TNF-α antagonists differentially induce TGF-β1-dependent resuscitation of dormant-like Mycobacterium tuberculosis

Abstract

TNF-α- as well as non-TNF-α-targeting biologics are prescribed to treat a variety of immune-mediated inflammatory disorders. The well-documented risk of tuberculosis progression associated with anti-TNF-α treatment highlighted the central role of TNF-α for the maintenance of protective immunity, although the rate of tuberculosis detected among patients varies with the nature of the drug. Using a human, in-vitro granuloma model, we reproduce the increased reactivation rate of tuberculosis following exposure to Adalimumab compared to Etanercept, two TNF-α-neutralizing biologics. We show that Adalimumab, because of its bivalence, specifically induces TGF-β1-dependent Mycobacterium tuberculosis (Mtb) resuscitation which can be prevented by concomitant TGF-β1 neutralization. Moreover, our data suggest an additional role of lymphotoxin-α-neutralized by Etanercept but not Adalimumab-in the control of latent tuberculosis infection. Furthermore, we show that, while Secukinumab, an anti-IL-17A antibody, does not revert Mtb dormancy, the anti-IL-12-p40 antibody Ustekinumab and the recombinant IL-1RA Anakinra promote Mtb resuscitation, in line with the importance of these pathways in tuberculosis immunity.

Fig 1. Human, in-vitro granulomas mimic dormant-like Mtb features.

(A) Representative bright-field microscopy pictures of 3D in-vitro granulomas formed 8 days post-infection with Mtb H37Rv compared to uninfected PBMCs. (B) Representative structure of in-vitro granulomas under bright-field (left panel) and fluorescence microscopy (right panel). Monocytes/macrophages were labeled in orange and CD4⁺ T cells in green. (C) Percentages of auramine-O- (green) and Nile red-positive (red) Mtb quantified by fluorescence microscopy (mean ± SEM from 7 independent donors) before (1 day post-infection) or after (8 days post-infection) granuloma formation. Statistical analysis was performed using a generalized linear mixed-effects model; ****, p<0.0001. (D) Relative expression values for icl1, gltA1, nuoB and ctaD after granuloma formation (8 days post-infection) determined by qRT-PCR (mean ± SEM from 4 independent donors). Results are expressed as fold change in log2 scale relative to an aerobically-grown, mid-log Mtb culture, using 16S rRNA as the endogenous control.

Fig 2. Human, in-vitro granulomas corroborate risk of LTBI reactivation linked to treatment with TNF-α-targeting biologics.

(A) Percentages of auramine-O- (green) and Nile red-positive (red) Mtb quantified by fluorescence microscopy (mean ± SEM from 7 independent donors) following 4 days of exposure with an isotype control (Iso), adalimumab (ADA), etanercept (ETA), ustekinumab (UST), anakinra (ANA) or secukinumab (SEK). Statistical analysis was performed using a generalized linear mixed-effects model; n.s., not significant; *, p<0.05; **, p<0.01; ****, p<0.0001. (B) Cytokine levels measured in supernatants of untreated in-vitro granulomas 8 days post-infection (median with interquartile ranges, minimum and maximum values for 5 independent donors).

Fig 3. Unequal interference of TNF-α-targeting biologics ADA and ETA with granuloma-induced Mtb dormancy does not relate to differential bioavailability.

(A) TNF-α concentration in supernatants on days 1, 4 and 8 post-infection with Mtb H37Rv (values for 7 independent donors are represented by line-connected circles and bars picture the mean concentration). Statistical analysis was performed using a Wilcoxon paired test. (B-D) At day 4 post-infection, granulomas were exposed for four additional days to an isotype control (Iso), adalimumab (ADA), or etanercept (ETA) at 10 ng/ml. (B) Percentages of auramine-O- (green) and Nile red-positive (red) Mtb quantified by fluorescence microscopy (mean ± SEM from 9 independent donors). (C) Mtb bacterial load quantified by CFU assessment (values for 4 independent donors are represented by line-connected circles and bars picture the mean concentration). Statistical analysis was performed using a paired t test. (D) Percentage of rifampicin (Rif)-tolerant Mtb quantified by CFU (mean ± SEM from 2 independent donors). (E) Percentages of auramine-O- (green) and Nile red-positive (red) Mtb quantified by fluorescence microscopy following four days of exposure to Iso (full squares), ADA (full circles/continuous line), or ETA (open circles/dotted line) at the indicated concentrations (mean ± SEM from 2 independent donors). Unless stated differently, statistical analysis was performed using a generalized linear mixed-effects model; n.s., not significant; *, p<0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001.

Fig 4. Late TNF-α production by Mtb-induced granulomas originates from T cells as well as macrophages.

Frequencies of TNF-α-producing macrophages (A) and CD4⁺ (B) or CD8⁺ T cells (C) from uninfected PBMCs (UI) or granulomas 6 and 8 days post-infection with Mtb H37Rv (median with interquartile ranges, minimum and maximum values for 4 independent donors). Statistical analysis was performed using Friedman test; *, p<0.05.

Fig 5. ADA specifically mediates TGF-β1-dependent Mtb resuscitation.

(A) Active TGF-β1 concentration in supernatants of granulomas 8 days post-infection with Mtb H37Rv and after 4 days of exposure to either adalimumab (ADA), etanercept (ETA) or an isotype control (Iso) (values for 4 independent donors are represented by line-connected circles and bars depict mean concentration). Statistical analysis was performed using Friedman test. (B-D) Averaged percentages of auramine-O- (green) and Nile red-positive (red) Mtb quantified by fluorescence microscopy after 4 days of exposure to: (B) 0, 8, 40 or 200 pg/ml of recombinant TGF-β1 in the absence (full circles/continuous line) or presence (open circles/dotted line) of a TGF-β1-blocking antibody (mean ± SEM from 3 independent donors); (C) an isotype control (Iso), adalimumab (ADA), or etanercept (ETA) in the absence or presence of a TGF-β1-neutralizing antibody (+ α-TGF-β1) (mean ± SEM from 5 independent donors); and (D) ADA or ADA Fab fragment (ADA-Fab) in the absence or presence of a TGF-β1-neutralizing antibody, ETA or Iso. Statistical analysis was performed using a generalized linear mixed-effects model; n.s., not significant; *, p<0.05, **, p<0.01; ***, p<0.001; ****, p<0.0001. For (B) all comparisons were performed against the untreated control. For (C-D) only the most relevant comparisons were plotted for clarity reasons but results from all combinations are available on S2 Table, panels A and B respectively.

Fig 6. ETA-specific interference with CD4⁺ and CD8⁺ T cell-derived LT-α sustains mild Mtb resuscitation.

(A) LT-α concentration in supernatants on days 1, 4 and 8 post-infection with Mtb H37Rv (values for 7 independent donors are represented by lined-connected circles and bars depict mean concentration). (B-C) Frequencies of LT-α-producing CD4⁺ (B) and CD8⁺ T cells (C) from uninfected PBMCs (UI) or granulomas 6 or 8 days post-infection with Mtb H37Rv (median with interquartile ranges, minimum and maximum values from 4 independent donors). (D) Percentages of auramine-O- (green) and Nile red-positive (red) Mtb quantified by fluorescence microscopy (mean ± SEM from 4 independent donors) after four days of exposure to either an isotype control (Iso), etanercept (ETA) or the Fab fragments from adalimumab (ADA-Fab) or an anti-LT-α antibody (α-LTα-Fab). Statistical analysis was performed using a generalized linear mixed-effects model; ***, p<0.001; ****, p<0.0001.