N-Linked Glycosylation of Folded Proteins by the Bacterial Oligosaccharyltransferase

See allHide authors and affiliations

Science  17 Nov 2006:
Vol. 314, Issue 5802, pp. 1148-1150
DOI: 10.1126/science.1134351


N-linked protein glycosylation is found in all domains of life. In eukaryotes, it is the most abundant protein modification of secretory and membrane proteins, and the process is coupled to protein translocation and folding. We found that in bacteria, N-glycosylation can occur independently of the protein translocation machinery. In an in vitro assay, bacterial oligosaccharyltransferase glycosylated a folded endogenous substrate protein with high efficiency and folded bovine ribonuclease A with low efficiency. Unfolding the eukaryotic substrate greatly increased glycosylation. We propose that in the bacterial system, glycosylation sites are located in flexible parts of folded proteins, whereas the eukaryotic cotranslational glycosylation evolved to a mechanism presenting the substrate in a flexible form before folding.

The addition of N-linked oligosaccharides to proteins is a vital process in eukaryotic cells, involved in, for example, control of protein folding, sorting and stability (1, 2), cell-cell and cell-virus interactions (3), and host-immune responses (4, 5). Synthesis of glycoproteins occurs in the lumen of the endoplasmic reticulum (ER) where a multisubunit enzyme, oligosaccharyltransferase (OTase), transfers a Glc3Man9GlcNAc2 moiety from a lipid-pyrophosphate donor to selected asparagines within nascent polypeptide chains (6). N-Glycan addition occurs on polypeptides entering the ER through the translocation pore, a process referred to as cotranslocational glycosylation. Because the acceptor protein is translocated in an extended conformation, the OTase is thought to recognize polypeptides in an unfolded state (7, 8). Consistent with this model, the OTase has been found to interact with the Sec translocase in vivo (9). However, for some proteins, glycosylation can occur after the entire polypeptide has been translocated into the ER (posttranslocational glycosylation) (10), potentially while folding is still incomplete.

The homologous biosynthetic pathway of N-linked glycoproteins in bacteria is less complex. The glycosylation machinery of Campylobacter jejuni is encoded by a single gene cluster called pgl (protein glycosylation) (11), sufficient for recombinant protein glycosylation in Escherichia coli (12). The C. jejuni OTase, PglB, is a single, integral membrane protein with high sequence similarity to the catalytic subunit of the eukaryotic OTase, STT3. Like its eukaryotic counterpart, PglB transfers oligosaccharides (GalNAc5GlcBac, where Bac is 2,4-diacetamido-2,4,6-trideoxy-d-glucopyranoside) from a lipid-pyrophosphate donor to asparagine side chains within proteins (11). The consensus sequence recognized on the polypeptide is also similar between the two systems (D-X-N-Z-S/T and N-X-S/T for the bacterial and eukaryotic systems, respectively, where X and Z are any amino acid except proline) (13, 14). The N-X-S/T motif, as well as the requirement for a 2′-acetamido group at the reducing end (15), are thought to play integral roles in catalysis in both eukaryotic and prokaryotic OTases.

To determine if bacterial glycosylation was coupled to translocation as in the eukaryotic system, we analyzed the time course of glycosylation for an endogenous model C. jejuni glycoprotein, AcrA, by pulse-chase experiments (Fig. 1). Cells expressing the functional pgl locus from C. jejuni and AcrA containing a Sec-dependent signal peptide (ssPelB) (12) were metabolically labeled with a short pulse of 35S-methionine followed by addition of an excess of nonlabeled methionine (chase). AcrA in glycosylation-deficient cells (Fig. 1A) showed a single band after 135 s, suggesting that AcrA was synthesized and the signal peptide processed. The same was true for AcrA with both glycosylation sequons mutated (N123Q and N273Q) (Fig. 1B). At an equivalent time point, three forms were detected when the wild-type AcrA protein with two glycosylation sites was analyzed (Fig. 1A). The two upper bands were the mono- and diglycosylated AcrA. The intensity of the doubly glycosylated form increased over time, suggesting a continuous increase in glycosylation per molecule over the entire 60 min of the chase (Fig. 1B). Glycosylation continued long after signal sequence cleavage, suggesting posttranslocational glycosylation. This was also observed in AcrA point mutants with only a single glycosylation site.

Fig. 1.

Kinetics of bacterial N-glycosylation of AcrA in vivo. E. coli cells were labeled with 35S-methionine for 15 s (A) or 30 s (B) and chased for the indicated times with nonlabeled methionine. (A) Autoradiographs of SDS-PAGE of immunoprecipitated AcrA expressed in glycosylation-competent (pgl) and -deficient (pglmut) cells. The positions of unprocessed (ssPelB), monoglycosylated (g1), and diglycosylated (g2) AcrA are indicated. (B) Glycosylation after completion of signal-peptide cleavage from 2 to 60 min. AcrA proteins containing two (wild type), one (N123Q or N273Q), or no glycosylation (N123Q, N273Q) sequons were expressed in glycosylation-competent cells and analyzed as in (A).

The posttranslocational progression of AcrA glycosylation suggested a translocation-independent reaction. To address the hypothesis, we exchanged the PelB signal sequence of the AcrA protein by the TorA signal (ssTorA), directing the fusion protein to the twin-arginine translocation (TAT) machinery that is able to transport folded and oligomeric proteins across membranes (1618). The ssTorA-AcrA fusion protein was expressed in E. coli cells harboring the pgl or pglmut locus. Periplasmic, glycosylated AcrA was only detected in glycosylation- and TAT-competent cells (Fig. 2A). Mass spectrometric (MS) analysis of the putative doubly glycosylated band (AcrAg2) confirmed that the same positions as those in the protein translocated by the Sec-system were N-glycosylated (N123 and N273) (fig. S1). In cells with an inactive TAT-secretion system due to a tatC mutation (16), the complete absence of AcrA in periplasmic extracts (Fig. 2A) demonstrated that ssTorA-AcrA translocated exclusively via the TAT and not the Sec system. ssPelB-AcrA transported via the Sec system was glycosylated to a similar extent in both the wild-type and tatC mutant (Fig. 2B). Thus, N-glycosylation was uncoupled from the translocation machinery in our experimental system, and PglB was able to glycosylate folded proteins exported via the TAT machinery.

Fig. 2.

N-Glycosylation is independent of the translocation process. Periplasmic extracts derived from cells expressing (A) ssTorA-AcrA fusion protein or (B) ssPelB-AcrA fusion protein, in glycosylation-competent (+), glycosylation-deficient (–), and wild-type (wt) or tatC cells, were separated by SDS-PAGE, immunoblotted, and analyzed with antiserum raised against AcrA and with the R12 serum predominantly reactive toward the oligosaccharide modification (12).

To address the folding state of acceptor proteins during glycosylation, we used an in vitro glycosylation assay. PglB was expressed in E. coli cells and purified from solubilized membrane fractions (19). We then incubated PglB with cytoplasmically expressed and purified AcrA and a crude lipid extract (cLLO) of E. coli cells synthesizing the C. jejuni LLO by the pglmut locus (20). SDS–polyacrylamide gel electrophoresis (PAGE) and immunoblotting revealed the presence of glycosylated AcrA (Fig. 3A). MS analysis of AcrA with higher molecular weight confirmed the modification of N123 and N273 in this product by the C. jejuni oligosaccharide. Thus, PglB was able to glycosylate an independently expressed and purified AcrA in vitro at multiple sites on the same molecule. Eukaryotic OTases, by contrast, do not possess any in vitro activity for folded polypeptides (21).

Fig. 3.

In vitro glycosylation of recombinant AcrA and GFP-21. Cytoplasmically expressed AcrA (cAcrA) (A) or GFP-21 (B) were used as acceptor proteins and in vitro glycosylated. Reaction products were separated by SDS-PAGE, immunoblotted, and analyzed with antiserum raised against AcrA and with the R12 serum predominantly reactive toward the oligosaccharide modification (12). The position of monoglycosylated (g1) and diglycosylated (g2) protein is indicated. Controls omitting either PglB (“no enzyme”), cLLO, or the acceptor protein are shown. Reactions from (B) were Coomassie stained (C), or their fluorescence was measured (D) (19). Relative fluorescence intensities are indicated. The high signal in the “no acceptor” control is likely an artifact due to light scattering by the cLLO.

To investigate the folding state of the acceptor protein during glycosylation directly, we grafted a 21–amino acid sequence containing the AcrA N123 glycosylation site into an insertion-tolerant loop present in the green fluorescent protein (GFP) (22). This GFP-derivative (GFP-21) was tested for accepting an N-glycan in our in vitro glycosylation assay (20). A higher molecular weight protein that was observed to cross-react with the glycosylation-specific serum band, GFP-21, was glycosylated to >90% at the single site in this reaction (Fig. 3C). The shift was due to a glycan attachment to the glycosylation site present in the grafted AcrA-derived loop (fig. S2). Moreover, the fluorescence of the product was similar to that of unglycosylated GFP (Fig. 3D), and thus the glycosylation reaction occurred on the folded GFP. Thus, PglB modified a peptide displayed on a folded protein, although it is likely that the grafted loop itself is relatively flexible.

To check the folding-dependent recognition of consensus sequons by PglB, we analyzed the PglB-dependent in vitro glycosylation of the eukaryotic glycoprotein bovine ribonuclease A (RNaseA). The native N-glycosylation site N34 of this protein is located in a structured domain. Different folding variants of the RNaseA allowed us to analyze the effect of unfolding on glycosylation. Consequently, RNaseA was expressed with a single point mutation (S32D), producing a bacterial consensus site at the native N-glycosylation site N34. Oxidative refolding of RNaseS32D from inclusion bodies (20) yielded enzymatically active protein (fig. S3), showing that RNaseS32D was able to fold into its native conformation despite the mutation.

Chemical treatments with denaturing, reducing, oxidizing, and alkylating agents yielded two RNaseS32D oxidation isomers (Fig. 4A). In reduced and alkylated RNaseS32D (RA), all four disulfide bonds were reduced in denaturing solution, and cysteines were alkylated to inhibit further disulfide bond formation. Rapid oxidation before alkylation led to another set of oxidation isomers, “scrambled” RNaseS32D (SC). Scrambled RNases contain randomly oxidized SS bonds, producing a heterogeneous population of proteins. A third RNaseS32D form was synthesized by limited proteolysis with subtilisin (20), which removes the N-terminal 21 amino acids of the protein. The resulting RNase S-protein (SP), like the RA and SC forms, was enzymatically inactive (fig. S3), but retained its native disulfide bonds and thus about 50% secondary structure, whereas the SC and RA forms appeared as random coils, as judged from far-ultraviolet circular dichroism (far-UV-CD) spectroscopy (Fig. 4B) (23). All four forms served as substrates for PglB (Fig. 4C). Glycosylation of the active RNaseS32D occurred with low efficiency (Fig. 4C), and the small amount of glycosylated RNaseS32D was active, as indicated by the zymogram assay (20) (Fig. 4C). The other forms were modified quantitatively (Fig. 4C). Thus, nonstructured protein domains are better substrates for PglB glycosylation than are folded ones. No glycosylation at all was observed with the same substrate proteins that lacked the S32D mutation (fig. S4B).

Fig. 4.

In vitro glycosylation of recombinant bovine RNaseS32D. (A) Synthesis scheme of the different RNase folding isomers. Inclusion bodies were produced in E. coli, refolded in vitro, and chemically treated as indicated. (B) Far-UV CD spectra of different RNaseS32D folding variants to measure α-helical and β-sheet structure of the active (Native), reduced-alkylated (RA), scrambled (SC), and S-protein (SP) versions of RNaseS32D. (C) In vitro glycosylation assay with equimolar amounts of RNaseS32D acceptor proteins incubated without (–) or with (+) PglB. Analysis was performed by immunoblotting with the indicated antisera and by zymogram analysis (20).

Our results show that completely folded proteins can be glycosylated both in vivo and in vitro. Bacterial OTase glycosylates native AcrA protein as well as an acceptor sequence grafted into the active GFP protein. In contrast, fully folded RNaseS32D was weakly glycosylated, whereas partial or complete unfolding strongly improved substrate activity.

The observation that the folding states of the acceptor protein affects glycosylation efficiency leads us to conclude that a specific substrate conformation must be adopted during the glycosylation process, most likely the Asn-turn (24). This makes potential acceptor sites present in a fixed environment suboptimal substrates for the bacterial OTase. We predict that native glycosylation sites in bacterial proteins will be located in locally flexible structures.

In contrast, the coupling of glycosylation and translocation in eukaryotes releases N-glycosylation from such structural constraints and, in combination with the less stringent primary sequence requirement, results in a more versatile and general glycosylation system.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S4

Table S1


References and Notes

View Abstract

Navigate This Article