Genomic markers of drug resistance in Mycobacterium tuberculosis populations with minority variants

ABSTRACT

Minority variants of Mycobacterium tuberculosis harboring mutations conferring resistance can become dominant populations during tuberculosis (TB) treatment, leading to treatment failure. Our understanding of drug-resistant within-host subpopulations and the frequency of resistance-conferring mutations in minority variants remains limited. M. tuberculosis sequences recovered from liquid cultures of culture-confirmed TB cases notified between January 2017 and December 2021 in New South Wales, Australia were examined. Potential drug resistance-conferring minority variants were identified using LoFreq, and mixed populations of different M. tuberculosis strains (≥100 SNPs apart) were examined using QuantTB. A total of 1831 routinely sequenced M. tuberculosis strains were included in the analysis. Drug resistance-conferring minority variants were detected in 3.5% (65/1831) of sequenced cultures; 84.6% (55/65) had majority strains that were drug susceptible and 15.4% (10/65) had majority strains that were drug resistant. Minority variants with high-confidence drug resistance-conferring mutations were 1.5 times more common when the majority strains were drug resistant. Mixed M. tuberculosis strain populations were documented in 10.0% (183/1831) of specimens. Minority variants with high-confidence drug resistance-conferring mutations were more frequently detected in mixed M. tuberculosis strain populations (2.7%, 5/183) than in single strain populations (0.6%, 10/1648; P = 0.01). Drug-resistant minority variants require monitoring in settings that implement routine M. tuberculosis sequencing. The frequency with which drug-resistant minority variants are detected is likely influenced by pre-culture requirement. Culture-independent sequencing methods should provide a more accurate reflection of drug-resistant subpopulations.

INTRODUCTION

The emergence and spread of drug-resistant (DR) Mycobacterium tuberculosis strains pose a major global public health challenge (1, 2). The World Health Organization (WHO) estimated that 450,000 new tuberculosis (TB) cases had rifampicin-resistant (RR) or multidrug-resistant (MDR) TB in 2021, although only 166,991 (37.1%) were formally reported (1). Undetected drug-resistant TB (DR-TB) is a major concern since it compromises patient care and facilitates DR-TB transmission (1).

The presence of drug-resistant M. tuberculosis strains as minority subpopulations in diagnostic samples has been documented (3 – 5). These minority subpopulations may represent either “mixed infection” (simultaneous co-infection of the host with more than one strain), or minority variants of the same strain, with heteroresistance defined as concomitant infection with both drug-resistant and drug-susceptible variants (either as mixed strain infections or minority variants). The presence of minority variants with drug resistance-conferring mutations could promote resistance amplification if patients receive a sub-optimal treatment (6), compromising patient outcomes and potentially facilitating the spread of these strains within the community (7 – 9).

The detection of drug-resistant M. tuberculosis can be achieved using either culture-based phenotypic drug-susceptibility testing (pDST) or genotypic drug-resistance prediction (gDST), which relies on rapid molecular tests or whole genome sequencing (WGS) to detect mutations associated with drug resistance (1, 10). Culture-based pDST can potentially detect drug-resistant subpopulations that are present at more than 1% of the cultured population (9, 11). Several rapid gDST methods have been endorsed by the WHO to detect drug resistance directly from clinical specimens, without initial culture. Both Xpert MTB/RIF and MTB/RIF Ultra are reported to be able to detect rifampicin resistance if present at more than 20% of M. tuberculosis population in the sample, while GenoTypeMTBDRplusv2.0 (LPA-Hain) detects isoniazid or rifampicin resistance if present in more than 5% of bacteria at the site of infection (4, 12, 13). All molecular tests are limited by the fact that they can only assess a small number of pre-determined mutations, which is not inclusive of many mutations associated with drug resistance listed in the WHO Catalogue (14).

WGS studies have demonstrated the presence of drug-resistant subpopulations in clinical samples (15, 16), confirming results from earlier molecular strain-typing methods (3, 4). However, WGS analysis of drug-resistance mutations usually focuses on the majority population present within the diagnostic specimen (17, 18), but recent validation of genome-wide variant calling tools for the detection of minority resistance-conferring variants has opened new opportunities to identify mixed strain populations and minority variants (18, 19). Specifically, LoFreq allows detection of minority variants harboring drug resistance-conferring mutations that may not be recognized by standard WGS data analysis pipelines (18, 19). In instances with multi-strain infection, QuantTB (20) quantifies the estimated mixture of constituent strains based on a strain specific single-nucleotide polymorphism (SNP) reference database and the number of SNP differences between strains.

This study leveraged routine prospective sequencing of all culture-confirmed TB cases in New South Wales (NSW), Australia, between 2017 and 2021. It aimed to investigate the incidence with which minority variants and mixed strain infections harbored drug resistance-conferring mutations not detected by standard bioinformatic pipelines. Accurate detection of minority variants with associated drug resistance is important for optimal clinical care and public health surveillance.

MATERIALS AND METHODS

Genomic sequences of M. tuberculosis cultures from TB cases notified between 1 January 2017 and 31 December 2021 in NSW, Australia, were examined together with their respective pDST results (Table S1).

All positive M. tuberculosis cultures were cultured either in BACTEC MGIT 960 (Becton Dickinson) or on Lowenstein Jensen (LJ) slopes. Phenotypic DST was conducted using the modified microdilution method in the BACTEC MGIT 960 system with WHO recommended critical concentrations as previously described (21). All M. tuberculosis isolates were prospectively sequenced in-house. WGS libraries were prepared with DNA extracted from MGIT or LJ cultures using the Nextera XT kit (Illumina, San Diego, California) as previously described (21). Sequencing was performed on an IIlumina NextSeq500 (Illumina, San Diego, California) instrument using 2 × 150bp paired-end chemistry. The quality of short reads was checked using FastQC v0.1 (https://github.com/s-andrews/FastQC). Sequenced reads were trimmed using Trimmomatic v0.36 (https://github.com/usadellab/Trimmomatic).

Detection of minority variants with drug resistance

The reference genome used for the genomic analysis was M. tuberculosis strain H37Rv (NCBI GenBank accession: NC_000962.3). Minority variants were defined as those with an allelic frequency of 3%–49% detected by LoFreq v2.1.1 (18, 19). Majority variants were defined as nucleobases with an allelic frequency of 50% or more, although 75% is frequently used at the cutoff (21, 22). We focused on the drug resistance-conferring minority variants in rpoB, katG, fabG1, inhA, pncA, panD, embB, embA, gyrA, gyrB, and rrs associated with resistance to first-line (isoniazid, rifampicin, pyrazinamide, and ethambutol) and key second-line TB drugs (fluoroquinolones and the injectables such as amikacin, capreomycin, and kanamycin). Drug resistance-conferring mutations in minority variants in genes Rv1979v, mmpL5, Rv0678, pepQ, mmpS5, fgd1, fbiAB, fbiC, ddn, Rv2983, rrl, and rplC reported in association with resistance to bedaquiline, pretomanid, and linezolid, now recommended for a short-course treatment by WHO (23, 24), were also examined.

Trimmed reads were aligned to the reference genome of M. tuberculosis strain H37Rv using BWA-MEM v0.7.17 (25). Potential drug resistance-conferring minority variants were called using LoFreq v2.1.1 (18, 19) from BAM files produced by BWA-MEM. The indel quality score was inserted into the BAM files and potential minority variants (SNPs and indels) were called. We used the strand bias filter to remove any minority variants with a read depth of less than five reads, an allelic frequency below 3% and a maximum strain variation over 0.01 (18, 19). Potential minority variants were annotated using SnpEFF v4.3 (26).

Drug resistance verification and classification

The potential drug resistance-conferring minority variants called by LoFreq were manually cross-checked with the 2021 WHO Catalogue (23, 24). Mutations that were classified as either in group 1 (Associated w R), group 2 (Associated w R - Interim), or group 3 (Uncertain significance) in the 2021 WHO Catalogue (23, 24) were analyzed. We defined “Associated w R” as high confidence resistance-associated variants, and “Associated w R - Interim” as interim confidence resistance-associated variants. The high confidence drug resistance mutations-associated minority variants were manually reviewed by direct visualization of the read alignments using Geneious Prime 2023.0.4.

Detection of mixed M. tuberculosis strain populations

Mixed population infection was defined as the presence of more than one strain of M. tuberculosis during the same disease episode. We detected mixed populations of M. tuberculosis in WGS data using QuantTB v1.0 (20) with default settings (>100 SNP differences between strains) and defined the population that was present in the highest proportion as the majority population.

Statistical analysis

Prism GraphPad v9.4.1 was used for statistical analyses. Differences in drug resistance between minority variants and mixed populations were assessed using the Fisher’s exact test with a significance level of P < 0.05.

RESULTS

A total of 1,831 M. tuberculosis strains sequenced during routine WGS were included in the analysis, which represent the majority of TB cases identified in NSW during the 5-year study period (Fig. 1; Table S1). All four major M. tuberculosis strain lineages were represented (Lineage 1 n = 571, Lineage 2 n = 544, Lineage 3 n = 285, and Lineage 4 n = 431). In this study, isolates were considered drug resistant if resistance was detected either on routine pDST, or if WHO-defined drug resistance-associated mutations were identified by WGS (if pDST was absent or isolates were reported susceptible by pDST).

Fig 1.

Detection of drug resistance-conferring minority variants in M. tuberculosis isolates using LoFreq. DR: drug resistant (majority variants resistant to any first-line TB drug); DS: drug susceptible (majority variants susceptible to all first-line TB drugs); DR-MVs: drug-resistant minority variants; NSW: New South Wales; PCR: M. tuberculosis-specific polymerase chain reaction; TB: tuberculosis. *Group 1 (“Assoc W R”) drug-resistant variants, as in 2021 WHO Catalogue (23, 24). **Group 2 (“Assoc W R - Interim”) drug-resistant variants, as in 2021 WHO Catalogue (23, 24).

Detection of drug resistance-conferring minority variants

Drug resistance-conferring minority variants were detected in 3.5% (65/1,831) of sequenced strains (Fig. 1); 70.8% of the 1,831 sequenced strains were cultured from respiratory specimens with roughly similar proportions of resistance-conferring minority variants identified in respiratory and non-respiratory specimens (Fig. S1a). Of the minority variants with drug-resistance mutations, 15/65 (23.1%) had high confidence mutations (Table 1) and 50/65 (76.9%) had interim confidence mutations (Table S2). In addition, we identified mutations in minority variants associated with resistance to BPaLM drugs, as well as those classified as group 3 (uncertain significance) in the 2021 WHO Catalogue (Table S3). All drug-resistance associated minority variants identified were supported by at least four overlapping reads with at least one forward and one reverse read; strand bias was less than 0.001. The allelic frequency of minority variants associated with drug resistance varied from 3.3% to 23.0% (Fig. 2). Minority variants with drug-resistance mutations were found in all major lineages (Fig. S1b), but more commonly associated with Lineage 1 compared to other lineages; however, this was not statistically significant (P = 0.08).

TABLE 1.

Overview of “high-confidence drug resistance-conferring mutations” ^a identified in M. tuberculosis minority variants, including mixed strain infections

Strain ID	Source	Smear microscopy	Lineage	gDST profile	pDST profile	Mixed strain population b (Y/N)
18-1773-0435	Pus mediastinal	Negative	Lineage 1	S	INH-R	Y
17-3391-0281	Lavage	Negative	Lineage 1	S	S	Y
17-3391-0008	Fluid - neck	Negative	Lineage 1	S	S	Y
17-3391-0083	Sputum	Negative	Lineage 2	fabG1 -15 c > t	INH-R	Y
18-1773-0007	Sputum	Negative	Lineage 3	S	S	Y
20-002-0186	Sputum	Negative	Lineage 1	fabG1 -15 c > t	INH-R	N
20-002-0197D1	Sputum	Positive	Lineage 1	S	S	N
21-002-0305	Pleura	Negative	Lineage 1	S	S	N
21-002-0237	Tissue - lung	Negative	Lineage 1	S	S	N
17-3391-0013	Lymph node	Negative	Lineage 2	S	S	N
17-3391-0018	Sputum	Negative	Lineage 2	S	S	N
21-002-0252D1	Sputum	Negative	Lineage 2	S	S	N
18-1773-0451	Uncertain	Negative	Lineage 4	S	S	N
20-002-0411	Washing	Negative	Lineage 4	S	S	N
21-002-0328	Pleura	Negative	Lineage 4	S	S	N

^a Group 1 (“Assoc W R”) drug-resistant variants, as in 2021 WHO Catalogue (23, 24).

^b Mixed strain population detected by QuantTB (20).

Fig 2.

Allelic frequency of drug resistance-conferring mutations detected in identified minority M. tuberculosis variants. DR: drug resistant (majority variants resistant to any first-line TB drugs); DS: drug susceptible (majority variants susceptible to all first-line TB drugs); DR-MVs: drug-resistant minority variants; LPA: line probe assay; pDST: phenotypic drug-susceptibility testing; WGS: whole genome sequencing. Horizontal red line (at 3%) and dashed lines indicate the limit of detection of minority variants for each method.

The majority of the 65 specimens with drug-resistant minority variants (84.6%, 55/65) had majority variants that were phenotypically and genomically drug susceptible. The remaining 10 (15.4%, 10/65) had majority variants with drug resistance along with minority variants with additional resistance (Fig. 1 and 2). High-confidence drug resistance-conferring mutations in minority variants were 1.5 times more common when majority variants were drug resistant (1.2%, 3/245) compared to drug susceptible (0.8%, 12/1,586; P = 0.44).

Identification of mixed M. tuberculosis strain populations

We identified mixed M. tuberculosis strain populations in 183 (10.0%, 183/1,831) specimens (Fig. 3); 177 specimens were determined to be mixed with two distinct populations (96.7%, 177/183), while six specimens showed signatures of three distinct populations (3.3%, 6/183) (Table S1). Specimens with mixed M. tuberculosis strain populations were distributed across all four major lineages, but significantly more common in Lineage 3 compared to other lineages (Fig. S2a) (P = 0.0002). The majority of specimens with mixed populations were cultured from respiratory specimens (124/183, 67.8%); 32.2% of non-respiratory specimens had mixed populations (Fig. S2b).

Fig 3.

Identification of mixed strain populations in M. tuberculosis isolates using QuantTB. DR: drug resistant (majority variants resistant to any first-line TB drugs); DS: drug susceptible (majority variants susceptible to all first-line TB drugs); Mixed: mixed strain population; NSW: New South Wales; PCR: M. tuberculosis-specific polymerase chain reaction; Single: single strain population; TB: tuberculosis.

Mixed population infections were 42.1% more common in strains where the majority population was drug resistant (13.5%, 33/245) compared to drug susceptible (9.5%, 150/1,586; P = 0.07). The combined results of LoFreq and QuantTB, to determine the proportion of strains with drug-resistant minority variants that represent mixed populations, are represented in Fig. 4. Drug-resistant minority variants were documented in 6.0% (11/183) of specimens with mixed populations, regardless of whether the dominant strain was considered drug resistant or susceptible (Fig. 4 and 5). Specifically, in 0.8% (2/245) where the majority strain was considered drug resistant and in 0.6% (9/1,586) where the majority strain was considered drug susceptible, while 3.3% (54/1,648) of specimens with a single strain population harbored drug-resistant minority variants.

Fig 4.

Combined assessment of drug-resistant minority variants and mixed M. tuberculosis strain populations detected in specimens with majority drug-resistant or drug-susceptible strains. DR: drug resistant (majority variants resistant to any first-line TB drugs); DS: drug susceptible (majority variants susceptible to all first-line TB drugs); DR-MVs: drug-resistant minority variants; MDR: multidrug resistant; Mixed: mixed strain population; NSW: New South Wales; PCR: M. tuberculosis-specific polymerase chain reaction; Single: single strain population; TB: tuberculosis. *Group 1 (“Assoc W R”) drug-resistant variants, as in 2021 WHO Catalogue (23, 24). **Group 2 (“Assoc W R - Interim”) drug-resistant variants, as in 2021 WHO Catalogue (23, 24).

Fig 5.

Overview of M. tuberculosis drug-resistant minority variants and mixed strain populations detected. DR: drug resistant (majority variants resistant to any first-line TB drugs); DS: drug susceptible (majority variants susceptible to all first-line TB drugs); DR-MVs: drug-resistant minority variants; Mixed: mixed strain population with/without drug-resistance mutations; Single: single strain population. Numbers (N) in the brackets next to legend indicate the total number of isolates per legend.

Assignment of drug resistance-conferring minority variants

The assignment of high-confidence drug-resistance mutations detected in minority variants in single strain population specimens is summarized in Table 2 (interim confidence mutations in Table S2). One specimen considered phenotypically and genotypically isoniazid mono-resistant (majority variant fabG1 -15 c > t) also had a rpoB_H445N mutation detected as a minority variant at 11.4% allelic frequency, indicating potential for selection of this MDR strain with inappropriate treatment (Table 2). Minority variants with high-confidence drug resistance-conferring mutations were more frequently detected in specimens with mixed strain populations (2.7%, 5/183) than in those with a single strain population (0.6%, 10/1,648) (P = 0.01). Due to limitations of the bioinformatics software and short-read sequencing technology used, minority drug-resistance mutations were unable to be assigned to either the minority or majority population within specimens with mixed population (Table 3).

TABLE 2.

“High-confidence drug resistance-conferring mutations” ^a identified in minority variants from specimens with a single M. tuberculosis strain (N = 1,648)

Strain ID	pDST profile	Majority drug-resistant variant	Minority drug-resistant variant						Predicted MDR c (Y/N)
			Mutation	POS	Nucleotide change	Read depth	AF (%)	Predicted DR b
20-002-0186	INH-R, RIF-S	fabG1 -15 c > t	rpoB_H445N	761139	C to A	70	11.4	RR	Y
18-1773-0451	S	S	rpoB_H445N	761139	C to A	83	12	RR	N
17-3391-0013	S	S	rpoB_H445N	761139	C to A	97	6.2	RR	N
20-002-0411	S	S	rpoB_H445N	761139	C to A	73	5.5	RR	N
20-002-0197D1	S	S	katG_S315T	2155168	C to G	57	8.8	INH-R	N
21-002-0305	S	S	katG_W328L	2155129	C to A	256	3.5	INH-R	N
21-002-0328	S	S	katG_W328L	2155129	C to A	135	3.7	INH-R	N
21-002-0252D1	S	S	katG_W328L	2155129	C to A	168	4.2	INH-R	N
21-002-0252D1	S	S	pncA_G97C	2288953	C to A	175	3.4	PZA-R	N
17-3391-0018	S	S	pncA_V180G	2288703	A to C	91	11	PZA-R	N
21-002-0237	S	S	pncA_A146T	2288806	C to T	255	6.7	PZA-R	N

^a Group 1 (“Assoc W R”) drug-resistant variants, as in 2021 WHO Catalogue (23, 24).

^b Based on the detection of mutations classified as group 1 (associated with resistance) in the WHO catalogue (23, 24).

^c Minority variants with genotypic rifampicin resistance detected in isoniazid resistant strain.

TABLE 3.

“High-confidence drug resistance-conferring mutations” ^a identified in minority variants from specimens with more than one M. tuberculosis strain (N = 183)

Strain ID	pDST profile	Majority drug-resistant variant	Mixed strain population ratio	Minority drug-resistant variant
				Mutation	POS	Nucleotide change	Read depth	AF (%)
17-3391-0083	INH-R, RIF-S	fabG1 -15 c > t	63%:37%	rpoB_H445N	761139	C to A	222	4.1
18-1773-0435	INH-R	S b	81.9%:18.1%	katG_S315T	2155168	C to G	62	11.3
18-1773-0007	S	S	80.4%:19.6%	katG_S315T	2155168	C to G	89	20.2
17-3391-0281	S	S	57.9%:42.1%	pncA_V128G	2288859	A to C	244	6.6
17-3391-0008	S	S	51.7%:48.3%	embB_D354A	4247574	A to C	140	8.6

^a Group 1 (“Assoc W R”) drug-resistant variants, as in 2021 WHO Catalogue (23, 24).

^b No isoniazid-associated resistance mutation detected.

DISCUSSION

This study is the first to quantify the frequency with which minority variants that harbor mutations conferring drug resistance are observed in routinely sequenced isolates from a low TB incidence setting. It is also the first to differentiate same strain minority variants from those representing mixed M. tuberculosis strain populations. From a programmatic perspective, the assessment of drug-resistant minority variants improves clinical risk management and should provide patient benefits, especially in settings where disease relapse is a concern.

Drug resistance-conferring minority variants are not routinely interrogated during WGS-based gDST, but may be detected if pDST is performed, leading to discrepancies between WGS and phenotypic data. Our analysis of WGS data from culture-confirmed clinical M. tuberculosis strains confirmed previous findings that minority strains carrying high-confidence drug-resistance mutations (such as katG_S315T at 11.3% allelic frequency) can account for occasional discrepancies between pDST and gDST (8, 27). Routinely assessing for minority variants with drug resistance-conferring mutations in clinical M. tuberculosis cultured specimens would enhance drug-resistance surveillance and have particular clinical relevance to prevent drug-resistance amplification and disease relapse.

Minority variants with drug resistance-conferring mutations, especially in strains demonstrating phenotypic drug resistance, may potentially develop additional drug resistance if exposed to sub-optimal treatment. Therefore, the presence of these minority variants provides an early risk indication for treatment failure (7, 8, 17, 28). The presence of minority variants with additional resistance to rifampicin or isoniazid is particularly important when majority variants have mono-resistance to either of these first-line drugs, as these strains have the potential to become MDR. In our analysis, we identified the presence of minority variants with additional rifampicin resistance in 5/245 (2.0%) of TB cases with isoniazid monoresistance. However, genomically and phenotypically undetectable MDR minority variants may be more common in high-incidence settings. The early detection of drug resistant minority variants should guide optimal treatment to reduce the risk of failure or relapse.

Our findings confirmed previous observations that minority M. tuberculosis variants with drug resistance-conferring mutations can be detected by WGS, even when this is not apparent on pDST (4). Unrecognized minority variants with drug resistance could be selected for during sub-optimal treatment, potentially leading to treatment failure and clinical relapse (29). The sensitivity of pDST has been reported to be as high as 1% of the strain population (9, 11); however, our gDST data reflected instances where 6.8% of strains had high-confidence drug-resistance mutations but still tested susceptible on pDST. In total, 3.5% (55/1,586) of strains that tested drug susceptible on pDST had minority variants above the 3% level with drug resistance-associated mutations. Instances of sub-optimal accuracy and sensitivity of pDST are well recognized (1, 23), especially when the drug-resistant subpopulations represent only a small proportion of strains. These discrepant results warrant further interrogation as undetected drug resistance has important laboratory reporting and clinical management implications.

Our observations indicate that minority variants with high-confidence drug resistance-conferring mutations were more frequently detected in cases of mixed infections. Infection with more than one strain (mixed-strain infection) will increase the risk of drug resistance-conferring mutations being present purely by chance. Infection with more than one strain is likely to occur in a high-transmission setting, where the risk of drug-resistant strains being present is elevated (30). The presence of minority variants from the same strain with drug-resistance mutations may be a marker of previous sub-optimal treatment exposure, which may have facilitated their selection. Although only a small number of patients had documented retreatment, exposure to over-the-counter TB drugs is not uncommon in South-East Asian countries from which many of the cases detected in Australia originate (31).

Important limitations of our analysis need to be acknowledged. In our analysis, the detection of minority variants with drug-resistance mutations was based on the WHO drug-resistant TB mutations catalogue released in 2021, which has well-recognized limitations. Use of the new WHO drug-resistant TB mutations catalogue, to be released in 2023, which includes additional mutations, should improve the accuracy of drug-resistance detection. Ongoing curation and improvement of TB drug-resistance databases should further enhance subpopulation analyses in future studies. Laboratory subcultures of M. tuberculosis can reduce clonal diversity, with culture selection leading to a loss of the information on resistant-minority populations in clinical samples (32, 33). The drug-resistant minority variants observed from culture-derived WGS data may not capture all relevant M. tuberculosis population dynamics (7). Culture-independent sequencing, without any pre-selection bias from culture, may be more reflective of the bacterial population in the clinical sample. Considering this, the incorporation of metagenomic sequencing or targeted next-generation sequencing (tNGS) directly from respiratory samples offers promising alternative approaches (34, 35). It would be valuable to have more data demonstrating the feasibility and accuracy of direct metagenomic sequencing or tNGS as alternatives to culture-based methods. Investigating and optimizing culture-independent WGS and the application of tNGS to detect drug-resistant minority variants may enhance patient management and surveillance of drug-resistant variants.

We employed two tools; LoFreq to determine minority variants associated with drug resistance and QuantTB to detect mixed M. tuberculosis strain infections. However, neither software was able to assign minority variants associated with drug resistance to majority or minority populations within the isolate. It is possible that the drug-resistant minority variant belongs to the minority population due to the fact that the vast majority of the strains tested were phenotypically and genotypically consistent. However, it is also possible that the minority drug-resistant variants arose from the majority strain population but were not detected using pDST. For instance, if the minority variant, rpoB_H445N, was present in the majority population of our sample with a majority variant fabG1 -15 c>t, there would be the potential risk of developing MDR-TB. In addition to the limitations with bioinformatic pipelines, the use of the short-read sequencing technology can also impact the assignment of drug resistance. It is unlikely to capture both drug-resistant minority variants and strain population differences in a single 150bp read. This highlights the need for enhanced bioinformatic pipelines with the ability to detect drug resistance-conferring mutations in both majority and minority populations and to assign these correctly, ideally together with M. tuberculosis strain/lineage identification. The false-positive detection of mixed strain co-infection by QuantTB remains a possibility but recent incorporation of additional validation steps and an updated SNP database should have increased its reliability. Determination of the (sub)lineages responsible for the mixed strain infection was not possible but might be beneficial to include in future software tools.

The allelic frequency of minority variants with resistance detected in our analysis ranged from 3.3% to 23.0%. Previous studies suggested that the presence of drug-resistant minority is associated with poor treatment outcomes (7, 36), especially if these minority variants have an allelic frequency of more than 5% (17, 36). However, some studies indicate that even when present at very low (<1%) allelic frequencies, these minority variants may be experiencing amplification during sub-optimal treatment, leading to establishment of resistance and a subsequent enhancement in fitness (15, 37). This study did not have the opportunity to access and integrate individual-level patient outcome data. We acknowledge this as a major limitation and hope to incorporate such data in future analyses to assess the impact of these minority variants on treatment outcomes.

In conclusion, our findings demonstrate that minority variants of M. tuberculosis with drug resistance-conferring mutations can be detected from routine WGS data using enhanced bioinformatic pathways. High-resolution interrogation of drug-resistant minority variants could improve clinical risk assessment and strengthen surveillance of drug resistant strains.

The findings of this study are equally applicable to other low incidence settings. The use of similar approaches also requires consideration in high TB incidence settings to assess its relevance and optimal implementation, in order to promote equitable access to the benefits associated with advanced pathogen genomics. However, low incidence settings do have an important role to explore and refine new applications to push the technological boundaries, while acknowledging that the appropriate use of different approaches and their relative value-add might be setting-specific.

DATA AVAILABILITY

Raw de-identified pathogen WGS data was deposited in the National Center for Biotechnology Information (NCBI) Short Read Archive (SRA) under BioProject number PRJNA899911.

ETHICS APPROVAL

The study was approved by the Western Sydney Local Health District (WSLHD) Human Research Ethics Committee (approval no. 2019/PID14240).

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/jcm.00485-23.

Figure S1. jcm.00485-23-s0001.tif.

M. tuberculosis strains with drug resistance-conferring minority variants; (a) the percentage of respiratory and non-respiratory specimens (b) the number and percentage of strains with minority variants detected in all four major lineages.

Figure S2. jcm.00485-23-s0002.tif.

M. tuberculosis strains with mixed populations; (a) the number and percentage of mixed strain populations detected in different M. tuberculosis lineages, (b) the percentage in respiratory and non-respiratory specimen sources.

Supplemental Figure S1 to S2 legends. jcm.00485-23-s0003.docx.

Figure S1: M. tuberculosis strains with drug resistance-conferring minority variants; (a) the percentage of respiratory and non-respiratory specimens (b) the number and percentage of strains with minority variants detected in all four major lineages; Figure S2: M. tuberculosis strains with mixed populations; (a) the number and percentage of mixed strain populations detected in different M. tuberculosis lineages (b) the percentage in respiratory and non-respiratory specimen sources.

Tables S1 to S3. jcm.00485-23-s0004.xlsx.

Supplementary Tables.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.