Assessing the propensity of TB clinical isolates to form viable but non-replicating subpopulations

Abstract

Current tuberculosis (TB) treatment is typically effective against drug-susceptible Mycobacterium tuberculosis, but can fail due to acquired drug resistance or phenotypic resistance. M. tuberculosis persisters, a subpopulation of viable but non-replicating (VBNR) antibiotic-tolerant bacteria, are thought to contribute to poor TB treatment outcomes. In this exploratory study, we investigated treatment-naïve drug-susceptible clinical isolates collected from people with TB, who subsequently had unsuccessful treatment outcomes. These were compared to isolates from cured individuals in terms of their ability to form VBNR subpopulations. Clinical isolates from individuals with unfavorable treatment outcomes form larger subpopulations of VBNR M. tuberculosis (2.67−13.71%) than clinical isolates from cured cases (0− 1.63%) following infection of THP-1 macrophages. All isolates were drug susceptible based on phenotypic and genotypic analysis. Whole genome sequencing identified 23 non-synonymous genomic variants shared by treatment failure clinical isolates, that were not present in isolates from cured cases. This exploratory study highlights the ability of treatment-naïve clinical isolates to form heterogeneous populations containing VBNR M. tuberculosis. We also demonstrate that clinical isolates from individuals with unsuccessful treatment outcomes form higher percentages of VBNR M. tuberculosis. The findings of this exploratory study suggest that an increased propensity to form VBNR subpopulations may impact TB treatment outcome.

Introduction

Mycobacterium tuberculosis exposure and tuberculosis (TB) infection can result in a spectrum of immunological and clinical outcomes including clearance, subclinical TB, latent TB and active TB disease¹. This simplified view of TB progression does not consider individuals known as “resisters” who remain immunologically naïve despite prolonged TB exposure^2,3. Another confounding factor is the diverse pathological presentation of disease in people with TB, which can include sterile tissue, caseous hypoxic lesions and liquefied cavities with a high bacterial burden¹. TB control strategies are largely focused on the identification and treatment of active TB, thereby neglecting the spectrum of latent or subclinical TB infections, potentially allowing the spread of disease². In addition to different host microenvironments giving rise to M. tuberculosis heterogeneity, inherent bacterial heterogeneity can also result in population diversity within the same environment⁴. M. tuberculosis heterogeneity in clinical isolates has been shown to include both genetic heterogeneity and phenotypic heterogeneity^5–11. Phenotypic heterogeneity develops in response to environmental stress and may result in a subpopulation of persister bacteria¹². Persister M. tuberculosis can prevent the sterilization of TB during treatment due to their viable but non-replicating (VBNR) and antibiotic tolerant characteristics, rendering most antibiotic treatments ineffective^13,14.

A previous study investigating sterilization of TB infections following treatment in a cohort of people with TB from the Western Cape, South Africa, demonstrated the detection of lung lesions using positron emission tomography and computerized tomography (PET–CT) imaging for majority of the study participants. Interestingly, the study also detected M. tuberculosis mRNA from sputum and bronchoalveolar lavage samples following the standard 6-month anti-TB treatment¹⁵. The authors speculated that viable M. tuberculosis, which can elicit a host immune response, persisted in the hosts following anti-TB treatment. Latent TB, classified by a positive tuberculin skin test or a QuantiFERON assay in the absence of TB symptoms, is in part thought to be caused by persister M. tuberculosis bacteria^16–18. A combination of host immune pressure and antibiotic exposure is thought to contribute to the formation of a persister subpopulation^16,19,20. This subpopulation can potentially convert to actively replicating bacteria under favorable conditions to cause disease^16,17.

We previously demonstrated the identification of VBNR and antibiotic tolerant persister M. tuberculosis using fluorescence dilution using a dual fluorescence replication reporter plasmid¹⁴. In this exploratory study, we applied this tool to investigate if M. tuberculosis isolated from people with TB who failed treatment or had recurrent disease would form a greater proportion of VBNR bacteria than M. tuberculosis obtained from cured cases. Since these individuals failed or had recurrent TB infections following six months of anti-TB treatment and were deemed drug-susceptible at the time of the study, we wanted to confirm that unsuccessful treatment was not because of drug resistance-conferring mutations as detected by whole genome sequencing (WGS). We further postulated that WGS may identify genomic variants unique to the genomes of clinical isolates from failed and recurrent TB infections when compared to that of clinical isolates obtained from cured cases.

Materials and methods

Bacterial strains and culturing

This study was approved by the Stellenbosch University Health Research Ethics Committee (HREC N10/01/13). Sputum samples were obtained in a parent study from people with TB at health clinics in Cape Town, Western Cape, South Africa¹⁵. Informed consent was obtained from all study participants¹⁵. All experiments were done in accordance with guidelines and regulation stipulated by the HREC committee at Stellenbosch University. All reagents were obtained from Sigma-Aldrich unless otherwise stated.

All bacterial strains and clinical isolates used in this study are described in Table S1. Twelve isolates were selected based on TB treatment outcome to represent cured (n = 6) and recurrent/failed treatment outcome groups (n = 6) (Table S1). Liquid cultures of mycobacterial clinical isolates and laboratory strains were cultivated in 7H9 (Becton Dickinson, NJ, United States) supplemented with 10% oleic acid-albumin-dextrose-catalase (OADC; Becton Dickinson, NJ, United States), 0.2% glycerol and 0.05% Tween 80 (7H9-OGT) and incubated at 37 °C. M. tuberculosis isolates were subjected to standard genetic characterization by restriction fragment length polymorphism (RFLP) analysis and spoligotyping as previously described^21,22.

Electrocompetent mycobacteria were prepared and transformed with the pTiGc dual reporter plasmid for the identification of VBNR M. tuberculosis as previously described^14,23. Transformed mycobacterial isolates were plated onto 7H10 solid media (Becton Dickinson, NJ, United States) supplemented with 10% OADC, 1% glycerol, and kanamycin (25 µg/mL) and cultured at 37 °C. For the identification of VBNR M. tuberculosis, fluorescence dilution was performed as previously described¹⁴. Briefly, pTiGc transformed M. tuberculosis was cultured with 4 mM theophylline for 7 days prior to macrophage infection to induce the expression of the TurboFP635 fluorescent protein.

Cell culture and macrophage infection

Cell culture and macrophage infections were performed as previously reported but with a THP-1 cell line (ATCC TIB-202)¹⁴. Cells were cultured in Roswell Park Memorial Institute-1640 medium (RPMI, Thermo Fisher Scientific), supplemented with 10% heat-inactivated fetal bovine serum (FBS, Thermo Fisher Scientific) (R10) at 37 °C in a 5% CO₂ atmosphere. Cells were passaged every 2–4 days. For infections, cells were seeded at 1.25 × 10⁵ per well in 96-well plates and differentiated with 50 ng/mL of phorbol-12-myristate-13-acetate (PMA) with incubation for 3 days at 37 °C with 5% CO₂ atmosphere. Following the incubation, the R10 media containing PMA was replaced with fresh R10 and the macrophages were allowed to recover for 24 h. Prior to infection, THP-1 cells were stimulated with R10 containing 100 ng/mL lipopolysaccharide (LPS) for 1 h at 37 °C. To prepare mycobacteria for infection, theophylline-induced cultures were sonicated in an ultrasonic bath at 37 kHz for 12 min at room temperature to disperse clumps, and thereafter filtered through a 40-µm cell strainer. Bacteria were washed in phosphate-buffered saline (PBS, Thermo Fisher Scientific) with 0.05% Tween 80 (PBS-T) and resuspended in R10 supplemented with 2 mM Theophylline. Bacteria were added to the macrophages at a multiplicity of infection (MOI) of 10:1 and incubated at 37 °C in 5% CO₂ for 3 h in the presence of 2mM Theophylline. Following uptake, the cells were washed once with PBS before replacing the media with R10 media containing 100 U penicillin/streptomycin. This was followed by incubation at 37 °C in 5% CO₂ for 1 h to kill any non-phagocytosed, extracellular bacteria. Cells were washed three times with PBS before adding fresh R10, containing 2 mM theophylline, to maintain expression of TurboFP635 for 24 h after infection. After 24 h, the R10 media containing 2 mM theophylline was replaced with R10. To recover mycobacteria for flow cytometry analyses, the macrophages were lysed by the addition of sterile distilled water. Intracellular bacteria were recovered from infected macrophages along with parallel in vitro bacterial cultures at day 0 and day 5.

Flow cytometry sample preparation, acquisition and analysis

Samples for flow cytometry analysis were prepared and analyzed as previously described¹⁴. Briefly, in vitro cultured bacteria or intracellular bacteria (from lysed macrophages), were pelleted and fixed in 4% formaldehyde for 30 min, washed once in PBS-T by centrifugation at 9400 xg for 5 min and resuspended in PBS-T before storing at 4 °C. Prior to flow analysis, samples were pelleted and resuspended in PBS and filtered through a 35-µM filter (Becton Dickinson, NJ, United States). Counting beads from the LIVE/DEAD BacLight Bacterial Viability kit (Thermo Fisher Scientific) were used to establish the effective MOI following infection. Briefly, 5 µL of 6.0 μm microsphere standard beads were added to each sample prior to flow cytometry analysis. Samples were analyzed on the FACSJazz flow cytometer (Becton Dickinson, NJ, United States). Forward scatter (FSC) and side scatter (SSC) data were captured, as well as the fluorescent intensity of GFP which was detected following excitation at 488 nm, using a 530/30 filter. TurboFP635 fluorescence intensity was measured following excitation at 561 nm, using a 610/20 filter. Compensation was carried out using single color and unlabeled controls in each experiment. For samples from both in vitro cultures and bacterial samples recovered from macrophages, 20 000 events were captured. Two biological replicates consisting of two to three technical replicates were included in this study.

FlowJo v10 software was used to analyze flow cytometric data. A primary gate was set according to FSC/SSC properties, followed by gating on the GFP-positive (live) population. The TurboFP635 fluorescence intensity of the population was then analyzed (Figure S3)¹⁴.

Determination of effective MOI within macrophages utilizing counting beads

Enumeration of bacterial uptake by and survival in THP-1 cells post infection was determined by exploiting the counting beads provided in the LIVE/DEAD Baclight Bacterial Viability and Counting kit according to the manufacturer’s specifications. The beads from the kit provided an accurate and more rapid alternative to using colony forming units (CFUs) for determining the effective MOI. Gating strategy for bacteria and beads demonstrated in Figure S4. The bacteria/mL were calculate based on the bacterial events captured, the number of bead events captured, the number of beads added, sample volume and the dilution factor, as follows:

Bacteria/mL = bacteria counted/(beads counted/bead concentration).

Effective MOI = ((Bacteria/mL)/sample volume/macrophages per well.

Determination of VBNR population frequency per isolate

To quantify the VBNR populations in the isolates obtained from cured and failed/recurrent individuals, we first applied a gating strategy as outlined in Fig. 1. Following the selection of viable bacteria (high GFP), TurboFP635 florescence was assessed in a histogram plot (Figure S3). A threshold gate was set based on the median TurboFP635 florescence intensity (MFI) of intracellular bacteria at 0 h for each isolate (Fig. 1a). To determine the frequency of a VBNR subpopulation, the top 50th percentile of TurboFP635 signal was selected and termed “high red”. This gate was used to determine the frequency of “high-red” VBNR bacteria in the in vitro bacteria and intracellular bacteria at 120 h post infection (Fig. 1b–c). The frequency of the macrophage-enriched VBNR subpopulation was determined by the following equation:

Fig. 1.

Calculating the VBNR subpopulation using flow cytometry data. The figure shows histogram data for Turbo FP635 fluorescence intensity measured at 0 h (red) and 120 h (orange) post THP-1 infection and for 120 h in vitro culture (dotted line). Histograms (a–d) is representative of a technical replicate for the isolate S153dx from the cured group. Histograms e-h is representative of a technical replicate for isolate S43dx from the failed/recurrent group. M. tuberculosis recovered from macrophages at 0 h post infection shows maximum red fluorescence intensity (a, e). Measured Turbo FP635 signal larger than the mean fluorescence intensity was deemed high red, as measured by the bar. This gate was applied to l20h Turbo FP635 intensities measured for M. tuberculosis recovered at 120 h post infection (c, g) and in vitro cultured M. tuberculosis (b, f). The macrophage-enriched VBNR subpopulation percentage was calculated by subtracting the percentage in vitro cultured M. tuberculosis from the percentage intracellular M. tuberculosis at 120 h. A histogram overlay for the cured sample (d) shows a complete overlap for Turbo FP635 intensity measured at 120 h post macrophage infection and in vitro cultured M. tuberculosis. This is indicative of a miniscule or absent VBNR subpopulation. The histrogram overlay for the representative failed isolate (h) shows that the fluorescence intensity signal measured at 120 h post infection does not fully overlap with that of the in vitro cultured bacteria, suggesting a high red fluorescent subpopulation of bacteria, suggestive of a VBNR population.

Statistical analysis was carried out in GraphPad prism v9.01. Differences between the means of cured vs. failed/recurrent VBNR formation was assessed using a Mann-Whitney U test.

Whole genome sequencing (WGS) and WGS analysis

DNA was obtained from all M. tuberculosis clinical isolates as previously described²⁴. Sequencing libraries for all isolates were constructed using the standard genomic DNA sample preparation kits from Illumina (Illumina, Inc, San Diego, CA), according to manufacturer’s specifications. The whole genomes of the M. tuberculosis isolates were sequenced using either the Illumina NextSeq or Illumina MiSeq platforms (Table S1).

Sequenced genomes were analyzed using the in-house Universal Sequence Analysis Pipeline (USAP)²⁵. In summary, quality assessment of the sequencing data was done using FASTQC²⁶. Trimming of reads was completed utilizing the sliding window approach in Trimmomatic v0.32²⁷. Reads were subsequently mapped to M. tuberculosis H37Rv genome (GenBank NC000962.3) with three mapping software packages Novoalign v3.02.13 (Novocraft), Burrows-Wheeler Aligner v0.6.2 (BWA), and SMALT v0.7.5^28,29. Single nucleotide polymorphisms (SNPs), small insertions and deletions were detected with two callers: SAMTools v1.3 and the Genome Analysis Tool Kit v3.5 (GATK)^30–32. In-house scripts written as part of the USAP pipeline were used to (1) annotate the identified variants, (2) calculate the resultant amino acid changes created by SNPs located within genes and (3) annotate the identified insertions and deletions. The genes in which the variants occur were classified based on their cellular function as reported in the TubercuList knowledgebase, Mycobrowser and Uniprot^33–35. Variants identified by both SAMTools v1.3 and the GATK v3.5 in three alignments were filtered to exclude variants in the pe/ppe gene family regions, repeat regions, insertion sequences and phages, and only variants with an allele frequency of > 0.95 were considered. A Python script was used to generate a connected sequence of all high confidence SNPs (n = 23861) recognized for each isolate in multi-FASTA format. Qualimap 2 was utilized for visual inspection of the depth of coverage of the alignments of all clinical isolates³⁶. Only 8 sequences passed quality control measures, with a depth of coverage of between 87,96 x – 140,06 x.

In parallel, TBProfiler (v4.4.2) was used to analyze the WGS data of clinical isolates for lineage identification and to identify known drug resistance conferring mutations³⁷. For a comparative analysis of genomic variants, an in-house automated pipeline for M. tuberculosis WGS analysis was adapted allowing for sequence data analysis as previously described³⁸.

Phenotypic drug susceptibility testing (pDST)

To determine the minimum inhibitory concentration (MIC) of all isolates against first line M. tuberculosis drug regimen the BACTEC MGIT 960 system was used as recommended by the manufacturer and FIND³⁹. This included isoniazid (INH), rifampin (RIF), ethambutol (EMB) and pyrazinamide (PZA). MGIT culture tubes containing 7 mL BBL media was supplemented with 800 µL OADC and inoculated with 500 µL untransformed isolates and incubated in the BACTEC MGIT 960 machine at 37 °C until cultures flagged positive. MGITs were sub-cultured into new MGIT tubes and incubated until the second day of positivity. MGIT tubes and PZA tubes containing PZA broth were prepared by supplementation with 800 µL OADC. MICs were determined utilizing the following concentrations respectively; INH (0.02, 0.05, and 0.1 µg/mL), RIF (0.06, 0.12, 0.25, 0.5 µg/mL), ETB (0.6, 1.25, 2.5 µg/mL) and PZA (25.0, 50.0, 75.0 and 100.0 µg/mL). Day 2 positive MGIT cultures were sub-cultured by adding 500 µL into each drug-containing MGIT tube and 500 µL of 1:100 saline diluted culture for all isolates was added as a drug-free growth control. Results were captured when 100 growth units (GU) were attained.

Results

Selection of clinical isolates

Twelve clinical isolates previously described were obtained from the Catalysis TB – Biomarker Consortium¹⁵. All M. tuberculosis isolates analyzed in this study were collected at the time of diagnosis, prior to treatment. Individuals who maintained negative sputum cultures after 6 months of treatment were classified as cured (n = 6). Individuals who did not convert to sputum culture negative after 6 months of treatment were classified as failed (n = 3). Participants who were sputum negative after 6 months, but who were diagnosed with recurrent TB within 2 years of completion of treatment, were classified as recurrent (n = 3) (Table 1)¹⁵. The cured group consisted of 5 males (90%) and 1 female (10%) with ages ranging from 19 to 42 years old. The failed/recurrent group consisted of 3 males (50%) and 3 females (50%) with ages ranging from 18 to 44 years. The prevalence of smokers overall was 11 (92%) (Table 1).

Table 1.

M. Tuberculosis isolates clinical information.

Sample ID	Age	Sex	Smoking Status	Outcome
S5dx	30	Male	Quit smoking	Cured
S29dx	42	Male	Smoking	Cured
S105dx	21	Male	Smoking	Cured
S126dx	39	Male	Quit smoking	Cured
S153dx	25	Male	Smoking	Cured
S159dx	19	Female	Never smoked	Cured
S43dx	18	Male	Quit smoking	Failed
S112dx	52	Male	Smoking	Recurrent
S137dx	44	Female	Smoking	Recurrent
S152dx	23	Female	Smoking	Recurrent
S163dx	25	Male	Smoking	Failed
S169dx	30	Female	Smoking	Failed

Increased VBNR formation in clinical isolates from individuals who had unsuccessful treatment outcome

To assess the replication dynamics of clinical isolates from the cured and failed/recurrent TB groups, we exploited a dual fluorescence reporter system as previously reported¹⁴. A macrophage infection model was selected to investigate the formation of VBNR bacteria as this infection model will recapitulate some of the same stressors experienced by M. tuberculosis during infection. The laboratory strain M. tuberculosis H37Rv carrying the reporter plasmid was used as a reference for M. tuberculosis replication dynamics.

The reporter plasmid constitutively expresses a green fluorescent protein (GFP) to measure cell viability and a red fluorescent protein (Turbo FP635) in the presence of the inducer (theophylline). Turbo FP635 mean fluorescence intensities (MFI) at 0 h and 120 h post THP-1 infection and at 120 h for in vitro cultures were used to estimate the percentage VBNR bacteria within the bacterial population (Fig. 1). The population with red fluorescence larger than the MFI at 0 h post THP-1 infection was deemed the high red fluorescence population (as measured by a bar in Fig. 1a and e). This gate was applied to in vitro cultured bacteria (Fig. 1b and f) and intracellular bacteria at 120 h post infection (Fig. 1c and g) to identify high-red fluorescence M. tuberculosis within the bacterial population. The macrophage-enriched VBNR subpopulation was calculated by subtracting the percentage MFI overlap of in vitro cultured M. tuberculosis from the percentage MFI overlap for intracellular M. tuberculosis at 120 h (Fig. 1d and h).

Following infection of the human cell line (THP-1), M. tuberculosis clinical isolates and M. tuberculosis H37Rv were expressing both GFP and Turbo FP635 fluorophores (0 h post infection) (Fig. 2 and Figure S1–2). The histogram profiles for Turbo FP635 varied among the clinical isolates and M. tuberculosis H37Rv (Fig. 2, Figure S2). Following the removal of the inducer at 24 h post infection, the red fluorescent signal in actively replicating bacteria is halved with every replication cycle, resulting in a lower fluorescence intensity. We previously showed that following macrophage infection, a heterogeneous population (with varying levels of Turbo FP635 fluorescence intensity) emerged after 48 h¹⁴. These data suggest that upon infection, M. tuberculosis will complete at least one to two replication cycles prior to the formation of a VBNR subpopulation. The emerging VBNR M. tuberculosis subpopulation will retain the majority of the red fluorescent signal.

Fig. 2.

Fluorescent intensities 120 H post THP-1 infection. Population-wide replication dynamics of baseline isolates obtained from cured treatment group (A–F). Population-wide replication dynamics of baseline isolates obtained from failed/recurrent treatment group (G–L). Intracellular bacteria lysed from macrophages 0 h (red), in vitro bacteria 120 h (dotted black line), intracellular bacteria lysed from macrophages 120 h (orange). Data is representative of two biological duplicates consisting of two to three technical replicates. A high level of red fluorescence was detected at 0 h (A–L), post induction, which decreased in isolates from cured individuals (A–F) in both in vitro cultures and in macrophage infections at 120 h, suggestive of replication of M. tuberculosis under these conditions. In contrast, red fluorescence intensity remained high in M. tuberculosis from failed/recurrent isolates which were recovered from macrophages at 120 h, suggesting slow or non-replication of these bacteria.

The red fluorescence intensity measured at 120 h post infection was lower than at 0 h for all samples, suggesting that bacterial replication occurred following macrophage infection (Fig. 2). However, in contrast to the clinical isolates from the failed/recurrent group, the isolates from the cured group showed a larger reduction in red fluorescence at 120 h post infection (Fig. 2). The red fluorescence intensity measured for isolates from the cured group and M. tuberculosis H37Rv largely overlapped with the fluorescence intensities of in vitro cultured controls at 120 h (Fig. 2a-f). These data suggest that bacteria recovered from macrophage infections for clinical isolates from the cured group as well as M. tuberculosis H37Rv was actively replicating, with little to no VBNR formation (Table 2; Fig. 2, Figure S2). In contrast, fluorescence intensities of clinical isolates from the failed/recurrent TB group at 120 h post infection did not fully overlap with the fluorescence intensities of in vitro failed/recurrent clinical isolates at 0 h (Fig. 2g–l). The VBNR subpopulations were calculated (as described in Fig. 1) from two to three technical replicates of two biological replicate experiments (Table S2, Table 2). M. tuberculosis isolates from the failed/recurrent TB group (estimated percentage VBNR formation of between 2.67 and 13.71%) were found to form a higher percentage of VBNR post infection when compared to the cured isolates (estimated percentage VBNR formation of 0-0.81%) (p-value 0.0022) (Fig. 3; Table 2).

Table 2.

WGS Clinical isolate phylogeny and VBNR formation.

Sample ID	Disease outcome	Family	Percentage VBNR formation (mean ± standard deviation)*
S5dx	Cured	L4.3.2.1/LAM	0% (+/−0)
S29dx	Cured	L4.3.2.1/LAM	0.52% (+/−1.17)
S105dx	Cured	L4.2/Ural	0% (+/−0)
S126dx	Cured	L4.1.1.3/X	0% (+/−0)
S153dx	Cured	L4.1.1.3/X	0.81% (+/−1.98)
S159dx	Cured	L4.9/T1	0% (+/−0)
S43dx	Failed	L2.2/Beijing	2,67% (+/−0.64)
S112dx	Recurrent	L2.2/Beijing	4,33% (+/−4.246)
S137dx	Recurrent	L2.2/Beijing	5,52% (+/−4.87)
S152dx	Recurrent	L2.2/Beijing	3,22% (+/−3.83)
S163dx	Failed	L2.2/Beijing	7,14% (+/−2.555)
S169dx	Failed	L4.1.2.1/X	13.71% (+/−12.10.14)
H37Rv::pTiGc	Control	L4.9	0% (+/−0)

* Percentage VBNR formation calculated from fluorescence intensity data as measured by flow cytometry.

Fig. 3.

VBNR frequency in isolates obtained from the cured and failed/recurrent patient groups following macrophage infections. Isolates from failed/recurrent patients exhibit a higher proportion of VBNR populations compared to the isolates from the cured patient group and M. tuberculosis H37Rv. Mean values of two independent biological replicate experiments consisting of two to three technical replicates plotted with a significant p-value of 0.0022 between M. tuberculosis isolates from the cured vs. failed/recurrent group.

Absence of drug resistance-conferring mutations and phenotypic drug resistance in failed/recurrent clinical isolates

The clinical isolates in this study were obtained on TB diagnosis prior to the initiation of standard anti-TB treatment. These isolates were reported as drug-susceptible by routine tests. We confirmed phenotypic drug susceptibility using pDST (Table 3). Isolates are scored as resistant if pDST exceeds a critical concentration (> 0.25 µg/ml for isoniazid, > 1.0 µg/ml for rifampin, > 4.0 µg/ml for ethambutol, and > 100 µg/ml for pyrazinamide)^40–42. All isolates selected for this study were scored as susceptible. We further examined genome sequences to confirm the absence of resistance-conferring mutations. Isolates were sequenced to a median average depth of coverage of 115.98x (min = 40.77x, max = 182.41x). For the majority of isolates, > 90% of the reads were mapped to the M. tuberculosis H37Rv reference genome. WGS analysis confirmed that all of the cured isolates belonged to lineage 4 and the majority of failed/recurrent clinical isolates belonged to lineage 2 (n = 3) or lineage 4 (n = 1), as previously reported by RFLP and spoligotyping (Table 2)¹⁵. In agreement with previous reporting, drug resistance prediction using TB-profiler identified no known resistance-causing mutations (Table S3)^15,37.

Table 3.

MIC reading after 7 days incubation.

Sample ID	Outcome	Isoniazid (ug/ml) (resistance > 0.25 ug/ml)	Rifampin (ug/ml) (resistance > 1 ug/ml)	Ethambutol (ug/ml) (resistance > 4 ug/ml)	Pyrazinamide (ug/ml) (resistance > 100 ug/ml)
S5dx	Cured	0,05	0,125	≤ 0,6	50
S29dx	Cured	0,05	0,06	1,25	≤ 25
S105dx	Cured	0,05	0,06	1,25	≤ 25
S126dx	Cured	0,05	0,125	1,25	50
S153dx	Cured	0,05	0,125	1,25	50
S159dx	Cured	0,05	0,06	0,6	50
S43dx	Failed	0,05	0,125	1,25	50
S112dx	Recurrent	0,05	0,06	1,5	25
S137dx	Recurrent	0,05	0,06	1,25	75
S152dx	Recurrent	0,05	0,06	0,6	25
S163dx	Failed	0,05	0,06	≤ 0,6	25
S169dx	Failed	0,05	0,125	1,25	≤ 25

High VBNR forming clinical isolates share non-synonymous variants

WGS data was inspected to ascertain if clinical isolates from the failed/recurrent group contained any genomic variants not present within the genomes from the cured group, which may contribute to the formation of a higher percentage of VBNR M. tuberculosis. In the four failed/recurrent clinical isolates, we identified 23 shared non-synonymous variants that were not detected in any of the cured clinical isolates (Table 4). Interestingly, we identified several instances of genomic variants in genes with gene ontologies related to membrane functions (13/23). These included two proteins with functions related to drug efflux (Rv1218c, Stp), an ESX-1 secretion system protein (EspK) and an ESAT-6-like protein, EsxP. Lipoproteins have been implicated in virulence and immunoregulatory processes⁴³. We identified genomic variants for two genes predicted to encode lipoproteins (LppA, LppB)⁴³. Genomic variants were also identified in genes encoding proteins that are required for phospholipid biosynthetic processes (Ino1, PssA) or lipid catabolic processes (Rv1592c). Further, we identified a variant in the transcriptional regulator Rv0324, which has previously been suggested to be a regulator of bedaquiline tolerance⁴⁴.

Table 4.

Shared non-synonymous mutations in failed/recurrent clinical isolates.

Rv annotation	Uniprot annotation	Gene name	Amino acid change	Protein length	Protein names
Rv0046c	P9WKI1	ino1	R190G	367	Inositol-3-phosphate synthase (IPS) (MIP synthase)
Rv0048c	P9WM87	rv0048c	V250A	289	Uncharacterized protein Rv0048c
Rv0324	O08446	rv0324	T168A	226	HTH-type transcriptional regulator
Rv0436c	P9WPG1	pssA	G167V	286	Phosphatidylserine synthase
Rv1127c	O06579	ppdK	G69E	490	Probable pyruvate, phosphate dikinase PpdK
Rv1218c	O86311	rv1218c	Q243R	311	Multidrug efflux system ATP-binding protein
Rv1448c	P9WG33	tal	T244A	373	Transaldolase
Rv1517	P71796	rv1517	L188F	254	Conserved hypothetical transmembrane protein
Rv1592c	P9WK89	rv1592c	I322V	446	Probable inactive lipase
Rv1915	O07718	aceAa	G179D	367	Putative isocitrate lyase subunit A (ICL)
Rv2109c	P9WHU1	prcA	R135P	248	Proteasome core protein PrcA
Rv2307c	P9WLC7	rv2307c	M24T	281	Uncharacterized protein
Rv2316	P71896	uspA	V127L	290	Probable sugar-transport integral membrane protein ABC transporter
Rv2333c	P9WG91	stp	D69Y	537	Multidrug resistance protein Stp (Spectinomycin tetracycline efflux pump)
Rv2347c	P9WNI5	esxP	T3S	98	ESAT-6-like protein EsxP
Rv2360c	O05838	rv2360c	A66T	142	Uncharacterized protein
Rv2439c	P9WHU9	proB	A226S	376	Glutamate 5-kinase
Rv2543	P9WK81	lppA	H37Q	219	Putative lipoprotein LppA
Rv2544	P9WK79	lppB	G36D	220	Putative lipoprotein LppB
Rv2756c	O33298	hsdM	L306P	540	site-specific DNA-methyltransferase
Rv3113	O05790	rv3113	G134E	222	Possible phosphatase
Rv3879c	P9WJC1	espK	C729S	729	ESX-1 secretion-associated protein EspK
Rv0619	Q79FY3	galTb	T174A	181	Probable galactose-1-phosphate uridylyltransferase

Given the apparent higher propensity for VBNR formation among the isolates from the failed/recurrent group, we wanted to investigate whether these isolates harbored any variants in genes previously suggested to contribute to M. tuberculosis persistence. All non-synonymous genomic variants identified in the failed/recurrent and subjected to WGS are described in Table S4-S7. Interestingly, the M. tuberculosis isolate S169, which formed the highest percentage of VBNR, had the most non-synonymous genomic variants (Table S7). We identified seven non-synonymous variants for genes previously reported to be associated with persistent infections (Table 5). The majority of the identified variants were for genes (including Rv2694c) that encode proteins associated with the membrane or the capsule⁴⁵.

Table 5.

Non-synonymous genomic variants in genes previously implicated in M. Tuberculosis persistent infections.

Rv annotation	Clinical isolate	Uniprot annotation	Gene name	Amino acid change	Protein length	Protein names	Role in persistence
Rv2438c	S43dx	P9WJJ3	nadE	L225R	679	Glutamine-dependent NAD(+) synthetase	M. tuberculosis is dependent on NAD synthesis for growth and survival during nonreplicating persistence. NadE is essential for persistence58.
Rv1736c	S43dx	P9WJQ1	narX	R154Q	652	Nitrate reductase-like protein	Anaerobic nitrate reduction is required for persistence of M. bovis BCG in the lungs, liver, and kidneys59.
Rv1212c	S112dx	P9WMZ1	glgM	A191T	387	Alpha-maltose-1-phosphate synthase (M1P synthase)	Isocitrate lyase, an enzyme essential for the metabolism of fatty acids, is essential for M. tuberculosis persistence in mice60,61.
Rv0126	S112dx	P9WQ19	treS	E252Q	601	Trehalose synthase/amylase (MTase)	M. tuberculosis requires TreS for virulence in a mouse model and drug-tolerance62,63.
Rv1194c	S169dx	O05296	rv1194c	F114L	421	Conserved protein	Upregulation of gene expression in M. tuberculosis hip mutants64.
Rv1130	S169dx	O06582	prpD	P415L	526	2-methylcitrate dehydratase (Aconitase)	Upregulation of gene expression under persister inducing conditions65.
Rv2694c	S169dx	O07196	rv2694c	L26V	122	Conserved protein	Upregulation of gene expression under persister inducing conditions65.

Discussion

TB infections have been shown to be genetically heterogeneous, compromising effective diagnosis and treatment^5–7. Genetic heterogeneity, or complex infections, have been associated with higher odds of continued culture positivity following treatment initiation⁸. Lengthy treatment regimens are therefore required for the treatment of these multi-strain infections. In recent years, phenotypic heterogeneity has also been shown to pose a threat to effective anti-TB treatment. M. tuberculosis generates phenotypically diverse subpopulations in response to environmental changes, promoting the emergence of a drug tolerant persister subpopulation without genetic mutagenesis^4,20. These persister subpopulations increase the likelihood of treatment failure and recurrence of the disease^4,12.

Our group previously demonstrated the identification of VBNR M. tuberculosis using a dual fluorescent reporter system which enables the visualization of bacterial replication dynamics using fluorescence dilution¹⁴. In the present exploratory study, the ability of clinical isolates carrying the replication reporter plasmid to form VBNR bacteria was investigated using a human cell line infection model to emulate the host environment during infection. Fluorescence dilution demonstrated M. tuberculosis VBNR formation and population heterogeneity for treatment-naïve clinical isolates from both cured and the failed/recurrent groups (Fig. 3). We found that clinical isolates from patients who failed or had recurrent TB infections formed a higher percentage of VBNR bacteria when compared to clinical isolates from cured cases and the laboratory reference strain (Fig. 3). These results suggest that the host environment does induce the VBNR phenotype and that the potential for phenotypic heterogeneity and the generation of M. tuberculosis VBNR phenotype probably exist prior to the initiation of treatment. Within the failed/recurrent group, we observed variation in the ability of these clinical isolates to form VBNR M. tuberculosis, suggesting variability in the propensity of these strains to form VBNR bacteria in response to environmental stress. Similar findings have been reported for Staphylococcus aureus infections, where moxifloxacin-susceptible clinical isolates from 36 patients who failed or had recurrent infections showed high persister formation for 17% of the isolates⁴⁶. The ability of clinical isolates to form persisters can adversely impact clinical outcome, even in the absence of drug resistance.

All cured isolates belonged to lineage 4 and most of the failed/recurrent isolates belonged to lineage 2 (Table 2). The lineage 2 and 4 strains have previously been shown to account for most TB infections in South Africa⁴⁷. Lineage 2 isolates were recently shown to have an increased propensity to form differentially culturable bacteria when compared to M. tuberculosis strains from lineage 4⁴⁸. Interestingly, although most of the isolates from the failed/recurrent group belonged to the lineage 2, the clinical isolate with the highest propensity to form VBNR bacteria, S169dx, belonged to lineage 4. WGS of TB clinical isolates confirmed the absence of known resistance-conferring mutations, indicating that canonical drug resistance has likely not contributed to treatment failure of the failed/recurrent group isolates⁴⁹. However, isolates which yielded larger subpopulations of VBNR M. tuberculosis did contain several genomic variants not detected in the genomes of isolates from the cured group (Table S4-S7). We identified 23 non-synonymous genomic variants in all failed/recurrent isolates that were not identified in any of the genomes of the M. tuberculosis isolates from individuals who were cured (Table 3). More than half of the genomic variants identified were in genes which encode proteins associated with the membrane, including lipoproteins (LppA and LppB) and proteins involved in drug efflux (Rv1218c, Stp) and secretion (EsxP and EspK). Lipoproteins are secreted membrane-anchored proteins which reportedly contribute to virulence and immunoregulatory processes⁴³. Likewise, the ESX-1 secretion system is known to contribute to M. tuberculosis virulence⁵⁰. Interestingly, we identified a predicted C729S mutation in EspK for all sequenced failed/recurrent isolates as well as a Thr90Lys variant in isolate S112 (Table S5) and a Leu39Trp variant in isolate S169 (Table S7). EspK is a chaperone protein required for the optimal secretion of the ESX-1 secretion substrate EspB⁵¹. WGS identified a predicted Arg135Pro amino acid change in the essential proteasome component PrcA (Rv2109c) for all failed/recurrent clinical isolates (Table 4)^52–55. The proteasome prcBA genes are essential for M. tuberculosis persistence in the chronic phase of infection in mice. Silencing of the PrcAB proteasome subunit results in increased susceptibility to reactive nitrogen intermediates but increased resistance to oxidative stress⁵⁶. The transcriptional regulator Rv0324 was also suggested to encode a Thr168Ala amino acid change in all failed/recurrent isolates (Table 4). Rv0324 has been found to be upregulated in response to bedaquiline, capreomycin and moxifloxacin exposure and has been suggested to be a regulator of bedaquiline tolerance^44,57. Antibiotic tolerance is strongly associated with persister TB bacteria and the impact of this predicted amino acid change requires further investigation.

Several genomic variants were identified for genes previously associated with persistent M. tuberculosis infections (Table 5). Persister formation can be induced by environmental stress, including nutrient limitation, acid stress, immune factors and exposure to immune cells¹². Most genomic variants identified in genes previously reported to be linked with mycobacterial persistence were associated with the membrane or the capsule, suggesting that these high VBNR forming isolates may respond differently to environmental stress than those from the cured group. All the genomic variants identified in this exploratory study require further investigation. Specifically, the impact of genomic variants of M. tuberculosis viability and the ability to respond to persister-inducing conditions should be investigated using mutagenesis or knockdown studies.

This exploratory study was limited by the small number of clinical isolates investigated, due to restricted sample availability, and consequently we could also not correct for clinical confounders. Despite these limitations, this exploratory study demonstrates differences in the ability of M. tuberculosis clinical isolates to form VBNR populations, irrespective of patient differences. The ability of clinical isolates from failed/recurrent infections to form VBNR subpopulations was not influenced by previous TB treatment exposure, as all clinical isolates included in this study was obtained from treatment-naïve patients at the time of diagnosis. Further studies need to assess the ability of TB isolates to form VBNR subpopulations in a larger cohort. WGS did not identify any known drug-resistance-conferring mutations which may have contributed to treatment failure by selecting for treatment resistant subpopulations. Several genomic variants were identified within the genomes of failed/recurrent isolates which were not detected in the genomes from the cured isolate group. However, the potential contribution of these genomic variants to VBNR formation remains to be investigated. Future studies investigating genomic contributions to VBNR formation should verify the presence of genomic variants found to be shared between isolates from the failed/recurrent TB group in a larger cohort. This exploratory study highlights the potential of VBNR M. tuberculosis to impact treatment outcome, despite the absence of known resistance-conferring mutations.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

JC, JM, AD and SS assisted in experimental conceptualization and JC and JM performed experimental work. JC, NK, AD and HL analyzed results. NK, JC drafted the manuscript, tables and figures. All authors contributed to the article and approved the submitted version.

Funding

This research was supported by the South African government through the National Research Foundation of South Africa (NRF), the South African Medical Research Council, and the Harry Crossley Foundation and the VALIDATE network. SS is funded by the South African Research Chairs Initiative of the Department of Science and Technology and National Research Foundation (NRF) of South Africa, award number UID 86539. NK and JM acknowledge salary support from the VALIDATE Network, which was funded by the UK’s Medical Research Council (MRC) and the Bill and Melinda Gates Foundation. AD acknowledges salary support from the Tuberculosis Omics Research Consortium headed by Prof. Annelies Van Rie, funded by the Research Foundation Flanders (FWO) under grant number G0F8316N (FWO Odysseus). Authors JC, NK, HL, JM, AD, SS are affiliated with the DSI-NRF Centre of Excellence for Biomedical Tuberculosis Research; South African Medical Research Council Centre for Tuberculosis Research; Division of Molecular Biology and Human Genetics, Faculty of Medicine and Health Sciences, Stellenbosch University, Cape Town.

Data availability

Flow cytometry data are available from the corresponding author upon request. The raw WGS data were deposited to the European Nucleotide Archive under accession number PRJEB67335. Data will be made publicly available upon acceptance of the manuscript or by request of the reviewers/editors.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval

Ethical approval was obtained from the Stellenbosch Human Research Ethics Committee (Registration Number N10/01/013) and the Biological and Environmental Safety committee (Registration Number BES-2023-13049).

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-024-79389-w.