were shipped for analysis approximately every 2 weeks, which could have resulted in loss of viability in a portion of them.

In conclusion, programmatic reliance on first-line molecular tests with a limited scope of gene regions interrogated has masked the extent of Hr-TB prevalence, resulting in a hidden epidemic of Hr-TB, potentially exacerbating the MDR-TB epidemic. Hr-TB prevalence in the Eastern Cape was nearly double that documented only 4 to 5 years prior, calling for improved surveillance. However, fewer than half of the isolates harbored a canonical resistance–conferring mutation. Future molecular TB assays should ideally incorporate high-impact noncanonical isoniazid resistance–conferring regions and have enhanced lower limits of detection for heteroresistant subpopulations. TDS may partially satisfy this need for a more comprehensive cover of potential resistance markers, with greater potential to detect underlying populations than most other molecular tests. However, the labor-intensive nature of the technique and the associated high cost is prohibitive for implementation in high-burden, low-income settings.