Engineered Mycobacterium tuberculosis triple-kill-switch strain provides controlled tuberculosis infection in animal models

Abstract

Human challenge experiments could accelerate tuberculosis vaccine development. This requires a safe Mycobacterium tuberculosis (Mtb) strain that can both replicate in the host and be reliably cleared. Here we genetically engineered Mtb strains encoding up to three kill switches: two mycobacteriophage lysin operons negatively regulated by tetracycline and a degron domain-NadE fusion, which induces ClpC1-dependent degradation of the essential enzyme NadE, negatively regulated by trimethoprim. The triple-kill-switch (TKS) strain showed similar growth kinetics and antibiotic susceptibilities to wild-type Mtb under permissive conditions but was rapidly killed in vitro without trimethoprim and doxycycline. It established infection in mice receiving antibiotics but was rapidly cleared upon cessation of treatment, and no relapse was observed in infected severe combined immunodeficiency mice or Rag^-/- mice. The TKS strain had an escape mutation rate of less than 10^-10 per genome per generation. These findings suggest that the TKS strain could be a safe, effective candidate for a human challenge model.

Fig. 1. The dual-lysin kill switch is effectively bactericidal in Mtb both in vitro and in vivo.

a, Regulatory scheme of aTc-inducible lysin expression. b, The kill curve of dual-lysin Mtb strain in vitro. Each dot shows the mean bacterial burden from biological triplicates. The error bars represent the standard deviation. The dashed line shows the CFU lower limit of detection (LLOD). c, The escape mutation rate of dual-lysin Mtb strain in the absence of aTc. The mutation rate was calculated from escape frequency of 20 independent cultures. The escape frequency is presented as the mean, with the error bar showing the standard deviation. d, The escape density plot shows mutations in L5-lysin kill switch and D29-lysin kill switch from 20 independent escape mutants. The mutation type and frequency in each individual kill switch are noted. e, The scheme of C57BL/6 mouse infection with Mtb dual-lysin strain. p.i., post infection. f, The clearance kinetics of Mtb dual-lysin strain in C57BL/6 mice with doxy depleted at day 14 or day 28. Each symbol shows the mean bacterial burden from five mice. The error bars represent the standard deviation. The dashed line shows the CFU LLOD. g, The scheme of doxy-dependent NHP infection with the dual-lysin kill-switch strain. h–m, The NHP infection outcome of the dual-lysin Mtb strain supplemented with doxy for 2 or 6 weeks; the outcome is judged by total thoracic CFU (h), lung CFU (i), lymph node (LN) CFU (j), gross pathology score (k), number of lung granulomas at necropsy (l) and number of involved LNs (m). Each symbol represents a macaque, and the line represents the median. Panels e–m created with BioRender.com, released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license. Source data

Fig. 2. Mycobacteria proteins fused with a ddTMP tag are subjected to TMP-dependent degradation via ClpC1-ClpP1P2 protease.

a–c, mCherry fused with ddTMP at the N-terminus (a), the C-terminus (b) or null (c). The ddTMP at either the N- or C-terminus manifests TMP-dependent degradation in Msm. d, ddTMP–mScarlet shows a TMP-dependent degradation in Mtb. e, Flow cytometry with clpC1- or clpX-knockout Msm indicates that the TMP-dependent degradation is mediated via protease ClpC1-ClpP1P2. The bottom row shows the RFP steady state measured in mid-log culture supplied with TMP. ClpX or clpC1 were knocked down via a gradient of aTc in the absence of TMP. f, The ClpC1-dependent degradation mechanism of the ddTMP tag in mycobacteria. All experiments have been repeated independently three times and show the representative results. Panel f created with BioRender.com, released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license. Source data

Fig. 3. TMP-dependent NadE degradation leads to Mtb death.

a, The genetic construct of tagging Mtb gene nadE with C-terminus ddTMP. b, The degradation kinetics of NadE-mScarlet-ddTMP were measured by quantifying mScarlet intensity per cell via microscope. The degradation curve is fitted into a one-phase exponential decay model. c, The growth curve of the Mtb nadE-ddTMP strain with various TMP concentrations in 7H9 media. Each dot represents the mean of biological triplicates. The data are presented as mean ± standard deviation. OD₆₀₀, optical density at 600 nm. d, The kill curve of the Mtb nadE-ddTMP strain in the absence of TMP. Each dot represents the mean of biological triplicates. The data are presented as mean ± standard deviation. e,f, The mutation rate of Mtb nadE-ddTMP strain (e) and its escape mutation density plot (f) are determined by whole-genome sequencing. The escape rate is determined via escape frequency of 20 independent cultures. The escape rate is presented as the mean, with error bars showing standard deviation. The dashed line indicates the LLOD. Source data

Fig. 4. The TKS strain manifests normal fitness at permission condition yet rapid killing and low escape rate at restrictive condition.

a, The growth curve of TKS and H37Rv strains at permissive condition. The dotted line shows the LLOD as OD 0.001. Each dot represents the mean of biological triplicates. The data are presented as mean ± standard deviation. b, The quantified bacterial morphology of strains H37Rv (n = 1,068) and TKS (n = 1,614). Cohen’s d was calculated to show effect size between H37Rv and TKS strains. c–f, The MIC of H37Rv and TKS strains against isoniazid (INH) (c), rifampin (RIF) (d), ethambutol (EMB) (e) and moxifloxacin (Moxi) (f) at permissive condition. Each dot represents the mean of biological triplicates. The curve shows a sigmoidal regression of relative viability as a function of antibiotics concentration on a logarithmic scale. g, The in vitro kill curve of TKS strain in the absence of aTc and/or TMP. The dotted line shows the LLOD as 10 CFU ml⁻¹. Each dot represents the mean of biological triplicates. The data are presented as mean ± standard deviation. h, The escape rate of the TKS strain in the absence of aTc and/or TMP from two independent experiments. Each experiment contains 20 independent bacteria culture for escape frequency quantitation. In each experiment, the escape rate is calculated from escape frequency of 20 culture and is presented as the mean, with the error bars showing standard deviation. The dashed line shows the LLOD as 3.8 × 10⁻¹⁰. The error bars represent standard deviation.

Fig. 5. The TKS strain is rapidly cleared in mouse without discernible relapse.

a, The scheme of clearance study in C57BL/6 mice. The mice were kept on doxy/TMP chow 3 days before infection until day 28 post infection. Lung and spleen CFU were enumerated at indicated timepoint. b,c, Lung (b) and spleen (c) CFU of the C57BL/6 mouse experiment. Each dot represents the CFU mean from n = 5 mice. The error bars denote standard deviations. The P values were calculated via two-tailed Student’s t-test compared between CFU in regular chow group and the doxy/TMP chow group at a given timepoint. d, The proportion of CD4⁺ T cells of the C57BL/6 mouse with activation markers at day 84. e, The proportion of CD4⁺ memory T cells of the C57BL/6 mouse at day 84. f, The proportion of CD8⁺ T cells of the C57BL/6 mouse with activation markers at day 84. g, The proportion of CD8⁺ memory T cells at day 84. The P values in d–g were calculated by a Student’s t-test with a multiple comparison corrected with Tukey method. The grey data are from uninfected mice, blue data are from mice switched to regular chow and green data are from mice with doxy/TMP chow. h,j, The scheme of relapse study in SCID mice (h) and Rag^−/− mice (j). The mice were kept on doxy/TMP chow 3 days before infection until day 14 post infection. The mice were then fed with regular chow until indicated timepoint. The doxy/TMP chow were later given to mice for 10 weeks for relapse assessment. Lung CFUs were enumerated at indicated timepoints. i,k, The kinetics of lung CFU enumerated in the SCID (i) and Rag^−/− (k) relapse experiment. The bacteria burden was assessed with n = 5 mice at each timepoint. Each dot represents the CFU from one mouse. The error bars denote the mean and standard deviation. The dotted line shows the CFU LLOD. Panels a, h and j created with BioRender.com, released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license.

Extended Data Fig. 1. Gating strategy for flow cytometry analysis of mouse lung cells.

(A) General gating strategy for CD4 and CD8 T cells. (B) Gating strategy for CD4 T cell memory populations. Central memory = CD62L⁺ CD44⁺ (Q2), effector memory = CD62L- CD44⁺ (Q1), resident memory = CD62L⁻ CD44⁺ CD69⁺. (C) Gating strategy for CD8 T cell memory populations. Central memory = CD62L⁺ CD44⁺ (Q2), effector memory = CD62L⁻ CD44⁺ (Q1), resident memory = CD62L⁻ CD44⁺ CD69⁺.

Extended Data Fig. 2. Spleen CFU in the SCID experiment.

The SCID mice were fed permissive chow for 14 days, followed by restrictive chow till day 98. After that, the mice received permissive chow till day 168. Spleen CFU were enumerated on 7H11 plates. Each time point include 5 mice. The average and standard deviation are shown on plots. Green indicates mice in permissive condition and blue indicates restrictive condition. The dashed line is the lower limit of detection.

Extended Data Fig. 3. TKS strain burden in extrapulmonary tissues in the Rag^−/− relapse experiment.

The Rag^−/− mice were fed permissive chow for 14 days, followed by restrictive chow till day 56. After that, the mice received permissive chow till day 126. The TKS bacteria burden were enumerated in multiple organs, including (A) spleen, (B) mediastinal lymph node, (C) liver and (D) bone marrow. Each time point include 5 mice. The average and standard deviation are shown on plots. Green indicates mice in permissive condition and blue indicates restrictive condition. The dashed line is the lower limit of detection.