Depletion of the DarG antitoxin in Mycobacterium tuberculosis triggers the DNA-damage response and leads to cell death.

Of the ~80 putative toxin-antitoxin (TA) modules encoded by the bacterial pathogen Mycobacterium tuberculosis (Mtb), three contain antitoxins essential for bacterial viability. One of these, Rv0060 (DNA ADP-ribosyl glycohydrolase, DarGMtb ), functions along with its cognate toxin Rv0059 (DNA ADP-ribosyl transferase, DarTMtb ), to mediate reversible DNA ADP-ribosylation (Jankevicius et al., 2016). We demonstrate that DarTMtb -DarGMtb form a functional TA pair and essentiality of darGMtb is dependent on the presence of darTMtb , but simultaneous deletion of both darTMtb -darGMtb does not alter viability of Mtb in vitro or in mice. The antitoxin, DarGMtb , forms a cytosolic complex with DNA-repair proteins that assembles independently of either DarTMtb or interaction with DNA. Depletion of DarGMtb alone is bactericidal, a phenotype that is rescued by expression of an orthologous antitoxin, DarGTaq , from Thermus aquaticus. Partial depletion of DarGMtb triggers a DNA-damage response and sensitizes Mtb to drugs targeting DNA metabolism and respiration. Induction of the DNA-damage response is essential for Mtb to survive partial DarGMtb -depletion and leads to a hypermutable phenotype.

INTRODUCTION

Toxin‐antitoxin (TA) systems are ubiquitously present in prokaryotic genomes and consist of a toxic protein that inhibits an essential cellular process and a counteracting antitoxin that binds to and neutralizes the toxin (Yamaguchi et al., 2011). TA systems were originally discovered due to their ability to prevent plasmid loss by post‐segregational killing (Ogura and Hiraga, 1983; Gerdes et al., 1986). They have subsequently been implicated in various cellular pathways including phage defense, genome stabilization, and bacterial persistence (Szekeres et al., 2007; Gerdes and Maisonneuve, 2012; Harms et al., 2018).

The phyletic distribution of TA modules indicates that Mycobacterium tuberculosis (Mtb), the causative agent of tuberculosis, harbors an expanded repertoire of 79 putative TA loci (Pandey and Gerdes, 2005; Sala et al., 2014; Slayden et al., 2018). Individual deletions of many of these TA loci exhibit few, if any phenotypic defects, likely due to functional redundancy (Singh et al., 2010; Tiwari et   al., 2015). However, 3 of the 79 TA modules in Mtb harbor antitoxins that are essential for viability of the organism, namely Rv0060, Rv1044, and Rv1990c (DeJesus et   al., 2017). Rv1044 is uncharacterized, and Rv1990c was recently identified as MbcA, an antitoxin that neutralizes a NAD+ phosphorylase toxin, MbcT (Freire et   al., 2019). Rv0060 (DarG), and its cognate toxin, Rv0059 (DarT), are the focus of this study.

Within mycobacteria, the darT‐darG locus is found only in species belonging to the Mtb complex (Kanehisa and Goto, 2000). However, orthologous genes have been identified in the extremophile Thermus aquaticus (Taq) and in enteropathogenic E. coli (EPEC) (Jankevicius et al., 2016; Lawaree et   al., 2020). In vitro experiments established that DarT, DarT, and DarT can ADP‐ribosylate single stranded DNA at thymidine residues (Jankevicius et   al., 2016; Lawaree et   al., 2020). In both E. coli and T. aquaticus, the antitoxin, DarG, physically interacts with DarT, leading to toxin neutralization. In addition, DarG, DarG, and DarG can enzymatically reverse the action of their cognate toxins by removal of the ADP‐ribose moiety (Jankevicius et   al., 2016; Lawaree et   al., 2020). Taken together, biochemical characterization demonstrates that DarT‐DarG mediate reversible ADP‐ribosylation of DNA (Jankevicius et   al., 2016).

While the biological roles of DarT‐DarG remain unknown, in vivo studies performed on the orthologous E. coli DarTG system find that ADP‐ribosylation by DarT halts DNA replication. The resulting DNA lesions are repaired by two DNA‐repair pathways: RecF‐mediated homologous recombination and nucleotide excision repair (Lawaree et   al., 2020). In this study, we characterize the Mtb orthologs of the DarTG system by delineating the cellular processes affected by genetic perturbation of the darT locus.

RESULTS

DarT and DarG form a toxin‐antitoxin pair that is dispensable for growth in vitro, growth in mice, persistence in mice, and resistance to various stresses

Tn‐seq studies predicted that darG is essential for growth in Mtb (DeJesus et   al., 2017). We asked if the essentiality of darG was dependent on the presence of darT. We attempted to generate a deletion strain of both darT and darG in Mtb by replacement of the native locus with a hygromycin resistance cassette. We failed to obtain mutant colonies, suggesting that darG could be essential even when darT was absent (Figure 1a, top row). To test this, we used an alternate strategy to obtain ΔdarT‐darG. First, we generated a strain that contained a second copy of darG on an attL5‐integrating plasmid with a streptomycin resistance cassette (Figure 1a, middle row). In this merodiploid we then successfully replaced the native darT‐darG locus with a hygromycin resistance cassette (Figure 1a, bottom row). Resident plasmids at the attL5 site can be efficiently switched with another attL5‐integrating plasmid containing a different antibiotic‐resistance cassette (Pashley and Parish, 2003). Hence, we transformed the ΔdarT‐darG::darG strain with an attL5‐integrating plasmid conferring zeocin‐resistance but not expressing any Mtb gene. We successfully obtained zeocin‐resistant colonies (Figure 1a, bottom row). We confirmed the resulting ΔdarT‐darG mutant by Southern blotting (Figure S1). ΔdarT‐darG exhibited no growth defect in standard 7H9 media (Figure 1b) and showed growth and survival comparable to WT following infection of C57B/6 mice (Figure 1c–1d). The mutant also failed to show a phenotype when subjected to various stressors including the antibiotics isoniazid, rifampicin, ciprofloxacin, levofloxacin, the anticancer drug mitomycin C (used as a DNA damaging agent; Iyer and Szybalski, 1963), starvation, nitric oxide, and H2O2 (Figure S2).

Figure 1

Generation and growth of ΔdarT in vitro and in vivo. (a) Schematic depicting generation of ΔdarT. (b) Growth of WT and ΔdarT in 7H9 media as measured by optical density. (c and d) Quantification of bacterial loads in (c) lungs and (d) spleens of C56BL/6 mice infected with WT or ΔdarT. Data are mean ± SD of four mice per group

These findings establish that the essentiality of darG is dependent on the presence of darT and confirm that the two form a toxin‐antitoxin pair. Toxicity from residual DarTMtb protein in the cytoplasm likely explains our failure to obtain the ΔdarT ‐ darG mutant by direct replacement in the WT strain.

DarG interacts with DarT and with proteins involved in DNA replication and repair

It is common for antitoxin proteins to inhibit their cognate toxins by direct protein–protein interactions (Yamaguchi et   al., 2011). Hence, we expressed and immunoprecipitated a FLAG‐tagged version of DarG in WT Mtb and identified interacting proteins by mass spectrometry. Indeed, we found that DarG bound to DarT (Figure 2a, Table 1). In addition, 9 of the top 20 hits were proteins related to DNA metabolism. Specifically, we identified interactions between DarG and members of the mycobacterial replisome including the replicative polymerases (DnaE1, PolA), helicase (DnaB), and primase (DnaG). DNA‐repair‐associated proteins such as RecA, RecB, RecF, Lhr, and AlkA were also part of the DarG interactome (Figure 2a, Table 1, and S1). We then tested if the association of DarG was dependent on an interaction with DNA or the ribosylation of DNA or proteins (Figure 2b). For example, co‐precipitation of DarG with any DNA‐binding protein could potentially be explained by the presence of complexes containing DNA bound independently to DarG and to resident DNA‐binding proteins (Figure 2b, Model I). To test this, we repeated the pull‐down on DNase‐treated lysates. We did not find a substantial difference in the protein‐binding profile of DarG with or without DNase treatment (Table 1), thus, ruling out Model I. Next, we conjectured that DarG might recruit DNA‐repair proteins on recognition and binding to an ADP‐ribosylated base (Figure 2b, Model II). Our results could also be explained by ADP‐ribosylation of the interacting proteins themselves (Figure 2b, Model III). We tested these possibilities by repeating the pull‐down in a strain lacking DarT, and therefore, presumably lacking ADP‐ribosylation. Interactions between DarG and DNA metabolism proteins persisted even in the absence of DarT. In short, DarG bound to DNA metabolism‐related proteins in a DarT‐independent manner.

Figure 2

DarG interacts with DarT and with proteins involved in DNA replication and repair. (a) Network map of selected interacting partners of DarG. DarT is marked in orange. (b) Schematics displaying possible modes of interaction between DarG (blue), DNA‐binding proteins (colored circles), and DNA. ADP‐ribosylation is represented by a brown line. The most plausible model is boxed. (c) Raw read counts from chromatin immunoprecipitation experiments for DosR‐FLAG, FLAG‐control, and DarG‐FLAG plotted against the Mtb H37Rv genome coordinate. Peaks within selected genes are annotated. Data are representative of three independent replicates

Table 1

DarG interacts with proteins involved in DNA replication and repair

Rank	Rv	Gene	Description	iP: DarGMtb in WT	iP: DarGMtb in WT + DNaseI	iP: DarGMtb in DKO
1	Rv3296	lhr	Probable ATP‐dependent helicase Lhr	22.75	18.5	18
2	Rv1547	dnaE1	Probable DNA polymerase III DnaE1	18.75	15	23
3	Rv0058	dnaB	Probable replicative DNA helicase DnaB	15.5	7.5	6
4	Rv2343c	dnaG	Probable DNA primase DnaG	20.25	15.5	9.5
7	Rv0059	darT	DNA ADP‐ribosyl transferase DarT	24	31	0.9
10	Rv1629	polA	Probable DNA polymerase I PolA	12	14	9.5
12	Rv0630c	recB	Probable exonuclease V (beta chain) RecB	11.25	5	5
16	Rv3051c	nrdE	Ribonucleoside‐diphosphate reductase NrdE	14.5	18.5	30
18	Rv2737c	recA	RecA protein (recombinase A)	33.5	27.5	33.5
22	Rv3208	rv3208	Probable transcriptional regulatory protein (probably TetR‐family)	8.5	11	9
30	Rv1701	rv1701	Probable integrase/recombinase	16.25	9	6.5
34	Rv1317c	alkA	Probable bifunctional regulatory protein and DNA‐repair enzyme AlkA	7.25	5.5	4.5
47	Rv1267c	embR	Probable transcriptional regulatory protein EmbR	15.25	17.5	12.5
57	Rv0823c	rv0823c	Possible transcriptional regulatory protein	7.75	8	5
61	Rv0003	recF	DNA replication and repair protein RecF	9.25	8	5.5
62	Rv2258c	rv2258c	Possible transcriptional regulatory protein	8.75	10.5	10.5
65	Rv3644c	rv3644c	Possible DNA polymerase	10.25	9	5.5
66	Rv1446c	opcA	Putative OXPP cycle protein OpcA	12	7	7
67	Rv3164c	moxR3	Probable methanol dehydrogenase transcriptional regulatory protein MoxR3	8	5.5	4
90	Rv2718c	nrdR	Probable transcriptional regulatory protein NrdR	8.25	8.5	9
91	Rv3246c	mtrA	Two component sensory transduction transcriptional regulatory protein MtrA	9	9	13

Rank, names, and average total spectrum counts of DNA metabolism‐related proteins identified to interact with DarG are displayed. DarG‐FLAG was immunoprecipitated (iP) from whole‐cell lysates of WT or ΔdarT (DKO) Mtb strains transformed with plasmids encoding DarG‐FLAG under a constitutive promoter. “DarG in WT + DNaseI” represents data from WT lysates treated with DNaseI post iP. Interacting proteins were identified by mass spectrometry. Data were calculated from four biological replicates (iP: DarG in WT) or biological duplicates (iP: DarG in WT + DNaseI, iP: DarG in DKO). Biological duplicates of WT Mtb and an experiment with WT Mtb overexpressing a FLAG tag were used as controls. Nonspecific binding peptides were removed from the results by setting the filter of “Total Spectrum Count” of each replicate to “≤4” in the control samples and “≥5” in “DarG in WT” samples. Hits were ranked in descending order based on the ratio of average total spectrum count of “DarG in WT” versus controls, with the corresponding rank indicated in the “Rank” column. Unfiltered raw counts are available in the Supporting Information (Table S1).

Finally, we examined if DarG either co‐localized with (Figure 2b, Model IV) or directly bound to DNA (Figure 2b, Model V). We performed microscopy using DarG‐mCherry and SYTO13‐labeled DNA (Figure S3). We observed variable localization of DarG‐mCherry. Specifically, 44% of bacteria contained DarG‐mCherry foci which did not overlap with SYTO13‐labeled DNA. The remaining bacteria contained diffuse mCherry signals (Figure S3). Since these results were inconclusive, we performed chromatin immunoprecipitation (ChIP‐seq) using an anti‐FLAG antibody on lysates of WT Mtb expressing DarG‐FLAG, DosR‐FLAG (positive control), or the FLAG tag alone (negative control). Compared to the FLAG tag control, there was little or no enrichment of read counts in the DarG‐FLAG ChIP (Figure 2c; no significant differentially bound sites detected using DiffBind (Stark and Brown, 2011)), with the exception of a peak mapping to darG. Since these read counts are not normalized to the ChIP input, this peak is likely an artifact originating from the multicopy episomal plasmid encoding DarG‐FLAG. In contrast, enriched peaks of known DosR targets, hspX and rv2628, were detected in the DosR ChIP (Figure 2c). While we cannot rule out short‐lived and/or nonsequence‐specific interactions with DNA that would be undetectable by ChIP‐seq, these data demonstrate that DarG likely forms a cytosolic complex with DNA‐repair proteins that assembles independently of an interaction with DNA (Figure 2b, Model VI).

DarG‐depletion triggers cell death which is prevented by its T. aquaticus ortholog

Next, we sought to understand the downstream effects of DarT expression. We were unable to overexpress DarT in Mtb, since its toxicity in E. coli precluded our attempts to clone the gene, similar to previous attempts (Jankevicius et   al., 2016). Instead, we generated an anhydrotetracycline (ATC)‐controlled knockdown strain of darG with the expectation that DarG‐depletion would phenocopy overexpression of darT. Briefly, we achieved proteolytic control of native DarG by its fusion to a carboxy‐terminal DAS‐tag. DAS‐tagged proteins are targeted to the ClpP protease by adapter protein SspB, which was expressed under the control of the reverse Tet repressor (Kim et   al., 2011). In the resulting darG‐TetON strain, removal of ATC led to a decrease in the expression of DarG which was accompanied by cell death (Figure 3a‐b). The toxicity associated with DarG‐depletion was fully rescued by constitutive expression of either DarG, or its ortholog from T. aquaticus, DarG (Figure 3c). Mutation of K80 in DarG abrogates the protein's de‐ADP‐ribosylation activity (Jankevicius et   al., 2016). In concordance with this, a K80A mutant of DarG failed to rescue ATC‐dependent growth of darG‐TetON (Figure 3c). Finally, we examined the consequences of overexpressing DarT in Mtb. We transformed ΔdarT‐darG with a plasmid encoding darT under an ATC‐inducible promoter. The resulting strain was unable to grow when exposed to ATC, a phenotype that was rescued by constitutive expression of DarG (Figure 3d). Overall, these data establish that the DarT‐DarG toxin‐antitoxin system is cross‐complemented by DarT‐DarG, thus signifying functional conservation.

Figure 3

DarG‐depletion triggers cell death which is prevented by its T. aquaticus ortholog. (a) Immunoblot of protein extracts from darG‐TetON grown with and without ATC. Blot was probed with DarG‐specific and DlaT‐specific (loading control) antisera (b) Growth of darG‐TetON quantified by CFU in 7H9 medium with or without ATC. (c) 7H10 agar plates cultured with darG‐TetON transformed with empty plasmid or plasmids encoding darG WT, darG WT, or darG mutant expressed from constitutive promoters. The disk in the center of the plate contains 100 ng of ATC; the concentration of ATC decreases from the center to the periphery of the plate. (d) 7H10 agar plates cultured with ΔdarT containing a plasmid encoding darT from an ATC‐inducible promoter transformed with or without a plasmid encoding darG from a constitutive promoter. The disk in the center of the plate contains 800 ng of ATC; the concentration of ATC decreases from the center to the periphery of the plate. Data in (b) are means ± SD from three independent experiments. Data in (c‐d) are representative of at least two independent experiments

Since we were unable to detect binding of DarG to DNA under baseline conditions (Figure 2c), we asked if DarG bound DNA upon induction of ADP‐ribosylation in the cell. We tested this by performing chromatin immunoprecipitation of FLAG‐tagged DarG from cells expressing ATC‐inducible DarT (Figure 3d). We found no differences in mapped reads between cells grown with and without ATC (Figure S4; no significant differentially bound sites detected using DiffBind (Stark and Brown, 2011)), indicating that DarG does not form stable and consistent interactions with DNA even on overexpression of an active DNA ADP‐ribosylase.

DarG‐depletion sensitizes Mtb to drugs targeting DNA metabolism and respiration

Next, we tested if targeting DarG sensitizes Mtb to a panel of antibacterial compounds. We measured the MIC of each drug to darG‐TetON while simultaneously varying the extent of DarG knockdown. DarG‐depleted (ATC‐low) Mtb were most susceptible to drugs targeting respiration (bedaquiline), gyrase (ciprofloxacin, levofloxacin), transcription (rifampicin, rifabutin), and causing DNA damage (mitomycin C, netropsin) (Figure 4b,c). In contrast, MICs of drugs inhibiting translation, and cell wall synthesis were largely unaffected by knockdown of DarG (Figure 4a,c). These data suggest that depletion of DarG pre‐sensitizes the cells to DNA‐damage inducing drugs. Thus, inhibition of DarG likely perturbs cellular pathways involved in respiration and DNA metabolism.

Figure 4

DarG‐depletion sensitizes Mtb to drugs targeting DNA metabolism and respiration. (a and b) Susceptibility of darG‐TetON to (a) chloramphenicol or (b) rifampicin. darG‐TetON was cultured in 7H9 medium without ATC for 6 days to decrease DarG expression before incubating with indicated concentrations of ATC and drug. Growth was measured after 14 days using optical density and normalized to that in the corresponding ATC concentration without drug treatment. (c) Heat‐map representation of the MIC50 shift of darG‐TetON to antimicrobial compounds. Experiments were performed as in (a) and (b). Values within each cell are the MIC50 for darG‐TetON grown in high ATC (H; 150 ng/ml) and low ATC (L; concentration of ATC that led to a ~70% growth defect in the absence of drug), normalized to high ATC. Data in (a–b) are representative of three independent experiments. Data in (c) are means calculated from three independent experiments. Symbols on the right indicate results from one‐sided t test (ns indicates non‐significant, * indicates p value ≤ 0.05, ** indicates p value ≤ 0.01, *** indicates p value ≤ .001)

DarG‐depletion induces the DNA‐damage response resulting in increased mutability

To further explore the consequences of DarG‐depletion, we used transposon mutagenesis followed by high‐throughput sequencing (Tn‐seq) in the darG‐TetON strain. We generated transposon mutant libraries in Mtb in DarG‐depleted (ATC‐low) and DarG‐replete (ATC‐high) conditions. We expressed differences as the log2 fold change (log2FC) of the mutant frequency between libraries exposed to ATC‐low and ATC‐high conditions. Mutants underrepresented in the ATC‐low condition show a negative log2FC and correspond to aggravating genetic interactions with darG (Figure 5a, left). In contrast, mutants with a positive log2FC are overrepresented in the ATC‐low condition and correspond to alleviating genetic interactions with darG (Figure 5a, right).

Figure 5

DarG‐depletion induces the DNA‐damage response resulting in increased mutability. (a) Volcano plot representing Tn‐seq data from darG‐TetON grown on 7H10 agar with low or high ATC. Tn‐seq log2FC (low ATC/high ATC) and false discovery rate‐adjusted p values (q‐values, q‐val) are plotted for each genetic locus. Loci with a q‐val ≤ 0.05 and a log2FC ≤ −1 are colored green. Loci with a q‐val ≤ 0.05 and a log2FC ≥ 1 are colored red. Selected mutants are annotated (b) Volcano plot representing RNA‐seq data from darG‐TetON grown in 7H9 medium with low or high ATC. Gene expression log2FC (low ATC/high ATC) and false discovery rate‐adjusted p values (q‐values, q‐val) are plotted for each gene. Color coding is identical that in to (b) (c) darG‐TetON was grown in 7H9 medium with or without ATC for 18 days before plating on 7H10 agar + ATC and with or without rifampicin (1 μg/ml). Plotted are the ratios of CFU in + rifampicin to ‐rifampicin conditions. Data from (a–c) are derived from three independent experiments

Aggravating genetic interactions were identified between darG and several genes involved in DNA repair including the master DNA‐damage responsive transcriptional activators (pafBC), members of the SOS DNA‐damage response pathway (recA and recO), and the helicase involved in nucleotide excision repair (uvrD1) (Figure 5a, Table S2) (Fudrini Olivencia et   al., 2017; Singh, 2017; Muller et al., 2018). Importantly, disruption of darT resulted in increased survival of the DarG‐depleted strain (Figure 5a, Table S2). In addition, mma4 and fecB exhibited alleviating genetic interactions with darG (Figure 5a, Table S2). Mma4, methoxy mycolic acid synthase 4, is a methyl transferase that modifies cell wall mycolic acids (Yuan and Barry, 1996). FecB is annotated as a putative iron dicitrate‐binding lipoprotein, and is a determinant of the intrinsic resistance of Mtb to antibiotics (Xu et   al., 2017). We reasoned that these mutants were favored due to their increased cell wall permeability which allowed greater uptake of ATC in the cell, and consequently, higher expression of DarG (Dubnau et   al., 2000; Xu et   al., 2017).

We also sought to elucidate the transcriptomic changes associated with DarG knockdown. In agreement with the Tn‐seq data, we found that targeting DarG resulted in a strong induction of genes involved in DNA metabolism (Figure 5b, Table S3). Specifically, we observed a 17‐fold upregulation of dnaE2, a DNA‐damage‐induced error‐prone translesion polymerase (Boshoff et al., 2003). Expression of this gene is associated with an increased mutation frequency and consequently, an increased rate of drug resistance (Boshoff et   al., 2003). Therefore, we evaluated if targeting DarG altered the mutation frequency of Mtb. We cultured darG‐TetON in 7H9 media either with (DarG‐replete) or without (DarG‐depleted) ATC before plating on ATC‐containing agar plates with rifampicin. We measured the frequency of rifampicin‐resistant Mtb as a proxy for the mutability of darG‐TetON. Indeed, we found that knockdown of DarG favored the emergence of rifampicin‐resistant Mtb (Figure 5c). Taken together, these results not only demonstrate that depletion of DarG induces a DNA‐damage response that decreases viability, but also increases mutability.

DISCUSSION

DarG is one of three putative antitoxins encoded in the Mtb genome that is essential for viability (DeJesus et   al., 2017). Here, we demonstrate that DarT‐DarG indeed form a functional toxin‐antitoxin system. Removal of DarG causes cell death, a phenotype that is rescued by simultaneous deletion of DarT (Figure 1a, 3b). The mechanism of action of DarT‐DarG is functionally conserved between Mtb and Taq, as evidenced by cross‐complementation experiments (Figure 3c–d). Further, DarT‐DarG physically interact within the cell (Figure 2a, Table 1). However, the role of the DarT‐DarG system in the physiology of Mtb remains unknown, as evidenced by the lack of phenotype of the ΔdarT‐darG mutant in a number of physiologically relevant stress conditions (Figure 1b–d, S2). While we cannot discount the possibility of a functionally redundant TA system, the essentiality of DarG and the uniqueness of DNA ADP‐ribosylation argues against this hypothesis. It has been suggested that some TA systems might primarily function to preserve genome integrity (Szekeres et   al., 2007; Ramage et al., 2009), which could be the case for DarT‐DarG. Finally, it is possible that the DarT‐DarG TA system plays a role under conditions that are yet untested.

We find that induction of the DNA‐damage response is essential for Mtb to survive DarG‐depletion (Figures 4, 5). Our data are in concordance with observations from EPEC showing that DarT‐mediated toxicity is aggravated by disruption of the RecFOR‐homologous recombination and nucleotide excision repair pathways (Lawaree et   al., 2020). Together, these observations support the claim that DarT primarily exerts its toxicity by ADP‐ribosylation of DNA (Jankevicius et   al., 2016). In addition, DarG forms cytosolic complexes with several proteins involved in DNA repair, either in a direct or indirect manner (Figure 2a–b, Table 1). This implies that the complex of DarG with DNA‐repair proteins is poised to be recruited to sites of DNA‐ADP‐ribosylation (Figure 2b, Model VI), where it mediates removal of the ADPr moiety followed by correction of the associated DNA damage. While we failed to detect DarG bound to DNA, including under conditions that induced ADP‐ribosylation (Figures 2c, S4), it is possible that these interactions are too transient or weak to be detected (Nebbioso et   al., 2017). Alternatively, the binding of DarG to DNA may lack sequence specificity, which would preclude detection of enriched peaks at a population level.

Importantly, the DNA‐damage response triggered by activation of DarT leads to increased mutability (Figure 5c), possibly due to the induction of DnaE2, a translesion polymerase implicated in the emergence of drug resistance in vivo (Figure 5b) (Boshoff et   al., 2003). Transient hypermutability can promote rapid adaptation to novel environments (Taddei et   al., 1997; Jolivet‐Gougeon et   al., 2011) and promote the evolution of drug resistance (Blazquez, 2003). Thus, induction of hypermutation by dysregulation of the DarT‐DarG complex could be beneficial to a larger population of Mtb under unfavorable conditions.

Activation of DarT has a bactericidal effect (Figure 3b), similar to induction of MbcT, the toxin of the only other characterized TA system in Mtb harboring an essential antitoxin (Freire et   al., 2019). In contrast, most other toxins studied in Mtb exert a bacteriostatic effect (Singh et   al., 2010; Tiwari et   al., 2015; Agarwal et   al., 2018; Sharrock et al., 2018; Tandon et al., 2019). This suggests that, among TA systems, DarG may be an attractive drug target. Indeed, peptides that disrupt the toxin‐antitoxin interface have been designed for other Mtb TA modules and inhibition of antitoxins could be a promising avenue for tuberculosis treatment in general (Williams and Hergenrother, 2012; Chan et al., 2015; Lee et   al., 2015; Kang et   al., 2017).

EXPERIMENTAL PROCEDURES

Bacterial culture conditions

M. tuberculosis H37Rv and derived strains were cultured in Middlebrook 7H9 medium (BD Difco) containing 0.2% of glycerol, 0.2% of dextrose, 0.5% of BSA (Roche), 0.085% of NaCl, and 0.05% of Tween‐80 or tyloxapol, or in Middlebrook 7H10 agar (BD Difco) containing 10% of OADC supplement (BD) and 0.5% of glycerol. Liquid cultures were incubated under static conditions at 37°C with 5% of CO2. Agar plates were incubated at 37°C. Selection antibiotics were used at the following concentrations: hygromycin (50 μg/ml), kanamycin (25 μg/ml), zeocin (25 μg/ml), and streptomycin (20 μg/ml). ATC was used at 500 ng/ml except where indicated otherwise. For liquid cultures, ATC was replenished 100% every 7 days.

Generation of strains

All plasmids were generated using Gateway cloning technology (Life Technologies). The ΔdarT‐darG strain was generated from WT M. tuberculosis H37Rv as described in the Results section using recombineering (Gee et   al., 2012; Murphy et al., 2015). For the DarG‐pull‐down and ChIP‐seq experiments, WT or ΔdarT‐darG strains were transformed with an episomal plasmid encoding for darG under the P750 promoter. WT Mtb transformed with an attL5‐integrating vector encoding for FLAG tag alone under the hsp60 promoter served as the negative control for the ChIP‐seq and the pull‐down experiments. WT Mtb transformed with an episomal vector encoding for dosR ‐FLAG under an ATC‐inducible promoter (obtained as a gift from Dr. Tige Rustad, Juno Therapeutics) served as the positive control for the ChIP‐seq. WT Mtb strains transformed with episomal vectors expressing mCherry alone or DarG‐mCherry under the P750 promoter were used for microscopy. The darG‐TetON strain was generated as described in (Johnson et   al., 2019). For cross‐complementation studies, the darG‐TetON strain was transformed with episomal plasmids expressing either darG, codon‐adapted darG , or codon‐adapted darG under the hsp60 promoter. Expression of darT was achieved by transforming a giles‐integrating plasmid encoding codon‐adapted darT under an ATC‐inducible promoter (P606) into ΔdarT‐darG or ΔdarT‐darG overexpressing DarG‐FLAG.

Mouse infections

The animal experiments were performed in accordance with National Institutes of Health guidelines for housing and care of laboratory animals and according to institutional regulations after protocol review and approval by the Institutional Animal Care and Use Committee of Weill Cornell Medicine (protocol number 0601441A). Female 7‐ to 8‐week‐old C57BL/6 mice (Jackson Laboratory) were infected with ∼100 CFU using an inhalation exposure system (Glas‐Col). CFU burden of lungs and spleens at each time point was determined by plating dilutions of organ homogenates on 7H10 agar. Four mice were euthanized at each time point for each group.

Pull‐down of DarG

About 150 ml of mid‐log phase Mtb culture was washed with PBS containing 0.05% of Tween 80 (PBST), resuspended in lysis buffer + protease inhibitors, lysed with 0.1 mm zirconia beads and incubated with anti‐Flag beads. We washed the beads three times with PBS before elution with FLAG peptides. The eluates were analyzed by mass spectrometry. For the DNaseI treatment, anti‐Flag beads were incubated with Mtb whole‐cell lysates overnight, washed five times with PBS, and then, treated with 25 units of DNaseI at 37°C under gentle shaking for 3 hr.

Microscopy

Mtb cultures were collected by centrifugation, washed with PBS containing 0.05% of Tween 80 (PBST), fixed with 4% of paraformaldehyde overnight prior to removal from BSL‐3 containment and incubated with 5 μM SYTO 13 (Thermo Fisher Scientific) for 5 min for nucleoid labeling. Single cell suspensions were prepared by centrifugation at 800 rpm for 10 min. After spreading on soft agar pads, bacteria were visualized with a DeltaVision image restoration microscope (GE Healthcare), a 100x oil objective and appropriate filter sets. Images were captured with a pco.edge scientific SCOS camera and analyzed with ImageJ (Schneider et al., 2012).

ChIP‐seq of DarG

About 50 ml cultures of Mtb overexpressing FLAG‐tagged proteins were grown to an OD of 0.8–1.2. For cultures expressing DosR‐FLAG, expression was induced by addition of ATC (100 ng/ml) for 4 days before addition of formaldehyde. For cultures expressing DarT , expression was induced by adding ATC (500 ng/ml) for 8 days before addition of formaldehyde. Chromatin immunoprecipitation was performed as in (Minch et   al., 2015). NGS library preparation was performed using NEBNext Ultra II DNA Library prep kit for Illumina. Samples were sequenced using standard Illumina protocols producing ~40 million 50‐bp single‐end reads. Reads were aligned to the reference genome using Bowtie 2.3 (Langmead and Salzberg, 2012). BAM alignment files were created, sorted and indexed using SAMtools (Li et   al., 2009) and viewed in the IGV viewer (Robinson et   al., 2011). Peak calling was performed using MACS2 (Zhang et   al., 2008). Computation of differentially bound sites was performed using DiffBind (Stark and Brown, 2011). For ChIP of DarG under baseline conditions, peaks from DarGMtb‐FLAG were compared against the FLAG control. For ChIP of DarG from cells overexpressing DarT, peaks from the +ATC condition were compared against the ‐ATC condition as control.

Western Blots

Rabbit polyclonal antibody for DarG was generated by GenScript. DlaT antibody (Bryk et al., 2002) was a gift from R. Bryk and C. Nathan at Weill Cornell Medicine. All secondary antibodies were purchased from LI‐COR biosciences. Protein lysates were prepared by mechanical lysis with 0.1 mm zirconia beads. Unbroken bacterial cells and beads were removed by centrifugation and supernatants were filtered using 0.22 μm spin‐X columns prior to removal from BSL‐3 containment. Protein lysates were separated using SDS‐PAGE and transferred to a nitrocellulose membrane. After washing and incubation with secondary antibodies, proteins were visualized using Odyssey Infrared Imaging System (LI‐COR Biosciences).

d arG ‐tetON antibiotic susceptibility

darG‐tetON cultures were grown to till mid‐log phase in 7H9 medium + ATC and washed twice in 7H9 medium ‐ATC. Washed cultures were diluted to an OD of 0.02 and grown for 6 days in 7H9 medium ‐ATC to pre‐deplete DarG. Pre‐depleted cultures were used to inoculate 384‐well black plates with clear flat bottoms that contained a range of ATC/drug concentrations at an OD of 0.01. Drugs were dispensed using an HP D300e Digital Dispenser (Hewlett Packard). The drug dispensing was randomized using the HP Digital Dispenser software (version 3.2.2), and the dimethyl sulfoxide (DMSO) concentration in each well was normalized to 1%–2%. After incubation for 14 days, the optical density (OD580) in each well was read using a SoftMax M2 plate reader. The data were de‐randomized using HP Digital Dispenser Data Merge software.

d arG ‐tetON Tn‐seq

darG‐tetON cultures were grown to mid‐log phase in 7H9 medium + ATC and washed twice in 7H9 medium ‐ATC. Washed cultures were diluted to an OD of 0.25 and grown for 3 days in 7H9 medium ‐ATC to pre‐deplete DarG. Pre‐depleted cultures were transduced with ΦMycoMarT7 phage as previously described (Long et   al., 2015; Xu et   al., 2017) and plated on 7H10 agar plates with 0.05% of tyloxapol and ATC at two different concentrations: 15 ng/ml ATC (ATC‐low) and 500 ng/ml (ATC‐high). Plates were incubated for 21 days before harvesting, extracting genomic DNA and sequencing as described previously (Long et   al., 2015; Xu et   al., 2017). Mapping and quantification of transposon insertions was done as described previously (Xu et   al., 2017). Differentially represented genes were identified using resampling in the TRANSIT analysis platform as described previously (DeJesus et al., 2015; Xu et   al., 2017). We defined genes having a q‐value of ≤.05 and a log2FC ≥ 1 or log2FC ≤ −1 as significant.

d arG ‐tetON RNA‐seq

darG‐tetON cultures were grown to mid‐log phase in 7H9 medium + ATC and washed twice in PBS‐tyloxapol 0.05%. Washed cultures were diluted to an OD of 0.015 and grown for 7 days in 7H9 medium ‐ATC (low ATC) or + ATC (high ATC). Total RNA was extracted as described in (Botella et al., 2017) and Illumina cDNA libraries were generated using the RNAtag‐Seq protocol as described in (Shishkin et   al., 2015) and (Botella et   al., 2017) and sequenced on HiSeq 4000 to generate 50 bases paired‐end reads. The samples from three independent replicates were processed in two rounds of library preparation and sequencing.

The sequencing reads were cleaned by trimming adapter sequences and low quality bases using cutadapt v1.9.1 (Martin, 2011), and were aligned to a modified M. tuberculosis reference genome using BWA v0.7.15 (Li and Durbin, 2009). The original M. tuberculosis genome (H37Rv) was retrieved from NCBI (https://www.ncbi.nlm.nih.gov/nuccore/NC_000962.3) and modified to add a 1,303 bp insertion before the stop codon of the gene darG. Raw read counts per gene were extracted using HTSeq‐count v 0.6.1 (Anders et al., 2015). Differential expression analysis was performed using the Agilent GeneSpring software.

Measuring mutability of darG ‐tetON

darG‐tetON cultures were grown till mid‐log phase in 7H9 medium + ATC and washed twice in 7H9 medium −ATC. Washed cultures were diluted to an OD of 0.05 and grown for 18 days in 7H9 medium with or without ATC. Cells were pelleted and plated on 7H10 agar plates with no rifampicin or 1 μg/ml rifampicin. Colonies were counted after ~3–4 weeks of incubation.

AUTHOR CONTRIBUTIONS

All authors contributed to the conception and design of the study. AZ, RW, LB, RS, LZ, JBW, NS, and RSJ contributed to the acquisition and analysis of data. AZ, SE, DS, and RW contributed to the writing of the manuscript. The authors declare no conflict of interest.

Fig S1‐4

Click here for additional data file.

Table S1

Click here for additional data file.

Table S2

Click here for additional data file.

Table S3

Click here for additional data file.