Bacterial diversity dominates variable macrophage responses of tuberculosis patients in Tanzania

The Mycobacterium tuberculosis complex (MTBC) comprises nine human-adapted lineages that differ in their geographical distribution. Local adaptation of specific MTBC genotypes to the respective human host population has been invoked in this context. We aimed to assess if bacterial genetics governs MTBC pathogenesis or if local co-adaptation translates into differential susceptibility of human macrophages to infection by different MTBC genotypes. We generated macrophages from cryopreserved blood mononuclear cells of Tanzanian tuberculosis patients, from which the infecting MTBC strains had previously been phylogenetically characterized. We infected these macrophages ex vivo with a phylogenetically similar MTBC strain (“matched infection”) or with strains representative of other MTBC lineages (“mismatched infection”). We found that L1 infections resulted in a significantly lower bacterial burden and that the intra-cellular replication rate of L2 strains was significantly higher compared the other MTBC lineages, irrespective of the MTBC lineage originally infecting the patients. Moreover, L4-infected macrophages released significantly greater amounts of TNF-α, IL-6, IL-10, MIP-1β, and IL-1β compared to macrophages infected by all other strains. While our results revealed no measurable effect of local adaptation, they further highlight the strong impact of MTBC phylogenetic diversity on the variable outcome of the host–pathogen interaction in human tuberculosis.


Single cell preparation
MTBC clinical isolates were expanded in 7H9 broth supplemented with 10% ADC (5% bovine albumin-fraction V, 2% dextrose, 0.003% catalase), 0.5% glycerol (PanReac AppliedChem) and 0.1% Tween-80 (Sigma-Aldrich) under gentle agitation at 37 °C until cultures reached early exponential growth phase (i.e., OD 600 comprised between 0.5 and 0.6) before single cell preparation.Bacterial cultures were pelleted at 3000 × g for 5 min and washed with an equal volume of phosphate buffer solution (PBS) containing 0.05% Tween-80 (PBST).Bacterial pellets were resuspended in complete cell culture medium (RPMI supplemented with 10% foetal bovine serum) and sonicated for 2 min, 100% power at 25 °C (Grant digital ultrasonic bath).Sonicated bacterial suspension was subjected to centrifugation at 260 × g for 5 min, and the supernatant recovered and cryopreserved at − 80 °C in aliquots containing 5% glycerol final.Single cell suspensions were subjected to ten-fold dilution in PBST in triplicate and plated onto Middlebrook 7H11 agar plates supplemented with oleic acid, albumin, dextrose and catalase (OADC, BD 211886) before incubation at 37 °C, 5% CO 2 for up to 6 weeks.Colony forming units (CFUs) were monitored weekly.We selected strains that yielded comparable high CFUs per OD ratios (> 1.10 7 CFU/ml at OD 600 = 0.4) as to prevent the confounding effect of bacterial clumping on host cell death during the subsequent experimental infections 39 .

Human monocyte derived macrophages
Monocytes were isolated using magnetic CD14 human microbeads and following manufacturer's recommendations (Miltenyi, 130-050-021).Isolated monocytes were stained with anti-human CD14-FITC (clone MϕP9, BD Biosciences) and CD16-PE (clone B73.1, BD Biosciences).Samples were then acquired on a MACSQuant analyzer 10 (Miltenyi), and data analysed using FlowJo_V10.Isolated monocytes were seeded in tissue culture treated dishes or plates at 5-6 × 10 6 cells/cm 2 in complete medium containing M-CSF at 50 ng/ml (Miltenyi, 130-096-491) for 6 days.Culture medium was removed and monocyte-derived macrophage (MDM) monolayer washed twice with PBS and detached by trypsin-EDTA (Sigma-Aldrich) treatment, scrapping and counting before seeding 1 × 10 5 cells per well in 96 well tissue-culture treated plates.For infection, culture medium was replaced by the same volume of complete medium containing the corresponding volume of MTBC single cell suspension to reach a multiplicity of infection (MOI) of 0.1.We used a low multiplicity of infection to ensure viability of infected cells and approach physiological in vivo conditions 40 .CFUs of all inoculums were assessed for each infection experiment and did not differ significantly (Supplementary Fig. 5).Supernatants were collected and CFU assessed after 1, 4-and 7-days post-infection.Cellular morphology was monitored across indicated time point for each experiment as depicted in supplementary Fig. 6.

CFU assessment
At the indicated time points, cell culture medium was removed, and 100 μl of 0.1% Triton X-100 (Sigma-Aldrich) in water was directly added onto the infected cells and incubated at 37 °C, 5% CO 2 for 20 min.The macrophage lysate was serially diluted in PBST before plating onto 7H11-OADC agar plates for incubation at 37 °C, 5% CO 2 and monitoring for up to 6 weeks.

Statistical analysis
The lineage distribution and patient characteristics were described using summary statistics: proportions for categorical variables, and for continuous variables means with standard deviations or medians together with the inter-quartile range (IQR).Friedman test with Post-hoc analysis (Conover test) was used to compare bacterial load and replication rate between lineages.A Wilcoxon test was used to compare bacterial load and replication rate between matched vs. mismatched infection across the patients groups.Analysis of variance (ANOVA) and Kruskal-Wallis test were used to compare means and medians between independent samples.Tukey's multiple comparison was applied to analyse variances of repeated measures.. Principal component analysis (PCA) was applied to deduce variation in cytokine and chemokines using prcomp function in R and visualized using factoextra package.GraphPad prism version 8.2.1 and R studio version 4.1.3were used for data analysis and graphics.

MTBC lineages differ in their replication capacity within macrophages from TB patients
Patients' MDMs were systematically infected in parallel with each individual MTBC-lineage representative.We observed a significant difference in mean CFU/well recovered in macrophages after 1-day post-infection (dpi) (Friedman, p < 0.0001, Fig. 2A).Statistical analysis adjusted for multiple comparisons revealed that intra-cellular bacterial counts were significantly reduced for L1 strain infections compared to infections with L2, L3 and L4 strains (Conover's, p adj = 0.0001, p adj = 0.01 and p adj < 0.0001, respectively).The bacterial load retrieved 24 h postinfection was at least equal or superior to the inoculum.Therefore, the lower bacterial load of L1 strains one day post-infection reflects a lag phase in the early stages of intra-cellular replication compared to other lineages.We also observed a significantly different replication rate across the infecting strains between 1 and 7dpi (Friedman, p = 0.0005, Fig. 2B).More specifically, infections with the L2 representative strain resulted in significantly higher replication rates compared to infections with L1 or L3 and L4 strains (Conover's, p adj = 0.0063, p adj = 0.0016, p adj = 0.0537, respectively).
We then further grouped our analysis by accounting for the genotype of the MTBC strain that had originally infected the TB patients from whom the MDMs were obtained (Fig. 3A).Matched infection of L1 patient MDMs at 1dpi resulted in a significantly lower bacterial load (Wilcoxon test, p = 0.0076) compared to that of mismatched infections (Fig. 3A).Overall, the replication rates between matched and mismatched infections did not differ significantly (Fig. 3B).However, upon further stratification across individual mismatched-infecting strains, macrophage infections by L2 resulted in a significantly higher replication rate in MDMs derived from L2-and L3-infected patients in particular (Fig. 3C).We quantitatively assessed the presence of 13 cytokines and chemokines as well as HGMB-1, a biomarker of cell death, in the supernatants of macrophages infected ex vivo.All analytes but IL-12p70 and IL-23 fell within the limit of quantification.To visualize the dominant variables driving the inflammatory responses of MDMs infected with the various endemic MTBC lineages, we performed a principal component analysis (PCA).The first two components captured 39.5% and 17.2% of the variance, respectively (Fig. 4A).When grouping data points by the lineage originally infecting the patient in vivo (Fig. 4B), no systematic pattern was identified.However, upon grouping by the lineage used to infect the MDMs ex vivo (Fig. 4C), a red cluster, representing L4, predominantly overlapped with the first and fourth quadrant.This indicates that the inflammatory responses induced by L4 infections is significantly different from other lineages (Fig. 4D-H) and captured by PC1 (Fig. 4C).The main predictor variables driving PC1, which all displayed positive loadings, are TNF-α, IL-6, MIP-1β, IL-10 and IL-1β (contribution > 10% of the variation, Fig. 4A and supplementary Fig. 4).Corroborating the PCA observations, supernatants of MDMs infected with L4 consistently induced higher levels of TNF-α, IL-6, MIP-1β, IL-10 and IL-1β than those infected by any other MTBC lineage (Fig. 4D-H).

Plasma inflammatory markers are dominated by the disease status
We next sought to investigate whether specific host-pathogen association signals may also translate in vivo into a differential accumulation of inflammatory mediators in the plasma of Tanzanian TB patients.For this, we analysed the host inflammatory responses from active TB patient plasma samples collected at the time of recruitment and before treatment initiation.Plasma samples from 92 active TB patients infected with the most dominant MTBC sub-lineages were analysed (n = 23 patients for each of the four main phylogenetic groups    cytokines and chemokines, IP-10, MIP-1β and MCP-1 fell within the limit of quantification and were statistically differentially accumulating between controls and patients.Compared to plasma levels from TB suspects from which TB diagnosis was excluded (n = 20; clinical characteristics provided as supplementary table 2), the plasma of TB patients had more IP-10 (Mann-Whitney, p = 0.0007) and less MIP-1β (p < 0.0001) as well as less MCP-1 (p < 0.0001) (Fig. 5A-C, upper panels).Moreover, the amount of MIP-1β and MCP-1 differed across patients when stratified by the infecting MTBC lineage (Kruskal-Wallis, p = 0.0141 and p = 0.0114 respectively, Fig. 5B, C, lower panels).Pair-wise comparisons revealed that L1-infected patients harboured more MIP-1β in their plasma compared to L3-infected patients (p adj = 0.017, Fig. 5B, lower panel), and that L2-infected patients accumulated more MCP-1 in their plasma compared to L1-infected patients (p adj = 0.012, Fig. 5C, lower panel).

Discussion
In this study, we recapitulated ex vivo the outcome of the first encounter between macrophages from TB patients with MTBC strains genetically either matching or not those that had infected the patients.We observed significant differences in the early recovery of bacteria from macrophages infected by the L1 strain compared to L2, L3 and L4 (Fig. 2A).This difference was further pronounced and remained significant between matched and mismatched L1infections (Fig. 3A).We also observed significant differences in the replication rate across lineages with L2 exhibiting the highest intra-cellular replication rate (Fig. 2B).This pattern also remained significant in matched but also mismatched L3 patients' infections (Fig. 3C).The cytokine production by macrophages was dominated by a higher inflammatory response to L4 infections compared to the three other MTBC lineages.Finally, we reported that the amount of MIP-1β and MCP1 accumulated most markedly in the plasma of L1-infected patients and L3-infected patients respectively.Our observations suggest that L1 strains are generally less prone to survive after invading human macrophages.This finding is consistent with the lower virulence of L1 strains compared to other lineages, as observed in guinea pigs 42 and mice models 23 .L1 infections in humans also showed shorter time-to-sputum conversion while resulting in a greater occurrence of extra-pulmonary TB compared to L2-4 27 .A lower virulence profile of L1 is also in line with the recent description of a longer incubation time before the apparition of TB symptoms in regions of the world where L1 strains are highly prevalent 43 .Our study included all three, so-called "modern" lineages, (L2, L3 and L4) for which an increase in virulence, including intra-cellular replication, has been associated to the loss of the genomic region known as "TbD1" 44 .That association goes in line with our observation of an increased survival and proliferation capacity of all three modern stains compared to L1 starting one day post infection.The highest intra-cellular replication of L2 is also consistent with previous studies linking the epidemiological success of L2 in different populations with a higher replication rate in various cellular models [44][45][46][47] .Altogether and using intra-cellular survival and replication as MTBC pathogenesis read-outs, the behaviour of L1 and L2 strains within macrophages derived from TB patients mostly corroborate previous reports that were based on animal models or human cells from healthy blood donors and cell lines.Moreover, our findings show that macrophages from TB patients infected in vivo with a particular MTBC genotype did not appear more susceptible or resistant to ex vivo infection by a strain belonging to the same MTBC genotype (Fig. 3A, B).We did not actively assess differences if the replication rates of the individual strains in broth cultures would correlate with their replication within macrophages.The bacterial clinical isolates were not passed and expanded directly from an aliquot obtained following culture of patients' sputa to minimise genetic drift and adaptation to artificial culture conditions.Time to culture recovery and exponential growth phase differed between strains yet variable sample concentration and viability may likely have contributed to this phenomenon.
It is well established that susceptibility to a given infectious disease is not necessarily reflected by the level of pathogen replication but instead can be primarily due to a detrimental modulation of the host inflammatory response, which in turn will depend on the host and pathogen genetic background.For instance, the phenotypic characterization of two independent MTBC outbreaks was associated immunologically to an inhibition of the innate immune cytokine response 48,49 .In contrast, a hyper-inflammatory syndrome lies behind most severe forms of viral infections 50 .In this context, we have compared the inflammatory responses characterizing the TB patients to their blood-derived macrophages in response to infection by matched or mismatched MTBC strains.The cytokine production by macrophages 1 day post infection was dominated by a higher inflammatory response to L4 infections compared to the three other MTBC lineages.This response was mediated by several proinflammatory cytokines, IL-1β, IL-6, TNF-α and the chemoattractant MIP-1β, as well as a very potent regulatory cytokine, IL-10.The higher inflammatory response induced by L4 corroborates findings from several studies 23,51 .Nonetheless, Sarkar et al. 52 reported higher levels of TNF-α production in response to MDMs infection by L3, while Wang et al. 51 reported that LAM strains (L4) induced comparable inflammatory response to that of Beijing strain (L2).It was somewhat unexpected that L4 infections would exhibit such a distinct phenotype compared to the other modern lineages (L2 and L3), as these have been associated with low inflammatory responses in other studies 26,49,53,54 .Although studies linking immunological phenotypes to MTBC genotypes are limited, it was reported that production of phenolic glycolipid is one of the bacterial factors the release of inflammatory mediators including TNF-α in MDMs infected with HN878, an L2 outbreak, compared to H37Rv, the prototypical L4 representative 48 .The most tangible explanation supporting these discrepancies across different studies lies on the increasing appreciation of intra-lineage diversity and as a consequence, lineage trends cannot be concluded from studying a single representative strain.Yet, when looking at inflammatory mediators accumulating in the plasma of our TB patients, MIP-1β and MCP-1 concentrations were reduced compared to controls.And while previous evidence already suggested this for MIP-1β 55 , we report here an even more pronounced reduction in L4-infected patients overall compared to L1-infected patients.This finding also adds on the report from Mihret et al. 56 .describing lower MIP-1β quantification in plasma from L4-compared to L3-infected patients 56 .These results all contrast the increased release of MIP-1β by L4-infected macrophages suggesting that the increased response observed at cellular level ex vivo does not necessarily translate in its accumulation at systemic level in vivo.Alternatively, it may be that other cellular sources such as T cells may contribute to the specific accumulation of MIP-1β during L1-mediated TB 57,58 .Importantly, MIP-1β has been shown to play a positive role in supressing MTBC intracellular growth and was associated with apparent resistance to MTBC infection in healthcare workers 59,60 .Taken together, the higher levels of MIP-1β in the plasma of L1 -infected TB patients could partially explain the increased time to progression to disease of TB patients in L1-endemic areas 43 .We also observed an overall accumulation of IP-10 that appeared specific of TB disease that was not statistically different across patients grouped by their infecting lineages and that further support the diagnostic potential of this non-sputum biomarker [61][62][63][64] .MTBC induced cell death can be either favourable or detrimental for the replication of mycobacteria in vitro and in vivo 65,66 .With a substantial detection of HMGB-1 in the supernatants of uninfected macrophages (supplementary Fig. 7), our attempt to assess this phenotype using this particular parameter was not conclusive and further experiment using alternative and more specific proxies of individual cell death pathways is warranted.
While previous studies looking at the phenotypic consequences of MTBC genetic variability have been using animal models, cell lines or cells derived from healthy individuals, we reported here the impact of MTBC lineages on the response of macrophages originating from TB patients themselves.Yet, the intra-cellular bacterial load or macrophage inflammatory responses were not significantly impacted by the fact that the MTBC lineage that originally infected the TB patients from whom the macrophage originated phylogenetically matched the strain used to infect the patients' cells ex vivo.Our results provide no evidence on a preferential susceptibility or resistance of macrophages from TB patients infected in vivo by a given MTBC genotype upon ex vivo re-infection by the same MTBC genotype.Instead, we found that the macrophage responses were uniformly dominated by the MTBC strains used for the ex vivo infection.This finding may reflect previous results, indicating that the main MTBC lineages circulating in Tanzania have only been introduced relatively recently (i.e. during the last 300 years), and might therefore have had insufficient time to measurably adapt to their human host population 10 .
Alternatively, subtle local adaptation might be masked by more fundamental and inherent differences in virulence characteristic of the main MTBC lineages which were repeatedly observed 2 .
Our study has limitations.First, our study would have benefited from the inclusion of a reference group made of MDMs from non-Tanzanians and/or non-endemic MTBC strains to assess host fitness in allopatric scenarios.For instance, we previously estimated that L2.2.1 was introduced around 20 years ago 10 and as such, may be considered an allopatric MTBC genotype.Second, we did not investigate the phenotypes exhibited by MDMs from Tanzanians who did not have a history of TB infection, which would have informed on macrophage fitness in the general population.This is particularly relevant as trained immunity following BCG vaccination or TB disease can affect monocyte response upon re-stimulation or re-infection 67,68 .Third, while the sample size sufficed to draw conclusions on the impact of lineages overall, stratification by patients' lineage led to small group sizes that may have limited our capacity to detect more subtle differences between patient groups.Finally, functional studies investigating the differences seen in this study would be manifold and fall beyond the scope and the capacity of this investigation for which additional sample material would have had to be collected.
In conclusion, local adaptation of MTBC strains to their human host population has been proposed to be at the basis of the MTBC phylogeography.However, at the level of the host macrophage at least, our results do not support this notion.Nevertheless, our study highlights the relevance of MTBC phylogenetic diversity on TB pathogenesis, with important implications for vaccine development.

Figure 3 .
Figure 3. Ex vivo matched and mismatched infections of TB patient-derived macrophages.(A) Bacterial load 1-day post-infection and (B, C) replication rate 7 days post-infection within TB patients' macrophages.Vertical dotted lines separate data according to each independent MTBC lineage that originally infected the TB patient in vivo.Data were further grouped by matched (+) or mismatched infection of the patients' macrophages (−) with pooled mismatched infections (B) or individual strain infection (C).Box plots present median with interquartile range and overlaying dots represent data from individual patient macrophage preparations (n = 28, L1 = 6, L2 = 8, L3 = 8, L4 = 6).

Figure 4 .
Figure 4. MTBC lineage 4 dominates cytokine and chemokine responses of TB patients' macrophages infected ex vivo.(A) Bar plot represents the proportion of variance explained by the top two principal components.Biplot represents the contribution of individual cytokines and chemokines to PC1 and PC2.Principal components PC1 & PC2 coloured by lineage originally infecting the patients (B) or by lineages used to infect the MDMs ex vivo (C).Smaller dots represent individual infections and bigger dots depict centroids for the respective clustering option.Box plots and individual data point depicting measured concentrations of (D) TNF-α, (E) IL-6, (F) IL-1β, (G) MIP-1β and (H) IL-10 secreted by patients' MDMs after 24 h of infection by the indicated representative lineage. 2 way ANOVA with Dunnett's multiple comparisons test, **** p < 0.0001.

Figure 5 .
Figure 5. IP-10 and MIP-1β characterizes host inflammatory response within TB patients' plasma.Box plots and individual data points depicting quantified concentrations of (A) IP-10, (B) MIP-1β and (C) MCP-1 that are differentially accumulating in plasma samples from active TB patients compared to non-TB symptomatic controls (Mann-Whitney, *** p < 0.001 and **** p < 0.0001).Respective lower panels depict that MIP-1β and MCP-1 but not IP-10 differentially accumulates in the plasma of TB patients infected with the different endemic strains (Kruskal-Wallis p = 0.0141 and p = 0.0114 respectively).MIP-1β most markedly accumulate in L1-infected patients compared to L4 and MCP-1 in L2-infected patients compared to L1 (Dunn's multiple comparisons test, p adj = 0.017 and p adj = 0.012 respectively).