Departamento de Biología Molecular y Biotecnología, Instituto de Investigaciones Biomédicas, Universidad Nacional Autónoma de México, Ciudad Universitaria, Ciudad de Mexico, DF, México
Correspondence: Luis Servín-González, Departamento de Biología Molecular y Biotecnología, Instituto de Investigaciones Biomédicas, UNAM, Apartado Postal 70228, Cd. Universitaria, Ciudad de Mexico, DF 04510, México. Tel.: +52 55 5622 8928;
A protein glycosylation system related to that for protein mannosylation in yeast is present in many actinomycetes. This system involves polyprenyl phosphate mannose synthase (Ppm), protein mannosyl transferase (Pmt), and lipoprotein N-acyl transferase (Lnt). In this study, we obtained a series of mutants in the ppm (sco1423), lnt1 (sco1014), and pmt (sco3154) genes of Streptomyces coelicolor, which encode Ppm, Lnt1, and Pmt, to analyze their requirement for glycosylation of the heterologously expressed Apa glycoprotein of Mycobacterium tuberculosis. The results show that both Ppm and Pmt were required for Apa glycosylation, but that Lnt1 was dispensable for both Apa and the bacteriophage φC31 receptor glycosylation. A bacterial two-hybrid assay revealed that contrary to M. tuberculosis, Lnt1 of S. coelicolor does not interact with Ppm. The D2 catalytic domain of M. tuberculosisPpm was sufficient for complementation of an S. coelicolor double mutant lacking Lnt1 and Ppm, both for Apa glycosylation and for glycosylation of φC31 receptor. On the other hand, M. tuberculosisPmt was not active in S. coelicolor, even when correctly localized to the cytoplasmic membrane, showing fundamental differences in the requirements for Pmt activity in these two species.
It is now well established that many bacteria are capable of carrying out different types of protein glycosylation, and recent studies have shown the importance of this protein modification (Nothaft & Szymanski, 2010). In some bacteria of the ε subdivision of the proteobacteria, such as Campylobacter jejuni, N-glycosylation of proteins has been shown to be an important factor for pathogenicity (Nothaft & Szymanski, 2013). A system homologous to that of protein O-mannosylation in yeast has been described in actinomycetes, including Streptomyces coelicolor and Mycobacterium tuberculosis (Lommel & Strahl, 2009; Espitia et al., 2010), and the crucial role of this protein modification in M. tuberculosis virulence has recently been demonstrated (Liu et al., 2013). This system involves polyprenyl phosphate mannose synthase (Ppm), homologous to dolichol phosphate mannose synthase of yeast; Ppm carries the GDP-mannose-dependent mannosylation of polyprenyl phosphate on the intracellular side of the cytoplasmic membrane. Mannosylated polyprenyl phosphate is then flipped to the extracytoplasmic side, and transfer of mannose to serine or threonine residues of protein substrates is then carried out by protein mannosyl transferase (Pmt), during secretion (VanderVen et al., 2005; Lommel & Strahl, 2009). In the case of M. tuberculosis, several mannoproteins important for pathogenesis have been identified (González-Zamorano et al., 2009), among them the 45- and 47-kDa antigen Apa, which is the best characterized mycobacterial glycoprotein in terms of the glycosylation sites and the configuration and number of sugar residues (Dobos et al., 1996; Espitia et al., 2010). However, there is little information on the specific proteins glycosylated by this system in S. coelicolor. Only the PstS protein has been shown to be a substrate for glycosylation (Wehmeier et al., 2009), and genetic evidence indicates that glycosylation of the phage φC31 receptor is required for infection by this phage (Cowlishaw & Smith, 2001, 2002). Streptomyces lividans, which is taxonomically closely related to S. coelicolor, has been shown to glycosylate the Apa antigenic protein of M. tuberculosis, and the resulting glycoprotein showed very similar antigenic properties to the native protein (Lara et al., 2004); this result is important, as even Apa produced from the fast-growing Mycobacterium smegmatis appeared to show differences in the glycosylation pattern (Horn et al., 1999). This opens the possibility for easier heterologous production of mycobacterial glycoproteins in the nonpathogenic and fast-growing streptomycetes. However, it has not been formally proven that glycosylation of mycobacterial proteins is carried out by the same yeast-like protein mannosylation system in streptomycetes.
Here, we show that the Apa protein is expressed and glycosylated by S. coelicolor, a strain that is taxonomically very close to S. lividans, but has the advantage of a well-developed system for genetic manipulation. Using a series of constructed null mutants, we demonstrate that Ppm and Pmt activities are essential for Apa glycosylation. We also show that Lnt1, the homologue of the D1 or Lnt domain of M. tuberculosis Ppm, is dispensable for glycosylation of the Apa protein and of the bacteriophage φC31 receptor and that, in contrast to mycobacteria, the homologous Lnt1 of S. coelicolor does not interact with the Ppm protein. Given the phylogenetic relationship between mycobacteria and streptomycetes, we also explored the functionality of M. tuberculosis Ppm and Pmt in S. coelicolor, as this might provide a way for production of mycobacterial glycoproteins by introducing a cognate glycosylation system in a heterologous host; we show that Ppm, but not Pmt, is functional when heterologously expressed.
Materials and methods
Media and growth conditions
Escherichia coli strains were grown in 2XYT medium (Sambrook & Russell, 2001). Growth of Streptomyces mycelium, preparation of spores, transformation with polyethylene glycol, conjugations, and phage propagation were carried out according to Kieser et al. (2000). For protein expression experiments, S. coelicolor was grown in LB broth containing 34% sucrose to obtain dispersed mycelial growth (Lara et al., 2004).
Construction of S. coelicolor mutants using PCR targeting
Unmarked deletion mutants were obtained by the PCR targeting procedure of Datsenko & Wanner (2000) on relevant cosmids carrying the cloned regions of interest of the S. coelicolor chromosome (Redenbach et al., 1996), followed by recombination of the mutations into the chromosome as described by Gust et al. (2004). All mutants were verified by PCR and sequencing to confirm replacement of the relevant gene with the 81-bp in-frame ‘scar’ sequence (Gust et al., 2004). The cosmids used were St6D7A, StE87, and 2StG2, which carry the cloned ppm, pmt, and lnt1 genes, respectively. Table 1 lists the strains, plasmids, and bacteriophage used in this study, while Supporting information, Table S1 lists the oligonucleotides used.
Table 1. Strains, plasmids, and bacteriophage used in this study
Plasmid construction and purification were carried out according to Sambrook & Russell (2001). DNA amplification was carried out using PfuUltra DNA polymerase AD and site-directed mutagenesis using the QuikChange kit (both from Agilent Technologies). A detailed description of plasmid construction is provided in Data S1.
Western blot analysis
For Western blot analysis of secreted Apa, proteins in the culture supernatant were concentrated by methanol/chloroform precipitation (Wessel & Flügge, 1984), subjected to 12% SDS-polyacrylamide gel electrophoresis and transferred to PVDF membranes. Detection was carried out using either 6A3 monoclonal antibody, which was raised against purified Apa (Lara et al., 2004), or concanavalin A (ConA) conjugated with peroxidase (Sigma), both at a 1 : 1000 dilution. For detection of the hemagglutinin epitope in tagged Pmt proteins, membrane fractions were subject to electrophoresis in 10% SDS-polyacrylamide gels, transferred to PVDF membranes, and incubated with 3F10 high-affinity anti-hemagglutinin-Peroxidase antibodies (Roche) at a 1 : 1000 dilution. Detection was carried out with the BM Chemiluminescence Western blotting kit (Roche).
Assay of Ppm synthase and Pmt activities
Purification of membrane fractions from Streptomyces mycelium was carried out as described by Kim et al. (2005), and fractionation of M. smegmatis membranes used for standardization of Ppm activity was carried out as described by Cascioferro et al. (2007). Assay of Ppm activity in membrane fractions was carried out as described by Gurcha et al. (2002), using GDP-[U-14C]mannose, 262 mCi/mmol (PerkinElmer). Pmt activity was determined in a coupled assay using the Apa-derived peptide A3 (Invitrogen) as described by Cooper et al. (2002). Detailed description is provided in Data S1.
The bacterial two-hybrid system of Karimova et al. (1998) was used, based on plasmids pKT25 and pUT18 with modified polylinker regions (Karimova et al., 2001). β-galactosidase activity was determined according to Miller (1972).
Results and discussion
Ppm activity is required for glycosylation of Apa by S. coelicolor
The sco1423 gene (ppm) encodes Ppm of S. coelicolor (PpmSco; Cowlishaw & Smith, 2002; Wehmeier et al., 2009). We constructed strain IB31 carrying a deletion of this gene in the J1928 background, which is wild type except for a pgl mutation that allows bacteriophage φC31 to form plaques (Table 1). As expected, φC31 was able to form plaques in J1928 (Fig. 1a, plate 1), but not in the Δppm mutant IB31 (Fig. 1a, plate 2; Table S2), confirming that PpmSco is required for infection by φC31. To determine whether PpmSco is required for glycosylation of the Apa protein of M. tuberculosis by S. coelicolor, we cloned the apa gene (Rv1860) under the control of the PtipA promoter in the integrative vector pRT802 and introduced the resulting plasmid (pBL1, Table 1) into the wild-type (J1928) and Δppm (IB31) strains. The Apa protein obtained from supernatants of J1928 carrying the cloned apa gene (in pBL1) could be seen as a clear band in Western blots, both when detection was based on a monoclonal antibody (Fig. 1b, lane 1) and when it was based on reaction to the ConA lectin (Fig. 1c, lane 1), meaning that S. coelicolor is able to express, secrete, and glycosylate the Apa protein, as has been previously shown for S. lividans; in contrast to S. lividans, the Apa protein secreted by S. coelicolor was subject to some degradation, as revealed by the presence of a faint faster migrating band not observed in S. lividans (Lara et al., 2004). This is consistent with the known observation that S. coelicolor is more proteolytic than S. lividans (Kieser et al., 2000; Jayapal et al., 2007). When supernatants of the Δppm mutant IB31 carrying the cloned apa gene (in pBL1) were analyzed, the Apa protein was still expressed and secreted, as evidenced by the presence of a clear band detected by the 6A3 monoclonal antibodies (Fig. 1b, lane 2), but it was not glycosylated, as indicated by the slightly lower mass observed and by the lack of reaction with ConA (Fig. 1c, lane 2). This result indicates that PpmSco is essential for glycosylation of M. tuberculosis Apa by S. coelicolor.
Complementation of a S. coelicolor ppm mutant by M. tuberculosis Ppm
To determine whether the S. coelicolor Δppm mutant IB31 could be complemented by M. tuberculosis Ppm (PpmMtu), the Rv2051c gene was amplified from M. tuberculosis H37Rv DNA and cloned under the control of the strong PtipA promoter (plasmid pBL10, Table 1); the S. coelicolor ppm gene (sco1423) and upstream flanking region were cloned in pSET152 as a control (plasmid pBL13, Table 1). Phage φC31 was able to form plaques in the S. coelicolor Δppm mutant IB31 carrying either pBL10 or pBL13, encoding PpmMtu and PpmSco, respectively (Fig. 1a, plates 3 and 4; Table S2). In addition, introduction of these same plasmids into the S. coelicolor Δppm mutant expressing Apa [IB31(pBL1)] restored glycosylation of this protein, as indicated by the presence of bands in Western blots detected with monoclonal antibodies (Fig. 1b, lanes 3 and 4), which showed restoration of ConA reactivity (Fig. 1c, lanes 3 and 4).
To demonstrate activity of PpmMtu in S. coelicolor, an in vitro assay was carried out to detect labeling of the membrane polyprenyl phosphate by GDP-[14C]mannose in purified membrane fractions. Streptomyces coelicolor harbors a single C45 membrane polyprenol (Wehmeier et al., 2009; Fig. S1), and clear labeling of this molecule was observed in membranes of wild-type S. coelicolor (J1928) as indicated by a single-labeled band (Fig. 2, lane 1), but not in membranes of the Δppm mutant (IB31; Fig. 2, lane 2). Complementation was confirmed by this in vitro assay, because labeling of the membrane polyprenyl phosphate was restored when either pBL13 (PpmSco) or pBL10 (PpmMtu) was introduced into the Δppm mutant (Fig. 2, lanes 3 and 4, respectively), confirming that PpmMtu is functional when expressed in S. coelicolor.
PpmMtu is a protein composed of two distinct domains. The N-terminal hydrophobic domain D1 (Met1-Tyr593) is responsible for lipoprotein N-acyltransferase (Lnt) activity, whereas the C-terminal domain D2 (Met594-Glu874) is the Ppm catalytic domain (Gurcha et al., 2002; Tschumi et al., 2009). We therefore decided to analyze whether the isolated D2 domain of PpmMtu was functional in S. coelicolor in the absence of the D1 domain. To do this, the portion of the Rv2051c gene encoding the D2 domain was cloned in pIJ6902 under control of the PtipA promoter (pBL11) and introduced into the S. coelicolor Δppm mutant IB31. Figures 1 and 2, and Table S2 show that expression of the isolated D2 domain of PpmMtu is sufficient for complementation of the Δppm mutation of IB31, both in vivo and in vitro, as evidenced by restoration of φC31 plaque formation (Fig. 1a, plate 5), ConA reactivity of the Apa protein (Fig. 1b, lane 5 and c, lane 5), and in vitro labeling of polyprenyl phosphate (Fig. 2, lane 5).
The S. coelicolor homologue of M. tuberculosis Lnt is not essential for Apa glycosylation by S. coelicolor
In mycobacteria and corynebacteria, Lnt and Ppm are either functional domains of the same protein (as in M. tuberculosis) or separate proteins encoded by contiguous genes that often exhibit translational coupling (Gurcha et al., 2002). All Streptomyces genomes sequenced to date reveal that the genes encoding Ppm are not preceded by those encoding homologues of the Lnt domain of PpmMtu; instead, Streptomyces genomes show the presence of two genes encoding homologues of Lnt located separately on the chromosome. It has recently been shown in Streptomyces scabies that Lnt1, the homologue exhibiting higher identity (44%) to Lnt of mycobacteria is functional, whereas the functionality of Lnt2, which exhibits only 26% identity, is still unclear (Widdick et al., 2011). We obtained a derivative of the wild-type J1928 with an in-frame deletion of the lnt1 gene (sco1014) and tested this strain (IB65, Table 1) for phage infection. Figure 3a (plate 3) and Table S2 show that φC31 was able to form plaques in the Δlnt1 mutant IB65; in addition, Apa protein obtained from this strain was recognized by ConA, indicating that it was glycosylated (data not shown).
Previous works have shown that the Lnt domain of PpmMtu is required for full Ppm activity and that it might anchor the catalytic domain (D2) to the membrane, in order to mannosylate the membrane polyprenyl phosphate (Gurcha et al., 2002). We therefore determined whether the PpmMtu D2 domain could complement the Δppm mutant in the absence of Lnt1. To do this, a double mutant was obtained with deletions of both the ppm and lnt1 genes (strain IB67, Table 1). Plasmids expressing PpmSco (pBL13) or only the D2 domain of PpmMtu (pBL11) were introduced into the Δppm Δlnt1 mutant IB67 and analyzed for their ability to restore phage infection. Results shown in Fig. 3a and Table S2 reveal that, as expected, φC31 was unable to form plaques in IB67 (Fig. 3a, plate 4) and that plaque formation in the double mutant was restored by complementation with either PpmSco (Fig. 3a, plate 5) or the PpmMtu D2 domain (Fig. 3a, plate 6), meaning that Lnt1 is dispensable for Ppm activity in S. coelicolor. Given this observation and the difference in gene arrangement between streptomycetes and mycobacteria (Fig. S2), we asked whether the domain interaction previously reported between the D1 (Lnt) and D2 (Ppm) domains of PpmMtu (Baulard et al., 2003) was also shown by Lnt1 and PpmSco. To answer this, the S. coelicolor lnt1 and ppm genes were cloned in the bacterial two-hybrid system of Karimova et al. (1998; pB18 and pB19, respectively; Table 1) and tested for interaction by measuring β-galactosidase activity in a cya mutant of E. coli; as a control, the D1 (Lnt) and D2 (Ppm) domains of PpmMtu were also cloned in the same system (pB16 and pB17, respectively; Table 1). The D1 and D2 domains of PpmMtu indeed interacted, as evidenced by the increase in β-galactosidase activity in cultures carrying both pB16 and PB17, when compared to the background levels observed with either one or both empty vectors (Fig. 3b). On the other hand, when the cultures carried pB18 (Lnt1) and pB19 (PpmSco), no significant increase in β-galactosidase activity above the background was observed (Fig. 3c), meaning that Lnt1 and PpmSco do not interact, a result consistent with the previous observation that Lnt1 is dispensable for Ppm function in S. coelicolor.
Streptomyces coelicolor protein mannosyl transferase (Pmt) is required for glycosylation of the Apa protein
The S. coelicolor pmt gene (sco3154) encodes a protein mannosyl transferase (PmtSco) that is essential for infection by φC31 and for glycosylation of the PstS protein (Cowlishaw & Smith, 2001; Wehmeier et al., 2009). PmtSco is a homologue of M. tuberculosis protein mannosyl transferase (PmtMtu). We therefore decided to analyze whether PmtSco was responsible for glycosylation of Apa by S. coelicolor. For this purpose, we obtained an S. coelicolor mutant carrying an in-frame deletion of the pmt gene (strain IB25, Table 1). Phage φC31 was unable to form plaques in IB25, as expected (Fig. 4a, plate 2; Table S2). In addition, the Apa protein produced from the Δpmt mutant IB25 carrying the cloned apa gene (in plasmid pBL1; Fig. 4b, lane 2) was not glycosylated, as indicated by its lack of reactivity to ConA (Fig. 4c, lane 2), compared with the same protein obtained from the wild-type J1928 (Fig. 4b lane 1 and c, lane 1). This result means that PmtSco (which is responsible for glycosylation of the φC31 receptor and of the PstS protein in S. coelicolor) is also responsible for glycosylation of the heterologously expressed Apa protein. We therefore asked whether PmtMtu could complement the null mutation in the Δpmt mutant IB25; heterologous expression of PmtMtu might be particularly important for synthesis of mycobacterial glycoproteins in Streptomyces, as this enzyme is the one responsible for recognition of sites in proteins targeted for glycosylation. In contrast to N-glycosylation, where a linear sequence constitutes a glycosylation site (Nothaft & Szymanski, 2013), there is no clear consensus of what constitutes a target site for O-glycosylation by the Pmt enzymes, although there appears to be a poorly defined sequence requirement, usually consisting of a threonine- and proline-rich region, which may point to a structural requirement (Lommel & Strahl, 2009; Espitia et al., 2010). If there are differences in recognition of sites targeted for glycosylation between Pmt enzymes, then the expression of PmtMtu in S. coelicolor might produce mycobacterial glycosylated proteins that are more similar to the native ones produced by M. tuberculosis. To answer whether PmtMtu is functional in S. coelicolor, the Rv1002c gene (which encodes PmtMtu) was amplified by PCR using appropriate oligonucleotides and cloned under the control of the PtipA promoter in vector pIJ6902 (a TTA leucine codon inside the Rv1002c gene was replaced by a CTG codon by site-directed mutagenesis, to ensure translation of PmtMtu, as this is a rare codon for Streptomyces) resulting in plasmid pBL9. No φC31 plaques were observed on the Δpmt mutant carrying the cloned Rv1002c gene for PmtMtu [IB25(pBL9)], whereas they could be observed when the Δpmt mutant carried an equivalent construct with the S. coelicolor pmt gene also under the control of PtipA [IB25(pBL12); Fig. 4a, plates 3 and 4; Table S2]. To explain this observation, we hypothesized that perhaps PmtMtu was functional, but failed to recognize the φC31 receptor. Therefore, plasmids pBL9 and pBL12 carrying the cloned genes for PmtMtu and PmtSco were also introduced into the S. coelicolor Δpmt mutant IB25 expressing the apa gene (from pBL1), and Apa produced by these strains was analyzed; only pBL12 carrying the gene for PmtSco complemented the ability to glycosylate the Apa protein (Fig. 4b and c, lane 3), whereas pBL9 did not (Fig. 4b and c, lane 4). Again a few degradation products were observed, and these were more apparent when Apa was not glycosylated, which is consistent with the notion that protection from degradation might be one of the functions for protein glycosylation. These results mean that the PmtMtu enzyme is unable to complement Pmt activity in the S. coelicolor mutant, even when the glycosylation target is Apa, a protein that, unlike the φC31 receptor, is normally recognized by PmtMtu.
One possibility to explain these results is that PmtMtu is not being correctly localized to the S. coelicolor membrane, unlike PmtSco. To test this, both PmtSco and PmtMtu were tagged at the C-terminus with a hemagglutinin epitope, to allow their identification using commercial anti-hemagglutinin antibodies, and cloned under the control of the PtipA promoter (pB14 and pB15, respectively; Table 1). Both plasmids were introduced into the Δpmt mutant IB25, and after induction of the cultures with thiostrepton, mycelium was harvested and subject to fractionation, and the cytoplasmic and membrane fractions were analyzed by Western blot using anti-hemagglutinin antibodies. Hemagglutinin-tagged PmtSco could only be found in the membrane fraction (Fig. 5, lane 1) and not in the cytoplasmic fraction (Fig. 5, lane 2), meaning that the hemagglutinin tag did not affect its correct localization. In addition, the hemagglutinin-tagged PmtSco was shown to complement the Δpmt mutant IB25 for the ability to form plaques when infected with φC31 (data not shown). These results show that the hemagglutinin tag did not affect either the correct localization or the functionality of PmtSco. Hemagglutinin-tagged PmtMtu was also found only in the membrane fraction (Fig. 5, lane 3) and not in the cytoplasmic fraction (Fig. 5, lane 4), meaning that it is correctly localized to the S. coelicolor membrane; therefore, incorrect localization is not the reason for lack of complementation of the Δpmt mutation in IB25. Even though both genes were expressed from the strong and inducible PtipA promoter, hemagglutinin-tagged PmtMtu appeared to be less abundant than hemagglutinin-tagged PmtSco, when expressed in S. coelicolor under full induction (to ensure that this fainter band was not due to a difference in the amount of protein loaded, the membrane was stained with Coomassie brilliant blue, Fig. S3). In addition, there appeared to be limited degradation of this protein, presumably related to the fact the S. coelicolor has an abundance of extracellular proteases (Jayapal et al., 2007). It is unlikely that this slightly lower abundance is the reason for lack of complementation, because hemagglutinin-tagged PmtSco was able to complement the Δpmt mutation for φC31 plaque formation even in the absence of inducer when expression relied on background PtipA transcription levels, revealing that even low levels of functional Pmt are sufficient for complementation (Fig. S4).
The previous result prompted us to look for differences between PmtSco and PmtMtu to search for clues to the nonfunctionality of PmtMtu in S. coelicolor. Protein mannosylation by PmtMtu requires Sec translocation, and it has been proposed that physical interactions between the Sec complex and Pmt explain this requirement (VanderVen et al., 2005); therefore, the nonfunctionality of PmtMtu in S. coelicolor could result from its inability to interact with the S. coelicolor Sec translocon. Upon alignment of the Pmt protein sequences from mycobacteria and Streptomyces species, it was clear that the main difference is the presence in the Streptomyces Pmt sequences, including that of S. coelicolor, of an N-terminal extension. According to the prediction for topology of mycobacterial Pmt, this N-terminal extension should be located on the intracellular side of the membrane (Lommel & Strahl, 2009; Fig. S5). Because this extension could prove important for Pmt function in S. coelicolor (if, for example, it is required specifically for interaction with the S. coelicolor Sec translocon), we constructed two modified versions of the Rv1002c gene to encode chimeric Pmt proteins and cloned them in pIJ6902; in the first construct (pBL20, Table 1), 55 amino acids of PmtSco were affixed to the N-terminus of PmtMtu, giving PmtMtu + 55, whereas in the second construct (pBL21, Table 1), 178 amino acids of PmtSco, which include the first extracellular loop where acidic residues essential for activity are localized (VanderVen et al., 2005), were substituted for the equivalent N-terminal region of PmtMtu (Fig. S5). When pBL20 was introduced into the Δpmt mutant IB25, no complementation was observed, either for φC31 plaque formation (Fig. 4a, plate 5) or for Apa glycosylation (Fig. 4b and c, lane 5). Therefore, PmtMtu + 55 was not functional, as the complete absence of φC31 plaques indicates that the phage receptor was not glycosylated by this chimeric construct, which was also unable to glycosylate the Apa protein. On the other hand, when pBL21 was introduced into the Δpmt mutant IB25, mycelium could only be grown under noninducing conditions (i.e. in the absence of thiostrepton) where no complementation of the Δpmt mutation was observed (Fig. S4); adding thiostrepton to the medium resulted in a lack of growth, either in liquid medium or on solid medium, meaning that expression of this chimeric Pmt protein was lethal (Fig. S6). Therefore, replacement of the N-terminal region containing the first extracellular loop of PmtMtu by that of PmtSco was apparently not innocuous. One possibility is that this construct results in a misfolded protein that is toxic; we consider this unlikely, given the structural conservation shown by both Pmt proteins (Fig. S5). Another possible explanation is that the lethal phenotype is the result of a nonproductive interaction between the chimeric Pmt and the Sec translocon, perhaps affecting Sec function to such an extent as to make it nonfunctional. We attempted to show specific interactions between components of the S. coelicolor Sec translocon and either PmtSco or PmtMtu using the bacterial two-hybrid system of Karimova et al. (1998), but no significant interactions could be observed (data not shown). It has also been recently suggested that interaction between corynebacterial Pmt and Lnt might be essential for Pmt function in these bacteria, as there was no detectable glycosylation of a lipoprotein that is normally glycosylated in the absence of Lnt (Mohiman et al., 2012). Mycobacteria are closely related to corynebacteria, so it is conceivable that PmtMtu is not functional in S. coelicolor because it is unable to interact with S. coelicolor Lnt1. Because our results show that Lnt1 is not required for PmtSco function, this might reveal a fundamental difference between these two bacterial groups.
Because PmtMtu failed to complement the Δpmt deletion of S. coelicolor in vivo, we wondered whether Pmt activity could be detected in vitro, using the assay previously described for glycosylation of the synthetic A3 peptide derived from the Apa protein with a purified membrane fraction (Cooper et al., 2002). As can be seen in Fig. S7, membranes of wild-type S. coelicolor J1928 were able to mannosylate the A3 peptide, whereas those of the Δpmt mutant IB25 were not. When plasmid pBL12 encoding PmtSco was introduced into the Δpmt strain IB25 in vitro activity was restored, but no Pmt activity was detected when pBL9 (encoding PmtMtu) was introduced into this strain, meaning that this enzyme is not functional when expressed in S. coelicolor. This result supports the idea that PmtMtu is not capable of forming a productive interaction with the S. coelicolor Sec translocon and that specific interaction with its cognate Sec apparatus (or the Lnt protein) is required for PmtMtu activity even in the absence of protein translocation. This specificity of PmtMtu functionality means that expression of M. tuberculosis glycoproteins will be better achieved by using a related host-like S. coelicolor with a homologous glycosylation system, rather than by attempting the heterologous expression of the M. tuberculosis glycosylation system.
We are grateful to Dr. Y. López-Vidal for the gift of M. tuberculosis H37Rv DNA, to Dr. Antonio Vallecillo for providing M. smegmatis mc2155 cells, to Dr. F. Bigi for providing the bacterial two-hybrid system, and to the Unidad de Biología Molecular of the Instituto de Fisiología Celular-UNAM for DNA sequencing. This work was supported by research grant 103214 from the SEP-CONACyT mixed fund and by a scholarship to L.E.C.-D. from Consejo Nacional de Ciencia y Tecnología (Mexico) to support her PhD studies at the Programa de Doctorado en Ciencias Biomédicas, Universidad Nacional Autónoma de México.