Cytokinin Signaling in Mycobacterium tuberculosis

ABSTRACT It was recently reported that the human-exclusive pathogen Mycobacterium tuberculosis secretes cytokinins, which had only been known as plant hormones. While cytokinins are well-established, adenine-based signaling molecules in plants, they have never been shown to participate in signal transduction in other kingdoms of life. M. tuberculosis is not known to interact with plants. Therefore, we tested the hypothesis that cytokinins trigger transcriptional changes within this bacterial species. Here, we show cytokinins induced the strong expression of the M. tuberculosis gene Rv0077c. We found that Rv0077c expression is repressed by a TetR-like transcriptional repressor, Rv0078. Strikingly, cytokinin-induced expression of Rv0077c resulted in a loss of acid-fast staining of M. tuberculosis. While acid-fast staining is thought to be associated with changes in the bacterial cell envelope and virulence, Rv0077c-induced loss of acid-fastness did not affect antibiotic susceptibility or attenuate bacterial growth in mice, consistent with an unaltered mycolic acid profile of Rv0077c-expressing cells. Collectively, these findings show cytokinins signal transcriptional changes that can affect M. tuberculosis acid-fastness and that cytokinin signaling is no longer limited to the kingdom Plantae.

Kan r pMV306.kan-Rv0078 plasmid with Trp100 changed to Arg This work pET24b(ϩ)-Rv0077cHis 6 Kan r ; for production of Rv0077c-His 6 in E. coli This work pET24b(ϩ)-Rv0078His 6 Kan r plasmid for production of Rv0078-His 6 in E. coli This work ( Fig. 1A; see Table S1 in the supplemental material). Rv0077c is conserved among many mycobacterial species, while Rv0076c and Rv0078 are present only in several mycobacterial genomes (Fig. 1B) (11). Notably, Mycobacterium smegmatis, a distant, nonpathogenic relative of M. tuberculosis, has a weak homologue of Rv0077c and no conspicuous Rv0078 homologue (Fig. 1B). mmpL6 is one of 13 mmpL (mycobacterial membrane protein large) genes in M. tuberculosis. In strain H37Rv, mmpL6 is predicted to encode a 42-kDa protein with five transmembrane domains and is truncated compared to the same gene in ancestral tuberculosis strains (12). Thus, it is unclear if mmpL6 encodes a functional protein in strain H37Rv. Rv0077c was by far the most strongly induced gene in M. tuberculosis upon iP treatment; therefore, we chose it for follow-up studies. We raised polyclonal antibodies to recombinant Rv0077c protein and showed that protein levels were increased in M. tuberculosis treated with iP for 24 h (Fig. 1C, first two lanes). Rv0077c was barely detectable in cell lysates of bacteria that had not been incubated with iP and was undetectable in a strain with a transposon insertion mutation in Rv0077c (Fig. 1C, center two lanes). Rv0077c protein was restored to WT levels in the mutant upon complementation with an integrative plasmid carrying Rv0077c expressed from its native promoter (Fig. 1C, last two lanes). We also found a dose-dependent induction of Rv0077c production using iP concentrations from 1 nM to 100 M (Fig. 1D).
We next synthesized and tested if the most abundantly produced cytokinin in M. tuberculosis, 2-methylthio-iP (2MeSiP) (8), could also induce Rv0077c production. 2MeSiP strongly induced Rv0077c production (Fig. 1E, lane 4). Importantly, we did not observe induction of Rv0077c when we incubated the bacteria with the appropriate cytokinin riboside (R) precursors iPR or 2MeSiPR (Fig. 1E, lanes 3 and 5). Similarly, AMP or the closely related molecule adenine could not induce Rv0077c expression (Fig. 1E,  lanes 6 and 7). We hypothesized while adenine could not induce Rv0077c expression, at high enough concentrations, it could possibly inhibit Rv0077c induction by com- In the primer sequences, changes to the normal sequence are in boldface. In the probe sequences, the 14-nucleotide oligomers are underlined.
peting with cytokinin for access to a transporter or receptor. Indeed, adenine reduced the induction of Rv0077c by iP in a dose-dependent manner (Fig. 1F).
Identification of an operator for the TetR-like transcriptional repressor Rv0078. Rv0077c is divergently expressed from Rv0078; the proposed translational start codons for these genes are separated by 61 bp. Based on its cytokinin-dependent expression and its proximity to Rv0077c, we hypothesized that Rv0078 encodes a repressor of both genes. We identified the promoter regions for each gene by performing rapid amplification of 5= cDNA ends (5= RACE) analysis of Rv0077c and Rv0078 and determined the likely start of transcription for one gene was within the 5= untranslated region of the other gene ( Fig. 2A). We sought to identify an operator sequence by using an electrophoretic mobility shift assay (EMSA) (Fig. 2B). We narrowed down a putative Rv0078 binding site to a region overlapping the proposed starts of transcription (ϩ1) of both Rv0077c and Rv0078 ( Fig. 2A, red box). TetR-like regulators generally bind to inverted repeat sequences; we identified a 21-bp sequence within positions Ϫ14/ϩ19 (relative to position ϩ1) of Rv0077c containing two 10-bp inverted repeat sequences. Mutagenesis of the probe in the repeat sequences disrupted Rv0078 binding to the DNA (Fig. 2B). Importantly, the binding sequence overlaps the putative transcriptional start sites of both genes, suggesting that expression of both genes is repressed by Rv0078. Notably, the addition of iP did not result in the release of Rv0078 from the DNA probe ( Fig. 2B; see Fig. S1 in the supplemental material). Therefore, cytokinins do not appear to directly bind to Rv0078 to induce gene expression.
Based on the 5= RACE analysis, we were able to delete and replace most of the Rv0078 gene with the hygromycin resistance gene (hyg) without disrupting the promoter of Rv0077c in M. tuberculosis H37Rv. A ΔRv0078::hyg strain displayed constitutively high expression of Rv0077c irrespective of the presence of cytokinin, supporting a model in which Rv0078 directly represses Rv0077c expression (Fig. 2C). A single copy of Rv0078 expressed from its native promoter restored iP-regulated control of Rv0077c in this strain (Fig. 2C, lanes 5 and 6). Interestingly, in the process of making the complementation plasmid, we acquired a random mutation (likely generated during PCR) in Rv0078 that changed a tryptophan to arginine (W100R); this allele was unable to restore cytokinin-dependent regulation of Rv0077c to the ΔRv0078::hyg strain (Fig. 2C, lanes 7 and 8).
Two Rv0078 dimers bind to one operator. To gain an understanding of how Rv0078 represses gene expression, we solved the crystal structure of Rv0078 to 1.85 Å by a single-wavelength anomalous dispersion method. As previously reported, Rv0078 forms dimers resembling other TetR-like proteins (13,14). Unlike the canonical TetR binding site, tetO, which is 15 bp long, the Rv0078 binding site is 21 bp long, suggesting Rv0078 binds to DNA differently than TetR. We performed an isothermal titration calorimetry (ITC) experiment and determined that the apparent K D (dissociation constant) of Rv0078 for operator DNA was 357 nM (Fig. 3A). The estimated stoichiometry of DNA duplex to Rv0078 dimer was 0.66, suggesting that two Rv0078 dimers bind to one operator sequence.
We next cocrystallized Rv0078 with a 23-bp DNA fragment that was extended 1 bp at each end of the 21-bp operator sequence identified in Fig. 2 (positions Ϫ8/ϩ13  (Table S1). Rv0078 was purified under native conditions from E. coli. (C) Deletion and disruption of Rv0078 result in the constitutive expression of Rv0077c. Total cell lysates were prepared and separated on a 10% SDS-PAGE gel. IB, immunoblot. The membrane was stripped and then incubated with antibodies to DlaT to confirm equal loading of samples. Molecular weight standards are indicated to the left of the blots and are in kilodaltons (kDa).
Cytokinins Induce Transcription in M. tuberculosis ® [relative to ϩ1] of Rv0077c) and solved the structure to a resolution of 3.0 Å ( Table 3). As predicted by the ITC experiments, we observed two Rv0078 dimers bound to a DNA duplex where each dimer bound on opposite sides of the DNA (Fig. 3B). The top or "central" dimer bound the DNA palindrome symmetrically, while the bottom or "peripheral" dimer bound to DNA off-center, staggered from the central dimer binding site by 7 bp. Thus, the longer-than-canonical (i.e., tetO) binding site is required for accommodation of two Rv0078 dimers, unlike TetR, which binds to tetO as a single dimer.
Rv0078 has an N-terminal DNA-binding domain (DBD) and a C-terminal ligandbinding domain (LBD) (Fig. 3B to D). DNA binding induced significant conformational changes across the Rv0078 dimer structure, with a root mean square deviation (RMSD) of 2.13 Å compared to the DNA-free dimer. In particular, the two ␣3 helices move toward each other by~6 Å, in order to reduce their distance to 36.4 Å and fit in the DNA major grooves (Fig. 3C). In the LBD, the ligand entry between helices ␣4 and ␣5 is open and the ligand-binding pocket is empty. It appears that changes initiated in the DBD regions upon binding to the DNA are transmitted to the LBD via the DBD-LBD interface, which involves extensive interactions, including two salt bridges and six H-bonds (Fig. 3D).
The ligand-binding pocket of Rv0078 is largely hydrophobic. Within this pocket, we observed an elongated density resembling a long aliphatic chain of a fatty acid (see Fig. S2 in the supplemental material). Gas chromatography-mass spectrometry (GC-MS) of the compounds extracted from Rv0078 purified from Escherichia coli revealed fatty acids commonly found in this organism, and a palmitate molecule fit well with the electron density ( Fig. S2A to E). A fatty acid carboxylate formed a hydrogen bond with Rv0078 Glu-70 ( Fig. 3E and Fig. S2C), and the long alkyl chain had numerous hydrophobic interactions within the extended ligand pocket. When we purified Rv0078 under denaturing conditions and refolded the protein to remove the lipid, we observed the same EMSA results as we observed with protein purified under native conditions (data not shown), suggesting this fatty acid is unlikely to be the native Rv0078 ligand; however, these data may suggest that the native ligand is fatty acid-like. Interestingly, we found that Trp100 faces the ligand-binding pocket (Fig. 3E), which suggests Trp100 interacts with the natural ligand. This hypothesis is supported by our data showing an Rv0078 W100R mutant was unable to complement the Rv0078 deletion mutant strain (Fig. 2C, lane 8).
In addition to characterizing the LBD, we identified 9 amino acids that were important for interacting with DNA ( Fig. 4A to E). Specifically, the hydroxyl groups of Arg48 is the only residue that interacted with DNA by recognizing a guanine base at a distance of 2.6 Å. To further examine the importance of these residues, we introduced single amino acid substitutions (T37V, T47V, R48M, and Y52F) or double mutations (T47V and R48M) into Rv0078 and performed EMSAs. While all of the mutant proteins were soluble and behaved like the WT protein in solution, Rv0078 R48M did not bind to DNA, and the other mutant proteins bound to DNA with reduced affinities (Fig. 4F).
To examine the DNA sequence specificity of Rv0078, we synthesized three EMSA probes by changing the Rv0078 Arg48-interacting guanines to adenines (G-to-A) in the central dimer binding region (probe S1), the peripheral dimer-binding region (probe S2), or both dimer-binding regions (probe S3). These G-to-A substitutions either abolished or compromised Rv0078 binding to DNA, affirming the critical role of guanines in the binding site (Fig. 4G). Notably, substitutions in probe S1 entirely abolished Rv0078 binding, while substitutions in S2 retained a partial gel shift at a high concentration (white arrowhead in Fig. 4G). This observation suggests that the binding of the two Rv0078 dimers is cooperative, with the central dimer likely the first one to bind to DNA. Constitutive expression of Rv0077c does not affect antibiotic susceptibility or virulence in mice. During our ongoing studies, a report was published on the identification of a small molecule of the spiroisoxazoline family, SMARt-420, which strongly induces the expression of the Rv0077c orthologue bcg_0108c in M. bovis bacillus Calmette-Guerin (15). Using X-ray crystallography and surface plasmon resonance techniques, the authors of this study found SMARt-420 binds to Rv0078 to derepress binding from the Rv0077c promoter (14,15). SMARt-420 was identified in a search for compounds that boost the efficacy of the second-line tuberculosis drug ethionamide (ETH). ETH is a prodrug that is activated by the mono-oxygenase EthA, which transforms ETH into highly reactive intermediates. Activated ETH and NAD form a stable adduct that binds to and inhibits InhA, an essential enzyme needed for mycolic acid synthesis in mycobacteria (16,17). While spontaneous inactivating mutations in ethA result in resistance to ETH, it was proposed that the induction of Rv0077c expression could bypass the need for EthA and transform ETH into its toxic form (15). Based on this study, we predicted that an Rv0078 mutant of M. tuberculosis, which expresses high levels of Rv0077c, should be hypersensitive to ETH compared to the parental strain H37Rv. However, we observed either little to no significant change in the 50% inhibitory concentration (IC 50 ) of ETH between the WT and ΔRv0078::hyg strains (Table 4) or in WT M. tuberculosis H37Rv treated with or without 100 M iP (IC 50 , 1.4). We also tested if the constitutive expression of Rv0077c changed the susceptibility of M. tuberculosis to other antibiotics, including two cell wall synthesis inhibitors. We observed no differences in the IC 50 s of these antibiotics between the WT and ΔRv0078:: hyg strains (Table 4).
We also tested if either the Rv0077c or Rv0078 mutant had growth defects in vivo compared to the WT H37Rv strain. We infected C57BL/6J mice by a low-dose aerosol route with the parental, mutant, and complemented mutant strains, as well as with the Rv0078 mutant transformed with the Rv0078 W100R allele. None of the strains revealed a difference in growth or survival compared to WT M. tuberculosis in mice, as determined by the recovery of CFU from the lungs and spleens ( Fig. 5; see Fig. S3 in the supplemental material).
Expression of Rv0077c alters acid-fast staining of M. tuberculosis. To gain insight into the function of Rv0077c, we performed metabolomic analysis of strains expressing Rv0077c to potentially determine how its presence alters bacterial physiology. We prepared total cell lysates of WT and Rv0077c mutant strains treated with or without 100 M iP for 24 h (see Materials and Methods). From a total of 337 detectable metabolites, we observed a significant change in 24 molecules after the addition of iP to WT M. tuberculosis. Seventeen metabolites showed a consistent difference between samples in which iP was added, and these changes disappeared in an Rv0077c-disrupted strain, suggesting the changes were specifically due to the presence of Rv0077c (Table S2, highlighted in yellow). We observed an increased abundance of several phospholipids and a decrease in a major precursor of peptidoglycan, N-acetylglucosamine-1-phosphate (GlcNAc1P). We therefore hy-  (Fig. 6a versus b), and this phenotype depended on the presence of Rv0077c (Fig. 6c versus d). Complementation of the mutation with Rv0077c alone restored the iP-induced loss of acid-fast staining ( Fig. 6e and f). Deletion of Rv0078 resulted in a constitutive loss of staining, irrespective of the presence of iP ( Fig. 6g and h). Complementation of the Rv0078 deletion with the WT gene restored iP control of loss of acid-fast staining ( Fig. 6i and j), but complementation with Rv0078 W100R , which could not fully derepress Rv0077c expression in the presence of iP (Fig. 2C), could not restore iP-induced loss of acid-fast staining ( Fig. 6k and l). Mixing and simultaneous staining of the Rv0077c and Rv0078 mutants further showed the staining differences were not a result of a technical artifact (Fig. 6m). We also tested whether or not a carbolfuchsin (the primary stain in ZN staining) could interact at all with bacteria expressing Rv0077c and found that both the Rv0078 mutant and the parental strain stained equally well ( Fig. 6n and o). This result suggested Rv0077cexpressing bacteria could bind to stain but not retain it after acid washing. Acid-fast staining is primarily thought to be associated with mycolic acids; alterations in mycolic acid synthesis result in negative effects on cell growth in vitro and in vivo (18,19). We used thin-layer and liquid or gas chromatography coupled to mass spectrometry to analyze the fatty acid, mycolic acid, and lipid contents of the WT and Rv0077c and Rv0078 mutant strains and observed no significant qualitative or quantitative differences between strains (see Fig. S4 and S5 and Table S3 in the supplemental material). In particular, the expression of Rv0077c did not conspicuously modify the chain length of mycolic acids, their cyclopropanation, or the relative abundance of keto-to ␣-mycolates and did not alter the wax ester and triglyceride contents of the strains that have previously been linked to acid-fast staining (19,20).

DISCUSSION
Our studies are the first to demonstrate CKs induce robust and specific transcriptional and physiological changes in a bacterial species. M. tuberculosis treated with CK expressed high levels of Rv0077c, which altered its metabolome and staining properties. Despite these changes, we did not observe an effect on the susceptibility of these bacteria to antibiotics of different classes or on survival in mice. In addition we found the transcriptional regulator Rv0078 represses Rv0077c expression in the absence of CK, and we defined the operator to which Rv0078 binds. We determined that two dimers of Rv0078 bind to inverted repeats in the intergenic region between Rv0077c and Rv0078, which likely represses the expression of both genes.
In plants, CKs are sensed by membrane receptors related to the sensors of bacterial two-component systems (TCSs) (21)(22)(23). In these systems, CKs interact with membrane receptors via a CHASE (cyclases/histidine kinases associated sensing extracellular) domain (24,25). As in bacteria, the ligand-receptor interaction stimulates a phosphorelay that ultimately leads to the phosphorylation of a response regulator protein, which either represses or activates gene expression. M. tuberculosis has at least 12 known TCSs, and none has a predicted CHASE domain (L. Aravind, February 2012, and I. Jouline, August 2017, personal communications). Thus, it remains to be determined if a TCS sensor protein is involved in CK signal transduction in M. tuberculosis.
It also remains to be determined what the natural ligand is for Rv0078; while SMARt-420 binds robustly to Rv0078 (14,15), the cytokinin iP did not bind to this TetR-like regulator in our studies. This result may not be surprising when considering SMARt-420 and cytokinins bear no resemblance to each other. Furthermore, we could not cocrystallize iP with Rv0078 (data not shown). We hypothesize that the interaction of CK with an unknown receptor or enzyme leads to the synthesis of a small molecule (e.g., a lipid) that binds to Rv0078, resulting in the derepression of Rv0077c. This possibility may be supported by our observation that there is an increase in several phospholipids in iP-treated M. tuberculosis. We are currently working to identify factors required for CK-mediated gene induction. Cytokinins Induce Transcription in M. tuberculosis ® Rv0077c is predicted to have an ␣/␤-hydrolase fold (26), which can have a variety of substrate types (27). Based on our metabolomic and microscopy studies, we predict Rv0077c targets one or more components of the cell envelope. The identification of the natural target of Rv0077c may help provide new insight into the elusive molecular basis of acid-fastness of this human-exclusive pathogen. While it is well known that perturbations in mycolic acid synthesis result in loss of acid-fast staining (18,19), our results may indicate that acid-fast staining can be affected by other changes in cell envelope chemistry. Alternatively, Rv0077c may have a subtle chemical effect that is undetectable using the available techniques for quantifying mycolic acids. Importantly, although our results show dramatic changes in acid-fast staining in vitro, we do not know if cytokinin signaling results in these alterations in vivo. It is possible that the production of Rv0077c only affects a partial pool of substrates within a subset of bacteria, the effects of which might not result in a total loss of staining as we observed in vitro.
While we did not observe any differences in bacterial burdens in mice infected with strains that either were disrupted for or constitutively expressed Rv0077c, it is possible that Rv0077c function might only be needed during very late or specific stages of infection. This idea may be supported by a previous observation that the cytokinin synthase log gene in Mycobacterium marinum is specifically expressed in late granulomas of infected frogs (28). Furthermore, mice might not provide an optimal model to observe a role for this pathway in tuberculosis.
A previous report suggested the expression of Rv0077c increases the sensitivity of M. tuberculosis to the antibiotic ETH. ETH must be activated by a monooxygenase, EthA, in order for it to be toxic to M. tuberculosis (16,17). We found the expression of Rv0077c did not confer increased susceptibility to ETH or any other antibiotic we tested, an observation consistent with the likelihood that Rv0077c is not a monooxygenase. It is possible that the effects of SMARt-420 are growth condition dependent, that this molecule affects another pathway to increase the susceptibility of M. tuberculosis to ETH, or that another SMARt-420-induced enzyme and Rv0077c synergize to activate ETH (A. Baulard, personal communication). Irrespective of these possibilities, it is unlikely that Rv0077c has a considerable role in the activation of ETH. While a previous report named Rv0077c and Rv0078 "EthA2" and "EthR2," respectively, we propose to rename them "LoaA" and "LoaR" for "loss of acid-fast staining A and repressor," respectively, due to the lack of association of these proteins with ETH susceptibility.
Finally, our studies have opened the door to the possibility that numerous commensal and pathogenic microbes (including fungi) use cytokinins for intra-or interspecies communication in complex systems such as the gut microbiome. The identification of one or more CK receptors and the signal transduction pathway that leads to the induction of Rv0077c expression will likely lay the foundation for understanding CK signaling in hundreds of bacterial species.

MATERIALS AND METHODS
Bacterial strains, plasmids, primers, chemicals, and culture conditions. The bacterial strains, plasmids, and primer sequences used in this study are listed in Table S1. All primers for cloning and sequencing were from Invitrogen, Inc. M. tuberculosis strains were grown in Middlebrook 7H9 broth (Difco) supplemented with 0.2% glycerol, 0.05% Tween 80, 0.5% fraction V of bovine serum albumin (BSA), 0.2% dextrose, and 0.085% sodium chloride ("7H9c"). M. tuberculosis cultures were grown without shaking in 25-or 75-cm 2 vented flasks (Corning) at 37°C. 7H11 agar (Difco) supplemented with 0.5% glycerol and BBL TM Middlebrook OADC enrichment (BD) was used for growth on solid medium ("7H11"). M. tuberculosis was transformed as described previously (29). E. coli strains used for cloning and expression were grown in LB-Miller broth (Difco) at 37°C with aeration on a shaker or on LB agar. E. coli strains were chemically transformed as previously described (30). The final concentrations of antibiotics used for M. tuberculosis growth were as follows: kanamycin, 50 g/ml; hygromycin, 50 g/ml; and streptomycin, 25 g/ml. The final concentrations for E. coli growth were as follows: hygromycin, 150 g/ml; kanamycin, 100 g/ml; and streptomycin, 50 g/ml. AMP, adenine, and iP were purchased from Sigma. iPR, 2MeSiP, and 2MeSiPR were synthesized as previously described (31). The purity of the synthesized cytokinin and derivatives was Ͼ98% for each, as determined by high-performance liquid chromatography and mass spectrometry.
The Rv0077c::MycoMarT7 mutant was isolated from a library of ordered transposon insertion mutants as previously described (3,32). The ΔRv0078::hyg mutant was made by deletion-disruption mutagenesis as described in detail elsewhere using pYUB854 (33,34). All mutants were confirmed by PCR using primers that annealed to sequence beyond the region used for making allelic exchange plasmids.
Protein purification and immunoblotting. DNA sequence encompassing the full-length Rv0078 or Rv0077c gene was cloned into pET24b(ϩ) vector using primers listed in Table S1. Recombinant proteins were produced in E. coli ER2566 and purified under native conditions for Rv0078 and denaturing conditions for Rv0077c according to the manufacturer's specifications (Qiagen). Polyclonal rabbit antibodies were raised by Covance (Denver, PA). For all immunoblots, cell lysates or purified proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), transferred to nitrocellulose, and incubated with rabbit polyclonal antibodies to the protein of interest at a 1:1,000 dilution in 3% bovine serum albumin in Tris-buffered saline-Tween (TBST: 25 mM Tris-HCl [pH 7.4], 125 mM NaCl, 0.05% Tween 20 [pH 7.4]). Equal loading was determined by stripping the nitrocellulose membranes with 0.2 N NaOH for 5 min, rinsing, blocking, and incubating the nitrocellulose with polyclonal rabbit antibodies to dihydrolipoamide acyltransferase (DlaT) (35). Horseradish peroxidaseconjugated anti-rabbit antibody (GE-Amersham Biosciences, Inc.) was used for chemiluminescent detection (SuperSignal West Pico; Thermo Fisher Scientific).
For crystallography studies of Rv0078-His 6 , bacteria were grown at 37°C to an optical density at 600 nm (OD 600 ) of 0.5 to 0.6 before being induced with 0.5 mM IPTG (isopropyl-␤-D-thiogalactopyranoside) and incubated at 16°C overnight. After being harvesting by centrifugation, cells were lysed by passing through a microfluidizer cell disrupter in a mixture of 10 mM potassium phosphate (pH 8.0), 10 mM imidazole, and 500 mM NaCl. The homogenate was clarified by spinning at 27,000 ϫ g, and the supernatant was applied to a HiTrap-Ni column (GE Healthcare) preequilibrated with the lysis buffer. Histidine-tagged protein was eluted with a 10 to 300 mM imidazole gradient in 10 mM potassium phosphate (pH 8.0) containing 300 mM NaCl. The Rv0078 fractions were further purified by a Superdex 75 column (16 by 1,000 mm; GE Healthcare) preequilibrated with 20 mM potassium phosphate (pH 8.0) and 300 mM NaCl. The purified Rv0078 was concentrated to 40 mg/ml for the crystallization screen.
RNA-Seq and 5= RACE. Three biological replicate cultures of WT M. tuberculosis were grown to an OD 580 of~1 and incubated in 100 M iP or DMSO (control) for 5 h. Cells were harvested and RNA was purified as described previously (33). Briefly, an equal volume of 4 M guanidinium isothiocyanate, 0.5% sodium N-lauryl sarcosine, and 25 mM trisodium citrate solution was added to cultures to arrest transcription. RNA was isolated with Trizol reagent (Invitrogen) and further purified using RNeasy Miniprep kits and DNase I (Qiagen). Transcriptome profiling by RNA-Seq was performed, and the results were analyzed as follows. RNA from M. tuberculosis cultures was extracted for library construction. Libraries were constructed and bar-coded with the Epicentre ScriptSeq Complete Gold Low Input (Illumina, Inc.) and sequenced on an Illumina HiSeq 2000 sequencer using version 3 reagents. Unique sequence reads were mapped to the corresponding reference genome, and reads per kilobase per million (RPKM) values were calculated in CLC (version 7.0.4). Genes with significantly different RPKM values were identified using the Significant Analysis for Microarray (SAM) statistical analysis component of MeV (36).
5= RACE was performed as described by the manufacturer (Invitrogen). Briefly, 1 g of RNA was used as the template for cDNA production using a reverse primer 150 to 300 bp downstream of annotated translational start sites. A 3= poly(C) tail was added to cDNA by recombinant Tdt. The cDNA was then amplified using a nested reverse primer and a primer that anneals to the poly(C) tail. Products were cloned and sequenced. Likely transcriptional start sites were selected based on the clones that had the most nucleotide sequence upstream of the start codon.
Crystallization and structure determination. DNA-free Rv0078 crystals were obtained by screening at 20°C using the sitting-drop vapor diffusion method. The C2 space group crystals were grown in 0.1 M sodium cacodylate, pH 6.4, and 1.3 M lithium-sulfate. SeMet substituted Rv0078 crystals with C2 space group were grown in a mixture of 0.1 M sodium cacodylate (pH 6.6), 1.3 M lithium sulfate, 0.2 M magnesium sulfate, and 2% polyethylene glycol 400 (PEG 400). Diffraction data to a resolution of 1.85 Å were collected at the Lilly Research Laboratories Collaborative Access Team (LRL-CAT) beamline of Advanced Photon Source (APS), Argonne National Laboratory, and were processed with Mosflm software (37). The program Hybrid-Substructure-Search in the Phenix package was used to locate the Se sites, and the initial phasing was carried out using the program Autosol of Phenix. The 2.7-Å map phased by the SAD method allowed us to build the Rv0078 model unambiguously. The native Rv0078 structure was subsequently determined by the program PHASER using SeMet-substituted Rv0078 as the initial search model. To obtain the Rv0078-DNA crystals, purified Rv0078 was cocrystallized with a 23-mer DNA duplex (5=-TTTACAAGCAGACTGCCGGTAAC-3=) at a molar ratio of 2:1 (protein dimer to DNA) in the presence of 150 mM MgCl 2 . The DNA-bound Rv0078 crystals were grown in the buffer containing only 0.2 M magnesium formate. Diffraction data to 3.0 Å were collected at the Life Sciences Collaborative Access Team (LS-CAT) beamline of APS and were processed with Mosflm. The Rv0078 DNA structure was determined by PHASER using DNA-free structure as the search model. All the refinements were performed using Phenix-refine (38). The statistics are provided in Table 3.