Export-Mediated Assembly of Mycobacterial Glycoproteins Parallels Eukaryotic Pathways

See allHide authors and affiliations

Science  05 Aug 2005:
Vol. 309, Issue 5736, pp. 941-943
DOI: 10.1126/science.1114347


Protein O-mannosylation is an essential and evolutionarily conserved post-translational modification among eukaryotes. This form of protein modification is also described in Mycobacterium tuberculosis; however, the mechanism of mannoprotein assembly remains unclear. Evaluation of differentially translocated chimeric proteins and mass spectrometry to monitor glycosylation demonstrated that specific translocation processes were required for protein O-mannosylation in M. tuberculosis. Additionally, Rv1002c, a M. tuberculosis membrane protein homolog of eukaryotic protein mannosyltransferases, was shown to catalyze the initial step of protein mannosylation. Thus, the process of protein mannosylation is conserved between M. tuberculosis and eukaryotic organisms.

Like eukaryotes, prokaryotes modify proteins with a variety of glycosyl residues that influence a number of biological events (1, 2). In Mycobacterium tuberculosis (Mtb), two fully characterized glycoproteins (Apa/Rv1860 and MPB83/Mb2898) possess threonine residues with linear α(1→2) and α(1→3) oligomannosides, a glycosylation pattern reminiscent of eukaryotic short-chain mannoproteins (35). Other proteins of Mtb are known to be glycosylated, presumably with mannose residues as defined by ConcanavalinA (ConA) reactivity (6, 7). As with other bacterial pathogens, the glycosylation of mycobacterial proteins influences host interactions; for example, antigen-specific T cell recognition of Apa requires mannosylation (8, 9), and LpqH, an O-mannosylated lipoprotein, is a Toll-like receptor 2 (TLR-2) agonist (10).

The spatial assembly of a few bacterial glycoproteins has been addressed (1113), and although most bacterial glycoproteins are extracytoplasmic (1), little evidence exists to link glycosylation with any of the bacterial protein secretion pathways. In Mtb, the biosynthetic pathway of protein glycosylation remains largely unknown. However, similarities exist between the O-mannosylation of Mtb proteins and protein O-mannosylation in Saccharomyces cerevisiae. Specifically, mycobacterial and yeast mannoproteins are exported in a Sec-dependent fashion (3, 4, 6, 7), and protein O-mannosyltransferase (PMT) activity of both organisms is membrane associated, requiring a lipid carrier to donate mannose (5, 14). These similarities between otherwise disparate organisms suggest that the enzymatic machinery for protein O-mannosylation might also be conserved. The PMTs of S. cerevisiae are integral membrane proteins sharing amino acid sequence identities of 57.5% (5). Bioinformatic approaches identified a single Mtb PMT homolog, designated Rv1002c. This putative 55.5-kD protein displayed 22 to 24% identity with the PMTs of S. cerevisiae and presented a similar hydropathy profile (fig. S1). More prognostic of potential PMT activity were two conserved Arg residues in transmembrane domains and an Asp-Glu motif in the first predicted extracytoplasmic domain of Rv1002c (Fig. 1A). The Arg and Glu residues are essential and form the putative active site in yeast PMTs (15). These invariant residues are also common to the PMTs of higher eukaryotes such as humans (16).

Fig. 1.

Membrane topology of Rv1002c and in vitro protein mannosylation activity. (A) Bioinformatic analyses predicted 11 transmembrane domains in Rv1002c and the presence of the conserved PMT active site residues R39, D55, E56, and R106. (B) In vitro PMT activity of cytoplasmic membrane preparations from log-phase recombinant M. smegmatis clones as determined through a composite assay that assesses the transfer of mannose from guanosine diphosphate mannose through mannosyl-phosphoryl-decaprenol (14) to a synthetic peptide (AAAPPAPATPVAPPPPHHHHHH) encompassing a single glycosylation domain of Apa. 1, vector control; 2, wild-type Rv1002c; 3, Rv1002c DE→AA mutant; 4, Rv1002c DE→AE mutant; 5, Rv1002c DE→DA mutant. Error bars indicate SD of three replicate experiments. Reverse transcription polymerase chain reaction analyses confirmed the expression of the recombinant forms of Rv1002c in M. smegmatis.

To demonstrate PMT activity, wild-type Rv1002c and three mutated forms of this gene conferring A55A56, A55E56, or D55A56 substitutions in the invariant D55E56 motif were overexpressed in Mycobacterium smegmatis, and membrane preparations were assayed for in vitro PMT activity. Overexpressed wild-type Mtb Rv1002c significantly increased (66.8%) PMT activity of the M. smegmatis membranes above that observed in the vector control (Fig. 1B). Only wild-type recombinant Rv1002c conferred increased PMT activity and, similar to the PMTs of S. cerevisiae, the invariant D55E56 motif of Rv1002c was critical to proper function (Fig. 1B). In particular, membranes containing recombinant Rv1002c with a D55E56 to A55E56 substitution resulted in decreased (56.0%) PMT activity compared with the M. smegmatis vector control. Thus, altered forms of Rv1002c apparently competed for substrate binding without catalysis of glycosylation.

The membrane location of the Mtb PMT (Rv1002c) and Sec-translocation of Mtb mannoproteins suggested a link between protein export and glycosylation. This hypothesis was partially supported by the reactivity of ConA to a subset of membrane and culture filtrate (CF) proteins but not cytosolic proteins (fig. S2). To evaluate the association between protein O-mannosylation and Sec-translocation, a recombinant form of FbpC/Rv0129c, a well-characterized Sec-translocated and nonglycosylated protein of Mtb (17), was fused with and without its signal peptide (SP) to a mannosylation cassette (MC) (DPEPAPPVPTTAASP) of Apa (3) and a hexa-His tag (Fig. 2A). These two chimeric proteins differentially localized on the basis of the presence or absence of a Sec-dependent translocation SP. SP(+)FbpC-MC localized to the membrane, cell wall (CW), and CF, but not the cytosol, of Mtb and M. smegmatis (Fig. 2B). N-terminal sequencing of the CW and CF forms of SP(+)FbpC-MC revealed the sequence (FSRPGLPVEY) of the mature amino terminus of FbpC (17), confirming that SP(+)FbpC-MC was processed through the Sec-translocase. In comparison, SP(–)FbpC-MCwasobservedinthecytosolof M. smegmatis as expected but was also found in the membrane and CW of this bacterium and was predominantly in the CW of Mtb, with a truncated form in the membrane (Fig. 2B). The occurrence of SP(–)FbpC-MC in membrane and CW fractions was probably due to mixing of protein pools during subcellular fractionation and the fatty acid binding domain of FbpC (18). Indeed, the absence of SP(–)FbpC-MC from the CF demonstrated that this construct was not actively secreted. Instability of SP(–)FbpC-MC in Mtb probably accounted for its absence from the cytosol and the apparent molecular weight shift in the membrane fraction.

Fig. 2.

Protein chimera structures and their localization in Mycobacterium spp. (A) Schematic representation of protein chimeras generated to evaluate the relation between O-mannosylation and translocation. Amino acid sequences shown indicate the trypsin-generated fragment containing the Apa MC, with the known glycosylation site (3) underlined (26). The heavy dark line depicts the SP of FbpC that targets proteins for Sec-translocation, the open box depicts the mature FbpC, the gray box represents the Apa MC and hexa-His tag, the dotted box depicts SodA, and the slashed box depicts ESAT-6. (B) Subcellular localization of each chimera produced in Mtb and M. smegmatis was determined by Western blotting using a monoclonal antibody to His6 as the probe. Lane 1, cytosol; lane 2, cell membrane; lane 3, CW; lane 4, CF.

The glycosylation status of Mtb- and M. smegmatis–derived SP(+)FbpC-MC and SP(–)FbpC-MC purified from the CF and CW was determined after affinity chromatography purification. SP(+)FbpC-MC was ConA reactive; however, SP(–)FbpC-MC did not bind ConA (fig. S3). Liquid chromatography–electrospray ionization–tandem mass spectrometry (LC-ESI-MS/MS) of the trypsin-digested Mtb SP(+)FbpC-MC yielded three major [M+H]3+ molecular ions that differed by 54 m/z (mass/charge ratio) (Fig. 3A) and corresponded to the MC with zero, one, and two hexose residues. MS/MS fragmentation of the largest molecular ion (1327.0 m/z) generated the [M+H]3+ daughter ions of 1273.7 and 1219.6 m/z arising from neutral losses of 54 m/z (Fig. 3B), a diagnostic property for the dissociation of hexose residues from a triply charged precursor peptide (3). The relative abundance of parent ions demonstrated in a semiquantitative manner that the majority of SP(+)FbpC-MC was glycosylated, with the most dominant form having a single hexose (Fig. 3A). Identical experiments with the M. smegmatis forms of this protein demonstrated similar patterns of glycosylation. In contrast, LC-ESI-MS/MS analysis of trypsin-digested SP(–)FbpC-MC recovered from the Mtb CW and from the M. smegmatis cytosol and CW revealed a single [M+H]3+ molecular ion (1218.9 m/z) that corresponded to the nonglycosylated MC (Fig. 3C). Thus, only fusion constructs with the Sec-dependent SP were presented in a manner that allowed glycosylation.

Fig. 3.

Mass spectrometry of peptides containing MCs derived from trypsin digestion of fusion proteins. (A) ESI-MS spectrum averaged over five scans corresponding to the LC elution containing the MC peptide of SP(+)FbpC-MC demonstrated primarily glycosylated forms of the peptide. (B) The 1327.0 m/z [M+H]3+ molecular ion marked by the asterisk in (A) was selected for ESI-MS/MS. The fragmentation pattern of the disaccharide linked to the peptide is shown in the inset. (C) Averaged ESI-MS spectrum of the LC elution containing the MC peptide of SP(–)FbpC-MC demonstrated an absence of glycosylation. (D) Averaged ESI-MS spectrum corresponding to the LC elution containing the MC peptide of SodA-MC. The region of the spectrum denoted by the asterisk is enlarged in the inset and reveals a minor [M+H]2+ molecular ion (1156.4 m/z) corresponding to the MC with a single hexose residue. (E) Averaged ESI-MS spectrum corresponding to the LC elution containing the MC peptide of ESAT-6-MC. The region of the spectrum denoted by the asterisk is enlarged in the inset and reveals a minor [M+H]3+ molecular ion (1312.5 m/z) for the MC with a single hexose residue. (F) Averaged ESI-MS spectrum corresponding to the LC elution containing the MC peptide of SP(+)ESAT-6-MC demonstrated a substantial level of glycosylation. (G) The dominant [M+H]3+ 1312.2 m/z molecular ion marked by an asterisk in (F) was selected for ESI-MS/MS. The fragmentation pattern of the monosaccharide linked to the peptide is shown in the inset.

Mtb produces several proteins that are secreted in a Sec-independent manner. Two such proteins, ESAT-6/Rv3875 and SodA/Rv3846, appear in the CF during early log-phase growth and lack SPs (19). To test the hypothesis that only Sec-translocation supports protein glycosylation, three additional recombinant fusion proteins were produced and assessed for glycosylation. The first two (SodA-MC and ESAT-6-MC) consisted of full-length SodA and ESAT-6 with C-terminal fusions to the Apa MC and a hexa-His tag (Fig. 2A). The final fusion SP(+)ESAT-6-MC had the SP and 34 additional amino acids of FbpC fused to the N terminus of ESAT-6-MC (Fig. 2A). This later chimera was designed to alter the natural export of ESAT-6 such that it was directed to the Sec-translocase. To further assure Sec-translocation of SP(+)ESAT-6-MC, a Mtb ΔRD1 mutant lacking ESAT-6 and its corresponding secretion machinery was used for recombinant expression (20). The SodA-MC and ESAT-6-MC fusions were abundant in all subcellular fractions of Mtb, including the CF and CW (Fig. 2B). SP(+)ESAT-6-MC, however, localized primarily to the cytoplasmic membrane and CW fractions of Mtb and Mtb ΔRD1 (Fig. 2B).

LC-ESI-MS/MS analyses of SodA-MC and ESAT-6-MC purified from the CF or CW of Mtb demonstrated low levels of glycosylation. Dominant molecular ions of [M+H]2+ 1075.5 m/z and [M+H]3+ 1258.4 m/z were observed and corresponded to the nonglycosylated MC peptides of SodA-MC and ESAT-6-MC (Fig. 3, D and E). Also observed were molecular ions (1156.4 m/z and 1312.5 m/z) for doubly and triply charged MC of SodA-MC and ESAT-6-MC that had single hexose residues. These molecular ions, however, were barely detectable compared with those of the nonglycosylated MCs (Fig. 3, D and E). The scarcity of glycosylation on SodA-MC and ESAT-6-MC was confirmed by the lack of ConA reactivity to purified SodA-MC and ESAT-6-MC (fig S3). The Sec-translocase-directed SP(+)ESAT-6-MC purified from the CW of Mtb and Mtb ΔRD1 was subjected to LC-ESI-MS/MS and produced two major [M+H]3+ molecular ions (1258.4 and 1312.2 m/z) corresponding to the MC with zero and one hexose residue (Fig. 3F). Moreover, and unlike ESAT-6-MC, the 1312.2 m/z molecular ion was clearly dominant, demonstrating relatively abundant glycosylation (compare Fig. 3, E and F), and this glycosylation was confirmed by MS/MS fragmentation of the 1312.2 m/z molecular ion (Fig. 3G).

These data support the hypothesis that, analogous to eukaryotic systems, Sec-translocation is required for protein mannosylation in Mtb. Secondary structures surrounding sites of O-glycosylation are important for glycosyltransferase recognition (5), and in prokaryotes only proteins maintained in an unfolded state are targeted for Sec-dependent translocation (21). Therefore, the requirement of Sec-translocation for protein O-mannosylation in Mtb may be associated with substrate structural requirements and the export of unfolded proteins. Protein folding before translocation may also explain the lack of glycosylation on the SodA and ESAT-6 fusion products exported by Sec-independent pathways (22, 23). Alternatively, efficient O-glycosylation of Sec-exported proteins could result from physical interactions between the Sec complex and a protein mannosyltransferase of Mtb. This would be similar to the proximity (30 to 40 Å) of the Sec61 translocon pore and the oligosaccharyltransferase (OST) complex and PMTs responsible for N- and O-glycosylation in eukaryotes (24, 25).

Our current data, along with the recent discovery of a eukaryotic-like OST N-glycosylation system in Campylobacter jejuni (13), demonstrate a previously unrealized evolutionary conservation between protein glycosylation pathways of eukaryotes and those of specific bacterial species. Moreover, the analogies to protein mannosylation of eukaryotes enable the establishment of a proposed mannoprotein assembly pathway in mycobacteria (fig. S4).

Supporting Online Material

Materials and Methods

Figs. S1 to S4


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article