GtrA Protein Rv3789 Is Required for Arabinosylation of Arabinogalactan in Mycobacterium tuberculosis

ABSTRACT

Mycobacterium tuberculosis possesses a thick and highly hydrophobic cell wall principally composed of a mycolyl-arabinogalactan-peptidoglycan complex, which is critical for survival and virulence. DprE1 is a well-characterized component of decaprenyl-phospho-ribose epimerase, which produces decaprenyl-phospho-arabinose (DPA) for the biosynthesis of mycobacterial arabinans. Upstream of dprE1 lies rv3789, which encodes a short transmembrane protein of the GtrA family, whose members are often involved in the synthesis of cell surface polysaccharides. We demonstrate that rv3789 and dprE1 are cotranscribed from a common transcription start site situated 64 bp upstream of rv3789. Topology mapping revealed four transmembrane domains in Rv3789 and a cytoplasmic C terminus consistent with structural models built using analysis of sequence coevolution. To investigate its role, we generated an unmarked rv3789 deletion mutant in M. tuberculosis. The mutant was characterized by impaired growth and abnormal cell morphology, since the cells were shorter and more swollen than wild-type cells. This phenotype likely stems from the decreased incorporation of arabinan into arabinogalactan and was accompanied by an accumulation of DPA. A role for Rv3789 in arabinan biosynthesis was further supported by its interaction with the priming arabinosyltransferase AftA, as demonstrated by a two-hybrid approach. Taken together, the data suggest that Rv3789 does not act as a DPA flippase but, rather, recruits AftA for arabinogalactan biosynthesis.

IMPORTANCE Upstream of the essential dprE1 gene, encoding a key enzyme of the decaprenyl phospho-arabinose (DPA) pathway, lies rv3789, coding for a short transmembrane protein of unknown function. In this study, we demonstrated that rv3789 and dprE1 are cotranscribed from a common transcription start site located 64 bp upstream of rv3789 in M. tuberculosis. Furthermore, the deletion of rv3789 led to a reduction in arabinan content and to an accumulation of DPA, confirming that Rv3789 plays a role in arabinan biosynthesis. Topology mapping, structural modeling, and protein interaction studies suggest that Rv3789 acts as an anchor protein recruiting AftA, the first arabinosyl transferase. This investigation provides deeper insight into the mechanism of arabinan biosynthesis in mycobacteria.

INTRODUCTION

Mycobacterium tuberculosis, the pathogen responsible for most cases of human tuberculosis (TB), is surrounded by a thick highly hydrophobic cell wall, which reduces permeability and confers resistance to many antibiotics. The cell wall core is composed of a covalently bound mycolyl-arabinogalactan-peptidoglycan (mAGP) complex, which is supplemented by proteins, free lipids, and immunomodulatory lipoglycans, such as lipomannan (LM) and lipoarabinomannan (LAM) (1–3). The mAGP complex and LAM are both essential for bacterial survival and virulence (4).

Arabinogalactan forms a major component of the mAGP complex. C-5 of three of the 30 alternating β(1→5) and β(1→6) galactofuranose (Galf) residues (positions 8, 10, and 12) are decorated with extensively branched arabinofuranose (Araf) residues to form mature arabinogalactan (5). Arabinan consists of a linear region of α(1→5) Araf and a branched terminal arabinan motif, [β-Araf-(1→2)-α-Araf]₂-3,5-α-Araf-(1→5)-α-Araf. Decaprenyl-phospho-arabinose (DPA) is the sole sugar donor required for arabinosylation in mycobacteria (6). It is formed by a unique metabolic pathway that includes the epimerization of decaprenyl-phospho-ribose (DPR), catalyzed by the concerted activity of the enzymes DprE1 and DprE2 (7, 8). Thereafter, membrane-associated arabinosyltransferases of the glycosyltransferase C (GT-C) superfamily catalyze arabinan polymerization (9–14).

A dedicated cell wall biosynthetic cluster comprising genes responsible for lipoarabinomannan, arabinogalactan, and mycolic acid biosynthesis exists in the M. tuberculosis genome (15, 16). The region also encodes enzymes of the DPA biosynthetic pathway and is conserved among pathogenic mycobacteria (Fig. 1). Located immediately upstream of dprE1 is rv3789, a gene encoding an integral membrane protein belonging to the GtrA family. Members of the GtrA family were proposed to have diverse functions, including translocation of undecaprenyl-phospho-glucose across the cell membrane during O-antigen modification in Shigella flexneri (17, 18). In a recent study, Rv3789 was implicated in the translocation of DPA across the plasma membrane, based on experiments performed in Mycobacterium smegmatis (19). However, in an M. tuberculosis rv3789 knockout mutant, arabinogalactan and lipoarabinomannan biosynthesis did not seem to be affected, based on the unaltered content and composition of these polymers (19). Thus, the actual function of Rv3789 warrants further investigation.

FIG 1.

Alignment of the rv3789-dprE1-dprE2 locus in different mycobacterial species. Genomic maps are adapted from MycoBrowser (44, 45). Red, rv3789 orthologs; black, dprE1 and dprE2 orthologs.

Here, we present the transcriptional analysis of rv3789 and investigate the topology of Rv3789. We also report the construction, characterization, and phenotypic analysis of an rv3789 deletion mutant in M. tuberculosis that was used to investigate arabinogalactan biosynthesis.

MATERIALS AND METHODS

Bacterial strains, culture conditions, and chemicals.

M. tuberculosis strain H37Rv (15) and derivatives were grown at 37°C in Middlebrook 7H9 broth (Difco) supplemented with 10% albumin-dextrose-catalase (ADC), 0.2% glycerol, and 0.05% Tween 80, or on Middlebrook 7H10 agar (Difco) supplemented with 10% oleic acid-albumin-dextrose-catalase (OADC) and 0.2% glycerol. M. smegmatis strain mc²155 and derivatives were grown at 37°C in Middlebrook 7H9 broth (Difco) supplemented with 2% ADC, 0.2% glycerol, and 0.05% Tween 80, or on Middlebrook 7H10 agar (Difco) supplemented with 2% OADC and 0.2% glycerol. Hygromycin (50 μg ml⁻¹), kanamycin (25 μg ml⁻¹), streptomycin (25 μg ml⁻¹), anhydrotetracycline (ATc) (200 ng ml⁻¹), or 2.5% sucrose was added when needed. For cloning procedures, Escherichia coli One Shot TOP10 cells (Invitrogen) were grown in Luria-Bertani (LB) broth or on LB agar with hygromycin (200 μg ml⁻¹), kanamycin (50 μg ml⁻¹), or spectinomycin (25 μg ml⁻¹). All chemicals were purchased from Sigma-Aldrich, unless otherwise stated.

Construction of recombinant plasmids.

Two fragments of about 900 bp homologous to the upstream and downstream regions of rv3789 were generated using the primer pairs U-fwd and U-rev and D-fwd and D-rev (see Table S2 in the supplemental material). The fragments were ligated in frame into the AvrII site and cloned into the PacI and AscI sites of pJG1100 vector (20, 21), kindly provided by J. McKinney, to yield the pGKH8 suicide vector (see Table S3 in the supplemental material).

To generate a conditional knockdown strain, pGKH23 was created by PCR amplification of rv3789 using the rv3789-E-fwd and rv3789-E-rev primers (see Table S2 in the supplemental material) and by cloning the PCR fragment between the AvrII and AscI sites of pGA44 (22) to obtain the pGKH23 integrative plasmid (see Table S3 in the supplemental material).

Transformation of M. tuberculosis and construction of mutants.

The deletion of rv3789 was accomplished by homologous recombination using pGKH8. After the transformation of H37Rv, the product of the first recombination was selected on 7H10 medium, which contained hygromycin and kanamycin. Colonies were screened by colony PCR using the primer pairs A-fwd and A-rev and B-fwd2 and B-rev (see Table S2 in the supplemental material). An rv3789 merodiploid strain was first generated by integrating pGKH23 at the attB site of transformants positive for the first crossover by selecting on 7H10 medium with hygromycin, kanamycin, and streptomycin. Colonies PCR positive for tetR were selected (using CS-235-tetRF and CS-236-tetRR primers). Finally, to generate the rv3789 conditional knockdown (cKD) mutant (see Table S1 in the supplemental material), deletion of the wild-type gene by allelic exchange was accomplished by plating on 7H10 medium supplemented with streptomycin and 2.5% sucrose. Hygromycin-susceptible (Hyg^s), kanamycin-susceptible (Kan^s), streptomycin-resistant (Str^r), and sucrose-resistant colonies were further tested by Southern blotting.

The rv3789 knockout (KO) strain was generated by replacing pGKH23 at the attB site of the rv3789 cKD strain with the empty pND255 plasmid, kindly provided by Neeraj Dhar. Transformants were selected on 7H10 plates containing hygromycin. The deletion of rv3789 was confirmed by quantitative PCR (qPCR) on the rv3789 KO 5 and rv3789 KO 9 mutants (see Table S1 in the supplemental material).

Two-hybrid system.

The protein interaction system was used as described by Singh and colleagues (23). Genes encoding proteins listed in Table 1 were cloned in frame, without the stop codon, into pUAB100 or pUAB200 at the 5′ end of domains 1 and 2 (F_1,2) or domain 3 (F₃) of dhfr, respectively. Genes were cloned at the 3′ end of F_1,2 or F₃ of dhfr in pUAB300 or pUAB400, respectively. The plasmid pairs were then cotransformed into M. smegmatis and selected on 7H11 medium containing hygromycin and kanamycin. Eight colonies were selected and grown overnight in 100 μl of 7H9 medium supplemented with antibiotics prior to replication on 7H11 plates containing 50, 25, and 13 μg ml⁻¹ trimethoprim (TMP) and incubated at 37°C for 3 to 4 days.

TABLE 1.

Interaction study of Rv3789 using the mPFC system

Location	Fusion protein 1a	Fusion protein 2b	TMP concn (μg ml−1)	Phenotype
Multimer	Rv3789-F1,2	Rv3789-F3	50	Growth
DPA pathway	DprE1-F1,2	Rv3789-F3	13	No growth
	F1,2-DprE1	Rv3789-F3	13	No growth
	Rv3789-F1,2	F3-DprE1	13	No growth
	Rv3789-F1,2	DprE2-F3	13	No growth
	F1,2-DprE2	Rv3789-F3	13	No growth
	Rv3789-F1,2	F3-DprE2	13	No growth
	DprE1-F1,2	DprE2-F3	13	No growth
Arabinosyltransferases	F1,2-AftA	Rv3789-F3	25	Growth
	F1,2-EmbB	Rv3789-F3	13	No growth
	Rv3789-F1,2	F3-EmbA	NAc	NA
	F1,2-AftB	Rv3789-F3	NA	NA

^aF_1,2, domains 1 and 2 of DHFR fused at either the N or C terminus of the protein of interest.

^bF₃, domain 3 of DHFR fused at either the N or C terminus of the protein of interest.

^cNA, results not available because no M. smegmatis transformants were obtained.

Details of the oligonucleotides and plasmids are available in Tables S2 and S3 in the supplemental material.

Growth curves.

To monitor the growth of the mutants, the rv3789 cKD, rv3789 KO, and H37Rv strains were grown to mid-logarithmic phase and then diluted to an optical density at 600 nm (OD₆₀₀) of 0.03. The OD₆₀₀ was recorded at different time points to obtain the growth curves.

RNA extraction and transcript analysis.

For the extraction of M. tuberculosis RNA, the cultures were harvested by centrifugation, flash-frozen in liquid nitrogen, and processed by bead beating in the presence of TRIzol (Invitrogen), as previously described (24, 25). The integrity of the RNA was checked by agarose gel electrophoresis, and the amount and purity of RNA were assessed using a NanoDrop instrument (Thermo Fisher Scientific).

For cotranscription analysis of rv3789, dprE1, dprE2, and aftA, the RevertAid first-strand cDNA synthesis kit (Fermentas) was used to generate specific primed cDNA from 400 ng of two independent RNA preparations, as recommended by the manufacturer. RNA was reverse transcribed using the rv3789-D-rev, dprE2-180-rev, and dprE1-B-rev reverse primers annealing downstream of rv3789, dprE1, and dprE2, respectively. PCR on cDNA fragments was then carried out to determine the region spanned by the transcript.

5′ rapid amplification of cDNA ends (5′-RACE) was carried out twice using 2 μg of RNA, as follows. Single-stranded cDNA was generated using the RevertAid kit and reverse primers specific for the −180 and −120 regions downstream of the rv3789 and dprE1 start codons, respectively (see Table S2 in the supplemental material). Single-strand cDNA was then purified using a High Pure PCR product purification kit (Roche), as recommended by the manufacturer. The poly(A) tail was added to the 3′ end of the first-strand cDNA using recombinant terminal transferase and ATP, as described in the 5′-/3′-RACE kit 2nd generation (Roche) protocol. In order to map the 5′ end of the transcript, the cDNA fragment was PCR amplified using the oligo(dT)-anchor primer specific for the poly(A) tail and reverse primers specific for the −60 and −120 regions downstream of the rv3789 and dprE1 start codons, respectively. PCR fragments were purified using ExoSAP-IT (Affymetrix) prior to sequencing.

Electron microscopy.

Cultures of M. tuberculosis H37Rv, rv3789 KO, and rv3789 cKD strains were grown to OD₆₀₀ of 0.5, and the cells were washed once with phosphate-buffered saline (PBS) and resuspended in 3 ml of PBS. Twenty microliters of each sample was spotted on a silicium grid for 2 min before submerging in a solution of 1.25% glutaraldehyde and 1% tannic acid in phosphate buffer (0.1 M [pH 7.4]) for 1 to 2 h. Washes and gradual dehydration were carried out as previously described (22) prior to coating with Au, Pd, and Os. Images were obtained with a Zeiss Merlin scanning electron microscope. Cell length was measured using ImageJ and then plotted as a percentage of the total number of cells as a function of length. Forty-four H37Rv, 147 rv3789 KO, and 92 rv3789 cKD cells were examined.

Isolation of arabinogalactan peptidoglycan.

M. tuberculosis mAGP was isolated according to Bhamidi et al. (26), with minor modifications, from cells grown to an optical density at 600 nm (OD₆₀₀) of 0.6 to 0.8. Briefly, about 100 mg of heat-inactivated (120°C for 40 min) lyophilized cells was suspended in 5 ml of phosphate-buffered saline (PBS) and disrupted three times using a One Shot cell disrupter (Constant Systems, Ltd., United Kingdom) at 8 × 10⁷ Pa, followed by centrifugation at 27,000 × g for 40 min at 10°C. The pellet was resuspended in 30 ml of PBS containing 2% sodium dodecyl sulfate (SDS) and stirred at room temperature overnight. The mixture was centrifuged for 30 min at 27,000 × g and 17°C before a second PBS-SDS treatment and stirring at 100°C for 1.5 h. The pellet was washed twice with 30 ml of deionized water, twice with 30 ml of 80% acetone, and once with 30 ml of deionized water and subsequently lyophilized.

In order to remove mycolates, about 10 mg of lyophilized mAGP was treated with 500 μl of 1 M sodium hydroxide at 95°C overnight. After cooling, the mixture was neutralized with glacial acetic acid and centrifuged for 30 min at 14,000 × g at room temperature. The supernatant was collected and dialyzed overnight against deionized water in dialysis tubing with a 6- to 8-kDa molecular mass cutoff. The resulting sample was dried in a vacuum concentrator and redissolved in 60 μl of deionized water. The soluble AGP concentration was determined by the phenol-sulfuric acid method (27).

Analysis of monosaccharide composition of the mAGP/AGP.

Alditol acetates were prepared from mAGP or AGP, as described previously (26), with minor modifications. Briefly, 10 μg of heptitol or scyllo-inositol, serving as the internal standards, was added to 0.1 to 1 mg of dry mAGP or AGP. The samples were hydrolyzed with 200 μl of 2 M trifluoroacetic acid (TFA) at 120°C for 2 h. After cooling to room temperature (RT), the samples were dried under N₂, resuspended in 1 ml water, and washed twice with 1 ml of chloroform to remove mycolic acids covalently attached to arabinosyl residues (26). This step was omitted if AGP was derivatized. The water phase was dried under N₂, and the monosaccharides released by TFA hydrolysis were reduced with 100 μl of 10-mg ml⁻¹ sodium borohydride in 1 M NH₄OH–50% ethanol for a minimum of 1 h at room temperature. Reduction was terminated with 3 or 4 drops of glacial acetic acid, and the sample was dried under N₂, washed twice by adding 10% acetic acid in methanol and twice by adding methanol only, followed by drying. Acetylation was carried out with 200 μl of acetic anhydride at 120°C for 2 h. After cooling to RT, the samples were partitioned with water-chloroform (1:1). The alditol acetates were recovered from the organic phase, dried under N₂, and redissolved in 100 μl of acetone.

Gas chromatography-mass spectrometry (GC-MS) measurements were performed on a 6890N gas chromatograph with a 5973Network mass-selective detector (Agilent Technologies, CA). The injection port was maintained at 300°C, and the sample of 1 μl was injected with splitless mode, followed by a purge 1 min after injection. The samples were separated using a 60-m by 0.25-mm (inside diameter) capillary column coated with a 0.25-μm-thick film of 50% cyanopropyl-methylpolysiloxane DB-23 as stationary phase (J&W Scientific, CA). The column temperature was 80°C initially, increased to 250°C at a ramp rate of 10°C min⁻¹, and held at the final temperature of 250°C for 10 min. Helium carrier gas with a constant flow of 1.7 ml min⁻¹ was used. The transfer line temperature was set at 250°C. The quadrupole conditions were electron energy of 70 eV and ion source temperature of 230°C. The target alditol acetates were measured based on SIM mode on the ions m/z 84, 115, 170, and 217. The data handling was carried out using the MSD ChemStation software G1701DA (Agilent Technologies).

Analysis of the lipid fraction.

Cultures of H37Rv, rv3789 KO, and complemented rv3789 KO strains were harvested and washed by centrifugation. The free lipids were extracted from the wet pellets at 70°C with 95% ethanol for 15 min (28). The ethanol fraction was evaporated to dryness under N₂ and partitioned between the organic and water phases with a mixture of chloroform-methanol-water (4:2:1). The lower organic phase was washed once and dried under N₂ prior to further treatment. Each sample was divided into 3 equal aliquots. After drying, two aliquots were subjected to mild alkali treatment after resuspension in 200 μl of chloroform-methanol (2:1) and then treated with 200 μl of 0.2 M sodium hydroxide in methanol at 37°C for 20 min. The reaction was stopped by adding glacial acetic acid to reach pH 7. Lipids were partitioned with chloroform-methanol-water (4:2:1), and the organic phase was dried under N₂. After mild alkali treatment, one sample was subjected to mild acid hydrolysis in a mixture of 100 μl of 1-propanol and 40 μl of 20 mM HCl at 60°C for 30 min. The reaction was neutralized by adding 4 μl of 0.2 M sodium hydroxide. The sample was partitioned with chloroform-methanol-water (4:2:1), and the bottom organic phase was dried under N₂. The untreated, mild-alkali-, and mild-alkali- and mild-acid-treated samples were recovered in chloroform-methanol (2:1), and all samples were spotted on a thin-layer chromatography (TLC) silica gel 60 F₂₅₄ plate (Merck), which was developed in chloroform-methanol-ammonium hydroxide–1 M ammonium acetate-water (180:140:9:9:23). Glycolipids were visualized with orcinol-sulfuric acid reagent (0.2% [wt/vol] orcinol in H₂SO₄-water [1:4, vol/vol] [29]).

The identity of the accumulated mild-acid-labile glycolipid in the mild-alkali-stable lipid fraction of the rv3789 KO strain was established by tandem mass spectrometry (MS/MS) and GC-MS. The mild-alkali-treated lipids, extracted from the rv3789 KO 9 mutant and its complemented form, were loaded on the preparative TLC. The regions corresponding to accumulated glycolipids were scraped off the plate, and lipids were extracted from silica with chloroform-methanol (2:1). For MS/MS, the lipid was mixed with an equal amount of dihydroxybenzoic acid (DHB) matrix, analyzed using matrix-assisted laser desorption ionization–time of flight MS (MALDI-TOF MS), and further fragmented using collision-induced dissociation. For GC-MS analysis, the dried samples were subjected to the procedure of the alditol acetate preparation and analyzed, as described above.

Linkage analysis of arabinogalactan.

AGP was subjected to methylation, according to Ciucanu and Kerek (30), with minor modifications. One milliliter of dimethyl sulfoxide (DMSO) and two pellets of crushed NaOH were added to the tube with 0.5 to 1 mg of dried AGP, and the mixture was stirred for 10 min at room temperature. Subsequently, 2 to 3 ml of iodomethane was added. After stirring the mixture for 30 min, the reaction was quenched by adding 2.5 ml of water, and methylated AG was extracted with 3 ml of chloroform. The chloroform phase was reextracted with water 4 times, dried under N₂, and subjected to the procedure of the alditol acetate preparation described above. GC-MS analysis was performed as described above but with setting the detector to full-scan mode.

Topology mapping.

Multiple sequence alignments were made using ClustalW (31). The C terminus of Rv3789 was fused genetically to green fluorescent protein (GFP) and the β-lactamase reporter protein ′Bla. Briefly, gfp was PCR amplified and cloned into pGKH23 using the AvrII and AflII sites. Full-length rv3789 and rv3789ΔCT were cloned into pMV261-derived plasmids containing E. coli bla under the control of the hsp60 promoter (pMZ102) using MscI and PstI and MscI and BamHI, respectively. pMZ102 was kindly provided by M. Zhang. Rv3789-GFP fusions were electroporated into the rv3789 KO mutant, while the β-lactamase fusions were transformed into the M. smegmatis Δbla strain PM759 (32).

Fluorescence microscopy.

M. tuberculosis cells were treated with 5% paraformaldehyde for 10 min, spotted on a microscope slide, and imaged with a Zeiss LSM 700 inverted microscope equipped with a Plan-Apochromat 63×/1.4-numerical-aperture oil objective and AxioCam MRm (B/W) camera. Images were acquired upon excitation at a 488 nm wavelength, and emission filters were set at 490 to 530 nm for GFP using the ZEN 2010B SP1 software. Images were processed through the ZEN 2011 software (Zeiss) and figures prepared using the ImageJ 1.47 software.

Molecular modeling.

Three-dimensional models were built with the I-TASSER server (33) using complementary information about interresidue contacts derived from coevolution data. More specifically, contacts were predicted from a direct information (DI) analysis of the extensive alignment available in Pfam family PF04138. DI plots were computed with EVfold, a Matlab program kindly provided by Andrea Pagnani (34). Evolutionary data corresponding to residue pairs with a direct information content of >0.7 were considered, and a contact map was plotted for Cα atoms with a cutoff of 8 Å. For each model, the appropriate sets of contacts were fed into the I-TASSER server; each top-scoring I-TASSER model was relaxed through short simulations in implicit solvent with the AMBER99SB force field (35) in NAMD 2.7 (36) and finally equilibrated in a 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) membrane solvated with TIP3P water and counter ions. Dimer formation was tested with the ClusPro server (37). Figures were generated with VMD and PyMOL.

RESULTS

Cotranscription of rv3789 with dprE1, dprE2, and aftA in M. tuberculosis.

Given the conserved position of rv3789 upstream of dprE1 in pathogenic mycobacteria (Fig. 1), we investigated whether rv3789 was coexpressed with dprE1 and the downstream genes (Fig. 2A and B) in M. tuberculosis. By reverse transcription of M. tuberculosis RNA using specific primers, a transcript spanning rv3789, the 39-bp intergenic region (IG) between rv3789 and dprE1, and the first 808 bp of dprE1 was identified, indicating that the two genes were cotranscribed. The last 461 bp of dprE1, dprE2, and the first 300 bp of aftA were also confirmed to be on the same transcript, supporting previous findings, which showed that aftA and embC were cotranscribed with dprE1 and dprE2 (13). Taken together, the data indicate the existence of an operon comprising rv3789, dprE1, dprE2, aftA, and embC, here termed the DPA operon.

FIG 2.

Transcriptional analysis of rv3789 and dprE1 in M. tuberculosis. (A) M. tuberculosis RNA was reverse transcribed using primers specific for the rv3789, dprE1, dprE2, and aftA genes. The top panels represent the PCR results with (+) or without (−) reverse transcription and the genomic DNA control (g). The PCRs span the 4235377 to 4236586 (a), 4234829 to 4236586 (b), 4236704 to 4237373 (c), and 4237168 to 4238264 (d) genomic sequences. (B) Schematic representation of the results obtained in panel A. In red are the cotranscribed regions, whereas the dashed line corresponds to separately transcribed genes. The red arrow indicates the promoter region. (C) Sequences of rv3789 (light gray) and the first 122 bp of dprE1 (dark gray). The guanine residue highlighted in red represents the TSS of the DPA operon, 64 bp upstream of the start codon of rv3789. The putative −10 and −35 promoter sequences are underlined, and the stop codon of rv3788 is in bold. IG, intergenic region. The numbers at left indicate the genome positions.

To localize the promoter of this gene cluster, we mapped the transcription start site (TSS) of the operon by 5′-RACE (Fig. 2C) to a guanine residue 64 bp upstream of the rv3789 start codon. This indicated the presence of a promoter upstream of rv3789, and potential −10 and −35 motifs, recognized by RNA polymerase, were identified (Fig. 2C). No TSS was pinpointed upstream of dprE1, thus ruling out the presence of a promoter between rv3789 and dprE1. Moreover, our results are consistent with the RNA sequencing (RNA-seq) profile of this region obtained from a deep transcriptional study of the M. tuberculosis genome (38) (see Fig. S1 in the supplemental material). Indeed, a sharp peak corresponding to the guanine residue identified here was noted, thereby reinforcing our findings that rv3789, dprE1, dprE2, aftA, and embC are all transcribed from a TSS located upstream of rv3789.

Deletion of rv3789 results in slow growth and altered cellular morphology in M. tuberculosis.

In order to investigate the effect of deleting Rv3789 in M. tuberculosis, we generated an unmarked rv3789 deletion mutant by removing the entire coding sequence while retaining the natural promoter upstream of dprE1 (see Fig. S2A in the supplemental material). When constructing M. tuberculosis mutants, we considered the possibility that rv3789 is essential. Thus, a complementing copy of rv3789 under the transcriptional control of a tetracycline-pristinamycin off system was initially introduced at the attB site of H37Rv, before removing rv3789 from the natural locus by homologous recombination, to yield an rv3789 cKD strain (see Table S1 in the supplemental material). Deletion of the wild-type gene was confirmed by Southern blotting, in which two fragments of 3.3 kb and 7.3 kb appeared in the rv3789 cKD, whereas an 11-kb fragment was present in the wild-type strain, as expected (see Fig. S2B in the supplemental material). Silencing of rv3789 was tested by growing the rv3789 cKD mutant in the presence of anhydrotetracycline (ATc). No growth arrest was observed, suggesting that the gene was not essential (data not shown). Consequently, the complementing copy of rv3789 was removed from the rv3789 cKD, and two rv3789 knockout (KO) mutants (5 and 9) were selected for further investigation (see Table S1).

The growth rates of the Rv3789 KO strains were compared with those of control strains, revealing that rv3789 KO mutants 5 and 9 grew more slowly than both wild-type H37Rv and the complemented strain (rv3789 cKD without ATc) (Fig. 3A). To exclude the possibility that this growth defect arose from the downregulation of dprE1 expression, we measured the transcription levels of rv3789, dprE1, and dprE2 in rv3789 KO mutant 5 and H37Rv (see Fig. S2C in the supplemental material). The levels of dprE1 and dprE2 mRNA in the rv3789 KO and rv3789 cKD strains were similar to those in H37Rv, confirming that the expression of dprE1 and dprE2 was not affected by the deletion of rv3789.

FIG 3.

Characterization of rv3789 KO in vitro. (A) The growth rates of the wild-type H37Rv (black), rv3789 KO mutant 5 (orange), rv3789 KO mutant 9 (red), and the complemented rv3789 cKD (gray) strains show a growth defect in rv3789 KO strains. (B) Scanning electron microscopy images of the H37Rv, rv3789 KO, and rv3789 cKD strains at different magnifications allowed the visualization of shorter and swollen rv3789 KO cells. Scale bar = 1 μm. (C) Length distribution of H37Rv (black), rv3789 KO (orange), and rv3789 cKD (gray) plotted as a percentage of the total number of cells. Bacteria lacking rv3789 are shorter than the wild-type or the complemented strain.

Observation of single cells by scanning electron microscopy revealed a morphological defect in the rv3789 KO mutant (Fig. 3B). M. tuberculosis lacking rv3789 was shorter and more swollen than H37Rv or the cKD strain. Indeed, 40% of the rv3789 KO cells measured <1 μm, and 11% were >2 μm long (Fig. 3C). In the H37Rv and rv3789 cKD strains, 18% and 16% of cells were <1 μm, and 34% and 48% were >2 μm, respectively. These observations therefore suggest that the deletion of rv3789 resulted in a defect in cell morphology and growth of M. tuberculosis, supporting the hypothesis that Rv3789 plays a role in cell wall biogenesis.

Deletion of rv3789 leads to inadequate arabinosylation of arabinogalactan with concomitant accumulation of DPA.

To gain deeper knowledge of the role of Rv3789 in M. tuberculosis, we analyzed the cell wall components, namely, the lipoglycan fraction comprising LAM and LM and the cell wall core-mAGP. While no significant differences were observed in the lipoglycan fraction (see Fig. S3A in the supplemental material), substantial and reproducible changes were found in the cell wall core. The glycosyl composition of mAGP purified from M. tuberculosis strains was established by GC-MS after the release of monosaccharides by hydrolysis with 2 M TFA and their conversion to alditol acetates. Compared to the wild-type strain, rv3789 KO mutants showed a 25% decrease in the Ara-to-Gal ratio (Fig. 4A). This change was explained by a reduction in arabinan content, as demonstrated by the 25% decline in the number of arabinose residues per rhamnose, while the number of galactose residues per rhamnose remained unaffected (see Fig. S3B in the supplemental material). It was shown that after hydrolysis of mAGP with TFA, mycolic acids remain attached to arabinose residues, and these are then not converted to alditol acetates (26). Thus, in order to determine more accurately the Ara-to-Gal ratio, we removed mycolates from mAGP by alkali hydrolysis before preparation of the alditol acetates. The reduction of the Ara-to-Gal ratio in the rv3789 KO strains compared to that in the wild type was even more profound, reaching 38% (see Fig. S3C in the supplemental material).

FIG 4.

Analysis of mAGP and lipid fractions of the M. tuberculosis cell wall. (A) Columns represent the analysis of three separate mAGP samples extracted from H37Rv (black), rv3789 KO mutant 5 (orange), rv3789 KO mutant 9 (red), and rv3789 cKD (gray) strains and demonstrate a decrease in the Ara-to-Gal ratio in M. tuberculosis depleted of Rv3789. ns, P value not significant; ***, P < 0.0001. Error bars indicate standard errors of the means. (B) TLC analysis shows the accumulation in the rv3789 KO strains of DPA (arrow), as confirmed by MS/MS and GC-MS (see Fig. S6 in the supplemental material).

Analysis of the glycosyl linkage composition of AG (Table 2) revealed a decrease in all types of Araf units in the rv3789 KO strains. This was accompanied by an increase in all types of Galf units, as the abundance of Araf and Galf species was calculated as a percentage of the total AG residues in this experiment. Therefore, the effect of rv3789 deletion on the biosynthesis of arabinan appeared to be general rather than due to the specific inhibition of the attachment of a particular Araf unit in AG biosynthesis. Accordingly, we observed a decline in the amount of mycolic acids bound to mAGP obtained from rv3789 KO strains (see Fig. S3D in the supplemental material), which is likely due to reduced arabinan content in the truncated AG.

TABLE 2.

Glycosyl linkage analysis of per-O-methylated AG^a

Residue	% (mean ± SD) or ratio in:
	H37Rv	rv3789 KO mutant 5	rv3789 KO mutant 9	rv3789 cKD mutant
5-Araf	38.0 ± 1.1	35.5 ± 1.6	33.1 ± 0.7	36.6 ± 1.3
3,5-Araf	11.9 ± 0.4	10.8 ± 0.5	10.1 ± 0.3	11.7 ± 0.4
2-Araf	11.2 ± 0.6	8.7 ± 0.5	7.8 ± 0.9	11.1 ± 0.1
t-Araf	9.9 ± 0.4	9.7± 0.6	9.9 ± 0.7	10.4 ± 0.9
5-Galf	16.7 ± 0.3	19.9 ± 1.5	22.4 ± 0.7	17.0 ± 0.8
6-Galf	8.3 ± 0.4	10.2 ± 0.8	11.2 ± 0.1	8.6 ± 0.2
5,6-Galf	2.5 ± 0.3	3.2 ± 0.5	3.4 ± 0.4	2.9 ± 0.2
t-Galf	1.5 ± 0.2	2.0 ± 0.6	2.1 ± 0.8	1.7 ± 0.4
Total Ara/Galb	2.44	1.83	1.55	2.3

^aPer-O-methylated AG was extracted from three different cultures of H37Rv, rv3789 KO mutant 5, rv3789 KO mutant 9, and the rv3789 cKD mutant. Percentages of the total arabinosyl-furanose and galactosyl-furanose residues are shown.

^bCalculated from the sum of Araf residues divided by the sum of Galf residues.

Finally, the decrease in arabinan content was accompanied by an accumulation of DPA in the rv3789 KO mutants (Fig. 4B). Indeed, analysis of the total extractable lipids revealed the presence of mild-alkali-stable and mild-acid-labile lipids, characteristic of the polyisoprenyl-phospho-sugar class of the precursors in the rv3789 KO strains. The regions corresponding to the accumulated metabolites in the mild-alkali-treated samples were extracted from the silica plate and analyzed by MS/MS and GC-MS (see Fig. S4 in the supplemental material). MS/MS revealed the presence of a species with an m/z ratio of 909.668, corresponding to the molecular ion of DPR or DPA as the major product (see Fig. S4A in the supplemental material). GC-MS identified arabinose as the main sugar component in this sample (see Fig. S4B in the supplemental material), leading to the conclusion that DPA is the polyisoprenyl-phospho-sugar that accumulates in M. tuberculosis depleted of rv3789.

Rv3789 is a four-helix transmembrane protein with cytoplasmic C terminus.

Sequence alignments of Rv3789 orthologs from different mycobacteria revealed considerable similarity and residue conservation among the pathogenic, opportunistic, and saprophytic species (Fig. 5A). Several topology prediction algorithms predicted the presence of four ordered transmembrane α-helices with cytosolic N and C termini (see Fig. S5 in the supplemental material). To confirm this prediction, the topology and orientation of the protein were investigated using two experimental approaches. In the first, a genetic fusion of Rv3789 with a C-terminal GFP was expressed in the rv3789 KO mutant. We also verified that the GFP fusion did not impair the function of Rv3789, since complementation restored the wild-type phenotype (see Fig. S6 in the supplemental material). In the second approach, using fluorescence microscopy, a positive GFP signal was observed along the cell boundaries (Fig. 5B), thereby validating the membrane localization of Rv3789 and confirming a cytoplasmic orientation of its C terminus.

FIG 5.

Topological analysis of Rv3789. (A) Multiple alignments of Rv3789 orthologs and their predicted transmembrane helices (in brown) from various mycobacteria. MtRv, M. tuberculosis H37Rv; Mbo, M. bovis; Mlp, M. leprae; MtKZN, M. tuberculosis strain KZN605; Mma, M. marinum; Mul, M. ulcerans; Mka, M. kansasii; Mav, M. avium; Mpt, M. paratuberculosis; MtCDC, M. tuberculosis strain CDC1551; Min, M. intracellulare; Mpa, M. parascrofulaceum; Mva, M. vanbaalenii; Mgi, M. gilvum; Msm, M. smegmatis; MAb, M. abscessus; C, cytoplasmic; P, periplasmic; t, transmembrane domain. (B) rv3789 KO complemented with rv3789-gfp shows membrane localization of GFP fluorescence. (C) M. smegmatis expressing truncated but not full-length Rv3789 β-lactamase fusion proteins confers carbenicillin resistance, indicating an extracytoplasmic orientation of helix 3 and a cytoplasmic orientation of the C terminus. Kan, kanamycin; Car, carbenicillin. (D) Topology map of transmembrane helices of Rv3789 based on C terminus fusions with GFP or β-lactamase.

As an additional means of verifying the topology of Rv3789, a previously described β-lactamase fusion system was used (32). Here, E. coli ′Bla, lacking its N-terminal signal sequence, was fused genetically to the C terminus of full-length Rv3789 and to Rv3789ΔCT, a truncated protein in which residues 89 to 121 corresponding to the 4th transmembrane helix were removed. These constructs were transformed into an M. smegmatis Δbla strain, and the strains were tested for their ability to grow on plates containing carbenicillin, indicating extracytoplasmic β-lactamase activity. Bacteria expressing the full-length Rv3789-BlaC fusion failed to grow on carbenicillin-containing medium, suggesting that the C terminus of Rv3789 is cytosolic, whereas the strain expressing Rv3789ΔCT-BlaC grew in the presence of carbenicillin, suggesting that the end of the 3rd helix is periplasmic (Fig. 5C and D), consistent with the GFP labeling results.

Rv3789 interacts with itself and with AftA.

Rv3789 appears to link the production of DPA and the biosynthesis of arabinan. Therefore, we tested the potential interactions of Rv3789 with enzymes involved in the last step of DPA production (DprE1 and DprE2) and with arabinosyltransferases involved in AG biosynthesis (AftA, AftB, EmbA, and EmbB) using the mycobacterial protein fragment complementation (mPFC) two-hybrid system (23). This system detects TMP resistance in M. smegmatis upon reconstitution of dihydrofolate reductase (DHFR) from two independent domains, F_1,2 and F₃, fused to the proteins under investigation. As reported in Table 1, fusion of the C-terminal end with domains F_1,2 and F₃ of DHFR confers resistance to 50 μg ml⁻¹ TMP, suggesting that Rv3789 self-associates. On the other hand, no interaction was detected between Rv3789 and DprE1 or DprE2 (Table 1).

In order to test the interaction of Rv3789 with the arabinosyltransferases, we fused the DHFR domains with the N termini of AftA, AftB, EmbA, and EmbB. M. smegmatis transformed with plasmids expressing Rv3789 and AftA fusion proteins was resistant to 25 μg ml⁻¹ TMP (Table 1), whereas the other transformants were either not viable (Rv3789 and EmbA, and Rv3789 and AftB) or did not confer TMP resistance (Rv3789 and EmbB). Therefore, the combined results suggest that Rv3789 interacts both with itself and with AftA.

Structural model of Rv3789.

Structural insight might elucidate the function of Rv3789 but, unfortunately, no three-dimensional structures were available for any GtrA family member. However, we were able to exploit the very large number of protein sequences available in databases to build reliable models ab initio using EVfold, a novel algorithm based on an analysis of sequence coevolution (Fig. 6). EVfold has proven to be particularly valuable for helical transmembrane proteins (39). Analysis of an alignment of >4,000 sequences predicts contacts between amino acid residues between helices 1 and 2, 2 and 3, 3 and 4, and 1 and 4 (see Table S4 in the supplemental material). The first three sets of contacts are between consecutive helices and thus correspond to intramonomer contacts, but the contacts between helices 1 and 4 might be satisfied either in a compact monomer or between monomers of a multimer, consistent with the results of our two-hybrid experiment. Thus, we built models of possible monomers and dimers. The resultant models are reasonable physically, consisting of a four-helix bundle per monomer, with a smooth exposed hydrophobic surface in the region expected to span the hydrophobic portion of the membrane and held by internal polar contacts (Fig. 6). The highly charged residues occur on the loops that are expected to be cytoplasmic, for instance between helices 2 and 3, consistent with von Heijne's “positive inside” rule (40).

FIG 6.

Structural models of Rv3789. (A) Map of the contacts between pairs of residues for which the global statistical modeling software EVfold computes a direct information (DI) value of >0.7 for connections between residues whose Cα atoms are separated by <8 Å. The dots signify contacts between residues predicted by DI. H1 to H4 represent the four transmembrane helical regions in Rv3789. Two possible structures of the monomer were modeled considering the interhelical contacts, as shown at right. (B) Model of monomeric Rv3789 assuming that contacts between helices 1 and 4 are intrachain contacts. (C) Dimer corresponding to panel B. (D) Model of monomeric Rv3789 considering that contacts between helices 1 and 4 are intermonomeric. The bold arrow indicates a small cavity. (E) Dimeric assembly of the monomer in panel D built to satisfy contacts between helices 1 and 4 between monomers. In all models, gray and cyan correspond to carbon, yellow corresponds to sulfur, blue corresponds to positively charged atoms, and red corresponds to negatively charged atoms. Nt and Ct denote the N terminus and C terminus, respectively.

DISCUSSION

The aim of this work was to explore the transcription of rv3789 and neighboring genes and to investigate the role of Rv3789, a small transmembrane protein, in M. tuberculosis. We identified a single mRNA produced from a TSS 64 bp upstream of rv3789, which spans rv3789, dprE1, dprE2, and aftA. We also showed that Rv3789 plays an important role at the interface between DPA synthesis and the arabinosylation of AG in M. tuberculosis.

When investigating transcriptional regulation in M. tuberculosis, a poor match to the −35 consensus sequence for RNA polymerase was identified upstream of rv3789, which may account for the low-level transcription. Indeed, this observation is supported by the results obtained in a deep transcriptional study of the M. tuberculosis genome, which reported that the respective RPM (number of reads per kilobase per million reads aligned) values of rv3789, dprE1, and dprE2 were 0.31, 0.34, and 0.86, respectively, in exponential phase. These values were considerably lower than those of other genes, including the housekeeping sigA and rpoB genes, which had RPM values of 5.1 and 6.23, respectively (38). Furthermore, the study also showed that gene expression in M. tuberculosis is directly correlated with the orientation of the coding sequence (CDS) with respect to the direction of DNA replication (38). In this context, it is noteworthy that rv3789, dprE1, dprE2, aftA, and embC are located on the lagging strand for DNA replication, as is the case in Mycobacterium bovis, Mycobacterium leprae, Mycobacterium marinum, and M. smegmatis, therefore implying low transcription levels in these species as well. The maintenance of low expression levels may be required physiologically, possibly to tally with the expression of other genes involved in the same pathway.

An unmarked rv3789 deletion mutant was then constructed to investigate the function of this GtrA protein while ensuring that expression of the downstream genes (dprE1, dprE2, aftA, and embC) occurred from the native promoter at the natural level. Indeed, we demonstrated that the deletion of rv3789 had no impact on their expression. This approach therefore gives us confidence that the phenotypic changes observed are due solely to a loss of Rv3789. Moreover, our attempt to minimize potential polar effects may explain the discrepancy with a previously published study, in which no differences were observed in an M. tuberculosis mutant in which rv3789 was disrupted by a kanamycin resistance cassette (19).

Two putative functions were initially proposed for Rv3789 as either a flippase for cytoplasmically synthesized DPA, or as an anchor protein involved in cell wall biogenesis by recruiting other proteins (19). The results presented here, along with recently published data and our ab initio structural models of Rv3789, are more consistent with the anchor protein hypothesis, as we discuss below.

First, the interaction of AftA with Rv3789 is compatible with the role of Rv3789 as an anchor protein that supports a complex, comprising other enzymes of the DPA pathway and the arabinosyltransferases, in order to facilitate efficient arabinosylation from the newly formed arabinosyl donor, DPA. AftA is the priming enzyme responsible for transferring the first arabinose residue onto the galactan chain, and its gene is coexpressed with rv3789, dprE1, and dprE2. Second, in agreement with the fact that Rv3789 interacts with AftA rather than with other Araf transferases, a global decrease in arabinose residues was found in AG branching analysis. Indeed, a reduction in the efficiency of AftA activity, due to the absence of Rv3789, would reduce the number of complete arabinan chains rather than affect a particular type of glycosidic linkage in the polymer.

Of note, Rv3789 is predicted to have a very high isoelectric point of 10.4, implying a highly positive charge at neutral pH that might facilitate the recruitment of proteins to specific sites in the membrane. Indeed, other recent studies support the notion of the existence of a multiprotein complex to maintain “the efficiency and fidelity of AG polymerization” (41). Rv3789 was proposed to stabilize the membrane-associated GlfT1 (19), and interactions between arabinosyltransferases and proteins of the lower DPA pathway were identified in an E. coli two-hybrid system (41). Therefore, the initiation of galactan elongation by GlfT1 and the initiation of arabinan biosynthesis by AftA are closely related and likely involve Rv3789.

Structural models of Rv3789, with its four transmembrane alpha helices, were helpful for functional understanding and are supported by our topological and two-hybrid data. The models of monomeric and dimeric Rv3789, in which contacts between helices 1 and 4 were considered to be intramonomeric, are both very compact, revealing no lumen that might accommodate a substrate for translocation. This also applies to the model of dimeric Rv3789, in which contacts between helices 1 and 4 are between monomers. These three structures are thus inconsistent with a flippase function. Only the monomer in the latter dimer (Fig. 6D) has a lumen of polar residues, albeit small, that might accommodate a molecule for translocation. However, the presence of a single negatively charged residue in the lumen of Rv3789 is most likely incompatible with the translocation of negatively charged DPA, since it has been reported that a cationic flippase lumen is required to translocate an anionic O-antigen subunit across the hydrophobic inner membrane (42).

A final and definitive piece of evidence suggesting a role for Rv3789 in protein recruitment rather than as a flippase comes from our very recent study showing that the epimerization of DPR to DPA takes place in the periplasm (43). Hence, since DPA is already located in the periplasm, a DPA flippase is unnecessary. To conclude, we confirm here previous findings that Rv3789 is involved in arabinan assembly in mycobacteria and propose a function for Rv3789 as an anchor for a larger multiprotein complex that synthesizes AG.