1. The Erp protein interacts with Rv1417 and Rv2617c in a bacterial two-hybrid system
We used a two-hybrid system developed by Ladant and co-workers [7], in which genes of interest are fused to T18 and T25, two complementary fragments that are essential for adenylate cyclase activity. If the corresponding fusion proteins interact, cAMP is produced in an endogenous adenylate cyclase-deficient E. coli strain (BTH101), and this functional complementation can be easily monitored by plating bacteria in minimal medium supplemented with lactose. In this work, we searched for Erp-binding proteins by screening an M. tuberculosis DNA expression library with full-length Erp using this bacterial two-hybrid system. The size of the library was approximately 105 clones. Out of 6 × 103 plated transformants, 10 cya+ clones which could grow in minimal medium supplemented with lactose were selected, indicating ten potential interactions. Enzymatic restriction analysis revealed that clones were unique (data not shown). In order to confirm the interactions and to exclude "false positives", plasmids were purified and used to retransform E. coli BTH101. In this second round, only three plasmid clones whose products were able to confer adenylate cyclase activity in E. coli BTH101 co-transformed with plasmid T25-Erp were selected. Sequence analysis of inserts revealed that two of these plasmids encoded Rv1417 and one Rv2617c. The plasmids encoding Rv1417 had a complete copy of the gene and they differed only in the length of the 5' region upstream of Rv1417, while in the plasmid encoding Rv2617c the first 60 bp of the gene were absent. Rv1417 and Rv2617c were annotated as membrane proteins, both of unknown function [8]. All sequenced fragments were in-frame with the ORF encoding T18. The fact that both plasmids encoding Rv1417 were independent clones confirms that non-redundant clones were present in the genomic library and reinforces the feasibility of Erp-Rv1417 protein interaction.
In order to examine the protein-protein interactions mentioned above, erp, Rv1417 and Rv2617c full-length genes were fused to both T18 and T25 gene sequences in pUT18c and pKT25 vectors, respectively. The efficiencies of functional complementation between hybrid proteins were determined by the number of colonies grown in M63 medium supplemented with lactose, and by β-galactosidase activity (see Additional file 1). The level of interaction between Erp and both Rv1417 and Rv2617c was significantly higher than that of the negative controls, independently of the adenylate cyclase fragments (T25 or T18) these proteins were fused to (Fig. 1). These interactions were confirmed by in vitro GST-Pull down assays (see Additional file 2).
2. Rv1417 and Rv2617c interact with each other
The binding of Erp to two different proteins raised the question whether these two proteins are able to interact with each other. In order to clarify this point, the binding between Rv1417 and Rv2617c was addressed by using the bacterial two-hybrid system. These experiments were facilitated by the availability of constructs with each gene in both vectors that had been prepared for the present work. Plasmids (encoding protein fusions of Rv1417 and Rv2617c with T25 and T18 polypeptides) were used to transform E. coli BTH101 cells. All plasmid combinations were subjected to a quantitative screening on selective medium plates. Each hybrid protein tested was able to associate with the other partners (Fig. 1 and Additional file 1). Indeed, both Rv1417 and Rv2617c exhibited strong self-associations, thus suggesting homodimer complex formation of these proteins (Fig. 1 and Additional file 3). The Erp fusions, however, were impaired in self-association (data not shown). None of the hybrid proteins gave a significant complementation signal when tested either with control T25 and T18 polypeptides or with unrelated proteins, like lipoprotein P27 [9].
3. The carboxy-terminal domain of Erp is not relevant for protein interactions
In an effort to map the Erp region involved in protein-protein interactions, we assessed the capacity of each Erp domain to bind both Rv1417 and Rv2617c. Firstly, we focused on determining whether the carboxy-terminal region of Erp, which is involved in the association with the cell wall [10], contained a binding domain. The full sequence of the erp gene was divided in two regions at the base number 528, and fusions of both regions to the T18 encoding sequence were generated. The resulting hybrid proteins were then tested in two-hybrid complementation assays with Rv1417 and Rv2617c fused to the T25 fragment. Deletion of the carboxy-terminal region of Erp did not affect its association with Rv1417 and Rv2617c, thus suggesting that this region of the protein is not essential for this interaction (Fig. 2).
As a second step, the interacting region of Erp was further divided at aa 80, and fusion proteins of the amino-terminal and central domains of Erp with T18 polypeptide were generated. When separated, the amino-terminal and the central domains were unable to interact with Rv1417 and Rv2617c (data not shown). Only the protein deleted in the carboxy-terminal region showed a level of interaction comparable to, or even higher than, that of the full-length protein.
4. Interaction of Erp members from Mycobacterium smegmatis and Mycobacterium leprae
As mentioned in the introduction section of this work, Erp carboxy- and amino-terminal domains are fully conserved, while the central region shows polymoprhism among mycobacterial species, with respect to the number and quality of repeats. While M. leprae has four repeats, M. smegmatis has twenty-six, half of which contain two mismatches [4]. Therefore, it was plausible that the interaction with Rv1417 and Rv2617c was affected by the number of repeats. In order to evaluate this assumption, the interaction of M. smegmatis and M. leprae Erp homologues with both Rv1417 and Rv2617c was assayed. M. leprae Erp (Ml Erp) was unable to associate with either Rv1417 or Rv2617c. M. smegmatis Erp (Ms Erp) showed interaction with Rv1417 but completely failed to bind Rv2617c (Fig. 3). Although we can not exclude the possibility that the lack of interaction of Rv1417 and Rv2617c with the Erp member from M. leprae was due to a misfolding of the T25-Ml Erp protein, the impossibility of this fusion protein and of the T25-Ms Erp protein to interact with Rv2617c is interesting since it correlates with the absence of a functional Rv2617c gene in the M. leprae and M. smegmatis genomes (see below). On the other hand, the interaction of Rv1417 with the Erp member of M. smegmatis, but not with the one from M. leprae, indicates that the number and sequences of the PGLTS repeats are relevant for the occurrence of such interaction.
5. Erp, Rv1417 and Rv2617c are located in close proximity
In order to determine the localization of the potential protein complex Erp-Rv1416-Rv2617c in mycobacterial cells, we first performed an in silico search for protein domains in Rv1417 and Rv2617c. Sequence analysis with InterProScan [11] recognized a signal-peptide domain in Rv1417 and Rv2617c. However, this was not confirmed by SOSUI or SignalP [12, 13], two different software programs that perform signal peptide predictions. Both SOSUI and TMHMM [14] predicted the presence of two and three transmembrane helices for Rv1417 and Rv2617c, respectively. The analysis performed with the SOSUI server showed two transmembrane helices (encompassing positions: 22–44 and 51–72) in Rv1417 and three transmembrane helices (encompassing positions: 19–41, 81–103 and 117–139) in Rv2617c. TMHMM predicted that the probability of an extracellular location for the C-terminal region of Rv1417 was 0.36 and 0.64 for a cytosolic orientation. The predicted probability for an outward orientation for the C-terminus of Rv2617c was approximately 0.82. These results indicate a probable membrane localization of these proteins. To confirm these predictions, we performed immunolocalization of the proteins in subcellular compartments and in culture supernatants by using specific antibodies. Rv1417 and Rv2617c were expressed as a fusion to the myc epitope (see Materials and Methods) to allow their detection in mycobacterial cells. Since attempts to detect Rv1417-Myc in M. tuberculosis were unsuccessful, probably due to a very low expression of the recombinant protein, the fused protein was expressed in the M. smegmatis strain mc2 155. Figure 4 shows the localization of Rv1417-Myc and Rv2617c-Myc in the membrane fraction of recombinant M. smegmatis and M. tuberculosis, respectively, but not in those transformed with the empty vector. Therefore, the co-localization of Rv1417 and Rv2617c suggests that protein-protein associations may take place in the cell envelope. In agreement with previous studies [1, 2], Erp was identified in the cellular wall fraction and culture supernatant of M. tuberculosis by using a monoclonal specific antibody [5]. The absence of reacting bands in a P36-deficient M. bovis strain verified the antibody specificity.
6. Characterization of interacting proteins
In a BLAST [15] comparison of the predicted amino acid sequences of Erp interacting proteins, Rv1417 appeared to be conserved among the Mycobacterium genus and showed similarity to hypothetical membrane proteins from other bacterial species, such as Rhodococcus sp (identity 48%, similarity: 67%), Corynebacterium ammoniagenes (identity: 37%, similarity: 58%, with RibX protein), Corynebacterium diphterae (identity: 35%, similarity: 59%), Streptomyces coelicolor (identity: 34%, similarity: 55%) and Streptomyces ambofaciens (identity: 34%, similarity: 57%). Conversely, the Rv2617c gene was observed to be disseminated only among members of the M. tuberculosis complex (MTC). A pseudogene similar to Rv2617c was observed to be present in the M. leprae genome. The deduced amino acid sequence of Rv2617c showed similarity to hypothetical membrane proteins from Rhodococcus sp (identity: 48%, similarity: 67%), Nocardoides sp (identity: 61%, similarity: 72%), and Arthrobacter sp (identity: 56%, similarity: 72%).
In order to experimentally analyse the distribution of Rv2617c and Rv1417 in the MTC, PCR assays using specific primers were performed on genomic DNA from MTC species. DNA fragments of expected size were obtained for each gene in all species studied (Fig. 5A). In addition, the transcription of Rv1417 and Rv2617c during the in vitro culture of M. tuberculosis was demonstrated by RT-PCR (Fig. 5B). These results suggest that Rv1417 and Rv2617c are functional genes conserved in the MTC.
We investigated the neighbourhoods of Rv1417 and Rv2617c, as well as of their orthologues, with the aim of obtaining clues regarding the biological role of these genes. While examination of the genomic location of the Rv2617c and its orthologues did not reveal any particular feature, we found that Rv1417 and its orthologues are upstream flanked by genes encoding for proteins involved in riboflavin synthesis. Riboflavin operons with a similar structure, containing an Rv1417-like gene, were identified in all Mycobacterium species whose genomes were sequenced, as well as in other species of Actinomycetales genera (Fig. 5C). In addition, RibX, whose gene is a putative member of the riboflavin operon of C. ammoniagenes, showed similarity to Rv1417 (E = e-22). Although the role of RibX in riboflavin synthesis remains elusive, a DNA fragment that includes part of the ribX gene was demonstrated to be involved in riboflavin production [16]. To determine whether Rv1417 is co-transcribed with RibH, a gene encoding a probable riboflavin synthase beta chain in M. tuberculosis and M. bovis strains, RT-PCR assays were performed by using primers that map across the two adjacent genes. The RT-PCR products shown in figure 6 indicate that Rv1417 and RibH are transcribed to a single mRNA molecule. Based on these results, we propose Rv1417 as part of the riboflavin operon in M. tuberculosis.