Mycobacterial STAND adenylyl cyclases: The HTH domain binds DNA to form biocrystallized nucleoids

Abstract

Mycobacteria harbor a unique class of adenylyl cyclases with a complex domain organization consisting of an N-terminal putative adenylyl cyclase domain fused to a nucleotide-binding adaptor shared by apoptotic protease-activating factor-1, plant resistance proteins, and CED-4 (NB-ARC) domain, a tetratricopeptide repeat (TPR) domain, and a C-terminal helix-turn-helix (HTH) domain. The products of the rv0891c-rv0890c genes represent a split gene pair, where Rv0891c has sequence similarity to adenylyl cyclases, and Rv0890c harbors the NB-ARC-TPR-HTH domains. Rv0891c had very low adenylyl cyclase activity so it could represent a pseudoenzyme. By analyzing the genomic locus, we could express and purify Rv0890c and find that the NB-ARC domain binds ATP and ADP, but does not hydrolyze these nucleotides. Using systematic evolution of ligands by exponential enrichment (SELEX), we identified DNA sequences that bound to the HTH domain of Rv0890c. Uniquely, the HTH domain could also bind RNA. Atomic force microscopy revealed that binding of Rv0890c to DNA was sequence independent, and binding of adenine nucleotides to the protein induced the formation of higher order structures that may represent biocrystalline nucleoids. This represents the first characterization of this group of proteins and their unusual biochemical properties warrant further studies into their physiological roles in future.

Significance

Unique domain fusions are seen in adenylyl cyclases from mycobacteria. Characterization of a split gene pair reveals that the adenylyl cyclase has very low activity, representing a pseudoenzyme. The second member of the gene pair harbors an NB-ARC domain that binds adenine nucleotides and a novel HTH domain that binds DNA and RNA in a sequence-independent manner. In the presence of adenine nucleotides, protein binding to DNA results in the formation of biocrystallized nucleoids. This family of enzymes, found only in slow-growing mycobacteria, have unusual properties warranting a closer study of their role in these bacteria.

Introduction

The prolonged coexistence of Mycobacterium tuberculosis with humans has resulted in the evolution of several distinctive features in the pathogen that presumably play a role in its complex interaction with the host. These include the increased frequency of genes devoted to lipid metabolism and the expansion of signaling pathways similar to those seen in eukaryotes, such as serine/threonine kinases (1) and adenylyl cyclases. M. tuberculosis encodes for a rich cAMP machinery, consisting of 16 putative adenylyl cyclases, 10 proteins that harbor cyclic nucleotide-binding domains, and a single characterized phosphodiesterase (2,3). The presence of multiple genes encoding for adenylyl cyclases is correlated with the ability of this organism to synthesize and secrete elevated amounts of cAMP, providing a mechanism for the pathogen to hijack signaling within the macrophage (3,4).

The diversity in domain organization and the biochemical properties of adenylyl cyclases argues against redundancy (5) in terms of function. Indeed, deletion of some individual adenylyl cyclase genes in the bacterium did not alter the virulence of the bacterium in macrophages (6, 7, 8). However, deletion of only rv0386 was reported to impair the virulence of M. tuberculosis in mice (8). Rv0386 has a complex, multidomain organization consisting of an adenylyl cyclase, a nucleotide-binding adaptor shared by apoptotic protease-activating factor-1, certain R gene products, and cell death protein-4 (NB-ARC), a tetratricopeptide repeat (TPR), and a tetrahelical LuxR-type helix-turn-helix (HTH) domain (C-A-T-H). This domain organization is shared by two other M. tuberculosis adenylyl cyclases, Rv1358 and Rv2488c, which are as yet uncharacterized (9). The C-A-T-H domain organization also places these proteins in a larger group of signal transduction ATPases with numerous domains (STAND) proteins, an emerging subgroup of AAA + ATPases that exhibit complex signaling paradigms by formation of oligomeric hubs (10,11). Although most STAND proteins have a single effector domain toward the N-terminus of the NB-ARC domain, Rv0386 and its paralogues present an interesting case of two potential effector domains at each terminus, an adenylyl cyclase and an HTH domain. Furthermore, this domain organization, in which a putative adenylyl cyclase domain is present along with an HTH domain in a single protein, is exclusive to this set of mycobacterial proteins and not found in any other organism (9).

The STAND adenylyl cyclases are large proteins and difficult to express and purify (12). Indeed, only the adenylyl cyclase domain of Rv0386 has been characterized to date. To characterize the properties of this group of unusual proteins from mycobacteria, we chose to work with the split gene pair rv0891c-rv0890c in which the gene encoding the adenylyl cyclase is separated from the adjacent A-T-H domains. We describe here the biochemical properties of this split gene pair (rv0891c-rv0890c) and find that although adenylyl cyclase activity is severely compromised, rendering Rv0891c a “pseudoenzyme,” the NB-ARC domain in Rv0890c binds ATP. Moreover, the HTH domain is able to bind DNA and forms biocrystallized nucleoids in the presence of adenine nucleotides. Our study indicates that the roles of these STAND adenylyl cyclases in mycobacteria must be viewed in the context of additional biochemical properties they possess and not solely on their ability to generate cAMP.

Materials and methods

Sequence alignments

Domain boundaries were identified using Pfam 29.0 (13), CDD (14), or Interpro Scan 5 (15). Sequence alignments were generated using Clustal Omega (16). Conserved motifs were identified using existing literature (5,10). All primers used in this study are listed in Table 1.

Table 1.

List of oligonucleotide primers used in the study

Name	Sequence (5′ to 3′)
Rv0891c BamHI fwd	CAGGATCCGTGCTCTT CAACGCAGTTC
Rv0891c T128STOP XbaI fwd	GTGTCGGAAGCTATCT AGAATCACGGTGGGG
Rv0890c TPR EcoR1 fwd	CCGGAATTCGCCACCG ACATCATCGC
Rv0890c TPR XbaI rvs	CCGGTCTAGATTATTCA CCGCGACCGCGTTGTGC
Rv0891c RT fwd	TCTATGTCGGCC CCACTATCAA
Rv0891c RT rvs	CGGCCAGTCACT AAATCACCTG
Rv0890c RT fwd	TGTCCGGAGC TGACGATCTT
Rv0890c RT rvs	GCAAACAATTCG ACTGCTTCG
Rv0891c-Rv0890c junction rvs	GCCAGACGGGTC TTGCCGACACC
Rv0891c T128STOP XbaI rvs	CCCCACCGTGATTCT AGATAGCTTCCGACAC
Rv0891c EcoR1 fwd	CGGAATTCGCTCTTCAAC GCAGTTCATAACAGC
Rv0891c90-285 BamHI fwd	TGCGGATCCCGGTGACGG TGACATTGCTCTTAGC
Rv0891c XhoI rvs	CGGTCTCGAGCCAACA GTGCCCGCACCTCAG
Rv0891c d90 EcoRI fwd	TGCGAATTCCGGTGACGGT GACATTGCTCTTAGC
Rv0891c d81 BamHI fwd	TGCGGATCCCGGTGGAC GTGAGCAGATTGCCGCC
Rv0890c EcoRI fwd	CTGGAATTCGGGCACTGTTG GCGCAGAACC
Rv0890c XbaI rvs	TGCCTCTAGAATCGT CACGAGGGTGAGCC
Rv0891c-Rv0890c ΔSTOP	CGTGCGCAAATAAGCGA GGTGCGGGCACTGTTGG
Rv0890c −60 StuI fwd	TCTAGGCCTATGTCGAT TCTCCCGGCGCAGTTC
T7 rvs	AATACGACTCACTATAGGGCG
Rv0890c R844A StuI fwd	CAAGGACATTGCAAAGGC CTTATTCGTCTCGCCGCGC
Rv0890c R850A NaeI fwd	CTTTTCGTCTCGCCGG CTACTGTGCAAACCCAC
SELEX linker 1(phosphorylated)	GATCCGGAATTCGA TTACATTAATTAAG
SELEX linker 2	CTTAATTAATGTA ATCGAATTCCG
Rv3133c BamH1fwd (DosR)	GTGGATCCATGGTAA AGGTCTTCTTGG
Rv3133c HindIII rev (DosR)	GGCAAGCTTTTGTC ATGGTCCATCACC

RNA isolation and reverse transcription PCR M. tuberculosis

H37Rv cells were grown in Middlebrook’s 7H9 broth containing 0.05% Tween-80 and supplemented with 0.2% glycerol, 0.2% glucose, or 0.2% propionate as indicated and cultured with shaking at 37°C till the exponential phase of growth was reached for each culture (4–8 days; OD 0.3–1.0 depending on the culture medium). Pelleted cells were resuspended in appropriate amounts of TRI reagent (Sigma-Aldrich, St. Louis MO) or RNAiso Plus (Takara Bio, Shiga, Japan). Approximately 800 μL were used per 5 × 10⁹ cells. Cells were lysed by bead beating (0.5-mm glass beads; BioSpec Products, Bartlesville, OK) in the Mini-Beadbeater-16 (BioSpec Products) followed by centrifugation at 16,000 rcf and 4°C for 10 min. The supernatant was mixed with 0.2 vol of chloroform and centrifuged at 16,000 rcf and 4°C for 10 min. The aqueous layer was mixed with an equal volume of isopropanol and centrifuged at 16,000 rcf and 4°C for 10 min. The obtained pellet was washed with 70% ethanol and dried and resuspended in water. A total of 5 μg of RNA was treated with RNase-free DNaseI (2 units). A total of 2 μg of DNase-free RNA was mixed with 200 ng of random hexamers followed by incubation at 70°C for 5 min and snap chilled on ice for 5 min. To this, 500 μM of dNTPs and an appropriate amount of RT buffer (Thermo Fisher Scientific, Waltham, MA) were added to a final volume of 18 μL, and the reaction was split into two equal halves. To one, 200 units of RevertAid reverse transcriptase was added. The tubes were incubated at 25°C for 10 min, followed by incubation at 42°C for 1 h. Enzyme inactivation was brought about by incubation at 70°C for 10 min. The cDNA was subjected to reverse transcription polymerase chain reaction (RT-PCR) using appropriate primers (Rv0890c RT rvs and Rv0890c RT fwd; Rv0891c RT fwd and Rv0891c RT rvs; Rv0891c RT fwd and Rv0891c-Rv0890c junction rvs) for testing the operon (Table 1).

Expression of Rv0891c-Rv0890c in Mycobacterium smegmatis

To generate a clone encompassing the rv0891c-rv0890c operon under the control of the sigA promoter, PCR was performed on M. tuberculosis H37Rv genomic DNA as the template using primers Rv0891c BamHI fwd and Rv0891c XhoI rvs (Table 1). The resulting 893 bp amplicon was digested with BamHI and XhoI and ligated with similarly digested pPROEX-HTb to generate pPRO-Rv0891c_start. The 899-bp-long BamHI-HindIII fragment obtained from pPRO-Rv0891c_start and a 340-bp-long KpnI-BamHI fragment from a plasmid containing the sigA promoter (amplified from genomic DNA) were ligated with the ∼3.8-kb KpnI-HindIII fragment from pMV 10–25 to generate pMV-sigA-Rv0891c. Next, the 3.1-kb NheI-HindIII fragment from pPRO-Rv0891c-Rv0890c (see below) was ligated with the ∼4.5-kb NheI-HindIII fragment from pMV-sigA-Rv0891c to generate pMV-sigA-Rv0891c-Rv0890c.

To generate a clone to investigate translational coupling, PCR was performed using primers Rv0891cBamHI fwd and Rv0891cT128STOP XbaI rvs. The resulting 400-bp amplicon was digested with BamHI and XbaI. PCR was also performed using primers Rv0891c T128STOP XbaI fwd and Rv0891c XhoI rvs, and the 480-bp amplicon was digested with XbaI and XhoI. Digested 400- and 480-bp amplicons were ligated with pBKSII digested with BamHI and XhoI to generate pBKS-Rv0891c T128STOP. The 3.1-kb NheI-HindIII fragment from pPRO-Rv0891c-Rv0890c (see below) were ligated with the 4.3-kb BamHI-HindIII fragment from pMV-sigA-Rv0891c-Rv0890c to generate pMV-sigA-Rv0891c-Rv0890c T128STOP.

Plasmids pMV-sigA-Rv0891c-Rv0890c and pMV-sigA-Rv0891c-Rv0890c T128STOP were electroporated into electrocompetent M. smegmatis cells and transformed cells were selected using 50 μg/mL hygromycin on Middlebrook’s 7H10 agar plates supplemented with 0.5% glycerol. Individual colonies were grown in 5 mL of Middlebrook’s 7H9 broth supplemented with 0.2% glycerol and 0.05% Tween-80 at 37°C till an OD600 of 1. Cells were lysed by bead beating (0.5-mm glass beads; BioSpec Products) in Mini-Beadbeater-16 (BioSpec Products) in 50 mM Tris-Cl (pH 8.0) at 4°C, 100 mM NaCl, 5 mM β-mercaptoethanol, 10% glycerol, 1 mM benzamidine, and 1 mM PMSF, and the protein concentration in the lysate was quantitated by Bradford’s method. A total of 50 μg of lysate were analyzed by Western blotting using anti-Rv0891c- and anti-Rv0890c-specific antibodies (see below).

Cloning of Rv0891c and Rv0890c

The Rv0891c gene was cloned using primers Rv0891c EcoRI fwd and Rv0891c XhoI rvs. PCR was carried out on the genomic DNA of M. tuberculosis H37Rv, and the product was digested with EcoRI and XhoI and cloned into similarly digested pPROEX-HTc to generate pPRO-Rv0891c.

Rv0891c_CHD was cloned using primers Rv0891c90-285 BamHI fwd and Rv0891c XhoI rvs. PCR was carried out on pPRO-Rv0891c, and the product was digested with BamHI and XhoI and cloned into similarly digested pPROEX-HTc vector to generate pPRO-Rv0891c_CHD.

Rv0891c_Δ81 (representing a protein with an additional nine amino acids at the N-terminus) was cloned using primers Rv0891c d81 BamHI fwd and Rv0891c XhoI rvs. PCR was carried out using plasmid pPRO-Rv0891c as template, and the product obtained was digested with BamHI and XhoI and cloned into similarly digested pPROEX-HTc vector to generate pPRO-Rv0891c_Δ81.

Rv0890c was cloned using primers Rv0890c EcoRI fwd and Rv0890c XbaI rvs. PCR was carried out on the genomic DNA of M. tuberculosis H37Rv, and the product was digested with EcoRI and XbaI and cloned into similarly digested pPROEX-HTb to generate pPRO-Rv0890c. Note that this and all clones containing Rv0890c harbor an Ala at position 866 instead of a Pro residue, as reported in the published H37Rv genome sequence. Inspection of a number of M. tuberculosis strains sequenced to date all show an Ala at position 866 (17). It is therefore possible that the strain used for obtaining the published genome sequence harbored a single nucleotide polymorphism that converted the amino acid to a Pro residue at this position or it represented a sequencing artifact.

Rv0891c-Rv0890c were cloned using Rv0891c EcoRI fwd and Rv0890c XbaI rvs. PCR was carried out on the genomic DNA of M. tuberculosis H37Rv, and the product was digested with EcoRI and XbaI and cloned into similarly digested pPROEX-HTc to generate pPRO-Rv0891c-Rv0890c.

Rv0890c₊₂₀ was generated by first performing deletion mutagenesis (18) using a PstI-BamHI fragment of Rv0891c-Rv0890c cloned in pBKSII as the template and Rv0891c-Rv0890c ΔSTOP as the primer to generate pBKS-Rv0891c-90c junction Δstop. The deletion of T856 would result in incorporation of a Glu residue (Fig. 7 A). This plasmid was used as template for PCR amplification with Rv0890c −60 StuI fwd and T7 rvs. The product was digested with BamHI and ligated with SmaI-BamHI-digested pBKSII vector to generate pBKS-Rv0890c₊₂₀ junction. The EcoRI-BamHI fragment from pBKS-Rv0890c₊₂₀ junction and the BamHI-HindIII fragment from pPRO-Rv0891c-Rv0890c were ligated with EcoRI-HindIII-digested pPROEX-HTa to generate pPRO-Rv0890c₊₂₀. This would generate a protein that encompasses residues S268–S285 of Rv0891c, followed by a Glu and then entire Rv0890c.

Figure 7.

(A) Volume of Rv0890c₊₂₀ alone (40 nM in a total volume of 20 μL) or in the presence of 5 mM MgATP, MgADP, and MgAMP-PNP as calculated from AFM images. Data points represent individual protein molecules seen with three independent protein preparations. Shown are sizes predicted for the monomer, dimer, and hexamer as estimated from (19). Black horizontal bar represents the mean volume. (B) Supercoiled DNA on mica surface (left) and DNA bound to Rv0890c₊₂₀ (right) (n > 3). Arrows indicate protein bound to DNA. (C) Conformation of DNA-protein complex in presence of 40 nM Rv0890c₊₂₀ with 40 nM DNA in presence of 5 mM MgATP (left) and 5 mM MgADP (right) (n > 3). (D) Conformation of DNA-protein complex in presence of 40 nM Rv0890c₊₂₀ WA with 40 nM DNA in presence of 5 mM MgATP (left) and 5 mM MgAMP-PNP (right) (n = 3). (E) Conformation of DNA-protein complex using pGEMT vector alone (40 nM) and Rv0890c₊₂₀ (40 nM) in presence and absence of 5 mM MgATP. All images were processed using Park XEI software, and data were analyzed using GraphPad Prism 8. The brightness palette at the bottom of the Figure indicates elevation of the sample from the surface for (B)–(E). Data are representative of experiments performed using three independent protein preparations.

For cloning Rv0891c-Rv0890c_FUS, Rv0891c was first recloned to obtain compatible sites. PCR was performed using pPRO-Rv0891c_CHD as template and primers Rv0891c d90 EcoRI fwd and Rv0891c XhoI rvs. The product was digested with EcoRI and XhoI and ligated into similarly digested pBKSII to generate pBKS-Rv0891c_CHD. Next, the EcoRI-PstI fragment from pBKS-Rv0891c_CHD was ligated with similarly digested pBKS-Rv0891c-90c junction Δstop. The EcoRI-BamHI fragment from the resulting plasmid and the BamHI-HindIII fragment from pPRO-Rv0891c-Rv0890c were ligated with pPROEX-HTb vector digested with EcoRI and HindIII, thus generating pPRO-Rv0891c-Rv0890c_FUS.

The TPR-HTH domain of Rv0890c was cloned using primers Rv0890c TPR EcoRI fwd and Rv0890c XbaI rvs. PCR was performed on genomic DNA of M. tuberculosis H37Rv, and the product was digested with EcoRI and XbaI and cloned into similarly digested pBKSII to generate pBKS-Rv0890c_TPR-HTH. The EcoRI-XbaI fragment from pBKS-Rv0890c_TPR-HTH was ligated with similarly digested pPROEX-HTa to obtain pPRO-Rv0890c_TPR-HTH.

The TPR domain of Rv0890c was PCR amplified using primers Rv0890c TPR EcoRI fwd and Rv0890c TPR XbaI rvs. PCR was performed on pBKS-Rv0890c_TPR-HTH, and the product was digested with EcoRI and XbaI and ligated with similarly digested pPROEX-HTa to obtain pPRO-Rv0890c_TPR.

For cloning Rv0891c-Rv0890c_FUS ΔHTH, the EcoRI-AgeI fragment from pPRO-Rv0891c-Rv0890c_FUS and the AgeI-HindIII fragment from pPRO-Rv0890c_TPR were ligated with pPROEX-HTb vector digested with EcoRI-HindIII to generate pPRO-Rv0891c-Rv0890c_FUS ΔHTH.

The region coding for DosR was PCR amplified using primers Rv3133c BamHI fwd and Rv3133c HindIII rvs on Mycobaterium bovis BCG genomic DNA as template. The amplicon was digested with BamHI and HindIII and ligated with similarly digested pPROEX-HTb to obtain pPRO-DosR.

Site-directed mutagenesis (18) was performed on pPRO-Rv0890c_TPR-HTH as template using primers Rv0890c R844A StuI fwd and Rv0890c R850A NaeI fwd to generate the respective single mutants.

All clones and mutants were confirmed by sequencing (Macrogen, Seoul, South Korea).

Generation of antibodies to Rv0891c and Rv0890c and Western blot analysis

Recombinant His-tagged Rv0891c_CHD and a fragment of Rv0890c (Rv0890c gene was digested with Nco1 followed by relegation; this protein contains regions of the NB-ARC and the TPR-HTH domains) were purified by Ni-NTA affinity chromatography followed by size-exclusion chromatography and used as an immunogen. The primary dose (400 μg of protein) was administered to rabbits in Freund’s complete adjuvant subcutaneously and booster doses were administered after 14, 28, and 42 days. Sera were collected on day 56.

Protein samples were electrophoresed on an SDS-polyacrylamide gel of appropriate percentage and transferred to PVDF membranes in transfer buffer (25 mM Tris, 192 mM glycine, 20% methanol (pH 8.3)) for 2 h at a constant current of 200 mA. Before probing with the appropriate antibody, membranes were blocked with 2.5% blocking agent (RP2109; GE Healthcare, Chalfont Saint Giles, UK) in TBS-T buffer (10 mM Tris-Cl (pH 7.5), 0.9% NaCl, and 0.1% Tween-20) for 1 h at room temperature. Blots were probed with primary antibody in TBS-T containing 0.2% BSA for 8–12 h at 4°C with gentle agitation. Membranes were washed thrice with TBS-T and incubated with anti-rabbit IgG conjugated to horseradish peroxidase (1:50,000; GE Healthcare) for 1 h at room temperature. Membranes were washed thrice with TBS-T, and bound antibody was detected using enhanced chemiluminescence with Luminata Crescendo (MilliporeSigma, Burlington, MA).

Expression and purification of proteins

For Rv0891c_CHD, Rv0891c-Rv0890c_FUS, Rv0891c-Rv0890c_FUSΔHTH, and Rv0890c₊₂₀, proteins were expressed in Escherichia coli BL21(DE3) pLysS endo⁻ strain following addition of 500 μM IPTG at 16°C for 16 h. Cells were lysed by sonication in lysis buffer (50 mM Tris-Cl (pH 8.0), 5 mM 2-mercaptoethanol (2-ME), 100 mM NaCl, 10% glycerol). A total of 1 mM benzamidine and 2 mM phenylmethylsulfonyl fluoride were added before sonication, followed by centrifugation at 30,000 × g. The supernatant was interacted with nickel-nitrilotriacetic acid beads (GE Healthcare) and bound protein was washed with wash buffer (50 mM Tris-Cl (pH 8.0), 5 mM 2-ME, 500 mM NaCl, 10% glycerol, 20 mM imidazole). Protein was eluted in lysis buffer supplemented with 300 mM imidazole and dialyzed into lysis buffer. For kinetic analysis of Rv0891c_CHD, the protein was dialyzed into buffer with 10 mM NaCl because this was found to potentiate activity. Aliquots of protein were stored at −70°C.

For purification of Rv0890c_TPR-HTH, the process was identical except 50 mM NaCl was maintained in all buffers during purification. Under these conditions, the protein had prebound RNA. Hence, the purification was altered to include 500 mM NaCl in the wash buffer while maintaining 50 mM NaCl elsewhere. This abrogated copurification with RNA. Hence, all other constructs of TPR-HTH domains and the DosR protein were purified in this manner.

Size-exclusion chromatography analyses were performed at a flow rate of 0.25 mL/min using a Superose 12 10/300 prepacked column (GE Healthcare) on an AKTA FPLC system (GE Healthcare). The column was equilibrated in buffer (50 mM Tris-Cl (pH 8.0) at 4°C, 50 mM NaCl, 5 mM β-mercaptoethanol, containing 10% glycerol) for three column volumes before loading the protein. The column was calibrated using commercially available gel filtration standards (Bio-Rad Laboratories, Hercules, CA) containing bovine thyroglobulin (670 kDa), bovine γ-globulin (158 kDa), chicken ovalbumin (44 kDa), horse myoglobin (17 kDa), and Vitamin B₁₂ (1.35 kDa).

Adenylyl cyclase assays

Adenylyl cyclase assays were performed in final volume of 50 μL. For comparisons between only cyclase domains and full-length proteins, Rv0891c_CHD (2 μg) or Rv0891c-Rv0890c_FUS (4 μg) were incubated in 50 mM Tris-Cl (pH 8.0), 100 mM NaCl, 5 mM 2-ME, 10% glycerol, 1 mM MnATP and 10 mM free Mn²⁺ at 37°C for 30 min. For kinetic analysis of Rv0891c_CHD (5 μg), increasing concentrations of MnATP were used with the free Mn²⁺ concentration constant at 10 mM, in 50 mM Tris-Cl pH 8.0, 10 mM NaCl, 5 mM 2-ME, 10% glycerol at 37°C for 15 min. The adenylyl cyclase reaction was initiated by addition of 1 mM MnATP and the proteins were incubated at 15°C for 3 h. All reactions were terminated by the addition of 450 μL of cold sodium acetate buffer (pH 4.75) and boiling of the samples for 10 min. Suitable volumes were used for cAMP estimation by radioimmunoassay.

ATPase assays

For ATPase assays, Rv0890c₊₂₀ were incubated in 50 mM Tris-Cl (pH 8.0), 100 mM NaCl, 5 mM 2-ME, 10% glycerol, 1 mM ATP, 10 mM MgCl₂, and ∼0.2 μCi of [γ-³²P]-ATP at 37°C for 30 min in a final volume of 20 μL. Approximately 3–4 μL of the reaction was spotted on PEI-cellulose plates (Merck, Darmstadt, Germany) and ran for 15 cm in a presaturated TLC chamber containing a solvent system of 0.5 M LiCl and 0.5 M formic acid. Plates were allowed to dry and then scanned using a phosphoimager (Typhoon FLA 9500; GE Healthcare).

Genomic systemic evolution of ligands by exponential enrichment M. bovis

BCG genomic DNA was digested with Sau3AI and ligated to systemic evolution of ligands by exponential enrichment (SELEX) linkers 1 and 2. For the first attempt of genomic SELEX, 4 μg of purified DNA library was precleared with Ni-NTA resin for 1 h at 4°C in SELEX buffer (50 mM Tris-Cl (pH 8.0) at 4°C, 50 mM NaCl, 5 mM 2-ME, 10% glycerol, 0.05% NP-40, 5 mM MgCl₂, 100 μM EDTA, 20 μg/mL poly dI-dC). Precleared DNA library was interacted with 2 μg of His-tagged Rv0890c_TPR-HTH in a final volume of 50 μL for 1 h at 4°C and protein-DNA complexes were pulled down by interaction with Ni-NTA resin. Beads were washed, and bound DNA was eluted by boiling in TE (10 mM Tris-Cl pH 8.0, 1 mM EDTA). The eluate was used as template for PCR with SELEX linker 2 and the resulting products were used as input for the next panning. In the second attempt at SELEX, 300 ng of protein was used at each panning, poly dI-dC was eliminated from the buffer, and the library was precleared using Rv0890c_TPR before interaction with Rv0890c_TPR-HTH. Three rounds of panning were performed in each attempt, and the final products were cloned into the TA vector (pGEMT) and sequenced. A clone pGEMT-S41 (which contained a fragment from 281,342 to 281,835 in M. bovis BCG genome) was used in further experiments because it exhibited a single shift with complete depletion of free DNA in electrophoretic mobility shift assays (EMSAs) (see below).

Sequences obtained from SELEX (Table S1) were analyzed in MEME (20) using all combinations of parameters specifying the frequency of occurrence of motifs in each sequence and the requirement for the motif to be palindromic. As a control, three to five sets of random sequences with identical size distributions were extracted from the M. tuberculosis H37Rv genome.

EMSA

A total of 500 ng of the indicated SELEX DNA fragment was prepared using PCR with SELEX linker 2 as primer and the corresponding SELEX fragment cloned into pGEMT vector as template. For radioactive EMSAs, double-stranded DNA was end-labeled using [γ-³²P]-ATP with T₄ PNK according to manufacturer’s instructions (Thermo Fisher Scientific). The labeled DNA was purified from the unincorporated nucleotide by passing it through a Sephadex G-25 column equilibrated with sterile MilliQ water.

For EMSA, 40 pmol of proteins were used except where indicated otherwise. Protein was incubated with 500 ng of DNA (for nonradioactive EMSAs) or 10⁵ cpm of [ γ-³²P]-labeled DNA (for radioactive EMSAs) in EMSA buffer (50 mM Tris-Cl (pH 8.0), 50 mM NaCl, 5 mM 2-ME, 10% glycerol, 5 mM MgCl₂, 0.05% NP-40, 100 μM EDTA, and 20 μg/mL salmon sperm DNA) for 2 h at 4°C in a total volume of 20 μL. The complexes were resolved on 5% polyacrylamide gels (29:1 acrylamide/bisacrylamide) containing TAE (40 mM Tris-acetate, 1 mM EDTA (pH 8.2)) and 5% glycerol and run in TAE at 4°C for 6–8 h at 50 V. Postrun, nonradioactive gels were stained with EtBr and visualized under 302 nm. Radioactive gels were dried and scanned using a phosphoimager (Typhoon FLA 9500; GE Healthcare).

The affinity of Rv0890c_TPR-HTH was estimated by densitometric scanning of the unbound DNA and subtracting that from the total DNA used in the reaction, to calculate the fraction bound.

For competition with unlabeled nucleic acids, S41 was used as double-stranded DNA and total RNA from M. bovis BCG as RNA (isolated as described previously (21)). A total of 13.2 μg of each (or indicated amounts of salmon sperm DNA) were preincubated with protein in EMSA buffer before addition of [ γ-³²P]-labeled DNA.

RNA pull-down

For RNA pull-down experiments, 30 μg of total M. smegmatis RNA was precleared on Ni-NTA resin at 4°C for 1 h in 50 mM Tris-Cl (pH 8.0), 50 mM NaCl, 5 mM 2-ME, and 10% glycerol. Precleared RNA was incubated with 660 pmol of protein at 4°C for 2 h (final volume 100 μL) followed by interaction with Ni-NTA resin at 4°C for 1 h. Beads and supernatant were separated, and beads were washed thrice with the interaction buffer. RNA was extracted by treatment with 2% SDS and 1 unit of proteinase K (Thermo Fisher Scientific) at 37°C for 1 h. Extracted RNA was treated with phenol/chloroform/isoamyl alcohol (25:24:1) and precipitated using 0.1 vol of 3 M sodium acetate (pH 5.2) and 2.5 vol of 100% ethanol. A total of 1/20th of the load and breakthrough samples and half of the pull-down samples were analyzed by denaturing agarose gel electrophoresis on a 1% agarose/formaldehyde gel.

Thermal shift assay

Proteins were purified in buffer containing 50 mM HEPES (pH 8.0) at 4°C, 150 mM NaCl, 10 mM β-mercaptoethanol, and 20% glycerol. The thermal shift assays were performed in HEPES buffer (50 mM (pH = 8.0) at 4°C) containing 150 mM NaCl, 20% glycerol, and 10 mM MgCl₂ in StepOnePlus real-time PCR machine (Applied Biosystems, Foster City, CA) using 1X SYPRO Orange dye (Sigma-Aldrich). The samples were heated from 19 to 95°C with an increase of 1°C per step and a hold of 2 min at initial and final temperature. The data from 37 to 55°C were taken for analysis. The initial temperature of the block of the PCR machine varied from 25 to 30°C. This resulted in fluctuations in intensities at initial temperature points. Hence, the data were always acquired from lower temperatures than that used for analysis. Protein concentration used in the assay was 4 μM. ATP or ADP was added with 10 mM MgCl₂ to achieve 5 mM MgATP or MgADP in the specified reaction conditions (calculated using MaxChelator).

Atomic force microscopy

Fresh DNA (SELEX sequences cloned into pGEMT vector) was isolated from E. coli DH10B cells using Unipro plasmid miniprep kit (Dr. KPC Life Sciences, West Bengal, India) and eluted in water. The DNA quality was checked by agarose gel electrophoresis and the band representing supercoiled DNA was eluted from the gel and quantitated. The reaction (total volume of 20 μL) was set up in ice with the required concentration of DNA and protein in atomic force microscopy (AFM) buffer containing 50 mM Tris-Cl (pH 8 at 4°C), 50 mM NaCl, 10% glycerol, 5 mM MgCl₂, and 5 mM β-mercaptoethanol and incubated at 4°C for 2 h. ATP, ADP, or AMP-PNP (5 mM) were added along with 10 mM MgCl₂.

Muscovite mica (V1 quality from Electron Microscopy Sciences, Hatfield, PA) was freshly cleaved using Scotch Tape. After incubation, the total reaction mix was deposited on the freshly cleaved mica and allowed to dry for 30 min. The mica surface was washed thoroughly with water and left to dry in a desiccator. Samples were then taken for imaging. All reagents used were autoclaved and filtered with a 0.22-μm filter to avoid debris. Imaging of samples on mica substrate was done via the noncontact mode in Park Systems XEI Atomic Force Microscope using 40 N/m ACTA cantilevers from Park Systems (Suwon, South Korea). From each sample, a minimum of six images were taken representing different parts of the mica surface. Images were acquired at 500 nm × 500 nm, 2 μm × 2 μm, or 5 μm × 5 μm according to the size of the DNA or DNA-protein complex being visualized. Image analysis was done using XEI software from Park Systems. Sizes of protein particles were calculated based on (19).

Statistical analysis

Statistical analysis was performed using Student’s t-test in GraphPad Prism 5.

Results

The rv0891c-rv0890c split gene pair transcript is expressed as an operon in M. tuberculosis

We began by confirming that rv0891c and rv0890c are expressed in M. tuberculosis. Scrutiny of transcription start sites (Fig. 1 A) as identified in a high-throughput analysis (22) indicated the presence of a site that would direct production of a messenger-RNA (mRNA)-encoding Rv0891c (marked as “a” in Fig. 1 A). Internal transcription start sites in rv0891c (“b” and “c”) were also identified, and an RNA transcript beginning at “c” could encode the entire coding sequence of Rv0890c. Because no transcription start site was identified close to the start site of rv0890c, and a number of mRNAs in M. tuberculosis are transcribed as leaderless mRNAs (22), it is possible that these two genes could represent an operon and be transcribed as a single RNA. RT-PCR using primers specific for the two genes indicated that these genes were expressed in M. tuberculosis grown in different media (Fig. 1 B), with glucose and propionate representing carbon sources that the bacterium is thought to experience within the host macrophage and granuloma (23,24). Moreover, a transcript was also detected using primers that would amplify an mRNA produced from an operon encoding rv0891c-rv0890c (Fig. 1 C).

Figure 1.

(A) A snapshot of the genomic region of rv0891c-rv0890c (shown as mustard bars) in M. tuberculosis, indicating the ribosome-binding sites detected by ribosome profiling (enrichment shown as red peaks) and transcription start sites (black arrows, direction indicating the direction of transcription). Marked are predicted amino acid sequences that could represent start sites for the protein (VSRL and/or VTLL) in the vicinity of a ribosome-enriched sequence of the mRNA. Also annotated are three transcription start sites that are found upstream (a) and within the rv0891c-coding sequence (b, c) and discussed in the text. The image is obtained from the web server https://mtb.wadsworth.org/. (B) PCR products obtained from cDNA prepared from RNA isolated from cultures of M. tuberculosis grown in the presence of different carbon sources. Primers used for PCR are contained within the coding sequence of the two individual genes. The plus and minus signs represent samples of RNA taken for PCR with or without reverse transcription being performed to monitor the extent of genomic DNA contamination that could generate a PCR product. Data shown are representative of experiments repeated in two independent experiments. (C) Primers spanning the two genes, as indicated by arrows in the schematic, were used for PCR with cDNA prepared from RNA obtained from glycerol grown cultures. The plus and minus signs represent samples of RNA taken for PCR with or without reverse transcription being performed, respectively. A PCR product of ∼1 kb would be produced from a single mRNA produced by the putative operon. (D) Schematic showing the domain organization of the STAND adenylyl cyclases. Vertical line indicates the split gene organization of Rv0891c and Rv0890c. Domain representations are not to scale. (E) Lysates were prepared from M. smegmatis cells harboring plasmids where the rv0891c-rv0890c operon (either wild-type or incorporating a stop codon after T128 in Rv0891c (indicated as a red star)) was transcribed under the control of the sigA promoter. Lysates were subjected to Western blot analysis and probed with antibodies raised to Rv0891c or Rv0890c as indicated. A Coomassie-stained gel of lysates taken for Western blotting is shown to normalize protein loading across samples. Arrow heads on the right indicate expression of Rv0891c (∼25 kDa; smaller than the size predicted from the annotated gene) and Rv0890c (∼90 kDa). Data shown are a representative blot from experiments repeated at least thrice.

Examination of the genomic locus encoding for Rv0891c revealed that rv0891c is separated by a single basepair from the downstream gene, rv0890c (Fig. 1 D; see also Fig. 8 A). Interestingly, rv0890c encodes for a protein containing NB-ARC, TPR, and HTH domains, making this a “split gene” pair (rv0891c-rv0890c) in comparison with the other STAND adenylyl cyclases, namely Rv0386, Rv1358, and Rv2488c (Fig. 1 D). Because rv0890c and rv0891c are expressed as a single mRNA, we asked if they could also be translationally coupled, whereby translation termination and reinitiation occurs at the junction of the two genes (25). We cloned the rv0891c-rv0890c operon under the strong sigA promoter (Fig. 1 E, left panel) and monitored expression of the two proteins in M. smegmatis, which does not harbor homologs of these two genes. As shown in Fig. 1 E right panel, a band of ∼25 kDa was detected specifically using antibodies raised to Rv0891c. This is a size smaller than the molecular weight of 31 kDa predicted from the annotated ORF. Ribosome profiling in M. tuberculosis (https://doi.org/10.1101/665208 and https://mtb.wadsworth.org/) does suggest that translation could initiate from transcripts originating from the transcription start sites “b” (Fig. 1 A) found downstream of the annotated start codon. Either of the two translated proteins from this transcript would encode shorter proteins of molecular weight <31 kDa.

Figure 8.

(A) Sequence alignment of genomic regions in M. tuberculosis and M. canetti, demonstrating the rationale for generation of Rv0891c-Rv0890c_FUS. (B) Thermal shift assay with Rv0891c- Rv0890c_FUS (4 μM) in the absence or presence of 5 mM MgATP. Data were analyzed and is representative of assays repeated thrice. Inset shows the purified protein used for assays. (C) Comparison of DNA-protein complexes formed by Rv0890c₊₂₀ (left) and Rv0891c-Rv0890c_FUS (right) in presence of 5 mM MgATP with 40 nM of DNA and protein (n = 2). All images were processed using Park XEI software. The brightness palette indicates elevation of the sample from the surface. (D) Indicated proteins (660 pmol) were interacted with 30 μg of total M. smegmatis RNA (load). Reaction mixtures were interacted with Ni-NTA agarose and RNA was extracted from the bound (pull down) and unbound (breakthrough) samples. 1/20th of the input and breakthrough samples and half of the pull-down samples were analyzed by denaturing agarose gel electrophoresis on a 1% agarose/formaldehyde gel and stained with EtBr. Data are representative of experiments performed using two independent protein preparations.

A band of ∼90 kDa was seen in cell lysates using antibodies to Rv0890c, indicating expression of this protein from the transcript expressed under the SigA promoter. The molecular weight predicted of the annotated Rv0890c is 94.5 kDa. Ribosome profiling (26) has identified ribosome binding sites at sequences near the start of the Rv0890c ORF (Fig. 1 A). Therefore, to monitor the extent of translational coupling in the expression of the two proteins encoded in the operon, we introduced a stop codon after T128 in Rv0891c, ∼700 bp upstream of the start codon of rv0890c, assuming that this would reduce translational coupling and/or ribosome skipping at the junction of the two genes. Although Western blot analysis revealed a complete loss of expression of Rv0891c, a band corresponding to Rv0890c could still be detected. We therefore conclude that translation initiation of Rv0890c within a single operonic mRNA can occur independently of translation of Rv0891c, which is in agreement with ribosome profiling data (Fig. 1 A).

The adenylyl cyclase domain of Rv0891c is a “pseudoenzyme”

A multiple-sequence alignment of Rv0891c with the adenylyl cyclase domains (CHD) of known mycobacterial adenylyl cyclases is shown in Fig. 2 A. The protein has a 90-amino-acid-long N-terminal extension preceding the adenylyl cyclase domain (residues colored in blue, Fig. 2 A). This extension did not bear significant similarity to known protein sequences, as inferred from a pBLAST analysis (27). Recombinant expression of Rv0891c as annotated in the genome resulted in its localization to inclusion bodies in E. coli (data not shown).

Figure 2.

(A) Multiple-sequence alignment of the adenylyl cyclase domains (residues colored in red) of Rv1900c, Rv1647, and Rv1625c with Rv0891c and Rv0386. The asterisk denotes metal-binding residues, arrows denote substrate-specifying residues, and inverted triangles denote transition-state-stabilizing residues. Residues colored in blue constitute the N-terminal extension in Rv0891c. Also indicated are the putative start codons represented by Valine 81 and Valine 91 (red circles). (B) Purified proteins used for adenylyl cyclase assays are shown. (C) Cyclic AMP production by Rv0891c_Δ81 and Rv0891c_CHD in the presence of 10 mM Mn²⁺ and increasing concentrations of MnATP. Values are mean ± SD of duplicate determinations from independent protein preparations.

As mentioned above, the nucleotide sequence encoding this N-terminal extension contains an internal transcription start site (“b” in Fig. 1 A) upstream of two putative start codons (GTG; represented by Valine 82 and Valine 91; red circles in Fig. 2 A) close to the start of residues that show similarity to the adenylyl cyclase domain of Rv0891c_. We expressed truncated proteins lacking either 81 residues from the N-terminus, (Rv0891c_Δ81) or 90 residues (Rv0891c_CHD, which represents the shortest protein encompassing the cyclase homology domain) in E. coli and were able to obtain substantial amounts of both proteins (Fig. 2 B, insets).

Although Rv0891c had conserved residues (indicated by asterisks in Fig. 2 A) for metal-binding (D97 and D142) and transition-state stabilization (open arrow heads, N193 and R197) seen in nucleotidyl cyclases (28), the protein lacked canonical residues for ATP recognition (depicted as vertical arrows), substituting an R138 and L186 in place of K and D seen usually in adenylyl cyclases. As revealed in the structure of the mycobacterial adenylyl cyclase Rv1900c (29), noncanonical substrate-specifying residues (N342 and D395; Fig. 2 A) are placed distant from the adenine ring and no base-specific interactions are seen at the active site. This results in the formation of an asymmetric active site, reminiscent of the mammalian adenylyl cyclases (29). We have previously reported that mutations in these ATP-binding residues abolishes dimerization of Rv1625c, another mycobacterial adenylyl cyclase (30,31), even though Rv1900c exists as a dimer in the absence of canonical substrate-specifying residues (29). Both Rv0891c_Δ81 and Rv0891c_CHD exhibited low V_max (0.18 mmol cAMP/min/mol) (Fig. 2 B) compared to ∼5 mol cAMP/min/mol of Rv1625c (30), and activity was detectable only because of the very sensitive assay used to measure cAMP (12,29,30,32). Adenylyl cyclase activity of the purified proteins displayed allostery kinetics with a K' for MnATP (480 μM) that was similar to other mycobacterial adenylyl cyclases (Fig. 2 B). Because the presence of an additional nine amino acids at the N-terminus did not increase catalytic activity substantially, we proceed to work with Rv0891c_CHD because this would represent the smallest protein that would harbor the catalytic domain.

Rv0891c_CHD was largely monomeric as determined by gel filtration and unable to convert GTP to cGMP, nor was activity increased in the presence of Mg²⁺ as the metal ion or in the presence of bicarbonate or detergents (data not shown). Thus, Rv0891c is a poorly active adenylyl cyclase under the conditions tested, possibly because of a reduced capacity to form stable and catalytically competent dimers.

ATP/ADP binding to the NB-ARC domain of Rv0890c

We next examined the properties of the second member of the split gene pair, Rv0890c, beginning with its N-terminal NB-ARC domain (Fig. 1 D). The NB-ARC domain of Rv0890c (Fig. 3 A) lacked a catalytic glutamate in the Walker B (the hhhhDD/E motif) and a conserved histidine in the MHD motif. These residues are essential for ATPase activity, suggesting that the protein may not hydrolyze ATP (33, 34, 35). However, the Walker A motif required for ATP binding was conserved (34). Rv0890c was cloned and expressed in E. coli, but the protein localized to inclusion bodies (Fig. 3 B).

Figure 3.

(A) Multiple-sequence alignment of the NB-ARC domains of the indicated proteins with predicted secondary structure for Rv0386. Blocks depict α-helices, and arrows depict β-strands. Purple, blue, and green colors in the secondary structure elements denote the nucleotide binding, helical domain 1, and winged-helix subdomains, respectively. Colored in red are residues from Rv0891c. Highlighted with orange boxes and text are the conserved motifs (x, any amino acid; h, hydrophobic amino acid; o, amino acid with alcoholic side chain). (B) Coomassie-stained SDS-polyacrylamide gel showing protein obtained from the purification of Rv0890c (left) and Rv0890c₊₂₀ (right). (C) [γ-³²P]-ATP was incubated alone (−) or in presence of Rv0890c₊₂₀, and the resulting products were separated by thin-layer chromatography. Calf intestinal alkaline phosphatase (CIAP) was used to identify the mobility of ATP and P_i.. (D) Thermal shift assay with 4 μM Rv0890c₊₂₀ or Rv0890c₊₂₀ WA in the absence or presence of 5 mM MgATP/MgADP (n = 3). Data are representative of experiments performed using three independent protein preparations. Error bars represent the mean ± SD of three experiments.

Based on the interrupted nature of the rv0891c-rv0890c gene pair, we hypothesized that a region of Rv0891c may be required to assist folding of Rv0890c to generate a soluble protein that would represent the domain seen in full-length homologs like Rv0386. The split between Rv0891c and Rv0890c occurs within a predicted α-helix (predicted using PSIPRED; Fig. 3 A; (36)), and the final 20 residues of Rv0891c include the motif hhGRExE, thought to be important for signal propagation across STAND proteins (10). Thus, we cloned Rv0890c with these additional 20 residues of Rv0891c fused at the N-terminus. This protein, Rv0890c₊₂₀, was expressed to high levels and could be purified efficiently (Fig. 3 B).

As predicted from the absence of important residues in the Walker B motif, the protein was unable to hydrolyze ATP (Fig. 3 C). We then asked if this domain could bind ATP. We used a fluorescence-based thermal shift assay, in which binding of a hydrophobic fluorescent dye increases as a protein unfolds, and the temperature at which this unfolding occurs could be altered in the presence of small-molecule ligands. We incubated Rv0890c₊₂₀ with MgATP and performed thermal shift assays. The melting profile of Rv0890c₊₂₀ in the absence of MgATP was broad, with initial higher intensity of SYPRO Orange fluorescence indicating poorly folded protein. In the presence of MgATP, the profile was biphasic and a new cooperatively melting species with a higher melting temperature was present. Similar shifts were observed in the presence of MgADP but not in the presence of AMP (Fig. 3 D; data not shown).

The Walker A motif (GxxGxGKT/S) has a conserved lysine that is involved in binding to the phosphates in nucleotides, and mutation of this lysine residue abrogates nucleotide binding. The melting profile of the Rv0890c_+20K80A (WA) mutant protein was not altered in the presence of the MgATP (Fig. 3 D), indicating that the shift observed in the melting profile of Rv0890c₊₂₀ in the presence of MgATP resulted from ATP binding to the Walker A motif of the NB-ARC domain.

The HTH domain of Rv0890c binds DNA

STAND proteins are known to engage in intramolecular interactions involving the NB-ARC domain with downstream regions (10,11), which, in the case of Rv0890c, is a C-terminal tetrahelical LuxR-type HTH domain (Fig. 1 D). To identify DNA sequences with which the HTH domain could interact, we purified the Rv0890c_TPR-HTH protein (Fig. 4 A, inset). The protein was found to elute on gel filtration as monomeric and dimeric species, and a construct without the HTH domain (Rv0890c_TPR; Fig. 4 A) showed similar oligomeric states. Using the Rv0890c_TPR-HTH protein, we performed a genomic SELEX using a library of Sau3AI-digested fragments from M. bovis BCG genomic DNA. We identified 46 independent sequences that bound to the protein that lay in both intergenic as well as intragenic regions of the genome (Table S1). These sequences did not share a common sequence motif as analyzed by MEME (20) (http://meme-suite.org).

Figure 4.

(A) Size-exclusion chromatography of Rv0890c_TPR-HTH and Rv0890c_TPR. Inset shows a Coomassie-stained SDS-polyacrylamide gel of the purified proteins. (B) EMSAs were performed by incubating double-stranded DNA sequences (500 ng) generated by PCR amplification of clones obtained from genomic SELEX, with native or heat-inactivated Rv0890c_TPR-HTH (2 μM total protein) followed by native polyacrylamide gel electrophoresis and EtBr staining. (C) EMSAs with ³²P-labeled 1-S41 double-stranded DNA with 2 μM of Rv0890c_TPR-HTH or Rv0890c_TPR. (D) EMSA with radiolabeled 1-S41 double-stranded DNA with increasing concentration of Rv0890c_TPR-HTH as indicated. (E) Affinity of DNA binding was estimated by densitometric quantitation of free and bound probe. “H” indicates the Hill coefficient. Data shown are mean ± SD of duplicate determinations from assays performed twice. (F) EMSAs with ³²P-labeled 1-S41 double-stranded DNA with 2 μM of Rv0890c_TPR-HTH and indicated amounts of unlabeled salmon sperm DNA in μg.

We confirmed that Rv0890c_TPR-HTH bound to several independent sequences using EMSA (Fig. 4 B). The DNA fragment 1-S41 was found to bind efficiently to Rv0890c_TPR-HTH with a clear single shift and complete depletion of free DNA. Hence, 1-S41 was used as the representative SELEX DNA fragment in further experiments. Heat inactivation of the protein abolished DNA binding, and Rv0890c_TPR showed no detectable binding to 1-S41 (Fig. 4 C). Increasing protein concentrations resulted in the formation of complexes with decreasing mobility (Fig. 4 D), indicating that multiple protein molecules could bind to the DNA fragment. Cooperative binding of Rv0890c_TPR-HTH for the ∼550-bp DNA sequence 1-S41, obtained from the genomic SELEX, was observed (Fig. 4 E). The DNA sequences used in these experiments do not possess any internal repeats, and analysis of the sequences obtained in the genomic SELEX showed a lack of sequence similarity. This suggested that the binding of the HTH domain was sequence independent. In agreement with this, incorporation of an excess amount of unlabeled nonspecific competitor salmon sperm DNA during the EMSA abolished DNA binding to Rv0890c_TPR-HTH (Fig. 4 F). Therefore, Rv0890c_TPR-HTH may bind to multiple sites on a single DNA molecule in a sequence-independent manner.

The HTH domain present in Rv0890c is similar to DosR, the response regulator in a two-component system involved in the regulation of genes that are induced under conditions of hypoxia and nitric oxide stress, with 46% sequence identity (37). The crystal structure of DosR along with bound DNA (38) has identified residues involved in DNA interaction (Fig. 5 A). Alignment of the HTH domain of Rv0890c and DosR (Fig. 5 B) reveals that a Lys residue in DosR (Lys179) is conservatively substituted with an Arg residue (Arg 850), whereas the other two residues in DosR that interact with DNA (Lys 182 and Asn 183) are replaced by Gln 853 and Thr 854 in Rv0890c. We mutated Arg 850 to an Ala residue in the Rv0890c_TPR-HTH construct of Rv0890c, purified the protein, and tested its DNA-binding ability. As shown in Fig. 5 C, the Arg850Ala mutant did not bind DNA, whereas the mutation of Arg 844 to Ala (corresponding to conserved Arg 173 in DosR that does not interact with DNA; Fig. 5, A and B) continued to interact with DNA. Purified DosR, which shows sequence-specific binding (37), did not bind to the DNA used for EMSA (Fig. 5 C) because this particular DNA did not contain sequences required for DosR binding.

Figure 5.

(A) Structure of DosR (blue) bound to DNA (Protein Data Bank, PDB: 1zlK), indicating amino acids discussed in the text that are important for DNA binding (K179, K182, and R183) as well as a residue exposed to solvent (R173). (B) Sequence alignment of DosR and the HTH domain of Rv0890c. Residues boxed in orange indicate those important for DNA binding in DosR. (C) EMSA performed by incubating 1-S41 double-stranded DNA as a PCR product (500 ng) with varying concentrations of Rv0890c_TPR-HTHR844A, Rv0890c_TPR-HTHR850A, or DosR. Samples were analyzed by native PAGE followed by EtBr staining.

The HTH domain of Rv0890c binds RNA

We hypothesized that the lack of sequence specificity in DNA-binding ability of Rv0890c_TPR-HTH may allow binding of the protein to RNA. We competed binding of the Rv0890c_TPR-HTH to DNA using double- or single-stranded DNA and RNA and noted that RNA was effectively able to compete for binding to the protein (Fig. 6 A).

Figure 6.

(A) EMSA with ³²P-labeled 1-S41 double-stranded DNA with 2 μM of Rv0890c_TPR-HTH in the presence of 13.2 μg of indicated unlabeled nucleic acids. (B) 660 pmol of indicated proteins was interacted with 30 μg of total M. smegmatis RNA. Bound complexes were pulled down by Ni-NTA agarose resin and bound (pull down) and unbound (breakthrough) samples were subjected to proteinase K and phenol/chloroform/isoamyl alcohol to remove residual protein. Nucleic acids were precipitated and analyzed by agarose/formaldehyde gel electrophoresis followed by EtBr staining. Data are representative of experiments performed using two independent protein preparations.

In an alternative approach, we interacted the purified protein with mycobacterial total RNA followed by immobilization of the protein-RNA complex on Ni-NTA beads. Beads were washed and bound nucleic acid detected by denaturing agarose gel electrophoresis followed by EtBr staining. As shown in Fig. 6 B, a significant amount of RNA could bind to the protein, and this binding was, again, dependent on the HTH domain. Interestingly, DosR failed to bind RNA (Fig. 6 B). The Arg850Ala mutant TPR-HTH protein did not bind RNA, whereas the Arg844Ala protein continued to bind RNA (Fig. 6 B). Therefore, residues critical for interacting with DNA are also essential to allow RNA binding. In summary, the HTH domain of Rv0890c is a nucleic-acid-binding domain that, despite its similarity to DosR and other LuxR-like DNA-binding motifs, can accommodate RNA. The nonsequence specificity of DNA binding is reminiscent of HU, a nucleoid-associated protein NAP from bacteria, including mycobacteria (39,40) and Dps from mycobacteria, which has been shown to organize the DNA into nucleoid-like structures (41).

AFM identifies DNA-protein complexes that form biocrystalline nucleoids in the presence of adenine nucleotides

We then studied the role of the HTH domain in association with the NB-ARC domain but found that the complexes formed with Rv0890c₊₂₀ and sequences obtained by SELEX did not enter acrylamide gels (data not shown) either in the presence or absence of ATP/ADP. We therefore asked if these proteins were oligomeric. We utilized AFM to investigate oligomeric states of Rv0890c₊₂₀ in the presence and absence of adenine nucleotides. Rv0890c₊₂₀ protein molecules alone had a globular shape of several sizes, with the volume spread across a wide range (Fig. 7 A), suggesting that the protein could exist as oligomers that may facilitate DNA/RNA binding. However, upon addition of ATP and AMP-PNP, a nonhydrolyzable analog of ATP, the volume of the protein molecules became more uniform and smaller. The mean volume of the protein molecules represented hexameric and dimeric forms in the presence of ATP and AMP-PNP, respectively. In the presence of ADP, particle volumes increased significantly (Fig. 7 A). These results demonstrate that the oligomeric status of the protein can be regulated by the presence of adenine nucleotides.

The apparent lack of sequence specificity of the HTH domain indicated a role for this class of proteins as being NAPs. DNA was prepared from the clone used for EMSA and imaged alone or after addition of protein. Rv0890c₊₂₀ protein molecules with globular shape of various sizes were found bound to the DNA (Fig. 7 B). No such structures were seen when only the TPR domain was used, in agreement with the EMSA (data not shown; Fig. 4 C).

Addition of ATP to the DNA-protein complex brought about dramatic changes and organized, assembled network-like structure of DNA could be seen (Fig. 7 C). To ensure that the formation of this higher order structure was because of binding of ATP or ADP by the NB-ARC domain, the Rv0890c₊₂₀ WA mutant was used in the presence of 5 mM MgATP or MgADP. Interestingly, the formation of the higher order organized DNA structure was lost, and the DNA-protein complexes showed similar conformation as those in the absence of nucleotides (Fig. 7 D). Thus, ATP or ADP binding by the NB-ARC domain brings about a conformational change across the protein that causes it to form such higher order structures with DNA.

Because both ATP and ADP binding resulted in the formation of organized higher order structures, we determined whether ATP hydrolysis after DNA binding is required. AMP-PNP, a nonhydrolyzable analog of ATP, was used with Rv0890c₊₂₀. In the presence of AMP-PNP, the formation of ordered structures can be seen, but the efficiency of packing the DNA into such structures was not as efficient as with ATP (Fig. 7 D), perhaps as a consequence of different conformational changes induced in the protein in the presence of ATP/ADP or AMP-PNP. Nevertheless, because organized structures were formed in the presence of AMP-PNP, ATP hydrolysis was not necessary to form complex DNA structures.

Finally, to confirm that DNA binding and nucleoid formation was not dependent on the sequence of DNA, we used pGEMT vector alone and observed that crystalloid formation was again seen in the presence of ATP (Fig. 7 E).

Characterization of the Rv0891c-Rv0890c fusion protein

In Mycobacterium canetti, the rv0891c-rv0890c genes are fused and encode a protein with all four domains present in a single polypeptide (Figs. 1 A and 8 A). Upon closer examination of the sequence, we noted that this fusion is a result of the absence of the T residue in the TGA codon of rv0891c thereby removing a stop codon and allowing the remaining basepairs to code for a Glu residue (Fig. 8 A). Other than this change, the amino acid sequences of the predicted M. canetti protein and Rv0891c-Rv0890c are identical. We therefore generated the corresponding deletion in rv0891c-rv0890c construct and were able to purify the recombinant fused protein (Rv0891c-Rv0890c_FUS) (Fig. 8 B, inset). Note that this fusion protein does not contain the first 90 amino acids of the predicted product from M. canetti, in which an additional two variations are seen at (V37G and P86L) when compared to Rv0891c-Rv0890c. This fusion protein would serve as a model to understand the features of other STAND adenylyl cyclases in M. tuberculosis, as well as the properties of the protein in M. canetti, if it too is produced from an internal transcription start site.

Rv0891c-Rv0890c_FUS showed a marginal increase in adenylyl cyclase activity compared to the isolated cyclase domain (∼40 μmol/min/mmol protein). The fusion protein showed no ATPase activity (data not shown), but thermal shift assays showed that it bound ATP (Fig. 8 B). However, strikingly, the curve showed a sharp denaturation, in contrast to the clearly biphasic curves seen with Rv0890c₊₂₀ (Fig. 3 D), indicating that the presence of the cyclase domain allowed the protein to form a single species which denatured at a lower temperature. The Rv0891c-Rv0890c_FUS protein bound to DNA, but complexes did not enter gels in an EMSA analysis (data not shown). As seen in AFM analysis, the fusion protein in the presence of ATP compacted the DNA more efficiently, in comparison to structures formed by Rv0890c₊₂₀ (Fig. 8 C). Importantly, RNA binding to the fusion protein was also observed, and this binding was dependent on the presence of the HTH domain (Fig. 8 D).

Discussion

In this study, we characterize a novel adenylyl cyclase pseudoenzyme and its adjacent gene, representing a “split gene” that belongs to the family of STAND adenylyl cyclases found in mycobacteria. The data we have presented indicate that the two genes are present in an operon in M. tuberculosis. Proteomic evidence indicated that both Rv0891c and Rv0890c are present in M. tuberculosis lysates (42), in agreement with observations made from ribosome profiling (Fig. 1 A; (26)). However, the antibodies we have raised were unable to detect endogenous expression of the proteins reproducibly (data not shown), but only when expressed in M. smegmatis under the strong sigA promoter (Fig. 1 D).

The cyclase domain of Rv0891c could be a vestigial remnant of a once active nucleotide cyclase after a gene splitting event. For example, the cyclase domain of Rv0386 has robust catalytic activity (12). Rv0891c has mutations at the catalytic center, suggesting it is a pseudoenzyme and indicating positive selection for loss of function (Fig. 2 A). The monomeric nature of the expressed and purified protein could also account for its poor activity, as nucleotide cyclases need to form dimers for activity (5). It is, however, possible that an extended N-terminus could facilitate dimerization, but such a protein could not be expressed in E. coli in a soluble form. As is seen in other pseudoenzymes, we speculate that the cyclase domain present in the M. cannetti ortholog for example, may allosterically regulate its associated domains after binding MgATP, as is seen in many pseudokinases (43). Indeed, fusion of these domains in a single protein could allow regions of the two domains to interact with each other robustly. Thermal inactivation assays showed no significant change in stability of the cyclase domain in the presence of ATP (data not shown). However, presence of the cyclase domain imparted thermal stability when fused to the NB-ARC-TPR-HTH domain in the fusion protein (Fig. 8 B). We have been unable to detect robust interactions between Rv0891c and Rv0890c₊₂₀ in vitro, but the products of the “split gene” may interact with each other in the cell, allowing cyclase domain regulation of nucleic acid binding by Rv0890c.

Rv0890c is an inactive ATPase (Fig. 3 C), but the Walker A motif present in the protein can facilitate ATP binding (Fig. 3 D). A catalytically important glutamate residue in ATPases is absent in Rv0890c (Fig. 3 A; (33)) and could account for the absence of ATP-hydrolysis activity. For several STAND proteins, ATP binding and hydrolysis is contingent on interaction with an inducer, which may be the case in Rv0890c (11). Alternatively, the STAND adenylyl cyclases as a group may function without ATPase activity, as exemplified by CED-4 and Dark (44,45). In fact, recent reports find that Dark can form apoptosomes even in the absence of ATP binding, highlighting the functional flexibility in nucleotide binding and hydrolysis by the NB-ARC module across STAND proteins (46).

Importantly, we show here that the HTH domain of mycobacterial STAND proteins act like NAPs and can organize DNA topology in the presence of adenine nucleotides. There are reports of histones forming such assembled DNA structures with λ-DNA (47) and DosR binds to DNA as a dimer (48). Gel filtration data demonstrated that the TPR domain in Rv0890c could determine oligomerization of the protein to facilitate DNA binding (Fig. 4 A). Inside a cell, ATP levels can change in response to external and internal stimuli such as starvation, hypoxia, etc. The [ATP]/[ADP] ratio can be sensed by Rv0890c, which may then bind the adenine nucleotides and promote the formation of a biocrystallized nucleoid, protecting the organism from imminent stress. Several nucleoid-associated proteins in prokaryotes including mycobacteria bring about diverse conformational changes in the nucleoid in response to external and internal stimuli. M. smegmatis adopts nucleoid “toroidal” and “coral-reef” structures in the stationary phase of growth (41). Such “toroidal” structures are similar to the structures formed by Rv0890c₊₂₀ and Rv0891c-Rv0890c_FUS with DNA as observed via AFM. These structures have been referred to as the “last resort for survival” of a bacterial species under stress and are formed because of a phenomenon called “biocrystallization” (49). Thus, under conditions when the dynamic order of life cannot be maintained, an equilibrium order is attained by reversible formation of tightly packed and highly ordered structures, within which vital components are protected through physical sequestration.

The sizes of structures formed by adenine-nucleotide bound protein and DNA seen using AFM are larger than the size of a mycobacterial cell, which is ∼3 μm in length with a diameter of ∼0.5–1 μm (50). However, the sizes we see in the AFM could be a consequence of the concentrations of DNA and protein used for experiments. For example, it is estimated that the concentrations of protein, RNA, and DNA of M. bovis BCG growing in a low-carbon chemostat approximate 72, 10, and 6 fg respectively, per cell (51), at which concentrations AFM would be difficult to perform. Alternatively, such large DNA-protein structures could form after lysis of cells in biofilms as has been reported in other bacteria, in which DNA-binding proteins are found in the biofilm matrix (52,53).

Rv0890c binds to both DNA and RNA, a property hitherto unattributed to LuxR-type HTHs. We engineered an artificial construct to characterize the properties of a protein which would represent the domain organization seen in other STAND adenylyl cyclases in M. tuberculosis. Thus, we posit that, in a protein like Rv0386, in which the absence of this gene has been shown to attenuate the infection seen in mice (8), the effects of this deletion may not rest solely on the reduction in cAMP levels in the cell but could also lie in abrogation of critical DNA-binding properties of this protein. This possibility in fact has not been addressed to date.

The lack of sequence specificity in nucleic acid recognition by the HTH domain of Rv0890c, and perhaps the HTH domain seen in other STAND proteins, may suggest that these proteins may function as NAPs and influence the global architecture of the DNA. Although such proteins typically function by bending, wrapping, or bridging DNA, recent reports have also implicated RNA binding by NAPs as a means of maintaining the nucleoid morphology (54). Noncoding RNAs associate with the nucleoid via the protein HU leading to DNA condensation (55). It is plausible that RNA binding to Rv0890c enhances or modulates interaction with DNA in a similar way. Although Rv0890c showed binding to ribosomal RNA (rRNA), rRNA is typically coated with a large number of other proteins in the cell, thereby decreasing its accessibility. Therefore, to access rRNA, Rv0890c may need to colocalize with the ribosome as has been demonstrated for the HTH of an archaeal protein (56). Alternatively, Rv0890c may be associated with leaderless RNAs or small RNAs produced by promiscuous transcription of the mycobacterial genome (22,57,58). Given the complex domain architecture of this group of STAND adenylyl cyclases, and their unique properties and presence almost exclusively in slow-growing mycobacterial species, it is anticipated that their roles in mycobacteria are complex and identifiable only under certain environmental conditions.

Author contributions

Conceptualization, investigation, and original draft preparation, A.Z.; conceptualization, investigation, and original draft preparation, A.B.; investigation, S.S.; investigation, A.R.; investigation, P.B.; conceptualization and investigation, A.R.S.; conceptualization, supervision, writing, reviewing, and editing, and funding acquisition, S.S.V.

Editor: Smita Patel.