Mtb isolates used in this study were collected from a cohort of participants with pulmonary TB (PTB) and were described previously (Vijay et al., 2017). One hundred and fifty three PTB patients were recruited from two district TB control units (4 and 8) in Ho Chi Minh City (HCMC) in southern Vietnam between January 2015 and October 2016. Patients had clinical symptoms of active PTB, which was confirmed by chest X-rays and sputum positive with acid fast bacilli by Ziehl-Neelsen stain. All were adults and HIV negative.

All patients recruited to the study were ≥18 years old and written informed consent was obtained from each patient or an accompanying relative if the patient was unable to provide consent independently, which was approved by Ethics Committees. The study protocol and informed consent form were approved by the institutional review boards of the Hospital for Tropical Diseases in Vietnam and the Oxford Tropical Research Ethics Committee in the United Kingdom (Vijay et al., 2017).

The whole blood was obtained from three healthy donors who are Vietnamese volunteers working at Oxford University Clinical Research Unit (OUCRU). The samples were part of an existing study at OUCRU on the influence of host genotype to disease susceptibility phenotype using a healthy cohort of Vietnamese adults. Participants enrolled to this study had consented to donate samples in future studies. These samples were anonymized by subject codes, without names or identifying information.

Sputum samples were collected from 153 PTB patients before the start of anti-TB therapy. The samples were decontaminated by N-acetyl-L-cysteine and 2% NaOH, inoculated in Mycobacteria growth indicator tubes and then on Lowenstein–Jensen (LJ) medium (Becton Dickinson (BD) at 37°C for 3–4 weeks to isolate Mtb. Single colonies were sub-cultured in 7H9 liquid medium and stored as glycerol stocks for further experiments. To prepare cultures for infection, Mtb isolates and H37Rv were inoculated in 7H9T medium (7H9 broth supplemented with 10% Oleic acid/Albumin/Dextrose/Catalase (OADC) enrichment, and 0.05% Tween 80, BD) at 37°C with shaking. The bacteria were ready for infection experiments when OD₆₀₀ reached 0.5–1.

THP-1 human monocytic cell line was obtained from ATCC (TIB-202). Cells were cultured in RPMI 1640 (Sigma) supplemented with 10% heat-inactivated fetal bovine serum (FBS, Sigma), 2 mM L-glutamine (Sigma) and 100 units of penicillin (Sigma). Cells were transferred to either 96-well plates (6 × 10⁴ cells per well) or 24-well plates (2.5 × 10⁵ cells per well) to establish a confluent monolayer. The cells were differentiated with 50 ng/ml phorbol 12-myristate 13-acetate (PMA, Sigma) for 2 days and then incubated in fresh media for at least 12 h before experiments.

To prepare human monocyte-derived macrophages (hMDM), peripheral blood monocyte cells (PBMCs) were separated from 20 ml heparinized whole blood by Lymphoprep (Axis-Shield) gradient centrifugation in accordance with the manufacturer's protocol. PBMCs were plated in cell-culture treated 60 mm × 15 mm petri dishes (Corning) with 6–8 × 10⁶ cells per dish in media without serum, containing RPMI-1640 (Sigma), 2 mM L-glutamine (Sigma), and 100 units of penicillin and streptomycin (Sigma). After 2 h, the non-adhered cells were washed off gently three times by warm phosphate buffered saline (PBS) with 3% FBS. Cells were re-suspended in 2 ml complete media containing 10 ng/ml human macrophage colony-stimulating factor (m-CSF, R&D system). On the following day, cells were seeded in 96-well plate (8 × 10⁴ cells per well). Complete media was changed at day 4, and macrophages were infected with Mtb at day 7.

Prior to the infection, Mtb was bead beaten to remove clumps. Macrophages were infected with Mtb in uptake buffer (0.5% BSA, 0.45% dextrose, 0.1% gelatin, 0.01 g CaCl₂, 0.01 g MgCl₂) at multiplicity of infection (MOI) of 1. Macrophages were incubated for 4 h, then washed twice and incubated for 6 days. The percentage of macrophage lysis was observed daily using light microscopy by two independent staff and blind sampling. Each experiment condition was triplicated and mean values determined.

The bacterial load in all sputum samples was measured using GeneXpert MTB/RIF assay (Cepheid, Sunnyvale, CA, United States) following the manufacturer's standard operating procedure. Briefly, 2 ml sample reagent was added to 200 μl of decontaminated sputum, vortexed for 30 sec, incubated at room temperature for 10 min, then added to the test cartridge, and finally loaded onto a GeneXpert instrument. Results were reported as cycle threshold (Ct) values, representing the number of PCR cycles required for the signal to reach a detection threshold. The average Ct values of five probes (excluding any delayed values due to rifampicin resistance) was used to estimate bacterial load (Blakemore et al., 2010).

Mtb DNA was extracted from cultures on LJ media by cetyltrimethyl ammonium bromide (CTAB) method. Lineages were identified by large sequence polymorphism (LSP) typing method as described previously (Caws et al., 2008; Vijay et al., 2017). Briefly, all the strains were first screened for either RD105 or RD239 deletions by PCR, as the majority of strains were anticipated to contain these deletions. Isolates bearing RD105 deletion were defined as East-Asia/Beijing genotype while those bearing RD239 deletion were defined as Indo-Oceanic genotype. Isolates without RD105 or RD239 deletions were further tested for Euro-American lineage by using PCR to detect the deletion of 7bp in the pks gene. For whole genome sequencing, Mtb DNA from each isolate was used to prepare a library using NEBNext Ultra II DNA Library Prep Kit for Illumina (New England BioLabs). The libraries were quantified using KAPA Library Quantification low ROX qPCR mix (Roche) and then sequenced with either 300-cycle NextSeq 500 Mid Output V2, or High Output V2 kits, or 600-cycle MiSeq Reagent V3 Kit (Illumina) with the standard Illumina procedure using Illumina NextSeq or MiSeq sequencers.

Mtb was infected to 2.5 × 10⁵ THP-1 cells in 24-well plate at MOI 1. After 4 h of incubation (considered as day 0), all extracellular bacteria were removed gently by washing and intracellular bacteria were harvested. At 6 days after infection, both extracellular bacteria released from macrophage lysed in supernatant and intracellular bacteria in intact cell layer were harvested. Bacteria at day 0 and day 6 were plated on Middlebrook 7H10 agar plates in triplicate, plates were incubated for 3–weeks at 37°C and CFU were counted. The growth index was defined as the number of CFU per ml at day 6 divided by the number of CFU per ml on day 0 (Wong et al., 2007).

Culture supernatants from uninfected and infected hMDM were harvested at 24 h post-infection and stored at −80°C. TNF-α, IL-6, and IL-1β concentrations were measured from stored supernatants at the same time using an enzyme-linked immunosorbent assay kit according to the manufacturer's instructions (DuoSet ELISA, R&D).

A linear trend test was used to calculate the differences in P-value across all three virulence phenotypes. The P ≤ 0.05 indicated an increasing or decreasing of bacterial load, growth, or cytokine response, associated with the increased level of bacterial virulence from low to intermediate to high. The P-value was computed using independence_test function of R package coin developed by Hothorn et al. (2008). The frequencies of Mtb lineages among different virulence phenotypes were compared by Chi-square test. P-values of ≤ 0.05 were considered statistically significant. Statistical analyses were performed using statistical package R v3.3.1 (R Core Team, 2016). Graphs were generated by GraphPad Prism v7.03 (GraphPad Software, San Diego, California, USA). For analysis of whole genome sequencing (WGS), FASTQ data were mapped against the H37Rv reference (NC_000962.3) using BWA mem (Li, 2013), SNPs were called using GATK v3.3.0 (McKenna et al., 2010) in unified genotyper mode. Core genome positions that had a high quality SNP (>90% consensus, coverage >5x, genotype quality >30, mapping quality >30) in at least one strain were extracted using snp-sites (Page et al., 2016) and used as the input for maximum likelihood phylogenetic analysis using IQ-TREE v1.6 (Nguyen et al., 2015) with automatic model selection by ModelFinder (Kalyaanamoorthy et al., 2017). Mtb lineage was predicted by Mykrobe (Bradley et al., 2015).

To assess virulence of Mtb strains, THP-1 macrophages were infected with 153 clinical isolates and the lab strain H37Rv (Figure 1). We examined the interaction of macrophages and Mtb every day for 10 days after infection. Macrophages were completely lysed (≥90% cell lysis) by a few isolates at day 4 and by half of isolates at day 7. From day 8 onwards uninfected macrophages were unhealthy due to old culture medium. Therefore, we chose day 6 after infection as an optimal time to characterize the virulence phenotype of Mtb isolates. The percentage of cells lysed was measured repeatedly for 6 days after infection (Figure 2). During the first 3 days, most of the infected macrophages still adhered and were intact (< 5% cell lysis). On later days, we observed significant variation in the level of cell lysis, ranging from 0 to 100%. For example, on day 4, infected macrophages remained intact in 80 (52.3%) isolates, 5–10% were lysed in 9 isolates, 10–30% in 30 isolates, 30–50% in 14 isolates, 50–70% in 8 isolates, 70–90% in 4 isolates, and completely (≥90%) in 8 isolates (Supplementary Table S1).

We grouped clinical isolates into three different phenotypes of virulence based on macrophage lysis at 6 days after infection: low virulence if < 5% infected cells were lysed, high if ≥90% infected cells were lysed, and intermediate for the remainder. Of the 153 clinical isolates tested, 34 (22.2%) had low virulence, 46 (30.1%) had high virulence, and 73 (47.7%) were intermediate (Figure 2). The lab strain H37Rv exhibited intermediate virulence in this model, causing ~50% macrophage lysis.

We examined whether Mtb virulence phenotypes were associated with bacterial load in the sputum of PTB patients before treatment (Figure 1). The linear trend test showed increased Mtb virulence was associated with increased bacterial load (P = 0.02). The median Ct values of Mtb isolates with intermediate and high virulence were 2 and 3 cycles lower, respectively, than those of low virulence isolates, meaning that the numbers of bacilli in sputum were ~4 and 8 times higher for intermediate and high virulence isolates, respectively, than for low virulence (Figure 3).

153 Mtb isolates were first genotyped by LSP typing. In line with previous reports (Caws et al., 2008; Thwaites et al., 2008), the three main Mtb lineages known in Vietnam were observed in our isolate collection. East Asian/Beijing lineage represented 78/153 (51.0%) of isolates, Indo-Oceanic lineage 50/153 (32.7%), and Euro-American just 18/153 (11.8%). The lineage of a small proportion of isolates (7/153, 4.5%) was undefined due to limitations of the typing method (Caws et al., 2008). To investigate whether virulence is associated with East Asian/Beijing lineage, we combined Indo-Oceanic, Euro-American (including lab strain H37Rv) and the unknown group as non-Beijing lineage to compare with East Asian/Beijing. We found that isolates with high virulence phenotype were likely to belong to East Asian/Beijing lineage rather than non-Beijing lineage (P = 0.002, OR = 4.32, 95% confidence interval (CI) 1.68–11.13) (Table 1).

A phylogenetic tree constructed from the whole genome sequencing data of 125/153 Mtb isolates (28 were not sequenced due to missing clinical data for these patients) showed similar frequencies of Mtb lineages to that from LSP typing. 66/125 (52.8%) isolates belonged to East Asian/Beijing, 40/125 (32%) to Indo-Oceanic, 15/125 (12%) to Euro-American, 2/125 (1.6%) to Delhi/Central Asia lineage, and 2/125 (1.6%) were undefined. Consistently, we observed a high distribution of highly virulent isolates among the East Asian/Beijing lineage (Figure 4).

After using THP-1 cell line to characterize the virulence phenotypic diversity among the strains when in the same host environment, we used hMDM from three different donors to validate virulence phenotype and also to investigate the interaction of human macrophages and Mtb isolates through host immune response (Figure 1). We randomly selected 28 isolates representative of different phenotypes, comprising 8 strains with low virulence, 14 strains with high, and 6 strains with intermediate virulence, and infected them together with the intermediately virulent H37Rv into hMDM. In comparison to THP-1 cells, the lysis of hMDM was slower, with complete destruction first being observed 12 days after infection. Hence, the virulence phenotype of clinical isolates in hMDM was categorized based on the same criteria as for THP-1 cells (low virulence if < 5% infected cells were lysed, high if ≥90%, intermediate for the remainder) but at day 12 (Figure 5).

Generally there was a similar trend in the virulence phenotype of Mtb isolates in the different donors even though host cell lysis was slightly different for 9/29 isolates, with slightly lower levels in donor 3 than in others (Figure 5). The differences could be due to variations in host response. Comparing the two macrophage models, 6/8 (75%) isolates with low virulence in THP-1 exhibited the same phenotype in hMDM while 1/8 (isolate 19, 12.5%) changed to intermediate. Among isolates with high virulence in THP-1, 11/14 (78.6%) remained high across all 3 donors; whereas 3/14 (21.4%) became intermediate. Less consistency was observed with isolates having intermediate virulence. However, overall the results showed that the majority (75%) of isolates conserved their low or high virulence phenotype in both THP-1 and hMDM, with none changing from low to high or vice versa.

We next examined whether the virulence phenotype of Mtb clinical isolates was associated with their survival and growth in macrophages (Figure 1). Twenty four clinical isolates and H37Rv, whose virulence phenotypes were validated in both THP-1 and hMDM models, were infected to THP-1 cells at MOI 1; the number of intracellular CFU was determined at day 0 and both intra- and extracellular CFU measured at 6 days after infection. At day 0, there was no difference in the bacillary inoculum of isolates with low, high, and intermediate virulence (Figure 6A). At day 6, greater virulence was associated with an increased number of bacteria (linear trend test, P = 3.6 × 10⁻⁵). Between day 0 and day 6, an almost hundredfold increase in CFU was observed in highly virulent isolates, while in isolates with low virulence the CFU decreased over tenfold (Figure 6B). There was no difference in the growth rate of the strains by virulence phenotypes in 7H9 broth in vitro (linear trend test, P = 0.12) (Figure 6C). Therefore, the high virulence phenotype as indicated by rapid host cell lysis is likely to be highly associated with their survival and replication ability in THP-1.

We hypothesized that the virulence phenotypes of Mtb isolates were driven by their ability to modulate host cytokine response. To test this hypothesis, we infected hMDM from three independent donors with each of the 29 Mtb isolates described above (including H37Rv) (Figure 1). Cell supernatants were collected after 24 h and concentrations of TNF-α, IL-6, IL-1β were determined.

It was noted that 8/29, 9/29, and 12/29 isolates showed low, intermediate, and high virulence phenotype in hMDM, respectively (Figure 5). Linear trend tests showed significant associations between virulence phenotype and concentrations of TNF-α, IL-6, and IL-1β in 2/3 donors (Figure 7). Increased virulence was associated with a reduced level of TNF-α [linear trend test, P = 0.009 (donor 2), P = 0.003 (donor 3)] (Figure 7A), and IL-6 [linear trend test, P = 0.002 (donor 2), P = 0.0009 (donor 3)] (Figure 7B), but intriguingly, with an increased level of IL-1β [linear trend test, P = 0.002 (donor 1), P = 0.0005 (donor 2)] (Figure 7C). Combined cytokine data from three donors showed significantly decreased concentrations of TNF-α and IL-6 but increased concentration of IL-1β in more virulent isolates (Figure 7D).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.