Introduction

Mycobacterium tuberculosis (Mtb) continues to stride as the number one killer among all infectious diseases, accounting for nearly 1.5 million deaths yearly. The aggravating situation is despite the clinical use of over 20 antibiotics and a century-old vaccine, BCG. The gradual rise in the emergence of increasingly drug-resistant strains and HIV-TB co-infection further highlights the urgency to identify newer attractive drug targets. Throughout the course of infection, Mtb is exposed to a continuum of dynamic host-induced stresses such as severe nutrient deprivation, acidified compartments, and toxic reactive oxygen species (ROS) and reactive nitrogen species (RNS) produced by its resident phagosomes. In turn, Mtb produces copious amounts of actinomycetes-specific mycothiol, the major antioxidant in actinomycetes that act as the functional equivalent of glutathione, to combat ROS and RNS. In addition to mycothiol, Mtb also produces ergothioneine, a low molecular weight thiol, and several enzymes that act concertedly to subvert host-induced redox stress. The redox-active group of both mycothiol and ergothioneine is derived from L-cysteine. Hence, genes involved in the biosynthesis of L-cysteine are upregulated in the host and in vitro upon oxidative and nutritional stress [1-5]. Notably, an increased expression of these genes is functionally crucial, as suggested by the attenuated survival of transposon mutants of many sulfur and L-cysteine biosynthesis genes within the host [6].

In mycobacteria, sulfur assimilation begins with the import of sulfate through a sulfate transporter composed of SufI.CysT.W.A. Intracellular sulfate is a substrate for APS synthase CysD.N.C, which adenylates and phosphorylates sulfate to form adenosine 5′-phosphosulfate (APS). APS sits at a metabolic branch point; it can either be converted into sulfolipids by consequent actions of multiple Stfs enzymes or reduced via SirA and SirH to sulfide (Figure S1). This pathway encompasses sulfide formation from sulfate, called the sulfur assimilation pathway. Mtb genome encodes three L-cysteine synthases – the canonical CysK1 and non-canonical CysM and CysK2 enzymes. Interestingly, humans do not possess L-cysteine synthases, raising the possibility of developing antibiotics without a homologous target in the host. CysK1 utilizes sulfide produced via the sulfur assimilation pathway and O-acetyl-L-serine produced from glycolytic intermediate 3-phosphoglycerate to produce L-cysteine. CysM, on the other hand, uses O- phospho-L-serine and a small sulfur carrier protein CysO as substrates. Like CysK1, CysK2 utilizes O-phospho-L-serine and sulfide as substrates (Figure S1). In addition, Mtb can also synthesize L- cysteine through a reverse transsulfuration pathway from L-methionine. This example of convergent metabolic redundancy raises several interesting questions: (1) Why would Mtb rely on multiple enzymes and pathways to produce the same biomolecule? (2) Are these “functionally redundant” enzymes dispensable, or are they required at a distinct cellular space, time, and condition? (3) Is the L-cysteine pool produced through a particular pathway functionally compartmentalized? That is, is it metabolized into a specific kind of downstream thiol?

To define these unsolved aspects of Mtb L-cysteine metabolism, we sought to investigate the interplay of non-canonical L-cysteine synthases of Mtb and elucidate their roles in abetting virulence. We aimed to decipher the relative contribution of CysM and CysK2 enzymes in alleviating host-induced stresses and promoting the survival of Mtb within the host. We also investigated their role in secondary metabolism, synthesizing low molecular weight thiols, such as mycothiol and ergothioneine, and understanding the consequential effects of their deletion on the global transcriptome of Mtb. Lastly, with the help of specific inhibitors, we evaluated their potential to serve as attractive drug targets for adjunct antibiotic therapy.

Results

Non-canonical L-cysteine synthases facilitate Mtb in combating host-induced stresses

Mtb is a generalist, a prototroph organism that can produce all 20 proteinogenic amino acids. In agreement with this notion, numerous microarray studies depict the upregulation of multiple amino acid pathways within the host [3, 4, 7], indicating a higher dependency of Mtb survival and virulence on amino acid biosynthesis and regulation. Beyond their role in protein synthesis, amino acids such as tryptophan, lysine, and histidine interact with the host immune cells to facilitate mycobacterial survival within the host [8, 9]. We sought to identify distinct host stresses that result in the transcriptional modulation of specific amino acid biosynthetic pathways with the help of RNA sequencing (RNA-Seq). We compared the transcriptional profile of H37Rv (Rv) grown in 7H9-ADC with the profiles obtained when bacilli were subjected to oxidative, nitrosative, starvation, and acidic stresses (Table S1). The volcano plot illustrates differentially expressed genes (DEGs) that were significantly upregulated (blue) and downregulated (red) under indicated stress conditions (absolute log₂ Fold change>1 and Padj<0.05) (Figure S2a-e). Exposure to starvation conditions resulted in drastic transcription modulation compared with other stresses, suggesting that nutrient deficiency is the primary driver of transcriptome remodeling. In contrast, we observed the lowest number of DEGs when the bacteria were subjected to mildly acidic conditions (pH 5.5). Heat maps of normalized DEGs depict that DEG changes were comparable across biological replicates within each sample set (Figure S2f-j).

To further refine our understanding of the DEGs, we grouped them into various functional categories and found that genes belonging to intermediary metabolism & respiration remained the most affected in all conditions highlighting their role of metabolic rewiring (Figure S3a). Pathway enrichment analysis of the most enriched Gene Ontology (GO) further revealed that, while, expectedly, metabolic pathways were found to be downregulated during starvation, we observed enrichment of nitrogen metabolism (Figure S3b & Table S2). Surfactant stress resulted in the upregulation of branched-chain amino acids/keto acids degradation pathways (Figure S3c), and nitrosative stress ensued up-regulation of fatty acid and lipid biosynthetic processes (Figure S3d). There were no changes observed in mildly acidic conditions (Figure S3e). Importantly, oxidative stress resulted in significant up-regulation of genes involved in sulfur metabolism (Figure S3f). We further analyzed the DEGs involved in sulfur and L-cysteine metabolism across sample sets and discerned an overlap of the genes affected under two or more conditions. Interestingly, we observed upregulation of sulfate transporters genes (subI, cysT, cysW, cysA1) across multiple stresses and sulfur assimilation (cysH, cysA2, cysD, cysNC) during starvation, oxidative and SDS stress. CysK2, a non-canonical L-cysteine synthase, was found to be up-regulated during all stresses except SDS stress (Figure 1a).

Fig 1

Thus RNA-Seq data suggest that sulfur assimilation and L-cysteine biosynthetic pathway genes are the most commonly upregulated pathways in Mtb exposed to multiple host-like stresses (Figure S2). To address the functional relevance of this observation, we deleted two non-canonical L-cysteine synthases-CysM [5] and CysK2, from the Mtb chromosome. CysK2 mutant, RvΔcysK2, was generated using the recombineering method, and the recombination at the native loci was confirmed with the help of multiple PCRs (Figure 1b). While deletion of cysM or cysK2 did not affect mycobacterial growth under in vitro nutrient-rich 7H9-ADC or 7H9-ADS (Figure S4a and S4b), it significantly compromised Mtb growth in nutrient-limiting Sauton’s media (∼2 log₁₀; Figure 1c), suggesting the importance of cysM- and cysK2-derived L-cysteine. Restoring cysM or cysK2 expression in the complementation strains, RvΔcysM::M and RvΔcysK2::K2, rescued the growth defects (Figure S4 & Figure 1c). Mtb is a metabolically versatile organism capable of utilizing a large variety of carbon and nitrogen sources [10]. Unlike 7H9, wherein glucose, glycerol, glutamate, and ammonia act as carbon and nitrogen sources, Sauton’s media contains glycerol and asparagine as the sole carbon and nitrogen sources, suggesting metabolic reprogramming aided by CysM and CysK2 enable Mtb to grow optimally in a limited nutritional environment. This observation was further recapitulated under nutrient starvation (PBS), wherein the survival of RvΔcysM and RvΔcysK2 was lower than parental Rv or RvΔcysM::M or RvΔcysK2::K2 strains (Figure 1d). Interestingly, when exposed to acidic conditions (pH 4.5), RvΔcysM and RvΔcysK2 survival were observed to be ∼0.85 and ∼0.24 log₁₀ lower, respectively, compared with Rv, suggesting CysM is relatively more important for bacillary survival under acidic stress (Figure 1e). The highest attenuation of RvΔcysM and RvΔcysK2 was observed upon the addition of cumene hydroperoxide (CHP), an organic hydroperoxide that, upon decomposition, generates free radicals. The relative survival of RvΔcysM and RvΔcysK2 was ∼2.04 and ∼1.31 log₁₀ lower, respectively, compared to Rv (Figure 1f). The addition of diamide, which results in thiol oxidation, also attenuated the survival of RvΔcysM and RvΔcysK2 by ∼1.33 and ∼1.26 log₁₀ compared with Rv (Figure S4c). When the mutant strains were subjected to nitrosative stress, the survival of RvΔcysM and RvΔcysK2 was ∼0.93 and ∼0.86 log₁₀ lower than Rv (Figure 1g). However, no significant attenuation was found during reductive and surfactant stress (Figure 1h-i). Collectively, RvΔcysM and RvΔcysK2 displayed increased susceptibility towards oxidative, nitrosative, mild acidification, and PBS starvation to varying degrees compared with Rv, RvΔcysM::M and RvΔcysK2::K2. The data suggests that L-cysteine, produced via CysM and CysK2, and its downstream products help mycobacteria thwart specific stresses that Mtb encounters within the host.

Distinct roles of CysM and CysK2.C

To decipher the mechanism through which CysM and CysK2 combat oxidative stress, we performed a global transcriptomic analysis of Rv, RvΔcysM, and RvΔcysK2 in the presence and absence of oxidative stress (CHP) (Table S3). Principal Component Analysis demonstrated clear separation of strains under different conditions. Intriguingly, while RvΔcysM and RvΔcysK2 were closely located on a PCA plot in the absence of any stress, oxidative stress resulted in a significant divergence between the two groups (Figure 2a). Deletion of cysM resulted in differential expression of 322 genes (159 downregulated and 163 upregulated) (Figure S5a-b), while deletion of cysK2 impacted 278 genes (155 downregulated and 123 upregulated) under regular growth conditions (Figure S5c-d) (absolute log₂ Fold change>1 and Padj<0.05). In contrast, upon treatment with CHP, nearly ∼33% and ∼53% of Mtb genes were differentially expressed in RvΔcysM and RvΔcysK2, respectively, compared with Rv (Figure 2b-c). To understand the individual contribution of CysM and CysK2 in combating oxidative stress, we compared the DEGs of RvΔcysK2 to RvΔcysM under regular growth and oxidative stress conditions. While DEGs between RvΔcysK2 and RvΔcysM were limited to 13 in untreated conditions (Figure 2d), CHP treatment resulted in differential expression of 1372 genes (Figure 2e and Figure S5e), highlighting unique transcriptional signatures associated with each L-cysteine synthase under oxidative stress. Subsequently, we compared the DEGs of Rv, RvΔcysM, and RvΔcysK2 obtained under oxidative stress with the help of a Venn diagram. This analysis revealed that CysM and CysK2 influence a shared repertoire of 1023 genes (529 upregulated and 494 downregulated) compared to Rv upon CHP treatment (Figure 2f). Importantly, CysM and CysK2 distinctively modulate 324 and 1104 genes, respectively, during oxidative stress (Figure 2f-g). Assortment of DEGs into functional classes revealed intermediary metabolism & respiration and cell wall pathways among the most impacted categories (Figure 2h). These results corroborate the recent finding that CysK2 alters the phospholipid profile of the Mtb cell envelope [11]. Pathway enrichment analysis of the most enriched Gene ontology biological process indicated that while most pathways are commonly affected by CysM and CysK2, these L-cysteine synthases also have a unique transcriptional footprint (Table S4). Pathways such as DNA binding, homologous recombination, translation, ribosome, etc., were commonly affected in RvΔcysM and RvΔcysK2 compared to Rv during oxidative stress (Figure S5f, g). Genes belonging to oxidative phosphorylation and quinone binding were upregulated, while DNA binding and cholesterol catabolism gene categories were found to be downregulated in RvΔcysM compared to RvΔcysK2 during oxidative conditions (Figure 2i). Together, our data point to the unique roles of CysM and CysK2, as their deletion differentially affects various cellular pathways under oxidative stress.

Fig 2

Key metabolites are differentially affected upon CysM and CysK2 deletion

L-Cysteine concentration in Mtb is notoriously low, usually 10-fold lower than most other amino acids [12]. As expected, L-cysteine levels were below the limit of detection/quantification (not shown). Therefore, we used key downstream metabolites to monitor L-cysteine biosynthesis and utilization. We followed pool size and labelling (³⁴S) of L-methionine, mycothiol/mycothione, and ergothioneine, made from L-cysteine, via three distinct and dedicated metabolic pathways (Figure S6). First, L-methionine, mycothiol, and ergothioneine concentrations vary from zero to 24 h, indicating that their pool sizes are not as stable as those of other core metabolites. This is consistent with their function, susceptibility to redox homeostasis, and L-methionine’s role in protein synthesis. Second, upon challenge with CHP for 24 h, L-methionine levels decrease in Rv, remain constant in RvΔcysK2, and increase in RvΔcysM (Figure 3a). Ergothioneine levels are identical in the three strains, and upon treatment with CHP, they are decreased in the parent strain but remain unchanged in the RvΔcysK2 and RvΔcysM mutant (Figure 3b). Mycothiol levels increase in the Rv, RvΔcysK2 strains and remain nearly constant in RvΔcysM (Figure 3c). Finally, mycothione levels increased slightly in the parent strain, decreased in the RvΔcysK2, and remained stable in the RvΔcysM strain (Figure 3d). These changes revealed a significant role of distinct L-cysteine biosynthetic routes on redox stress and homeostasis.

Fig 3

Employing ³⁴S labelling, we observed a reduced rate of synthesis of mycothiol, mycothione and ergothioneine in RvΔcysM compared to Rv and RvΔcysK2 (Figure 3e-g). Interestingly, ergothioneine levels were found to be reduced in both RvΔcysM and RvΔcysK2 compared to Rv, possibly underlying one of the reasons for their enhanced sensitivity to oxidative stress.

This result further indicates that these two routes are partially redundant; that is, they generate the same end product, L-cysteine. Yet, pool size measurements demonstrate significant changes, highlighting that while these pathways produce the same metabolite, their complex biological roles and requirements (co-substrates, pathways, regulation, etc.) are not fully redundant. This interpretation is in accordance with transcriptomics and phenotypic results described above. These non-overlapping metabolic requirements (e.g., O-acetyl-L-serine vs. O-phospho-L-serine vs. CysM) are likely the source of the different metabolic phenotypes observed. Therefore, even producing the same end-product, genetic disruption of different L-cysteine synthases differently affects bacterium metabolism and fitness, leading to the distinct phenotypes observed between RvΔcysK2 and RvΔcysM in vitro, in cellulo, and in vivo.

CysM and CysK2 alleviate the toxicity of host-produced ROS and RNS

To define whether the attenuation of mutant strains observed during defined in vitro host-like conditions can be recapitulated within the host cells, we compared the survival of parental, mutants, and complementation strains in murine peritoneal macrophages (Figure 4a). Survival of RvΔcysM and RvΔcysK2 was ∼1.22 and ∼0.85 log₁₀ lower, respectively, compared with Rv at 96 h post-infection (p.i) (Figure 4b). Notably, the addition of L-cysteine alleviated survival differences (Figure 4c), suggesting that reduced L-cysteine levels in mutant strains are responsible for their attenuated survival. To further validate that CysM and CysK2 support Mtb survival by detoxifying peroxides generated by the host, we compared the survival of strains in the peritoneal macrophages extracted from wild-type C57BL/6 (WT) mice that are proficient in eliciting oxidative and nitrosative stress or murine strains that are deficient in their ability to produce hydrogen peroxide (H₂O₂) (phox^-/-) or both oxidative and nitrosative stress (IFNγ^-/-). As anticipated, regardless of the mice genotype, there was no difference in the survival of Rv and complementation strains (Figure 4d-e). On the other hand, attenuation of RvΔcysM and RvΔcysK2 observed in wild-type macrophages is partially reversed in the macrophages obtained from phox^-/- mice (Figure 4d) and completely nullified in macrophages isolated from IFNγ^-/- mice (Figure 4e). To further dissect the relative contribution of CysM and CysK2 in combating ROS and RNS independently or collectively, the survival of Mtb strains was examined in peritoneal macrophages isolated from WT and phox^-/- in the presence or absence of iNOS inhibitor. While the inhibition of ROS or RNS independently only partially alleviated the attenuated survival RvΔcysM and RvΔcysK2, inhibition of both ROS and RNS by treating peritoneal macrophages isolated from phox^−/− with iNOS inhibitor completely nullified survival defects of the mutants (Figure 4f). These results suggest that non-canonical L-cysteine synthases, CysM and CysK2, play an essential role in combating host-induced redox stress.

Fig 4

CysM and CysK2-derived L-cysteine support mycobacterial survival in vivo

Given the importance of non-canonical L-cysteine synthases in mitigating host-induced redox stress, we next sought to examine whether CysM and CysK2 independently facilitate mycobacterial survival in a murine infection model. We followed disease progression by enumerating colony forming units (CFU) of the lung and spleen of the infected mice at day 1, 4, and 8 weeks p.i (Figure 5a & d). CFUs enumerated on day 1 showed equal bacillary deposition across strains in the lungs (Figure 5b & e). Compared with Rv, survival of RvΔcysM was ∼1.25 log₁₀ lower at 4 weeks p.i, which further attenuated to ∼1.71 log₁₀ at 8 weeks p.i in the lungs (Figure 5b). Similarly, dissemination of RvΔcysM in the spleen was ∼1.63 log₁₀ lower at 4 weeks p.i and ∼1.72 log₁₀ lower at 8 weeks p.i compared with Rv (Figure 5c). Similarly, RvΔcysK2 was ∼1.67 log₁₀ and ∼1.87 log₁₀ lower at 4 and 8 weeks p.i, respectively, compared with Rv in the lungs (Figure 5e) and ∼1.63 log₁₀ and ∼1.72 log₁₀ lower at 4 and 8 weeks p.i, respectively in the spleens of infected mice (Figure 5f).

Fig 5

To understand whether the ability of CysM to mitigate oxidative and nitrosative stress is linked to its role in facilitating mycobacterial survival in vivo, we infected WT, phox^-/- and IFNγ^-/- mice with Rv, RvΔcysM and RvΔcysM::M and analyzed bacillary load at 4 weeks p.i (Figure 5g). As shown in Figures 5h & 5i, the survival of RvΔcysM is impaired in WT mice that produce both ROS and RNS, compared with phox^-/- and IFNγ^-/- mice. The survival defect observed due to the absence of CysM in the lungs was partially rescued in phox^-/- mice. Due to the lack of ROS and RNS in IFNγ^-/- mice, RvΔcysM showed higher bacillary load than in phox^-/- mice (Figure 5h). RvΔcysM displayed better survival in phox^-/- and IFNγ^-/-mice in the spleen than lungs, which could either be because of relatively lesser ROS and RNS stress or imperiled response in clearing Mtb infection (Figure 5i). In contrast to the peritoneal macrophage infection experiment (Figure 4), attenuated survival of RvΔcysM was not completely salvaged in IFNγ^-/^- mice, suggesting that CysM may also be involved in alleviating additional stresses such as nutrient deprivation and/or IFNγ independent ROS/RNS produced by the host. Data presented suggests that L-cysteine produced through non-canonical pathways are independently important for mycobacterial survival in vivo.

L-Cysteine synthase inhibitors can effectively kill Mtb within the host

Data presented above demonstrated that CysM and CysK2 might serve as clinically-important targets for adjunct TB therapy. Brunner et al. screened a compound library to identify inhibitors of mycobacterial L-cysteine synthases-CysK1, CysK2, and CysM. We selected three compounds - Compound 1 (C1) (named compound 2 in [13]), which inhibits all three synthases; Compound 2 (C2) (compound 6 in [13]), which inhibits both non-canonical L-cysteine synthases and Compound 3 (C3) (compound 31 in [13]), which selectively inhibits CysK1 (Figure 6a & b). To examine the therapeutic potential of these inhibitors, we first tested their effect on the survival of Rv within host peritoneal macrophages. Treatment with C1 resulted in ∼1 log₁₀ attenuation compared with the untreated cells. Killing mediated by C2 or C3 was marginally lower compared with C1 (Figure 6c). Next, we examined the ability of these drugs to enhance the bactericidal activity of INH, a drug whose activity depends on redox state of the cell [14], and therefore could be affected by disruption of L-cysteine downstream pathways. Towards this goal, we first measured the MICs of INH (0.06 µg/ml) and C1 (0.6 mg/ml), C2 (0.6 mg/ml), and C3 (0.15 mg/ml) independently. As expected, all the compounds were ineffective in killing Mtb because of redundant roles and relatively less requirement of L-cysteine synthases during regular growth conditions. To examine the combinatorial effect, we selected sub-MIC values of INH (0.03 µg/mL, encircled orange) and increasingly two-fold diluted concentrations of C1, C2, or C3, all below MIC values (Figure 6c-e; blue encircled denotes the starting concentration) and assessed bacterial survival upon combinatorial treatment with INH and the three L-cysteine synthase inhibitors. The addition of either of the inhibitors below MIC rendered Mtb highly susceptible to INH, highlighting the potency of these compounds to serve as an adjunct therapy to combat mycobacterial infection.

Fig 6

Discussion

Throughout its course of infection, Mtb must persist in a hostile, oxidizing, and nutrient-deprived environment of host macrophage. Understanding the dynamic metabolic interactions between the pathogen and its host is imperative to identify its weaknesses which can be exploited to design new-age chemotherapy. Our study provides convincing evidence that the genes involved in the L-cysteine biosynthetic pathway are attractive targets for the design of anti-mycobacterial drugs. Genes involved in sulfur assimilation and L-cysteine biosynthesis were found to be upregulated in the transcriptome profile of Rv subjected to various host-like stresses. The sulfur metabolism pathway was particularly enriched upon the addition of oxidizing agent, CHP (Figure 1, S2 & S3). This observation has been consistently reported by multiple studies demonstrating the up-regulation of sulfur assimilation and L-cysteine biosynthetic genes in response to oxidative stress, nutrient deprivation, and macrophage infection [1-5]. Besides being involved in protein synthesis, L-cysteine is also important for the biosynthesis of L-methionine, S-adenosyl methionine, coenzyme A, and iron-sulfur clusters. As a thiol-containing molecule, L-cysteine contributes to the intracellular redox state directly and through the production of major antioxidants like mycothiol and ergothioneine. These numerous essential functions of L-cysteine prompted us to further examine its roles in the context of Mtb cellular metabolism and virulence.

Unlike MtbΔcysH, which was reported to be an L-cysteine auxotroph and thus required L-methionine or glutathione supplementation to grow in vitro (from which L-cysteine can be generated catabolically)[15], neither deletion of cysK2 nor cysM impacted Mtb growth kinetics in vitro suggesting that these genes are functionally redundant during rich growth conditions [5] [11] (Figure S4). Interestingly, the induction of various host-like stresses attenuated the growth of these mutants compared to the parental strain, pointing towards the possibility of an enhanced requirement of L-cysteine-derived antioxidants and other biomolecules to subvert these stresses (Figure 1). In agreement with this hypothesis, we previously showed that the cellular thiol levels are higher during oxidative stress in Mtb [5]. Similarly, various MSH and ERG mutants display enhanced sensitivity to oxidative stress caused by treatment with H₂O₂, cumene hydroperoxide, or O^2•− [16-20].

While the transcriptomic profiles of mutants were highly similar to each other during regular growth conditions, the addition of CHP resulted in differential expression of >30% Mtb genes, indicating that cues and downstream effects of the two non-canonical L-cysteine synthases are partially non-overlapping, further underlying their importance for the bacillary growth inside the host (Figure 2). We also found that the steady-state levels (pool sizes) of key L-cysteine derived antioxidants, ergothioneine, and mycothiol, are significantly different at 24 h (Figure 3). The rate of synthesis of mycothiol, mycothione and ergothioneine were observed to be lower in RvΔcysM compared to Rv and RvΔcysK2.

The addition of L-cysteine in oxidatively stressed cells nullified the compromised survival of the mutants indicating that Mtb cells are able to uptake L-cysteine from the extracellular medium, as shown previously [5], and reduced levels of L-cysteine in the mutants are chiefly responsible for their attenuated growth in the presence of stresses (Figure 4). Various Mtb mutants in mycobacterial sulfur metabolism genes are severely compromised to persist within the host and cause disease [15, 21-26]. Inorganic sulfate is present at 300-500 μM in human plasma [26]. However, the inability of host-derived sulfur/L-cysteine to compensate for attenuated survival of mutants is suggestive of either inaccessibility of sufficient L-cysteine in Mtb niches or inefficient expression, function, or uptake by transporters in vivo. Importantly, compromised survival of RvΔcysM and RvΔcysK2 in vivo was largely mitigated in IFNγ-/- and Phox-/-, indicating that the non-canonical L-cysteine synthases, CysM and CysK2, independently facilitate mycobacterial survival during immune-mediated redox stress (Figure 5).

Using network analysis of the Mtb protein interactome, a flux balance analysis of the reactome and randomized transposon mutagenesis data, along with sequence analyses and a structural assessment of targetability, Raman et al., reported CysK2 and CysM as high confidence drug targets [27]. Similarly, inhibitors of CysM showed mycobactericidal activity in a nutrient-starvation model of dormancy [28]. Interestingly, humans do not reduce sulfur to produce L-cysteine; they rather synthesize L-cysteine through SAM-dependent transmethylation followed by transsulfuration of L-methionine. Owing to their complete absence in humans, mycobacterial L-cysteine biosynthetic genes and their regulators represent unique, attractive targets for therapeutic intervention [29]. We found that both RvΔcysM and RvΔcysK2 displayed enhanced antibiotic sensitivity in vitro and within the host (data not shown). Importantly, a combination of L-cysteine synthase inhibitors with front-line TB drugs like INH, significantly reduced the bacterial load inside the host (Figure 6). Altogether, this study demonstrates for the first time that the two non-canonical L-cysteine synthases have non-redundant biological functions. Deletion or biochemical inhibition of CysM or CysK2 perturbs redox homeostasis of Mtb and allow for the maximal effect of host macrophages antibacterial response, and thus increased elimination of virulent Mtb.

Methods

Bacterial strains and culturing conditions

Mtb culturing conditions were performed as described previously [5, 30].

Generation of RvΔcysK2 mutant and complementation strain

We generated gene replacement mutants through the recombineering method [31] as previously described [5, 30]. Briefly, 671bp upstream of 176^th nucleotide from 5’ end (5’ flank) and 643 bp downstream of 943^rd nucleotide from 3’ end (3’ flank) of cysK2 were PCR amplified from Rv genomic DNA using Phusion DNA polymerase (Thermo Scientific). The amplicons were digested with BstAPI and ligated with compatible hygromycin resistance (hyg^r) cassette and oriE+cosλ fragments [31], to generate the allelic exchange substrate (AES). AES was digested with SnaBI to release the LHS-hyg^r-RHS fragment, and the eluted fragment was electroporated into the recombineering proficient Rv-ET strain [5]. Multiple hyg^r colonies were examined by PCRs to screen for legitimate recombination at cysK2 locus. RvΔcysK2, thus generated, was cured of pNit-ET through negative selection on LB Agar plates containing 2% sucrose. To generate the complementation strain, full length cysK2 gene was PCR amplified from genomic DNA isolated from Rv as the template and Phusion DNA polymerase (ThermoFischer Scientific). The amplicon was digested with NdeI-HindIII (NEB), cloned into the corresponding sites in pNit-3F and the resultant plasmid was electroporated into RvΔcysK2 to generate the RvΔcysK2::cysK2 strain. RvΔcysM and RvΔcysM::cysM strains were generated in the lab previously using a similar method [2].

Ex vivo infection experiments

Balb/c, C57BL/6 (B6), phox^-/- (B6.129S6-Cybbtm1Din/J; JAX# 002365) or IFN-γ^-/- (B6.129S7-Ifngtm1Ts/J; JAX#002287) mice were procured from The Jackson Laboratory. 4% thioglycolate (Hi-Media) was injected into the peritoneum cavity of 4- to 6-weeks old mice and peritoneal macrophages were extracted four days post-injection and seeded and processed as described [5]. In specific cases, the peritoneal macrophages were treated with 10 ng/ml IFNγ (BD Biosciences) overnight and 10 ng/ml lipopolysaccharide (LPS, Sigma) for 2 h for activation. Where indicated, cells were further pretreated with 100 µM 1400 W (Sigma) overnight to inhibit iNOS or pretreated with 0.2 mM L-cysteine before infection. Single-cell Mtb suspensions were used for infection at 1:10 (host cells: bacteria) MOI. 4 h post-infection (p.i), cells were washed thrice and replenished with complete RPMI containing IFNγ + LPS, 1400 W, or L-cysteine, as required. The infected host cells were washed thrice with PBS, lysed using 0.05% Dodecyl sulfate sodium salt (SDS), and serial dilutions were plated on 7H11-OADC to enumerate bacillary survival.

In vivo infection experiments

Mice (4–6-weeks) housed in ventilated cages at the Tuberculosis Aerosol Challenge Facility at the International Centre for Genetic Engineering and Biotechnology (New Delhi, India) were infected via aerosol route with ∼ 100 bacilli using the Madison Aerosol Chamber (University of Wisconsin, Madison, WI). At 24 h post-infection, mice (n=2; per group) were euthanized to determine the bacterial deposition. At 4/8 weeks p.i, lungs and spleen were homogenized and plated on 7H11+OADC containing PANTA.

RNA isolation and qRT–PCRs

Mtb strains were cultured in triplicates in 7H9-ADS till O.D. reached 0.3-0.4. One set was left untreated (control), the other was treated with 50 μM of CHP for 6 h. A culture equivalent to 10 O.D.₆₀₀ was resuspended in TRIzol (Invitrogen). Zirconium beads were added to the cell-Trizol mix to facilitate lysing Mtb cells with the help of a bead beater (MP FastPrep system, MP Biomedicals) and RNA was extracted and analyzed as described [5]. Data was plotted as 2^(−ΔΔCt) wherein the gene expression was normalized with respect to 16s rRNA (rrs gene), followed by normalization with control strain/condition/group.

RNA sequencing and analysis

Total RNA was isolated from two-three biological replicates of indicated Mtb strains, and their concentrations were checked by Qubit (ThermoFischer Scientific) followed by quality assessment through Agilent 2100 BioAnalyzer (Agilent RNA 6000 Nano Kit). Both sets of RNA seq represented in Figures 1 and 5 were processed and run at the same time, the same set of control (Rv) triplicates were used for the analysis of both figures. Samples with RIN values > 7 were processed for RNA sequencing using the Illumina NovaSeq 6000 Platform (CSIR-CCMB central facility; read length of 100 bp, 20 million paired-end reads / sample).

Illumina adapters and low-quality reads were discarded and those with quality scores < 20 and smaller than 36 bp were eliminated from raw sequencing reads using cutadapt [32]. Processed reads were mapped to the Mtb H37Rv, (https://ftp.ncbi.nlm.nih.gov/genomes/refseq/bacteria/Mycobacterium_tuberculosis/reference/G CF_000195955.2_ASM19595v2/), using hisat2 with default parameters [33]. Uniquely aligned reads were counted with the help of feature Counts of Subread package [34] and those with total read count < 10 across all the samples were removed, and the rest were used for further downstream analysis. Differentially expressed genes (DEGs) were identified using DESeq2 [35] and those with adjusted p-value < 0.05 and absolute log₂ Fold change >1 or 0.5 were considered. The raw read counts were rlog normalized the raw read counts for PCA plot and heat map with the DESeq2 package.

Functional enrichment analysis

Functional enrichment analysis was performed with DAVID web services [36]. We specifically used GO terms and KEGG pathways for this analysis. Only top 10 enriched hits based on gene counts were plotted.

Preparation of samples for metabolomics

Mtb strains were grown in 7H9 media until an OD 1 and then inoculated onto 0.22um nitrocellulose filters and grown on 7H10 plates containing 0.5g/L BSA Fraction V, 0.2% dextrose and 0.085% NaCl (ADS) for 5 days. The filters were then transferred to 7H10 plus ADS plates containing Sodium sulfate-34S (Merck 718882) containing either 50uM CHP or no CHP for 24 and 48 h. The metabolites were extracted by mechanical lysis in cold acetonitrile/methanol/water (2:2:1) containing 0.1mm acid washed Zirconia beads. The lysates were clarified by centrifugation and filtered through a 0.22um Spin-X column (Costar). The lysates were mixed 1:1 with acidified acetonitrile (0.1% formic acid). The samples were normalized to protein concentration using the BCA assay kit (Thermo).

Liquid chromatography – mass spectrometry metabolomics

LC–MS analysis was done in an Agilent 1290 Infinity II HPLC connected to a 6230B time-of-flight (ToF) mass spectrometer using a Dual AJS ESI ionization source. Compounds were separated in a Cogent Diamond Hydride Type C silica column (2.1 × 150 mm). Solvent A was LC– MS grade H2O + 0.1% (v/v) formic acid and solvent B was acetonitrile + 0.1% (v/v) formic acid. The gradient was from 85% B to 5% B over 14 min. Flow rate was 0.4 mL/min. The ion source parameters were as follows: Gas temperature 250 °C, Drying gas flow rate 13 l/min, nebulizer pressure 35 psig, sheath gas temperature 350 °C, sheath gas flow 12 l/min, capillary voltage 3500 V and nozzle voltage 2000 V, The ion optic voltages were 110 V for the fragmentor, 65 V for the skimmer and 750 for the octopole radio frequency voltage. MS data were analyzed with the MassHunter suite version B0.7.0.00.

MIC analysis

MIC values were assessed using Alamar Blue assay, as described previously [37]. Briefly, 100μl 7H9-ADS medium without Tween 80 was added to each well of a 96-well plate. First well of each column were filled with 100μl of the test drug/antibiotic, which was then serially diluted across the column. Mtb cells corresponding to 0.01 A₆₀₀ were diluted in 100μl 7H9-ADS medium and were added to each well. Two rows one in which the drug was not added (no drug [ND]), and the other wherein Mtb cells were replaced with 7H9-ADS (no cells [NC]) acted as controls. The 96 well plate was sealed with parafilm and kept at 37°C. After 5 days, 20μl of 0.25% filter-sterilized resazurin was added to each well, and color development was captured after 24 h.

Statistical analysis

Unless otherwise specified, experiments were performed in triplicates and repeated independently at least twice. CFU results were plotted, and significance of the datasets was calculated using one-way ANOVA followed by a post hoc test (Tukey test) on GraphPad Prism 5. Figures were customized using Adobe Illustrator version 26.3.1. Statistical significance was set at P values < 0.05 significant (*, P < 0.05; **, P < 0.005; ***, P < 0.0005). Source datasets can be requested from the corresponding author.

Data availability

RNA Seq data is available at the NCBI Gene Expression Omnibus Database, accession no. GEO GSE225792, the link to the database is https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE225792 and the reviewer token for private access is ipmnekailfmlbcj.

Acknowledgements

We thank Dr. Apruva Sarin for phox^−/− mice and thoughtful discussion; ICGEB for access to their TACF; CCMB for access to their sequencing facility. MZK was supported by Research Associateship from SERB (CRG/2018/001294). Research reported in this publication was supported by the core grant of the National Institute of Immunology; DBT grant BT/PR13522/COE/34/27/2015; SERB grant CRG/2018/001294, and JC Bose fellowship JCB/2019/000015 to VKN. Work in LPSC’s laboratory was supported by the Francis Crick Institute, which receives its core funding from Cancer Research UK Grant FC001060, UK MRC Grant FC001060, and Wellcome Trust Grant FC001060.

Author contributions

MZK, LPSC and VKN conceptualized the study; MZK contributed to generation of constructs and strains; BS and MZK were involved in the preparation of RNA sequencing samples, while NKS and DTS performed the analyses. MZK and BS performed animal studies at ICGEB; MZK contributed to in vitro stress susceptibility assays, macrophage assays and inhibitor studies; DMH and LPSC performed the metabolomic studies; DVSP and DS synthesized chemical inhibitors; MZK curated the data. MZK, YK, LPSC and VKN wrote the manuscript; VKN and LPSC acquired funding for the study. All authors read and approved the manuscript.

Ethics statement

Protocols for animal experiments were preapproved by the Animal Ethics Committee of the National Institute of Immunology, New Delhi, India (IAEC numbers 409/16 and 462/18) as per standard institutional guidelines.

Supplementary Figure 1

Supplementary Figure 2

Supplementary Figure 3

Supplementary Figure 4

Supplementary Figure 5

Supplementary Figure 6

Table S1: Differentially Expressed Genes (DEG) analysis Rv under stress vs Rv.

Table S2: Pathway enrichment analysis of DEGs from Table S1.

Table S3: DEG analysis of RvΔcysM or RvΔcysK2 vs Rv with or without CHP stress & DEG analysis RvΔcysM +CHP vs RvΔcysK2 + CHP.

Table S4: Pathway enrichment analysis of DEGs from Table S3.